TO THE SERIES
The aim of the Handbooks in Economics series is to produce Handbooks for various
branches of economics, each of which is a definitive source, reference, and teaching
supplement for use by professional researchers and advanced graduate students. Each
Handbook provides self-contained surveys of the current state of a branch of economics
in the form of chapters prepared by leading specialists on various aspects of this
branch o f economics. These surveys summarize not only received results but also
newer developments, from recent journal articles and discussion papers. Some original
material is also included, but the main goal is to provide comprehensive and accessible
surveys. The Handbooks are intended to provide not only useful reference volumes for
professional collections but also possible supplementary readings for advanced courses
for graduate students in economics.
KENNETH J. ARROW and MICHAEL D. 1NTRILIGATOR
PUBLISHER'S
NOTE
For a complete overview of the Handbooks in Economics Series, please refer to the
listing at the end of this volume.
CONTENTS O F T H E HANDBOOK
VOLUME 1A
PART 1 - E M P I R I C A L A N D H I S T O R I C A L P E R F O R M A N C E
Chapter 1
Business Cycle Fluctuations in US Macroeconomic Time Series
JAMES H. STOCK and MARK W WATSON
Chapter 2
Monetary Policy Shocks: What Have we Learned and to What End?
LAWRENCE J. CHRISTIANO, MARTIN EICHENBAUM and CHARLES L. EVANS
Chapter 3
Monetary Policy Regimes and Economic Performance: The Historical Record
MICHAEL D. BORDO AND ANNA J. SCHWARTZ
Chapter 4
The New Empirics of Economic Growth
STEVEN N. DURLAUF and DANNY T. QUAH
PART 2 - M E T H O D S O F D Y N A M I C A N A L Y S I S
Chapter 5
Numerical Solution of Dynamic Economic Models
MANUEL S. SANTOS
Chapter 6
Indeterminacy and Sunspots in Macroeconomics
JESS BENHABIB and ROGER E.A. FARMER
Chapter 7
Learning Dynamics
GEORGE W. EVANS mad SEPPO HONKAPOHJA
Chapter 8
Micro Data and General Equilibrium Models
MARTIN BROWNING, LARS PETER HANSEN and JAMES J. HECKMAN
vii
viii
PART 3 - MODELS
OF ECONOMIC
GROWTH
Chapter 9
Neoclassical Growth Theory
ROBERT M. SOLOW
Chapter 10
Explaining Cross-Country Income Differences
ELLEN R. McGRATTAN and JAMES A. SCHMITZ, Jr.
VOLUME
1B
PART 4 - CONSUMPTION
AND INVESTMENT
Chapter 11
Consumption
ORAZIO E ATTANASIO
Chapter 12
Aggregate Investment
RICARDO J. CABALLERO
Chapter 13
Inventories
VALERIE A. RAMEY and KENNETH D. WEST
PART 5 - M O D E L S
OF ECONOMIC
FLUCTUATIONS
Chapter 14
Resuscitating Real Business Cycles
ROBERT G. KING AND SERGIO T. REBELO
Chapter 15
Staggered Price and Wage Setting in Macroeconomics
JOHN B. TAYLOR
Chapter 16
The Cyclical Behavior of Prices and Costs
JULIO J. ROTEMBERG and MICHAEL WOODFORD
Chapter 17
Labor-Market Frictions and Employment Fluctuations
ROBERT E. HALL
Chapter 18
Job Reallocation, Employment Fluctuations and Unemployment
DALE T. MORTENSEN and CHRISTOPHER A. PISSAR/DES
VOLUME 1C
PART 6 - F I N A N C I A L M A R K E T S A N D T H E M A C R O E C O N O M Y
Chapter 19
Asset Prices, Consumption, and the Business Cycle
JOHN Y. CAMPBELL
Chapter 20
Human Behavior and the Efficiency of the Financial System
ROBERT J. SHILLER
Chapter 21
The Financial Accelerator in a Quantitative Business Cycle Framework
BEN S. BERNANKE, MARK GERTLER and SIMON GILCHRIST
PART 7 - M O N E T A R Y A N D F I S C A L P O L I C Y
Chapter 22
Political Economics and Macroeconomic Policy
TORSTEN PERSSON and GUIDO TABELLINI
Chapter 23
Issues in the Design of Monetary Policy Rules
BENNETT T. McCALLUM
Chapter 24
Inflation Stabilization and BOP Crises in Developing Countries
GUILLERMO A. CALVO and CARLOS A. VI~GH
Chapter 25
Government Debt
DOUGLAS W. ELMENDORF AND N. GREGORY MANKIW
Chapter 26
Optimal Fiscal and Monetary Policy
V.V. CHAR/and PATRICK J. KEHOE
ix
Purpose
The Handbook of Macroeconomics aims to provide a survey of the state of knowledge
in the broad area that includes the theories and facts of economic growth and economic
fluctuations, as well as the consequences of monetary and fiscal policies for general
economic conditions.
Progress in Macroeconomics
Macroeconomic issues are central concerns in economics. Hence it is surprising that
(with the exception of the subset of these topics addressed in the Handbook of
Monetary Economics) no review of this area has been undertaken in the Handbook
of Economics series until now.
Surprising or not, we find that now is an especially auspicious time to present such a
review of the field. Macroeconomics underwent a revolution in the 1970's and 1980's,
due to the introduction of the methods of rational expectations, dynamic optimization,
and general equilibrium analysis into macroeconomic models, to the development of
new theories of economic fluctuations, and to the introduction of sophisticated methods
for the analysis of economic time series. These developments were both important
and exciting. However, the rapid change in methods and theories led to considerable
disagreement, especially in the 1980's, as to whether there was any core of common
beliefs, even about the defining problems of the subject, that united macroeconomists
any longer.
The 1990's have also been exciting, but for a different reason. In our view, the
modern methods of analysis have progressed to the point where they are now much
better able to address practical or substantive macroeconomic questions - whether
traditional, new, empirical, or policy-related. Indeed, we find that it is no longer
necessary to choose between more powerful methods and practical policy concerns.
We believe that both the progress and the focus on substantive problems has led to
a situation in macroeconomics where the area of common ground is considerable,
though we cannot yet announce a "new synthesis" that could be endorsed by most
scholars working in the field. For this reason, we have organized this Handbook around
substantive macroeconomic problems, and not around alternative methodological
approaches or schools of thought.
xi
xii
Preface
The extent to which the field has changed over the past decade is considerable, and
we think that there is a great need for the survey of the current state ofmacroeconomics
that we and the other contributors to this book have attempted here. We hope that the
Handbook of Macroeconomics will be useful as a teaching supplement in graduate
courses in the field, and also as a reference that will assist researchers in one area of
macroeconomics to become better acquainted with developments in other branches of
the field.
Overview
The Handbook of Macroeconomics includes 26 chapters, arranged into seven parts.
Part 1 reviews evidence on the Empirical and Historical Performance of the
aggregate economy, to provide factual background for the modeling efforts and
policy discussion of the remaining chapters. It includes evidence on the character
of business fluctuations, on long-run economic growth and the persistence of crosscountry differences in income levels, and on economic performance under alternative
policy regimes.
Part 2 on Methods of Dynamic Analysis treats several technical issues that arise in
the study of economic models which are dynamic and in which agents' expectations
about the future are critical to equilibrium determination. These include methods for
the calibration and computation of models with intertemporal equilibria, the analysis
of the determinacy of equilibria, and the use of "learning" dynamics to consider the
stability of such equilibria. These topics are important for economic theory in general,
and some are also treated in the Handbook of Mathematical Economics, The Handbook
of Econometrics, and the Handbook of Computational Economics, for example, from
a somewhat different perspective. Here we emphasize results - such as the problems
associated with the calibration of general equilibrium models using microeconomic
studies - that have particular application to macroeconomic models.
The Handbook then tunas to a review of theoretical models of macroeconomic
phenomena. Part 3 reviews Models of Economic Growth, including both the
determinants of long-run levels of income per capita and the sources of cross-country
income differences. Both "neoclassical" and "endogenous" theories of growth are
discussed. Part 4 treats models of Consumption and Investment demand, from the
point o f view of intertemporal optimization. Part 5 covers Models of Economic
Fluctuations. In the chapters in this part we see a common approach to model
formulation and testing, emphasizing intertemporal optimization, quantitative general
equilibrium modeling, and the systematic comparison of model predictions with
economic time series. This common approach allows for consideration of a variety
of views about the ultimate sources of economic fluctuations and of the efficiency of
the market mechanisms that amplify and propagate them.
Part 6 treats Financial Markets and the Macroeconomy. The chapters in this part
consider the relation between financial market developments and aggregate economic
Preface
xiii
activity, both from the point of view of how business fluctuations affect financial
markets, and how financial market disturbances affect overall economic activity. These
chapters also delve into the question of whether financial market behavior can be
understood in terms of the postulates of rational expectations and intertemporal
optimization that are used so extensively in modern macroeconomics-an issue of
fundamental importance to our subject that can be, and has been, subject to special
scrutiny in the area of financial economics because of the unusual quality of available
data.
Finally, Part 7 reviews a number of Monetary and Fiscal Policy issues. Here we
consider both the positive theory (or political economics) of government policymaking
and the normative theory. Both the nature of ideal (or second-best) outcomes according
to economic theory and the choice of simple rules that may offer practical guidance
for policymakers are discussed. Lessons from economic theory and from experience
with alternative policy regimes are reviewed. None of the chapters in this part
focus entirely on international, or open economy, macroeconomic policies, because
many such issues are addressed in the Handbook of International Economics.
Nevertheless, open-economy issues cannot be separated from closed-economy issues
as the analysis of disinflation policies and currency crises in this part of the Handbook
of Maeroeconomics, or the analysis of policy regimes in the Part I of the Handbook
of Macroeconomics make clear.
Acknowledgements
Our use of the pronoun "we" in this preface should not, of course, be taken to
suggest that much, if any, of the credit for what is useful in these volumes is due
to the Handbook's editors. We wish to acknowledge the tremendous debt we owe to
the authors of the chapters in this Handbook, who not only prepared the individual
chapters, but also provided us with much useful advice about the organization of the
overall project. We are grateful for their efforts and for their patience with our slow
progress toward completion of the Handbook. We hope that they will find that the
final product justifies their efforts. We also wish to thank the Federal Reserve Bank of
New York, the Federal Reserve Bank of San Francisco, and the Center for Economic
Policy Research at Stanford University for financial support for two conferences on
"Recent Developments in Macroeconomics" at which drafts of the Handbook chapters
were presented and discussed, and especially to Jack Beebe and Rick Mishkin who
made these two useful conferences happen. The deadlines, feedback, and commentary
at these conferences were essential to the successful completion of the Handbook.
We also would like to thank Jean Koentop for managing the manuscript as it neared
completion.
Stanford, California
Princeton, New Jersey
John B. Taylor
Michael Woodford
Chapter 11
CONSUMPTION
ORAZIO P. ATTANASIO
Contents
Abstract
Keywords
1. Introduction
2. Stylised facts
742
742
743
745
746
750
2.2.1. Nature of the data sets and their comparability with the National Account data
2.2.2. Life cycle profiles
4. Aggregation issues
4.1. Aggregation across consumers
4.2. Aggregation across commodities
751
752
760
761
762
765
770
772
777
780
781
781
782
783
784
785
785
787
* A preliminary draft of this chapter was presented at a conference at the New York Fed., February 27-28
1997, where I received useful comments from my discussant, Chris Carroll and several participants.
Tullio JappeUi provided many careful and insightful comments for which I am very grateful. I would
like to thank Margherita Borella for research assistance and James Sefton for providing me with the UK
National Accounts Data. Material from the FES made available by the ONS through the ESRC Data
Archive has been used by permission of the Controller of HMSO. Neither the ONS nor the ESRC Data
Archive bear any responsibility for the analysis or interpretation of the data reported here.
742
5.2.3. Conditional second (and higher) moments
5.3. Micro data: some evidence
6. Where does the life cycle model stand?
7. Insurance and inequality
8. Intertemporal non-separability
8.1. Durables
8.2. Habit formation
9. Conclusions
References
O.P. Attanasio
788
789
791
795
798
799
802
804
805
Abstract
Consumption is the largest component of GDP. Since the 1950s, the life cycle and
the permanent income models have constituted the main analytical tools to the study
of consumption behaviour, both at the micro and at the aggregate level. Since the
late 1970s the literature has focused on versions of the model that incorporate the
hypothesis of Rational Expectations and a rigorous treatment of uncertainty. In this
chapter, 1 survey the most recent contribution and assess where the life cycle model
stands. My reading of the evidence and of recent developments leads me to stress two
points: (i) the model can only be tested and estimated using a flexible specification
of preferences and individual level data; (ii) it is possible to construct versions of the
model that are not rejected by the data. One of the main problems of the approach
used in the literature to estimate preferences is the lack of a 'consumption function'.
A challenge for future research is to use preference parameter estimates to construct
such functions.
Keywords
consumption, life cycle model, household behaviour
JEL classification: E2
Ch. 11:
Consumption
743
1. Introduction
In most developed economies, consumption accounts for about two thirds of GDP.
Moreover, it is from consumption that, in all likelihood, utility and welfare are in
large part determined. It is therefore natural that macroeconomists have devoted
a considerable amount of research effort to its study. In modern macroeconomics,
consumption is typically viewed as part of a dynamic decision problem. There is
therefore another sense in which an tmderstanding of consumption is central for
macroeconomics. Consumption decisions are also saving decisions from which the
funds available for capital accumulation and investment arise. Therefore, consumers
attitudes to saving, risk bearing and uncertainty are crucial to understand the behaviour
of capital markets, the process of investment and growth and development. It is
not by chance that modern consumption theory is also used to characterise asset
prices equilibrium conditions. The desire consumers might have to smooth fluctuations
over time determines the need for particular financial instruments or institutions.
Understanding recent trends in consumption and saving is crucial to the study, both
positive and normative, of the development of financial markets, of the institutions that
provide social safety nets, of the systems through which retirement income is provided
and so on.
One of the main themes of this chapter is that consumption decisions cannot be
studied in isolation. Exactly because consumption and saving decisions are part of a
dynamic optimisation problem, they are determined jointly with a number of other
choices, ranging from labour supply to household formation and fertility decisions,
to planned bequests. While modelling all aspects of human economic behaviour
simultaneously is probably impossible, it is important to recognise that choices are
taken simultaneously and to control for the effects that various aspects of the economic
environment in which consumers live might have on any particular choice. This is
particularly true if one wants to estimate the parameters that characterise individual
preferences.
Implicit in this argument is another of the main themes of this chapter: consumption
decisions should be modelled within a well specified and coherent optimisation model.
Such a model should be flexible and allow for a variety of factors. Indeed, I think it is
crucial that the model should be interpreted as an approximation of reality and should
allow for a component of behaviour that we are not able to explain. However, such a
model is crucial to organise our thinking and our understanding of the data. Without
a structural model it is not possible to make any statement about observed behaviour
or to evaluate the effect of any proposed change in economic policy.
This, however, is not a call for a blind faith in structural models. Inferences should
always be conditional on the particular identification restrictions used and on the
particular structural model used. Such models should also be as flexible as possible
and incorporate as much information about individual behaviour as is available. It
should be recognised, however, that without such models we cannot provide more than
a statistical description of the data.
744
0.17. Attanasio
The other main theme of the analysis in this chapter is that to understand aggregate
trends it is necessary to conduct, in most situations, a detailed analysis of individual
behaviour. In other words, aggregation problems are too important to be ignored. This
obviously does not mean that the analysis of aggregate time series data is not useful.
Indeed, I start the chapter with a brief summary of the main time series properties of
consumption. Estimation of structural models of economic behaviour, however, cannot
be performed using aggregate data only.
This chapter is not an exhaustive survey of the literature on consumption: such a
literature has grown so much that it would be hard even to list it, let alone summarise
all the contributions. What I offer, instead, is a discussion of the current status of our
knowledge, with an eye to what I think are the most interesting directions for future
research. In the process of doing so, however, I discuss several of the most important
and influential contributions. Omissions and exclusions are unavoidable and should
not be read as indicating a negative judgement on a particular contribution. At times,
I simply chose, among several contributions, those that most suited my arguments and
helped me the most to make a given point. Moreover, notwithstanding the length of the
chapter, not every sub-fields and interesting topic has been covered. But a line had to
be drawn at some point. There are four fields that I did not included in the chapter and
over which I have agonised considerably. The first is asset pricing: while much of the
theoretical material I present has direct implications for asset prices, I decided to omit a
discussion of these implications as there is an entire chapter of this Handbook devoted
to these issues. The second is the axiomatisations of behaviour under uncertainty
alternative to expected utility. There are several interesting developments, including
some that have been used in consumption and asset pricing theory, such as the KrepsPorteus axiomatisation used by Epstein and Zin (1989, 1991) in some fascinating
papers. The third is the consideration of within-household allocation of resources.
There is some exciting research being developed in this area, but I decided to draw the
line of 'macro' at the level of the individual household. Finally, I do not discuss theories
of consumption and saving behaviour that do not assume optimising and fully rational
behaviour. Again, there is some exciting work in the area of social norms, mental
accounting, time varying preferences, herd behaviour and so on. In the end, however,
I decided that it would not fit with the rest of the chapter and rather than giving just
a nod to this growing part of the literature I decided to leave it out completely.
The chapter is organised as follows. In Section 2, I start with a brief description
of some stylised facts about consumption. These include both facts derived from
aggregate time series data and from household level data. Throughout the section,
I use in parallel data from two countries: the USA and the UK.
In Section 3, I discuss at length what I think is the most important model of
consumption behaviour we have, the life cycle model. In that section, I take a wide
view of what I mean by the life cycle model: definitely not the simple textbook
version according to which the main motivation for saving is the accumulation of
resources to provide for retirement. Instead, I favour a flexible version of the model
where demographics, labour supply, uncertainty and precautionary saving and possibly
745
bequests play an important role. In other words, I consider the life cycle model
as a model in which consumption decisions are determined within an intertemporal
optimisation framework. What elements of this model turn out ot be more important
is largely an empirical matter. Indeed, even the presence of liquidity constraints, or
borrowing restrictions, can and should be incorporated within this framework.
In Section 4, I discuss aggregation problems. In particular, I focus on two different
kinds of aggregation: that across consumers and that across commodities. The aim of
this section is not just to give lip service to the aggregation issues and proceed to sweep
them under the carpet. With the development of computing and storage capability and
with the availability of increasing large number of micro data sets, it is important
to stress that scientific research on consumption behaviour cannot afford to ignore
aggregation issues.
In Section 5, I consider the empirical evidence on the life cycle model and discuss
both evidence from aggregate time series data and evidence from micro data. In this
section I also address a number of econometric problems with the analysis of Euler
equations for consumption. In Section 6, I take stock on what I think is the status of
the life cycle model, given the evidence presented in Section 5.
In Section 7, I address the issues of insurance and inequality. In particular, I present
some of the tests of the presence of perfect insurance and discuss the little evidence
there is on the evolution o f consumption inequality and its relationship to earning
inequality.
Most of the models considered up to this point assume time separability of
preferences. While such a hypothesis is greatly convenient from an analytical point
of view, it is easy to think of situations in which it is violated. In Section 8, I discuss
to forms of time dependence: that induced by the durability of commodities and habit
formation. Section 9 concludes the chapter.
2. Stylised facts
In this section, I document the main stylised facts about consumption behaviour
using both aggregate and individual data. I consider two components of consumption
expenditure: on non-durable and services and on durables. In addition I also consider
disposable income. While most of the facts presented here are quite well established,
the evidence in this section constitute the background against which one should set
the theoretical model considered in the rest of the chapter.
The data used come from two western countries: the United States and the United
Kingdom. I have deliberately excluded from the analysis developing or less developed
countries as they involve an additional set of issues which are not part of the present
discussion. Among the developed countries I have chosen the USA and the UK both
because data from these two countries have been among the most widely studied and
because the two countries have the best micro data on household consumption. For
O.P. Attanasio
746
4000
UK
USA - disposable income and consumption
~------150000
100000
2000
50000
J
J
97.1
2319
55
6's
75
as
&
65
75
85
95
Fig. 1. Disposable income (top curve) consumption, divided into durables (bottom curve) and nondurables (middle curve).
the UK, in particular, the Family Expenditure Survey runs for 25 consecutive years,
giving the possibility of performing interesting exercises.
2.1. Aggregate time series data
In this section, I present some of the time series properties of consumption expenditure
and of disposable income. While the models considered in the following sections
refer to household behaviour, typically the consumption aggregates considered in the
National Account statistics include outlays of a sector that, together with households,
includes other entities, such as charities, whose behaviour is unlikely to be determined
by utility maximisation. While this issue is certainly important, especially for structural
tests of theoretical models of household behaviour, in the analysis that follows I ignore
it and, instead of isolating the part of total expenditure to be attributed to households,
I present the time series properties of National Account consumption. Seslnick (1994)
has recently stressed the importance of these issues.
In Figure 1, I plot household (log) disposable income along with consumption
divided into durables and non-durables and services for the UK and the USA. The
series have quarterly frequency and run from 1959:1 to 1996:3 for the USA and
from 1965:1 to 1996:2 for the UK. The data are at constant prices and are seasonally
adjusted. From the figure, it is evident that non-durable consumption is smoother than
disposable income. Durable consumption, on the other hand, which over the sample
accounts, on average, for 13% of total consumption in the USA and around 14% in
the UK, is by far the most volatile of the three time series. This is even more evident
in Figure 2 where I plot the annual rate of changes for the three variables. In Table 1,
I report the mean and standard deviation of the three variables. These figures confirm
and quantify the differences in the variability of the three variables considered.
In Tables 2 and 3, I consider two alternative ways of summarising the time series
properties of the three series I analyse for both countries. In Table 2, 1 report the
estimates of the coefficient of an MA(12) model for the same series. The advantage
of such an un-parsimonious model is that it avoids the sometimes difficult choice
among competing ARMA representations. Furthermore, its impulse response function
Ch. 11:
747
Consumption
0.2
0.2
0.1
0.1
-0.1
-0.1
-0.2
-0.2
55
65
75
85
95
65
75
85
95
Disposable income
Nondurable consumption
Durable expenditure
UK
Mean
St. dev.
Mean
St. dev.
0.032
0.023
0.048
0.025
0.018
0.069
0.026
0.017
0.043
0.026
0.021
0.112
can be easily read from the estimated coefficients. I also purposely decided to be
agnostic about the presence o f random walks in the time series consumption or income,
even though this has implications for the so called 'excess smoothness' puzzle briefly
discussed below. In Table 3, instead, I report the M a x i m u m Likelihood estimates o f a
parsimonious A R M A m o d e l for the first differences o f the log o f the three variables.
While in some cases there were alternative specifications that fitted the data as well as
those reported in the table, the latter all pass several diagnostic tests. The Q-statistics
reported in the table indicates that the representations chosen capture adequately the
dynamic behaviour o f the series over the period considered.
The time series properties o f the rate o f growth o f the three variables are remarkably
different. Notice, in particular, the fact that both in the U K and in the USA, the sum o f
the M A coefficients for non-durable consumption is positive, while that for durables
is negative. The time series properties o f non-durable consumption differ remarkably:
in Table 2 the sum o f the first 12 M A coefficient is much larger in the U K than in the
USA. Furthermore, while the US data are well represented by an M A ( 3 ) (with the first
and third lag large and very strongly significant), the U K require an AR(2) model 1.
I The presence of an MA(3) effect in the non-durable series for the USA is evident even in the MA(12)
representation but it is not very robust. If one truncates the sample to 1990 or dummies out the few
quarters corresponding to the 1990-91 recession, 03 is estimated non-significantly different from zero
748
O.P Attanasio
Table 2
MA(12) representation a
US
0i
01
02
03
04
05
06
07
08
09
010
011
012
~ ) 2 I 0i
UK
Disposable
income
Non-durable
consumption
Durable
consumption
Disposable
income
Non-durable
consumption
Durable
consumption
-0.30
(0.091)
0.15
(0.094)
0.092
(0.094)
-0.092
(0.088)
-0.15
(0.087)
0.11
(0.088)
-0.13
(0.087)
-0.17
(0.088)
0.38
(0.088)
0.20
(0.095)
-0.06
(0.096)
-0.27
(0.091)
-0.25
0.41
(0.096)
0.18
(0.103)
0.43
(0.104)
0.12
(0.110)
-0.057
(0.108)
0.100
(0.108)
0.11
(0.107)
-0.20
(0.107)
0.05
(0.109)
-0.03
(0.100)
0.05
(0.099)
0.08
(0.092)
1.23
-0.092
(0.088)
-0.035
(0.089)
0.063
(0.089)
0.084
(0.086)
-0.16
(0.085)
0.15
(0.077)
-0.45
(0.077)
-0.021
(0.085)
-0.23
(0.085)
-0.03
(0.088)
0.005
(0.087)
-0.23
(0.086)
-0.95
-0.10
(0.094)
0.12
(0.095)
-0.06
(0.092)
-0.18
(0.088)
-0.19
(0.089)
0.22
(0.088)
0.21
(0.088)
0.14
(0.087)
-0.14
(0.086)
-0.20
(0.087)
0.05
(0.088)
-0.05
(0.086)
-0.18
0.005
(0.094)
0.20
(0.093)
0.004
(0.093)
0.28
(0.092)
0.19
(0.093)
0.19
(0.094)
0.09
(0.094)
0.22
(0.092)
-0.11
(0.090)
0.23
(0.091)
0.18
(0.091)
0.03
(0.093)
1.51
-0.29
(0.094)
-0.14
(0.096)
0.24
(0.097)
-0.45
(0.099)
0.15
(0.106)
0.05
(0.107)
-0.07
(0.106)
-0.18
(0.104)
-0.08
(0.098)
0.02
(0.095)
-0.20
(0.094)
-0.02
(0.089)
-0.97
both in the MA(12) and in the MA(3) model. The same result is obtained if one excludes services from
this series.
Ch. 11:
749
Consumption
Table 3
ARMA representation a
Variable
US
Disposable
income
'[Pl
--
--
'10 2
--
01
-0.19
(0.088)
0 2
--
03
UK
-0.77
(0.293)
-
0.684
(0.339)
0.38
(0.083)
0.18
0.015
(0.087)
0.28
(0.087)
-
-1.09
(0.098)
-0.45
(0.082)
0.85
(0.077)
0.85
(0.077)
7.22
(0.30)
(0.088)
Q-stat
(p-value)
13.40
(0.10)
0.39
(0.082)
7.35
(0.28)
10.09
(0.26)
0.684
(0.339)
11.79
(0.11)
11.49
(0.12)
2 For durable consumption in the UK, the best model is an ARMA(2,1), by far the most complex model
I fitted to these data.
750
0.17 Attanasio
structural interpretation to this type of results. This does not mean, however, that
aggregate time series studies are not useful.
The careful specification of a flexible time series model for consumption and other
variables can be quite informative, especially if the dynamic specification allows for
the type of dynamic effects implied by the microeconomic behaviour. Several of the
studies by David Hendry and his collaborators are in this spirit; one of the most widely
cited examples of this literature is the paper by Davidson et al. (1978).
The approach taken in these papers, which received a further motivation by the
development of cointegration techniques, is to estimate a stable error correction
model which relates consumption to other variables. The statistical model then allows
to identify both short run and long run relationships between consumption and its
determinants. While the theory can be informative on the choice of the relevant
variables and even on the construction of the data series, it does not provide explicit
and tight restrictions on the parameters of the model. A good example of a creative
and informative use of this type of techniques is Blinder and Deaton (1985). While it
is difficult to relate this type of models to structural models and therefore they cannot
be directly used for evaluating economic policy, they constitute useful instruments for
summarising the main features of the data and, if used carefully, for forecasting. Often
the lack of micro economic data makes the use of aggregate time series data a necessity.
The only caveat is that these studies cannot be used to identify structural parameters.
3 Both data sets containvery detailed information on the expenditure on individual commodities. Some
of this information can be used to attribute some items to some household members. For many items,
however, such attribution is difficult both in practice and conceptually. Browning (1987) has imputed
expenditure on alcoholand tobaccoto the adults to checkwhether predictedchangesin householdincome
and composition(such as the arrival of children with consequent- at least temporary- withdrawal from
the labour force of the wife) cause changes in consumption. Gokhale, Kotlikoffand Sabelhaus (1996) in
their study of saving behaviourhave attempted to impute all of consumptionto the individual household
members.
751
This approach reflects the fact that the theoretical discussion in the next sections will
be focused around the life cycle model.
2.2.1. Nature of the data sets and their comparability with the National Account
data
The FES is now available for 25 consecutive years. Each year around 7000 households
are interviewed and supply information on their consumption patterns as well as their
demographic characteristics and several other economic variables such as employment
status, income, education and so on. Each household stays in the sample for two weeks,
during which it fills a diary in which all expenditure items are reported. At the end o f
the two week period an interviewer collects the diaries and asks additional information
on durables acquired during the previous three months and on all major expenditure
items reported in the diary and periodic expenditures such as utilities.
The CEX is available on a continuous and roughly homogeneous basis since
1980. Each year about 7000 different households are interviewed for 4 subsequent
interviews, with quarterly frequency 4. Each month new households enter the survey to
replace those that have completed their cycle o f interviews. During each interview the
household is asked to report expenditure on about 500 consumption categories during
each o f the three months preceding the interview 5. The panel dimension of the CEX
is unfortunately very short: because each household is only interviewed four times,
seasonal variability is likely to dominate life cycle and business cycle movements. In
what follows, I do not exploit the panel dimension o f the survey.
There have been several discussions about the quality o f survey data and the
importance o f measurement error and about their ability to reproduce movements in
aggregate consumption. Several studies, both in the U S A and the UK, have addressed
the issue 6. It should be stressed that the aggregated individual data and the National
Account aggregate should be expected to differ for several reasons. First o f all, for
many consumption categories, the definitions used in the surveys and in the National
Accounts are quite different. Housing, for instance, includes imputed rents in the
National Accounts data but does not in the surveys. In the CEX, health expenditure
4 In total there are data for over 20000 interviews per year. Each household is in fact interviewed
five times. However, the Bureau for Labor Statistics does not release information on the first (contact
interview). The Bureau of Labor Statistics also rtms a separate survey based on diaries which collects
information on food consumption and 'frequently purchased items'.
s Unfortunately,the monthly decomposition of the quarterly expenditure is not very reliable. For several
commodities and for many households, the quarterly figure is simply divided by three. Given the rotating
nature of the sample, the 'quarters' of expenditure do not coincide perfectly. For instance, somebody
interviewed in December will report consumption in September, October and November, while somebody
interviewed in November will report consumption in August, September and October.
6 See, for instance, Seslnick (1992) and Paulin et al. (1990) for comparisons between the aggregate
Personal Consumption Expenditure and the CEX in the USA and the papers in Banks and Johnson
(1997) for comparisons on the FES and the UK National Accounts.
752
O.P. Attanasio
7 There are substantial differences in this ratio between the early CEX surveys (1960-61 and 1972-73)
and those of the 1980s, probably due to the differences in the methodology employed. In the FES the
one commodity for which a (downward) trend in the ratio is apparent is tobacco.
8 Ghez and Becker (1975) use observations on individual of different ages to study life cycle behaviour.
However, as they use a single cross section, they do not control for cohort effects as Browning et al.
(1985) do. Deaton (1985) and, more recently, Moffitt (1993) have studied some of the econometric
problems connected with the use of average cohort techniques. Heck~manand Robb (1987), MaCurdy
and Mroz (1989) and Attanasio (1994) discuss identification issues.
Ch. 11:
Consumption
753
individuals as they age. Groups can be defined in different ways, as long as the
membership o f the group is constant over time 9. Within the life cycle framework,
the natural group to consider is a 'cohort', that is individuals (household heads) born
in the same period. Therefore, to compute the life cycle profile o f a given variable, say
log consumption, one splits the households interviewed in each individual cross section
in groups defined on the basis o f the household head's year o f birth. This involves, for
instance, considering all the individuals aged between 20 and 24 in 1980, those aged
between 21 and 25 in 1981 and so on to form the first cohort; those aged between 25
and 29 in 1980, between 26 and 30 in 1981 and so on to form the second cohort, etc.
Having formed these groups in each year in which the survey is available, one can
average log consumption and therefore form pseudo panels: the resulting data will
have dimension Q x T, where Q is the number o f groups (cohorts) formed and T is
the number o f time periods m. Even if the individuals used to compute the means in
each year are not the same, they belong to the same group (however defined) and
one can therefore study the dynamic behaviour o f the average variables. Notice that
non-linear transformations o f the variables do not constitute a problem as they can be
computed before averaging.
The resulting age profiles will not cover the entire life cycle o f a given cohort, unless
the available sample period is longer than any o f the micro data set commonly used.
Each cohort will be observed over a (different) portion o f its life cycle.
These techniques can be and have been used both for descriptive analysis and
for estimating structural models. Their big advantage is that they allow to study
the dynamic behaviour o f the variables o f interest even in the absence o f panel
data. Indeed, in many respects, their use might be superior to that o f panel data ~1.
Furthermore, as non-linear transformations o f the data can be handled directly when
forming the group means, they allow one to solve various aggregation problems that
plague the study o f structural models with aggregate time series data.
In what follows, I define groups on the basis o f the year o f birth and educational
attainment o f the household head. The length o f the interval that defines a birth
9 Group membership should be fixed over time so that the sample is drawn from the same population
and the sample mean is a consistent estimator of the mean of the same population. Attanasio and Hoynes
(1995) discuss the implications of differential mortality for the use of average cohort techniques. Other
possible problems arise, at the beginning of the life cycle, from the possible endogeneity of household
formation and, more generally, from migration.
~0 Here I am implicitly assuming that the pseudo panel is a balanced one. This is not always the case
as each group might be observed for a different number of time periods. Suppose, for instance, to have
data from 1968 to 1994. One might want to follow the cohort born between 1965 and 1970 only from
the late 1980s or the early 1990. On the other hand, at some point during the 1980s one might want to
drop the cohort born between 1906 and 1910.
l~ Time series of cross sections are probably less affected by non-random attrition than panel data.
Furthermore, in many situation, averaging across the individuals belonging to a group can eliminate
measurement error and purely idiosyncratic factors which are not necessarily of interest. As most
grouping techniques, average cohort analysis has an Instrumental Variable interpretation.
O.P. Attanasio
754
Table 4
Cohort definition and cell size
Cohort
Year of birth
Cell size
US
Average size
Years in sample
Averagesize
UK
Years in sample
1895-1999
338
1968-1977
1900-1904
459
1968-1982
1905-1909
526
1968-1987
1910-1914
232
1980-1992
560
1968-1992
1915-1919
390
1980-1992
519
1968-1992
1920-1924
333
1980-1992
653
1968-1992
1925-1929
325
1980-1992
572
1968-1992
1930-1934
317
1980-1992
546
1968-1992
1935-1939
345
1980-1992
562
1968-1992
10
1940-1944
420
1980-1992
594
1968-1992
11
1945-1949
566
1980-1992
652
1968-1992
12
1950-1954
657
1980-1992
547
1971-1992
13
1955-1959
734
1980-1992
508
1976-1992
14
1960-1964
463
1981-1992
15
1965 1969
334
1986-1992
cohort is chosen taking into account the trade-off between cell size and within-cell
homogeneity. Table 4 contains the definition o f the cohorts and the average sample
size for both surveys.
We start, in Figures 3 and 4, with the life cycle profile of (log) consumption and
disposable income at constant prices for both countries. The units o f measurement
for income and consumption are chosen so that the two graphs would be roughly
in the same scale, enabling to stress the differences in the shape of the age profile.
In the figures, I plot the average cohort (log) consumption at each point in time,
against the median age o f the household head. Each connected segment represent the
behaviour o f a cohort, observed as it ages, at different points in time. As each cohort
is defined by a five year interval, and both surveys cover a period longer than five
years, at most ages we observe more than one cohort, obviously in different years.
It might be tempting to attribute the differences between adjacent cohorts observed
at the same age, to 'cohort effects'. It should be remembered, however, that these
observations refer to different time periods and might therefore be reflecting business
cycle effects. The plotted profiles reflect age, time and cohort effects 12 that, without
755
UK - disposable income
25000
~'250
200
20000
150
15000
lOO
8 50
10000
20
40
Age
60
70
20
40
Age
60
70
60
70
Fig. 3.
USA - total consumption
25000
20000
UK - total consumption
250
~2o0
"~150
~o
&loo
15000
05
10000
0
20
40
Age
60
70
50
20
40
Age
Fig. 4.
an arbitrary normalisation or additional information from a structural model, cannot
be disentangled.
Several considerations are in order. First of all, both consumption and income age
profiles present a characteristic 'hump'. They both peak in the mid 40s and decline
afterwards. The picture seems, at first glance, to contradict the implications of the life
cycle model as stressed in the typical textbook picture which draws a 'hump shaped'
income profile and a flat consumption profile. For total disposable income, the decline
around retirement age is faster in the UK than in the USA, but approximately of the
same magnitude. This probably reflects the more synehronised retirement of British
individuals. The consumption profiles, however, present some strong differences. The
most notable is the fact that UK consumption declines much more at retirement than
US consumption. Total consumption at age 70 is roughly 35% of the peak in the UK
and above 50% in the USA. I discuss the decline of consumption at retirement below.
In the UK consumption profile, the consumption boom of the late 1980s, followed
by the bust of the early 1990s, is quite apparent. Notice, in particular, the fact that the
aggregate consumption boom is accounted for mainly by the youngest cohorts. I have
discussed elsewhere how to interpret that episode. It is worth stressing, however, that
the analysis of the cross sectional variability of consumption can be useful to shed
some light on the nature of episodes that the analysis of the time series data cannot
0.17. Attanasio
756
Table 5
Variability of consumption and income
Standard error (%)
Variable
Total consumption
Total consumption per adult equivalent
Non-durable consumption
Non-durable consumption per adult equivalent
Durable consumption
Non-durable consumption (from levels)
Income
USA
(CEX)
UK (FES);
age < 81
UK (FES);
10 cohorts,
year < 86
2.94
2.39
2.60
1.95
15.79
2.58
3.68
2.46
2.65
2.62
2.30
2.49
9.54
2.64
1.88
2.05
8.54
2.31
3.05
1.86
3.60
explain. Information about which groups in the populations where mainly responsible
for a determinate episode can be informative about alternative hypotheses 13
It is not obvious how to assess the time series volatility o f (log) consumption and
income. The main reason for this is that a large part o f the variation o f consumption
over the life cycle is very predictable and can be explained by age and cohort effects.
Furthermore, given the limited size o f our samples, the year to year variation in the
average cohort data reflects both genuine time series variation and the measurement
error induced by sample variation. As Deaton (1985) has stressed, some information
about the size o f the measurement error can be gained using the within-cell variability
o f the variables used. Using this information, one might correct for that part o f
variability accounted for by sampling variation and attempt to isolate the genuine time
variation. In an attempt to isolate this component, I run a regression o f log consumption
and income on a fifth order polynomial in age and cohort dummies and consider
the deviations o f the observed profiles from such a profile. The standard deviation
o f the changes in these deviations, corrected for that part which can be attributed to
sampling error, is my measure o f time variability 54. These estimates o f volatility for
(log) income and consumption are reported in Table 5 along with those for the other
variables considered. The first column refers to the USA, while the second and third
columns are computed using the U K data. The former includes the whole sample,
i3 See Attanasio and Weber (1994). Groups do not need to be formed on the basis of age. In Attanasio
and Banks (1997) that analysis is extended considering not only the variability across cohorts but also
across regions.
14 The sample mean is distributed around the population mean as a random variable with variance
given by o2/N, where N is the cell size and cr is the within-cell variance. The latter can be estimated
from the available micro data. These estimates can be used to correct our estimates of volatility.
Ch. 11:
Consumption
757
UK - non-durable consumption
per household and per adult equivalent
15000
200-
10000
100E
5000
20
40
Age
60
70
5020
40
Age
60
70
Fig. 5.
while the latter truncates it to 1986 to remove the effect of the consumption 'boom
and bust' of the last part of the sample.
As in the case of aggregate time series, total consumption appears less volatile than
disposable income, both in the UK and in the USA. In particular, the standard deviation
of changes in total disposable income at the cohort level is above 3% in both countries.
That of total consumption is between 0.6% and 0.95% less.
It may be argued that the differences in the consumption profiles for the two
countries are due to the differences in the definitions used in the two surveys. For this
reason, I next focus on a narrower definition of consumption which excludes a number
of items which might be recorded in different fashion in the two countries. In particular,
in Figure 5 I plot (log) expenditure on non-durables and services against age. This
definition excludes from total consumption durables, housing, health and education
expenditure. The other advantage of considering consumption of non-durables and
services, is that I avoid the issue of durability and the more complicated dynamics
that is linked to durables. The main features of the two profiles, however, including the
larger decline observed in the UK, are largely unaffected. In Table 5, the volatility of
non-durable consumption is considerably less than that of total consumption, especially
in the UK when data up to 1986 are used.
An important possible explanation for the life cycle variation of consumption over
the life cycle (and between the two countries considered), is the variation in needs
linked to changes in family size and composition. To control for this possibility,
I have deflated total household expenditure by the number of adult equivalents in the
household. For such a purpose, I use the OECD adult equivalence scale 15. The most
evident result is that the life cycle profile of consumption looks much flatter now. In
this sense, we can say that a large proportion of the variability of consumption over the
life cycle is accounted for by changes in needs. This result is perhaps not surprising
15 No adult equivalence scale is perfect. Different alternatives, however, do not make much difference
for the point I want to make here. The OECD scale gives weight 1 to the first adult, 0.67 to the following
adults and weight 0.43 to each child below 19.
O.P. Attanasio
758
USA - durables
UK - durables
5000-
4000-
looo-
30002000
1000-
~0
4o
6o
~0
Age
~0
40
6'o
8S
1;0
Age
Fig. 6.
if one considers that the life cycle profile of the number of adult equivalents (or of
family size) is also 'hump-shaped'. It may be argued that changes in needs are, to a
large extent, predictable. In terms o f the measure o f volatility in Table 5, it is greatly
reduced for the USA, while is slightly increased for the UK.
While the profile for non-durable consumption per adult equivalent is quite flat in
the first part of the life cycle, a marked decline is still noticeable in the last part.
It seems that the decline corresponds roughly to the time of retirement. In the UK,
where retirement is much more synchronised than in the USA, the decline is much
more rapid. The fact that per adult equivalent consumption declines with retirement
suggests that this might be due to a link between labour market status and consumption.
A possibility, for instance, is that some components of consumption are linked to
labour market participation. More generally, it is possible that consumption and leisure
are non-separable and, therefore, need to be analysed jointly. These issues have been
recently discussed by Banks et al. (1998).
Finally, it is of some interest to consider the life cycle profile of expenditure on
durables. The life cycle profiles for durables are plotted in Figure 6. Consistently with
the findings in aggregate time series data, the life cycle profiles for durable expenditure
are much more volatile than those for non-durables and services. The measure in
Table 5 for durables is 5 times as large as that o f total consumption for the USA
and almost 4 times as large for the U K 16.
Several variables are likely to be important determinants of, or determined jointly
with consumption. I have already stressed the important role which is likely to be
played by demographics and retirement behaviour in shaping the life cycle profiles
o f consumption. Similar considerations can be made for other labour supply variables
16 Because durable expenditure can be zero at the individual level I do not compute the average of
the log. Therefore, the deviations from the life cycle profiles are not percentage deviations, but are
measured in constant dollars. Because of this, in Table 5, in the row corresponding to durables I report
the coefficient of variation, rather than the standard deviation. For comparison, I adopt the same procedure
for non-durable consumption, in the following row.
Ch. 11:
Consumption
759
such as the participation rate of females to the labour market and the total number of
hours of work. A characterisation of the life cycle patterns of these variables and their
differences between the U K and the USA would go beyond the scope of this section 17.
However, it is important to stress that, as I argue in Section 5 and 6 below, one cannot
test any model of consumption without controlling for these factors, that, for the most
part, can only be analysed using household level data.
In Table 5, I only report the variability of the various components of consumption
and of disposable income. As with the aggregate time series data, it would be
interesting to characterise the autocorrelation properties of these variables and their
covariances. This analysis could be quite informative about the plausibility of
alternative structural models 18.
One of the implications of the textbook version of the life cycle model I discuss
in Section 3, is that consumption and current income should not be related. And yet,
comparing Figures 3 and 4, one cannot help noticing the similarity in the shape of
the two life cycle profiles. This similarity was interpreted as a failure of the life cycle
model by Thurow (1969) and reinterpreted in terms of non-separability of consumption
and leisure by Heckman (1974).
To pursue this issue, in Figure 7, I plot the life cycle profile of (log) disposable
income and non-durable consumption for four education groups in the USA defined
on the basis of the educational attainment of the household head: high school dropouts,
high school graduates, some college and college graduates. An interesting feature of
this figure is that the differences across groups in the shape of the income profiles
are mirrored in differences in the consumption profiles. In particular, notice that
both income and consumption profile of better educated individuals present a more
pronounced hump; not only are their income and consumption higher, but the profiles
are also much steeper in the first part of the life cycle. These differences where
interpreted within a life cycle model by Ghez and Becker (1975), but have interpreted
as a failure of the model by Carroll and Summers (1991) in an influential paper. An
interesting question, addressed below, is whether a version of the life cycle model
I discuss could generate these profile and account for the differences across education
groups.
In Table 6, I compute the variability of income, consumption and its components
as in Table 5, but splitting the sample by education. The most interesting feature of
this table is the fact that the only large difference in volatility among the groups is in
durable consumption. Expenditures on durables by high school dropouts is twice as
variable as that of college graduates, while the figure for high school graduates is in
the middle.
17 Interested readers can find the life cycle profiles for several variables in Attanasio (1994), Banks,
Blundell and Preston (1994) and Attanasio and Banks (1997).
a8 MaCurdy (1983) and Abowd and Card (1989) perform analyses of these kinds for earnings and hours
of work and use the results to assess the plausibility of different structural model. No similar analysis
exists for consumption and/or its components.
O.P. Attanasio
760
log i n c o m e a n d n o n d u r a b l e c o n s u m p t i o n
by e d u c a t i o n
high school graduates
40000l
30000
20000
-l
10000
3--~
5459.04
some college
college graduates
30000
20000
10000
5459,04
20
910
410
50
age
Fig. 7.
Table 6
Variability of consumption and income by educational group
Variable
Total consumption
Non-durable consumption
Durable consumption
Income
2.88
2.40
22.85
5.98
2.74
2.93
16.58
6.53
2.88
2.25
9.66
5.17
761
maximises utility over time. The various versions of the model will then differ for their
assumptions about optimisation horizon, uncertainty, curvature of the utility function,
assumptions about separability and so on. Which of these various versions is the most
relevant is in part a matter of taste and, above all, an empirical matter.
3.1. The simple textbook model
The main attractiveness of the life cycle-permanent income model, developed during
the 1950s in a number o f seminal contributions 19, is the fact that consumption
decisions are treated as part of an intertemporal allocation problem. The allocation
of consumption over time is treated in a fashion similar to the allocation of total
expenditure among different commodities in demand analysis. The model recognises,
therefore, that intertemporal prices and the total amount of resources available to an
individual are bound to be important determinants of consumption. This approach
immediately gives the study of consumption solid microfoundations and constitutes
a discontinuous jump with respect to the Keynesian consumption function which
assumed consumption to be a simple function of current disposable income.
The main difference between the life cycle and the permanent income model in
their original formulation lies in the time horizon considered. The life cycle model is,
almost by definition, a finite horizon model, while in the permanent income model the
horizon is infinite. In both cases, however, consumers decide how much to consume
keeping in mind their future prospects. If no uncertainty is introduced in the model, its
predictions are quite straightforward: concavity of the utility function implies a desire
to smooth consumption over time; the main motivation for saving is to smooth out
fluctuations in income; consumption increases with current income only if that increase
is a permanent one. In the case of the life cycle model, the explicit consideration of
retirement, that is a period in which income declines considerably, generates the main
motivation for saving: households accumulate wealth to provide for their consumption
during retirement.
An interesting implication of the life cycle model in its simplest incarnation is the
way in which aggregate saving is generated. It is quite obvious that in a stationary
life cycle economy with no growth aggregate saving is zero: the younger generations
will be accumulating wealth, while the older ones will be decumulating it. Aggregate
saving, however, can be generated in the presence of growth. I f the amount of resources
available over the life cycle to younger generations is larger than that available to
older ones, it is possible that the amount accumulated at a point in time exceeds the
amount that is decumulated. This introduces a relationship between aggregate saving
and growth that Modigliani has stressed in several studies. It should be stressed,
however, that such a relationship depends on a number of factors including the life
19 Modiglianiand Brumberg(1954) contains the first formulation of the life cycle model. The permanent
income model was sketched in Milton Friedman's 1957 volume [Friedman (1957)].
762
O.P. Attanasio
cycle profile, the way in which growth is generated and who benefits from it and so
on.
3.2. Quadratic preferences, certainty equivalence and the permanent income model
One of the problems with the life cycle-permanent income model is that the
dynamic problems that consumers are assumed to solve can be quite complex. As a
20 Carroll and Kimball (1996) provide quotes from Keynes' General Theory in which he suggested a
concave consumptionfimction.
21 Friedman(1957) essentially approximatedpermanent income with a distributed lag of current income.
Modigliani and Ando (1963) stressed the role played by wealth (in addition to disposable income) in
aggregate consumptionequations. Both the Modigliani and Ando paper and Friedman's book contained
interesting discussions of the aggregationproblems that were absent, for a long time, from subsequent
empirical studies.
22 This fact is still true: data from the Consumer Expenditure Surveyfrom 1980 to 1992 confirm that
the saving rates of household headed by a black are systematicallyhigher, for any interval of income,
than those of household headed by a non-black.
Ch. 11:
Consumption
763
(1)
where k= 1 if the (fixed) interest rate equals the subjective discount factor, and
permanent income YP is defined as
j=0 (1 +r)J
where r is the fixed interest rate, ;~=r/(1 +r), At is the value of current wealth, and
y is disposable labour income.
The main attraction of Equations (1) and (2) is that they provide a straightforward
relationship between the stochastic process that generates income and consumption.
These relationship give rise to a number of testable implications that have been studied
at length in the literature.
Flavin (1981) and Sargent (1978) were the first studies to exploit the fact that
Equations (1) and (2), together with the hypothesis that expectations about future
labour income are rational, imply cross equation restrictions on the bivariate VAR
representation of consumption and disposable income. Flavin (1981), in particular,
estimated such a system using US time series data and rejected the restrictions
implied by Equations (1) and (2). Flavin finds some evidence of excess sensitivity
of consumption to income.
Campbell (1987) proposes a slightly different interpretation of Flavin's results. From
Equations (1) and (2) it is possible to obtain the following expression for saving 23:
OO
st = )~ ~
j-0
EtYt+j -Yt
(1 + r)J
(3)
23 s on the left-hand side of Equation (9) coincideswith saving (i.e. income minus consumption),only
when k in Equation (7) is equal to 1. Otherwise,s = y - c/k.
764
0.17. Attanasio
The restrictions implied by Equation (3) are the same as those implied by Equations (1)
and (2). The nice thing about Equation (3), however, is its interpretation. The fact
that consumers smooth consumption over time is reflected in Equation (3) in the fact
that saving anticipates expected decline in disposable income. It is for this reason that
Equation (3) has been dubbed as the 'saving for a rainy day' equation. Formally, the
implication of Equation (3) can be written as saying that actual saving should equal the
best forecast of labour income declines. Consistently with Flavin's findings, Campbell
(1987) rejects the implications of the model. Campbell, however, finds that the time
series pattern of actual saving is not far from that implied by the model. He claims that
excess sensitivity of consumption to income within this framework "is more naturally
interpreted as insufficient variability of saving than as a correlation between changes
in consumption and lagged changes in income" (p. 1272).
Related to the tests of excess sensitivity discussed above, and using the same
framework, are those papers discussing the issue of 'excess smoothness' of consumption. Campbell and Deaton (1989) were the first to stress that, because Equations
(1) and (2) can be used to derive the relationship between changes in consumption and
innovations to the process generating income, the relationship between the volatility
of consumption (or permanent income) and that of current income depends on the
stochastic properties of the process generating the latter. In particular, if labour income
is difference stationary (rather than trend stationary), permanent income, and therefore
consumption, will be more volatile than current income. Intuitively, this result follows
the fact that if labour income is not stationary, current innovations are persistent
and will therefore imply a permanent revision to permanent income. Therefore the
observation that consumption growth is less volatile than current disposable income
growth contradicts the permanent income hypothesis 24. This result is ironic as one
of the original motivations for the development of the permanent income model
was, indeed, the observation that consumption is smoother than income. The most
problematic issue with this branch of the literature is the well known difficulty in
distinguishing between trend stationary and difference stationary models 25.
The version of the model with quadratic preferences has also been used to introduce
further refinements to the model. Goodfriend (1992) and Pischke (1995), for instance,
consider the implications of the lack of complete information on contemporaneous
aggregate variables. Pischke, in particular, explains the excess sensitivity results
typically obtained with aggregate data with this type of phenomena 26.
In a recent paper, Blundell and Preston (1998) use the assumption of quadratic
preferences to devise a clever way of decomposing transitory and permanent
components of income shocks. The idea is quite simple: under the permanent income
24 For a clear discussion of these issues see chapters 3 and 4 in Deaton (1992).
25 See, for instance, Christiano and Eichenbaum(1990)
26 Deaton(1992) also discusses the possibilitythat the informationset used by individual agents differs
from that available to the econometrician.
Ch. 11:
Consumption
765
max Et ~
j-0
(4)
N
subject to Z
i-1
766
o.P. Attanasio
(i)
Equation (4) assumes that utility is separable over time. This is a strong
assumption and rules out at least two important phenomena: habit formation
and durable commodities. The marginal utility o f consumption in any given time
period does not depend on consumption expenditure in any other period. We also
assume that expenditure coincides with consumption. One obvious possibility to
rationalise this model without excluding the existence o f durables is to assume that
the instantaneous utility function is additively separable in non-durables and in
the services provided by durables. In this case, a term for expenditure on durables
should be added to the intertemporal budget constraint.
(ii) Utility is derived from an homogeneous consumption good. The conditions under
which intertemporal choices can be summarised by a single price index are seldom
discussed. The two situations that are treatable are the absence o f changes in
relative prices, so that one can construct a Hicks composite commodity, or that
preferences take the Gorman polar form. These issues are discussed below.
(iii) Labour income is exogenous. No labour supply choices are considered. One
can reconcile a situation in which labour supply is endogenous with the model
discussed above, assuming that the instantaneous utility function is additively
separable in leisure and consumption. In this case, one should modify only the
budget constraint o f the problem (4) above.
(iv) The duration o f life is certain. This assumption is easily relaxed to assume an
uncertain life time. Davies (1981) has shown that this equivalent to assuming a
discount factor fi that varies with age as a consequence of a varying probability
o f survival. Utility at future ages is discounted not only because it accrues in
the future but also because its accrual is uncertain. I will not discuss this issue
any further, except when I discuss some o f the issues relevant for the analysis of
consumption based on numerical methods.
(v) The rate o f return on assets does not depend on the net position on that asset or
on the total level o f wealth held by the consumer. The model, however, can easily
accommodate a situation in which several assets are subject to various kinds of
constraints, as long as there is at least one asset in which is possible to borrow
and lend at the same rate 27
(vi) For simplicity, I have not considered explicitly the presence o f inflation. Obviously
the presence of (uncertain) absolute price changes is simply accommodated in the
model above by changing appropriately the budget constraint and the definition
o f interest rates.
Under these assumptions it is possible to derive an extremely useful first-order
condition for the intertemporal maximisation problem described above. I f we denote
27 This is the condition under which the first-order condition derived below holds. If the rate of return
on a given asset changes with the net position in that asset in a continuous and differentiable fashion, that
is if the intertemporal budget constraint is concave and does not present kinks, the first-order condition
derived below can be easily modified. More complicated is the situation in which there are discontinuities
and kinks for all assets at some level of net worth (for instance zero).
767
with )~t the multiplier associated to the intertemporal budget constraint at time t, it can
be shown that two of the first-order conditions for the problem in Equation (4) are
OU(Ct, zt, 03
OCt
(5)
-- i~t,
i= l,...,m.
(6)
Equation (6) holds for the m (m ~<N) assets for which is possible to borrow and lend
at the same rate and for which the consumer is not at a corner.
Equation (5) states that the marginal utility of consumption is equal to the marginal
utility of wealth at time t. The latter term summarises all the information about
the future. Equation (6) is the Euler equation that corresponds to the problem in
Equation (4) and states that the intertemporal maximisation problem (4) implies that
the discounted value of the marginal utility of wealth is kept constant over time. In
other words, the marginal utility of wealth is a martingale 28.
If we substitute Equation (5) into Equation (6) we obtain
Uct = E t [U~,~/3(1 +R~+l) ] ,
i = 1,...,m,
(7)
28 MaCurdy (1981) uses this frameworkto construct his 'A-constant' labour supply function.
0.17. Attanasio
768
The assumption of rational expectations and the fact that expectations in Equation (8)
are conditional to the information available at time t can be used to find valid
instruments to identify the structural parameters of Equation (8), which, once again,
can be considered for several rates of returns. If we denote with q the dimension
of the vector of parameters [in Equation (8), q = 2], with m the assets for which (8)
holds and with k the number of instruments considered, Equation (8) yields m k - q
overidentifying restrictions that can be used to test the model. The results obtained by
Hansen and Singleton on aggregate data indicate that the model is strongly rejected
whenever several returns are considered simultaneously. On the other hand, when one
asset is considered in isolation, the overidentifying restrictions are not violated, but
the preference parameters are estimated at somewhat implausible values.
Hansen and Singleton (1983) considered a log-linear version of Equation (8). If one
assumes that the rate of returns and consumption growth are joint log-normal, from
Equation (8) one can derive the following expression:
Alog(Ct+l) = 1 [log/3 + g2vart(Alog(Ct+l )) + vart (log(1 + Rit+l))
(9)
+2Cov(Alog(Ct+l), log(1 +R)+I))] +
where et+l is an expectational error uncorrelated with all the information available
at time t. The advantages of Equation (9) are that it is (log) linear and some of its
coefficients, as the one on the interest rate, have a natural and interesting interpretation.
When the utility function in Equation (4) is isoelastic, its curvature parameter y
plays a double rote. On the one hand it is equal to the coefficient of relative risk
aversion and therefore summarises consumer's attitude towards risk. On the other,
Ch. 11:
Consumption
769
(10)
where y is labour income and r an interest rate. Both o f the variables on the right hand
side are instrumented. According to the model the parameter )~ should be zero and the
residual term should be orthogonal to the information available at time t.
Equations (8) and (9) illustrate one of the main advantages o f the Euler equation
approach. Even if it is not possible to obtain a closed form solution for consumption, it
29 For a discussion of the interpretation of such a parameter and the links between intertemporal
substitution and risk aversion see Hall (1988), Attanasio and Weber (1989) and Epstein and Zin (1989,
1991).
3o Carroll et al. (1994) argue for the explicit consideration of the MA(1) process that would allow
the use of the orthogonality conditions with instruments dated t - 1 and therefore yield more precise
estimates and more powerful tests.
770
O.P.Attanasio
31 Alternatively, an equation similar to Equation (6) can be obtained by a Taylor expansion of the
marginal rate of substitution, as in Dynan (1993). Banks, Blundell and Brugiavini (1997) use an
approximation of the Euler equation developedby Blundell and Stoker (1999).
Ch. 11:
Consumption
771
motive for saving. As Browning and Lusardi (1996) stress, the fact that the effect that
the conditional variance has on the rate of growth of consumption is scaled by the
coefficient of relative risk aversion is an artefact of the CRRA utility function in which
a single parameter controls both risk aversion and prudence (as well as the elasticity of
intertemporal substitution). More generally, Carroll and Kimball (1996) have proved
the concavity of the consumption function under precautionary saving.
It should be noticed that the variance of consumption is not an exogenous variable.
While it is likely to depend on the overall uncertainty facing the consumer, it is
not obvious how it reacts to the arrival of new information or how it is related to
uncertainty in income and interest rates. To evaluate this relationship it would be
necessary to solve for the level of consumption and establish how consumption at
t+l reacts to changes in the economic environment which, under CRRA preferences,
is not possible.
Caballero (1990a,b, 1991) using the exponential utility function as a parametrization
of within-period utility obtains, with some additional assumptions, a closed form
solution for consumption. In an interesting example, he expresses consumption as a
function of permanent income (as in the case of certainty equivalence), minus a term
which summarises the effect of the precautionary motive. He then goes on to evaluate
the effect that precautionary saving is likely to have in reality.
While the results obtained with the exponential utility give useful insights about
the potential importance of the precautionary motive, such a parametrization of the
utility function is not exempt from criticism. It is therefore important to evaluate
the importance of precautionary saving using more general functional forms for
the utility function. Unfortunately, when one uses different utility functions is not
possible to obtain a closed form solution for consumption. It is therefore necessary
to use numerical methods to obtain solutions and additional assumptions about
the nature of the problem faced by consumers. This is the approach taken in a
number of studies, such as those of Skinner (1988), Zeldes (1989b), Deaton (1991),
Carroll (1994) and Hubbard, Skinner and Zeldes (1994, 1995). A disturbing feature
of these studies is that the quantitative importance of the precautionary motive
depends crucially on the properties of the distribution of the income process in the
left tail. Bounding away the income process from zero (or from arbitrarily small
realisations) greatly reduces the precautionary motive. While the reason for this
is clear (consumers will want to insure themselves against disastrous events), the
realism of such a mechanism is questionable. It might be worth investigating how the
precautionary motive would change in the presence of insurance mechanisms other
than self insurance (such as a safety net supplied either by society or by family and
relations).
Carroll (1994, 1997a) has strongly advocated the precautionary saving motive (or
'buffer stock saving') as an explanation of most empirical puzzles in consumption,
including the tracking of expected consumption and income. I discuss the empirical
findings on the precautionary motive and more generally on the Euler equation in
Section 5. From a theoretical point of view, however, it should be stressed that while the
772
O.P Attanasio
32 Individuals are asked to divide 100 probabilitypoints over several intervals of income growth.
Ch. 11:
ConsumpHon
773
occur. Therefore, excess sensitivity tests and, more generally, Euler equations are
not the best way to identify the presence o f liquidity constraints. Furthermore,
as we discuss below, there are several reasons why predicted income and consmnption might be related that have nothing to do with the presence of liquidity
constraints.
With these considerations, I do not want to dismiss the possibility of liquidity
constraints as unimportant or unrealistic 33. Especially for some groups in the
populations, they might be quite relevant and have an important effect on aggregate
consumption. Early contributions to the literature on the policy implications of the
permanent income-life cycle models, such as the papers by Flemming (1973) and
Tobin and Dolde (1971), were quite aware of the importance o f liquidity constraints.
It is crucial, however, that the presence o f liquidity constraints is incorporated in the
optimising framework sketched above.
The definition o f liquidity constraints I use in what follows appeal to partial
equilibrium considerations: the interest rate schedule is taken as given. General
equilibrium considerations, however, even though they are rarely made, can be quite
important. In a world o f identical consumers, the possibility o f smoothing consumption
over time will be constrained by the technology available to transfer resources over
time. In equilibrium, interest rates and asset prices will adjust depending on the
demand for borrowing (saving) and on the available technology. In such a situation,
nobody is liquidity constrained at the current interest rates. With heterogeneous
consumers, it is possible that some will want to borrow and others will be saving.
The equilibrium interest rates will then reflect these factors. While the focus o f most
o f the considerations in this section is on the partial equilibrium effects, it is worth
keeping in mind that aggregate fluctuations, for instance a recession, might have effects
on asset prices that reduce the demand for loans relative to what would be observed
under constant interest rates.
The first step in the integration of borrowing restrictions in the model above must
be their exact definition. There are several possibilities. The first and most general
alternative is to allow the interest rate paid on assets to depend on the net asset position.
As a simple example o f this alternative, consider the possibility o f a difference between
borrowing and lending rate. Such a wedge induces a kink in the intertemporal budget
constraint. One can then consider Euler equations for individuals who are net borrower
and net savers: the interest rate relevant for the two groups will be different, but the
Euler equation still holds as an equality. For the individuals that will cluster at zero net
assets, however, the Euler equation (with either interest rate) holds as an inequality.
33 As Hayashi (1996) stresses, most specifications of preference implicitly imply a form of borrowing
restrictions. If the marginal utility of consumption is infinite a zero consumption, consumers will not
want to borrow more than the present discounted value of the minimum realisation of income, even
though the probability of this event is very low. This is because they want to avoid the possibility of
zero consumption even with a very small probability.
O.P. Attanasio
774
It is also possible that the interest rate varies continuously with the quantity borrowed
or saved. I f that is the case, at any point in which the function is differentiable, it is
still possible to write down an Euler equation which, relative to Equation (8), contains
an additional term referring to the derivative of the interest rate with respect to the
asset position 34:
OAf ,'/3
(l +
Et
= 1
(al)
At those points of the intertemporal budget constraint where the interest rate changes
discontinuously (such as zero), Equation (11) will be replaced by an inequality.
An alternative to the consideration of interest rates varying with the amount
borrowed (or saved) is the assumption that individuals face a limit to the amount they
can borrow (which can be zero). Obviously this can be interpreted as a case o f the
previous situation, with the borrowing rate being infinite at the limit. Even in this
case, however, there are several alternatives. It is possible, for instance, that the limit
an individual can borrow is not fixed but a function of some variables which, in turn,
can be endogenous 35. Finally, it is possible to consider the existence of collateralizable
loans.
When liquidity constraints take the form of a limit to borrowing, it is still possible
to write the Euler equation for consumption, as long as the constraint is not binding.
When it is binding, instead, the Euler equation will hold as an inequality, or as an
equality with the addition of a slack variable (a Kuhn-Tucker multiplier). Equation (8)
becomes
E,
[ C'+Yl(1
= 1 +~tt,
(12)
where At is the marginal utility of wealth and Pl is the relative price of commodity i.
Notice that gt appears in the first-order conditions for all commodities, so that any
Ch. 11:
Consumption
775
two of these equations can be used to eliminate it. This implies that the intratemporal
first-order conditions hold regardless of the presence of liquidity. Meghir and Weber
(1996) used this intuition to distinguish between liquidity constraints and intertemporal
dependence in preferences. They consider three different non-durable commodities and
stress that the presence of dynamic effects in the Euler equation can be rationalised
with intertemporal non-separabilities only if the same dynamic effects are found in the
intratemporal first-order conditions. On the contrary, if one finds that the intratemporal
conditions do not show any sign of dynamic effects while these appear in the Euler
equations, one should interpret this evidence as a sign of binding liquidity constraints.
Some studies have explored the possibility that some commodities, such as durables,
can be used as collateral to relax the severity of borrowing restrictions. Because
durables can be used as collateral, when liquidity constraints are binding, they become
relatively more attractive than non-durables. This fact has implications for the withinperiod allocation of resources between durables and non-durable consumption. In the
absence of liquidity constraints, this would depend only on the relative price and
on their marginal rate of substitution (where the relevant price for durables would
be their user cost). In the presence of liquidity constraints, however, an additional
term has to be added to the first-order condition to reflect the fact that it is possible
to use durables to borrow against future resources. Therefore, as Chah, Ramey and
Starr (1995), Brugiavini and Weber (1994) and Alessie, Devereux and Weber (1997)
have noted, the possibility of using durables as collateral and the presence of liquidity
constraints distorts the intratemporal allocation of resources between durables and nondurables.
At several points in the discussion so far ! have stressed that even when liquidity
constraints are present, if they are not binding, the Euler Equation (8) will hold.
Therefore, such an equation and the mis-specification tests conducted on it (such as
tests of 'excess sensitivity') are likely to be a poor tool to identify the presence of
borrowing restrictions. In addition to the power considerations just made, in what
follows I also stress that evidence on excess sensitivity can often be interpreted as
evidence of non-separability between consumption and leisure.
The fact that the Euler Equation (8) holds whenever the borrowing restrictions
are not binding does not mean that these constraints have no effect on the level of
consumption. Indeed, as discussed clearly by Hayashi (1987), the presence of potential
liquidity constraints has the same effect as that of a shortening of the horizon relevant
for current choices. Alternatively, one can interpret the presence of liquidity constraints
(when they are not binding) as an increase in the discount factor.
From a policy perspective, liquidity constraints are important because of their
effect on the level of consumption, rather than on its changes. In other words, what
matters is how consumption reacts to unexpected changes in the economic environment
(including policy changes). Euler equations are not informative about this.
Deaton (1991) provides one of the first analysis of the effect of liquidity constraints
on the level of consumption. By solving the Euler equation numerically, he shows
that the behaviour induced by the presence of liquidity constraints is similar to that
776
O.R Attanasio
equation 36.
From a theoretical point of view, the considerations above indicate that a profitable
research strategy is one which aims at characterising the response of consumption to
various news when liquidity constraints are relevant (regardless of whether they are
binding). Deaton (1991) constitutes a first important step in this direction. On a more
specific level, the analysis in Hubbard, Skinner and Zeldes (1994, 1995) constitutes
another good example of how a consistent theoretical model incorporating borrowing
restrictions can be used and be informative about important policy issues 37
Several other interesting problems remain to be modelled and understood. If labour
supply choices are endogenous, the presence of liquidity constraints might induce
female labour force participation. These effects might be strengthened if the ability
to borrow is linked to earnings 38. All these issues are examples of the need to be able
to use the model to make statements about consumption levels that I discuss again
below.
Identifying the presence and the relevance of borrowing restrictions is not easy. One
possibility is to use direct questions on the matter. Jappelli (1990) used this strategy and
analysed the answers to some questions contained in the Survey of Consumer Finances
in the USA. The households interviewed in 1983 were asked whether they were denied
credit or whether they did not applied because they felt they would be denied. The
main problem with this research strategy is that very few household surveys contain
questions similar to those analysed by Jappelli.
777
The presence of (binding) liquidity constraints has some direct implications for
the demand for loans. Its analysis constitute therefore an interesting possibility for
the identification of borrowing restriction and for evaluating their importance. If
consumers are subject to binding borrowing restrictions they will be at a kink (or in a
relatively steep section) of an intertemporal budget constraint. Their demand for loans,
therefore, will not be much affected by changes in the interest rate. On the contrary, if
consulners are not at a corner (or in a relatively fiat part of the budget constraint) their
demand for loans will be elastic to the interest rate. The effects of maturity, however,
should be opposite. A consumer who is not affected by borrowing restrictions will
be indifferent to changes in maturity. On the other hand, a constrained consumer, for
whom an increase in maturity will effectively increase his ability to borrow by reducing
the size of the payments to maturity, will increase its demand for loan. Juster and
Shay (1964) were the first to use this intuition using semi-experimental data. They
asked a sample of consumers whether they would finance the hypothetical purchase
of an automobile when faced with different packages of interest rates and maturity.
The packages of interest rate and maturity were randomised to the individuals in the
sample so to enable the estimation of the interest rate and maturity elasticity of the
demand for loans. In a recent paper, Attanasio and Goldberg (1997) develop the work
of Attanasio (1995b) and perform a similar exercise but on a sample of households who
actually purchased (and occasionally financed) automobiles. The main advantage of the
Attanasio and Goldberg exercise is that their data refer to actual choices. The main
disadvantage relative to the work of Juster and Shay (1964) is that for the households
that decided not to finance their car purchases neither interest rates nor maturity are
observed. This poses a number of econometric problems that are similar to those that
labour economists deal with when analysing participation choices and labour supply.
o.P. Attanasio
778
Given these considerations, it is important, especially if one wants to bring the model
to the data, to consider a specification that allows for these factors. While it is true, as
shown in Section 2, that consumption life cycle profiles often mirror income profiles, it
is not obvious that one should interpret this as evidence of the empirical failure of the
model without considering explicitly the possibility of important demographic effects
and that consumption and leisure are not separable. These are not new arguments and
I discuss them at length in Section 5, where I interpret the available evidence. The
point of this section, however, is to stress that the consideration of these factor is not
only important, but also relatively simple. If one does not want to model explicitly
these variables, controlling for them does not jeopardise the empirical tractability of
the model illustrated by Equations (8) and (9).
I re-write a slight modification of Equation (4) for convenience:
T-t
(4')
where the superscript nd indicates that I am now modelling explicitly only non-durable
consumption. As before, the vector of observable variables z indicates variables that are
relevant for the intertemporal optimisation problem. The variable u is unobservable to
the econometrician. The latter variable is sometimes referred to as 'taste shift'. More
generally, it represents all the unobservable factors that affect consumption choices
and that we do not model or control for. The variables included in the vector z may
range from demographic variables, to labour supply variable, to other components of
consumption (such as services from durables). Notice that these variables may be either
exogenous to the choice problem (age is a good example), or chosen simultaneously
with non-durable consumption (such as labour supply). Even if one does not wish
to model these latter variables explicitly, it is possible to identify, using Equation (4I),
conditional preferences and the associated parameters under very mild conditions 39.
If the z variables in Equation (4 ~) enter the utility function in an additive separable
fashion there is no need to consider them. If, on the other hand, they affect the marginal
utility of non-durable consumption, they will enter the corresponding Euler equation.
As an example, let us consider the following parametrization:
U ( C t d, zt, ot) -
(cpd)l-7 exp(O'zt +
1-y
ut).
(13)
A possible interpretation of Equation (13) is that the discount factor varies with
variables z and v. Such an interpretation is particularly attractive for demographic
variables. In this case, the vector of parameters 0 implicitly represents an equivalence
scale.
39 See the discussion in Browningand Meghir (1991) on demand systemsconditionalon labour supply.
Ch. 11:
779
Consumption
Under the parametrization used in Equation (13), the Euler equation for non-durable
consumption will be
E, [ \ ~ - j
[(,q,+a)
(1 + R ~ + , ) / 3 e x p [ O ' ( z t + l - z , ) +
v t + x - vt]
= 1.
(14)
It should be stressed that Equation (14) holds even if some of the variables included in z
are endogenous choice variable, regardless of the way in which they are determined.
Modelling labour supply or durable consumption, for example, can be quite difficult
because of corner solutions, intertemporal (non)-separability and transaction costs.
These problems do not need to be tackled explicitly and Equation (14) can be
considered as an equilibrium relationship that holds at the optimal values of the
endogenous z, regardless of how these are obtained.
Equation (14) can be linearised in the same fashion as Equation (8) to obtain a
log-linear expression similar to Equation (9). Neglecting the second moments, that
I incorporate into the constant of the equation, we have:
Alog(C~dl) = const. + oR~+1 + OtAZt+ 1 + Avt+l + et+l.
(15)
Notice that because of the presence of the term representing unobserved heterogeneity
and taste shocks, the residual of equation has now two components, one which
represents an expectational error and, by the assumption of rational expectations, is
likely to be uncorrelated over time, and another that has an MA(1) structure if taste
shocks are i.i.d, over time.
If the unobserved heterogeneity term v has a time invariant component which is
individual specific, it will be eliminated in the first differences. On the other hand, it
is possible that such a term is persistent (rather than white noise) but not fixed and/or
that the 'pure' discount factor/3 is individual specific. In the former case the term Av
in Equation (15) would have a structure more complex than a simple MA(1), while
in the latter, Equation (15) would have a fixed effect. In Section 5, I discuss these
econometric problems.
The particular parametrization considered in Equation (13) is just an example. More
complex structures may be and have been considered. It is possible, for instance,
to allow some of the variables in the vector z to affect the curvature of the utility
function as well as the rate at which utility is discounted. Blundell, Browning and
Meghir (1994) and Attanasio and Browning (1995) have used this approach and found
significant deviations of estimated differences from the simple isoelastic case. The
issue is particularly important in that preferences of this kind can allow for systematic
differences in the elasticity of intertemporal substitution across consumers.
Attanasio and Browning (1995) also stress the fact that to estimate an Euler equation
and its parameters, it is not necessary to specify explicitly the within-period utility
function. It is possible and analytically convenient to start from a flexible specification
780
0.17. Attanasio
for the marginal utility of consumption and, if needed, obtain the corresponding utility
function by integration 4.
According to the simplest version of the life cycle model sketched at the beginning of
the section, the main motivation for individual saving is to provide resources during
the last part of the life cycle, when, following retirement, income is low. In such
a situation aggregate wealth can be generated by and associated with productivity
growth, if the generations that save (the young) are relatively wealthier than those
that dis-save (the old). I f the bequest motive is operative, on the other hand, the
mechanism through which aggregate wealth is accumulated is quite different. A lively
exchange on whether a large or small proportion of aggregate wealth in the USA is
accounted for by bequeathed wealth or retirement saving wealth developed in the 1980s
between Kotlikoff and Summers [Kotlikoff and Summers (1981), Kotlikoff (1988)]
on one side and Modigliani (1988) on the other. The result of that debate remained
somewhat ambiguous, as the answer seems to depend mainly on whether one considers
the interests earned on bequeathed wealth as originating from bequests or not.
A bequest motive can be simply added to the basic model (1) by considering a term
which is a function of the bequest left. As it is obvious such a term does not affect
the Euler equation for consumption in subsequent periods. It will, however, affect the
level of consumption (and saving).
One of the implications of the simplest version of the life cycle model is that
wealth is decumulated in the last part of the life cycle. While the rate at which
wealth is decumulated depends on the parameters of the model and in particular about
beliefs about longevity, the result that wealth should decline seems to be robust. The
evidence on this point is mixed, in that several studies do not find strong evidence of
decumulation of wealth by the elderly 41.
When bequests motives are operative, from a theoretical point of view, the wealth
age profile can take different shapes in the last part of the life cycle. In an important
paper Hurd (1989) has characterised several of these profiles. Hurd also showed that
for several realistic sets of parameters, the wealth age profile under bequests is also
declining in the last part of the life cycle. From an empirical point of view, Hurd (1989)
stresses the importance of conditioning on the labour force status of the individuals
in the sample. In particular, whether individuals are retired or not seems to be crucial
for the decumulation of their wealth.
40 If one follows such a procedure, one should check the integrability conditions.
41 JappeUiand Modigliani (1997) have forcefully argued that pension benefits should be considered as
decumulation of pension wealth.
781
4. Aggregation issues
The models considered in Section 3 refer to the dynamic optimisation problem faced
by an individual consumer (or household). The early contributors to the life cyclepermanent income model were quite aware of the aggregation issues involved with
the empirical implementation of the theory of intertemporal optimisation. These issues,
however, were largely ignored by the literature of the late seventies and early eighties
which focused mainly on the rigorous introduction of uncertainty in the model.
Aggregation problems, however, cannot be ignored when the model is tested and when
is estimated to evaluate structural parameters. In this section, I consider briefly two
problems: the aggregation across consumers and that across commodities.
4.1. Aggregation across consumers
The Euler equations for (non-durable) consumption that can be derived from the
problem in Equation (1) are, for most specifications of preferences, non-linear. As they
refer to individual households, their aggregation is problematic. A number of papers
have shown that the dynamics of aggregate consumption implied by the Euler equation
for individual consumption cannot be described simply by the first moments of the
cross-sectional distribution of consumption. Attanasio and Weber (1993), Blnndell,
Pashardes and Weber (1993b) and Blundell, Browning and Meghir (1994) have shown
that higher order moments can play an important role. Attanasio and Weber (1993), in
particular, have stressed that the use of aggregate data to estimate an Euler equation is
equivalent to omit high moments of the cross sectional distribution of consumption and
might cause systematic biases in the estimation of structural parameters and lead to
rejections of the over-identifying restrictions 42. These contributions will be discussed
when I evaluate the empirical evidence, at this point I only wanted to stress that
exact aggregation is in general impossible when considering the Euler equation for
consumption. This implies that aggregate data cannot be used to estimate structural
preference parameters and/or to test the model.
Individual panel data on consumption might be very difficult to obtain. Partly
because of this, many of the papers I discuss in Section 5 use the average cohort data
that I have used in Section 2 to describe the individual data. The use of grouped data
allows one to use time series of repeated cross section to study dynamic models. In
this sense, average (or synthetic) cohort data are particularly useful to study life cycle
models. It is important to stress, however, that synthetic cohort data are aggregate
data; the difference relative to National Accounts data is that the aggregation process
is controlled directly. As a cohort ages, the researcher can follow the evolution of
the variables of interest, that range from consumption to income, to family size and
42 Blundell, Pashardes and Weber (1993b) make a similar point for a demand system. Similar results
were obtained in an asset pricing frameworkby Constantinides and Duffle (1996).
782
O.P. Attanasio
Another important aspect which is often neglected in the literature on the Euler
equation is the aggregation of expenditure on different commodities. In principle, one
should consider simultaneously the allocation of total expenditure across time periods
and the allocation within each period across different commodities. I f one had detailed
enough data on expenditure on individual commodities and was willing to specify the
form of the direct utility function, the problem could be addressed in a straightforward
manner. One could consider the Euler equation defined in terms of the marginal utility
of each individual commodity.
The issue, however, is to determine under what conditions one can consider the
within-period utility function defined in terms of total consumption and assume that
the allocation of expenditure over time can be determined with the help of a single
price index. One obvious and not particularly interesting answer is when relative
prices do not change so that it is possible to construct a Hicks composite commodity.
Gorman (1959) was the first to provide a more interesting answer in that he derived
43 Estimation of the effects of demographic variables can be interpreted as the estimation of adult
equivalent scales and is therefore important for a variety of reasons.
783
the conditions that the utility function has to satisfy so that it is possible to consider a
single commodity 44. In particular, it is necessary that preferences take the generalised
Gorman Polar form, that is, i f X is the K x 1 vector of commodities andp the vector
of corresponding prices, the indirect utility function has to take the form
(16)
where b(p) and a(p) are functions homogeneous of degree 1 and degree 0 respectively
and depend on within-period preferences.
From the specification in Equation (16) one can derive an Euler equation for
total consumption expenditure: two-stage budgeting can be used to separate the
intertemporal from the within-period allocation. To implement Equation (16) it is
necessary, however, to estimate the price functions a(p) and b(p), which requires
the specification and estimation of a within-period demand system. In an important
paper, Blundell, Browning and Meghir (1994) estimate a demand system and use
the resulting price indexes to estimate the Euler equation for total consumption
expenditure. A remarkable result they obtain is that a Stone price index (which does not
require the knowledge of preference parameters) constitutes a good approximation to
the price indexes in Equation (16). This result is important because points to a simple
way to implement Euler equations for non-durable consumption without estimating the
entire demand system. The results in Blundell, Browning and Meghir are generalised to
a demand system with quadratic Engel curves by Banks, Blundell and Preston (1994).
Attanasio and Weber (1995) with a different parametrization of the indirect utility
function find a statistically significant role for the zero-homogeneous price index a(p)
on US data.
If one is willing to specify a utility function defined over several commodities, one
can derive an Euler equation for each of these commodities. Under the assumption
of additive separability, each of these Euler equation depends only on expenditure on
that commodity, nominal interest rates, changes in the commodity-specific price index
and on the relevant controls. On the other hand, when additive separability does not
hold, the individual commodity Euler equation depends also on the consumption of
other commodities. This is an important point in that creates problems for the use of
data sources which contain information only on some components of total expenditure,
such as the US PSID which gives only information on food consumption. Attanasio
and Weber (1995) show that the consideration of food consumption in isolation can
yield very misleading results.
5. Econometric issues and empirical evidence
I now move to consider the empirical evidence on the life cycle-permanent income
model. In the process of doing so, I also discuss some of the econometric problems
44 See Deaton and Muellbauer (1980) for a clear discussion of these issues.
784
O.P. Attanasio
45 Because they worry about time aggregation, Campbell and Mankiw use instruments dated t - 2 and
earlier.
785
More recently, in the spirit of the Campbell and Mankiw studies, several papers have
studied the time series properties of aggregate consumption and have tried to interpret
its relation to a number of other variables in terms of the life cycle-permanent income
model or as alternative deviations from it. Carroll et al. (1994), for instance, relate
consumption to the index of consumer confidence and provide some interpretation of
the high predictive power that such a variable has. Ludvigson (1997), on the other
hand, relates consumption to consumer credit and finds that 'excess sensitivity' of
consumption to such a variable is even more marked than that to labour income.
In the past, some of the papers that used aggregate time series data have also tried to
incorporate in the model the possibility that consumption and leisure are not separable
in the utility function. Bean (1986), Eichenbaum and Hansen (1988) and Mankiw,
Rotemberg and Summers (1985) are examples of these attempts. The presence of
corner solutions in labour supply, such as those observed in the case a spouse does
not participate to the labour force or in the case of retirement, make the aggregation
issues even more complicated.
(17)
786
o.P. Attanasio
46 A problem similar to the one just discussed is the issue of the presence of individual fixed effects in
the Euler equation for consumption. These could arise, for instance, if the discount factors contain an
individual specific component. In this case, the use of weakly exogenous instrument would be invalid.
For a discussion of the issues related to this problem see Keane and Runkle (1992) and the comments
to that paper in the Journal of Business and Economic Statistics.
787
individual specific. In the former case the term Ao in Equation (9 ~) would have a
structure more complex than a simple MA(1), while in the latter, Equation (9 ~) would
have a fixed effect. In such a situation the choice of instruments is not trivial, in that
lagged individual variables are, in principle, correlated with the residuals of the Euler
equation.
The problem might be less severe if one uses average cohort data: grouping the
individuals belonging to a given cohort averages out the individual fixed effects and
leaves only the cohort specific ones. Here the availability of a long time period is once
again crucial: it is possible to estimate cohort specific fixed effects (say in discount
factors) by introducing cohort dummies in the Euler equation.
The necessity of a 'large T' to obtain consistent estimates of the parameters of an
Euler equation follows from the nature of the residuals of such an equation which
incorporate expectational errors. The same argument does not apply to the residuals
of intratemporal (within-period) first-order conditions. Indeed, if expectational errors
are the only component of the residuals, such equations should have a perfect fit. It
is for this reason that the consideration of unobserved heterogeneity is crucial in the
specification of preferences (see Equation 13 above).
788
O.P. Attanasio
The structure of the errors is even more complicated when one considers more
cohorts simultaneously. In this case, one has to keep into account the possibility
that the residuals of different cohorts in the same time period are correlated. The
variance covariance matrix of the residuals is therefore quite complex. Its computation
is, however, important to make correct inferences 49.
Besides adjusting the estimated standard errors for the presence of an MA(1) error
and the correlation among the expectationat errors of different cohorts, it is also
possible to use a GLS type of scheme to improve the efficiency of the estimator. One
should be careful, however, in filtering the data so to avoid the inconsistency caused
by the correlation of lagged expectational errors with the instruments 5o.
5.2.3. Conditional second (and higher) moments
49 A technical problem arises from the fact that is not easy to guarantee that the estimated variance
covariance matrix is positive definite in finite samples.
50 See Hayashi and Sims (1983).
Ch. 11:
Consumption
789
second moments are significant in the Euler equation; and second, the estimate of some
important structural parameters, such as the elasticity of intertemporal substitution, is
not much affected by the introduction of the conditional second moments.
Yet another alternative to the log-linearization procedure is the one proposed in
Attanasio and Browning (1995) who start with a flexible functional form for the
(log) marginal utility of consumption so to avoid the necessity of approximating an
Euler equation. Should one be interested in the utility function that generates such a
function, one can obtain it by integration. Not only does this method avoid dealing with
the issue of linearization, but gives the possibility of estimating much more flexible
functional forms than the isoelastic one. This is not to say that precautionary motives
are unimportant, but that their relevance is going to depend on the curvature of the
marginal utility of consumption and is, in the end, an empirical matter.
The alternative of estimating a non-linearised Euler equation (such as Equation 8)
is particularly unappealing because it would imply assuming away measurement error
and would, in any case, make the use of average cohort techniques much harder, if
not impossible.
5.3. Micro data: some evidence
In this section, I do not discuss in detail all the papers that have estimated and tested
Euler equations for consumption using micro data. Rather, I summarise some of the
main contributions with an eye to the overall evaluation of the evidence which I give
in the next sub-section.
One of the first papers to consider the implications of the permanent incomelife cycle models with micro data is Hall and Mishkin (1982) in which the authors
used PSID data to test and reject the implication that consumption changes were
uncorrelated with lagged values of current income. Hall and Mishkin (1982) evidence
was later criticised by Altonji and Siow (1987) because it did not allow for
measurement error.
Hall and Mishkin (1982) focused on the permanent income hypothesis and did not
consider explicitly the Euler equation that one can get from the consumer intertemporal
optimisation problem. One of the first, and probably the most influential article to take
such an equation to the data, was the paper by Zeldes (1989a), who implemented
ideas very similar to those in Runkle (1991). In both articles, the Euler equation
is fitted to observations on food consumption from the PSID. In both articles, the
authors consider explicitly the possibility of liquidity constraints and the possibility
that these constraints affect different group in the population differently. Effectively,
Zeldes (1989a) and Runkle (1991) estimated versions of Equation (12) for different
groups of the populations that had, on the basis of an observable variable, different
probabilities of having/tt = 0.
The results they get, however, are quite different. Zeldes (1989a), in particular,
splits the sample according to the wealth held and finds that the rate of growth of
consumption is related with the lagged level of income for the low wealth sample. The
790
O.P. Attanasio
same result does not hold for the high wealth sample. Zeldes interprets this results as
evidence of binding liquidity constraints for a large fraction of the population. Runkle
(1991), who uses a different extract of the PSID, and different econometric techniques,
obtains different results 51.
Since Zeldes (1989a) and Runkle (1991), many papers, including Hayashi (1985a,b),
Garcia, Lusardi and Ng (1997), Lusardi (1996), Mankiw and Zeldes (1994) and Shea
(1995), have used PSID data, either alone or in conjunction with other data, to estimate
and test various versions of the model. Shea (1995), in particular, identifies a subsample
of PSID households for whom he can track their union wage contracts and therefore
construct a good measure of their expected wage growth. As many others, Shea (1995)
tests (and marginally rejects) the hypothesis that consumption growth is not related to
wage growth. Shea (1995), however, also notices that liquidity constraints imply an
asymmetry between households who expect a decline and a rise in wages, because the
former should be saving rather than borrowing. As he fails to identify these effects,
his results cast doubts about the plausibility of liquidity constraints as an explanation
of the excess sensitivity of consumption growth to income or wage growth.
One of the main problems with the PSID is that the measure of consumption that
is included in the survey refers only to food consumption. I have stressed above the
theoretical problems with such a measure. Attanasio and Weber (1995) show how the
use of such a measure of consumption can lead to misleading results. In that paper,
my co-author and I use average cohort data constructed from the CEX for the period
1980-1992 to show that some of the excess sensitivity results obtained on micro data
can be accounted for by the non-separability of food and other consumption.
The other two issues on which Attanasio and Weber (1995) focuses are the effects
of aggregation over consumers and over commodities. As far as the aggregation over
commodities is concerned, Attanasio and Weber (1995) find significant effects of the
zero-homogeneous price index in Equation (14) that, however, are not quantitatively
important. This evidence is consistent with that for the UK FES reported in Blundell,
Browning and Meghir (1994) and Banks, Blundell and Preston (1994).
About the aggregation over individuals, Attanasio and Weber (1995) report evidence
which shows that the effect of the non-linearities in the Euler equation can be quite
important. This is consistent with the evidence reported in Attanasio and Weber
(1993) on FES data. In the latter paper, my co-author and I show that aggregating
the individual data so to obtain an aggregate conceptually similar to the National
Accounts statistics (i.e. taking the log of the arithmetic mean of consumption) one
obtains results that are quite similar to those obtained with aggregate times series
data. In particular, it is possible to obtain excess sensitivity to predicted income, a low
estimated elasticity of intertemporal substitution and rejection of the overidentifying
51 Keane and Rtmkle (1992) have recently re-estimated Zeldes' equations on his data but using
different econometrictechniques. Jappelli, Pischke and Souleles (1997) match SCF data, which contain
information on self reported 'liquidity constrained' status with PSID data.
Ch. 11:
Consumption
791
restrictions. The results obtained aggregating the data in the proper way (i.e. taking the
average of log consumption), however, are much more consistent with the theoretical
model. In particular, there is no rejection of the overidentifying restrictions. The
difference between the log of the average and the average of the log is a measure
of inequality which varies over the business cycle and is likely to be correlated with
the instruments used in estimating the Euler equation.
The last group of papers I have cited [Attanasio and Weber (1993, 1995), Blundell,
Browning and Meghir (1994), Banks, Blundell and Preston (1994), Attanasio and
Browning (1995)] all stress the importance of controlling for demographic factors
and for labour supply effects. These papers, which use average cohort data for the
UK and the USA (the data presented in Section 2) show that once one controls for
the influence that demographics and labour supply might have on the marginal utility
of consumption, there is no evidence of excess sensitivity of consumption to income
or rejection o f the overidentifying restrictions. Female labour force participation and
family size seem to be particularly important in this respect 52.
If one believes that the estimation of the Euler equations discussed in this section
yields consistent estimates, the parameter on the real interest rate can be interpreted
as the elasticity of intertemporal substitution. Such a parameter is of some importance
for a number of policy issues and its size has been discussed at length. Hall (1988) in
particular, claims that the use of the correct instruments delivers very low estimates of
this elasticity. However, the evidence that emerges from the micro studies which use
an isoelastic specification of preferences [such as Attanasio and Weber (1993, 1995),
and Blundell, Browning and Meghir (1994)], is that, both in the UK and in the USA,
the elasticity of intertemporal substitution of consumption (EIS) is just below 153.
From the discussion above, it is clear that in addition to the EIS, a number of other
parameters, measuring the effect of leisure, that of demographics or other variables
affecting the marginal utility of consumption are likely to be extremely important
in determining the level of consumption. Their interpretation, however, is not easy
without a solution for the level of consumption.
It is now time to take stock on the empirical relevance of the models discussed so far.
My reading of the evidence briefly presented in Section 5 is that the life cycle model,
enriched to account for the effect of demographic and labour supply variables, is not
rejected by the available data. Alternatively, and perhaps more accurately, one could
52 Demographicvariablesmight be capturing the effect of the conditional second (and higher) moments
ignored in the log linearization procedure. Lusardi (1996) criticises the use of income data in the CEX
as they might be affected by measurement error.
53 As mentioned above, the use of aggregate rather than individual data and aggregation problems can
explain part of the differences in the size of the estimatesof the elasticity of intertemporal substitution.
792
0.t?. Attanasio
say that the life cycle model can be made complex enough not to be inconsistent with
the available data. This view is not widely accepted in the profession so that it deserves
some discussion.
Before discussing these issues, however, a number of caveats are in order. First, the
model has been fitted with success only to households in the middle of their life cycle.
While these account for a large fraction of aggregate consumption, additional work is
needed to understand the behaviour of young and elderly households.
The behaviour of retirees in particular, can be quite difficult to model. I f leisure
and consumption are non-separable in the utility- function, a radical change in labour
supply could be linked to a change in consumption. Furthermore, a number of other
important factors, ranging from family size, to health status, to the probability of death,
changes dramatically in the years after retirement. In section 2, we have seen that both
in the USA and especially in the UK, consumption drops substantially at retirement.
This could be related to insufficient savings and a misperception about the amount
of (public and private) pension benefits. Alternatively, the drop in consumption could
be explained within the optimisation framework of the life cycle model if considered
together with all the changes mentioned above. This important topic is studied in a
recent paper by Banks, Blundell and Tanner (1998) who find that, even ignoring health
status and mortality issues, two thirds o f the drop in consumption observed around
retirement can be rationalised by an optimisation model.
Second, while the endogeneity of labour supply choices can be controlled for in
the empirical analysis, the joint modelling of consumption and of labour supply is
extremely valuable and is still at the beginning 54. Similarly, expenditure on durables,
which is an important component of total expenditure, needs to be modelled in a
different way, as discussed in Section 8.
Third, the fact that a particular specification of preferences fits the data reasonably
well and that one can obtain consistent estimates of the structural parameters, does not
mean that borrowing restrictions and liquidity constraints are unimportant. The Euler
equation can be a very poor tool for identifying the presence of such phenomena, as
the equation holds whenever the constraints are not binding. For the notion of liquidity
constraints to be useful, however, it is necessary to model the behaviour of constrained
individual explicitly and determine what the aggregate implications of these constraints
might be.
A literal interpretation of the papers mentioned in the last part of Section 5 is that
a flexible version of the life cycle model, which allows for demographic and labour
supply effects in the utility function, cannot be rejected by the data. Neither excess
sensitivity tests, nor tests of overidentifying restrictions reject the null. The strength
54 Surprisinglyfew studies analyse consumption and labour supply choicesjointly. An almost exhaustive
list includes MaCurdy (1983), Browning, Deaton and Irish (1985), Hotz et al. (1988), Altug and Miller
(1990), Blundell, Meghir and Neves (1993a) and Attanasio and MaCurdy (1997). Of these only the last
two papers consider female and male labour supply jointly.
Ch. 11:
Consumption
793
of a result that does not reject the null, however, obviously depends on the power of
the test used.
The power of the tests of overidentifying restrictions or of the excess sensitivity
tests that fail to reject within a sufficiently complex version of the model might
be questioned. If an Euler equation is saturated with demographic and labour
supply variables, it might be hard to measure precisely the coefficient on expected
income or, more generally, reject overidentifying restrictions. This problem might
be particularly serious if some of these variables, such as labour supply, are
correlated with expected income. In this sense, this criticism is equivalent to state the
difficulty in distinguishing hypotheses about preferences from hypotheses on budget
constraints.
A possible response to this kind of criticism is to check whether the preferences
parameters of the model that is not rejected by the data are sensible and, most
importantly, whether the model is able to explain facts other than those that have been
used to fit it. An example will make this argument clear.
In Attanasio et al. (1996), my co-authors and I consider a relatively simple version
of the life cycle model where the within-period utility function includes demographic
variables and whose parameters are assumed to be the same across education groups.
Such a specification of the model is estimated on US average cohort data and it is
not rejected by the data. The elasticity of intertemporal substitution is estimated to
be just below 1. Given some hypotheses about the stochastic process that generates
income, interest rates and demographic variables, and given an assumption on terminal
conditions, one can solve numerically for the consumption function at each age. We
do so and then simulate the model for income and demographic profiles calibrated on
different education groups.
With this technique we are able to reproduce some of the main feature observed in
the data, that is, that the consumption profiles are steeper for the groups that have a
steeped income profiles. This feature of the data has been interpreted as a failure of the
life cycle model by Carroll and Summers (1991). Our exercise, however, shows that
a flexible version of the life cycle model incorporating some realistic features, such
as the effect of family composition on utility, can explain it. The result is remarkable
because there is nothing, at the estimation level, that fits the particular feature of the
data (differences in consumption age profiles by education groups) that the estimated
preferences (assumed to be the same across groups) and the differences in demographic
profiles are able to explain. In this sense the exercise constitute a genuine 'out-ofsample' verification of the ability of the model to fit the data.
As the utility function is of the CRRA type, the model also incorporates the
precautionary motive linked to income uncertainty which Carroll (1997a) has recently
advocated as the most likely explanation of the relationship between the shape of
consumption profiles and that of income profiles. The simulations show, however,
that this effect is quantitatively less important than that generated by differences in
demographic profiles. Indeed, most of the 'tracking' of consumption and income
profiles is generated by corresponding similarities in demographics. This argument
794
O.P. Attanasio
is also consistent with the evidence presented in Section 2 where consumption 'per
adult equivalent' profiles are quite flat over the life cycle 55.
The simulation techniques used in Attanasio et al. (1996) are similar to those used
by Deaton (1991) and by Hubbard, Skinner and Zeldes (1994), among others, to obtain
a closed form solution for consumption. These techniques are the only way, if one is
not willing to assume quadratic utility or CARA utility [as in Caballero (1990a)] 56, to
obtain a consumption function in the presence of uncertainty. They are quite expensive
as they are numerically intensive for any problem which goes beyond the simplest
assumptions and require the researcher to take a stand on any detail of the optimisation
problem (from the terminal condition to the nature of the income process). The payoff
one can obtain can, however, be quite large.
The Euler equation approach, first used by Hall (1978) in the consumption literature,
was a major innovation. It allowed the rigorous consideration of uncertainty in a
complicated dynamic decision problem while preserving the empirical tractability of
the model. The beauty of the approach consists in the fact that exploits the main
implication of the life cycle-permanent income model, that is that an optimising
consumer keeps the discounted expected value of the marginal utility of consumption
constant, to eliminate the unobservable 'fixed effect', the marginal utility of wealth,
which incorporates the influence of expectations and all the variables that are relevant
for the consumer optimisation problem. The price that one pays by differencing out
the marginal utility of wealth is, however, non-negligible. While it is true that one
can use the Euler equation to estimate structural parameters, one looses the ability of
saying anything about the l e v e l of consumption. This implies that without additional
information and structure, it is not possible to answer a number of extremely important
questions. In other words, the Euler equation for consumption is n o t a consumption
function and therefore it does not tell us anything about how consumption reacts to
news about the economic environment in which economic agents operate.
The symptom of this unsatisfactory state of affairs are apparent in the literature. If
one reads, for instance, the literature on saving, which in recent years has dealt with a
number of interesting issues that range from the decline in personal saving rates in the
USA to the effects of tax incentives on personal and national saving, one rarely finds
a systematic use of Euler equation estimates or even references to the Euler equation
literature. Indeed, the use of structural models of consumption behaviour is quite rare.
On the other hand, the same literature suffers from a number of identification problems
that arise from the inability or unwillingness to put more structure onto the descriptive
analysis. A good example is the debate on the effectiveness of fiscal incentives to
retirement saving.
55 It can be argued that the demographics in the Euler equation for consumption are picking up the
effect of the omitted conditional higher moments.
56 Both in the case of quadratic and CARA utility a number of additional assumptions about interest
rates are necessary.
795
It seems that, 40 years after the publication o f Friedman's book, we still have
not been able to construct a 'consumption function' which incorporates the main
features of the dynamic optimisation model without loosing analytical and empirical
tractability. The development o f the numerical solution techniques referred to above
is one possible avenue. The main problem with these techniques, however, is that
they require a complete specification of the economic environment (including terminal
conditions) in which the agent lives. Therefore, to make the problem treatable,
even with the use o f powerful computation methods, very strong assumptions are
needed. Furthermore, we cannot use the conditioning arguments which simplify the
estimation o f an Euler equation. While at the estimation level variables such as durables
and labour supply can be rigorously treated as endogenous without modelling them
explicitly, at the level o f numerical solutions, they cannot be ignored.
796
O.P Attanasio
From a theoretical point of view, the basic proposition can be derived if one
considers the allocation problem faced by a central planner who maximises, given a set
of Pareto weights, the average of individual utilities given the resource constraint. This
framework is particularly useful because it can be used to control easily for a number
of factors and in particular for the presence of multiple commodities, non-separable
leisure and so on. The basic implication is that the rate of change of the marginal utility
of consumption (which might depend on leisure and other commodities) is equalised
across individuals.
Townsend (1994) was one of the first to consider the empirical implications of
complete market, that is that under certain conditions "... individual consumptions are
determined by aggregate consumption, no matter what the date and history of shocks,
and so individuals' consumption will move together" (p. 540) 57. While Townsend
(1994) considered data from India, other studies, such as Mace (1991), Cochrane
(1991) and Hayashi et al. (1996), performed similar exercises using US data, that is
either the CEX or the PSID.
Both Cochrane (1991) and Hayashi et al. (1996) reject the implications of the risk
sharing hypothesis. Cochrane (1991) finds that food consumption changes in the P SID
are related to changes in health and employment status. Hayashi et al. (1996) use the
same data but control for the possibility that the relationship between changes in wages
and consumption is generated by the non-separability of leisure and consumption in
the utility function. By following spin-offs in the PS1D, they also test (and reject) the
hypothesis that risk is shared among families.
Attanasio and Davis (1996) have matched data from the Current Population Survey
and the CEX to test the hypothesis that movements in relative wages are reflected
into movements in relative consumption or, to be precise, in marginal utilities of
consumption. Attanasio and Davis consider education and year of birth cohorts. Most
of the variability in relative wages comes from the shifts in the wage distribution
across education groups that occurred during the 1980s. They find that such movements
were mirrored in movements in the consumption distribution, indicating a 'spectacular
failure' of the perfect insurance hypothesis 58. The analysis of Attanasio and Davis is
particularly damning for the perfect insurance paradigm because it focuses on economy
wide, well observed shifts, and therefore the relationship between consumption and
income cannot be explained with models which consider private information such as
those in Phelan and Townsend (1991).
Attanasio and Davis (1996) also try to evaluate the welfare cost of the lack of
an institutional framework that allows for the diversification of idiosyncratic risk.
The evaluation of such a cost obviously depends on preference parameters and on
57 Townsend (1994) credits Diamond (1967) and Wilson (1968) with the first versions of this
proposition.
58 As Hayashi et al. (1996), Attanasio and Davis (1996) use lead as well as lags of wages as instruments.
Both studies consider short and long tun shocks.
797
the nature o f the shocks considered. Attanasio and Davis (1996) under a set of
plausible assumptions and a coefficient o f relative risk aversion of two, evaluate the
cost of the lack of insurance to 'group specific shocks' at 2.5% of consumption. This
estimate is considerably larger than the cost of business cycle fluctuations evaluated
by Lucas (1987) and is more in line with the results reported by Imrohoroglu (1989).
It should also be stressed that Attanasio and Davis (1996) estimate completely ignores
'within-groups' shocks and focuses mainly on relative low (rather than business cycle)
frequencies.
Whilst the analysis of Attanasio and Davis (1996) can be framed in terms of a
test of the perfect insurance hypothesis, it can also be interpreted as documenting the
evolution in inequality in consumption and wages in the USA during the 1980s. The
increase in inequality in wages in the USA has been analysed in many studies by
labour economists. On the other hand, the evolution o f inequality in consumption has
not received much attention until quite recently. Little is known about the evolution
of consumption inequality over the business cycle and over different stages of the
process of economic growth. In addition to Attanasio and Davis (1996), Cutler and
Katz (1991) analyse the evolution of inequality in consumption during the 1980s using
CEX data. Goodman, Johnson and Webb (1997) provide an exhaustive analysis of the
recent trends in both income and expenditure inequality in the UK.
The lack of evidence on consumption inequality is somewhat disconcerting,
especially if one compares it to the amount of evidence on wage and income inequality.
From a theoretical point of view, it is not completely obvious which of the two
measures of inequality is more interesting. Blundell and Preston (1998) discuss the
advantages and drawbacks of both: consumption inequality is more likely to reflect
the cross sectional variability of permanent income, while income inequality is more
affected by temporary shocks. The relationship between the two, however, depends on
a number of factors that range from the specification of preferences to the nature of
the income shocks, to the institutions and instruments available to households to (self)
insure against idiosyncratic shocks.
More generally, a theoretically consistent analysis of consumption inequality cannot
avoid the discussion of the implications of different preference specifications and
market institutions. Suppose, for instance, that individual households are prevented
from borrowing. If labour supply behaviour is considered exogenous, these households
will engage in precautionary saving to avoid the effects of extremely negative
shocks to income. On the other hand, if one models labour supply choices and
in particular female participation in the labour market, it is quite possible that
the role of precautionary saving is taken by labour supply. That is to say, it is
possible that households, rather than consuming less goods, decide to consume less
leisure.
Another important element, which has been only recently analysed, is the availability
of means tested safety nets. Hubbard, Skinner and Zeldes (1994) consider the
effect that liquidity constraints and means tested safety nets have on the pattern of
consumption using numerical techniques to solve the consumption function.
798
O.P. Attanasio
Browning and Crossley (1997) have recently considered another potentially important channel that some households might use to smooth the marginal utility of
consumption over time: what they call small durables. The idea is intuitive: individuals
that have no access to savings or loans to smooth out fluctuations in income, might
postpone the replacement of small durables. The features that characterise the 'small
durables' in Browning and Crossley's problem are the durability and the irreversibility
(due to the lack of a second hand market) of such items, as well as the fact that
scrappage decisions are not only determined by the physical depreciation but also
by the economic situation faced by the agent. Browning and Crossley (1997) present
both some theoretical results and some empirical evidence on a sample of Canadian
unemployed supporting the importance of this type of smoothing behaviour. The
evidence I presented in Table 6, about the relative variability of durable expenditure
for groups of households with different levels of education is consistent with the idea
that poorer individuals might use durables to smooth out fluctuations in income.
Browning and Crossley (1997) focus on items that cannot be used as collateral (such
as clothes or small appliances) to obtain loans. It would be interesting to consider
the implications of the same ideas for the replacement of items for which reasonably
efficient second hand markets exists and which are collateralizable such as automobiles
and, to a certain extent, 'white goods'. Greenspan and Cohen (1996) have stressed
the important role that scrappage of old cars plays in forecasting expenditure on
automobiles which, in turn, is a very important indicator of the status of the business
cycle.
A few studies, recently, have started to study the evolution of inequality over the
life cycle. The idea is quite simple: if consumption follows a random walk, the cross
sectional distribution of individual consumption should 'fan out' over the life cycle.
That is, the cross sectional variance of individual consumption should be increasing,
on average, over long periods of time. This idea was first exploited by Deaton and
Paxson (1994) who analysed average cohort profiles for the variance of (log) nondurable consumption in three different countries, the USA, the UK and Taiwan.
8. Intertemporal non-separability
In the discussion so far, I have assumed that preferences are separable over time. While
this is an extremely convenient assumption, it is easy to think of situations in which it is
violated. In this section I consider the implications of the fact that current expenditure
might have lasting effects. After sketching the general problem, I discuss in detail two
particular examples of time non-separability of some importance: that of durability
and of habit formation. As is clear from the discussion, the complication implied by
time dependence of preferences is only one of the things that make the treatment of
these phenomena extremely hard.
The general problems can be stated in reasonably simple terms if one states the
problem in terms of a flow of services derived from a 'stock' and assumes that there
799
are no costs in changing the 'stock'. The simple problem in Section 3, can therefore
be re-written as:
T-t
max
Et v(s,+j)
j=0
IAt+j+l~l+rt+j)At+j+Yt+j-st+j,
subject to [St+j = Z
(18)
= 1
k=0
where St is the 'stock' from which utility is derived, st is expenditure and the
coefficients ak define the type of time dependency. As we see below, such a model
can easily incorporate durability as well as habit formation. It is also straightforward
to consider the case in which S is a vector whose components have different degrees
of durability 59.
A rational individual, in choosing the level of expenditure s will take into account
the effect that this has on the level of the 'stock' in the current as well as in the future
period. It is straightforward to verify that the first-order condition for such a problem
is going to be
(19)
where m t = ~-~'~k=0~ku1(St+k) ak" Notice that if the ak s are different from zero,
mt is not a variable known at time t as it depends on the future marginal utility of the
stock S. In general, therefore, it will not be possible to express Equation (19) in terms
of the rate o f growth of mt as is usually done [see, for instance, Equation (8)]. Notice
also that the marginal utility of expenditure mt depends on a potentially very large
number of terms. To make an equation such as (19) operational it will be necessary
to simplify it by some algebraic manipulation whose nature depends, once again, on
the pattern of the coefficients ak.
The two models of non-separability I consider below (durables and habits) differ on
the nature of the time dependence. In the first case we have substitutability, while in
the latter we have complementarity of expenditure over time.
8.1. Durables
As illustrated in Section 2, expenditure on durables, both at the aggregate level and
at the cohort level, is by far the most volatile component of consumption. In addition,
the dynamic of durable expenditure seems much more complex of that of the other
components o f consumption. A simple model of durable consumption can be obtained
O.P Attanasio
800
from the system in Equation (18) if it is assumed that ak = (1 - 6) k. In this case S can
be interpreted as the value of the stock of durables, and 6 as the depreciation rate, and
the third equation in system (18) becomes
St = (1 - 6)St_1 +st.
(20)
Assuming quadratic utility, Mankiw (1982) generalised Hall's random walk model
to durable expenditure. In particular, he showed that under these circumstances the
model implies that changes in expenditure should follow an MA(1) process with
coefficient equal to ( 1 - 6). This implication is strongly rejected by the aggregate
data, as shown, for instance, in Section 2. Dunn and Singleton (1986) have used a
more attractive preference specification and considered the Euler equation for durable
consmnption and its implications for asset pricing. The most comprehensive treatment
of durability and of the Euler equations associated with a variety of preferences is
found in Eichenbaum and Hansen (1990) who use a Gorman-Lancaster technology that
converts expenditure into stocks and stocks into services over which utility is defined.
The aggregation issues I discussed in Section 4 are obviously relevant for durables
as well. The aggregation problems are actually even more complicated because of the
possibility of corner solutions, i.e. households that decide not to own a durable (a car
for instance), that can be safely ruled out for non-durables under the assumption that
the marginal utility goes to infinity at zero or low levels of consumption 6.
A possibility for the deviations of the dynamics of aggregate durable expenditure
from that implied by the simple model is that of the existence of adjustment costs. It
seems natural to assume that adjusting the stock of durables involves costs, motivated
both by the existence of imperfections in the second hand market and by 'pure'
adjustment costs (such as search costs, taxes and similar).
The literature first studied convex (typically quadratic) costs of adjustment. A typical
example is Bernanke (1984, 1985), who estimated models with quadratic costs both
with aggregate and individual data. Eichenbaum and Hansen (1990) also incorporate
the possibility of adjustment costs in their technology. Such attempts, however, have
turned out to be unsuccessful. The main reason for this is that convex adjustment
costs predict a smooth adjustment towards an equilibrium. To avoid increasing costs
households will adjust their stock of durables often and by small amounts. Even
casual observation suggests that this is not a very accurate description of household
behaviour.
The next step is then to consider non-convex costs of adjustment; that is either fix
or proportional costs. The characterisation of optimal behaviour under this type of
costs is obviously much harder because for many households in many periods the
optimal policy involves no adjustment. The first paper to characterise the optimal
6o Bernanke (1984) and Hayashi (1985a,b) are among the few studies that have used individual data to
analyse durable expenditure.
801
adjustment policy under non-convex costs was Grossman and Laroque (1990) who
studied a problem of a consumer deriving utility from a single durable and solving a
dynamic optimisation problem involving the choice of the optimal size for the durable
and the optimal portfolio investment for her financial wealth. Grossman and Laroque
(1990) proved that the optimal strategy involves an (S, s) rule of the kind considered
in the optimal inventory literature of the 1950s. Specifically, they proved that the value
function associated to the consumer problem is a function of a single state variable,
the ratio of the value of the durable to financial wealth, and that the durable is adjusted
to a 'target' level when the state variable crosses a lower or an upper bound. When the
state variable is within the two bounds the optimal policy is not to adjust the durable. In
addition, Grossman and Laroque (1990) characterised the size of the inaction band as
a function of the parameters of the problem, as well as the optimal portfolio strategy.
One of the problems of the Grossman-Laroque model is that a solution can be
obtained only if the problem can be expressed as a function of a single state variable.
This precludes, for instance, the consideration of labour income. More recently, Eberly
(1994) and Beaulieu (1993) have extended the Grossman-Laroque model in various
directions. Eberly (1994), has showed that changes in the durable stock between periods
in which an adjustment is performed obey to an Euler equation. Beaulieu (1993) has
reformulated the Grossman-Laroque model and expressed the (S, s) rule in terms of
the ratio of durables to non-durables.
The attractiveness of the (S,s) model of adjustment lies in its implications that,
in most periods, consumers do not adjust their stock of durables and when they do
adjust, they usually make substantial adjustments. These features (large and infrequent
adjustments), that can be generated by the (S, s) model, are the exact contrary of those
of the quadratic adjustment model which yields small and frequent adjustments.
The consideration of large and infrequent adjustments poses an entire new category
of aggregation problems. Caplin (1985) was the first to consider them. Caplin and
Spulber (1987), in the context of a model of price adjustment, provided (stringent)
conditions under which lumpy individual adjustment could result in a smooth aggregate
adjustment. Bertola and Caballero (1990) analysed the implications of the (S,s) type
of adjustment for the dynamics of aggregate durable expenditure and stressed that the
assumptions needed to generate the type of neutrality results obtained by Caplin and
Spulber are indeed very fragile. (S, s) models can generate very rich and complicate
dynamics that seem not to be inconsistent with the observed autocorrelation of durable
expenditure. Caballero (1990a,b, 1993), in particular, has stressed how (S,s) models
can generate patterns of MA coefficients that are similar to those I reported in
Section 2. Caballero and Engel (1991) have studied more generally the aggregate
properties of (S, s) economies. Caballero (1993) has tried to estimate the parameters of
the (S, s) model by fitting the highly non-linear model resulting from the aggregation
of a simple (S, s) rule to aggregate time series data.
The estimation of (S, s) models with individual data is not an easy task for several
reasons. First of all, the data requirements can be quite formidable: information on
the value of the stock of durables before and after the adjustment is necessary. It is
802
o.P Attanasio
also desirable to follow households over some time to bound the range of inaction by
the households that are observed not to adjust. In addition, the numerical problems in
estimation can be quite formidable.
The first attempt at estimating such a model with individual data was by Lam (1991)
who estimated by maximum likelihood an (S, s) rule defined in terms of the ratio of the
stock of automobiles over permanent income. Eberly (1994), also estimated the width
of the (S, s) rule for the subset of consumers who were observed to adjust the stock of
their automobiles in her sample 6~. B eaulieu (1992) also uses individual observations
but follows a different approach. He considers a single cross section and computes
the ergodic cross sectional distribution to which the economy would converge in the
absence of aggregate shocks. The parameters of this distribution will depend on the
parameters of the (S, s) rule. He can then estimate them from the parameters of the
observed cross sectional distribution which is assumed to coincide with the ergodic
distribution.
Like Lain (1991), I have estimated the parameters of the (S,s) rule by maximum
likelihood on individual data in Attanasio (1995a). However, I formulate the (S, s) rule
in terms of the ratio of the stock of automobiles to non-durable consumption, allow
for observed and unobserved heterogeneity in both the target and the band width
and consider a more flexible stochastic specification. The consideration of unobserved
heterogeneity in both the target level and the band width makes the model equivalent
to a recent formulation of lumpy adjustment by Caballero and Engel (1991) who
consider, rather than a strict (S, s) rule, a 'hazard function' which gives the probability
of adjusting as a function of the difference between the current stock and its target
level.
While substantial progress in the understanding of the behaviour of models with
lumpy adjustment has been made (not only in consumption but also in investment
and labour demand), a lot of research is still necessary to establish the empirical
relevance of these models and to quantify their parameters. The aggregate dynamics
of an (S, s) system depends crucially on the properties of several stochastic processes
about which we still have little or no information. These include the process by which
the state variable changes when no adjustment occurs, the process by which the target
level changes, the degree of heterogeneity in target levels and band width, the degree of
persistence of individual shocks to band width (cost of adjustment) and the correlations
among these variables.
Ch. 11:
803
Consumption
coefficients a in Equation (18). For instance, Heaton (1995) and Constantinides (1990)
use the following specification:
O<3
St = ct - a(1 - 0) ~
OJet a~;
O~a~l,
0~0~1.
(21)
j-0
The term (1 - 0) ~ ) ~ 00Jct-l-j is referred to as the 'stock of habits', which depreciates
according to the parameter 0. A particularly simple specification is obtained when
0 = 0, in which case the stock of habits is simply given by last period consumption.
As with the cases considered above, the marginal utility of time t expenditure is a
function of both present and future variables, so that it is now known at time t. The
habit formation model has a long history, dating back at least to Gorman (1967), Pollak
(1970) and Houthakker and Taylor (1970) 62. Spynnewin (1981) presents an ingenious
reparametrization of the optimisation problem which, by an appropriate redefinition of
prices and quantities, allows the maximisation of an intertemporally additive function.
However, it is only in the 1980s that habits models are coupled with the hypothesis
of rational expectations and preferences incorporating temporal persistence are used
to derive Euler equations analogous to Equation (19) above. Examples of this practice
include Eichenbaum and Hansen (1988) (for a model with habit forming consumption
and leisure), Eichenbaum and Hansen (1990), Novales (1990) and Constantinides
(1990), while Hotz, Kydland and Sedlacek (1988) and Kennan (1988) consider habit
formation in leisure. Browning (1991) uses a dual approach to derive the equivalent of
'Frisch' consumption functions in the presence of intertemporal non-separability 63.
More recently, Heaton (1993, 1995) has considered the interplay between time
aggregation and habit formation. His argument is that at high frequency consumption
seems to exhibit substitutability, while at lower frequencies, there is evidence of
complementarities. Heaton argues that the local substitutability of consumption could
be explained by time aggregation. In general, he presents preferences that are flexible
enough to accommodate the 'slow' formation of habits.
Most of the work on habit models so far has been done on aggregate time series data.
One of the reasons for this is the fact that very few panel data contain information on
consumption. The CEX, which I used in Section 2, contains only up to four quarterly
observations per households. The average cohort analysis that is used in the study of
dynamic models of consumption, cannot be used in the analysis of habit formation,
because these models involve the covariance of subsequent consumption observations
f o r the s a m e household. In other words, we cannot aggregate the product C h
t C ht k over
the household index h, unless we have observations for times t and t - k for the same
households.
62 Browning (1991), however, has citations from Marshall and Haavelmo.
63 Browning's simple structatre can encompass both substitutability and complementarities over time.
The main problem with his approach is that he can only introduce uncertainty by considering points
expectations about the future.
804
O.P. Attanasio
The only paper that studies time dependence in preferences using micro data is, to
the best of my knowledge, Meghir and Weber (1996), that I discussed in the section
on liquidity constraints. Another novelty of that paper is that it considers different
components of consumption and allows for different levels of persistence depending on
the commodity considered. The interesting evidence is that, when durable commodities
are controlled for, there is no evidence of persistence in the 3 equations demand system
that Meghir and Weber (1996) consider.
9. Conclusions
Rather than summarising the various sections that compose the chapter I prefer to
conclude the chapter comparing the status of our knowledge and understanding of
consumption behaviour to that of twenty years ago. It is fair to say that substantial
progress has been made. In the last twenty years we have learned how to deal with
uncertainty in a rigorous fashion and have recognised the importance that this may
have for consumption and saving decision. Indeed an entire branch of the literature,
that on precautionary saving, has developed to deal with these issues.
While the emphasis given to the proper treatment of uncertainty and the lack of
appropriate data sources has meant that many studies have focused on aggregate data,
it has now become clear that it is very hard, if not impossible, to estimate preference
parameters from aggregate time series data. Aggregation issues are important in
many dimensions. In addition to the standard aggregation problems created by the
non-linearity of the relevant theoretical relationships, there are other ways in which
aggregation issues become important. Corner solutions and participation decisions (in
labour and financial markets), inertial behaviour and transaction costs, all add a new
dimension of complexity and make the cross sectional distribution of the variables
under study relevant for the dynamics of the aggregate.
The Euler equation has been the main instrument to analyse consumption both in
micro and macro data, to estimate preference parameters and to test the overidentifying
restrictions implied by the consumers' optimisation problem. In the process we have
learned a lot about the econometric problems of estimating Euler equations. In
particular, we have learned what are the identifying assumptions that one needs to
make to get consistent estimates from the available data.
The Euler equation is a powerful tool in that it allows the consideration of complex
and flexible preference specifications without loosing the empirical tractability. As long
as the variable over which one is optimising can be adjusted without cost and is not
at a corner, one can derive an Euler equation which, even if it includes the values
of other endogenous variables, delivers orthogonality restrictions that can be used to
estimate preference parameters. The empirical research has used Euler equations for
non-durable consumption and its components that have become, in the attempt to fit
the observable data, increasingly complicated. This level of complexity is probably
unavoidable if one is serious about taking the model to the data.
805
As I stressed several times during the chapter, the price one pays in dealing
with Euler equations is not negligible: one loses the ability of saying anything
about the level of consumption. The empirical tractability of the Euler equation is
obtained differencing out the marginal utility of wealth, and therefore one of the
main determinants of consumption levels. The challenge for future research is to
construct a consumption function that incorporates the insights of the Euler equation
and yet allows us to say something about the levels o f consumption and about how
consumption reacts to changes in the economic environment. Such a consumption
function is necessary to make predictions about future consumption, about saving
behaviour, and about the effects of alternative policy measures.
One possibility that is being explored is the use of numerical techniques. They
have proved to be useful both to improve our understanding of the dynamic
optimization problems typically postulated and to characterise the implications of
different preference structures and economic environments on consumption and saving
behaviour. They can also be used to validate, indirectly, the preference specifications
implied by the Euler equation estimates that best fit the consumption growth data.
Their main limitation, however, is an important one: they can only be used to analyse
extremely simplified models and they require the full characterisation of the agents'
economic environment.
Several other areas of research are important and exciting. I will just mention two.
Our understanding of consumption smoothing and of the evolution of inequality is still
at the beginning. The development of this understanding and of the importance that
different institutional frameworks and financial instruments have for these issues is an
extremely important research agenda both from a positive and from a normative point
of view. Related to this is also the issue of the time series properties of individual
consumption: unlike for earnings and hours of work, next to nothing is known about
this. Our understanding of durable expenditure is still very limited and yet it is an
extremely important issue, both because durables are the most volatile component of
consumption and because they have a number of features (the fact that they can be
used as collateral, transaction costs, the durability itself) that make them difficult to
model.
References
Abowd, J., and D. Card (1989), "On the covariancestructureof earnings and hours changes",Econometrica
57:411-445.
Alessie, R., B. Melenberg and G. Weber (1989), "Consumption, leisure and earnings related liquidity
constraints: a note", Economics Letters, 27:101-104.
Alessie, R., M. Devereux and G. Weber (1997), "Intertemporal consumption, durables and liquidity
constraints: a cohort analysis", European Economics Review 41:37-60.
Altonji, J.G., and A. Siow (1987), "Testingthe response of consumption to income changes with (noisy)
panel data", Quarterly Journal of Economics 102:293-328.
Altug, S., and R.A. Miller (1990), "Household choices in equilibrium", Econometrica58(3):543 570.
806
O.P. Attanasio
Arellano, M., and C. Meghir (1992), "Female labour supply and on the job search. An empirical model
estimated using complementary data sets", Review of Economics Studies 59:537-557.
Atkeson, A., and M. Ogaki (1996), "Wealth-varying intertemporal elasticities of substitution: evidence
from panel and aggregate data", Journal of Monetary Economics 38:507334.
Attanasio, O.E (1994), "Personal saving in the US", in: J. Poterba, ed., International Comparisons of
Household Savings (NBER/Chicago University Press) 57-124.
Attanasio, O.E (1995a), "Consumer durables and inertial behavior: estimation and aggregation of (S, s)
rules", Working Paper No. 5282 (NBER).
Attanasio, O.E (1995b), "The intertemporal allocation of consumption", Carnegie-Rochester Conference
on Public Policy, Supplement to the Journal of Monetary Economics.
Attanasio, O.E, and J. Banks (1997), "Trends in household saving", Economic Policy, No. 27, 549-583.
Attanasio, O.P., and M. Browning (1995), "Consumption over the life cycle and over the business cycle",
American Economic Review 85:1118-1137.
Attanasio, O.E, and S.J. Davis (1996), "Relative wage movements and the distribution of consumption",
Journal of Political Economy 104:1227-1262.
Attanasio, O.E, and EK. Goldberg (1997), "On the relevance of borrowing restrictions. Evidence from
car loans in the US", mimeograph (University College London).
Attanasio, O.E, and H.W Hoynes (1995), "Differential mortality and wealth accumulation", Working
Paper No. 5126 (NBER); Journal of Human Resources, forthcoming.
Attanasio, O.E, and T.E. MaCurdy (1997), "Interactions in household labor supply and their implications
for the impact of EITC", mimeograph (Stanford University).
Attanasio, O.E, and G. Weber (1989), "Intertemporal substitution, risk aversion and the Euler equation
for consumption", Economic Journal 99(Suppl.):59-73.
Attanasio, O.E, and G. Weber (1993), "Consumption growth, the interest rate and aggregation", Review
of Economics Studies 60:631-649.
Attanasio, O.E, and G. Weber (1994), "The UK consumption boom of the late 1980s: aggregate
implications of microeconomic evidence", Economic Journal 104:1269 1302.
Attanasio, O.P., and G. Weber (1995), "Is consumption growth consistent with intertemporal optimization?
Evidence from the consumer expenditure survey", Journal of Political Economy 103:1121-1157.
Attanasio, O.P., J. Banks, C. Meghir and G. Weber (1996), "Humps and bumps in lifetime consumption",
Working Paper No. 5350 (NBER); Journal of Business and Economic Statistics, forthcoming.
Banks, J., and EG. Johnson, eds (1997), How Reliable is the Family Expenditure Survey? Trends in
Income and Expenditures over Time (The Institute for Fiscal Studies, London).
Banks, J., R. Blundell and L Preston (1994), "Life-cycle expenditure allocations and the consumption
cost of children", European Economic Review 38:1234-1278.
Banks, J., R. Blundell and A. Brugiavini (1997), "Risk pooling and consumption growth", mimeograph
(University College London).
Banks, J., R. Bhmdell and S. Tanner (1998), "Is there a retirement saving puzzle?", American Economic
Review, forthcoming.
Bean, C.R. (1986), "The estimation of 'surprise' models and the 'surprise' consumption function",
Review of Economics Studies 53:497-516.
Beaulieu, J.J. (1992), "Aggregate movements in durable goods with fixed costs of adjustment",
mimeograph (MIT).
Beaulieu, J.J. (1993), "Optimal durable and non-durable consumption with transaction costs", Federal
Reserve Board Washington, Finance and Economics Discussion Series, 93-12.
Bernanke, B.S. (1984), "Permanent income, liquidity, and expenditure on automobiles: evidence from
panel data", Quarterly Journal of Economics 99:587-614.
Bernanke, B.S. (1985), "Adjustment costs, durables, and aggregate consumption", Journal of Monetary
Economics 15:41 68.
Bertola, G., and R.J. Caballero (1990), "Kinked adjustment costs and aggregate dynamics", in:
Ch. 11:
Consumption
807
O.J. Blanchard and S. Fischer, eds., NBER Macroeconomics Annual (MIT Press, Cambridge, MA)
237-288.
Blinder, A., and A. Deaton (1985), "The time-series consumption revisited", Brookings Papers on
Economic Activity 1985:465321.
Blundell, R., and I. Preston (1998), "Consumption inequality and income uncertainty", Quarterly Journal
of Economics 103:603-640.
Blundell, R., and T. Stoker (1999), "Consumption and the timing of income risk", European Economic
Review 43(3):475-508.
Blundell, R., C. Meghir and E Neves (1993a), "Labor Supply and lntertemporal Substitution", Journal
of Econometrics 59:137 160.
Blundell, R., E Pashardes and G. Weber (1993b), "What do we learn about consumer demand patterns
from micro data?", American Economic Review 83:570-597.
Blundell, R., M. Browning and C. Meghir (1994), "Consumer demand and the lifetime allocation of
consumption", Review of Economics Studies 61:57-80.
Browning, M. (1987), "Eating, drinking, smoking, and testing the lifecycle hypothesis", Quarterly Journal
of Economics 102:329-345.
Browning, M. (1991), "A simple nonadditive preference structure for models of household behavior over
time", Journal of Political Economy 99:607-637.
Browning, M., and T. Crossley (1997), "Socks and stocks", mimeograph (McMaster University).
Browning, M., and A. Lusardi (1996), "Household saving: micro theories and micro facts", Journal of
Economic Literature 34:17921855.
Browning, M., and C. Meghir (1991), "The effect of male and female labour supply on commodity
demands", Econometrica 59:925-951.
Browning, M., A. Deaton and M. Irish (1985), "A profitable approach to labor supply and commodity
demand over the life cycle", Econometrica 53:503-544.
Brugiavini, A., and G. Weber (1994), "Durable and non durable consumption: evidence from Italian
household data", in: A. Ando, L. Guiso and I. Visco, eds., Saving and the Accumulation of Wealth
(Cambridge University Press, Cambridge) 305-329.
Caballero, R.J. (1990a), "Consumption puzzles and precautionary savings", Journal of Monetary
Economics 25:113-136.
Caballero, R.J. (1990b), "Expenditure on durable goods: a case for slow adjustment", Quarterly Journal
of Economics 105:727-743.
Caballero, R.J. (1991), "Earnings uncertainty and aggregate wealth accumulation", American Economic
Review 81:859-871.
Caballero, R.J. (1993), "Durable goods: an explanation for their slow adjustment", Journal of Political
Economy, 101:351-384.
Caballero, R.J., and E. Engel (1991), "Dynamic (S, s) economies", Econometrica 59:1659-1686.
Campbell, J.Y. (1987), "Does saving anticipate declining labor income? An alternative test of the
permanent income hypothesis", Econometrica 55:1249-1273.
Campbell, J.Y., and A.S. Deaton (1989), "Why is consumption so smooth?", Review of Economic Studies
56:357074.
Campbell, J.Y., and N.G. Mankiw (1989), "Consumption, income and interest rates: reinterpreting the
time series evidence", in: O.J. Blanchard and S. Fischer, eds., NBER Macroeconomies Annual 1989
(MIT Press, Cambridge, MA) 185-216.
Campbell, J.Y., and N.G. Mankiw (1991), "The response of consumption to income: a cross-country
investigation", European Economic Review 35:715-721.
Caplin, A.S. (1985), "The variability of aggregate demand with (S, s) inventory policies", Econometrica
53:1395-1409.
Caplin, A.S., and D. Spulber (1987), "Menu costs and the neutrality of money", Quarterly Journal of
Economics 102:703-726.
808
O.P. Attanasio
Carroll, C.D. (1994), "How does future income affect current consumption?", Quarterly Journal of
Economics 109:111-147.
Carroll, C.D. (1997a), "Buffer-stock saving and the life cycle/permanent income hypothesis", Quarterly
Journal of Economics 112:1-55.
Carroll, C.D. (1997b), "Death to the log-linearised Euler equation", mimeograph (Johns Hopkins
University).
Carroll, C.D., and M. Kimball (1996), "On the concavity of the consumption function", Econometrica
64:981-992.
Carroll, C.D., and L.H. Summers (1991), "Consumption growth parallels income growth: some new
evidence", in: B.D. Bernheim and J.B. Shoven, eds., National Saving and Economic Performance
(University of Chicago Press, Chicago, IL) 305-343.
Carroll, C.D., J.C. Fuhrer and D.W. Wilcox (1994), "Does consumer sentiment forecast household
spending? If so why?", American Economic Review 84:1397 1408.
Chah, E.Y., V.A. Ramey and R.M. Starr (1995), "Liquidity constraints and intertemporal consumer
optimization: theory and evidence from durable goods", Journal of Money, Credit and Banking
27:272-287.
Chamberlain, G. (1984), "Panel data", in: Z. Griliches and M.D. Intriligator, eds., Handbook of
Econometrics, vol. 2 (North-Holland, Amsterdam) 1247-1318.
Christiano, L.J., and M. Eichenbaum (1990), "Unit roots in real GNP: do we know and do we care?",
Carnegie-Rochester Conference Series on Public Policy 32:7-61.
Cochrane, J.H. (1991), "A simple test of consumption insurance", Journal of Political Economy 99:
957-976.
Constantinides, G.M. (1990), "Habit formation: a resolution of the equity premium puzzle", Journal of
Political Economy 98:519-543.
Constantinides, G.M., and D. Duffle (1996), "Asset pricing with heterogeneous consumers", Journal of
Political Economy 104:219-240.
Cutler, D.M., and L.E Katz (1991), "Macroeconomic performance and the disadvantaged", Brookings
Papers on Economic Activity 1991(1): 1-74.
Davidson, J., D. Hendry, E Srba and S. Yeo (1978), "Econometric modelling of the aggregate time-series
relationship between consumers' expenditure and income in the UK", Economic Journal 88:661-692.
Davies, J.B. (1981), "Uncertain lifetime, consumption, and dissaving in retirement", Journal of Political
Economy 89:561-577.
Deaton, A. (1985), "Panel data from time series of cross sections", Journal of Econometrics 30:109-126.
Deaton, A. (1991), "Saving and liquidity constraints", Eeonometrica 59:1221-1248.
Deaton, A. (1992), Understanding Consumption (Oxford University Press, Oxford).
Deaton, A., and J. Muellbauer (1980), Economics and Consumer Behavior (Cambridge University Press,
Cambridge).
Deaton, A., and C. Paxson (1994), "Intertemporal choice and inequality", Journal of Political Economy
102:437-467.
Diamond, P. (1967), "The role of a stock market in a general equilibrium model with technological
uncertainty", American Economic Review 57:759-776.
Dreze, J., and E Modigliani (1972), "Consumption decisions under uncercainty", Journal of Economics
Theory 5:308-335.
Dunn, K.B., and K.J. Singleton (1986), "Modeling the term structure of interest rates under non-separable
utility and durability of goods", Journal of Financial Economics 17:27-55.
Dynan, K.E. (1993), "How prudent are consumers?", Journal of Political Economy 101:1104~1113.
Eberly, J.C. (1994), "Adjustment of counsumers' durables stocks: evidence from automobile purchases",
Journal of Political Economy 102:403-436.
Eichenbaum, M., and L.E Hansen (1988), "A time series analysis of representative agent models of
consumption and leisure under uncertainty", Quarterly Journal of Economics 103:51-78.
Ch. 11:
Consumption
809
Eichenbaum, M., and L.E Hansen (1990), "Estimating models with intertemporal substitution using
aggregate time series data", Journal of Business and Economic Statistics 8:53-69.
Epstein, L.G., and S.E. Zin (1989), "Substitution, risk aversion, and the temporal behavior of consumption
and asset returns: a theoretical framework", Econometrica 57:937-969.
Epstein, L.G., and S.E. Zin (1991), "Substitution, risk aversion, and the temporal behavior of consumption
and asset returns: an empirical analysis", Journal of Political Economy 99:263-286.
Flavin, M. (1981), "The adjustment of consumption to changing expectations about future income",
Journal of Political Economy 89:974-1009.
Flemming, J.S. (1973), "The consumption function when capital markets are imperfect. The permanent
income hypothesis reconsidered", Oxford Economic Papers 25:160-172.
Friedman, M. (1957), A Theory of the Consumption Function (Princeton University Press,
Princeton, N J).
Garcia, R., A. Lusardi and S. Ng (1997), "Excess sensitivity and asymmetries in consumption", Journal
of Money, Credit and Banking, forthcoming.
Ghez, G., and G.S. Beeker (1975), The Allocation of Time and Goods over the Life Cycle (NBER/
Columbia University Press, New York).
Gokhale, J., L.J. Kotlikoff and J. Sabelhaus (1996), "Understanding the postwar decline in U.S. saving:
a cohort analysis", Brookings Papers on Economic Activity 1996(1):315-390.
Goodfriend, M. (1992), "Information aggregation bias", American Economic Review 82:508-5 l 9.
Goodman, A., R Johnson and S. Webb (1997), Inequality in the UK (IFS, London).
Gorman, W.M. (1959), "Separable utility and aggregation", Econometrica 27:469-481.
Gorman, W.M. (1967), "Tastes, habits and choices", International Economic Review 8:218522.
Greenspan, A., and D. Cohen (1996), "Motor vehicle stocks, scrappage and sales", Federal Reserve
Board Finance and Economics Discussion Paper, No. 96-40.
Grossman, S.J., and G. Laroque (1990), "Asset pricing and optimal portfolio choice in the presence of
illiquid durable consumption goods", Econometrica 58:25-51.
Guiso, L., T. Jappelli and D. Terlizzese (1992), "Earnings uncertainty and precautionary savings", Journal
of Monetary Economics 30:307-337.
Guiso, L., T. Jappelli and D. Terlizzese (1996), "Income risk, borrowing constraints and portfolio choice",
American Economic Review 86:158-372.
Hall, R.E. (1978), "The stochastic implications of the life cycle-permanent income hypothesis: theory
and evidence", Journal of Political Economy 86:971-987.
Hall, R.E. (1988), "Intertemporal substitution in consumption", Journal of Political Economy
96:339-357.
Hall, R.E., and ES. Mishkin (1982), "The sensitivity of consumption to transitory income: estimate from
panel data on households", Econometrica 50:461-481.
Hansen, L.E, and K.J. Singleton (1982), "Generalized instrumental variables estimation of nonlinear
rational expectation models", Econometrica 50:1269-1286.
Hansen, L.P., and K.J. Singleton (1983), "Stochastic consumption, risk aversion and the temporal behavior
of asset returns", Journal of Political Economy 91:249-265.
Hayashi, E (1985a), "The effects of liquidity constraints on consumption: a cross section analysis",
Quarterly Journal of Economics 100:183-206.
Hayashi, E (1985b), "The permanent income hypothesis and consumption durability: analysis based on
Japanese panel data", Quarterly Journal of Economics 100:1083-1116.
Hayashi, E (1987), "Tests for liquidity constraints: a critical survey and some new observations", in:
T.E Bewley, ed., Advances in Econometrics: Fifth World Congress, vol. Ii, 91-120.
Hayashi, E (1996), "A review of empirical studies of household saving", mimeograph (Columbia
University).
Hayashi, E, and C.A. Sims (1983), "Nearly efficient estimation of time series models with predetermined
but not exogenous instruments", Econometrica 51:783-792.
810
O.P. Attanasio
Hayashi, E, J. Altonji and L. Kotlikoff (1996), "Risk-sharing between and within families", Econometrica
64:261~94.
Heaton, J. (1993), "The interaction between time-nonseparable preferences and time aggregation",
Econometrica 61:353-386.
Heaton, J. (1995), "An empirical investigation of asset pricing with temporally dependent preferences",
Econometrica 63:681 718.
Heckman, J.J. (1974), "Life-cycle consumption and labour supply: an exploration of the relationship
between income and consumption over the life cycle", American Economic Review 64:188-194.
Heckman, J.J., and R. Robb (1987), "Using longitudinal data to estimate age period and cohort effects
in earnings equations", in: W.M. Mason and S.E. Fienberg, eds., Cohort Analysis in Social Research
(Springer, Berlin).
Hotz, V.J., EE. Kydland and G.J. Sedlacek (1988), "Intertemporal preferences and labor supply",
Econometrica 56:335-360.
Houthakker, H.S., and L.D. Taylor (1970), Consumer Demand in the United States: Analyses and
Projections, 2rid edition (Harvard University Press, Cambridge, MA).
Hubbard, R.G., J.S. Skinner and S.P. Zeldes (1994), "The importance of precautionary motives in
explaining individual and aggregate saving", Carnegie-Rochester Conference Series on Public Policy
40:59-125.
Hubbard, R.G., J.S. Skinner and S.P. Zeldes (1995), "Precautionary saving and social insurance", Journal
of Political Economy 103:360-399.
Hurd, M.D. (1989), "Mortality risk and bequests", Econometrica 57:779-813.
Imrohoroglu, A. (1989), "Cost of business cycle with indivisibilities and liquidity constraints", Journal
of Political Economy 100:118-142.
Jappelli, T. (1990), "Who is credit constrained in the US economy?", Quarterly Journal of Economics
105:219-234.
Jappelli, T., and E Modigliani (1997), "Is the age-saving profile consistent with the life cycle hypothesis?",
mimeograph (MIT).
Jappelli, T., and M. Pagano (1994), "Saving, growth and liquidity constraints", Quarterly Journal of
Economics 109:83-110.
Jappelli, T., J.-S. Pischke and N. Souleles (1997), "Testing for liquidity constraints using complementary
data sources", Review of Economics and Statistics, forthcoming.
Juster, ET., and R.P. Shay (1964), "Consumer sensitivity to finance rates: an empirical and analytical
investigation", NBER Occasional Paper No. 88.
Keane, M.P., and D. Rankle (1992), "On the estimation of panel-data models with serial correlation
when instruments are not strictly exogenous", Journal of Business and Economic Statistics 10:1 9.
Kennan, J. (1988), "An econometric analysis of fluctuations in aggregate labor supply and demand",
Econometrica 56:317-334.
Kimball, M.S. (1990), "Precautionary saving in the small and in the large", Econometrica 58:53-73.
Kotlikoff, L.J. (1988), "Intergenerational transfers and savings", Journal of Economic Perspectives
2:41-58.
Kotlikoff, L.J., and L.H. Summers (1981), "The role of intergenerational transfers in aggregate capital
accumulation", Journal of Political Economy 89:706-732.
Lam, P.S. (1991), "Permanent income, liquidity and adjustment of automobile stocks: evidence from
panel data", Quarterly Journal of Economics 106:203-230.
Lucas Jr, R.E. (1987), Models of Business Cycles (Blackwell, New York).
Ludvigson, S. (1997), "Consumption and credit: a model of time-varying liquidity constraints",
mimeograph (Federal Reserve Bank of New York).
Ludvigson, S., and C. Paxson (1997), "Approximation bias in linearized Euler equations", mimeograph
(Princeton University).
Lusardi, A. (1996), "Permanent income, current income and consumption: evidence from two panel data
sets", Journal of Business Economics and Statistics 14:81-90.
Ch. 11:
Consumption
811
Mace, B.J. (1991), "Full insurance in the presence of aggregate uncertainty", Journal of Political Economy
99:928-956.
MaCurdy, T.E. (1981), "An intertemporal model of portfolio choice and human capital accumulation
under uncertainty with extensions incorporating taxes, consumer durables, imperfections in capital
markets and non separable preferences", Working Paper E-81-18 (Hoover Institution).
MaCurdy, T.E. (1983), "A simple scheme for estimating an intertemporal model of labor supply and
consumption in the presence of taxes and uncertainty", International Economic Review 24:265-289.
MaCurdy, T.E., and T.A. Mroz (1989), "Measuring macroeconomic shifts in wages from cohort
specifications", mimeograph (Stanford University).
Manldw, N.G. (1982), "Hall's consumption hypothesis and durable goods", Journal of Monetary
Economics 10:417-425.
Mankiw, N.G., and S.P. Zeldes (1994), "The consumption of stock owners and the consumption of non
stock owners", Journal of Financial Economics.
Manldw, N.G., J.J. Rotemberg and L.H. Summers (1985), "Intertemporal substitution in macroeconomics",
Quarterly Journal of Economics 100:225-253.
Meghir, C., and G. Weber (1996), "Intertemporal non-separability or borrowing restrictions?
A disaggregated analysis using a U.S. consumption panel", Econometrica 64:1151 1181.
Modigliani, E (1988), "The role of international transfers and life cycle saving in the accumulation of
wealth", Journal of Economic Perspectives 2:15-40.
Modigliani, E, and A. Ando (1963), "The life cycle hypothesis of saving: aggregate implications and
tests", American Economic Review 53:55-84.
Modigliani, E, and R. Brumberg (1954), "Utility analysis and the consumption function: an interpretation
of cross-section data", in: K.K. Kufihara, ed., Post-Keynesian Economics (Rutgers University Press,
New Brunswick, NJ) 128-197.
Moffitt, R. (1993), "Identification and estimation of dynamic models with a time series of repeated
cross-sections", Journal of Econometrics 59:99-124.
Novales, A. (1990), "Solving nonlinear rational expectations models: a stochastic equilibrium model of
interest rates", Econometrica 58:93-111.
O'Brien, A.M., and C.B. Hawley (1986), "The labor force participation behavior of married women
under conditions of constraints on borrowing", Journal of Human Resources 21:267-278.
Paulin, G., M. Boyle, R. Branch and R. Cage (1990), "Comparison of 1987-1988 CE integrated survey
and PCE annual estimates", mimeograph (BLS).
Phelan, C., and R.M. Townsend (1991), "Computing multi-period information constrained optima",
Review of Economic Studies 58:853-881.
Pischke, J.-S. (1995), "Individual income, incomplete information, and aggregate consumption",
Econometrica 63(4, July):805-840.
Pissarides, C.A. (1978), "Liquidity considerations in the theory of consumption", Quarterly Journal of
Economics 92:279-296.
Pollak, R.A. (1970), "Habit formation and dynamic demand functions", Journal of Political Economy,
78:745-763.
Runlde, D.E. (1991), "Liquidity constraints and the pernaanent-income hypothesis: evidence from panel
data", Journal of Monetary Economics 27:73-98.
Sargent, T.J. (1978), "Rational expectations, econometric exogeneity and consumption", Journal of
Political Economy 86:673-700.
Seslnick, D. (1992), "Aggregate consumption and saving in the postwar United States", Review of
Economics and Statistics 585-597.
Seslnick, D. (1994), "Are our data relevant to the theory? The case of aggregate consumption",
mimeograph (Rice University).
Shea, J. (1995), "Union contracts and the life-cycle/permanent- income hypothesis", American Economic
Review 85:186-200.
812
O.P. Attanasio
Skinner, J.S. (1988), "Risky income. Life cycle consumption and precautionary savings", Journal of
Monetary Economics 22:237-255.
Spynnewin, E (1981), "Rational habit formation", European Economic Review 15:91-109.
Thurow, L. (1969), "The optimum lifetime distribution of consumption expenditures", American
Economic Review 59:371-396.
Tobin, J., and W.C. Dolde (1971), "Wealth, liquidity and consumption", Reserve Bank of Boston,
Monetary Conference Series No. 5.
Townsend, R.M. (1994), "Risk and insurance in village India", Econometrica 62:539-592.
Weber, G. (1993), "Earnings related borrowing restrictions: empirical evidence from a pseudo panel for
the UK", Annales d'Economie et de Statistique 29:157-173.
Wilson, R. (1968), "The theory of syndicates", Econometrica 36:119-132.
Zeldes, S.P. (1989a), "Consumption and liquidity constraints: an empirical analysis", Journal of Political
Economy 97:305-346.
Zeldes, S.E (1989b), "Optimal consumption with stochastic income", Quarterly Journal of Economics
104:275-298.
Chapter 12
AGGREGATE INVESTMENT
RICARDO J. CABALLERO
MIT and NBER
Contents
Abstract
Keywords
1. Introduction
2. Basic investment t h e o r y and findings
2.1. Pre 1990s: Dismay
2.2. "Econometrics": Cost of capital and q matter
2.2.1. Long-run
2.2.2. Short-run
3. L u m p y and irreversible investment
3.1. Plant/firm level
3.1.1. Microeconomic adjustment: characterization
3.1.2. "Representative" problem
3.1.3. A detour: q-theory and infrequent investment
3.1.3.1. On the fragility of marginal q
3.1.3.2. When does q-theory work?
3.1.3.3. Taking stock
3.1.4. Another detour: Several misconceptions about irreversible investment
3.1.5. Adjustment hazard
3.2. Aggregation
3.3. Empirical evidence
.
.
.
.
.
3.3.1. Microeconnmic data
3.3.2. Aggregate data
3.3.3. Pent-up demand
3.4. Equilibrium
4.
814
814
815
816
816
818
820
821
822
822
823
824
828
828
830
832
832
835
837
838
838
839
841
842
844
844
847
* I am grateful to Andrew Abel, Steve Bergantino, Olivier Blanchard, Jason Cummins, Esther Duffio,
Eduardo Engel, Austan Goolsbee, Luigi Guiso, Kevin Hassett, Glenn Hubbard, John Leahy, Kenneth
West, and Michael Woodford for many useful comments. I thank the NSF for financial support.
Handbook of Macroeconomics, Volume 1, Edited by JB. Taylor and M. WoodJbrd~
1999 Elsevier Science B.V. All rights reserved
813
814
5. Inefficient investment
5.1. Informationalproblems
5.2. Specificityand opportunism
5.2.1. Genericimplications
5.2.2. Credit constraints
6. Conclusion and outlook
References
R.J. Caballero
848
849
851
854
856
857
858
Abstract
The 1990s have witnessed a revival in economists' interest and hope of explaining
aggregate and microeconomic investment behavior. New theories, better econometric
procedures, and more detailed panel data sets are behind this movement. Much of the
progress has occurred at the level of microeconomic theories and evidence; however,
progress in aggregation and general equilibrium aspects of the investment problem also
has been significant. The concept of sunk costs is at the center of modern theories. The
implications of these costs for investment go well beyond the neoclassical response to
the irreversible-technological friction they represent, for they can also lead to firstorder inefficiencies when interacting with informational and contractual problems.
Keywords
sunk costs, irreversible investment, (S, s) models, adjustment hazard, aggregation,
non-linearities, private information, incomplete contracts, pent-up investment
815
1. Introduction
Aggregate investment is an important topic. Cotmtries and firms are often judged by
their performance along this dimension, since investment is viewed as providing hope
for future prosperity. It is not surprising, therefore, that much has been written about
investment. It is even less surprising that many surveys, and surveys of surveys, already
exist ~. Rather than surveying the surveys of surveys, as one would expect from a
handbook chapter, I have chosen to focus most of my discussion on that which is
relatively new. The cost of this, of course, is that most of the theories I will discuss
have not yet passed the test of time and are often only half the distance toward full
development.
Most, but not all, of the subjects I plan to discuss relate directly to investment
in equipment and structures. Investing means trading the present for the future; as
is the case, for example, when a firm purchases equipment, builds structures, trains
its workers, restructures production, spends resources on R&D, hoards labor during a
recession; or when a worker leaves a job to search for another one, invests in human
capital; or when a country undergoes a structural adjustment, a trade liberalization or
a fiscal reform. The more theoretical sections of this survey apply to most of these
examples. Further, except for specific empirical results, a large part of the discussion
about equipment and structures also applies to other forms of investment.
The style of this review article is mostly empirical in early sections and mostly
theoretical in later ones. This ordering is highly correlated with the chronology of
research on investment. It follows that I am implicitly advocating for more empirical
work on the newest theories.
The layout of the chapter is as follows. Section 2 is rather traditional in content.
It describes the basic investment theory and findings, taking the view that the pre
1990s empirical literature was in disarray with respect to finding a role for the cost
of capital in investment equations. During the 1990s, however, we have learned from
long run relationships and "natural experiments" that the cost of capital does indeed
have significant effects on investment, although it is probably not the most important
explanatory variable. Neither, I should add, is measured q.
Section 3 describes what has been well known but largely ignored until recently:
that microeconomic investment is lumpy and mostly sunk. It turns out that changes
in the degree of coordination of lumpy actions play an important role in shaping the
dynamic behavior of aggregate investment. The old concept of pent-up demand is back.
This section contains a more detailed description of models and techniques than the
others. It also attempts to clarify several misconceptions about the implications of these
models.
Section 4 is about equilibrium interactions and scrapping. It describes the consequences of free entry and different assumptions about the elasticity of the supply
1 See, for example, Chirinko (1993), Hassett and Hubbard (1996b), for excellent surveys of traditional
investment equations.
816
R.J. Caballero
Since very early on, economists have attempted to explain investment behavior using
both scale and relative price variables, and since very early on, the former have been
more successful than the latter.
One of the first "theories" of investment was the accelerator model [Clark (1917)].
Scarcely a theory, the accelerator model is derived by inverting a simple fixed
proportion production function and taking first differences. Unable to account for the
serial correlation of investment beyond that o f output growth, this model was soon
transformed into the flexible accelerator model [Clark (1944), Koyck (1954)]:
n
I, = Z/3~AKT-r'
(2.1)
"~-0
where I denotes investment, the/3r's are distributed lag parameters, and K* is the
desired, as opposed to actual, level of capital. In the simple fixed proportions world,
K* can be written as a linear function of the output level, Y:
K* = a Y
where a is a parameter.
817
The absence of prices (the cost o f capital, in particular) from the right-hand side of
the flexible accelerator equation has earned it disrespect despite its empirical success.
Jorgenson's (1963) neoclassical theory o f investment intended to remedy this situation.
Starting from the optimization problem o f a perfectly competitive firm facing no
adjustment costs, myopic expectations, and constant returns Cobb-Douglas technology,
Jorgenson obtained the standard static first-order condition:
K = aY/Ck,
where Ck stands for the cost of capital and a is now the share o f capital in a simple
Cobb-Douglas production function. As with the accelerator model, this model was
unable to account for the serial correlation o f investment, and so gave way to the
flexible neoclassical model o f Hall and Jorgenson (1967), where
K* = aY/Ck,
(2.2)
818
R.J. Caballero
The elegant theoretical contributions o f Abel (1979) and Hayashi (1982) connected
the existing theories and partial theories. They showed that the neoclassical model
with convex adjustment costs yields a q-model. This q, known as marginal q,
should be interpreted as the marginal value o f an installed unit o f capital, which
corresponds to the shadow value o f a trait o f capital in the firm's optimization problem.
Further, Hayashi showed that (for price-taking firms) when the production function and
adjustment cost function are linearly homogeneous in capital and labor, marginal and
average q are equal. This is an important result from an empirical standpoint because
marginal q is unobservable to the econometrician whereas average q is, in principle,
observable to the econometrician 6.
Soon, however, the q-model, along with expanded and ad-hoc "flexible-q" models
(i.e. with additional lags o f q on the right-hand side), joined models based on the cost
o f capital in their lack o f empirical success. Scale variables such as cash-flows always
seemed to matter more in investment equations than q which, in principle, should have
been a sufficient statistic 7.
Figure 2.1, which reproduces figures 1 and 3 o f Hassett and Hubbard (1996b), helps
us understand the statistical reasons for the problem. The bottom line is clear: In
aggregate US data (which is probably representative o f many other data sets for this
purpose) the unconditional correlation between cost o f capital and investment is low,
and so is that between average q and investment. On the other hand, cash flows and
sales's growth closely track aggregate investment.
The 1980s discontent with respect to investment equations is probably well captured
in Blanchard's (1986) discussion o f Shapiro's (1986) investment paper at Brookings:
"... it is well known that to get the user cost to appear at all in the investment equation, one
has to display more than the usual amount of econometric ingenuity, resorting most of the time
to choosing a specification that simply forces the effect to be there ..." [my emphasis]
Today, the first emphasized statement still holds, but the second one probably does
not. This takes me to the next subsection.
2.2. "Econometrics": Cost o f capital a n d q matter
Econometric "ingenuity" eventually pays off, although this often means isolating that
part o f the relationship which conforms with the theory, rather than explaining a
6 I have mixed feelings about this equivalence result, however. Not about its theoretical derivation,
which is elegant and useful; rather about its abuse in empirical work. Too often, it is used to justify
substituting average for marginal q on the right-hand side of investment equations, even though the
assumptions required for the equivalence between the two are not nearly satisfied in the industry or
firms studied (e.g. Compustat). This does not mean that average q should not be used, but it says that
we should not pretend that the foundation for its use is beyond the basic intuition provided by Tobin
(1969), and that the additional properties that hold for marginal q are to be expected from average q
(e.g. sufficiency).
7 Fazzari et al. (1988) started a large literature documenting the role of these variables, even after
conditioning for average q. I will return to the interpretation of these regressions later in the survey.
819
o
-
0
o
cq
"/"
:'.::::::5 :
~:~
--
.~- ~ ~
ixl
:+:+:.:.:.:.~
. . . . ........................
.:.:.:::.:::::~ .~:::::::::::::::::::::::::::,
k
]~.
v,
o
. ..
!~!!!iiii:: ;
i!: ~
::i ~
~
o
I g :i
,'~
"
e~
820
R.J Caballero
substantial fraction of the movements of the left-hand side variable, or even relating
a significant fraction of the volatility of the right-hand side variables to that of the
left-hand side variable. In my view, this is the type of payoff obtained from the recent
incarnations of the "traditional" line. Still, it is progress.
Going from less to more ambitious, there are two generic developments I wish
to discuss. First, ignoring high and medium frequency variations, we have come to
the realization that the low frequency aspects of the data are not inconsistent with
theories that assign an important role to the cost o f capital in determining the rate of
capital accumulation. Second, there are distinctive episodes during which changes in
the cost of capital are sufficiently dramatic that it becomes possible to demonstrate the
importance of the cost of capital at higher frequencies as well.
Other recent developments within the traditional line include the use of Euler
equation procedures. In my view, and unlike the case in basic finance and consumption
applications, these procedures are a form of morphine rather than a remedy: their lack
of statistical power allows us to sometimes not see the problem. Since my goal is to
discuss progress, I will skip results obtained with these procedures s.
2.2.1. L o n g - r u n
Many of the problems with investment equations have to do with the presence of
complex and not well understood dynamic issues (more on this in the next sections).
From early on, researchers have found it useful to think about investment in two
steps: first, derive some simple expression for a "target" stock of capital, which I have
called K* here; and second, model dynamics as a, possibly complex, function of
contemporaneous and lagged changes in K*. It seems sensible, therefore, to start by
asking whether the first step resembles what we expect before going into the difficult
issues of timing.
Taking logs on each side of Equation (2.2), disregarding constants, and relaxing the
unit elasticity constraint on the cost of capital, yields
k* - y
= yck,
(2.3)
where lower case letters denote logarithms and y is the parameter of interest.
This expression cannot be estimated, of course, because k* is not observed. There
is a simple argument based on cointegration, or a close small sample "cousin," which
allows us to get around this observability problem, however. The whole purpose of
deriving k* is to then model k as trying to keep pace with it. Thus, differences
between these two variables should only be transitory (up to constants). I f k* and k are
See Oliner, Rudebusch and Sichel (1995) for a damaging evaluation of the statistical properties of
these procedures.
821
sufficiently volatile (ideally with unit roots, in large samples), then we can "ignore"
the discrepancy between these two variables in estimating y. Let
k = k* + e,
with c a stationary residual that captures transitory discrepancies between the two
variables due to adjustment costs. Substituting this expression into Equation (2.3)
yields an equation that can be estimated:
k - y = yck + c.
(2.4)
Estimating this equation by OLS (the simplest o f the cointegration procedures) yields,
for aggregate US data, an estimate o f y o f -0.4; significantly different from zero 9.
We can do better, however. In any small sample, the cointegration argument will not
take its full bite, and the estimates o f ~/will be affected by the correlation between regressors and e. Caballero (1994a) argues that this is particularly serious and systematic
in models with slow adjustment (e.g. due to adjustment costs). The intuition behind
this idea is simple. A partial adjustment mechanism implies that, in any finite sample,
the variance o f K/Y ought to be less than the variance o f K * / Y , which means the lefthand side o f Equation (2.4) ought to be less volatile than the right-hand side of Equation (2.3), or yck lo. However, by the normal equations o f OLS, the estimated counterparts o f yck and e on the right-hand side o f Equation (2.4) must be orthogonal, so that
the variance o f k - y is greater than the variance o f )ck, which is equal to the variance
o f the estimated '* - y . Since this inequality is in contradiction with what is implied by
adjustment cost mechanisms, we conclude that the estimate o f y is biased toward zero.
Using Monte Carlo simulations, I showed in that paper that this bias can be
substantial, and then proceeded to correct it using Stock and Watson's (1993)
procedure. I obtained an estimate o f ~/close to minus one, very near the neoclassical
benchmark 11.
2.2.2. Short-run
9 This estimate of g was obtained using US quarterly NIPA data for the period 1957:1-1987:4. Capital
corresponds to equipment capital and cost of capital is constructed as in Auerbach and Hassett (1992).
l0 Note that if adjustment costs are non-convex it is possible, at the microeconomic level, and in a
sufficiently short sample, to have these relative volatilities reversed. This is not an issue for the aggregate
data results discussed here. See the next section for more on non-convex adjustment cost models.
11 Similar estimates were obtained by Bertola and Caballero (1994) and Caballero, Engel and
Haltiwanger (1995) with different data sets.
822
R.J Caballero
experiments, such as periods of tax reform, which present the econometrician with
more accurate measures of (often substantial) changes in the cost of capital and q.
Measures of q, for example, are not only very noisy because of the substitution of
average q for marginal q, but also because there may be substantial "non-fundamental"
movements in the value of firms, making average q mismeasured as well. However,
there are certain episodes (e.g., periods of tax reform) when the movements in q are
likely to be large, in a predictable direction, and for the "right reasons". As with
cointegration, during those episodes problems with the residual can be more or less
disregarded.
The movement from aggregate to microeconomic data, by itself, has not done
much to improve affairs. Although microeconomic data has improved precision,
coefficients on the cost of capital and q in investment equations have remained
embarrassingly small. Combined with the use of natural experiments, however,
emphasis on microeconomic data has had much higher payoffs. The work of Cummins,
Hassett and Hubbard (1994, 1996a) is salient in this regard. They isolate periods
with important tax reforms and find that the coefficient on q is much larger in those
episodes. Most recently, using firm level data for 14 developed countries, they find that
while using standard instrumental variable procedures yields coefficients on q which
range from 0.03 to 0.1, when contemporaneous tax reforms are included among the
instruments, the estimates jump to a range between 0.09 to 0.8, with median and mean
not very far from 0.5. In the USA, for example, the estimate of the coefficient on q
jumps from 0.048 to 0.6512
Although these empirical results represent significant progress, there is still plenty
of work needed to retrace the steps back to the aggregate and we must not forget that
a substantial component of the variation in aggregate and microeconomic investment
remains unexplained. The next sections describe progress on both fronts.
The most basic form of friction occurs at the level of microeconomic units, and goes
under the general heading of adjustment costs.
12 See Caballero (1994b) for a discussion of their results, interpretations and procedures.
823
There are essentially three basic types o f adjustments observed at the establishment
level: (a) ongoing frictionless flow (maintenance); (b) gradual adjustments (e.g. refinements and training dependent improvements); (c) major and infrequent adjustments 13
The structural literature o f the 1980s and before, based explicitly or implicitly on
convex adjustment cost models (the quadratic adjustment cost model, in particular)
dealt with (a) and (b). The implicit "hope" was that the smoothness brought about
by aggregation would make disregarding the importance o f infrequent adjustments
for individual units, unimportant for aggregate phenomena. Instead, the idea was to
derive aggregate investment equations as coming from the solution to the optimization
problem of a fictitious agent facing adjustment costs which only led to smooth
adjustments o f type (a) and (b). Many authors disagreed with this strategy [e.g.
Rothschild (1971)]; but for most the relative simplicity of the quadratic model was
too enticing to resist.
A combination of factors eventually led economists to revisit and reevaluate some
o f the shortcuts which were in widespread use by the end o f the 1980S 14. First,
there was frustration with the disappointing empirical results described above. Second,
techniques which could handle models of lumpy investment became part of the modern
economist's tool kit. And third, microeconomic data made the obvious even more
apparent: microeconomic investment is extremely lumpy, and this lumpiness is unlikely
to fully "wash out" at the aggregate level.
The work o f Doms and Dunne (1993) was instrumental in stressing the last point.
They documented investment patterns o f 12 000 plants in US manufacturing over the 17
year period, 1972-1989. Their findings are many, of which I have chosen to emphasize
those that are most closely related to the purpose o f this survey.
For each establishment, Doms and Dunne constructed a series o f the proportion o f
the total equipment investment o f the establishment (over the 17-year period) made
in each year. They found that on average the largest investment episode accounts for
more than 25 percent o f the 17 year investment o f an establishment and that more than
half o f the establishments exhibited capital growth close to 50 percent in a single year.
They also note that the second largest investment spike often came next to the largest
investment spike (right before or right after) suggesting that both spikes correspond to
a single investment episode. 15 Combining the two primary spikes, they find that nearly
R.J Caballero
824
0 < ~ < 1,
(3.1)
where K is the firm's stock of capital; 0 is a profitability index that combines demand,
productivity and wage shocks; r is the discount rate; and 7 represents the elasticity
16 Cooper, Haltiwanger and Power (1994) go one step further in characterizing infrequent lumpiness.
Using a data set similar to that of Doms and Dtmne, they show that the probabilityof a firm experiencing
a major investment spike is increasing in the time since the last major spike. This feature of the data is
highly consistent with the implications of the models reviewed later in this section.
17 Where aggregate investmentcorrespondsto the investmentof all the establishmentsin their sample.
18 Dixit (1993) provides an excellent discussion of the basic problem and the mathematicaltechniques
needed to solve it.
825
o f gross profits with respect to capital. It is less than one as long as the firm exhibits
some degree o f decreasing returns or market power, which ! assume to be the case.
For convenience, capital does not depreciate.
It will also facilitate things to assume that increments in the logarithm o f 0 are
i.i.d., and that time and the sample paths of 0 are "ahnost" continuous (i.e. At is
small and changes in the value of 0 over an interval o f time At are small). I make
these assumptions so I can, informally, use all the convenience o f Ito's lemma and
Brownian motions. I choose to depart from strict continuous time, on the other hand,
because discrete time will allow me to present this section in a more ratified manner.
As in the previous section, we can find an expression for the static optimum of the
stock o f capital, or "desired" capital:
I
I-y
K* = argmaxxH(K, 0) =
(3.2)
K
K*
'
also inherits the geometric random walk process, for any given K. Substituting this
expression into Equation (3.1) and using Equation (3.2) to solve for 0 yields
0 < 7 < 1.
(3.3)
In order to generate infrequent actions, the cost of adjusting the stock o f capital must
increase sharply around the point of no adjustment. A cost proportional to the size of
adjustment is enough to do so. Lumpiness requires a little more, for there must be an
advantage in bunching adjustment; increasing returns in the adjustment technology is
the standard recipe, o f which a fixed cost is the simplest. Let C(rI, K*) denote the cost
o f adjusting the capital stock by K ' t / :
C(rI, K . ) = { A K ~ e O } K .
(cf+c~rl
/ cf-Cprl
if
if
t/>0,
t/<0,
(3.4)
where the K* term ensures that the relative importance of adjustment costs remains
unchanged over time 19. Figure 3.1 illustrates an example of C(., .)/K*.
19 This goal would also be accomplished by K, but K* yields slightly simpler mathematical expressions
at a low cost in terms of substantive issues.
Also, I have allowed proportional costs to differ with respect to upward and downward adjustments in
order to talk later about the irreversible investment case; for this purpose, I could have done it equally
well through asymmetric fixed costs. Allowing for both forms of asymmetries simultaneously is a trivial
but uninteresting extension.
826
R.J. Caballero
\
J
Cf
<
)rl
The problem of the firm can be characterized in terms of two functions of Z and
K*: V(Z,K*) and ~'(Z, K*). The function V(Z,K*) represents the value of a firm with
imbalance Z and desired capital K* if it does not adjust in this period, and V(Z,K*)
is the value of the firm which can choose whether or not to adjust. Thus,
v ( z , , I,:?) = rt(z,, I<t)At + (1 - rat) Et [~(Zt+A~, K,~At)],
(3.5)
(3.6)
and:
The nature of the solution of this problem is now intuitive. Given the function
V(Z, K*), Equation (3.6) provides most of what is needed to characterize the solution.
First, since C is positive even for small adjustments, it is apparent that when Z is
near that value for which V(Z,K*) is maximized, the first term on the right-hand side
of Equation (3.6) is larger than the second term; that is, there is a range of inaction.
Second, since both adjustment costs and the profit function are homogeneous of degree
one with respect to K*, so are V and ~'. Thus, it is possible to fully characterize the
solution in the space of imbalances, Z. Among other things, this implies that the range
of inaction described before, is fixed in the space of Z. Let L denote the minimum value
of Z for which there is no investment, and U the maximum value for which there is no
disinvestment; thus the range of inaction is (L, U). Third, conditional on adjustment,
changes must not only be large enough to justify incurring the fixed cost, but also the
(invariant) target points must satisfy
Vz(/) = Cp
(3.7)
and
Vz(u) = -Cp,
where Vz is the derivative of
(3.8)
827
el
I ..../........
V ( z, K*) / K*
i"
<
>
L
Vz(L) = Cp
(3.9)
and
v z ( u ) = -Cp,
(3.10)
(3.11)
(3.12)
which simply say that since the investment rule (optimal or not[) dictates that once
a trigger point is reached, adjustment must occur at once, the only difference in the
value of being at trigger and target points must be the adjustment cost of moving from
the former to the latter.
Figure 3.2 illustrates the value function. Smooth pasting says that the tangents at
L and l have slope Cp, while those at U and u have slope -Cp. Value matching says
828
R.J. Caballero
that the value function evaluated at the target minus the value function evaluated at
the trigger point is equal to the variable cost paid at adjustment plus the fixed cost (all
these normalized by K* in the figure).
There are a few particular cases which are worth highlighting because they appear
often in the literature:
(1) If there is no variable cost of investment, once the adjustment decision has
been taken, adjustment from both sides is complete since the marginal cost of
adjustment is zero. Thus, the (L, l, u, U) rule reduces to an (L, c, U) rule, where
c is the common target for investment as well as disinvestment, and is that value
which maximizes V(Z, K*) for any K* 2o.
(2) If there are variable costs but no fixed cost, there is no reason for adjustment
to be lumpy, for there are no increasing returns in the adjustment technology 21.
Once the boundaries of the inaction range, L and U, are reached, the firm adjusts
just enough to avoid crossing outside the inaction range; that is L and U become
reflecting barriers.
(3) If there is a large (not necessarily infinite) cost to disinvestment, then investment
becomes irreversible. In the absence of investment costs, the investment rule
reduces to a single reflecting barrier L, which is to the left of one (reluctance
to invest). This is the standard irreversible investment case.
3.1.3.1. On the fragility of marginal q. It is apparent from the lack of global concavity
o f the value function in Figure 3.2, that traditional q-theory is not likely to work in the
presence o f jumps. Caballero and Leahy (1996) develop the argument in detail, which
I summarize below.
20 Note that in general c ~ 1. That is, the optimal dynamic target is generally different from the static
one.
21 And we have already assumed that shocks are "small" in any given At.
829
(a)
lqw...-....-..---.---.....~
J ~ ~ q m
1+ C+p
S
qm(z)
il-c<
(b)
qm
1 + C+p
1 - c-p
>
<
U=u
L=I
(3.13)
Figure 3.3a plots qM against the imbalance m e a s u r e Z 23. Smooth pasting implies that
qM must be the same at trigger and target points (because Vz must be the same at
trigger and target points); if there are jumps, these are points very far apart in state
830
R.J. Caballero
space. Two points with the same value of qM lead to very different levels of investment
(zero and large). Moreover, since the value function becomes linear outside the inaction
range, all points outside the inaction range (on the same side) have the same q~t, and
all of them lead to different levels of investment. It is apparent, therefore, that the
function mapping qM into investment no longer exists. Worse, in between trigger and
target points, the relation between qM and Z is not even monotonic.
What is happening? Marginal q is the expected present value of the marginal
profitability o f capital. Far from an adjustment point, it behaves as usual with respect
to the state o f the firm: if conditions improve, future marginal profitability of capital
rises, and so does q~. Close to the investment point, on the other hand, the effect of
a change in the state of the firm over the probability o f a large amount of investment
in the near future dominates. An abrupt increase in the stock of capital brings about
an abrupt decline in the marginal profitability o f capital as long as the profit function
is concave with respect to capital 24. Thus, an improvement in the state of the firm
makes it very likely that it invests in the near future, reducing the expected marginal
profitability o f capital in the near future, thus lowering the value of an extra unit of
installed capital.
Caballero and Leahy (1996) show that adding a convex adjustment cost to the
problem does not change the basic intuition of the mechanism described above. They
also show, somewhat paradoxically, that average q, which is often thought o f as a
convenient albeit inappropriate proxy for marginal q, turns out to be a good predictor
of investment even in the presence of fixed costs, although it is no longer a sufficient
statistic, except for very special assumptions about the stochastic nature of driving
forces.
3.1.3.2. When does q-theory work?. The failure o f q-theory described above is rooted
in the presence of increasing returns in the adjustment cost function (3.4). This feature
of the adjustment technology is responsible for the loss of global concavity of the value
function, which is behind the non-monotonicity o f marginal q 25.
Monotonicity of qM inside the inaction range is recovered by dropping the fixed cost
from Equation (3.4), as was done in cases 2 and 3 in Section 3.1.2. Figure 3.3b portrays
this scenario. Adjustment at the trigger points no longer involves large projects, thus
proximity to these triggers no longer signal the sharp changes in future marginal
profitability of capital which were responsible for the "anomalous" behavior of qM 26
There is still the issue that in the (very) rare event that a firm finds itself outside the
inaction range it will adjust immediately to the trigger, at a constant marginal cost, so
831
different levels of investment are consistent with the same value o f qM. This is easily
remedied by adding a convex component to the adjustment cost function27:
c(.,K*) =
> 1.
(3.14)
This is essentially what Abel and Eberly (1994) do 28. Absent the advantage of lumping
adjustment brought about by the presence o f fixed costs, standard q-theory is recovered
whenever the firm invests. Provided adjustment takes place, the firm equalizes the
marginal benefit of adjustment and the marginal cost o f investing, which is now an
increasing function o f adjustment:
qM = 1 + sgn(r/)
(Cp +/3Cqlr/l~-l) ,
27 Which, at the same time, makes transitions outside the inaction range less rare.
28 Needless to say, it is trivial to add asymmetries to the adjustment cost function. But that is beside
the point of this section.
29 Alternatively, if one assumes perfect competition and constant returns to scale, the profit function
becomes linear with respect to capital (if the other factors of production can be adjusted at will), so
changes in investment do not feed back into q. In this extreme case, the modified (i.e. with an inaction
range) q-theory works well even in the presence of traditional fixed costs.
832
R.J Caballero
range relevant for different types of investments could explain the negative BarnettSakellaris finding.
3.1.3.3. Taking stock. One may be inclined to conclude from this section that before
going ahead with q-theory one should check whether investment literally exhibits
jumps or not. This is not the lesson I draw, however.
For once, this is not right. It is not difficult to add a time to build mechanism such
that a lumpy project is decomposed into a fairly smooth flow, without altering the
argument of why marginal q fails in the presence of fixed costs. But more importantly,
I suspect the main lesson is one of modesty. I doubt that researchers will often find the
required data and/or patience to determine whether one scenario or the other holds. In
this case, we might as well acknowledge that the relationship between marginal q and
investment is not robust, and that average q is unlikely to be a sufficient statistic for
investment. Of course it is important to include variables that capture knowledge of
the future on the right-hand side of investment equations, but we should avoid reading
"too much" from these regressions.
3.1.4. Another detour." Several misconceptions about irreuersible investment
As I mentioned before, when describing the special case of irreversible investment,
the regulation barrier, L, is to the left of one. That is, investment occurs only when the
stock of capital is substantially below the frictionless stock of capital. Alternatively,
investment occurs when the marginal profitability of capital is substantially above the
cost of capital. This is the famous "reluctance to invest" result.
There are several misconceptions about the implications of this "reluctance" result.
I will mention three of them. It is often said that, (a) reluctance implies that, in the
presence of irreversibility, the firm accumulates less capital; (b) since reluctance rises
with uncertainty (the regulation point moves further to the left), more uncertainty
implies less capital; and (c) standard present value techniques are inappropriate because
reluctance reflects the value of the "option to wait" for more information before
irreversibly sinking resources and this is not taken into account by the standard
formulae.
In order to show the fallacious nature of the first statement, it is useful to go back
to our canonical problem and simulate the path o f the (log of) stock of capital of a
firm facing no irreversibility constraint. Panel (a) in Figure 3.4 does so for a random
realization of the path of 0. Panel (b) in the figure shows the corresponding path of
the marginal profitability of capital, which is equal to the constant - frictionless - cost
of capital, r 3.
Imagine now imposing an irreversibility constraint on the firm, but assume that the
firm does not modify its "frictionless" investment rule whenever it can invest. This is
833
(b)
[I
~r
~L
IlL
o;
~L
~o
s
lo
rs
20
~o
40
4S
~0
t~ne
P~E~ ]
tlme
(c)
(d)
I~rg. Prof.
~o
,;
io ,',
tlme
'i
| .............
i
Is
i ............
I?'--
t
Io
(f)
Al
i
~
tlme
(e)
i
zo
r
zls
time
i
ao
i
~
- -
In Kf
- - -
In K d
--
InK
i
4o
tl
4s
c
--
c+h
Mar'g.
so
,_
.:
I0
~o
J
2'~
time
t
~
~
4o
Prof
i
45
"
so
834
R.J Caballero
portrayed in panel (c) of the figure. The solid and dashed lines represent the actual
and frictionless stocks o f capital, respectively. It is apparent that the firm would, on
average, have too much capital, for it would have the same stock o f capital in good
times, but too much in bad times. The counterpart o f this is in panel (d), which shows
that on average the marginal profitability of capital is below the cost o f capital.
Reluctance to invest in good times is an optimal response attempting to offset the
natural tendency to over-accumulate capital induced by the irreversibility constraint.
Panel (e) illustrates this point. The solid and dashed lines represent the same variables
as in panel (c), while the dotted line illustrates the target stock of capital when the firm
behaves optimally. The counterpart o f the negative value of ln(Ka/Kf ) is a positive
constant h in the marginal profitability o f capital required for investment to take place
(panel f). It is apparent that whether the stock o f capital is on average higher or lower
than without the irreversibility constraint is unclear; the firm has too little capital
during good times but too much during very bad times. A precise answer depends
on things about which we know little, and which may turn out to yield only secondorder effects 31.
It is now easy to see the fallacious nature of the second statement. More uncertainty
raises reluctance precisely because it raises the need to reduce the extent o f excessive
capital during the now deeper recessions. Without raising reluctance, an increase in
uncertainty would raise the average stock o f capital in the presence of irreversibility
constraints. This occurs because there would now be greater capital accumulation
during extremely good times which would not be offset by large disinvestment during
extremely bad times 32.
The third misunderstanding is o f a different nature. In my view, it is the result of
insightful but, unfortunately, abused language. First, what is right: there is nothing
mysterious about irreversibility constraints as a mathematical problem. Dynamic
programing works, in the same way it does with other, more traditional, adjustment
frictions. This means that present value formulae, using the correct calculation of
future marginal profitability of capital also work. O f course such calculations must
be performed along the optimal investment path, constraints included! What is wrong:
the standard analysis must be modified to consider the value of the "option to wait".
3a See Bertola (1992), Caballero (1993a), Bertola and Caballero (1994) for early discussions of this
issue and of the related uncertainty-investment misconception. More recently, Abel and Eberly (1996b)
have formalized these claims and made them more precise.
32 This does not mean that one cannot construct scenarios where an increase in uncertainty reduces
investment. For example, if there is an increase in perceived future uncertainty, the investment threshold
may jump today - i.e., before the variance of shocks does - resulting in an unambiguous decline in
investment.
Also, one should not confuse changes in uncertainty with changes in the probability of a bad event.
The latter links increases in uncertainty to a reduction in expected value, an entirely different and more
straightforward effect on investment. One can find traces of this confusion in the (informal) credibility
literature.
835
As we have seen, there is no need to do so. However, one may choose to follow
an alternative path, in which one starts by evaluating the future marginal profitability
o f capital without considering the effect o f future optimal investment decisions on
marginal profitability. This "mistake" can then be "corrected" with a term that has
an option representation. This alternative way o f doing things is akin to the arbitrage
approach in finance, and it was nicely portrayed in Pindyck (1988). The confusion
arises, in my view, from mixing the language in the two approaches 33.
A related claim exists for a once and for all project (as opposed to incremental
investment). It is said that the simple positive net present value rule used in business
schools to decide whether a project should be implemented does not hold because it
does not consider the option to wait and decide tomorrow, when more information
is available. Since I have never taught at a business school I cannot argue directly
against that claim. However, if the issue is whether to invest today or tomorrow, the
right criterion has never been invest if NPV is positive - at least that is what we
teach economics undergraduates. This is a case of mutually exclusive projects, thus
the fight criterion has always been to compare their net present value and take that
with the highest NPV, provided it is positive. If investment is irreversible, the project
invest tomorrow has a lower bound at zero (because investment will not occur if
N P V looks negative tomorrow), which the project invest today does not. Thus, other
things equal, irreversibility necessarily makes investing tomorrow more attractive than
investing today.
3.1.5. Adjustment hazard
At a qualitative level, the (L, l, u, U) models described above capture well the nonlinear
nature o f microeconomic adjustment. Maintenance expenditures aside, investment is
mostly sporadic and often lumpy; scarcely reacting to small changes in the environment
but abruptly undoing accumulated imbalances when they become sufficiently large, and
with possibly significant asymmetries between investment and disinvestment.
At an empirical level, however, these characterizations are too stark. For reasons,
some o f which we understand and most of which we do not, firms respond differently
to similar imbalances over time and across firms. Caballero and Engel (1999) propose
a probabilistic instead o f a deterministic adjustment rule. Rather than having a clear
demarcation between regions o f adjustment and inaction, they model a situation where
large imbalances are more likely to trigger adjustment than small ones 34.
33 See Bertola (1988) for one of the first discussions of this issue in the economics literatme. There
is also a related discussion in applied mathematics; see, for example, E1 Karoui and Karatzas (1991).
Abel et al. (1996) have recently revisited and expanded the discussion on the relation between the two
approaches.
34 Another advantage of this approach is that it nests linear models as the probability of adjustment
becomes independent of Z.
R.J. Caballero
836
There are many formal motivations for such an assumption. A particularly simple
one, pursued by Caballero and Engel (1999), is to assume that cf in the adjustment
cost fimction (3:4) is an i.i.d, random variable, both across firms and time. Although
technically more complex, the nature of the problem is not too different from that of
the simpler (L, l, u, U) model. Let ~o denote the random fixed cost, and G(w) its time
invariant distribution. It is possible to characterize the problem of the firm in terms
of two functions similar to those used before: V(Z, K*) and ~'(Z, K*, w), the value of
a firm with imbalance Z, desired capital K*, and realization of fixed adjustment cost
w. In particular, V(Z,K*) is the value o f the firm provided it does not adjust, while
[/(Z,K*, w) is the value o f the firm when it is left free to choose whether or not to
adjust. Thus,
V(Zt,K; ,~o)-max
(3.15)
(3.16)
Not surprisingly, the nature of the solution is not too different from that o f the
(L, l, u, U) case. Indeed, conditional on c~ it is an (L, l, u, U) rule, although there are
additional intertemporal considerations, since the firm weighs the likelihood of drawing
higher or lower adjustment costs in the future. Without conditioning on ~o, it is a
probabilistic rule in the space of imbalances.
In order to simplify the exposition, I will suppress the proportional costs. Thus,
conditional on adjustment, the target point is the same regardless of whether the firm
is adding or subtracting to its stock of capital (i.e. l = u = c). Moreover, let me define
a new imbalance index centered around zero:
x - In (Z/c).
The probability of adjustment rises with the absolute value of x because there are
more realizations of adjustment costs which justify adjustment. This is the sense in
which the (S, s) nature of the simpler models is preserved. Let A(x) denote the function
describing the probability of adjustment given x, and call it the adjustment hazard
function [see Caballero and Engel (1999)].
Given an imbalance x, it is no longer possible to say with certainty whether or not
the firm will adjust, but the expected investment by the firm is given by
E
11, ]
K/t x = (e ~ -
1)A(x) ~ - x A ( x ) ,
(3.17)
which is simply the product of the adjustment if it occurs, and the probability that
adjustment occurs 35. Aggregation is now only a step away.
35 Caballero and Engel (1999) refer to A(x) as the "effective hazard" to capture the idea that, through a
normalization, it also captures scenarios where adjustment, if it occurs, is only a fraction of the imbalance
X.
837
3.2. Aggregation
Unlike microeconomic data, aggregate investment series look fairly smooth. Large
microeconomic adjustments are far from being perfectly synchronized. The question
arises, and this was the maintained hypothesis during the 1980s, as to whether
aggregation eliminates all traces of lumpy microeconomic adjustment. The answer is
a clear no. Doms and Dunne's evidence on the role of synchronization of primary
spikes in accounting for aggregate investment, and on the high time series correlation
between aggregate investment and a Herfindahl index of microeconomic investments,
as well as the more structural empirical evidence reviewed in the next section, support
this conclusion,
With the setup at hand, aggregation proceeds in two easy steps. To simplify things
further, I will define the aggregate as the behavior of the average, rather than the
weighted average 36.
Both steps rely on having a large number of establishments, so that laws of
large numbers can be applied. In the first step, one takes as the average investment
rate (i.e. the ratio of investment to capital) of establishments with more or less
the same imbalance of capital, x, the conditional expectation of this ratio given in
Equation (3.17):
It ,~x = -xA(x),
(3.18)
g)
where the superscript x denotes the aggregate for plants with imbalance x.
The second step just requires averaging across all x. Let f(x, t) denote the cross
sectional density of establishments' capital imbalances just before investment takes
place at time t. Then the aggregate investment rate at time t, (It/Kt) A, is
( It ~A = - f xA(x)f(x,t)dx.
g)
(3.19)
This is an interesting equation, with macroeconomic data on the left and microeconomic data on the right-hand side. An example serves to illustrate this aspect of the
investment equation: If the adjustment hazard is quadratic,
_ 2X(3),
(3.20)
where Xt(1), Xt(2) and Xt(3) denote, respectively, the first, second and third moments of
the distribution of establishments' imbalances.
36 Using microeconomicdata, Caballero, Engel and Haltiwanger (1995) show that in US manufacturing
this approximationis. See Caballero and Engel (1999) for a detailed discussion of the issue.
838
R.J. Caballero
If t l l = ,~2 = 0, the model only has aggregate variables, both on the right and left hand side. Indeed this case corresponds to the celebrated partial adjustment model, and
it also coincides with the equation obtained from a quadratic adjustment cost model
with a representative agent [e.g. Rotemberg (1987) and Caballero and Engel (1999)]. If
either )!.1 or/~2 is different from zero, however, information about the cross sectional
distribution o f imbalances is needed on the right-hand side. All the microeconomic
models discussed in this section yield situations where higher moments o f the cross
sectional distribution play a role.
3.3. Empirical evidence
There are two polar empirical strategies used to estimate Equation (3.19), with a
continuum o f possibilities in between. At one extreme, one can use microeconomic
data to construct all the elements on the right-hand side; in particular one can construct
the path o f the cross sectional distribution and estimate the adjustment hazard as an
accounting identity, or estimate a parametric version o f it. At the other extreme, one
can attempt to learn about the adjustment hazard from aggregate data only, by putting
enough structure on the stochastic processes faced by firms and by starting with a
guess on the initial cross sectional distribution. Both avenues have been explored, with
similar results along dimensions they can be compared.
3.3.1. Microeconomic data
Caballero, Engel and Haltiwanger (1995) use information on approximately seven
thousand US manufacturing plants from 1972 to 1988 to empirically recreate the steps
described in the previous section 37. The figures below were constructed with data from
that paper 3s.
The procedure used by Caballero, Engel and Haltiwanger is essentially accounting,
except for the first step, which requires estimating a series of frictionless capital for
each establishment, and, from this, a m e a s u r e ofxit (an index of the capital imbalance
o f firm i at date t).
The series o f frictionless capital were constructed using a procedure similar to
that described in Section 2, but cointegration regressions were run at the individual
establishment level 39. The average estimate of the long run elasticity of capital with
37 As in Doms and Dunne (1993), we used data from the Longitudinal Research Datafile (LRD). The
LRD was created by longitudinally linking the establishment-level data from the Annual Survey of
Manufacturing. The data used in that paper is a subset of the LRD, representing all large, continuously
operating plants over the sample. The data sets include information on both investment and retirement
of equipment (i.e. the gross value of assets sold, retired, scrapped, etc.
38 Warning: x in that paper corresponds to ~ in this survey.
39 The results reported there constrained the coefficient on the elasticity of capital with respect to its
cost to be equal across two-digit sectors, but all principal results were robust to different constraints and
specifications.
839
0.4
0.3
0.2
0.1
I
-2
-1
o.o
respect to its cost was close to minus one, with substantial heterogeneity across sectors.
The measures ofxit, up to a constant, correspond to the difference between actual and
estimated frictionless capital4.
There are two results from that paper which seem particularly relevant for this
section of the survey. One on the shape o f the adjustment hazard, and the other on
the consequences o f this shape for aggregate dynamics. I discuss the former here
and the latter after the next subsection. Figure 3.5 reports the average adjustment
hazard constructed from simply averaging the investment rates o f establishments in
a small neighborhood o f each x, divided by minus the corresponding x. The hazard
is clearly increasing for positive adjustment (i.e. expected investment rises more
than proportionally with the shortage o f capital), as one would expect from the
nonlinearities implied by (L, 1, u, U) type models, and unlike the linear models which
imply a constant hazard. The estimated hazard is also very low for negative changes,
suggesting irreversibility 41 .
Following a similar procedure, Goolsbee and Gross (1997) have studied very
detailed and high quality microeconomic data on capital stock decisions in the US
airline industry. They found clear evidence of behavior consistent with non-convex
adjustment costs.
40 The establishment specific constants were estimated as the average gap between their respective kit
and k/t for the five points with investment closest to their median (broadly interpreted as maintenance
investment).
41 Retirements include assets sold, scrapped or retired. It is possible that observations are very noisy
on this side. The right-hand side of the figure should therefore be viewed with some caution.
840
R.J Caballero
The basic operations affecting the evolution o f f ( x , t) are quite simple. Given the
density, or histogram, at time t - 1, there are three basic operations in its transformation
into f ( x , t). First, aggregate shocks and common depreciation shift everybody's x in
the same direction; second, given the adjustment hazard, the density at each x is split
into those that stay there and those that adjust and move to some other position in
the state space (in the simplest case, they move to x = 0, but this is not necessary);
and third, idiosyncratic shocks hit, which amounts to a convolution o f the density
resulting from the second step and that o f idiosyncratic shocks. Making distributional
assumptions about idiosyncratic shocks and the initial cross sectional distribution,
is enough, therefore, to keep track o f the evolution o f the cross sectional density,
conditional on aggregate shocks and for a given adjustment hazard.
In continuous time, and assuming Brownian motions for aggregate and idiosyncratic
shocks, Bertola and Caballero (1994) estimated the irreversible investment model, and
Caballero (1993b) did so for the (L,/, u, U) model. 42
In discrete time but continuous state space, Caballero and Engel (1999), estimated
the more general adjustment hazard model described in the previous sections, assuming
that both idiosyncratic and aggregate shocks were generated by log-normal processes.
We did so for US manufacturing investment in equipment and structures (separately)
for the 1947-1992 period 43. The results were largely consistent with those found
with microeconomic data by Caballero, Engel and Haltiwanger (1995). There is clear
evidence o f an increasing hazard model; that is, the expected adjustment o f a firm
grows more than proportionally with its imbalance 44. An important point to note is that
since only aggregate data were used, these microeconomic nonlinearities must matter
at the aggregate level, for otherwise they would not be identified. The improvement
in the likelihood function from estimating this non-linear model rather than a simple
linear model (including the quadratic adjustment cost model) was highly significant,
and so was the improvement in the out-of-sample forecasting accuracy 45.
42 See Bertola and Caballero (1990) for a discrete time and space model and estimation procedure.
43 Another important difference between this and the previous papers is that estimation was done
by a single step maximum likelihood procedure, which did not require estimating frictionless capital
separately.
44 We did not allow for asymmetries between ups and downs but this turned out not to matter much
because given the strong drift induced by depreciation and the small value we found for the hazard in
an interval around zero, the model effectively behaves as if investment is irreversible (i.e. It is very
asymmetric around the median value of x and with a very small hazard for values of x much higher
than that.).
45 For within sample criteria, we ran Vuong's [Rivers and Vuong (1991)] test for non-nested models,
and we rejected strongly the hypothesis that both models (linear and non-linear) are equally close to
the true model against the hypothesis that the structural (non-linear) model is better. For out-of-sample
criteria, we dropped the last ten percent of the observations and evaluated the Mean Squared Error of
the one step ahead forecasts for these observations [see Caballero and Engel (1999)].
841
What is the aspect of the data that makes these models better than linear ones at
explaining aggregate investment dynamics?
The simplest answer comes from an example. Suppose that a history of mostly
positive aggregate shocks displaces the cross sectional distribution of imbalances
toward the high part of the hazard. Such a sequence of events will not only lead to more
investment along the path but also to more pent-up investment demand; indeed, the
cross sectional distribution represents unfulfilled investment plans. But as unfulfilled
demand "climbs" the hazard, more units are involved in responding to new shocks;
incremental investment demand is more easily boosted by further positive aggregate
shocks, or depressed by a turnabout of events. This time-varying/history-dependent
aggregate elasticity plays a very important role for aggregate investment dynamics. It
captures the aggregate impact of changes in the degree of synchronization of large
adjustments; already an important explanatory variable in Doms and Dunne's less
structural study. In particular, their observation that the Herfindahl of investment rises
during episodes of large aggregate investment matches well this mechanism.
Using the path of cross sectional distributions and hazards described at the beginning
of this subsection, Caballero, Engel and Haltiwanger (1995) found an important role
for the mechanism described above. Figure 3.6 depicts the relative contribution of the
time-varying aggregate elasticity for aggregate investment dynamics. A positive value
reflects an amplification effect (micro-nonlinearities exacerbate the economy's response
to aggregate shocks), while a negative value reflects an offsetting effect. The impact
of the time-varying elasticity appears to be especially large after the tax-reform of
1986 (when tax-incentives for investment were removed). The decline in investment
was 20 percent greater than it would have been under a linear model.
Fraction
0.15
0.10 /
0.05
0
- 0.05
-0.10
-0.15
I
1976
1978
1980
1982
1984
1986
842
R.J. Caballero
d
CD
LO
0
[___
c)
46 Note that just allowing for skewness and kurtosis in shocks, although it improves the performance of
linear models, is not nearly enough to make the linear model as good as the non-linear one. In Caballero
and Engel (1999) we compared the structural model with normal shocks (to the rate of growth of desired
capital) with a linear model which flexibly combined normal and log-normal shocks (which allows for
skewness and kurtosis). We found that Vuong's test still favored the non-linear model very clearly.
Moreover, in Caballero, Engel and Haltiwanger (1995) we found no evidence that would allow us to
reject the hypothesis that shocks have a normal distribution.
843
by firms and the dimension of the state space. It is this last observation which has
inhibited progress in constructing general equilibrium versions of these models. In
principle, the entire cross sectional distribution is needed to forecast future prices
faced by any particular firm, which means that actions today, and therefore equilibrium
determination, depend on these complex forecasts, and so on.
We are, however, beginning to see progress along this dimension. Much of this has
occurred in models with active extensive margins, and will be discussed in the next
sections, together with the reasons why the presence o f an extensive margin (entry
and/or exit) may facilitate rather than complicate the solution of the model.
However, there has also been recent progress along the lines of the intensive
margin models discussed up to now. Krieger (1997) embeds the heterogeneous agents
irreversible investment model of Bertola and Caballero (1994) into a more or less
standard Real Business Cycle model. He deals with the curse of dimensionality by
arguing that, except for very high frequency aspects of the data, expectations can be
well approximated by keeping track of a finite (and not too large) number of statistics of
the Fourier representation of the cross sectional distribution. I suspect that the quality
of this approximation is facilitated by the fact that, in Krieger's model, aggregate
shocks occur only infrequently. Nonetheless, I view his as an important step forward.
At this stage, the primary effect of general equilibrium is not surprising. It brings
important sources of aggregate convexity into the problem, smoothing further the
response of aggregate investment to aggregate shocks. How important are aggregate
sources of convexity? I suspect that, together with time to build considerations, they
are among the main sources of convexity in the short run. On one hand, we have
already presented substantial evidence on microeconomic lumpiness, which is largely
inconsistent with a dominant role for generalized convexity at the microeconomic level.
On the other, not only is it well known that estimated partial adjustment coefficients
grow with the degree of disaggregation of the data, but we also have direct evidence
on the importance of bottlenecks. Goolsbee (1995a) provides interesting evidence on
the latter. He exploits the variation across time and assets (capital) in investment tax
incentives, as instruments for short-run investment demand. He shows that the price
of assets is highly responsive to ITCs: A 10 percent increase in ITCs leads to an
average increase in the price of capital goods of about 6 percent. This price effect
slowly vanishes over the following three years 47.
Equilibrium considerations will play a central role in the sections that follow. In
particular, the issue of the elasticity of the supply of capital, generally interpreted, as
well as that of other bottlenecks will be revisited often.
47 In further work, Goolsbee (1997) concludes that an important fraction of the increase in short rnn
marginal cost is due to an increase in the wages of workers who produce capital goods. In the last part
of Section 5 I will discuss the connection between sunk investment and payments to complementary
factors.
Questioning the robustness of Goolsbee's(1995a) findings, Hassett and Hubbard (1996a), find evidence
of a positive effect of tax credits on prices of capital goods before 1975 but not after that.
R.J Caballero
844
Yt =
/o
Sit d i = N~,
(4.1)
where Nt is the measure of firms at time t. Given Art, the industry price is determined
from the demand equation:
Pt
= V t Y t -1/~1 =
VtNt 1/~,
(4.2)
48 See Greenspan and Cohen (1996), for a discussion of the importance of considering endogenous
scrappage to forecast sales of new motor vehicles in the USA.
49 See Hopenhayn (1992) for an elegant characterization of the steady state properties of a competitive
equilibrium model of entry and exit.
845
(4.3)
Using Fubini's Theorem (i.e. moving the expectation with respect to the idiosyncratic
shocks inside the integral) allows us to remove the idiosyncratic component from
Equation (4.3), yielding
F>~EtlftP,.er(St)dsJ.
(4.4)
Given Nt, the industry price is exclusively driven by the aggregate demand shock. Thus,
absent entry, the right-hand side of Equation (4.4) is an increasing function of Pt, call
it j~(P). Entry, however, cannot always be absent, for that would occasionally violate
the free entry condition. Indeed, as soon as J~(P) > F, there would be infinite entry
which, in turn, would lower the equilibrium price instantly. There is only one price, call
it/5o, such that the free entry condition holds with equality. Once this price is reached,
enough entry will occur to ensure that the price does not cross this upper bound; but,
to be justified, entry must not occur below that bound either. Entry, therefore, changes
the stochastic process of the equilibrium price from a Brownian Motion to a regulated
Brownian Motion. This change in the price process, however, means that.]~ is no longer
the right description of the expression on the right-hand side of Equation (4.4). There
is a new function, Ji(P), which is still monotonic in the price, but which satisfies
Ji(P) < J~(P) for all P because of the role of entry in preventing the realization of
high prices. This, in turn, implies a new reservation/entry price P1 > P0, which leads
to a new function )~(P), such that j~ > y~ > Ji, which leads to a new regulation
point in between the previous ones, and so on until convergence to some equilibrium,
~(p),p)51.
Thus, through competitive equilibrium, we have arrived at a solution like that of the
irreversible investment problem at the individual level, but now for the industry as a
whole. Periods of inaction are followed by regulated investment (through entry) during
favorable times. The constructive argument used to illustrate the solution isolates
50 Adding an aggregate productivity shock is straightforward. The Brownian Motion assumption is not
needed, but it simplifiesthe calculations.
51 Needless to say, this iterative procedure is not needed to obtain the solution of this problem.
846
R.J Caballero
847
short run (remember Goolsbee's evidence), we return to a scenario in which the "curse
of dimensionality" appears. Caballero and Hammour (1994, 1996a) have dealt with this
case in scenarios where aggregate shocks follow deterministic cycles 52. Besides the
specific issues addressed in those papers, the main implication for the purpose of this
survey is that investment by potential entrants becomes less responsive to aggregate
shocks, which also means a break down of perfect insulation and therefore a more
volatile response of the scrapping and intensive margins.
Krieger (1997) also discusses equilibrium interactions between creation and destruction margins, although he obtains positive rather than negative comovement between
investment and scrapping. In his model, a permanent technology shock leads to a short
term increase in interest rates which squeezes low productivity units relative to high
productivity ones. The ensuing increase in scrapping frees resources for new higherproductivity investment. Similarly, Campbell (1997) studies the equilibrium response
of entry and exit to technology shocks embodied in new production units. He argues
that the increase in exit generated by positive technological shocks is an important
source of resources for the creation of new production sites.
4.2. Technological heterogeneity and scrapping
Scrapping is an important aspect of the process of capital accumulation. Understanding
it is essential for constructing informative measures of the quantity and quality
of capital at each point in time. Nonetheless, the scrapping margin is seldom
emphasized, I suspect, mostly because of the difficulties associated with obtaining
reliable data53. As a result, many time series comparisons of capital accumulation
and productivity growth (especially across countries) are polluted by inadequate
accounting of scrapping. Effective capital depreciation must surely be higher in
countries tmdergoing rapid modernization processes.
Partly to address these issues, vintage capital and putty-clay models have regained
popularity lately. Benhabib and Rustichini (1993), for example, describe the investment
cycles that follow scrapping cycles in a vintage capital model. While Atkeson and
Kehoe (1997) argue that putty-clay models outperform standard putty-putty models
with adjustment costs in describing the cross sectional response of investment and
output to energy shocks. Gilchrist and Williams (1996), on the other hand, embody
the putty-clay model in an otherwise standard RBC model and document a substantial
gain over the standard RBC model in accounting for the forecastable comovements of
economic aggregates. And Cooley et al. (1997) describe the medium/low frequency
52 In work in progress [Caballero and Hammour (1997b)], we have obtained an approximate solution
for the stochastic case, in a context where the sources of convexity are malfimctioninglabor and credit
markets.
53 See Greenspan and Cohen (1996) for sources of scrapping data for US motor vehicles.
848
R.J Caballero
aspects of a multisectoral vintage capital economy, and show how tax policy can have
significant effects on the age distribution of the capital stock 54.
The technological embodiment aspect of these models captures well the creativedestruction component of capital accumulation and technological progress 55. Salter's
(1960) careful documentation of the technological status of narrowly defined US and
UK industries is very revealing with respect to the simultaneous use of different
techniques of production and the negative correlation between productivity ranking
and the technological age of the plant 56. For example, his table 5 shows the evolution
of methods in use in the US blast furnace industry from 1911 to 1926. At the
beginning of the sample, the "best practice" plants produced 0.32 gross tons of pigiron per man-hour, while the industry average was 0.14. By the end of the sample,
best practice plants productivity was 0.57 while the industry average was 0.30. While
at the beginning of the sample about half of the plants used hand-charged methods of
production, only six percent did at the end of the sample.
As mentioned above, obsolescence and scrapping are not only driven by slowly
moving technological trends, but also by sudden changes in the economic environment.
Goolsbee (1995b) documents the large impact ofoil shocks on the scrapping of old and
fuel-inefficient planes. For example, he estimates that the probability of retirement of a
Boeing 707 (relatively inefficient in terms of fuel) more than doubled after the second
oil shock. This increase was more pronounced among older planes. Once more, the
endogenous nature of the scrapping dimension must be an important omitted factor in
our accounting of capital accumulation and microeconomie as well as macroeconomic
performance.
The sunk nature of technological embodiment is a source of lumpy and discontinuous actions at the microeconomic level. The (S, s) apparatus, with its implications for
aggregates, is well suited for studying many aspects of vintage and putty-clay models.
In particular, episodes of large investment which leave their technological fingerprints,
and remain in the economy, reverberating over time.
5. Inefficient investment
Fixed costs, irreversibilities and their implied pattern of action/inaction, have microeconomic and aggregate implications beyond the mostly technological (and
neoclassical) ones emphasized above. Indeed, they seed the ground for powerful
inefficiencies. This section describes new research on the consequences o f two of
849
850
R.J. Caballero
the industry. Caplin and Leahy show that the price process x(t) obeys the following
differential equation:
where F is the fixed entry cost paid by the firm and r is the real interest rate. This
equation has a natural interpretation which captures the idea that competitive firms
are indifferent between entry today and entry tomorrow. The left-hand side represents
the loss in current revenue incurred by a firm which delays entry for a brief instant
beyond t s7. The right-hand side captures the expected gain from this delay. The term
rF reflects the gain due to the postponement of the entry cost, while the last term
represents the saving due to the possibility that delay will reveal the true industry's
marginal cost, aborting a wasteful investment sS.
In equilibrium, entry is delayed and price declines slowly; "gradualism" maintains
prices high enough for sufficiently long so as to offset (in expectation) the risk incurred
by investors who act early rather than wait and free-ride off of others' actions s9.
Caplin and Leahy (1994) characterize the opposite extreme, one of delayed exit. The
key connection with the previous sections is that the problem of information revelation
arises from the fact that, as we have seen, fixed costs of actions make it optimal not
to act most of the time. Thus, information that could be revealed by actions remains
trapped.
Their model is one of time-to-build. Many identical firms simultaneously start
projects which have an uncertain common return several periods later (e.g. a real estate
boom). Along the investment path, firms must continue their investment and receive
private signals on the expected return. The nature of technology is such that required
investment is always the same if the firm chooses to continue in the project. The firm
has the option to continue investing ("business as usual"), to terminate the project,
or to suspend momentarily, but the cost of restarting the project after a suspension is
very large. Project suspension reveals (to others) negative idiosyncratic information; if
nobody suspends, it is good news. However, the costly nature of suspension delays it,
and therefore information revelation is also delayed. Bad news may be accumulating
but nobody suspends, because everybody is waiting for a confirmation of their bad
signals by the suspension of other people. Eventually, some firms will receive enough
bad signals to suspend in spite of the potential cost of doing so (i.e., if they are wrong
57 At the time when the industry starts, potential investors' priors are that the price is distributed
uniformly on [0, I]. As entry occurs and the price declines, the priors are updated. If convergence has
not happened at time t, marginal cost is assumed uniformly distributed on [0,x(t)]. The expected cost
of waiting is, therefore, equal to the price minus the expected marginal cost, x(t).
58 Here ~ d t is the probabilitythat price hits marginal cost during the next dt units of time.
59 Even though entrants make zero profits in expectation, ex-post, early entrants earn positive profits,
while late entrants lose money.
851
in their negative assessment o f market conditions). Since the number of firms in their
model is large, the number of firms that suspend for the first time fully reveals future
demand: if demand is low, everybody exits; if it is high, all those that suspended
restart.
If it were not for the interplay between inaction (investment as usual) and private
information, the fate of the market would be decided correctly after the first round of
signals. Information aggregation does not take place until much later, however. Thus,
substantial investment may turn out to be wasted because the discrete nature of actions
inhibits information transmission. The title of their paper beautifully captures the expost feeling: Wisdom after the fact.
The "classic" paper from the literature on information and investment is due to
Chamley and Gale (1994). In their model all (private) information arrives at time zero;
the multiple agent game that ensues may yield many different aggregate investment
paths, including suboptimal investment collapses. In reviewing the literature, Gale
(1995) illustrates the robustness of the possibility of an inefficient investment collapse
(or substantial slowdown and delay). He notices that in order for there to be any value
to waiting to see what others do before taking an action (investing for example) it must
be the case that the actions of others are meaningful. That is, the action taken in the
second period by somebody who chose to wait in the first period must depend in a non
trivial way on the actions of others at the first date. If a firm chooses to wait this period,
possibly despite having a positive signal, it will only invest next period if enough other
firms invest this period. It must therefore be possible for every firm to decide not to
invest next period because no one has invested this period, even though each firm may
have received a positive signal this period, in which case, investment collapses.
This is a very interesting area of research for those concerned with investment issues
and is wanting for empirical developments.
5.2. Specificity and opportunism
852
R.J Caballero
controlled by economic agents with the will and freedom to behave opportunistically.
In a sense, this is a property rights problem, and as such it must have a first-order effect
in explaining the amount and type o f capital accumulation and, especially, differences
in these variables across countries.
Thus, the window for opportunism arises when part o f the investment is specific to
an economic relationship, in the sense that if the relationship breaks up, the potential
rewards to that investment are irreversibly lost. Further, such opportunism is almost
unavoidable when this "fundamental transformation" from uncommitted to specialized
capital is not fully protected by contract [Williamson (1979, 1985)] 60
Specificity, that is, the fact that factors o f production and assets may be worth
more inside a specific relationship than outside o f it, may have a technological or an
institutional origin. Transactions in labor, capital and goods markets are frequently
characterized by some degree o f specificity. The creation of a job often involves
specific investment by the firm and the worker. Institutional factors, such as labor
regulations or unionization also build specificities.
There is a very extensive and interesting microeconomic literature on the impact of
unprotected specificity on the design o f institutions, organizations and control rights.
Hart (1995) reviews many o f the arguments and insights. For the purpose o f this survey,
however, the fundamental insight is in Simons (1944), who clearly understood that
hold-up problems lead to underinvestment:
... the bias against new investment inherent in labor organizations is important .... Investors
now face ... the prospect that labor organizations will appropriate most or all of the earnings
.... Indeed, every new, long-term commitment of capital is now a matter of giving hostages to
organized sellers of complementary services.
More recently, Grout (1984) formalized and generalized Simons' insight, and
Caballero and Hammour (1998a) studied, at a general level, the aggregate consequences o f opportunism 61. Here, I borrow the basic model and arguments from that
paper to discuss those aspects o f the problem which are most relevant for aggregate
investment.
Everything happens in a single period 62. There is one consumption good, used as
a numeraire, and two active factors o f production, 1 and 2 63. Ownership o f factors
1 and 2 is specialized in the sense that nobody owns more than one type of factor.
8 5 3
There are two modes o f production. The first is joint production, which requires, in
fixed proportions, xl and x2 units of factors 1 and 2, respectively, to produce y units
o f output. Let E denote the number of joint production units, so Ei = x i E represents
employment o f factor i in joint production. The other form o f production is autarky
where each factor produces separately, with decreasing returns technologies Fi(Ui),
and where Ui denotes the employment of factor i in autarky, such that Ei + Ui = 1.
The autarky sectors are competitive, with factor payments, pi:
Pi = F;(Ui).
(5.2)
For now, there are no existing units. At the beginning of the period there is mass
one o f each factor o f production. There are no matching frictions so that, in the
efficient/complete contracts economy, units move into joint production (assuming
corners away) until
y-p~xl
+p2x2,
(5.3)
wixi = (1
(5.4)
- q}l)plXl
- (1 - q}2)p2x2.
(5.5)
854
R.J. Caballero
(5.6)
Substituting Equations (5.4) and (5.5) into Equation (5.6), transforms factor i's
participation condition into
y ~ p l x t +p2x2 + Ai,
(5.7)
Z~i ~ ~)iPiXi -- ~ j p j x j ,
(5.8)
with
which measures the net sunk component of the relationship for factor i. In other words,
it is a measure of the "exposure" of factor i to factor j. When Ai is positive, part of
factor i's contribution to production is being appropriated by factor j 65.
5.2.1. Generic implications
Figure 5.1 characterizes equilibrium in both efficient and incomplete contract economies. The two dashed curves represent the right-hand side of condition (5.7) for factors
1 and 2. They are increasing in the number of production units because the opportunity
cost of factors of production (the pis) rise as resources are attracted away from autarky.
The thick dashed curve corresponds to that factor of production (here factor 1) whose
return in autarky is tess responsive to quantity changes 66. If one thinks of capital and
labor, arguably capital is this factor; which is a maintained assumption through most
of this section. The horizontal solid line is a constant equal to y, which corresponds to
the left-hand side of condition (5.7). Equilibrium in the incomplete contracts economy
corresponds to the intersection of this line with the highest (at the point of intersection)
of the two dashed lines. In the figure, the binding constraint is that of capital.
An efficient equilibrium, on the other hand, corresponds to the intersection of the
horizontal solid line with the solid line labeled Eft. The latter is just the sum of the exante opportunity costs of factors of production [the right-hand side of Equation (5.3)].
This equilibrium coincides with that of the incomplete contracts economy only when
both dashed lines intersect; that is, when net appropriation is z e r o (A i = - A j = 0).
There are several features of equilibrium which are important for investment (or
capital accumulation). First, there is underinvestment; equilibrium point A is to the
left of the efficient point A*. Because it is being appropriated, capital withdraws into
autarky (e.g. consumption, investment abroad, or investment in less socially-valuable
855
(2) Eft.
B ~ y , ~ (1)
~
l /
.'i//
//
///
U. / / /
~.
rium.
activities) 67. Second, the withdrawal of capital constrains the availability of jobs and
segments the labor market 68. In equilibrium, not only are there fewer joint production
units, but also the right-hand side of condition (5.7) for labor is less than y, reflecting
the net appropriation of capital; outside labor cannot arbitrage away this gap because
its promises are not enforceable. Third, investment is more volatile than it would be
in the efficient economy 69. Changes in y translate into changes in the number of joint
production units through capital's entry condition (thick dashes), which is clearly more
elastic (at their respective equilibria) than the efficient entry condition ("Eft" line).
If profitability in joint production is high enough, equilibrium is to the right
of the balanced specificity point, B. In that region, it is the labor entry condition
which binds. In principle, problems are more easily solved in this region through
contracts and bonding. I f not solved completely, however, there are a few additional
conclusions o f interest for an investment survey. First, there is underinvestment since
the complementary factor, labor, withdraws (relative to the first best outcome) from
joint production. Second, capital is now rationed, so privately profitable investment
projects do not materialize. Third, investment is now less volatile than in the efficient
economy. Changes in y translate into changes in the number of joint production units
through labor's entry condition (thin dashes), which is clearly less elastic than the
efficient entry condition ("Eft" line).
67 See Fallick and Hassett (1996) for evidence on the negative effect of union certification on firm level
investment.
68 This holds even in the extreme case where capital and labor are perfect substitutes in production.
See Caballero and Hammour (1998a).
69 In a dynamic model, this translates into a statement about net capital accumulation rather than,
necessarily, investment. The reason for the distinction is that the excessive response of the scrapping
margins and intertemporal substitution effects on the creation side may end up dampening actual
investment. See Caballero and Hammour (1996a).
856
R.J Caballero
There is by now a large body of evidence supporting the view that credit constraints
have substantial effects on firm level investment. Although there are a number of
qualifications to specific papers in the literature, the cumulative evidence seems
overwhelmingly in favor of the claim that investment is more easily financed with
internal than external funds 72. I will not review this important literature here because
there are already several good surveys 73.
70 We argue that this is a plausible factor behind the large increase in capital/labor ratios in Europe
relative to the USA.
71 Rights to sell, to rent, to bequeath, to pledge, to mortgage, etc.
72 For a dissenting view, see e.g. Kaplan and Zingales (1997) and Cummins, Hassett and Oliner
(1996b).
73 See e.g. Bernanke et al. (1996, 1999) and Hubbard (1995) for recent ones.
857
While there are extensive empirical and theoretical microeconomic literatures, the
macroeconomics literature on credit constraints is less developed. Notable exceptions
are: Bernanke and Gertler (1989, 1990), Kiyotaki and Moore (1997) and Greenwald
and Stiglitz (1993)74. Although the exact mechanisms are not always the same, many
of the aggregate insights of this literature can be described in terms of the results in
the preceding subsections.
Changing slightly the interpretation of factor 2, from labor to entrepreneurs, allows
us to use Figure 5.1 to characterize credit constraints. Rationing in the labor market
becomes rationing of credit available to projects. To the left of point B, which is the
region analyzed in the literature, net investment is too responsive to shocks; there is
more credit rationing as the state of the economy declines; and there is underinvestment
in general.
Internal funds and collateralizable assets reduce the extent of the appropriability
problem by playing the role of a bond, and introduce heterogeneity and therefore
ranking of entrepreneurs. Since the value of collateral is likely to decline during a
recession, there is an additional amplification effect due to the decline in the feasibility
of remedial "bonding" 75.
This survey started by arguing that the long run relationship between aggregate capital,
output and the cost of capital is not very far from what is implied by the basic
neoclassical model: in the US, the elasticity of the capital-output ratio with respect
to permanent changes in the cost of capital is close to minus one.
In the short run things are more complex. Natural-experiments have shown that, in
the cross section, the elasticity of investment with respect to changes in investment
tax credits is much larger than we once suspected.
How to go from these microeconomic estimates to aggregates, and to the response
of investment to other types of shocks is not fully resolved. We do know, however,
that these estimates represent expected values of what seems to be a very skewed
distribution of adjustments. A substantial fraction of a firm's investment is bunched
into infrequent and lumpy episodes. Aggregate investment is heavily influenced by the
degree of synchronization of microeconomic investment spikes. For US manufacturing,
the short run (annual) elasticity of investment with respect to changes in the cost of
capital is less than one tenth the long run response when the economy has had a
depressed immediate history, while this elasticity can rise by over 50 percent when
the economy is undergoing a sustained expansion.
74 Also see Gross (1994) for empirical evidence and a model integrating financial constraints and
irreversibility.
75 See e.g. Kiyotaki and Moore (1997).
858
R.J Caballero
Still, the mapping from microeconomics to aggregate investment dynamics especially equilibrium aggregate investment dynamics - is probably more complex
than just the direct aggregation o f very non-linear investment patterns. Informational
problems lead to a series o f strategic delays which feed into and feed off o f the natural
inaction o f lumpy adjustment models. This process has the potential to exacerbate
significantly the time varying n a c r e o f the elasticity o f aggregate investment with
respect to aggregate shocks.
Moreover, sunk costs provide fertile ground for opportunistic behavior. In the
absence o f complete contracts, aggregate net investment is likely to become excessively
volatile. The lack o f response o f equilibrium payments to complementary - and
otherwise inelastic - factors (e.g. workers), exacerbates the effects o f shocks
experienced b y firms. Also, the withdrawal o f financiers' support during recessions
further reduces investment. Thus, capital investment seems to be hurt at both ends:
workers that do not share fairly during downturns, and financiers that want to limit
their exposure to potential appropriations from entrepreneurs which cannot credibly
commit not to do so during the recovery.
The last two themes, equilibrium outcomes with informational problems and
opportunism, are wanting for empirical work. I therefore suspect that we will see plenty
o f research filling this void in the near future.
References
Abel, A.B. (1979), Investment and the Value of Capital (Garland, New York).
Abel, A.B., and J.C. Eberly (1994), "A unified model of investment under uncertainty", American
Economic Review 84(December):1369-1384.
Abel, A.B., and J.C. Eberly (1996a), "Investment and q with fixed costs: an empirical analysis",
mimeograph (Wharton School, University of Pennsylvania, January).
Abel, A.B., and J.C. Eberly (1996b), "The effects of irreversibility and uncertainty on capital
accumulation", mimeograph (Wharton School, University of Pennsylvania, May).
Abel, A.B., A.K. Dixit, J.C. Eberly and R.S. Pindyck (1996), "Options, the value of capital, and
investment", Quarterly Journal of Economics 111(3, August):753-777.
Acemoglu, D. (1996), "A microfoundation for social increasing returns in human capital accumulation",
Quarterly Journal of Economics 111(3, August):779-804.
Atkeson, A., and EJ. Kehoe (1997), "Models of energy use: putty-putty vs. Putty-clay", Federal Reserve
Bank of Minneapolis, Research Department, Staff Report 230 (March).
Auerbach, A.J., and K.A. Hassett (1992), "Tax policy and business fixed investment in the United States",
Journal of Public Economics 47(2):141-170.
Barnett, S., and E Sakellaris (1995), "Non-linear response of firm investment to Q: testing a model of
convex and non-convex adjustment costs", mimeograph (University of Maryland, August).
Benhabib, J., and A. Rustichini (1993), "A vintage model of investment and growth: theory and evidence",
in: R. Becker, ed., General Equilibrium, Growth and Trade, vol. II (Academic Press, New York).
Bernanke, B.S., and M. Gertler (1989), "Agency costs, net worth, and business fluctuations", American
Economic Review 95:14-31.
Bernanke, B.S., and M. Gertler (1990), "Financial fragility and economic performance", Quarterly
Journal of Economics 105:87-114.
859
Bernanke, B.S., M. Gertler and S. Gilchrist (1996), "The financial accelerator and the flight to quality",
Review of Economics and Statistics 78:1-15.
Bernanke, B.S., M. Gertler and S. Gilchrist (1999), "The financial accelerator in a quantitative business
cycle framework", ch. 21, this Handbook.
Bertola, G. (1988), "Adjustment costs and dynamic factor demands: investment and employment under
uncertainty", Ph.D. Dissertation (MIT, Cambridge, MA).
Bertola, G. (1992), "Labor turnover costs and average labor demand", Journal of Labor Economics
10(4):389-411.
Bertola, G., and R.J. Caballero (1990), "Kinked adjustment costs and aggregate dynamics", in:
O.J. Blanchard and S. Fischer, eds., NBER Macroeconomics Annual (MIT Press, Cambridge, MA)
237-288.
Bertola, G., and R.J. Caballero (1994), "Irreversibility and aggregate investment", Review of Economic
Studies 61(2, April):223-246.
Besley, T. (1995), "Property rights and investment incentives: theory and evidence from Ghana", Journal
of Political Economy 103(5, October):903-937.
Blanchard, O.J. (1986), "Investment, output, and the cost of capital: a comment", Brookings Papers on
Economic Activity 1986(1):153-158.
Blanchard, O.J., and M. Kremer (1996), "Disorganization", mimeograph (MIT, October).
Brainard, W.C., and J. Tobin (1968), "Pitfalls in financial model building", American Economic Review
58(May):99-122.
Caballero, R.J. (1993a), "On the dynamics of aggregate investment", in: L. Serven and A. Solimano,
eds., Striving for Growth After Adjustment, The Role of Capital Formation (The World Bank,
Washington, DC) 81-106.
Caballero, R.J. (1993b), "Durable goods: an explanation for their slow adjustment", Journal of Political
Economy 101(2, April):351-384.
Caballero, R.J. (1994a), "Small sample bias and adjustment costs", Review of Economics and Statistics
76(I, February):52-58.
Caballero, R.J. (1994b), "A reconsideration of investment behavior using tax reforms as natural
experiments: a comment", Brookings Papers on Economic Activity 1994(2):62 68.
Caballero, R.J., and E. Engel (1999), "Explaining investment dynamics in U.S. manufacturing:
a generalized (S, s) approach", Econometrica 67(4, July).
Caballero, R.J., and M.L. Hammour (1994), "The cleansing effect of recessions", American Economic
Review 84(5, December):1350-1368.
Caballero, R.J., and M.L. Hammour (1996a), "On the timing and efficiency of creative destruction",
Quarterly Journal of Economics 111(3, August):805-852.
Caballero, R.J., and M.L. Hammour (1996b), "On the ills of adjustment", Journal of Development
Economics 51(1, October):161-192.
Caballero, R.J., and M.L. Hammour (1997b), "Improper churn: financial constraints and factor markets",
mimeograph (MIT, May).
Caballero, R.J., and M.L. Hammour (1998a), "The macroeconomics of specificity", Journal of Political
Economy 106(4, August):724-767.
Caballero, R.J., and M.L. Hammour (1998b), "Jobless growth: appropriability, factor substitution, and
unemployment", Carnegie-Rochester Conference Series on Public Policy 48(June):51-94.
Caballero, R.J., and J. Leahy (1996), "Fixed costs: the demise of marginal q", Working paper No. 5508
(NBER, March).
Caballero, R.J., and R.S. Pindyck (1996), "Uncertainty, investment, and industry evolution", International
Economic Review 37(3, August):641-662.
Caballero, R.J., E. Engel and J. Haltiwanger (1995), "Plant-level adjustment and aggregate investment
dynamics", Brookings Papers on Economic Activity 1995(2):1-54.
Campbell, J.R. (1997), "Entry, exit, embodied technology, and business cycles", Working Paper No. 5955
(NBER, March).
860
R.J Caballero
Campbell, J.R., and J.D.M. Fisher (1996), "Aggregate employment fluctuations with microeconomic
asymmetries", mimeograph (University of Rochester, August).
Caplin, A., and J. Leahy (1993), "Sectoral shocks, learning, and aggregate fluctuations", Review of
Economic Studies 60(4, October):777-794.
Caplin, A., and J. Leahy (1994), "Business as usual, market crashes and wisdom after the fact", American
Economic Review 84(3, June):54%565.
Chamley, C., and D. Gale (1994), "Information revelation and strategic delay", Econometrica 62:
1065-1085.
Chirinko, R.S. (1993), "Business fixed investment spending: a critical survey of modelling strategies,
empirical results, and policy implications", Journal of Economic Literature 31 (December): 1875-1911.
Clark, J.M. (1917), "Business acceleration and the law of demand: a technical factor in economic cycles",
Journal of Political Economy 25(March):217-235.
Clark, J.M. (1944), "Additional note on business acceleration and the law of demand", in: American
Economic Association, Readings in Business Cycle Theory (Blackiston Company, Philadelphia, PA).
Cooley, T.E, J. Greenwood and M. Yornkoglu (1997), "The replacement problem", Working paper
No. 444 (Rochester Center for Economic Research, August).
Cooper, R., J. Haltiwanger and L. Power (1994), "Machine replacement and the business cycle: lumps
and bumps", mimeograph (Boston University).
Cummins, J.G., K.A. Hassett and R.G. Hubbard (1994), "A reconsideration of investment behavior using
tax reforms as natural experiments", Brookings Papers on Economic Activity 1994(2):1-59.
Cummins, J.G., K.A. Hassett and R.G. Hubbard (1996a), "Tax reforms and investment: a cross country
comparison", Journal of Public Economics 62(1/2, October):237-273.
Cummins, J.G., K.A. Hassett and S.D. Oliner (1996b), "Investment behavior, internal funds, and
observable expectations", mimeograph (New York University, October).
Dixit, A. (1989), "Entry and exit decisions under uncertainty", Journal of Political Economy 97:620-638.
Dixit, A. (1993), The Art of Smooth Pasting (Harwood Academic Publishers, Langhorns, PA).
Doms, M., and T. Dunne (1993), "An investigation into capital and labor adjustment at the plant level",
mimeograph (Center for Economic Studies, Census Bureau).
Eisner, R. (1969), "Tax policy and investment behavior: a comment", American Economic Review
59(June):379-388.
E1 Karoui, N., and I. Karatzas (1991), "A new approach to the Skorohod problem, and its applications",
Stochastics 34:57-82.
Fallick, B.C., and K.A. Hassert (1996), "Investment and union certification", Discussion paper
No. 1996-43 (FED, November).
Fazzari, S.M., R.G. Hubbard and B.C. Petersen (1988), "Financing constraints and corporate investment",
Brookings Papers on Economic Activity 1988(1):141-195.
Gale, D. (1995), "What have we learned from social learning?", mimeograph (Boston University,
August).
Gilchrist, S., and J.C. Williams (1996), "Putty-clay and investment: a business cycle analysis",
mimeograph (Boston University, May).
Goolsbee, A. (1995a), "Investment tax incentives and the price of capital goods", mimeograph (Chicago
GSB).
Goolsbee, A. (1995b), "Factor prices and the retirement of capital goods", mimeograph (Chicago GSB,
July).
Goolsbee, A. (1997), "The incidence of investment tax subsidies: to the workers go the spoils?",
mimeograph (Chicago GSB, February).
Goolsbee, A., and D.B. Gross (1997), "Estimating adjustment costs with data on heterogeneous capital
goods", mimeograph (Chicago GSB, September).
Greenspan, A., and D. Cohen (1996), "Motor vehicles stocks, scrappage, and sales", Working paper
No. 1996-40 (FED, October).
Ch. 12:
Aggregate Investment
861
Greenwald, B., and J. Stiglitz (1993), "Financial market imperfections and business cycles", Quarterly
Journal of Economics 108(1, Febrnary):77-114.
Gross, D. (1994), "The investment and financing decisions of liquidity constrained firms", mimeograph
(Chicago GSB).
Grout, RA. (1984), "Investment and wages in the absence of binding contracts: a Nash bargaining
approach", Econometrica 52(2, March):449-460.
Hall, R.E., and C.I. Jones (1996), "The productivity of nations", mimeograph (Stanford University,
August).
Hall, R.E., and D.W Jorgenson (1967), "Tax policy and investment behavior", American Economic
Review 57(3, June):391-414.
Hart, O. (1995), Firms, Contracts and Financial Structure, Clarendon Lectures in Economics (Oxford
University Press, Oxford).
Hassett, K.A., and R.G. Hubbard (1996a), "New evidence concerning the openness of the world market
for capital goods", mimeograph (Board of Governors of the Federal Reserve System, June).
Hassett, K.A., and R.G. Hubbard (1996b), "Tax policy and investment", Working Paper No. 5683 (NBER,
July).
Hayashi, E (1982), "Tobin's marginal Q and average Q: a neoclassical interpretation", Econometrica
50(1, January):213-224.
Hopenhayn, H.A. (1992), "Entry, exit, and firm dynamics in long run equilibrium", Econometrica
60:1127-1150.
Hubbard, R.G. (1995), "Capital-market imperfections and investment", mimeograph (Columbia
University).
Jorgenson, D.W.. (1963), "Capital theory and investment behavior", American Economic Review
53(2, May):247-259.
Jovanovic, B. (1997), "Obsolescence of capital", mimeograph (New York University, February).
Kaplan, S.N., and L. Zingales (1997), "Do investment-cash flow sensitivities provide useful measures of
financing constraints?, Quarterly Journal of Economics 112(1, February):169 216.
Kiyotaki, N., and J. Moore (1997), "Credit cycles", Journal of Political Economy 105(2, April):211~48.
Koyck, L.M. (1954), Distributed Lags and Investment Analysis (North-Holland, Amsterdam).
KriegeL S. (1997), "The general equilibrium dynamics of investment, scrapping and reorganization in
an economy with firm level uncertainty", mimeograph (Chicago, July).
Leahy, J. (1993), "Investment in competitive equilibrium: the optimality of myopic behavior", Quarterly
Journal of Economics 108:1105-1133.
Meyer, J.R., and E. Kuh (1957), The Investment Decision: An Empirical Study (Harvard University
Press, Cambridge, MA).
Nickell, S.J. (1978), The Investment Decisions of Firms (Cambridge University Press, Oxford).
Oliner, S.D., G.D. Rudebusch and D. Sichel (1995), "New and old models of business investment: a
comparison of forecasting performance", Journal of Money, Credit and Banking 27:806-826.
Pindyck, R.S. (1988), "Irreversible investment, capacity choice, and the value of the firm", American
Economic Review 78(5, December):969-985.
Ramey, G., and J. Watson (1996), "Contractual fragility, job destruction and business cycles", mimeograph
(University of California at San Diego, June).
Rivers, D., and Q.H. Vuong (1991), "Model selection tests for nonlinear dynamic models", Working
paper No. 91-08 (University of Toulouse).
Rotemberg, J.J. (1987), "The new Keynesian microfoundations", in: O.J. Blanchard and S. Fischer, eds.,
NBER Macroeconomics Annual 1987 (MIT Press, Cambridge, MA) 6~104.
Rothschild, M. (1971), "On the cost of adjustment", Quarterly Journal of Economics 85(November):
605-622.
Salter, W.E.G. (1960), Productivity and Technical Change (Cambridge University Press, Cambridge).
Shapiro, M.D. (1986), "Investment, output, and the cost of capital", Brookings Papers of Economic
Activity 1986(1):111 152.
862
R.J. Caballero
Simons, H.C. (1944), "Some reflections on syndicalism", Journal of Political Economy 52:125.
Stock, J.H., and M.W Watson (1993), "A simple MLE of cointegrating vectors in higher order integrated
systems", Econometrica 61(4, July):783-820.
Tinbergen, J. (1939), "A method and its application to investment activity", in: Statistical Testing of
Business Cycle Theories, vol. 1 (Economic Intelligence Service, Agathon Press, New York).
Tobin, J. (1969), "A general equilibrium approach to monetary theory", Journal of Money, Credit and
Banking 1:15-29.
Williamson, O.E. (1979), "Transaction-cost economics: the governance of contractual relations", Journal
of Law and Economics 22(2, October):233-261.
Williamson, O.E. (1985), The Economic Institutions of Capitalism (Free Press, New York).
Chapter 13
INVENTORIES *
VALERIE A. RAMEY
University of Wisconsin
Contents
Abstract
Keywords
Introduction
1. Sectoral and secular behavior of inventories
2. Two stylized facts about inventory behavior
2.1. Procyclical inventory movements
2.1.1. Illustrative evidence
2.1.2. A survey of results
2.2. Persistent movements in the inventory-sales relationship
2.2.1. Illustrative evidence
2.2.2. A survey of results
4. Decision rule
4.1. Introduction
4.2. Derivation of decision rule
4.3. Persistence in the inventory-sales relationship
4.4. Summary on persistence in the inventory-sales relationship
864
864
865
868
872
873
873
875
877
877
880
882
882
882
885
887
887
887
888
891
892
893
* We thank the National Science Foundation and the Abe Foundation for financial support; Clive
Granger, Donald Hester, James Kahn, Anil Kashyap, Linda Kole, Spencer Krane, Scott Schuh, Michael
Woodford and a seminar audience at the University of Wisconsin for helpful comments and discussions;
and James Hueng and especially Stanislav Anatolyev for excellent research assistance. Email to:
kdwest @ facstaff, wisc. edu; vrameyOweber, ucsd. edu.
864
6. Dynamic responses
7. Empirical evidence
7.1. Introduction
7.2. Magnitude of cost parameters
7.3. Shocks
7.4. Interpretation
8. Directions for future research
8.1. Introduction
8.2. Inventories in production and revenue functions
8.3. Models with fixed costs
8.4. The value of more data
9. Conclusions
Appendix A. Data Appendix
Appendix B. Technical Appendix
B.1. Solution of the model
B.2. Computation of E(Q2 - S2)
B.3. Estimation of
B.4. Social planning derivation of the model's first-order conditions
References
894
902
902
903
906
906
909
909
909
910
911
912
913
914
914
919
919
919
920
Abstract
We review and interpret recent work on inventories, emphasizing empirical and
business cycle aspects. We begin by documenting two empirical regularities about
inventories. The first is the well-known one that inventories move procyclically. The
second is that inventory movements are quite persistent, even conditional on sales.
To consider explanations for the two facts, we present a linear-quadratic model. The
model can rationalize the two facts in a number o f ways, but two stylized explanations
have the virtue o f relative simplicity and support from a number o f papers. Both
assume that there are persistent shocks to demand for the good in question, and that
marginal production cost slopes up. The first explanation assumes as well that there
are highly persistent shocks to the cost o f production. The second assumes that there
are strong costs of adjusting production and a strong accelerator motive.
Research to date, however, has not reached a consensus on whether one of
these two, or some third, alternative provides a satisfactory explanation o f inventory
behavior. We suggest several directions for future research that promise to improve
our understanding o f inventory behavior and thus o f business cycles.
Keywords
J E L classification: E22, E32
Ch. 13:
Inventories
865
Introduction
In developed countries, inventory investment typically averages less than one-half of
one percent of GDP, whereas fixed investment averages 15% of GDP and consumption
two-thirds. Perhaps with these fractions in mind, macroeconomists have concentrated
more on the study of consumption and fixed investment than on inventories. Inventories
generally do not appear as separate variables in dynamic general equilibrium models,
nor in exactly identified vector autoregressive models.
It has long been known, however, that other ways o f measuring the importance
of inventories suggest that inventories should receive more attention, especially in
business cycle research. Half a century ago, Abramowitz (1950) established that US
recessions prior to World War II tended to be periods of inventory liquidations. Recent
experience in the G7 countries indicates this regularity continues to hold, and not
just for the USA. In six of the seven G7 countries (Japan is the exception), real
GDP fell in at least one recent year. Line 2 of Table 1 shows that in five of those
six countries (the United Kingdom is now the exception), inventory investment also
declined during the period of declining GDP, accounting in an arithmetical sense for
anywhere 12-71% of the fall in GDE And Table 1 's use of annual data may understate
the inventory contribution: Table 2 indicates that for quarterly US data, the share is 49
rather than 12% for the 1990-1991 recession, with 49 a typical figure for a post-War
US recession.
Such arithmetical accounting of course does not imply a causal relationship. But
it does suggest that inventory movements contain valuable information about cyclical
fluctuations. In this chapter, we survey and interpret recent research on inventories,
emphasizing empirical and business cycle aspects. Among other points, we hope
to convince the reader that inventories are a useful resource in business cycle
analysis. They may be effective in identifying both the mechanisms of business cycle
propagation and the sources o f business cycle shocks.
Our chapter begins by documenting two facts about inventories. The first is the
well-known one that inventories move procyclically. They tend to be built up in
expansions, drawn down in contractions. The second, and not as widely appreciated,
fact is that inventory movements are quite persistent, even conditional on sales. In
many data sets, inventories and sales do not appear to be cointegrated, and the firstorder autocorrelations of supposedly stationary linear combinations of inventories and
sales are often around 0.9, even in annual data.
To consider explanations for the two facts, we use a linear quadratic/flexible
accelerator model, which is the workhorse for empirical research on inventories. In
our model, one source of persistence is from shocks to demand for the good being
put in inventory - "demand" shocks. ("Demand" is in quotes because we, and the
literature more generally, do not attempt to trace the ultimate source of such shocks;
for example, for an intermediate good, the shocks might be driven mainly by shocks
to the technology of the industry that uses the good in production.) But even if this
shock has a unit root, our model yields a stationary linear combination of inventories
866
Table 1
Arithmetical importance of inventory change in recessions of the 1990s (annual data)a
Country
Canada France
West
Italy Japan
Germany
UK
USA
1989
1991
1992
1993
1992
1993
1992
1993
n.a.
1990
1992
1990
1991
50
71
19
30
n.a.
-0.
12
a The figures are based on annual real data. The inventory change series is computed by deflating the
annual nominal change in inventories in the National Income and Product Accounts by the GDP deflator;
see the Data Appendix.
b The trough year was found by working backwards from the present to the last year of negative real
GDP growth in the 1990s. There were no such years in Japan. The peak year is the last preceding year
of positive real GDP growth.
c Computed by multiplying the following ratio by 100:
inventory change in trough year-inventory change in peak year
GDP in trough year- GDP in peak year
By construction, the denominator of this ratio is negative. A positive entry indicates that the numerator
(the change in the inventory change) was also negative. The negative entry for the United Kingdom
indicates that the change in the inventory change was positive.
Table 2
Arithmetical importance of inventory changes in post-war US recessions (quarterly data) a
Peak quarter-trough quarter
1948:4-1949:2
130
1953:2-1954:2
41
1957:1-1958:1
21
1960:1-1960:4
122
1969:3-1970:1
127
1973:4-1975:1
59
1980:1-1980:3
45
1981:3-1982:3
29
1990:2-1991:1 b
49
a The figures are based on quarterly real data. See the notes to Table 1 for additional discussion.
b The figure for the 1990-1991 recession differs from that for the USA in Table 1 mainly because
quarterly data were used. It also differs because in this table the inventory change is measured in
chain weighted 1992 dollars, whereas Table 1 uses the nominal inventory change deflated by the GDP
deflator.
867
and sales. This stationary linear combination can be considered a linear version of
the inventory-sales ratio. We call it the inventory-sales relationship. And our second
inventory fact is that there is persistence in this relationship.
While the model is rich enough that there are many ways to make it explain the two
facts, we focus on two stylized explanations that have the virtue of relative simplicity,
as well as empirical support from a number of papers. Both explanations assume a
upward sloping marginal production cost (a convex cost function). The first explanation
also assumes that fluctuations are substantially affected by highly persistent shocks to
the cost of production. Cost shocks will cause procyclical movement because times of
low cost are good times to produce and build up inventory, and conversely for times of
high cost. As well, when these shocks are highly persistent a cost shock that perturbs
the inventory-sales relationship will take many periods to die off, and its persistence
will be transmitted to the inventory-sales relationship.
The second explanation assumes that there are strong costs of adjusting production
and a strong accelerator motive. The accelerator motive links today's inventories to
tomorrow's expected sales, perhaps because of concerns about stockouts. Since sales
are positively serially correlated, this will tend to cause inventories to grow and
shrink with sales and the cycle, a point first recognized by Metzler (1941). As well,
with strong costs of adjusting production, if a shock perturbs the inventory-sales
relationship, return to equilibrium will be slow because firms will adjust production
only very gradually.
Both explanations have some empirical support. But as is often the case in empirical
work, the evidence is mixed and ambiguous. For example, the cost shock explanation
works best when the shocks are modelled as unobservable; observable cost shifters,
such as real wages and interest rates, seem not to affect inventories. And the literature
is not unanimous on the magnitude of adjustment costs.
While the literature has not reached a consensus, it has identified mechanisms and
forces that can explain basic characteristics of inventory behavior and thus of the
business cycle. We are optimistic that progress can continue to be made by building
on results to date. Suggested directions for future research include alternative ways
of capturing the revenue effects of inventories (replacements for the accelerator),
alternative cost structures and the use of price and disaggregate data.
The chapter is organized as follows. Section 1 presents some overview information
on the level and distribution of inventories, using data from the G7 countries, and
focussing on the USA. We supply this information largely for completeness and to
provide a frame of reference; the results in this section are referenced only briefly in
the sequel.
Section 2 introduces the main theme of our chapter (business cycle behavior of
inventories) by discussing empirical evidence on our two facts about inventories.
Procyclical movement is considered in Section 2.1, persistence in the inventory-sales
relationship in Section 2.2. In these sections, we use annual data from the G7 countries
and quarterly US data for illustration, and also summarize results from the literature.
868
France
West
Germany
Italy
Japan
UK
USA
2.32
37.4
12.3
12.3
2.41
1.81
23.6
3.91
40.1
12.7
9.8
1.44
3.04
21.6
721
6882
2608
1351
453
584
6066
131
n.a.
411
n.a.
71
104
971
a The inventory change series is computed by deflating the annual nominal change in inventories in the
National Income and Product accounts by the GDP deflator; see the Data Appendix. Units for all entries
are billions (trillions, for Italy and Japan) of units of own currency, in 1990 prices.
b Sample periods are 1957-1994 for West Germany and 1960-1994 for Italy, not 1956-1995.
c GDP entries for Italy and Germany are for 1994, not 1995.
d The "level" entries for Canada, West Germany, Japan and the UK are computed by deflating the
nominal end of year value by the GDP deflator; see the Data Appendix. The entry for the US is the
Department of Commerce constant (chained 1992) dollar value for non-farm inventories, rescaled to
a 1990 from a 1992 base with the GDP deflator.
869
Table 4
Sectoral distribution of US non-farm inventories a,b
(1) Percent of total
level, 1995
100
21.4
(22.5)
3.5
(3.5)
37
7.0
(11.6)
2.8
(4.2)
Finished goods
13
2.5
(4.4)
3.0
(4.8)
Work in process
12
2.3
(5.9)
2.8
(6.0)
Raw materials
12
2.2
(5.4)
2.6
(6.2)
52
12.2
(13.4)
4.4
(4.5)
Retail
26
5.9
(10.3)
4.2
(6.7)
Wholesale
26
6.2
(7.3)
4.5
(4.8)
11
2.2
(5.1)
3.1
(5.8)
Total
Manufacturing
Trade
Other
870
I
,I
06
o6
.9
current dollars
.8
,7
,'\ ,..,,,,',,,",
,,.
.6
"t " ~ ""
""~
'
,5
56
6b
66
76
year
ob
871
.pi
.S
I
20t.8
3513
'
17A3.97
Fig. 2. Quarterly inventories and sales, 1947:1-1996:4, in billions of chained 1992 dollars.
Readers familiar with the monthly inventory-sales ratios commonly reported in the
US business press may be surprised at the absence of a downward secular movement.
Such monthly ratios typically rely on nominal data. The solid line in Figure 1
shows that the ratio of nominal non-farm inventories to nominal sales of domestic
product indeed shows a secular decline. Evidently, the implied deflator inventories
has not been rising as fast as that for final sales. We do not attempt to explain the
differences between the nominal and real series. We do note, however, the nominal
ratio shows persistence comparable to that of the real ratio. The estimate of the firstorder autocorrelation of the ratio is 0.97 whether or not we allow a different mean
inventory-sales ratio for the 1947:I-1973:IV and 1974:I-1996:IV subsamples.
To return to the secular behavior of the real series: we see from column 3 in Table 4
that the rough constancy of the overall ratio hides some heterogeneity in underlying
components. In particular, raw materials, and to a lesser extent, work in progress,
have been growing more slowly than the aggregate, implying a declining ratio to final
sales. This fact was earlier documented by Hester (1994), who noted that possible
explanations include just-in-time inventory management, outbasing of early stages of
manufacturing to foreign countries, and a transitory response to transitory movements
in costs.
In the sequel we do not attempt to explain secular patterns in inventory-sales ratios;
see Hester (1994) for a discussion of US data, for retail as well as manufacturing, West
(1992a) and Allen (1995) for discussions of Japanese data. Instead we hope that the
reader will take the message away from these tables that inventories and sales are
positively related in the long run: they tend to rise together. This is illustrated quite
872
strikingly in Figure 2, which is a scatterplot o f the inventory and sales data. A second
message in the tables and the autocorrelations reported above is that while inventory
movements are small relative to GDP, they are volatile and persistent. Characterizing
and explaining the stochastic, and especially business cycle, behavior o f inventories is
the subject o f the rest o f this chapter.
1 When we summed the AH t series, we initialized with H 0 _: 0. Given the tinearity of our procedures,
the results would be identical if we instead used the AH t series to work forwards and backwards from
the 1995 levels reported in Table 3. The reader should be aware that when prices are not constant, a
series constructed by our procedure of summing changes typically will differ from one that values the
entire level of stock at current prices. Those with access to US sources can get a feel for the differences
by comparing the inventory change that figures into GDP (used in the G7 data, and in NIPA Tables
5.10 and 5.11) and the one implied by differencing the series for the level of the stock (used in our
quarterly US data and NIPA Tables 5.12 and 5.13).
2 We repeated some of our quarterly calculations using final sales of goods and structures, which differs
from total final sales because it excludes final sales of services. There were no substantive changes in
results.
873
All of these measures are linked by the identity production = sales + (inventory
investment), or
Qt = st +AHt,
(2.1)
874
Table 5
Relative variability of output and final sales a-e
Country
Period
(1)
corr(S, AH)
(2)
var(Q)/var(S)
(3)
var(AQ)/
var(AS)
(4)
1 + [E(Q 2 - S 2)/
var(AS)]
Canada
1956 1995
1974-1995
1956 1995
1974-1995
1957-1994
1974-1994
1960-1994
1974-1994
1956-1995
0.14
0.17
0.17
0.32
0.12
0.13
0.13
0.11
0.23
1.16
1.21
1.36
1.63
1.10
1.08
1.30
1.27
1.07
1.53
1.55
1.65
2.09
1.36
1.27
1.81
1.83
1.10
1.41
1.24
1.68
1.41
1.01
1.03
1.12
1.08
1.30
1974-1995
1956-1995
1974-1995
1956-1995
1974-1995
1947:I-1996:IV
1974:I-1996:IV
0.51
0.28
0.26
0.26
0.25
0.30
0.14
1.15
1.21
1.17
1.19
1.21
1.26
1.13
1.08
1.52
1.38
1.48
1.50
1.39
1.40
1.12
1.10
1.04
1.12
0.98
1.41
1.48
France
West Germany
Italy
Japan
UK
USA
USA
a "var" denotes variance, "corr" correlation, Q = output, S = final sales, A H - change in inventories. The
variables are linked by the identity Q = S + AH.
b In all but the last row, data are annual and real (1990 prices), with Q=real GDP, S=real final sales,
AH=real change in aggregate inventories. In the last row the data are quarterly and real (1992 prices),
with S=final sales of domestic business goods and structures, AH=ehange in non-farm inventories,
and Q _=S + AH. See the text and Data Appendix for sources.
c In colmnns 1 and 2, Q and S were linearly detrended, with the full sample estimates allowing a shift in
the constant and trend term in 1974 (1974:I in the last row); AH was defined as the difference between
detrended Q and S. In columns 3 and 4, AQ and AS were simply demeaned, again with the full sample
estimates allowing a shift in the mean in 1974 (1974:I).
d In column 4, the term E(Q 2 - S 2) essentially is the difference between the variance of Q and the
variance of S, computed in a fashion that allows for unit autoregressive roots in Q and S. See the
Technical Appendix for further details.
e The post-1973 sample, as well as the post-1973 shifts in the full sample estimates, were included to
allow for the general slowdown in economic activity.
875
I
Ig
2b
2~
lg
2b
2a
8.7077
v,%.
o~
01
/
o
inventories
,--4
/
i ...........................................
quarter
5.01083
f_
(7}
01
quarter
As many readers no doubt are aware, similar findings have been reported for many
though not all data sets. A brief summary of estimates o f variance inequalities in
studies using aggregate data: Fukuda and Teruyama (1988) report comparable results
876
for industrialized countries, but also conclude that by contrast in less developed
countries GDP tends to be smoother than final sales 3. Beaulieu and Miron (1992)
report that in manufacturing in some industrialized countries, seasonals in production
are no less variable than those in sales.
For US quarterly economy-wide or monthly two-digit manufacturing data, the
pattern is as pronounced as in Table 5. This applies first of all to demeaned or
detrended data such as is reported in Table 5. For example, Blinder (1986a) reports
var(Q)/var(S) > 1 for 18-20 two-digit manufacturing industries. It also applies to
deterministic seasonals [West (1986), Miron and Zeldes (1988), Cecchetti et al.
(1997)]: in US manufacturing, seasonal variation in production tends to be larger than
seasonal variation in sales 4.
Finally, for both deterministic and stochastic terms, studies that have taken sampling
error into account sometimes but not always find it quite unlikely that sampling error
alone accounts for the lack o f evidence production smoothing [West (1986, 1990b)].
Does aggregation substantially account for the lack of production smoothing
evident in these studies, either because o f measurement error in aggregate data sets,
or heterogeneity across firms? Probably not. It has been argued persuasively that
disaggregate data measured in physical units are more accurate than the aggregate data
used in most studies [see Fair (1989), Krane and Braun (1991), and Ramey (1991)] 5
But as summarized below, studies with disaggregate data still find production more
variable than sales in many cases.
For evaluation of the effects o f firm heterogeneity, analytical arguments are not
particularly helpful. I f var(Q) > var(S) for a single firm, the inequality may be shown
analytically to apply to an aggregate o f firms if each individual firm solves a
linear quadratic problem such as the one presented below with identical parameters,
regardless o f the correlation o f demand shocks across firms [West (1983)]. But if
the cost functions are different for different firms, analytical results appear not to
be available. Lai (1991) and Krane (1994) show by example that aggregate variance
ratios might differ substantially from individual firm ratios. And even if we assume
identical parameters across firms, which as just noted implies that var(Q)/var(S)<~ 1 in
the aggregate, this aggregate ratio may be larger (or smaller) than that o f individual
firms, with the direction o f the bias depending on the correlation across firms of
demand shocks.
So to consider possible biases from aggregation we must turn from analytical to
empirical studies of data disaggregated to the firm or perhaps physical product level.
3 So far as we know, there have been no systematic attempts to explain this finding. It is possible that
measurement error plays a large role.
4 Carpenter and Levy (1998) report a related finding: for manufacturing, the spectra of inventory
investment and production show very high coherence at seasonal frequencies.
5 And the G7 data that we use for illustration are among the worst measured. In some countries, the
change in inventories apparently is constructed at least initially as the difference between product and
income estimates of GDP, and thus includes the statistical discrepancy. See West (1990a).
877
The disaggregate picture is broadly similar though less striking than the aggregate one,
at least in the relatively well-studied USA. Production smoothing is markedly absent in
the automobile industry [Blanchard (1983), Kashyap and Wilcox (1993)]. But Krane
and Braun (1991) found production is less variable than sales in about two thirds of
a set of 38 physical products. Finally, Schuh (1996) found deseasonalized production
less variable than deseasonalized sales in only one fourth of 700 manufacturing firms;
deterministic seasonals in production were less variable than those in sales in about
half the firms. For the deseasonalized data, Schuh reports that the median ratio of
production to sales variance was about 1.1, which is consistent with the US figure
reported in column 2 of Table 5.
2.2. Persistent movements in the inventory-sales relationship
Our second stylized fact is that the inventory-sales relationship is highly persistent.
While there may be a steady state linear relationship between inventories and sales,
movement towards that steady state is very slow. This characteristic may be more
recognizable to inventory experts if it is stated as a "slow speed of adjustment"; a link
between persistence in the inventory-sales relationship and the speed of adjustment is
demonstrated formally in Sections 4 and 5 below. Section 2.2.1 illustrates this with the
annual G7 and quarterly US data used above, Section 2.2.2 documents it with citations
to various papers.
2.2.1. Illustrative evidence
To illustrate persistence, we use a standard technique described in the Technical
Appendix to attempt to find a stationary relationship between the levels of inventories
and sales. This yields Ht - ~OSt for a parameter 0 estimated from the data. Those
familiar with the literature on cointegration will recognize Ht - OSt as the (estimated)
error-correction term if inventories and sales have unit autoregressive roots and are
cointegrated. We refer to H t - OSt as the inventory-sales relationship, sinee it is a
generalized linear version of the inventory-sales ratio. More precise terminology for
this variable is "deviation from the long-run inventory-sales relationship." In our view,
the disadvantages of the length of this term outweigh the advantages of being more
precise.
After obtaining 0, we estimate the first two autocorrelations in the putatively
stationary variable Ht - "OSt, to gauge the extent of persistence in the relationship. We
emphasize that our aim is merely to document quickly evidence of high persistence,
not to work towards a complete time series model, nor even to endorse the use of unit
roots as a modeling device; among other tasks, the latter would require testing for unit
roots in Ht and St, perhaps allowing for a one-time shift in mean in 1974, and so on.
Rather, this is a way of organizing facts and theories about inventories that has some
rough plausibility for a wide range of data.
878
Table 6
Persistence in stationary linear combinations of inventories and sales a,b
Country
Period
Canada
1956-1995
0.16
0.92
0.82
France
1956-1995
0.31
0.95
0.89
West Germany
1957 1994
0.27
0.93
0.82
Italy
1960 1994
0.45
0.88
0.80
Japan
1956-1995
0.22
0.97
0.91
UK
1956-1995
0.20
0.95
0.87
USA
1956-1995
0.24
0.88
0.82
USA
1947:I-1996:IV
0.68
0.94
0.88
USA
1947:~1996:IV,
H and S in logs
1.12
0.95
0.90
6 The technique used is not invariant to normalization. For the annual data, we re-estimated with
S t on the left instead of H t. Positive estimates of (1/0) resulted. On the other hand, if time trends are
879
P-_ B.7077
B.7077
P-_
I/
-4.58753
inventory-sales
relationship
V'
quapteP
Fig. 5. Response to sales equation shock, quarterly VAR.
the whole sample but allowing for different means in the 1947:I-1973:IV and 1974:I-1996:IV sample.
This reduced the estimate only slightly, to 0.89 and 0.94 (vs. the 0.94 and 0.95 values in Table 6).
Rossana (1998) carefully investigates the question of stability in two-digit US data. He finds evidence
of instability, and lack of cointegration within subsamples.
s Now seems an appropriate time to comment on the use of logs versus levels of the data, in response to
comments by two of the readers of an earlier version of this paper. Much empirical work in inventories,
and a distinct majority of work using intertemporal dynamic models, has relied on levels rather than
logarithms of the variables in question. We follow that convention in most of this paper. Working in
levels has the advantage of preserving the identity that links inventories, production and sales [see
Equation (2.1)]. In addition, inventory investment, as reported by the National Income and Product
Accounts, is defined as the change in the level of inventories, and is frequently negative, which further
discourages the use of logarithms. In a few cases, such as in the last row of Table 6, we do use logarithms.
That little turns on levels versus logs is suggested by the similarity of the results in the last two rows
in Table 6, and, more generally of the review of the literature that we are about to present: flexible
accelerator studies typically use data in logs, while linear quadratic studies typically use levels. Both
find great persistence.
880
I
3.74996
~ . ,,'~'~
~4
"0
t~
t~
O~
~4
quapter,
Fig. 6. Response to inventoryequation shock, quarterlyVAR.
3 and 4. The inventory-sales relationship is computed using the estimated value of
= 0.68 from Table 6 to form a linear combination of the impulse response functions
for sales and inventories.
In response to a positive one standard deviation shock to the sales equation,
inventories rise, but less rapidly than do sales (see Figure 3). This induces the fall in
the inventory-sales relationship depicted in Figure 5. Eventually, inventories build up
more rapidly, leading to overshooting and an increase relative to the initial position.
Since both inventories and sales are close to their pre-shock values after six years
(see Figure 3), the inventory relationship returns as well, as Figure 5 shows. By
contrast, a shock to the inventory equation leads to a complicated pattern displaying
more persistence. The inventory-sales relationship rises initially by construction, and
then declines erratically for eight quarters before briefly rising again and then slowly
decaying. At the end of six years, a return to the pre-shock value is not evident:
these unrestricted VAR estimates suggest great persistence in the inventory-sales
relationship.
2.2.2. A suruey o f results
Table 6 and Figures 5 and 6 suggest that there is little mean reversion in economywide inventory-sales relationships. Congruent evidence comes from two sources,
unstructured tests such as in Table 6, and structural estimates of what is called a "speed
of adjustment".
Consider first the unstructured tests. Using quarterly economy-wide data, West
(1990b, 1992b) cannot reject the null of no cointegration for the USA and for
881
Japan. Using monthly US data, Granger and Lee (1989) find only mild evidence
for cointegration in monthly two-digit manufacturing and trade series; fewer than a
third of their 27 data sets reject the null of no cointegration at the 5% level, and the
median first-order autocorrelation of the putatively stationary linear combination of
inventories and sales is about 0.99. Rossana (1993, 1998) uses similar data and allows
a vector of cost variables W t to enter the cointegrating relationship. He uses regression
techniques to search for a combination H t - OSt - otI W t that is stationary. Wt is defined
to include real wages, real materials prices, nominal interest rates and inflation. In the
end, however, he finds mixed evidence of cointegration across H t , St and Wt. (He does
not report autocorrelations o f stationary linear combinations.)
In addition, a large literature has used structural inventory models to estimate the
"speed of adjustment". We shall describe such models below. For the moment, what
is relevant is that under conditions described below, a slow speed of adjustment is
equivalent to persistence in the inventory-sales relationship.
Let p be the largest autoregressive in H t - OSt, or H t - OSt - ct~Wt, with the latter
variable the relevant one if, as in the Rossana (1993, 1998) papers cited above, a vector
of cost variables is entered into the cointegrating relationship. The large empirical
literature has estimated inventory equations, with results implying that ~ is near 1.
A typical value in quarterly data is around 0.8-0.95.
The following are some examples. Using quarterly economy-wide data for some
industrialized countries, Wilkinson (1989) found ~ - 0.75-1.0, with W t including
one or more of: sales shocks, inflation, wages, raw materials prices and capacity
utilization. Similar results have repeatedly been found using monthly or quarterly US
manufacturing data. Examples include Maccini and Rossana (1981), Blinder (1986b),
and Haltiwanger and Maccini (1989). In these papers, Wt included one or more of:
factor prices such as wages, raw materials prices and real and nominal interest rates,
other factors of production such as labor, and sales expectational errors.
Does aggregation across heterogeneous firms substantially account for this high
serial correlation? We are not aware of analytical arguments establishing genera]
conditions under which aggregate estimates of p will be higher than typical firmspecific estimates, though no doubt such arguments could be constructed. Simulations
in Lovell (1993) and a limited amount of empirical evidence suggests that aggregation
does impart a bias. Schuh (1996) reports that for monthly data for 700 manufacturing
firms, the median estimate of p is about 0.6 (Wt includes a real interest rate); for
quarterly data for some publicly owned firms, and with Wt including measures of
credit conditions, ~ is reported to be about 0.6-0.8 by Carpenter et al. (1994, 1998),
although a somewhat higher value of about 0.9 is found by Kashyap et al. (1994)10
9 The autocorrelation was computed from the Durbin-Watson statistic reported in the last column of
Table II in Granger and Lee, using 2(1 -~) = d.w.
10 A small literature has discussed how estimation ofp is affected when the decision interval for firms
is smaller than the sampling interval of the data [Christiano and Eichenbaum (1989), Jorda (1997)].
Such time aggregation does not appear capable of explaining the high persistence.
882
Of course, tests for cointegration have notoriously low power. And there is
heterogeneity of estimates o f p implied by the flexible accelerator literature. But these
results suggest that there is little evidence that mean reversion of Ht - OSt towards its
mean takes place rapidly, and considerable evidence of persistence. To interpret such
persistence, as well as the procyclical behavior discussed in the previous section, we
now present a standard inventory model.
The linear quadratic inventory model dates back to Holt et al. (1960). Although this
book is written in an operations research style that suggests that its main aim was to
provide practical advice to managers, it may still be profitably reviewed by economists
interested in inventory behavior. In fact, Holt et al. (1960) develop models more general
than those in many of the applications we review here: they allow for stockout costs,
order backlogs, and stochastic variation in costs.
Early uses of a linear quadratic model to interpret macro data include Childs
(1967) and Belsley (1969). Tools developed in the 1970s and 1980s to estimate and
interpret decision rules and first-order conditions from rational expectations models
[e.g., Hansen and Sargent (1980), Hansen and Singleton (1982)] were subsequently
used by many authors to estimate one or another version of the model. Some papers
used aggregate data (sometimes economy wide, sometimes at the two-digit level); some
used data at the level of individual firms or physical products.
It is this more recent literature that we review in this and the next four sections.
The focus of this literature, and of our discussion, is on how production parameters
and constraints influence the intertemporal interaction between production, sales and
inventories. While different papers of course vary in the details of the model, there is
sufficient commonality that the model presented in Section 3.2 may fairly be described
as representative. Section 3.3 derives a first-order condition. Section 3.4 discusses
whether the model is applicable to a wide range of inventories.
Table 7 lists notation, and may be a useful reference in this and subsequent sections.
3.2. A model
We assume that a firm maximizes the present value of future cash flows. In macro
applications this will be a representative firm. In the formal statement of the model
about to be given, we omit constant, linear and trend terms for notational simplicity.
Earlier work provides tediously detailed treatment o f such terms, allowing for trends
that are either arithmetic [West (1983)] or geometric [working paper versions of West
(1988, 1990b)].
Ch. 13:
Inventories
883
Table 7
Variable and parameter definitions a
H t -H i
H t -- OS t
inventory-sales relationship
Qt
production in period t
St
sales in period t
Uct
uct
Udt
Wt
Mnemonic
Description
a0
Section b
3.2
a1
cost o f production
3.2
a2
3.2
a3
3.2
discount factor
3.2
4.2
3.3
3.2
0c
4.2
0d
4.2
qSw
4.2
= a 3 [al(1-b)/ba2] , coefficient on S t in H i
3.3
:~H
4.2
:v t, :v2
4.3
a All variables and parameters are scalars, with the exceptions of ewt, W t , a, ~t and q~w. The variables
in panel A are introduced in Section 3.2.
b Section where parameter is introduced.
W e u s e t h e f o l l o w i n g n o t a t i o n : P t is real p r i c e (say, r a t i o o f o u t p u t p r i c e to t h e w a g e ) ,
St
real e n d o f p e r i o d i n v e n t o r i e s , Ca real p e r i o d
c o s t s , b a d i s c o u n t f a c t o r , 0 ~< b < 1, a n d E t m a t h e m a t i c a l e x p e c t a t i o n s c o n d i t i o n a l o n
884
max l i m r ~
Et Z
j-o
subject
(3.1)
( Qt = St + He -
He 1.
The scalar Uct is a cost shock, and, as discussed below, may depend on both observable
and unobservable variables. The term Pt +jSt +j is revenue. The analysis in this section
does not depend on specification of demand or market structure, so we defer discussion
o f P t + j S t +j until the next section.
The cost function Ct allows two possible roles for inventories. One is a production
smoothing role, in which inventories facilitate intertemporal allocation of production.
(N.B.: in much of the inventory literature, the phrase "production smoothing"
references smoothing from demand shocks; we use it to reference smoothing from
cost shocks as well.) This role is reflected in the terms in AQ 2 and Q2. The second
role is a revenue role, in which inventories allow a firm to satisfy demand that cannot
be backlogged. This role is reflected in the " a 3 S t " term in (Ht-1 - a3St)2; (He 1 - a 3 S t ) 2
induces an accelerator motive. In our discussion o f these terms, we assume for the
moment that a~ and a2 are positive, ao and a3 nonnegative.
The first production smoothing term, aoAQ2t, captures increasing costs of changing
production and of production. This represents, for example, hiring and firing costs.
Not all authors include this term, and, as discussed in Section 7, some empirical tests
find estimates of ao insignificantly different from zero.
The second production smoothing term, a~Q 2, reflects costs of production. It can
be interpreted as the second order term in a quadratic approximation to an arbitrary
convex cost function associated with a decreasing returns to scale technology. In data
with trends, this approximation would likely be around a growth path. (Recall that
constant and trend terms are omitted for notational simplicity.)
The accelerator term a 2 ( n t - 1 - a 3 S t ) 2 embodies inventory holding and backlog
costs. Consider first when a3 = 0, so that the term becomes a2H2t 1. Then this can
be interpreted as the second order term in a quadratic approximation to an arbitrary
convex inventory holding cost function. When a3 ~ 0, the term is intended to reflect
backlog (stockout) and batch as well as inventory holding costs, and thus captures a
revenue-related motive for holding inventories. Stockout costs arise when sales exceed
the stock on hand, perhaps entailing lost sales, perhaps entailing delayed payment if
orders instead are backlogged. Batch costs vary inversely with the stock of inventories,
since fewer production runs and larger lot sizes imply larger inventory levels on
average.
Ceteris paribus, the higher the stock of inventories, the less likely is a stockout and
the lower are stockout costs. As well, higher stocks result when the number of batches
885
falls, since lot sizes rise. On the other hand, higher stocks entail higher inventory
holding costs. This quadratic term approximates the tradeoff between the two costs,
with a3 rising as stockout costs rise relative to backlog costs. Holt et al.'s (1960) formal
derivation o f this time invariant approximation to this inherently nonlinear, and timevarying, cost is presented in Section 8. Note in any case that in many applications,
inventories are strictly positive and large relative to sales in all time periods, perhaps
in some data sets because o f aggregation over heterogeneous firms. So in such data,
careful treatment of nonlinearity may have limited empirical payoff. Section 8 below
discusses alternative approaches.
The final term in the cost function is UotQt. This captures exogenous stochastic
variation in costs. (We omit terms o f the form cost shock x A Q t and cost shock H t
to avoid needless algebraic complications.) In some applications Uct = 0 and this shock
is absent [West (1986)]; in others it is not observed by the economist and follows an
exogenous process [Eichenbaum (1989)]; in still others it has both observable and
unobservable components [Ramey (1991)]. To cover all three cases, we write
Uct = a' W t + Uct.
(3.2)
A n optimizing firm will not expect to increase cash flow by producing one more
unit this period, putting the unit in inventory, and decreasing production by one unit
next period, all the while holding revenue constant. Formally, upon differentiating the
objective function (3.1) with respect to H t we obtain 11
E t [a0(AQ, - 2bAQt +1 + b 2zxQt+2) + a l (Qt - bQt+ l) + ba2 (Ht - a3S,+ 1) + Uct
b Uct+i ] = O.
(3.3)
For discussion of the second o f our stylized facts (persistence in the inventory-sales
relationship), it will be helpful to note a low frequency implication o f Equation (3.3),
namely, that inventories and sales are cointegrated if St is I(1) and Uct is I(0). [Here,
11 The first-order condition derived assuming that [O(pt+jSt+j)/OHt]- O. This assumption is not
particularly appealing, since a 3 > 0 is motivated in part by stockout costs, and presumably increases
in H t will decrease stockouts and increase revenues (increase p,St). As noted above, our modeling of
stockout costs is crude but given data constraints in some applications the gains from more sophisticated
treatments perhaps are small. See Section 8 below for alternative treatments.
886
we use standard time series notation: a variable that is integrated o f order 0 - I(0),
for short - is one that is covariance stationary, with a spectrum that is finite and
strictly positive at all frequencies. An "integrated" variable - I(1), for short - is one
whose arithmetic difference is I(0). I(1) variables are sometimes called "difference
stationary", or as having unit autoregressive roots (a term that we used in Section 2).]
To see the cointegration result, write the term in brackets in Equation (3.3) as xt+2,
so that (3.3) is Etxt+2 = 0. If we replace expectations with realizations, we can write
(3.4)
Note that even if some or all o f the variables that comprise xt+2 are integrated, ~t+2 will
still be stationary. Kashyap and Wilcox (1993) observe that xt+2 may be rewritten
Xt+2 =
Agt+l) + all'r-It
(3.5)
al(1 - b )
O~a
ba2
~
=~
H t = OSt,
Ht -
1-b
O = a3 - al - ba2 '
OSt + a ' W t ,
a =_
(3.6)
l-b_
--a.
ba2
Observe that this result does not require parametrization of the demand curve, or
specification o f market structure. Rossana (1995, 1998) assumes exogenous revenue
12 A variable xt is 1(0) around trend if Ext = m o + m lt for some m l ~ 0 and xt -Ex~ = I(0). Such variables
are sometimes called trend stationary.
887
and uses the resulting decision rule to provide a complementary proof that Uct ~ I(O)
[Uct ~I(0)] implies cointegration between H t and St (between H t , St and Wt).
3.4. Whose inoentories?
4. Decision rule
4.1. Introduction
Section 4.2 briefly reviews alternative treatments of demand, and solves for a decision
rule under a simple specification. Section 4.3 derives the process followed by the
13 "Production to stock" industries are ones that typically sell finished goods off the shelf. By contrast,
"production to order" industries are ones that maintain order backlogs, often deferring final assembly
until orders are in hand. See Abramowitz (1950) and Belsley (1969).
888
inventory-sales relationship, and may be skipped without loss of continuity. Section 4.4
summarizes the implications of Section 4.3 for persistence in the inventory-sales
relationship. We present a detailed treatment o f the inventory-sales relationship
because this process has not received much direct attention in existing literature.
Discussions o f procyclicality in subsequent sections will cite analytical results in
existing literature.
4.2. Derioation o f decision rule
Pt = - g S t + g [ f dt ~ -gSt + Udt.
(4.1)
In (4.1), Udt and Udt are stochastic. The demand curve is written in this form so
that exogenous sales is a special case of Equation (4.1), implemented by letting the
parameter g --+ oo and specifying ~rdt as exogenous; with g ---+ oo, St = Udt. In
practice, one needs to allow for serial correlation in Udt. In principle one might want
to rationalize such serial correlation with (say) costs of adjustment on the part of
purchasers, or with observable shifters o f the demand curve [West (1992b)]. But since
the model focuses on production, and, moreover, is typically not used to study the
effects of a hypothetical intervention or change in regime, taking such serial correlation
as exogenous is a useful simplification that will be maintained here.
Finally, Christiano and Eichenbaum (1989) and West (1990b), building on Sargent
(1979, ch. XVI) derive the linear demand curve (4.1) in general equilibrium. Both
papers assume a representative consumer whose per period utility is quadratic in St
and linear in leisure. The disturbance Udt is a shock to the consumer's utility. There is
no capital; the only means o f storage is inventories. See the cited papers for detail.
14 TO prevent confusion, we note that the first-order condition (3.3) also results if one assumes cost
minimization. So if one aims to use condition (3.3) to estimate model parameters, one can motivate the
equation by reference to cost minimization without taking a stand on the how revenue is determined
(apart from the caveat stated in Footnote 11).
889
Here, we do not derive Equation (4.1) in general equilibrium but take (4.1) as given.
We do not attempt to trace the demand shock back to preferences or other primitive
sources. We therefore caution the reader that despite the label "demand", Udt should
not be thought of as literally a nominal or monetary shock, since it (like all our
variables) is real. As well, one can imagine scenarios in which Udt reflects forces
typically thought of as supply side. If the good in question is an intermediate one, for
example, one can imagine that shocks to the technology of the industry that produces
the good dominate the movement of Uat.
Whatever the interpretation of Udt, we derive an industry equilibrium assuming a
representative firm. Even so, to obtain a decision rule, we must be specific about market
structure and the structure of the demand and cost shocks. We assume here that the
market is perfectly competitive, and normalize the number of firms to one. If the firm
is a monopolist, the reduced form is identical, but with a certain parameter being the
slope of the marginal revenue curve rather than the slope of the demand curve.
We assume that Wt, Udt and Uct follow exogenous AR processes, possibly with
unit autoregressive roots. By "exogenous" we mean "predictions of Wt, Udt and uot
conditional on lagged Ws, Uas and uos are identical to those conditional on lagged
Ws, Uds, uos, and industry-wide H s and Ss: Wt, Uat and uot are not Granger-caused
by industry-wide Ht or St". (Of course in general equilibrium, such exogeneity of Wt
is doubtful.)
Finally, for notational simplicity, and to make contact with the literature on the
"speed of adjustment" (see the next section), we tentatively assume that
a0 = 0.
(4.2)
This assumption is arguably not a good one empirically, and we will relax it below.
Under the assumption of perfect competition, there are two equivalent methods for
deriving the decision rules for inventories and sales. The first method, which studies the
decentralized optimization problem, derives the individual firm's first-order conditions
and then incorporates those into the industry equilibrium. The second method, which
uses a social planning approach, derives the first-order conditions for the social planner
problem and obtains decision rules for those. Both methods yield identical answers.
We exposit here the decentralized method, and present in the Technical Appendix
the social planner approach. With a0 = O, the first-order condition for sales St for the
representative firm is
Pt - Et[alQt - a2a3(Ht-i - a3St) + Uct] = O.
(4.3)
In the absence of inventories, this would simply tell our competitive firm to
set marginal revenue Pt equal to marginal cost a l Q t + U c t . The additional term
aaa3(Ht 1 - a3St) is the effect on inventory holding costs of an additional unit produced
for sale.
Upon using Pt = - g S t + Udt in Equation (4.3) and Qt =St + A H t in Equation (4.3)
and the inventory first-order condition (3.3), we obtain a pair of linear stochastic
890
difference equations in Ht and St. This two equation system is solved in the Technical
Appendix. The resulting decision rule is
Ht = YgHHt-1 + distributed lag on uct, Udt and Wt,
(4.4a)
(4.4b)
In (4.4a), srH is the root to a certain quadratic equation, ]~H[ < 1. Both ~ / a n d ~s
depend on b, g, al, a2 and a3. (Note two differences from the relatively well-understood
case of exogenous sales. Even when as, a2 > 0, i f g < oc: (a) it is in principle possible
to have :v~ ~<0, and (b) the accelerator coefficient a3 affects ~/t.) The distributed
lag coefficients on Uct, Udt and Wt depend on b, g, al, a2 and a3 as well as the
autoregressive parameters governing the evolution of the uct, Udt and Wt. In the
empirically relevant case of z~H > 0, ~H increases with marginal production costs al
and decreases with marginal inventory holding costs a2. The signs of OJvH/Oa3 and
OYfH/Og a r e ambiguous.
The solution when revenue is exogenous (g --+ c~) is obtained by replac~g Udt
with gUdt [see Equation (4.1)] and letting g ~ c~. In this case, :Vs =0, St = Udt and
the solution (4.4) may be written in the familiar form
H t = ~HHt I + distributed lag on St and on measures of cost,
(4.5a)
(4.5b)
On the other hand, when revenue is endogenous, :rs ~ 0 and we see in Equation (4.4b) that inventories Granger-cause sales. The intuition is that forward looking
firms adjust inventories in part in response to expected future conditions. Thus
industry-wide stocks signal future market conditions, including sales. This signalling
ability is reflected in Equation (4.4b).
These same results can be obtained directly from the social planner problem that
maximizes consumer surplus plus producer surplus, which is equal to the area between
the inverse demand and supply curves. See the Technical Appendix.
In empirical application, matching the data might require allowing shocks with
rich dynamics. Such dynamics may even be required to identify all the parameters
of the model. Blanchard (1983), for example, assumes that the demand shock follows
an AR(4). For expositional ease, however, we assume through the remainder of this
section that all exogenous variables - Udt, uct, Wt - follow first-order autoregressive
processes (possibly with unit roots). Specifically, assume that
Et-1 Wt = (I-)wWt 1,
Wt = CrAwWt-1 + ewt,
Et_lewt = O,
I I - ~wZl = 0 ~ Izl/> 1,
Et l Uct = q~cuct-1,
Et-1 Udt = ~ d U d t - 1 ,
Et-leot = 0,
Et ledt = O,
[q~c[ ~< 1,
]d[
~< 1.
(4.6)
Ch. 13:
Inventories
891
(Given the growth in the number of symbols, it may help to remind the reader that
Table 7 summarizes notation.) The Technical Appendix shows that the distributed lags
in Equation (4.4) are all first order, and Equation (4.4) is
H t = YfHHt-1 +f~/w W t + f HcUct + f HdUdt,
(4.7a)
St = ~sgt-I
(4.7b)
+ f t s w W t + f scUct + f sdUdt.
See the Technical Appendix for explicit formulas for the ')"s in terms of b, g, al, a2,
a3, qJw, q~c and q~d. Of course, if ~ = 0 so that Uct =uct, thenfHw = f s w =0.
4.3. Persistence in the inventory-sales relationship
(4.8)
Recall that if Uct is stationary, H t - OSt is stationary as well, where the cointegrating
parameter 0 is defined in Equation (3.5).
When there are no observable cost shifters (~ = 0 ~ fHw = f s w = 0), tedious
manipulation o f Equation (4.7) yields
H t - H t =~ H t - OSt = ~ H ( H t - l
(4.9)
where moo, ml~ and m0d depend on O, f H o , f H d , f S c and f S d (see the Technical
Appendix). Let "L" be the lag operator. Since ( 1 - ~ c L ) u c t = e c t , it follows from
Equation (4.8) that
(1 - ~,vL)(1 - (bcL)(Ht - H t ) = vt,
vt = mocect + m~cect-i + mOdedt -- ~)cmOdedt-1 ~ MA(1).
(4.10)
Thus, H t - H t ~ ARMA(2, 1) with autoregressive roots Jr/4 and ~bc. (This presumes that
the moving average root in vt does not cancel an autoregressive root in H t - H i, which
generally will not happen.) Note that the innovation edt, rather than the shock Udt,
appears in Equation (4.9) and thus in Equation (4.10). With q~d ~ 1, however, the right
hand side of Equation (4.10) would include a linear combination of Udt and Udt-I
that would not reduce to a linear function of edt, and ~d would also be one of the
autoregressive roots of H t - H i . In this case, if ~d ~ 1, then Ht - H i would also have a
moving average root that would approximately cancel the autoregressive root of ~bd.
Similarly, when there are observable cost shifters (a ~ 0), it may be shown that
Equations (4.6) and (4.7) imply
H t - H i = H t - OSt - TIWt = :rH(Ht-i - OSt-1 - a1Wt-1) + disturbance,
disturbance
(4.11)
Once again, persistence in H t - H 2 is induced by ;r,q and q~c.
892
(4.12)
It is well known that when revenue is exogenous (g ---+ oc), costs of adjusting
production put additional persistence in inventories [Belsley (1969), Blanchard
(1983)]: in this case Equation (4.5a) becomes
H t = Y g H I H t I + J~H2Ht 2 q- distributed lag on S t and on measures of cost,
(4.13)
with Y~H2~ 0. Unsurprisingly, inventory decisions now depend on Qt-1 =St-1 +Ht-1 H t - 2 and thus on Ht-2, even after taking into account Ht-t and the sales process.
As one might expect, the presence of costs of adjusting production has a similar
effect even when sales and revenue are endogenous, and on the inventory-sales
relationship as well as inventories. The Technical Appendix shows that a0 ~ 0 puts an
additional autoregressive root in Ht - H t , which now follows an ARMA(3, 2) process.
One autoregressive root is Oc. We let Jrl and Y~2 denote the two additional (possibly
complex) roots. These are functions of b, a0, al, a2, a3 and g. Intuition, which
is supported by the simulation results reported below, suggests that increases in a0
increase the magnitude of these roots.
4.4. S u m m a r y
on persistence
in t h e i n v e n t o r y - s a l e s
relationship
We summarize the preceding subsection as follows: assume the shocks follow the
AR(1) processes given in Equation (4.6), with the additional restriction (4.8) that the
demand shock and observable cost shifters follow random walks. Then
ao = 0 ~
(4.14)
The root ~H is a function of b, g and the ai, but not the autoregressive parameters of
the shocks, and is increasing in the marginal production costs at. In addition,
ao ~ 0 ~
Ht - Ht
=-- H t - OSt - a t W t ~
0c=0,
Ht-H
ARMA(3, 2),
(4.15)
t~ARMA(2,1)
893
Technical Appendix and simulations reported in Section 6 indicate that the modulus
o f the larger o f the roots increases with a0 and al 15
Thus the persistence documented in Section 2.2 above follows i f there are sharply
increasing production costs (a0 and/or al are sufficiently large) and/or serially
correlated cost shocks. In addition, it is important to observe that qualitatively similar
reduced forms are implied b y the following two scenarios: (1) serially correlated cost
shocks with no costs o f adjusting production, and (2) serially uncorrelated cost shocks
and sharply increasing costs o f adjusting production. We shall return to this point
below.
O f course persistence m a y also follow if we put different dynamics into the shocks
Wt, Uct and Udt.
(5.1)
In (5.1), v > 0 is the weight o f the second cost relative to the first, and ut is an
exogenous unobservable disturbance 16. The first-order condition is then
Ht-Ht-1
= [1/(1 + v ) ] ( H ; - H t - 1 ) -
[1/(1 + v)]ut.
(5.2)
The coefficient 1/(1 + v) is the fraction o f the gap between target and initial inventories
closed within a period. I f v is big (cost o f adjusting inventories is big), the fraction o f
~5 Under the present set of assumptions, then, the parameter called "p" in Section 2.2 is max{~H, $c}
if a 0 = 0, max{l~ l I, 1~2I,Oc} if a 0 0.
16 Ht and St are sometimes measured in logs [e.g., Maccini and Rossana (1981, 1984)], and the
variable ut is sometimes split into a component linearly dependent on the period t surprise in sales
and a component unobservable to the economist [e.g., Lovell (1961), Blinder (1986b)]. We slur over
differences between regressions in levels and logs, which in practice are small (see Footnote 8), and omit
a sales surprise term in the inventory regression, which in practice has little effect on the coefficients
that are central to our discussion.
894
the gap expected to be closed is, on average, small. To make this equation operational,
target inventories H~ must be specified. Let
(5.3)
H t = OS, + a ' W t .
Here, Wt is a vector of observable cost shifters [as in Section 2.2.2 and Equation (3.2)].
Notation has been chosen because of link about to be established with 0 and a ~W t as
defined earlier. Suppose
S t = St-1 + edt,
W t = W t - 1 + ewt,
(5.4)
(5.5)
6. Dynamic responses
To develop intuition about how the model works, and what the two stylized facts
suggest about model parameters and sources of shocks, this section presents some
impulse responses. Specifically, we present the industry equilibrium response of (1) H t ,
St and Qt, or (2) H t , St and H t - OSt, to a shock to Udt or Uct, for various parameter
sets, with no observable cost shifters (a = Wt = 0). While the parameter values we use
are at least broadly consistent with one or another study, we choose them not because
we view one or more of them as particularly compelling, but because they are useful
in expositing the model.
Table 8 lists the parameter sets. It may be shown that the solution depends only
on relative values o f g, ao, al and a2; multiplying these 4 parameters by any
nonzero constant leaves the solution unchanged. [This is evident from the first-order
conditions (3.3), (B.4) and (B.5): doubling all these parameters leaves the first-order
conditions unchanged, apart from a rescaling of the shocks.]
Our choice of a2 = 1 is simply a normalization. We fix g = 1 in part because some
of the properties documented below can be shown either to be invariant to g [see
West (1986, 1990b) on procyclicality of inventories] in part because a small amount of
Ch. 13:
895
Inventories
Table 8
Parameter sets a
(1)
Mnemonic
(2)
g
(3)
a0
(4)
a1
(5)
a2
(6)
a3
(7)
~c
(8)
~d
n.a. b
0.7
n.a.
0.7
0.7
-0.1
n.a.
AI
n.a.
An
0.7
n.a.
n.a.
n.a.
a See Table 7 for parameter definitions. The behavior of the model depends only on the scale of the
parameters g, a0, a~ and a2; doubling all these leaves behavior unchanged. The discount factor b is set
to 0.99 in all experiments.
b "n.a." means that the autoregressive parameter is irrelevant for the impulse responses plotted in
Figures 7-13: the response is for a shock to the other variable, whose AR(1) parameter is given.
e x p e r i m e n t a t i o n indicated little sensitivity to g. To p r e v e n t possible confusion, we note
explicitly that the p a r a m e t e r a3 is identified in absolute terms and not just relative to
other parameters. Throughout, w e set the discount factor b = 0.99, and interpret the t i m e
p e r i o d as quarterly. To facilitate discussion, in the graphs we set vertical tick marks
labelled " - 1 " , " 1 " , "2", and so on, but this (or any other) choice o f traits to measure the
response is arbitrary. [In actual application, the units u s e d w o u l d o f course be m o n e t a r y
(e.g., billions o f 1992 dollars in the impulse responses in Section 2 above)].
The p r o d u c t i o n s m o o t h i n g aspect o f the m o d e l is m o s t clearly e v i d e n t w h e n shocks
are m e a n reverting. We therefore begin with three p a r a m e t e r sets illustrating the
response to an innovation in a stationary A R ( 1 ) d e m a n d shock Ud~, w i t h A R parameter
~d = 0 . 7 . Since q~d = 0 . 7 is probably far e n o u g h f r o m unity to m a k e the notion o f
cointegration b e t w e e n H t and St unappealing, we plot the responses o f Qt, St and H t
but not those o f H t - O S t 17.
Parameter set A illustrates the production s m o o t h i n g model. F i g u r e 7 presents the
response to a d e m a n d shock. A s may be seen, w h e n there is a , p o s i t i v e innovation to
demand, sales o f course rise. But part o f the increase in sales is m e t by drawing d o w n
inventories, thereby b u f f e r i n g production f r o m the d e m a n d shock. A s sales return to
17 Naturally, even though we do not include the plots here we did examine them ourselves. As it
turned out, H t - O S t showed persistence. From Equation (B.11) in the Technical Appendix, we see
that H t - OSt has an autoregressive root of Od that is cancelled by a moving average root only when
0d --4 1. This autoregressive root apparently explains the persistence. In our view such persistence is not
particularly interesting: in a stationary model, 0 = a3 - [al (1 - b)/ba 2 ] is not the parameter corresponding
to a projection o f H t onto St, and thus H t - OS t does not correspond to the quantity displaying persistence
in, for example, Table 6.
896
5
43-
"-1
2.t-
O-
~,
inventories
quarter
the steady state, inventories are gradually built back up. It may be seen in the graph
that production is smooth relative to sales.
The intuition is straightforward: given increasing marginal costs (al >0), it is
cheaper to produce at a steady rate than to produce sometimes at a high rate, sometimes
at a low rate. So the increased demand is met partly with inventories, and production is
smoothed relative to sales. (Note that this logic applies even for a competitive firm that
can sell as much as it wants at the prevailing market price.) Inventory movements are
countercyclical, in the sense that they covary negatively with sales. It may be shown
analytically that such cotmtercyclical behavior will obtain when a l > 0, a3 = 0 and there
are no cost shocks [West (1986) for a stationary model, working paper version o f West
(1990b) for a model with unit roots].
One can obtain procyclical movements when costs are convex if the accelerator
term is operative (a3 > 0) and is sufficiently strong to offset the production smoothing
motive. In Figure 8, which shows results when a3 = 1 rather than a3 = 0, inventories
initially rise along with sales when there is a positive innovation to the stationary
demand shock. So production rises even more than does sales, and is more variable.
All three variables then fall smoothly back towards the steady state.
Some algebra may help with intuition: if ao=a] =0, and Uct=O, the firstorder condition for Equation (3.1) is simply Ht=a3EtSt+l. Thus inventories will
covary positively with expected sales, and thus with sales themselves since St is
positively serially correlated in equilibrium. With a0 0, al ~ 0, inventory movements
will reflect a balance of accelerator and production smoothing motives. I f the
897
I
-t-I
e-
0
I
quarter
accelerator motive dominates, as it does in this parameter set, inventories will move
procyclically IS.
Another way to obtain procyclical movements in response to demand shocks, is with
nonconvex production costs [Ramey (1991)]. Parameter set C captures this with a small
negative value for al. [The linear quadratic problem will still be well-posed, and lead
to an internal solution, as long as the nonconvexity is not too marked; in the present
context, this essentially demands that a2 and g be sufficiently large relative to ]al I.
See Ramey (1991).] We see in Figure 9 that a positive innovation to demand causes
inventories to rise (though by a small amount - an artifact o f our choice o f parameters):
with al < 0 it is cheaper to bunch rather than smooth production. Thus, firms build up
inventories when sales are high. If there is a cost o f changing production (a0 ~ 0),
marginal production costs are (1 +b)ao +al. Ramey (1991) has noted that al < 0 may
induce a tendency to bunch production even if (1 + b)ao + a~ > 0 [see West (1990b) for
a particular set o f parameters for which this happens].
We now turn to parameter sets with a unit root in the demand shock (q~d= 1).
Figure 10 plots the response o f inventories, sales and the inventory-sales relationship
18 Recall that the accelerator term is motivated in part by stockout costs. Kahn (1987) rigorously shows
that when nonnegativity constraints are imposed, demand uncertainty (which implies uncertainty about
whether a stockout will occur) will lead to procyclical movements if demand is serially correlated.
898
~.2
10
8
4
2
0
2_
_'~-~_ 2 _
_"-2"~-_ _ - _ ' _
_ _-_,_
_ ~ =
4
quartsr
19 To prevent confusion: sales and revenue are endogenous in this experiment (g<ec). Although
St looks like a random walk in the figure, in fact ASt does have a little bit of serial correlation,
and is Granger-caused by inventories.
899
I
sales
3
e-
]
inventory-sales relationship
/ J_/
inventories
quarter
Fig. 10. Response to a unit root demand shock; parameter set A ~.
of all three parameter sets, however, was rapid mean reversion in H t - OSt. The
autoregressive root x~/ [see Equation (4.9)] was 0.270 (parameter set A), 0.269 (B)
and -0.148 (C).
The negative sign of xH in parameter set C perhaps deserves a word of mention.
Recall that al < 0 in this parameter set, and that downward sloping marginal costs
induces production bunching. So bunching generates a negative autocorrelation, since
high activity in one period tends to be followed by low activity in the next.
In parameter set A", we continue to use the same technology parameters, but now
plot responses to a cost rather than demand shock. The cost shock is stationary, with
an AR parameter of q~o= 0.7. We see in Figure 11 that inventories move procyclically
in response to a negative cost shock: the shock causes both inventories and sales to
rise, and makes production (not depicted) more variable than sales. (Recall that we are
studying industry equilibrium, and sales of course change as costs change.)
The intuition once again is straightforward: a firm with a convex cost function will
use periods of low cost to produce a lot and to build up inventory stocks (as in the
figure), and use periods of high cost (not depicted) to produce little and instead sell out
of inventory stocks that have already been built up. In this case, inventories serve to
buffer production from cost shocks. (Reminder: the phrase "production smoothing" is
conventionally used only to describe smoothing from demand shocks but not, as in the
present paragraph, smoothing from cost shocks.) Inventory movements are procyclical.
It may be shown analytically that such procyclical behavior will obtain when al > 0,
900
4
sales
=~1
r-
4
quarter
20 Of course, for the computed response of H t - OSt to give us insight into the empirical behavior
of H t - OSt, there must be unit roots or near unit roots in H t and St, and hence this experiment and
parameter set cannot provide the whole explanation for the two stylized facts.
Ch. 13:
901
Inventories
I
5
43-
B
.H
t--
2~.-
~--i
O-
....
inventories
I
quarter
Fig. 12. Response to a unit root demand shock; parameter set D.
5
4
/ ~
inventories
3
2
i
-I
/
-2
-3
\//
inventory-sales relationehlp
quarter
Fig. 13. Response to a unit root demand shock; parameter set E.
902
Table 9
Possible explanations of stylized facts a
Inventories
procyclically
Persistence in
inventory sales
relationship b
no
yes
yes
yes
i"10
yes
yes
Inove
no
7. Empirical evidence
7.1. I n t r o d u c t i o n
The analytical results in West (1986, 1990b) and Section 4 and the simulations in
Section 6 suggest at least two different ways o f rationalizing the procyclicality o f
inventory movements and the persistence o f the inventory-sales relationship. One is
a demand-driven model with rapidly increasing marginal production costs (marginal
production costs a0 and/or al are large relative to marginal inventory holding costs a2),
together with a strong accelerator motive (a2a3 large relative to a0 and a 0 . The second
is a cost-driven model, with increasing marginal production costs; such a model may
or may not have a role for the accelerator. For simplicity we somewhat loosely refer to
these as our d e m a n d - d r i v e n and our c o s t - d r i v e n explanations. We do so with some
reservations: please recall that our demand shock Udt may in some data basically
reflect supply side forces.
These two do not exhaust the possibilities, and many economists (including us)
would expect both cost and demand shocks to be important over samples o f reasonable
length. Our own work, for example, has emphasized the possibility o f declining
marginal production costs [Ramey (1991)]. In combination with highly persistent
cost shocks, both procyclicality o f inventories and persistence o f the inventory-sales
relationship m a y result. A n d West (1990b) finds both stylized facts explicable with a
model with strong costs o f adjusting production and a substantial role for both cost
and demand shocks, but with no accelerator.
903
But our demand-driven and cost-driven explanations have the virtue of simplicity,
and both have support from a number of papers: as summarized in this section and the
next, most aggregate studies, and the limited microeconomic evidence available, do not
point to declining marginal cost, and do find a role for the accelerator. In citing such
support we do not cast a wide net but instead selectively cite representative papers.
In addition, after some introductory remarks on papers using the flexible accelerator
model (Section 7.2), we focus on papers that explicitly use the linear quadratic model,
for ease of exposition.
Section 7.2 reviews parameter estimates from the linear quadratic literature,
Section 7.3 discusses sources of shocks, and Section 7.4 provides an interpretation.
We remind the reader that the behavior of inventories depends only on the relative
values of g, a0, al and a2. All statements referencing "large" values of one of these
parameters should be understood to mean "large relative to another parameter or
linear combination of parameters". The normalization involved will be clear from the
context.
904
10-20% o f the gap between actual and target inventories is closed in a quarter. Ever
since Carlson and Wehrs (1974) and Feldstein and Auerbach (1976), many observers
have found such estimated speeds puzzling and perhaps not well-rationalized by the
flexible accelerator model. One reason is that even the largest quarterly movements
in inventories amount to only a few days production. This suggests to Feldstein and
Auerbaeh (1976, p. 376) and others that costs o f adjusting inventories [v, in the
notation o f Equation (5.1)] cannot be very large 21 .
To interpret this second result with the linear quadratic model, recall that we set
a0-= 0 when we established a mapping from the flexible accelerator to the linear
quadratic model (3.1). With a0-= 0, an arbitrarily slow speed o f adjustment results
when al is arbitrarily large. It is not clear to us how large a value o f al is implausibly
large. But we take from the flexible accelerator literature the message that many find
this simplest version o f the model unappealing [see Blinder and Maccini (1991) for a
recent statement].
Accordingly we consider the other sources o f persistence isolated above: costs of
adjustment (discussed in this subsection), and serial correlated cost variables (discussed
in the next subsection). To focus the discussion o f costs of adjustment, we highlight
estimates from some recent linear quadratic studies using two-digit manufacturing data
from the USA. Different studies present estimates o f a0, al and a2 relative to different
parameters or linear combinations o f parameters. To display results from various
studies in consistent form, we restate published estimates of ao, a~ and a2 relative
to a c o m m o n linear combination o f the published estimates of those parameters. This
linear combination is
c ~ (1 + 4 b + b 2 ) a o + ( 1 + b ) a l +ba2,
(7.:)
with b ~ 0.99. Here, "c" is the second derivative o f the objective function (3.1) with
respect to Ht; the Legendre-Clebsch condition states that e > 0 is a necessary condition
for an optimal solution. [See Stengel (1986, p. 213) or Kollintzas (1989, p. 11).]
Note that the estimates we discuss will therefore not be comparable to those used
in the simulations in the previous section and in Table 8, and often are not as easily
interpreted as those expressed relative to a single parameter. We nonetheless use this
normalization since studies sometimes report negative estimates o f a0, al or a2, which
can make interpretation o f estimates relative to one o f those parameters problematic.
Most authors examine more than one specification. Table 10 presents results for a
specification that seemed to be preferred by the author(s). For the preferred specification, columns 2-6 present the median point estimate o f ao/c, a l/c, [(1 + b)ao + a l ]/c
21 The logic apparently is that it should be easy to make inventory movements rapid if firms are
beginning from a starting point in which current movements are small relative to production. But small
inventory movements seem to be exactly what one would associate with slow adjustment speeds, if costs
of adjustment determine both the size of movements and the adjustment speeds; if, instead, the slow
adjustment speeds were accompanied by large movements in inventories, there would be a puzzling
contrast between regression results and basic data characteristics.
Ch. 13:
905
Inventories
Table 10
Median point estimates of model parameters a-d
(1) Reference e
(2)
ao/C
(3)
al/c
(4)
[(1 +b)a 0 +al]/
c
a2/c
(5)
(6)
a3
(7)
Number
of
industries
0.
0.43
0.43
0.15
0.55
0.
0.21
0.21
0.58
1.15
-0.16
0.83
0.64
-0.09
1.14
0.15
-0.63
-0.43
1.69
0.40
0.13
0.12
0.38
0.00
0.67
0.05
0.34
0.44
0.01
1.12
10
a In the column definitions, e -- (1 + 4b + b2) a 0 + (1 + b) a 1 + ba2, b =- 0.995. Note that the magnitudes
in columns 2-4 are therefore not comparable to those in columns 4-6 in Table 8.
b Different papers expressed point estimates relative to different linear combinations of parameters. For
each paper, the reported point estimates were restated relative to c. The Legendre-Clebsch condition
states that c > 0 is a necessary condition for an interior solution of the optimization problem. The table
reports the median of the restated estimates. When a0 = 0 (lines 1 and 2), or when there is only one
industry (line 5), the column 4 entry for marginal production cost is by construction equal to: (1 +b)
times column 2, plus column 3.
c All the studies used two-digit manufacturing data from the USA. The exact data, sample period,
specification and estimation technique vary from paper to paper.
d Most papers present more than one set &results. We chose the specification that seemed to be favored
by the author(s).
e Sources by reference: (1) Table 7 (p. 85), entries labelled "Table 3"; (2) Table 2 (p. 861); (3) Tables 1 6
(pp. 77-80), columns labelled "random walk"; (4) Table 1 (p. 323), excluding autos; (5) Table 4 (p. 128),
entry labelled "FIML-endogenous sales"; (6) Table 4 (p. 391).
( = m a r g i n a l production cost, taking into account costs o f adjusting production), a 2 / c
a n d a3. The m e d i a n is c o m p u t e d across the datasets considered b y the author; the
n u m b e r o f datasets is given i n c o l u m n 7.
A skim o f the table suggests a broad consensus on a3 ( c o l u m n 6). A s well, there
is relatively little d i s a g r e e m e n t on the sign o f the slope o f marginal production costs
( c o l u m n 4); with the exception o f R a m e y (1991), the studies find a n upward slope to
m a r g i n a l production cost. There is, however, some variation in the extent to which the
cost o f adjustment a0 contributes to this upward slope. Consistent with the d e m a n d driven explanation, F u h r e r et al. ( t 9 9 5 ) (line 5) and to a lesser extent West (1986)
(line 6) find that a0 contributes to the upward slope.
Some studies with other datasets have found an even stronger role for the cost o f
adjustment a0, with a0 positive and significant but with estimates o f the production
cost al negative [consistent with R a m e y (1991)], or e c o n o m i c a l l y or statistically
906
indistinguishable from zero. For example, Kashyap and Wilcox's (1993) study of the
automobile industry in the 1920s and 1930s yielded median estimates of parameters
as follows:
ao/c
al/c
[(1 + b)ao + al]/c
a2/c
a3
(7.2)
0.20
-0.11
0.29
0.03
0.72"
Similar results are reported for the modern automobile industry by Blanchard (1983)
and Ramey (1991), and for US aggregate inventories by West (1990b).
On the other hand, we see in lines 1 and 2 that the preferred specifications in the
Eichenbaum (1989) and Durlauf and Maccini (1995) set the cost of adjustment to
zero. In these two papers, the estimates of al tended to be positive but perhaps not
so large as to imply a speed of adjustment that Feldstein and Auerbach (1976) would
find implausibly slow. In part these papers set a0 to zero - because in a setup similar
to that of Kollintzas (1995) in line 3, negative and insignificant point estimates of a0
tended to result.
Rounding out the cost-driven story requires finding substantial persistence from
stochastic variation in costs. This is discussed in the next subsection.
7.3. Shocks
There is much circumstantial evidence that serially correlated cost shifters have
important effects on inventory behavior. In particular, the data often seem happy with
specifications in which the unobservable disturbance uct is highly autocorrelated [e.g.,
Eichenbaum (1989), West (1990b), Ramey (1991)].
One's confidence that this unobservable disturbance really reflects stochastic
variation in production costs would be increased if inventories could be shown to
respond aggressively to observable measures of costs. Unfortunately, this appears not
to be so. In practice, factor prices and interest rates usually are insignificant (in both
economic and statistical terms), and sometimes have effects opposite of the theoretical
predictions. For statistical significance, Table 11 shows a selection of results using cost
variables, from studies of two-digit manufacturing in the USA, and now including
flexible accelerator as well as linear quadratic studies.
It may be seen in columns 1-4 that a finding of a statistically significant effect of
observable measures of costs is rare: only 2 entries are "y"s, indicating that in only
two of the 21 studies did significance at the 5% level characterize at least three-fourths
of the coefficients estimated in a given study. 11 entries are "n"s, indicating that in
these 11 studies fewer than one-fourth of the coefficients were significant. On the other
hand, in column 5 it may be seen that for the unobservable disturbance, three of the
6 entries are "y"s, and that two of these "y"s are for studies that also included some
observable measures of costs (lines 6 and 8); none of the 6 entries are "n"s.
7.4. Interpretation
We showed in Section 4 that the demand-driven and cost-driven explanations put
two large autoregressive roots in the inventory sales relationship H t - O S t ; in fact,
907
Table 11
Statistical significance of cost variables ~c
Reference d
Wage
Materials
prices
Energy
prices
Interest
rate
Unobservable
shock
n
y
?
ao/c
al/c
0.03
0.42
a2/c
a3
0.01
2.51'
(7.3)
908
median estimate of a0 was higher. In fact, the estimate o f a0 was higher in 5 o f the 6
datasets 22.
An interpretation is that a model that fits his data would imply two autoregressive
roots. When serial correlation in the cost shock was suppressed, the positive values
for a0 rationalized a second autoregressive root; when serial correlation in the cost
shock was imposed, large (or even positive) values o f a0 would imply an autoregressive
structure too elaborate for the data, and accordingly the regression yielded diminished
values o f a0.
Discriminating between the two explanations thus means distinguishing between
costs o f adjustment [when a0 0 and the serial correlation of the cost shock is zero
(~bc= 0)] and exogenous serial correlation (when a0 = 0 and q~c is near one). In principle
this may be done, using either cross-equation restrictions, or additional variables such
those in the Wt vector. But in both inventory and non-inventory contexts this has
proved difficult [e.g., Blinder (1986b), McManus et al. (1994), Surekha and Ghali
(1997)].
And in any case, our discussion so far perhaps has understated the extent o f conflict
across empirical results. There is a range of estimates of most parameters (including
some wrong-signed or otherwise implausible ones), and we have pushed papers into
one o f just two camps in the interest o f summarizing a complex set o f results: while
in principle it may be possible to pin down important macroeconomic parameters and
sources o f shocks by simply estimating linear inventory models with aggregate data,
this tantalizing idea has not proved true in practice so far.
The conflict across papers, or the range o f estimates, may be no worse than in
empirical work in other areas. For example, those familiar with the real business
cycle literature will probably not be surprised that it is difficult to find observable
counterparts to unobservable cost shocks. And Lovell (1994) shows that the estimated
speed o f adjustment o f Ht towards H~ is in fact no slower than those o f some other
variables. As well, part o f the conflict across papers no doubt results from econometric
problems related to sample size or estimation technique [West and Wilcox (1994,
1996), Fuhrer et al. (1995)]. Finally, it may be that careful analysis would reveal that
seemingly disparate conclusions in fact result mainly from the use o f different sample
periods, datasets, and observable cost shiflers ("Wt", in the notation o f the previous
sections).
But pointing out (perhaps unfairly!) that other literatures have similar problems will
not advance our knowledge about inventories. Nor, most likely, will sharp estimates
be produced by even the most refined econometric technique, at least when applied to
familiar data. We therefore suggest some alternative approaches.
22 Such statements potentially are sensitive to how the parameters are expressed (relative to "c", as in
Table 10, or some other linear combination of parameters). But in this case the statement applies not
only with respect to the normalization we have used, but also with respect to the normalization used by
Kollintzas, which was relative to a0(1 + b)+ a l.
909
8.1. Introduction
In this section we offer what we believe to be fruitful directions for future research.
Sections 8.2 and 8.3 describe alternative modelling strategies. Some of our suggestions
are based on alternatives to the basic linear quadratic production smoothing model;
others extend the basic model. All build on the insights delivered by the basic model:
that procyclical movements result when inventories facilitate sales (a force captured
in the basic model with the accelerator term), and that the shape of production
costs influences both the character of cyclical movements and the persistence of
the inventory-sales relationship. In addition, all seem intuitively capable of helping
explain either or both of our stylized facts (procyclicality of inventories, persistence
o f the inventory-sales relationship), although, of course, research to date involving
these suggestions has its share of blemishes (e.g., wrong-signed parameter estimates).
Finally, Section 8.4 describes how the use of different data may help understand
inventory behavior.
8.2. Inventories in production and revenue functions
The potential importance of the accelerator term (a3) in explaining both the business
cycle and long-run behavior of inventories suggests that the relationship between
inventories and sales deserves more study. Consider first Holt et al.'s (1960) original
motivation for this formulation as az(Ht-a3St+l) 2. As discussed on pp. 56-57 of
their book, their initial model of optimal inventory holdings used lot-size formulas,
where the optimal batch size, the number of batches and optimal inventory levels all
increase with the square root o f the sales rate. They used two approximations to capture
the costs and benefits associated with inventory holdings. First, they approximated
the square-root relationship with a linear relationship between inventories and sales
(e.g., H 7 =a3St+l). Second, they approximated all costs and benefits associated with
inventories with a quadratic in which costs rise with the square of the deviation of
inventories from the optimal level. This generates the accelerator term a 2 ( H t - H t ) 2.
While this tractable formulation provides a plausible mechanism for procyclical inventory movements, there are two potential problems with it. First, the approximations
may be inadequate. As we will discuss below, the approximations used imply that the
cost of a marginal reduction in inventories is linear in the stock of inventories, whereas
at least one paper found significant convexities. Second, inventories may directly affect
revenue in a way that is not well captured by including the accelerator term in the cost
function.
One alternative strand of the literature has modelled inventories as factors of
production, or considered interrelationships between inventories and other factors of
production. Christiano (1988), Ramey (1989), Galeotti et al. (1997) and Humphreys
et al. (1997) are examples. Ramey (1989) argues that since inventories at all stages
910
We next consider arguments and evidence that a key shortcoming of the linear
quadratic model is that it fails to account for fixed costs facing firms. Blinder (1981),
Caplin (1985), Mosser (1988), Blinder and Maccini (1991), and Fisher and Hornstein
(1996) all argue that fixed costs of ordering may be very important for understanding
the behavior of retail and wholesale inventories as well as manufacturers' materials and
supplies. In some environments the aggregation argument presented in Section 3.4 will
not apply, and research to date has shown that under certain conditions such fixed costs
may lead to (S, s) type of decision rules. In their review article, Blinder and Maccini
(1991) recommend that future inventory research concentrate on the (S, s) model. This
will require resolution of difficult problems of aggregation, perhaps partly through
the use of simulations [Lovell (1996)]. While the results look suggestive at the level
of a single-product firm, the implications for a multi-product firm, let alone for an
911
Finally, we discuss how the addition of more data may help narrow the estimates
obtained from the linear quadratic model, as well as shed light on the unobserved
cost shocks. We will argue that there are several available sources of data that have
the potential to clear up ambiguities.
One possible explanation for the range of estimates obtained from the production
smoothing model is the data are not sufficient for distinguishing the relative values
of the parameters. One way to glean more information from macroeconomic data
is to use information contained in prices, something done in a handful of papers
912
including Eichenbanm (1984), Blanchard and Melino (1986) and Bils and Kahn
(1996). Pindyck's (1994) results using futures prices provides additional evidence of
the information contained in prices.
A second possible use of new data is to measure the stochastic variation in cost. As
Table 11 indicates, a number of authors have experimented with several observable cost
shifters, but generally do not find effects. Another possible source of cost shocks that
has been studied in a few papers is credit conditions. We remarked in Section 2.2 that
Kashyap et al. (1994) and Carpenter et al. (1994, 1998) still find persistence in the
inventory-sales relationship after including measures of credit conditions. But they
also regularly find that credit conditions affect inventory holding behavior of small
firms, across various specifications. If these credit conditions are serially correlated
(which they are likely to be), and if small firms are important enough to substantially
affect industry- and economy-wide aggregates, credit conditions may ultimately help
explain our two stylized facts.
Finally, we advocate more plant and firm-level studies, although gathering such data
requires substantial work. Schuh (1996), for example, uses panel data from the M3LRD
to calibrate biases from aggregation. And consider Holt et al.'s (1960) study o f six
firms ranging from a paint producer to an ice cream maker and Kashyap and Wilcox
(1993) and Bresnahan and Ramey's (1994) studies of the automobile industry. They use
not only firm-level data on production, inventories and sales, but also company reports
and industry press, which provide valuable insights into the cost structure facing firms.
For example, Bresnahan and Ramey (1994) were able to categorize the cause of every
plant shutdown using information from Automobile News, which chronicled drops in
demand and cost shocks such as strikes and model year change-overs.
9. Conclusions
913
assumes that there are strong costs of adjusting production and a strong accelerator
motive.
Our review o f the empirical evidence, however, indicates that the range of estimates
of key parameters and of the relative importance of cost versus demand shocks is
too wide to allow us to endorse one of the two or some third explanation. But while
the literature has not reached a consensus, it has identified mechanisms and forces that
can explain basic characteristics of inventory behavior. We believe that several research
strategies, and use of different data, promise to continue to improve our understanding
of inventory movements and therefore of business cycle fluctuations.
914
Table 3.2.9. Agriculture (line 2) was subtracted from total (line 1), and the result
was deflated by the GDP deflator. Japan: Economic Planning Agency, Annual Report
on National Accounts, 1997, table on "Closing Stocks". The nominal figure for
total stocks was deflated by the GDP deflator. United Kingdom: Office for National
Statistics, United Kingdom National Accounts: The Blue Book, 1996, Table 15.1.
Agriculture (series DHIE) and government (AAAD) were subtracted from total
(DHHY), and the result was deflated by the GDP deflator.
Data sources for sectoral distribution of US inventories: broad sectoral categories
were obtained from Citibase, and manufacturing inventories by stage of processing
were obtained from the BEA. The stage of processing inventories were converted from
monthly to quarterly data by sampling the last month of the quarter.
We assume throughout that al, a2, g > 0 and a0, a3/> 0. See Ramey (1991) for solutions
when al < 0.
We begin by working through in detail the solution discussed in Section 4, when
a0 = 0 and the forcing variables follow first-order autoregressions. For simplicity, for the
most part we set ~ ---- 0 as well. Thus Uot =Uct [see Equation (3.2)], Et-luct =(&u~t-l
and Et 1Udt = (9,JUdt-l. To insure a unique stable solution, we assume that either (B.la)
or (B.lb) holds:
g > a2a3(1 - a3),
a2a3(1 - a3) > g >
(B.la)
2(1 + b 1)ala2(a3 - 0.5)(a3 - b(1 + b) -1)
a2 +2al(1 + b 1)
(B.lb)
Note that the right-hand inequality in (B. l b) follows if a3 falls outside (b(1 + b) -I , 0.5),
a narrow range when b ~ 1. There will also be a stable solution when a2a3(1 - a 3 ) = g .
But to allow us to divide by g - a2a3(1 - a 3 ) at certain stages in the derivation, we rule
this out for conciseness.
915
Use P t = - g S t + U d t ,
(B.2) becomes
(B.2)
Qt=St+AHt,
(B.3)
al
St = - ~ H t
(B.4)
al + a2a3 .
+ ~ - - H t
l-~Uct+
1
~Udt'
Use (B.4) and (B.4) led one period to substitute out for St and St+l in H t ' s first-order
condition (3.3) (with a0 =- 0). After some rearrangement, the result may be written
0 = bEtHt+i - (1 + b + m ) H t + H t 1 + gtlcuct + grid
Udt
'
(B.5)
grid --
al [g + a2a3(a3 - 1)]
al - bq~d(al + a2a3)
a l [ g + a2a3(a3 - 1)]'
It can be shown that inequality (B. 1) guarantees that there is exactly one root less than
one to the polynomial
bx 2 - t/x + 1 = 0.
(B.6)
if
~/>0,
0 . 5 b - 1 [ ~ l + ( t 1 2 - 4 b ) 1/2]
if
~/<0.
=0.5b-l[r]-(r]2-4b)
~z4
(B.7)
Using techniques from Hansen and Sargent (1980) it follows that the solution to
problem (B.5) is
H t = ~ H H t - 1 + f HcUct + f Hd Udt,
f Hc =-- [YgH/(1 --b~14Oc)]gHc,
f Hd =~ [~H/(1 --bYgHOd)]gHd.
(B.8)
Upon substituting Equation (B.8) into Equation (B.4) and rearranging, we obtain
St = ~ s H t - I + f scUct + f sdUdt,
fSc =
1 + alfHc
2
'
al + a2a 3 + g
~S =~
al(1 - ~ H ) + a2a3
al + a2a 2 + g
(B.9)
1 -- a l f H d
f S d =-
al + a2a~ + g
916
Let L be the lag operator, "adj" the adjoint of a matrix. From Equations (B.7) and (B.8),
a representation for the bivariate (Ht, H t - O S t f =-- Y t process is
Yt = AYt-I +BUt,
Jr.
A ----
C14 - 0Cs
fH
BI ~
0 '
Yt = (I-AL)-IBUt
file--
c0
f Sc
fHd
f Hd -- O f sd
adj ( I - A L ) ~ .
~
(B.lO)
t~t-/t
1I - A L l Yt : a d j ( I - A L ) B U t .
This m a y be used to solve for the univariate process for H t - OSt, which is
(1 - CHL)(Ht - OSt) = ( f Hc - O f so) uct + O ( : V H f Sc -- Y f S f Hc)Uct-1
03.11)
+ ( f H d -- O f s d ) Udt + O(gHfSd -- ~ S f H o ) U d t - 1 .
:=k
(1 -
(B.12)
Thus, when Od = 1, H - OSt N ARMA(2, 1) with autoregressive roots eVIl and 0c.
Now suppose that a 0, so that Uct = "aIWt + Uct, with E t a Wt = q)wWt 1. Algebra
similar to that used above may be used to conclude that the first-order condition (B.5)
and the decision rules (B.8) and (B.9) become
/
+ f HcUct + f HdUdt,
(B.13)
( g + a 2 a 2 ) I -- b [ g + a2a3(a3 - 1)]q~/w ~,
al [ g + aza3(a3 - 1)]
f S w --
"a + a l f i-lw
2
al + a2a 3 + g
Ch. 13:
Inventories
917
edt
x14L)(Ht-OSt
--
a ' W t ) - --
(B.15)
Now allow ao ~ 0, as well as arbitrary autoregressive processes for Wt, Uct, and Udt.
When a0 ~ 0, the first-order condition for St (B.2) becomes
-gSt
+ Udt -
Et[(aoAQt - baoAQt+l ) +
alQt
- a2a3(Ht-l
- a 3 S t ) + U c t ] = O,
(B.16)
where Pt = - g S t + Udt has been used to substitute out for Pt. The solution is most
concisely derived if one uses St = Q t - A t t t to remove the St from the first-order
condition (equivalently, if one makes Qt and H t rather than St and H t the choice
variables). After so doing, Equations (B.16) and (3.3) may be written
Et[bAtlXt+l + A o X t +A1Xt-1 +BoUt +BlUr+l] = O, X t = (Ht, Q t f .
(B.17)
The (21) vector Ut is (Udt, U c t f . The matrices A1, Ao and Bo and B1 are (2 2),
with
Ao =A 1 =--
-[g-a2a3(1-a3)]
[ g - a2a3(1 -
~'
0
Vt~CVt.
--a0
Bo =
B1 =
--
'
I:l
0
"
(B.18)
Suppose that Vt ~ (uet, Udt, W~) f N AR(p) (possibly with unit autoregressive roots),
Et 1Vt =q)lVt l + ' " . + ~ p V t ~ . (There may be many zeros in the q)i if there is lots
more dynamics in say Wt than in either uct or Ud). Note for the future that
Ut = 10 01
a3)]
- ( g + a2a23)
g + a~ + a2 a2 + (1 + b)ao '
(B.19)
918
(B.20)
(B.21a)
(B.Zlb)
Equation (B.21a) requires bA'l R2 +AoR+A1 =0. Given model parameters, and thus
knowledge o f A1 and A0, this equation may be used to solve for a stable matrix R.
(The matrix equation will have multiple solutions, just as does the scalar quadratic
Equation (B.5). Restrictions similar to those in (B. 1) will insure that there is a unique
stable solution.) It may help to note that the solution matches the earlier one if we
reimpose the assumption that there are no costs o f adjusting production. With a0 = 0,
the second column o f A1 is zero, from which it follows that the second column o f R is
zero. Further, R(1, 1) ~ rl 1 = JrH, r21 = JrH + ;rs - 1 for JrH and ;rs defined in Equations
(B.7) and (B.9).
Whether or not a0 = 0, once R is recovered from Equation (B.21 a), Equation (B.2 lb)
may then be used to solve for the Gi. For example, suppose that Wt and the shocks
are first-order autoregressive, so that p = l : q~l is block diagonal with ~bc in its
(1, 1) element, Cd in its (2, 2) element and q~w in the block in its lower right hand
corner. Then Equation (B.21b) implies
bA'I (RGo + Goq~l) + AoGo + BoC + B1C ~ I = 0,
(B.22)
which can be used to solve linearly for Go in terms o f the model parameters and q~l.
Return to the solution (B.20). Let R = [r/j]. Transform from a solution in (Ht, Qt)'
to one in (Ht, St)' =- Zt, using Qt = St + A H t . The result is
Zt = 111Zt-I + H2Zt-2 + Fo Vt + " + Fp 1Vt p+l,
//1 =
rll +r12
1 +r22+r21 - ( r l l +r12)
Fi = MGi,
M =
lO
1 1
r12
r22-r12
112 =
'
-r12
-(r22-r12 )
0
0 '
"
(B.23)
Thus St 2 does not appear in the reduced form, and the first column o f / / 2 is the
negative o f the second column o f H~. To repeat an earlier point: if ao = 0, then
rl2=r22 =0, rl~ = ~ H , and 1 + r 2 ~ - r l l = ~ s .
Finally, suppose we use Equation (B.23) to derive the autoregressive process for
(Ht, H t - OSt)', and then solve for the univariate process for H t - OSt, using the method
Ch. 13.
919
Inventories
that led to Equation (B.11). When a0 s 0 , the relevant determinant [the analogue
of ] I - A L I in Equation (B.10)] is a second-order polynomial in the lag operator.
Specifically, for/'/1 defined in Equation (B.23), let II1 = [~/j]. Then this second order
lag polynomial is
1 - (Jr11 + 3g12)L+ (3"gl13"g22-}-,7112 -- gg213"gi2)L 2.
(B.24)
The roots to this polynomial are called ~l and :r2. Equation (B.22) may be used to
show that the moving average component of H t - OSt is first order when q)c = 0 and
Od = 1.
fo
UdtS, - 0.5gS~.
(B.25)
The area under the supply curve is equal to the cost function presented as part of
Equation (3.1). Thus, the competitive equilibrium solution is equivalent to the solution
920
to the following social planner problem that maximizes the difference between these
two functions:
MaxH,s V = E, ~_~ b j IUdt+jSt+j - 0.5gSet+j j-O
O.5aoAQ2/ - O.5al Q~+j - O.5a2 (Ht +j-i - a3St +j)2
- U c t +jQt +j]
03.26)
subject to the inventory identity that Qt +j = St +j + A H t +j. The first-order c o n d i t i o n for
inventories is identical to the one o b t a i n e d from the firm-level problem. The first-order
c o n d i t i o n for sales is identical to the one obtained w h e n the industry d e m a n d curve is
substituted into the firm-level first-order c o n d i t i o n for sales.
References
Abramowitz, M. (1950), Inventories and Business Cycles (National Bureau of Economic Research, New
York).
Allen, D.S. (1995), "Changes in inventory management and the business cycle", Federal Reserve Bank
of St. Louis Review 77:17-26.
Beaulieu, J.J., and J.A. Miron (1992), "A cross country comparison of seasonal cycles and business
cycles", The Economic Journal 102:772-788.
Belsley, D. (1969), Industry Production Behavior: The Order-Stock Distinction (North-Holland,
Amsterdam).
Bils, M., and J.A. Kalm (1996), "What inventory behavior tells us about business cycles", manuscript
(University of Rochester).
Blanchard, O.J. (1983), "The production and inventory behavior of the American automobile industry",
Journal of Political Economy 91:365-400.
Blanchard, O.J., and A. Melino (1986), "The cyclical behavior of prices and quantifies: the case of the
automobile market", Journal of Monetary Economics 17:379-408.
Blinder, A.S. (1981), "Retail inventory behavior and business fluctuations", Brookings Papers on
Economic Activity 1981(2):443-505.
Blinder, A.S. (1986a), "Can the production smoothing model of inventory behavior be saved?", Quarterly
Journal of Economics 101:431-453.
Blinder, A.S. (1986b), "More on the speed of adjustment in inventory models", Journal of Money, Credit
and Banking 18:355-365.
Blinder, A.S, and L.J. Maccini (1991), "Taking stock: a critical assessment of recent research on
inventories", Journal of Economic Perspectives 5:73-96.
Bresnahan, T.E, and V.A. Ramey (1994), "Output fluctuations at the plant level", Quarterly Journal of
Economics 109:593-624.
Caplin, A.S. (1985), "The variability of aggregate demand with (S, s) inventory policies", Eeonometrica
53:1395-1409.
Carlson, J.A., and W. Wehrs (1974), "Aggregate inventory behavior", in: G. Horwich and E Samuelson,
eds., Trade, Stability and Growth: Essays in Honor of Lloyd A. Metzler (Academic Press, New York).
Carpenter, R.E., and D. Levy (1998), "Seasonal cycles, business cycles and the comovement of inventory",
Journal of Money, Credit and Banking 30:331 346.
921
Carpenter, R.E., S.M. Fazzari and B.C. Petersen (1994), "Inventory investment, internal-finance
fluctuations, and the business cycle", Brookings Papers on Economic Activity 1994(2):75-138.
Carpenter, R.E., S.M. Fazzari and B.C. Petersen (1998), "Financing constraints and inventory investment:
a comparative study using high frequency data", Review of Economics and Statistics 80:513-519.
Cecchetti, S.G., A.K. Kashyap and D.W Wilcox (1997), "Interactions between the seasonal and business
cycles in production and inventories", American Economic Review 87:884-892.
Childs, G.D. (1967), Inventories and Unfilled Orders (North-Holland, Amsterdam).
Christiano, L.J. (1988), "Why does inventory investment fluctuate so much?", Journal of Monetary
Economics 21:247-280.
Christiano, L.J., and M.S. Eichenbaum (1989), "Temporal aggregation and the stock adjustment model
of inventories", in: T. Kollintzas, ed., The Rational Expectations Inventory Model (Springer-Verlag,
New York) 70-109.
Durlauf, S.N., and L.J. Maccini (1995), "Measuring noise in inventory models", Journal of Monetary
Economics 36:65-90.
Eichenbaum, M. (1984), "Rational expectations and the smoothing properties of finished goods", Journal
of Monetary Economics 14:71-96.
Eichenbaum, M. (1989), "Some empirical evidence on the production level and production cost smoothing
models of inventory investment", American Economic Review 79:853-864.
Fair, R.C. (1989), "The production smoothing model is alive and well", Journal of Monetary Economics
24:353070.
Feldstein, M.S., and A.J. Auerbach (1976), "Inventory behavior in durable goods manufacturing: the
target adjustment model", Brookings Papers on Economic Activity 1976(2):351-396.
Fisher, J.D.M., and A. Hornstein (1996), "(S, s) inventory policies in general equilibrium", working paper
WP-96-24 (Federal Reserve Bank of Chicago).
Fuhrer, J.C., G.R. Moore and S. Schuh (1995), "Estimating the linear-quadratic inventory model:
maximum likelihood versus generalized method of moments", Journal of Monetary Economics 35:
115-157.
Fukuda, S.-i., and H. Ternyama (1988), "Some international evidence on inventory fluctuations",
Economics Letters 28:225-230.
Galeotti, M., L. Guiso, B. Sack and E Schiantarelli (1997), "hwentories, production smoothing and the
shape of the cost function", manuscript (Universit~t degli Studi di Bergamo).
Granger, C.WJ., and T.H. Lee (1989), "Investigation of production, sales and inventory relationships
using multicointegration and non-symmetric error correction models", Journal of Applied Econometrics
4:S145-S159.
Hall, G. (1996), "Non-convex costs and capital utilization: a study of production and inventories at
automobile assembly plants", manuscript (Federal Reserve Bank of Chicago).
Haltiwanger, J.C., and L.J. Maccini (1989), "Inventories, orders, temporary and permanent layoffs: an
econometric analysis", Carnegie-Rochester Series on Public Policy 30:301-366.
Hansen, L.P, and T.J. Sargent (1980), "Formulating and estimating dynamic linear rational expectations
models", Journal of Economic Dynamics and Control 2:7-46.
Hansen, L.E, and K.J. Singleton (1982), "Generalized instrumental variables estimation of nonlinear
rational expectations models", Econometrica 50:1269-1286.
Hester, D.A. (1994), "Changing relations between inventories and bank loans", in: R. Fiorito, ed.,
Inventory Cycles and Monetary Policy (Springer-Verlag, Berlin) 125-147.
Holt, C.C., E Modigliani, J.E Muth and H.A. Simon (1960), Planning Production, Inventories and Work
Force, (Prentice Hall, Englewood Cliffs, NJ).
Humphreys, B.R., L.J. Maccim and S. Schuh (1997), "Input and output inventories", manuscript (Federal
Reserve Bank of Boston).
Jorda, O. (1997), "Random-time aggregation in partial adjustment models", manuscript (University of
California at Davis).
922
Kahn, J.A. (1987), "Inventories and the volatility of production", American Economic Review 77:
667-680.
Kahn, J.A. (1992), "Why is production more volatile than sales? Theory and evidence on the stockoutavoidance motive for inventory holding", Quarterly Journal of Economics CVII:481-510.
Kashyap, A.K., and D.W Wilcox (1993), "Production and inventory control at the General Motors
Corporation during the 1920s and 1930s", American Economic Review 83:383-401.
Kashyap, A.K., O.A. Lamont and J.C. Stein (1994), "Credit conditions and the cyclical behavior of
inventories", Quarterly Journal of Economics CIX:565-592.
Kollintzas, T. (1989), "The linear rational expectations equilibrium inventory model: an introduction",
in: T. Kollintzas, ed., The Rational Expectations Equilibrium Inventory Model (Springer-Verlag, New
York) 1-32.
Kollintzas, T. (1995), "A generalized variance bounds test with an application to the Holt et al. inventory
model", Journal of Economic Dynamics and Control 19:59-89.
Krane, S.D. (1994), "The distinction between inventory holding and stockout costs: implications for
target inventories, asymmetric adjustment, and the effect of aggregation on production smoothing",
International Economic Review 35:117-136.
Krane, S.D., and S.N. Braun (1991), "Production smoothing evidence from physical product data",
Journal of Political Economy 99:558-581.
Lai, K.S. (1991), "Aggregation and testing of the production smoothing hypothesis", International
Economic Review 32:391-403.
Lovell, M.C. (1961), "Manufacturers' inventories, sales expectations, and the acceleration principle",
Econometrica 29:293-314.
Lovell, M.C. (1993), "Simulating the inventory cycle", Journal of Economic Behavior and Organization
21:147-179.
Lovell, M.C. (1994), "Researching inventories: why haven't we learned more?", International Journal of
Production Economics 35:33-41.
Lovell, M.C. (1996), "Macroeconomic implications of S, s versus accelerator finished goods inventory
strategies", International Journal of Production Economics 45:55-64.
Maccini, L.J., and R.J. Rossana (1981), "Investment in finished goods inventories: an analysis of
adjustment speeds", American Economic Review 71:17-22.
Maccini, L.J., and RJ. Rossana (1984), "Joint production, quasi-fixed factors of production, and
investment in finished goods inventories", Journal of Money, Credit and Banking 16:218-236.
McManus, D.A., J.C. Nankervis and N.E. Savin (1994), "Multiple optima and asymptotic approximations
in the partial adjustment model", Journal of Econometrics 62:91-128.
Metzler, L.A. (1941), "The nature and stability of inventory cycles", The Review of Economics and
Statistics XXIII:113-129.
Miron, J.A., and S.P. Zeldes (1988), "Seasonality, cost shocks and the production smoothing model of
inventories", Econometrica 56:877-908.
Mosser, P.C. (1988), "Empirical tests of the (S, s) model for merchant wholesalers", in: A. Chikan and
M. Lovell, eds., The Economics of Inventory Management (Elsevier, Amsterdam) 261-284.
Pindyck, R.S. (1994), "Inventories and the short-rtm dynamics of commodity prices", The RAND Journal
of Economics 25:141-159.
Ramey, V.A. (1989), "Inventories as factors of production and economic fluctuations", American
Economic Review 79:338-354.
Ramey, V.A. (1991), "Nonconvex costs and the behavior of inventories", Journal of Political Economy
99:306-334.
Rossana, R.J. (1990), "Interrelated demands for buffer stocks and productive inputs: estimation for two
digit manufacturing industries", Review of Economics and Statistics 72:19-29.
Rossana, R.J. (1993), "The long-run implications of the production smoothing model of inventories: an
empirical test", Journal of Applied Econometrics 8:295-306.
Ch. 13:
Itwentories
923
Rossana, R.J. (1995), "Technology shocks and cointegration in quadratic models of the firm", International
Economic Review 36:5-17.
Rossana, R.J. (1998), "Structural instability and the production smoothing model of inventories", Journal
of Business and Economic Statistics 16:206-215.
Rotemberg, J.J., and G. Saloner (1989), "The cyclical behavior of strategic inventories", Quarterly Journal
of Economics 104:73-98.
Sargent, T.J. (1979), Macroeconomic Theory (Academic Press, New York).
Schuh, S. (1996), "Evidence on the link between firm-level and aggregate inventory behavior", Finance
and Economics Discussion Paper No. 1996-46 (Federal Reserve Board of Governors).
Stengel, R.E (1986), Stochastic Optimal Control: Theory and Applications (Wiley, New York).
Stock, J.H., and M.W. Watson (1993), "A simple estimator of cointegrating vectors in higher order
integrated systems", Econometrica 61:783-820.
Surekha, K., and M. Ghali (1997), "The speed of adjustment and production smoothing", manuscript
(Western Washington University).
West, K.D. (1983), "Inventory models and backlog costs", Ph.D. Thesis (Massachusetts Institute of
Technology).
West, K.D. (1986), "A variance bounds test of the linear quadratic inventory model", Journal of Political
Economy 94:374401.
West, K.D. (1988), "Order backlogs and production smoothing", in: A. Chikan and M. Lovell, eds., The
Economics of Inventory Management (Elsevier, Amsterdam)305-318.
West, K.D. (1990a), "Evidence from seven countries on whether inventories smooth aggregate out-put",
Engineering Costs and Production Economics 19:85-90.
West, K.D. (1990b), "The sources of fluctuations in aggregate inventories and GNP", Quarterly Journal
of Economics CV:939 972.
West, K.D. (1992a), "A comparison of the behavior of US and Japanese inventories", International
Journal of Production Economics 26:115-222.
West, K.D. (1992b), "Sources of cycles in Japan, 1975 1987", Journal of the Japanese and International
Economies 6:71-98.
West, K.D. (1995), "Inventory models", in: M. Pesaran and M. Wickens, eds., Handbook of Applied
Econometrics, vol. I, Macroeconometrics (Basil Blackwell, Oxford) 188-220.
West, K.D., and D.W. Wilcox (1994), "Some evidence on the finite sample behavior of an instrumental
variables estimator of the linear quadratic inventory model", in: R. Fiorito, ed., Inventory Cycles and
Monetary Policy, (Springer-Verlag, Berlin) 253-82.
West, K.D., and D.W Wilcox (1996), "A comparison of alternative instrumental variables estimators of
a dynamic linear model", Journal of Business and Economic Statistics 14:281-293.
Wilkinson, M. (1989), "Aggregate inventory behavior in large European economies", European Economic
Review 33:181-194.
Chapter 14
Contents
Abstract
Keywords
1. Introduction
2. Stylized facts o f a g g r e g a t e activity
2.1. Measuring business cycles with the HP filter
2.2. Some stylized facts of US business cycles
2.3. Some stylized facts of economic growth
2.4. Implications of stylized facts
3. The basic neoclassical m o d e l
3.1. The structure
3.2. Steady-state growth and transforming the economy
3.3. Optimal capital accumulation
3.4. The nature of the steady state
3.5. Transitional dynamics
3.6. The (Un)importance of capital formation
3.7. Constructing dynamic stochastic models
4. The Real Business C y c l e s h o c k
4.1. The driving process
4.2. Calibrating and solving the model
4.3. Business cycle moments
4.3.1. Simulations of US business cycles
4.4. The importance of capital accumulation
4.5. Early successes and criticisms
928
928
929
931
932
934
941
941
942
942
944
946
947
948
950
951
952
952
953
956
958
960
960
* We benefited from the comments of the editors as well as from those of Rui Albuquerque, Robert
Barro, Marianne Baxter, Satyajit Chatterjee, Aubhik Khan, Daniele Coen Pirani, Henry Siu, Julia
Thomas, and Michelle Zaharchuk. Dorsey Farr provided detailed comments and excellent research
assistance. Support from the National Science Foundation is gratefully acknowledged.
Handbook of Macroeconomics, Volume 1, Edited by J.B. Taylor and M. Woodford
1999 Elsevier Science B.V. All rights reserved
927
928
5. The central role o f productivity shocks
5.1. Productivity shocks must be large and persistent
5.2. The influence of productivity persistence
5.3. Why not other shocks?
6. Extensions o f the basic neoclassical model
6.1. The supply of labor
6.1.1. Estimated and assumed labor supply elasticities
6.1.2. Implications of varying the aggregate labor supply elasticity
6.1.3. Modeling the extensive margin
6.2. Capacity utilization
7. Remeasuring productivity shocks
8. Business cycles in a high substitution economy
8.1. Specification and calibration
8.2. Simulating the high substitution economy
8.3. How does the high substitution economy work?
8.4. What are the properties of the shocks?
8.5. How sensitive are the results?
9. Conclusions
Appendix A. Dynamic theory
A.1. Assumptions on preferences and technology
A.2. The dynamic social planning problem
A.3. A dynamic competitive equilibrium interpretation
A.4. The welfare theorems
References
963
963
969
973
974
974
975
975
976
980
982
984
984
986
986
988
990
993
995
995
997
999
1001
1002
Abstract
The Real Business Cycle (RBC) research program has grown spectacularly over the last
decade, as its concepts and methods have diffused into mainstream macroeconomics.
Yet, there is increasing skepticism that technology shocks are a major source o f
business fluctuations. This chapter exposits the basic RBC model and shows that it
requires large technology shocks to produce realistic business cycles. While Solow
residuals are sufficiently volatile, these imply frequent technological regress. Productivity studies permitting unobserved factor variation find much smaller technology
shocks, suggesting the imminent demise o f real business cycles. However, we show that
greater factor variation also dramatically amplifies shocks: a RBC model with varying
capital utilization yields realistic business cycles from small, nounegative changes in
technology.
Keywords
macroeconomics, business cycles
JEL class~cation: El0, E32
Ch. 14:
929
1. Introduction
Business cycle research studies the causes and consequences of the recurrent
expansions and contractions in aggregate economic activity that occur in most
industrialized countries. Over the last century, exploration of real business cycles the idea that economic fluctuations are caused primarily by real factors - has itself
undergone periods of intense activity and relative dormancy. In the 1920s, real theories
played a leading role: economists sought to use new microeconomic tools to learn
about the aggregate consequences of shifts in demand and supply of goods and
productive factors. However, the Great Depression of the 1930s had a dramatic effect
on business cycle research. Economists began to believe that microeconomic theory
was an inadequate basis for understanding business cycles. Real factors came to be
less stressed, with greater weight given to monetary conditions and the psychology of
households and firms. Government management of the economy came to be seen as
not only desirable but essential.
The rise of Keynesian macroeconomics to a position of orthodoxy in aggregate
economics meant that it took half a century for a revival of interest in equilibrium
business cycle models. The breakdown in the performance of macroeconometric
models in the 1970s and the associated rational expectations revolution pioneered
by Lucas (1976) set the stage for a vigorous recovery, since the logic of rational
expectations ultimately required general equilibrium analysis 1. Kydland and Prescott
(1982) and Long and Plosser (1983) first strikingly illustrated the promise of this
approach, suggesting that one could build a successful business cycle model that
involved market clearing, no monetary factors and no rationale for macroeconomic
management. It is now perhaps hard to recall that this idea was met with surprise and
disbelief.
By the end of the 1980s there was a central and controversial finding of real business
cycle (RBC) research, as this line of work came to be called. Simple equilibrium
models, when driven by shifts in total factor productivity measured using Solow's
(1957) growth accounting approach, could generate time series with the same complex
patterns of persistence, comovement and volatility as those o f actual economies.
Writing a survey of RBC research at that time, it was difficult to find sufficient material
so we settled for expositing the basic model and forecasting future developments 2. A
decade later, our task in this chapter is substantially different: it is time to take stock
of a decade of research, to assess criticisms, and to evaluate the health of the research
program.
The first observation is that it has been a decade of spectacular growth: so many
theoretical and empirical articles use the RBC approach that a full bibliography would
Sargent (1982).
2 King,Plosserand Rebelo (1988a,b) surveyedthis area when a single conferenceprogram could include
most participants in the RBC research program.
930
likely exhaust the generous page constraint on our contribution to this volume. 3 Real
business cycle analysis now occupies a major position in the core curriculum of nearly
every graduate program. 4 At a recent NBER conference, a prominent Cambridge
economist of the New Keynesian school described the RBC approach as the new
orthodoxy of macroeconomics, without raising a challenge from the audience.
Continuing on the positive side of the ledger, the methods of the RBC research
program are now commonly applied, being used in work in monetary economics,
international economics, public finance, labor economics, asset pricing and so on. In
contrast to early RBC studies, many of these model economies involve substantial
market failure, so that government intervention is desirable. In others the business
cycle is driven by shocks to the monetary sector or by exogenous shifts in beliefs. The
dynamic stochastic general equilibrium model is firmly established as the laboratory
in which modern macroeconomic analysis is conducted.
At the same time, there has been increasing concern about the mechanism at the
core of standard RBC models: the idea that business cycles are driven mainly by large
and cyclically volatile shocks to productivity, which in turn are well represented by
Solow residuals as in the provocative study of Prescott (1986). A key difficulty is
that typical estimates of Solow residuals imply a probability of technical regress on
the order of 40%, which seems implausible to most economists. Recent studies have
corrected the Solow residual for mismeasurement of inputs - notably, unobserved effort
and capacity utilization - and inappropriate assumptions about market structure. These
remeasurements have produced technology shocks with more plausible properties:
notably, productivity growth is much less likely to be negative. In effect, these
studies have caused productivity shocks to grow smaller and less cyclically volatile
by introducing elements which respond sympathetically to economic activity [see, for
example, Burnside, Eichenbaum and Rebelo (1996)].
Since the standard RBC model requires large and volatile productivity shocks, this
remeasurement research is typically interpreted as indicating that our chapter should
be a first draft of the obituary of the RBC research program. In fact, most of our survey
does read like a chronicle of the life and death of RBC models. We begin in Section 2
by discussing the measurement of the business cycle as well as reviewing facts about
growth and business cycles that have motivated the construction of aggregate models.
We next turn in Section 3 to the basic neoclassical model of capital accumulation, as
initially developed by Solow (1956) and others for the purpose of studying economic
growth but now used more widely in the study of aggregate economic activity. We
then celebrate the early victories of the RBC program in Section 4 and discuss early
criticisms.
3 One valuable monitor of this ever-expanding literature is provided by Christian Zimmermann's web
page (http : //ideas. uqam. ca/QMRBC/index, html).
4 One manifestation of the breadth of this intellectual impact is that Hall (1999) cites Berkeley's
David Romer (1996) and Harvard's John Campbell (1994) for authoritative presentations of the basic
RBC model.
931
There has been a substantial amomlt of research on real business cycles, but we
organize our discussion around three main points in the next three sections. The central
role of large and persistent productivity shocks in the basic model is discussed in
Section 5. Important improvements to the basic RBC framework are highlighted in
Section 6. The remeasurement of productivity shocks, which has fostered concern
about the health of real business cycles, is reviewed in Section 7.
In Section 8 we argue that the incipient demise o f the real business cycle is
hardly as likely as suggested by conventional wisdom. In fact, rather than weaken
the case for a real theory of the cycle, our view is that the recent remeasurement
of productivity actually strengthens it. To make this point, Section 8 describes a
very simple variant of the basic RBC model, but one that is different on two key
dimensions. The economy has indivisible labor, which is one of the key improvements
reviewed in Section 6, and has costly variation in capital utilization, which is
one of the structural features that makes the standard Solow residual depart from
productivity. There is a substantial remeasurement of productivity shocks mandated
by this economy: when we do the necessary correction, the standard deviation of
productivity growth drops to less than one-fifth of the standard deviation of the growth
rate of the Solow residual. Productivity regress occurs in less than 1% of the postwar quarterly observations, even though the measured Solow residual shrinks 37% of
the time. Yet these small shocks can generate empirically reasonable business cycles
because our model features substantial amplification o f productivity shocks: readily
variable capital utilization and a highly elastic labor supply lead small changes in
productivity to have major effects on macroeconomic activity. When we drive our
model with such small measured productivity shocks, there is a remarkable coincidence
between actual US business cycles and simulated time paths of output, consumption,
investment, and labor input. The same structural features that lead the Solow residual
to dramatically overstate productivity fluctuations also lead the economy to greatly
amplify productivity shocks.
s However,there is some recent interest in these methods. Watson (1994) uses the Burns and Mitchell
methodology to contrast inter-war and post-war US business cycles. Simkins (1994) and King and
Plosser (1994) show that RBC models produce artificial data that the Burns and Mitchell methods
would recognize as having similar characteristics to US data.
932
found the Burns-Mitchell facts intact, lurking underneath almost half a century of
accumulated dust. As stressed by Lucas (1977), the finding that "business cycles are
all alike" suggested that the nature o f macroeconomic fluctuations does not hinge on
institutional factors or country-specific idiosyncrasies, so that one can hope to construct
a unified theory of the business cycle.
2.1. M e a s u r i n g business cycles with the H P f i l t e r
Most real quantities, such as US real national output in the top panel o f Figure 1,
grow through time. Hence, the statistical measurement o f business cycles necessarily
involves some way o f making the series stationary, which is most commonly done by
the removal o f a secular trend. In their study o f quarterly post-war US data, Hodrick
and Prescott (1980) detrended their variables using a procedure now widely known as
the HP filter. In essence, this method involves defining cyclical output y~ as current
output Yt less a measure o f trend output yg, with trend output being a weighted average
o f past, current and future observations:
J
Y; = Yt - Yg = Yt - Z ajyt4.
j=-j
(2.1)
Figure 1 displays how cyclical output is constructed. In the first panel, the logarithm
o f current output is the more variable series and trend output is the smoother series.
The HP cyclical component o f output is the dotted line in the second panel, defined
from the elements o f the first panel as y~ = Yt - Y t g except that we have multiplied by
100 so that cyclical output is a percentage 6. Aggregate output displays business cycles
in that there are alternating periods o f high and low output, but these episodes are of
unequal duration and amplitude.
To see how the cyclical output measure produced by the HP filter compares with
those from other detrending methods, we can look at the second and third panels of
{y, },=0
For quarterly data, the standard value chosen for the smoothing parameter it is 1600. When it = oo
the solution to this problem is a linear trend, while with it = 0 the trend coincides with the original
series. In a finite sample context, the weights aj in Equation (2.1) depend on the length of the sample,
so that the text expression is a simplification. King and Rebelo (1993) discuss additional properties of
this detrending procedure, including derivation of filter weights and frequency response fimctions for
the case in which the sample is infinitely large. They establish that the HP filter has strong detrending
properties, in the sense that it can make stationary series up through four orders of integration.
933
Output
and
HP Trend
8.6-
8.4-
i
o,
............... Output
o
HP Trend
7.8-
7.6-
7.4-
f
7.2
47
I
52
I
57
',
:
62
:
67
',
:
72
l
77
~
82
I
87
I
92
Date
HP and
6.
Band
Pass
(6,32)
Cyclical
Components
42-
~. -2.
-4.
i.J
~'
,.~
~'~. . . . . . .
HP
-6-8 ,
47
',
',
',
',
52
',
',
',
57
',
',
',
62
',
',
',
',
67
',
1
72
t
77
',
',
82
',
',
',
87
',
',
',
92
',
',
',
',
',
',
Date
HP
15
/
-10 ~t
-15 I
47
Growth
Component
and
Linear
Trend
'~
~/
I
I
52
I
57
Residual
i
62
I
67
',
72
',
',
',
',
77
',
',
',
',
82
',
',
87
',
',
',
',
92
Date
Fig. 1. Trend and business cycle in US real output. Sample period is 1947:1-1996:4.
934
Figure 1. First, if we simply subtract a linear trend from the log o f output, the resulting
business cycle component would be the solid line in the third panel o f Figure 1. This
alternative measure o f cyclical output is much more persistent than the HP measure:
for example, output is high relative to its linear trend during most o f the 1960s and
1970s. The dashed line in the third panel is the gap between the HP trend and a linear
trend, thus indicating that the HP filter extracts much more low frequency information
than a simple linear trend.
It is also useful to compare the HP cyclical component with the business cycle
measures resulting from the band-pass (BP) filter procedure developed by Baxter and
King (1995), since such measures are presented by Stock and Watson (1999) in their
extensive compilation o f business cycle facts elsewhere in this volume 7. In the second
panel o f Figure 1, the HP measure o f cyclical output is accompanied by a BP measure
o f cyclical output: this procedure makes the cyclical component mainly those parts
o f output with periodicities between 6 and 32 quarters. For series like output, which
contain relatively little high frequency variation, Figure 1 shows that there is a minor
difference between these alternative cyclical measures.
There has been some controversy about the suitability of the HP filter for business
cycle research. Prescott (1986) notes that the HP filter resembles an approximate highpass filter designed to eliminate stochastic components with periodicities greater than
thirty-two quarters. Adopting that perspective, we are simply defining the business
cycle in a fairly conventional way: it is those fluctuations in economic time series that
have periodicity o f eight years or less s. At the same time, the third panel o f Figure 1
reminds us that there are slow-moving stochastic components o f economic time
series omitted by this definition, which may have substantial positive and normative
significance.
7 Since an exact bandpass filter contains an infinite number of moving average terms, a practical
bandpass filter cannot be produced exactly but involves approximations. Baxter and King (1995) derive
the relevant formulas, imposing the constraint that the sum of the filterweights must be zero; they also
compare the BP filter to several other detrending methods including the HP filter.
8 Eightyears corresponds to the longest reference cycle that Burns and Mitchell (1946) tmcoveredusing
very different methods. Stock and Watson (1999) adopt an alternative view of the interesting business
cycle periodicities (six to twenty four quarters), but this is a difference of degree rather than kind.
9 This data set covers the period 1947 (first quarter) to 1996 (fourth quarter).
935
Nondurables
2|
i, i '
.,I
.
sD
~
!X
'
~i~
-8,
:
47
Consumption
and Output
i~"
~,
~':"
in!
ii
"d(
..
ki
xj ....
- - -
~6
: :
52
: :
57
: :
62
: :
67
: :
72
: :
77
: :
82
Output
Consumption (ND)
: :
87
: :
92
Date
30
Durables
Consumption
and Output
20
Output
....
15.L
I~
o[~t
/ .p,(~,
!/\.!,k
p l , l ' , # i { i "~,"J
I J
IJ
!,lA
1/.1 " / i . q "
l'I
-5
i?
....
j~.<,h,,~,~
I"
""i
t,,?i
/
'. . . . . . . . . . . . . .
47
52
57
62
67
/,/i'~ . . . .
7. j~
IJ
7""~
72
c .....
pt~on ( D )
ft,i'l.IJll~NY't\
....
i'J.\
f
I t" e l
77
f:4't,.,.~,~
I'"
82
-;d"
87
"
92
Date
Investment
15-
and Output
105-
if'<./
~0
+, .
i\
zJ
<<1"!,
,,N~
i,,
~ ~
.... .,
~ -5-10
-154
.....
,,
,,
', :
52
', ,, :
57
', ,, ,, ,, ,, :
62
67
', :
72
v .--7W y ont
. . . . . . . . . . . . . . . . . .
77
82
87
92
-20 -
Government
1510-
Spending
and Output
i~
50
k,v:'li i
m-1O
,g
'71
-5-
,l
,jv'i ~
,~/.-..-~
""','<l
" ; ':'
,.~A/.~
-254
-25
Output
-- -- - - Gov't Spending
!~
-15 :
'
-vv ,.;.--,'w-v
. . . . . . .
; ;
52
', ;
57
: :
62
; ;
57
: :
72
: :
77
: :
82
: :
87
: :
92
Date
Fig. 2. Cyclical components of US expenditures. Sample period is 1947:1-1996:4. All variables are
detrended using the Hodrick-Prescott filter.
936
6
Total Hours
and Output
4.
2.
"<
!f
-..
'!d
-; ;--~ 2',,?'.o~
~I
-6-8,
47
: :
52
: :
57
: :
62
', :
67
',
: :
72
: :
77
: :
82
: :
87
: :
92
Date
Capital
:t
i~ ~!
'ii
"
and Output
i:
"
i t
i '
, :.,:
lJ
47
: ',
52
; :
57
: :
82
: :
67
I I
72
!, i
i/
i I
77
/
............. pu,
~I
..........
I I
82
I I
87
I I
92
Date
Capital
10-
Utilization
and Output
5-
J',~
:'~,,~
! ~' t;i
t I
~'4
;I
t'l
"~1
-lO -15,
fl, h
i,!
:%2
it "v'~e
VI
.D
/jl
\,-/
W7
II
I/J
W/
'~ ~,, t
I ,
57
62
67
72
-" "
~V'
~ I
~,I
~l
52
m ....
'~ .t
:,
"~,..d u
'~ u
"
Out ut
- - -- - - Capital UtilizaSon
77
82
87
92
Date
!4
Productivity
1.',J
J. ~-
47
"~/
~/
[J
it
.o
(Solow
,t
"
residual)
'(.,'~Lt
/b/
?!
52
57
and Output
~,
62
67
72
I*:
'~/
.............o=,
~"
77
82
- - - - - - Productivity
87
92
Date
Fig. 3. Cyclical component of US factors of production. Sample period is 1947:1-1996:4. All variables
are detrended using the Hodrick-Prescott filter.
937
it
o ~
~t
~.i<
~,
._,~,,i.~
~"
iv;
.,
t I
~
'~
7
i;
tY
i,
,ok.
; .,,,
~ ,,
"
"v"
,.A!,,,x.<,~
~ . ~ ^ ]
;~"
i/
~,
I >
t;
,1
\ f
........................
Output
- - - - - - Hoursper Worker
Date
,~',
:-
-4
V~
-6
47
i. 7 '
52
(;
iY
kV'
","
ij
57
62
67
72
Datll
77
o.o.,
Employment
----
82
87
92
,
,,
,:,
P.
h,,>, ,2,~v.,~,"
~l\w.
t /
i l;~ .ACd~,..."~_,
i, ii
- -
h~..- .,
A,e,.g. P,odoot
7
-8
47
52
57
62
67
72
Date
77
82
87
92
i"'"v',d'".'hv '-'v"
i!
t 7
/ i
I}
t/
{/
I'
. 47
. . . . . . 52
. . . . . . 57. . . . . 62
'..~; '4,;',41
"i'/
ti
t~
6]i 7I
~/
/ )
7
V
712 . . . . 77
. . .
v.>.-/'- ~.~'>,",-./"
/
i
" ~
\,J
," .4
......... Output
-- - - - Rea Wages
812111817
I 912111
Date
Fig. 4. Cyclical component of US labor market measures. Sample period is 1947:1-1996:4. All variables
are detrended using the Hodrick-Prescott filter.
938
Table 1
Business cycle statistics for the US Economy
Standard deviation
Relative standard
deviation
First-order
autocorrelation
Contemporaneous
correlation with output
1.81
1.00
0.84
1.00
1.35
0.74
0.80
0.88
5.30
2.93
0.87
0.80
1.79
0.99
0.88
0.88
Y/N
1.02
0.56
0.74
0.55
0.68
0.38
0.66
0.12
0.30
0.16
0.60
-0.35
0.98
0.54
0.74
0.78
a All variables are in logarithms (with the exception of the real interest rate) and have been detrended
with the HP filter. Data sources are described in Stock and Watson (1999), who created the real rate
using VAR inflation expectations. Our notation in this table corresponds to that in the text, so that Y is
per capita output, C is per capita consumption, I is per capita investment, N is per capita hours, w is
the real wage (compensation per hour), r is the real interest rate, and A is total factor productivity.
easily gauge the relative volatility o f the series i n question and its c o m o v e m e n t with
output. S u m m a r y statistics for selected series are provided in Table 1 10
Volatility: Economists have long b e e n interested in understanding the e c o n o m i c
m e c h a n i s m s that underlie the different volatilities o f key m a c r o e c o n o m i c aggregates.
The facts are as follows, working sequentially w i t h i n each figure and using the notation
panel 2-1 to denote panel 1 o f Figure 2 and so forth:
C o n s u m p t i o n o f non-durables is less volatile t h a n output ( p a n e l 2-1);
C o n s u m e r durables purchases are m o r e volatile t h a n output ( p a n e l 2-2);
I n v e s t m e n t is three times more volatile than o u t p u t ( p a n e l 2-3);
G o v e r n m e n t expenditures are less volatile than output ( p a n e l 2-4);
Total hours worked has about the same volatility as output ( p a n e l 3-1);
Capital is m u c h less volatile than output, b u t capital utilization in m a n u f a c t u r i n g is
more volatile than output ( p a n e l s 3-2 and 3-3)11;
E m p l o y m e n t is as volatile as output, while hours per worker are m u c h less volatile
than output (panels 4-1 and 4-2), so that most o f the cyclical variation in total hours
worked stems from changes in employment;
Labor productivity (output per m a n - h o u r ) is less volatile than output ( p a n e l 4-3);
10 For a discussion of open economy stylized facts see Baxter and Stockman (1989), Back-us and Kehoe
(1992) and Baxter (1995).
1~ This measure of capacity utilization, constructed by the Federal Reserve System, is subject to
substantial measurement errol, see Shapiro (1989).
939
The real wage rate is much less volatile than output (panel 4-4).
Comovement: Figures 2 through 4 show that most macroeconomic series are
procyclical, that is, they exhibit a positive contemporaneous correlation with output.
The high degree of comovement between total hours worked and aggregate output,
displayed in panel 3-1, is particularly striking. Three series are essentially acyclical wages, government expenditures, and the capital stock - in the sense that their
correlation with output is close to zero 12.
Persistence: All macroeconomic aggregates display substantial persistence; the firstorder serial correlation for most detrended quarterly variables is on the order of 0.9.
This high serial correlation is the reason why there is some predictability to the
business cycle.
In presenting these business cycle facts, we are focusing on a small number of
empirical features that have been extensively discussed in recent work on real business
cycles. For example, in the interest of brevity, we have not discussed the lead-lag
relations between our variables. In choosing the series to study, we have also left
out nominal variables, whose cyclical behavior is at the heart of m a n y controversies
over the nature o f business cycles 13. However, we do report the cyclical behavior
of a measure of the expected real rate of interest from Stock and Watson (1999) in
Table 1. This real interest rate is constructed by subtracting a forecast o f inflation
from the nominal interest rate on US treasury bills. There is a negative correlation
of the real interest rate with real output contemporaneously and, indeed, this negative
relationship is even stronger between real output and lagged real interest rates. Many
modern macroeconomic models, including real business cycle models, have difficulty
matching this feature of business cycles ~4
12 The observation that the real wage is not tightly related to the business cycle goes back to Dunlop
(1938) and Tarshis (1939) who stressed that this was at odds with Keynesian models. This finding is
somewhat dependent on precisely how the real wage is constructed,depending on whether the numerator
(the wage) includes various compensation items and on the index in the denominator (the price level).
Two particular features of wage measurement that affect its cyclical behavior are as follows. First, firms
pay for overtimehours in an expansion and layoffregular hours in a recession. Second, there is a cyclical
composition bias in the labor force lowerquality workers are hired in expansions - which suggests
that the real wage per efficiencyunit of labor effort is procyclical.
13 See Stock and Watson (1999, Sections 3.4, 3.6, and 4.1) for a discussion of literature and empirical
results.
14 King and Watson (1996) find this negative "leading indicator" relationship between the real interest
rate and real activity, using BP filtered data. They also show that a number of modem macroeconomic
models, including the basic RBC model, are unable to match this fact even when driven by complicated
forcing processes that allow them to match most other features of the business cycle. However, while this
result is provocative, it is important to stress that the behavior of this real interest rate involves assuming
that the inflation forecasting equation is temporallystable and that agents know this forecasting structure
in advance.
940
0.6.
L a b o r ' s Share of O u t p u t
0.575.
s',
/ t
k/
'. / ~j/
,,~
0.65
0.525-
~i
> ?
"W
<'j
i~ ;
=~" j
i l t!j
b!
.... 4 7
"2
'
. . 62
. . .
7~
67
'
"
'
'2
'
"
'
92
'
'
Date
0.25 -
0.2-
0.15 -
0.1-
0.o5 -
oi
'
47
'
52
'
"~
'
'
"2
'
7,
'
"2
'
6',
72
'
Daia
6>75
0.7
\J'lt.
0.65
,,.#
0.6
0.55
o%
.
52
. . .
57
'2
'
'
'
72
'
7,
. . .82 .
"2
87
Date
11oo
j~
1000
~J
<ji
660
\/
000
850
. . . . .
47
52
62
72
Date
77
67
'
941
15 Klein and Kosobud (1961) produced an early test of the stability of the "great ratios" in the USA.
More recently, King, Plosser, Stock and Watson (1991) drew attention to how the constancy of these
great ratios was a cointegration implication about the logarithms of the variables, which they tested for
the USA. Evidence on cointegration for other OECD countries is contained in Neusser (1991).
942
between hours worked and aggregate output has led some economists to believe that
understanding the labor market is key to understanding business fluctuations. Finally,
the relatively small variability of real wages and the lack of a close correspondence of
wages with aggregate output, has led some economists to conclude that the wage rate
is not an important allocative signal in the business cycle. The growth facts suggest the
importance of building models that feature a common trend in most real aggregates.
In the 1950s and 1960s, aggregate economic activity was analyzed with two very
different types of dynamic macroeconomic models. The trend components of aggregate
economic activity were studied with "growth models" that stressed three sources
of dynamics: population growth, productivity growth and capital formation. The
business-cycle components were studied with Keynesian macroeconomic models,
which stressed the interaction of consumption and investment but downplayed the
importance of capital accumulation and productivity growth. While there were attempts
to synthesize these developments towards the end of this period, the study of growth
and business cycles most frequently involved very disparate models. Relative to this
traditional macroeconomic approach, the real business cycle literature took a very
different point of view. Its core is a neoclassical growth model of the form developed
by Solow (1956), Cass (1965) and Koopmans (1965). It then follows Brock and
Mirman (1972) in making this growth model stochastic, by positing that the production
technology is buffeted by random aggregate shocks to productivity.
Our introduction to RBC analysis thus naturally begins by reviewing the "basic
neoclassical model", which has implications for both growth and business cycles. In
this section, we focus on the structure of the model, trace some of its implications for
capital accumulation, and discuss Solow's work on productivity measurement. In the
next section, we discuss its business cycle properties.
Eo ~
btu(Ct, Lt),
b > 0,
(3.1)
t=O
where b denotes the discount factor, Ct represents consumption and Lt leisure. The
symbol E0 denotes the expectation of future values of C and L based on the information
943
available at time zero 16. The infinite horizon assumption, which greatly simplifies the
mathematical analysis o f economic growth and business cycles, is usually justified
by appealing to the presence o f altruistic links across generations [Barro (1974)].
However, it can be viewed as an approximation to an economy with many long-lived
agents 17. In our exposition o f this model, we treat the population as constant for
simplicity, although we discuss the consequences of relaxing this assumption at various
points below.
The momentary utility function u(Ct, Lt) in Equation (3.1) is assumed concave and
obeys regularity conditions discussed in the Appendix. It implies a preference for
smooth profiles o f consumption and leisure. It also implies a willingness to substitute
across time if interest rates and wage rates imply differing costs of consumption and
leisure at different dates. Thus, the neoclassical model imbeds a form of the permanent
income hypothesis of Friedman (1957).
Endowments: The fundamental endowment that individuals have is their time, which
can be split between work (Nt) and leisure activities (Lt). Normalizing the total amount
o f time in each period to one, the time constraint is:
Nt + Lt = 1.
(3.2)
We abstract from other endowments of resources since the production from unimproved
land and from nonrenewable resources is a small fraction o f output in most developed
countries.
Technology: The output o f the economy is assumed to depend on a production
function that combines labor and capital inputs. To capture the upward trend in output
per capita that is shown in Figure 1, the basic neoclassical model incorporates secular
improvement in factor productivity. In particular, output (lit) depends on the amounts o f
capital (Kt) and labor (Nt) according to a constant returns to scale production function
which satisfies regularity conditions discussed in the Appendix.
Yt = AtF(Kt, NtXt),
(3.3)
y > 1.
(3.4)
16 TO simplify, we adopt a dating convention that does not distinguish between "planning time" for the
individual and "calendar time" for the economy. Alternative presentations that emphasize this distinction
would write the objective as E t }-~)~obJu(Ct+j,Lt+j), where t is calendar time andj is planning time.
17 Rios-Rull (1994) finds that an overlapping generations model calibrated to the age structure of the
US population has business cycle properties that are similar to an infinite horizon model.
944
(3.5)
I1, = c , + 1,1
This equation corresponds to the basic national i n c o m e accounting identity for a closed
economy with no government. The stock of capital evolves according to:
Kt+ l = It + (1 - 6)Kt,
(3.6)
where 6 is the rate o f depreciation. This formula coincides with the one used in practice
to estimate the stock o f capital according to the "perpetual inventory method ''18.
The form o f the production function (3.3) is motivated by the growth facts and
was widely employed in growth models after Phelps (1966) showed that steady-state
growth - a situation in which all variables grow at a constant rate - required that the
deterministic component o f technology be expressible in labor augmenting form in
economies with Equations (3.5) and (3.6) 19~ In fact, in the feasible steady states of
this model consumption, investment, output and capital all grow at the same rate - the
rate o f trend technical progress - so that the great ratios are stationary.
Initial conditions: The economy starts out with a capital stock K0 > 0. It also
begins with a level o f the technology trend X0 > 0, which we set equal to unity for
convenience, and an initial productivity shock A0 > 0.
u(C) = ~-L-~[C
1-o
- 1],
(3.7)
where cr > 0. This utility function insures that the marginal rate o f substitution between
consumption at dates t and t + 1 depends only on the growth rate o f consumption.
18 In practice the perpetual inventory method allows the depreciation rate to vary through time according
to empirical measures of economic depreciation schedules. Ambler and Paquet (1994) study a RBC model
with depreciation shocks.
19 Three types of technical progress frequently discussed in the literature can be represented in a general
production function:
Yt = X y F (KtXK, NtX,).
The variable X'tH represents total factor augmenting (Hicks-neutral) technical, progress, YtK capital
augmenting technical progress, and Xt labor augmenting (Harrod-neutral) technical progress,. When
the production fimction is Cobb-Douglas these different forms of technical progress are interchangeable
and, hence, they are all consistent with balanced growth. For all other production functions, the only
form of technical progress consistent with steady-state growth is labor augmenting,
Ch. 14:
945
In the basic neoclassical model of growth and business cycles, which features
endogenous labor supply, a steady state also requires that hours per person be
invariant to the level of productivity. King, Plosser and Rebelo (1988a,b) show that
the momentary utility function must be expressible as
b/(C, L) = ~
{ [ C v ( L ) ] 1-g _ 1},
(3.8)
which also implies exactly offsetting income and substitution effects of wage changes
on labor supply 2. The function v(.) satisfies regularity conditions discussed in the
Appendix.
When these restrictions are imposed, it is possible to transform the economy - so
that steady-state growth is eliminated - by scaling all of the trending variables by the
initial level of X: Using lower case letters to denote these ratios, for example yt = Yt/Xt,
we can then write the optimal growth problem as maximizing the transformed utility
fimction:
C~
(3.9/
~-~/3tu(ct,Lt)
t=0
with/3 = by 1-~ being a modified discount factor satisfying 0 < /3 < 1. Utility is
maximized subject to the transformed constraints:
Nt
= 1 - Lt,
(3.1 O)
Yt
= AtF(kt,Nt),
(3.11)
Yt
= Ct + it,
(3,12)
ykt+ l = it (1 - 6)kt.
(3.13)
20 That is, suppose that Equation (3.8) is maximized subject tothe static budget constraint C <~ w(1 -L).
The equality of the real wage with the marginal rate of substitution between leisure and consumption
implies
CDv(L) _ [w(1-L)]Dv(L)
W
v(L)
v(L)
'
with the latter equality following from eliminating consumption using the budget constraint. Changes in w
have no effect on the optimal level of leisure and labor supply.
21 By leaving out population growth, we have essentially proceeded in this manner. However, since
productivity is labor-augrnenting~ we can reinterpret the stationary transformation as.one that involves
dividing through by both the population and the productivity of labor. Under this interpretation, ~' is the
growth rate of population and productivity.
946
(3.14)
The optimal path of capital accumulation can be obtained by choosing sequences for
consumption {ct}~0, leisure {Lt}~0, labor {Nt}~-0 and the capital stock {kt+ l}~0
to maximize Equation (3.9) subject to conditions (3.10) and (3.14). For this purpose,
we form the "Lagrangian"
O<3
L = Z
fitu(ct'Lt)
t=O
+ ~ffXt[AtF(kt,Nt)
+ (1 - b)kt - ct - 7kt+ 1]
t=O
oG
t=O
: DlU(Ct,Lt) = L ,
Lt
: D2u(ct,Lt) = cot,
Aft
: L A t D 2 F ( k t , N t ) = cot,
(3.15)
(3.16)
(3.17)
(3.18)
where we use the notation Diu(c, L) to denote the partial derivative of the function
u(c,L) with respect to its ith argument. The first pair of these efficiency conditions
dictate that the marginal utility of consumption be set equal to its shadow price
[associated with the constraint (3.14)] and that the marginal utility of leisure be set
equal to its shadow price (associated with the time constraint Nt + Lt = 1). The
second pair of efficiency conditions dictate that the utility value of goods produced
with a marginal unit of work (the marginal value product &tAtD2F(kt, Nt)) equal its
utility denominated cost (cot) and that the present value of the future product of capital
([~t+ 1[At+ 1DlF(kt+ 1,Nt+ 1) + 1 - 6]) equal its current utility cost (T),t). An optimal
consumption, leisure, work and capital plan - sequences {ct}~0, {Lt}~0, {Nt}~0,
and {kt}~ 0 - satisfies these first-order conditions, the original constraints, the initial
condition requirement on kt and the transversality condition, l i m t ~ fit,~tkt +j = O.
Optimal capital accumulation in the basic neoclassical model is a "general
equilibrium" phenomenon in three ways. First, the choices of consumption, labor
and capital accumulation are interdependent at each point in time and across time:
947
a solution for optimal capital accumulation involves specifying sequences for all three
of these variables. Second, the requirement that the optimal decisions must respect the
resource constraints of the economy is signalled by the shadow prices (cot and )~t). An
optimal capital accumulation plan thus also involves specification of sequences of these
prices. Third, if these shadow prices were market prices for individual households,
then they would similarly signal these agents to supply and demand the optimal
quantities. That is, there is an equivalence between the optimal quantities chosen by
the social planner and those in a dynamic competitive general equilibrium in the basic
neoclassical model. We will thus move between optimal and market outcomes in our
discussion of the basic model as it seems useful in the next several sections. We will
return later to discuss how work on real business cycles relates to other developments
in dynamic stochastic general equilibrium modeling.
AD1F(k, N) = (r + 6),
(3.19)
where r = ~ - 1 is the stationary state real interest rate and r + 0 is the stationary
state rental price of capital. Given that the production function is constant returnsto-scale, the marginal product of capital depends on the capital-labor ratio k rather
than on the levels of the factors. Accordingly, Equation (3.19) determines the capitallabor ratio as a function of productivity and the rental rate. Second, given this capitallabor ratio and the level of A, the marginal product of labor is also determined, since
ADzF(k, N) = ADzF(~, 1). Thus, there is a real wage rate w = co/;. that is determined
independently o f the total quantity of labor:
w=~=AD2F(k,1).
Third, there are unique levels of work, consumption and the shadow price of
consumption that satisfy the remaining equations.
We know that the variables in the original, untransformed economy are related to
those of the transformed economy by a simple scaling procedure, Yt = ytXt, etc. Hence,
if the transformed economy is in a stationary state, then the original economy will be
in a steady state with many variables - including, consumption, capital output and real
wages - growing at the same rate. Other variables will be constant, notably work effort
and the real interest rate.
948
Output
Capital
0-
0.
-0.05'-
-0.2.
-0.1 -0.4.
~=~o ~ j
-0.15 ~
-0.6.
-0.0
,o.25 -
~'~
. . . . . . .
-0.3 :
-0.351
-t.2,
3 4
2 0 4
5 6 7 8 9 10 11 12 13 14 15 16 ~7 ,18 19 20
',
0,1.
-0.2 -
O.6
0.
1.2
Consumption
and Labor
1.4
0.8
Quarters
Quarters
Investment
5 0 ? 8 9 1011121314151617181020
.... c,
"" ~
~0.3
~ -0,4.
0,4
4/.5
0.2,
-O.6
2
2
9 10
14 15 ~6 17 18 19 20
5 0
7 8 9 10 11 1213 14 15 16 17 18 19 20
-O.7.
Quarters
16
Real Wage
14
.
12
-0.1
.-0.05
... I
-0.15
~
~,~
~8
" ~
-0.2
O~
9 10 11 12 13 14 15 16 17 18 19 20
Quarters
9 10 11 12 13 14 15 16 17 18 19 20
Ouartem
Fig. 6. Transitional dynamics: Basic RBC model (stars); fixed labor model (circles); dashes in mid
left-hand panel represent labor response.
22 In the transformed economy, this is the movement from an initial level of capital k0 to the stationary
level k.
949
capital. Working to establish this stability theorem, Cass and Koopmans were hampered
by the absence of an explicit solution to the model, which stemmed from the presence
of many interdependent choices for consumption and capital at different dates. Thus,
they were able to establish the stability property, rather than ascertaining how the pace
of the transition process depended on the underlying preferences and technology.
More precise descriptions required the detailed specification of preferences and
technology, with near steady-state linear approximations sometimes being used to
evaluate the nature of the global transitional dynamics. Figure 6 illustrates the nature of
these local transitional dynamics of capital, as well as the related movements in output,
investment and consumption. There are two sets of paths in each of the panels: the
' o ' path describes the fixed-labor model used by Cass and Koopmans and the ' . ' path
describes the variable-labor model used in RBC analysis. In these panels, all variables
are displayed as percentage deviations from their corresponding stationary values. For
both models, the economy is assumed to start off with a capital stock that is one percent
lower than its stationary value, so that both of the paths in the upper left panel have
an entry o f - 1 at the first date.
Looking first at the 'o' paths that describe the transitional dynamics of the CassKoopmans model, we see that capital accumulates through time toward its stationary
level, i.e., the (negative) deviation from the stationary level becomes smaller in
magnitude through time. Since capital is low relative to the steady state, the second
panel shows that output is also low relative to the stationary state. Capital is built up
through time by individuals postponing consumption. In a market economy, a high real
rate of return is the allocative signal that makes the postponement of consumption
occur, with the rate of return given by rt = [ A D 1 F ( k t + I , N ) - 6] in the fixed labor
economy. Using the preference specification (3.7), in fact, one can show that the growth
rate of consumption is given by
l o g \( O ct
r + ' j) -- - ~1 log ( ~ t l )
~ -1~ [ r t - r ] ,
(3.20)
so that a high real interest rate fosters consumption growth along the transition path 23.
The initial level of consumption is set by the wealth o f individuals or, equivalently,
so that the efficient path of consumption will ultimately be at the stationary level.
In general, as the economy saves to accumulate the new higher stationary level of
capital, net investment it - 6kt is positive along the transition path. In this figure
gross investment is also higher than its stationary level during the process of capital
accumulation.
23 In fact, the value of cr used in constructing Figure 6 is one, so that consumption growth and the rate
of return are equal. The time unit of these graphs is a quarter of a year, however, and the interest rate
is expressed at an annual rate, so that the slope of the consumption path is one fourth of the annualized
interest rate shown in the next to last panel. A basis point is one hundredth of a percentage point, so that
an annual interest rate that is 0.10 percentage points above the steady-state level causes consumption to
grow from about -0.600 to about -0.575 in the second panel.
950
When work effort is endogenous, as in the ' , ' path of Figure 6, these transitional
dynamics change significantly, although there is still a period of capital accumulation
toward the stationary level. The extra margin matters for the speed of the transition.
When capital is initially low individuals work harder and produce more output that is
used for capital accumulation. This extra effort occurs despite the fact that the real
wage (wt = AD~F(kt,Nt)) is low relative to its stationary level. Again, the allocative
signal is a high rate of return that makes it desirable to forgo leisure (as well as
consumption) during the transition process. In fact, there is a subtle general equilibrium
channel that enhances the magnitude of the effect on the rate of return. With more
future effort anticipated, the rate of return rt = [ADtF(kt+I, Nt+ 1) - 6] is higher than
in the fixed labor case, because additional effort raises the marginal product of capital.
Further, this higher rate of return stimulates current work effort. On net, variable effort
raises the speed of transition and mitigates the effects of initially low capital on output
and consumption, while enhancing the investment response.
3.6. The (Un)importance o f capital formation
The capital accumulation mechanism at the heart of the basic neoclassical model
is sometimes viewed as relatively unimportant for growth and business cycles. In
the growth area, Solow followed his (fixed saving rate) growth model (1956) with
a celebrated demonstration that productivity was much more important to economic
growth than was capital formation. It is easiest to exposit this result if we assume that
the production function is Cobb-Douglas:
Yt = AtK~t-a(NtXt) a,
(3.21)
with the parameter a being measurable as labor's share of national output (an idea
which we discuss more below) and thus being constrained to be between zero and
one 24. Then, using time series for output, labor and capital, we can compute the Solow
residual as:
logSRt = log Yt - a logNt - (1 - a) log Kt.
(3.22)
Abstracting from measurement error in outputs or inputs, the Solow residual can be
used to uncover the economy's underlying productivity process,
log SRt = log(At) + a log(Xt).
(3.23)
951
(the average of log(Yt/Nt) - log(Yt 1/Nt-l)) was divided between growth in productivity
(the average of log(SRt) - log(SRt_l)) and growth in capital per unit of labor input
(the average of log(KJNt) - log(Kt i/Nt 1)). The result was a surprising one and must
also have been disappointing in view of his just completed work on the dynamics
of capital accumulation [Solow (1956)]: capital accumulation accounted for only one
eighth of the total, with the remainder attributable to growth in productivity. Thus,
the transitional dynamics of capital formation turned out to be not too important for
understanding economic growth.
Moreover, the transitional dynamics of Figure 6 do not display the positive
comovement o f output, consumption, investment and work effort that take place during
business cycles. Labor and investment are higher than in the steady state when capital
is low while consumption and output are below the steady state. Further, consumption
is much more responsive to a low capital stock than either labor or output, which is
inconsistent with the evidence on relative volatilities reviewed earlier.
Sometimes these results are interpreted as indicating that one should construct
macroeconomic models which abstract from capital and growth, since the introduction
of these features complicates the analysis without helping to understand business
cycle dynamics. However, real business cycle analysis suggests that this conclusion is
unwarranted: the process of investment and capital accumulation can be very important
for how the economy responds to shocks.
3.7. Constructing dynamic stochastic models
In this section, we have concentrated on describing the steady state and the transitional
dynamics of the basic neoclassical model, as an example of the type of dynamic general
equilibrium model now used in RBC analysis and other areas of macroeconomics.
There is now a rich tool kit for studying the theoretical properties of stochastic
equilibrium in these models, such as the advances described by Stokey, Lucas and
Prescott (1989). A systematic analysis of the Brock and Mirman (1972) stochastic
growth model, modified to include variable labor supply as above, calls for the
application of these methods. We review these developments in the Appendix, using
the basic RBC framework to highlight two important issues. First, we characterize
the optimal decision rules for consumption, capital, output, investment and labor
using dynamic programming. Second, we demonstrate that the outcomes of the
optimal growth model are the same as the outcomes of a dynamic stochastic general
equilibrium model, in which firms and workers trade goods and factors in competitive
markets. This equivalence requires that firms and workers have rational expectations
about future economic conditions. Another notable result for this stochastic model,
first established by Brock and Mirman (1972), is that the stationary state is replaced
by a stationary distribution, in which the economy fluctuates in response to shocks.
We have also discussed the local transitional dynamics of the basic neoclassical
model illustrated in Figure 6. The development of real business cycle models and
dynamic stochastic general equilibrium theory has also heightened interest in methods
952
for solving and simulating dynamic equilibrium models. In this survey, we rely on
now-standard loglinear approximation methods for solving the various real business
cycle models that we construct 25. These methods have been shown to be highly
accurate for the basic RBC model 26. The application of these methods contains
essentially two steps. First, it is necessary for us to specify the utility function, the
production fixnction, the depreciation rate and so forth so that we can solve for
the steady state of the model economy, working much as we did in Section 3A.
Second, we take loglinear approximations to the resource constraints (3.10)-(3.13)
and the efficiency conditions (3.15)-(3.18). We then assume that these approximate
equations hold in expected value - a certainty equivalence assumption - and solve the
resulting expectational linear difference equation system. This yields a system of linear
difference equations forced by random shocks, f r o m which moments and simulations
can b e easily computed.
(4.1)
and exploiting the fact that log(Xt) = log(Xt_l)+ log(y), it is possible to estimate the
stochastic process for productivity and, in particular, the persistence parameter p and
25 There are versions of the basic RBC model that can be solved analytically but require restrictive
assumptions on preferencesand technology. See, e.g., Radner (1966), Long and Plosser (1983), Devereux,
Gregory and Smith (1992), and Rebelo and Xie (1999).
26 See, e.g., Danthine, Donaldson and Mehra(1989), Christiano (1990), and Dotsey and Mao (1992).
953
the standard deviation of the innovation et. Todo this, one fits a linear trend to logSRt
in order to compute y. Then, one uses the residuals from this regression to estimate p
and standard deviation of~t. For our quarterly data set, the resulting point estimates
are 0.979 for p and .0072 for the Standard deviation o f et. The high estimated value of
p reflects the substantial serial correlation in panel 4 o f Figure 3, where the variable
described as productivity is the Solow residual.
27 This calibration approach is commonlyassociated solelywith the RBC program, but it was also used
in the early 1980s by researchers studying the effects of nominal contracting on economicfluctuation,
such as Blanchard and Taylor. The calibration approach had been previously used in other areas of
economics, such as public finance and international trade, which employedcomplicated, though static,
general equilibrium models. Calibration is now routine in a wide range of macroeconomicareas, although
it was controversial in the late 1980s because of Kydland and Prescott's (1991) insistence that it should
be used instead of standard econometricmethodology.A nontechnical review of the interaction between
the quantitative theory approach and econometricsis provided by King (1995).
954
two-thirds which is a standard value for the long run US labor income share 28. Both
X0 (the initial level o f technical progress) and the m e a n value o f A are parameters
which affect only the scale o f the economy, and hence can be normalized to one.
The growth rate o f technical progress is chosen to coincide with the average growth
rate o f p e r capita output in the U S A during the post war period (1.6% per annum),
which implies a quarterly gross growth rate o f technical progress o f 7 = 1.004 29.
The rate o f depreciation is chosen to be 10% p e r annum (6 = 0.025) 3. Taking all
o f this information together, we can solve for the capital-labor ratio k/N, using the
requirement that r + 6 = A D 1 F ( k , N ) . In the C o b b - D o u g l a s case we obtain:
k
[ r+6
"
In turn, this implies that the steady-state value o f the capital-output ratio is
k/y = ( k / N ) / ( y / N ) since the average product o f labor y / N = A ( k / N ) l-a. We also thus
compute the steady-state ratios i/y = ( y - 1 + 6 ) ( k / y ) and c/y = 1 - (i/y), as well as the
steady-state real wage rate w = aA(k/N) l-a.
P a r a m e t e r i z i n g utility: The constant elasticity class o f utility functions (3.8) is
motivated b y having steady-state growth in productivity lead to steady-state growth
in consumption and a constant average level o f hours per person. In our discussion o f
this basic model, we use the momentary utility function
u ( c t , L t ) = log(ct)+ _1- ~ 0 (L~- r l - 1),
(4.2)
Once we specify the parameter which governs the labor supply elasticity (r/) we
choose 0 to match steady state N , which is about 20% o f the available time in the
U S A in the postwar period 31 . The studies o f Prescott (1986) and Plosser (1989) used
the " l o g - l o g " case, which makes utility u(Ct, Lt) = l o g ( C t ) + 0 log(Ll), motivating this
form b y arguing that a range o f microeconomic and asset-pricing evidence suggests a
coefficient o f risk aversion o f ~r = 132. We begin with this case (in which ~/= 1) and
then consider some alternative values in Section 6 below.
28 This is higher than the value of a = 0.58 used in King, Plosser and Rebelo (1988a,b). This number
is somewhat sensitive to the treatment of the government sector and of proprietor's income. See Cooley
and Prescott (1995) for a discussion.
29 Many calibration studies ignore growth all together, as we ignore growth in population. Incorporating
population growth would raise the appropriate value of 7 to 1.008.
30 Here we use a conventional value for 6, but there is some evidence that it should be lower. The ratio
of capital consumption allowances to the capital stock (excluding consumer durables and government
capital) for the USA in the post war period takes values of the order of 6 percent [see Stokey and
Rebelo (1995, Appendix B)]. The average investment share is very sensitive to y and 6, but the near
steady-state dynamics are not.
c which can be solved for 0.
3~ That is, the condition w = D2u(c,L)
Dlu(c.r) = ~Oc implies that a = - ~ - ~ON;,
32 There is substantial uncertainty about ~r, which tends to be estimated with a very large standard error,
see Koeherlakota (1996).
955
~/
ge
0.984
3.48
1.004
0.667
0.025
0.979
0.0072
Collecting these results, we have Table 2 listing the parameters that are used in the
baseline model in the remainder o f this section.
Loglinearizing the model economy: The next step in solving the model is to
approximate its various equations, which is most frequently done so as to produce
log-linear relations. Sometimes, as with the utility specification (4.2), it happens that
the relations o f interest are exactly log-linear to start. For example, the consumption
efficiency condition (3.15) is At = DlU(Ct,Lt) = (ct) -I and the leisure efficiency
condition (3.16) is cot = ~twt = D2u(ct,Lt) = L t ~, so that:
-ct
= ,
(4.3)
(4.4)
(4.5)
i.e., the real wage is raised by productivity and by increases in the capital labor ratio.
Other equations o f the m o d e l are not exactly log-linear and so must be approximated.
The time constraint is Nt + Lt = 1 so that small changes in labor and leisure satisfy
dNt + dLt = 0 and thus )~"v d NN' + L @ = 1. Since log(-~) ~ ~dNt
- , we conclude33:
(N)Nt + (L)Lt = 0.
(4.6)
33 An alternative derivation of this and other results involves assuming that the behavioral equation
depends on A/t = log(Nt/N) etc. and then taking a linear approximation in the hatted variables. For
example, the time constraint is
N exp()~,)+ L exp(I,t) = 1,
so tha~t is approximately Equation (4.6), given that a first-order Taylor series approximation to exp(~t)
about Nt = 0 is Nt.
956
A 1.(1 a)~'t
7tra so that a
Since the constraint on uses of goods takes the form ct + it = ~t'~t
mixture of the two methods used above yields
(4.7)
Other equations such as (3.6) and (3.18), which contain variables at different dates,
can be similarly approximated. The result is a loglinear dynamic system that can be
solved numerically.
Interpreting aspects o f the model economy: One benefit of this solution strategy is
that the researcher maybe able to interpret certain aspects of the model economy prior
to obtaining its numerical solution. For one example; Equation (4.5) can be interpreted
as a description of "labOr demand", so as to discuss the influence of productivity,
the real wage rate and the stock of capital on the quantity of labor. For another,
combining Equations (4.4) and (4.6), we arrive at a "labor supply schedule" that relates
the quantity of labor to the real wage and the shadow price of goods, Le.,
L
so that a higher value of r/ lowers the labor supply elasticity. Individually, these
equations describing "labor supply" and "labor demand" can be used to evaluate
the consistency of the macroeconomic model with microeconomic evidence. Taken
together, they provide an explanation of how the quantity of labor and the real wage
rate respond to variations in productivity, the capital stock, and the shadow price.
4.3. Business cycle moments
One way of evaluating the predictions of the basic RBC model is to compare moments
that summarize the actual experience of an economy with similar moments from the
model. On the basis of such a moment comparison, Prescott (1986) argued that the
basic RBC model predicts the observed "large fluctuations in output and employment"
and, more specifically, that "standard theory.., correctly predicts the amplitude o f . . .
fluctuations, their serial correlation properties, and the fact that investment is about six
times as volatile as consumption."
Table 3 reports summary statistics on HP cyclical components of key variables
for simulations of the basic neoclassical model driven by productivity shocks. These
statistics are comparable to those reported in Table 1 for the US economy.
Volatility o f output and its components: Productivity shocks produce a model
economy that is nearly as volatile as the actual US economy. More specifically,
comparing the ratio of model and empirical standard deviations, Kydland and Prescott
(1991) have argued that the real business cycle model explains the dominant part
957
Table 3
Business cycle statistics for basic RBC model a,b
Standard deviation
Relative standard
deviation
First-order
autocorrelation
Contemporaneous
correlation with output
Y
C
I
1.39
0.61
4.09
1.00
0:44
2.95
0.72
0.79
0.71
1.00
0.94
0.99
N
Y/N
w
r
A
0.67
0.75
0.75
0,05
0.94
0.48
0.54
0.54
0.04
0.68
0.71
0.76
0.76
0.71
0.72
0.97
0.98
0.98
0.95
1.00
All variables have been logged (with the exception of the real interest rate) and detrended with the
HP tilter.
b The moments in this table are population moments computed from the solution of the model. Prescott
(1986) produced multiple simulations, each with the same number of observations available in the data,
and reported the average HP-filtered moments across these simulations.
a
o f business cycles 34. For the numbers in Tables 1 and 3, the Kydland-Prescott
variance ratio is 0.77 = (1.39/1.81) 2, suggesting that the RBC model explains 77%
o f business fluctuations. Using a variation on the basic m o d e l which introduces costs
o f moving labor out o f the business sector, Kydland and Prescott (1991) argued that
"technology shocks account for 70 percent o f business cycle fluctuations". Using a
slightly different version o f the model Prescott (1986) h a d previously attributed 75%
o f output fluctuations to productivity shocks.
The real business cycle m o d e l is consistent with the observed large variability o f
investment relative to output, as indicated by the relative standard deviations reported
in the second columns o f Tables 1 and 3. In particular, investment is about three times
more volatile than output in both the actual economy (where the ratio o f standard
deviations is 5.30/1.81=2.93) and the model economy (where the ratio o f standard
deviations is 2.95). Consumption is substantially smoother than output in both the
model and actual economies. In our basic model, however, consumption is only about
one-third as volatile as output while it is over two thirds as volatile as output in the
US economy. We return to discussion o f this feature o f the economy in Section 6
below.
Persistence and comovement with output. Business cycles are persistently high or
low levels o f economic activity: one measure o f this persistence is the first-order serial
correlation coefficient. Table 3 shows that the persistence generated by the basic model
34 See Eichenbaum (1991) for a criticism of this interpretation of the variance ratio.
958
is high, but weaker than in the data (see Table 1). The relative standard deviations
also provide a measure of the limited extent to which the basic RBC model amplifies
productivity shocks: in terms of its business cycle behavior, output is 1.48 times as
volatile as productivity.
Business cycles also involve substantial comovement of aggregate output with inputs
(such as labor) and the components of output (such as consumption and investment).
Accordingly, Table 3 reports the contemporaneous correlation of output with the other
four measures. All of these correlations are quite high, indicating the basic RBC
model captures the general pattern of comovement in the data. However, the empirical
correlations of output with labor, investment and productivity are substantially smaller
than their model counterparts.
From this battery of statistics, we can see that the RBC model produces a
surprisingly good account of US economic activity. However, there are also evident
discrepancies. Notably, consumption and labor input in the basic model are each much
less volatile than in the data. Further, the basic RBC model produces a strongly
procyclical real wage and real interest rate, which does not accord well with the
US experience summarized in Table 1.
4.3.1. Simulations o f US business cycles
Figure 7 depicts US data together time series generated by simulating the model with
the innovations to the actual US Solow residual. On the basis of results similar to
those in this figure, Plosser (1989) argued that "the simple (RBC) model appears to
replicate a significant portion of the behavior of the economy during recessions and
during other periods" 35. Indeed, looking at the first panel of Figure 7, it is clear that
the basic RBC model gives quite a good account of the quarter-to-quarter variation
in the output time series. The correlation between these series is 0.79; the model also
works well in all major recession and expansion episodes.
Turning to the individual components of output, the performance of the RBC model
is also surprisingly good for such a simple model. Consumption in the model and
the data are strongly positively associated (the contemporaneous correlation is 0.76),
although the model's series in the bottom panel of Figure 7 is much less volatile than
the actual experience, as suggested by the previous discussion of moments above.
Investment in the model and the data also move together in the third panel of
Figure 7, although model investment appears to lead actual investment by one to two
quarters. One measure of this lead is that the contemporaneous correlation of model
and actual investment is 0.63 and the correlation between actual investment and past
model investment is 0.73 at one lag and 0.69 at two lags. While the volatility of labor is
35 This section uses a model that is essentially the same as that in Plosser (1989), but our simulated
time series are slightly different due to (i) differences in data; and (ii) differences in filtering. Plosser's
use of the first-differencefilter emphasizes higher frequency components of time series relative to our
use of the HP filter.
959
6-
Output
g--2-
51
56
61
66
71
76
81
86
91
96
Date
Labor
2.
~-2
Input
l' l V 1 -
g.
i tf
"I/
V"
-6.
, ; ,1 , , , o , ,
-8.
, 1, , , , o ,6, , ,
"
"
___~2~o,
96
Date
Investment
2,;t
-15 ~-
.....
51
56
61
66
71
76
81
86
91
96
Date
Consumption
654,
Data
3i
2-
7,-
~- o
-1-2-
-3 i
I
I I
51
I I
56
I I
61
] I
66
I I
71
I I
76
I :
81
',
',
: :
86
',
', ',
91
',
:
96
Date
Fig. 7. Basic model: simulated business cycles. Sample period is 1947:2-1996:4. All variables are
detrended using the Hodrick-Prescott filter.
960
broadly similar in the data and in the model, there is much less of a period-to-period
correspondence between labor and output in the simulations in the second panel of
Figure 7 than there is in the US economy.
4.4. The importance of capital accumulation
The process of capital accumulation is central to business cycles in the RBC
framework. To highlight this importance, consider the effect of a positive productivity
shock under the assumption that investment is zero at all dates. Then, higher
productivity would raise the level of the production function and the marginal product
of labor schedule, with each increasing proportionately. From the standpoint o f the
representative individual, this would work just like a secular rise in the real wage
rate, with exactly offsetting wealth and substitution effects on labor. Thus, hours per
worker would be invariant to productivity, with consumption moving one-for-one with
output 36. By contrast, in the RBC framework, investment increases in response to a
positive productivity shock, i.e., the representative household optimally saves some
fraction of the higher current output. Thus, it is also efficient to lower consumption
and raise work effort relative to the fixed investment case, producing the cyclical
comovements that we see in the actual economy.
However, the introduction of capital as a factor of production in the basic model
tends to make it difficult to match the behavior of output and labor input that we saw
in the first panel of Figure 3, where labor is nearly as volatile as output. Fundamentally,
this reflects the fact that capital is not particularly variable over the business cycle (see
the second^panelof Figure 3). More specifically, the Solow decomposition indicates
that 33t At + aNt + (1 a)tct, so that we expect a one percent change in labor to
produce an a = ~ per cent change in output. It also works to mitigate the volatility
of labor input, since it makes the marginal product of labor decline with the quantity
of labor, given that the capital stock is essentially fixed over the business cycle. That
is, despite substantial business cycle changes in investment, these do not have a large
effect on the capital stock.
=
36 With a total factor productivity shock, this exact offset requires that there be a Cobb-Douglas
production function, although it holds for any production function if the shock is labor augmenting.
While it clearly holds if there is no capital, it also holds if capital is present, as may readily be verified
using the line of argument in footnote 20. A key part of this more general result is that capital income
increases with the productivity shock.
961
clearly was important additional work to done. Prescott (1986) summarizes moment
implications as indicating that "the match between theory and observation is excellent,
but far from perfect". Plosser ('1989) summarizes the model simulations as indicating
that "the whole idea that such a simple model with no government, no market failures
of any kind, rational expectations, no adjustment costs could replicate actual experience
this well is very surprising". Rogoff (1986) warns of the potential power of the RBC
model: " T h e . . . real business cycle results.., are certainly productive. It has been said
that a brilliant theory is one which at first seems ridiculous and later seems obvious.
There are many that feel that (RBC) research has passed the first test. But they should
recognize the definite possibility that it may someday pass the second test as well."
One notable part of the RBC program is its insistence on the construction of
dynamic stochastic general equilibrium models, which is now the accepted approach
to macroeconomic analysis across a wide range of research areas and perspectives.
Even those who are skeptical o f the central role of productivity shocks have accepted
the idea that "the basic methodological approach ... (is) ... relevant to models in
which monetary disturbances play a greater role" as Rogoff (1986) forecasted that
they would.
But it is the other component of the RBC approach that was immediately
controversial and remains so to this day: that technology shocks are the dominant
source of fluctuations. The striking performance of the basic RBC model drew a
strong critical reaction from macroeconomists working in the Keynesian tradition
[Summers (1986), Mankiw (1989)]. Their criticisms focused on three main points.
First, they questioned some of the parameter values used in the calibration of
the model. In particular, they stressed that the model's performance required an
empirically unreasonable degree of intertemporal substitution in labor supply. Second,
they emphasized the model's counterfactual implications for some relative and absolute
prices. The critics observed that the strongly procyclical character of the model's real
wage rate was inconsistent with the findings of numerous empirical studies. They
also pointed to Mehra and Prescott's (1985) earlier finding that standard preferences,
such as Equation (4.2), are incompatible with the equity premium, i.e. the difference
between the average rate of return to equities and the risk free rate. 37 In addition,
t h e y suggested that a productivity shock theory of the cycle should imply a strongly
countercyclical price level. Third, they argued that the use of the Solow residual was
highly problematic, leading to excessively volatile productivity shocks.
In retrospect, the first two criticisms of RBC analysis had a small impact on the RBC
program. Rather than being fragile, the model's performance is surprisingly resilient
to variations in its parameters. Much of the model's performance is anchored on three
single ingredients: a highly persistent technology shock that is sufficiently volatile,
a sufficiently elastic labor supply, and empirically reasonable steady-state shares of
962
consumption and investment in output 38. The RBC model does not need to rely on a
high degree of intertemporal substitution in labor choice. In fact some RBC models
[e.g. Greenwood, Hercowitz and Huffman (1988)] assume that this elasticity is zero.
However, either intertemporal or intratemporal substitution must be strong enough to
produce realistic labor movements, a point to which we will return in Section 6 below.
RBC researchers have produced a battery of models that lead to a relatively high
elasticity of labor supply. The model's predictions for the real wage can be improved
if we step away from the assumption of spot labor markets and incorporate contracts
between firms and workers that allow for wage smoothing [Gomme and Greenwood
(1995), Boldrin and Horvath (1995)] or other forms of labor contracts [Danthine and
Donaldson (1995)]. It is hardly surprising that the assumption of spot labor markets
produces unreasonable implications for the real wage. And while research on the equity
premium puzzle continues, we now have models that are consistent with some aspects
of the equity premium while maintaining the business cycle performance of the basic
model 39. Lastly, the studies of Kydland and Prescott (1990) and Cooley and Ohanian
(1991) have concluded that the price level is indeed strongly countercyclical during
most time periods.
It is the final criticism - that the Solow residual is a problematic measure of
technology shocks - t h a t has remained the Achilles heel of the RBC literature. The key
issues in this area involve quantitative rather than qualitative disagreements. With the
exception of the two oil shocks, it is hard to identify the macro shocks that produce the
productivity variations suggested by the Solow residual. If these shocks are large and
important why can't we read about them in the Wall Street Journal? Also, the Solow
residual often declines suggesting that recessions are caused by technological regress.
Finally there are several measurement problems that can make the Solow residual a
bad measure of productivity at cyclical frequencies. Summers and Mankiw emphasized
the importance of labor hoarding, that is, unmeasured variation in labor effort over
the business cycle. Perhaps even more important than labor hoarding is the cyclical
variability in capital utilization. Solow-residual based measures of technology shocks
that do not account for unmeasured variations in labor and capital will tend to be more
volatile and procyclical than true shocks to technology.
These difficulties arose as well in the earlier literature on growth accounting, where
the Solow residual had its origins. The stated goal of that literature was to measure the
long run evolution of disembodied technical progress, not the short run behavior of
productivity. Its hidden agenda was to make the Solow residual negligible, that is, to
measure production inputs well enough that all growth in output could be accounted for
by movements in factors of production. For this reason the residual was often referred
38 Given the high correlation between investment and output it is not surprising that the model cannot
display enough investment volatility if its share is unreasonably high. As the steady-state investmentoutput ratio increases the volatility of investment has to converge to that of output.
39 See Boldrin, Christiano and Fisher (1995) and Christiano and Fisher (1995). These models employ
a two-sector structure and use preferencesthat feature habit formation.
963
to as a "measure of our ignorance". Growth accountants were horrified when they saw
the measure of their ignorance recast as the main impulse to the business cycle. For
now we will put the problems associated with the Solow residual as a measure of
technology shocks on the back burner. But we return to them in Sections 7 and 8.
In the standard RBC model, productivity shocks are central to the nature of business
cycles. In this section, we will discuss three major aspects o f this relationship. First, we
explain that the standard RBC model requires large and persistent productivity shocks,
by considering how the comparative dynamics of the model change as productivity
persistence is altered. Second, we show how the assumption that agents have rational
expectations matters to the nature of real business cycles. Third, we discuss why other
shocks cannot easily generate real business cycles in the standard model.
964
an innovation standard deviation that is 0.0012 or about ~ times as large as the Solow
residual. The result of this is shown in panels 1 and 2 of Figure 8: real business cycles
explain a very small fraction of output and labor volatility. Since the standard RBC
model is approximately linear, changes in the standard deviation of the innovations,
std(e), simply work to rescale the model's fluctuations. We will return later to discuss
more of the details of the computation of these alternative shocks, but at present
it is sufficient to note that they were not chosen arbitrarily. Rather, they arise from
correcting the Solow residual for the effects of varying capital utilization in ways that
we will discuss further in Sections 6-8 below.
Why does the standard model require persistent shocks? By saying that the
variations in productivity must be persistent, we mean that the series generated from
the standard RBC model will display autocorrelation similar to the US data only if
p is near one. To discuss this, we consider in detail how the standard RBC model's
implications depend on the extent of serial correlation in productivity. We begin by
discussing the response of the economy to a serially uncorrelated productivity shock,
i.e., the solution of the model when p = 0. While the dynamic responses to this
shock shown in Figure 9 are the result of a complex set of factors - the preferences
of households for consumption and labor supply, the production function and the
mechanism for accumulating capital, and the interaction of households and firms in
general equilibrium - the key mechanisms can be easily described.
Productivity is assumed to increase by one percent (e = 1) in the initial period
(date 1). Given the rise in the marginal product of labor resulting from the increase in
productivity, the representative household faces an unusually high opportunity cost of
taking leisure in this initial period. While there are offsetting income and substitution
effects, the model's preferences were chosen so that a permanent increase in the
real wage (such as the one associated with the trend in technical progress) generates
exactly offsetting income and substitution effects so that labor and leisure are left
unchanged. An implication of this result is that N has to rise in response to a temporary
productivity increase. With a temporary shock, there is a much smaller income effect
and there is greater incentive to substitute intertemporally, since the current wage is
high relative to expected future wages. On net, the positive labor response amplifies the
productivity shock: the impact effect on output in Figure 9 is 2%. Half of this response
is due to the direct effect of the productivity shock and half due to the increase in
labor.
The representative agent must choose what the economy will do with all this
additional output. One possibility is to consume it all in period one 40. However, this
would be inefficient given that the marginal utility of consumption is decreasing,
thus inducing a preference for smooth consumption paths. It is optimal to increase
consumption both today and in the future. In fact, given that there are many future
40 Our figures make the impact date of the shock period 1, while the earlier theoretical analysis made
the initial period zero. The discussion in the text follows the dating conventionin the figures.
965
64-
2"E o
"?
g.-~,
-4-
~T'W"
-651
56
61
66
71
76
81
86
91
96
Date
4T
Labor Input w i t h s m a l l A s h o c k s
Date
O u t p u t w i t h s m a l l e r labor elasticity
6,
4
2
~o
~.-2.
51
56
61
66
71
Date
76
81
86
91
96
it
TI ? qI ~ ~
V
" w "?'
"T/" w -
~
51
56
....
61
66
71
Date
76
81
o"='
86
91
96
Fig. 8. Consequences of smaller shocks and smaller labor elasticity. Sample period is 1947:2-1996:4.
All variables are detrended using the Hodrick-Prescott filter.
R.G.
966
1.2-
productivity
Capital
1~ 0.8-
~o.~.
0.40.20
Quartem
Quarte~
9.
Investment
8.
7.
co:
~. 0.06
~04
0.02
_;
Quarters
10 11 12 13 14 15 16 17 18 19 2 o
Quarters
1.6.
Output
Labor
Input
IA.
1.2.
1,
0,8,
g. o.o.
0.4.
0.2,
0
Quaae~
Quar~
Real Wages
0.6
0-
Real Interest
Rate
-0.5,
0.5
-1
0,4
-ili
o.~
g.
0.2
10
11
Quarte~
12 13
:
14
15 16
17
18
~I
~...................
~ ; ~ ~ '~ ~ ~ ;o ;~ ,'~;~ ;~ ;~ ;~ ;~ ;~ ;o~'o
19 2 0
Qua~e~
967
periods, only a small fraction of the output windfall will be consumed at time 1; most of
it will be invested. Thus investment rises by 8% in response to a 2% increase in output.
It is interesting to note that the high volatility of investment, which Keynes ascribed
to "animal spirits", arises naturally in this economy as the flip side of consumption
smoothing.
In the future, which begins with period 2 in Figure 9, productivity retums to
its original benchmark level. The only difference relative to period zero is that the
economy has accumulated some capital and only a relatively small amount since the
productivity shock lasted just for one period. In line with the transitional dynamics that
we discussed in Section 3 above, the optimal policy for the economy is to gradually
reduce this excess capital by enjoying higher levels of consumption and leisure. The
real interest rate again signals individuals to adopt these consumption and leisure
paths: with a purely temporary change in productivity, the real interest rate falls in the
impact period and in all future periods, making it desirable for individuals to choose
consumption profiles that decline through time toward the steady-state level.
The impulse response makes it clear that there will be no tendency for a period of
high output and work effort to be followed by another period which has similarly high
output. That is: the basic neoclassical model does not produce substantial internal
propagation of temporary productivity shocks, a point which has been stressed by
Cogley and Nason (1995). The effects of the one-time shock are propagated over
time: the large investment in period 1 leads to high values of the capital stock that
keep output above its steady-state level in the following periods. But this propagation
mechanism is very weak. This weakness, together with the fact that Solow residuals
display substantial persistence led most RBC studies to focus on specifications in which
the persistence is inherited from the shock process.
What happens with serial correlation in productivity consistent with Solow
residuals? The solid line in Figure 10 depicts the effect of a serially correlated
productivity shock using the estimate discussed in Section 3 above ( p = 0.979).
In this case, the different series exhibit realistic persistence, which is inherited from
the shock. The same mechanisms are at work as in the case of a purely temporary
shock, but these effects are now drawn out over time. We now have an extended
interval in which productivity is above normal. During this interval, workers respond by
increasing their labor supply and most of the additional output is invested. Interestingly,
high productivity is now initially associated with a high real interest rate, since the
marginal product of capital schedule [At+ iD1F(kt+l,Nt+ 1) - 6] is shifted upwards
by the productivity shock and by a higher level of future labor input, with capital
responding only gradually via the accumulation of investment. However, later in the
impulse responses, the rate of return is below its steady-state level because the capital
stock has been built up while the stimulative effects of the productivity shock and
labor input have dissipated. This leads consumption to initially grow through time and
then subsequently to decline back toward the stationary level. Later in the impulse
responses, as productivity converges slowly to its normal level, labor supply actually
drops below the steady-state level as the economy enters a phase that resembles the
968
Productivity
1.2
Capital
14
1.2
1
1
0.8
0.8
:0.6
O.S
0.4
0,4
&2
0
-O.2
-0.2
1
26
,
,0
26
lO{)
,
51
76
~00
Qual'~e~
Q.a~m
Consumpf~l
I.vestillent
5-
14
4~
O.8,
0.62-
~" 0.4
0.2
26
26
-O.2
51
IO0
76
100
Q.aae~
Quar~'l
Output
Labor Input
1.4
0;7
1.~
0.6
0.5
&8
0.4
o.e
o,3
0.4
0.2
0.2
0.1
-O.2
1
26
51
Quarters
76
=0,1 I
1
lOO
Real Wages
I
26
I
51
Quarters
76
I
lOO
1.2
0.8
15
06
!:
,o
== 0.4
-0.2
26
51
Quart~m
76
100
Quarters
Fig. 10. Comparative dynamics to more persistent productivity shock. Circled lines are impulse responses
filtered with the Hodrick-Prescott filter,
969
transitional dynamics discussed above. Investment also eventually drops below the
steady state, as the economy runs down the capital that was accumulated during the
initial expansion.
As with the case of the purely temporary shock discussed above, the early part of
the impulse responses is dominated by the fact that the productivity shock raises the
desirability of work effort, production, investment and consumption; the latter part of
the impulse response function is dominated by the transitional dynamics, i.e., reduction
o f capital back toward its stationary level.
These impulse responses govern the autocovariances of productivity, output and
other variables. With many periods of high output, there will be positive correlation
between output and its past values: expansions and recessions will persist for many
periods.
Since the HP filter is so widely used in the real business cycle literature, it is
worthwhile investigating its effects on the impulse response function, as an indication
o f the effects that it has on the moments of the different variables: In Figure 10, the HPfiltered impulse responses are given by the generally lower paths that are highlighted
with the 'o' symbol. One notable feature of this filtered impulse response is that there
is less tendency for series to remain above or below their normal levels, i.e., filtering
reduces the persistence of the various series: This effect is particularly noticeable for
output and for productivity. Filtering also flattens the response of consumption and the
real wage, at the same time that it makes the capital stock largely acylical.
970
Productivity
Capital
1.6
1.4
1,
E~:~--~
1.2
.....
0.~.
~ O.S
o.~,
o.4.
O.4
o.2.
0.2
0
2
1
Quam~
13 14 15 I 6 17 18 19 20
10 1~ ~2 13 14 ~5 16 17 18 19 20
Q~aers
Investment
Consumption
4.~
aS
O.2
c.
:
2
:
3
:
4
:
5
:
6
:
8
0.5
o.
: : : : : : : : : : : :
9 10 11 12 13 14 15 16 17 18 19 20
10 11 12 13 14 15 16 17 18 19 20
Quar~
Ouarm~
Laborlnput
Output
0.7
1.2
E
1
"
:
2
:
3
----*--*~*--~*----,~.o
o.~ ~
o.4
o.2
0.2
c,
1
:
4
:
5
:
6
:
7
:
8
"~2~,:~:p.~
~~
~,:~ j,;~,~
,,
;T ..................
: : : : : : : : : : : :
9 10 11 12 13 14 15 16 17 18 19 20
9 10 11 12 13 14 15 16 17 18 19 2(1
Quarle~
Quart~m
Real Wages
1.6
2~T
2O '
e.
1
9 10 11 12 13 14 15 16 17 18 19 20
Quar~
"
~; t , ,
-s
I
4
I
5
I : : "
:
"~~"~
6
: :
: ;
:
; :
:
:
:
:
10 11 12 13 t4 15 16 17 18 19 2(1
Quariers
Fig. 11. Comparative dynamics to fully permanent shock. Standard model, p = 0.979 (circles); p = 1
(stars).
971
972
C o n s u m p t i o n : Wealth Effect
o.+t
o.e+ T
.:~- -+.p- +:+- ~ p -+:+~ - ~:P -~:p ~: P ~::~::~':+ ++P+,~++- -.+:<+~z+~z- +,:<~+P -,++
o.~
o:1
0 m5 ~l..=l_=,l__l+l=+l~l+l__l+l+l+l__l+l__l__l+l__l__l__
+,ia
~+"
,/.,-+~+~ ~ . ~ .
o.2
0*15
"~-~,~,
0.1
0.0
26
76
lOO
+0.8 1
1
2+
~1
57+
100
ole
0.7 ,-+, . . , + + , . + 4 . - , - , - + , + . - , . . , . , . + , . . , . + , + , . - ,
,+,
0+
0..I
o.+ I
0.3
0.2
o.1
o
1
I
76
-, + + , - . , + ,
0.2
I
51
2s
--"
0:1
16
I1
loo
.I6
1100
Quarle~
0.6
+m+m+mmmmm++zm+:mmmmm
..m++'
0.4'
o,25-
=--,--,--,--,--,--,--,--,=,
0,2
+-,+--,--, -+=.-,+.+.=-+,-+,+-,-.,-~=
02.
0.15
-0,2
0.1-
.0.4
-06
0.05
o
~6
Sl
16
4.8
5~
10o
'
76
i~
0.6
+=~,
0.4
,~,~,~,I,~,
002
o
~+'~'+
0,2
0
-0,2
'\
-0.4
,- + ' ~ "
+~,_,_=,=,_,_,_,_~,_,_,
,~ . + : ~ : + . . ~ ~ ~ - - A : ' + . - - . + ~
"
, / + . . ++ ' '::+- ~ -~
m,o2
+3.o+
"%i~
+"+I/l"
f,1"
+o.+m "i"
,+"
.~"~]'+"
-0.6
-0.8
-1
"+ m m + m + + + , z - , ~ . m
51
o+.~am
+.14 .p/
-0.++,
m ram+ram+++
71s
loo
o,+I
++
7+
1oo
Quarl~
973
determine the consequences for consumption and leisure at date O, which we call the
wage effect. This effect is analogous to the Hicksian effect of the wage on consumption
and leisure, in that it holds utility fixed, tracing out a substitution response. However,
in our general equilibrium model, the productivity shock implies that wages change in
all periods, {wt}t-0. Thus, the wage effect in Figure 12 takes into account the entire
change in the time path of wages, combining static and intertemporal substitutions.
When p = 0.979, the representative household correctly understands that productivity
will raise the path of wages at date 0 and in many future periods, but that the long-run
level of the wage will be unchanged. Accordingly, the household plans to consume
more at date 0. Leisure hardly changes at all because the current period is about
"average"; this conclusion depends on the particular p value. However, this pattern
is sharply altered when p = 1, for then the household recognizes that the current wage
is below the long-run wage and leisure rises due to the wage effects that stem from a
positive productivity shock 42.
In the general equilibrium of our RBC model, there is one additional channel:
interest rate effects that induce intertemporal substitutions of consumption and leisure.
In general, these intertemporal price effects are a powerful influence, but one that is not
much discussed in informal expositions of the comparative dynamics of RBC models.
In particular, permanent increases in productivity lead to high real interest rates and
these induce individuals to substitute away from date 0 consumption and leisure as
shown in Figure 12.
We are now in a position to describe why a permanent shift in productivity (arising
when p = 1) has a smaller effect on labor than a persistent but ultimately temporary
shock ( p = 0.979). When the shock is temporary, there is a small wealth effect
that depresses labor supply but temporarily high wages and real interest rates induce
individuals to work hard. When the shock is permanent, there are much larger wealth
effects and the pattern of intertemporal substitution in response to wages is reversed
since future wages are high relative to current wages. However, labor still rises in this
case in response to productivity shocks due to very large intertemporal substitution
effects of interest rates.
5.3. Why not other shocks?
We have just seen that the basic real business cycle model driven by persistent
technology shocks can produce realistic business cycle variation in real quantities.
Do these same patterns emerge when the economy is buffeted by other disturbances?
Shocks to fiscal and monetary policy have been long standing suspects in the search
42 The wage effect on consumptionis constant acrosstime in each case because the separablemomentary
utility function implies that efficient consumption plans do not depend on the amount of work.
Equivalently, with this utility function, there is a general substitution effect on consumption at all
dates that works much like a wealth effect.
974
for the causes o f business cycles. It is thus natural to ask what are the effects o f these
shocks in the standard RBC model.
Shocks to government spending cannot, by themselves, produce realistic patterns
o f comovement among macroeconomic variables 43. This result stems from the fact
that an increase in government expenditures (financed with lump sum taxes) gives
rise to a negative wealth effect that induces consumption to fall at the same time that
labor and output rise. Thus, if government spending were the only shock in the model,
consumption would be countercyclica144.
Changes in labor and capital income taxes have effects that are similar to
productivity shocks. However, these taxes change infrequently making them poor
candidates for sources o f business cycles fluctuations.
Monetary policy shocks have small effects in this class o f models both in versions
in which money is introduced via a cash-in-advance constraint [Cooley and Hansen
(1989)] and in models that stress limited participation [Fuerst (1992), Christiano and
Eichenbaum (1992b)]. Many researchers are also currently investigating the nature of
business cycles in models that start with the core structure of an RBC framework
but also incorporate nominal rigidities o f various forms 45. This research has not yet
produced a business cycle model that performs at the same level as the RBC workhorse
described in Section 4.
Since the basic RBC model contains explicit microeconomic foundations, part o f the
literature has tried to improve its predictions for individual behavior. Other researchers
have sought to improve the fit between model and data, focusing on moments and
sample paths o f macroeconomic time series. In this section, we discuss two strands of
this research: work on labor supply and on capital utilization.
6.1. The supply of labor
There is a substantial body of work that focuses on the labor supply and, more
generally, on the labor market in RBC models. This research is motivated by four
difficulties encountered by the basic model on micro and macro dimensions. In most
43 There is a large literature that investigates the effects of fiscal policy in an RBC context. References
include Wyrme (1987), Christiano and Eichenbaum (1992a), Rotemberg and Woodford (1992), Baxter
and King (1993), Braun (1994), McGrattan (1994), and Cooley and Ohanian (1997).
44 For an early discussion of this difficulty, see Barro and King (1984). There is actually some evidence
that in historical periods dominated by large shocks to government expenditures consumption was
countercyclical, see Correia, Neves and Rebelo (1992) and Wynne (1987).
45 Examples include Cho and Cooley (1995), Dotsey, King and Wolman (1996), and Chari, Kehoe and
McGrattan (1996).
975
RBC models the implied labor supply elasticity to wage changes is very large, relative
to micro studies. All of the variation in aggregate hours in the model arise due to
movements in hours-per-worker, while the US experience is that most of the action
comes from movements of individuals in and out of employment. Labor in the model
lacks a close correspondence to labor in data (see Figure 6). Finally, labor input and
its average product are very highly correlated in the model, but not in the data.
(6.1)
976
choosing 1-N 1 which is the upper bound suggested by Pencavel's estimates, rather
than ~-N = 4 as in the model of Section 4. There is an important reduction in the
volatility of output in the third panel of Figure 8. However, the model loses most of
its ability to produce fluctuations in labor (see the fourth panel of Figure 8). In terms
of moments, the standard deviation of output falls from 1.39 to 1.16 with the smaller
labor supply elasticity and the standard deviation o f labor falls from 0.67 to 0.33.
=
46 In actual economies,variationsin aggregate hours reflect changes at both the intensive and extensive
margins. In a model where workers have different fixed costs of going to work, Cho and Cooley (1994)
have captured both of these responses. Such a framework appears necessary to explain the differing
cyclical patterns of employmentand hours-per-workerin the USA and Europe that are documented by
Hansen and Wright (1992).
977
choice set. By entering a lottery an agent can choose to work a fraction p of his days
remaining unemployed a fraction (1 - p ) of his time. Let us use the subscript 1 to
denote those agents who are assigned to work by the random lottery draw and the
subscript 2 to refer to the unemployed agents. The expected utility of an individual
prior to the lottery draw is
p u ( c l , 1 - H ) + (1 - p)u(c2, 1),
(6.2)
(6.3)
where c is per-capita consumption. Maximizing Equation (6.2) subject to condition (6.3), we find that marginal utility of consumption must be equated across types,
i.e.,
DlU(Cl, 1 - H ) = DlU(C2, 1),
(6.4)
(6.5)
where L = 1 - p H is the average number of hours of leisure in the economy and where
vl = v(1 - H ) and v2 = v(1).
There are three notable features of this economy. First, even though each individual
agent has a finite elasticity of labor supply, the macroeconomy acts as if it were
populated by agents with a more elastic supply of labor. In particular, the standin representative agent for this economy has preferences that are linear in leisure,
implying a infinite )>constant elasticity of labor supply [see Equation (6.1) with
t / = 0], a feature whose consequences we explore further below. Second, contrary to
conventional wisdom, this is an economy in which it is optimal to have unemployment.
Finally, agents actually choose to bear uncertainty by entering the lottery arrangement
instead of working a fixed number of hours in every period.
It is interesting to explore further why the individual elasticity of labor supply differs
from that of the economy as a whole and the consequences of this difference for the
978
determination of output and labor. The individual elasticity of labor supply answers the
question "how many more hours would you work in response to a 1% raise in salary?".
But the answer to this question is irrelevant because the number o f hours worked is not
flexible, it is either H or zero. In other words, the intensive margin is not operative and
hence its elasticity of response is irrelevant. Proceeding to the consequences for the
determination of labor, the preferences o f the stand-in representative agent (6.5) imply
that small changes in wages and prices can lead to very large effects on quantities. To
see this, consider an isolated individual maximizing
[3tIlg(ct)+(1-Lt)llgQ~
) + log(v2) 1 t=0
Along the relevant intertemporal budget constraint, suppose that the discounted cost of
a unit o f leisure is [3t)~twt. Then, for the individual to work part of the time (0 < Lt < 1)
But, if this condition is
in each period, it must be the case that )~twt = 1 1og(vl/V2)47.
satisfied, the individual is indifferent across all sequences of leisure which imply the
same level of ~ _ 0 /3t [(1 - Lt ) ~ log(vl/v2 )]: there is an infinite intertemporal elasticity
of substitution in work.
One implication of this labor supply behavior is that it is the demand side o f the
labor market which determines the quantity o f employment and work effort in the
equilibrium o f the indivisible labor model. From this perspective, firms choose the
quantity o f labor that equates its marginal product to the real wage, with the position
of the demand schedule being shifted by the level o f productivity and the capital
stock. Since the capital stock and the multiplier )~t are endogenously determined, this
labor market equilibrium picture is incomplete, but it is a useful partial equilibrium
description.
The indivisible labor model with more general preferences: When the indivisible
labor model is generalized, as in Rogerson and Wright (1988), there are interesting
new conclusions. To develop these, we use the utility function (3.8), with a 1.
Efficient risk-sharing condition implies that consumption allocations must satisfy
Cl ~C2
(6.6)
47 If AtW~ < 1 1og(v~/O2), our agent spends all available time at t in leisure (L = 1). If
,~twt > log(vl/V2), our agent devotes no time to leisure (L = 0).
4s This conclusion makes use of the fact that v2 = v(l) > vI = v(1 -H), which follows from the fact
that v is an increasing function.
979
u(c,r) =
{cl-av*(L) l - a -
1}
(6.7)
where
o
v*(L)=
v1~ +
1-
v ~-) i ~ .
There are two points about this expression. First, the stand-in's utility function inherits
the long-run invariance o f hours to trend changes in productivity from the underlying
utility function (3.8). Second, the stand-in's utility function inherits the original utility
function's properties with respect to effects of changes in leisure on the marginal utility
of consumption. In particular, when a > 1, the marginal utility of consumption is
decreasing in leisure.
Let us again think about an isolated individual maximizing lifetime utility,
~ _ o f i t u ( c t , L t ) , but with the new momentary utility function (6.7). As with our
discussion of the representative worker in Section 4 and as with our previous discussion
in this section, the stand-in agent equates the marginal utility of consumption and the
marginal utility o f leisure to the shadow values along the economy's resource constraint
(Dlu(ct,Lt) = )~t and Dzu(ct, Lt) = ~twt = ),tAtDzF(kt,Nt)). These conditions must
always hold if there is an interior optimum for work effort, i.e., 0 < Lt < 1 in each
period. Taking loglinear approximations to this pair of conditions, we find
- a ~ t + (1 - a)xL,
= ~t,
LDo* (L)
v*(r)
(6.8)
(6.9)
49 There are two steps to this demonstration. First, one shows that efficient risk-sharing implies that
expected utility is proportional to:
1 {
1-a
[ (1 a)
ct a [ pol~
( ]-o'~]0 }
+(l-P)V2'J-1
if a l ,
) - Lw - L (wN/y)
-LD2u(c,L
CDlu(c,L)
c
N (c/y)"
980
This set of equations reveals that there is infinitely elastic labor supply even when
the preference specification is not separable. That is, the pair of equations implies that
0 = ~, + a~,
which is the statement that the stand-in will supply any amount of work at a
particular real wage. But because preferences are nonseparable, variations in work
require variations in consumption. When a > 1, in particular, workers require more
consumption than nonworkers and aggregate consumption is negatively related to
leisure, i.e.,
Thus, this model involves a modified form of the permanent income hypothesis, which
includes the effects of changes in work effort on the marginal utility of consumption.
Baxter and Jermann (1999) have argued that this type of preference nonseparability
will arise in any model with household production; they have also stressed that this
specification can make consumption more cyclically volatile.
6.2. Capacity utilization
In the standard version of the neoclassical model, there is a dramatic contrast between
the short run and long run elasticities of capital supply. The short run elasticity of
capital supply is zero: there is no way for the economy to increase the capital stock
inherited from the previous period. In contrast, the long run elasticity of capital supply
is infinity: there is only one real interest rate consistent with the steady state o f the
economy. This difference between short run and long run elasticities stems from the
assumption that capital services are proportional to the stock of capital. This is an
assumption we make every time we write a production function as Y = F ( K , N ) .
While this assumption may be suitable for some purposes, it is clearly problematic
for business cycle analysis. The third panel of Figure 3 suggests that capacity
utilization displays pronounced cyclical variability. The fact that equipment and
machinery are used more intensively in booms than in recessions is corroborated by the
procyclical character of electricity consumption in manufacturing industries [Burnside,
Eichenbaum and Rebelo (1995)] and by the fact that expansions are accompanied by
the use of two and three shifts in manufacturing industries [Shapiro (1993)]. All this
evidence suggests that the flow of capital services is high in expansions. In contrast,
recessions are times when capital tends to lie idle, thus producing a small service flow.
Several authors have extended the basic RBC model to incorporate variable capital
utilization. Kydland and Prescott (1988) showed that introducing time-varying capital
utilization enhanced the amplification capability o f their 1982 model. Greenwood,
Hercowitz and Huffman (1988) introduced variable utilization in a model that features
981
shocks to the productivity o f new investment goods 51. Finn (1991) used a similar
framework to study the interaction between capital utilization and energy costs. In
her model, more intensive capital use accelerates the depreciation o f capital and raises
marginal electricity consumption. Burnside and Eichenbaum (1996) explored a model
with both capital utilization and labor hoarding. They showed that these two features
significantly enhance the ability of the model to propagate shocks through time 52.
Modeling variable utilization. Most studies o f variable utilization assume that
depreciation is an increasing function of the utilization rate 53. The benefits from
variable capital utilization can be incorporated into the production function as follows:
Yt = AtF(ztKt,NtXt) = At(ztKt) l a(NtXt)a.
where zt denotes the utilization rate 54. The costs o f variable capital utilization are
imbedded in the following law o f motion for the capital stock:
K t + 1 : ~ + [1 - b ( z , ) ] K t ,
(6.10)
which is the requirement that the marginal benefit in terms o f additional output
produced be equated to the marginal cost in terms o f additional units o f capital being
worn out.
The consequences o f variable utilization. To explore how efficient variation in the
utilization rate affects the linkages in the economy, we linearize Equation (6.10) to
51 These shocks tend to make consumption and investment move in opposite directions. Introducing
capital utilization eliminates this counterfactual correlation between consumption and investment.
52 Their model is also capable of producing a humped shape response of investment to technology
shocks - a feature that is common in empirical impulse response functions estimated using VAR
techniques.
53 An exception is Kydland and Prescott (1988).
54 For simplicity, we use the Cobb-Douglas form throughout our discussion of capital utilization.
55 Thus, it has a positive first derivative Dr(.) and a positive second derivative De6(.).
982
obtain an expression for ~t and substitute this result into the linearized production
function:
33t = 3t + a-~t + (1 - a)(]c~ + ~t)
= A t + (1 - a)/ct + aNt + ~1 - a ( A t - a ] c t + a N t )
(6.11)
~t:(f:,-Nt):3t+(1-a)[ct+(a-1)Nt+
1-a
(At-alct+aNt)
(6.12)
and, as ~ is driven toward zero, the response o f the real wage approaches wt = FAt.1
^
In other words, the labor demand schedule drawn in (w,N) space "flattens" as
depreciation becomes less costly on the margin (~ falls). When ~ is driven to zero,
the labor demand curve becomes completely flat.
983
(7.1)
where log(z/) is the log o f electricity use. They find that when electricity use is
employed as a proxy for capital services the character o f the Solow residual associated
with the manufacturing sector changes dramatically: (i) there is a 70% drop in the
volatility o f the growth rate o f productivity shocks relative to output, implying that a
successful model must display much stronger amplification than the basic RBC model;
(ii) the hypothesis that the growth rate o f productivity is uncorrelated with the growth
rate o f output cannot be rejected; and (iii) the probability o f technological regress
assumes much more plausible values, dropping to 10% in quarterly data and to 0% in
annual data. These corrections to the Solow residual significantly reduce the fraction
o f output variability that can be explained as emanating from shocks to technology 58.
Using the model to measure capacity utilization: A n alternative strategy is to
use the model's implications for efficient utilization to solve for the unobserved
57 Several variants of this proxy strategy have been used to shed indirect light on the presence of labor
hoarding. Bils and Cho (1994) use time and motion studies to document the presence of variability in
effort. Shea (1992) uses data on on-the-job accidents to construct an indirect measure of labor hoarding.
Burnside, Eichenbaum and Rebelo (1993), Sbordone (1997), and Basu and Kimball (1997) postulated
a model of labor hoarding that they proceeded to use to purge the Solow residual of variations in the
level of effort.
s8 Aiyagari (1994) proposed a method to compute a lower bound on the contribution of technology
shocks to output volatility. His procedure relies on knowledge of two moments in the data: the variability
of hours relative to the variability of output and the correlation between hours worked and labor
productivity (which is essentially zero in the data). Unfortunately, his method is not robust to the
presence of labor hoarding or capacity utilization.
984
utilization decision, i.e., zt. In essence, this empirical strategy corresponds to our
theoretical method in the previous section, when we solved out for zt in order to
derive Equation (6.11), which describe how output responds to changes in labor, capital
and productivity when utilization is efficiently varied. One possibility would be to
exactly follow this strategy, substituting observed variations in labor and capital into
Equation (6.11) to compute the productivity residual, but we use a more "reduced
form" approach that we describe more fully in the next section.
The specification and calibration o f the model follows the same general approach that
we used in Section 4, but with some relatively minor modifications.
Restrictions on the steady state: First, we know that the production side o f the basic
model determines most aspects o f the steady state and that continues to be true with
variable capital utilization. The efficiency condition for utilization in the steady state
determines a steady-state utilization rate such that r + b(z) = D6(z), with the remainder
o f the steady-state relative prices and great ratios then adjusted to reflect the fact that
the flow o f capital services is z K rather than K.
59 The approach was suggested by Mario Crucini in unpublished research many years ago, so perhaps
we should call these "Crucini residuals". Another application is contained in Burnside, Eichenbaum and
Rebelo's (1993) study of unobserved effort (labor hoarding). Ingram, Kocherlakota and Savin (1997)
use a similar procedure to infer information on observed shocks to the home production sector.
985
Table 4
Calibration of high substitution economy
~r
(Ye
0.984
1.004
0.667
0.025
0.1
0.9892
0.0012
Yt = : ykkt + ZyAA,.
Using this decision rule together.with data for output and capital (which we logged
and linearly detrended), we can compute an initial guess about the time series for
technology shocks 61:
1
_ ,ryk
60 There is no unique way of computing this shock process, but rather any of the model's decision
rules could be used in this way or these rules could be combined with other relationships in the
model. For example, one could exploit the decision rule for utilization as in Burnside, Eichenbaum
and Rebelo's (1993) analysis of labor hoarding, ~t = JVyklct+ ~yA~lt, and combine this with the modified
Solow decomposition(7.1). This alternativemethod would produce a different shock process, which lead
to broadly similar, but somewhat less dramatic results. The difference between these two productivity
measures lies in whether labor in Equation (7.1) is taken from the data or from the model.
61 We should not use the empirical capital stock series since these are flawed in the eyes of the model:
they are computed assuming constant rates of depreciation. This can be circumventedby using a second
decision rule to compute the "true" capital stock series. In practice this has little impact on the results.
986
This guess is not exactly fight because the serial correlation coefficient (p) for this zit
series need not match that used to solve the model and to construct the Jr coefficients.
Therefore, once we obtain a time series for zit, we compute its persistence (p) and use
this new value to solve the model again. Using the new decision rule, we recompute
~it and once again its calculate its persistence. We continue this process until the new
and old estimates for the serial correlation of~it are the same. This iterative procedure
yielded an estimate of 0.9892 for the first-order serial correlation and 0.0012 for the
standard deviation of the et.
8.2. Simulating the high substitution economy
With a series of productivity shocks in hand, we simulated our model economy's
response to these shocks just as we previously did for the standard RBC model.
Figure 13 displays the results, which we think are dramatic. Panel 1 shows the model
and actual paths for output, which are virtually identical. In part, this is an artifact of
our procedure for constructing the technology shock, which is a weighted average of
output and capital as we just discussed. For this reason, we think that the performance
of the model should not be evaluated along this dimension. Instead, the model has
to be judged by its predictions for other variables of interest. The remaining panels
of Figure 13 display the model's implications for total hours worked, consumption
and investment, with all of these series detrended with the HP filter. The correlation
between the empirical and the simulated series is 0.89 for labor, 0.74 for consumption
and 0.79 for investment! This remarkable correspondence leads to three sets of
questions, similar to those which arose in the analysis of the standard RBC model.
First, how do small variations in productivity have such dramatic effects? Second, what
are the properties of the technology shocks? Third, how sensitive are the results?
8.3. How does the high substitution economy work?
The high substitution economy contains four mechanisms that substantially amplify
productivity shocks and lead to strong comovements of output, labor, consumption
and investment. To begin, variable capacity utilization makes output respond more
elastically to productivity shocks in Equation (6.11), which we repeat here for the
reader's convenience:
Since utilization of capital increases when there is a positive productivity shock, there
is a direct effect which is part of the amplification mechanism. In the limiting case of
= 0 for example, a labor's share of a = 2 implies that the productivity shock raises
output by ~1 or 3 times its direct effect. We use a value of ~ = 0.1 in constructing our
simulations, so that the effect with a = 2 is 1 + ~l-a = 1+33
~_~ = 1.43. Thus, variable
utilization helps create amplification, but only in a modest manner.
987
Output
51
56
61
66
71
76
81
86
91
96
Date
Input
Labor
4.
2,
0
4-2,
#.
-4.
"
--
........
Data
-6,
I
51
-8
56
61
66
71
76
8I
86
91
96
Date
15
Investment
Mode[
10
5
~- -5
-10
-15
51
56
61
66
76
71
81
86
91
96
-20
6.
5.
4.
3.
2.
Consumption
~..~
Model
--,-~-- - Data
~.
~o
-1.
-2.
-3.
51
56
61
66
71
76
81
86
91
96
Date
Fig. 13. Capacity utilization model: simulated business cycles. Sample period is 1947:2-1996:4. All
variables are detrended using the Hodrick-Prescott filter.
988
Relative to the standard RBC model that we discussed in Section 4, most of the
increased amplification in the model of this section comes from greater elasticity of
the labor demand and labor supply schedules. Highly elastic labor supply is due to
indivisible labor: work effort is highly responsive to small changes in its rewards.
In fact, we have previously argued that it is the demand side which approximately
determines this quantity in indivisible labor economies. Variable capacity utilization
makes the labor demand more elastic. As discussed above, labor demand is implicit
in the equation:
~t = ( ) t _ ~ t ) = ~ t + ( l _ a ) ~ + ( a _
1)~t + 1 - a
(At-aict+aNt)
In the model without variable utilization (or with ~ = c~), a one percent increase
in labor quantity causes the real wage to fall by 0.333 percent when a = 2, since
the coefficient on Nt is (a - 1). At the other extreme, as ~ is driven toward zero,
the response of the real wage to a productivity shock approaches ~t = FAt,
1 ^ i.e., the
labor demand schedule becomes more elastic until it is completely elastic in the limit.
With variable utilization, the combined coefficient on labor is (a - 1) + ~1-a
a . Using
a = ~ and ~ = 0.10, as in our simulations, we find that the combined coefficient is
(0.67 - 1) + 0.33
n 67 = -0.043: a one percent change in labor requires a decline in the
0.77 ....
wage that is an order of magnitude smaller than in the standard model. With indivisible
labor and variable utilization, a small productivity shock shifts up labor demand and
calls forth a large increase in labor supply. In order to determine the exact size o f this
change, however, it is essential that we simultaneously determine the path of capital
(kt) and the multiplier ()~t).
The final structural feature that is important for the simulated time series is the
nonseparable form of the utility function. In the standard Hansen-Rogerson case
of log utility, most of the model's change in output goes into investment rather
than consumption. However, since the efficient plan calls for the allocation of more
consumption to employed individuals when ~r > 1, the high substitution economy
displayed in Figure 13 involves more volatile consumption that corresponds closer to
the data. We return to a discussion of this feature in the context of impulse responses
later in this section.
989
7.7-
3 7.67.5. j ...........- , -
7.47.3-
7.2
51
56
61
66
71
76
81
86
91
99
Da~
0.9
0.8
0.7
0.6
0.5
0.4
9.3
9.2
9.1
0
-0.1
I
51
I
56
I
61
I
66
71
I
76
I
81
I
86
I
91
I
96
Date
4
3
2
1
i,
:L,V
. . . .
;'..
;'~
'
".
:~',
'
-1.
-2.
-3,
-4
I
51
I
56
I
61
I
66
I
71
I
76
I
81
I
86
I
91
I
96
Date
990
Table 5
Sensitivity analysis to different ~ values
value
ec
1
l
1
I
1
7
11o
Standard deviations
I
N
A
1.36
1.39
1.40
1.40
1.41
1.41
1.42
1.01
0.94
0.92
0.91
0.91
0.90
0.89
2.62
2.86
2.94
2.99
3.05
3.09
3.15
0.90
1.07
1.13
1.16
1.20
1.23
1.26
0.79
0.45
0.34
0.28
0.22
0.19
0.15
Persistence
parameter (p)
Likelihood of
technical regress
0.0061
0.0034
0.0026
0.0021
0.0017
0.0014
0.0012
0.9783
0.9798
0.9822
0.9841
0.9866
0.9880
0.9892
0.1859
0.1106
0.0653
0.0352
0.0101
0.0050
0.0050
We next discuss the sensitivity o f our results to the choice of parameters and to the
measurement o f output.
Sensitivity to parameterization. The value chosen for the parameter ~ is a key
ingredient in the results. This is not surprising since we know that when ~ equals
infinity the model with capital utilization reduces essentially to the standard model.
Table 5 shows how some key model statistics change with different values for ~.
For every value of ~ we used the iterative process described above to ensure that
the stochastic process assumed for A~ is in fact consistent with the properties o f the
technology shock implied by the model. In every case we report the persistence of
the shock ( p ) and the standard deviation of the innovation (e) as well as the implied
probability o f technological regress. Low probabilities o f technological regress can be
1
obtained for values o f ~ that are lower than 3"
As an alternative check on the sensitivity o f the model to ~, Figure 15 depicts the
impulse response for this model for three values o f ~: oe, I and 1 . To simplify the
comparison between these impulse responses we did not adjust the stochastic process
for the technology shock. All three responses were computed with the same standard
deviation o f innovation (a~, = 0.0072) and same persistence ( p = 0.979).
The three impulse response functions depicted in this figure have similar dynamic
properties, but vary mostly in the degree of amplification. The solid line is a fixed
capital utilization model (~ = oc) like the basic RBC model of Section 4, but with
indivisible labor. In this model, a productivity shock has a larger effect on output than
in the standard RBC model: when there is a one percent productivity shock, output rises
by just less than two percent on impact with fixed utilization (~ = ec). However, the
Ch. 14:
991
Productivity
1.2
Capital
5.
45,
1.
4.
0.8,
3.5,
. . . .
~0.6.
0,4.
,.,="
............................
3.
2.S.
2.
1.5.
=
1,
O.2.
.~..1
,. ................................................
O.5', ', :
2 3 4
o.
:
5
:
6
:
7
:
8
:
9
:
:
:
:
:
'. :
:
:
:
:
10 11 12 13 14 15 16 17 18 19 2O
0
2
Ouarte~
=~
10 11 12 13 14 15 16 17 18 19 20
Quarters
Consumption
35
Investment
7
6
~2ot
.......
o,
1
10 11 12 13 14 15 16 17 18 19 20
:
1
'.
3
'.
4
'.
5
'.
6
'.
7
'.
8
'.
',
'.
'.
'.
', :
', ', ',
:
10 11 12 13 14 15 16 17 18 19 20
'.
9
Quarters
14 T
Quarters
Output
14 112
~8
10 1I
12 13 ~4 15 16 17 18 19 20
Labor
.............
.
1
Quarters
1'i I
..
10 11 12 13 14 15 16 17 18 19 20
Quartg~
iit
Real Wages
Input
...
n,.r.,.a,o
0.6 ..............
0.4
0.2
0
. . . . . . . . . . . . . . . . . .
1
10 11 12 13 14 15 16 17 18 19 20
Quartar~
10 11 12 13 1,4 15 16 ~7 18 19 20
Quarters
Fig. 15. Sensitivity analysis to alternative utilization costs (~): solid line, ~ = e~; diamonds, ~ = ;
squares, ~ = ~ .
992
increase in amplification is small relative to what happens when indivisible labor and
capacity utilization are introduced simultaneously. A one percent productivity shock
has an impact effect on output o f 8 percent when ~ = 0.2 and o f 13 percent when
= 0.1. From various experiments with this model economy, it is clear that values
of ~ less than one are important to obtain substantial amplification. For example, if
there is a value of ~ = 1 then there continues to be on average a regress in the level of
productivity every ten quarters (see Table 5 above). It is important that econometric
evidence be produced on the cost o f varying capital utilization, so as to determine the
extent to which this high substitution economy is realistic 62.
One specific feature o f the impulse response in Figure 15 is worth some additional
discussion. In all o f the cases, the real interest rate increases in response to a positive
productivity shock, at least for the first twenty quarters shown in the graph. In all cases,
the level o f consumption broadly resembles the level o f output and the growth rate of
consumption is negative, even though the real interest rate is high by comparison to
its steady-state level. This behavior o f consumption reflects the fact that aggregate
consumption is the sum o f consumptions by individuals that are working and those
who are not. Since working agents have more consumption, an increase in the fraction
o f individuals working makes aggregate consumption rise and fall with aggregate
employment 63.
Sensitivity to the measurement o f output. We have seen that variable utilization and
indivisible labor produce an economy in which (i) small productivity shocks have large
effects on output; (ii) the standard Solow residual is substantially mismeasured; and
(iii) labor and output move together on an approximately one-for-one basis. In this
economy, however, there is an important sense in which output is mismeasured. There
is a standard line o f intuition which suggests that "intermediate" activities such as
utilization should not be too important for economic activity and, in this case, suggests
that the large effects o f productivity on output and the strong eomovement o f output
and labor are simply artifacts of output mismeasurement. To explore these ideas, output
net o f depreciation can be defined as
Ot = Yt - 6(zt)kt = AtF(ztkt, Nt) - 6(zt)k~
(8.1)
and this expression can be used to make four important points. First, output is also
mismeasured in the standard neoclassical model, i.e., even in the absence of a variable
depreciation rate. Second, with efficient utilization, changes in net output are
dot = F(ztkt, Nt) dA, + AtO2F(ztkt, Nt) dN,
+AtDlF(ztkt, Nt)(kt dzt + zt dkt) - D6(zt) ]~tdzt --
(~(Zt)
dkt
= F(ztkt, Aft) dAt + AtO2F(ztkt, Nt) dNt + AtOlF(ztkt, Nt) zt dkt - 6(zt) dkt,
62 Basu and Kimball (1997) provide an estimate of a parameter that is essentially our ~. Their point
estimate is about unity, but the parameter is very imprecisely estimated.
63 Baxter and Jermann (1999) stress that equilibrittm models with nonseparable preferences can generate
apparent excess sensitivity of consumption to income, working in a model where labor supply variation
is on the intensive rather than extensive margin.
993
where the latter equality follows from AtD2F(ztkt, Nt) kt dzt - D6(zt) ktdzt = 0 when
utilization is efficient. Thus, there is a sense in which the standard intuition is correct
because net output does not respond to utilization. Third, near the steady state, the
Solow decomposition for net output is
(8.2)
where m = ~ = [1 @]-l. This modification takes into account the fact that the
net production function :is more labor intensive and the fact that productivity shocks
affect gross output but not depreciation. Thus, for example, if depreciation investment
is 10% of output, then m = 1.11. Thus, if output is measured as net of depreciationindependent of whether capacity utilization affects depreciation - then this will tend
to strengthen the magnitude o f labor's effect on output. Fourth, most importantly for
our purposes, the net production function qSt = AtF(Nt, ztkt) - 6(zt)kt has the same
marginal product schedule for labor, A t D I F ( N t , ztkt) as does the standard production
function. Thus, our analysis of the "labor demand" consequences of efficient utilization
are unaffected by whether depreciation costs are deducted from output or whether
they are not. Returning to Equations (6.11) and (6.12), we can thus see that a "net
output" measurement requires that we replace Equation (6.12) with the modified
growth accounting expression (8.2), but that we need not change the labor demand
schedule (6.11) at all. Further, it is a highly elastic labor demand that is the key force
behind the great amplification present in our high substitution economy.
-
9. Conclusions
64 Just as this first round of problems is set to rest, new challenges arise for the RBC model regarding
the comovement between productivity and economic activity. In a recent paper, Gali (1996) argues,
using VAR techniques, that technology shocks actually reduce input usage in the aggregate economy.
994
amplification o f productivity shocks requires highly elastic labor supply and readily
variable capital utilization 65.
Although we have concentrated on the one sector neoclassical model, which has
been the central laboratory for most work on real business cycles, the next stages
o f RBC research will likely use richer frameworks, as we discuss next. However, we
believe that the exploration o f these richer frameworks will require consideration of
the structural features that we have stressed in this chapter.
One exciting research direction is the exploration o f models with multiple sectors,
i.e., a long overdue continuation o f the trail scouted by Long and Plosser (1983).
Interesting recent work on these models retains most o f the assumptions on preferences
and production opportunities commonly incorporated in the one-sector RBC model
[Horvath (1997), Dupor (1998)]. The one-to-one movement between hours and output
observed in aggregate data also holds at a sectoral level. To us, this suggests
the importance o f introducing variable capacity utilization into sectoral production
structures. Another promising direction is work on models with heterogeneous agents
[Krusell and Smith (1998), den Haan (1993)]. This work seems particularly important
for enriching labor market dynamics and modeling unemployment 66. Fleshing out
labor market dynamics is important on its own terms. But this work may also provide
us with an alternative way o f obtaining a highly elastic aggregate labor supply which
appears necessary for RBC modeling. A third interesting research direction seeks to
tmderstand the industry dynamics that seem intimately related to the business cycle
[Hopenhayn and Rogerson (1993), J.R. Campbell (1998)]. Finally, there are many
aspects o f microeconomic activity in addition to employment in which discrete choice
seems very important. It has frequently been suggested, for example, that the volatility
o f investment is related to the fact that much o f firm investment is lumpy in character
label and Eberly (1995), Caballero and Engel (1994)]. The incorporation o f lumpy
investment decisions into the RBC model and its implications for aggregate dynamics
is an exciting new direction o f research on which some initial progress has been made
[Veracierto (1996), Thomas (1997)]. The interaction o f lumpy investment with costly
capacity utilization seems a particular important topic o f investigation. All four of
This finding receives indirect support from two sources that do not rely on the VAR methodology:
Burnside, Eichenbaum and Rebelo (1996) document that a sectoral capital-utilization adjusted measure
of technology shocks is essentially tmcorrelated with production in 2-digit SIC manufacturing industries.
Basu, Fernald and Kimball (1997) show that, for this same set of industries, input usage is negatively
correlated with technology shocks. One interpretation of these facts is that they reflect the presence of
nominal rigidities that keep nominal aggregate demand fixed and lead inputs to contract in response
to a productivity increase [Gali (1996)]. An alternative flexible-price explanation for these same facts
involves a multisector model in which goods are complements so that a technology shock to an individual
sector does not necessarily warrant an expansion of input usage in that sector.
65 All interesting and open question is whether these same mechanisms can amplify other shocks besides
productivity shocks sufficiently that these can produce realistic business cycles.
66 Some recent examples include Andolfatto (1996), Merz (1995), Gomes, Greenwood and Rebelo
(1997) and den Haan, Ramey and Watson (1997).
995
these lines o f inquiry involve enriching the RBC model in ways that seemed virtually
impossible a decade ago.
While we think that economists may have prematurely dismissed the idea that the
business cycle m a y originate from real causes, we also think that m a n y o f the lessons
drawn from current and future RBC research are likely to be independent o f the main
source o f business fluctuations. This is one important reason why the RBC literature
has been a positive technology shock to macroeconomics.
j_d[Cv(L)]1-G
l
l-or
if
o > 0,
if
o = 1.
cr ~e 1,
(A.1)
u(C,L) =
It is easy to verify two properties o f these specifications. First, i f agents have a budget
constraint for goods and leisure o f the form e + wL <. w, where w is the real wage
rate and 1 is the time endowment, then there is invariance o f L to the level o f w 68.
67 Another possibility is that utility depends on leisure in efficiency units, i.e., on leisure augmented
by technological progress (L~Xt). In this case it is sufficient to assume that u(C, LX) is homogeneous,
of class C2, and concave. The dependency of utility on leisure measured in efficiency units can be
justified by introducing home production into the model. See Greenwood, Rogerson and Wright (1995,
pp. 161-162) for a discussion.
68 This invariance extends to a setting where the budget constraint includes nonwage income which
grows at the same rate as the real wage.
996
Second, u s i n g U H o p i t a l ' s rule, the s e c o n d case is the limiting expression o f the first
as a --+ 1. We require that utility be sufficiently differentiable as well as c o n c a v e and
increasing in c o n s u m p t i o n and leisure; this implies restrictions that m u s t be p l a c e d on
v w h i c h d e p e n d on the value o f a 69. Differentiability allows us to characterize efficient
allocations u s i n g variational methods. W h e n c o m b i n e d w i t h convexity o f the constraint
set, c o n c a v i t y o f preferences insures that the s o l u t i o n to the planner's p r o b l e m is
unique, w h e n e v e r lifetime utility ( U ) is finite 7. Since, as we will see shortly, the
c o m p e t i t i v e e q u i l i b r i u m under rational expectations c o i n c i d e s with the solution to the
planner's p r o b l e m , this guarantees that the c o m p e t i t i v e equilibrium is also unique.
T e c h n o l o g y : The p r o d u c t i o n function F ( . ) is also t w i c e continuously differentiable,
concave and h o m o g e n e o u s o f degree one. Constant returns to scale implies that the
n u m b e r o f firms in the competitive e q u i l i b r i u m is undetermined. W i t h increasing
returns to scale a competitive e q u i l i b r i u m does not exist because it w o u l d entail
negative profits for all firms 71. In contrast, with d e c r e a s i n g returns to scale we w o u l d
see an infinite n u m b e r o f infinitesimal firms w h o s e total output w o u l d be infinite 72.
Alternatively, firms w o u l d earn e c o n o m i c profits if, for some reason, entry were
limited.
We a s s u m e that F ( . ) satisfies the f o l l o w i n g l i m i t i n g conditions, often referred to as
Inada conditions 73:
lim DIF(K,N)
K --~ c,o
= 0,
lim DtF(K,N)
K-+O
= cxz.
These c o n d i t i o n s ensure the existence o f a steady state in which the level o f capital is
strictly positive. One can also show that they i m p l y that labor is essential in production:
F ( K , O) = O.
69 More specifically, we assume that the fimctions vi are twice continuously differentiable. If a - 1,
then concavity requires that the function log(v) must be increasing and concave. If a is not equal to 1,
then v 1-a must be increasing and concave if a < 1 and decreasing and convex if a > 1. In addition we
need - a v ( L ) v"(L) > (1 - 2a)[v'(L)] 2 to assure the overall concavity of u.
70 Whenever there is one path that yields infinite utility it is always possible to construct other paths (in
fact a continuum of paths) that also yield infinite utility. Thus, to ensure that there is only one solution
to the planner's problem we need to constrain the discount factor so that life-time utility, U, is finite.
The requirement (by I a < 1) involves the interaction of preferences and technology. See Alvarez and
Stokey (1999) for a discussion of this type of conditions.
71 See Hornstein (1993), Rotemberg and Woodford (1995), and Chatterjee and Cooper (1993) for a
discussion of models that move away from perfect competition and incorporate increasing returns to
scale.
72 Suppose, for example, that the production function is Cobb-Douglas and that there is a stock of
capital K and a number of labor hours N which will be divided equally among n fu-ms. Total production
will be given by Y = nA(K/n) a~(N/n) a2 - AKa~Na2n 1 a~ a2. With decreasing returns to scale a I + a 2 < 1
and lim, _~~ Y = 0<~.
73 As ill the main text we use the notation DiF(.) to refer to the partial derivative of F(-) with respect
to its ith argument. We use DF(.) to refer to the total derivative of a fimction of a single variable.
997
Let us consider first the case in which allocation decisions are made by a benevolent
planner who maximizes the welfare of the representative agent. The solution to this
problem will be a symmetric Pareto-optimum in which all agents receive the same
consumption and leisure allocations.
The stationary economy: In the steady state of a deterministic version of this
economy Y, C, I, and K all grow at rate ~, i.e., the model captures the Kaldor growth
facts. This suggests that it is useful to write the planner's problem for this economy
in terms of variables that are constant in the steady state: y = Y/X, c = C/X, i = I / X ,
k = K/X. Using these stationary variables the planner's problem is given by
oo
max Eo ~
f l u ( c , 1 - Nt)
(A.2)
t-0
subject to:
Yt
= AtF(k,,Nt),
(A.3)
Yt
= ct + it,
(A.4)
~/kt + 1 = it + (1 - b)k,,
(A.5)
k0
(A.6)
> 0,
where fi =_ b y ~-.
In a deterministic environment the solution to the problem of maximizing Equation (A.2) subject to conditions (A.3)-(A.5) would be a sequence of consumption,
labor supply and capital accumulation decisions: {ct}t=0,~ {N~}t=0,ooand {]~t}t~l.~ These
decisions could be made at time zero, since no relevant information is revealed later
on. In contrast, in a stochastic economy agents learn over time the realizations of
the random shocks that affect their environment. It would be inefficient to ignore this
information that will be available later on and cast in stone the consumption and leisure
decisions at time zero. For this reason, the solution to the utility maximization problem
is a set of contingency rules, which specify how much to consume and work at each
point in time as a function of the state of the economy in that period. Since the state
of the economy can be, at any point in time summarized by two variables, the value
of At, which influences current output and helps predict future productivity, and the
value of the stock of capital. Thus contingency rules take the form c = c ( k , A ) and
N = N(k,A).
D y n a m i c programming: To use this approach, we write the planner's problem in
recursive form as
V ( k , A ) = max{u(c, 1 - N ) + f l E V ( k ' , A ' ) } ,
(A.7)
c,N,k I
subject to:
c + y k ' - (1 - 6)k = A F ( k , N ) .
(A.8)
where we use primes ( ) to denote the value of a variable in the next period. The
value function V ( k , A ) represents the expected life-time utility of the representative
998
(A.9)
(A. 10)
V(k',A')-
dN
d~-
dk'
dk"
Using the same logic as in the derivation of the "envelope theorem" in demand theory,
this equation can be greatly simplified by using the first-order conditions (A.9) and
(A.10) to set the two bracketed terms equal to zero. Intuitively, given that the values
of N and U were optimally chosen, there are zero net benefits from the adjustments
in these quantities that will arise from a change k. Thus D~ V(k,A) can be simplified
to
D1V(k',A') = )~ [A'D1F(U,N') + (1 - 6)].
(A.11)
Finding the decision rules: Conditions (A.9) and (A. 10) can be used to solve for c
and N as a function of )t, k and A. These functions are not quite the decision rules for
999
consumption and labor, since they depend on ,~ which we have not yet determined. We
specify the resulting functions as N = N(k, ).,A) and c = ~(k, )~,A). To find the decision
rule for capital, we proceed as follows. Using N = N(k, )~,A) and c = ~(k, X, A), we
can express the optimization conditions as a first-order system o f nonlinear stochastic
difference equations in ;. and k:
)~7
(A.12)
A*DIF(k*,N*)+(1-6)
= 7/3,
A*D2F(k*, N*)D1 u(c*, 1 - N*) = D2u(c*, 1 - N*),
(A.13)
7k*
(A.15)
= A * F ( k * , N * ) - c* + (1 - 6 ) k * .
(A.14)
We use an asterisk to denote the steady-state values o f the different variables. This
system of equations is recursive. Equation (A.13) determines the value of k*/N*;
recall that F is homogeneous o f degree one and thus D1F is homogeneous of degree
zero implying that DlF(k*, N*) = D1F(k*/N*, 1). Equations (A.14) and (A. 15)jointly
determine c* and N*. We will return below to discussing the nature of the steady state
in the competitive economy.
1000
and the stock of capital and make three inter-related decisions: how much labor to
supply (Ns), how much capital to accumulate (k~), and how much to consume (c). In
this decentralization scheme, households have to take into account the law of motion
for the wage rate (w) and for the rental price of capital (R). Both of these prices are
a function of the state of the economy, as summarized by the productivity level A and
the aggregate capital stock k:
w = w(k,A),
(A.16)
R = R(k,A).
(A.17)
To forecast these prices, agents have to know the functions w and R and the law of
motion for A and k. The variable A evolves according to H ( A ' , A ) , while the law of
motion for the aggregate capital stock will be described as
k'=g(k,A).
The H o u s e h o l d Problem. With these preliminaries in place we can now write the
subject to:
c + yk~ = w(k,A)N~. + (1 + R ( k , A ) - b)k~ + zc.
(A.18)
where v is the value function of the household and :v denotes the firms' profits, which,
as we will see in a moment, are always equal to zero.
It is useful to define the real interest rate as the rental price of capital net of
depreciation:
r ( k , A ) = R ( k , A ) - 6.
= c(ks, k,A),
Ns = N(ks, k,A).
(A.19)
(A.20)
Ch. 14:
1001
The Firm "s Problem. The firms in this economy solve a static problem. They have
to decide how much capital and labor to hire in the spot competitive markets for both
of these factors:
max
k,t , Nd
Jr =
AF(kd, Nd)
wNd
Rkd.
(A.21)
(A.22)
Given that the production function exhibits constant returns to scale, profits will always
be equal to zero:
= AF(kd, Nd) - AD2F(kd, Nd)Nd -- ADIF(kd, Nd)kd = O.
Market Clearing. There are three markets in this economy: spot markets for capital,
labor and output. By Walras's law if two of these markets are in equilibrium the third
market will also have to be in equilibrium. Thus we can state the equilibrium conditions
limiting ourselves to the factor markets:
kd =k,=k,
Nd=Ns.
To ensure that this is a competitive equilibrium under rational expectations, the law
of motion conjectured by households for the competitive equilibrium has to coincide
with the actual aggregate law of motion for this variable:
ks(k, k , A ) = g(k,A).
The steady state in the market economy: If we treat N as fixed for the moment, we
can interpret Equation (A. 13) as equating the long run demand and supply for capital.
The real rate of return to capital in a decentralized version of this economy is given
by r = A D 1 F ( k , N ) - 6. This can be seen as a demand schedule; given the value of
r (and the value of N) it tells us the value of k that the economy would choose. The
long run supply of capital is given by r = y/fi - 1 and is thus perfectly elastic: the
capital stock of the economy always adjusts so that the steady-state real interest rate
is r = y / f i - 1 .
A.4. The welfare theorems
To show heuristically the connection between the competitive equilibrium and the
Pareto Optimum we can now compare the first-order conditions of the competitive
equilibrium with those of the planner's problem to show that they coincide. Replacing
1002
Equations (A.21) and (A.22) in (A. 18) and exploring the fact that F(.) is homogeneous
of degree one, we obtain a resource constraint that is equivalent to the one implied
by Equations (A.3)-(A.5). Making use of Equations (A.21) and (A.22) it can also
be readily shown that (A.19)-(A.20) are equivalent to (A.9)-(A.11). Notice that the
assumption o f rational expectations is crucial for this comparison. The equivalence
between the conditions that characterize the two problems underlies the two welfare
theorems that apply to this economy: the competitive equilibrium is Pareto Optimal
and a Pareto Optimal allocation can be decentralized as a competitive equilibrium.
The fact that the competitive equilibrium can be solved as a solution to a planning
problem has important technical implications. Since the planners problem involves
maximizing a continuous function defined over a compact set we know that a solution
to the problem exists. Furthermore, since the planner's problem is strictly concave, its
solution is unique. Thus, the existence and uniqueness o f the competitive equilibrium
can then be established by exploring its equivalence to the planner's problem.
There are many instances in which we may want to explore economies where the first
welfare theorem does not hold. Examples include economies with distortionary taxes,
externalities, or monopolistic competition. Rarely can the competitive equilibrium for
these economies be mapped into a concave planning problem TM. We can still linearize
the system o f equations that characterizes the competitive equilibrium to explore some
of its properties. However, we no longer have the guarantee that the equilibrium
exists or that it is unique. This is the reason why the multiple equilibrium literature
discussed in Farmer (1993) focuses on economies in which the competitive equilibrium
is suboptimal.
References
Abel, A.B., and J.C. Eberly (1995), "Investment and Q with fixed costs: an empirical analysis", Working
paper (University of Pennsylvania).
Aiyagari, S.R. (1994), "On the contribution of technology shocks to business cycles", Federal Reserve
Bank of Minneapolis Quarterly Review 18:22-34.
Alvarez, E, and N. Stokey (1999), "Dynamic programming with homogeneous fimctions", Journal of
Economic Theory, forthcoming.
Ambler, S., and A. Paquet (1994), "Stochastic depreciation and the business cycle", International
Economic Review 35:101 116.
Andolfatto, D. (1996), "Business cycles and labor market search", American Economic Review 86:
112-132.
Backus, D.K., and P.J. Kehoe (1992), "International evidence on the historical properties of business
cycles", American Economic Review 82:864-88.
Barro, R.J. (1974), "Are government bonds net wealth?", Journal of Political Economy 82:1095-1117.
Barro, R.J., and R.G. King (1984), "Time separable preferences and intertemporal substitution models
of business cycles", Quarterly Journal of Economics 99:812839.
1003
Basu, S., and M.S. Kimball (1997), "Cyclical productivity with unobserved input variation", Working
Paper No. 5915 (NBER).
Basu, S., J.G. Fernald and M.S. Kimball (1997), "Are technology improvements contractionary?",
Working Paper (University of Michigan).
Baxter, M. (1995), "International trade and business cycles", in: G. Grossman and K. Rogoff, eds.,
Handbook of International Economies, vol. 3 (Elsevier Science, Amsterdam) 1801-1864.
Baxter, M., and U.J. Jermann (1999), "Household production and the excess sensitivity of consumption
to current income", American Economic Review, forthcoming.
Baxter, M., and R.G. King (1993), "Fiscal policy in general equilibrium", American Economic Review
83:315-334.
Baxter, M., and R.G. King (1995), "Measuring business cycles: approximate band-pass filters for
economic time series", Working Paper No. 5022 (NBER); Review of Economics and Statistics,
forthcoming.
Baxter, M., and A.C. Stockman (1989), "Business cycles and the-exchange-rate regime: some international
evidence", Journal of Monetary Economics 23:377-400.
Bils, M., and J.O. Cho (1994), "Cyclical factor utilization", Journal of Monetary Economics 33:31%354.
Boldrin, M., and M. Horvath (1995), "Labor contracts and business cycles", Journal of Political Economy
103:972-1004.
Boldrin, M., L.J. Chffstiano and J. Fisher (1995), "Asset pricing lessons for modeling business cycles",
Working Paper (Northwestern University).
Braun, R.A. (1994), "Tax disturbances and real economic activity in the postwar United States", Journal
of Monetary Economics 33:441-462.
Brock, WA., and L.J. Mirman (1972), "Optimal economic growth and uncertainty: the discounted case",
Journal of Economic Theory 4:47%513.
Burns, A.E, and W.C. Mitchell (1946), Measuring Business Cycles (National Bureau of Economic
Research, New York).
Burnside, C., and M. Eichenbaum (1996), "Factor hoarding and the propagation of business cycle
shocks", American Economic Review 86:1154-1174.
Burnside, C., M. Eichenbanm and S.T. Rebelo (1993), "Labor hoarding and the business cycle", Journal
of Political Economy 101:245-273.
Burnside, C., M. Eichenbaum and S.T. Rebelo (1995), "Capital utilization and returns to scale", NBER
Macroeconomics Annual 1995 (MIT Press, Cambridge, MA) 67-110.
Burnside, C., M. Eichenbaum and S.T. Rebelo (1996), "Sectoral Solow residuals", European Economic
Review 40:861-869.
Caballero, R.J., and E. Engel (1994), "Explaining investment dynamics in U.S. manufacturing:
a generalized (S, s) approach", Working Paper (MIT).
Campbell, J.R. (1998), "Entry, exit, embodied technology, and business cycles", Review of Economic
Dynamics 2:371-408.
Campbell, J.Y. (1994), "Inspecting the mechanism: an analytical approach to the stochastic growth
model", Journal of Monetary Economics 33:463-506.
Campbell, J.Y. (1999), "Asset prices, consumption, and the business cycle", ch. 19, this Handbook.
Cass, D. (1965), "Optimum growth in an aggregative model of capital accumulation", Review of
Economic Studies 32:223-240.
Chaff, V.V., P.J. Kehoe and E.R. McGrattan (1996), "Sticky price models of the business cycle: can the
contract multiplier solve the persistence problem?", Working Paper No. 5809 (NBER).
Chatterjee, S., and R. Cooper (1993), "Entry and exit, product variety and the business cycle", Working
Paper No. 4562 (NBER).
Cho, J.O., and T.E Cooley (1994), "Employment and hours over the business cycle", Journal of Economic
Dynamics and Control 18:411-432.
Cho, J.O., and T.E Cooley (1995), "The business cycle with nominal contracts", Economic Theory,
13-33.
1004
Cho, J.O., and R. Rogerson (1988), "Family labor supply and aggregate fluctuations", Journal of Monetary
Economics 21:233-245.
Christiano, L.J. (1990), "Solving the stochastic growth model by linear quadratic approximation and by
value function iteration", Journal of Business and Economic Statistics, 8:23-26.
Christiano, L.J., and M. Eichenbaum (1992a), "Current real business cycle theory and aggregate labor
market fluctuations", American Economic Review 82:430-450.
Christiano, L.J., and M. Eichenbaum (1992b), "Liquidity effects and the monetary transmission
mechanism", American Economic Review 82:346-353.
Christiano, L.J., and J. Fisher (1995), "Tobin's q and asset returns: implications for business cycle
analysis", Working Paper (Northwestern University).
Cogley, T., and J.M. Nason (1995), "Output dynamics in real business cycle models", American Economic
Review 85:49~511.
Cooley, T.E, and G.D. Hansen (1989), "The inflation tax in a real business cycle model", American
Economic Review 79:733-748.
Cooley, T.E, and L.E. Ohanian ( 1991), "The cyclical behavior of prices", Journal of Monetary Economics
28:25-60.
Cooley, T.E, and L.E. Ohanian (1997), "Post-war British economic growth and the legacy of Keynes",
Journal of Political Economy 105:439-472.
Cooley, T.E, and E.C. Prescott (1995), "Economic growth and business cycles", in: T.E Cooley, ed.,
Frontiers of Business Cycle Research (Princeton University Press, Princeton, NJ) 1-38.
Correia, I., J. Neves and S.T. Rebelo (1992), "Business cycles 1850-1950: new facts about old data",
European Economic Review 36:459-467.
Danthine, J.-P., and J.B. Donaldson (1985), "A note on the effects of capital income taxation on the
dynamics of a competitive economy", Journal of Public Economics 28:255-265.
Danthine, J.-P., and J.B. Donaldson (1995), "Non-Walrasian economies", in: T.E Cooley, ed., Frontiers
of Business Cycle Research (Princeton University Press, Princeton, NJ) 217-242.
Danthine, J.-P., J.B. Donaldson and R. Mehra (1989), "On some computational aspects of real business
cycle theory", Journal of Economic Dynamics and Control 13:449-470.
den Haan, W.J. (1993), "Interaction between heterogeneous agents in intertemporal stochastic models: a
solution algorithm and applications", Working Paper (University of California at San Diego).
den Haan, W.J., G. Ramey and J. Watson (1997), "Job destruction and propagation of shocks", Working
Paper (University of California at San Diego).
Devereux, M., A.W. Gregory and G.W. Smith (1992), "Realistic cross-country consumption correlations
in a two-country, equilibrium, business-cycle model", Journal of International Money and Finance
11:3-16.
Dotsey, M., and C.S. Mao (1992), "How well do linear approximation methods work?", Journal of
Monetary Economics 29:25-38.
Dotsey, M., R.G. King and A.L. Wolman (1996), "State dependent pricing and the general equilibrium
dynamics of money and output", Working Paper (University of Virginia).
Dunlop, J.T. (1938), "The movement of real and money wage rates", Economic Journal 48:413-434.
Dupor, B. (1998), "Aggregation and irrelevance in multisector models", Journal of Monetary Economics,
forthcoming.
Eichenbauna, M. (1991), "Real business cycle theory: wisdom or whimsy?", Journal of Economic
Dynamics and Control 15:607-626.
Evans, C. (1992), "Productivity shocks and real business cycles", Journal of Monetary Economics
29:191-208.
Farmer, R. (1993), The Macroeconomics of Self-Fulfilling Prophecies (MIT Press, Cambridge,/VIA).
Finn, M. (1991), "Energy, price shocks, capacity utilization and business cycle fluctuations", Institute
for Empirical Macroeconomics Discussion Paper 50 (Federal Reserve Bank of Minneapolis).
Friedman, M. (1957), A Theory of the Consumption Function (Princeton University Press, Princeton,
N J).
1005
Fuerst, T. (1992), "Liquidity, loanable funds and real activity", Journal of Monetary Economics 29:3~4.
Gali, J. (1996), "Technology, employment and the business cycle: do technology shocks explain aggregate
fluctuations", Working Paper No. 5721 (NBER).
Gomes, J., J. Greenwood and S.T. Rebelo (1997), "Equilibrium unemployment", Working Paper
(University of Rochester).
Gomme, E, and J. Greenwood (1995), "On the cyclical allocation of risk", Journal of Economic Dynamics
and Control 19:91 124.
Greenwood, J., Z. Hercowitz and G.W. Huffman (1988), "Investment, capacity utilization and the business
cycle, American Economic Review 78:402-417.
Greenwood, J., R. Rogerson and R. Wright (1995), "Household production in real business cycle theory",
in: T.E Cooley, ed., Frontiers of Business Cycle Research (Princeton University Press, Princeton, NJ)
15~174.
Hall, R.E. (1988), "The relation between price and marginal cost in U.S. industry", Journal of Political
Economy 96:921-947.
Hall, R.E. (1999), "Labor-market frictions and employment fluctuations", ch. 17, this Handbook.
Hamilton, J. (1994), Time Series Analysis (Princeton University Press, Princeton, NJ).
Hansen, G.D. (1985), "Indivisible labor and the business cycle", Journal of Monetary Economics
16:309-327.
Hansen, G.D., and R. Wright (1992), "The labor market in real business cycle theory", Federal Reserve
Bank of Mhmeapolis Quarterly Review 16:2-12.
Hodrick, R., and E.C. Prescott (1980), "Post-war business cycles: an empirical investigation", Working
Paper (Carnegie-Mellon University).
Hopenhayn, H., and R. Rogerson (1993), "Job turnover and policy evaluation: a general equilibrium
analysis", Journal of Political Economy 10:915 938.
Hornstein, A. (1993), "Monopolistic competition, increasing returns to scale and the importance of
productivity shocks", Journal of Monetary Economics 31:299-316.
Horvath, M. (1997), "Sectoral shocks and aggregate fluctuations", Working Paper (Stanford University).
Ingram, B., N. Kocherlakota and N. Savin (1997), "Using theory for measurement: an analysis of the
cyclical behavior of home production", Journal of Monetary Economics 40:435-456.
Kaldor, N. (1957), "A model of economic growth", Economic Journal, 67:591-624.
King, R.G. (1991), "Value and capital in the equilibrium business cycle program", in: L. McKenzie and
S. Zamagni, eds., Value and Capital Fifty Years Later (MacMillan, London).
King, R.G. (1995), "Quantitative theory and econometrics", Federal Reserve Bank of Richmond Economic
Quarterly 81:53-105.
King, R.G., and C.I. Plosser (1994), "Real business cycles and the test of the Adelmans", Journal of
Monetary Economics 33:405-438.
King, R.G., and S.T. Rebelo (1993), "Low frequency filtering and real business cycles", Journal of
Economic Dynamics and Control 17:207-231.
King, R.G., and M.W Watson (1996), "Money, prices, interest rates and the business cycle", Review of
Economics and Statistics 78:35-53.
King, R.G., C.I. Plosser and S.T. Rebelo (1988a), "Production, growth and business cycles: I. The basic
neoclassical model", Journal of Monetary Economics 21:195-232.
King, R.G., C.I. Plosser and S.T. Rebelo (1988b), "Production, growth and business cycles: I. The basic
neoclassical model, Technical Appendix", Working paper (University of Rochester).
King, R.G., C.I. Plosser, J.H. Stock and M.W Watson (1991), "Stochastic trends and economic
fluctuations", American Economic Review 81:819-840.
Klein, L., and R. Kosobud (1961), "Some econometrics of growth: great ratios of economics", Quarterly
Journal of Economics 75:173-198.
Kocherlakota, N. (1996), "The equity premium: it's still a puzzle", Journal of Economic Literature 34:
42-71.
Koopmans, T. (1947), "Measurement without theory", Review of Economics and Statistics 29:161 172.
1006
Koopmans, T. (1965), "On the concept of optimal economic growth", in: The Econometric Approach to
Development Planning (Rand-McNally, Chicago, IL).
Krnsell, E, and A.A. Smith (1998), "Income and wealth heterogeneity in the macroeconomy", Journal
of Political Economy 106:876-896.
Kuznets, S. (1973), "Modern economic growth: findings and reflections", American Economic Review
63:247-258.
Kydland, EE., and E.C. Prescott (1982), "Time to build and aggregate fluctuations", Econometrica
50:1345 1370.
Kydland, EE., and E.C. Prescott (1988), "The workweek of capital and its cyclical implications", Journal
of Monetary Economics 21:343-360.
Kydland, EE., and E.C. Prescott (1990), "Business cycles: real facts and a monetary myth", Federal
Reserve Bank of Minneapolis Quarterly Review 14:3-18.
Kydland, EE., and E.C. Prescott (1991), "The econometrics of the general equilibrium approach to the
business cycles", Scandinavian Journal of Economics 93:161-178.
Long, J., and C.I. Plosser (1983), "Real business cycles", Journal of Political Economy 91:39-69.
Lucas, R. (1976), "Econometric policy evaluation: a critique", Carnegie-Rochester Conference Series
on Public Policy 1:19-46.
Lucas, R. (1977), "Understanding business cycles", Carnegie-Rochester Conference Series on Public
Policy 5:7-29.
Lucas, R. (1980), "Methods and problems in business cycle theory", Journal of Money, Credit and
Banking 12:696-715.
MaCurdy, T.E. (1981), "An empirical model of the labor supply in a life cycle setting", Journal of
Political Economy 89:1059-1085.
Mankiw, N.G. (1989), "Real business cycles: a new Keynesian perspective", Journal of Economic
Perspectives 3: 79-90.
McGrattan, E.R. (1994), "The macroeconomic effects of distortionary taxation", Journal of Monetary
Economics 33:573-601.
Mehra, R., and E.C. Prescott (1985), "The equity premium: a puzzle", Journal of Monetary Economics
15:145-161.
Merz, M. (1995), "Search in the labor market and the real business cycle", Journal of Monetary
Economics 36:269-300.
Nelson, C.R., and C.I. Plosser (1982), "Trends and random walks in macroeconomic time series: some
evidence and implications", Journal of Monetary Economics 10:139-167.
Neusser, K. (1991), "Testing the long run implications of the neoclassical growth model", Journal of
Monetary Economics 27:3-37.
Pencavel, J. (1986), "Labor supply of men: a survey", in: O. Ashenfelter and R. Layard, eds., Handbook
of Labor Economics (North-Holland, Amsterdam) 3-102.
Phelps, E. (1966), Golden Rules of Economic Growth (Norton, NY).
Plosser, C.I. (1989), "Understanding real business cycles", Journal of Economic Perspectives 3:51-77.
Prescott, E.C. (1986), "Theory ahead of business-cycle measurement", Federal Reserve Bank of
Minneapolis Quarterly Review 10:9-22.
Radner, R. (1966), "Optimal growth in a linear-logarithmic economy", International Economic Review
7:1-33.
Rebelo, S.T., and D. Xie (1999), "On the optimality of interest rate smoothing", Journal of Monetary
Economics 43:263-282.
Rios-Rull, J. (1994), "On the quantitative importance of market completeness", Journal of Monetary
Economics 34:463-496.
Rogerson, R. (1988), "Indivisible labor, lotteries and equilibrium", Journal of Monetary Economics
21:3 16.
Rogerson, R., and R. Wright (1988), "Involuntary unemployment in economies with efficient risk
sharing", Journal of Monetary Economics 22:501-515.
1007
Chapter 15
Contents
Abstract
Keywords
1. Introduction
2. A n empirical guide to price and wage setting in m a r k e t e c o n o m i e s
2.1. General observations about wage and price setting
2.1.1. Wage setting
2.1.2. Price setting
2.2. Individual firm and worker evidence
2.2.1. Direct evidence on wage setting
2.2.2. Indirect evidence on wage setting
2.2.3. Direct evidence on price setting
2.3. Summary of findings about price and wage setting
3. Market-clearing and expected-market-clearing approaches
3.1. Market-clearing models
3.1.1. Empirical tests
3.2. Expected-market-clearing models
4. Staggered contracts m o d e l s
4.1. A simple price setting model
4.2. More general staggered wage- and price-setting models
4.2.1. Fixed duration models
4.2.2. Random duration models
4.2.3. State-dependent duration models
5. Bolstering the theoretical foundations o f staggered contracts m o d e l s
5.1. Deriving the optimal price: the role of market power
5.2. Towards dynamic optimizing models of staggered price and wage setting
5.3. Staggered price and wage setting in general equilibrium models
1010
1010
1011
1013
1013
1013
1014
1015
1015
1017
1018
1020
1021
1021
1022
1024
1027
1027
1029
1029
1030
1031
1032
1033
1033
1035
* This research was supported by the Center for Economic Policy Research at Stanford University.
I wish to thank VV Chari, Christopher Erceg, Robert Hall, Ellen McGrattan, Akila Weerapana, and
Michael Woodford for useful comments and assistance.
Handbook of Macroeconomics, Volume 1, Edited by J.B. Taylor and M. Woodford
1999 Elsevier Science B.V. All rights reserved
1009
1010
5.4. Explanationsof why price and wage setting is staggered
5.5. Indexingand optimal contractlength
6. Persistence puzzles and possible resolutions
6.1. Inflationpersistence
6.2. Real output persistence
6.3. Changes in stabilityand nominal rigidity overtime
7. Concluding remarks on policy applications and future research
References
JB. Taylor
1037
1038
1038
1039
1040
1042
1042
1044
Abstract
This chapter reviews the role of temporary price and wage rigidities in explaining of
the dynamic relationship between money, real output, and inflation. The key properties
to be explained are that monetary shocks have persistent, but not permanent, effects
on real output, and that the correlation between current output and inflation is positive
for leads of inflation and negative for lags of inflation.
The paper begins with a short empirical guide to price- and wage-setting behavior
in market economies. It then compares alternative price- and wage-setting theories and
argues that staggered contracts models continue to provide the most satisfactory match
with the key macroeconomic facts. It then examines the microeconomic foundations
of staggered contracts models and reviews some of their extensions and applications.
Research in this area has been very active in the 1990s with a remarkable number
of studies using, estimating , or testing models o f staggered price and wage setting.
A new generation of econometric models incorporating staggered price and wage
setting with rational expectations has been built. Researchers have begun to incorporate
staggered wage and price setting into real business cycle models. Close links have been
discovered between the parameters of people's utility functions and the parameters of
staggered price- and wage-setting equations. There is now a debate about whether
standard calibrations of utility functions prevent staggered price models, at least those
with frequent price changes, from explaining long persistence of real output.
A theme of the paper is that the advent of rational expectations in the 1970s led to
models of price and wage rigidities which were more amenable to empirical testing
than earlier models, and this is one reason for the recent controversies and debates.
There is much to be discovered from these debates and from the future research they
stimulate.
Keywords
staggered price and wage setting, staggered contracts model, contract multiplier,
money, monetary policy, monopolistic competition, market clearing, expected market
clearing, time-dependent pricing, state-dependent pricing
JEL classification: E32, El0
1011
1. Introduction
Why does a change in the money supply cause real output and employment to change
in the short run, but not in the long run? This is one of the oldest questions in
macroeconomics, yet it persists as both the most difficult and the most practical of
all. From David Hume in the 18th century to Milton Friedman in the 20th, economists
have had a common answer: there are temporary price and wage rigidities in the
economy. In other words, in the short run, price and/or wage levels do not change as
much as the money supply changes. Thus, if the money supply increases, then real
money balances rise, stimulating production and employment. As described by David
Hume more than 200 years ago, "by degree the price rises, first of one commodity, then
of another", or as stated more recently by Milton Friedman "prices are sticky" [see
Rotwein (1955, p. 38) and Friedman (1982, p. 64)]. Except in unusual circumstances,
it takes time for price and wage levels to fully adjust; as prices and wages gradually
rise, real money balances return to their original level and in the long run the real
economy is unaffected.
The purpose of this chapter is to review the current state of knowledge about the
nature of price and wage rigidities and their ability to explain the dynamic relationship
between money, real output, and inflation. I survey recent research, classify the major
facts and models, and draw some conclusions and suggestions for future research.
Because both price rigidities and wage rigidities are important for macroeconomic
dynamics, both types of rigidities are reviewed.
The dynamic stochastic properties of money, output, and inflation that we would
like the theories to explain have been carefully documented using modern time series
techniques in the chapters by Stock and Watson (1999) and by Christiano, Eichenbaum
and Evans (1999) in this Handbook. In addition to the property that money shocks have
a short-run impact on output and a long-run impact on inflation, three other important
properties of the money, real output, and inflation relation have attracted attention.
These properties are (1) that a monetary shock has a persistent effect on real output a propagation effect that lasts well beyond the initial impact effect of a monetary shock,
(2) that there is a positive correlation between real output and future inflation, and
(3) that there is a negative correlation between inflation and future real output. [See
Stock and Watson (1999), Table 2, lines 41 and 54, and Christiano, Eichenbaum and
Evans (1999), Figures 2 and 3, which compare the impact of alternative measures
of money shocks.] Properties (2) and (3) can also be characterized using vector
autoregressions and the concept of Granger-causality: property (2) is that real output
Granger-causes inflation positively and property (3) is that inflation Granger-causes
real output negatively. This "reverse dynamic" cross correlation was shown to hold
in the USA and other countries in Taylor (1980a, 1986). The ability to explain such
robust findings represents an important measure of success for any theory of price and
wage rigidities.
It was in the early 1970s that the rational expectations revolution in macroeconomics
first began to take hold. Of course, price and wage rigidities were a big part of
1012
J.B. Taylor
macroeconomics before the early 1970s; indeed, they were central to both Keynesian
and monetarist ideas. However, by forcing macroeconomists to think in economywide (general equilibrium) terms, the rational expectations revolution fundamentally
changed how economists model price and wage rigidities. It was at this time that the
staggered contracts model was first proposed. This model emphasizes the overlapping
of individual prices and wages due to staggered price and wage setting. It could
explain not only why changes in money impact real output, but also why the other
properties mentioned above are found to hold in many countries. Fron~ the start
the staggered contract model was explicit enough that testable hypotheses could be
formulated and, as a result, the model underwent extensive empirical and theoretical
scrutiny, generating much debate. This explicimess was necessitated, in my view, by
the discipline of the rational expectations approach to modeling.
During the 1980s, researchers extended and modified models of staggered price and
wage setting, fitted them to data from many different countries, and began to give them
a more specific theoretical footing. Models of staggered price and wage setting were
frequently used as a source of monetary nonneutrality in rational expectations policy
studies of monetary policy rules. By using these theories of price and wage rigidities
for policy analysis and by confronting them with practical tests using real world data both microeconomic and macroeconomic - much has been learned about the process
of wage and price formation in market economies.
In the 1990s research on staggered price and wage setting has shown no signs
of slowing down. Modifications motivated by a need to explain inflation dynamics
more accurately or to reduce model complexity have been introduced and used in
econometric models. A new generation of econometric models incorporating price and
wage rigidities with rational expectations has been put in place for policy analysis at
many central banks including the Federal Reserve Board. Perhaps even more exciting,
researchers have begun to incorporate theories of staggered wage and price setting into
what would otherwise be real business cycle models. This research has uncovered close
links between the parameters of people's utility functions or firm's production functions
and the parameters of staggered price and wage setting equations. The research has also
uncovered a puzzle: in some models, standard calibration of utility function parameters
indicates substantially less persistence of real output than had been found in earlier
estimates of staggered contract models that had not made this formal link to individual
utility maximization. Several ways to resolve this puzzle have been proposed and
are discussed here. On balance, the staggered wage and price setting models still
seem to be consistent with the broad features o f the data, but there is much to be
discovered in future research. With the price and wage setting theories so amenable
to empirical testing, it is not surprising that controversies and debates about price and
wage rigidities continue.
I begin the chapter with a short guide to price and wage setting behavior in market
economies based on direct and indirect observations. I then review alternative price
and wage setting models and show how models based on staggered price and wage
setting match the facts mentioned above, a match that explains, in my view, why
1013
the models have proved useful to macroeconomists doing applied econometrics and
policy work. I then go on to explore the microeconomic foundations for these models,
discussing several intriguing puzzles and their possible resolutions. Finally, I discuss
several historical and policy applications of econometric models based on staggered
price and wage setting.
As research on staggered price and wage setting has developed, several surveys
of research on price and wage rigidities have appeared, including Taylor (1985),
Btanchard (1990), Roberts (1995), and Goodfriend and King (1997), as well as several
good graduate textbook treatments including Blanchard and Fischer (1989, Chapter 8)
and Romer (1996, Chapter 6). In this survey I try to focus on topics that are either more
recent or that have not been emphasized in these other surveys and reviews. I also try
to trace out the gradual evolution of the staggered contracts model over time.
Before discussing this quantitative evidence, I think it is useful to first review some
general observations about wage and price setting which are simply part of our
everyday experience as participants in market economies. One might criticize these
types of observations as arbitrary and subjective, but in fact such informal case studies
have informed the theoretical research in this area much as David Hume's casual
observations informed his theory.
2.1.1. Wage setting
One observes a great variety of ways in which wages are determined through the
interaction of workers and firms in a market economy. These mechanisms evolve over
time especially when there are large changes in the economic environment such as a
change in the tax law or a rise in the average rate of inflation. There are, however,
some important common features of these wage setting arrangements.
For most workers employed in medium to large sized firms, wages (including
benefits) are normally adjusted at rather long discrete intervals, most commonly once
per year. The wage adjustment is typically associated with an extensive performance
and salary review. A large fraction of the wage payment is usually stated in a fixed
amount of dollars (or other currency) per unit of time (hour, week or month), but
1014
J.B. Taylor
overtime pay, bonuses, profit sharing, piece rates, and other incentive arrangements
are not uncommon for part of workers' pay. Indexing of wages to macroeconomic
variables, such as the inflation rate, occurs, but is rare in wage setting arrangements
of one year or less.
It is important to emphasize that setting nominal wages at a fixed level for more than
several quarters and usually for as long as a year appears to be just as prevalent for
workers who are not in unions, or who do not have formal employment contracts, as
for union workers with employment contracts. It is a mistake to limit studies o f how
wages are set to formal employment contracts. As we will see below there is more
quantitative information about indexing and contract length for workers and firms in
the union sectors, but in the USA only a relatively small fraction of workers are in
unions, so the evidence is less relevant.
Throughout the economy wage setting is not usually synchronized in any one period.
Rather wage adjustments occur at different times for different firms throughout the
year, much as the dates for firms' fiscal years vary throughout the year. There are
exceptions to this nonsynchronization. In Japan, for example, there has been a Shunto,
or spring offensive, during which wages at most large firms are set.
Wage decisions are clearly influenced by wages paid to other workers in a
community or in similar occupations. When hiring new engineers or business school
graduates, firms must consider the prevailing wage in the market for similarly educated
workers when designing a wage package. Pattern bargaining among labor unions where
the wage set at one union becomes a strong influence at another union is common.
2.1.2. Price setting
There are even greater varieties of price setting than of wage setting. Prices change
continuously in auction markets for commodities and financial instruments. The idea
of price rigidities seems ludicrous when applied to such markets.
On the other hand, prices change infrequently in posted-price customer markets for
goods and services. [Okun (1981) used the term "customer markets" to distinguish
these types of markets from "auction markets".] Prices of final goods in customer
markets seem to be more responsive to changes in the costs of intermediate inputs to
production than they are to changes in demand. In other words, changes in markups
seem less of a source of price change than changes in costs. But even within socalled customer markets, prices - and markups of prices over costs - can and do
change rapidly. For example, the price of airline tickets changes on a day to day basis
depending on estimates of demand. Some airline tickets are now even auctioned on
the Internet.
Like wages, the prices of goods and services appear to be staggered over time and
set taking the prevailing price of competing producers (or monopolistically competing
producers) into account. News accounts suggest, for example, that a decision to
change the price of Big Macs is closely watched by Burger King and other fast food
companies. Case studies of pricing at large supermarket chains by Levy, Dutta, Bergen
1015
and Venable (1998) show that obtaining information about the prices of competitors
is the first step in deciding on how much to change prices; they find that "people are
sent on a weekly basis to competitors" stores to gather and record (usually by hand)
their price ... ". Clearly some degree of non-synchronized or staggered price setting
is necessary if such surveys of competitors' prices are to be useful. The policy o f
matching a competitor's price is common among discount stores. The prevailing price
is even an important consideration in price setting in auction markets. For example,
knowing the price of recently auctioned items influences the reservation price of buyers
and the minimum price of sellers in auctions for art and other rare goods.
Because surveys of competitors and other steps in the process of changing prices
(such as changing and verifying tag changes) are costly to undertake, Levy, Bergen,
Dutta and Venable (1997) find that the costs to a retail store of changing prices are
quite large. Examining several large US supermarket chains, they find that the fixed
cost is about $0.52 per price change; such costs are "fixed" in the sense that they do
not depend on the size of the price change. (Because the cost of changing a menu also
does not depend on the size of the price change, such costs are usually called "menu
costs".) Using a different methodology Slade (1996) finds the fixed costs of a retail
price change to be even larger.
2.2. Individual f i r m and worker evidence
One of the great accomplishments of research on wage and price rigidities in the 1980s
and 1990s is the bolstering of these case studies and casual impressions by quantitative
evidence from thousands of observations of price and wage setting collected at the
firm, worker, or union level. By carefully studying these data one can learn much
about the nature of wage and price rigidities in the USA and other countries. Some of
the conclusions I draw from these data are summarized in Section 2.3 below, but it is
worthwhile to consider in detail several of the individual research studies.
2.2.1. Direct evidence on wage setting
1016
J.B. Taylor
Cecchetti (1984) computed a time series of average duration of wages in the USA
for the large union sector. He found that the average period between wage changes was
slightly less than 2 years (7 quarters) in the 1950s and 1960s when inflation was low,
but then fell to about one year (four quarters) in the 1970s during the high-inflation
period. Hence, there is some evidence that the length of the period between wage
changes is sensitive to the underlying inflation rate. However, even in the years of the
great inflation wages were set at fixed nominal levels for one year on average. Cecchetti
(1984) also found evidence that the amount of indexing increased as the inflation rate
rose from the 1960s to the 1970s.
Of course, quantitative observations on wage setting at the microeconomic level is
not confined to the USA. For example, Fregert and Jonung (1986) provide evidence
from Sweden. Their data show that contract length decreased in Sweden as inflation
and monetary uncertainty rose. However, average contract length never dropped below
one year. They also found that wage setting showed little synchronization until the
1950s when centralized wage bargaining was introduced. Recall that Japan is another
country where much union wage setting occurs at the same time each year.
Lebow, Stockton and Wascher (1995), McLaughlin (1994), and Card and Hyslop
(1997) provide evidence about wage rigidity in the USA using individual wage data
from the Panel Study of Income Dynamics. Because these studies, which reach similar
conclusions, are not restricted to union workers, they are an important addition to
the earlier studies on unions. However, because the wage is only sampled once per
year, these studies do not provide information about whether the average duration of
fixed wages is less than one year. Card and Hyslop (1997), who also examine data
from the US Current Population Survey form 1979 to 1993, report that between 6 and
15 percent of workers (among those who do not change jobs) have their nominal wages
unchanged from one year to the next, suggesting a relatively small number of workers
whose wages change less frequently than once per year. Card and Hyslop (1997) also
report that the frequency of wage adjustment increases with inflation: during the highinflation period of the late 1970s, only 6 to 10 percent of workers with the same job
had unchanged wages from one year to the next, while during the lower-inflation period
of the 1980s the fraction of workers with unchanged wages was close to 15 percent.
This confirms the results observed by Cecchetti (1984) using union wage data in the
USA and Fregert and Jonung (1986) for Sweden. Card and Hyslop (1997) also report
information on downward wage rigidity (again for annual wage changes) and find that
a significant fraction of workers (15 to 20 percent) experience nominal wage reductions
from one year to the next. This suggests that downward wage rigidities are no more
significant for macroeconomics than upward wage rigidities. In fact, few of the models
of price and wage rigidities used empirically in macroeconomic models make use of the
distinction between upward and downward rigidity - in particular asymmetric rigidities
have little relevance for the key empirical correlations mentioned in the introduction.
1017
1018
J.B. Taylor
Carlton (1986, 1989), using actual transactions price data that originated with Stigler
and Kindahl (1970), has documented in great detail the extent of price rigidity for a
wide variety of products in the USA. [Carlton (1989) also provides a useful summary
of studies of price rigidity prior to Stigler and Kindahl (1970). One of the most
interesting early findings, attributed to EC. Mills, is that the empirical frequency
distribution of price change is U-shaped: goods with low-frequency and high-frequency
price changes are more common than goods with medium-frequency price changes].
For most product groups, Carlton (1989) found that the time between adjustment of
transactions prices seems remarkably long: ranging from about 1 years for steel,
cement, and chemicals to about g1 year for plywood and nonferrous metals. Carlton
found little evidence for a fixed cost of adjusting prices, but observed a positive
correlation between industry concentration and price rigidity.
Cecchetti (1986) collected data on US magazine prices and also found price changes
to be remarkably infrequent. In the 1950s the average length of time between price
changes was 7 years; in the high-inflation years in the 1970s the time between price
change was much lower at 3 years, but still remarkably infrequent given the changes in
demand and costs during such long periods. Similarly, Kashyap (1995) found that mailorder companies keep their catalog prices fixed for periods ranging from six months
to two years.
Surveys of pricing policies at firms by Blinder (1994) and by Blinder et al.
(1998) confirm the findings of Carlton and Cecchetti. Blinder (1994) found the mode
frequency of price adjustment for firms in the survey to be one year. About 40 percent
of firms tend to change their prices once per year, while only 10 percent change
prices more frequently than once per year; the remaining 50 percent leave their prices
unchanged for more than a year. Blinder (1994) also reports that the mean lagged
response of price changes to changes in demand or costs is about three months and is
1019
1020
JB. Taylor
1021
be a good approximation to this more complex world, but most likely some degree
of heterogeneity will be required to describe reality accurately. It is worth noting
that the aggregate time series studies of Christiano, Eichenbaum and Evans (1999)
and Stock and Watson (1999) also show considerable heterogeneity in price and
wage setting: for example, commodity prices seem to respond more quickly to
money shocks than general price indices or than wages.
(3) Neither price setting nor wage setting is synchronized. Except for cases of national
union bargaining, wage setting is staggered through time, whether one looks at
union or non-union workers. Price setting is also staggered through time. These
facts, which are apparent in casual observation, are confirmed in studies of wages,
retail prices, and industrial prices.
(4) The frequency of wage and price changes depends on the average rate of inflation.
Data from union workers and non-union workers in different countries show that
the frequency of wage setting increases with the average rate of inflation. Similarly,
prices at small businesses, industrial prices, and even the prices of products like
magazines are adjusted more quickly when the rate of inflation is higher. This
dependency of price and wage setting on events in the economy is one of the
more robust empirical findings in the studies reviewed here. However, it should
be emphasized that for the range of inflation rates observed in the developed
economies in the 1970s the average duration of wages and prices remained high.
Moreover, we have no empirical evidence that anything other than a change in the
inflation rate would change the frequency of price and wage adjustment, though
one would expect legal or technological changes that increase the cost of changing
prices would reduce price adjustment frequency. For a given average inflation rate,
constant frequencies of price adjustment may not be a bad assumption to make in
an empirical or policy model.
t022
JB. Taylor
of prices and thereby explain the impact of a change in money on the economy a la
Hume and Friedman, it is necessary to posit some lack of information by people
which slows down the adjustment of prices endogenously. Imperfect information about
whether a change in the money supply or some other factor is the source of the shift
in a firm's demand curve, for example, may cause prices to change by less than they
would if the firms were fully informed. Thus, prices become temporarily rigid or sticky
in an otherwise perfectly flexible price model.
Robert Lucas's (1972) path-breaking "expectations and the neutrality of money"
paper - which introduced rational expectations and dynamic optimization into
economy-wide monetary models - takes this approach. McCallum (1984) provides
a good review of this model and Lucas (1996) provides a recent survey of the
general imperfect-information approach in comparison with other approaches. The
Lucas (1972) model is an overlapping-generations model with rational expectations
in which trading takes place in decentralized markets. The overlapping-generations
model assumption is used to create an endogenous demand for money. The old, who
are neither able to store goods they produce when young nor produce goods when old,
use money to pay the young for the goods. When the old exchange money for goods a
price is determined. Under certain uniqueness assumptions, a single equilibrium price
level can be found in each period. If the money supply is constant, and distributed
evenly to the old, then the price level will be a constant. When everyone is perfectly
informed, the price level will be proportional to the money supply. Thus, with perfect
information there are no price rigidities as defined here: a once and for all increase in
the money supply will lead to a proportional increase in the price level and employment
and production will be tmchanged.
Price rigidities arise as a result of limits on information that prevent a realization of
the money stock from being revealed to both the young and the old. The mechanism
for limiting information is that trading takes place in two distinct markets. Then an
increase in the price in one of the markets can signal either (1) that the money stock has
increased, in which case there is a general increase in all prices and no need to change
production, or (2) that there are fewer suppliers in the market and that the relative price
has increased, in which case it makes sense to increase production. Hence, with limited
information about the source of the price rise, suppliers must solve a signal extraction
problem; the solution to the problem is to increase supply when the price rises, but by
an amount that is less than the increase in the money supply. The response of prices to
the increase in the money supply depends on the relative variability of money supply
changes and market specific shocks. Thus, in this case there is a price rigidity: the price
level rises by less than the increase in the money supply and production increases.
3.1.1. Empirical tests"
Lucas (1973) provided the first test of this model; he looked at the price-output
behavior in several countries and found evidence that an increase in the variability of
the general price level tended to raise the response of prices to changes in aggregate
1023
demand as predicted by the theory. Later, Ball, Mankiw and Romer (1988) provided
further support. For example, in a sample of 43 countries Ball, Mankiw and Romer
(1988) find a significant negative relationship between the response coefficient of real
output to aggregate nominal demand shocks (a more general measure of demand than
the money supply) and the variability of nominal demand.
Soon after the Lucas model appeared, Barro (1977) endeavored to test the theory
by empirically distinguishing between unanticipated and anticipated changes in money,
finding that the effects of unanticipated money were larger, as predicted by the theory.
Barro's study triggered an enormous amount of research and debate about whether
anticipated money had any effect at all. Because his tests could not completely
distinguish between limited information models and other models with rational
expectations (such as the ones discussed later in this paper) that also predicted a
difference between expected and unexpected money, questions were raised about the
relationship between his test and the Lucas theory. See Mishkin (1982) for a good
summary and econometric assessment of this debate.
Three types of empirical results, however, raised questions about market clearing
models with imperfect information as an explanation of the impact of money on the
economy. First, contrary to the theory, actual measures of mis-perceived changes in
money had little impact on production; Barro and Hercowitz (1980), for example,
provided evidence for this by comparing preliminary and revised data on the money
supply in the USA; the difference between the preliminary and the revised data was
taken as a measure of mis-perceived money. One can question the Barro-Hercowitz
method as taking the Lucas model too literally - clearly many people are unaware
of the money supply even after it is revised. Perhaps a better test of the spirit of the
model would use a more general measure of aggregate demand as in Ball, Mankiw and
Romer (1988). Nevertheless, the information limitations assumed in the Lucas model
do seem strong without direct evidence for their existence.
A second empirical result that raised questions about the market-clearing approach
with limited information, is that unanticipated changes in prices seem to explain only
a small fraction of the changes in production over the business cycle, again contrary to
the causal mechanism of the model when taken literally. Sargent (1976), for example,
showed that unanticipated changes in prices had little effect on production using data
for the USA. However, like the Barro-Hercowitz tests, these tests may focus too much
on the details of the Lucas model and not enough on its more general implications. For
example, in models with sequential purchase restrictions [such as Lucas and Woodford
(1994) and Eden (1994b)], an unanticipated increase in demand causes an increase in
sales without an immediate increase in prices to signal more production.
A third empirical shortcoming relates to the persistent effects of a money shock
on real output. In market-clearing limited information models, the duration of the
effect of a money shock on real output is no longer than the time it takes to resolve
the uncertainty about the source of the shock. It is unlikely that such information
would take longer than a few months let alone longer than a year to obtain. Hence,
without adding other sources of persistence the market-clearing models with limited
1024
J.B. Taylor
information have difficulty explaining a key fact about the money, real output, inflation
relationship. To be sure, it is possible to add in such persistence effects through
adjustment costs or inventories, but doing so leaves other aspects of the business
cycle - such as the reverse dynamic cross correlations between inflation and output
discussed in the introduction - unexplained. It is because of these difficulties in
explaining persistence, in my view, that most estimated economy-wide monetary
models do not utilize the perfectly flexible-price, market-clearing, assumption, even
with limited information.
In any case, there has been relatively little research in recent years on imperfect
information as a source of price rigidities in models with perfectly flexible prices.
However, for reasons discussed below it is likely that a complete theory of price
rigidities will eventually involve elements of the limited information theory, though
perhaps in conjunction with the staggered wage and price setting formulations
discussed below. One rationale for infrequent price changes stems from the information
value that a stable price conveys to customers. Moreover, as discussed below, the
microeconomic foundations of the staggered contracts models - and in particular why
staggering even exists - involves imperfect information about whether a shock is
temporary or permanent or local or economy-wide.
1025
the real effects of money in rational expectation models with perfectly flexible prices
assumes that people must learn about monetary policy during a "transition" to rational
expectations [Taylor (1975), Brunner, Cukierman and Meltzer (1980)]. Models with
learning have developed considerably in the 1980s and 1990s as discussed in the
chapter by Evans and Honkapohja (1999) in this Handbook.
Incorporating sticky prices or wages directly into an economy-wide rational
expectations model meant assuming that prices or wages are set in advance of the
market period when they apply and then are fixed at that level during the market period.
For example, automobile firms might set their price pt for the quarter t at the start of
quarter t - 1 and then keep the price at that value throughout the market period. Or
firms might set the wage wt for year t at the start of period t - 1 and set the wage Wt+l
for the year t + 1 at the start of period t, and so on. To make this model operational in
an internally consistent rational-expectations model of the economy, one might assume
that prices or wages are set in such a way that markets are expected to clear during the
period in which the price or wage applies. If we let St(pt) represent supply in period t
and Dt(pt) represent demand in period t, then expected market clearing simply means
that the price pt which is set in period t - j is chosen so that
Etq(St(pt)) = Et-j(Dt(pt)),
(3.1)
where Et_j is the conditional expectation given information through periods t - j . This
is the approach taken by Fischer (1977), Gray (1976) and Phelps and Taylor (t977)
to incorporate price and wage rigidities into economy-wide models. In applying this
approach Phelps and Taylor (1977) focussed on the aggregate price level; they assumed
that firms set prices one period in advance so that expected aggregate demand for
goods (a negative function of the price level) equals expected aggregate supply. Fischer
(1977) and Gray (1976) assumed that wages were set in advance in an analogous
fashion with the price level perfectly flexible. They assumed that the wage was set
so that the expected quantity of labor supplied equaled the expected quantity of labor
demanded. In other words, in the Fischer and Gray models the wage is set in advance
and D and S refer to the economy-wide labor market during the period when the wage
applies, while in the Phelps and Taylor (1977) model prices are set in advance and
D and S refer to the aggregate goods market during the period when the price applies.
This is also the approach taken in more recent work by Cho (1993), Cho and Cooley
(1995), and Rankin (1998) to incorporate price or wage rigidities into real business
cycle models, though the D and S functions in these models are far more complex
than in the relatively simple macro models in which price and wage rigidities were
inserted in the mid-1970s.
When setting prices or wages, agents in the model are assumed to make forecasts
of supply and demand conditions in the future. Because the forecasts of supply and
demand depend on the price in the next period, it is possible to find a price so that
the expected quantity supplied equals the expected quantity demanded as shown in
Equation (3.1). This price is the expected equilibrium price level; it is this value
1026
J.B. Taylor
which is assumed to be set in advance of the market period and then not changed.
Actual supply and demand conditions will differ from these expected values because of
unforeseen shocks. But because the price is fixed, the actual quantity demanded will not
equal the actual quantity supplied - hence there is an excess demand or excess supply.
The convention is then simply to assume that the quantity demanded determines the
supply. For example, firms are assumed to supply whatever is demanded at the fixed
price for that period. In the meantime the price is set for the next period and the model
proceeds through time.
Of course, this expected-market-clearing mechanism describing how the price is
determined is not necessarily a realistic description of how firms actually set prices, no
more than the standard market-clearing assumption in perfectly flexible-price models
is meant to be a realistic description of how firms actually set prices. In the equilibrium
models, it is hoped that the market-clearing approach gives good predictions and so it
is with the "expected equilibrium" models.
One advantage of this approach to sticky prices or wages is that the longrun neutrality of monetary policy is always preserved, a point first emphasized by
McCallum (1982) in an earlier survey. However, there is a serious disadvantage:
the expected-market-clearing approach provides no explanation of the persistence
of monetary shocks. The persistence caused by this type of sticky wage or price
mechanism could last no longer than the longest lead time for wage and price setting.
Recall that the review of the studies of wage and price setting in the previous section
showed that the average duration of nominal wages and prices was no longer than
one year. Yet the persistence of money shocks lasts well beyond one year. In other
words, the expected-market-clearing approach has the same empirical shortcoming as
the market-clearing approach, and trying to remedy these shortcomings by introducing
other sources of persistence leads to a counterfactual implication for the inflation and
output dynamics.
It was in trying to apply the expected-market-clearing approach in constructing an
estimated econometric model that I realized how serious this disadvantage was. The
empirical problem with such a formulation was that real output jumped back to the full
employment level much too quickly and sharply after a money shock. I found that in
building an empirical model I could not use the expected-market-clearing formulations
that had proved useful in building the simple theoretical models of Phelps and Taylor
(1977), Gray (1976), or Fischer (1977). It is interesting that Yun (1994, Chapter 2)
had a similar experience when endeavoring to introduce sticky prices into an empirical
real business cycle model; he found that setting a price for a single market period in
advance could lead to no longer persistence than the length of the longest lead time
in price setting, and for this reason he eventually adapted to an alternative approach
used prior to the development of real business cycles - a version of the staggeredprice-setting approach described in the next section.
In retrospect, it is interesting to observe that the persistence problems of the
expected-market-clearing models with prices or wages set in advance are similar
to those of the market-clearing model with perfectly flexible prices and limited
1027
information. The persistence of a shock to money would not last beyond the length
of the longest lead time for price setting or beyond the longest time for learning the
source of an observed shock to aggregate demand.
The staggered contract model developed in Taylor (1979a, 1980a) was explicitly
designed to have these features; it is a simple model of price or wage setting designed
to highlight certain key properties of real-world price and wage setting. The equations
are essentially the same for wage setting and price setting. Consider the case of price
setting. In the basic model prices are set for a fixed number (N > 1) of periods and are
not changed during the length of this N-period "contract period". Each period, 1/N of
the firms change their "contract prices". At any moment of time, the prevailing price
would be an average of the N outstanding contract prices determined in the current
and the last N - 1 periods. When setting the current price, firms would take account
of both future and past price decisions of other firms because these would be part
of the prevailing price. Thus the equations have both forward-looking and backwardlooking terms which are implied automatically from elementary considerations about
how prices are set. These staggered contract equations had the feature that there is
no long-run trade off between inflation and unemployment: regardless of the steadystate inflation rate the unemployment rate would equal the natural rate, although this
property requires that future prices are not discounted when setting today's price, a
property that may be a good approximation for short contracts of one year or less.
1028
J.B. Taylor
To be concrete, suppose that N = 2, and that Pt is the log of the average price
prevailing in period t, xt is the log of the price set in period t to apply to period t
and t + 1, and Yt is the log of aggregate real output. Then, the basic staggered price
setting model is
Pt = 0.5(xt +xt-1),
(4.1)
(4.2)
where et is a shock to price setting. The first equation states that the current price is
the average of the two outstanding "contract prices". Equation (4.2) posits that this
contract price will depend on prices prevailing during the contract period and on a
measure of total demand in the economy during the two-period contract length. In
Taylor (1979a, 1980a) I viewed the role of Yt in this equation as representative of
excess demand in the markets during the periods when xt applied, but as discussed
below this is only one interpretation. The model can be closed by assuming a simple
demand-for-money function such as (rot - P t ) = Yt, where mt is the log of the money
supply, and by specifying a stochastic process for the money supply.
Suppose for example that mt = rot-1 + th, where t/t and et are serially uncorrelated
random variables. Then, by substituting for Yt in Equation (4.2) using the demand
for money and then substituting for pt using Equation (4.1) one can easily derive an
autoregressive moving average process for Yt in which the autoregressive parameter
(the coefficient onyt_~) is a = e - (e 2 - 1)-J , where c = (1 + ]/)(1 - )-1 and the moving
average terms depend on the shocks to money and the price-setting equation. The
autoregressive parameter a is inversely related to the parameter y. For small y, the
parameter a will be large and there will be a lot of persistence. Hence, 7 is a
key parameter.
The autoregressive part of the process for Yt arises because the price xt set by one
firm partly depends on the price xt-1 set at other firms - as can be seen by substituting
Equation (4.1) into (4.2). Because of the autoregression, a shock to the money supply
has a long drawn out effect on output and the price level. The autoregressive term
is analogous to a dynamic multiplier and for this reason I used the term "contract
multiplier" to describe this persistence effect. This autoregressive component is why
the effect on output lasts much longer than the length of the longest contract (2 periods
here).
West (1988) and Phaneuf (1990) showed with more detailed models and data from
the USA and other countries, that the persistence could be large enough to explain the
near unit-root behavior that had been associated with real business cycle models. In
other words they showed that near unit-root behavior was consistent with a monetary
theory of the business cycle with relatively short-lived staggered prices and wages.
Moreover, as I showed in Taylor (1980b), there is also a pattern of reverse dynamic
cross correlations implied by this model in which a higher level of real output is
followed by a higher price level, while a higher price level is followed by a lower level
of real output - in other words real output Granger-causes the price level (positively)
1029
while the price level Granger-causes real output (negatively). This reverse correlation
is found in the data from many countries as reported in Taylor (1980b) and mentioned
in the introduction to this chapter. Hence, even this highly stylized model is capable
of explaining key facts of the dynamic relation between money, output, and the price
level. Note, however, that these facts pertain to the behavior of the price level rather
than the rate of inflation, an important issue to which I will return in the discussion
o f inflation persistence below.
Pt = Z
~,xt-,,
(4.3)
s=0
which is drawn from Taylor (1979b); Equation (4.2) can also be generalized in an
analogous fashion with ~ weights replacing 1/N or 1 in the case o f N = 2. In the special
case where ar0=a~ =0.5 and the rest of the a~'s are zero, Equation (4.3) reduces to
Equation (4.1), the two-period price-setting case. Alternatively, if there were an equal
number of prices with durations of one through four quarters, then the a~ weights would
decline linearly. In the next several sections, I consider versions of Equation (4.3) that
have proved useful in research.
1030
JB. Taylor
(1983) I calibrated the ~ weights using union wage data in the USA; the observed
contract distribution varied in length from 4 quarters to 12 quarters. The resulting
model was then used to simulate different disinflation paths for the USA to follow to
bring the inflation rate down from the high levels of the late 1970s and early 1980s.
Finally, in an estimated multicountry model [Taylor (1993a)] I estimated implied
distributions of contract length for the largest seven industrialized countries, with the
allowance of more synchronization in Japan, where observation of the wage setting
process suggested there would be synchronization. (These estimates were discussed in
the section on indirect observations of wage setting above.)
Blanchard (1983, 1987) significantly extended the idea of unsynchronized price and
wage setting to a complete stage-of-process model in which the price of inputs affects
the price of outputs which then affects the price of an input to another firm and so on.
The process of passing through price changes at each stage of production generates
staggered price setting with the dynamics depending on the input-output structure
of the economy. Hence, this provides another way to calibrate, or at least interpret,
staggered price- and wage-setting models. Gordon (1981) also places great emphasis
on stage-of-process effects in models of aggregate price dynamics.
Christiano's (1985) extension of the staggered price and wage setting model also
proved useful. He allowed for adjustment of contracts more frequent than the time
interval for data collection and estimation. For example, Christiano's generalization
could be used to estimate a model in which some contracts last only one quarter, but
the data are annual. Using this approach Christiano was able to improve the goodness
of fit of the simple staggered contract model.
Buiter and Jewitt (1981) generalized the staggered contract model to allow wage
setters to take the real wages of other workers into account rather than nominal wages.
They showed that this change preserved many of the dynamic properties of the basic
model, but allowed for additional effects because different price indices might be
relevant for workers and firms. This is especially relevant in international economics
where distinguishing between consumer and producer prices allows one to consider
the important implications of exchange-rate pass-through.
4.2.2. Random duration models
Calvo (1982, 1983) developed a simple, but useful, version of Equation (4.3) by
assuming that the Jr weights had a simple geometric form: Jr~. = a ~' for a < 1. Calvo's
original suggestion was to convert the staggered contract model to continuous time and
thus assume an exponential distribution. Moreover, Calvo (1982) provided a stochastic
interpretation of the staggered contracts model: in his words "we basically adopt the
same assumptions [as the standard staggered contracts model] except, to simplify
the mathematics, we suppose that contract length is stochastic and independent and
identically distributed across contracts" rather than described by a fixed distribution
of contracts of different lengths. Calvo (1982) suggested that the equations could be
interpreted as implying that the contracts ended randomly according to a geometric (or
1031
The exogeneity of the price- and wage-change intervals in both the fixed and
random-duration staggered price-setting models has been one of their most criticized
assumptions. Making the price change or contract termination decision endogenous
is important for policy or empirical work, especially if exogeneity is a poor
approximation. Fortunately, a number of recent studies have begun to develop models
in which the duration of price and wage decisions depends on the state of the economy;
this approach is called state-dependentpricing. In contrast the simple staggered priceand wage-setting model is called time-dependentpricing because prices change at fixed
or randomly selected times.
Caplin and Spulber (1987) developed a widely-discussed model in which all prices
are completely state dependent; that is, there is no explicit dependence on time as in
the staggered contract models. With state-dependent pricing each firm is faced with
fixed costs of price adjustment and uses an (S, s) policy to determine whether the
price will change and by how much [see Sheshinski and Weiss (1988) for more on
(S, s) policies in price adjustment]. Because not every firm will change its price in
every period the resulting pattern of price adjustment looks just like staggered timedependent pricing. However, Caplin and Spulber (1987) find the switch from pure
time-dependent pricing to pure state-dependent pricing greatly reduces the effects of
staggered wage and price setting on the macroeconomy. In particular they find that
money can be completely neutral in such a model. The reason is that if all price setters
are following a (S, s) policy, with a wide enough band, then they can all change prices
by the full amount of a monetary shock as soon as the shock appears. In contrast with
the time-dependent pricing in the staggered price-setting model discussed above, some
firms will not change their price so that the aggregate price level adjusts slowly.
Tsiddon (1991, 1993) shows that slow aggregate price adjustment and the nonneutrality of money reappear if the changes in the money supply are highly persistent or
1032
J.B. Taylor
have large swings. In such cases the size of the firm's price change is reduced leading to
smaller changes in the aggregate price than the money supply. Conlon and Liu (1997)
show that if firms change prices in response to other things than price misalignment
(a product upgrade, a new model or a new product mix), then the nonneutrality of
money reappears. Essentially this change results in a mixture of time-dependent and
state-dependent pricing, a situation which probably better reflects reality than either
extreme.
An important recent application of state-dependent pricing is found in a general
equilibrium model developed by Dotsey, King and Wolman (1996), who modified the
geometric staggered contract framework of Calvo (1982) to allow for state-dependent
pricing. In the Dotsey, King and Wolman (1996) model the fraction of firms that are
changing their prices in any one period increases when the inflation rate rises. This
prediction of the model is supported by many of the papers surveyed in Section 2. An
important advantage of the Dotsey, King and Wolman (1996) paper is that they embed
state-dependent pricing into an economy-wide model and preserve some degree of
time dependence. They find that the money, output, and price dynamics resulting from
their state-dependent model are not too dissimilar from the dynamics of the purely
time-dependent model discussed above.
Another example of state-dependent pricing is the model of Caballero and Engel
(1993). Like the Caplin and Spulber (1987) model, only a fraction of prices will be
adjusted each period in the Caballero and Engel model. Thus staggered price setting
emerges. However, Caballero and Engel (1993) assume that the probability that an
individual price will adjust depends on both the size and the sign of the deviation of
the price from some desired price. Caballero and Engel (1993) look at the implications
of their model for time-series behavior of the aggregate price level focussing on
the producer price index. The detrended log of the index itself is described by a
second-order vector autoregression with coefficients of 1.68 and -0.76. The Caballero
and Engel (1993) model with its emphasis on first-order adjustment has difficulty
mimicking this humped-shape behavior implicit in a second-order process, but at least
the model generates the persistence or stickiness of the aggregate price levels found
in the pure time-dependent staggered contract model.
As described above in Section 4.1, the basic staggered contract model was constructed
to be consistent both with certain observed features of price- and wage-setting
behavior and with basic microeconomic principles about the operation of competitive
or imperfectly competitive markets. In particular, the idea that the price and wage
decisions of firms depend on the prevailing prices and wages - and thus on the
price and wage decisions at other firms - is an essential characteristic of staggered
wage- and price-setting equations. In a sense these equations endeavor to describe
a price discovery process in markets with posted prices, much like the equilibrium
1033
1034
JB. Taylor
The combination is achieved in two steps. The first step is to find the optimal price from
a monopolistically competitive model without staggered price setting, such as Svensson
(1986) or Blanchard and Kiyotaki (1987). For example, a typical optimal pricing
rule would say that the price should depend on the average of other monopolistically
competitive firms prices (p) and on a constant times aggregate output (y), which would
represent a demand shift. A linear version of such a rule would look like
x = p + yy,
(5.1)
where x is the firm's price, p is the average of other firms prices, and y is aggregate
output.
Now, suppose that pricing is staggered with 2 period contracts so that the price xt
must last for two periods: t and t + 1. Then it seems reasonable that firms would set
their price to be the average of the optimal price during the two periods during which
the price applies. This reasoning leads to the second step which sets
xt = 0.5(pt + YYt) + 0.5(Et lPt+l + ]/Et-lYt+l).
(5.2)
Note that Equation (5.2) is identical in functional form (ignoring the random shock)
to Equation (4.2). Hence, pricing under monopolistic competition gives a more formal
underpinning of the staggered price setting model. However, now the role of y
is to shift firms' demand functions rather than to serve as a measure of (excess)
demand pressure in the market. Blanchard and Fischer then went on to describe
various persistence properties when Equation (5.2) is imbedded in a model. Because
Equation (5.2) is identical to Equation (4.2), the properties are identical to those
I discussed above. Romer (1996) presents a very useful textbook treatment of this
type of derivation of the staggered price setting equation, providing useful details of
the derivations and fin'ther discussion. Rotemberg's (1987) derivation of a staggered
price setting mentioned above is similar to the one of Blanchard and Fischer (1989)
described here, except that rather than using the example in Equation (4.2) he uses the
version of Equation (4.2) with geometric declining weights on future p's and y's as in
Calvo's (1982) geometric version of the staggered contracts model [see the discussion
following Equation (4.3)].
Although the above derivation of the staggered price setting equations provides
a helpful microeconomic interpretation, it is still not a fully optimizing treatment.
In work that began in the mid-1980s, Akerlof and Yellen (1991) developed a
dynamic model in which Equation (4.2) or (5.2) emerged directly from a monopolistic
competition model without the two steps described above. They showed how the simple
staggered contract equations could be derived from a maximization problem in which
monopolistically competitive firms interact.
Consider a simple two-firm model in which each firm faces a downward-sloping
demand curve and thus has a degree of market power. The demand curve depends on
real income in the economy, its own price, and the price charged by its rival. (The
1035
process determining real income is left unspecified, but is eventually related to the
money supply as in other models.) One firm sets prices in even periods and the other
firm sets prices in odd periods. The prices are fixed at the same nominal value for
two periods. The rival firm can be thought of as a monopolistic competitor that sets a
constant nominal price every other period. Akerlof and Yellen (1991) show that if each
firm chooses a price to maximize expected profits taking as given (1) the price set by
the rival firm (a Bertrand price setter) and (2) income in the economy, then the optimal
price is given by the simple staggered price setting model in Equations (4.1) and (4.2).
To get linearity in the decision rule, Akerlof and Yellen (1991) approximate the profit
function by a quadratic.
5.3. Staggered price and wage setting in general equilibrium models
Although Akerlof and Yellen (1991) derived staggered price- and wage-setting
equations from first principles, they did not endeavor to embed the equations in a fully
optimizing model of the economy - that is, a model that includes utility functions for
representative households. During the late 1980s and 1990s there has been a great
amount of research aimed at doing just that. Nelson (1997) provides an excellent
review of the most recent part of this research, but some discussion of earlier work
is useful too. In one of the earliest studies along these lines, Deborah Lucas (1985,
1986) developed a full optimizing model in which some prices are determined in spot
markets and some are determined in contract markets. The prices in spot markets
are determined in the usual market-clearing way. The wage contracts are assumed
to last two periods; 50 percent of the contracts are set in odd periods and the other
50 percent are set in even periods. The contract specifies a fixed nominal wage for
two periods: the current period and the next period. Interestingly, Lucas (1985, 1986)
developed a wage-setting mechanism for the contracts in competitive markets, so that
the results do not depend on the market power of firms. The terms of the contract are
determined in a market-clearing fashion. As stated by Lucas (1986), "In the process
of competing for the N workers available to a given sector, the economy equilibrates
so that the marginal utility gained from the wages paid over the contract period equals
the marginal disutility of labor over the contract period". Note that this approach to
modeling wage determination in the contracts is analogous to that used in the optimal
contract literature [see Azariadis (1975)]; it is different from both the monopolistic
competition assumption and the expected-market-clearing approach mentioned above.
Deborah Lucas (1985, 1986) used a cash-in-advance approach to money demand
and has prices set for two periods, with decisions being made every other period.
Simulations demonstrated the effect of the nominal rigidities on the effect of monetary
policy. In her model, the amplitude of cycles was proportional to the fraction of markets
with wage contracts compared with spot pricing.
Levin (1989, Chapter 2 and 1990) also developed an optimizing model in which
wages are set in a staggered manner and determined optimally. Levin also obtained
estimation and policy results. He estimated the model using maximum likelihood
1036
JR Taybr
methods and US data. He then stochastically simulated the model to derive optimal
feedback rules for monetary policy, and calculated the tradeoff between output stability
and price stability. Because the parameters of Levin's model were estimated he could
obtain quantitative estimates of the policy rule and the policy tradeoff.
Yun (1994, 1996) developed a general equilibrium model with staggered price
setting, both with a fixed and random duration. Because Yun uses monopolistic
competition, his analysis of staggered price setting has similarities to that of Blanchard
and Fischer (1989) above, but it is not a two-step approach. Staggered pricirrg equations
of the form (5.2) emerge from the optimization problem without requiring that one first
find the optimal price without staggering and then inserting that price into the basic
staggered price-setting equations. A difference in the resulting equations is that prices
in furore periods are discounted relative to current prices; that is (pt + Yyt) would get
a larger weight than (Et lPt+1 + yEt lYt+l) in Equation (5.2) because the firms' profits
in the second period would be discounted. Yun's approach shows explicitly how the
single price in a staggered price-setting model must balance out profits in different
periods because the price cannot be at different levels in different periods.
Kimball (1995) discusses several results that emerge when he places formally
derived staggered price-setting equations in a general equilibrium model. Kimball
(1995) noted that the parameters of the resulting equations may give less persistence
than earlier estimated staggered price-setting models (for reasons similar to those
mention in my discussion of Blanchard and Fischer) and discussed several factors that
could lengthen persistence that have proved useful in later work. In commenting on
Kimball (1995), Woodford (1995) gives a nice comparison of the Yun (1994, 1996)
and Kimball (1995) models.
Compared to other research in this area, the work by King and Wolman (1996)
focusses more explicitly on policy analysis. King and Wolman (1996) develop a utility
maximization model with price and wage rigidities. Like Deborah Lucas (1985, 1986),
Levin (1989), and Yun (1994) they add price rigidities to the model, using a staggered
price-setting model. Money enters their model through a transactions technology
device in which monetary services allow for higher levels of consumption. As with
Levin (1989) and Deborah Lucas (1985, 1986), money has a real effect in the model
because of the nominal rigidities. King and Wolman (1996) examine inflation targeting
procedures and other important issues in the design of monetary policy rules.
By the mid-1990s there were also a number of papers that added other forms of price
and/or wage rigidities to a general equilibrium model. Hairault and Portier (1993),
Kim (1995), and Ireland (1997) used a quadratic cost of price adjustment approach
as suggested by Rotemberg (1982), while Cho and Cooley (1995) assumed that prices
were set in advance in such a way that the overlapping features of staggered price
setting would not play a role in producing persistence. Other work in this general
area has been motivated by various policy, empirical, and methodological issues and
includes Leeper and Sims (1994), Bernanke, Gertler and Gilchrist (1999), Ohanian
and Stockman (1994), Chari, Kehoe and McGrattan (1998), and others. The paper by
Chari, Kehoe and McGrattan (1998) raised some puzzles about the ability of staggered
1037
price and wage equations to deliver persistent dynamics and is taken up along with
several responses later in this chapter.
Although based on empirical observations, the staggering of wage and price setting
is simply assumed in all the models discussed thus far in this chapter, from Taylor
(1979b) through Chaff, Kehoe and McGrattan (1998). Why is price and wage setting
staggered? This question has been pursued by many researchers in the 1980s and 1990s
and is still the subject of debate. The question goes to the heart of the price discovery
process in a market economy and well beyond macroeconomics.
Fethke and Policano (1984, 1986) develop a model in which wages must be set
several periods in advance and then are fixed without contingencies. The question is
whether in such a world the wage settings should be staggered or all occur at one time.
Fethke and Policano (1984) proved that staggering is a good way for the economy to
adjust to sector-specific shocks. When disturbances are primarily due to relative, as
distinct from aggregate, shocks, staggering of decisions is optimal because adjustments
of some prices enable the sectors that are locked into fixed wages or prices to partially
adjust. Parkin (1986) also shows that the degree of staggering depends on the relative
size of aggregate shocks versus sector specific shocks. The analysis of monetary policy
in this type of model is considered in Fethke and Policano (1987). They derive a
Nash equilibrium where timing of monetary policy intervention and synchronization
of contracts are simultaneously decided upon.
An entirely different explanation for the existence of staggering comes from
informational considerations. Ball and Cecchetti (1988) show that staggering allows
firms to obtain information about what is going on in other markets. By observing
the price in other markets firms are able to extract information about whether shocks
are aggregate or relative. In addition, Ball and Romer (1989) find that there is a
more rapid adjustment to sector (idiosyncratic) shocks with staggering. As shown
by Ball (1987), these microeconomic advantages are offset, at least in part, by
macroeconomic disadvantages of slow aggregate price adjustment, a concept he refers
to as "externalities from contract length". Sheshinski and Weiss (1988) consider the
question of staggering in oligopolies where timing is endogenous. Lau (1996, 1997)
explores the strategic issues between price setters in an oligopoly game. The strategic
rationale for staggering was explored by Matsukawa (1986) in the case of wage
setting.
In a recent paper, Bhaskar (1998) derives endogenous price staggering in a
model with heterogeneous firms. Firms within an industry have stronger strategic
complementarities than firms in different industries. This results in an equilibrium in
which there is synchronization within industries but not across the economy. De Fraja
(1993) also utilizes such strategic considerations in a model in which staggered price
setting can be endogenous.
1038
J.B. Taylor
1039
models such as Equation (4.3) above do pass goodness of fit tests like those proposed
by Ashenfelter and Card (1982).
Another empirical criticism of the staggered contract model is found in Dezhbakhsh
et al. (1984). They argued that the statistical Phillips curve correlations reported in
Taylor (1980b) arose after supply shocks, not demand shocks, and that correlations
were actually based on forward looking derivatives (differences). While the point about
the supply shocks is correct, the more typical statistical Phillips curve correlations
would arise if demand shocks were correlated. A related point was made by Phaneuf
(1987a,b). He pointed out that uncorrelated demand shocks showed a negative effect
after the length of the longest contract. This property can be avoided with serially
correlated demand shocks.
6.1. Inflation persistence
A widely discussed econometric problem with the staggered contract model is its
apparent inability, without serial correlation or other sources of dynamics, to generate
the persistence (or inertia) of inflation observed in the data. For example, Ball (1994)
showed how it was possible to reduce inflation without recession - indeed with the
right policy to have a boom! Phelps (1978), Taylor (1983), and Abraham (1987)
also examined costless disinflations with rational expectations, but they did not relate
this finding to empirical defects with the model. Though the inconsistency between
these results and the observed costly disinflations can be easily explained by learning
or by lack of credibility, the inconsistency raises some doubts about the ability of
staggered contracts models with rational expectation to generate inflation persistence.
The apparent inconsistency has led some empirical researchers to use a modification,
proposed by Fuhrer and Moore (1995a,b), of the staggered contracts model.
Fuhrer and Moore (1995a,b) present autocorrelation plots that nicely document
some of the difficulties with the ability of the staggered contract model to produce
inflation persistence. They show that the cross autocorrelation functions based on
actual inflation and output data were not closely matched with the simulations of
the basic staggered contract model. Note that the tests reported in Taylor (1980b)
mentioned in Section 4 above compared price levels rather than rates of inflation;
hence, the reason for one test passing and the other failing.
To remedy this problem Fuhrer and Moore (1995a,b) modified the staggered contract
model. Rather than current price levels being based on expected future price levels
as in the standard staggered contract models, their modified model has current price
inflation being based on expected future price inflation. The replacement of levels by
rates of change generated a model with inflation persistence rather than simply pricelevel persistence. However, the theoretical foundations of the staggered contract model
are based on levels not rates of change [recall the discussion of the derivation by
Akerlof and Yellen (1991) above].
To be sure, Fuhrer and Moore motivated their alternative formulation by restating
the price decision in "real" terms. To see this most simply, set 7 = et = 0 and replace
1040
J.B. Taylor
rational expectations by perfect foresight in Equation (4.2). Note that after substitution
from Equation (4.1) we get
xt = 0.5(x,_l +xt+l),
(6.1)
which illustrates how the current contract price depends on future and lagged contract
prices. Now suppose that rather than using xt in Equation (6.1) we formulate the
price decision in terms of xt - P t , which looks like a "real" price: the current (log)
contract price is deflated by the (log) aggregate price. Replacing xt+~ with Xt+s-Pt+s
for s = 0, 1 and -1, in Equation (6.1), we get (after some algebraic manipulation)
Axt = 0.5(Axt_l + Axt+l),
(6.2)
where Axt = x t - x t 1. Thus, all the properties stated above in terms of price levels are
now restated in terms of the inflation rate. However, this definition of the real price
effectively deflates by a price that does not apply to the full period (t and t + 1) of the
contract price; that is, pt applies to only 1 of the period that xt applies to. Replacing
xt in Equation (6.2) with x t - 0.5(pt +Pt+1) may seem to result in a more appropriate
real price. Thus, while the Fuhrer-Moore formulation may work in macroeconomic
empirical applications, it leaves puzzles about the microfoundations.
Rotemberg (1997), in commenting on Fuhrer (1997), argues that there is nothing
wrong with appealing to other sources of persistence of inflation within the staggered
contracting approach. It is possible, of course, that inflation persistence could be due
to serial correlation of money, but since one of the aims of these models is to explain
persistence, leaving all the persistence of inflation to exogenous serial correlation is
not a completely satisfactory conclusion either.
In a recent paper Roberts (1997) re-examines Fuhrer and Moore's (1995a,b)
findings. By exploring alternative expectations-formation mechanisms, Roberts (1997)
demonstrates that a small amount of"imperfect information" about the determinants of
inflation when combined with staggered price setting is enough to explain the observed
serial correlation of inflation. In my view, Roberts' (1997) results indicate why it is
likely that a full understanding of price and wage rigidities will eventually involve both
imperfect information and staggered contracts of some form. Gertler (1981, 1982)
developed a model of wage rigidities and wage inertia that is based on imperfect
information. Rudin (1987) developed a formal model of staggered price setting in
which there are diverse expectations on the part of firms giving rise to a situation
where firms' expectations depend on other firms' expectations, and so on.
6.2. Real output persistence
The recent study by Chari, Kehoe and McGrattan (1998) emphasizes another potential
persistence problem with the staggered price- and wage-setting model. Their study and
the responses it has stimulated nicely illustrate the potential benefits of using dynamic
Ch. 15:
1041
optimizing models to study wage and price stickiness. As mentioned above, Chari, Kehoe, and McGrattan (1996) place a staggered contracts mechanism into an optimizing
model in which price-setting rules are derived assuming monopolistic competition;
they assume that money is in the utility function. They then examine the dynamic
properties of the model and compare the properties with the staggered contract mechanism when placed in a model with less formal optimization; that is, with Equations
(4.1) and (4.2) above. Through a pairwise comparison of the models, they find that,
with reasonable parameter values calibrating the optimizing model, they cannot get
coefficients on the staggered price setting equations that are large enough to generate
empirically realistic serial correlation; they find virtually no persistence beyond the
length of the longest contract. In terms of Equation (4.2), the value of 7 they get from
their calibration exercise is much too large. Hence, the persistence is much shorter than
the kind of persistence that West (1988), Blanchard (1990), and Phaneuf (1990) have
found with the staggered contract model. Ascari (1998) develops an optimizing model
of staggered wage setting which also implies that the y parameter is way too large.
Several interesting papers have already been written in reaction to the findings
of Chari, Kehoe and McGrattan (1998). Gust (1997) shows that restricting capital
mobility between sectors can increase persistence in the Chari, Kehoe and McGrattan
(1998) model. Kiley (1997) and Jeanne (1997) show that increasing the size of real
rigidities [in the sense of Ball and Romer (1990)] can increase persistence in the model.
A common theme of these papers is that there needs to be some neighborhood effects
between price setters, so that one firm pays close attention to the price decision of
the next firm and the most recent firm, thereby linking the price decision of one firm
to another and causing the persistence effects. Gust's (1997) model illustrates this by
tracing out in detail the effects of shocks with and without capital mobility; in his
model capital plays a role in affecting the linkage between price decisions in different
markets.
The Chari, Kehoe and McGrattan (1998) model assumes complete wage flexibility.
Alternatively, Erceg (1997) shows how including staggered wage setting along with
staggered price setting increases persistence and enables a calibrated optimizing model
with staggered contracts to generate the persistence observed in the data. Bergen and
Feenstra (1998) introduce a more general functional form for the demand curve facing
the monopolistic competitors, which leads to a lower value for 7. The effect of the
constant elasticity demand functions on persistence is also noted by Kimball (1995).
It is also important to note that Rotemberg and Woodford (1999) and King and
Wolman (1998) in a similar type of modeling framework find long persistence of
monetary shocks with relatively short price contracts. Rotemberg and Woodford (1999)
assume price-setting equations which are geometric after a time delay, thereby coming
close to the microeconomic empirical estimates discussed in Section 2 where the most
common length of price contracts seems to be about three or four quarters. Rotemberg
and Woodford (1999) conclude that their model generates realistic macroeconomic
persistence of money shocks. King and Wolman (1998) assume two-period contracts
and also generate realistic macroeconomic persistence.
1042
J.B. Taylor
Another possible resolution of the Chari, Kehoe and McGrattan persistence puzzle is
that the monopolistic competition model used to derive the price adjustment equation
may not be adequate. Recall that Deborah Lucas (1985, 1986) used a competitive
contracting mechanism to derive the price rule (actually a wage rule), and found that
the impacts of shocks were quite persistent. The inconsistency pointed out in the Chari,
Kehoe and McGrattan (1998) study raises issues about monopolistically competitive
pricing that need to be investigated further. As Arrow (1959) argued, the market power
a firm has when setting its price is temporary and may be quite different thar~ the market
power in a full monopolistic competition model. If so, then it is a mistake to tie the
price adjustment parameter to demand-elasticity parameters. As mentioned in the
discussion following Equation (4.1) above, prices may be responding to excess demand
and not simply be moving along a demand curve as is assumed in the monopolistic
model of price setting. Thus, the findings of Chari, Kehoe and McGrattan (1998) may
indicate that the monopolistic competition (stationary market power) model may not
be sufficient as a microeconomic foundation.
6.3. Changes in stability and nominal rigidity over time
Ch. 15:
1043
1044
J.B. Taylor
A n important advantage o f the newer models that include both utility maximization
and staggered price setting is that monetary policy can be evaluated using the standard
tools o f public finance. Welfare measures such as compensating variations or equivalent
variations thus replace cruder quadratic loss functions in terms o f aggregate output or
inflation. Rotemberg and Woodford (1999) evaluate the effect o f different monetary
policy rules using the welfare function o f the representative agents in their model.
They find that the parameters o f the staggered price-setting equations in their model
have a significant effect on their welfare calculations.
In sum, the form, interpretation, and parameter values o f staggered price and wage
setting models are highly relevant not only for explaining the impact o f monetary
policy, but also for evaluating its welfare consequences. Understanding these models
more thoroughly takes one well beyond macroeconomics into the heart o f the price
discovery and adjustment process in competitive and imperfectly competitive markets.
Further research on the empirical robustness and microeconomic accuracy o f staggered
contracts models is thus both interesting and practically important.
References
Abraham, J. (1987), "Income redistribution during a disinflation", Journal of Macroeconomics
9:203521.
Akerlof, G.A., and J.L. Yellen (1991), "How large are the losses from rules of thumb behavior in models
of the business cycle", in: W. Brainard, W. Nordhaus and H. Watts, eds., Money, Macroeconomics,
and Economic Policy:Essays in Honor of James Tobin (MIT Press, Cambridge, MA).
Arrow, K. (1959), "Toward a theory of price adjustment", in: M. Abramovitz et al., The Allocation
of Economic Resources: Essays in Honor of Bernard Francis Haley (Stanford University Press,
Stanford, CA).
Ascari, G. (1998), "Optimizing agents, staggered wages and persistence in the real effects of money
shocks", unpublished paper (University of Warwick).
Ashenfelter, O., and D. Card (1982), "Time series representations of economic variables and alternative
models of the labor market", Review of Economic Studies 49:761-781.
Azariadis, C. (1975), "Implicit contracts and underemployment equilibria", Journal of Political Economy
83:1183-1202.
Backus, D. (1984), "Exchange rate dynamics in a model with staggered wage contracts", Discussion
Paper No. 561 (Queen's University, June).
Ball, L. (1987), "Externalities from contract length", American Economic Review 77:615 629.
Ball, L. (1994), "Credible disinflation with staggered price setting", American Economic Review 84:
282-289.
Ball, L., and S.G. Cecchetti (1988), "Imperfect information and staggered pricing", American Economic
Review 78:99%1018.
Ball, L., and D. Romer (1989), "The equilibrium and optimal timing of price changes", Review of
Economic Studies 56:179-198.
Ball, L., and D. Romer (1990), "Real rigidities and the nonneutrality of money", Review of Economic
Studies 57:183-203.
Ball, L., N.G. Mankiw and D. Romer (1988), "The new Keynesian economics and the output-inflation
tradeoff", Brookings Papers on Economic Activity 1988(1):1 65.
Barro, R.J. (1977), "Unanticipated money growth and unemployment in the United States", American
Economic Review 67:101-115.
Ch. 15:
1045
Bah'o, R.J., and Z. Hercowitz (1980), "Money stock revisions and unanticipated money growth", Journal
of Monetary Economics 6:257-267.
Benabou, R., and C. Bismut (1987), "Wage bargaining and staggered contracts: theory and estimation",
Discussion Paper No. 8810 (CEPREMAP, Paris, France).
Bergen, ER., and R.C. Feenstra (1998), "Staggered price setting and endogenous persistence", Working
Paper No. 6429 (NBER).
Bernanke, B.S., M. Gertler and S. Gilchrist (1999), "The financial accelerator in a quantitative business
cycle framework", ch. 21, this Handbook.
Bhaskar, V. (1998), "On endogenously staggered prices", University of St. Andrews Discussion Paper
Series, No. 9806 (St. Andrews, Scotland).
Blanchard, O.J. (1983), "Price asynchronization and price level inertia", in: R. Dornbush and M. Simonsen,
eds., Inflation, Debt, and Indexation (MIT Press, Cambridge, MA) 3-24.
Blanchard, O.J. (1987), "Aggregate and individual price adjustment", Brookings Papers on Economic
Activity 1987(1):57-122.
Blanchard, O.J. (1990), "Why does money affect output? A survey", in: B. Friedman and E Hahn, eds.,
Handbook of Monetary Economics (North-Holland, Amsterdam) 779-835.
Blanchard, O.J., and S. Fischer (1989), Lectures on Macroeconomics (MIT Press, Cambridge, MA).
Blanchard, O.J., and N. Kiyotaki (1987), "Monopolistic competition and the effects of aggregate demand",
American Economic Review 77:647-666.
Blinder, A. (1994), "On sticky prices: academic theories meet the real world", in: N.G. Mankiw, ed.,
Monetary Policy (University of Chicago Press, Chicago, IL) 117-150.
Blinder, A.S., E.D. Canetti, D.E. Lebow and J.B. Rudd (1998), Asking About Prices: a new approach to
understanding price stickiness (Russell Sage Foundation, New York).
Brayton, E, A. Levin, R. Tryon and J.C. Williams (1997), "The evolution of macro models at the Federal
Reserve Board", in: B.T. McCallum and C.I. Plosser, eds., Carnegie-Rochester Conference Series on
Public Policy 47:43 81.
Brunner, K., A. Cukierman and A.H. Meltzer (1980), "Stagflation, persistent unemployment, and the
permanence of shocks", Journal of Monetary Economics 6:467-492.
Bryant, R.C., E Hooper and C.L. Mann (1993), Evaluating Policy Regimes: New Research in Empirical
Macroeconomics (Brookings Institution, Washington, DC).
Buckle, R.A., and J.A. Carlson (1995), "Price duration with two-sided pricing rules", in: K.H. Oppenlander
and G. Poser, eds., Business Cycle Surveys: Forecasting Issues and Methodological Aspects (Avebury,
Aldershot) 99-118.
Buiter, W., and I. Jewitt (1981), "Staggered wage setting with real wage relatives: variations on a theme
of Taylor", The Manchester School 49:211-228.
Caballero, R.J., and E. Engel (1993), "Microeconomic rigidities and aggregate price dynamics", European
Economic Review 37:691-711.
Calvo, G.A. (1982), "Staggered contracts and exchange rate policy", in: J.A. Frankel, ed., Exchange
Rates and International Macroeconomics (University of Chicago Press, Chicago, IL).
Calvo, G.A. (1983), "Staggered prices in a utility maximizing framework", Journal of Monetary
Economics 12:3-398.
Caplin, A.S., and D. Spulber (1987), "Menu costs and the neutrality of money", Quarterly Journal of
Economics 102:703-725.
Card, D., and D. Hyslop (1997), "Does inflation 'grease the wheels of the labor market'?", in: C. Romer
and D. Romer, eds., Reducing Inflation (University of Chicago Press, Chicago, IL) 71-114.
Carlton, D.W (1986), "The rigidity of prices", American Economic Review 76:637~558.
Carlton, D.W. (1989), "The theory and the facts of how markets clear: is industrial organization valuable
for understanding macroeconomics?", in: R. Schmalensee and R.D. Willig, eds., Handbook of Industrial
Organization, vol. 1 (North-Holland, Amsterdam) 909-946.
Cecchetti, S.G. (1984), "Indexation and incomes policy: a study of wage adjustment in unionized
manufacturing", Journal of Labor Economics, 5:391-412.
1046
J.B. Taylor
Cecchetti, S.G. (1986), "The frequency of price adjustment: a study of newsstand prices of magazines",
Journal of Econometrics 31:255-274.
Chadha, B. (1987), "Is increased price inflexibility stabilizing?", Journal of Money, Credit and Banking
21:481-497.
Chaff, VV, EJ. Kehoe and E.R. McGrattan (1998), "Sticky price models of the business cycle: can
the contract multiplier solve the persistence problem?", Research Department Staff Report No. 217
(Federal Reserve Bank of Minneapolis) revised April 1998.
Cho, J.O. (1993), "Money and business cycles with one period nominal contracts", Canadian Journal of
Economics 26:638 659.
Cho, J.O., and T.E Cooley (1995), "The business cycle with nominal contracts", Economic Theory
6:13-33.
Christiano, L.J. (1985), "A method for estimating the timing interval in a linear econometric model, with
application to Taylor's model of staggered contracts", Journal of Economic Dynamics and Control
9:363-404.
Christiano, L.J., M. Eichenbaum and C.L. Evans (1999), "Monetary policy shocks: What have we learned
and to what end?", ch. 2, this Handbook.
Conlon, J.R., and C.Y. Liu (1997), "Can more frequent price changes lead to price inertia? Nonneutralities
in a state-dependent pricing context", International Economic Review 38:893-914.
Davis, D., and C.A. Holt (1997), "Price rigidities and institutional variation in markets with posted
prices", Economic Theory 9:63-80.
De Fraja, G. (1993), "Staggered versus synchronized wage setting in oligopoly", European Economic
Review 37:1507-1522.
DeLong, J.B., and L.H. Summers (1986), "Is increased price flexibility stabilizing?", American Economic
Review 78:1031-1044.
Dezhbakhsh, H., A.A. Haug, J.H. McCulloch, G.S. Poonia and L.-T. Wang (1984), "The statistical
Phillips Curve in Taylor's staggered contracts model", unpublished paper.
Domberger, S., and D.G. Fiebig (1993), "The distribution of price changes in oligopoly", The Journal
of Industrial Economics 41:295-313.
Domowitz, I., R.G. Hubbard and B.C. Petersen (1986), "Business cycles and the relationship between
concentration and price-cost margins", Rand Journal of Economics 17:1-17.
Dornbusch, R. (1982), "PPP exchange rate rules and macroeconomic stability", Journal of Political
Economy 90:158-165.
Dotsey, M., R.G. King and A.L. Wolman (1996), "State dependent pricing and the dynamics of business
cycles", unpublished paper.
Driskill, R.A., and S.M. Sheffrin (1986), "Is price flexibility destabilizing?", American Economic Review
76:802 807.
Durra, S., M. Bergen and D. Levy (1997), "Price flexibility in channels of distribution: evidence from
scanner data", unpublished paper ( Emory University, September 14).
Eden, B. (1994a), "Time rigidities in the adjustment of price to monetary shocks: an analysis of micro
data", Discussion paper No 94.16 (Bank of Israel).
Eden, B. (1994b), "The adjustment of prices to monetary shocks when trade is uncertain and sequential",
Journal of Political Economy 102:493-409.
Erceg, C. (1997), "Nominal wage rigidities and the propagation of monetary disturbances", unpublished
paper (Federal Reserve Board).
Evans, G.W, and S. Honkapohja (1999), "Learning dynamics", oh. 7, this Handbook.
Fethke, G., and A. Policano (1984), "Wage contingencies, the patterns of negotiation and aggregate
implications of alternative contract structures", Journal of Monetary Economics 14:151-170.
Fethke, G., and A. Policano (1986), "Will wage setters ever stagger decisions?", Quarterly Journal of
Economics 101:867-877.
Fethke, G., and A. Policano (1987), "Monetary policy and the timing of wage negotiations", Journal of
Monetary Economics 19:87-105.
1047
Fischer, S. (1977), "Long-term contracts, rational expectations, and the optimal money supply rule",
Journal of Political Economy 85:191~05.
Fregert, K., and L. Jonung ( 1986), "Monetary regimes and file length of wage contracts", paper presented
at the 1986 Konstanz Conference.
Friedman, M. (1982), Column, p. 64, Newsweek, July 12.
Fuhrer, J.C. (1997), "Towards a compact, empirically-verified rational expectations model for monetary
policy analysis", in: B. McCallum and C. Plosser, eds., Carnegie-Rochester Conference on Public
Policy 47:197-231.
Fuhrer, J.C., and G.R. Moore (1995a), "Monetary policy tradeoffs and the correlation between nominal
interest rates and real output", American Economic Review 85:219-239.
Fuhr'er, J.C., and G.R. Moore (1995b), "Inflation persistence", Quarterly Journal of Economics 110:
127-159.
Gertler, M. (1981), "Long-term contracts, imperfect information, and monetary policy", Journal of
Economic Dynamics and Control 3:197-216.
Gertler, M. (1982), "Imperfect information and wage inertia in the business cycle", Journal of Political
Economy 90:967-987.
Goodfriend, M., and R.G. King (1997), "The new neoclassical synthesis and the role of monetary
policy", in: B.S. Bernanke and J.J. Rotemberg, eds., NBER Macroeconomics Annual (MIT Press,
Cambridge, MA) 493 530.
Gordon, R. (1981), "Output fluctuations and gradual price adjustment", Journal of Economic Literature
19:493-530.
Gray, J.A. (1976), "Wage indexation: a macroeconomic approach", Journal of Monetary Economics
2:221~35.
Gray, J.A. (1978), "On indexation and contract length", Journal of Political Economy 86:1-18.
Gust, C. (1997), "Staggered price contracts and factor immobilities: the persistence problem revisited",
unpublished paper (Northwestern University).
Hairault, J.-O., and E Portier (1993), "Money, new Keynesian macroeconomics, and the business cycle",
European Economic Review 37:1533 1568.
Ireland, EN. (1997), "A small structural quarterly model for monetary policy evaluation", CarnegieRochester Conference Series on Public Policy 47:83-108.
Jeanne, O. (1997), "Generating real persistent effects of monetary shocks: how much nominal rigidity
do we really need?", Working Paper No. 6258 (NBER).
Kashyap, A.K. (1995), "Sticky prices: new evidence from retail catalogues", Quarterly Journal of
Economics 110:245-274.
Kiley, M.T. (1997), "Staggered price setting, partial adjustment, real rigidities, and sunspots", unpublished
paper (Federal Reserve Board).
Kim, J. (1995), "Monetary policy in a stochastic equilibrium model with real and nominal rigidities",
unpublished paper (Yale University).
Kimball, M.S. (1995), "The quantitative analysis of the basic neomonetarist model", Journal of Money
Credit and Banking 27:1241-1277.
King, R.G., and A.L. Wolman (1996), "Inflation targeting in a St. Louis model of the 21st century",
Federal Reserve Bank of St. Louis Review 78:93-107.
King, R.G., and A.L. Wolman (1998), "What should the monetary authority do when prices are sticky",
in: J.B. Taylor, ed., Monetary Policy Rules (University of Chicago Press, Chicago, IL).
Lach, S., and D. Tsiddon (1992), "The behavior of prices and inflation: an empirical analysis of
disaggregated data", Journal of Political Economy 100:349-389.
Lach, S., and D. Tsiddon (1996), "Staggering and synchronization in price-setting: evidence from
multiproduct firms", American Economic Review 86(December): 1175 1196.
Lau, S.H.E (1996), "Aggregate pattern of time-dependent adjustment rules I: a game-theoretic analysis
of staggered versus synchronized wage setting", Economic Journal 106:1645-1658.
1048
J.B. Taylor
Lau, S.H.R (1997), "Aggregate pattern of time-dependent adjustment rules II: strategic complementary
and endogenous nonsynchronization", Working Papers in Economics and Econometrics, No. 317 (The
Australian National University, January).
Lebow, D.E., D.J. Stockton and W. Wascher (1995), "Inflation, nominal wage rigidity and the efficiency
of labor markets", unpublished paper (Federal Reserve Board).
Leeper, E.M., and C.A. Sims (1994), "Toward a modern macroeconomic model usable for policy analysis",
in: S. Fischer and J.J. Rotemberg, eds., NBER Macroeconomics Annual (MIT Press, Cambridge, MA)
81-118.
Levin, A. (1989), "The theoretical and empirical relevance of staggered wage contract models",
Ph.D. Dissertation (Stanford University).
Levin, A. (1990), "Monetary stabilization policy in a general equilibrium model with wage contracts",
Working Paper No. 90-hyphen;43 (University of California at San Diego, December).
Levin, A. (1991), "The macroeconnmic significance of nominal wage contract duration", Working Paper
No. 91-08 (University of California at San Diego, February).
Levy, D., M. Bergen, S. Dutta and R. Venable (1997), "The magnitude of menu costs: direct evidence
from large U.S. supermarket chains", Quarterly Journal of Economics 114:791-825.
Levy, D., S. Dutta, M. Bergen and R. Venable (1998), "Price adjustment at multiproduct retailers",
Managerial and Decision Economics, forthcoming.
Lucas, D.J. (1985), "Price and interest rate dynamics induced by multiperiod contracts", Working Paper
(Northwestern University).
Lucas, D.J. (1986), "Rigid wages as a transmission mechanism for monetary shocks", Working Paper
(Northwestern University).
Lucas Jr, R.E. (1972), "Expectations and the neutrality of money", Journal of Economic Theory 4:
103-124.
Lucas J1; R.E. (1973), "Some international evidence on output inflation tradeoffs", American Economic
Review 63:326-334.
Lucas Jr, R.E. (1976), "Econometric policy evaluation: a critique", Carnegie-Rochester Conference Series
on Public Policy 1:19-46.
Lucas Jr, R.E. (1996), "Nobel lecture: monetary neutrality", Journal of Political Economy 104:661 682.
Lucas Jr, R.E., and M. Woodford (1994), "Real effects of monetary shocks in an economy with sequential
purchases", Working Paper No. 4250 (NBER).
Marschak, J. (1953), Economic Measurement for Policy and Prediction, Studies in econometric method,
CoMes Commission for Research in Economics, Monograph No. 14 (Yale University Press, New
Haven, CT).
Matsukawa, S. (1986), "The equilibrium distribution of wage settlements and economic stability",
International Economic Review 27:415-437.
McCallum, B.T. (1982), "Macroeconomics after a decade of rational expectations: some critical issues",
Federal Reserve Bank of Richmond Economic Review 68:3-12.
McCallum, B.T. (1984), "A linearized version of Lucas's neutrality model", Canadian Journal of
Economics 17:138-145.
McCallum, B.T. (1988), "Robustness properties of a rule for monetary policy", Carnegie-Rochester
Conference Series on Public Policy 29:173-203.
McCallum, B.T. (1999), "Issues in the design of monetary policy rules", ch. 23, this Handbook.
McLaughlin, K.J. (1994), "Rigid wages?", Journal of Monetary Economics 34:383-414.
Mishkin, F.S. (1982), "Does anticipated monetary policy matter? an econometric investigation", Journal
of Political Economy 90:22-51.
Montgomery, E. (1983), "A note on aggregate dynamics and staggered contracts: a test of the importance
of spillover effects", Working paper 82-6 (Carnegie Mellon School of Urban and Public Affairs).
Nelson, E. (1997), "A framework for analyzing alternative models of nominal rigidities", unpublished
paper (Carnegie Mellon University).
Ch. 15:
1049
Ohanian, L.E., and A.C. Stockman (1994), "Short run effects of money when some prices are sticky",
Federal Reserve Bank of Richmond Economic Review 80:1~3.
Okun, A.M. (1981), Prices and Quantities: A Macroeeonnmic Analysis (Brookings Institution,
Washington, DC).
Parkin, M. (1986), "The output-inflation tradeoff when prices are costly to change", Journal of Political
Economy 94:200-224.
Phaneuf, L. (1987a), "Can contract-based models explain business cycles?", working paper (University
of Montreal).
Phaneuf, L. (1987b), "Propri6t6s dynamiques des modules du cycle/t contrats echelonn6s", Canadian
Journal of Economics 20:123-139.
Phaneuf, L. (1990), "Wage contracts and the unit root hypothesis", Canadian Journal of Economics
23:580-592.
Phelps, E. (1978), Disinflation without recession: adaptive guideposts and monetary policy,
Weltwirtschaftliches Archiv 114:783 809.
Phelps, E., and J.B. Taylor (1977), "Stabilizing powers of monetary policy under rational expectations",
Journal of Political Economy 85:163-190.
Prescott, E.C. (1975), "Efficiency of the natural rate", Journal of Political Economy 83:1229 1236.
Rankin, N. (1998), "Nominal rigidity and monetary uncertainty", European Economic Review 42:
185-199.
Roberts, J.M. (1995), "New Keynesian economics and the Phillips curve", Journal of Money, Credit and
Banking 27:975 984.
Roberts, J.M. (1997), "Is inflation sticky?", Journal of Monetary Economics 39:173-196.
Romer, D. (1996), Advanced Macroeconomics (McGraw-Hill, New York).
Rotemberg, J.J. (1982), "Sticky prices in the United States", Journal of Political Economy 90:1187-1211.
Rotemberg, J.J. (1987), "The new Keynesian foundations", in: S. Fischer, ed., NBER Macroeconomics
Annual 1987 (MIT Press, Cambridge, MA) 69 104.
Rotemberg, J.J. (1997), "A comment", Carnegie-Rochester Conference Series on Public Policy 47:
231-243.
Rotemberg, J.J., and M. Woodford (1997), "An optimization-based econometric framework for the
evaluation of monetary policy", in: B.S. Bernanke and J.J. Rotemberg, eds., NBER Macroeconomics
Annual (MIT Press, Cambridge, MA).
Rotemberg, J.J., and M. Woodford (1999), "The cyclical behavior of prices and costs", ch. 16, this
Handbook.
Rotwein, E., ed. (1955), David Hume's Writings on Economics (University of Wisconsin Press,
Madison, WI).
Rudin, J. (1987), "Diverse expectations in an empirical model: an extension of the staggered contracts
model", Chapter 2 of "Diverse expectations: policy and empirical implications", Ph.D. Dissertation
(Stanford University).
Sargent, T.J. (1976), "A classical maeroeconomic model for the United States", Journal of Political
Economy 84:207-237.
Sargent, T.J., and N. Wallace (1975), "'Rational' expectations, the optimal monetary instrument, and the
optimal money supply rule", Journal of Political Economy 83:241-254.
Sheshinski, E., and Y. Weiss (1988), "Staggered and synchronized price polices and multiproduct
monopolies", Working paper No. 24-78 (Foerder Institute for Economic Research, Tel-Aviv, July).
Slade, M.E. (1996), "Optimal pricing with costly adjustment: evidence from retail-grocery prices",
unpublished paper (University of British Columbia).
Stigler, G., and J. Kindahl (1970), The Behavior of Industrial Prices, NBER General Series, No. 90
(Columbia University Press, New York).
Stock, J.H., and M.W. Watson (1999), "Business cycle fluctuations in US macroeconomic time series",
ch. 1, this Handbook.
1050
JB. Taylor
Svensson, L.E.O. (1986), "Sticky goods prices, flexible asset prices, monopolistic competition, and
monetary policy", Review of Economic Studies 52:385~405.
Taylor, J.B. (1975), "Monetary policy during a transition to rational expectations", Journal of Political
Economy 83:1009-1021.
Taylor, J.B. (1979a), "Estimation and control of an econometric model with rational expectations",
Econometrica 47:1267 1286.
Taylor, J.B. (1979b), "An econometric business cycle model with rational expectations: some estimation
results", unpublished paper (Columbia University); condensed version reprinted in chapter 2 of
Taylor (1993a).
.
Taylor, J.B. (1980a), "Aggregate dynamics and staggered contracts", Journal of Political Economy
88:1-22.
Taylor, J.B. (1980b), "Output and price stability: an international comparison", Journal of Economic
Dynamics and Control 2:109-132.
Taylor, J.B. (1983), "Union wage settlements during a disinflation", American Economic Review 73:
981-993.
Taylor, J.B. (1985), "Rational expectations models in macroeconomics", in: K. Arrow and S. Honkapohja,
eds., Frontiers of Economics (Basil Blackwell, Oxford).
Taylol, J.B. (1986), "Improvements in macroeconomie stability: the role of wages and prices", in:
R.J. Gordon, ed., The American Business Cycle. Continuity and Change (University of Chicago Press,
Chicago, IL).
Taylor, J.B. (1993a), Macroeeonomic Policy in the World Economy: From Econometric Design to
Practical Operation (W.W. Norton, New York).
Taylor, J.B. (1993b), "Discretion versus policy rules in practice", Carnegie-Rochester Conference Series
in Public Policy 39:195-214.
Tsiddon, D. (1991), "On the stubborness of sticky prices", International Economic Review 32:69-75.
Tsiddon, D. (1993), "The (mis)behavior of the aggregate price level", Review of Economic Studies
60:889-902,
Warner, E.J., and R. Barsky (1995), "The timing and magnitude of retail store markdowns: evidence
from weekends and holidays", Quarterly Journal of Economics 110:321-352.
West, K.D. (1988), "On the interpretation of near random-walk behavior in GNP", American Economic
Review 78:202-209.
Woodford, M. (1995), "Comment on the quantitative analysis of the basic neomonetarist model", Journal
of Money, Credit and Banking 27:1277-1284.
Yun, T. (1994), "Monetary policy, nominal price rigidity and business cycles", Ph.D. Dissertation
(Department of Economics, University of Chicago).
Yun, T. (1996), "Nominal price rigidity, money supply endogeneity, and business cycles", Journal of
Monetary Economics 37:345-70.
Chapter 16
T H E C Y C L I C A L B E H A V I O R OF P R I C E S A N D C O S T S
JULIO J. ROTEMBERG
Harvard Business School
MICHAEL WOODFORD
Princeton University
Contents
Abstract
Keywords
1. Introduction: Markups and the Business Cycle
2. The cyclical behavior of markups
2.1. Cyclical behavior of the labor share
2.2. Corrections to the labor-share measure of real marginal cost
2.2.1. A non-Cobb-Douglas production function
2.2.2. Overhead labor
2.2.3. Marginal wage not equal to the average
2.2.4. Costs of adjusting the labor input
2.2.5. Labor hoarding
2.2.6. Variable utilization of capital
2.3. Alternative measures of real marginal cost
2.3.1. Intermediate inputs
2.3.2. Inventory fluctuations
2.3.3. Variation in the capital stock
2.4. The response of factor prices to aggregate shocks
2.5. Cross-sectional differences in markup variation
1052
1052
1053
1057
1058
1064
1064
1065
1069
1072
1076
1079
1081
1081
1084
1086
1086
1092
1093
1094
1094
1100
1104
1112
1113
1119
1119
1120
1052
1123
1125
1126
1129
1131
1131
Abstract
Keywords
JEL classification: E3, D4, D3
1053
1 Many authors instead define the "markup" or "price-cost margin" as (P- MC)/P. The two quantities
are obviouslymonotonic transformations of one another.
I054
JJ Rotembergand M. WoodJord
theory "a problem still remains: Why cannot businessmen defend their profit margins
against the threatened encroachment of costs by marking up their selling prices?" As
we shall see, a variety of theories of imperfectly competitive behavior by firms explain
why they may choose not to do so. Such imperfectly competitive behavior thus plays
an important role in accounting for the character of aggregate fluctuations.
Fluctuations in markups are an important factor, in our view, for a reason somewhat
different than that emphasized by Mitchell. In Mitchell's analysis, the squeezing of
profit margins late in booms is what brings the boom to an end, as reduced profitability
dampens investment demand and hence sales. This suggests that an improvement in
firms' power to set prices above marginal cost would extend the boom. But this neglects
the fact that firms cannot all raise their relative prices. Let the marginal cost of each
firm i be given by Pc(yi), where Yi is the quantity supplied and P is the general price
level. (Marginal costs are proportional to the general price level because the variable
factors of production are supplied at relative prices that depend upon the quantity
demanded of those factors.) Let us suppose furthermore that c(y) is an increasing
fimction, for the sort of reasons cited by Mitchell. Then an increase in the quantity
supplied by industry i, if not associated with any shift in the marginal cost schedule,
wilt be associated with an increase in marginal cost. In the case of an individual firm
or industry, this need not be associated with any change in markups; it might simply
be associated with an increase in the relative price P i / P 2.
But if we consider a uniform increase in the quantity produced by each sector, all
relative prices Pi/P would have to increase by the same amount and this is not possible.
In fact, in a symmetric equilibrium, one must have
1
-
c(Y),
(1.1)
where Y is the common (and hence aggregate) level of output, and/~ the common
(and hence average) markup. It follows from Equation (1.1) that an increase in output
Y is possible only insofar as either the real marginal cost schedule c shifts, or the
markup falls. If firms allow their markups to decline, this will mean a higher level of
equilibrium output than would otherwise be possible, given the current real marginal
cost schedule. Thus if markups decline in the later stages of economic expansions, as
Mitchell argues, this is not something that brings the expansion to an end; rather, it
makes the expansion stronger (and possibly more prolonged) than is justified by cost
conditions alone 3.
Equation (1.1) suggests that a useful question about fluctuations in aggregate activity
is to ask to what extent they result from variations in average markups, as opposed
2 For example, in the case of a competitive industry, the industry supply curve is simply given by
Pi - Pc(Fi). An increase in industry demand results in a movement up this curve, to a higher relative
Ch. 16:
1055
to shifts in the real marginal cost schedule of the typical firm. A related question
that has received more attention in the literature is the extent to which the historical
record actually suggests that markups were indeed low when output was high. This
question is related to the first because, if the data suggested that markups were constant
(or procyclical), independent movements in markups could not possibly account for
a significant portion o f output fluctuations. Thus, our survey o f empirical evidence
starts in Section 2 with the simpler question of the extent to which measured markups
are countercyclical. Since marginal cost cannot be measured directly, all o f these
measures are indirect and rely on theories o f the cost function facing individual firms.
Once one has made these assumptions, however, one need only make a few additional
suppositions to obtain an estimate o f the derivative of the function c with respect to Y.
This then allows one to infer the output fluctuations induced by changes in measured
markups. The result is that one obtains a decomposition of output in terms o f output
movements due to markup changes and output movements due to shifts in the marginal
cost schedule. We consider this decomposition in Section 3.
Section 4 is devoted to a brief survey o f models o f variable markups 4. The above
decomposition makes it clear that these models can serve two separate purposes. First,
they can affect the extent to which shifts in the marginal cost schedule affect output.
If, for example, reductions in marginal cost lower markups, their effect on output is
magnified. What is perhaps more interesting still is that these theories allow shocks
other than shocks to the marginal cost schedule to affect output as long as these
shocks affect markups. In particular, allowing for endogenous markup variations adds
a channel through which demand disturbances may affect output and employment.
This does not mean that the part of output variations that is due to shifts in the
real MC schedule is due to "supply shocks", and the part due to markup variations
with the part due to "demand shocks". There are various ways in which demand
disturbances might, in principle, shift the real MC schedule 5. Similarly, as we have
already mentioned, the models of endogenous markup variation discussed below imply
that "supply shocks" as well as "demand shocks" may cause markups to vary. For
example, in Rotemberg and Woodford (1996b) it is shown how an endogenous increase
in markups following an oil price increase can increase the contractionary impact o f
such a shock, even though the oil shock would also contract output (by shifting up the
real MC schedule) to a lesser extent in a perfectly competitive model.
4 See Rotemberg and Woodfurd (1995) for further discussion of several of the leading models, with
greater attention to the structure of general equilibrium models incorporating these mechanisms, and to
issues such as calibration and numerical solution of such models.
5 Well-knownproposals include nominal wage rigidity, as in Keynes (1936), as a result of which inflation
lowers the real wage and hence real marginal cost; and variations in the household labor supply schedule
due to wealth effects and intertemporal substitution effects, as in Barro's (1981) analysis of the effects
of government purchases. Evaluation of their importance is beyond the scope of this survey, though it
is important to remember that these proposals require that real wages fall (by as much as the marginal
product of labor) for output to expand.
1056
6 The crucial role of markup variations in explainingthe real effects of purely nominal disturbances is
stressed in particular by Kimball (1995).
Ch. 16:
1057
variations in desired markups - and not simply variations in markups resulting from a
discrepancy between actual and desired prices - are of some importance in accounting
for aggregate fluctuations. A consideration of the determinants of desired markups is
therefore required.
Our survey proceeds as follows. In the next section, we discuss the evidence on
cyclical variation in markups. Most of the evidence reviewed relates the slightly
procyclical behavior of real wages to cyclical movements in the marginal product o f
labor. The key issue in Sections 2.1 and 2.2 is whether, as a result of technical progress,
the marginal product of labor is as procyclical as real wages. While the marginal
product of labor cannot be measured directly, we provide a number of reasons to believe
that it is substantially more countercyclical than the relevant real wage so that, indeed,
markups are countercyclical. In Section 2.3 we consider measures of markup variation
that are not based on wage variations; these involve cyclical variations in the use of
intermediate inputs and in inventory accumulation. We then proceed to study responses
of markups to particular shocks. Insofar as we are able to identify nontechnological
disturbances, the analysis of markup changes is considerably simplified because we
do not have to worry about the effect of technical progress on the real marginal cost
schedule. Section 2 closes with an analysis of the differences in markup variations
across industries.
Section 3 then turns to the consequences of markup variations for macroeconomic
variables of interest. First we deal with the effect of markup changes on productivity
and profits. We show that, under a variety of circumstances, increases in output that are
caused by reductions in markups are associated with increases in profits and measured
productivity. Since both productivity and profits are known to be procyclical, this is
important in making sure that it is not implausible for changes in markups to be behind
movements in output. Section 3 concludes with a method for decomposing output
movements into those caused by shifts in the marginal cost schedule and movements
due to markup changes (which induce movements along particular marginal cost
curves). Section 4 is devoted to a survey of theories of markup variation, and Section 5
concludes.
7 Note, however,that the studies of individual industries discussed in Section 2.4 do attemptto measure
industry markups, rather than levels of real marginal cost.
1058
Y = F ( K , zH),
(2.1)
where K is the capital stock, H the number o f hours worked, and z an index o f laboraugmenting technical progress, then the markup o f price over marginal cost # is given
by 8
PZFH(K, zH)
/~ -
(2.2)
This equation provides an approach to the measurement of markup (or real marginal
cost) variations. It also highlights two reasons for the real marginal cost schedule
referred to in Section 1 to be upward-sloping. The first is that, holding constant other
determinants o f labor supply, the real wage must presumably rise to induce more people
to work. The second is that, if one makes the standard assumption that the production
function F is concave and one fixes both the capital stock and the state of technology z,
the marginal product FH is a decreasing function o f the labor input.
Whether or not typical increases in employment are in fact associated with markup
declines depends, however, upon whether they are associated with increases in K or z,
or decreases in the real wage W/P, sufficient to offset the effects o f the increase in the
labor input on FH. In general, real wages do not move countercyclically, and in fact,
there is clearlyprocyelical variation in the real wages received by individuals, once one
corrects for cyclical variation in the composition o f the workforce 9. This is the famous
8 Here and below, we use the notation FH to mean the partial derivative ofF with respect to its second
argument, the effective labor input zH, rather than with respect to H.
9 Kydland and Prescott (1988); Solon et al. (1994). This is not true of all industry wages, however. See
Chirinko (1980), Rotemberg and Saloner (1986) and Solon et al. (1994). For a review of this issue, see
Abraham and Haltiwanger (1995).
Ch. 16:
1059
criticism raised by Dunlop (1938) and Tarshis (1939) against the theory of aggregate
supply of Keynes (1936), which is essentially just Equation (2.2), under the assumption
of constant technology and a markup equal to one. Keynes (1939) recognized the
appeal of a hypothesis of countercyclical markup variation as a resolution of the puzzle,
which is the interpretation that we propose.
An alternative explanation, of course, is that real wages are procyclical because
fluctuations in activity are caused by variations in technical progress. [Real business
cycle models are sometimes criticized for predicting real wage movements that are t o o
procyclical, so that Kydland and Prescott (1988) take their findings as support for the
technology-shock hypothesis.] The need to correct for possible variations in the rate
of technical progress means that one needs to measure the variation in both the labor
input and in the quantity produced. The required calculation is especially simple if,
following Bils (1987), we specialize the production function (2.1) to the case
Y = g ( K ) ( z H ) a,
(2.3)
(2.4)
where SH is the labor share W H / P Y . Under these assumptions, markup variations are
simply the inverse of the variations observed in the labor share. In the case of a
Cobb-Douglas production function (or the slightly more general form assumed above),
marginal cost is proportional to average labor cost so that a valid measure of markup
variations is given by fluctuations in the ratio of price to "unit labor cost" W H / Y - a
measure of variations in price-cost margins often referred to in empirical studies of
business cycles such as those of Moore (1983)10.
We first consider the evidence regarding cyclical variation in this simple measure.
The price P with which firms are concerned is the price they receive for their products.
This means that the relevant labor share is not the ratio of labor compensation to
the value of output conventionally measured in national income accounts, but rather
the ratio to the revenue received by firms, which equals the value of output minus
indirect taxes 11. We consider cyclical variation in three different measures of this
labor share, for the whole economy, the corporate sector and the nonfinancial corporate
sector respectively. The first of these measures is less satisfactory than the others for
two reasons. First, it includes the government, many of whose services are not sold
in markets. Second, it includes income of proprietors in the denominator, and this
10 The ratio of price to unit labor cost is also used as an empirical proxy for the markup in studies such
as Phelps (1994).
11 The denominator is thus obtained by adding depreciation (the difference between GNP and NNP) to
the conventional concept of "national income".
1060
0.68
La
0.66
Nonfinancial
/~ v\/~ A
crpirate,
0.64
0.62
- 0.78
Overall
/[~j/!^<2~i ~
'
-I~,
k4,
~2 -C "4
,~
- 0.76
~ A.
~-~J'-~.~,
,:7
"
%,/j
,'.',
,, ~
v"
v.,'
v~',
0,74
- 0~72
0.60
- 0.70
0.58
- 0.68
Corporate
0.56
- 0.66
50
55
60
65
70
75
80
85
90
t2 Our sample stops in 1993 because, at the time these calculations were made, the pre-1960 data were
not comparable to the more recent revised NIPA data. The results from 1970 onwards were the same
for the two data sets, however.
Ch. 16:
0.76 -
1061
,,-7
,,
-~
,:
ii
i i
i
i
[ ]
i
i I
i
i I
!!
i
i i
i
i
i i
li r '
II
i
I
i
,,,,
'
''
i i
1 L
i i
:III
i
i
'11
,I
-0.8
i
1
0.74 i
1.0
q
i!
-0.6
i'
0.72 i
r
i
0.70 -
0.68
i
~
i
ii
i i
i
1
i
i
i
i
i
i
0.66
ii
ii
ii
ii
ii ii
i
~i
i
I
]
ii
ii
ii
ii
i
i
1
i i
I I
ii
Ii
ii
ii
~1
1
i
i
i
-0.4
-0,2
-0.0
50
55
60
65
70
75
80
85
F i g . 2. L a b o r s h a r e i n n o n f i n a n c i a l c o r p o r a t e s e c t o r a n d N B E R
90
recessions.
to remember that, for the labor share to be perfectly procyclical, its peaks ought to
be aligned with the business cycle peaks themselves. The labor share ought then to
decline between peak and trough and then increase during the recovery that takes place
after the trough. It is this latest implication of a procyclical labor share that is closest
to being true, as the labor share increases in the recoveries after the 1954, 1974 and
1982 recessions.
The actual correlation between the change in the labor share and a dummy variable
that takes the value of one between peaks and troughs and a value of zero otherwise
is 0.17, which suggests, albeit very weakly, that the labor share rises when output is
declining. However, the correlation between the labor share and the two quarter lagged
value of this dummy variable is -0.28. Thus, as the plot itself suggests, the labor share
tends to rise late in expansions and to fall late in recessions. This basic pattern: a
weak slightly negative relation between the labor share and contemporaneous cyclical
indicators and a much stronger positive relation between the labor share and slightly
lagged indicators of cyclical activity is what comes out of a more formal analysis as
well.
For this slightly more formal statistical analysis, we considered four indicators of
the business cycle. A popular indicator of this sort is obtained by detrending real GDP
using the Hodrick-Prescott filter and then using this detrended series to be a measure
of the business cycle. This is, however, a rather arbitrary procedure (since there is no
obvious reason to choose one value rather than another of the weighting parameter that
1062
determines the degree to which the trend is smoothed) 13. An alternative that we find
appealing is to follow Beveridge and Nelson (1981) and equate the "cyclical" level of
GDP at time t with the expected decline in GDP from time t onwards. This captures the
intuitive idea that cyclical movements are temporary, so that a cyclically low level of
output corresponds to a high expected rate of growth of output. The difficulty with this
approach is that it only becomes meaningful when one specifies an information set that
can be used to forecast GDP growth. Following on the steps of an extensive literature,
Rotemberg and Woodford (1996a) show that the linearly detrended level of hours spent
working in nonagricultural establishments and the ratio of consumer expenditures on
nondurables and services to GDP are particularly useful in this respect.
A simpler cyclical indicator is the linearly detrended level of hours worked. We
prefer to linearly detrend hours rather than GDR since once a trend is included in the
regression, the Dickey-Fuller test strongly rejects the hypothesis that the logarithm
of hours worked in non-agricultural establishments has a unit root. This measure of
detrended hours is in fact one of the main components of the Rotemberg-Woodford
measure of forecastable output movements; low levels of hours worked (which are
closely related to high unemployment), imply that output can be expected to grow and
are thus a good indicator of recessions. For purposes of comparison, we also consider
an hours series that has been detrended using the Hodrick-Prescott filter.
The first three rows of Table 1 report correlations of our three measures of the
labor share with the various cyclical indicators, over the sample period 1947:1 through
1993:1. The first column shows correlations with the predicted declines in output
over 12 quarters considered in Rotemberg and Woodford (1996a), the second column
shows the correlations with the Hodrick-Prescott filtered level of output, the third
column uses linearly detrended hours while the last uses the hours series detrended
using the H - P filter. Except the correlations with linearly detrended hours (which are
small and positive), the other correlations are small and negative; suggesting weak
countercyclical movements in the labor share.
These results are consistent with Boldrin and Horvath (1995), Gomme and
Greenwood (1995) and Ambler and Cardia (1998) who also report negative correlations
of the labor share with output. The correlations they report are larger in absolute
magnitude because they use the H - P filter to detrend the labor share as well as to
detrend output. Using the H - P filter to detrend the labor share seems problematic,
however, because the large movements at relatively high frequencies of the resulting
"trend" displayed in Figure 1 are difficult to interpret 14.
The last twelve rows of Table 1 report the correlations of the labor share in the
nonfinancial corporate sector with leads and lags of the four cyclical indicators. The
correlations with the lags are uniformly positive and attain their biggest value when the
13 For further discussion of the properties of this filter, see King and Rebelo (1993).
14 We also considered labor shares detrended with a linear trend. This had only a negligible effect on
our results.
Ch. 16:
1063
Table 1
Correlations of selected variables with cyclical indicators a
Predicted declines
in GDP
H-P filtered
GDP
Linearly
detrended hours
H-P filtered
hours
Long sample
Overall
-0.070
0.095
0.055
-0.023
Corporate
-0.080
-0.188
0.031
-0.044
Nonf. Corp.
-0.014
-0.158
0.072
-0.014
Private
-0.009
-0.192
0.178
-0.192
Overall
-0.230
-0.403
-0.189
-0.015
Corporate
-0.077
-0.273
-0.010
0.064
0.066
-0.169
0.103
0.184
-0.334
-0.466
-0.293
-0.156
Sample: 1969:1-1993:1
Nonf. Corp.
Private
Correlation of private labor share with leads and lags of cyclical indicator
Lead six quarters
-0.437
-0.108
-0.218
-0.136
-0.521
-0.579
-0.176
-0.312
-0.211
-0.270
-0.406
-0.276
-0.5828
-0.360
-0.461
-0.314
-0.564
-0.429
-0.454
-0.304
-0.509
-0.477
-0.407
-0.256
-0.157
-0.283
-0.100
0.015
0.026
0.110
0.063
0.162
0.075
0.023
0.180
0.270
0.149
0.138
0.258
0.346
0.177
0.194
0.303
0.388
0.213
0.222
0.317
0.406
aThe long sample for all correlations except those involving either the labor share in the private sector
or predicted declines in private GDP is 1947:1 to 1993:1. The sample for the correlations involving
the labor share in the private sector starts in 1952:1. That for the correlations of predicted output
declines with the other labor share starts in 1948:3 because these predicted declines are drawn from
Rotemberg and Woodford (1996a). The correlations with leads and lags of output are based on data
from 1969:1 to 1993:1.
1064
cyclical indicator is lagged four quarters. Thus a high level of activity is associated with
subsequent increases in the labor share. Interestingly, this result also extends to the case
where we study the cross-correlogram of H-P filtered output and the H-P filtered labor
share. The correlations of the labor share with the leads of our cyclical indicators are
uniformly negative. This means that the labor share peaks before the peak in hours.
2.2. Corrections to the labor-share measure o f real marginal cost
While the labor share (or equivalently, the ratio of price to unit labor cost) is a familiar
and easily interpretable statistic, it represents a valid measure of markup variations only
under relatively special assumptions. In this section, we briefly discuss a number of
corrections to this measure that would arguably be required to obtain a more realistic
measure of real marginal cost. As we shall see, several of these corrections imply that
real marginal cost is more procyclical than the labor share. For any of several reasons,
then, the measurements discussed above may understate the degree to which cyclical
variations in output and employment are due to markup variations as opposed to shifts
in the real marginal cost schedule. We take the possible corrections up in sequence.
2.2.1. A non-Cobb-Douglas production function
Suppose that the aggregate production function is of the general form (2.1), but that
F is not necessarily of the Cobb-Douglas form [or, more precisely, isoelastic in the
labor input, as in Equation (2.3)]. Equation (2.2) can still equivalently be written
# = t/Hs~1,
(2.5)
where t/H --= zHFH(K, zH)/F(K, zH) is the elasticity of output with respect to the
(effective) labor input. [Equation (2.5) reduces to (2.4) in the case of a constant
elasticity.]
The effect of the additional variable factor in Equation (2.5) depends upon the nature
of cyclical variations, if any, in the elasticity of output with respect to the labor input.
In the case that F exhibits constant returns to scale, the elasticity t/H can be expressed
as a function of the effective labor-to-capital ratio, zH/K, or equivalently as a function
of the output-to-capital ratio:
t/~q = t/H(y),
(2.6)
where y - Y/K. In the case that the elasticity of substitution between capital
and (effective) labor inputs is less than one, the function t/H(y) is monotonically
decreasing. This would seem the most likely direction of deviation from the CobbDouglas case (a constant elasticity of substitution exactly equal to one), as the CobbDouglas specification is widely regarded as a reasonable representation of long-run
substitution opportunities, whereas short-run factor substitutability (which is relevant
1065
for the present calculation) might well be less (for example, because technology is
"putty-clay"). In this case, because y is a procyclical variable (this follows directly from
the fact that the capital stock evolves slowly relative to the length of business-cycle
fluctuations), Equation (2.6) implies that the additional factor t/H in Equation (2.5)
imparts additional countercyclical variation to the implied markup series, roughly
coincident with the cyclical component of output or hours. A correction of this kind
thus leads to the conclusion that markups fall more in booms than is suggested by the
simple labor share measure, and that markup declines coincide more closely in time
with increases in output and hours.
The size of this correction can be quantified as follows. Assuming constant returns
to scale, the elasticity of t/H with respect to y is given by a - (1 - e K 1 ) ( r ] H 1 - - 1), where
c ~ is the (Hicks-Allen) elasticity of substitution between capital and labor inputs. A
log-linear approximation to the markup series implied by Equation (2.5) is then given
by
= @ - s~4,
(2.7)
where hats denote deviations of the logarithm of a stationary variable from its average
value. A quantitative estimate of the elasticity a requires values for exH and for the
average value of OH (i.e., the value of this elasticity in the case of the "steady-state"
factor ratio arotmd which one considers perturbations). Using Equation (2.5), the latter
parameter may be calibrated from the average labor share, given an estimate of the
average markup, resulting in a = (1 - e~q)(l~-lsTi 1 - 1). (In this last expression, all
symbols refer to the average or steady-state values of the variables.) With a markup
kt near one, a labor share of 0.7 and an elasticity o f substitution eK~ of 0.5, this
formula gives a value of a equal to -0.4. Table 2 reports the resulting correlations
between fi and predicted declines in GDP for markups based on both the nonfinancial
corporate and the private labor shares. Not surprisingly, markups are now much
more cotmtercyclical. However, the contemporaneous correlation of the markup with
detrended output is still smaller in absolute value than the correlation with lagged
output. Also, as in the case where we do not adjust the labor share, the correlations with
leads of output are greater than the contemporaneous correlation. These are actually
positive for output led more than 3 quarters 15.
2.2.2.
Overhead
labor
15 As is true of all the results of Table 2, similar results obtain when we use detrended hours as our
cyclical indicator.
1066
Table 2
Correlation of markup based on private labor share with leads and lags of expected declines in GDP a
a = -0.4,
b,c:0
b - -0.4,
a,c=0
c = 4,
a,b=0
c = 8,
a,b=0
0.355
0.370
0.136
-0.058
0.323
0.342
0.067
-0.169
0.256
0.289
0.048
-0.316
0.135
0.189
-0.203
-0.478
-0.001
0.075
-0.321
-0.594
-0.163
-0.050
-0.418
-0.670
Contemporaneous
-0.402
-0.212
-0.372
-0.542
-0.504
-0.312
-0.235
-0.319
-0.522
-0.344
-0.162
-0.181
-0.503
-0.337
-0.124
-0.095
-0.451
-0.301
-0.066
-0.001
-0.355
-0.226
-0.003
0.079
-0.278
-0.164
0.011
0.110
a Markup is based on Equations (2.14) and (2.15) and uses the labor share in the nonfinancial corporate
business sector.
that e a c h f i r m ' s p r o d u c t i o n f u n c t i o n 16 is o f t h e f o r m Y = F(K, z ( H - / z / ) ) , w h e r e F
is h o m o g e n e o u s o f d e g r e e o n e as b e f o r e , a n d f / ~> 0 r e p r e s e n t s " o v e r h e a d l a b o r "
that m u s t b e h i r e d r e g a r d l e s s o f the q u a n t i t y o f o u t p u t that is p r o d u c e d . N o t e that
a n o v e r h e a d l a b o r r e q u i r e m e n t i m p l i e s i n c r e a s i n g r e t u r n s ( a v e r a g e cost e x c e e d i n g
m a r g i n a l cost), a l t h o u g h m a r g i n a l cost r e m a i n s i n d e p e n d e n t o f scale 17.
16 Once we depart from the assumption of constant returns to scale, it is important to distinguish between
firm production functions and the relation that exists between aggregate inputs and outputs. We now
assume that each firm is the sole producer of a differentiated good, so that the overhead costs cannot be
reduced by simply concentrating all production in a single firm. In a symmetric equilibrium, where the
same quantity is produced of each good using the same factor inputs, then this equation also indicates
the relation that exists among aggregate output and aggregate factor demands.
17 There are other ways of modeling increasing returns. In particular, one might suppose that marginal
cost declines with output; an econometric specification of this kind is estimated, for example, by Chirinko
and Fazzari (1994). The notion that marginal cost declines with output is problematic, however. For many
firms, increasing output involves an increase in either the number of machines that are employed or an
increase in the number of hours for which a given number of machines are used. Both of these seem
inconsistent with declining marginal cost since more efficient machines would presumably be used first.
More generally, firms whose technology has a declining marginal cost over some range would benefit by
bunching production so that their plants are idle some of the time, and output, when positive, is always
at a level sufficiently large that marginal cost is not declining in output.
Ch. 16:
1067
Replacing Equation (2.1) by this, implies that Equation (2.5) should become instead
where t;H now refers to the elasticity of output with respect to the effective nonoverhead labor input. Under this definition, ~,/ is again constant in the case that
F takes the Cobb-Douglas form, and countercyclical if the elasticity o f substitution
between capital and labor is less than one. The new factor in Equation (2.8), H--~' is a
monotonically decreasing function of H i f / 7 / > 018. Allowing for overhead labor thus
provides a further reason to regard markups as more countercyclical than is indicated
by the labor share alone. A similar conclusion is reached if one assumes fixed costs
in production that do not take the form of overhead labor alone, e.g., if one assumes
a production function o f the form
Y= F(K, zH)
45,
(2.9)
The elasticity b is obviously non-positive. Its size depends on the average fraction o f
labor which constitutes overhead labor. A useful bound on this can be obtained by
relating So to the degree o f returns to scale. Let the index o f returns to scale p be
defined as the ratio of average cost to marginal cost o f production. Measured at the
steady-state factor inputs, one obtains
p = 1 + t//4
so that instead o f calibrating So, one may equivalently calibrate p. In terms of this
parameter, we obtain t/H = #sH -- (p -- 1) for the steady-state elasticity o f output with
respect to non-overhead labor, and b = - ( / 9 - 1)/[#SH -- (p -- 1)].
It is easily seen that one must have p ~< #, in order for there to exist non-negative
profits in the steady state. This allows one to bound the possible size o f the elasticity
18 Here we assume that the overhead labor requirement is acyclical. This depends upon an assumption
that entry of either firms or plants is slow, as in Rotemberg and Woodford (1995) and Ambler and Cardia
(1998), and so can be neglected at business cycle frequencies. The consequences of variable entry are
considered further below, in Section 3.
1068
b, given an estimate o f the average markup. On the other hand, the same consideration
provides a reason for supposing that overhead costs are non-negligible, if one believes
that prices do exceed marginal cost on average. For with constant returns to scale,
prices higher than marginal cost would imply the existence of pure profits (in addition
to the competitive return to capital); for example, a markup o f 25% (bt = 1.25) would
imply that pure profits should make up 20% o f total revenues. This is rather large given
the scant evidence for the existence o f pure profits in US industry. Indeed, Hall (1988)
finds (using stock market re~rns to construct a user cost for capital) that pure profits
in US industry are close to zero. It furthermore makes sense that profits should be zero
in the steady state, due to entry, which one should expect to eliminate persistent profits
in the long run, even if entry does not respond quickly enough to eliminate cyclical
fluctuations in profits. If we assume this, we can impose p =/~, so that there is only
a single parameter to calibrate, that describes both the degree o f returns to scale and
the degree o f market power. With # = 1.25 and a labor share o f 0.7, the parameter b
is then -0.4. Table 2 shows that, even letting a equal zero, such a value o f b leads to
markups that are strongly countercyclical though the correlations with lagged output
remain higher in absolute value 19.
The significance that one attaches to such findings obviously depends upon the size
of the average markup (or degree o f returns to scale) that one is willing to assume.
Here it is worth remarking that a value of ~ equal to 1.6 need not mean that any
individual firm marks up its costs by 60%. The reason for this is that firms do not
just mark up their labor costs but also their materials cost. To see what this implies
about the markup, suppose that, as in Rotemberg and Woodford (1995), materials are a
fixed proportion SM o f aggregate output while value added constitutes only a fraction
(1 - S M ) o f total costs. The marginal cost o f producing one unit o f gross output is
then
(1 - sM) W
zFH
q- SM,
and the markup o f the price o f gross output over total marginal cost ktc is given by
1
btGO
- (1
1
--SM)~
+SM,
(2.10)
where tt vA is the "value-added markup" that satisfies Equation (2.2). If the materials
share equals 0.6 (as is typical of US manufacturing), then a/~vA o f 1.6 [the "baseline
19 Rotemberg and Woodford (1991) use a variant of this method to construct series for markup changes
using aggregate US data. Assuming an average markup of 1.6 and an elasticity of substitution equal to 1
(their baseline case), they find that markups fall by about 1% when hours increase by 1%. The constructed
markup series is also strongly negatively correlated with fluctuations in aggregate hours worked. Portier
(1995) uses the same method on French data, and assumes an average markup of 1.373 and an elasticity
of substitution equal to 1. His estimates imply that a 1% increase in GDP is associated with about a
1.5% reduction in markups. Thus markups would appear to be more countercycliealfor France (a finding
that is especially striking given the lower assumed returns to scale).
1069
case" o f Rotemberg and Woodford (1991)] requires that the typicalfirm "s price be only
18% higher than its marginal cost.
A related correction would assume, instead o f overhead labor, a "setup cost" for each
employee, as is considered in Basu and Kimball (1997). Suppose that the production
function is Y = F ( K , z(h - h)N)), where now N represents the number of employees
and h the number of hours worked by each. We again assume that F is homogeneous o f
degree one; the "set-up cost" h > 0 represents a sort o f per-employee fixed cost. (The
observed preference for full-time employees observed in many lines o f work makes
the existence o f such costs plausible 20.) If we consider the marginal cost of increasing
output solely on the employment margin (holding fixed hours per week), we again
obtain Equation (2.8), but with H and { / r e p l a c e d by h and h in the first factor. We
correspondingly again obtain Equation (2.9), but with if/replaced by h. Since hours
per employee are also a strongly procyclical variable, the first factor in Equation (2.8)
is again a source of further countercyclical movement in implied markups. Basu and
Kimball suggest that So = 0.25 should be an upper bound on the importance of such
set-up costs (as full-time wage premia should otherwise be larger); but this value would
still allow the elasticity in Equation (2.9) to be as large as b = -0.3.
2.2.3. Marginal wage not equal to the average
Thus far, we have assumed wage-taking behavior on the part o f firms, meaning that
they regard themselves as being able to hire additional hours o f work, at the margin, at
a wage which is also the wage paid for each o f the hours that they do hire - so that the
relevant marginal wage is also the average wage that is paid. Suppose, however, that
this is not true, and that the firm's wage bill is W ( H ) , a function that is increasing, but
not necessarily linear in H 2~. In this case, marginal cost depends upon the marginal
wage, W t ( H ) , so that Equation (2.5) becomes
# = 09-1 tlHS~i1,
(2.1 1)
where o) = H W t ( H ) / W ( H ) is the ratio o f the marginal wage to the average wage. This
might vary cyclically for several reasons.
One reason might be monopsony power in the labor market. Suppose that each firm
faces an upward-sloping firm-specific labor supply curve, and takes this into account
in its hiring and production decisions. (The wage that the firm must pay may also
20 One might ask, if such costs exist, why finns do not minimize costs by hiring all of the time of
those employees that they hire at all. The answer must be that finns face a wage schedule that is not
simply linear in the number of hours worked by a given employee, as discussed below. Note that this
hypothesis about individual wages is of no consequence for the marginal cost calculation considered in
this paragraph.
21 A marginal wage that is increasing in the number of hours hired is, for example, allowed for in such
studies as Abel (1978), Shapiro (1986), Bils (1987), and Basu and Kimball (1997).
1070
depend upon other variables such as the overall level of employment in the economy,
but these factors are taken as given by the individual firm, and can simply be treated
as time-variation in the location of the firm-specific labor supply curve.) If w(H) is the
wage that the firm must pay if it hires H hours of work, then W(H) = Hw(H), and
-l where CHw is the elasticity of the firm-specific labor supply curve. This
co = 1 +CHw,
might be either increasing or decreasing with increases in hours hired by the firm.
The most plausible assumption, however, would probably be that the elasticity of labor
supply decreases as the hours hired by the firm increase (it is hard to induce people
to work more than a certain number of hours, even at very high wages, while on
the other hand the opportunity cost of their time tends not to fall below a certain
level even when the number of hours worked is small). Under this assumption, the
factor m is an increasing function of H , and Equation (2.9) again holds, with b < 0.
This would imply that real marginal costs would actually be more procyclical (and
markups more countercyclical) than would be suggested by consideration only of the
terms in Equation (2.5).
Alternatively, one might imagine that firms first hire a certain number of employees,
and that they initially contract with them about a wage schedule which determines the
wage as a function of hours worked. Subsequently, perhaps after receiving additional
information about current demand conditions, the firms determine the hours of work.
If all employees are asked to work the same number of hours at this stage, we
may interpret W(H) in Equation (2.11) as the wage schedule negotiated with each
employee. Now if the number of employees is chosen ex ante so as to minimize
the cost of the number of hours that the firm expects to use, then ex ante expected
hours per worker will be the level H* that minimizes the average wage W ( H ) / H 22. At
this point, the marginal wage should equal the average wage, and (assuming a unique
minimum) in the case of small fluctuations in H around the value H*, co should be
increasing in H. Again this would imply markups more countercyclical than would be
suggested by Equation (2.5).
Most observed wage contracts do not involve wages that increase continuously with
the number of hours that the employee is asked to work. On the other hand, if one
supposes that contractual wages are not the true shadow price of additional labor to a
firm, because of the presence of implicit contracts of the kind assumed, for example,
by Hall (1980), then one might suppose that the true cost to the firm rises in proportion
to the employee's disutility of working, even if the wages that are paid in the current
period do not. This would be a reason to expect the effective wage schedule W(H) to
be convex, so that the above analysis would apply.
Bils (1987) observes that in many industries, a higher wage is paid for overtime
hours (i.e., hours in excess of 40 hours per week). He thus proposes to quantify the
extent to which the marginal wage rises as firms ask their employees to work longer
22 This conclusion depends upon an assumption that only person-hours enter the production ftmction,
rather than employmentor hours per employeemattering separately.
1071
hours, by measuring the extent to which the average number o f overtime hours per
employee, V, rises with increases in the total number o f hours worked per employee
H , and then assuming that W ( H ) = wo[H + p V ( H ) ] , where w0 is the straight-time
wage a n d p is the overtime p r e m i u m (0.5 according to the US statutory requirement)23.
For example, he finds that when average hours per employee rise from 40 hours p e r
week to 41 hours, the average number o f overtime hours worked per employee rises b y
nearly 0.4 hours, while when they rise from 41 to 42 hours per week, overtime hours
rise by another 0.5 hours. This increase in the fraction o f hours that are overtime
hours as average hours increase means not only that the marginal wage exceeds the
average wage, but that the ratio o f the marginal wage to the average wage rises
as hours increase. A s s u m i n g p = 0.5, Bils finds that an increase in average hours
from 40 to 41 hours increases the average wage b y about 0.5%, but increases the
marginal wage by 4.6%. O n average, he finds that the factor co in Equation (2.11)
has an elasticity o f 1.4 with respect to variations in average hours a4. Thus a loglinear approximation to Equation (2.11) is again o f the form (2.9), where in Bils'
w o r k / 2 / r e f e r s to fluctuations in average hours per worker 25, and b = -1.4.
Since average hours worked in US manufacturing are strongly procyclical, taking
into account this factor m a k e s the implied markup significantly more countercyclical.
Indeed, when Bils regresses his constructed markup series [using Equation (2.9)]
on a measure o f cyclical employment 26, he finds that markups decline, on average,
by 0.33% for each one-percent increase in employment. O f this cyclical variation, a
0.12% decline is implied by the increase in the labor share (which is mildly procyclical
in his sample), while the remaining 0.21% decline comes from the increase in the ratio
o f the marginal wage to the average wage.
One may question whether the statutory premium p a i d for overtime hours represents
a true cost to the firm; some argue, for example, that the opportunity to work overtime
is in fact dispensed as a reward for exemplary behavior at other times. Bils answers
23 The fact that V(H) is modeled as a fraction that rises continuously with H, rather than being zero
for all H ~<40 hours per week and one for all H > 40 hours per week requires that not all employees
work the same number of hours. The nature and consequences of this heterogeneity are not explicitly
modeled.
24 This average elasticity is slightly smaller than the elasticity of 1.6 indicated by the figures given in
the text relating to an increase from 40 to 41 hours per week.
25 Bils studies the variations of production-worker hours in manufacturing, and computes the marginal
cost of increasing output through an increase in production-worker hours only, holding other inputs
fixed, including non-production-worker hours. Thus in Equation (2.9), s~4 refers to fluctuations in the
share of production-worker wages. Because he assumes a production function which is isoelastic in
production-worker hours, holding fixed the other inputs, a = 0 in his calculations.
26 His cyclical indicator is the difference between current production-worker employment and a moving
average of that series. Note that Bils does not assume, as in the simple analysis above, that employment
is fixed in advance and that all short-run variation in hours occurs on the hours-per-employee margin.
In fact, in his "second method" of computing the cyclical variability of the marginal wage, he explicitly
considers substitution between the employment and hours-per-employee margins.
1072
this objection by pointing out that if one assumes that because of sophisticated implicit
contracts, the true cost to the firm is proportional to the worker's disutility of working
o(H), then one might well obtain estimates of the degree of procyclical movement
in the ratio of the marginal wage to the average that are as large as those obtained
using his method. Under the assumption suggested above about the steady-state level
of hours, the coefficient b in Equation (2.9) would in that case equal -o"/1t*ol, or
-C,w, where eHw is now the Frisch (or intertemporal) elasticity of labor supply by a
wage-taking household in a competitive spot market. A value of b less negative than
Bils' value of -1.4 would then be obtained only if one assumed preferences implying
an elasticity of labor supply greater than 0.7, whereas many microeconomic studies of
labor supply estimate a lower elasticity than that.
27 This assumption is more appealing in the case that H is interpreted to refer solely to productionworker hours, as in Bils's (1987) work, rather than total hours.
Ch. 16:
1073
for the necessary changes in the inputs used in the adjustment process at both dates
t and t + 1. In this case, Equation (2.5) becomes 28
/x = g2 1 r#HSTs1,
(2.12)
where
2t =
1 + K't {[O(rHt) +
Wt
rHtOt(]lHt)]
-- E t [ R t , t + l
2 10, (]/Ht+l)] },
'Y1ct+l]/~/t+
(2.13)
in which in turn Ym - H t / H t 1, Yrt = tot~tot-1 29, and Rt,t+l is the stochastic discount
factor by which firms discount random income at date t + 1 back to date t. [Here we
have written Equation (2. t 3) solely in terms o f variables that we expect to be stationary,
even i f there are unit roots in both H and w, to indicate that we expect Y2 to be a
stationary random variable.
I f ~b is strictly convex (i.e., i f there are non-zero adjustment costs), the cyclical
variation in the factor g2 changes the nature o f implied markup fluctuations. Because
~ is positive when the labor input is rising and negative when it is falling, 12 should be
a procyclical factor, though with a less exact coincidence with standard business cycle
indicators than the cyclical correction factors discussed thus far. I f we take a log-linear
approximation to Equation (2.13), near a steady-state in which the variables H , to~w, yx,
and R are constant over time, we obtain
D, =
~[~'H, -/3E~9~,+,],
(2.14)
where here the coefficient c > 0 denotes q~"(1) times the steady state value o f to~w,
and fl denotes the steady-state value o f Ryr, the discount factor for income streams
measured in units o f the adjustment-cost input. This can then be substituted into the
log-linear approximation to Equation (2.12),
= a ~ - s 7 4 - {2,
(2.15)
28 In this equation, sH refers to wH/PY as before. In order for this to correspond to labor compensation
as a share of value added, one must assume that the adjustment-cost inputs are not purchased from outside
the sector of the economy to which the labor-share data apply. However, to a first-order approximation,
it does not matter whether the adjustment costs are internal or external, as discussed below.
29 More generally, we shall use the notation Yxtto denote the growth rate xt/xt_l, for any state variable x.
1074
to fall in the future), and correspondingly low when they are temporarily low. Thus,
it tends to increase the degree to which implied markups are countercyclica130.
More precisely, the factor g2 tends to introduce a greater negative correlation
between measured markups and future hours. Consider, as a simple example, the case
in which hours follow a stationary AR(1) process given by
~It = p~It l -}- ~t,
where 0 < p < 1, and c is a white-noise process. Then ~2t is a positive multiple of
f / t - )~/~t-l, where )~ ~ (1 -fl(1 - p ) ) - l , and cov(g2t, Art+j)is of the form C(1 -)~p)pJ for
all j ~> 0, where C > 0, while it is of the form C(1 - ) , p 1)p4 for all j < 0. One
observes (since p < )~ < 1/p) that the correlation is positive for all leads j ~> 0,
but negative for all lags j < 0. Thus this correction would make the implied markup
series more negatively correlated with leads of hours, but less negatively correlated
with lags of hours. The intuition for this result is that high lagged levels of hours
imply that the current cost of producing an additional unit is relatively low (because
adjustment costs are low) so that current markups must be relatively high. Since, as
we showed earlier, the labor share is more positively correlated with lags of hours
(and more negatively correlated with leads of hours) this correction tends to make
computed markup fluctuations more nearly coincident with fluctuations in hours. To
put this differently, consider the peak of the business cycle where hours are still rising
but expected future hours are low. This correction suggests that marginal cost are
particularly high at this time because there is little future benefit from the hours that
are currently being added.
The last two columns of Table 2 show the effect of this correction for c equal to 4 and
8 while fi is equal to 0.99. To carry out this analysis, we need an estimate of Et~m+l.
We obtained this estimate by using one of the regressions used to compute expected
output growth in Rotemberg and Woodford (1996a). In particular, the expectation at t
of Ht+l is the fitted value of a regression of/~/t+i on the values at t and t - 1 of/~/, the
rate of growth of private value added and the ratio of consumption of nondurables and
services to GDE Subtracting the actual value of/~/t from this fitted value, we obtain
Et)'m+l. This correction makes the markup strongly countercyclical and ensures that
the correlation of the markup with the contemporaneous value of the cyclical indicator
is larger in absolute value than the correlation with lagged values of this indicator. On
the other hand, the correlation with leads of the indicator is both negative and larger
still in absolute value, particularly when c is equal to 8.
The same calculations apply, to a log-linear approximation, in the case that the
adjustment costs take the form of less output from a given quantity of labor inputs.
30 Even though they allow for costs of changing employment, Askildsen and Nilsen (1997) do not
find any industries with countercyclicalmarkups in their study of Norwegian manufacturing industries.
However, their adjustment-cost parameter is often estimated to have the wrong sign and one would
expect the markups computed on the basis of these estimates to be procyclical.
Ch. 16:
1075
Suppose that in the above description o f production costs, H refers to the hours that
are used for production purposes in a given period, while Hq~ indicates the number o f
hours that employees must work on tasks that are created by a firm's variation o f its
labor input over time. (In this case, t =_ w.) Equations (2.12) and (2.13) continue to
apply, as long as one recalls that H and sH now refer solely to hours used directly in
production. Total hours worked equal AH instead, and the total labor share equals AsH,
where A = 1 + q~(yu). But in the log-linear approximation, we obtain zi = 0, and so
Equations (2.14) and (2.15) still apply, even if )/t and s~r refer to fluctuations in the
total labor inputs hired by firms.
A more realistic specification of adjustment costs would assume costs of adjusting
employment, rather than costs of adjusting the total labor input as above 31. Indeed,
theoretical discussions that assume convex costs o f adjusting the labor input, as above,
generally motivate such a model by assuming that the hours worked per employee
cannot be varied, so that the adjustment costs are in fact costs o f varying employment.
In the data, however, employment variations and variations in total person-hours are
not the same, even if they are highly correlated at business-cycle frequencies. This
leads us to suppose that firms can vary both employment N and hours per employee
h, with output given by F(K, zhN), and that costs o f adjusting employment in period t
are given by lftNtO(Nt/Nt 1)- If, however, there are no costs o f adjusting hours, and
wage costs are linear in the number o f person-hours hired Nh, firms will have no need
ever to change their number o f employees (which is clearly not the case). If, then,
one is not to assume costs o f adjusting hours per employee 32, one needs to assume
some other motive for smoothing hours per employee, such as the sort o f non-linear
wage schedule discussed above. We thus assume that a firm's wage costs are equal to
W(h)N, where W(h) is an increasing, convex function as above.
One can then again compute the marginal cost o f increased output at some date,
assuming that it is achieved through an increase in employment at that date only,
holding fixed the number o f hours per employee h at all dates, as well as other inputs.
One again obtains Equation (2.12), except that the definition o f g2 in Equation (2.13)
must be modified to replace YH by YN, the growth rate of employment, throughout.
[In the modified Equation (2.13), w now refers to the average wage, W(h)/h.]
Correspondingly, Equation (2.15) is unchanged, while Equation (2.14) becomes
~'~t = C[YNt -- ~Et YNt+l ],
(2.16)
31 Bils and Cho (1994) assume a convex cost of adjusting the employee-to-capital ratio, interpreting this
as a cost of changing the organization of production, rather than a cost of hiring and firing employees.
Because most variations in the employment-to-capital ratio at business-cycle frequencies are due to
variations in employment, the consequences of such a specification are similar to those of the more
familiar assumption of convex costs of changing the number of employees.
32 Studies that estimate separate adjustment costs for variations in employment and in the number of
hours worked per employee, such as Shapiro (1986), tend to find insignificant adjustment costs for
hours.
1076
Thus one obtains, as in the simpler case above, a correction to Equation (2.5) that
results in the implied markup series being more countercyclical (since employment is
strongly procyclical, just as with the total labor input).
Alternatively, one could compute the marginal cost o f increased output, assuming
that it is achieved solely through an increase in hours per employee, with no change
in employment or in other inputs. In this case, one obtains again Equation (2.11),
but with H everywhere replaced by h in the first factor on the right-hand side.
There is no contradiction between these two conclusions. For the right-hand sides o f
Equations (2.11) and (2.12) should be equal at all times; cost-minimization requires
that
W'(ht) = wt+tCt[~)(Ymt)+ YNtO'(YNt)] -Et[Rt,t+l~:t+lY~t+l
2
, (YNt+l)]},
(2.17)
which implies that g2 = ~o. Condition (2.17) is in fact the Euler equation that Bils
(1987) estimates in his "second method" of determining the cyclicality of the marginal
wage; he uses data on employment and hours variations to estimate the parameters o f
this equation, including the parameters of the wage schedule W(h)33. An equivalent
method for determining the cyclicality o f markups would thus be to determine the
importance o f employment adjustment costs from estimation o f Equation (2.17), and
compute the implied markup variations using Equations (2.15) and (2.16). Insofar as
the specification (2.17) is consistent with the data, both approaches should yield the
same implied markup series. It follows that Bils' results using his second method give
an indication o f the size o f the correction that would result from taking account o f
adjustment costs for employment, if these are o f the size that he estimated. His estimate
of these adjustment costs imply an elasticity o f 12 even greater than the value o f 1.4
discussed above.
2.2.5. Labor hoarding
Suppose now that not all employees on a firm's payroll are used to produce current
output at each point in time. For example, suppose that o f the H hours paid for by
the firm at a given time, Hm are used in some other way (let us say, maintenance o f
the firm's capital), while the remaining H - H m are used to produce the firm's product.
Output is then given by Y = F ( K , z ( H - H m ) ) rather than Equation (2.1). We can again
33 Bils is able to estimate this equation by assuming parametric ftmctional forms for the functions
W~(h) and O(YN), and assuming that tt is a constant multiple of the straight-time wage. He also notes
that the term wt should refer not simply to the average hourly wage, but to total per-employee costs
divided by hours per employee; the numerator thus includes the costs of other expenses proportional
to employment but independent of the number of hours worked per employee, such as payments for
unemployment insurance. In fact, identification of the parameters in Equation (2.17) is possible only
because wt is assamed not to be given by a time-invariant fimction W(ht)/ht, but rather by (W(ht) + Ft)/ht,
where the shift term F t representing additional per-employment costs is time-varying in a way that is
not a function of h t.
Ch. 16:
1077
compute the marginal cost o f increasing output by hiring additional hours, holding Hm
fixed (along with other inputs). One then obtains instead of Equation (2.5)
Iz = u ) l tlHS741,
(2.1 8)
where UH ---- ( H - H m ) / H is the fraction of the labor input that is utilized in production.
Note that this conclusion is quite independent o f how we specify the value to the firm
o f the alternative use to which the hours Hm may be put. It suffices that we believe
that the firm is profit-maximizing, in its decision to allocate the hours that it purchases
in this way, as in its other input decisions, so that the marginal cost of increasing
production by shifting labor inputs away from maintenance work is the same as the
cost o f increasing production by hiring additional labor.
The fraction uH is often argued to be procyclical, insofar as firms are said to "hoard
labor" during downturns in production, failing to reduce payrolls to the extent o f the
decline in the labor needed to produce their output, so as not to have to increase
employment by as much as the firms' labor needs increase when output increases again.
For example, the survey by Fay and Medoff (1985) finds that when output falls by 1%,
labor hours used in production actually fall by 1.17%, but hours paid for fall only by
0.82% 34,
Insofar as this is true, it provides a further reason why markups are more
countercyclical than would be indicated by Equation (2.5) alone 35. I f the Fay and
Medoff numbers are correct, and we assume furthermore that nearly all hours paid
for are used in production except during business downturns, they suggest that
UH falls when output falls, with an elasticity o f 0.35 (or an elasticity of about 0.4
with respect to declines in reported hours). Thus this factor alone would justify setting
b = - 0 . 4 in Equation (2.9).
A related idea is the hypothesis that effective labor inputs vary procyclically more
than do reported hours because ofprocyclical variation in work effort. We may suppose
in this case that output is given by Y = F ( K , zeH), where e denotes the level o f effort
exerted. If, however, the cost o f a marginal hour (which would represent e units o f
effective labor) is given by the reported hourly wage W, then Equation (2.5) continues
to apply. Here the presence o f time-variation in the factor e has effects that are no
different than those o f time-variation in the factor z, both of which represent changes
in the productivity o f hours worked; the fact that e may be a choice variable of the
34 Of the remaining hours paid for, according to survey respondents, about two-thirds represent an
increase in employee time devoted to non-production tasks, while the other third represents an increase
in employee time that is not used at all. Fair (1985) offers corroborating evidence.
35 Models in which output fluctuations result from changes in firms' desired markups can also explain
why labor hoarding should be counter-cyclical, as is discussed further in Section 2.3. At least some
models in which fluctuations in output result from shifts in the real marginal cost schedule have the
opposite implication: periods of low labor costs should induce increases both in the labor force employed
in current production and in the labor force employed in maintenance tasks.
1078
firm while z is not has no effect upon this calculation. Note that this result implies
that variations in the relation between measured hours of work and the true labor
input to the production due to "labor hoarding" are not equivalent in all respects to
variations in effort, despite the fact that the two phenomena are sometimes treated as
interchangeable 3 6.
I f we allow for variation in the degree to which the measured labor input provides
inputs to current production (either due to labor hoarding or to effort variations), one
could also, in principle, measure marginal cost by considering the cost of increasing
output along that margin, holding fixed the measured labor input. Consideration of
this issue would require modeling the cost of higher utilization of the labor input
for production purposes. One case in which this does not involve factors other
than those already considered here is if higher effort requires that labor be better
compensated, owing to the existence of an effort-wage schedule w(e) of the kind
assumed by Sbordone (1996). In this case the marginal cost of increasing output by
demanding increased effort results in an expression of the form (2.11), where now
co =_ ew'(e)/w(e). If, at least in the steady state, the number of hours hired are such
that the required level of effort is cost-minimizing, and that cost-minimizing effort
level is unique, then (just as in our discussion above of a schedule specifying the
wage as a function of hours per employee) the elasticity co will be an increasing
function of e, at least near the steady-state level of effort. The existence of procyclical
effort variations would then, under this theory, mean that implied markup variations
are more countercyclical than one would conclude if the effort variations were not
taken into account.
This does not contradict the conclusion of the paragraph before last. For in a
model like Sbordone's, effort variations should never be used by a firm, in the
absence of adjustment costs for hours or employment (or some other reason for
increasing marginal costs associated with increases in the measured labor input, such
as monopsony power). In the presence, say, of adjustment costs, consideration of the
marginal cost of increasing output through an increase in the labor input leads to
Equation (2.12), rather than to Equation (2.5); this is consistent with the above analysis,
since a cost-minimizing choice of the level of effort to demand requires that
co(e) = f2
(2.19)
at all times. It is true (as argued two paragraphs ago) that variable effort requires
no change in the derivation of Equation (2.12). But observation of procyclical effort
variations could provide indirect evidence of the existence of adjustment costs, and
hence of procyclical variation in the factor Q.
A further complication arises if the cost to the firm of demanding greater effort
does not consist of higher current wages. Bils and Kahn (1996), for example, assume
36 For example, models of variable effort are sometimesreferred to as models of "labor hoarding", as
in Burnside et al. (1993).
1079
that there exists a schedule w(e) indicating the effective cost to the firm of demanding
different possible effort levels, but that the wage that is actually paid is independent
of the current choice of e, due to the existence of an implicit contract between firm
and worker o f the form considered in Hall (1980). They thus suppose that the current
wage equals w(e*), where e* is the "normal" (or steady-state) level of effort. In this
case, Equation (2.12) should actually be
1
w(e*)
t~ = ~2 ~OHSH w(e)
(2.20)
I f effort variations are procyclical, the factor w(e)/w(e*) is procyclical, and so this
additional correction makes implied real marginal costs even more procyclical. In their
empirical work Bils and Kahn (1996) relate w(e)/w(e*) to variations in the energy
consumption per unit of capital and show that this correction makes marginal cost
significantly procyclical in four of the six industries they study. Interestingly, these
four industries have countercyclical marginal costs when they ignore variations in the
cost of labor that result from variations in effort.
JJ Rotembergand M. Woodford
1080
= ~o-l(nH +zox)~h ~,
(2.22)
where OK is the elasticity o f output with respect to the effective capital input. However,
while the presence o f ~ > 0 in Equation (2.20) is o f considerable importance for
one's estimate o f the average level o f the markup (it increases it), it has less dramatic
consequences for implied markup fluctuations. In the Cobb-Douglas case, r/H and r/K
are both constants, and implied percentage variations in markups are independent
of the assumed size o f )~. Thus this issue has no effect upon the computations o f
Bils (1987).
If we maintain the assumption o f constant returns but depart from the Cobb-Douglas
case by supposing that ~//t is countercyclical (because eKH < 1), then allowance for
0 < 3~ ~< 1 makes the factor ~//~+ )o/K less countercyclical. This occurs for two reasons;
first, the factor z/H + )~r/K decreases less with decreases in ~/H (and in the limit of
~, = 1, it becomes a constant), and second, the factor y/uK (upon which 0/4 depends)
is again less procyclical. Nonetheless, even if we assume that all countercyclical
37 They provide evidence of a statistical correlation between hours per worker and other proxies for
capital utilization. Their econometric results are consistent with an assumption that capital utilization is
proportional to hours per employee, a result that also has a simple interpretation in terms of a common
work-week for all inputs. On the other hand, as Basu and Kimball (1997) note, this correlation need not
indicate that firms are forced to vary the two quantities together.
38 More generally, belief that ~. should take a significant positive value, perhaps on the order of 1,
reduces the significance of variations in r/n as a contribution to implied markup variations, since both
y and h are strongly procyclical. It is not plausible, however, to suppose that ~. should be large enough
to make ~ - ,b~ a significantly countercyclicalfactor.
1081
variation in this factor is eliminated, implied markup variations will still be as strongly
countercyclical as they would be with a Cobb-Douglas production function.
To sum up, there are a number of reasons why the simple ratio of price to unit
labor cost is likely to give an imprecise measure of cyclical variations in the markup.
As it happens, many of the more obvious corrections to this measure tend to make
implied markups more countercyclical than is that simple measure. Once at least some
of these corrections are taken account of, one may easily conclude that markups vary
countercyclically, as is found by Bils (1987) and Rotemberg and Woodford (1991).
2.3. Alternative measures o f real marginal cost
Our discussion in Sections 2.1 and 2.2 has considered for the most part a single
approach to measuring real marginal cost (or equivalently, the markup), which
considers the cost of increasing output through an increase in the labor input. However,
as we have noted, if firms are minimizing cost, the measures of real marginal cost that
one would obtain from consideration of each of the margins along which it is possible
to increase output should move together; thus each may provide, at least in principle,
an independent measure of cyclical variations in markups. While cyclical variation
in the labor input is clearly important, cyclical variations in other aspects of firms'
production processes are observed as well. We turn now to the implications of some
of these for the behavior o f real marginal cost.
2.3.1. Intermediate inputs
Intermediate input use (energy and materials) is also highly cyclical. Insofar as the
production technology does not require these to be used in fixed proportions with
primary inputs [and Basu (1995) presents evidence that in US manufacturing industries
it does not], this margin may be used to compute an alternative measure of real
marginal cost. Consideration of this margin is especially attractive insofar as these
inputs are not plausibly subject to the kind of adjustment costs involved in varying
the labor input [Basu and Kimball (1997)], so that at least some of the measurement
problems taken up in Section 2.2 can be avoided.
Suppose again that gross output Q is given by a production function Q(V,M),
where V is an aggregate o f primary inputs, and M represents materials inputs. Then,
considering the marginal cost of increasing output by increasing materials inputs alone
yields the measure
-
PQM(V,M)
PM
(2.23)
by analogy with Equation (2.2). [Note that in Equation (2.23),/~ refers to the "grossoutput" markup which we called/~c in Equation (2.10). Also note that P now refers
to the price of the firm's product, and not a value-added price index as before.] Under
1082
the assumption that Q exhibits constant returns to scale 39, QMis a decreasing function
o f M / V , or equivalently of the materials ratio m = M/Q. In this case, log-linearization
of Equation (2.23) yields
=fr~ -/3v,
(2.24)
39 This assumption allows for increasing returns, but requires that they take the form of increasing
returns in the value-addedproduction function V(K, zH).
40 This is shown in the fourth row of his Table 5. He regresses the percentage change in m on the
percentage change in Q, for each of 21 two-digit US manufacturing industries. He instruments output
growth using the Ramey-Hall instntments for non-technological aggregatedisturbances. He also shows
that intermediate inputs rise more than does a cost-weightedaverage of primaryinputs (labor and capital),
using the same instruments; as one should expect, the regression coefficientin this case is much larger.
Ch. 16:
1083
(1995) estimate of the response o f m to changes in the relative price o f primary and
intermediate inputs suggests an elasticity o f substitution half that size 41. Thus it seems
most likely that i n s t e a d f < - 1 in Equation (2.24). If the materials ratio rn is procyclical
as found by Basu, it follows that real marginal costs are actually more procyclical than
is indicated b y the materials share alone.
A related measure is used by Domowitz, Hubbard and Petersen (1986), who measure
"price-cost margins" defined as the ratio o f price to "average variable cost". They
measure this as a the ratio o f industry revenues to the sum o f labor and materials costs,
which is to say, as the reciprocal o f the sum o f the labor and materials shares. This
should correspond to the markup as we have defined it only under relatively special
circumstances. I f the production function is isoelastic in both labor inputs and materials
inputs, then real marginal cost is proportional to the labor share (as explained in
Section 2.1), and also proportional to the materials share (as explained in the previous
paragraph). It then follows that these two shares should move in exact proportion to one
another, and hence that their s u m is a multiple o f real marginal cost as well. Domowitz
et al. report that this sum is somewhat countercyclical for most industries, and as a
result they conclude that p r i c e - c o s t margins are generally procyclical. However, the
conditions under which this measure should correspond to variations in the markup o f
price over marginal cost are quite restrictive, since they include all o f the conditions
required for the labor share to be a valid measure o f real marginal cost, a n d all o f
those required for the materials share to be a valid measure. We have reviewed in
Section 2.2 a number o f reasons why the labor share is probably less procyclical than
is real marginal costs. Similar considerations apply in the case o f the materials share,
although the likely quantitative importance o f the various corrections is different in
the two cases; in the case o f materials, the elasticity o f substitution below unity is
probably a more important correction, while adjustment costs are probably much less
important. Nonetheless, one must conclude, as with our previous discussion o f the
labor share alone, that real marginal cost is likely to be significantly more procyclical
than is indicated by the Domowitz et al. measure o f "average variable cost" 42.
41 The last line of his Table 5 indicates an increase in rn of only 0.12% for each 1% increase in the
relative price of primary and intermediate inputs. His estimates of the cyclicality of materials input use
indicate three times as large an elasticity for M/V as for M/Q (comparing lines 2 and 4 of that table),
though the estimated elasticity of M/V is reduced when labor hoarding is controlled for. This would
suggest an increase in M/V of at most 0.36% for each percent increase in the relative price of inputs.
42 Similar issues arise with the study of Felli and Tria (1996) who use the price divided by overall
average cost as a measure of the markup. They compute this by dividing total revenue by total cost
including an imputed cost of capital (which depends on a measure of the real interest rate). Leaving
aside the difficulties involved in measuring the cost of capital, it is hard to imagine that adding together
the shares of labor, materials and capital is appropriate for computing markups unless each share in
isolation is appropriate as well. In addition, the existence of adjustment costs of capital probably make
the marginal cost that results from producing an additional unit by adding capital considerably more
procyclical than average capital cost. These adjustment costs may also rationalize the dynamic relation
they find between their ratio of average cost to output and output itself.
1084
Ct+l
(2.25)
where now Pt,t+~ is the corresponding discount factor for real income streams. Given an
assumption about the form of the marginal benefit function b(I, Z), observed inventory
accumulation then provides evidence about real marginal costs in an industry - more
precisely, about the expected rate o f change in real marginal costs.
The early studies in this literature [e.g., Eichenbaum (1989), Ramey (t991)] have
tended to conclude that real marginal cost is countercyclical. The reason is that they
assume that the marginal benefit of additional inventories should be decreasing in
the level of inventories (or equivalently, that the marginal cost of holding additional
inventories is increasing); the finding that inventories are relatively high in booms
then implies that b is low, from which the authors conclude that real marginal costs
1085
must be temporarily low 43. Eichenbaum interprets the countercyclical variation in real
marginal costs as indicating that output fluctuations are driven by cost shocks, while
Ramey stresses the possibility that increasing returns to scale could be so pervasive
that marginal cost could actually be lower in booms. Regardless o f the explanation,
if the finding o f countercyclical real marginal costs is true for the typical sector, it
would follow that markups in the typical sector must be procyclical. This is indeed
the conclusion reached by Kollman (1996).
Bils and Kahn (1996) argue, instead, that real marginal cost is procyclical in
each o f the six production-for-stock industries that they investigate. The differing
conclusion hinges upon a different conclusion about cyclical variation in the marginal
benefits of additional inventories. They begin by observing that inventory-to-sales
ratios do not vary secularly. This suggests that the function b is homogeneous o f
degree zero in inventories and sales; specifically, they propose that b is a decreasing
function, not o f I alone, but o f I/S 44. A similar conclusion follows from noticing that
inventory-to-sales ratios are fairly constant across different models o f automobiles at
a given point in time, even though these models differ dramatically in the volume o f
their sales.
But this implies that b is actually higher in booms. The reason is that, as Bils
and Kahn show, the ratio o f inventories to sales is strongly countercyclical; while
inventories rise in booms, they rise by less than do sales. Thus, the marginal value
o f inventories must be high in booms and, as a result, booms are periods where real
marginal costs are temporarily high.
This conclusion is consistent both with the traditional view that diminishing returns
result in increasing marginal costs, and with the view that business cycles are not
primarily due to shifts in industry cost curves. As noted earlier, Bils and Kahn also
show that their inventory-based measures of real marginal cost covary reasonably
closely with a wage-based measure of the kind discussed above, once one corrects the
labor cost measure for the existence ofprocyclical work effort as in Equation (2.20). If
their conclusion holds for the typical industry, and not just the six that they consider,
it would have to imply countercyclical markup variations 45.
43 This aspect of inventory behavior has been much discussed as an embarrassment to the "production
smoothing" model of inventory demand, which implies that inventories should be drawn down in booms
[e.g., Blinder (1986)]. That prediction is obtained by adjoining to Equation (2.25) the assumptions that
b is decreasing in I and that real marginal cost is increasing in the level of production Q.
44 A theoretical rationale for this is provided in terms of a model of the stockout-avoidance demand for
inventories.
45 The price data for the particular industries considered by Bils and Kahn are ambiguous in this regard;
they find that (given their measures of variations in marginal cost) markups are countercyclical in some
industries but procyclical in others. This means that certain of their sectors have strongly procyclical
relative prices for their products - something that cannot be true of industries in general.
1086
A final way in which output can be increased is by increasing the stock o f capital 46.
Thus
t~ -
PFK(K, z H )
E(r)
'
(2.26)
where E ( r ) is the expected cost o f increasing the capital stock at t by one unit while
leaving future levels o f the capital stock unchanged. Assuming that the capital stock
at t can actually be changed at t but also letting there be adjustment costs, rt equals
PK,t + CI,t - Rt,t+l(1 - 6)(PK,t+I + cI,t+l)
where PK,t is the purchase price o f capital at t, c~,t is the adjustment cost associated
with increasing the capital stock at t by one unit, 6 is the depreciation rate. It then
becomes possible to measure changes in/~ by differentiating Equation (2.26). This is
somewhat more complicated than the computation o f marginal cost using either labor
or materials because the rental rate o f capital r cannot be observed directly; it must
be inferred from a parametric specification for c~.
A related exercise is carried out by Galeotti and Schiantarelli (1998). After
specifying a functional form for Cl and making a homogeneity assumption regarding F,
they estimate Equation (2.26) by allowing/~ to be a linear function of both the level
of output and o f expected changes in output. Their conclusion is that markups fall
when the level of output is unusually high and when the expected change in output is
unusually low. As we discuss further in Section 3, this second implication is consistent
with certain models o f implicit collusion.
2.4. The response o f Jhctor prices to aggregate shocks
Thus far we have discussed only the overall pattern o f cyclical fluctuations in markups.
Here we take up instead the degree to which markup variations play a role in the
observed response o f the economy to particular categories of aggregate shocks. We are
especially interested in shocks that can be identified in the data, that are known to be
non-technological in character and that are thus presumptively statistically independent
of variations in the rate o f technical progress 47. These cases are o f particular interest
46 We have considered separately each of these different ways in which firms can increase their output
and their associated marginal cost. An alternative is to postulate a relatively general production (or cost)
function, estimate its parameters by assuming that firms minimize costs, and thereby obtain estimates
of marginal cost that relate to many inputs at once. One could then compare this "average" estimate of
marginal cost to the price that is actually charged. Morrison (1992) and Chirinko and Fazzari (1997)
follow a related approach.
47 In taking this view, of course, we assume that variations in technical progress are essentially
exogenous, at least at business-cycle frequencies.
1087
because we can then exclude the hypothesis o f shifts in supply costs due to changes
in technology as an explanation for the observed response o f output and employment.
This allows us to make judgments about the nature o f markup variations in response
to such shocks that are less dependent upon special assumptions about the form of the
production function than has been true above (where such assumptions were necessary
in order to control for variable growth in technology).
In particular, in the case o f a variation in economic activity as a result of a nontechnological disturbance, if markups do not vary, then real wages should move
countercyclically. In our basic model, this is a direct implication o f Equation (2.2),
under the assumption o f a diminishing marginal product of labor 48. For in the short
run, the capital stock is a predetermined state variable, and so increases in output
can occur if and only if hours worked increase, as a result o f which the marginal
product of labor must decrease; this then requires a corresponding decrease in the real
wage, in order to satisfy Equation (2.2). In the case o f such a shock, then, the absence
o f countercyclical real wage movement is itself evidence of countercyclical markup
variation.
Before turning to the evidence, it is worth noting that the inference that procyclical
(or even acyclical) real wages in response to these shocks imply countercyclical
markups is robust to a number o f types of extension o f the simple model that leads to
Equation (2.2). For example, the presence of overhead labor makes no (qualitative)
difference for our conclusion, since the marginal product o f labor should still be
decreasing in the number o f hours worked. A marginal wage not equal to the average
wage also leads to essentially the same conclusion. If, in particular, we assume that
the firm's wage bill is a nonlinear function o f the form W(H) = woo(H), where the
function o(H) is time-invariant though the scale factor w0 may be time-varying 49, then
w(H), the ratio of the marginal to the average wage, is a time-invariant function. Since
the denominator of Equation (2.2) should actually be the marginal wage, when the
two differ, our reasoning above actually implies that/~o must be countercyclical. But
as we have explained above, w(H) is likely to be an increasing function (if it is not
constant), so that/~ should vary even more countercyclically than does the product/too
(which equals the ratio o f the marginal product o f labor to the average wage). If there
are convex costs of adjusting the labor input, one similarly concludes that #f2 must
be countercyclical. But since the factor (2 [defined in Equation (2.13)] will generally
48 Note that the latter assumption is necessary for equilibrium, if we assume that markups do not vary
because product markets are perfectly competitive. In the case of market power but a constant markup
(as in a model of monopolistic competition with Dixit-Stiglitz preferences and perfectly flexible prices see below), a mildly increasing marginal product of labor schedule is theoretically possible, but does
not seem to us appealing as an empirical hypothesis.
49 For example, Bils (1987) assumes a relationship of this kind, where w0 represents the time-varying
straight-time wage, while the function v(H) reflects the nature of the overtime premium, which is timeinvariant in percentage terms.
1088
vary procyclically, this is again simply a reason to infer an even stronger degree of
countercyclical variation in markups than is suggested by Equation (2.2).
I f there is labor hoarding, it can still be inferred in the case of an increase in output
due to a non-technological disturbance that H - H m must have increased; and then, if
real wages do not fall, Equation (3.8) implies that markups must have declined. In the
case of variable capital utilization, the situation is more complicated. Condition (2.2)
generalizes to
/~ =
PzFH(uI(K, zH)
W
(2.27)
F1((uxK, zH)
# = V'((1 - 6(ux))K)6'(uir)"
(2.28)
Now if zH/uKK decreases when output expands, it follows that Fir declines.
Furthermore, this requires an increase in uK, so that, under our convexity assumptions,
both ~-/ and 6 ~ must increase. Thus Equation (2.28) unambiguously requires the
markup to decrease. Alternatively, if zH/uirK increases, FH declines, and then, if there
is no decline the real wage, Equation (2.27) requires a decline in the markup. Thus
under either hypothesis, markup variations must be countereyclical, if real wages are
not 51"
We turn now to the question of whether expansions in economic activity associated
with non-technological disturbances are accompanied by declines in real wages. There
are three important examples of identified non-technological disturbances that are often
used in the literature. These are variations in military purchases, variations in the world
1089
oil price, and monetary policy shocks identified using "structural VAR" methods. At
least in the USA, the level of real military purchases has exhibited noticeable variation
over the post-World War II period (as a result of the Korean conflict, Vietnam, and the
Reagan-era military build-up). The causes of these variations are known to have had to
do with political events that have no obvious connection with technical progress. (We
consider military purchases rather than a broader category of government purchases
exactly because this claim of exogeneity is more easily defended in the case of military
purchases.) Similarly, the world oil price has been far from stable over that period
(the two major "oil shocks" of the 1970s being only the most dramatic examples of
variation in the rate of increase in oil prices), and again the reasons for these variations,
at least through the 1970s, are known to have been largely external to the US economy
(and to have had much to do with political dynamics within the OPEC cartel)52. In
the case of monetary policy shocks, the identification of a time series for exogenous
disturbances is much less straightforward (since the Federal funds rate obviously
responds to changes in economic conditions, including real activity and employment,
as a result o f the kind of policies that the Federal Reserve implements). However,
an extensive literature has addressed the issue of the econometric identification of
exogenous changes in monetary policy 53, and we may therefore consider the estimated
responses to these identified disturbances. In each of the three cases, the variable in
question is found to be associated with variations in real activity, and these effects are
(at least qualitatively) consistent with economic theory, so that it is not incredible to
suppose that the observed correlation represents a genuine causal relation.
We turn now to econometric studies of the responses to such shocks, using relatively
unrestricted VAR models of the aggregate time series in question. Rotemberg and
Woodford (1992) show that increases in real military purchases raise private value
added, hours worked in private establishments and wages deflated by the relevant value
added deflator. Ramey and Shapiro (1998) show that the effect on this real wage is
different when revised NIPA data are used and that, with revised data, this real wage
actually falls slightly. They argue that this response can be reconciled with a two-sector
constant markup model. Whether a one-sector competitive model can be reconciled
with their evidence remains an open question.
Christiano, Eichenbaum and Evans (1996) show, using a structural VAR model to
identify monetary policy shocks, that output and real wages both decline in response to
the increases in interest rates that are associated with monetary tightening. This again
suggests that the contraction in output is associated with an increase in markups. An
increase in the federal funds rate by one percent that leads to a 0.4% reduction in
output reduces real wages by about 0.1%. I f one supposes that hours fall by about the
same percent as output, the effective increase in the markup is about 0.2%.
52 These first two series have been widely used as instruments for non-tectmologicalsources of variation
in US economic activity, following the precedent of Hall (1988, 1990).
53 For a recent survey, see Leeper, Sims and Zha (1996).
1090
Rotemberg and Woodford (1996b) look instead at the response of the US economy
to oil price increases. They show that during the pre-1980 OPEC period, such increases
lowered private value added together with real wages. Specifically, a one percent
unexpected increase in oil prices is shown to lead to a reduction of private value added
by about a quarter of a percent after a year and a half, and to a reduction of the real
wage (hourly earnings in manufacturing deflated by the private value-added deflator)
by about 0.1%, with a similar time lag. This combination of responses again suggests
that markups increase, especially during the second year following the sfiock.
The inference is, however, less straightforward in this case; for one might think that
an increase in oil prices should have an effect similar to that of a negative technology
shock, even if it does not represent an actual change in technology. In fact, Rotemberg
and Woodford show that this is not so. Let us assume again the sort of separable
utility function used to derive Equation (2.23), but now interpret the intermediate input
"M" as energy. In this case, consideration of the marginal cost of increasing output by
increasing labor inputs yields
/~ =
PQv(V,M) VH(K,zH)
w
(2.29)
Comparison of Equation (2.29) with (2.23) allows us to write a relation similar in form
to Equation (2.2),
PVH(K,zH)
-
Vl
'
(2.30)
P Y - ~PMM
V(K, zH)
(2.31)
Thus if we deflate the wage by the proper price index P, it is equally true of an energy
price change that a decrease in labor demand must be associated with an increase in
the real wage, unless the markup rises. [Note that the situation is quite different in the
case of a true technology shock, since the relation (2.30) is shifted by a change in z.]
Under the assumption of perfect competition (/~ = 1), the price index defined in
Equation (2.31) is just the ideal (Divisia) value-added deflator. Thus a competitive
model would require the value-added-deflated real wage to rise following an oil
shock, if employment declines 54; and the observation that real wages (in this sense)
decline would suffice to contradict the hypothesis of perfect competition. The results
of Rotemberg and Woodford do not quite establish this; first, because their privatevalue-added deflator is not precisely the ideal deflator, but more importantly, because
54 This result is discussed extensively by Bruno and Sachs (1985), who use it to assert that the
unemploymentfollowingthe oil shocks was due to real wage demands being too high.
1091
their measure of private value added includes the US energy sector, whereas the
above calculations refer to the output of non-energy producers (that use energy as
an input). Still, because the energy sector is small, even the latter correction is not too
important quantitatively; and Rotemberg and Woodford show, by numerical solution
of a calibrated model under the assumption of perfect competition, that while small
simultaneous declines in their measure of output and of the real wage would be possible
under competition, the implied declines are much smaller than the observed ones 55
Similar reasoning allows us to consider as well the consequences of changes in the
relative price of intermediate inputs other than energy. We ignored materials inputs in
our discussion above of the inferences that may be drawn from the response of real
wages to identified shocks. As before, however, Equation (2.2) [and similarly (2.27)]
can be interpreted as referring equally to a production technology in which materials
inputs are used in fixed proportions with an aggregate of primary inputs, under the
further assumption that the relative price of materials is always one, because materials
and final goods are the same goods. But the relative prices of goods differing by "stage
of processing" do vary, and so a more adequate analysis must take account of this.
When one does so, however, one obtains Equation (2.30) instead of (2.2). It is still the
case that the failure of real wages to rise in the case of a non-technological disturbance
that contracts labor demand indicates that markups must rise, as long as the real wage
in question is w/fL
What, instead, if one observes only the behavior o f w/P? Then the failure of this
real wage to rise might, in principle, be explained by a decline in P/P, consistent with
a hypothesis of constant (or even procyclical) markups. However [referring again to
Equation (2.29)], this would require a decline in Qv(V, M). Under the assumption that
Q is homogeneous degree one, this in turn would require a decline in M/V, hence an
increase in QM (V, M). I f markups are constant or actually decreasing, this would then
require an increase in the relative price of materials, PM/P, by Equation (2.23). Thus
we can extend our previous argument to state that if one observes that neither w/P
nor PM/P increases in the case of a non-technological disturbance that leads to reduced
labor demand, one can infer that markups must increase. In fact, Clark (1996) shows,
in the case of a structural VAR identification of monetary policy disturbances similar
to that of Christiano et al., that a monetary tightening is followed by increases in the
price of final goods relative to intermediate goods and raw materials. This, combined
with the evidence of Christiano et al. regarding real wage responses, suggests that a
monetary tightening involves an increase in markups.
A possible alternative explanation of declines in real wages and the relative price
of materials inputs at the same time as a contraction of output and employment is an
increase in some other component of finns' marginal supply cost. Christiano et al.
55 Finn (1999), however,finds larger declines in the case of a competitivemodel that allows for variable
utilization of the capital stock.
1092
propose that an increase in financing costs may be the explanation o f their findings 56.
As they show, in a model where firms require bank credit to finance their wage bill, the
interest rate that must be paid on such loans also contributes to the marginal cost o f
production; and it is possible to explain the effects o f a monetary tightening, without
the hypothesis o f markup variation, as being due to an increase in marginal cost due
to an increase in the cost o f credit. But while this is a theoretical possibility, it is
unclear how large a contribution financing costs make to marginal costs o f production
in reality 57. This matter deserves empirical study in order to allow a proper'quantitative
evaluation o f this hypothesis.
2.5. Cross-sectional differences in markup variation
In this subsection we survey the relatively scant literature that investigates whether
markups are more countercyclical in industries where it is more plausible a priori that
competition is imperfect. This issue is of some importance because countercyclical
markups are less plausible in industries where there is little market power. For markups
below one imply that the firm can increase its current profits by rationing consumers
to the point at which marginal cost is no higher than the firm's price. But if markups
never fall below one, there is little room for markup variation unless average markups
are somewhat above one. In addition, the theoretical explanations we present for
cotmtercyclical markups in section 3 all involve imperfect competition. A consideration
of whether the measures o f markup variation that we have proposed imply that markup
variation is associated with industries with market power is thus a check on the
plausibility o f our interpretation o f these statistics. Quite apart from this, evidence
on comparative markup variability across industries can shed light upon the adequacy
of alternative models o f the sources o f markup variation.
The most straightforward way o f addressing this issue is to compute markups for
each sector using the methods discussed in section 2, and compare the resulting markup
movements to output movements. In Rotemberg and Woodford (1991), we carry out
this exercise for two-digit US data, treating each o f these sectors as having a different
level o f average markups and using Hall's (1988) method for measuring the average
markup in each sector 58. We show that the resulting markups are more negatively
56 The same explanation is offered by Clark for the behavior of the relative prices of goods at different
stages of processing.
57 Interruptions of the supply of bank credit certainly can significantly affect the level of economic
activity, but the most obvious channel through which this occurs is through the effects of financing costs
upon aggregate demand. Financing costs are obviously important determinants of investment demand,
the demand for consumer durables, and inventory accumulation; but a contraction of these components
of aggregate demand can easily cause a reduction of equilibrium output, without the hypothesis of an
increase in supply costs.
58 For a more elaborate analysis of the evolution of cyclical markups in four relatively narrowly defined
(four digit) industries, see Binder (1995). He finds that these four industries do not have a common
pattern of markup movements, though none of them has strongly countercyclical markups.
1093
correlated with GNP in sectors whose eight-digit SIC sector has a higher average
four-firm concentration ratio. Thus, assuming this concentration is a good measure
of market power, these results suggest that sectors with more imperfect competition
tend to have more countercyclical markups.
One source of this result is that, as shown earlier by Rotemberg and Saloner
(1986), real product wages Wi/Pi are more positively correlated with GNP, and even
with industry employment, in more concentrated industries. By itself, this is not
sufficient to demonstrate that markups are more countercyclical since zFH could be
more procyclical in these sectors. However, the analysis of Rotemberg and Woodford
(1991) suggests that this is not the explanation for the more procyclical real product
wages in more concentrated sectors.
As we discussed earlier, Domowitz, Hubbard and Petersen (1986) measure markup
changes by the ratio of the industry price relative to a measure of "average variable
cost". They show that this ratio is more procyclical in industries where the average
ratio of revenues to materials and labor costs is larger, and see this as suggesting that
markups are actually more procyclical in less competitive industries. As we already
mentioned, this method for measuring markup variation imparts a procyclical bias
for a variety of reasons. This bias should be greater in industries with larger fixed
(or overhead) costs [because of Equation (2.8)], and these are likely to be the more
concentrated industries. In addition, the ratio of revenues to labor and materials costs
is a poor proxy for the extent to which a sector departs from perfect competition,
because this indicator is high in industries that are capital-intensive, regardless of the
existence of market power in their product markets.
Domowitz, Hubbard and Petersen (1987) use a different method for measuring
industry markup variations and obtain rather different results. In particular, they run
regressions of changes in an industry's price on changes in labor and materials cost as
well as a measure of capacity utilization. Using this technique, they show that prices
are more countercyclical, i.e., fall more when capacity utilization becomes low, in
industries with higher average ratios of revenues to materials and labor costs. If the
relation between capacity utilization and marginal cost were the same across industries,
and if one accepted their method for deciding which industries are less competitive,
their study would thus show that markups are more countercyclical in less competitive
industries.
1094
profits. We show that these procyclical variations contain very little information on
the importance of markup changes because markup variations induce such procyclical
responses.
We next take up a more ambitious attempt at gauging the role of markup fluctuations
in inducing cyclical fluctuations in economic activity. In particular, we study the extent
to which the markup changes that we measured in Sections 2.1 and 2.2 lead to output
fluctuations. Any change in output that differs from that which is being induced by
changes in markups ought naturally to be viewed as being due to a shift in real marginal
costs (for a given level of output). Thus, this approach allows us to decompose output
changes into those due to markup changes and those due to shifts in the marginal
cost curve. What makes this decomposition particularly revealing is that, under the
hypothesis that markups are constant all output fluctuations are due to shifts in real
marginal costs.
1095
total factor productivity growth will be found. As Hall (1988) notes, the Solow residual
involves a biased estimate of just this kind, if firms have market power. Consider a
production function of the form (2.1), where F is not necessarily homogeneous of
degree 1. Differentiation yields
(3.1)
As noted before, Equation (2.2) implies that r/,q = /~sH; similar reasoning (but
considering the marginal cost of increasing output by increasing the quantity of capital
used) implies that ~/K = #Sx. Thus under perfect competition (so that # = 1), the
elasticities correspond simply to the factor shares, and a natural measure of technical
progress is given by the Solow residual
e slw -
5'~ - s K p ~ -
s.5,..
More generally, however, substitution of Equation (3.1) (with the elasticities replaced
by/* times the corresponding factor income share) yields
eSolow = # - 1 Yr +sH#/>
/Z
(3.2)
In the case of perfect competition, only the second term is present in Equation (3.2),
and the Solow residual measures growth in the technology factor z. But in the presence
of market power (# > 1), increases in output will result in positive Solow residuals
(and decreases in output, negative Solow residuals), even in the absence of any change
in technology. In particular, output fluctuations due to changes in the markup will result
in fluctuations in the Solow residual, closely correlated with output growth.
Hall (1990) points out that in the case that the production function exhibits constant
returns to scale, this problem with the Solow residual can be eliminated by replacing
the weights sK,sH by the shares of these factor costs in total costs, rather than their
share in revenues. Thus he proposes a "cost-based productivity residual"
C a l l -= ~?r - ~K ~?K - ~,v 9H,
where ~/~ = SH/(SK + sH), and sK = 1 - ~H. In terms of these factor shares, the
production function elasticities are given by ~/H = p~H, ~/x = p~,v, where p = ~K + r/~/
is the index of returns to scale defined earlier. Similar manipulations as are used to
derive Equation (3.2) then yield
cHal I
p-
1^
. ^
Yr + s~/yz.
(3.3)
P
Even if kt > 1, as long as p = 1, Hall's "cost-based" residual will measure the growth
in z. One can show, in fact, that this measure of productivity growth is procyclical
1096
to essentially the same degree as is the Solow residual 59. But again this need not
indicate true technical change. For if there are increasing returns to scale (p > 1), due
for instance to the existence of overhead labor as discussed above, then increases in
output will result in positive Solow residuals even without any change in technology.
This explanation for the existence of procyclical productivity in the absence of cyclical
changes in technology is closely related to the previous one, since we have already
indicated that (given the absence of significant pure profits) it is plausible to assume
that/~ and p are similar in magnitude.
The quantitative significance of either of these mechanisms depends upon how
large a value one believes it is plausible to assign to /~ or p. Hall (1988, 1990)
argues that many US industries are characterized by quite large values o f these
parameters. He obtains estimates of/~ that exceed 1.5 for 20 of the 26 industries for
which he estimates this parameter. Within his 23 manufacturing industries, 17 have
estimates of/~ above 1.5 while 16 have estimates of p that are in excess of 1.5. His
evidence is simply that both productivity residuals are positively correlated with output
movements, even those output movements that are associated with non-technological
disturbances. In effect, he estimates the coefficients on the first terms on the righthand sides of Equations (3.2) and (3.3) by instrumental-variables regression in using
military purchases, a dummy for the party of the US President, and the price of oil
as instruments for non-technological disturbances that affect output growth. However,
even assuming that the correlations with these instruments are not accidental, this
merely establishes that some part of the procyclical productivity variations that are
observed are not due to fluctuations in true technical progress; since explanations
exist that do not depend upon large degrees of market power or increasing returns,
one cannot regard this as proving that # and p are large.
A second possible mechanism is substitution o f intermediate for primary inputs,
as discussed by Basu (1995). Suppose that materials inputs are not used in fixed
proportions, but instead that each firm's gross output Q is given by a production
function Q = Q ( V , M ) , where M represents materials inputs and V is an index of
primary input use (which we may call "economic value added"), and the function Q is
differentiable, increasing, concave, and homogeneous of degree 1. As before, economic
value added is given by a value-added production function V = F ( K , z H ) . Now
consider a symmetric equilibrium in which the price of each firm's product is the same,
and this common price is also the price of each firm's materials inputs (which are the
products of other firms). Consideration of the marginal cost of increasing output by
increasing materials inputs alone then yields
t~ = Q M ( V , M ) .
(3.4)
59 Because, as Hall notes, pure profits are near zero for US industries, sK + SH has a value near one for
a typical industry; hence the two types of factor shares, and the two types of productivity residuals, are
quantitatively similar in most cases.
Ch. 16:
1097
(3.5)
Furthermore, as long as firms have some degree o f market power (~ > I),
Equation (3.4) implies that QM > 1. Hence Q(1, m) - m will be increasing in m, and
Equation (3.5) implies that Y / V , the ratio of measured value added to our index o f
"economic value added", will be a decreasing function of g.
This implies that a decline in markups would result in an increase in measured value
added Y even without any change in primary input use (and hence any change in V).
This occurs due to the reduction of an inefficiency in which the existence of market
power in firms' input markets leads to an insufficiently indirect pattern of production
(too great a reliance upon primary as opposed to intermediate inputs). I f one's measure
of total factor productivity growth is based upon the growth in Y instead of V, then
markup variations will result in variations in measured productivity growth that are
unrelated to any change in technology. Since a markup decline should also increase
the demand for primary factors of production such as labor, it will be associated
with increases in employment, output, and total factor productivity - where the latter
quantity increases because of the increase in Y / V even if the measurement problems
stressed by Hall (relating to the accuracy of one's measure of the increase in V that
can be attributed to the increase in primary factor use) are set aside.
The quantitative importance of such an effect depends upon two factors, the
elasticity of the function m and the elasticity of the function Q(1, m) - m. The first
depends upon the degree to which intermediate inputs are substitutable for primary
inputs. Basu (1995) establishes that materials inputs do not vary in exact proportion
with an industry's gross output; in fact, he shows that output growth is associated
with an increase in the relative use of intermediate inputs, just as Equation (3.4)
would predict in the case of an output increase due to a reduction in markups. The
second elasticity depends upon the degree of market power in the steady state (i.e.,
the value o f / t around which we consider perturbations), because as noted above,
the derivative of Q(1, m) - m equals/t - 1. Thus while Basu's mechanism is quite
independent of Hall's, it too can only be significant insofar as the typical industry
possesses a non-trivial degree of market power.
An alternative mechanism is "labor hoarding"; indeed, this is probably the most
conventional explanation for procyclical productivity variations. I f only H - Hm hours
are used for current production, but productivity growth is computed using total payroll
hours H as a measure of the labor input, then a tendency of Hm to decline when
H - Hm increases will result in spurious proeyclical variations in measured productivity
1098
growth. Furthermore, this is exactly what one should expect to happen, in the case o f
fluctuations in activity due to markup variations.
Suppose that the value to a firm (in units o f current real profits that it is willing
to forego) of employing Hm hours on maintenance (or other non-production) tasks is
given by a function v(Hm). It is natural to assume that this function is increasing but
strictly concave. Then if the firm is a wage-taker, and there are no adjustment costs for
varying total payroll hours H , the firm should choose to use labor for non-production
tasks to the point at which
(3.6)
Let us suppose furthermore that the real wage faced by each firm depends upon
aggregate labor demand, according to a wage-setting locus of the form
(3.7)
w/P = v ( H ) ,
(3.8)
Substituting for Hm in the numerator the decreasing function of H just derived, and
substituting for w in the denominator using Equation (3.7), the right-hand side o f
Equation (3.8) may be written as a decreasing function of H. It follows that a reduction
in the markup (not associated with any change in the state of technology, the value
of non-production work, or the wage-setting locus) will increase equilibrium H and
reduce equilibrium Hm. The result will be an increase in output accompanied by an
increase in measured total factor productivity.
If the firm faces a wage that increases with the total number of hours that it hires
(due to monopsony power in the labor market, the overtime premium, or the like), then
the resulting procyclical movements in measured productivity will be even greater. In
this case, Equation (3.6) becomes instead
J ( H m ) = og(H)w/P,
(3.9)
where ~o(H) is the ratio of the marginal to the average wage, as in Equation (2.11). We
have earlier given several reasons why ~o(H) would likely be an increasing function,
60 If we imagine a competitive auction market for labor, then Equation (3.6) is just the inverse of the
labor supply curve. But a schedule of the form (3.6) is also implied by a variety of non-Walrasian models
of the labor market, including efficiency wage models, union bargaining models, and so on. See, e.g.,
Layard et al. (1991), Lindbeck (1993), and Phelps (1994) for examples of discussions of equilibrium
employment determination using such a schedule.
1099
at least near the steady-state level of hours. Hence the specification (3.9) makes the
right-hand side an even more sharply increasing function of H than in the case of (3.6).
Similarly, if there are convex costs of adjusting the total number of hours hired by the
firm, Equation (3.6) becomes instead
ot (Hm) = g-2w/P,
(3.10)
where g2 is again the factor defined in Equation (2.13). Again, this alternative
specification makes the right-hand side an even more procyclical quantity than in
the case of (3.6). Thus either modification of the basic model with labor hoarding
implies even more strongly countercyclical movements in Hm, and as a result even
more procyclical variation in measured productivity.
A related explanation for cyclical variation that results from markup variations in
measured productivity is unmeasured variation in labor effort. If, as in the model of
Sbordone (1996), the cost of increased effort is an increase in the wage w(e) that must
be paid, and there are convex costs of varying hours, then the cost-minimizing level
of effort for the firm is given by Equation (2.19). As discussed earlier, this implies
that effort should co-vary positively with fluctuations in hours (albeit with a lead),
since the factor 2 will be procyclical with a lead, while the function ~o(e) will be
increasing in e. Furthermore, consideration of the marginal cost of increasing output
by demanding increased effort implies that 61
PzFL,(K,zeH)
-
(3.11)
w'(e)
Since wt(e) must be increasing in e (at least near the steady-state effort level, as
a consequence of the second-order condition for minimization of the cost w/e of
effective labor inputs), Equation (3.11) requires that a reduction in markups result in an
increase in e H (to lower the numerator), an increase in e (to increase the denominator),
or both. Since e and H should co-vary positively as a consequence of Equation (2.19),
it follows that a temporary reduction of markups should be associated with temporary
increases in effort, hours, and output. Countercyclical markup fluctuations would
therefore give rise to procyclical variations in measured productivity.
Another related explanation is marneasured variation in the degree of utilization
of the capital stock. The argument in this case follows the same general lines. I f
markups fall, firms must choose production plans that result in their operating at a
point of higher real marginal costs (which quite generally means more output). Costminimization implies that real marginal costs increase apace along each of the margins
available to the firm. Thus if it is possible to independently vary capital utilization, the
real marginal cost of increasing output along this margin must increase; under standard
61 As noted earlier, this implies that Equation (2.11) holds with co replaced by co(e).
1100
assumptions, this will mean more intensive utilization o f the firm's capital. But the
resulting procyclical variation in capital utilization will result in procyclical variation
in measured productivity, even if there is no change in the rate o f technical progress.
Similar conclusions are obtained when capital utilization is a function o f hours
worked per employee. Consider again the case in which there is an interior solution
for hours because the wage schedule W(h) is nonlinear in hours per employee, and in
which hours per employee nonetheless vary because o f convex adjustment costs for
employment. Then the cost-minimizing decision regarding hours per employee satisfies
the first-order condition 62
O=~o(h)( ~/~ nH
)
+ )~/K "
(3.12)
t~ =
Business profits are also well-known to vary procyclically [e.g., Hultgren (1965)];
corporate profits after taxes have long been a component of the NBER's index o f
coincident business cycle indicators. This is sometimes thought to make it implausible
that business expansions are associated with declines in markups, since reduced
markups should lower profits. Indeed, Christiano, Eichenbaum and Evans (1996) report
calculations intended to show that a model in which expansions are due to markup
declines will almost inevitably make the counterfactual prediction that profits must
decline when output expands 63.
62 Note that this follows from the fact that both Equations (2.12) and (2.22) apply in this case.
63 They present their analysis as a criticism of sticky-price models of the effects of monetary policy;
but in fact their criticism relates simply to the fact that the model is one in which output increases due
to a reduction in markups.
Ch. 16:
1101
This implication is, however, less direct than it might at first seem. There are a
number of reasons why profits might well rise when markups fall. Many of these
have been introduced above as reasons why the inverse of the labor share need not
move countercyclically to the same extent as the markup. The connection between
these two issues is simple. The cyclical variation in (real) profits is essentially
determined by the cyclical variation in the amount by which the value of output
exceeds the wage bill, Y - ( w / P ) H . (This is because the remaining deductions
involved in the calculation of accounting profits, such as interest payments and
depreciation allowances, are relatively less cyclical.) Now if the labor share w H / P Y
is n o t procyclical, it follows that when output increases, w H / P increases no more
than proportionally to output, which surely means l e s s in absolute magnitude, since
labor compensation is on average only three-quarters of the value of output. Hence
Y - ( w / P ) H will increase. Thus any model that does not predict a procyclical labor
share will a f o r t i o r i not predict countercyclical profits. And indeed, parameter values
that imply procyclical variation in profits in response to markup variations are not hard
to find.
Consider first our simplest model, in which firms pay the same wage regardless of
the number of hours they hire, there are no adjustment costs, and the measured capital
and labor inputs are all that matter for a firm's output. Then equilibrium output Y ,
hours t l , and real wage w / P are determined by Equations (2.1), (2.2), and (3.7), given
the capital stock K, the state of technology z, and the markup #. Let us consider the
effects of markup variations, holding fixed the other two parameters (and the functions
F and v). If we neglect changes in interest and depreciation, the change in profits is
given by
d H = d(Y - v H ) = (ZFH - v) d H - H d v
= (t~ - 1 - c , , ) v d H ,
(3.13)
(3.14)
Now this is certainly possible; under the hypothesis of market power in the product
market (which we require in order to suppose that markup variations are p o s s i b l e ) ,
/~ > 1, so it is simply necessary that e~ be small enough.
This may not, however, seem empirically plausible; essentially, Christiano et al.
argue that it would require a greater degree of market power than is plausible for
most US industries. Their proposed value for e,, however (their "baseline" calculation
assumes e~ = 1), is based not upon the observed degree of cyclicality of wages, but
upon what they regard as a plausible specification of household preferences, given
an interpretation of Equation (3.7) as the labor supply schedule of representative
1102
household. In fact, the average wage is observed to be relatively acyclical, and even i f
this is a puzzle for the theory o f labor supply, there is no reason to assume a stronger
real wage response to increases in labor d e m a n d in calculating the effect on profits
o f an increase in output associated with a decline in markups. For example, Solon,
Barsky and Parker (1994) find an elasticity o f the average real wage with respect to
hours worked o f about 0.3 64; thus an average markup in excess o f 1.3 would suffice to
account for procyclical profit variations. A n d again, this is the "value-added markup"
that must exceed 1.3; for this, the typical supplier's markup need not be' much more
than 10 percent.
In any event, procyclical profits do not require even as large an average markup
as this, i f we make the model more realistic, in any o f the several ways discussed
above. Consider first the possibility that the marginal wage paid by a firm varies
with the number o f hours that it hires, and not only with aggregate labor demand
(as assumed above), due, for example, to m o n o p s o n y power in the labor market. Let
us write the firm's wage bill as W(Hi;H), where H i represents hours hired b y firm i,
and H represents aggregate hours hired. Then in a symmetric equilibrium, the average
wage v is given by W(H; H)/H, and the ratio co o f the marginal wage to the average
wage is given by HWI(H; H)/W(H; H). In this case, Equation (3.13) generalizes to
dH = d(Y-
W ( H ; H ) ) = (zFi~ - W1 -
W2)dH
(3.15)
Since, as explained earlier, there are a number o f reasons for co to be larger than one,
the markup need not be as large as is required b y Equation (3.14) in order for profits
to be procyclical. If, for example, we assume that co = 1.2, as Bils (1987) estimates 65,
and c~ = 0.3, it suffices that/~ = 1.1 (which means a gross-output markup o f 4%).
64 Solon et al. find a considerably larger elasticity for the wage of individuals, once one controls for
cyclical changes in the composition of the workforce. However, for purposes of the cyclical profits
calculation, it is the elasticity of the average wage that matters; the fact that more hours are low-wage
hours in booms helps to make profits more procyclical.
65 This is what Bils' estimates imply for the ratio of marginal wage to average wage when the margin
in question is an increase in weekly hours per employee, and the derivative is evaluated at a baseline of
40 hours per week. (As noted above, Bils finds that this ratio rises as hours per employee increase.) In
applying this ratio to Equation (3.15), we assume that the marginal cost of additional hours is the same
whether they are obtained by increasing hours per employee or by increasing the number of employees,
as must be true if firms are cost-minimizing.
Ch. 16:
1103
H - Hm
with respect to
(3.16)
If labor hoarding is countercyclical, 0 > 1, and Equation (3.16) also requires a smaller
markup than does Equation (3.14). The findings of Fay and Medoff (1985), discussed
above, would suggest a value of 0 on the order of 1.4. This would be enough to satisfy
Equation (3.16) regardless o f the size o f the markup.
Similar results are obtained in the case of variable labor effort or variable capital
utilization. The implied modification of Equation (3.14) is largest if the costs of
higher effort or capital utilization do not show up in accounting measures of current
profits. For example, suppose that effective capital inputs are given by u x K , where
the utilization rate uK is an independent margin upon which the firm can vary its
production process, and suppose that the cost of higher utilization is faster depreciation
of the capital stock (but that this is not reflected in the depreciation allowance used
to compute accounting profits). As explained above, we should expect a decline in
markups to be associated with a simultaneous increase in real marginal costs along
each margin, so that firms choose to increase ux at the same time that they choose to
increase labor inputs per unit of capital. Let )~ denote the elasticity o f uK with respect
to H as a result of this cost-minimization on the part o f firms 66. Then Equation (3.13)
becomes instead
d H = d(Y -
vii)
and Equation (3.14) again takes the form (3.16), where now 0 - (t/H + ,~K)/~7~I. If
capital utilization and hours co-vary positively (as we have argued, and as is needed
in order to interpret procyclical productivity variations as due to cyclical variation
in capital utilization), then 0 > 1, and again a smaller markup than is indicated by
Equation (3.14) will suffice for procyclical profits. If, for example, ;~ = 1, as argued
66 Note that we do not here assume a structural relation between the two variables.
1104
by Bils and Cho (1994), then 0 > 1:3, and Equation (3.16) is satisfied no matter how
small the average markup may be.
3.2. Identifying the output fluctuations due to markup variation
We now describe the consequences of alternative measures of marginal costs for one's
view of the sources of aggregate fluctuations. We propose to decompose the log of
real GDP Yt as
Yt = Y t +Yt~,
(3.17)
where the first term represents the level of output that is warranted by shifts in the
real marginal cost curve introduced in Section 1 (for a constant markup), while the
second is the effect on output of deviations of markups from their steady-state value,
and hence represents a movement along the real marginal cost schedule. We then use
this decomposition to investigate the extent to which changes in y are attributable
to either term. Because there is no reason to suppose that changes in markups are
independent of shifts in the real marginal cost curve, there is more than one way in
which this question can be posed. First, one could ask how much of the fluctuations in
aggregate activity can be attributed to the fact that markups vary, i.e., would not occur
if technology and labor supply varied to the same extent but markups were constant
(as would, for example, be true under perfect competition). Alternatively, one might
ask how much of these fluctuations are due to markup variations that are not caused
by shifts in the real marginal cost schedule, and thus cannot be attributed to shifts in
technology or labor supply, either directly or indirectly (through the effects of such
shocks on markups).
The first way of posing the question is obviously the one that will attribute the
greatest importance to markup variations. On the other hand, the second question
is of particular interest, since, as we argued in Section 1, we cannot attribute much
importance to "aggregate demand" shocks as sources of business fluctuations, unless
there is a significant component of output variation at business-cycle frequencies that
can be attributed to markup variations in the more restrictive sense.
Mere measurement of the extent to which markup variations are correlated with the
cycle - the focus of our discussion in Section 2, and the focus of most of the existing
literature - does not provide very direct evidence on either question. I f we pose the first
question, it is obviously necessary that significant markup variations exist, if they are
to be responsible for significant variation in economic activity. But the relevant sense
in which markup variations must be large is in terms of the size of variations in output
that they imply. The size of the correlation of markup variations with output is thus
of no direct relevance for this question. Moreover, markup variations could remain
important for aggregate activity in this first sense even if markups were procyclical
as a result of increasing whenever real marginal costs decline. In this case, markup
variations would dampen the effects of shifts in real marginal costs.
1105
If, instead, we ask about the extent to which markup variations contribute to output
movements that are independent of changes in real marginal cost, the correlation of
markups with output plays a more important role. The reason is that these orthogonal
markup fluctuations lead output and markups to move in opposite directions and thus
induce a negative correlation between output and markups. However, markups could
be very important even without a perfect inverse correlation since, as we show below,
the dynamic relationship between markup variations and the employment and output
variations that they induce is fairly complex in the presence of adjustment costs.
Furthermore, even neglecting this, a strong negative correlation between markups
and activity would be neither necessary nor sufficient to establish the hypothesis that
orthogonal movements in markups contribute a great deal to output fluctuations. The
negative correlation might exist even though the business cycle is mainly caused
by technology shocks, if those shocks induce countercyclical markup variations that
further amplify their effects upon output and employment. And the negative correlation
might be weak or non-existent even though shocks other than changes in real marginal
cost are important, if some significant part of aggregate fluctuations is nonetheless due
to these cost shocks, and these shocks induce procyclical markup variations (that damp,
but do not entirely eliminate or reverse, their effects upon output).
In this section, we try to settle these questions by carrying out decompositions
of the sort specified in Equation (3.17) and analyzing the extent to which y*, ~
and the part o f ~ ~ that is orthogonal to y* contribute to movements in y. We do this
for two different measurements of/Tt, which imply different movements in ~ . The first
measurement of/Tt we consider is based on Equation (2.9) while the second is based
on the existence of a cost of changing the level of hours worked. Because of space
constraints, we are able to give only a cursory and illustrative analysis of these two
cases.
We start with the case where markups are given by Equation (2.9), for which we
gave several interpretation above. To compute how much output rises when markups
fall, we must make an assumption about the extent to which workers demand a higher
wage when output rises. We thus assume that, in response to changes in markups,
wages are given by
~t = t/w/2/,.
(3.18)
Thus, t/v/ represents the slope of the labor supply curve along which the economy
moves when markups change. Obviously, this simple static representation is just a
simplification. We again let t/H represent the elasticity of output with respect to hours
when hours are being changed by markup movements. Using Equation (3.18) in (2.9)
together with the assumption that changes in output induced by markup changes equal
t/H times [/, it follows that
=-
(1-b-~H(1-a) *Iw)~
~H
~-H Y '
(3.19)
where the term in parentheses is positive because ~/H is smaller than one and a and b
are nonpositive. This formula allows us to compute #~ once we have measured/~ as
1106
(3.20)
where Zt now represents the current and lagged values of the change in private value
added, the ratio of nondurables and services consumption to output, and detrended
hours worked in the private sector. To obtain the ratio of per capita compensation to per
capita output that we use in Equation (3.20) we divided the labor share by the deviation
of hours from their linear trend. Since this same deviation of hours is an element of
the Zt vector, we would have obtained the same results if we had simply projected the
labor share itself. For this included level of hours (and output) to be comparable to the
labor share we use to construct (w - y ) , this labor share must refer to the private sector
as a whole. We thus use only this particular labor share in this section. Because of the
possibility that this labor share does not follow a single stationary process throughout
our sample, we estimated Equation (3.20) only over the sample 1969:1 to 1993:1.
Equation (3.20) allows us to express ( w - y) as a linear fimction of Z. Given that
a is zero, the only other determinant of the markup in Equation (2.9) is the level
of hours/2/, which is also an element of Z. Thus, our estimate of ~ is now a linear
function of Z. Equation (3.19) then implies that fit~ is a linear function of Zt as well.
1107
It is not the case, however, that y[ is a linear function o f Zt. The reason for this is that
Z includes only stationary variables and therefore does not include y. On other hand,
the change in private value added, Ay, is an element of Z. This means that, armed with
the stochastic process for Z that we estimated in Rotemberg and Woodford (1996a),
Zt =AZt 1 +et,
(3.21)
we can construct the innovations in ~u and in y*. These are linear functions of the
vector et which, given Equation (3.21), equals (Zt - A Z t - 1 ) so that these innovations
depend only on the history of the Z's. Similarly, the vector (Zt - A Z t - l ) together with
the matrix A in Equation (3.21) determines how the expectation of future values of Z is
revised at t. This means that we can use Equation (3.21) to write down the revisions at t
^u and y*t+k as linear functions of the history of the Z's.
in the expectations of Yt+k, Yt+k
Finally, the variance covariance matrix of the c's (which can be obtained from A and
the variance covariance matrix of the Z's) then implies variances and covariances for
both the innovations and revisions in the y's, the ~U's and the y*'s.
Table 3 focuses on some interesting aspects o f these induced variances and
covariances. Its first row focuses on innovations so that it shows both the variance of
the innovation in y* and in ~u as ratios of the innovation variance in y. The subsequent
rows focus on revisions at various horizons. The second row, for example, gives the
population variances of the revisions at t of y[+s and ~ut+5 as ratios to the variance of
the revision of yt+s. All these revisions correspond to output changes one year after
the effect of the corresponding et's is first felt. The next row looks at innovations two
years after the innovations first affect output and so on.
We see from Table 3 that this measure of the markup has only a very modest effect
on one's account of the source of aggregate fluctuations in output. The variances
of revisions in y* are almost equal to the corresponding variances of y for all the
horizons we consider. The innovation variance of y* is actually bigger which implies
that innovations in ~u that are negatively correlated with y* dampen the effect of these
short-run movements of y* on y. The last column in Table 3 looks at the variances of
the component o f ) u that is orthogonal to y*. This variance is equal to the variance of
flu times (1 -/3 2) where p represents the correlation between flu and y* and where this
correlation can easily be computed from Equation (3.21). To make the results clearer,
we again present the variance of this orthogonal component of flu as a fraction of
the corresponding variance of y. It is apparent from this column that this orthogonal
component explains very little of the variance o f y at any of the horizons we consider.
Thus, even though this measure of the markup is negatively correlated with our cyclical
indicators, it induces movements in output that are much smaller than the actual ones.
Overturning this finding appears to require implausible parameters. To make
output more responsive to markup changes requires that the term in parenthesis in
Equation (3.19) be smaller. We could achieve this by making r/r1 smaller or t/w bigger
but, given the values that we have chosen, large changes of this sort would be
unreasonable. An alternative way of lowering this coefficient is to make b smaller
1108
Table 3
Fractions of the variance o f y accounted by fi~ and y* a
varAy*
VarAy
varA))t~
VarAy
Innovation variances
1.43
0.05
0.01
0.88
0.06
0.,06
0.86
0.08
0.08
0.90
0.07
0.07
0.90
0.05
0.05
0.91
0.05
0.05
0.91
0.04
0.04
Val'dy
b = -0.4, a , c = O
c=8, a , b - O
Innovation variances
2.38
2.89
0.97
0.55
1.28
0.97
0.65
1.13
0.89
0.66
1.13
0.90
0.61
1.03
0.86
0.59
0.91
0.81
0.58
0.81
0.75
0.84
0.21
0.21
a Calculations based on projecting ( w - y ) on Z for period 1969:1 1993:1 and using properties of
stochastic process in Equation (3.21) where this stochastic process is estimated from 1948:3 to 1993:1.
in absolute value. The problem is that, as we saw before, this makes the markup less
cyclical. Thus, it does not help in making ~ track more of the cyclical movements
in output. By the same token, setting a equal to a large negative number makes the
markup more countercyclical but raises the term in parenthesis in Equation (3.19)
thereby reducing the effect of the markup on 3~.
We now turn to the case where adjustment costs imply that deviations of the markup
from the steady state are given by Equation (2.15). We follow Sbordone (1996) in that
we also let output vary with employee effort and this, as we saw, is consistent with
Equation (2.15). Letting a be zero and using Equation (2.14), Equation (2.15) can be
rewritten as
(3.22)
Allowing for variable effort is useful because it relaxes the restriction that the short run
output movements induced by markup variations are perfectly correlated with changes
1109
in hours. Thus, as in our earlier discussion of her model, we suppose that output is
given by Y = F(K, zeH). As a result, we have
f = t/H(/2/+ b).
(3.23)
We suppose as before that the wage bill is given by Hfvg(e), where ~ captures all the
determinants of the wage that are external to the firm and g is an increasing ftmction.
This leads once again to Equation (2.19) which, once linearized, can be written as
et = c
[(f/t -/2L 1 ) -
Et(Iqt+l-/2/t)]
(3.24)
(3.25)
Et Lfi-(1-
- }~2L)/~, = -~(fi, -
i1L)(1
trot),
where L is the lag operator while ~.1 and ~.2 are the roots of
/3/~,2 _ [l + / 3 +
013,+ 1 = 0
and
0--
~P+e~-t/H c'
~=
ca)
~P+e~o--t/H c
Noting that ~.1/3 is equal to 1/~.2 and letting ~. be the smaller root (which is also smaller
than one as long as 1 + t/v/ > 0H and ~p + e,o > 0), the solution of this difference
equation is
OO
OO
(3.26)
k=0j-0
The deviations of hours from trend that are due to changes in markups,/~/~', can
then be obtained by simply ignoring the movements in Wo in Equation (3.26). We can
1110
then compute the deviations o f output from trend that are due to markup variations,
~ , by combining Equations (3.23) and (3.24) to yield
A~)]) .
(3.27)
This implies that, as before, )3~ is a linear function o f current and past values Zt. To
see this, note first that Equation (3.22) implies that we can write/2t as a function of Zt.
The reason for this is that (w - y) is a function o f Zt, Ht is part of Zt and, as a
result o f Equation (3.21), Et/2/t+l is the corresponding element o f AZt. Therefore,
using Equation (3.21) once again, the expectation at t o f future values of/~ must be a
ftmction o f Zt. Past expectations of markups which were, at that point, in the future are
therefore fimctions of past Z's. The result is that we can use Equation (3.26) to write
) ~ as a fimction of the history o f the Z's 67. Finally, we use Equation (3.27) to write
the component of output that is due to markup changes as a function of the Z's.
We require several parameter values to carry out this calculation. First, we set c equal
to 8 to calculate ~t in Equation (3.22). To compute the connection between y~ and the
Z's we need three more parameters. It is apparent from Equations (3.26) and (3.27)
that this calculation is possible if one knows ~, ~ and cr~ in addition to c (which
is needed to compute markups anyway). For illustrative purposes, we set these three
parameters equal to 0.79, 0.13 and 3 respectively. The parameters ~ and ~ are not as
easy to interpret as the underlying economic parameters we have used to develop the
model. In addition to c and co) these include t/v, t/w, ~. Because the number o f these
parameters is larger than the number o f parameters we need to compute .~, there is a
one dimensional set of these economically meaningful parameters that rationalizes our
construction of33~. In particular, while this construction is consistent with supposing
that t/H, t/w, and ~0 equal 0.7, 0.25 and 0.1, respectively, it is also consistent with
different values for these parameters 6s.
We use our knowledge o f the relationship between ~ and the Z's for two purposes.
As before, we compute the variances o f the innovations and revisions in )3~ as well as
of y*. Second, we look at the resulting sample paths of)3~ and y*. The second part o f
Table 3 contains the variances, which correspond to the ones we computed before. The
results are quite different, however. In particular, the variance of the component of33 ~
that is orthogonal to y* now accounts for about 90% o f the variance of the revisions in
output growth over the next two years. Thus, independent markup movements are very
important in explaining output fluctuations over "cyclical" horizons. Moreover, if one
67 Because we later want to compute the sample values of~ ~ we truncate k so that it rtms only between
zero and eighteen. Given that our ,~ equals 0.79, this truncation should not have a large effect on our
results.
68 Note that we have made t/w, the elasticity of the wage with respect to hours along the aggregate
labor supply curve, somewhat smaller than before because our use of a positive ~p implies that wages
rise with output not only because hours rise but also because effort rises.
1111
8.6
8.5
8.4
8.3
8.2
8.1
"~jl eli
8.0
Fig. 3. Constant-markupand actual output.
takes the view that the movements o f y that are genuinely attributable to y* are those
which are not due to the component offi ~ that is orthogonal to y*, the movements in y*
account for only about 10% of the movements in y. Movements in y* have essentially
no cyclical consequences. It is not that the revisions in the expectations of y* are
constant. Rather, upwards revisions in y* over medium-term horizons are matched
by increases in markups that essentially eliminate the effect of these revisions on y.
This cannot be true over long horizons since the markup is assumed to be stationary
so that flu is stationary as well. Thus, changes in y* that are predicted 20 years
in advance account for about 80% of the revisions in output that are predicted
20 years in advance.
An analysis of the sample path o f ~ ~ (and the corresponding path of y*) delivers
similar results. Such sample paths can be constructed since ~ depends on the Z's which
are observable. Admittedly, Equation (3.26) requires that the entire history of Z's be
used. Data limitations thus force us to truncate k at 18 as explained in footnote 67. The
result is that ~ depends on 18 lags of Z. To make sure that the lagged expectations
of markups which enter Equation (3.26) are computed within the period where the
labor share remains a constant stationary function Zt, we construct this sample path
starting in 1973:2. The resulting values of y* and the log of output y are plotted in
Figure 3. It is apparent from this figure that the episodes that are usually regarded as
recessions (which show up in the Figure as periods where y is relatively low) are not
reflected in movements of y*. Figure 4 plots instead y~ against the predicted declines
of output over the next 12 quarters. These series are nearly identical so that, according
to this measure of the markup, almost all cyclical movements in output since 1973
1112
0.06
Predicted decline
f~
0.04
0.02
!"i
0.00
"-\/
i~,
i_
-0.02
-0.04-
[.i
/
-0.06 -
-0.08-0.10-
74
76
78
80
82
84
86
88
90
92
4. M o d e l s o f variable m a r k u p s
69 For another setting where inferences regarding markups are significantly affected by supposing that
there are costs of adjusting labor, see Blanchard (1997).
1113
then any source of variations in the markup that are independent of variations in the
marginal cost schedule itself will result in inverse variations in the level of output, and
so a negative correlation between the markup and economic activity. Thus theories of
why markups should vary as functions of interest rates or inflation (rather than of
the current level of economic activity) might well be successful explanations of the
cyclical correlations discussed in Section 2. In fact, a theory according to which the
markup should be a function of the level of economic activity is, in some respects,
the least interesting kind of theory of endogenous markup variation. This is because
substitution of/~ =/~(Y) into Equation (1.1) still gives no reason for equilibrium output
Y to vary, in the absence of shifts in the marginal cost schedule. (Such a theory, with
/~ a decreasing function o f Y, could however serve to amplify the output effects of
shifts in that schedule.)
Care is also required in relating theories of pricing by a particular firm or industry,
as a function of conditions specific to that firm or industry, to their implications for
aggregate output determination. For example, a theory according to which a firm's
desired markup is an increasing function of its relative output, /d = /t(yi/y) with
/~ > 0, might be considered a theory of "procyclical markups". But in a symmetric
equilibrium, in which all firms price according to this rule, relative prices and outputs
never vary, and there will be no cyclical markup variation at all. If instead (as discussed
in section 4.3 below) not all firms continuously adjust their prices, the fact that
adjusting firms determine their desired markup in this way can reduce the speed of
overall price adjustment; and this increase in price stickiness can increase the size
of the countercyclieal markup variations caused by disturbances such as changes in
monetary policy.
The models we look at fall into two broad categories. In the first class are models
where firms are unable to charge the price (markup) that they would like to charge
because prices are sticky in nominal terms. Monetary shocks are then prime sources of
discrepancies between the prices firms charge and the prices they would like to charge.
This leads to changes in markups that change output even if desired markups do not
change. In the second class of models, real factors determine variations in desired
markups, even in the case of complete price flexibility. Finally, we briefly consider
interactions between these two types of mechanisms.
4.1. Sticky prices
We do not provide a thorough survey of sticky price models since that is taken up in
Taylor (1999). Rather, our aim is threefold. First, we want to show how price stickiness
implies markup variations, and so may explain some of the findings summarized in
our previous sections. Second, we want to argue that markup variations are the crucial
link through which models with sticky prices lead to output variations as a result
of monetary disturbances. In particular, such models imply a link between inflation
and markups which is much more robust than the much-discussed link between
inflation and output. Thus viewing these models as models of endogenous markup
1114
variation may help both in understanding their consequences and in measuring the
empirical significance of the mechanisms they incorporate. Finally, we discuss why
sticky prices alone do not suffice to explain all o f the evidence, so that other reasons
for countercyclical markups also seem to be needed.
It might seem at first peculiar to consider output variations as being determined by
markup variations in a model where prices are sticky. For it might be supposed that if
prices are rigid, output is simply equal to the quantity demanded at the predetermined
prices, so that aggregate demand determines output directly. However, this' is true only
in a model where prices are absolutely fixed. It is more reasonable to suppose that
some prices adjust, even over the time periods relevant for business cycle analysis.
The issue then becomes the extent to which prices and output adjust, and, as we shall
see, this is usefully tmderstood in terms of the determinants of markup variation.
We illustrate the nature of markup variations in sticky-price models by presenting a
simple but canonical example, which represents a discrete-time variant of the model of
staggered pricing of Calvo (1983), the implications of which are the same (up to our
log-linear approximation) as those o f the Rotemberg (1982) model of convex costs of
price adjustment. First, we introduce a price-setting decision by assuming monopolistic
competition among a large number o f suppliers o f differentiated goods. Each firm i
faces a downward-sloping demand curve for its product of the form
Y/=D(P~Yt,
\PtJ
(4.1)
where Pj is the price of firm i at time t, Pt is an aggregate price index, I(t is an index o f
aggregate sales at t, and D is a decreasing function. We suppose that each firm faces
the same level of (nominal) marginal costs Ct in a given period 7. Then neglecting
fixed costs, profits of firm i at time t are given by
( e; ) r,.
\PtJ
Et ~-" j~ .1-1i+j
2._., a ~xt,t+1Pt+j'
j=0
where Rt,t+j is the stochastic discount factor for computing the present values at t of
a random level of real income at date t +j. (The factor ctJ represents the probability
70 Note that we have discussed above reasons why this need not be so, for example when a firm's
marginal wage differs from its average wage. As Kimball (1995) shows, deviations from this assumption
may matter a great deal for the speed of aggregate price adjustment, but we confine our presentation to
the simplest case here.
1115
that this price will still apply j periods later.) Denoting by Xt the new price chosen
at date t by any finns that choose then, the first-order condition for this optimization
problem is
e,
aJe,,.
j=o
O'
t+J
X,
p~+j ~
eD(Xt/Pt+j)
C,./]
Xt J = O,
(4.2)
where eD(x) =-- - x D ' ( x ) / D ( x ) is the elasticity of the demand curve (4.1).
For now, we further simplify by assuming that the elasticity of demand is a positive
constant [as would result from the kind of preferences over differentiated goods
assumed by Dixit and Stiglitz (1977)]. This means that a firm's desired markup, in the
case of flexible prices, would be a constant,/~* = eo/(eo - 1). In this way we restrict
attention to markup variations due purely to delays in price adjustment. It is useful to
take a log-linear approximation of the first-order condition (4.2) around a steady state
in which all prices are constant over time and equal to one another, marginal cost is
similarly constant, and the constant ratio of price to marginal cost equals g*. Letting
xt, ~t and ct denote percentage deviations of the variables Xt/Pt, Pt/Pt i and Ct/Pt,
respectively, from their steady-state values, Equation (4.2) becomes
j=0
, -
O,
(4.3)
k=l
represents the relative price in period t + j of the firm that chooses new price Xt in
period t, and so Equation (4.3) says, essentially, that the firm's price is expected to be
proportional to its marginal cost of production on average, over the time that the price
chosen at date t applies. This equation can be solved for the relative price ~t of firms
that have just changed their price, as a function of expected future inflation and real
marginal costs. The resulting relation can be quasi-differenced to yield
2ct = ot[3Etfrt+l + (1 - ot[3)~t + ot[3Et~t+b
(4.4)
We suppose that the price index Pt is a symmetric homogeneous degree one function
of the prices of the individual goods. Then near the steady state, it can be approximated
to first order by the geometric average of the prices. Since each period a fraction a
1116
of the prices remain unchanged, while the others all change to the common value Xt,
the rate of increase of the index satisfies
(4.5)
where t = (1 - a/3)(1 - a ) / a and fit = -ct denotes the percentage deviation of the
average markup ktt =- Pt/Ct from its steady-state value of/~*.
This equation relates the average markup at any date to current and expected future
inflation. To obtain the behavior of equilibrium output, one must use Equation (1.1)
along with this. If we log-linearize the real marginal cost schedule as ~t = z/c~t, where
~'t denotes the percentage deviation of output from trend, then Equation (1.1) implies
- f t t = ~/~Yt. Substitution of this into Equation (4.5) then yields an aggregate supply
relation of the form
frt = OYt +/3Et~t+l,
(4.6)
Ch. 16:
1117
technology shocks relative to what would happen under perfect competition; as a result,
input demand may actually decline in response to a favorable technology shock 71 .
We have seen that a sticky-price model of this kind involves endogenous variation
in the average markup. But is it useful to think of the endogenous markup variations as
central to the way that nominal variables have real effects in this model? We believe
that there are several advantages to viewing the model in this way (in addition, of
course, to the fact that it helps one in relating the predictions o f the sticky-price
model to the kinds o f facts discussed in Sections 2 and 3). First, if one is willing (as
seems reasonable) to abstract from the effects of monetary frictions upon the relations
(labor supply, labor demand, and so on) that underlie the real marginal cost schedule,
then the effects of monetary policy upon the determination o f real variables may be
reduced entirely to its effects upon the average markup. A general equilibrium model
of the effects o f monetary policy may then be usefully decomposed into three distinct
parts, each derived from largely separate microeconomic foundations: (i) a theory of
equilibrium output determination given the markup, essentially a more elaborate
version of Equation (1.1); (ii) an equation relating the markup to inflation variations,
Equation (4.5); and (iii) a theory of how monetary policy affects nominal aggregate
demand. An advantage o f viewing the structure of one's macroeconomic model this
way is that part (i) of the model involves no specifically monetary elements, and may
(except for the allowance for a time-varying markup) be identical to the equations
o f a real business cycle model, while part (iii) does not involve the specification
o f aggregate supply relations, and so may be directly adapted from conventional
Keynesian or monetarist models of the effects o f monetary policy on aggregate
demand. The theory of endogenous markup variation thus provides the crucial link
that allows the concerns o f real business cycle models and conventional monetary
models to be synthesized 72.
Second, understanding the markup variations that are associated with variations in
real activity in a sticky-price model is important to understanding when and how those
fluctuations in activity are inefficient, since the markup directly measures the extent to
which a condition for efficient resource allocation fails to hold. This perspective can
be a source o f important insights into the welfare losses from price-level instability
and the nature of optimal monetary policy.
And third, recognizing that Equation (4.5) is a more fundamental prediction of
the model o f price-setting than is a relation such as (4.6), which also depends upon
one's specification of wage-setting behavior and the like, may allow more accurate
71 This is what Gali (1999), Basu, Fernald and Kimball (1998) and Kiley (1996) find to be true in
US data, using a variety of quite different methods. Shea (1998), who identifies productivity shocks
from data on R&D spending and patents, does not find this contractionary effect upon input demand,
though his identified shocks also have little impact on long run output.
72 See Kimball (1995) and Goodfriend and King (1997) for more detailed sketches of this program,
which the latter authors term "the New Neoclassical Synthesis". Goodfriend (1997) also stresses the
importance of markup variations in accounting for the real effects of monetary policy.
1118
empirical estimation of the speed of aggregate price adjustment. Sbordone (1998) tests
the accuracy of Equation (4.5) as a model of aggregate price dynamics in the USA by
first estimating the evolution of unit labor cost (assumed to be proportional to marginal
cost) using a VAR. Using this evolution of unit labor costs, she then computes the
equilibrium path of the price index implied by Equation (4.5). She finds that this simple
model accounts quite well for the evolution of the private GDP deflator in the USA,
at the quarterly frequency, over the period 1960-1997. In the case of her best-fitting
value for a 73, the variance of the discrepancy between the actual price aeries and the
one that would be predicted on the basis of the unit labor cost process is reduced to
only 12% of what it would be in the absence of price rigidity 74, while the variance
of the discrepancy between the actual and predicted inflation series is reduced to
only 4% of what it would be according to the flexible-price (constant-markup) model.
It is especially striking that the model fits this well without any need for complications
such as stochastic disturbances to the pricing equation; this suggests that Equation (4.5)
is indeed more descriptive of the data than the aggregate supply relation (4.6). This
would suggest that the stochastic disturbances to this aggregate supply relation, which
require Roberts (1995) to add additional terms to his estimated equation and to estimate
it using instrumental variables, represent mainly disturbances to the real marginal cost
schedule, rather than disturbances to the pricing relation (4.5).
Despite the impressive success of this simple model as an explanation of much of
the cyclical variation in prices relative to labor costs, there is some reason to doubt that
this model of markup variation is completely adequate. In particular, the implication
that the output effect of supply shocks is muted in sticky-price models is problematic
given that, as suggested by Hamilton (1983) economic activity has tended to fall in the
aftermath of pre-1986 oil-price increases. If the principal effect of oil-price increases is
to increase marginal costs, then a sticky-price model (by implying that prices should
rise less than the increase in marginal costs, so that markups fall) will imply even
less of a contraction of equilibrium output than one should expect in the case of
a flexible-price model. However, the size of the observed contractionary effects of
oil-price shocks on the US economy is already rather larger than makes sense under
competitive pricing, owing to the relatively small share of energy costs in total marginal
costs of production. For this reason, Rotemberg and Woodford (1996b) propose that
oil price increases lead to increases in desired markups. With this motivation, we turn
to a brief review of models where desired markups vary.
73 This value is about 0.75 for her quarterlymodel, which implies an averagetime betweenprice changes
of approximately 14 months. This represents less frequent price adjustment than is observed in most
sectors of the US economy, according to the survey evidence presented in Blinder et al. (1998). The
coefficient I estimated by Sbordone can be reconciled, however, with more frequent price adjustments
if one hypothesizes variations in desired markups, as discussed in Section 4.3.
74 This means that the model of markup variation (4.5), combined with the simple measure (2.4) of
marginal costs, can account for 88% of the observed variability of the log ratio of price to unit labor
cost (or equivalently, of the log labor share) over this period.
1119
For simplicity, in this section we assume completely flexible prices. We also simplify
by making all firms fully symmetric so that, in equilibrium, they all charge the same
price. A number of types of theories of this kind have been considered in the literature.
4.2.1. Varying elasticity o f demand
Probably the simplest and most familiar model of desired markup variations attributes
them to changes in the elasticity of demand faced by the representative firm. There
are two important ways in which one might allow for variations in the elasticity of
demand at a symmetric equilibrium where all relative prices are equal to one.
The first is to suppose that the utility and/or production functions that define buyers'
preferences over differentiated goods are not homothetic, so that changes in aggregate
purchases Yt change the elasticity of demand. This is not an entirely satisfactory
assumption, however, because it is unappealing to assume that growth should lead
to secular changes in the elasticity of demand and in markups. One may, however,
avoid this implication by complicating the model, for example by assuming that growth
is associated with an increase in the number of differentiated goods rather than any
secular increase in the scale of production of any individual goods
A more appealing way of obtaining changes in thi s elasticity is to follow Gali (1994)
and Heijdra (1995) and assume that there are several different kinds of purchasers 75.
Each of these purchases all of the goods that are produced, but the different types
each have different preferences over differentiated goods. Suppose, for example, that
two groups 1 and 2 each care only about the amount they obtain of a composite good
defined by a symmetric, homogeneous degree 1 aggregate of all goods purchases, but
that the aggregator functions H~ and H2 are different for the two groups. Then the
demand for good i by groupj can be written as Yj,tDj(P[/pj,t), where Yj,t is the quantity
purchased by group j of its composite good, and pj,t is the price of that composite
good (a homogeneous degree 1 index of the individual goods prices). Total demand
for good i is then
r/= D1 kPl,t ]
\P2,t J
y2,,
(4.7)
where Dj(1) = 1 for both groups. At a symmetric equilibrium, all prices are the
same and the amount purchased of each good is the same, so that Yj,t is simply the
1120
ZtO11(1)+(1 -Zt)D~2(1)
where
Zt -
Yl,t
Yl,t + Y2,t"
Therefore, an increase in the share o f group 1 purchases in total purchases makes the
overall elasticity of demand more similar to D~, the elasticity of the demand by group 1.
An important feature of business cycles is that, as noted in Campbell (1987) and
Cochrane and Sbordone (1988), the ratio of consumption to GDP is high in recessions
and low in booms. Exactly the converse behavior applies to the ratio of investment
to GDE Thus, as Gall (1994) points out, the assumption that firms have more elastic
demands than consumers can provide one explanation for countercyclical markups.
Moreover, an exogenous increase in the fraction of output demanded for investment
purposes would increase yU.
Another variable that varies more cyclically than GDP is the purchase of durables.
This has led Bils (1989) and Parker (1996) to argue that increased purchases of
durables in booms reduce markups in these periods. This idea is closely related to
a proposal of Robinson (1932), who argued that people who purchased durables in
downturns were predominantly replacing durables that had ceased functioning and
that, as a result, demand in downturns was less elastic than demand in booms, which
consisted largely of demand by new purchasers.
To ensure that a story of this sort also leads to reduced markups when the
government expands its own purchases of goods and services, as would be needed
in order to account for the expansionary effects of government purchases other than
through an effect on labor supply, one must assume that the government has a relatively
elastic demand 76. The main disadvantage of this general type of explanation is that
aggregate demand, as such, has no direct role in lowering markups and thereby
increasing output. Rather, it is the composition of demand that affects aggregate output;
increases in aggregate demand only raise output if they happen to shift demand towards
sectors with more elastic demand. This means that at least some kinds of disturbances
that increase some important component of current spending must be contractionary
rather than expansionary (e.g., an increase in consumer demand, in Gali's model). It
is hard to think of empirical support for this kind of prediction.
76 See Heijdra (1995) for an analysis where government purchases may affect markups through this
channel.
1121
future sales. Probably the best-known model of this type is the "customer market"
model of Phelps and Winter (1970).
The customer market model is a model of monopolistic competition, in that each
firm maximizes profits with respect to its own price (markup) taking the price (markup)
of all other firms as given. But it differs from the standard model of monopolistic
competition [e.g., the model of Dixit and Stiglitz (1977)] in introducing a dynamic
element into the response of demand to prices. A firm that lowers its current price not
only sells more to its existing customers, but also expands its customer base. Having a
larger customer base leads future sales to be higher at whatever price is charged then.
One simple formulation that captures this idea involves writing the demand for firm i
at time t as
k t ~ t l mr
< 0,
t/(1) = 1,
(4.8)
where/~ is the markup of price over marginal cost implicit in the price charged by
firm i at time t, and the "market share" m~ is the fraction of average demand Yt that
goes to firm i if it charges the same price as all other firms. The market share depends
on past pricing behavior according to the rule
mt+l = g \[At I m t
< 0,
g(1) = 1,
(4.9)
so that a temporary reduction in price raises firm i's market share permanently.
Equations (4.8) and (4.9) are intended to capture the idea that customers have
switching costs, in a manner analogous to the models of Gottfries (1986), Klemperer
(1987), and Farrell and Shapiro (1988)77. A reduction in price attracts new customers
who are then reluctant to change firms for fear of having to pay these switching costs.
One obvious implication of Equations (4.8) and (4.9) is that the long-run elasticity
of demand, i.e., the response of eventual demand to a permanent increase in price, is
larger than the short-run elasticity of demand. In our case, a firm that charges a higher
price than its competitors eventually loses all its customers, though this is not essential
for our analysis.
The firm's expected present discounted value of profits from period t onward is
thus
j=O
-ff--+j/mHg
z=O
77 For a survey of much of this theoretical literature and its applications, see Klemperer(1995).
1122
Firm i chooses/~ to maximize this expression, taking as given the stochastic processes
{/~t} and {Yt} that define aggregate demand conditions. Therefore
(4.10)
I r~t+y I
Fir
= O,
where subscripts denote partial derivatives. At a symmetric equilibrium where all firms
charge the same price, each has a share mI = 1, and g equals 1 in all periods. So the
expectation term in Equation (4.10) is equal to the common present discounted value
of future profits, which we denote by Xt. Solving Equation (4.10) for the markup, we
obtain
/~t=/t
X
~ t ) ---
rf(1)
(4.11)
1 + ~/'(1) + g ' ( 1 ) ~
Because ~(1) and g~(1) are both negative, the derivative of/~ with respect to X / Y
is negative 7s. An increase in X / Y means that profits from future customers are high
relative to profits from current customers so that each firm lowers its price in order
to increase its market share. Thus, in this model, expansionary fiscal policy (which
raises real interest rates and thus lowers X/Y) raises markups and lowers output 79. On
the other hand, this is a model that can potentially amplify the expansionary effects
of loose monetary policy in the presence of sticky prices. The reason is that loose
monetary policy lowers real interest rates if prices are rigid and this raises X / Y s0.
A rather different view of the determinants of markups and output is obtained if
the customer market model is combined with the assumption that financial markets
are imperfect, as in Greenwald, Stiglitz and Weiss (1984) and Gottfries (1991). With
imperfect capital markets, shocks that raise the shadow cost of funds by making it
more difficult to borrow (such as reductions in asset values that lower the value of
firm's collateral) can lower X / Y and thereby lower output.
Chevalier and Scharfstein (1995, 1996) provide some evidence for this financeconstrained version of the customer market model. Chevalier and Scharfstein (1996)
consider pricing by supermarkets and pay particular attention to the prices charged
in states hit hard by the oil-price decline of 1986. They ask whether, within these
78 Felli and Tria (1996) argue that their proposed markup series is consistent with this implication.
79 Phelps (1994) emphasizes that this can be overturned in open economies under flexible exchange
rates. Expansionary fiscal policies then tend to appreciate the exchange rate, thereby forcing domestic
firms to lower their markups in order to compete effectivelywith foreign firms.
80 The model as expounded here and in the literature, however, involvesflexibleprices. The extension
of the theory to allow for delays in price adjustment would seem a high priority for future research.
Ch. 16:
1123
oil states, supermarkets that belonged to national chains (and who thus could rely
on externally provided cash to some extent) lowered their prices relative to those of
local supermarkets, who were presumably more strapped for cash. They find that they
do, suggesting that national supermarkets were more willing to invest in customers
at this point, presumably because they had lower discount rates as a result of their
access to cash. Chevalier and Scharfstein (1995) shows more generally that industries
with a relatively large fraction of output produced by small firms tend to have more
countercyclical markups if one controls for total concentration. The idea is that small
firms have less access to capital markets and so should be more strapped for cash in
recessions. This induces them to invest less in customers and raise their markups in
recessions. Controlling for concentration creates problems of interpretation because, as
discussed further below, highly concentrated sectors (in which large firms are clearly
important) have more countercyclical markups.
4.2.3. Implicit collusion
An alternative intertemporal model, where the same variable X / Y again turns out to
be the crucial determinant of the equilibrium markup, is the implicit collusion model
presented in Rotemberg and Woodford (1992). We consider an economy with many
industries, each of which consists of n firms. The n firms in each industry collude
implicitly in the sense that there is no enforceable cartel contract, but only an implicit
agreement that firms that deviate from the collusive understanding will be punished.
On the other hand, the firms in each industry, even when acting in concert, take
other industries' prices, the level of aggregate demand, and the level of marginal cost
as given. Abusing the language somewhat, we can view industries as monopolistic
competitors in the usual sense, while the firms within each industry collude implicitly.
Keeping this distinction in mind, we write the demand for firm i in industry j as
The function
D i
is syrmnetric in its first n arguments except the ith, and the functions
D i (for i = 1 , . . . , n) are all the same after appropriate permutation of the arguments.
The resulting profits of firm i in industry j if all other firms in the economy charge a
markup/h and it charges a markup #ij are
rt;
)= \
,I
j .
(4.13)
If the firm goes along with the collusive agreement at t and charges the same
iJ
markup as all other firms, it gets H~ (#t,/h) which we denote by H~(th)Yt. I f it
deviates and the punishment is as strong as possible, it earns some higher profits at t
1124
but it can expect to earn a present value of zero thereafter. In this case, a deviating
firm simply maximizes Equation (4.13) with respect to #y and its resulting profits
are /-/ta(#t)Yt. It is easy to show that the difference Hta(/zt) - / ( f ( / ~ ) is increasing
in #t. Intuitively, it should be clear that this difference is zero at the markup that
corresponds to the equilibrium where each firm behaves like a monopolistic competitor
and takes other firm's prices as given. If firms in the industry charge higher markups,
deviating by cutting prices is more attractive. Because this difference is increasing in
the markup, a profit-maximizing collusive oligopoly which is unable to sustain the
monopoly outcome for the industry will agree upon a markup that keeps firms just
indifferent between charging the collusive markup and deviating. Such a collusive
optimum implies that
z/f(~t)-/V(~,)
- x,.
(4.14)
This equation can again be solved for an equilibrium markup function of the form
~t = # ( x # r t ) .
In Rotemberg and Woodford (1992) we give the conditions under which there exists
an equilibrium where Equation (4.14) is binding near a deterministic steady state.
Because the left-hand side of Equation (4.14) is increasing in the markup #t the
equilibrium markup is increasing in X / Y . An increase in X, the expected present value
of future collusive profits, makes firms want to go along with the collusive price so
that this price can be higher. An increase in current output, by contrast, tends to reduce
the markup that can be sustained without breaking the collusive agreement. The result
is that tight fiscal policy, which raises real interest rates, raises markups and lowers
output. Temporary oil-price increases also raise X relative to Y and thus also reduce
output according to this model.
Rotemberg and Woodford (1991) provide evidence that, if asset price data are used
to compute X, markups are not just decreasing in Y but are also increasing in X. This
fits well with the finding of Galeotti and Schiantarelli (1998) that markups fall when
the expected rate of growth of output is high. Such a high rate of growth raises X
since profits are procyclical and this should lead to an increase in markups according
to this model.
A striking confirmation that high levels of X raise current markups is provided
by Borenstein and Shepard (1996) in their analysis of retail gasoline markets. Their
analysis looks at retail gasoline prices in 59 cities over 72 months and takes advantage
of the fact that the relationship between expected future demand and current demand
varies across cities because they experience different seasonal cycles. Similarly, there
are cross-city differences in expected future costs. Borenstein and Shepard show that,
consistent with this model, high expected future demand and low expected future costs,
both of which raise X, raise current markups. A similar finding, though the evidence
in this ease is so weak that one cannot reject the hypothesis of no effect, is reported
by Ellison (1994). He shows that a railroad cartel operating in the 1880s, the Joint
1125
Executive Committee, tended to have low prices when demand was low relative to
expected future demand.
The dependence of markups on X leads Bagwell and Staiger (1997) to conclude
that this model actually implies procyclical markups. This conclusion follows from
identifying booms with periods where the rate of growth of output is high and
identifying recessions with periods where the rate of output growth is low. Given
that the rate of growth of output is positively serially correlated, periods where output
growth is high are actually periods where a crude computation of X/Y (one that only
took note of the positive serial correlation of output growth) is high and the conclusion
follows. As noted by several authors [see Evans and Reichlin (1994), Rotemberg and
Woodford (1996a) and the papers cited therein] there are variables other than current
output growth that are useful for forecasting future output growth. As Evans and
Reichlin (1994) and Rotemberg and Woodford (1996a) show, once these other variables
are taken into account when computing expected output growth, recessions as defined
by the NBER are actually periods where expected output growth is high. Once this is
recognized, the model does indeed predict that markups should be high in periods that
are generally regarded as recessions.
Because past output growth is nonetheless also somewhat useful in forecasting future
output growth, it follows that expected output growth just after business-cycle troughs
(when output has already started to increase) is higher than expected output growth
just before these troughs. Thus, X / Y is higher just after business-cycle troughs than just
before. The model is thus consistent with some interesting observations made by Baker
and Woodward (1994). They compare the price charged by firms some time before an
industry trough (the reference month) with the price charged after the trough in the first
month in which output is no smaller than output in the reference month. They report
that, for some industries, the latter price is much higher than the former. Moreover,
the size of this price increase is larger in more concentrated industries. This suggests
that concentrated industries, where this theory is more likely to apply, are ones where
the markup is more likely to vary positively with X / Y .
One open question about this model (and the customer market model) is whether
they can explain the reduction in inputs that seems to accompany periods of genuine
technical progress. What determines which of these two models can explain this fact
is whether genuine technical progress raises or lowers X/Y. If the progress raises
mainly output in the future, one might expect X to rise relative to Y except that this
effect might be offset by an increase in the rate of interest (which reduces the present
value X ) . I f X / Y nonetheless rises with technical progress, the implicit collusion would
also imply that such shocks tend to raise markups and reduce output relative to what
would occur under frictionless perfect competition.
4.2.4. Variable entry
A final theoretical reason for markups to vary with cyclical variables is that entry
is procyclical. An advantage of this explanation is that it is undoubtedly true that
1126
more new firms incorporate in booms, as noted by Chatterjee, Cooper and Ravikumar
(1993) for the USA, and documented by Portier (1995) for France. Moreover, as long
as profits are procyclical, it makes sense that entry should be procyclical. As we saw in
Section 3, such proeyclical profits are possible even if output fluctuations are entirely
due to changes in markups, rather than to shifts in the real marginal cost schedule.
Suppose that, as in Chatterjee, Cooper and Ravikumar (1993) or Pottier (1995)
firms behave in Cournot fashion so that each industry contains several firms producing
perfect substitutes and these firms take the output of all other firms as'given when
deciding on their own level of output. In this model, the addition of new firms
cause markups to fall 81. The biggest problem with this explanation for countercyclical
markups is that technical progress would lead markups to fall both in the short run and
in the long run. As long a technological progress does not increase the fixed cost q~,
such long term progress increases the number of firms and thereby reduces markups.
One way of avoiding this difficulty is to assume that entry simply leads to an
increased number of goods being produced by monopolistically competitive firms, as
in Devereux, Head and Lapham (1996) or Heijdra (1995). These authors assume that
the monopolistically competitive finns produce intermediate goods that are purchased
by firms which combine them into final goods by using a Dixit-Stiglitz (1977)
aggregator. The result is that increased entry does not change the ratio of price to
marginal cost. It does, however, reduce the price of final goods relative to the price of
intermediate goods, because final goods can be produced more efficiently when there
are more intermediate goods. This reduction in the price of final goods effectively
raises real wages and, particularly if it is temporary, leads to an increase in labor supply.
Thus, Devereux, Head and Lapham (1996) show that, in their model, an increase in
government purchases raises output together with real wages. The increase in output
comes about because the increased government purchases make people feel poorer and
this promotes labor supply; this results in a shift out of the real marginal cost schedule.
The real wage in terms of final goods then rises because of the increase in the number
of intermediate goods firms.
4.3. Interactions between nominal rigidities and desired markup variations
Finally, we briefly consider the possibility that markups vary both because of delays in
price adjustment and because of variations in desired markups, for one or another of
the reasons just sketched. The possibility that both sources matter is worth mentioning,
since interactions between the two mechanisms sometimes lead to effects that might
not be anticipated from the analysis of either in isolation.
For example, variations in desired markups may amplify the effects of nominal
rigidities, making aggregate price adjustment even slower, and hence increasing the
81 Portier (1995) also considers a model where markups fall not only because entry occurs in booms
but also because the threat of entry leads incumbentfirms to lower their prices.
Ch. 16:
1127
output effects of monetary policy, even if the model of desired markup variation would
not imply any interesting effects of monetary policy in the case of flexible prices. To
illustrate this, let us consider again the discrete-time Calvo model of Section 4.1, but
now drop the assumption that the function D has a constant elasticity with respect to
the relative price. In this case, log-linearization of Equation (4.2) yields
Et
j=0
[ ^ des
+ ct+j]
(4.t5)
= 0,
k=l
~,/des - -
CD,t -- 1
from its steady-state value. The elasticity CD, and hence the desired markup, is a
function of the relative price of the given firm i, or equivalently of the firm's relative
sales yi/y. Letting the elasticity of the desired markup with respect to relative sales
be denoted ~, we obtain
t --
f'gt+k
[ ^ des
k=l
- c , +^j ,
= (1
(4.16)
k=l
(4.17)
A number of authors have proposed reasons why one might have ~ > 0, i.e., an
elasticity of demand decreasing in the firm's relative price. Kimball (1995) shows how
to construct aggregator functions that lead to arbitrary values of ~. Thus, this model can
rationalize extreme price stickiness even when the fraction of firms that change prices
is relatively high. Woglom (1982) and Ball and Romer (1990) suggest that search costs
provide an alternative rationale for a positive ~. The idea is that search costs imply
that relatively small price increases lead many customers to depart while small price
reductions only attract relatively few customers. A smoothed version of this kinked
demand curve gives the variable elasticity just hypothesized.
Equation (4.17) implies that ~ > 0 makes t~ a smaller positive quantity, for any
given assumed average frequency of price changes. This affects the quantitative form
of the markup equation (4.5), and hence the aggregate supply curve (4.6), in the same
1128
way as would a larger value of a 82. In particular, it implies that a given size permanent
increase in nominal aggregate demand (due, for example, to a monetary policy shock)
will result in both a larger and a more persistent increase in output. Thus allowing
for variation in desired markups of this kind can increase the predicted real effects of
monetary policy (including the size of the countercyclical markup variations caused
by monetary shocks).
To gain some intuition for this result, imagine an increase in aggregate demand
which increases marginal cost by increasing the demand for labor by >firms whose
prices are fixed. A firm which is free to change its prices would thus normally choose
a price above that charged by other firms. If, however, having a price that is relatively
high implies that demand is relatively elastic then such a firm would have a relatively
low desired markup and would thus choose a relatively low price. The effect of this
is that prices do not rise by as much on impact so that output increases by more.
In subsequent periods, the same logic leads those firms who can change their prices
to raise them to only a limited extent. Thus, the effects of the increase in nominal
aggregate demand are drawn out.
This occurs despite the fact that the hypothesis of a demand elasticity decreasing
in a firm's relative price does not, in itself, provide any reason for monetary policy to
have real effects. Indeed, under the hypothesis that all prices are perfectly flexible, it
provides no reason for equilibrium markups to vary at all. For with flexible prices, we
would expect a symmetric equilibrium in which all firms' prices are always the same,
so that the elasticity of demand faced by a firm (and hence its desired markup) would
never vary in equilibrium. Thus this hypothesis is much more interesting, both as an
explanation of markup variations and as a channel for real effects of shocks other than
cost shocks, when combined with the hypothesis of nominal price rigidity than it is
on its own.
It is also interesting to note that this hypothesis requires that desired markups be
low when the firm's relative price is high, i.e., when its own sales are low relative to
those of its competitors. Thus, its desired markups are positively correlated with its
own output relative to that of its competitors. At the firm or industry level, one might
well observe procyclical markups, if one measures the correlation with own output; yet
as shown above, the hypothesis is one that can increase the size of the countercyclical
markup variations at the aggregate level that occur as a result of aggregate demand
variations.
Inflation, search and markups are also linked in the work surveyed in Benabou
(1992). The idea in this research is that price rigidity in the face of inflation leads to
more price dispersion and this price dispersion makes search generally more valuable
to consumers. This, in turn, makes demand more elastic for all producers and thus
82 Thus it may help to reconcile the estimate of t by Sbordone (1998), based on the comovementof
aggregate indices of prices and labor costs, with microeeonomic evidence on the frequency of price
changes.
Ch. 16:
1129
exerts downwards pressure on markups. This theory implies that inflation ought to be
generally negatively related to markups. As Benabou (1992) shows, this implication
is confirmed in US data on the retail trade sector.
Variations in desired markups that are uniform across goods (rather than depending
on firms' relative demands) also interact in interesting ways with nominal rigidity.
For example, Kiley (1997) develops a model which combines staggered price-setting
with Gali's (1994) assumption of differential demand elasticities for consumption and
investment purchases. Monetary expansions then increase investment disproportionately and this temporarily lowers desired markups for all firms. This means that firms
that revise their prices do not raise them as much as they otherwise would (given the
increase in marginal cost) so that output rises more. This mechanism increases the
degree of strategic complementarity among different firms' pricing decisions. If other
firms raise their prices less, a given change in nominal rates (or in the money supply)
has a bigger effect on real rates of interest thereby affecting investment demand more.
This, in turn, implies that any given firm wants to raise its price by less. The greater
degree of strategic complementarity implies a slower adjustment of the aggregate price
level and hence a more persistent effect of the monetary expansion. Thus, while Gali's
(1994) model of markup variation does not directly imply that monetary shocks affect
output, it increases both the size and the persistence of the output effects of monetary
disturbances in the presence of sticky prices.
These illustrations demonstrate that a combination of endogenous variation in
desired markups and price stickiness can yield further channels through which
disturbances other than cost shocks affect the level of economic activity. This relatively
unexplored topic surely deserves further research.
5. Conclusions
The main benefit of allowing for markup variations is that it expands the range of
types of disturbances that can affect aggregate economic activity s3. Without variable
markups, output can only increase if real marginal cost falls, for example due to a
change in the effective labor supply to firms, or as a result of technological progress.
With variable markups, monetary and fiscal shocks can have effects other than those
that result from changes in the real wages at which workers are willing to work.
In addition, the output effects of certain supply shocks (like variations in the rate
of technical progress) may be muted, while other supply shocks (such as oil-price
increases) can lead to larger output movements.
83 In focusing on the effect of markup variations (rather than the effect of the average level of the
markup) we are assigning to imperfect competitiona role in macroeconomicsthat is quite different from
the one which Carlton (1996) argues is unimportant. For a discussion of the effect of the markup level,
see also Rotemberg and Woodford (1995) and the references cited therein.
1130
Ch. 16:
1131
Acknowledgment
We w i s h to thank M a r k Bils, Susanto Basu, R o b e r t Chirinko, M i l e s Kimball, and
A r g i a S b o r d o n e and J o h n Taylor for c o m m e n t s , and the N S F and the H a r v a r d Business
School D i v i s i o n o f R e s e a r c h for research support.
References
Abel, A.B. (1978), "Investment and the value of capital", Ph.D. Dissertation (MIT).
Abraham, K.G., and J.C. Haltiwanger (1995), "Real wages and the business cycle", Journal of Economic
Literature 33:1215-1264.
Ambler, S., and E. Cardia (1998), "The cyclical behavior of wages and profits under imperfect
competition", Canadian Journal of Economics 31:148-164.
Askildsen, J.E., and 0.A. Nilsen (1997), "Markups, business cycles and factor markets: an empirical
analysis", mimeograph (University of Bergen).
Bagwell, K., and R.W Staiger (1997), "Collusion over the business cycle", Rand Journal of Economics
28:82-106.
Baker, J.B., and P.A. Woodward (1994), "Market power and the cross-industry behavior of prices around
a business cycle trough", mimeograph (Federal Trade Commission).
Ball, L., and D. Romer (1990), "Real rigidities and the nonneutrality of money", Review of Economic
Studies 57:183-203.
Barro, R.J. (1981), "Output effects of government purchases", Journal of Political Economy
89:1086-1121.
Basu, S. (1995), "Intermediate inputs and business cycles: implications for productivity and welfare",
American Economic Review 85:51~531.
Basu, S., and M.S. Kimball (1997), "Cyclical productivity with unobserved input variation", Working
Paper No. 5915 (NBER, February).
Basu, S., J.G. Fernald and M.S. Kimball (1998), "Are technology improvements contractionary?",
mimeograph (University of Michigan).
Benabou, R. (1992), "Inflation and markups: theories and evidence from the retail trade sector", European
Economic Review 36:566 574.
Beveridge, S., and C.R. Nelson (1981), "A new approach to the decomposition of economic time series
into permanent and transitory components with particular attention to the measurement of the 'business
cycle'", Journal of Monetary Economics 7:151-174.
Bils, M. (1987), "The cyclical behavior of marginal cost and price", American Economic Review
77:838 857.
Bils, M. (1989), "Pricing in a customer market", Quarterly Journal of Economics 104:699 718.
Bils, M., and J.O. Cho (1994), "Cyclical factor utilization", Journal of Monetary Economics 33:319-354.
Bils, M., and J.A. Kahn (1996), "What inventory behavior tells us about business cycles", Working Paper
no. 428 (Rochester Center for Economic Research).
Binder, M. (1995), "Cyclical fluctuations in oligopolistic industries under heterogeneous information:
an empirical analysis", mimeograph (University of Maryland).
Blanchard, O.J. (1997), "The medium term", Brookings Papers on Economic Activity 1997(2):89-141.
Blinder, A.S. (1986), "Can the production smoothing model of inventory behavior be saved?", Quarterly
Journal of Economics 101:431453.
Blinder, A.S., E.D. Canetti, D.E. Lebow and J.B. Rudd (1998), Asking About Prices: a new approach to
understanding price stickiness (Russell Sage Foundation, New York).
Boldrin, M., and M. Horvath (1995), "Labor contracts and business cycles", Journal of Political Economy
103:972-1004.
1132
Borenstein, S., and A. Shepard (1996), "Dynamic pricing in retail gasoline markets", Rand Journal of
Economics 27:429-451.
Bruno, M., and J, Sachs (1985), Economics of Worldwide Stagflation (Harvard University Press,
Cambridge, MA).
Burnside, C., M. Eichenbaum and S.T. Rebelo (1993), "Labor hoarding and the business cycle", Journal
of Political Economy 101:245-273.
Calvo, G.A. (1983), "Staggered prices in a utility-maximizing framework", Journal of Monetary
Economics 12:383-398.
Campbell, J.Y. (1987), "Does saving anticipate declining labor income? An alternative test of the
permanent income hypothesis", Econometrica 55:124%1273.
Carlton, D. (1996), "A critical assessment of the role of imperfect competition in macroeconomics",
Working Paper No. 5782 (NBER).
Chatterjee, S., R. Cooper and B. Ravikumar (1993), "Strategic complementarity in business formation:
aggregate fluctuations and sunspot equilibria", Review of Economic Studies 60:795-812.
Chevalier, J.A., and D. Scharfstein (1995), "Liquidity constraints and the cyclical behavior of markups",
American Economic Review Papers and Proceedings 85:390-396.
Chevalier, J.A., and D. Scharfstein (1996), "Capital-market imperfections and countercyclical markups:
theory and evidence", American Economic Review 86:703 725.
Chirinko, R.S. (1980), "The real wage rate over the business cycle", Review of Economics and Statistics
62:459-461.
Chirinko, R.S., and S.M. Fazzari (1994), "Economic fluctuations, market power and returns to scale:
evidence from firm-level data", Journal of Applied Econometrics 9:47-69.
Chirinko, R.S., and S.M. Fazzari (1997), "Market power, inflation, and product market structure",
mimeograph (Emory University).
Christiano, L.J., M. Eichenbaum and C.L. Evans (1996), "Sticky price and limited participation models
of money: a comparison", Working Paper No. 5804 (NBER).
Clark, T.E. (1996), "The response of prices at different stages of production to monetary policy shocks",
Research working paper No. 96-12 (Federal Reserve Bank of Kansas City, December).
Cochrane, J., and A. Sbordone (1988), "Multivariate estimates of the permanent components of GNP
and stock prices", Journal of Economic Dynamics and Control 12:255-296.
Devereux, M.B., A. Head and B.J. Lapham (1996), "Monopolistic competition, increasing returns and
the effects of government spending", Journal of Money, Credit and Banking 28:233-254.
Dixit, A., and J. Stiglitz (1977), "Monopolistic competition and optimum product diversity", American
Economic Review 67:297-308.
Domowitz, I., R.G. Hubbard and B.C. Petersen (1986), "Business cycles and the relationship between
concentration and price-cost margins", Rand Journal of Economics 17:1-17.
Domowitz, I., R.G. Hubbard and B.C. Petersen (1987), "Oligopoly supergames: some empirical evidence
on prices and margins", Journal of Industrial Economics 35:379-398.
Dunlop, J.T. (1938), "The movement of real and money wage rates", Economic Journal 48:413-434.
Eichenbattm, M. (1989), "Some empirical evidence on the production level and the production cost
smoothing models of inventory investment", American Economic Review 79:853-864.
Ellison, G. (1994), "Theories of cartel stability and the Joint Executive Committee", Rand Journal of
Economics 25:37-57.
Evans, G.W., and L. Reichlin (1994), "Information, forecasts and measurement of the business cycle",
Journal of Monetary Economics 33:233-254.
Fair, R.C. (1985), "Excess labor and the business cycle", American Economic Review 75:239-245.
Farrell, J., and C. Shapiro (1988), "Dynamic competition with switching costs", Rand Journal of
Economics 19:123-137.
Fay, J.A., and J.L. Medoff (1985), "Labor and output over the business cycle", American Economic
Review 75:638-655.
1133
Felli, E., and G. Tria (1996), "Markup pricing strategies and the business cycle", mimeograph (University
of Rome).
Finn, M. (1999), "Perfect competition and the effects of energy price increases on economic activity",
Journal of Money, Credit and Banking, forthcoming.
Galeotti, M., and E Schiantarelli (1998), "The cyclicality of markups in a model with adjustment costs:
econometric evidence for U.S. industry", Oxford Bulletin of Economics and Statistics 60:121-142.
Gali, J. (1994), "Monopolistic competition, business cycles, and the composition of aggregate demand",
Journal of Economic Theory 63:73-96.
Gali, J. (1999), "Technology, employment and the business cycle: do technology shocks explain aggregate
fluctuations?", American Economic Review 89:249-271.
Gomme, P., and J. Greenwood (1995), "On the cyclical allocation of risk", Journal of Economic Dynamics
and Control 19:91-124.
Goodfriend, M. (1997), "A framework for the analysis of moderate inflations", Journal of Monetary
Economics 39:45-65.
Goodfriend, M., and R.G. King (1997), "The new neoclassical synthesis and the role of monetary
policy", NBER Macroeconomics Annual, 231-282.
Gottfries, N. (1986), "Price dynamics of exporting and import-competing firms", Scandinavian Journal
of Economics 88:417-436.
Gottfries, N. (1991), "Customer markets, credit market imperfections and real price rigidity", Economica
58:317-323.
Greenwald, B., J. Stiglitz and A. Weiss (1984), "Informational imperfections in the capital market and
macroeconomic fluctuations", American Economic Review Papers and Proceedings 74:194-199.
Hall, R.E. (1980), "Employment fluctuations and wage rigidity", Brookings Papers on Economic Activity
1980(1):91-123.
Hall, R.E. (1988), "The relation between price and marginal cost in U.S. industry", Journal of Political
Economy 96:921-947.
Hall, R.E. (1990), "Invariance properties of Solow's productivity residual", in: P.A. Diamond, ed.,
Growth, Productivity and Unemployment: Essays to Celebrate Bob Solow's Birthday (MIT Press,
Cambridge, MA) 71 112.
Hamilton, J.D. (1983), "Oil and the macroeconomy since World War II", Journal of Political Economy
91:228~48.
Heijdra, B.J. (1995), "Fiscal policy multipliers: the role of market imperfection and scale economies",
mimeograph (University of Amsterdam).
Hultgren, T. (1965), Costs, Prices and Profits: Their Cyclical Relations (National Bureau of Economic
Research, New York).
Kalecki, M. (1938), "The determinants of the distribution of national income", Econometrica 6:97-112.
Keynes, J.M. (1936), The General Theory of Employment, Interest and Money (Macmillan, London).
Keynes, J.M. (1939), "Relative movements of real wages and output", Economic Journal 49:34-51.
Kiley, M.T. (1996), "Labor productivity in U.S. manufacturing: does sectoral comovement reflect
technology shocks?", mimeograph (Federal Reserve Board).
Kiley, M.T. (1997), "Staggered price setting and real rigidities", mimeograph, (Federal Reserve Board).
Kimball, M.S. (1995), "The quantitative analytics of the basic neomonetarist model", Journal of Money,
Credit and Banking 27:1241-1277.
King, R.G., and S.T. Rebelo (1993), "Low frequency filtering and real business cycles", Journal of
Economic Dynamics and Control 17:207-231.
Kiemperer, P.D. (1987), "Markets with consumer switching costs", Quarterly Journal of Economics
102:375-394.
Klemperer, P.D. (1995), "Competition when consumers have switching costs: an overview with
applications to industrial organization, macroeconomics and international trade", Review of Economic
Studies 62:515-539.
1134
Kollman, R. (1996), "The cyclical behavior of markups in U.S. manufacturing and trade: new empirical
evidence based on a model of optimal storage", mimeograph (University of Montreal).
Kydland, EE., and E.C. Prescott (1988), "Cyclical movements of the labor input and its real wage",
Working Paper 413 (Research Department, Federal Reserve Bank of Minneapolis).
Layard, R., S. Nickell and R. Jackman (1991), Unemployment (Oxford University Press, Oxford).
Leeper, E.M., C.A. Sims and T. Zha (1996), "What does monetary policy do?", Brooldngs Papers on
Economic Activity 1996(2):1-63.
Lindbeck, A. (1993), Unemployment and Macroeconomics (MIT Press, Cambridge, MA).
Means, G.C., et al. (1939), The Structure of the American Economy (U.S. National Resource Committee,
Washington, DC).
Mills, E (1936), Prices in Recession and Recovery (National Bureau of Economic Research, New
York).
Mitchell, WC. (1941), Business Cycles and their Causes (University of California Press, Berkeley, CA).
Moore, G.H. (1983), Business Cycles, Inflation and Forecasting, 2nd edition (Ballinger,
Cambridge, MA).
Morrison, C.J. (1992), "Markups in U.S. and Japanese manufacturing: a short-rnn econometric analysis",
Journal of Business and Economic Statistics 10:51-63.
Murphy, K.M., A. Shleifer and R.W Vishny (1989), "Building blocks of market clearing business cycle
models", NBER Macroeconomics Annual, 24~86.
Parker, J.A. (1996), "The timing of purchases, market power and economic fluctuations", mimeograph
(Princeton University).
Phelps, E.S. (1994), Structural Slumps: The Modern Equilibrium Theory of Unemployment, Interest,
and Assets (Harvard University Press, Cambridge, MA).
Phelps, E.S., and S.G. Winter (1970), "Optimal price policy under atomistic competition", in: E. Phelps,
ed., Microeconomic Foundations of Employment and Inflation Theory (W.W. Norton and Co, New
York) 309-337.
Pindyck, R., and J.J. Rotemberg (1983), "Dynamic factor demands and the effects of energy price
shocks", American Economic Review 73:1066-1079.
Plosser, C.I. (1989), "Understanding real business cycles", Journal of Economic Perspectives 3:51-78.
Portier, E (1995), "Business formation and cyclical markups in the french business cycle", Annales
d'l~conomie et de Statistique 37:411-440.
Ramey, VA. (1991), "Non-convex costs and the behavior of inventories", Journal of Political Economy
99:306-334.
Ramey, V.A., and M.D. Shapiro (1998), "Costly capital reallocation and the effects of government
spending", Carnegie-Rochester Conference Series on Public Policy 48:145-194.
Roberts, J.M. (1995), "New Keynesian economics and the Phillips curve", Journal of Money, Credit and
Banking 27:975-984.
Robinson, J. (1932), The Economics of Imperfect Competition (Macmillan, London).
Rotemberg, J.J. (1982), "Sticky prices in the United States", Journal of Political Economy 90:1187-1211.
Rotemberg, J.J. (1996), "Prices, output, and hours: an empirical analysis based on a sticky price model",
Journal of Monetary Economics 37:505-533.
Rotemberg, J.J., and G. Saloner (1986), "A superganae-theoretic model of price wars during booms",
American Economic Review 76:390-407.
Rotemberg, J.J., and M. Woodford (1991), "Markups and the business cycle", NBER Macroeconomics
Annual 63-129.
Rotemberg, J.J., and M. Woodford (1992), "Oligopolistic pricing and the effects of aggregate demand
on economic activity", Journal of Political Economy 100:1153-1207.
Rotemberg, J.J., and M. Woodford (1995), "Dynamic general equilibrium models with imperfectly
competitive product markets", in: T.E Cooley, ed., Frontiers of Business Cycle Research (Princeton
University Press, Princeton, N J) 243-293.
Ch. 16:
1135
Rotemberg, J.J., and M. Woodford (1996a), "Real business cycle models and the forecastable movements
in output, hours and consumption", American Economics Review 86:71-89.
Rotemberg, JJ., and M. Woodford (1996b), "Imperfect competition and the effects of energy price
increases on economic activity", Journal of Money, Credit and Banking 28:549-577.
Sbordone, A.M. (1996), "Cyclical productivity in a model of labor hoarding", Journal of Monetary
Economics 38:331-362.
Sbordone, A.M. (1998), "Prices and unit labor costs: a new test of price stickiness", IIES Seminar paper
No. 653 (Stockholm University, October).
Shapiro, M.D. (1986), "The dynamic demand for capital and labor", Quarterly Journal of Economics
101:513-542.
Shea, J. (1998), "What do technology shocks do?", NBER Macroeconomics Annual, 275-310.
Solon, G., R. Barsky and J.A. Parker (1994), "Measuring the cyclieality of real wages: how important
is composition bias", Quarterly Journal of Economics 109:1-25.
Tarshis, L. (1939), "Changes in real and money wage rates", Economic Journal 19:150-154.
Taylor, J.B. (1999), "Staggered price and wage setting in macroeconomics", ch. 15, this Handbook.
Woglom, G. (1982), "Underemployment equilibrium with rational expectation", Quarterly Journal of
Economics 97:89-107.
Chapter 17
Contents
Abstract
Keywords
1. Introduction
2. The baseline neoclassical model
2.1. Failure of amplification in the baseline neoclassical model
2.2. Evidence about technology impulses
2.3. Failureof the baseline neoclassical model to explain persistence
2.4. Absence of unemployment from the baseline neoclassical model
3. Amplification
3.1. Elastic conventionallabor supply
3.2, Empirical research
3,3. Unemployment
3.3.1. Mechanism design and labor contracts
3.3.2. The modem strategic view of the employmentrelationship
3,3,3. Efficiencywages
3.3.4. Job destruction
3.3.5. Reorganizationand reallocation
4. Persistence
4.1. Time-consumingmatching in the labor market
4.2. The importance of secondary job loss for persistence
5. Conclusion
Acknowledgments
References
1138
1138
1139
1140
1141
1141
1142
1143
1145
1145
1148
1150
1154
1157
1157
1158
1160
1162
1162
1163
1167
1167
1167
1138
R.E. Hall
Abstract
The labor market occupies center stage in modern theories of fluctuations. The most
important phenomenon to explain and understand in a recession is the sharp decline in
employment and jump in unemployment. This chapter considers explanations based on
frictions in the labor market. Earlier research within the real business cycle paradigm
considered frictionless labor markets where fluctuations in the volume of work effort
represented substitution by households between work in the market and activities at
home. A preliminary section of the chapter discusses why frictionless models are
incomplete - they fail to account for either the magnitude or persistence of fluctuations
in employment. And the frictionless models fail completely to describe unemployment.
The evidence suggests strongly that consideration of unemployment as a third use of
time is critical for a realistic model. The two elements of a theory of unemployment
are a mechanism for workers to lose or leave their jobs and an explanation for the time
required from them to find new jobs. Theories o f mechanism design or of continuous
re-bargaining of employment terms provide the first. The theory of job search together
with efficiency wages and related issues provides the second. Modern macro models
incorporating these features come much closer than their predecessors to realistic and
rigorous explanations of the magnitude and persistence of fluctuations.
Keywords
Ch. 17:
1139
1. Introduction
The bulk of modern fluctuations theory fits into the broad framework of impulsesamplification-persistence. In this framework, fluctuations begin with a random
impulse. These can be shifts in technology or preferences, shifts in monetary or fiscal
policy, or spontaneous movements in consumption, investment, or other components
of spending. Most impulses of realistic size have small effects on employment in a
standard neoclassical model. To explain the observed volatility of employment and
other important aggregates, some form of amplification must occur. Modern thinking
about sources of amplification has focused primarily on highly elastic labor supply.
Labor supply is elastic because workers have economically valuable alternative uses of
their time. Though non-market alternatives lie behind the standard view of labor supply
and the view of labor supply in the original real business cycle model, the alternative
that is most prominent in recent models is job search. In fact, more than anything
else, this chapter is about the successful integration of unemployment theory into
formal dynamic general-equilibrium models. Not only do the new general-equilibrium
models comprehend the important phenomenon of unemployment, but it turns out that
unemployment is key to amplification and persistence.
Without consideration of unemployment, earlier dynamic general-equilibrium models explained persistence in employment largely through persistence in driving forces.
The models themselves did not contain much in the way of persistence mechanisms.
Where unemployment is considered explicitly, persistence arises naturally from the
time-consuming process o f placing unemployed workers in jobs following an adverse
impulse.
This chapter does not consider the origins of impulses, nor the propagation of
fluctuations across industries, though that propagation may interact with amplification
and persistence. The chapter focuses on frictions in the labor market, some of
which amplify impulses and others of which result in persistence, especially of
unemployment.
Most of the research I will consider here is developed within dynamic stochastic
general equilibrium (DSGE) models. By placing amplification and persistence mechanisms in formal general equilibrium models, contributors to modern fluctuations
research achieve a degree of clarity missing from earlier macroeconomics. Confusing
notions from earlier work, such as "aggregate demand" and "supply shocks," are giving
way to clearer general equilibrium counterparts, such as monetary and fiscal impulses
and shifts in the terms of trade. Moreover, research in the impulses-amplificationpersistence mode has come close to eliminating the traditional polarization of macro
researchers. Impulses come from technology, policy, and spontaneous shifts, a list
broad enough to include almost any earlier idea about the sources of fluctuations.
Old-fashioned Keynesian ideas such as wage rigidity, new Keynesian ideas such as
efficiency wages, and ideas about imperfect information all compete on an equal
footing to explain amplification, and are no longer assigned to warring schools of
thought.
R.E. Hall
1140
n(t))] dt.
(2.1)
Log consumption insures that static labor supply has zero wage elasticity - models
typically adopt this specification to match the positive trend in real wages to the zero
trend in annual hours per worker 1.
In principle, the baseline model could be driven by almost any kind of impulse shifts in technology or preferences, changes in policy regimes, or random elements
of policy. Kydland and Prescott's (1982) pioneering exploration of fluctuations using
the baseline model and alternatives focused on vibrations of the aggregate production
function as the single driving force. Much o f the ensuing literature retained that
focus, though changes in government purchases - an important topic before Kydland
and Prescott's formalization of the baseline model - remain a second important
driving force in that literature [see, for example, Aiyagari, Christiano and Eichenbanm
(1992)].
Empirical measures of aggregate technology - generally obtained by calculating the
Solow (1957) residual - suggest that changes in technology are quite persistent. In fact,
a random walk is not a bad approximation to the stochastic properties of aggregate
technology or for the technologies of particular industries. To put it differently, the
year-to-year Solow residual, which measures the change in technology, is close to
white noise. The custom has developed in the literature on DSGE models to model the
stochastic process of aggregate technology as first-order autoregressive with a serial
J See King and Rebelo (1999) for further discussion of this point.
1141
1142
R.E. Hall
Solow residual corrected for increasing returns is less correlated with output and is
hardly correlated with employment.
A substantial subsequent literature has questioned Hall's finding of high markups
matched by equally high returns to scale. New work has substituted better measures
of labor and capital input [Basu (1996), Burnside, Eichenbaum and Rebelo (1993)]
and corrected aggregation bias [Basu and Fernald (1997)]. With all corrections in
place, there is no remaining evidence of a correlation of the Solow residual with
employment. Thus, there are two reasons to question the existence of teclinology shifts
that are correlated with employment changes. First, what appears to be shifts in the
simple Solow residual are actually artifacts of imperfect competition or increasing
returns. Second, the correlation is the result of errors of measurement. Either finding
is troublesome for the view that technology shocks are an important driving force.
Hall (1997) investigates the role of the technology impulse in a semi-econometric
general equilibrium framework. The model adopts the premise of the baseline model
mentioned in the previous section - by making the kernel of utility depend on the
log of consumption, the model excludes any direct effect of a technology shift on
labor supply. A direct effect is one that occurs even in an economy without capital
or other methods for shifting resources between time periods. The absence of direct
effects occurs when the income and substitution effects in the corresponding static
labor supply function offset each other.
Hall observes that all of the effects of the technology impulse on employment must
operate through the intertemporal channel, that is, through investment - a positive
technology innovation sets off an investment boom. Hours of work rise, GDP rises,
and consumption falls, as the economy moves quickly to take advantage of higher
productivity. Hall examines the empirical relation between the investment/GDP ratio
and the simple uncorrected Solow residual. He finds a robust positive relation a one-percent shift of the production function causes about a one-percent increase in
the investment/GDP ratio. But the fraction of employment volatility explained by the
technology impulse is essentially zero. Further, the use of a more refined version of
the Solow residual would probably eliminate what little role the technology impulse
is found to have.
1143
The baseline model considers only two uses of time - employment and leisure.
A strong consensus has emerged in macroeconomic thinking that a realistic model
needs to consider a third use of time - job search or unemployment. Hall (1997)
takes the following approach to demonstrating the need for explicit consideration
of unemployment: He considers a neoclassical model without unemployment, but
one where shifts in household preferences drive fluctuations along with shifts
in technology and changes in government purchases. Because he considers three
aggregate variables - output, consumption, and hours of work - he is able to solve
for the values of the three impulses from the values of the three observed variables,
based on standard values for the parameters of the neoclassical model. Almost all
the explanatory power is assigned to the preference shift. Changes in government
purchases have a small role because the observed changes are small and not generally
associated with booms or recessions. Technology shifts also receive little weight
because they operate solely through the intertemporal investment channel and because
they should cause employment to change much more than consumption.
Hall models the preference shift as a random variable that determines the marginal
rate of substitution contemporaneously between consumption and leisure. Shifts in
the variable cause employment and consumption to move in the same direction. Hall
uses an empirical approach to determine the relative explanatory powers of the driving
forces for employment fluctuations. Almost all the credit goes to the preference shift.
But his conclusion is not that preference shifts are actually a major driving force.
Rather, another use of time - unemployment - is left out of the model. Periods of higher
unemployment are times when employment and consumption are both low. A better
way to explain the positive correlation of employment and consumption at businesscycle frequencies is to bring unemployment explicitly into the model.
Rotemberg and Woodford (1996) demonstrate the failure of the baseline model in
a rather different way. They focus on the joint time-series properties of employment,
output, and consumption. They show that the data contain a business cycle in the
weak sense defined by Beveridge and Nelson (1981) - the data tend to return to a
long-run trend whenever they deviate from that trend in the short run. The forecast
of employment growth is unusually high, for example, if employment is below its
trend because a recession occurred recently. In the baseline model driven by a
technology shock that follows a strict random walk [not the AR(1) process with a serial
correlation of 0.95 1 discussed earlier], employment and other variables lack almost any
tendency to return to normal. They derive measures of that tendency from a 3-variable
vector autoregression. When there is a tendency to return to normal, the VAR forecasts
future values for the variables that are different from the current values. They measure
this forecasting power at various horizons. Their findings are summarized in Table 1.
In the baseline model, the current value of output is close to the best value of the
forecast of future values at any horizon. There is a very slow-moving forecastable
component in the baseline model, associated with capital accumulation, so the standard
1144
R.E. H a H
Table 1
Failure of the baseline modela
Standard deviation of forecastable componentof output (percent)
Baseline model
Actual data with VAR
8 quarters
ahead
12 quarters
ahead
24 quarters
ahead
Infinitely
ahead
0.17
2.95
0.23
3.22
0.36
3.05
0.53
3.06
deviation o f the forecastable component rises as the horizon lengthens. In the actual
data for the USA, there is a pronounced rebound from abnormal conditions in a year
or two. The standard deviation of the change in output forecasted by the VAR is
about 3 percent at all horizons. In addition to being much larger than the forecastable
component in the baseline model, the time profile of the forecasting power is quite
different in the actual US economy - the forecasted change occurs almost entirely in
the first 8 quarters. In the baseline model, the forecasting power grows quite a bit
after 8 quarters.
Related failures of the baseline model are revealed in the correlations of the
foreeastable components. In the baseline model, the forecasted change in work effort
should have the opposite sign from the forecasted change in output. When a shock has
caused the economy to be at a point below its steady-state capital stock, work effort
will be above its steady state while output will be below. As capital is accumulated,
output rises and work effort falls. In fact, forecasted movements in output and work
effort are in the s a m e direction. After a recession, both hours of work and output rise
more rapidly than normal.
It appears that a reasonable explanation for the failure of the baseline model in
Rotemberg and Woodford's work is the absence o f unemployment in the model. The
forecastable rebound that occurs in the US economy following a recession occurs
during the period when workers displaced during the recession are making their
way back into long-term employment. In the baseline model, there is no burst of
unemployment in the first place and no two-year period of rematching. Both anomalies
reported by Rotemberg and Woodford are resolved by adding unemployment to the
model - the augmented model has much more predictable recovery from bad (or good)
shocks, and output and work effort move in the same direction during recoveries.
Not only does consideration of unemployment provide a more sensible interpretation
of correlations among key macro variables, but modern ideas about both amplification
and persistence often involve job destruction and job search, key ideas in the modern
theory of unemployment.
1145
3. Amplification
Amplification occurs when the response of employment to a driving force is stronger
than in the baseline neoclassical model. Macro research in the DSGE framework has
recognized the need for amplification mechanisms since Kydland and Prescott's (1982)
paper launched the framework. The mechanisms I discuss in this section all involve
elastic labor supply, either in the conventional sense or in the sense that there is another
activity - j o b search - that is a substitute for work.
The earliest amplification mechanisms invoked elastic labor supply in the standard
setting where workers choose between work effort and leisure. I f labor supply is
more elastic - for example, if the labor part of the kernel of the utility function is
(~ - n) s instead of log(~ - n) - the response of employment to a technology shock
is almost twice as large; see Campbell (1994, Table 3). Then a favorable impulse to
technology sets off the process that is the signature of the real business cycle model a burst of extra employment and a decline in consumption resulting in vigorous capital
accumulation.
In addition to the simple assertion of elastic labor supply, the literature proposing
fluctuations theories based on that hypothesis has offered three supporting ideas.
First was Kydland and Prescott's (1982) use of non-time-separable utility. Second
was Rogerson's (1988) observation that workers facing a binary choice between not
working and working full time may behave as if they had linear utility and perfectly
elastic labor supply. Third was B enhabib, Rogerson and Wright's (1991) consideration
of substitution between work in the market and work at home. Their paper marked
the beginning of the investigation of margins other than labor-leisure within DSGE
models.
A convenient family of non-time-separable preferences follows suggestions of
Sargent (1979, p. 371) and Kydland and Prescott (1982) [my discussion is taken from
Hall (1991b)]. Let zt be the accumulated stock of current and past work effort, with
persistence factor co:
t
zt = (1 - co) ~
~oSnt_s.
(3.1)
s=0
1146
R.E. H a l l
model, the worker orders work schedules with a utility function that is separable over
time in the cumulation variable, z:
(3.2)
-
t-O
1~---1
"
Define effective leisure as ~ - zt and actual leisure as B - nt. The parameter o is both the
intertemporal elasticity of substitution in effective leisure and the long-ruff elasticity of
substitution in actual leisure (where the long run is enough time so that the distributed
lag feature does not matter). In the short run, the elasticity of substitution in actual
leisure is greater than a by an amount that is controlled by the memory parameter, ~o.
The parameter ~ is the number of weeks physically available for work.
A worker with a high cr will suffer little from a work schedule involving many
weeks of work per year in one decade and few weeks per year in another decade,
in comparison to putting in the same number of lifetime weeks with no variation
from decade to decade. In a situation with free choice of weeks, such a worker will
concentrate weeks disproportionately during the years of highest wages.
On the other hand, a worker with low intertemporal substitution (low o) but high
memory persistence, o) (that is, close to one), will tolerate short-term fluctuations
in weeks of work but resist decade-to decade movements. Kydland and Prescott use
preferences that are slightly more general - current work can have a role in the utility
function beyond the role implicit in the variable zt.
To illustrate the difference between the short-run and medium-run responses of labor
supply to wage changes, consider the following question: let 2N be the number of
periods considered to define the medium run, which might be 24 months. Suppose a
worker increases weeks of work by 1 percent in periods t - N , . . . , t . . . . . t + N. By what
percent does the supply price of a week of work in period t increase? The elasticity
of labor supply over the 2N + 1-period run is the ratio of the two numbers.
It is convenient to use the ~-constant or Frisch labor supply schedule to answer this
question. Let ,~ be the Lagrangian multiplier associated with the worker's intertemporal
budget constraint. The first-order condition associated with labor supply is
O U (hi . . . . . n t , . . . , lilT)
Ont
= )~wt.
(3.3)
Here wt is the real wage in period t stated in period-0 prices, that is, in prices
discounted to period 0. The Frisch inverse labor supply function is simply the marginal
disamenity of work stated in wage units:
10U(nl
)~
.....
nt,. .. ,nT)
Ont
(3.4)
When U is additively separable in labor, this can be solved to give current labor
supply as a fimction of the current wage. Absent separability, it states the supply price
1147
of work in one period as a function of the level o f work in that and other periods.
Keeping & constant has two interpretations. First, Equation (3.4) gives the supply price
of labor at different points in time along the same labor-supply trajectory. Under this
interpretation, statements about the response of the supply price to different levels of
work are comparisons of the supply price at different points in time; the change in
the level o f work is fully anticipated. Second, the supply price conditional on ,~ has a
comparative statics interpretation when the change has little or no effect on ,~. Under
this interpretation, Equation (3.4) is very similar to (but not quite the same as) the
compensated labor supply schedule.
The Frisch labor supply function associated with the preferences considered here
is
Wt =
1
~(1
T
(3.5)
dx
1 (~_n)_l/o_
~o
2coN+I)
(3.6)
l+co
2coN+I)
1
T +~ J "
(3.7)
n-n
/,/
e(oo) = a - -
(3.8)
The elasticity E(cxD) controls labor supply over the life cycle. A worker with an
e(oo) of 1 will work twice as many weeks at age 40 as at age 20 if the wage at age 40
is double its level at age 20 (and this doubling was known to be in the offing at age 20).
Life-cycle variations in weeks of work do not show an elasticity anywhere near 1 - the
evidence appears to favor values of 0.1 to 0.2. If 7-,
is 5/47 = 0.11 and the mediumn
run elasticity of labor supply is 0.15, then ~r is 0.15/0.11 = 1.4. Here I am considering
anticipated life-cycle changes in the wage or changes of short enough duration that
feedback through & can be neglected.
R.E. Hall
1148
(3.9)
Ch. 17:
1149
Table 2
Seasonal averageddeviations from trenda
Quarter
Number of workers employed
Weekly hours
Total hours
-1.45
-0.87
-2.32
0.07
-0.14
-0.07
0.30
0.81
1.11
1.08
0.20
1.28
and workers. Workers could take extra weeks off by quitting one job and delaying
taking another job, but that step dissipates the value o f job-specific capital. The finding
of small reductions in weeks of work in the negative income tax experiments is
not inconsistent with the hypothesis that much larger reductions can occur when the
marginal revenue product o f labor declines in a downturn. One is unprecedented and
unfamiliar, completely new to the environment under which employment arrangements
have evolved; the other is exactly within the historical experience that shaped those
arrangements.
Another reason that panel studies, both survey and experimental, are not good
evidence against elastic short-run supply is the amount of variability they reveal in
annual work effort. According to MaCurdy (1981), the standard deviation of annual
hours of work around the predictions of his labor supply function is several hundred
hours, a significant fraction of the normal level of around 2000 hours. Most of this
noise is variation over time around a worker's own normal level of work. I f the
intertemporal elasticity of labor supply is as low as the numbers in Pencavel's survey,
with respect to substitution between one year and the next, then the deadweight burden
of the unexplained variability of work is extremely high. A more reasonable conclusion
is that the low elasticities apply to life-cycle influences but that much higher elasticities
operate at year-to-year frequencies.
A low intertemporal elasticity of substitution in the short run should also make
workers averse to predictable seasonal variations in their volume of work. Table 2,
taken from Barsky and Miron (1989, Table 2), presents the seasonal averages in percent
deviations from trend by quarter found for the private non-agricultural sector of the
US economy. The United States has a recession every winter comparable to businesscycle recessions. There is a boom in the summer and fall. Although one might suppose
that part of the seasonal movements in hours of work reflects seasonal variations in
preferences for work and leisure, it is hard to see how that would result in more
work in the summer, during the vacation season, and less work in the winter. If
workers had a strong aversion to uneven work schedules, institutions would develop to
smooth employment over the season. The seasonal data suggest reasonable amounts
of intertemporat substitution among the quarters of the year.
1150
R.~HaH
As I noted earlier, the most conspicuous shortcoming of the baseline model is its failure
to understand unemployment. The mechanism by which workers lose jobs in response
to adverse shocks is a promising area to find amplification, and the slow process of
re-employment is surely part of the story of persistent periods of slack.
The baseline neoclassical model fails to deal with unemployment in two ways. First,
it assumes that the labor market clears instantaneously. Even if workers are leaving
some jobs and taking others, the process takes no resources and no time. Second,
the model recognizes no heterogeneity in workers or jobs. The model contains no
ingredients that would suggest that workers should change jobs - that a worker is more
productive in a new job than in the current one. Unemployment cannot be grafted on
to the baseline model. A new model, unfortunately much more complicated, is needed
to deal effectively with unemployment. New work on job destruction, job creation, and
job search has made important advances in this area. Only recently have these ideas
been incorporated in DSGE models. Newly developed models achieve employment
Ch. 17:
1151
Table 3
Alternative measures of quarterly rates of job loss a
Source
Permanent separations, UI system data
CPS tenure survey, 1981
All separations, Current Population Survey
Gross employment reductions, LRD
Permanent layoffs, PSID, 1985
Displaced workers survey, all workers, 1991 1993
Displaced workers survey, workers on the job for at least 3 years, 1991-1993
Quarterly rate of
job loss (%)
t7.23
10.04
8.29
5.66
1.81
0.61
0.59
1152
R.E. Hall
percent per quarter. A large fraction of total job separations arise from temporary or
short-term work.
Rates of job loss rise dramatically at the onset of a recession. Davis and
Haltiwanger's rate of job destruction in manufacturing reached peaks of 11 percent
and 9 percent per quarter in the recessions of 1975 and 1982, from an average level
of 5.7 percent [Davis, Haltiwanger and Schuh (1996)].
The magnitude of job loss when an adverse shock hits the economy is puzzling in
some respects. The great majority of workers have been on the job for 5 years or more
and expect to remain in the same job for many more years [Hall (1982)]. Higher-tenure
workers may have accumulated substantial amounts of job-specific capital, measured as
the difference in the expected present discounted value of earnings at their current jobs
and the values conditional on departing. Evidence in Ruhm (1991) discussed in Hall
(1995) suggests that the typical layoff of a high-tenure worker costs the victim about
1.2 years of earnings, in the form of multiple spells of unemployment and reduced
hourly wages. I f the anticipated value of job-specific capital is divided evenly between
worker and employer, then the typical level of the capital is 2.4 years of earnings, or
around $100 000.
It should take a substantial adverse shock to merit the dissipation of $100 000 of
specific capital. Labor-market institutions should evolve to protect specific capital
against shocks of all kinds, including aggregate ones. Of course, not every job has
the typical amount of match capital. Workers with low tenure or in failing businesses
may be close to the point where separation would be efficient - it would not lower the
joint value of employers and worker. But the evidence at least creates the suspicion that
many of the workers who lose their jobs in a recession do not fall into this category.
As a general matter, it appears that firms tend to lay workers off despite opportunities
to preserve still-valuable job-specific capital. A number of authors have taken this
hypothesis as a point of departure for theories of amplification. In addition, recent work
has considered the role of heterogeneity in the values of job matches - separations are
most likely in the matches whose values are in the lower tail of the distribution, thanks
to idiosyncratic factors.
The basics of the theory of job termination are well developed in labor economics 2.
A core question is the efficiency of terminations - efficiency, as usual, means the
maximization of joint value. Figure 1 displays the analysis of efficient terminations.
The horizontal axis shows earnings available from the next best job in the open
market, net of search costs. The vertical axis shows the worker's marginal product
at this employer. Separation should occur below the 45 line. Whether the separation
is initiated by the worker as a quit or by the employer as a layoff depends on the details
of the employment arrangement 3.
2 For example, Hashimoto and Yu (1980), Hall and Lazear (1984), and McLaughlin (1991).
3 See McLaughlin (1991).
Ch. 17:
1153
Marginalproduct
at firm
45 line
Workerstays,
efficientlY~worke r
departs, "Y
Wageat bestalternativejob
Fig. 1. Efficientseparation.
4 Hall and Lilien (1979) discussed efficient employment arrangements with unilateral information private
to employers.
1154
R.E. Hall
employer's location. Firms accumulate valuable knowledge about their workers' skills.
More subtle employment practices may be needed to protect investments in specific
capital.
When the efficiency of the continuation of the match is a live issue, protection of
specific investments becomes a serious challenge. In that case, some kind of joint or
unilateral procedure is needed to determine if a match should continue or end. I f either
party has the power to end the job (the worker to quit or the employer to terminate),
the party can use that power to deprive the other party of the expected 'return to the
investment. For example, an employer might attract a worker to make an expensive
move by offering a high salary. A year later, the employer might approach the worker
and say that the worker would be terminated unless the worker accepted a much lower
salary. The worker would accept the salary reduction as long as the salary remained
above the value of the next best job, which might involve another expensive move. An
employment arrangement can include severance pay to limit this type of opportunistic
behavior by employers.
3.3.1. Mechanism design and labor contracts"
The discussion of the design of the labor contract has been strongly influenced by the
literature on mechanism design derived from Mirrlees's (1971) famous paper. A key
idea in this literature is that contracts can only be contingent on measures that are
verifiable - it is not enough that the measures be observable. Hart (1983) discusses
the first round of thinking along these lines, where separation or other employment
decisions are made unilaterally by worker or firm, subject to a contract determined in
advance. A more recent elaboration of the theory of the employment relationship in
the mechanism design framework is in the work of Charles Kahn and Gur Huberman
(1988). In their model, the worker's productivity is observed only by the employer,
but the productivity depends on an investment in specific capital observed only by
the worker. Absent both of these information limitations, simple contracts would
give the first-best outcome. If productivity were verifiable, then the wage would be
contingent on actual productivity, and the worker would have the right incentive to
make the investment. If the investment were itself observable, the employer would
reward the worker for making the investment. With both unobservable, the following
more complicated contract delivers the efficient outcome: The parties agree in advance
on a wage to be paid after the investment is made. Upon observing the worker's
productivity later, the employer can either keep the worker and pay the wage, or
discharge the worker. The worker does in fact make the investment and is retained,
which is the efficient outcome.
Gilson and Mnookin (1990) argue that the up or out rule common in law firms is
the result of suppression of renegotiation. In order to induce associates to make firmspecific investments, the firm promises not to offer the associate a salary just above
the best outside salary. Instead, at a predetermined time, the firm chooses between
offering partnership or terminating the associate.
Ch. 17:
1155
Although Kahn and Huberman do not stress the point, suppression of renegotiation
is central to the success of their contract. After the worker has made the firm-specific
investment, the employer could say, "If I have to pay you the wage we agreed upon,
I won't keep you. But if you agree to a lower wage, I will keep you." There is no
violation of the contract in this offer. But if the worker anticipates that the employer is
free to make this offer, the worker will not make the investment and the scheme will
fail.
Considered as a game played only once, the Kahn-Huberman contract fails the test
of credibility - it is not subgame perfect. Suppression of renegotiation requires the
employer to commit not to take a step later that would be rational and permitted under
the terms of the contract. The problem is the same as the one studied extensively
by monetary economists (a central bank needs some way to commit not to create a
monetary surprise later, when such a surprise would be rational later) and in public
finance (the tax authorities need some way to commit not to levy a capital tax later,
when such a tax is the ideal, neutral lump-sum tax later)5.
Although in both the monetary and fiscal settings, there are no formal institutions to
enforce commitments, experience - especially recently - suggests that something like
the favorable equilibrium with commitment can be achieved anyway. Reputations of
policymakers and institutions seem to be an important part of the story - see Barro and
Gordon (1983) in the analogous context of monetary policy. Reputation may explain
the credibility of the suppression of renegotiation as well. If an employer is expected to
remain in business permanently, it will pay for it to develop a reputation for adhering
to policies of not renegotiating. The concept of reputation can be explained in models
of games of repeated play, or in other frameworks 6.
Suppression of renegotiation seems to be an important part of the cultural norms of
the labor market as well. The offer to retain an employee by departing from previously
announced standards of compensation is seen as morally wrong. Standards of ethical
conduct support up-or-out rules in universities and professional practices. It is wrong
to extend a non-tenured faculty member's appointment after denial of tenure, even
though both sides favor it.
Truman Bewley's extensive field study of employment relationships in a depressed
local labor market documents the absence of renegotiation 7. By far the most
common reason given by employers and their advisers for not rewriting employment
arrangements in order to preserve jobs is that lowering wages would destroy morale.
In other words, workers see a departure from the established compensation patterns
as a violation of the rules of the workplace. They think it is wrong to depart from
the principle that employers unwilling to pay promised levels of compensation should
discharge their workers.
5 See Fischer (1980). On the general issue of the value of commitment in games, see Fudenberg and
Tirole (1991, pp. 74-77).
6 See Carmichael (1984).
7 Bewley (1994).
1156
R.E. Hall
Value at firm
Inefficient
45 line
i: qult /
i
i /
Efficient
Efficient
retention
W
............. i ............................................
Inefficient
layoff
/
/
/
/
i
!
i
Efficient layoffquit
Effiaent i
layoff
Contract value,
w
Value at best alternative job
1157
8 The recent literature on contracts in general has gone through a similar transformation see, for
example, Segal and Whinston (1996, 1997).
1158
R.E. Hall
Ch. 17:
Valueat f i r m
1159
X
45 line
in accord with the efficiency condition discussed earlier. The value of each match
begins at a positive level, reflecting match capital created by the worker's search
effort and the employer's recruiting effort. Thereafter, it evolves as idiosyncratic and
aggregate shocks perturb product demand, productivity, and the worker's alternative
opportunities. In their general form, this type of model is hard to handle because the
state of the economy at each moment includes the distribution of matches by current
value, which has a dimension equal to the number o f matches.
Gomes, Greenwood and Rebelo (1997) tackle the hard problem of studying a DSGE
model without the simplifying assumptions of earlier authors. Their workers do not
enjoy the perfect unemployment insurance assumed in earlier models. The value of
a match evolves according to a stochastic process with memory, where every match
is also influenced by an aggregate variable, so the distribution of workers by current
match productivity cannot be simplified.
Efficiency wages have been brought back into the picture in the DSGE framework by
G. Ramey and Watson (1997). Figure 3 strips their model down to its bare essentials.
Workers can enjoy a benefit, X, if they misbehave - for example, X might be the
amount they could steal. Misbehavior is detected with certainty by the employer, but
cannot be proven in court, so it cannot be a contingency in a contract. Unless workers
have a personal value from continuing in the job of at least X, they will take X and
then find another job.
The zone of inefficient separations in the Ramey-Watson model is the area between
the 45 line and the line that is X above the 45 line. A job could have substantial
joint value, say at J in the figure. But a small shock could move that job below the
upper line, causing a separation. The employment relationship is unnecessarily fragile
(compared to its full-information version) because it breaks up whenever the joint value
achieved by the match falls below X,, rather than surviving unless it falls to zero.
In what sense is there amplification in the Ramey-Watson model? Jobs are at
risk to small shocks even though they have positive amounts of joint job-specific
1160
R.E. Hall
Ch. 17:
1161
1162
R.E. HaH
a contraction is a period of matching frenzy, with both job destruction and job creation
at abnormally high levels.
Evidence cited by Hall (1991a) on the cyclical behavior of job-finding rates is mixed.
Blanchard and Diamond (1990) report that a recession that raises unemployment
by two percentage points reduces the job-finding rate from a normal level of
24.0 percent per month to 21.8 percent per month. On the other hand, Hall reports
a regression relating the Davis-Haltiwanger measure of the volume of job-worker
matching to the level of unemployment. He finds an increasing margimll benefit from
the stock of unemployment on the flow of new matches.
Although the amplification mechanism based on endogenous improvements in
search efficiency during recessions is on uncertain ground, the evidence just reviewed
raises serious doubts about the opposite (and conventional) view that recessions are
times when jobs become much harder to find. A reasonable intermediate view is that
search efficiency is about the same at high and low unemployment. This disposes of
a potential attenuation mechanism - job matches would be more stable in recessions
than normal times if job search became more costly in recessions.
4. Persistence
The time-series properties of the principal macro variables are reasonably well
understood. Unemployment is stationary - it returns about one third of the way to
its normal level each year after a shock displaces it 9. Output and employment have
both cyclical and highly persistent - possibly integrated - components. The persistence
mechanisms in a fluctuations model need to be able to explain the stationary but serially
correlated movements of unemployment and the corresponding cyclical movements of
output and employment. The highly persistent components of employment and output
derive from slow-moving changes in preferences and technology and are not in the
domain of the persistence mechanism of the fluctuations model.
Although a number of authors have identified sources of persistence other than the
mechanics of job search [such as Burnside and Eichenbaum (1996) and Saint-Paul
(1996)], I will focus mainly on this single topic, which dominates current thinking
about persistence.
4.1. Time-consuming matching in the labor market
One of the most interesting and successful recent developments in the labor side
of macroeconomics has been the development of modern models of job search.
Diamond (1982b) and Mortensen (1982) are the starting points. Good summaries are
Ch. 17:
1163
in Mortensen (1986), Pissarides (1990), and Romer (1996, chapter 10). My discussion
will be brief because Mortensen's chapter in this volume covers this area in detail.
In the standard matching model, a random meeting occurs between a job seeker and
an employer. A match occurs if it increases the joint value of the two parties, in which
case they divide the joint surplus. The simplest model has a constant probability that a
job-seeker will be matched. The persistence parameter for aggregate unemployment the serial correlation coefficient - is controlled by the job-finding probability. The
matching model provides a simple and elegant persistence mechanism for a general
equilibrium macro model.
From the start, it has been clear that it is an uphill battle to use the matching model
to explain the actual persistence of unemployment in the USA. To see the relation
between the job-matching rate and the serial correlation of unemployment, consider the
following elementary model. Let dt be job destruction, nt be employment, ~ be the fixed
supply of labor, ut = ~ - nt be unemployment, and f the per-period job-finding rate.
Then employment this period consists of employment last period plus those among
the unemployed who found new jobs less the number of jobs destroyed:
nt =nt 1 + f u r - d r
(4.1)
ut = (1 - f ) u t
(4.2)
or
1 +dt.
Thus, if job destruction is white noise, unemployment follows an AR(1) process with
serial correlation 1 - f .
As I noted earlier, the average job-finding rate is about 24 percent per month. The
monthly serial correlation of unemployment is 0.988, which would imply a job-finding
rate of only 1.2 percent per month. There is a discrepancy of a factor of 20 between
the time-series properties of unemployment and the job-finding rates experienced by
individuals. Cole and Rogerson (1996) have studied this discrepancy and concluded,
"Our main finding is that the [matching] model can account for the business cycle facts,
but only if the average duration of a non-employment spell is relatively high - about
nine months or longer." With an average job-finding rate of 24 percent per month, the
average duration is actually much less. Something is missing from the simple model.
4.2. The importance o f secondary j o b loss f o r persistence
Hall (1995) suggests that the missing element is induced secondary job loss. The first
job that a recently discharged worker finds may be an explicitly temporary job, or it
may turn out to be a bad match once the worker joins the firm, or the match may break
soon because, in its early stages, it has little job-specific capital. There is evidence of
large amounts of secondary spells of unemployment following an initial impulse.
R.E. Hall
1164
12
10
8
6
o
4
2
0
72
73
74
75
76
77
78
79
80 81
Quarters
82
83
84
85
86
87
88
10 As Davis, Haltiwanger and Schuh note, plant level employment is highly persistent; it is essentially
a random walk. Hence the flow of reductions is close to white noise.
tl The data come from the Current Population Survey and are published in Employment and Earnings.
They refer to workers who became unemployed as a result o f permanent layoff, whose unemployment
began within five weeks of the survey.
Ch. 17:
1165
3.5
2.5
2
0.5 l
0
~llll 1111 II 1IN II I'~ 1111',~'111111 fill llllll Ill ll'~Ill',II~ll lllllIlllllll~
Quarters
whereas new unemployment is measured economy-wide 12. A systematic lag of nonmanufacturing behind manufacturing would explain some part of the lag. Second,
many workers who lose their jobs do not become unemployed- they move immediately
to other jobs or leave the labor force. During the period of slack labor markets
following a burst of employment reductions, a larger fraction of job-losers become
unemployed. Third, permanent job loss has important delayed effects. Many of the
workers who move quickly to other jobs have taken temporary work, either jobs with
predetermined short terms, or those with naturally high turnover. Those who left the
labor force upon loss of a long-term job often re-enter the labor force.
The micro and macro evidence suggests strongly that terminations beget later
terminations. When an event breaks a set of long-term employment relationships, the
workers released into the labor market will form new relationships. Many o f the new
jobs will prove to be short-lived.. First, it.may make sense for an individual totake a
temporary job while looking for a new permanent job. Second, a worker long out of
the market may experiment with alternative types of work before finding a good longterm match. Third, employers may have explicit policies of hiring many candidates
and keeping only the fraction who prove to be well matched. Fourth, immediately after
being hired, the typical worker will be close to the margin for discharge, either by the
standards o f the efficient separation model or the models of suppressed renegotiation
or efficiency wages. Both the systematic accumulation of match-specific capital and
the random accumulation o f rent will have had little time to occur. Low-tenure workers
1166
R.E. Hall
are the logical candidates for separation - last hired, first fired is the rational separation
rule under broad conditions.
A specific adverse event will create an immediate burst of terminations, followed
by the second, third, and subsequent rounds of terminations. Induced subsequent job
losses seem to be a promising explanation of persistence. Following a single adverse
shock, employment will be depressed and unemployment elevated by subsequent
rounds of adjustment in the labor market.
A glance at the data show that a simple model of transitions between jobs and search
cannot be faithful to even the most conspicuous features of the market's dynamics.
Rates of separation from jobs decline sharply with tenure on the job, and job-finding
rates fall with the duration of unemployment. Part of the duration dependence is
genuine and part reflects the sorting of heterogeneous workers 13. Moreover, previous
history appears to influence transition rates. For example, workers terminated from
long-term jobs have lower job-finding rates than do other searchers, are more likely to
lose subsequent jobs than are other short-tenure workers, and have lower job-finding
rates in subsequent spells of unemployment.
Some basic properties of job loss have emerged in this review of the evidence.
Microeconomic studies of serious job loss show significant downstream effects on
the subsequent experiences of individuals in the labor market. Loss of a long-term
job leads to a period of episodic employment, periods of job search or time out of
the labor market, and lower earnings when working. The effects extend for at least
four years. In the macroeconomic evidence, bursts of gross employment reductions
coincide with abnormal levels of serious job loss. The downstream effects visible in
time series data for unemployment are similar to the effects found in micro data for
individuals.
The macro data show occasional sharp disruptions of employment followed
by long periods of rebuilding of employment relationships. This rebuilding may
be an important part of the propagation mechanism of the business cycle. The
length of time that the economy takes to recover from an adverse shock has
perplexed macroeconomists for many years. Rebuilding may help solve this puzzle
of persistence.
Den Haan, Ramey and Watson (1997) have developed a DSGE model with realistic
persistence in which efficient job destruction interacts with capital formation. They
provide an alternative explanation of induced secondary job loss. A key property
of their model is that the idiosyncratic shock at the level of the plant or individual
job match is unpredictable white noise. An aggregate shock results in a first rotmd
of job destruction. There follows a period of high interest rates during which the
threshold value for the idiosyncratic shock changes so as to increase the probability
of job destruction. Until the aggregate shock wears off, job destruction continues
at abnormally high levels. The model is successful in explaining the persistence of
Ch. 17:
1167
job destruction and unemployment, without invoking unrealistically low rates o f job
finding. On the other hand, it relies on highly persistent technology shocks (with a
quarterly serial correlation of 0.95) in order to generate persistent changes in interest
rates. The model's assumption that the idiosyncratic component of job match value is
white noise is also intrinsic to the model's success in explaining persistence. Under the
more realistic assumption o f a random walk for the idiosyncratic component, all of the
job destruction triggered by an aggregate shift in technology would occur immediately
and there would be no persistent subsequent job destruction.
5. Conclusion
In the economies of the USA and other modem countries, large responses, especially
recessions, seem to result from small impulses. Their effects on the economy
must operate through an amplification mechanism. The fragility of the employment
relationship seems to underlie that sensitivity. Despite substantial job-specific capital
in the majority of jobs, millions of workers are released into the labor market during
each contraction. The resulting unemployment is persistent. Not only does it take time
for workers displaced by a recession to find new jobs, but the average one has to find
several new jobs, a process that stretches over about four years.
DSGE models have come a long way since Kydland and Prescott (1982) in incorporating labor-market frictions and giving correspondingly more realistic portrayals of
the economy. Recognition of the heterogeneity of workers and jobs has been central
to this improvement in macro modeling.
Acknowledgments
This research was supported by the National Science Foundation under grant SBR9410039 and is part of the NBER's research program in Economic Fluctuations and
Growth. I am grateful to the editors for helpful comments.
References
Aghion, E, M. Dewatripont and P. Rey (1990), "On renegotiation design", European EconomicReview
34:322-329.
Aiyagari, S.R., L.J. Christiano and M. Eichenbaum (1992), "The output, employment, and interest rate
effects of government consumption", Journal of Monetary Economics 30:73-86.
Akerlof, G.A., and J.L. Yellen (1990), "The fair wage-efforthypothesis and unemployment",Quarterly
Journal of Economics 105:255-283.
Andolfatto, D. (1996), "Business cycles and labor-market search", American Economic Review 86:
112-132.
Barro, R.J., and D.B. Gordon (1983), "Rules, discretion, and reputation in a model of monetarypolicy",
Journal of Monetary Economics 12:101-121.
1168
R.E. Hall
Barskry, R.B., and J.A. Miron (1989), "The seasonal cycle and the business cycle", Journal of Political
Economy 97:503-534.
Basu, S. (1996), "Procyclical productivity: increasing returns or cyclical utilization?", Quarterly Journal
of Economics 111:719-751.
Basu, S., and J.G. Fernald (1997), "Returns to scale in U.S. production: estimates and implications",
Journal of Political Economy 105:249-283.
Benhabib, J., R. Rogerson and R. Wright (1991), "Homework in macroeconomics: household production
and aggregate fluctuations", Journal of Political Economy 99:1166-1187.
Beveridge, S., and C.R. Nelson (1981), "A new approach to the decomposition of economic time series
into permanent and transitory components with particular attention to the measurement of the 'business
cycle'", Journal of Monetary Economics 7:151-174.
Bewley, T. (1994), A Field Study on Downward Wage Rigidity (Yale University).
Blanchard, O.J., and P.A. Diamond (1990), "The cyclical behavior of the gross flows of U.S. workers",
Brookings Papers on Economic Activity 1990(2):85-155.
Burnside, C., and M. Eichenbaum (1996), Factor hoarding and the propagation of business cycle shocks",
American Economic Review 86:1154-1174.
Burnside, C., M. Eichenbaum and S.T. Rebelo (1993), "Labor hoarding and the business cycle", Journal
of Political Economy 101:245-273.
Caballero, R.J., and M.L. Hammour (1996), "On the timing and efficiency of creative destruction",
Quarterly Journal of Economics 111:805-852.
Caballero, R.J., and M.L. Hammour (1998), "Jobless growth: appropriability, factor substitution, and
unemployment", Carnegie-Rochester Conference Series on Public Policy 48(June):51 94.
Caballero, R.J., E. Engel and J. Haltiwanger (1997), "Aggregate employment dynamics: building from
microeconomic evidence", American Economic Review 87:115-137.
Campbell, J.Y. (1994), "Inspecting the mechanism: an analytical approach to the stochastic growth
model", Journal of Monetary Economics 33:463-506.
Card, D. (1994), "Intertemporal labor supply", in: C.A. Sims, ed., Advances in Econometrics: Sixth
World Congress of the Econometric Society, vol. II (Cambridge University Press) 49-78.
Carmichael, H.L. (1984), "Reputations in the labor market", American Economic Review 74:713-725.
Cogley, T., and J.M. Nason (1995), "Output dynamics in real-business-cycle models", American Economic
Review 85:492-511.
Cole, H.L., and R. Rogerson (1996), "Can the Mortensen-Pissarides matching model match the
business cycle facts?", Staff Report 224 (Federal Reserve Bank of Minneapolis, Research Department,
December).
Danthine, J.-E, and J.B. Donaldson (1995), "Non-Walrasian economies", in: T.E Cooley, ed., Frontiers
of Business Cycle Research (Princeton University Press, Princeton) 217-242.
Davis, S.J., J.C. Haltiwanger and S. Schuh (1996), Job Creation and Destruction (MIT Press,
Cambridge, MA).
den Haan, W.J., G. Ramey and J. Watson (1997), Job Destruction and Propagation of Shocks (University
of California, San Diego).
Devine, T.J., and N.M. Kiefer (1991), Empirical Labor Economics: The Search Approach (Oxford
University Press, New York).
Diamond, P.A. (1982a), "Wage determination and efficiency in search equilibrium", Review of Economic
Studies 29:217-227.
Diamond, EA. (1982b), "Aggregate demand management in search equilibrium", Journal of Political
Economy 90:881-894.
Fischer, S. (1980), "Dynamic inconsistency, cooperation and the benevolent dissembling government",
Journal of Economic Dynamics and Control 2:93-107.
Fudenberg, D., and J. Tirole (1991), Game Theory (MIT Press, Cambridge, MA).
Gilson, R.J., and R.H. Mnookin (1990), "The implicit contract for corporate law firm associates: ex post
opportunism and ex ante bonding", in: M. Aoki, B. Gustafsson and O. Williamson, eds., The Firm
1169
as a Nexus of Treaties (Swedish Collegium for Advanced Study in Social Sciences Series, London)
209-236.
Gomes, J., J. Greenwood and S.T. Rebelo (1997), Equilibrium Unemployment (University of Rochester,
Rochester, NY).
Hall, R.E. (1982), "The importance of lifetime jobs in the U.S. economy", American Economic Review
72:716-724.
Hall, R.E. (1988), "The relation between price and marginal cost in U.S. industry", Journal of Political
Economy 96:921-947.
Hall, R.E. (1990), "Invariance properties of Solow's productivity residual", in: P. Diamond, ed.,
Growth/Productivity/Unemployment: Essays to Celebrate Robert Solow's Birthday (MIT Press,
Cambridge, MA) 71-112.
Hall, R.E. (1991a), "Labor demand, labor supply, and employment volatility", National Bureau of
Economic Research Annual: 17-46.
Hall, R.E. (1991b), "Substitution over time in consumption and work", in: L. McKenzie and S. Zamagni,
eds., Value and Capital Fifty Years Later (MacMillan) 239-267.
Hall, R.E. (1995), "Lost jobs", Brookings Papers on Economic Activity 1995(1):221-273.
Hall, R.E. (1997), "Macroeconomic fluctuations and the allocation of time", Journal of Labor Economics
15 :$223-$250.
Hall, R.E., and E.P. Lazear (1984), "The excess sensitivity of layoffs and quits to demand", Journal of
Labor Economics 2:233-257.
Hall, R.E., and D.M. Lilien (1979), "Efficient wage bargains under uncertain supply and demand",
American Economic Review 69:868-879.
Hart, O. (1983), "Optimal labour contracts under asymmetric information: an introduction", Review of
Economic Studies 50:3-35.
Hashimoto, M., and B. Yu (1980), "Specific capital, employment contracts, and wage rigidity", Bell
Journal of Economics 11:536-549.
Heckman, J.J., and B. Singer (1985), Longitudinal Analysis of Labor Market Data (Cambridge University
Press, New York).
Kahn, C., and G. Huberman (1988), "Two-sided uncertainty and 'up-or-out' contracts", Journal of Labor
Economics 6:423-444.
Killingsworth, M.R., and J.J. Heckman (1986), "Female labor supply: a survey", in: O. Ashenfelter and
R. Layard, eds., Handbook of Labor Economics, vol. 1 (North-Holland, Amsterdam) ch. 2, 103-204.
King, R.G., and S.T. Rebelo (1999), "Resuscitating real business cycles", ch. 14, this Handbook.
Kydland, EE., and E.C. Prescott (1982), "Time to build and aggregate fluctuations", Econometrica
5:1345-1370.
Lilien, D.M. (1982), "Sectoral shifts and cyclical unemployment", Journal of Political Economy 90:
777-793.
Lucas, R.E. (1972), "Expectations and the neutrality of money", Journal of Economic Theory 4:103 124.
MacLeod, W.B., J.M. Malcomson and P. Gomme (1994), "Labor turnover and the natural rate of
unemployment: efficiency wage versus frictional unemployment", Journal of Labor Economics 12:
276 315.
MaCurdy, T.E. (1981), "An empirical model of labor supply in a life cycle setting", Journal of Political
Economy 89:1059-1085.
McLaughlin, K.J. (1991), "A theory of quits and layoffs with efficient turnover", Journal of Political
Economy 99:1-29.
Merz, M. (1995), "Search in the labor market and the real business cycle", Journal of Monetary
Economics 36:269-300.
Mirrlees, J.A. (1971), "An exploration in the theory of optimum income taxation", Review of Economic
Studies 38:175-208.
Mortensen, D.T. (1982), "Property rights and efficiency in mating, racing, and related games", American
Economic Review 72:968-979.
1170
R.E. Hall
Mortensen, D.T. (1986), "Job search and labor market analysis", in: O. Ashenfelter and R. Layard, eds.,
Handbook of Labor Economics (North-Holland, Amsterdam) 849-919.
Mortensen, D.T., and C.A. Pissarides (1994), "Job creation and destruction in the theory of
unemployment", Review of Economic Studies 61:397-415.
Mulligan, C.B. (1995), "The intertemporal substitution of work - what does the evidence say", University
of Chicago Population Research Center Discussion Paper 95-3.
Pencavel, J. (1986), "Labor supply of men: a survey", in: O. Ashenfelter and R. Layard, eds., Handbook
of Labor Economics (North-Holland, Amsterdam) 3-102.
Phelps, E.S. (1994), Structural Slumps: The Modern Equilibrium Theory of Unemployment, Interest,
and Assets (Harvard University Press, Cambridge, MA).
Picard, P. (1993), Wages and Unemployment: A Study in Non-Walrasian Macroeconomics (Cambridge
University Press, Cambridge, MA).
Pissarides, C.A. (1990), Equilibrium Unemployment Theory (Blackwell, Oxford).
Ramey, G., and J. Watson (1997), "Contractual fragility, job destruction, and business cycles", Quarterly
Journal of Economics 112:873411.
Rogerson, R. (1988), "Indivisible labor, lotteries, and equilibrium", Journal of Monetary Economics
21:3-16.
Romer, D. (1996), Advanced Macroeconomics (McGraw-Hill, New York).
Rotemberg, J.J., and M. Woodford (1996), "Real-business-cycle models and the forecastable movements
in output, hours, and consumption", American Economic Review 86:71-89.
Ruhm, C. (1991), "Are workers permanently scarred by job displacements?", American Economic Review
81:31%324.
Saint-Paul, G. (1996), "Efficiency wages as a persistence mechanism", in: H.D. Dixon and N. Rankin,
eds., The New Macroeconomics: Imperfect Markets and Policy Effectiveness (Cambridge University
Press, Cambridge) 186-205.
Sargent, T.J. (1979), Macroeconomic Theory (Academic Press, New York).
Segal, I.B., and M.D. Whinston (1996), "Naked exclusive contracts and buyer coordination", Discussion
Paper 1780 (Harvard Institute of Economic Research).
Segal, I.B., and M.D. Whinston (1997), Exclusive Dealing and Specific Investments (Harvard University
and University of California, Berkeley).
Shapiro, C., and J.E. Stiglitz (1984), "Equilibrium unemployment as a worker discipline device", American
Economic Review 74:433-444.
Solow, R.M. (1957), "Technical change and the aggregate production function", Review of Economics
and Statistics 39:312-320.
Woodford, M. (1994a), "Notes on dynamic efficiency wage models". Unpublished.
Woodford, M. (1994b), "Structural slumps", Journal of Economic Literature 32:1784-1815.
Wright, R. (1986), "Job search and cyclical unemployment", Journal of Political Economy 94:38 55.
Chapter 18
Contents
Abstract
Keywords
Introduction
1. O E C D facts
2. The equilibrium rate o f unemployment
2.1. Job destruction and job creation conditions
2.2. Generalized Nash bargaining
2.3. Fundamental determinants of unemployment
3. Employment fluctuations
3.1. Stochastic equilibrium
3.2. The Beveridge curve
3.3. Job creation and job destruction flows
3.4. Quits and worker flows
4. Explaining the data
4.1. Explaining job flows data
4.2. Capital accumulation and shock propagation
5. Technological progress and job reallocation
5.1. Disembodied technology
5.2. Adoption through "creative destruction"
6. O E C D unemployment differences
6.1. 'Skill-biased' technology shocks
6.2. Mean-preserving shocks to idiosyncratic productivity
1172
1172
1173
1174
1183
1185
1188
1192
1194
1194
1196
1197
1198
1200
1201
1203
1207
1208
1210
1213
1215
1218
1172
6.3. Other influences
7. Concluding remarks
Appendix A. Mathematical appendix
A.1. Mean-preservingshifts in productivity
A.2. Labor's bargaining strength
References
1220
1223
1223
1223
1224
1225
Abstract
The purpose of this chapter is twofold. First, it reviews the model of search and
matching equilibrium and derives the properties of employment and unemployment
equilibrium. Second, it applies the model to the study of employment fluctuations and
to the explanation of differences in unemployment rates in industrialized countries.
The search and matching model is built on the assumptions of a time-consuming
matching technology that determines the rate of job creation given the unmatched
number of workers and jobs; and on a stochastic arrival of idiosyncratic shocks that
determines the rate of job destruction given the wage contract between matched firms
and workers. The outcome is a model for the flow of new jobs and unemployed workers
from inactivity to production (the 'job creation' flow) and one for the flow of workers
from employment to unemployment and of jobs out of the market (the 'job destruction'
flow). Steady-state equilibrium is at the point where the two flows are equal.
The model is shown to explain well the employment fluctuations observed in the
US economy, within the context of a real business cycle model. It is also shown that
the large differences in unemployment rates observed in industrialized countries can
be attributed to a large extent to differences in policy towards employment protection
legislation (which increases the duration of unemployment and reduces the flow into
unemployment) and the generosity of the welfare state (which reduces job creation).
It is argued that on the whole European countries have been more generous in their
unemployment support policies and in their employment protection legislation than the
USA. The chapter also surveys other reasons given in the literature for the observed
levels in unemployment, including mismatch and real interest rates.
Keywords
J E L classification: J63, J64, J65, J68, E24, E32, J41
1173
Introduction
Market economies experience large employment fluctuations and average unemployment rates that are often different from those experienced by apparently similar
economies. The search and matching framework provides a convenient lens through
which to view explanations of such differences. Our purpose in this chapter is twofold:
first, to present the essential concepts that underlie the framework and, second, to use
the framework to suggest answers to the questions posed by the data.
Existing employment relationships command monopoly rents because of search
and recruiting investments, hiring and firing costs, and other forms of match-specific
human capital formation. The surplus that accrues is allocated between the parties
to the employment relationship by a wage contract. Given a particular wage rule,
employers provide jobs and recruit workers while workers search for employment. At
the same time, an existing employer-worker match ends when sufficiently bad news
arrives about their expected future. These job creation and job destruction decisions
generate worker flows into and out of employment which depend on the current value
of the employed stock. When the two flows differ, employment dynamics are set in
motion which, under a reasonable set of conditions, lead to a unique steady-state
employment level. These properties characterize the equilibrium model of job creation
and job destruction applied in the chapter.
The search and matching approach owes its origins to the pioneering works of
Stigler (1962), Phelps (1968) and Friedman (1968) and was already at an advanced
state when the Phelps et al. (1970) volume was published. The equilibrium analysis
of the current vintage of models, however, did not start until the early 1980s, when
models by Diamond (1982a,b), Mortensen (1982a,b) and Pissarides (1984a,b) explored
the properties of two-sided search and characterized the nature and welfare properties
of market equilibrium. Despite a flurry of activity since then, there are still many
important questions that are unexplored. One such question is the dynamics of worker
movement in and out of the labor force, of which, despite its empirical importance
[Clark and Summers (1979), Blanchard and Diamond (1989)] and some attempts to
model it by Burdett et al. (1984), Pissarides (1990, Chapter 6) and Andolfatto and
Gomme (1996), our knowledge is still scant.
Virtually all search equilibrium models assume an exogenous labor force, which
is used to normalize all aggregate quantities, and model either the equilibrium
employment or unemployment rate. It is simple enough to superimpose on this
structure a neoclassical labor-supply decision, as is done, for example, by Andolfatto
(1996) and Merz (1995), but still the worker flow from the labor force to out
of the labor force is ignored. Given this restriction, we can interchangeably talk
either about employment equilibrium or about unemployment equilibrium. In the
latter case, the equilibrium is often referred to as a "natural rate" equilibrium,
following Friedman's introduction of the term in 1968. Indeed, the equilibrium that
we shall describe corresponds closely to the one advocated by Friedman (1968) and
Phelps (1967, t968).
1174
1. O E C D facts
What are the main facts about employment and unemployment that the search and
matching approach can help explain? In Figures 1 and 2, labor force and unemployment
time series from 1960 to 1995 are illustrated for the USA, Japan and a weighted
average of the four largest European economies, Germany, France, Italy and the United
Kingdom.
The four European countries are grouped together because their unemployment
and labor force experiences have been sufficiently similar to each other. Comparable
data can also be found for most of the other members of the European Union, in
particular Spain, the Benelux, and Scandinavian countries. The experience of Spain
and the Scandinavian countries, however, has been different from that of the big
four, essentially for labor-market policy reasons. In Spain, excessive safeguarding of
the rights of workers through the legal system created a sharp distinction between
insiders and outsiders and led to low aggregate job creation. In the Scandinavian
countries large-scale active labor market policies held unemployment artificially low
until recently. Since we shall not address issues of labor market policy in any detail, we
decided not to aggregate those countries with the four large economies. The experience
1175
66\\\ \ \ / / f - ~ - ~ \
. /"
.-"'USA
64
\\~
'
62
.g
Japan
/"
60
Europe
58 -t~
56
54
60
//
65
70
75
80
85
90
95
Year
Fig. 1. Labor force p~ticipation rates, 1960-1995.
of the Benelux countries is sufficiently close to that of France and Germany, to the
extent that data from them will not add to the information given here.
Figure 1 shows the sharply contrasting participation experience of the USA on
the one hand and Europe and Japan on the other. Whereas in the early 1960s the
participation rates in Europe and the USA were essentially the same, since then
participation in the USA has been on an upward trend and in Europe on a downward
trend. The upward trend in the USA was driven largely by the female participation
rate, whereas in Europe, where female participation rates have also increased, the
downward trend was driven by early retirements among men and by later school
leaving. In Japan the participation rate is uniformly above the European rate but its
dynamic behavior since 1960 has been very similar to the European rate. The figure
also shows some evidence of cyclical variations in the participation rate: it is these
cyclical movements that the search and matching approach could in principle handle,
but has so far ignored.
The trend changes, in the USA in particular, are more likely the outcome of lifetime
labor supply decisions that are independent of the labor market frictions that underlie
the search and matching approach. In Europe, however, much of the decline in laborforce participation has been the result of policy incentives or of private responses to the
rise in unemployment ("discouragement"). The policy to encourage early retirement
was also largely in response to the rising unemployment, so the fall in participation
can be partly attributed to the same factors that increased unemployment during this
period. But exogenous labor supply changes have also played an important role, as
comparison of Figures 1 and 2 shows. The trend decline in labor force participation
1176
10
/- ~ J Europe
I \
/
/
\
]II
/I /
\\ //
?~/1~\1
l
[
//\\
I \j\ \\\
\\\
'\
IilI\ \\\1 I/
\~
\\
1/
\X
\\ ~ / / U S A /
Japan
t
60
65
70
75
i - Y - F q ~ - i
80
85
90
95
Year
Fig. 2. Standardizedunemploymentrates, 1960-1995.
began before the unemployment rise and it was accompanied by a fall in annual hours
of work for those that remained in the labor force.
We shall briefly return to the question of participation changes in the final section. In
the remainder of this section we look at the behavior of unemployment and at gross job
creation and job destruction flows normalized by the labor force. The aim is to point
out key features of the data ("stylized facts") that will guide the model presentation
in the rest of the chapter.
Figure 2 plots the unemployment rates for the three country groups, as far as possible
adjusted to the same (US) definition. Table 1 gives data for more countries for two
periods that were approximately in the same cyclical phase. The contrast is clear.
Whereas in the USA and Japan unemployment is a cyclical variable without trend,
in Europe the biggest changes in unemployment over the last thirty years were due
to changes in the average level of unemployment across cycles. This latter feature
of the European time series led most who analyzed this problem to conclude that the
changes in European unemployment are changes in the "natural rate", not changes in its
cyclical component [Layard et al. (1991), Phelps (1994), Blanchard and Katz (1997)].
The approach that we describe in this chapter is motivated by this observation and is
especially suitable for the analysis of changes in the natural rate. Inflation, expectations
errors and other nominal influences are ignored.
The net changes in employment over the cycle conceal large movements in gross job
creation and job destruction, as well as worker turnover for other reasons. Information
on this feature of labor markets sheds light on the appropriate flow models that should
be used to analyze aggregate labor market changes. This feature of labor markets has
been emphasized by Davis and Haltiwanger (1992) in particular, but also by others
1177
Table 1
OECD unemployment,1974-79 to 1986~90a,b
Country
Unemploymentrate 1974-1979
1.5
81.8
Europe
Austria
Belgium
6.3
41.1
Denmark
5.5
44.7
Finland
4.4
-2.3
France
4.5
77.8
Germany
3.2
61.2
Ireland
7.6
75.7
Italy
4.6
51.5
Netherlands
5.1
54.5
Norway
1.8
66.5
Spain
5.3
126.1
Sweden
1.5
12.5
Switzerland
1.0
64.2
United Kingdom
5.1
54.5
6.7
-14.4
Others
USA
Cmlada
7.2
14.2
Australia
5.0
36.4
Japan
1.9
27.4
since then. Recently Contini et al. (1995) have assembled data on job reallocation for
several countries. Their summary Table is shown in Table 2 and the results are also
summarized in Figure 3.
For most countries the job flow data are calculated from establishment level flows,
though for some only firm-level data were available. Annual gross job creation reflects
employment change only in the establishments or firms that are new entrants or that
have experienced an increase in employment over the period. The job creation rate
is defined as the sum o f the gross increase in employment expressed as a percentage
of the total labor force. Similarly, gross job destruction includes only units that have
experienced a decrease in employment and the j o b destruction rate is equal to the
gross decrease in employment as a percentage of the employment level. By definition
1178
Table 2
Net and gross job flows, OECD, late 1980sa
Country
Period
Japan
UK
Germany
Finland
Italy
USA
Canada
France
Sweden
Denmark
New Zealand
1985-1992
1985-1991
1983-1990
1986-1990
1984-1993
1984-1991
1983-1991
1984-1992
1985-1992
1983-1989
1987-1992
Job creation
8.64
8.70
9.00
10.40
11.90
13.00
14.50
13.90
14.50
16.00
15.70
3.39
2.10
1.50
- 1.60
0.81
2.60
2.60
0.70
-0.10
2.20
-4.10
Gross
reallocation
13.89
15.30
16.50
22.40
22.99
23.40
26.40
27.10
29.10
29.80
35.50
a Source: Contini et al. (1995), Table 3.1 [derived mainly from OECD Employment Outlook (1987,
1994)].
then, the net growth rate in employment is the difference between the job creation rate
and the job destruction rate. International comparisons of data of this kind are fraught
with difficulties and Contini et al. (1995, p. 18) warn that the numbers for Japan and
the United Kingdom are probably understated and for France and New Zealand are
overstated. So if anything, the small differences shown in Table 2 are likely to be
overstated.
Notwithstanding the statistical problems, the results show that Japan has low gross
job creation and job destruction rates, despite high net job creation. The United
Kingdom and Germany, also with positive net job creation, have low gross flows. But
the rest of the countries have high gross job flows, comparable to those of the USA.
There does not seem to be any relation between the volume of gross reallocation and
the net employment change, and the USA does not appear unusually turbulent when
compared to other countries. These findings are also illustrated in Figure 3.
Some regularities emerge from the international comparison of job creation and job
destruction rates. These findings apply to comparisons of economy-wide job creation
and job destruction flows but are also consistent with the more detailed analysis of
Davis et al. (1996) for US manufacturing flows.
First, the flow data always exclude the public sector, where job reallocation is small.
In some European countries the public sector employs a large fraction of the labor force
(8% in Japan, 8.5% in the USA, 7.9% in Germany, 11% in the UK, 22.5% in Italy;
the highest share in the European Union is in Denmark, 31%).
1179
25
20
15
CA
USA
i<
10 UK +GE//~ +FI
/
5
1~0
15
'
20,
Job destruction rate (%)
25
Second, gross job reallocation is inversely correlated with capital intensity: service
jobs create and destroy more jobs than manufacturing does.
Third, smaller and younger establishments create and destroy more jobs than larger
and older plants; about one-third of job creation and job destruction is due to plant
entry and exit. So in international comparisons countries with a larger fraction of
smaller firms (e.g. Italy) are likely to have a larger job reallocation rate than countries
with larger firms (e.g. the USA).
Fourth, at the individual level, the main cause of job turnover is idiosyncratic shocks,
i.e. shocks that do not appear correlated with common economy-wide or sector-specific
shocks, or with other common characteristics across firms. The implication of this
fact is that the regularities listed above, as well as the business cycle, explain less
than half the variance of gross job creation and job destruction across production
units. Aggregate and cyclical shocks explain a small fraction of the variance, about
10 percent. Measurable firm characteristics, such as size and age, explain more, but
still less than half.
Fifth, although younger plants are more likely to create and destroy jobs, there is
large persistence in job creation and job destruction. The idiosyncratic shocks that
cause job reallocation do not reverse shortly after they occur. In both the USA and
Italy (the only two countries with comparable data on this issue), about 70 percent of
1180
jobs created in one year are still active the next year and about 55 percent are active
two years later. Persistence rates for job destruction are slightly higher.
The cyclical properties of job flows, which is of primary concern in the analysis
of employment fluctuations, are not clear-cut in the empirical data so far assembled.
A fact that seems to be universal is that job creation and job destruction flows are
negatively correlated with each other. Thus, recessions are times when job destruction
rates rise and job creation rates fall, and vice versa for expansions. More controversial,
but potentially more interesting, is the finding that job destruction is more "volatile",
in the sense that even when abstracting from growth, the length of time when job
destruction is the dominant flow is shorter than the length of time when job creation
is the dominant flow. Since on average over the cycle job destruction and job creation
rates must be equal, it follows that job destruction rates must peak at higher values
than job creation rates, which are more flat. This asymmetry is consistent with the
observation that recessions are on average of shorter duration than booms and has
attracted a lot of attention in the empirical literature, where, following Davis and
Haltiwanger (1990), it is often reported as a negative correlation between gross job
reallocation and net job reallocation. However, the negative correlation, although a
strong feature of the US manufacturing data, is not universal. The "asymmetry" of
job creation and job destruction rates here is simply taken to mean that the difference
between job destruction and job creation when positive is larger and of shorter duration
than when it is negative.
One final observation on the international comparison of job flows is of interest.
There does not appear to be a significant correlation across countries either between
the level of unemployment on the one hand and the gross job reallocation rate on the
other or between labor productivity growth and the job reallocation rate. There does
seem, however, to be a correlation between the gross job reallocation rate and the rate
of long-term unemployment: countries with lower job reallocation rates seem to have,
on average, longer unemployment durations [Garibaldi et al. 1997)].
Comparative data on worker flows are even less reliable than comparative data on
job flows, even though the definition of worker flows can be a lot less ambiguous
than the definition of job flows. The gross flow of workers in and out of employment,
defined analogously to the gross flow of jobs, is necessarily larger than the job flow. The
difference is, however, large. Contini et al. (1995, p. 108) report that in both the USA
and the major European economies, the worker flow is about three times as big as the
job flow. There is some evidence that worker flows are bigger for the USA than for the
European countries or Japan, and also that in the USA there is more movement in and
out o f unemployment and the labor force. The latter claim, however, may be based
on the different kind of question that is often asked about participation in national
surveys. Two interesting aggregate facts that have emerged from the study of worker
flows, bearing in mind the paucity of the data, are that gross unemployment flows rise
in recession and fall in the boom, whereas flows into employment are strongly procyclical and separations mildly pro-cyclical or neutral. Of course, because the stock
of unemployment rises in recession as well, the average rate at which workers leave
1181
CA
+
USA
2.0
H+
DE+
1.5
Austria
NO
+
AU+ +NZ
1.0
UK +
PO +
SW +
~D
GE +
SP+ IT+
0.5
IR+~
GR +
SWI
BE+ +
+JA
NE
0.0
I-
10
20
30
40
Outflow rate (outflows/unemployment)
50
unemployment goes down, even though the gross number of exits goes up. The finding
about employment flows is explained by the fact that in the boom job creation is up and
voluntary job-to-job quits are also up, leading to more inflows; whereas in recession
quits are sufficiently down but job destruction up giving rise to conflicting influences
on separations.
Still, substantial systematic cross-country differences between unemployment inflow
and outflow rates do exist, reflecting underlying differences in unemployment incidence
and duration between Europe and the USA. In Figure 4, borrowed from Martin (1994),
inflow-outflow rate combinations in 1992 are plotted for the OECD countries. These
plots show that although the average length of unemployment spells (the inverse of
the outflow rate) is much longer in the typical EU country than in the USA, the
probability of job loss (to the extent reflected by the inflow rate) is much smaller.
Hence, long spells of unemployment rather than more frequent spells is the reason for
higher unemployment in the EU relative to the USA.
The contrasting experience of unemployment in the USA and Europe is reflected
in contrasting experience in wage growth. The fall in US real earnings at the bottom
end of the wage distribution, in contrast to growth in Europe, has been documented
by many writers and by the OECD in its official publications [see, e.g. OECD (1994),
Chapter 5]. We show in our Figure 5 a feature of wage and unemployment behavior
1182
100
/-x
80
~ F R
NO
60
+
IT
40
UK
~ .
AU +
.=~
JA
20
\--~ CA
SW+
\
USA
-20
-20
-10
10
20
30
that should be explainable within the search and matching framework, though to our
knowledge there are as yet no models that claim to explain it fully. We make an attempt
to explain it in Section 5.1 [see also Mortensen and Pissarides [1999)]. Thus, for twelve
OECD countries with comparable data on wage inequality, there appears to be a close
correlation between the percentage change in wage inequality during the 1980s and
the percentage rise in unemployment. Wage inequality is measured by the ratio of the
earnings of the most educated group in the population to the least educated [usually,
university graduates versus early school leavers; see OECD (1994), p. 160-1). Other
measures of inequality, however, give similar results [e.g. OECD (1994), p. 3; the
results in Galbraith (1996), are also consistent with our claim, despite his claim to the
contrary, if one measures the change in inequality by the change in the Gini coefficient
of the wage distribution].
Figure 5 shows that the USA, Canada and Sweden experienced the biggest rises
in inequality and the smaller rises in unemployment (fall in the USA). Japan and
Australia come next, with moderate rises in both, and the European countries follow,
with small rises or falls in inequality but big rises in unemployment. The only country
that does not conform to this rule is the United Kingdom, which experienced NorthAmerican style increase in inequality and European-style increase in unemployment
over the sample of the chart. Recently, however, unemployment in the UK has fallen
substantially, giving support to the view that the reforms of the 1980s moved the United
1183
Kingdom closer to a US style economy but had their impact first on inequality and
only more recently on unemployment.
t Of course, that at least some unemploymentis due to "frictional" factors has alwaysbeen recognized.
Lilien (1982) was among the first to claim that even "cyclical"unemploymentwas of this kind. Although
his results have been criticized, e.g. by Abraham and Katz (1986) and Blanchard and Diamond (1989),
the modern approach to unemploymentgroups all kinds of unemploymentinto one, as we do here.
1184
that this function is increasing in both arguments but exhibits decreasing marginal
products to each input. Constant returns, in the sense that
m(v,u)=m
(u)
1,-~ v = q(O) v
where
0-
O
U
(2.1)
1185
of the model and the assumption that future match product evolves according to a
Markov process with persistence, all matches are equally productive initially, until a
shock arrives 4.
Under these assumptions, an existing match starts life with x = 1 but is eventually
destroyed when a new value of x arrives below some reservation threshold, another
endogenous variable denoted as R. Unemployment incidence 3.F(R), the average
rate of transition from employment to unemployment, increases with the reservation
threshold.
As all workers are assumed to participate, the unemployed fraction evolves over time
in response to the difference between the flow of workers who transit from employment
to unemployment and the flow that transits in the opposite direction, i.e.,
/t = AF(R)(1 - u) - Oq(O) u,
(2.2)
where 1 - u represents both employment and the employment rate. The steady-state
equilibrium unemployment rate is
u=
)tF(R)
)~F(R) + Oq(O)"
(2.3)
4 Generalizing the model to realistically allow for productivity heterogeneity across vacancies and for
the fact that a random sample of new job-worker matches initially improve in average productivity are
still problems at the research frontier.
1186
JR 1[J(z) -
(2.4)
where r represents the risk free interest rate, V is the value of a vacancy, a n d p T denotes
a firing cost borne by the employer, represented as forgone output. We multiply the
termination cost by p to show that it is generally more expensive to fire a more skilled
worker than a less skilled one. The termination cost is assumed to be a pure tax and
not a transfer payment to the worker and to be policy-determined. For example, it may
represent the administrative cost o f applying for permission to fire, as is the case in
many European countries. O f course, T ~> 0 and none o f the fundamental results are
due to a strictly positive T.
Condition (2.4), that the return on the capital value of an existing job-worker
match to the employer is equal to current profit plus the expected capital gain or
loss associated with the possible arrival of a productivity shock, is a continuous-time
5 Note that contracts of this form are instantly "renegotiated" on the arrival of a new idiosyncratic
shock. MacLeod and Malcomson (1993) persuasively argue that the initial wage need not be adjusted
until an event occurs that would otherwise yield an inefficient separation. Contracts of this form may
well generate more realistic wage dynamics but job creation and job destruction decisions are the same
under theirs and our specification. Hence, for the purpose at hand, there is no relevant difference.
1187
Bellman equation. A n analogous relationship implicitly defines the asset value o f the
same match to the worker involved, W ( x ) . Namely,
r W ( x ) = w ( x ) + ;~
fR [W(z) -
(2.5)
(2.6)
(2.7)
where p c is the recruiting cost flow per vacancy held, and p C is a fixed cost of hiring
and training a new worker plus any other match-specific investment required. Here
these costs are indexed by the aggregate productivity parameter to reflect the fact that
the forgone output that these costs represent is larger when labor is more productive.
The value of unemployment solves
r U = b + Oq(O)[Wo - U],
(2.8)
6 See Mortensen (1978) for an early analysis of this issue within the search equilibrium framework. For
alternative approaches to the modeling of the job destruction flow, see Bertola and Caballero (1994), who
model a firm with many employees moving between a high-employment and a low-employment state,
and Caballero and Hammour (1994), who analyze the implications of sunk costs and appropriation.
1188
Given an initial wage equal to wo, the by now familiar asset pricing relations imply
that the initial value o f a match to employer and worker respectively satisfy
//
//
rJo = p - Wo + )~
and
rWo = w0+,~
[ W ( z ) - W0] d F ( z ) + ) ~ F ( R ) [ U - W0],
(2.9)
(2.10)
:~ ~ + C qtv)
Jo
(2.11)
As the expected number o f periods required to fill a vacancy is 1/q(O), the condition
equates the cost o f recruiting and hiring a worker to the anticipated discounted
future profit stream. The fact that vacancy duration is increasing in market tightness
guarantees that free entry will act to equate the two.
2.2. Generalized Nash bargaining
The generalized axiomatic Nash bilateral bargaining outcome with "threat point"
equal to the option o f looking for an alternative match partner is the baseline wage
specification assumption found in the literature on search equilibrium 7. Given that
the existence o f market friction creates quasi-rents for any matched pair, bilateral
bargaining after worker and employer meet is the natural starting point for an
analysis 8.
7 See Diamond (1982b), Mortensen (1978, 1982a,b), and Pissarides (1985, 1990).
s Binmore, Rubinstein and Wotinsky (1986), Rubinstein and Wolinsky (1985) and Wolinsky (1987)
applied Rubinstein's strategic model in the search equilibrium framework. The analyses in these papers
imply the following: If the worker searches and the employer recruits at the same intensities and if
/3 is interpreted as the probability that the worker makes the wage demand (1 -/3 is the probability that
the employer makes an offer) in each round of the bargaining, then the unique Markov perfect solution
to the strategic wage bargaining is the assumed generalized Nash solution. If neither searches but there
is a positive probability of an exogenous job destruction shock during negotiations, the solution is again
the one assumed but with/3 = . However, if neither seeks an alternative partner while bargaining
and there is zero probability of job destruction, the strategic solution divides the joint product of the
match Jo - p C + Wo subject to the constraint that both receive at least the option value of searching
and recruiting, U and V, rather than the net surplus, as we assumed. As these bargaining outcomes
generate the same job creation and job destruction decisions, we consider only the former case with a
/3 between 0 and 1.
1189
Given the notation introduced above, the starting wage determined by the generalized Nash bargain over the future joint income stream foreseen by worker and employer
supports the outcome
wo = arg max {[Wo - U]/3 [So -(Wo - U)] l-is }
subject to the following definition of initial match surplus,
So =- J o - p C -
(2.12)
V + Wo- U
In the language of axiomatic bargaining theory the parameter/3 represents the worker's
relative "bargaining power." Analogously, the continuing wage contract supports the
outcome
w(x) = arg max { [ W ( x ) - U] t~ [S(x) - (W(x) - U)] l-Is},
(2.13)
V +pT.
The difference between the initial wage bargain and subsequent renegotiation arises
for two reasons. First, hiring costs are "sunk" in the latter case but "on-the-table" in
the former. Second, termination costs are not incurred if no match is formed initially
but must be paid if an existing match is destroyed.
The solution to these two different optimization problems satisfy the following firstorder conditions
/3(Jo - V - p C )
= (1 - / 3 ) (w0 - u ) ~
w0 - u =/3SO
(2.14)
and
/3(J(x) - V + p T ) = (1 - / 3 ) ( W ( x ) - U) ~ W ( x ) -
U =/3S(x).
(2.15)
As a preliminary step in solving for the match surplus function and the continuing wage contract that supports the bargaining solution, first rewrite Equations
(2.4) and (2.5) as follows:
(r + )0 (J(x) - V + p T ) = p x - w(x) - r ( V - p T )
1
+ 2, ~ [J(z) - V + p T ] dE(z)
(2.16)
1190
and
i" 1
(2. 17)
By summing these equations, one obtains the following functional equation which the
surplus function must solve
S(x)
p x - r ( U + V - p T ) + 3. f l S(z) dF(z)
(2.18)
r+J,
(2.19)
(z-R)dF(z)=r(U+V-pT).
The reservation product, pR, plus the option value of continuing the match attributable
to the possibility that match product will increase in the future, the left-hand side,
equals the flow value of continuation to the pair, the right-hand side of the equation.
As the left- and right-hand sides of Equation (2.16) multiplied by 1 -/3 respectively
equal the left- and right-hand sides of Equation (2.17) when multiplied by /3
given (2.15), the continuing match product specific wage that supports the bargaining
outcome is
(2.20)
Note that this result is the generalized Nash outcome in a continuous bargain over
match output px given a "threat point" equal to the flow values of continuing the
match, namely (r(V - p T ) , rU).
Analogously, by summing equations (2.9) and (2.10), one obtains
(r + )OSo = (r + X ) ( J o - V - p C +
= p-r(U
Wo - U)
+ V)-(r + X)pC-lpT
+ ,~
S(z)dF(z)
(2.21)
(2.22)
1191
The logic of the derivation of the initial wage is similar to that used to obtain the
continuing wage function. First, rewrite Equations (2.9) and (2.10) as
(r + ~) (Jo - V - p C )
= p - wo - r V - (r + ) t ) p C - )tpT
+ 3,
[J(z) - V + p T ] dE(z)
and
( r + ) O ( W o - U) = wo - r U + ) ~
JR1[W(z)-
U] dF(z).
Second, multiply both sides of the first equation by 1 -/3, both sides of the second by
/3, and then apply Equations (2.14) and (2.15) to obtain
wo = r U +/3 [p - r ( V + U) - (r + ) O p C - )~pT].
(2.23)
Note that the initial wage equals the worker's share of the initial match flow surplus
p - r ( V + U + p C ) less the sum of hiring and firing costs amortized over the initial
period prior to the arrival of a subsequent match specific shock ){p(C + T). In short,
the worker share of both the quasi-rents and match specific investments required to
both create and end the match is the market power parameter/3.
To complete the derivation of the equilibrium conditions, we use the fact that the free
entry condition (2.22), the surplus sharing rule (2.14), and the value of unemployment
equation (2.8) imply that the flow value of unemployment is linear and increasing in
market tightness.
rU=b+/3Oq(O)So=b+(P--~/3)O.
By direct substitution into Equations (2.23) and (2.20), the equilibrium wage contract
can be written as
w0 =/3p [1 + cO - (r + )0 C - )~T] + (1 -/3) b
(2.24)
(2.25)
and
+ ~
(x-R)dF(x)
= r(g-pT)
(2.26)
=b-rpT+(lfi~fi)pcO.
1192
R
R*
>
C
D
O*
(R*, 0").
An equilibrium solution is any pair (R*, 0") that solves the job creation condition
(2.22) and the job destruction condition (2.26). The associated starting wage w0,
continuing wage function w(x), and steady-state unemployment rate u are those
specified in Equations (2.24), (2.25), and (2.3). Because the relation defined by the job
creation condition (2.22) is downward sloping, as illustrated by the line CC in Figure 6,
while the job destruction condition (2.26) can be represented as the upward sloping
line DD, there is a single equilibrium solution to the two equations 9. The equilibrium
pair is strictly positive if the product of a new match, p, less the opportunity cost of
employment, b, is sufficient to cover recruiting, hiring, and anticipated firing costs.
2.3. Fundamental determinants o f unemployment
Figure 6 provides insight into how the various parameters of the model affect the
steady-state unemployment rate. For this purpose, it is useful to remember that the
job creation line CC reflects the standard dynamic demand requirement that the cost
of hiring and training a worker is equal to the expected present value of the future profit
attributable to that worker over the life of the job. It is downward sloping because a
higher reservation threshold implies a shorter expected life for any new match. The
upward slope of the job destruction line DD reflects the sensitivity of the reservation
product threshold to general wage pressure as reflected by market tightness.
Now it is clear from Equation (2.22) that given R neither p nor b influence
equilibrium 0. Thus, general productivity and the supply price of labor do not shift CC.
By dividing Equation (2.26) byp, we find that b andp enter the equilibrium conditions
as a ratio b/p. Hence, the influence of general productivity and the opportunity cost of
employment is due entirely to the fact that the latter is independent of the former. If
for whatever reason the opportunity cost of employment b was proportional to general
productivity [as in the long-run equilibrium model of Phelps (1994), through wealth
9 Note in passing that the equilibrium pair is stationary even out of steady state because there is no
feedback from currentemploymentto expectationsabout future match output. This fact is an implication
of the linear specificationof both agent preferences and productiontechnologyand of the absence of
memory in the idiosyncraticshock process. A change in any one of these specificationassumptions
substantially complicatesbut enriches the model.
1193
accumulation], general productivity changes would not influence the equilibrium rate
of unemployment.
Given our specification and the interpretation of the two lines in Figure 6, an
increase in the supply price of labor, b, or a fall in general productivity p, shifts
the D D line up but has no direct effect on CC. As a consequence, the equilibrium
value of the reservation threshold increases and the equilibrium value of market
tightness falls with b/p. Hence, steady-state unemployment increases because both
unemployment duration and incidence increase in response.
The other parameters of the model have more complicated effects on equilibrium
unemployment and at the analytical level we can only derive unambiguous results
for unemployment duration and incidence, but not for the stock of unemployment.
Inspection of equations (2.22) and (2.26) shows that the only other parameter that
shifts only one of the lines is the job creation cost C. An increase in C shifts C C to
the left and so implies lower R and 0: unemployment duration rises but incidence
falls. The intuition behind the result is that higher job creation costs reduce job
creation, increasing the duration of unemployment, but also reduce job destruction,
to economize on the job creation costs that are incurred if the firm is to re-enter the
market. The effect on unemployment is ambiguous.
A similar ambiguity arises from changes in job termination costs. Higher T shifts
the CC line to the left and the D D line to the right. Although the effect on 0 appears
ambiguous, a formal differentiation of the equilibrium conditions yields a negative
net effect on both R and 0. Once again, job destruction falls, because it is now more
expensive to fire workers, implying less unemployment incidence. Job creation falls
because over its lifetime the job will pay the termination cost with probability 1,
implying a longer duration of unemployment.
Other parameters of the model have even more complicated effects on unemployment duration and incidence. The rate of discount, r, and the rate of arrival of
shocks, )~, both shift the job creation line down, because, in the case of r, future
product is discounted more heavily and in the case of ;~, the expected life of the job
falls. But the job destruction line also shifts. Differentiation of the two equilibrium
conditions shows that both r and X reduce market tightness, and so increase the
duration o f unemployment. The arrival rate of idiosyncratic shocks also reduces the
reservation threshold, reducing the incidence of unemployment but the rate of discount
has ambiguous effects on the threshold.
Finally, an increase in the worker's share of match surplus as reflected in an increase
in the "market power" parameter/3 shift CC downward but D D upward in Figure 6.
The result is a negative effect on equilibrium market tightness but the sign of the
resultant change in the reservation product is indeterminate. Differentiation of the
equilibrium conditions shows that the effect of/3 on R has the sign of/3 - t/, where
t/is the elasticity of the matching function with respect to unemployment. Interestingly,
if/3 = t/ the search externalities are internalized by the wage bargain, and it is a
useful benchmark case in simulations with search equilibrium models [Hosios (1990),
Pissarides (1990)].
1194
3. Employment fluctuations
The negative co-movement between aggregate measures of vacancies and unemployment, known as the Beveridge curve, has long been an empirical regularity of interest in
the literature on labor market dynamics 10. Generally, high vacancies and low levels of
unemployment characterize a "tight" labor market in which workers find jobs quickly
and higher wage rates prevail. Time-series observations suggest that job vacancy
movements lead unemployment changes both in the sense that drops ~n job vacancy
rates herald downturns in employment and that employment recoveries follow jumps
in vacancies. These observations also suggest that fluctuations in derived demand for
labor, as reflected in vacancy movements, rather than labor supply shocks are the
principal driving force behind cyclical unemployment dynamics.
The empirical work of Davis and Haltiwanger (1990, 1992) and Davis, Haltiwanger
and Schuh (1996) has stimulated general interest in the components & n e t employment
change, which they call job creation and job destruction flows. As we saw in Section 1,
the job creation and job destruction rates move in opposite directions over the business
cycle but are always both large and positive at every level of industry and regional
disaggregation. These facts suggest that employment reallocation across economic
activities is a significant and continual process that accounts for a large measure of
unemployment.
Mortensen and Pissarides (1994), Mortensen (1994b), Cole and Rogerson (1996),
and den Haan, Ramey and Watson (1997) claim that an extended version of the
equilibriurn unemployment model, one that allows for an aggregate shock to labor
productivity, can explain the stylized facts of the job creation and job destruction
flows that we listed in Section 1. To recall, apart from the negative correlation between
them just noted, job destruction is more volatile than job creation (which, at least for
US manufacturing, shows up as negative correlation between the sum and difference
of the job creation and job destruction flows) and quit rates are procyclical, i.e. there
is a positive correlation between quit rates and the difference between job creation and
job destruction. The purpose of this section is to present a version of the model that
allows for employment fluctuations which can be used to illustrate these claims.
3.1. Stochastic equilibrium
l0 For an interesting early treatment, see Hansen (1970). For more recent search-based analyses, see
Pissarides (1986) and Blanchard and Diamond (1989).
1195
productivity p induce the cycle. On the argument that recruiting and hiring costs
represent forgone output, we continue assuming as well that these costs are indexed
by the productivity parameter p.
According to real business cycle theory, economic fluctuations are induced by
exogenous persistent shocks to aggregate labor productivity 11. Whether exogenous
technical change is the cause or not, labor productivity is procyclical in fact and our
model's implications for wage and employment responses are an implication of that
fact whatever its cause. Given sufficient persistence, one would expect these shocks
to induce cyclical effects on the market tightness and the reservation idiosyncratic
product which are similar to those associated with a permanent change in the level of
aggregate productivity.
For the sake of a simple presentation, assume that aggregate productivity fluctuates
between a high value Ph and a low value Pl, where the continuous time transition rate
or frequency is t/. For this specification, the autocorrelation coefficient of the p-process
given a short time interval of length A is 2e -r/a - 1. Indeed,
rSi(x) = p i x - r ( g ; + g i - p i T ) + ,1
['
d Ri
rSo; = p ; x o - r(U~ + Vi - p i T ) - (r + ,l + ~ ) p ; ( C + T )
+,1
(3.1)
11 For example, see Kydland and Prescott (1982) and Lucas (1987).
1196
where the aggregate state contingent values o f a vacancy and unemployment solve
rVi = q ( O i ) ( l -[~)Soi q- ~ (Vj - Vi) - p i c ,
rUi = b + Oiq ( Oi) flSoi + 71 ( Uj - Ui) .
(3.2)
and
S~(R~)=0,
i E {I,h}.
Market tightness is procyclical and market tightness and the reservation product
threshold move in opposite directions in response to aggregate shocks if the shock
is sufficiently persistent. Formally, a unique equilibrium exists with the property that
Ph > Pl
Rh < Rt f o r all ~ while a critical value cc > ~ > 0 exists such that
As just demonstrated, "boom" and "bust" in this simple model are synonymous with
the prevalence of the "high" and "low" average labor productivity when the aggregate
shock is persistent. Unemployment dynamics in each aggregate state are determined
by the law of motion
/t = )~F(R,)(1 - u) - Oiq(Oi) u.
(3.3)
Hence, the unemployment rate tends toward the lower of the two aggregate state
contingent values, represented by
.
)tF(Ri)
ui = )~F(Ri) + Oiq(Oi)'
i E {I, h},
(3.4)
uh Uz
1197
Oh
Oh
Ot
v;
- -
.2
.7
The observation that actual vacancies and unemployment time series are negatively
correlated is consistent with this model under appropriate conditions, a fact illustrated
in Figure 7. In the figure, the two rays from the origin, labeled 0l and Oh, represent the
vacancy-unemployment ratios in the two aggregate states when Oh > 0l. The negatively
sloped curves represent the locus of points along which there is no change over time in
the unemployment rate, one for each of the two states. Because the curve for aggregate
state i is defined by
vq(v/ui)
- - tlF(Ri),
1 - ui
Rh < Rl implies that uh < uz for every v as drawn in Figure 7. Finally, the two steadystate vacancy-unemployment pairs lie at the respective intersections of the appropriate
curves, labeled L and H in the figure. Provided that the curve along which/l = 0 doesn't
shift in too much when aggregate productivity increases, v~ > v~ as well as u~ < u~.
However, sufficient persistence, in the form of a low transition frequency, is necessary
here. Indeed, the points L and H lie on a common ray when persistence is at the critical
value tl = ~ since 0l = Oh by definition.
3.3. Job creation and j o b destruction flows
In our simple model, the notion of a job is equivalent to that of an establishment, plant,
or firm given the linear technology assumption. Consequently, the job creation flow,
the employment changes summed across all new and expanding plants over a given
period of observation, can be associated with the flow of new matches in the model.
Analogously, job destruction, the absolute sum o f employment reductions across
contracting and dying establishments, is equal to all matches that either experience
an idiosyncratic shock that falls below the reservation threshold or were above the
1198
threshold last period but are below it this period. The fact that market tightness
and the reservation product move in opposite directions in response to an aggregate
productivity shock implies negative co-movements in the two series, as observed.
Furthermore, a negative productivity shock induces immediate job destruction while a
positive shock results in new job creation only with a lag. This property of the model is
consistent with the fact that job destruction "spikes" are observed in the job destruction
series for US manufacturing which are not matched by job creation "spurts" 13. As in
the OECD data, cyclical job destruction at the onset of recession is completed faster
than cyclical job creation at the onset of a boom.
1)i
ui +si(1 - ui)"
Once employed, workers have an incentive to move from lower to higher paying jobs.
Suppose that employed workers can search only at an extra cost, ~, interpreted as
foregone leisure, a reduction in b. As search is jointly optimal for the pair if and only
if the expected return, equal to the product of the job-finding rate and the gain in
13 These points are discussed in more detail in Mortensen and Pissarides (1994) and Mortensen
(1994b).
1199
match surplus realized, exceeds the cost, all workers employed at x equal to or less
than some critical value, denoted as Qi, will search where 14
Oiq(Oi) [S/(1) - Si (Qi)] = a,
i c {l, h}.
(3.5)
si = F ( Q i ) - F(Ri).
Because a quit represents an employment transition for the worker and the loss of a
filled job for the employer, the surplus value equation under joint wealth maximization
is
rSi(x) = p i x - o - r(Ui + Vi - T) + ~
+ t/[Sj(x) - S i ( x ) ] + Oiq(Oi)(Si(1)-Si(x))
(3.7)
V x < Qi.
Because the worker does not search when x >7 Qi and this condition always holds when
x = 1, Equations (3.1) continue to hold in this range. To the extent that market tightness
is procyclical, Equation (3.5) implies Qh > Ql. Hence, the quit flow is procyclical for
two separate reasons. First, because Q is higher and R is lower in the high aggregate
productivity state, the fraction of employed workers who search is procyclical, i.e.,
sh > sl. Second, because Oh > 01 when the aggregate shock is sufficiently persistent,
the rate at which searching workers meet vacancies Oq(O) is also larger in the high
aggregate product state.
Worker reallocation across different activities is represented by both the direct
movement from one job to another via quits and by movements through unemployment
induced by job destruction and subsequent new job creation. Davis, Haltiwanger
and Schuh (1996) estimate that between 30% and 50% o f worker reallocation is
attributable to the job destruction and creation process. Given the procyclicality of
the quit flow and the flow of hires, the sum o f job creation and quits is highly
procyclical, while the separation flow, the sum of job destruction and quits, is acyclical.
Hence, the reallocation o f workers across activities is procyclical relative to the more
countercyclical reallocation of jobs across activities both in fact and according to the
model.
The quit process also interacts with job creation and job destruction in more
complicated ways that are not explicitly modeled here. For example, when a worker
14 Although the decision to maximize the sum of the pair's expected future discounted income by the
appropriate choice of the worker's search effort is individually rational under an appropriate contract,
both costless monitoring and enforcement of the contract is generally necessary to overcome problems
of dynamic inconsistency. Indeed, otherwise the worker will search if and only if the personal gain
exceeds cost, i.e., iff W/(1)- W/(x) =/~[Si(1)- Si(x)] > o" which would imply too few quits.
1200
quits an existing job to take a new one, the employer can c h o o s e to search for a
replacement. If the decision is not to replace the worker, the quit has induced the
destruction of a job with no net change in either the number o f jobs or unemployment.
I f the decision is to declare the job vacant, a new job was created by the original
match but there will be no net reduction.in unemployment unless the old job vacated
is filled by an unemployed worker. O f course, if filled by an employed worker, the
employer left by that worker must decide whether o r n o t to seek a replacement. This
sequential replacement process by which a new vacancy leads to an ever/tual hire from
the unemployment pool, known in the literature as a vacancy chain, propagates the
effects o f job creation shocks on unemployment [see Contini and Revelli (1997) and
Akerlof, Rose and Yellen (1998)].
Also, quit rates are high in the first several months after the formation of new
matches and then decline significantly with match tenure, presumably as a consequence
o f learning about the initially unknown "quality" o f the fit between worker and
job 15. This source o f quits is o f significant magnitude and it represents the primary
form o f quits to unemployment. Because this "job shopping" process implies that an
unemployed worker typically tries out a sequence o f jobs before finding satisfaction,
a job destruction shock is likely to be followed by a drawn-out period o f higher
than normal flow into and out o f unemployment 16. Were the job shopping process
incorporated in the model, job reallocation shock effects on worker flows would be
prolonged and amplified, features that should also improve the model's fit to the data.
15 There is an extensive labor economics literature on this point initiated by the seminal theoretical
development by Jovanovic (1979). See Farber (1994) for a recent analysis of the micro-data evidence on
tenure effects on quit rates and the extent to which these are explained by the job shopping hypothesis.
Pissarides (1994) explains these facts within a search model with learning on the job.
J6 Hall (1995) argues that this effect is apparent in the lag relatioships between the time series
aggregates.
17 For attempts to estimate structural forms of the matching model see Pissarides (1986) and Yashiv
(1997).
18 When used to calibrate the business cycle facts the models are often offered as alternatives and
compared with Hansen's (1985) indivisible labor model.
Ch. 18:
1201
Cole and Rogerson (1996) conduct an analysis of the extent to which the rudimentary
Mortensen-Pissarides model can explain characteristics of the time series observations
on employment and job flows in US manufacturing. For this purpose, they construct the
following stylized approximation to the continuous time formulation sketched above:
Job creation in period t, ct, is equal to the matches that form during the observation
period and survive to its end. As one can ignore the possibility that a job is both
created and destroyed when the observation period is sufficiently short, approximate
job creation in period t is
ct = as, ~(1 - nt 1),
1-
(4.1)
a s t = e -O~tq(O~),
Ot =
where
6i = 1 - e-zF(Ri),
6o =
~ (F(Rl) -
(4.2)
F
(Rh))
where
2"glh = 3"ghl = 375 =
1 - e -rl
~t+160) nt
1202
given the Markov forcing process {st} defined on the state space {l,h} and
characterized by the symmetric probability of transition Jr.
Obviously, the employment, job creation, and job destruction processes are interrelated and fully characterized by the set of reduced form parameters {al, ah, 6t, 6h, 60, Jr}.
The question asked by Cole and Rogerson (1996) is whether an appropriate choice of
these parameters will replicate the salient features of the Davis-Haltiwanger-Schuh
observations, which they summarize in the following useful way:
(1) Volatility: Job creation is roughly four times as volatile as employment, and job
destruction is more than six times as volatile.
(2) Persistence: The series for job creation, job destruction and employment display
strong positive autocorrelation, but the autocorrelation for employment, which is
0.9, is nearly twice that for the other two series.
(3) Contemporaneous Correlations: Creation and destruction have a fairly large negative correlation. Destruction is (weakly) negatively correlated with employment,
whereas creation is virtually uncorrelated with employment.
(4) Dynamics: Creation is negatively correlated with lagged employment, and positively correlated with fuWxe employment. The opposite pattern holds for destruction.
To answer their question, Cole and Rogerson simulate the model above for trial
parameter values, compute the associated simulation statistics, and then adjust the
parameter values to obtain a better match. They conclude that the model can replicate
observations in their sense when the probability of finding a job is not too large.
Specifically, the model simulation for the parameter set
{al, ah, 6l, 6h, 6o, zc} = {0.21, 0.30, 0.069, 0.044, 0.01,0.2}
generates their preferred results which are not only consistent with their qualitative
characterization of the data but are quite close in quantitative terms as well. Given that
the two job destruction rates 6l and 6h are set to match the average of 0.055 reported in
the Davis-Haltiwanger-Schuh data, one potential problem which Cole and Rogerson
emphasize and discuss are the low values of the probabilities of finding employment.
To see the significance of the point, simply note that the two state contingent steadystate unemployment rates associated with this parameter set are
ul
6l+al-0"25'
uh
6h + ah
0 . 1 3 ,
two numbers that yield an average unemployment rate of 19%. Nonetheless, the authors
argue that these numbers are reasonable given the following observations reported
by Blanchard and Diamond (1990): First, although the simple model ignores nonparticipants, in fact the flow to employment from this stock is roughly equal to the
flow from those officially categorized as unemployed. Second, the number of workers
classified as out-of-the-labor-force who report they want jobs is also roughly equal to
1203
the number classified as unemployed. Including these individuals in the pool of the
unemployed would rationalize the low average value of a, especially if these workers
search at lower intensities.
4.2. Capital accumulation and shock propagation
Merz (1995) and Andolfatto (1996) each construct different but related syntheses
of the neoclassical stochastic growth model and the Pissarides (1990) model of
frictional unemployment. The contributions of these authors include a demonstration
that the "technology shocks" responsible for business cycles in the real business
cycle (RBC) model will also induce negative correlation between vacancies and
unemployment, the Beveridge curve, and a positive correlation between flows into
and out of unemployment in a version of the model with a labor market characterized
by a matching process. However, like the earlier simpler RBC models, the amended
models fail to propagate productivity shocks in the manner suggested by the observed
persistence in actual output growth rates.
Recently, den Haan, Ramey and Watson (1997) have constructed, calibrated,
and simulated a synthesis of the Mortensen and Pissarides (1994) model of job
creation and job destruction with the neoclassical stochastic capital accumulation
model. As in the Merz and Andolfatto models, job creation is governed by a
matching function whose inputs include vacancies and unemployed workers. In
addition, a job destruction margin is introduced by supposing that existing job
matches experience idiosyncratic productivity shocks orthogonal to the aggregate
shock to match productivity as described above. They find that interaction between
the household saving decision and the job destruction decision play a key role in
propagating aggregate productivity shocks. As a consequence, their synthesis provides
an explanation for the observed autocorrelation in output growth rates as well as the
correlation patterns observed in job flows with themselves and employment, those
matched by Cole and Rogerson (1996).
Den Haan et al. (1997) explicitly formulated the model in discrete time with
each period equal to one quarter. Following Merz (1995) and Andolfatto (1996),
idiosyncratic variation in labor income attributable to unemployment spells is fully
insured through income pooling. Hence, the existence of a representative household
can be invoked; one assumed to have additively separable preferences over future
consumption streams represented by ~ t ytu(Ct) where t is the time period index,
y is the time discount factor, and u(C) is one period utility expressed as a concave
function of consumption. A single consumable and durable asset, capital, exists which
also serves as a productive input. The sequence of future market returns for holding
the asset, denoted {rt}, is an endogenous stochastic process. Hence, the optimal
consumption plan must satisfy the usual Euler equation
u'(Ct) = gEt{u'(Ct+l)(1 - 6 + rt+l)},
(4.3)
where the expectation is taken with respect to information available in period t and
6 is the rate of physical capital depreciation.
1204
The surplus value of a new match is another endogenous stochastic process, denoted
{S+1}. When an unemployed worker and job vacancy meet at the beginning of
period t + 1, Nash bargaining takes place. The outcome allocates the share fiStl
to the worker and the remainder (1 -fl)Sl to the employer, where as above fi
represents worker market power. The anticipated bargaining outcome motivates search
and recruiting effort by unemployed workers and employers with vacancies during
period t. The flow return to unemployed search is the sum of home production while
unemployed, b, and the expected gain attributable to finding a match: '
b+Otq(Ot)[3Et{
u ( t+l
~
)S?+1) .
(4.4)
The expected capital gain, the second term, is the product of the probability of finding
a job and the expected value of the worker's share of match surplus given information
available in period t appropriately discounted back to the present by a factor which
accounts for any difference in the marginal utility of consumption in the next and
the current period. Similarly, free entry of vacancies requires zero profit in the sense
that recruiting cost per vacancy posted, ptc, equals expected return, the product of the
probability that the employer finds a match and the employer's share of its expected
discounted surplus value:
(4.5)
The aggregate productivity shock, the process {pt}, is Markov with the transition
probability kemel assumed to be common knowledge. For simplicity, den Haan et
al. (1997) assume that the match-specific process, represented by {xt}, is i.i.d, with
c.d.f. F(x) 19. Still, the idiosyncratic shock is expected to persist for the duration of the
current period. The output of an existing match in period t is ptxtf(kt) where kt is the
amount of capital per worker rented during the period at rate rt, andf(k), normalized
output per worker, is an increasing concave function. Because the option value of
continuing the match is zero for the employer and equal to the flow value of search for
the worker, b + [3ptcO/(1 -[3) from Equations (4.4) and (4.5), the joint match surplus
conditional on idiosyncratic productivity xt is
{ yut(Ct+l)
(4.6)
}
,
where the last term reflects appropriate discounting of next-period surplus and the
option to destroy the match next period if need be.
19 Otherwise,the distributionof idiosyncraticproductivityacross existingmatchesis a decisionrelevant
state variable.Theyclaimthatthe modelloses no essentialpropertyas a consequenceof this abstraction.
1205
By implication of the optimal capital choice, the current period demand for rented
capital by an existing match characterized by idiosyncratic productivity xt = x is
k[(x)= { d ~p,
0
if
(4.7)
otherwise,
Kt=[fRid(~pt) dF(x)lNt,
(4.8)
where (Kt,Nt) is the given aggregate capital stock and employment pair as of the
beginning of period t.
As the current reservation value Rt solves St(R) = 0, Equation (4.6) implies
t~c (4.9)
max{ptRtf(k)-rdQ+lZttr ]lbll(Ct+l)
~
max {St(xt+l), 0}} = b+ ~_fiw./3pt
Given that xt ~ F(x), it follows that expected ex ante match surplus conditional on
knowledge of (Pt,Rt) is
--',
--
- rtk}
/3ptc tl
=b+~_flvt
- Et
ut(Ct)
t+l
(4.11)
- rt+lk})dg(x) [ .
J
Finally, because x = 1 for a new match, So = St(l). Hence, Equations (4.6) and (4.10)
imply that Equation (4.5) can be written as
1206
Note that Equations (4.11) and (4.12) are generalizations of the job destruction and job
creation conditions. Indeed, in the non-stochastic case with linear utility and no capital,
these equations are equivalent to Equations (2.26) and (2.22) since Equation (4.3)
implies 7 = 1/(1 + rt) for all t and the discrete time specification and the assumption
that the idiosyncratic shock persists for one period imply that the duration of any shock
is unity, i.e., tt = 1. However, a complete characterization of general equilibrium also
requires that the equilibrium conditions of the neoclassical stochastic growth model,
Equations (4.3) and (4.8), and the laws of motion hold.
The laws of motion for capital and employment are
dF(x) l Nt
(4.13)
- cptOtq(Ot)(1 - Art) - Ct
and
(4.14)
respectively 2. The first equation reflects the effects of job destruction and capital
demand decisions made at the beginning of the period on output and the consumption
decision while the second reflect the outcomes of current period job creation and
destruction decisions. As the information relevant state of the economy is a triple
composed of the capital stock, the employment level, and the aggregate shock, a
dynamic stationary general equilibrium is a vector function that maps the state
variable triple (N, K,p) to the four endogenous variables (C, r, R, 0); one that solves
the Euler equation (4.3), the capital market clearing condition (4.8), the job destruction
condition (4.11), and the job creation condition (4.12) under the laws of motion (4.13)
and (4.14).
Den Haan et al. (1997) derive the properties of the equilibrium by solving and
simulating a particular parameterization of the model numerically. The qualitative
properties they report are intuitively suggested by the known implications of the
two models married in this synthesis. For example, a positive aggregate shock
stimulates current investment in both job creation and physical capital which augment
employment and productive capacity in the next period. In the short run, these
investments must be financed with an output increase induced by a lower than
normal reservation productivity choice and by a reduction in consumption. However,
because of the consumption smoothing motive, the limited ability to increase output by
increasing utilization through reductions in job destruction, and the complementarity
of physical capital and labor, more investment of both types is made in subsequent
periods as well, i.e., the shock is propagated.
20 Followingthe literature, home production b cannot be used to create capital by assumption. It is
simply consumed.
1207
A negative shock has an immediate and sharp negative effect on output along the
job destruction margin. Although the effect is cushioned by the reallocation of existing
capital to those jobs that continue, rental rates fall on impact in response to the decrease
in demand for capital induced by job destruction and will be expected to fall further
in the future as a consequence of the persistence in the shock. The result is a reduction
in capital formation and job creation which has the effect of reducing output further in
the future. Again the consumption smoothing motive interacting with the job creation
and destruction process propagates the shock into the future.
As a consequence of the adjustment mechanisms described above, the simulated
model implies strong first- and second-order autocorrelation in output growth rates,
substantial persistence in the response of physical capital to negative productivity
shocks, and a substantial magnification of the effects of productivity shocks on
aggregate output. Neither the RBC model nor the augmented model featuring job
matching but exogenous job destruction, like those of Merz (1995) and Andolfatto
(1996), explain these features of the aggregate time-series data. As in Cole and
Rogerson's (1996) reduced form analysis of the Mortensen and Pissarides job creation
and destruction model, the calibrated version of the extended model studied by den
Haan et al. (1997) also reproduces all the job flow time series stylized facts.
1208
(1990), Chapter 2]. The effect o f faster growth on j o b destruction is, however, o f
indeterminate sign. We then consider the vintage model in the sense that "new capital"
is assumed to be embodied only in newly created jobs. We show that under the
assumption that the same worker cannot be m o v e d from an old j o b to a new one
without intervening unemployment, steady-state unemployment is higher at faster rates
o f technological progress [as in A g h i o n and Howitt (1994)] 21.
-g
<r.
(5.1)
P
We treat r as a constant independent o f g 22. The other restrictions made are the same as
in the basic model o f Section 2.1, with the additional assumption that unemployment
income is also a function o f time. We assume for simplicity that b(t) = bp(t). This
assumption is needed to ensure the existence o f a steady-state growth equilibrium
and is plausible in a long-run equilibrium when p(t) is an aggregate productivity
parameter 23.
The j o b creation and j o b destruction conditions o f Section 2.1 change in an obvious
way. Because all parameters in the value expressions (2.4), (2.5), (2.7) and (2.8) are
multiplied by p(t), and the wage equation still satisfies either (2.20) or (2.23), there
is an equilibrium where all value expressions grow at constant rate g. Intuitively, the
firm that has a j o b with value J(x, t) at time t, expects to make a capital gain o f
dJ(x, t)/dt ==-J(x) = gJ(x) on it. The same holds true for the value o f a j o b to the
worker, W(x, t), and the value o f unemployment, U(x, t), where the capital gain is,
respectively, gW(x) and gU(x). But the value o f a vacant job, V(t), because it is zero
21 Mortensen and Pissarides (1998) consider a more general case of adoption of the new technology at
a cost and show that the two cases that we consider here are two limiting cases, the first case approached
when the adoption cost tends to zero and the second when the adoption cost tends to infinity. The main
result of the paper is that there is a critical level of the adoption cost below which the dominant influences
on job creation and job destruction are those described here under disembodied technology and above
which the dominant influences are those described under embodied technology. See also Aghion and
Howitt (1998, chapter 4) for more analysis of this issue.
22 Eriksson (1997) embeds the model in an optimizing (Ramsey) growth model and shows that the
restriction that the effective discount rate decline with the rate of growth can be violated by feasible
parameter values. He also considers the effects of growth on unemployment in an endogenous growth
framework.
23 Making b(t) a proportional function of the equilibrium wage rate would not change the results.
1209
by the free entry condition, does not change. It is this asymmetry between V(t) on the
one hand and the other asset values on the other that creates the capitalization effect
of faster growth.
We do not reproduce all the value expressions with growth but show instead the
value of a continuing job to the firm, (2.4):
(5.2)
The capital gain to the firm is shown as an addition to revenues from continuing the
job. Replacing the capital gain by its steady-state value, we get
1
(5.3)
+ ~ F ( R ) [ V ( t ) - p ( t ) T - J ( x , t)].
The main result of the introduction of growth can be seen from Equation (5.3).
Because all value expressions grow at the constant rate g, wages will also grow at the
constant rate g, and so all time-dependent variables in Equation (5.3) can be written
as proportional functions of p(t). Letting then J ( x , t) = p(t) J ( x ) and using similar
notation for the other time-dependent variables, we can re-write Equation (5.3) in the
same form as Equation (2.4), except that the discount rate r is replaced by r - g .
It is straightforward to work through the model of Section 2 with the assumption
that all time-dependent variables are proportional functions of aggregate productivity
and show that there is a solution for the job creation and job destruction flows that
replicate the solution shown in Figure 6 but with r replaced by r - g . Hence, under
the assumption that r - g falls monotonically in g, we find that faster disembodied
technological progress increases market tightness 0 but has ambiguous effects on the
reservation productivity R. Therefore, faster growth increases job creation, decreases
the duration of unemployment but has ambiguous effects on job destruction and
the incidence of unemployment in general. However, much of the literature on the
effects of growth on unemployment concentrates on the obsolescence effects of new
technology on job destruction (see the next section) and ignores the idiosyncratic
reasons for job destruction. This assumption, also adopted in Pissarides (1990,
Chapter 2), is justified in the long-rma context by the fact that most variations in
the job destruction rate in the data are high-frequency, with, at least in the European
context where there have been substantial changes in the unemployment rate, virtually a
constant job destruction flow across business cycles. This fact justifies a 0, 1 restriction
on the support of the distribution of idiosyncratic shocks. In this case, variations in R
do not influence the job destruction rate, which is equal to 2~, and so the effect of faster
growth is to increase job creation and reduce unemployment.
1210
New technology cannot always be adopted by existing jobs. Much of public discussion
and a large body of literature deals with the situation where the adoption of new
technology requires the creation of new jobs with new capital equipment. This process
of implementation is referred to in the literature as "creative destruction", because old
jobs have to be destroyed to release the resources for the creation of new jobs [see
Aghion and Howitt (1992, 1994), Grossman and Helpman (1991), and Caballero and
Hammour (1994)]. In this section we assume that the process of creative destruction
induces a transition of the worker to unemployment and search for a new job. We
demonstrate that more rapid technological progress under these assumptions induces
more labor reallocation and so higher unemployment because of both lower job
creation rate and higher job destruction rate.
In order to emphasize the new element of the model we abstract from idiosyncratic
productivity shocks. Instead, heterogeneity in productivity arises because older jobs
embody less productive technology and a job is destroyed when the technology
embodied becomes obsolete.
Given that current technological improvements affect only productivity in newly
created jobs, we need to distinguish between the date at which a job is created, its
vintage v, and the current date, denoted as t. The expected present value of both
future profit J and wage income W for a given job-worker match depends on the
job's vintage and the current date. These value functions solve the following asset
pricing equations:
rY(v, t) = p ( v ) x - w(v, t) - 6Y (v, t) + J (v, t),
(5.4)
where x represents job match productivity, w(v, t) is the wage paid on a job of vintage v
at date t, 6 > 0 represents an exogenous job separation rate, and U(t) is the value of
unemployed search at t.
The fixed cost of investment in a new job, denoted as p(t)C, is incurred when the
match forms. The investment is specific to a job, i.e., it is "irreversible" with no outside
option value once the match forms. The recruiting costs, p(t) c, are modelled as a cost
per vacancy posted. New vacancies enter at every date until market tightness is such
that the value of creating a vacancy, V(t), is zero, i.e.
r V ( t ) = q(O)[J(t, t) - p ( t ) C ] - cp(t) = 0,
(5.5)
where q(O) is the rate at which vacancies are filled. Similarly, the value of
unemployment solves the asset pricing equation
r U ( t ) = p(t) b + Oq(O)[W(t, t) - U(t)] + U(t),
(5.6)
where p ( t ) b represents the opportunity cost of employment and where Oq(O) is the rate
at which workers find vacancies. As before, recruiting costs, the investment required to
1211
create a match, and the opportunity cost of employment grow at rate g by assumption
to assure the existence o f a balanced growth path equilibrium solution to the model.
We assume that the wage bargain divides the surplus value o f a continuing match
in fixed proportion, i.e.,
(5.7)
where/3 represents the worker's share 24. Because Equations (5.4) and (5.6) imply
(5.8)
right reflects the
value outside the
but the former is
(5.4), we obtain
(5.9)
Indeed, Equation (5.9) holds only for t - v ~< r where r is the optimal economic lifespan of a job. The employer's choice of a job's economic life maximizes its value,
i.e.,
J(v't)=max~r ~JtF+~[(1-/3)P(V)x-p(s)[(1-/3)b+/3(cO+Oq(O)C)]]
(5.10)
e-(r+6)(s-t) ds}.
The maximal value o f a new job at time t is the special solution to this equation
satisfying the balance growth equation J(t,t)
= J(O)p(t) where, given the
normalization p(0) = 1,
J(O) =- J(O, O)
= m a x { f 0 T [ ( 1 - / 3 ) x - e gs [(1 -/3) b + / 3 ( c 0 + Oq(O)C)]]e -(r+~)s ds}.
(5.11)
24 Here workers do not share the cost of initial investment by accepting a lower starting wage for an
initial period of employment as assumed in Section 2. Instead, the initial wage is equal to the continuing
wage at initial productivity. Although equilibrium market tightness will be too low relative to a social
optimum initially, the qualitative behavior of a model under a jointly efficient wage bargain would be
much the same. See Caballero and Hammour (1994, 1996) for more discussion of this issue.
1212
The first-order condition for a positive optimal choice of the economic life of a
job equates stationary match product with the rising opportunity cost of continuing an
existing match, i.e.
(1 -/3) x - [(1 -/3) b +/3(c0 + Oq(O) C)] egr = 0.
(5.12)
Since J(t, t) = J(O)p(t), the free entry condition (5.5) can be written as
c = q(O)[J(O)- C].
(5.13)
A search equilibrium is characterized by any market tightness and age at job
destruction pair (0", r*) that solves Equations (5.12) and (5.13).
Because the right-hand side of Equation (5.13) is strictly decreasing in 0, equilibrium market tightness is unique. Of course, given 0", the associated equilibrium value
of the optimal job age at destruction, r*, is the unique solution to Equation (5.12).
Since Equations (5.12), (5.13) and (5.11) imply
c + C = J(O*) = (1 -/3)x for* [1 - eg(s_r.)] e_(r+~)xds '
q(O*)
(5.14)
a necessary but hardly sufficient condition for the existence of a positive equilibrium
pair (0", *) is that match productivity x exceed the opportunity cost of employment b.
Indeed, given this condition, an economically meaningful equilibrium exists only if
both recruiting and creation costs, c and C, are sufficiently small.
Because the surplus value of a match decreases with the rate of technological
progress, g, for every value of market tightness by virtue of Equation (5.1 l) and the
envelope theorem, namely
OJ o
Og
and because both the value of a job and the rate at which vacancies are filled decrease
with market tightness, the free entry condition (5.13) implies that market tightness
falls with the growth rate, i.e.,
00" _ (
q(0*) 2
OJ
Og
~,cq'(O*)+q(~O*)2oJ--~ ~ g
<0.
Because the left-hand side of (5.14) is decreasing in g and the right-hand side is
increasing in both g and z*, it follows that the economic life of a new job also falls
with the rate of growth, i.e.,
Or* _
Og
<0.
To derive the implications of these facts for unemployment and job flows, first note
that job creation at time t is
1213
of each cohort that survive to age ~ is e -~r given the exogenous destruction hazard is
6, the job destruction flow at time t is
(5.16)
Hence, the steady-state unemployment rate that equates job creation and job destruction flows through time is
u* =
6
di + (1 - e-~*)O*q(O*)"
(5.17)
It increases with the rate o f embodied technical progress because both market tightness
and the economic life of an existing job decline with g and because the unemployment
duration hazard Oq(O) is increasing in 0.
Technological progress in this model adversely effects worker flows into and out of
employment for two reasons. The first is a restriction that we have imposed on the
model, namely, that when a machine is replaced because of obsolescence the worker
that was employed on that machine is also replaced. This assumption also underlies
the work of Aghion and Howitt (1994) and Caballero and Hammour (1996) and is
derived from Schumpeter's notion of "creative destruction". The idea is that when a
job is destroyed it is replaced by a technologically more advanced one, with positive
effects on factor productivity. The second is a particular assumption about the timing
of job creation costs.
The implication of the first restriction for the job destruction flow is straightforward
enough: faster technological progress necessitates more job destruction. Job creation
also fails in our model when there is faster technological progress because as the life of
a job becomes shorter, the expected present value o f future profit attributable to a job
falls. It may turn out to be surprising that even when the interest rate is independent of
growth faster growth does not have a countervailing effect on the present discounted
value of profits. Since in the expressions that we have derived for the surplus from a
job the effective discount rate is r - g, profits are discounted at lower rate. So faster
growth has a "capitalization" effect on the profits stream. Our results, however, show
that this capitalization effect is dominated by the negative influence on the present
value calculation implied by the shorter life of a job.
Aghion and Howitt's (1994) model of the adoption of new technology is essentially
the same as the one in this section, yet it has a bigger capitalization effect that is not
always dominated by the shorter life of the job. This effect is implied by the assumption
that there are job set-up costs that have to be paid before the firm begins the recruiting
process. In this case the profit stream is discounted more heavily, since the zero profit
restriction requires that the present discounted value of profits at the date the vacant
job is created must equal to the set-up costs.
1214
because with increasing openness and trade, and with the global oil and material shocks
of the 1970s, the shocks affecting OECD countries cannot have been very different
in different countries. The different experience of OECD countries is most likely due
to a different response of each country to common shocks, due to different market
structures, or to differences in policy.
The most frequently discussed contrast in OECD experience is that between the
USA and "Europe". Although the contrast is often exaggerated, especially in the more
popular discussions, there is some truth in the basic argument, that whereas wages at
the lower end of the wage distribution fell in the USA with unemployment remaining
the same on average, in most of Europe wages increased but unemployment increased
too. We saw in Section 1 that there appears to be a trade off between the increase in
wage inequality and the increase in unemployment experienced by OECD countries.
Figure 4 shows that over the 1980s the USA experienced a bigger increase in inequality
and a smaller increase in unemployment (in fact, a decrease) than the major European
countries.
The experience documented in Figure 4 is most likely a response to a heterogeneous
aggregate productivity shock that can be decomposed into two parts, one that shifted
the productivity distribution to the right and one that widened the range of the
distribution for given mean. There has been a long debate in the literature as to whether
the second component of the shock, the one that worsened prospects for unskilled
workers but improved them for skilled ones, was due to a technology shock, associated
for example with computerization, or to a trade shock, associated with the expansion
of trade with newly industrialized nations in South East Asia and Latin America. Our
analysis, and more generally the search and matching framework, is one that can be
used to analyze the consequences of the shocks, whatever their source.
In this section we survey the key influences that have been identified in the literature
as the causes of the experience of OECD countries summarized in Section 1. In a
discussion of this kind, it is difficult to avoid a discussion of labor market policy,
especially if one were to discuss the unemployment experiences of countries like Spain
and Sweden and why they have been different from the median European experience 25.
The detailed modelling and discussion of labor market policies, however, will take us
beyond the scope of our chapter. We mention instead policy influences in passing, using
parameters that we already have in our analysis to represent the effects of policy. Two
parameters in particular are relevant to our discussion, unemployment income b, which
we take also to represent the generosity of the unemployment insurance system, and
the firing cost T , which we take to stand for employment protection legislation. The
active labor market policies pursued by Sweden, and to a lesser degree by some other
2s There has been a large literature on the unemployment experience of each of these cotmtries. For
Spain, see for example, Blanchard et al. (1995), Dolado and Jimeno (1997) and Marimon and Zilibotti
(1998). For Sweden, see Calmfors (1995) and Ljungqvist and Sargent (1995). See also Scarpetta (1996)
for a cross-countryOECD study.
1215
countries, can be shown in the model by reductions in the job creation cost C, though
this does not do justice to the complexity and sophistication of some of the targeted
policies in operation.
6.1. 'Skill-biased' technology shocks
As noted above, changes in technology that raise the productivity of skilled workers
relative to that of unskilled is one of the explanations given for the recent increase in
US wage dispersion. It has been argued that these same shocks may have generated
the observed increases in European unemployment [see the OECD Jobs Study
(1994), Krugman (1994) and others] 26. The reason for the different response is a
different labor market policy regime. In Europe, where higher level of unemployment
compensation, minimum wages, and employment protection restrict accommodation
through downward wage adjustment, the response is likely to be higher unemployment,
particularly among the unskilled. The purpose of this section is to explore this
hypothesis within the equilibrium search and matching framework.
In the Mortensen-Pissarides model, a producing unit is a job-worker match. To
capture skill differences across workers, one can simply reinterpret the parameter p as
an efficiency unit measure of the worker's skill. Given two workers in an identical
match, the relative product per time period of the second worker is equal to the
ratio P2/P~ where pi, i = 1,2, represents the "skill" of each.
Let P, a set of real numbers, represent the set of skill types and G : P --+ [0, 1]
denote the distribution of the labor force population over these types. Given this
formalization, a pure skill-biased shock to technology can be interpreted as a meanpreserving increase in the spread of G. In this section we argue that such a shock will
increase unemployment in the Mortensen-Pissarides framework and that the extent
of the increase is likely to be larger for economies with high level of unemployment
compensation and stringent employment protection laws.
Given the assumption that skill differences are observable, as say they would be
if associated with different levels of education, we can consistently assume that the
labor market is segmented along skill lines. Across markets the reservation levels of
the idiosyncratic shock and market tightness can differ. In the sequel, let R(p) and 0(p)
characterize equilibrium relationships between these two endogenous variables and
worker skill. Obviously, these functions, which satisfy the job creation and job
destruction conditions Equations (2.22) and (2.26), and the steady-state Beveridge
condition (2.3), determine the equilibrium relationship between unemployment and
skill of interest in this section.
The qualitative differences in both market tightness 0 and the reservation value of
the idiosyncratic component of match product R at two different skill levels are readily
26 Acemoglu (1996, 1998) explains changes in unemployment and wage inequality in terms of
endogenoustechnologychanges and changes in labor supply.
1216
Table 3
Baseline parameter values
Parameter
Symbol(s)
Value
Discount rate
Matching elasticity
Recruiting cost
Creation cost
Productivity shock frequency
Minimum match product
Value of leisure
Worker's share
Firing Cost
r
tl
m~o)
C
A
y
b
/3
T
predicted by the model. For a more skilled worker, one characterized by a higher
value of general productivity parameter p, the relative opportunity cost of employment,
the ratio b/p, is lower. Given the assumption that hiring and firing costs increase
proportionately with p, the job destruction relation, D D in Figure 6, is lower given
a higher value o f p and the job creation relation C C is unaffected by variation in p.
As a consequence, markets for the more skilled are tighter, unemployment durations
are shorter. Furthermore, the reservation value of idiosyncratic productivity is lower
in markets for the more skilled and, consequently, the incidence of unemployment is
lower. These inferences are consistent with empirical findings to the extent that the
level of education is a good indicator of skill.
As the unemployment rate is a positive number by definition, the fact that it declines
with the skill parameter p implies that the unemployment-skill profile is convex, at
least on average. To the extent that the relationship has this shape, any increase in the
mean-preserving spread of the distribution of relative productivity, defined above as a
'skill-biased' technology shock, will increase unemployment. This effect can explain
the run up in European unemployment rates relative to those in the USA if European
labor policies increase the convexity of the unemployment-skill profile. In short, if
unemployment compensation and employment protection has a larger relative impact
on the unemployment of unskilled workers, then the same 'skill-biased' technology
shock increases unemployment more in countries with these policies.
To ascertain whether this explanation has force, we calibrate a simple version of
the model and then use it to compute the implied unemployment-skill profile for
different policy regimes. A matching function of the Cobb-Douglas form is assumed
with elasticity with respect to unemployment equal to t/, i.e., ln(q(0)) = -t/ln(0). The
distribution of idiosyncratic shocks is assumed to be uniform on the support [7, 1], i.e.,
F ( x ) = (x - y)/(1 - 7)Vx c [7, 1]. The baseline parameters used in the computations are
reported in Table 3. Except for value of income while unemployed b and the minimum
1217
.25
0.2
u(0.77, 1.1,p)
u(0.77, 0,p)
u(0.62, 0,p) 0.1
0 0
0.8
0.9
1.1
1.2
1.3
Fig. 8. Unemployment-skillprofiles.
match product },, which are chosen so that the steady-state unemployment rate of a
worker of average skill (p = 1) is 6.5% and the average duration of an unemployment
spell for such a worker is 3 months, values which reflect experience in the USA over
the past twenty years, the parameter values are similar to those assumed and justified
in Mortensen (1994b) and Millard and Mortensen (1997).
To obtain parameters that reflect the European experience, we maintain the same
values of all parameters except for unemployment income b and firing cost T which
are chosen to yield the same average unemployment rate but an average spell duration
of 9 months. The results, b = 0.77 and T = 1.1, are consistent with the fact that
unemployment compensation and the implicit cost of employment protection are both
substantially higher is Europe than in the USA and the fact that unemployment spells
are much longer in Europe.
The computed unemployment-skill profiles for three different policy parameter
combinations are illustrated in Figure 8. Specifically, each curve is a plot of the
equilibrium unemployment function u(b, T , p ) for value of the skill parameter p. The
flattest profile corresponds to low unemployment compensation and no employment
protection policy, the base line case of (b, T) = (0.62, 0). Given a more generous
unemployment compensation but still no employment protection, (b, T) = (0.77, 0),
the profile lies everywhere above the original but is substantially more convex, i.e.,
the steady-state unemployment rate of the less skilled is more responsive to the level
of unemployment compensation. Adding employment protection, as illustrated by the
solid curve representing the case (b, T) = (0.77, 1.1), actually lowers the unemployment
rate of the more skilled but raises that of the unskilled. In short, employment protection
policy induces even more convexity into the unemployment-skill profile.
1218
x(h) = x + h(x - 2 ) ,
(6.1)
where h ~> 0 is a parameter and 2 is the mean of the distribution. We shall consider
the effect of a shift in h on the steady-state equilibrium, evaluated in the neighborhood
of the old equilibrium, h = 0. In order to make the analysis more meaningful
for the question in hand, we assume that p2 >~ rU, i.e. that the reservation wage
of the unemployed job seekers is below mean productivity. This ensures that the
multiplicative shock reduces the productivity of at least some active low-productivity
jobs.
Reworking the job creation and job destruction conditions with x(h) replacing x is
straightforward. The job creation condition (2.22) becomes
c -(1-[3)((l+h)(1-R)
q(O)
r + )~
C- T
(6.2)
1219
cO.
(6.3)
Equilibrium is still shown by the two lines of Figure 6. Higher h shifts the DD line
up, implying higher reservation productivity at all levels of market tightness, because
the productivity of the marginal job is now worse. But higher h also shifts the CC line
to the right, because, for given reservation productivity greater than zero, the benefits
from the higher productivity of jobs above the mean outweigh the costs from the lower
productivity of jobs below the mean, the tail of which is truncated. Thus, job creation
unambiguously goes up at given unemployment stock but the effect of the higher h on
job destruction is ambiguous from the diagram alone. Differentiation of Equations (6.2)
and (6.3) with respect to h, however, shows that at h = 0, the reservation productivity
rises unambiguously (see Appendix A). So both job creation and job destruction rise
at given unemployment when there is a multiplicative productivity shift.
The effect of this shift on unemployment is ambiguous. On impact, unemployment
rises, because job destruction leads job creation, but whether unemployment rises or
falls in steady state depends on whether the direct impact on job destruction or job
creation dominates. The effect on wage inequality is also ambiguous, because, although
the range of productivities falls, the productivity of the marginal job may rise or fall.
The impact on the productivity of the marginal job, when evaluated at h = 0, is given
by
OR(h)
Oh
OR
Oh (2- R).
(6.4)
1220
when/3 is in the neighborhood of the elasticity of the matching function with respect to
unemployment, or the elasticity of q(0)]. Therefore, countries with more powerful labor
organization when hit by a shock that increases inequality are likely to experience more
unemployment, through less job creation, than countries with less powerful labor. It is
often asserted that labor is more powerful in Europe than in the USA, either because of
more powerful trade unions or because of legislation that favors labor. So this could be
one factor behind the different unemployment experience of the two continents. With
regard to inequality, however, the model does not have strong predictions 27.
6.3. Other influences
Several other influences on the equilibrium unemployment rate have been investigated
in the empirical literature, in search of the elusive explanation for the rise in European
unemployment. Virtually all the determinants of the equilibrium rate discussed in
Section 3 have been, at one time or another, listed as possible causes of higher
unemployment in Europe. This includes, in addition to unemployment income and
trade union power discussed above, the real rate of interest, taxes on wages, which
reduce the net surplus from a job match, "mismatch", by which is usually meant more
heterogeneity in the labor market and which is represented by a shift of the aggregate
matching function, employment protection legislation, which increases the costs of job
destruction, and on the positive side "active labor market policies", which reduce the
job creation costs and costs of labor to the firm.
As we saw in Section 2.3, higher real rate of interest reduces market tightness but
has ambiguous effects on the reservation productivity. At given unemployment rates
job creation falls. In terms of the Beveridge diagram, real interest rates have ambiguous
impact on the Beveridge curve but rotate the job creation line down. It has been argued,
however, that empirically higher real interest rates have depressed employment in the
OECD, i.e. that the job creation effect dominates over the job destruction effect [Phelps
(1994)].
Taxes on employment reduce the net surplus from the job, so whether they reduce
job creation or not depends on their influence on non-employment income. If nonemployment income is not taxed, their effects on the equilibrium of the model is similar
to a rise of non-employment income, i.e. they reduce job creation and increase job
destruction at given unemployment rate. Taxes, however, may also have distortionary
effects if they are not proportional to incomes, a topic that would take us beyond the
scope of our chapter 28.
27 One prediction is that the lowest wage is almost certain to rise when the multiplicativeshock arrives,
because of the increase in the reservationproductivityand in market tightness. Then, it becomes likely
that the cross effect of h and/3 on the lowest wage is also positive, so countries with less powerful labor
experience more increase in inequality. (These results are not proved here.)
28 Pissarides (1998), Mortensen (1994a), and Millard and Mortensen (1997) all study tax effects using
search equilibrium models. See also Daveri and Tabellini (1997), who explain the slowdown in growth
and rise in unemploymentin Europe by tax increases on labor.
1221
Mismatch can arise in our framework in the following sense. The aggregate
matching function conceals a lot of heterogeneity in the labor market. It is a convenient
modelling device when our interest is in aggregate changes rather than individual
employment histories. Out of all the interactions between the many heterogeneous
groups in the population, a stable relationship emerges between the job matching rate
and the stocks of aggregate unemployment and vacancies. But if conditions are such
that the type of workers and jobs available change, either in skill requirement or in
location, the aggregate outcome from the interaction between those groups is also
likely to change. An increase in mismatch shifts the aggregate matching function down
at all levels of vacancies and unemployment.
Mismatch bears some relationship to the more commonly discussed, in the
US literature, "sectoral shifts hypothesis", though it is more general [Lilien (1982)].
It also bears some relationship to the older view of "structural" unemployment, which
was thought to be unemployment arising from fast structural change in the economy
as a whole. In Europe, mismatch has been proposed by Jackman et al. (1989), Layard
et al. (1991) and others as a cause of the rise in European unemployment. The
oil, technology and other real shocks of the 1970s and 1980s increased the speed
with which unemployed workers needed to adapt to the changing requirements of
employers. This led to increased mismatch, which increased unemployment at given
vacancies. Although neither the sectoral shifts hypothesis in the USA, nor the mismatch
hypothesis in Europe, has had much success in accounting for a large fraction of
employment fluctuations, we look here at the implications of the mismatch hypothesis
within the search and matching framework.
The argument is that because labor in Europe is less mobile than in the USA, a
problem aggravated by the longer durations of unemployment in Europe, the changing
requirements of jobs lead to bigger and more prolonged shifts of the aggregate
matching function. Mismatch in the formal model is shown as a fall in the productivity
of the aggregate matching process, i.e. a downward shift of the transition rate q(O) at
all values of 0. This shifts the job creation line in Figure 6 down, reducing both
market tightness and the reservation productivity. But in addition, mismatch has the
implication that for given market tightness, the rate of job matching is lower. This
implies, in our model, a shift of the Beveridge curve out, over and above any effects
that there might be through job creation and job destruction. It is this additional shift
in the Beveridge curve that has attracted most attention in the discussions of mismatch
in the search literature 29.
It is clear that the overall effect of increased mismatch on equilibrium unemployment
is uncertain, because of the three interacting effects: less job entry at given
29 See Jackman et al. (1989) for the United Kingdom, Abraham and Katz (1986) and Blanchard and
Diamond (1989) for the USA and Jackman et al. (1990) for an international comparison of Beveridge
ctuve shifts. Andolfatto (1996) incorporates a stochastic shift parameter in the matching fucntion in his
calibrations of the search and matching model.
1222
unemployment, less job destruction and less job creation at given vacancies and
unemployment. The empirical literature, however, invariably takes the latter effect,
shown in the diagrams by the outward shift of the Beveridge curve, as the one that
dominates on unemployment. Of course, a sufficient condition for this is that for
given O, the fall in q(O) due to the direct effect of increased mismatch dominate the
fall in )~F(R) due to the indirect effect from the fall in the reservation productivity. But
since the job creation line in the Beveridge curve diagram (Figure 7) rotates down when
mismatch increases, the effect of increased mismatch on equilibrium job vacancies is
thought in the empirical literature to be unimportant. It is this latter property (higher
unemployment at given vacancies), which has been a feature of the 1980s rise in
European unemployment, that has attracted research in this area.
Countries with more restrictions in job separations are ones that have higher values
for the firing cost T. We saw that those countries should experience less job creation
and job destruction at given unemployment rate, through lower R and 0. The effect on
equilibrium unemployment is ambiguous but the effect on job reallocation is negative.
This result might explain why job reallocation rates differ across countries. In an
analysis of the data on job reallocations given in Section 1 and the employment
protection provisions in different countries as constructed by the OECD, Garibaldi et
al. (1997) found a clear relationship between employment protection legislation and
job reallocation. Given, however, the Beveridge curve equation that defines equilibrium
unemployment, there is no reason to expect a correlation between job reallocation and
equilibrium unemployment 30.
Firing costs might also explain, to some extent, the differences between the job
reallocation rates between small and large firms. Usually large firms in Europe are
subject to many more restrictions on firing workers, imposed either by legislation or
by trade unions. In Italy, where there are severe restrictions on job separations in large
firms, many more small firms come into operation and job reallocation rates in those
small firms are comparable to those in the USA [see Contini et al. (1995)].
Finally, lower job creation costs lead to more job creation at given unemployment
and more job destruction. Once again, the effects on equilibrium unemployment
are ambiguous. Many European governments, however, have tried to encourage job
creation by giving incentives which reduce job creation costs. One of the criticisms
levelled against such policies is that they encourage the creation of "unstable" jobs
that do not stay in operation for long periods. This argument is valid in our analysis
but still hiring subsidies may be justified, particularly if the worker's effective share
of match-specific investments in training and information are less than their share of
continuing match surplus [Mortensen (1996)].
30 Bertola and Rogerson(1997) claim that job reallocationrates do not differ as much across countries
as they should be expectedto do, given differencesin firing costs. They explain this by the differences
in wage inequalitythat characterizescountries, and which tends to reducejob reallocationrates.
1223
remarks
We have demonstrated that the search and matching approach provides a rich
framework for the analysis of aggregate employment fluctuations and of the observed
differences in average unemployment rates across countries. Calibrations o f the
models track the cyclical fluctuations in the job creation and job destruction flows
reasonably well. The framework provides a convenient medium for the analysis
of policy influences on unemployment, which lie at the heart of the explanations
of average unemployment differences across countries. Although there is still no
consensus on the causes of the higher unemployment rates in Europe than in the
USA, we have shown how policy influences, in particular the unemployment insurance
system and employment protection legislation, can contribute to the differences in both
unemployment rates and wage inequality.
Wages in the models that we have examined are determined by a fixed rule that
shares the economic rents that each employer-worker match creates. Other methods
of wage determination are also consistent with our framework and some promising
work is being done in this area of research. We discuss some of this work in our
companion chapter for the Handbook of Labor Economics. Another promising area
of current research is the interaction between technology, capital and labor in markets
with frictions. This area of research provides a natural framework for the analysis of
hold-up problems and problems of obsolescence and growth. We discussed some work
in this area in this chapter but much remains to be done.
Appendix
A. Mathematical
appendix
I-~[I-F(R)] N=(~-R)-~
(z-R)dF(z)-~cN. (A.1)
c~l 0 0 _ 1-[3 [ I _ R _ O R 1
Oq(O) Oh r + )~
~ "
(1.2)
Substitution of OR~Oh from Equation (A.1) into (A.2) reveals that the sign of O0/Oh
is the same as the sign of
1- R -
2 - R - v ~ z f / ( z - R) dF(z)
1- ~
2, [1 - F ( R ) ]
(A.3)
1224
Multiplying out the denominator o f Equation (A.3) and collecting terms, we find that
the sign of the terms in (A.3) is the same as the sign of
1 - 2 - -r +- ~ ,
(1 - z) dF(z),
(1.4)
J01
(1 - z) dF(z).
(1.5)
3~
1-~--~[1-F(R)]
OR_
0/3
1-~
o0]
cO + /3c-~
(A.6)
ctl O0
Oq(O)O/3
c
q(0)(1-/3)
1 OR
(1 - / 3 ) r +)~ 0/3'
(A.7)
Substitution of 00/0/3 from Equation (A.7) into (A.6) reveals that the sign of 0R/0/3
is the same as t/-/3. So R reaches a unique maximum at/3 = ~/, which is also the
efficient point, when the search externalities are internalized [see Hosios (1990)].
Although there is no reason why the two parameters should be equal, the usual
restriction on/3 in symmetric bargaining situations is fi = and the empirical evidence
on t/suggests that it is close to 0.5 so the restriction/3 = t/is a convenient simplification
that may be adopted. We shall do so in the derivations in this Appendix. Under the
restriction then that
OR
o/3
- 0,
(A.S)
0
t/(1 -/3)'
(A.9)
Turning now to the question of the cross partials of h and/3 on R and 0, i.e.
on the response of reservation productivity and market tightness to a multiplicative
1225
OhO/3
00
7/(1 -/3) Oh
< 0.
(A.10)
00
1 - fi oh
020
/3-OhO/3"
(A.11)
Making use o f Equations (A.2) and (A.10), we easily find that the sign o f Equation (A.11) is the same as
/3(1 - r/)
(1 - / 3 ) r/'
(A.12)
References
Abraham, K.J., and L.E Katz (1986), "Cyclical unemployment: sectoral shifts or aggregate disturbances?",
Journal of Political Economy 94:507-522.
Acemoglu, D. (1996), "Changes in unemployment and wage inequality: an alternative theory and some
evidence", Discussion Paper No. 1459 (CEPR, July).
Acemoglu, D. (1998), "Why do new technologies complement skills? Directed technical change and
wage inequality", Quarterly Journal of Economics 113:1055-1089.
Aghion, P., and P. Howitt (1992), "A model of growth through creative destruction", Econometrica
60:323-350.
Aghion, P., and P. Howitt (1994), "Growth and unemployment", Review of Economic Studies 61(July):
477-494.
Aghion, P., and P. Howitt (1998), Endogenous Growth Theory (MIT Press, Cambridge, MA).
Akedof, G.A., A. Rose and J.L. Yellen (1998), "Job switching and job satisfaction in the U.S. labor
market", Brookings Papers on Economic Activity 1998(2):495-582.
Andolfatto, D. (1996), "Business cycles and labor market search", American Economic Review 86:
112-132.
Andolfatto, D., and P. Gomme (1996), "Unemployment insurance, labor market dynamics, and social
welfare", Carnegie-Rochester Conference Series on Public Policy.
Arrow, KJ. (1965), "The theory of risk aversion", lecture 3 of K.J. Arrow, Aspects of the Theory of
Risk-Bearing (Yrjo Jahnssoninsaatio, Helsinki).
Bertola, G., and R.J. Caballero (1994), "Cross-sectional efficiency and labour hoarding in a matching
model of unemployment", Review of Economic Studies 61 (July):435-456.
Bertola, G., and R. Rogerson (1997), "Institutions and labor reallocation", European Economic Review
41(June): 1147-1171.
1226
Binmore, K.G., A. Rubinstein and A. Wolinsky (1986), "The Nash bargaining solution in economic
modelling", Rand Journal of Economics 17:176-188.
Blanchard, O.J., and EA. Diamond (1989), "The Beveridge curve", Brookings Papers on Economic
Activity 1989(1):1-60.
Blanchard, O.J., and EA. Diamond (1990), "The cyclical behavior of the gross flows of U.S. workers",
Brookings Papers on Economic Activity 1990(2):85-143.
Blanchard, O.J., and L.W. Katz (1997), "What we know and do not know about the natural rate",
Economic Perspectives 11:51-72.
Blanchard, O.J., J.E Jimeno, J. Andres, C. Bean, E. Malinvand, A. Revenga, D. ,Snower, G. SaintPaul, R.M. Solow, D. Taguas and D. Toharia (1995), "Spanish unemployment: is there a solution?",
(CEPR).
Burdett, K., N.M. Kiefer, D.T. Mortensen and G.R. Neumann (1984), "Earnings, unemployment, and
the allocation of time over time", Review of Economic Studies 51:559-578.
Burdett, K., D.T. Mortensen and R. Wright (1996), "Search and matching equilibrium", manuscript, in
process.
Caballero, R.J., and M.L. Hammour (1994), "The cleansing effect of recession", American Economic
Review 84:1350-1368.
Caballero, R.J., and M.L. Hammour (1996), "On the timing and efficiency of creative destruction",
Quarterly Journal of Economics 111:805-852.
Calmfors, L. (1995), "Will high unemployment in Sweden become persistent? An introduction", Swedish
Review of Economic Policy 2:3-14.
Clark, K.B., and LH. Summers (1979), "Labor market dynamics and unemployment: a reconsideration",
Brookings Papers on Economic Activity 1979(1): 13-60.
Cole, H.L., and R. Rogerson (1996), "Can the Mortensen-Pissarides matching model match the
business cycle facts?", Staff Report 224 (Federal Reserve Bank of Minneapolis, Research Department,
December).
Contini, B., and R. Revelli (1997), "Gross flows vs. net flows in the labor market: what is there to be
learned?", Labour Economics 4:245-263.
Coutini, B., L. Pacelli, M. Filippi, G. Lioni and R. Revelli (1995), "A study of job creation and
job destruction in Europe", Study for the Commission of the European Communities (file soc
94-1023318-05003 - item B34010).
Daveri, E, and G. Tabellini (1997), "Unemployment, growth and taxation in industrial countries",
Discussion Paper 1681 (CEPR, July).
Davis, S.J., and J.C. Haltiwanger (1990), "Gross job creation and destruction: microeconomic evidence
and macroeconomic implications", NBER Macroeconomics Annual 5:123-168.
Davis, S.J., and J.C. Haltiwanger (1992), "Gross job creation, gross job destruction, and employment
reallocation", Quarterly Journal of Economics 107:819-863.
Davis, S.J., J.C. Haltiwanger and S. Schuh (1996), Job Creation and Destruction (MIT Press, Cambridge,
MA).
den Haan, W.J., G. Ramey and J. Watson (1997), "Job destruction and propagation of shocks", Working
Paper 97-23 (University of California at San Diego).
Diamond, P.A. (1982a), "Aggregate demand management in search equilibrium", Journal of Political
Economy 90:881-894.
Diamond, P.A. (1982b), "Wage determination and efficiency in search equilibrium", Review of Economic
Studies 49:217-227.
Dolado, J.J., and J.E Jimeno (1997), "The causes of Spanish unemployment: a structural VAR approach",
European Economic Review 41(July): 1281 1307.
Eriksson, C. (1997), "Is there a trade-off between employment and growth?", Oxford Economic Papers
49(January):77-88.
Farber, H. (1994), "An analysis of interRrm worker mobility", Journal of Labor Economics 12(4):
554-593.
1227
Friedman, M. (1968), "The role of monetary policy", American Economic Review 58:1-17.
Galbraith, J.K. (1996), "Inequality and unemployment: an analysis across time and countries",
mimeograph (University of Texas at Austin).
Garibaldi, R, J. Konings and C.A. Pissarides (1997), "Gross job reallocation and labour market policy",
in: D.J. Snower and G. de la Dehesa, eds., Unemployment Policy: Government Options for the Labour
Market (Cambridge University Press, Cambridge) 467-489.
Grossman, G.M., and E. Helpman (1991), "Quality ladders and product cycles", Quarterly Journal of
Economics 106:557-586.
Hall, R.E. (1995), "Lost jobs", Brookings Papers on Economic Activity 1995(1):221 256.
Hansen, B. (1970), "Excess demand, unemployment, vacancies, and wages", Quarterly Journal of
Economics 84(February): 1~3.
Hansen, G.D. (1985), "Indivisible labor and the business cycle", Journal of Monetary Economics
16(November) :309-327.
Hosios, A.J. (1990), "On the efficiency of matching and related models of search and unemployment",
Review of Economic Studies 57:279-298.
Howitt, R (1988), "Business cycles with costly search and recruiting", Quarterly Journal of Economics
103 (February): 147-166.
Jackman, R., R. Layard and C.A. Pissarides (1989), "On vacancies", Oxford Bulletin of Economics and
Statistics 51 (November):377-394.
Jackman, R., C.A. Pissarides and S. Savouri (1990), "Labour market policies and unemployment in the
OECD", Economic Policy 11:449-490.
Jovanovie, B. (1979), "Job matching and the theory of turnover", Journal of Political Economy 87:
972-990.
Krugman, R (1994), "Past and prospective causes of high unemployment", in: Reducing Unemployment:
Current Issues and Policy Options (Federal Reserve Bank of Kansas City, August 1994).
Kydland, EE., and E.C. Prescott (1982), "Time to build and aggregate fluctuations", Econometrica
50:1345 1370.
Layard, R., S. Nickell and R. Jaekman (1991), Unemployment: Macroeconomic Performance and the
Labour Market (Oxford University Press, Oxford).
Lilien, D.M. (1982), "Sectoral shifts and cyclical unemployment", Journal of Political Economy 90:
777 793.
Ljungqvist, L., and T. Sargent (1995), "The Swedish unemployment experience", European Economic
Review 39:1043-1070.
Lucas Jr, R.E. (1987), Models of Business Cycles (Basil Blackwell, New York).
MacLeod, W.B., and J.M. Malcomson (1993), "Investments, holdup, and the form of market contracts",
American Economic Review 83:811-837.
Marimon, R., and E Zilibotti (1998), "'Actual' versus 'virtual' employment in Europe. Is Spain different?",
European Economic Review 42:123-153.
Martin, J.R (1994), "The extent of high unemployment in OECD countries", in: Reducing Unemployment:
Current Issues and Policy Options (Federal Reserve Bank of Kansas City, August 1994).
Merz, M. (1995), "Search in the labor market and the real business cycle", Journal of Monetary
Economics 36:269-300.
Millard, S.P., and D.T. Mortensen (1997), "The unemployment and welfare effects of labour market
policy: a comparison of the U.S. and U.K.", in: D. Snower and G. de la Dehesa, eds., Unemployment
Policy: Government Options for the Labour Market (Cambridge University Press, Cambridge).
Mortensen, D.T. (1978), "Specific capital and labor turnover", The Bell Journal of Economics 9(2):
572-586.
Mortensen, D.T. (1982a), "The matching process as a noncooperative/bargaining game", in: J.J. McCall,
ed., The Economics of Information and Uncertainty (University of Chicago Press, Chicago, IL)
233-254.
1228
Mortensen, D.T. (1982b), "Property rights and efficiency in mating, racing, and related games", American
Economic Review 72:968-979.
Mortensen, D.T. (1994a), "Reducing supply side disincentives to job creation" in: Reducing
Unemployment: Current Issues and Policy Options (Federal Research Bank of Kansas City).
Mortensen, D.T. (1994b), "The cyclical behavior of job and worker flows", Journal of Econorrfic Dynamics
and Control 18:1121-1142.
Mortensen, D.T. (1996), "The unemployment and income effects of active labor market policy: the case
of the U.K.", in: H. von Christian Drager, E Pissulla, and A.W. yon Czege, eds., More Competition,
More Jobs Is Full Employment an Illusion? (Nomos Verlagsgesellschaft, BadenrBaden).
Mortensen, D.T., and C.A. Pissarides (1994), "Job creation and job destruction in the theory of
unemployment", Review of Economic Studies 61(July):397-415.
Mortensen, D.T., and C.A. Pissarides (1998), "Technological progress, job creation and job destruction",
Review of Economic Dynamics 1:733-753.
Mortensen, D.T., and C.A. Pissarides (1999), "Unemployment responses to 'skill-biased' technology
shocks: the role of labor market policy", Economic Journal 109(April):242-265.
OECD (1994), Jobs Study (OECD, Paris).
Phelps, E.S. (1967), "Phillips Curves, expectations of inflation and optimal unemployment over time",
Economica 34:254-281.
Phelps, E.S. (1968), "Money-wage dynamics and labor-market equilibrium", Journal of Political Economy
76:679-711.
Phelps, E.S. (1994), Structural Slumps (Harvard University Press, Cambridge, MA).
Phelps, E.S., et al. (1970), Microeconomic Foundations of Employment and Inflation Theory (Norton,
New York).
Pissarides, C.A. (1984a), "Search intensity, job advertising and efficiency", Journal of Labor Economics
2(January):128-143.
Pissarides, C.A. (1984b), "Efficient job rejection", Economic Journal Conference Papers 94:97-108.
Pissarides, C.A. (1985), "Short-rnn equilibrium dynamics of unemployment, vacancies and real wages",
American Economic Review 75(September):676-690.
Pissarides, C.A. (1986), "Trade unions and the efficiency of the natural rate of unemployment", Journal
of Labor Economics 4(October):582-595.
Pissarides, C.A. (1990), Equilibrium Unemployment Theory (Basil Blackwell, Oxford).
Pissarides, C.A. (1994), "Search unemployment with on-the job search", The Review of Economic
Studies 61 (July):457-477.
Pissarides, C.A. (1998), "The impact of employment tax cuts on unemployment and wages; the role of
unemployment benefits and tax structure", European Economic Review 42(January): 155-183.
Rubinstein, A., and A. Wolinsky (1985), "Equilibrium in a market with sequential bargaining",
Econometrica 53:1133-1150.
Scarpetta, S. (1996), "Assessing the role of labour market policies and institutional settings on
unemployment: a cross-country study", OECD Economic Studies No. 26 (OECD, Paris).
Solow, R.M. (1956), "A contribution to the theory of economic growth", Quarterly Journal of Economics
70(1):65-94.
Solow, R.M. (1959), "Investment and technical progress", in: Mathematical Methods in the Social
Sciences (Stanford University Press, Stanford, CA).
Stigler, G.J. (1962), "Information in the labor market", Journal of Political Economy 70:94-105.
Wolinsky, A. (1987), "Matching, search, and bargaining", Journal of Economic Theory 42(August):
311-333.
Yashiv, E. (1997), "The determinants of equilibrium unemployment", Working Paper CR 600/1997
(HEC).
AUTHOR INDEX
Bacchetta, P. 1344
Bacchetta, E, s e e Feldstein, M. 1637
Bachelier, L. 1316
Backus, C.K. 549
Backus, D. 1017, 1031, 1270, 1405, 1414,
1415
Backus, D.K. 9, 42, 45, 938, 1316, 1708
I-1
I-2
Bade, R. 1432, 1438
Bagehot, W. 155, 1485, 1515
Bagwell, K. 1125
Bagwell, K., s e e Bernheim, B.D. 1647
Bailey, M.J. 1643
Bairoch, P. 719, 724
Baker, J.B. 1125
Balasko, Y. 427, 506
Balassa, B.A. 705
Balke, N.S. 6, 61, 114, 204, 205, 221
Ball, L. 42, 72, 199, 1023, 1037, 1039, 1041,
1127, 1415, 1499, 1504, 1542, 1632, 1650,
1651
Ball, R. 1321
Ballard, C. 1639
Banerjee, A., s e e Aghion, R 1377
Bange, M.M., s e e De Bondt, W.E 1321
Banks, J. 751, 758, 759, 770, 783, 788,
790-792
Banks, J., s e e Attanasio, O.P. 756, 759, 793,
794
Bannerjee, A.V 1332
Bansal, R. 1255
Barberis, N. 1294, 1322
Barclays de Zoete Wedd Securities 1238
Barkai, H. 1572
Barnett, S. 831
Barnett, W. 538, 540
Barone, E. 702
Burro, R.J. 101, 157, 158, 173, 237, 245, 246,
252, 269, 271,272, 277-281,284, 643, 651,
657, 659, 671, 675, 681, 683-685, 688,
689, 691-694, 696, 943, 974, 1023, 1055,
1155, 1404, 1405, 1411, 1412, 1414, 1415,
1425, 1438, 1439, 1466, 1485-1489, 1637,
1641, 1642, 1645, 1662, 1675, 1702, 1705,
1707
Barsky, R. 43, 558, 564, 565
Barsky, R., s e e Solon, G. 579, 1058, 1102,
1106
Barsky, R., s e e Warner, E.J. 1019
Barsk-y, R.B. 182, 215, 216, 1149, 1237, 1277,
1294-1296, 1653
Barth, J.R. 1657
Bartle, R.G. 76
Barueci, E. 525
Basar, T. 1449
Basu, S. 399, 402, 433, 983, 992, 994, 1069,
1080-1082, 1096, 1097, 1117, 1142
Bates, D.S. 1310, 1324
Baumol, W.J. 252, 269
Author
Index
Author
Index
I-3
Boadway, R. 1463
Bodnar, G. 1318
B6hrn, V. 475, 646
Bohn, H. 1465, 1622, 1650, 1691
Boldrin, M. 362, 399, 400, 506, 962, 1062,
1284, 1297, 1465
Bolen, D.W. 1325
Bollerslev, T. 1236, 1280
BoRon, E, s e e Aghion, E 1377, 1450, 1454,
1465
Bona, J.L. 313
Boothe, P.M. 1658
Bordo, M.D. 152, 155-160, 162, 164-167, 182,
184, 185, 194, 202-204, 207-209, 211,215,
217~21, 1404, 1438, 1590
Bordo, M.D., s e e Bayoumi, T. 161
Bordo, M.D., s e e Belts, C.M. 217
Borenstein, S. 1124
Boschan, C., s e e Bry, G. 8
Boschen, J.E 139
Boskin, M.J. 618
Bossaerts, E 454
Bosworth, B., s e e Collins, S. 653
Bourguignon, E, s e e Levy-Leboyer, M. 222
Bovenberg, A.L., s e e Gordon, R.H. 1637
Bovenberg, L., s e e Beetsma, R. 1411
Bowen, W. 619
Bowman, D. 1313
Boyd, W.H., s e e Boleu, D.W. 1325
Boyle, M., s e e Paulin, G. 751
Boyle, E 380
Boyle, P.P., s e e Tan, K.S. 334
Brainard, W.C. 817
Brauch, R., s e e Paulin, G. 751
Braun, R.A. 974
Braun, S.N., s e e K r a n e , S.D. 876, 877
Bray, A. 1290
Bray, M. 454, 463, 465, 466, 473-475, 527
Brayton, E 1043, 1344, 1485
Brayton, E, s e e Hess, G.D. 1485, 1509
Breeden, D. 1246
Breiman, L. 289
Bresnahan, T.E 911,912
Bretton Woods Commission 208
Broadbent, B. 1412
Broadbent, B., s e e Barro, R.J. 1412
Broadie, M., s e e Boyle, P. 380
Brock, W.A. 319, 407, 455, 528, 532, 547, 552,
556, 942, 951, 1507
Brown, C. 585
Brown, E, s e e Ball, R. 1321
1-4
Brown, S. 1242
Browning, E. 1463
Browning, M. 598, 606, 607, 610-612, 750,
752, 771,778, 787, 792, 798, 803
Browning, M., s e e Attanasio, O.P. 607, 608,
610, 611,613, 779, 789, 791, 1655
Browning, M., s e e Blundell, R. 611,612, 779,
781,783, 790, 791
Broze, L. 487, 488
Brugiavini, A. 775
Brugiavini, A., s e e Banks, J. 770, 788
Brumberg, R., s e e Modigliani, E 761
Brumelle, S.L., s e e Puterman, M.L. 336, 338
Brunner, A.D. 104
Brunnel, K. 179, 183, 191, 1025, 1491
Bruno, M. 471, 1090, 1496, 1538, 1539, 1543,
1553
Bry, G. 8
Bryant, R.C. 1043, 1491, 1497, 1516-1518
Bryant, R.R. 1313
Buchanan, J.M. 1631, 1642
Buchholz, T.G. 1643
Buckle, R.A. 1019
Bufinan, G. 1543
Buiter, W. 1030, 1521
Bulirsch, R., s e e Stoer, J. 334
Bull, N. 1675, 1711
Bullard, J. 466, 507, 509, 515, 526
Bullard, J., s e e Arifovie, J. 527
Bulow, J. 1448, 1449
Burdett, K. 1173, 1196
Bureau of the Census 1618, 1619
Burns, A.E 5, 8, 931,934
Burns, A.E, s e e Mitchell, W.C. 8, 44
Burnside, C. 399, 930, 980-985, 994, 1078,
1142, 1162
Burfless, G. 618, 620
Butkiewiez, J.L. 1621
Caballe, J. 578
Caballero, R.J. 399, 749, 771, 794, 801, 802,
821-823, 828, 830, 832, 834-838, 840-842,
844, 846, 847, 852, 855, 856, 994, 1032,
1157, 1158, 1160, 1187, 1210, 1211, 1213,
1472
Caballero, R.J., s e e Bertola, G. 801,821, 834,
840, 843, 1187
Cagan, E 157, 161,203, 1534
Cage, R., s e e Paulin, G. 751
Calmfors, L. 1214
Author
Index
Author
Index
1-5
Christiano, L.J. 43, 67-70, 83, 84, 89, 91-94,
99, 108, 109, 114, 115, 124, 137, 143, 144,
314, 329, 330, 339, 347, 349, 350, 355,
362, 364, 367, 369, 370, 376, 377, 379,
426, 504, 547, 764, 881, 888, 909, 952,
962, 974, 1011, 1017, 1018, 1021, 1030,
1038, 1089, 1100, 1296, 1365, 1369, 1708,
1736
Chffstiano, L.J., s e e Aiyagafi, S.R. 1140
Chffstiano, L.J., s e e Boldrin, M. 962, 1284,
1297
Christiano, L.J., s e e Charl, VV 72, 1449, 1673,
1675, 1676, 1691, 1699, 1708 1710, 1720,
1723
Chung, K.L. 299
Clarida, R. 95, 96, 136, 422, 1364, 1368,
1486
Clark, D., s e e Kushner, H. 476
Clark, J.M. 816
Clark, K.B. 602, 1173
Clark, RB., s e e Mussa, M. 208
Clark, T.A. 173
Clark, T.E. 1091, 1485
Cochrane, J. 1120
Cochrane, J.H. 101, 211, 796, 1234, 1246,
1249, 1296
Cochrane, J.H., s e e Campbell, J.Y. 1237, 1251,
1284, 1286
Coe, D.T. 265
Cogley, T. 211,395, 547, 967, 1142, 1503
Cohen, D. 271
Cohen, D., s e e Greenspan, A. 798, 844, 847
Cob_n, R., s e e Modigliani, E 1321
Cole, H.L. 576, 1163, 1194, 1201-1203, 1207,
1446, 1449, 1603
Cole, H.L., s e e Chaff, V.V. 1459
Coleman, T. 601
Coleman, W.J. 367, 380
Coleman II, W.J. 114
Coleman II, W.J., s e e Bansal, R. 1255
Collins, S. 653
Conference Board 43
Congressional Budget Office 1618, 1619, 1621,
1624-1627, 1639, 1640, 1660
Conley, J.M., s e e O'Barr, W.M. 1332
Conlon, J.R. 1032
Constantinides, G.M. 559, 567, 781,803, 1237,
1284, 1291, 1293
Constantinides, G.M., s e e Ferson, W.E. 1284
Contini, B. 1177, 1178, 1180, 1200, 1222
Cook, T. 194, 195, 1493
I-6
Cooley, T.E 42, 69, 97, 101, 115, 124, 137,
376, 380, 408, 411, 549, 847, 954, 962,
974, 1376, 1463, 1736
Cooley, T.E, s e e Cho, J.O. 974, 976, 1025,
1036
Cooper, R. 204, 398, 824
Cooper, R., s e e Azariadis, C. 395
Cooper, R., s e e Chatterjee, S. 996, 1126
Cootner, EH. 1316
Corbo, V. 1543, 1554
Correia, I. 974, 1537, 1675, 1720, 1733
Cossa, R. 584
Council of Economic Advisers 1639
Cox, D. 705
Cox, W.M. 1621
Cox Edwards, A., s e e Edwards, S. 1543, 1554,
1555, 1575
Crawford, V.E 475
Crossley, T., s e e Browning, M. 610, 798
Croushore, D. 1485, 1653
Crucini, M.J. 178, 705
Crucini, M.J., s e e Baxtel, M. 1296
Cukierman, A. 1404, 1414, 1415, 1432, 1437,
1438, 1450, 1456, 1463, 1465
Cukierman, A., s e e Alesina, A. 1424, 1426
Cukierman, A., s e e Brunner, K. 1025
Cummings, D., s e e Christensen, L.R. 673,
688
Cummins, J.G. 822, 856, 1344
Cunliffe Report 161
Currie, D. 454, 504
Cushman, D.O. 95, 96
Cutler, D.M. 797, 1290, 1320, 1321, 1624
Cyrus, T., s e e Frarlkel, J.A. 280
Dahlquist, G. 337
Daniel, B.C. 1647
Daniel, K. 1322
Danthine, J.-P. 329, 370, 952, 962, 1002,
1157
Darby, M.R. 166
Dasgupta, P. 655, 656
d'Autttme, A. 487
DaVanzo, J. 618
Daveri, E 1220
Davidson, J. 750
Davies, J.B. 766
Davis, D. 1033
Davis, P.J. 333
Davis, S.J. 1151, 1152, 1160, 1161, 1176,
1178, 1180, 1194, 1199
Author
Index
Author
Index
I-7
Duffy, J., s e e Bullard, J. 526
Duguay, E 215
Dumas, B. 561, 564
Dunlop, J.T. 939, 1059
Dunn, K.B. 800, 1284
Dunne, T., s e e Doms, M. 823, 838
Dupor, B. 994
Durkheim, 1~. 1331
Durlauf, S.N. 254, 262-264, 268, 270, 271,
287, 289, 303, 550, 905 907
Dm-lauf, S.N., s e e Bernard, A.B. 254, 271,287,
288
Dutta, EK. 380
Dutta, S. 1019, 1020
Dutta, S., s e e Levy, D. 1014, 1015, 1019
Dutton, J. 156
Dyl, E.A. 1334
Dynan, K.E. 770
I-8
Eichengreen, B. 152, 154-157, 160, 162 164,
168, 178, 185, 187, 189, 204, 208, 209,
211,219, 1449, 1465, 1590
Eichengreen, B., s e e Bayolmai, T. 211, 216,
217, 219
Eichengreen, B., s e e Bordo, M.D. 162
Eichengreen, B., s e e CaseUa, A. 1463, 1465
Eijffinger, S. 1404, 1432, 1438
Eisner, R. 817, 1310, 1621, 1622
Ekeland, I. 1689
E1 Karoui, N. 835
Elias, V.J. 673
Ellison, G. 475, 1124
Ellson, R.E., s e e Bordo, M.D. 157
Elmendorf, D.W. 1439
Elmendorf, D.W., s e e Ball, L. 1650, 1651
Elmendorf, D.W., s e e Feldstein, M. 1656
Emery, K.M. 215
Emery, K.M., s e e Balke, N.S. 114
Engel, E., s e e Caballero, R.J. 801, 802, 821,
835-838, 840-842, 994, 1032, 1158
Engelhardt, G. 1344
Engle, R., s e e Bollerslev, T. 1280
Engle, R.E 50
Engle, R.E, s e e Chou, R.Y. 1236, 1280
Englund, E 9
Epstein, L.G. 556, 558, 564, 565, 744, 769,
1250, 1256
Erceg, C. 1041
Erceg, C.J., s e e Bordo, M.D. 182
Eriksson, C. 1208
Erlieh, D. 1314
Ermoliev, Y.M., s e e Arthur, W.B. 476
Escolano, J. 1718
Esquivel, G., s e e Caselli, E 277-279, 283,284,
286
Esteban, J.-M. 264
Estrella, A. 43, 1281, 1485
Evans, C. 982
Evans, C.L. 105
Evans, C.L., s e e Bordo, M.D. 182
Evans, C.L., s e e Christiano, L.J. 67, 68, 70, 83,
84, 89, 91-94, 99, 108, 137, 143, 144, 1011,
1021, 1038, 1089, 1100, 1365, 1369
Evans, C.L., s e e Eiehenbaum, M. 83, 94, 96,
137
Evans, G.W. 425, 426, 453-455, 461-465, 468,
470, 472-478, 480, 481, 483, 484, 487,
489-492, 495-497, 500, 502, 504-507,
509-513, 516, 518-521,526-528, 530-532,
1025, 1125
Author
Index
Evans, M. 182
Evans, P. 283, 1635, 1647, 1656-1659
Faig, M. 1675, 1720
Fair, R. 1416, 1425
Fair, R.C. 876, 1077, 1491
Fair, R.C., s e e Dominguez, K. 182
Falcone, M. 326
Fallick, B.C. 855
Fama, E.E 1235, 1280, 1281, 1307, 1316,
1320-1323
Farber, H. 1200
Farmer, R. 662, 1002
Farmer, R.E. 391,395, 396, 411-414, 427-430,
434, 437, 500, 505
Farmer, R.E., s e e Benhabib, J. 395, 399-402,
408, 412-414, 417, 425,427, 431,433-435,
442, 505
Farrell, J. 1121
Faust, J. 69, 217, 1416, 1425, 1437
Fanvel, Y. 1573
Favaro, E. 1554, 1555
Fay, J.A. 1077, 1103
Fazzari, S.M. 818, 1344
Fazzari, S.M., s e e Carpenter, R.E. 881,912,
1344
Fazzari, S.M., s e e Chirinko, R.S. 1066, 1086
Featherstone, M. 1332
Federal Reserve Board 176
Feenberg, D. 60
Feenstra, R. 1569
Feenstra, R.C., s e e Bergen, ER. 1041
Feiwel, G.R. 535
Feldman, M. 474
Feldstein, M. 44, 197, 1485, 1497, 1498, 1622,
1631, 1633, 1636, 1637, 1639, 1656, 1660
Feldstein, M.S. 904, 906
Felli, E. 1083, 1122
Fellner, W. 641,657
Ferejohn, J. 1425
Fernald, J.G., s e e Basu, S. 399, 402, 433, 994,
1117, 1142
Fernandez, R. 1543, 1562
Ferris, S.P. 1314
Ferson, W.E. 1284
Festinger, L. 1314
Fethke, G. 1037
Fiebig, D.G., s e e Domberger, S. 1019
Filippi, M., s e e Contini, B. 1177, 1178, 1180,
1222
Fillion, J.E 1498
Author
Index
I-9
French, K.R., s e e Fama, E.E 1235, 1281, 1320,
1323
Frenkel, J.A. 203, 1630
Frenkel, J.A., s e e Aizenman, J. 1497
Frennberg, P. 1238
Friedman, B.M. 43, 44, 1632, 1642
Friedman, D. 475
Friedman, J.H, s e e Breiman, L. 289
Friedman, M. 46, 48, 61, 137, 154, 160, 162,
168, 172, 176, 179, 180, 185, 189, 195, 203,
222, 275, 376, 572, 761, 762, 943, 1011,
1173, 1325, 1485, 1488, 1496, 1537, 1674,
1720
Froot, K. 1266, 1316
Frydman, R. 453, 454, 474, 528, 536, 539
Fuchs, G. 464, 474
Fudenberg, D. 455, 475, 1155
Fudenberg, D., s e e Ellison, G. 475
Fuerst, T. 99, 974, 1378
Fuerst, T.~ s e e Carlstrom, C, 1348, 1357, 1368~
1378, 1379
Fuhrer, J.C. 454, 905, 908, I039, 1040, 149I,
1518
Fuhrer, J.C., s e e Carroll, C.D. 769, 785
Fukuda, S.-i, 875
Fuller, W.A., s e e Dickey~ D.A. 53, 54, 212
Fullerton, D. 576, 588, 616
Funldaouser, R. 699
Futia, C. 299
Galbraith, J.K. 1182
Gale, D. 389, 475, 849, 851, 1376
Gale, D., s e e Chamley, C. 851
Gale, W.G. 1646
Galeotti, M. 909, 1086, 1124
Gali, J. 395, 405~407, 426, 429, 434, 993, 994,
1117, 1119, 1120, 1129
Gali, J. 67, 69, 217
Gali, J., s e e Benhabib, J. 424
Gali, J., s e e Clarida, R. 96, 136, 422, 1364,
1368, 1486
Gallarotti, G.M. 154
Gallego, A.M. 321,322
Galor, O. 262, 263, 272, 660
Gandolfi, A.E., s e e Darby, M.R. 166
Garber, P.M. 165, 1323, 1543
Garber, EM., s e e Eichengreen, B. 187, 189
Garber, P.M., s e e Flood, R.P. 408, 1595, 1596
Garcia, R. 790
Garibaldi, P. 1180 1222
Garratt, A. 504
1-10
Garratt, A., s e e Carrie, D. 454, 504
Garriga, C. 1675, 1718
Gaspar, J. 324, 369
Gastil, R.D. 689
Gatti, R., s e e Alesina, A. 1432
Gavin, W. 1485
Geanakoplos, J.D. 395, 458, 1322
Gear, C.W. 346
Geczy, C.C., s e e Brav, A. 1290
Gelb, A., s e e De Melo, M. 1535, 1551
Genberg, H. 165, 1428
Geoffard, RY., s e e Chiappori, RA. 391
Gerlach, S., s e e Baechetta, E 1344
Gersbach, H. 1376
Gertler, M. 83, 92-94, 1040, 1343, 1348, 1366,
1373, 1374, 1376-1378
Gertler, M., s e e Aiyagari, S.R. 1293, 1631
Gertler, M., s e e Bernanke, B.S. 92, 144, 183,
856, 857, 1036, 1345, 1346, 1352, 1357,
1365, 1369, 1371, 1373, 1376-1378, 1578
Gertler, M., s e e Clarida, R. 95, 96, 136, 422,
1364, 1368, 1486
Geweke, J. 34, 334
Geweke, J., s e e Barnett, W. 540
Geweke, J.E 89
Ghali, M., s e e Surekha, K. 908
Ghez, G. 615, 752, 759
Ghezzi, E 1572
Ghosh, A.R. 202, 207, 208
Giavazzi, E 167,203, 1438, 1446, 1449, 1580
Giavazzi, E, s e e Missale, A. 1450
Gibson, G.R. 1307
Gigerenzer, G. 1308, 1318
Gilbert, R.A. 195
Gilchrist, S. 847, 1344
Gilchrist, S., s e e Bernanke, B.S. 856, 1036,
1345, 1373, 1376
Gilchrist, S., s e e Gertler, M. 83, 92-94, 1366,
1373, 1374, 1376
Gill, EE. 329
Gilles, C., s e e Coleman II, W.J. 114
Gilson, R.J. 1154
Giovannini, A. 156, 158, 160, 166, 169, 380
Giovannini, A., s e e Giavazzi, F. 167
Gizycki, M.C., s e e Gruen, D.K. 1316
Glasserman, E, s e e Boyle, E 380
Glazer, A. 1456, 1465
Glomm, G. 712, 1472
Glosten, L. 1280
Goetzmann, W., s e e Brown, S. 1242
Author
Index
Author
Index
1-11
Haberler, G. 185
Hahn, E 661
Hahn, T., s e e Cook, T. 194, 1493
Hahn, W. 479
Hairault, J.-O. 1036
Haldane, A.G. 1432, 1438, 1485, 1495, 1497
Haley, W.J. 585
Hall, G. 911
Hall, R.E. 399, 556, 573, 595, 607, 608, 673,
679, 680, 683-686, 702, 765, 767-769,
784, 789, 791, 794, 817, 856, 930, 982,
1068, 1070, 1079, 1089, 1092, 1095, 1096,
1141-1143, 1145, 1151-1153, 1157, 11601164, 1200, 1261, 1485, 1493, 1498, 1655,
1656
Hall, S., s e e Currie, D. 454, 504
Hall, S., s e e Garratt, A. 504
Hailerberg, M. 1460, 1465
Haltiwanger, J., s e e Caballero, R.J. 821, 837,
838, 840-842, 1158
Haltiwanger, J., s e e Cooper, R. 824
Haltiwanger, J.C. 881
Haitiwanger, J.C., s e e Abraham, K.G. 1058
Haltiwanger, J.C., s e e Davis, S.J. 1151, 1152,
1160, 1161, 1176, 1178, 1180, 1194, 1199
Hamermesh, D. 577
Hamilton, A. 1659
Hamilton, J. 963
Hamilton, J.D. 12, 72, 80, 182, 1118, 1265
Harnmerlin, G. 344
Hammour, M.L., s e e Caballero, R.J. 846, 847,
852, 855, 856, 1157, 1158, 1160, 1187,
1210, 1211, 1213, 1472
Hannerz, U. 1332
Hansen, B. 1194
Hansen, B.E. 38, 39
Hansen, G.D. 547, 55l, 602, 976, 977, 1200
Hansen, G.D., s e e Cooley, T.E 69, 97, 101,
115, 124, 137, 380, 408, 411,974, 1736
Hansen, L.P. 547, 555, 556, 558, 572-574, 768,
769, 784, 882, 915, 1234, 1246, 1249, 1250,
126l, 1294, 1295
Hansen, L.P., s e e Anderson, E.W. 368, 369
Hansen, L.P., s e e Cochrane, J.H. 1234, 1246,
1249
Hansen, L.P., s e e Eichenbaum, M. 549, 550,
785, 799, 800, 803
Hanson, J. 1543
Hansson, B., s e e Frennberg, P. 1238
Harberger, A.C. 1554, 1590
1-12
Harden, I., s e e von Hagen, J. 1439, 1460,
1465
Hardouvelis, G.A. 1281
Hardouvelis, G.A., s e e Estrella, A. 43, 1281
Harris, R., s e e Cox, D. 705
Harrison, A. 277, 279, 280
Harrison, S.G., s e e Christiano, L.J. 426
Harrison, S.H. 402
Harrod, R. 640
Hart, O. 852, 1154
Hartwick, J. 656
Harvey, A.C. 9
Harvey, C.R. 1236, 1280
Hashimoto, M. l 152
Hassett, K.A. 815, 818, 843, 1344
Hassett, K.A., s e e Auerbach, A.J. 821
Hassett, K.A., s e e Cummins, J.G. 822, 856,
1344
Hassett, K.A., s e e Fallick, B.C. 855
Hassler, J. 9, 1238
Haug, A.A., s e e Dezhbakhsh, H. 1039
Haugen, R.A., s e e Ferris, S.P. 1314
Hause, J.C. 569
Hausman, J. 620
Hausman, J., s e e Burtless, G. 620
Hawley, C.B., s e e O'Brien, A.M. 776
Hayashi, E 773, 775, 776, 785, 788, 790, 796,
800, 818, 1649
Head, A., s e e Devereux, M.B. 1126
Heal, G., s e e Dasgupta, E 655, 656
Heal, G.M., s e e Ryder Jl, H.E. 1284
Heaton, J. 380, 547, 569, 803, 1242, 1255,
1284, 1293
Heckman, J.J. 576, 578, 579, 582, 584-587,
590, 592, 593, 595, 601-603, 605, 615-617,
620-624, 752, 759, 1166
Heckman, J.J., s e e Ashenfelter, O. 618
Heckman, J.J., s e e Cameron, S. 589
Heckman, J.J., s e e Cossa, R. 584
Heckman, J.J., s e e Killingsworth, M.R. 550,
601, 1148
Heijdra, B.J. 1119, 1120, 1126
Heinemann, M. 495, 525
Hellwig, M., s e e Gale, D. 1376
Helpman, E. 203, 1580
Helpman, E., s e e Coe, D.T. 265
Helpman, E., s e e Drazen, A. 1580
Helpman, E., s e e Grossman, G.M. 264, 639,
672, 715, 1210, 1464
Hendershott, RH. 1333
Henderson, D.W. 1497
Author
Index
Author
Index
1-13
Inman, R., s e e Bohn, H. 1465
Intriligator, M., s e e Griliches, Z. 541
Ireland, P.N. 129, 194, 1036, 1492, 1494,
1497
Irish, M., s e e Browning, M. 611, 612, 752,
787, 792
Irons, J., s e e Faust, J. 1416, 1425
Irwin, D.A. 178
Isard, E, s e e Flood, R.P. 158, 1429, 1438
Islam, N. 283-285, 287, 653
Ito, T. 1425
Iwata, S., s e e Hess, G.D. 9
Jackman, R. 1221
Jackxnan, R., s e e Layard, R. 1098, 1176, 1177,
1221
Jackwerth, J.C. 1310
Jaeger, A., s e e Harvey, A.C. 9
Jaffee, D.M. 1376
Jagannathan, R., s e e Glosten, L. 1280
Jagannathan, R., s e e Hansen, L.E 547, 1234,
1246, 1249
James, H., s e e Bernanke, B.S. 183, 184
James, W. 1330
Janis, I. 1332
Jappelli, T. 776, 780, 790, 1344
Jappelli, T., s e e Guiso, L. 772
Jeanne, O. 156, 1041
Jeanne, O., s e e Bensaid, B. 1446, 1449
Jefferson, P.N. 1485, 1509
Jegadeesh, N. 1321
Jegadeesh, N., s e e Chan, L. 1321
Jensen, H. 1415, 1427
Jensen, H., s e e Beetsma, R. 1436, 1438
Jensen, M. 1344
Jeon, B.N., s e e von Fttrstenberg, G.M. 1333
Jermann, U.J. 1296
Jermalm, U.J., s e e Alvarez, E 575
Jermann, U.J., s e e Baxter, M. 980, 992
Jewitt, I., s e e Buiter, W. 1030
Jimeno, J.E, s e e Blanchard, O.J. 1214
Jimeno, J.E, s e e Dolado, J.J. 1214
John, A., s e e Cooper, R. 398
Johnson, H.G. 702, 704, 705
Johnson, E, s e e Goodman, A. 797
Johnson, P.A., s e e Durlauf, S.N. 254, 263, 264,
270, 271,289, 303
Johnson, EG., s e e Banks, J. 751
Johnson, S.A. 345, 381
Jones, C.I. 237, 264, 290, 292, 672, 696,
714-716, 718, 719
1-14
Jones, C.I., s e e Hall, R.E. 673, 679, 680,
683-686, 702, 856
Jones, L.E. 245, 257, 261, 380, 672, 709,
711-713, 720, 1675, 1711
Jones, L.E., s e e Chaff, V.V. 715, 1578
Jones, M. 1540
Jonsson, G. 1404, 1411, 1415, 1426, 1438
Jonung, L. 159, 1485
Jonung, L., s e e Bordo, M.D. 152, 215, 217,
220, 221
Jonung, L., s e e Fregert, K. 1016
Jorda, O. 881
Jorgenson, D. 664
Jorgenson, D.W. 817
Jorgenson, D.W., s e e Christensen, L.R. 673,
688
Jorgenson, D.W., s e e Hall, R.E. 817
Jorion, P., s e e Goetzmann, W.N. 1242, 1252,
1320
Jovanovic, B. 702, 848, 1200
Jovanovic, B., s e e Greenwood, J. 664, 692
Judd, J.P. 1485, 1487, 1512, 1516
Judd, K. 590, 1652
Judd, K., s e e Bizer, D. 380
Judd, K.J., s e e Gaspar, J. 324, 369
Judd, K.L. 314, 324, 340, 343, 347, 348, 350,
354, 1673, 1675, 1694
Judson, R. 663
Judson, R., s e e Porter, R. 1509
Juhn, C. 569, 619
Jun, B. 474
Juster, ET. 777
Juster, T., s e e Barsky, R. 558, 564, 565
Kaas, L., s e e B6hm, V 646
Kafka, A. 1543
Kahaner, D. 329, 333
Kahn, C. 1154
Kahn, C.M., s e e Blanchard, O.J. 391,504
Kahn, J., s e e Crucini, M.J. 178, 705
Kahn, J.A. 897, 910
Kahn, J.A., s e e Bils, M. 910, 912, 1053, 1078,
1079, 1085
Kahneman, D. 1308, 1309, 1311
Kahneman, D., s e e Thaler, R.H. 1313
Kahneman, D., s e e Tversky, A. 1308, 1315,
1319, 1330
Kalaba, R., s e e Bellman, R. 340
Kaldor, N. 237, 238, 240, 640, 941
Kalecki, M. 1054
Kallick, M. 1325
Author
Index
Kamihigashi, T. 428
Kaminsky, G.L. 1550, 1553, 1590
Kandel, S. 1235, 1252, 1253, 1265, 1270,
1272
Kandoff, M. 475
K a n e , A . , s e e Chou, R.Y. 1236, 1280
Kaniovski, Y.M., s e e Arthur, W.B. 476
Kaplan, S.N. 856, 1344
Karatzas, I., s e e E1 Karoui, N. ' 835
Karras, G., s e e Cecchetti, S.G. 217
Kashyap, A.K. 137, 877, 88l, 886, 903, 906,
912, 1018, 1344, 1374, 1376
Kashyap, A.K., s e e Cecchetti, S.G. 876
Kashyap, A.K., s e e Hoshi, T. 1344
Kashyap, A.K., s e e Hubbard, R.G. 1344
Katz, L. 577, 578
Katz, L., s e e Autor, D. 577
Katz, L.E, s e e Abraham, K.J. 1183, 1221
Katz, L.E, s e e Cutler, D.M. 797
Katz, L.W., s e e Blmlchard, O.J. 1176
Kaufman, H. 1344
Keane, M.P. 608, 609, 786, 790
Keefer, P., s e e Knack, S. 1466, 1471
Kehoe, EJ., s e e Atkeson, A. 847, 1675, 1718,
1720
Kehoe, P.J., s e e Backus, C.K. 549
Kehoe, P.J., s e e Baekus, D.K. 9, 42, 45, 938,
1708
Kehoe, P.J., s e e Chaff, V.V. 124, 397, 422, 672,
697, 698, 700, 701, 709, 720, 722, 723,
974, 1036, 1037, 1040-1042, 1371, 1448,
1449, 1488, 1489, 1673-1676, 1691, 1699,
1708-1710, 1720, 1723
Kehoe, P.J., s e e Cole, H.L. 1449
Kehoe, T.J. 314, 380, 389, 391,574, 575
Kehoe, T.J., s e e Cole, H.L. 1446, 1449, 1603
Kehrer, K.C., s e e Moffitt, R.A. 618
Kelly, M. 271
Kemmerer, E.W. 173
Kendrick, D.A., s e e Amman, H.M. 535
Kenen, EB. 165, 1496
Kennan, J. 803
Kessler, D. 1646
Keynes, J.M. 158, 161, 1055, 1059, 1537
Kiefer, J. 476
Kiefer, N.M., s e e Burdett, K. 1173
Kiefer, N.M., s e e Devine, T.J. 1166
Kiguel, M. 1535, 1543, 1546, 1554, 1555
Kihlstrom, R.E. 563
Kiley, M.T. 422, 423, 1041, 1117, 1129
Killian, L. 87
Author
Index
1-15
Kocherlakota, N., s e e Ingram, B. 984
Kocherlakota, N.R. 271,673, 694
Kochin, L., s e e Benjamin, D. 161
Kollintzas, T. 904-907
Kollman, R. 1085
Kon-Ya, E, s e e Shiller, R.J. 1316
Konings, J., s e e Garibaldi, P. 1180, 1222
Koopmans, T. 931,942, 948
Koopmans, T.C. 244, 246, 247, 295, 643, 649,
1673
Koopmans, T.J. 9
Kormendi, R.C. 278-281,671, 1656, 1657
Komai, J. 703
Kortum, S., s e e Eaton, J. 719
Kosobud, R., s e e Klein, L. 941
Kosters, M.H. 618
Kotkin, B., s e e Bellman, R. 340
Kotlikoff, L. 1448, 1449, 1465
Kotlikoff, L., s e e Hayashi, E 796
Kotlikoff, L.J. 780, 1624, 1646
Kotlikoff, L.J., s e e Auerbach, A.J. 380, 549,
576, 588, 590, 591, 593, 616, 1624, 1634,
1635, 1639, 1652, 1718
Kotlikoff, L.J., s e e Gokhale, J. 750
Koyck, L.M. 816
Kramer, C., s e e Flood, R.R 1596
Krane, S.D. 876, 877
Kxemer, M., s e e Blanchard, O.J. 852
Kremer, M., s e e Easterly, W. 277, 278, 281,
675
Kreps, D.M. 540, 557, 1256
Kreps, D.M., s e e Bray, M. 474
Kreps, D.M., s e e Fudenberg, D. 475
Krieger, S. 380, 843, 847
Krishnamurthy, A. 1376, 1378
Kroner, K.E, s e e Bollerslev, T. 1236, 1280
Krueger, A., s e e Autor, D. 577
Krueger, A.O. 673, 679, 699
Krueger, J.T. 104, 105
Krugman, P. 1215, 1536, 1590, 1592, 1594,
1596, 1601, 1605, 1606, 1632
Krusell, E 380, 547, 566, 567, 994, 1293,
1445, 1473
Krusell, P., s e e Greenwood, J. 664
Kuan, C.-M. 476
Kugler, P. 1281
K u h , E . , s e e Meyer, J.R. 817
Kumhof, M. 1596
Kurz, M. 474
Kushner, H. 476
Kushner, H.J. 476
1-16
Kusko, A.L. 1327
Kuttner, K., s e e Evans, C.L. 105
Kuttner, K.N., s e e Friedman, B.M. 43, 44
Kuttner, K.N., s e e Krueger, J.T. 104, 105
Kuznets, S. 941
Kuznets, S., s e e Friedman, M. 572
Kwiatkowski, D. 212
Kydland, EE. 9, 42, 158, 428, 547, 549, 578,
929, 953, 956, 957, 962, 980, 981, 1058,
1059, 1140, 1141, 1145, 1167, 1195, 1400,
1405, 1415, 1449, 1485, 1486, 1488, 1557,
1561, 1673, 1708
Kydland, EE., s e e Backus, C.K, 549
Kydland, EE., s e e Backus, D,K, 1708
Kydland, EE., s e e Bordo, M.D, 158, 160, 185,
215, 1438
Kydland, EE., s e e Hotz, V.J. 792, 803
Kyle, A,S., s e e Campbell, J.Y. 1290
La Porta, R. 1240, 1320
Labadie, P., s e e Giovannini, A. 380
Labadie, P.A., s e e Coleman II, W.J. 114
Lach, S. 1019
Ladron de Guevara, A. 317
Laffont, J., s e e Gourieroux, C. 487
Laffont, J.J., s e e Kihlstrom, R.E. 563
Laffont, J.-J, 538
Lahiri, A. 1539, 1571, 1578, 1579, 1597
Lai, K.S. 876
Laibson, D. 1653
Laidler, D. 1485
Lakonishok, J. 1323
Lakonishok, J., s e e Chan, L. 1321
Lam, P.-S., s e e Cecehetti, S.G. 1251, 1265,
1270, 1272, 1294, 1296
Lam, P.S. 802
Lambert, J.D. 346
Lambertini, L. 1457, 1465
Lamo, A.R. 290
Lamont, O.A., s e e Kashyap, A.K. 881, 912,
1344, 1374
Landi, L., s e e Barucci, E. 525
Lane, P. 1472
Langer, E.J. 1329
Lansing, K . , s e e Guo, J.-T. 416
Lapham, B.J., s e e Devereux, M.B. 1126
Laroque, G., s e e Fuchs, G. 464, 474
Laroque, G., s e e Grandmont, J.-M. 464, 474,
475, 481,507
Laroque, G., s e e Grossman, S.J. 801
Lau, L. 664
Author
Index
Author
Index
1-17
Loury, G.C. 299
Lovell, M.C. 881,893,908, 910
Lown, C., s e e Bernanke, B.S. 1343
Lucas, DA. 1035, 1036, 1042
Lucas, D.J., s e e Heaton, J. 380, 547, 569, 1255,
1293
Lucas, R. 398, 424, 425, 641, 651,929, 932,
953
Lucas, R.E. 46, 50, 380, 1158, 1446, 1449
Lucas, R.E., s e e Stokey, N.L. 314, 318-321,
346, 951,998, 999
Lucas J1, R.E. 67, 88, 158, 238, 245, 264, 265,
293, 454, 457, 463, 474, 545, 547, 554,
559, 561, 575, 578, 582, 583, 615, 616,
672, 710-715, 720, 797, 102~1024, 1043,
1195, 1268, 1489, 1490, 1495, 1500, 1592,
1673, 1675, 1699, 1711, 1723, 1728
Lucas Jr, R.E., s e e Atkeson, A. 575
Lucas Jr, R.E., s e e Stokey, N.L. 271,299
Ludvigson, S. 785, 788, 1344, 1652
Lundvik, E, s e e Hassler, J. 9, 1238
Lusardi, A. 608, 790, 791
Lusardi, A., s e e Browning, M. 606, 771
Lusardi, A., s e e Garcia, R. 790
Luttmer, E.G.J. 575
Lyons, R.K., s e e Caballero, R.J. 399
1-18
Maddison, A. 288, 673-675, 677, 678, 720,
721
Madison, J. 1659
Mailath, G.J., s e e Kandori, M. 475
Makhija, A.K., s e e Ferffs, S.E 1314
Malcomson, J.M., s e e MacLeod, W.B. 1157,
1186
Malinvaud, E., s e e Blanchard, O.J. 1214
Malkiel, B. 1316
Mankiw, N.G. 135, 158, 159, 173, 216, 244246, 252-255, 269-271,277-279, 289, 397,
567, 653, 655, 660, 673, 679-686, 694,
749, 785, 790, 800, 961, 1281, 1290, 1292,
1638, 1702, 1742
Mankiw, N.G., s e e Abel, A.B. 1266, 1651
Mankiw, N.G., s e e Ball, L. 42, 1023, 1632,
1650, 1651
Mankiw, N.G., s e e Barro, R.J. 1637
Mankiw, N.G., s e e Barsky, R.B. 1653
Mankiw, N.G., s e e Campbell, J.Y. 769, 784,
1261, 1264, 1290, 1655
Mankiw, N.G., s e e Elmendorf, D.W. 1439
Mankiw, N.G., s e e Hall, R.E. 1485, 1493,
1498
Mankiw, N.G., s e e Kimball, M.S. 1653
Mann, C.L., s e e Bryant, R.C. 1043, 1491,
1497, 1516-1518
Manuelli, R.E., s e e Chari, VV. 715, 1578
Manuelli, R.E., s e e Jones, L.E. 245, 257, 261,
380, 672, 709, 711-713, 720, 1675, 1711
Man, C.S., s e e Dotsey, M. 370, 952
Marcet, A. 314, 326, 348, 351,454, 455, 464,
465, 468, 473-476, 480, 494, 499, 525,
528-530, 532, 1675, 1705, 1707
Marcet, A., s e e Canova, E 283
Marcet, A., s e e den Haan, W.J. 347, 354, 369
Margarita, S., s e e Beltratti, A. 524, 525
Margaritis, D. 474
Mariano, R.S., s e e Seater, J.J. 1656, 1657
Mariger, R.R 1344
Marimon, R. 455, 464, 472, 475, 523, 531,
1214
Marimon, R., s e e Evans, G.W. 483, 509, 527,
528, 531
Marion, N., s e e Flood, R.P. 1429, 1438
Mark, N.C., s e e Cecchetti, S.G. 1251, 1265,
1270, 1272, 1294, 1296
Marris, S. 1632
Marschak, J. 582, 1043
Marshall, A. 203
Marshall, D.A., s e e Bekaert, G. 1281
Author
Index
Author
Index
1-19
Mills, T.C., s e e Capie, E 163, 1438
Mincer, J. 581,592, 684
Minehart, D., s e e Bowman, D. 1313
Mirman, L., s e e Levhari, D. 1450, 1465
Mirman, L.J., s e e Brock, W.A. 319, 547, 552,
556, 942, 951
Miron, J.A. 173, 216, 876, 907, 1242
Miron, J.A., s e e Barsky, R.B. 1149
Miron, J.A., s e e Beaulieu, J.J. 876
Miron, J.A., s e e Feenberg, D. 60
Miron, J.A., s e e Mankiw, N.G. 173, 216,
1281
Mirrlees, J.A. 1154
Mirrlees, J.A., s e e Diamond, P.A. 1684
Mishkin, ES. 101, 183,216, 1023, 1380, 1432,
1438
Mishkin, ES., s e e Bernanke, B.S. 1495
Mishkin, ES., s e e Estrella, A. 1485
Mishkin, ES., s e e Hall, R.E. 607, 608, 789,
1655
Mishra, D. 1416, 1425
Missale, A. 1450
Mitchell, B.R. 222
Mitchell, W.C. 8, 44, 1053
Mitchell, W.C., s e e Burns, A.E 5, 8, 931,934
Mitra, K. 530, 532
Mnookin, R.H., s e e Gilson, R.J. 1154
Modiano, E.M. 1543
Modigliani, E 761,762, 780, 1321, 1343, 1646,
1656, 1657
Modigliani, E, s e e Dreze, J. 770
Modigliani, E, s e e Holt, C.C. 882, 885, 888,
909, 910, 912
Modigliani, E, s e e Jappelli, T. 780
Modigliani, E, s e e Samuelson, P.A. 643
Moffitt, R. 752, 787
Moffitt, R.A. 618
Moler, C., s e e Kahanel, D. 329, 333
Mondino, G. 1540
Monfort, A., s e e Gourieroux, C. 487
Monro, S., s e e Robbins, H. 476, 478
Montgomery, E. 1017, 1018
Montiel, P. 1539
Montiel, P., s e e Ag6nor, ER. 1543
Montrucchio, L. 330
Montrucchio, L., s e e Boldrin, M. 362
Moore, B.J. 455, 475, 496
Moore, G.H. 1059
Moore, G.H., s e e Zarnowitz, V. 40
Moore, G.R., s e e Fuhrer, J.C. 905, 908, 1039,
1040, 1518
1-20
Moore, J., s e e Kiyotaki, N. 852, 857, 1353,
1356, 1376, 1378, 1379
Moreno, D. 481
Morgan, D. 1374
Morrison, C.J. 1086
Mortensen, D.T. 1157, 1158, 1162, 1163, 1173,
1182, 1183, 1187, 1188, 1194, 1198, 1203,
1208, 1217, 1220, 1222
Mortensen, D.T., s e e Burdett, K. 1173, 1196
Mortensen, D.T., s e e Millard, S.P. 1217, 1220
Morton, T.E. 338
Mosser, RC. 910
Motley, B., s e e Judd, J.P. 1485, 1487, 1512,
1516
Mroz, T.A. 618
Mroz, T.A., s e e MaCurdy, T.E. 592, 752
Muellbauer, J., s e e Deaton, A. 783
Mueller, D. 1464
Mulligan, C.B. 346, 1150
Mundell, R.A. 1496
Murphy, K. 581
Murphy, K., s e e Juhn, C. 569, 619
Murphy, K., s e e Katz, L. 577, 578
Murphy, K.M. 262, 278, 1082
Murray, C.J., s e e Nelson, C.R. 11
Murray, W., s e e Gill, EE. 329
Musgrave, R.A. 1631, 1661
Mussa, M. 208, 1404, 1637
Mussa, M., s e e Flood, R.R 152, 202, 1428
Mussa, M.L., s e e Frenkel, J.A. 203
Muth, J.E 457, 473, 484
Muth, J.E, s e e Holt, C.C. 882, 885, 888, 909,
910, 912
Myerson, R. 1459
Nakamura, A. 618
Nakamura, M., s e e Nakamura, A. 618
Nalebuff, B., s e e Bliss, C. 1461, 1465
Nance, D.R. 1318
Nankervis, J.C., s e e McManus, D.A. 908
Nash, S., s e e Kahaner, D. 329, 333
Nason, J.M., s e e Cogley, T. 395, 547, 967,
1142, 1503
Natanson, I.P. 342
NBER 8
Neale, M.A., s e e Northcraft, G.B. 1315
Negishi, T. 559
Nelson, C.R. 11, 211, 213, 264, 969, 1264,
1320
Nelson, C.R., s e e Beveridge, S. 1062, 1143
Nelson, D.B. 182
Author
Index
Nelson, E. 1035
Nerlove, M. 283, 284
Neumann, G.R., s e e Burdett, K. 1173
Neusser, K. 941
Neves, J., s e e Correia, I. 974
Neves, P., s e e BlundeU, R. 792
Ng, S., s e e Garcia, R. 790
Nickell, S., s e e Layard, R. 1098, 1176, 1177,
1221
Niekell, S.J. 823
Nicolini, J.P., s e e Marcet, A. 455, 530, 532
Niederreiter, H. 334
Nilsen, O.A., s e e Askildsen, J.E. 1074
Nishimura, K., s e e Benhabib, J. 403-405, 425,
435
Nordhaus, W. 1400, 1425
North, D. 1449
Northcraft, G.B. 1315
Novales, A. 803
Nurkse, R. 163, 203
Nyarko, Y. 465, 474
O'Barr, W.M. 1332
O'Brien, A.M. 776
O'Brien, A.E 181
Obstfeld, M. 159, 164, 165, 407, 1411, 1415,
1429, 1438, 1449, 1507, 1571, 1588, 1590,
1592, 1630
Obstfeld, M., s e e Froot, K. 1266
O'Connell, S.A. 1650
Odean, T. 1314, 1323
O'Driseoll, G.P. 1643
OECD 1181, 1182, 1215, 1620
Office of Management and Budget 1622
Officer, L. 155
Ogaki, M., s e e Atkeson, A. 610, 786
Ohanian, L.E. 1036
Ohanian, L.E., s e e Cooley, T.E 42, 962, 974
O'Hara, M., s e e Blume, L.E. 321,322
Ohlsson, H., s e e Edin, D.A. 1457
Okina, K. 1508
Okun, A.M. 1014, 1541
Oliner, S.D. 137, 820, 1374, 1376
Oliner, S.D., s e e Cummins, J.G. 856
Olsder, G., s e e Basar, T. 1449
Olshen, R.A., s e e Breiman, L. 289
Oppers, S. 154
Orphanides, A. 198, 1485
Ortega, E., s e e Canova, E 376, 377, 379
Ortigueira, S., s e e Ladron de Guevara, A. 317
Ostry, J. 1568
Author
Index
1-21
1456, 1459, 1460, 1465, 1466, 1469, 1470,
1490
Persson, T., s e e Englund, E 9
Persson, T., s e e Hassler, J. 9, 1238
Persson, T., s e e Horn, H. 1415
Persson, T., s e e Kottikoff, L. 1448, 1449,
1465
Persson, T., s e e Persson, M. 1447, 1449
Pesaran, H. 487
Pesaran, M.H., s e e Binder, M. 271
Pesaran, M.H, s e e Im, K. 283
Pesaran, M.H., s e e Lee, K. 284
Pestieau, EM. 1718
Petersen, B.C., s e e Carpenter, R.E. 881,912,
1344
Petersen, B.C., s e e Domowitz, I. 1020, 1083,
1093
Petersen, B.C., s e e Fazzari, S.M. 818, 1344
Petterson, E 1457
Pflug, G., s e e Ljung, L. 476
Phaneuf, L. 1028, 1039, 1041
Phelan, C. 380, 575, 796
Phelan, C., s e e Atkeson, A. 1298
Phelps, E. 944, 1025, 1026, 1039
Phelps, E.E., s e e Frydman, R. 453, 454, 474,
528, 536, 539
Phelps, E.S. 46, 168, 1059, 1098, 1121, 1122,
1157, 1173, 1176, 1192, 1220, 1537, 1538,
1720, 1724
Philippopoulus, A., s e e Lockwood, B. 1415
Phillips, A.W. 1510
Phillips, A.W.H. 46
Phillips, L.D., s e e Lichtenstein, S. 1318
Phillips, EC.B., s e e Kwiatkowski, D. 212
Pieard, P. 1157
Pieper, EJ., s e e Eisner, R. 1621
Pierce, J.L. 195
Piketty, T., s e e Aghion, E 1377
Pindyck, R. 1072
Pindyek, R.S. 835, 910, 912
Pindyck, R.S., s e e Abel, A.B. 835
Pindyck, R.S., s e e Caballero, R.J. 844
Pippenger, J. 156
Pischke, J.-S., s e e Jappelli, T. 790
Pischke, J.-S. 764
Pissarides, C.A. 774, 1163, 1173, 1183, 1184,
1188, 1193, 1194, 1200, 1203, 1207, 1209,
1220
Pissarides, C.A., s e e Garibaldi, P. 1180, t222
Pissarides, C.A., s e e Jacl~anan,R. 1221
1-22
Pissarides, C.A., s e e Mortensen, D.T. 1158,
1182, 1183, 1194, 1198, 1203, 1208
Plosser, C.I. 952, 954, 958, 961, 963, 1094,
1658
Plosser, C.I., s e e King, R.G. 9, 54, 369, 391,
429, 435, 549, 929, 931, 941, 945, 954,
995
Plosser, C.I., s e e Long, J. 929, 952, 953, 994
Plosser, C.I., s e e Nelson, C.R. 11,211, 213,
264, 969
Plutarchos, S., s e e Benhabib, J. 437
Polemarchakis, H.M., s e e Geanakoplos, J.D.
395, 458
Policano, A., s e e Fethke, G. 1037
Pollak, R.A. 803
Pollard, S. 161
Poole, W. 192, 1514, 1515
Poonia, G.S., s e e Dezhbakhsh, H. 1039
Popper, K. 376
Porter, R. 1509
Porter, R.D., s e e LeRoy, S.F. 1235, 1319
Porteus, E.L., s e e Kreps, D.M. 557, 1256
Portier, E 1068, 1126
Portier, E, s e e Hairault, J.-O. 1036
Posen, A. 1404, 1426, 1432, 1438
Posen, A., s e e Mishkin, ES. 1432, 1438
Poterba, J.M. 159, 1235, 1320, 1465, 1648,
1655
Poterba, J.M., s e e Cutler, D.M. 1290, 1320,
1321
Poterba, J.M., s e e Feldstein, M. 1633
Poterba, J.M., s e e Kusko, A.L. 1327
Power, L., s e e Cooper, R. 824
Pradel, J., s e e Fotu'geaud, C. 454, 465, 473,
475
Praschnik, J., s e e Hornstein, A. 549
Prati, A. 162
Prati, A., s e e Alesina, A. 1446, 1449
Prati, A., s e e Drudi, E 1450
Prescott, E.C. 178, 365, 545, 675, 700, 702,
930, 934, 952, 954, 956, 957, 961, 963,
982, 1033, 1296, 1488, 1489, 1710
Prescott, E.C., s e e Chari, V.V 1488, 1489,
1674
Prescott, E.C., s e e Cooley, T.F. 376, 549, 954
Prescott, E.C., s e e Hansen, G.D. 602
Prescott, E.C., s e e Hodrick, R. 9, 12, 34, 428,
931,932
Prescott, E.C., s e e Kydland, EE. 9, 42, 158,
428, 547, 549, 929, 953, 956, 957, 962,
980, 981, 1058, 1059, 1140, 1141, 1145,
Author
Index
Author
1-23
Index
1-24
Rogoff, K., s e e Obstfeld, M. 407, 1507, 1590,
1630
Rojas-Suarez, L. 1575
Roland, G., s e e Persson, T. 1460
Roldos, J. 1578
Roll, R. i328
Romer, C.D. 6, 69, 92, 137, 183, 187, 204,
205, 1618
Romer, D. 237, 643, 649, 651,661,930, 1013,
1034, 1140, 1157, 1163, 1635, 1661
Romer, D., s e e Ball, L. 1023, 1037, 1041,
1127
Romer, D., s e e Frankel, J.A. 280, 281
Romer, D., s e e Mankiw, N.G. 244-246, 252255, 269-271,277-279, 289, 653, 655, 660,
673, 679-683, 685, 686, 1638
Romer, D.H., s e e Romer, C.D. 69, 92, 137
Romer, EM. 238, 245, 260, 261,264, 265, 271,
278, 280, 398, 424, 425, 641, 651, 665,
672, 705-707, 715-717, 719, t638
Romer, EM., s e e Evans, G.W. 425, 426, 506,
521
Rose, A., s e e Akerlof, G.A. 1200
Rose, A.K., s e e Eichengreen, B. 1590
Rose, A.K., s e e Frankel, J.A. 1590
Rosen, A., s e e Meehl, E 1319
Rosen, S. 584, 5 8 5 , 976
Rosensweig, J.A. 1659
Rosenthal, H., s e e Alesina, A. 1425, 1426
Roseveare, D. 1626
Ross, L. 1319
Ross, S., s e e Brown, S. 1242
Ross, S.A. 1331
Rossmla, R.J. 879, 881,886, 907
Rossana, R.J., s e e Maccini, L.J. 881, 893, 894,
903, 907
Rossi, EE., s e e Jones, L.E. 380, 672, 711-713,
1675, 1711
Rotemberg, J.J. 67, 68, 395, 397, 406, 407,
423, 429, 434, 838, 910, 974, 996, 1020,
1033, 1034, 1036, 1040, 1041, 1043, 1044,
1055, 1056, 1058, 1062, 1063, 1067-1069,
1074, 1081, 1082, 1088-1090, 1092, 1093,
1106, 1t07, 1114, 1116, 1118, 1123-1125,
1129, 1143, 1144, 1365, 1464, 1492, 1494,
1497
Rotemberg, J,J., s e e Mankiw, N.G. 785
Rotemberg, J.J., s e e Pindyck, R. 1072
Rotemberg, J.J., s e e Poterba, J.M. 159
Rothschild, M. 823
Rotwein, E. 1011
Author
Index
Author
1-25
Index
1-26
Schwert, G.W. 1236, 1280
Schwert, G.W., s e e French, K. 1280
Seater, I I 1621, 1654, 1656, 1657
Sedlacek, G., s e e Heckman, I I 578, 579
Sedlacek, G.J., s e e Hotz, VJ. 792, 803
Segal, I.B. 1157
Senhadji, A.S., s e e Diebold, EX. 11
Sentana, E., s e e King, M. 1333
Seppala, J., s e e Marcet, A. 1675, 1705, 1707
Seslnick, D. 746, 751
Shafir, E. 1316, 1324, 1329
Shafir, E., s e e Tversky, A. 1324
Shapiro, C. 1157
Shapiro, C., s e e Farrell, J. 1121
Shapiro, M. 938, 980
Shapiro, M., s e e Barsky, R. 558, 564, 565
Shapiro, M.D. 138, 818, 1069, 1075, 1655
Shapiro, M.D., s e e Dominguez, K. 182
Shapiro, M.D., s e e Mankiw, N.G. 135
Shapiro, M.D., s e e Ramey, VA. 67, 1089
Sharma, S., s e e Masson, P.R. 1554, 1588
Sharpe, S. 1344
Shaw, E.S., s e e Gurley, J.G. 1507
Shaw, K. 584
Shay, R.P., s e e Juster, ET. 777
Shea, J. 402, 608, 790, 983, 1117
Sheffrin, S.M., s e e Driskill, R.A. 1042
Shefrin, H. 1313, 1317, 1321, 1330
Shell, K. 389, 391,516
Shell, K., s e e Balasko, Y. 427
Shell, K., s e e Barnett, W. 540
Shell, K., s e e Cass, D. 389, 516, 662
Shepard, A., s e e Borenstein, S. 1124
Sherali, D.H., s e e Bazaraa, M.S. 331
Sheshinski, E. 1031, 1037
Sherry, C.M., s e e Bazaraa, M.S. 331
Shiller, R.J. 173, 1234, 1235, 1238, 1249, 1290,
1316, 1317, 1319, 1320, 1323, 1324, 1327,
1330-1332
Shiller, R.J., s e e Campbell, J.Y. 1235, 1265,
1280, 1320
Shiller, R.J., s e e Case, K.E. 1323
Shiller, R.J., s e e Grossman, S.J. 1242, 1246,
1268, 1291
Shin, M.C., s e e Puterman, M.L. 339
Shin, Y., s e e Im, K. 283
S l f m , Y . , s e e Kwiatkowski, D. 212
Shleifer, A. 1317, 1324
Shleifer, A., s e e Barberis, N. 1294, 1322
Shleifer, A., s e e Bernheim, B.D. 1646
Shleifer, A., s e e DeLong, J.B. 1290, 1324
Author
Index
Author
Index
1-27
Stephen, P., s e e Ryder, H. 587
Sterling, A., s e e Modigliani, E 1656, 1657
Stigler, G. 1018
Stigler, G.J. 1173
Stigler, S.M. 275
Stiglitz, J., s e e Dixit, A. 1115, 1121, 1126
Sfiglitz, J., s e e Greenwald, B. 857, 1122,
1377
Stiglitz, J., s e e Jaffee, D.M. 1376
Stiglitz, J.E. 1675, 1696, 1718
Stiglitz, J.E., s e e Atkinson, A.B. 1673, 1676,
1680, 1682, 1718
Stiglitz, J.E., s e e Shapiro, C. 1157
Stock, J.H. 9, 11, 39, 43, 45, 50-54, 821,
878, 919, 934, 938, 939, 1011, 1021, 1404,
1674
Stock, J.H., s e e Feldstein, M. 44, 1485, 1497,
1498
Stock, J.H., s e e King, R.G. 54, 941
Stock, J.H., s e e Staiger, D. 49, 50
Stockman, A. 1578
Stockman, A.C. 549
Stockman, A.C., s e e Baxter, M. 203, 938,
1404
Stockman, A.C., s e e Darby, M.R. 166
Stockman, A.C., s e e Gavin, W. 1485
Stockman, A.C., s e e Ohanian, L.E. 1036
Stocks, Bonds, Bills and Inflation 1639
Stockton, D.J., s e e Lebow, D.E. 215, 1016
Stoer, J. 334
Stoker, T., s e e Blundell, R. 770, 788
Stokey, N., s e e Alvarez, E 996
Stokey, N., s e e Lucas Jr, R.E. 559, 561
Stokey, N., s e e Milgrom, R 1322
Stokey, N.L. 271,299, 314, 318-321,346, 578,
583, 672, 705, 709, 711, 714, 951, 954,
998, 999, 1674
Stokey, N.L., s e e Lucas, R.E. 380, 1446, 1449
Stokey, N.L., s e e Lucas Jr, R.E. 158, 1673,
1675, 1699, 1723, 1728
Stone, C.J., s e e Breiman, L. 289
Strang, G. 82
Strongin, S. 83-85, 87, 114
Strotz, R.H. 1653
Strotz, R.H., s e e Eisner, R. 1310
Stroud, A.H. 334
Stuart, A. 1485
Stulz, R.M. 1317
Sturzenegger, F., s e e Dornbusch, R. 1543
Sturzenegger, E, s e e Guo, J.-T. 427
Sturzenegger, E, s e e Mondino, G. 1540
1-28
Suarez, J. 1378
Subrahmanyam, A., s e e Daniel, K. 1322
Sugden, R., s e e Loomes, G. 1313
Suits, D., s e e Kallick, M. 1325
Summers, L.H. 961
Summers, L-.H., s e e Abel, A.B. 1266, 1651
Summers, L.H., s e e Alesina, A. 1432
Summers, L.H., s e e Bernheim, B.D. 1646
Summers, L.H., s e e Blanchard, O.J. 416,
1635
Summers, L.H., s e e Carroll, C.D. 759, 793,
1655
Summers, L.H., s e e Clark, K.B. 602, 1173
Summers, L.H., s e e Cutler, D.M. 1290, 1320,
1321
Summers, L.H., s e e DeLong, J.B. 279, 695,
1042, 1290, 1324
Summers, L.H., s e e Easterly, W. 277, 278, 281,
675
Summers, L.H., s e e Kotlikoff, L.J. 780, 1646
Summers, L.H., s e e Mankiw, N.G. 785
Summers, L.H., s e e Poterba, J.M. 1235, 1320,
1648
Summers, R. 238, 301, 640, 673-675, 677,
680, 681,689, 720
Sun, T. 1270
Sundaram, R.K., s e e Dutta, RK. 380
Sundaresan, S.M. 1284
Sunder, S., s e e Marimon, R. 455, 472, 531
Surekha, K. 908
Sussman, O., s e e Suarez, J. 1378
Svensson, J. 1466, 1471, 1472
Svensson, L.E.O. 156, 197, 417, 1033, 1034,
1273, 1411, 1432, 1434, 1489, 1493, 1494,
1498, 1504
Svensson, L.E.O., s e e Englund, E 9
Svensson, L.E.O., s e e Kotlikoff, L. 1448, 1449,
1465
Svensson, L.E.O., s e e Leiderman, L. 1432,
1438, 1495
Svensson, L.E.O., s e e Persson, M. 1447, 1449
Svensson, L.E.O., s e e Persson, T. 1449, 1450,
1454, 1456, 1465
Swagel, P., s e e Alesina, A. 277-279, 1460,
1466, 1471
Swan, T.W. 244, 246, 247, 643
Sweeney, J., s e e Kneese, A. 656
Swoboda, A., s e e Genberg, H. 165
Symansky, S.A., s e e Bryant, R.C. 1491, 1497,
1516
Szafarz, A., s e e Adam, M. 500
Author
Szafarz, A.,
see
Index
Author
Index
1-29
Townsend, R.M., s e e Phelan, C. 380, 575,
796
Traub, J.E 338
Traub, J.E, s e e Papageorgiou, A. 334
Trehan, B. 159
Tria, G., s e e Felli, E. 1083, 1122
Triffin, R. 157, 165
Trostel, RA. 1652
Tryon, R., s e e Brayton, E 1043, 1344, 1485
Tsiddon, D. 1031
Tsiddon, D., s e e Lach, S. 1019
Tsitsiklis, J.N., s e e Chow, C.-S. 326, 334
Tsutsui, Y., s e e Shiller, R.J. 1316
Tullio, G. 156
Tullio, G., s e e Sommariva, A. 222
Tullock, G., s e e Grier, K.B. 253
Tuncer, B., s e e Krueger, A.O. 699
Yurnovsky, S. 474
Tversky, A. 1308, 1315, 1319, 1324, 1330
Tversky, A., s e e Kahneman, D. 1308, 1309,
1311
Tversky, A., s e e Quattrone, G.A. 1329
Tversky, A., s e e Shafir, E. 1316, 1324, 1329
Tversky, A., s e e Thaler, R.H. 1313
Tybout, J., s e e Corbo, V. 1543
Tylor, E.B. 1331
Uhlig, H. 70
Uhlig, H., s e e Lettau, M. 524, 1297
Uhlig, H., s e e Taylor, J.B. 314
United Nations 681
Uppal, R., s e e Dumas, B. 564
Uribe, M. 1539, 1578, 1589
Uribe, M., s e e Benhabib, J. 419, 421,423
Uribe, M., s e e Mendoza, E. 1571, 1579
Uribe, M., s e e Schmitt-Groh~, S. 416, 418,
431
US Bureau of the Census 585
Uzawa, H. 578, 651,710
Valdes, R., s e e Dornbusch, R. 1590
Valdivia, V, s e e Christiano, L.J. 504
Van Huyck, J.B., s e e Grossman, H.J. 158, 1415,
1449
van Wineoop, E., s e e Beaudry, E 1264
Van Zandt, T., s e e Lettau, M. 470, 472
Vasicek, O. 1270
V6gh, C., s e e Guidotti, EE. 1675, 1720
V6gh, C.A. 1535, 1538, 1542, 1543, 1546,
1550, 1554, 1588
V6gh, C.A., s e e Bordo, M.D. 158
1-30
V6gh, C.A., s e e Calvo, G.A. 1428, 1535, 1538,
1539, 1546, 1554, 1557, 1563, 1564, 1568,
1571, 1572, 1582, 1587-1589, 1597, 1605
Vdgh, C.A., s e e De Gregorio, J. 1546, 1551,
1573, 1575, 1577
V6gh, C.A., s e e Edwards, S. 1578-1580
V6gh, C.A., s e e Fischer, S. 1538, 1547, 1561
V6gh, C.A., s e e Guidotti, EE. 1537, 1588,
1603
V6gh, C.A., s e e Hoffiaaaister, A. 1561, 1589
V6gh, C.A., s e e Lahiri, A. 1597
V~gh, C.A., s e e Rebelo, S.T. 1546, 1568, 1578,
1579, 1581, 1606
V~gh, C.A., s e e Reinhart, C.M. 1545, 1546,
1551, 1553, 1561, 1572, 1573
V6gh, C.A., s e e Sahay, R. 1535
Vela, A., s e e Santaella, J. 1543
Velasco, A. 416, 1446, 1449, 1450, 1459, 1465,
1540
Velasco, A., s e e Sachs, J. 1590, 1591
Velasco, A., s e e Tommasi, M. 1540
Velasco, A., s e e Tomell, A. 1466, 1472, 1590
Venable, R., s e e Levy, D. 1014, 1015, 1019
Venegas-Martinez, E 1571
Ventura, G., s e e HuggeR, M. 380
Veracierto, M. 994
Verdier, T., s e e Saint-Paul, G. 1472
Vetterling, WT., s e e Press, WH. 329-334, 343,
348, 356, 365
Viana, L. 1543
Vickers, J. 1414, 1415
Vigo, J., s e e Santos, M.S. 321,322, 326, 327,
335
Vinals, J., s e e Goodhart, C.E.A. 1438, 1495
Vishny, R.W., s e e Barberis, N. 1294, 1322
Vishny, R.W., s e e La Porta, R. 1240
Vishny, R.W, s e e Lakonishok, J. 1323
Vishny, R.W, s e e Murphy, K.M. 262, 278,
1082
Vishny, R.W, s e e Shleifer, A. 1324
Visscher, M., s e e Prescott, E.C. 700
Vives, X. 474, 532
Vives, X., s e e Jun, B. 474
Volcker, P.A. 1630
von Furstenberg, G.M. 1333
von Hagen, J. 1439, 1460, 1465
von Hagen, J., s e e Eiehengreen, B. 1465
von Hagen, J., s e e Fratianni, M. 1431
yon Hagen, J., s e e Hallerberg, M. 1460, 1465
von Weizs/icker, C. 641,650, 657
Vredin, A.E., s e e Bergstr6m, V. 538
Author
Index
Author
Index
1-31
Wilcox, D. 1242
Wilcox, D., s e e Kusko, A.L. 1327
Wilcox, D.W 1655
Wilcox, D.W., s e e Carroll, C.D. 769, 785
Wilcox, D.W., s e e Cecchetti, S.G. 876
Wilcox, D.W, s e e Kashyap, A.K. 137, 877,
886, 903, 906, 912
Wilcox, D.W, s e e Orphanides, A. 198, 1485
Wilcox, D.W., s e e West, K.D. 908
Wildasin, D., s e e Boadway, R. 1463
Wilkinson, M. 881
Williams, J.C., s e e Brayton, E 1043, 1344,
1485
Williams, J.C., s e e Gilchrist, S. 847
Williams, J.C., s e e Wright, B.D. 347, 348
Williamson, J. 1597
Williamson, O.E. 852
Williamson, S. 1376
Willis, R., s e e Heckrnan, J.J. 602, 623
Wilson, B., s e e Saunders, A. 181
Wilson, C.A. 408
Wilson, R. 554, 796
Winter, S.G., s e e Phelps, E.S. 1121
Woglom, G. 1127
Wohar, M.E., s e e Fishe, R.P.H. 173
Wojnilower, A. 1344
Wolf, H., s e e Dornbusch, R. 1543
Wolf, H., s e e Ghosh, A.R. 202, 207, 208
Wolff, E. 664
Wolfowitz, J., s e e Kiefer, J. 476
Wolinsky, A. 1188
Wolinsky, A., s e e Binmore, K.G. 1188
Wolinsky, A., s e e Rubinstein, A. 1188
Wolman, A.L., s e e Dotsey, M. 974, 1032,
1043
Wolman, A.L., s e e King, R.G. 1036, 1041,
1043, 1364, 1367
Wolters, J., s e e Tullio, G. 156
Wong, K.-E 108
Wood, G.E., s e e Capie, E 163, 1438
Wood, G.E., s e e Mills, T.C. 204
Woodford, M. 389, 395, 406, 407, 409, 418,
421-423, 439, 454, 473-476, 481,483, 507,
516, 518, 521,662, 1036, 1157, 1507, 1509,
1518-1520, 1537, 1630, 1675, 1676, 1720,
1731
Woodford, M., s e e Bernanke, B.S. 1361, 1363
Woodford, M., s e e Boldrin, M. 506
Woodford, M., s e e Farmer, R.E. 395, 396
Woodford, M., s e e Guesnerie, R. 439, 454,
460, 465, 474, 475, 506, 511,516, 526
1-32
Woodford, M., s e e Kehoe, T.J. 380
Woodford, M., s e e Lucas Jr, R.E. 1023
Woodford, M., s e e Rotemberg, J.J. 67, 68,
395, 406, 407, 429, 434, 974, 996, 1020,
1041, 1043, 1044, 1055, 1056, 1062, 1063,
1067-1069, 1074, 1081, 1082, 1088-1090,
1092, 1093, 1106, 1107, 1118, 1123-1125,
1129, 1143, 1144, 1365, 1492, 1494, 1497
Woodford, M., s e e Santos, M.S. 1266
Woodward, P.A., s e e Baker, J.B. 1125
Wooldridge, J., s e e Bollerslev, T. 1280
Wozniakowski, H., s e e Traub, J.E 338
Wright, B.D. 347, 348
Wright, M.H., s e e Gill, RE. 329
Wright, R. 1158
Wright, R., s e e Benhabib, J. 402, 550, 1145
Wright, R., s e e Boldrin, M. 399
Wright, R., s e e Burdett, K. 1196
Wright, R., s e e Greenwood, J. 550, 995
Wright, R., s e e Hansen, G.D. 976
Wright, R., s e e Kiyotaki, N. 524
Wright, R., s e e Parente, S.L. 702
Wright, R., s e e Rogerson, R. 978
Wurzel, E., s e e Roseveare, D. 1626
Wynne, M. 974
Wyrme, M.A., s e e Huffman, G.W 437
Wyplosz, C., s e e Eichengreen, B. 168, 1590
Xie, D. 425
Xie, D., s e e Benhabib, J. 425
X i e , D . , s e e Rebelo, S.T. 952
Xu, Y. 344
Yashiv, E. 1200
Yellen, J.L., s e e Akerlof, G.A. 397, 1034, 1035,
1039, 1157, 1200
Yeo, S., s e e Davidson, J. 750
Yi, K.-M., s e e Kocherlakota, N.R. 271
Yin, G.G., s e e Kushner, H.J. 476
Yong, W., s e e Bertocchi, G. 474
Yorukoglu, M., s e e Cooley, T.E 847
Author
Index
SUBJECT INDEX
1-34
borrowing constraint 566, 575, 593, 595, 597,
598, 772, 775, 1293
see also capital market imperfections;
credit market imperfections; liquidity
constraints
Boschen-Mills index 139-142
bottlenecks 842, 843
bounded rationality 454, 464
BP (band-pass) filter 12, 933, 934
Bretton Woods 152, 153, 163-168, 188, 190,
192, 199, 202-204, 206~09, 211,213, 215,
218-220
Brownian motion 825, 845
regulated 845
bubble-free solution 1524
bubble solutions 1522
bubbles 499
explosive 499
budget deficit 1619
budget surplus 1619
buffer-stock saving 771, 1653, 1654
building permits 45
Burns-Mitchell business cycle measurement
932
business cycles 865, 927-1002, 1620, 1621,
1659
s e e also cycles; fluctuations in aggregate
activity
facts about 934, 938, 939, 956
general equilibrium models 67
in RBC model 968
measuring 932
persistence of 939
table of summary statistics 956, 957
US facts 934
USA 935-938, 956
Cagan model of inflation 497
calculation equilibrium 462
calibration 545, 550, 567, 601,614, 616
Canada 45
capacity utilization 41,427, 431,930
modeling of 980
rate of 981
steady-state rate of 984
capital 1617, 1687
broad measure 701
desired 816, 842
frictionless 832, 838
human 673, 678, 679, 681-687, 701, 710,
713, 714, 716-718, 720, 732, 734
Subject I n d e x
see also human capital
organizational 700, 701
physical 678-683, 701,710, 713, 714, 721,
732
specific 1154
stock of 1629, 1630, 1632, 1633, 1636-1638,
1648, 1652, 1656
target 820
unmeasured 701,702
vintage 702
capital accumulation 942, 1203
general equilibrium nature of 946
optimal 946
perpetual inventory method 944
capital budgeting 1623
capital controls 1588
capital imbalances, establishments' 837
capital intensities 641,644, 679, 680, 682, 685,
686
capital investment decision 1349
capital/labor substitution 856
capital market imperfections 1648, 1649
s e e also borrowing constraint; credit market
imperfections
capital taxation 1661, 1708
optimality of zero 1693
capital utilization 848
CARA utility 794
cash-in-advance constraint 397, 1722
cash-credit model 1720, 1721
"catching up with the Joneses" 1284
certainty equivalence 762
Chamley result 1698
characteristics model 578, 579, 582, 602
characterization of equilibria 487, 489
Cholesky factor 80
classification 262, 289, 303
classifier systems 465, 523
closed economy 1714
closed-form solution 769
club-convergence 660
Cobb-Douglas production function in RBC
model 944, 950
"cobweb" model 456
coefficient of relative risk aversion 1249
cohort data 78t
cohort effects 576, 577, 590-592, 617, 753,
754
cointegration 50, 750, 820, 838, 877-881,
885-887, 903, 1266
collateral 857
1-35
Subject lndex
1-36
debt neutrality 1644
debt-income ratio 1630
debt--output ratio 1619
decentralized economy 547, 575, 576, 602
decision rule 888-890
deficits 1617
nominal 1621
real 1621
demand shocks 865, 884, 889 892, 895, 898,
1055
demographic transition 658
demographic variables 793
demographics 547, 551-615, 744
and retirement behavior 758
depreciation 642, 1633
detrending and business cycle measurement
932
difference models of habit 1284
difference-stationary models 764
difference-stationary process 211,215, 1497
differential equation 472
diminishing returns 639
separately to capital and augmented labor
653
dirty floating 1587
discount factor 548, 555-557, 561,567, 588,
595, 606, 607, 609, 610, 616
disinflation, output costs of 1542
disjunction effect 1324
disparity in GDP 675
disparity in incomes 674
distribution dynamics 263, 290-295, 299
distribution of country incomes 674
distribution of relative GDP 674
dividend growth 1233, 1242, 1276
dollarization 1589
domestic debt 1595, 1601
domestic policy regime 153, 202
Dornbusch-type model 502
DSGE, s e e dynamic stochastic general
equilibrium models
durability 798, 1242
durable goods 549, 746, 799, 1550, 1552, 1573,
1575
dynamic economic models 312, 313
Dynamic New Keynesian (DNK) framework
1346
dynamic programming 834
dynamic stochastic general equilibrium (DSGE)
models 930, 1139, 1145, 1150, 1157,
1166
Subject Index
1158
1-37
Subject Index
1-38
general equilibrium 543-625, 888
generational accounting 1624
genetic algorithms 465, 521,525
Germany 45, 1631
global culture 1332, 1333
GLS 788
gold standard 153-190, 199-220
Golden Rule 1650
Gorman-Lancaster technology 800
government
budget constraint 1687, 1719
consumption 671,691,694, 1687, 1736
rate to GDP 688, 689
debt 1617, 1687
production 672, 695, 701
production of investment 699
purchases 41
purchases and markups 1120
share 692
in GDP 671,689
in investment 695, 696
in manufacturing output 696
in output 693
gradual adjustments 823
gradualism 849
Granger causality 34
Great Depression 153, 163, 175, 178, 180-184,
199, 200, 213, 1343
Great Inflation of the 1970s 153
great ratios of macroeconomics 939, 940
gross domestic product (GDP)
per capita 674
per worker 671
gross substitutes 1731
growth
cycles 9
growth accounting 678, 687, 688
growth miracles in East Asia 709
growth-rate targets 1524
maximum growth rate 677, 726, 728, 732
habit formation 798, 802, 1237, 1284
habits 564, 802
Harrod-Domar models 640
hazard rate
constant 839
effective 836
increasing 840
hedging demand 1275
Herfindahl index 824
heterogeneity 546, 547, 552, 553
Subject lndex
in firms 1366
in learning 527
in values of job matches 1152
of preferences 545, 558, 563
unobserved 779, 831
heterogeneous agents 843, 1237, 1290
heterogeneous consumers 1686
Hicks composite commodity 766
Hicksian demand decomposition in RBC model
971
hiring rate 1161
histogram 840
historical counterfactual simulations 1523
history-dependent aggregate elasticity 841
Hodrick-Prescott filter, see HP filter
hold-up problems 852
home production 402, 417, 431,702
home sector 435
homotheticity 1725, 1728, 1733
Hotelling's rule 657
HP (Hodrick-Prescott) filter 12, 932, 933
human capital 527, 546, 547, 576, 577,
583-592, 594, 639, 653, 1638, 1712
hump-shaped impulse responses 405, 436,
1374
hump-shaped profiles 755
hyperinflation (seignorage) 509, 520, 531,
1631
hysteresis and threshold effects 455, 530
i.i.d, model 1739
identification problem 75-78
global identification 76, 77
local identification 76
underidentification 76, 77
idiosyncratic risk 795, 1290
idiosyncratic shocks 840
in productivity 1183
imbalances 826
imperfect competition 665
implementability constraint 1677, 1689, 1719,
1729
implicit collusion 1123
imports 41
impulse 1140
impulse response
measure of comparative dynamics 967
to productivity in RBC model 967
impulse response functions 74, 81, 85, 86, 90,
98, 100, 102, 107, 110, 112, 133, 140, 397,
411,430, 431,880, 894
Subject Index
inaction range 832
Inada conditions 645
income distribution, cross-country 671
income elasticity 1681
income inequality 797
income processes 569, 574, 610
income tax 672
income uncertainty 1652
incomplete contracts 853, 854, 856
incomplete markets 566-576, 1742
indeterminacy 491, 494, 506, 1161, 1506,
1691
nominal 418, 1506, 1524
of price level 215, 216, 415, 417, 419, 423
real 413, 415, 416, 418, 419, 423
indicator, cyclical 1062
indivisible labor model, role in RBC model
977
industry equilibrium 888, 889
inequality 745, 795
infinite-horizon consumption program 647
inflation 42, 1534, 1536, 1630
and business cycles 939
and markups 1128
inertia 1562
level 198
persistence 166, 211,213-215
rate 1738
autocorrelation 1738, 1739
variability 207
inflation correction 1621
inflation forecast targeting 1504
inflation-indexed bonds 1271
inflation-indexed consol 1269
inflation targeting 1499, 1505
vs. price-level targeting 1497
inflation tax 1538, 1720
inflationary expectations 1281
information externality 849
information pooling 849
information set 455
informational problems 849-851,858
infrequent actions 825
instability 481,519
of interest rate pegging 514
of REE 507
institutional factors 852
instrument feasibility 1507
instrument instability 1517
instrument variable 1492, 1524
instrumental variables (IV) estimator 787
1-39
instrumental variables (IV) regression 1261
insurance 745, 795
integrated world capital market 1297
interest rate 43, 1620, 1621, 1629, 1630,
1634, 1635, 1637, 1639, 1648, 1652, 1653,
1657-1659
nominal 1524
interest rate instrument 1514
interest rate policy 1596
interest rate smoothing 1509
intermediate-goods result 1684, 1720, 1733
intermediate-goods taxation 1676
intermediate input use 1081
internal habit models 1284
international capital flows 1636-1638
International Financial Statistics (IFS) 1238
international reserves 1594
intertemporal allocation 761
intertemporal budget 555, 561,647, 661
intertemporal budget constraint 1259, 1268
intertemporal CAPM 1275
intertemporal channel 1142
intertemporal elasticity
of labor supply 1149
of substitution in leisure 1147
intertemporal marginal rate of substitution
1245
intertemporal non-separabilities 775
intertemporal optimization 745
intertemporal substitution 1055, 1150
"intervention" policy 1587
intradistribution dynamics 274, 292
intratemporal first-order conditions 775
inventories, target 894
inventories of finished goods 887
inventory fluctuations 1084
procyclical 872-882, 898, 900, 909
inventory investment 865
inventory sales ratio 871
inventory-sales relationship 867
investment 40, 641
collapse 851
competitive equilibrium 844
delays 1365
distortions 672, 695 698
empirical 1344
expected 839
frictionless 832
lumpy 822, 823
share in output 693, 699
spike 823, 824, 857
1-40
investment (cont'd)
tax incentives 843
US manufacturing 840
investment episode 823
investment-output ratio 714
irrational expectations 1237, 1293
irregular models 490, 493, 505
irreversibility cons~aint 832
irreversible investment 822, 828, 832
iso-elastic utility function 606, 607, 610
Italy 1619
Ito's lemma 825
Japan 45
Jensen's inequality 1247
job-finding rate, cyclical behavior of 1162
job loss 1151
job search 1143, 1150, 1158, 1162
job-specific capital 1152
job to job flows 1198, 1200
job-worker separations 1184
jobs
creation 846, 1150, 1158, 1161, 1173, 1176,
1178, 1185, 1201, 1219
cost 1187, 1193, 1215, 1222
creation and destruction, international
comparison 1178
destruction 846, 1150, 1158, 1160, 1166,
1173, 1176, 1178, 1185, 1197, 1201,
1219
rate 1151, 1152
flow 1197
international comparison 1180
reallocation 1222
termination 1152
cost 1193
joint production 853
joint surplus 1157
just-in-time 871
Kaldor facts about economic growth 941
Keynes, J.M. 1660
Keynesian analysis 1628
Keynesian consumption function 761
Kreps-Porteus axiomatization 744
Krugman model 1592
Kulm-Tucker multiplier 774
labor 1687
bargaining strength 1219
labor-augmentation 651
Subject Index
Subject Index
1-41
market clearing 1021-1024, 1026, 1035
expected 1021, 1024-1027
market imperfection 390, 405, 424, 426, 433
market structure 546, 553, 558, 575, 598
market tightness 1185
market work 550, 594, 601
Markov chain 1708, 1736
Markov process 1264
markup 399, 400, 406, 407, 426, 429, 431,
1053
average 1068
countercyclical 406, 1113
for France 1068
cyclical 1092
desired 1056
measurement 1058
models of variation 1055, 1112
procyclical 1113, 1128
variable 406, 407
variation in desired 1129
Marshallian demands 597
martingale 767
martingale difference sequence 487
match capital 1152
matching function 1183
matching model 1163
Maximum Principle 650
measure of financial development 691
measurement error 518, 546, 561, 572-574,
609, 616, 1242
"mechanieal" approach 1560
mechanism design 1154
Medicare 1622, 1626
men 550, 552, 607, 615, 620
mental compartments 1317
menu costs 397
microeconomic data 543-625, 745
microeconomic lumpiness 824
microfomadations 761
military purchases 1088
Mincer model 568, 569, 581,582, 584, 592
minimal state variable solutions, see MSV
solutions
mismatch 1221
mismeasurement of average inflation 1254
Modigliani-Miller theorem 1343
monetary accommodation 1539
monetary base 44, 1507, 1524
monetary economies 1720
monetary model with mixed datings 500
1-42
monetary policy 692, 695, 715, 1012, 102Zl~
1037, 1281, 1630, 1660, 1720
optimal, cyclical properties of 1736
monetary policy rule 1364
monetary policy shocks 65-145
effect 69
on exchange rates 94-96
on US domestic aggregates 91 94
on volatility 123-127
identification schemes 68-70, 1369
Bernanke-Mihov critique 115-123
Bernanke-Mihov test 119-121
empirical results 121 123
Coleman, Gilles and Labadie 114, 115
narrative approach 136 141
s e e also Romer and Romer shock
pitfalls 134-136
plausibility 100-104
assessment strategies 114-123
problems 143-145
interpretations 71 73
non-recursive approaches 127-134
output effects 1129
recursiveness assumption 78-127
s e e also recursiveness assumption
responses to 1368
monetary regimes 153, 168, 178, 202, 204,
211,216, 220
money 44, 1011-1013, 1020-1029, 1031-1033,
1035, 1036, 1040, 1041
money anchor 1588
money-based stabilization 1535, 1543, 1554,
1558, 1582
money demand 50, 598, 1603, 1736
consumption elasticity of 1725
interest elasticity of 1736
money growth rate 1738
money-in-the-ufility-ftmction model 1720,
1728
money supply 1536
money velocity 1588
s e e also M1 velocity
monopolies 695
monopolistic competition 1033-1036, 1041,
1042
monotonicity 830
Morgan Stanley Capital International (MSCI)
1238
MSV (minimal state variable) solutions 488,
493, 502
and learning 503
Subject Index
1-43
Subject Index
overparametrization 473
overreaction 1319-1322
overtaking 650
overvaluation 1563
panel data 275, 283-287, 295, 781
Pareto weights 559-564, 796
partial adjustment model 821,838
participation 574, 601, 1218
path dependence of adaptive learning dynamics
455
peacetime 1699
Penn World Table 674, 680
pent-up demand 841
perceived law of motion (PLM) 466, 472, 490,
511
perceptron 524
perfect competition 831
perfect foresight 650
perfect insulation 846
perfect-insurance hypothesis 796
periodic or chaotic dynamics 646
see also cycles
permanent-income hypothesis 749, 1641,
1662
permanent shocks 216-219
perpetual inventory method 680
persistence 870-882, 891, 893, 900, 902, 904,
1142, 1162, 1166, 1739
of business cycles, see persistence under
business cycles
of fluctuations 527
of inflation 1537
peso problem 1252
pessimism 1295
Phillips curve 46, 1056, 1363, 1542
planner's problem in RBC model 997, 1002
policy 455
affecting labor markets 672
distorting investment 695
impeding efficient production 672
policy accommodation 1538
policy function 320-381
political rights 671, 689
political stability 671,688, 692
Ponzi scheme 1650
population aging 1625, 1640
population growth 941
endogenous 639
power utility 1249
1-44
precautionary saving 744, 770, 1253, 1288,
1653
preference parameters 550, 555, 556, 558, 567,
601, 605
preferences 546-550, 552, 553, 556-558, 564,
565, 567, 572, 582, 593, 601, 604, 605,
607, 608, 610, 614, 616, 617, 623
additive 594
conditional 778
functional forms 550
Gorman polar 766, 783
heterogeneity 545, 552, 558-565, 567, 593,
594, 609, 621,623
homogeneity 553-556, 577
of representative agent in RBC model 942
quadratic 762, 770
present-value model of stock prices 1264
log-linear approximation 1265
present-value neutrality 573
price elasticity 1681
price functions 1723
price puzzle 97-100
price rules 1688
price-cost margin 1053
s e e also markup
price-dividend ratio 1265, 1266, 1276
prices 42
of machinery 696
of raw materials 1082
pricing, equilibrium 555, 602, 845
primal approach 1676
primary budget 1619
principal-agent problems 1345
principles of optimal taxation 1676
private and public saving 1629
private information 574-576, 849
production costs, non-convex 897, 911
production economy 1686
production efficiency 1684, 1735
production function 548-550, 578, 579, 581,
583-586, 588, 590, 591,594
non-Cobb-Douglas 1064
production possibilities surface 401
production smoothing 876, 877, 884, 895,
1085
production to order 887
production to stock 887
productivity 552, 553, 566, 583, 602, 1057
cyclical 938, 1094
deterministic growth of 943
general 1192, 1193
Subject lndex
growth of 942
shocks 930, 943, 965, 972
amplification of 963
modeled as first-order autoregressive process
963
persistence of (serial correlation) 952,
963
RBC model's response to 964
remeasurement of 982,
slowdown 664
profit function 830
profits 1057
cyclical 1100
projection facility (PF) 480
propagation of business cycles 865
propensity to consume 762
property rights 852, 856
proportional costs 825
proportional taxes 1687
prospect theory 1308-1313
protection of specific investments 1154
"provinces" effect 1540
proxies for capital utilization 1080
prudence 771
PSID 783
public consumption 1581
public debt 1601, 1603
public finance 1676
public saving 1629, 1641
putty-clay models 847, 848
q-theory 817
s e e also Tobin's q
average q 817, 818
"flexible q" 818
marginal q 818
fragility of 828
quadratic adjustment cost model 823, 838
Quandt Likelihood Ratio (QLR) 34
quantitative performance 1578, 1581
quantitative theory 671-673, 695-719
see also dynamic stochastic general
equilibrium models
quasi-magical thinking 1329, 1330
Ramsey allocation problem 649, 1679, 1691,
1692, 1713, 1719, 1723, 1729
Ramsey equilibrium 1678, 1688, 1723, 1729,
1732
Ramsey growth model 1651, 1652
Ramsey prices 1679
1-45
Subject Index
1-46
recursiveness assumption (cont'd)
relation with VARs 78-83
REE (rational expectations equilibria) 452
cycles 458
multiple 454, 467
reduced order limited information 529
unique 484
reflecting barriers 828
regime switching 426
regression tree 289
regular models 490
regulation barrier 832
relative price of investment to consumption
696-698, 700, 701
reluctance to invest 828, 832
renegotiation 1153, 1155
renewable/nonrenewable resources 655, 656
rental prices 588, 590, 592
of capital 1000
reorganization 1160, 1161
representative agent 556, 557, 560, 561,563,
587, 601, 838, 1249, 1259, 1268
in RBC model
altered preferences in indivisivle labor
977
altruistic links 943
preferences of 942
representative household 643
representativeness heuristic 1319, 1322, 1327
reproduction 522
research and development (R&D) 664, 672,
692, 695, 708, 709, 715-719
residence-based taxation 1715
restricted perceptions equilibrium 529
restrictions in job separation 1222
restrictions on government policy 1707
retailers 869
retirements 839
returns to scale 639
decreasing 656
increasing 652, 653, 664, 828, 830, 1066
social 460, 509, 521
Ricardian equivalence 418, 1617, 164~1659,
1661
Ricardian regime 418
Ricardo, D. 1640
risk 546, 547, 552, 554-558, 563-567, 569,
572, 575, 593, 606
risk adjustment 555, 557, 558
risk aversion 547, 552, 556-558, 564-566, 606,
771
Subject Index
risk premium 1246, 1247, 1250
risk price 1236, 1280
risk-sharing in indivisible labor version of RBC
model 977
riskfree rate puzzle 1235, 1252
robustness approach 1491, 1523
Romer and Romer shock 137-142
rule-like behavior 1487, 1522
rule-of-thumb decision procedure 524
rules 152-154, 156, 158, 160, 166, 168, 184,
200, 208, 219, 220
rules vs. discretion 1485
Rybczinski theorem 404
S u b j e c t Index
tests 611
separation rate 1151
Sharpe ratio 1249
shock absorber 1699, 1710, 1739
shock propagation 1203
shocks and accommodation 1539
shopping-time model 1720, 1732
shopping-time monetary economy 1732
short-term bonds 1280
short-term maturity debt 1603
~r-convergence 659
Sims-Zha model 128-134
empirical results 131 134
skill-biased technology shock 1215, 1216,
1218
skills 546, 547, 569, 576-579, 581,582, 584,
586-588, 590-594, 623
slow adaption 480
slow speed of adjustment 877, 894
small durables 798
small open economy 1715
small sample 820
small versus large firms 1373
smooth pasting conditions 827
Social Security 1619, 1622, 1624, 1626, 1635
Solow residual 930, 1140, 1141
as productivity measure 962
in growth accounting 962
mismeasurement 962
solvency conditions 575
specificity 851,852, 856
spectral analysis 11
SSE, s e e stationary sunspot equilibria
stability conditions 454
stabilization 1534, 1562
stabilization goals 153
stabilization time profiles 1547
stable equilibrium point 481
stable roots 393
staggered contracts model 1012, 1013, 1024,
1027, 1030, 1032, 1039
staggered price and wage setting 1012, 1013,
1027, 1030, 1031, 1033, 1035-1037, 1040
staggered price setting 397, 422, 423, 1129,
1363
staggered-prices formulation 1582
standardized employment deficit 1621
state-contingent claims 555, 602
state-contingent returns on debt 1687, 1699
state-dependent pricing 1031, 1032
state dynamics 477
1-47
state prices 1294
stationary distribution of RBC model 999
stationary sunspot equilibria (SSE) 408, 517
c-SSE 517
near deterministic solutions 520
steady states 468, 507, 525, 549-551,576, 592,
598, 639
of RBC model 944
sterilization 1595
sticky price models 503, 1113
stochastic approximation 468, 475, 476
stochastic discount factor 1234, 1245
log-normal 1246
stochastic growth model 546-577, 592
stochastic simulations 1516, 1523
stock market 1310, 1312, 1313, 1315, 1316,
1320-1328, 1331, 1333
stock market volatility puzzle 1235, 1236,
1268, 1276, 1280
stock prices 43
stock return 1233, 1240
stockout costs 884, 885
Stolper-Samuelson theorem 404
Stone price index 783
storage technologies 574, 575
strategic complementarity 1129
strategic delays 858
strong rationality 464
structural model 462
structural shifts 530
structures 840
subgame perfection 1679
subjective discount factor 548, 552, 561,593,
595, 609, 616
subsistence wage 657
substitutes 577, 590, 591,613, 616
sunk costs 858
sunspot equilibria 454, 515
sunspot paths 662
sunspot solutions 495
s e e also learning sunspot solutions
sunspots 489, 515
supply of capital 846
supply price of labor 1192, 1193
supply shocks 1129
supply-side responses 1577
surplus 853
surplus consumption ratio 1286
survivorship bias 1242
sustainability 1597
1-48
T-mapping 467, 471,512
Tanzi effect 1741
target points 826
target variables 1492, 1523
tariff 672, 695, 703-707
taste shift 778
tax
see also labor tax rate; capital taxation
distortionary 1651, 1652, 1654
on capital income 1686
on employment 1220
on international trade 703
policy 672, 708
rate 1441
on private assets 1709
reforms 822
smoothing 1655, 1659, 1662, 1705
intertemporal 1617
source-based 1715
system 1679
Taylor expansion 1265
Taylor rule 1364
technological change 1708
technological embodiment 848
technological progress 641, 1207, 1213
disembodied 1207, 1208
endogenous 639
Harrod-neutral, Hicks-neutral 944
labor-augmenting 944
purely labor-augmenting 650
technological regress, probability of in RBC
models 930
technology adoption 672, 708
technology shocks 1141, 1142, 1736
temporariness hypothesis 1569, 1572
temporary shocks 216
temporary work 1165
term premium 1255
term structure of interest rates 1270
termination costs 708
thick-market externality 1161
threshold externalities 527
thresholds 258-262, 276, 289
time-additive utility function 661
time aggregation 881
time-consistent behavior 1488
time dependency 799
time-dependent pricing 1031, 1032
time-inconsistent behavior 1653
time preference 547, 588
time preference rate 1253
Subject Index
1-49
Subject lndex
1256, 1280