JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms
American Economic Association is collaborating with JSTOR to digitize, preserve and extend
access to Journal of Economic Literature
This content downloaded from 148.206.159.132 on Fri, 15 Sep 2017 03:42:34 UTC
All use subject to http://about.jstor.org/terms
Journal of Economic Literature
Vol. XXII (March 1984), pp. 58-76
By Kevin D. Hoover
Nuffield College, Oxford University
58
This content downloaded from 148.206.159.132 on Fri, 15 Sep 2017 03:42:34 UTC
All use subject to http://about.jstor.org/terms
Hoover: Two Types of Monetarism 59
monetarists, but rather more closely re- its most eminent proponent, Friedman,
lated to the Austrian school (Kantor, by himself, while letting several econo-
1979). He rejects Hahn's classification as mists' views present the new classical doc-
misleading, because it is precisely theoret- trine. One reason for doing so is that parts
ical differences which, he believes, sepa- of Friedman's work are a foundation for
rate the new classicals from the monetar- much of the new classical doctrine (espe-
ists. cially Friedman 1968; Lucas 1972a,
In this essay, I do not want to enter the 1981b). Another reason is that Friedman
debate over titles-"monetarist" or not has contributed to economic methodol-
"monetarist." I do want to clarify the rela- ogy, as well as to monetarist thought. As
tionship between some sorts of monetar- a result, since the principal distinction be-
ism and the new classical school. To do tween Friedman and the new classicals
this I will compare Milton Friedman, in- examined in Parts IV and V is method-
disputably a monetarist, with Robert Lu- ological, Friedman's writings give us an
cas, Thomas Sargent, Neil Wallace and explicit formulation from which to work.
others, as representative of the new classi- Part of the imbalance can be set right
cal school. The principal theme of the es- by defining the new classical economics.
say is that, although we may wish to clas- It has rarely, if ever, been explicitly de-
sify the new classicals as monetarists (for fined by its adherents. The territory, how-
Tobin's or for Hahn's reasons) or we may ever, has been surveyed (Sargent, 1979,
not (for Laidler's reasons), Friedman, as 1982; Michael Beenstock, 1980; Willem
one important monetarist, differs from theBuiter, 1980). And the subsidiary element,
new classicals on a fundamental point of rational expectations, has been exten-
methodology: he is a Marshallian; they are sively covered (Robert Shiller, 1978; Da-
Walrasians. vid Begg, 1981; Lucas and Sargent, 1981;
Part II attempts to define the new classi- Rodney Maddock and Michael Carter,
cal economics and to clarify its relation 1982).
to the rational expectations hypothesis. Three tenets are keys to the new classi-
Part III explores the practical and theoret- cal doctrine. First, agents' real economic
ical relations between Friedman and the decisions for example, about savings,
new classicals that give plausibility to To- consumption or investment-are based
bin's title, "monetarist mark II," and inter- solely on real, not nominal or monetary
est to the debate over its- aptness. Part factors. Second, agents are, to the limits
IV introduces the fundamental methodo- of their information, consistent and suc-
logical distinction between Friedman and cessful optimizers; i.e., they are continu-
the new classicals. Part V shows how this ously in equilibrium. Third, agents make
distinction underlies and accounts for the no systematic errors in evaluating the eco-
differences between Friedman's and the nomic environment: i.e., they hold ra-
new classical views on equilibrium and dy- tional expectations (Lucas, 1977; Sargent,
namics. Finally, Part VI summarizes the 1979, Ch. 16).
argument. The rational expectations hypothesis is
perhaps the most striking feature of the
II. The New Classical Economics Defined new classical doctrine. The universally ac-
cepted formulation is due to John Muth
In view of the difficulties of definition (1961, p. 316): "Expectations . . . tend to
and classification mentioned in the Intro- be distributed, for the same information
duction, it may seem odd to represent the set, about the prediction of the theory (or
older monetarism by the views of even the 'objective' probability distribution of
This content downloaded from 148.206.159.132 on Fri, 15 Sep 2017 03:42:34 UTC
All use subject to http://about.jstor.org/terms
60 Journal of Economic Literature, Vol. XXII (March 1984)
outcomes)."1 Muth claims that this hy- agents are consistent and successful op-
pothesis involves three assertions: first, timizers.
information is scarce and the economic The importance of the rational expecta-
system does not waste it; second, expec- tions hypothesis is to carry these features
tations are formed from the specific struc- over to the dynamic problem. If agents
ture of the relevant system describing are to optimize over their future behavior,
the economy; and, third, public predic- their expectations of the future are bound
tions can have no substantial effects unless to be important. Rational expectations im-
there is "inside" information; i.e., a (true) plies that what they do expect is (within
economic forecast does not give anyone a serially uncorrelated error) what the
a special opportunity to profit from it if true model says they should expect. This
it is known to everyone. guarantees that they will be consistent
Muth's notion of an information set can and successful.
be taken broadly, but if taken too broadly Not everyone who uses the rational ex-
it is of little use. Typically the information pectations hypothesis should be classified
set is defined to include all the exogenous as new classical. The principle of rational
variables, all past values of endogenous expectations can be employed by anyone
variables and, crucially, the structure of seeking a convenient and, in some sense,
the model (i.e., Muth's "relevant system"). neutral way of introducing endogenous
This amounts to claiming that the model expectations into an economic model.
adequately captures those features of the One may adhere to the rational expecta-
world relevant to the formation of expec- tions hypothesis, yet violate the other te-
tations, and that agents act as if they know nets by, for example, holding that prices
the model when forming their expecta- are not flexible or that agents do not op-
tions. timize (Stanley Fischer, 1977; Edmund
While the rational expectations hypoth- Phelps and John Taylor, 1977; Tobin,
esis is a fundamental part of the new classi- 1980). Fischer, for example, constructs a
cal economics, it is, nevertheless, indepen- model in which long-term wage contracts
dent of the other tenets. A new classical produce nominal wage rigidity, which in
economist necessarily believes in rational turn permits monetary policy to have real
expectations. But a belief in rational ex- effects, since real wages then depend on
pectations by itself is not sufficient for one price movements. This violates the first
to be a new classical. In his survey of the tenet of the definition of new classical eco-
new classical macroeconomics, Sargent nomics. Lucas (1981b) criticizes Fischer's
(1982) characterizes it as going beyond ad model for supposing wage rigidity, rather
hoc supply and demand curves. He argues than explaining it as the outcome of an
that, in order to explain the behavior of optimizing decision consistent with the
macroeconomic aggregates, we must go second tenet (Lucas and Sargent, 1979).
back to the underlying objective functions
and the constraints that agents face. In a
static framework, this requires the first III. Friedmanian Roots, New Classical
two tenets of the new classical doctrine- Conclusions
that only real phenomena count and that
The new classical economics grew up
1 Muth's paper presents rational expectations as ainresponse to the perceived failure of
a supply/demand system, aimed at overthrowing the
modern-day Keynesian macroeconomics,
"cobweb theorem" analysis. Recently, however, ra-
tional expectations figure most prominently in mac- particularly as a result of the apparent
roeconomics. breakdown of the Phillips curve-ever
This content downloaded from 148.206.159.132 on Fri, 15 Sep 2017 03:42:34 UTC
All use subject to http://about.jstor.org/terms
Hoover: Two Types of Monetarism 61
This content downloaded from 148.206.159.132 on Fri, 15 Sep 2017 03:42:34 UTC
All use subject to http://about.jstor.org/terms
62 Journal of Economic Literature, Vol. XXII (March 1984)
equations, provided there is embedded in each producer (or worker) at first per-
them the actual structural characteristics ceives (falsely) a general price (wage) rise
of the labor and commodity mar- as a favorable shift of relative prices.
kets.... He also recognizes a natural The aggregate supply version of the
real rate of interest and, presumably, a Phillips curve reverses Friedman's adjust-
natural rate of output as well. This notion ment mechanism: quantities (labor or out-
of a natural rate of unemployment accords put) respond to price signals. The shift in
well with the tenets of new classicism (Lu- relative prices is only apparent not actual.
cas, 1972a, 1973; Sargent, 1973).2 No appeal is made to wage stickiness. As
The importance of the natural rate of with the new classical analysis of neutral-
unemployment and the neutrality of ity, agents are continuously in equilibrium
money is clearly seen in the expectations- given their information. Rational expecta-
augmented Phillips curve (Friedman, tions guarantee that, except for random
1968; also, Phelps, 1967, 1968). Friedman shocks, this information is correct. Hence,
argues that a sudden monetary expansion an inverse relation of unemployment to
temporarily expands output and depresses inflation is observed, but cannot be ex-
the rate of unemployment below the natu- ploited deliberately and consistently, even
ral rate. It also produces an acceleration in the shorter run, because it reflects the
in the rate of inflation. Prices rise fasterunsystematic or random component of the
than wages, so that the real wage is cut. relation between changes in individual
Observing that the volume of employ- workers' wage levels (or producers' prices)
ment is greater than they would wish at and the general rate of inflation.
this real wage, workers endeavor to re- In later work, Friedman sometimes of-
store real wages to their old level, and fers an aggregate supply explanation of
the actual rate of unemployment rises to- the Phillips curve (1974a; Friedman and
ward the natural rate. Prices (i.e., wages) Schwartz, 1982).3 Nevertheless, he does
here respond to quantity (i.e., employ- not take over the full new classical posi-
ment) signals. tion. He does not, as the new classicals
The Phillips curve in this version cannotdo, rule out prices adjusting to quantities.
be exploited in the long run, for once Indeed, his adjustment functions for both
workers anticipate any constant rate of in- prices and real income have actual and
flation, they adjust their wage claims in anticipated income as arguments (Fried-
line with it. Nevertheless, there is a real man, 1974b; Friedman and Schwartz,
shorter-run tradeoff between unemploy- 1982, Ch. 2). Furthermore, in Monetary
ment and inflation. Trends, as in his presidential address to
As with the neutrality of money, the the American Economic Association
new classicals deny even this shorter-run (1968), he still maintains that expectations
tradeoff. They recast the Phillips curve as of inflation are slow to develop (Friedman
the so-called "Lucas short-run aggregate and Schwartz, 1982, Ch. 10). The process
supply function" (Lucas, 1972a; Sargent has sped up over the past two decades,
and Wallace, 1976; Laidler, 1981, 1982). but it still takes years, not weeks or
Aggregate supply in this presentation is months, for expectations of inflation to ad-
positively related to the deviation of ac- just fully to actual changes.
tual from expected prices (or wages) as
This content downloaded from 148.206.159.132 on Fri, 15 Sep 2017 03:42:34 UTC
All use subject to http://about.jstor.org/terms
Hoover: Two Types of Monetarism 63
3.2 Government Policies: Rules versus inflation are slow to develop. He believes
Discretion that holding the growth of the money
stock to a constant X percent rate will
Following Henry Simons (1936),
result in a steady (and, if X is low enough,
Friedman (1948, 1959b, 1974a) has long
zero) rate of inflation over the longer run.
preferred rules over discretion or authori-
Once agents have adjusted to this steady
ties in the conduct of monetary policy (and
rate, no important and unpleasant sur-
other policies as well-e.g., tax and expen-
prises await them as long as the rule con-
diture policies.)4 His advocacy of fixed
tinues to be followed.
rules is based on the authorities' ignorance
Friedman's X percent rule receives
of the sources and timing of economic dis-
weak support from the new classical
turbances. Discretionary policies may or
school (Sargent and Wallace, 1975; Lucas,
may not be perverse or adverse on aver-
1981b). Their argument is derived from
age, but they do increase noise in the eco-
their belief, examined in Section 3.1, that
nomic system, which interferes with
monetary policy cannot systematically
prices acting as efficient signals (Fried-
trade off inflation for output or employ-
man, 1961, 1977).
ment. At a formal level any determinate
Lucas (1980b, 1981b) comes down
(i.e., nonrandom) rule will have no real
squarely in Friedman's camp, acknowl-
effects, because agents with rational ex-
edging the same intellectual debt. Fur-
pectations will understand the rule and
thermore, the signal extraction problem
not be fooled by changes in the money
in the context of rational expectations has
stock into moving away from the natural
been developed with great theoretical
rate of unemployment or output. Random
nicety by the new classical school (Sar-
policies may have real effects since agents
gent, 1979, Ch. 5). Sargent and Wallace
cannot anticipate them; but, of course,
(1976, p. 169) declare somewhat tenden-
they cannot be systematically pursued;
tiously: "there is no longer any serious de-
and, hence, are hardly policies at all.
bate about whether monetary policy
It is, then, the determinateness or non-
should be conducted according to rules
randomness of Friedman's rule that the
or discretion."5
new classicals support, not its particular,
Friedman has advocated a particular
simple form. Yet the new classical argu-
rule for monetary policy: namely, that
ment is that agents act as if they know
money should be allowed to grow at a con-
the structure of the economy. Of course,
stant X percent per year, where X is de-
they must infer most features, including
termined to equal the secular rate of GNP
money supply rules, from past experience.
growth after allowing for secular changes
Simple rules may be easier to infer and,
in the velocity of circulation of the money
therefore, may be preferred (Lucas,
stock. Authorities should not use money
1980b).
in an attempt to offset cyclical movements
in economic activity. This rule is related 3.3 A Paradox: Empirical Agreement,
to Friedman's belief that expectations of Theoretical Difference
Friedman (1974b) argues that what
4"Discretion" is now the more usual term, but
"authorities" is the more venerable, having been separates monetarists from Keynesians
coined by Simons (1936). are differences of empirical judgment, not
5For a contrary view, see Buiter (1980) and Ste-
of theoretical principle.6 His relation to
phen Goldfeld (1982). Sargent and Wallace (1981)
note a circumstance under which Friedman's rule
may not be an appropriate policy for securing a
lower rate of inflation. This possible objection, never- 6Friedman's view of the matter is controversial.
theless, does not alter the new classicals' support for For supporters, see, for example, Mayer (1978); for
rules over discretion in general. an opponent see, for example, Hahn (1971).
This content downloaded from 148.206.159.132 on Fri, 15 Sep 2017 03:42:34 UTC
All use subject to http://about.jstor.org/terms
64 Journal of Economic Literature, Vol. XXII (March 1984)
the new classicals appears to be just the will examine this difference in approach
reverse: their empirical judgments are more closely in Part V. Before that, how-
broadly similar, while their theoretical ever, we must examine the even more
paths to those judgments are, at times, fundamental distinction between Fried-
strikingly different. man's Marshallian incrementalism and
There are, of course, important theoret- the new classicals' Walrasian globalism.
ical similarities-e.g., the supposition of a
natural rate of unemployment. But it is IV. Walrasian and Marshallian
the similarity of their policy prescriptions Economics
which leads many to classify the new clas-
sicals as "monetarist." And it is the radical Friedman's most famous and controver-
nature of those prescriptions which leads sial contribution to economic methodol-
observers such as Tobin to distinguish ogy is his essay, "The Methodology of Posi-
them from older monetarists by labels like tive Economics" (1953). Also well known,
"mark II." if less frequently discussed, is the earlier
It is commonly believed that the ra- essay "The Marshallian Demand Curve"
tional expectations hypothesis is the (1949). Here Friedman introduces the key
source of new classical radicalism. But this distinction that separates him, as it turns
is only partly true. The examples in the out, from the new classicals. The first sec-
two preceding sections show that the tion of this part states and illustrates Fried-
characteristic new classical conclusions re- man's distinction. The relation between
quire both the other tenets of their doc- the two methodological essays is, of
trine-only real quantities matter for real course, not the main theme. Nevertheless,
decisions and continuous optimization. Friedman's views on positive economics
Indeed, Friedman (and Schwartz, 1982, are too important to be avoided. The sec-
Ch. 9) takes the rational expectations hy- ond section explores some aspects of that
pothesis to be equivalent to the proposi- relation. The third section discusses the
tion that agents make no consistent mis- new classicals as Walrasians. The final sec-
takes about real variables in the long run. tion reexamines the three defining tenets
Later Friedman writes: of the new classical doctrine in light of
the methodological distinction developed
The formalization in the theory of rational ex- in the first three sections.
pectations of the ancient idea that economic
actors use available information intelligently in 4.1 The Cournot Problem
judging future possibilities is an important and
valuable development. But it is not the open Friedman (1949) argues that the com-
sesame to unraveling the riddle of dynamic mon view that Marshall deals with partial
change that some of its more enthusiastic pro-
equilibrium, while Walras deals with gen-
ponents make it out to be [Ch. 12 p. 630].
eral equilibrium is false. Rather both deal
This quotation gets to the nub of the with general equilibrium. Partial equilib-
matter. The new classicals wish to analyze rium must be conceived of as but a special
economic dynamics using the rational ex- case of general equilibrium. Marshall, too,
pectations hypothesis. To do this they is an advocate of the view that in econom-
must collapse the long run into the short ics everything depends on everything
run, appealing to the other tenets of the else.
new classical doctrine. Friedman wishes In place of this common and, he says,
to analyze economic dynamics by retain- erroneous distinction between Marshall
ing the Marshallian distinction between and Walras, Friedman proposes another.
the market, short and long periods. We He argues that Marshall and Walras con-
This content downloaded from 148.206.159.132 on Fri, 15 Sep 2017 03:42:34 UTC
All use subject to http://about.jstor.org/terms
Hoover: Two Types of Monetarism 65
This content downloaded from 148.206.159.132 on Fri, 15 Sep 2017 03:42:34 UTC
All use subject to http://about.jstor.org/terms
66 Journal of Economic Literature, Vol. XXII (March 1984)
macromodel that includes a money supply lem. For the point of the Cournot problem
equation as well as a money demand equa- and Marshall's (and Friedman's) solution
tion, in his empirical work he estimates to it is that, whatever the economic prob-
only money demand, making the practical lem, any practically significant analysis of
judgment that the money supply is too it requires that reality be partitioned. The
unstable to model and may be treated as most important bits with respect to the
exogenous (Friedman 1959a). Similarly, problem at hand are analyzed in detail;
he holds that the whole range of implicit the rest are summarized in less detail (but
returns on assets might affect money de- not forgotten, of course).
mand or consumption, but these are re- A Marshallian approach does, however,
duced in his applied work to portmanteau rule out one sort of criticism of theories.
variables or ignored altogether (Friedman It is pointless to attack a theory as "unre-
1956a, b, 1959a). Likewise, he estimates alistic" or "incomplete" solely on the
expectations or permanent values of varia- grounds that it partitions reality: all useful
bles by distributed lags on their own actual theories must do so.
past values, even though they may in prin- The question naturally arises about how
ciple depend in part on the present and this defense of "unrealism" relates to
expected future values of the dependent Friedman's (1953) famous claim-the so-
variables of the equations in which they called F-twist-that the less realistic a the-
appear as independent variables (Fried- ory is, the more useful it is. There is con-
man 1956b, for the example of permanent siderable debate about what Friedman
income in the consumption function). It means by "unrealism," and it is not to our
is not that Friedman must use this method purpose to try to decide the matter here.
in every particular case. It is just that only It is enough to note that if "unrealism"
a practical advantage of another method, means nothing more than partitioned,
say rational expectations, in generating then Friedman's advocacy of unrealistic
successful predictions would obligate him theories follows naturally from his Mar-
to abandon it. shallian method. Of course, the Marshal-
lian method requires an appropriate de-
4.2 Positive Economics
gree of partition, not necessarily the
In drawing the distinction between greatest possible degree. So if this is what
the Marshallian and Walrasian ap- Friedman means by "unrealism," the F-
proaches, Friedman makes much of his twist may be an exaggerated claim. If "un-
belief, developed at length in his "The realism" means something more than par-
Methodology of Positive Economics" titioned, then Friedman's claim must be
(1953), that success at prediction is the sole supported on other grounds. A supporter
criterion for judging theories. Predictive of the Marshallian method might then re-
success is the standard of efficacy by which ject the F-twist. (On Friedman's positiv-
the engines of economic analysis are to ism and the F-twist, see Lawrence Boland,
be evaluated. 1979, 1982.)
The question of the appropriate stan-
4.3 The New Classical Economics'
dard of efficacy is, nevertheless, separable
Walrasian Solution
from the Marshallian response to the
Cournot problem. Even if there were- The desirability of predictive success
contrary to Friedman's view-another is not at issue between Friedman and the
standard of efficacy for the empirical, pol- new classicals. Again and again they praise
icy problems that interest him, it would predictive success as a standard of efficacy
not affect his analysis of the Cournot prob- (Lucas, 1977, 1981b; Muth, 1961). Oddly
This content downloaded from 148.206.159.132 on Fri, 15 Sep 2017 03:42:34 UTC
All use subject to http://about.jstor.org/terms
Hoover: Two Types of Monetarism 67
Sargent
enough, the importance of and Wallace tract-
practical (1975) strongly criti-
ability implicit in the Marshallian ap- cize their own and Keynesian macro-
proach is also not at issue, in some re- models for the lack of consistent assump-
spects. Lucas' preference for quantities tions about firms' and individuals'
adjusting to prices over prices adjusting objective functions. Friedman, while be-
to quantities is based on the fact that the lieving in rational optimizing man to a first
theory of the former is better developed approximation, could not sustain that sort
and easier to handle than the theory of of objection if the model yielded accurate
the latter (1977). Lucas accepts that theo- predictions, whatever its theoretical as-
retical techniques develop over time and, sumptions.8
consequently, that it is misguided to criti- An explicit insistence on a general equi-
cize our forerunners for not developing librium foundation is the basis for the fa-
theories which would have been intracta- mous "Lucas critique" of macroecono-
ble, given the then-available techniques metric models (Lucas, 1976; Lucas and
(1980a). Sargent, 1979). Expectations terms cannot
On the surface, then, Friedman and the be observed. So in most econometric mod-
new classicals agree about the desirability els they are replaced by some function
of predictive success and analytical tract- of the observable terms upon which ex-
ability. Friedman, however, anticipated pectations are formed. The rational expec-
this essentially false concord. "Most mod- tations hypothesis implies that the struc-
ern theorists would accept [the Marshal- ture of the model itself is part of the
lian view] of the objectives of economic information upon which expectations are
theory. But our work belies our profes- formed. If some part of the structure
sions" (Friedman, 1949, p. 490). Friedman changes, say, the money supply rule, then
is, in his own view, primarily an empirical the proxy function for the price expecta-
economist, who uses a few deeply held tions term will itself change, even if the
principles to sift through facts in search underlying behavioral relations remain
of predictions. The "new classicals" recog- fixed. A model based on the old function
nize the challenge of the facts, but see will no longer forecast well. (A particularly
the problem more as one of consistently straightforward illustration of this argu-
reconciling the facts with their world ment can be found in Thomas Turner and
view. Charles Whiteman, 1981.)
The new classicals are Walrasians in two
senses. First, they advocate general equi-
8 There is a third sense of "Walrasian" which can-
librium analysis (Lucas, 1980a). Friedman not be ascribed to the new classicals. This is perhaps
admires the Walrasian system for ". . . its the most common but not the most useful, sense
indicating those who pursue purely mathematical
beauty, its grandeur, its architectonic
general equilibrium theory. E. Roy Weintraub (1983,
structure . . . ," but he does not expect p. 37) concludes his historical review of the literature
to obtain useful predictions from it (1955, on the existence of equilibrium: "The [Walrasian]
'equilibrium' story is one in which empirical work,
p. 905). The new classicals do. Lucas (1977)
ideas of facts and falsifications, played no role at all."
argues that, given business cycles in the The Austrian precursors of the new classicals were
data, the theorist's challenge is to recon- sceptical of empirical economics (Friedrich A.
Hayek, 1933, Ch. 1; 1979, Ch. 4 and 6). The new
cile them with general equilibrium theo-
classicals themselves are sceptical of empirical eco-
ries. nomics that does not pay sufficient heed to general
The second sense in which the new clas- equilibrium but not of empirical economics in gen-
eral (Lucas, 1976; Sargent, 1982). The acceptance
sicals are Walrasian is obviously related
of the importance of empirical economics may be
to the first: partition is a legitimate ground
the one thing that distinguishes the new classicals
on which to criticize an applied theory. most clearly from the Austrians.
This content downloaded from 148.206.159.132 on Fri, 15 Sep 2017 03:42:34 UTC
All use subject to http://about.jstor.org/terms
68 Journal of Economic Literature, Vol. XXII (March 1984)
The new classical alternative to conven- In his latest work, Friedman concedes
tional, i.e., structurally noninvariant, the principle of the "Lucas critique," that
econometric models is to posit the model a function of past values of a variable may
with its expectational terms in place. not be a stable proxy for its expected value
These are then solved out using the ra- (Friedman and Schwartz, 1982, Ch. 2).
tional expectations hypothesis in such a Nevertheless, his Marshallian method
way that the interrelationships between does not force him to take account of it
the coefficients of different equations be- unless it proves to be a practical barrier
come explicit. Now when a policy rule to accurate prediction in a specific case.
changes, the appropriate changes to the Even then his preferred method is to work
expectations function and, therefore, to backwards incrementally to the minimum
the solutions of the full model are auto- necessary level of complication, rather
matic. So far, this has been done only for than to impose a fully interdependent
relatively simple models. structure on the problem, as Lucas sug-
To see that the "Lucas critique" result gests.10
follows from the insistence on general in- In addition to such explicit applications
terdependence, and not from the rational of the Walrasian method, it is applied im-
expectations hypothesis itself, consider plicitly in other cases: partition is taken
Friedman's (1956b) permanent income as a sufficient ground for criticism. For ex-
hypothesis. Permanent income is the flow ample, the rational expectations models
of income from wealth defined as the sub- of Fischer (1977) and Phelps and Taylor
jective expectation of the discounted (1977) are criticized because they impose
stream of the agent's future income. In unexplained wage or price rigidities (Lu-
order to apply the permanent income hy- cas and Sargent, 1979; Lucas, 1981b). It
pothesis, Friedman supposes that the sub- is only if these rigidities (e.g., long term
jective expectation equals what in fact ac- labor contracts) can be deduced from first
tually happens on average. This is a principles-say, from a general equilib-
simple form of rational expectations. Per- rium system with contingent contracts-
manent income can, thus, be represented that the models can be accepted, even if
as an optimal forecast. Friedman repre- they fit the data. For only then can their
sents it by a geometrically declining lag predictive success rest on more than the
on past income, which has since been accidental constancy of a theoretically
shown by Muth (1961) to be statistically variable bit of reality (Lucas, 1977).
optimal under some circumstances. One
4.4 The Tenets of New Classicism
of the circumstances is that there are no
Reexamined
policy changes. If there are policy
changes-say, a change in taxation affect- In Part II we suggested that three te-
ing income-then the predictions are no nets characterize the new classical doc-
longer optimal (Lucas, 1976). Thus Fried- trine-namely, that agents attend to real
man implicitly uses the rational expecta- factors only in making real economic deci-
tions hypothesis, but still falls victim of sions, that they are consistent and success-
the "Lucas critique" for failing to account ful optimizers, and that they hold rational
for interdependence with the policy rule.9 expectations. So far, we have argued in
9 Buiter (1980) correctly observes that all that is 10 Lucas (1981a, Intro.) qualifies his position some-
needed for the "Lucas critique" to go through is a what. How far back one must go to secure invariance
direct effect of government policy on private expec- depends in part on the particular problem at hand.
tations. Rational expectations is but one way of get- Nevertheless, he would generally prefer to take only
ting it. tastes and technology as given.
This content downloaded from 148.206.159.132 on Fri, 15 Sep 2017 03:42:34 UTC
All use subject to http://about.jstor.org/terms
Hoover: Two Types of Monetarism 69
this part that the most important distinc- der unspecified constraints. The Marshal-
tion between Friedman's doctrine and lian method is in large measure the view
that of the new classicals is that, in Fried- that it is legitimate not to analyze those
man's own usage, they are Walrasians, constraints explicitly when the longer-run
while he is a Marshallian. Now let us con- behavior is itself largely independent of
sider how this fundamental distinction re- the shorter run and when one is primarily
lates to the tenets of new classicism. interested in the longer run.
The first tenet, that only real factors It might be argued that the contrast be-
matter for real economic decisions, is tween Friedman's view that agents are
not-at least for longer time horizons- optimizing under unspecified constraints
fundamentally at issue. It is the basis for and the new classical view that agents op-
Friedman's and the new classicals' general timize to the limits of their information
agreement about the neutrality of money, draws a distinction without a difference.
the natural rate of unemployment and the This would, however, confuse the agent's
Phillips curve discussed in Section 3.1. point of view with the observer's (or econ-
The differences that remain between omist's) point of view. Friedman argues
them on these matters should be ascribed that we can have faith that agents do their
to the second tenet. best, but that, as observers, we need not
The proposition that economic agents specify how they do it. The new classicals,
are consistent and successful optimizers on the other hand, require that the infor-
is in some sense agreed by all those-both mational limits which constrain agents be
Friedman and the new classicals in- specified precisely if the observer is to un-
cluded-imbued with the so-called Chi- derstand their behavior at all.
cago tradition in economics (Melvin Friedman is a pessimist about solving
Reder, 1982). Precisely in what sense to the Cournot problem. His Marshallian
take the proposition is the fundamental method is a way of pursuing economic
point at issue between them. Friedman analysis in the face of the problem, but
takes consistent optimization as an article it does not dissolve it. The new classicals,
of faith. Because of the Cournot problem, on the other hand, are optimists with re-
he cannot detail every aspect of agents' spect to the Cournot problem. They be-
economic behavior. Thus, he applies the lieve that a Walrasian program of fully
optimization proposition to a part of the specifying the optimization problem
problem, while retaining a faith that which agents face is a real possibility (Sar-
agents do optimize with respect to those gent, 1982; Lucas, 1977). To return once
parts not worked out in detail. For exam- more to the Phillips curve example, they
ple, in his analysis of the Phillips curve, argue that when it is cast as an optimiza-
he sets out agents' longer-run behavior in tion problem with the constraints on
order to show that, if they are optimizers, agents fully set out, the possibility of an
any tradeoff between inflation and unem- exploitable shorter-run tradeoff vanishes.
ployment will not endure. He does not The new classicals' conclusion about the
analyze their shorter-run behavior in de- Phillips curve hinges, as we observed in
tail: he uses the Marshallian method to Section 3.1, on the assumption of rational
partition the problem. Nevertheless, he expectations. We can now see why this
does not assert at any point that agents third tenet is a necessary, but subsidiary,
are not optimizing in the shorter run. If element of the new classical doctrine. Ex-
agents' shorter-run behavior differs from pectations become important as soon as
their longer-run, optimal behavior, he one turns to a dynamic problem. The Wal-
supposes that they are still optimizing un- rasian interpretation of the second tenet-
This content downloaded from 148.206.159.132 on Fri, 15 Sep 2017 03:42:34 UTC
All use subject to http://about.jstor.org/terms
70 Journal of Economic Literature, Vol. XXII (March 1984)
This content downloaded from 148.206.159.132 on Fri, 15 Sep 2017 03:42:34 UTC
All use subject to http://about.jstor.org/terms
Hoover: Two Tuves of Monetarism 71
mand for money function. Then, with re- uncertainty, economic reasoning will be
spect to that particular function, an agent of no value" (1977, p. 15).
may be in short-run disequilibrium. The Friedman implicitly defends the impor-
essence of much of Friedman's work is tance of uncertainty. He points out the
that much may be known about the long difficulty of giving precise meaning to the
run, yet little about the short run. For ex- identification of the subjective with the
ample, there is, he argues, a stable long- objective probability distribution and to
run relation between money and prices; the ambiguity of the time limit over which
yet the lags are long and variable (Fried- forecast errors are supposed to be uncor-
man and Schwartz, 1963b; Friedman, related (Friedman and Schwartz, 1982,
1974b). Ch. 12). In a counter-example, he distin-
The new classical response to the vacu- guishes between anticipations of inflation
ousness of the general definition of equi- during the 1890s owing to the rise of the
librium is as predictably Walrasian as free silver movement, reflected in high
Friedman's is Marshallian. It is to impose rates of interest, and the realized fall in
a fully specified theory on the data and the rate of inflation. Agents made persis-
to use the rational expectations hypothesis tent, serially correlated errors in fore-
to solve the problem that the future is un- casting inflation; but whether their
known (Lucas and Sargent, 1979; Lucas, expectations coincided with the objective
1980a). This is, in effect, to collapse Mar- probability distribution is a moot point,
shall's long run into the short run, as in because a historical situation cannot be re-
the examples of Part III. peated in a controlled experiment in or-
der to ascertain the objective frequency
5.2 Dynamic Equilibrium versus Long-
of it falling out one way or the other.
run Equilibrium
Whether or not the United States would
In Section 4.4 we saw that the rational go on to free silver was a classic example
expectations hypothesis was an implica- of uncertainty. Observing their past fore-
tion of applying the Walrasian interpreta- cast errors, agents might (rationally) not
tion of the consistent optimization propo- revise their current forecasts so long as
sition-i.e., the second tenet of the new they believed a move to free silver was
classical doctrine-to a dynamic problem. a real possibility.
Agents are subject to error arising from The distinction between risk and uncer-
their own ignorance of the future. But tainty suits Friedman's pessimism about
their ignorance can be characterized as solving Cournot's problem along Walra-
risk in Frank Knight's sense (1937). Over sian lines. It is a Marshallian distinction.
any length of time or over many indepen- It partitions the agent's ignorance into
dent agents the errors that result from risk that which is quantifiable (risk) and that
take on a pattern that may be summed which is not (uncertainty). In the free sil-
up in an objective probability distribution. ver illustration, Friedman also partitions
One can insure against such errors. time into the Marshallian short and long
The rational expectations hypothesis runs. Thus, he argues, that if the time unit
amounts to identifying the objective prob- is long enough-say, twenty years-fore-
ability distribution of forecast errors with cast errors are uncorrelated, which is to
the agent's own subjective assessments. say characterized by risk. In the short run,
This rules out Knight's other category however, uncertainty is relevant, at least
of ignorance-uncertainty or uninsurable in retrospect when we explain an episode
error. Uncertainty, says Lucas, has no like that of the 1890s historically.
place in economic analysis: "In cases of The new classicals' view of uncertainty
This content downloaded from 148.206.159.132 on Fri, 15 Sep 2017 03:42:34 UTC
All use subject to http://about.jstor.org/terms
72 Journal of Economic Literature, Vol. XXII (March 1984)
This content downloaded from 148.206.159.132 on Fri, 15 Sep 2017 03:42:34 UTC
All use subject to http://about.jstor.org/terms
Hoover: Two Types of Monetarism 73
Lucas' model is an elaboration of a neo- optimists about solving the Cournot prob-
classical growth model in which, in a lem: Lucas predicts that an adequate equi-
steady-state with certainty, real quanti- librium model may be ". . . five but not
ties, the money stock and prices would twenty-five years off" (1977, p. 25).
grow at constant rates. In order to illus- Friedman does not believe that Cour-
trate his explanation of business cycles, not's problem will be so easily solved. In
suppose that there is uncertainty and, in the meantime, the Marshallian method
particular, that there is an unanticipated and its accompanying distinction between
increase in the money supply which the long run and the short run are needed
agents first perceive as increased demand. for practical purposes. Lucas writes: "The
They expand output and, if they are igno- idea that an economic system in equilib-
rant of the shock's transient nature, in- rium is in any sense 'at rest' is simply an
crease their rate of investment in order anachronism" (1980a, p. 708). Friedman's
to adjust the capital stock to the higher approach to economic dynamics requires
level of demand. By rational expectations, no such assumption. He needs merely to
they soon learn of their mistake; but, since maintain that our ignorance of the fine
capital endures, they are stuck with a real details of economic processes is such that
change in their economic environment. we know more about the secular than the
Even if there are no further shocks, a cyclical behavior of the economy. It is,
cycle has been generated. For now agents then, more useful to detail the secular in
see that their capital stock is too high, so our description of the Marshallian long-
they choose a lower rate of investment run equilibrium, which may well be mov-
in order to optimally reduce the capital ing, and to summarize cyclical behavior
stock to its optimal level. Prices cycle as as a short-run adjustment towards this
well: the initial effect of the increased sup- equilibrium."1
ply of money is that prices rise faster than Indeed, Friedman's method of treating
they otherwise would have. Once in- the business cycle has not altered much
stalled, however, the increased capacity in the past twenty years. In the earlier
resulting from the new investment retards empirical examination of the role of
the rate of increase of prices below the money in business cycles (Friedman and
steady-state rate until the optimal capital Schwartz, 1963b) as well as in the theoreti-
stock is restored. Other real quantities- cal sections of Monetary Trends (1982),
employment, output and so forth-con- Friedman imagines the economy in long-
nected by agents' plans to their decisions run equilibrium (an "Elysian state of mov-
about the capital stock and investment cy- ing equilibrium") before and after a mone-
cle as well. Agents do not make persistent tary shock. He then deduces general
mistakes; their mistakes may nonetheless properties of the economic dynamics that
have persistent consequences, even when must obtain if the economy is to move
they act optimally. from one long-run equilibrium to the
Lucas is not happy with the details of other. The picture he paints of the mecha-
the particular model just sketched (1981a, nism by which monetary shocks are trans-
Intro.). He recognizes that a competitive mitted to cycles in real quantities remains
equilibrium theory of the business cycle impressionistic and does not extend to the
must solve the Cournot problem by ade- full specification of the agents' economic
quately characterizing the constraints that
""Secular" is used here in the usual, modern
agents face, and he is careful not to claim
sense, which differs from Marshall's usage in which
success too early. Nevertheless, as we have the secular is a longer horizon than the long run
already observed, the new classicals are (Marshall 1930, Book 5, Ch. 5).
This content downloaded from 148.206.159.132 on Fri, 15 Sep 2017 03:42:34 UTC
All use subject to http://about.jstor.org/terms
74 Journal of Economic Literature, Vol. XXII (March 1984)
environment that is required by the new ful to divide the problem up into Marshal-
classical standards of adequacy. lian "runs," they insist that nothing less
The difference between Friedman and than a full equilibrium approach, in which
the new classicals here as elsewhere is not the distinction between short run and the
over the nature of the economy: both long run is abolished, will do.
agree that at root it is a complex, interde-
pendent system. Rather it is that, in the REFERENCES
face of the Cournot problem, Friedman
believes that only the Marshallian method BEENSTOCK, MICHAEL. A neoclassical analysis of
macroeconomic policy. Cambridge: Cambridge U.
of partitioning problems produces fruitful
Press, 1980.
results; while the new classicals believe BEGG, DAVID K. The rational expectations revolu-
that only the Walrasian method of fully tion in macroeconomics: Theories and evidence.
Deddington, Oxford, England: Philip Allan, 1982.
specifying the optimization problem that
BoLAND, LAWRENCE A. "A Critique of Friedman's
agents face produces secure results. Critics," J Econ. Lit., June 1979, 17(2), pp. 503-
22.
. The foundations of economic method. Lon-
VI. Summary don: George Allen and Unwin, 1982.
BUITER, WILLEM. "The Macroeconomics of Dr. Pan-
There has been a lively debate over gloss: A Critical Survey of the New Classical Mac-
roeconomics," Econ. J., Mar. 1980, 90(357), pp.
whether or not Lucas, Sargent, Wallace
34-50.
and their fellow adherents to the new clas- COURNOT, AUGUSTIN. Researches into the mathe-
sical doctrine are best thought of as mone- matical principles of the theory of wealth. Trans.:
NATHANIEL T. BACON. NY: Macmillan, [1838]
tarists. The central argument of this essay
1927.
is simply that, whatever we may choose FISCHER, STANLEY. "Long Term Contracts, Rational
to call them, these new classicals must be Expectations and the Optimal Money Supply
Rule," J Polit. Econ., Feb. 1977, 85(1), pp. 191-
distinguished from an older strand of mon-
205.
etarism represented by Friedman. FRIEDMAN, MILTON. "A Monetary and Fiscal Frame-
We have seen that the strongest case work for Economic Stability," Amer. Econ. Rev.,
for calling the new classicals "monetarists" June 1948, 38(3), pp. 245-64.
. "The Marshallian Demand Curve," J. Polit.
is based on the similarity of their policy Econ., Dec. 1949, 57(6), pp. 463-95.
prescriptions to Friedman's. Nevertheless, . "The Methodology of Positive Economics,"
Essays in positive economics. Chicago, IL: U. of
their conclusions have a more radical tone
Chicago Press, 1953, pp. 3-46.
than his. This is the result of a different . "Leon Walras and His Economic System: A
theoretical analysis. Review Article," Amer. Econ. Rev., Dec. 1955,
45(5), pp. 900-09.
An essential part of the theoretical dif-
.__ "The Quantity Theory of Money: A Restate-
ference is a difference in methodology: ment," in Studies in the quantity theory of money.
they are Walrasians, while he is a Marshal- Ed.: MILTON FRIEDMAN. U. of Chicago Press,
1956, pp. 3-24.
lian. The characteristic new classical re-
_ A theory of the consumption function.
sults follow from the attempt to analyze Princeton, NJ: Princeton U. Press, 1957.
the economy as an interdependent whole. . "The Demand for Money: Some Theoretical
Friedman recognizes interdependence, and Empirical Results," J Polit. Econ., Aug. 1959a,
67(3), pp. 327-51.
but insists that, for practical purposes, . A program for monetary stability. NY: Ford-
problems must be partitioned into parts ham U. Press, 1959b.
analyzed in detail and parts summarized. . "The Lag in the Effect of Monetary Policy,"
in The optimum quantity of money. By MILTON
An important aspect of the difference FRIEDMAN. London: Macmillan, [1961] 1969a, pp.
between the Marshallian and Walrasian 237-60.
. "The Role of Monetary Policy," Amer. Econ.
approaches is the different treatment that
Rev., Mar. 1968, 58(1), pp. 1-17.
Friedman and the new classicals give of . The optimum quantity of money. London:
economic dynamics. While he finds it use- Macmillan, 1969a.
This content downloaded from 148.206.159.132 on Fri, 15 Sep 2017 03:42:34 UTC
All use subject to http://about.jstor.org/terms
Hoover: Two Types of Monetarism 75
This content downloaded from 148.206.159.132 on Fri, 15 Sep 2017 03:42:34 UTC
All use subject to http://about.jstor.org/terms
76 Journal of Economic Literature, Vol. XXII (March 1984)
Real Rate of Interest, and the Natural Rate of Un- Dynamic Structure of Macroeconomic Models: A
employment," Brookings Pap. Econ. Act., 1973:2, Critical Review," J. Monet. Econ., 1976, 4(1), pp.
pp. 429-72. 1-44.
. "A Classical Macroeconometric Model of the SIMONS, HENRY C. "Rules versus Authorities in Mon-
United States," J. Polit. Econ., Apr. 1976, 84(2), etary Policy," J. Polit. Econ., Feb. 1936, 44(1), pp.
pp. 207-37. 1-30.
. Macroeconomic theory. NY: Academic Press, STEIN, JEROME, ed. Monetarism. Amsterdam: North
1979. Holland, 1976.
. "Beyond Demand and Supply Curves in . Monetarist, Keynesian and new classical eco-
Macroeconomics," Amer. Econ. Rev., May 1982, nomics. Oxford: Blackwell, 1982.
72(2), pp. 382-89. TOBIN, JAMES. Asset accumulation and economic ac-
AND WALLACE, NEIL. "'Rational Expecta- tivity. Oxford: Blackwell, 1980.
tions,' the Optimal Monetary Instrument and the . "The Monetarist Counter-Revolution To-
Optimal Money Supply Rule," J. Polit. Econ., Apr. day-An Appraisal," Econ. J., Mar. 1981, 91(361),
1975, 83(2), pp. 241-54. pp. 29-42.
. "Rational Expectations and the Theory of TURNER, THOMAS H. AND WHITEMAN, CHARLES H.
Economic Policy," I. Monet. Econ., Apr. 1976,2(2), "Econometric Policy Evaluation Under Rational
pp. 169-83. Expectations," Fed. Res. Bank Minn. Quart. Rev.,
. "Some Unpleasant Monetarist Arithmetic," 1981, 5(2), pp. 6-15.
Fed. Res. Bank Minn. Quart. Rev., 1981, 5(3), pp. WEINTRAUB, E. Roy. "On the Existence of a Com-
1-17. petitive Equilibrium: 1930-1954," J. Econ. Lit.,
SCHILLER, ROBERT. "Rational Expectations and the Mar. 1983, 21(1), pp. 1-39.
This content downloaded from 148.206.159.132 on Fri, 15 Sep 2017 03:42:34 UTC
All use subject to http://about.jstor.org/terms