John H. Cochrane
Over the last century, the these issues. I start with the statistical analysis
stock market in the United of past stock returns. The long-term average
States has yielded impressive return is in fact rather poorly measured. The
returns to its investors. For standard statistical confidence interval extends
example, in the postwar period, from 3 percent to 13 percent. Furthermore,
stock returns have averaged 8 percentage points average returns have been low following times
above Treasury bills. Will stocks continue to of high stock prices, such as the present. There-
give such impressive returns in the future? Are fore, the statistical evidence suggests a period
long-term average stock returns a fundamental of quite low average returns, followed by slow
feature of advanced industrial economies? Or are reversion to a poorly measured long-term average,
they the opposite of the old joke on Soviet agri- and it cautions us that statistical analysis alone
culture—100 years of good luck? If not pure leaves lots of uncertainty.
good luck, perhaps they result from features of Then, I survey economic theory to see if
the economy that will disappear as financial standard models that summarize a vast amount
markets evolve. of other information shed light on stock returns.
How does the recent rise in the stock mar- Standard models do not predict anything like
ket affect our view of future returns? Do high the historical equity premium. After a decade
prices now mean lower returns in the future? of effort, a range of drastic modifications to the
Or have stocks finally achieved Irving Fisher’s standard models can account for the historical
brilliantly mistimed 1929 prediction of a “per- equity premium. But it remains to be seen
manently high plateau?” If stocks have reached whether the drastic modifications and a high
a plateau, is it a rising plateau, or is the market equity premium, or the standard models and a
likely to bounce around its current level for many low equity premium will triumph in the end. In
years, not crashing but not yielding returns much sum, economic theory gives one further reason
greater than those of bonds? to fear that long-term average excess returns
These questions are on all of our minds as will not return to 8 percent, and it details the kind
we allocate our pension plan monies. They are of beliefs one must have about the economy to
also important to many public policy questions. reverse that pessimistic view.
For example, many proposals to reform social
security emphasize the benefits of moving to John H. Cochrane is the Sigmund E. Edelstone
a funded system based on stock market invest- Professor of Finance in the Graduate School of
Business at the University of Chicago and a con-
ments. But this is a good idea only if the stock sultant to the Federal Reserve Bank of Chicago.
market continues to provide the kind of returns His research is supported by a grant from the
in the future that it has in the past. National Science Foundation, administered by the
National Bureau of Economic Research, and by
In this article, I summarize the academic, the University of Chicago. The author would like
and if I dare say so, scientific, evidence on to thank George Constantinides, Andrea Eisfeldt,
and David Marshall for many helpful comments.
TABLE 1
Annual real returns 1947–96
VW S&P500 EW GB CB TB
(- - - - - - - - - - - - - - - - - - percent - - - - - - - - - - - - - - - - - - -)
4 ECONOMIC PERSPECTIVES
However, the risk of the bet (total return) is TABLE 2
much larger: It only takes an average number
How risk and return vary with
of heads equal to 0.499 (that is, 499/1,000) to investment horizon
lose a dollar; if the average number of heads is
0.490, still very close to 0.5, you lose $10. Just E (R e ) σ(R e ) 1 E (R e )
Horizon h
as we care about dollars, not the fraction of (years) h h h σ (R e )
heads, we care about total returns, not annual-
ized rates. 1 8.6 17.1 0.50
To address the short-run price fluctuation 2 9.1 17.9 0.51
fallacy directly, table 2 shows that mean returns 3 9.2 16.8 0.55
and standard deviations scale with horizon just 5 10.5 21.9 0.48
about as this independence argument suggests, e
Notes: R = value weighted return less T-bill rate.
out to five years. Column one shows average excess return divided
by horizon. Column two shows standard deviation
(In fact, returns are not exactly indepen- of excess return divided by square root of horizon.
Column three shows Sharpe ratio divided by square
dent over time. Estimates in Fama and French root of horizon. All statistics in percent.
[1988a] and Poterba and Summers [1988]
suggest that the variance grows a bit less slowly
than the horizon for the first five to ten years,
on substantial risk. What happens to a funded
and then grows with horizon as before, so
social security system if the market goes down?
stocks are in fact a bit safer for long horizons
than the independence assumption suggests. Means versus standard
However, this qualification does not rescue the deviations—Sharpe ratio
annualized return fallacy. Also bear in mind Figure 1 presents mean returns versus their
that long-horizon statistics are measured even standard deviations. In addition to the portfoli-
less well than annual statistics; there are only os listed in table 1, I include ten portfolios of
five nonoverlapping ten-year samples in the NYSE stocks sorted by size. This picture
postwar period.) shows that average returns alone are not a
The stock market is like a coin flip, but it particularly useful measure. By taking on more
is a biased coin flip. Thus, even though mean risk, one can achieve very high average re-
and variance may grow at the same rate with turns. In the picture, the small stock portfolio
horizon, the probability that one
loses money in the stock market
does decline over time. (For exam- FIGURE 1
ple, for the normal distribution, tail Mean vs. standard deviation of real returns, 1947–96
probabilities are governed by E(r)/ average return, percent
σ(r), which grows at the square 16
root of horizon.) However, portfolio
advice is not based on pure proba-
bilities of making or losing money; 12
but on measures such as the mean Equally weighted
and variance of return. Based on S&P500
Value weighted
such measures, there is not much 8
presumption that stocks are dramati-
cally safer for long-run investments.
I cannot stress enough that the 4
high average returns come only as Corporate bonds
compensation for risk. Our task Treasury bill Government bonds
below is to understand this risk and 0
people’s aversion to it. Many dis- 0 4 8 12 16 20 24 28 32
standard deviation of return, percent
cussions, including those surrounding
Notes: Triangles are equally weighted and value weighted NYSE; S&P 500;
the move to a funded social security three-month Treasury bill; ten-year government bond; and corporate bond
returns. Unmarked squares are NYSE size portfolios.
system, implicitly assume that one
gets the high returns without taking
6 ECONOMIC PERSPECTIVES
Selection and crashes banking panics, no depressions, no civil wars,
Two important assumptions behind the no constitutional crises, the cold war was not
standard error calculation, however, suggest ways lost, and no missiles were fired over Berlin,
in which the postwar average stock return might Cuba, Korea, or Vietnam. If any of these things
still have been largely due to luck. Argentina had happened, there undoubtedly would have
and the U.S. looked very similar at the middle been a calamitous decline in stock values. The
of the last century. Both economies were un- statistical problem is nonnormality. Taking the
derdeveloped relative to Britain and Germany standard deviation from a sample that did not
and had about the same per capita income. If include rare calamities, and calculating average
Argentina had experienced the U.S.’s growth return probabilities from a normal distribution
and stock returns, and vice versa, this article may understate the true uncertainty. But inves-
would be written in Spanish from the Buenos tors, aware of that uncertainty, discount prices
Aires Federal Reserve Bank, with high Argentine and hence leave high returns on the table.
stock returns as the subject. We can cast the issue in terms of funda-
The statistical danger this story points to is mental beliefs about the economy. Was it clear
selection or survival bias. If you flip one coin to people in 1945 (or 1871, or whenever the
ten times, the chance of seeing eight heads is sample starts) and throughout the period that
low. But if you flip ten coins ten times, the the average return on stocks would be 8 percent
chance that the coin with the greatest number greater than that of bonds? If so, one would
of heads exceeds eight heads is much larger. expect them to have bought more stocks, even
Does this story more closely capture the 50-year considering the risk described by the 17 per-
return on U.S. stocks? Brown, Goetzmann, and cent year-to-year variation. But perhaps it was
Ross (1995) present a strong case that the uncer- not in fact obvious in 1945, that rather than
tainty about true average stock returns is much slipping back into depression, the U.S. would
larger than σ / T suggests. As they put it, experience a half century of growth never be-
“Looking back over the history of the London fore seen in human history. If so, much of the
or the New York stock markets can be extraor- equity premium was unexpected; good luck.
dinarily comforting to an investor—equities
appear to have provided a substantial premium Time varying expected returns
over bonds, and markets appear to have recov- Regressions of returns on price/dividend ratios
ered nicely after huge crashes. . . . Less com- We are not only concerned with the aver-
forting is the past history of other major mar- age return on stocks but whether returns are
kets: Russia, China, Germany, and Japan. Each expected to be unusually low at a time of high
of these markets has had one or more major prices, such as the present. The first and most
interruptions that prevent their inclusion in natural thing one might do to answer this ques-
long term studies” [my emphasis]. tion is to look at a regression forecast. To this
In addition, think of the things that didn’t end, table 4 presents regressions of returns on
happen in the last 50 years. There were no the price/dividend (P/D) ratio.
TABLE 4
OLS regressions of excess returns and dividend growth on VW P/D ratio
Rt→t+k = a + b(Pt /Dt) Dt+k/Dt = a + b(Pt/Dt)
Horizon k
(years) b σ( b ) R 2
b σ( b ) R2
8 ECONOMIC PERSPECTIVES
deviation. If there were a 50 percent R2 at a the out-of-sample forecasts, I paired the regres-
daily horizon, one could make a lot of money; sions from table 4 with an autoregression of P/Dt,
but not so at a five-year horizon.
The slow movement of the P/D ratio also P Dt +1 = µ + ρ P D t + δ t +1 .
means that on a purely statistical basis, one can
cast doubt on whether the P/D ratio really Then, for example, since my data run
forecasts returns. What we really know, look- through the end of 1996, the forecast returns
ing at figure 2 (figure 4 also makes this point), for 1997 and 1998 are
is that low prices relative to dividends and
earnings in the 1950s preceded the boom mar-
E( R1997 ) = a + b( P D1996 )
ket of the early 1960s; that the high P/D ratios
of the mid-1960s preceded the poor returns of E( R1998 ) = a + b(µ + ρ P D1996 ),
the 1970s; and that the low price ratios of the
mid-1970s preceded the current boom. We also and so on.3
know that price ratios are very high now. In any The one-year return forecast is extraordi-
real sense, there really are three data points. I narily pessimistic. It starts at a –8 percent excess
do not want to survey the extensive statistical return for 1997, and only very slowly returns to
literature that formalizes this point, but it is the estimated unconditional mean excess return
there. Most importantly, it shows that the t-statis- of 8 percent. In ten years, the forecast is still –5
tics one might infer from regressions such as percent, in 25 years it is –1.75 percent, and it is
table 4 are inflated; with more sophisticated still only 2.35 percent in 50 years. The five-year
tests, return predictability actually has about a return forecasts are similarly pessimistic.
10 percent probability value before one starts Of course, this forecast is subject to lots
to worry about fishing and selection biases. of uncertainty. There is uncertainty about what
actual returns will be, given the forecasts. This
What about repurchases? P/E
and other forecasts will always be true: If one could precisely
Is the P/D ratio still a valid signal? Per- forecast the direction of stock prices, stocks
haps increasing dividend repurchases mean would cease to be risky and would cease to
that the P/D ratio will not return to its histori- pay a risk premium. There is also a great deal
cal low values; perhaps it has shifted to a new of uncertainty about the forecasts themselves.
mean so today’s high ratio is not
bad for returns. To address this
issue, figure 2 plots the S&P 500 P/E FIGURE 3
ratio along with the P/D ratio. The
Actual and forecast one-year excess returns
two measures line up well. The P/E
percent, one-year excess return
ratio forecasts returns almost as 50
well as the P/D ratio. The P/E ratio,
price/book value, and other ratios
are also at historic highs, forecasting
25
low returns for years to come. Yet
they are of course immune to the
criticism that the dividend–earnings
0
relationship might be fundamentally
different from the past.
Return forecasts -25
In-sample forecast
What do the regressions of table
Actual returns
4 say, quantitatively, about future Out-of-sample forecast
returns? Figure 3 presents one-year -50
returns and the P/D ratio forecast. 1950 ’60 ’70 ’80 ’90 2000 ’10 ’20
Figure 4 presents five-year returns Notes: One-year excess returns on the value weighted NYSE and
forecast from a regression on the P/D ratio. Returns are plotted on the
and the P/D ratio forecast. I include day of the forecast. For example, 1995 plots a + b × P/D 1995 and the
1996 return. The out-of-sample forecast is made by joining R t+1 = a + bP/Dt
in-sample and out-of-sample fore- with P/D t+1 = µ + ρP/D t.
10 ECONOMIC PERSPECTIVES
advice are now unemployed. Doesn’t this decline, or the P/D ratio must never return to its
mean that the P/D signal should no longer historical average. Which of the three options
be trusted? holds for our stock market?
To answer this criticism, look at the figures Historically, virtually all variation in P/D
again. They make clear that the returns for 1995 ratios has reflected varying expected returns.
and 1996 and even another 20 percent or so At a simple level, table 4 makes this point with
return for 1997 are not so far out of line, despite regressions of long-horizon dividend growth
a pessimistic P/D forecast, that we should throw on P/D ratios to match the regressions of returns
away the regression based on the previous 47 on P/D ratios. The dividend-growth coefficients
years of experience. To return to the analogy, are much smaller, typically one standard error
if it is ten degrees below zero in Chicago, that from zero, and the R2 values are tiny. Worse,
means spring is coming. But we can easily have the signs are wrong. To the extent that a high
a few weeks of 20 degree below weather before P/D ratio forecasts any change in dividends, it
spring finally arrives. The graphs make vivid seems to forecast a small decline in dividends.
how large a 17 percent standard deviation To be a little more precise, the identity
really is, and to what extent the forecasts based
on the P/D ratio mark long-term tendencies Pt +1 + Dt +1
+1 Rt +1 = Rt +1
1 = Rt–1 –1
that are still subject to lots of short-term swings Pt
rather than accurate forecasts of year-to-year
booms or crashes. yields, with a little algebra, the approximate
Another source of uncertainty about the identity
forecast is how persistent the P/D ratio really ∞
is. If, for example, the P/D ratio had no persis-
tence, then the low return forecast would only
1) p t − d t = const. + ∑ ρ j(∆ d t + j − r t + j) +
j =1
last a year. After that, it would return to the
unconditional mean of 9 percent (8 percent lim ρ j ( pt + j − d t + j ),
over Treasury bills). Now, given a true value j→∞
ρ = 0.98 in P/Dt = µ + ρP/D t-1 + δt, the median
ordinary least squares (OLS) estimate is 0.90, where ρ = P/D/(1 + P/D) is a constant of approxi-
as I found in sample. That is why figure 3 uses mation, slightly less than one and lowercase
the value ρ = 0.98. However, given this true letters denote logarithms (Campbell and Shiller,
value, the OLS estimate lies between 0.83 and 1988). Equation 1 gives a precise meaning to
0.94 only 50 percent of the time and between my earlier statement that a high P/D ratio must
0.66 and 1.00 for 95 percent of the time. Thus, be followed by high dividend growth ∆ d, low
there is a huge range of uncertainty over the true returns r, or a bubble.
value of ρ. The best thing that could happen to Bubbles do not appear to be the reason for
the forecast is if the P/D ratio were really less historical P/D ratio variation. Unless the P/D
persistent than it seems. In this case, the near- ratio grows faster than 1/ρ j, there is no bubble.
term return forecast would be unchanged, but It is hard to believe that P/D ratios can grow
the long-term return forecast would return to 9 forever. Empirically, P/D ratios do not seem
percent much more quickly. to have a trend or unit root over time. 5
This still leaves two possibilities: are high
Variance decomposition prices signals of high dividend growth or low
When prices are high relative to dividends returns? To address this issue, equation 1 implies6
(or earnings, cash flow, book value, or some
other divisor), one of three things must be ∞
true: 1) Investors expect dividends to rise in the 2) var ( pt − d t ) = cov( pt − d t , ∑ ρ j∆ d t + j ) −
future. 2) Investors expect returns to be low in the j =1
12 ECONOMIC PERSPECTIVES
economic growth and dynamic micro, macro, relative constancy of real risk-free rates over
and international economics for close to 25 time and across countries, and the relatively
years. Does a large equity premium make sense low correlation of stock returns with consump-
in terms of such standard economics? Did people tion growth. This is a tough assignment, which
in 1947, and throughout the period, know that is only now starting to be accomplished.
stocks were going to yield 8 percent over bonds Then I survey alternative views that do
on average, yet were rationally unwilling to promise to account for the equity premium,
hold more stocks because they were afraid without (so far) wildly counterfactual predic-
of the 17 percent standard deviation or some tions on other dimensions. Each modification
other measure of stocks’ risk? If so, we have is the culmination of a decade-long effort by a
“explained” the equity premium. If so, statis- large number of researchers. (For literature
tics from the past may well describe the future, reviews, see Kocherlakota, 1996, and Cochrane
since neither people’s preferences nor the riski- and Hansen, 1992). The first model maintains
ness of technological opportunities seems to the complete and frictionless market simplifi-
have changed dramatically. But what if it makes cation, but changes the specification of how
no sense that people should be so scared of people feel about consumption over time, by
stocks? In this case, it is much more likely that adding habit persistence in a very special way
the true premium is small, and the historical that produces a strong precautionary saving
returns were in fact just good luck. motive. The second model abandons the perfect
The answer is simple: Standard economic markets simplification. Here, uninsurable indi-
models utterly fail to produce anything like the vidual-level risks are the key to the equity pre-
historical average stock return or the variation mium. I will also discuss a part of an emerging
in expected returns over time. After ten years view that the equity premium and time-variation
of intense effort, there is a range of drastic of expected returns result from the fact that few
modifications to standard models that can people hold stocks. This view is not flushed
explain the equity premium and return predict- out yet to a satisfactory model, but does give
ability and (harder still) are not inconsistent some insight.
with a few obvious related facts about con- Both modifications answer the basic ques-
sumption and interest rates. However, these tion, “why are consumers so afraid of stocks?”
models are truly drastic modifications; they in a similar way, and give a fundamentally
fundamentally change the description of the different answer from the standard model’s
source of risk that commands a premium in view that expected returns are driven by risks
asset markets. Furthermore, they have not yet to wealth or consumption. The modifications
been tested against the broad range of experi- both say that consumers are really afraid of
ence of the standard models. These facts must stocks because stocks pay off poorly in reces-
mean one of two things. Either the standard sions. In one case a recession means a time
models are wrong and will change drastically, when consumption has recently fallen, no matter
or the phenomenon is wrong and will disappear. what its level. In the second case a recession is
I first show how the standard model utterly a time of unusually high cross-sectional (though
fails to account for the historical equity premium not aggregate) uncertainty. In both cases the
(Sharpe ratio). The natural response is to see raw risk to wealth is not a particularly impor-
if perhaps we can modify the standard model. tant part of the story.
I consider what happens if we simply allow a
The standard model
very high level of risk aversion. The answer
To say anything about dynamic economics,
here, as in many early attempts to modify the
we have to say something about how people
standard model, is unsatisfactory. While one
are willing to trade consumption in one moment
can explain the equity premium, easy explana-
and set of circumstances (state of nature) for
tions make strongly counterfactual predictions
consumption in another moment and set of
regarding other facts. The goal is to explain the
circumstances. For example, if people were
equity premium in a manner consistent with the
always willing to give up a dollar of consump-
level and volatility of consumption growth (both
tion today for $1.10 in a year, then the economy
about 1 percent per year), the predictability of
would feature a steady 10 percent interest rate.
stock returns described above, the relative lack
It also might have quite volatile consumption,
of predictability in consumption growth, the
14 ECONOMIC PERSPECTIVES
return is uncorrelated with consumption growth (1985), as reinterpreted by Hansen and Jagan-
corr(∆c,r) = 0. If it yields any more, the consumer nathan (1991). The failure is quantitative not
can buy just a little bit of the security, and come qualitative, as Kocherlakota (1996) points out.
out ahead because the risk is perfectly diversifi- Qualitatively, the right-hand side of equation 6
able. Readers familiar with the capital asset does predict a positive equity premium. The
pricing model (CAPM) will recognize the intu- problem is in the numbers. This is a strong
ition; replacing wealth or the market portfolio advertisement for quantitative rather than just
with consumption gives the most modern and qualitative economics.
general version of that theory.
Can we change the numbers?
The equity premium puzzle The correlation of consumption growth
To evaluate the equity premium, I transform with returns is the most suspicious ingredient
equation 5 to in this calculation. While the correlation is
undeniably low in the short run, a decade-long
rise in the stock market should certainly lead to
E( r ) − r f
6) = γ σ ( ∆c ) corr ( ∆c, r ). more consumption. In fact, the low correlation
σ( r ) is somewhat of a puzzle in itself: Standard (one-
shock) models typically predict correlations of
The left-hand side is the Sharpe ratio. As I 0.99 or more. Marshall and Daniel (1997) find
showed above, the (unconditional) Sharpe ratio correlations in the data up to 0.4 at a two-year
is about 0.5 for the stock market, and it is robust horizon, and by allowing lags. But even plugging
to leveraging or choice of assets. The right-hand in a correlation of corr(∆c,r) = 1, σ(∆c) = 0.01
side of equation 6 says something very impor- and γ < 10 implies a Sharpe ratio less than 0.1, or
tant. A high Sharpe ratio or risk premium must one-fifth the sample value.
be the result of 1) high aversion to risk, γ, or 2) A large literature has tried to explain the
lots of risk, σ(∆c). Furthermore, it can only occur equity premium puzzle by introducing frictions
for assets whose returns are correlated with the that make T-bills “money-like,” which artifi-
risks. This basic message will pervade the follow- cially drive down the interest rate (for example,
ing discussion of much-generalized economic Aiyagari and Gertler, 1991). The highest Sharpe
models. If the right-hand side of equation 6 is ratio occurs in fact when one considers short-
low, then the consumer should invest more in term risk-free debt and money, since the latter
the asset with return r. Doing so will make the pays no interest. Perhaps the same mechanism
consumption stream more risky and more corre- can be invoked for the spread between stocks
lated with the asset return. Thus, as the consumer and bonds. However, a glance at figure 1 shows
invests more, the right-hand side of equation 6 that this will not work. High Sharpe ratios are
will approach the left-hand side. pervasive in financial markets. One can recover
The right-hand side of equation 6 is a pre- a high Sharpe ratio from stocks alone, or from
diction of what the Sharpe ratio should be. It stocks less long-term bonds.
does not come close to predicting the historical
equity premium. The standard deviation of Time-varying expected returns
aggregate consumption growth is about 1 percent The consumption-based view with
or 0.01. The correlation of consumption growth u′(C) = C-γ also has trouble explaining the fact
with stock returns is a bit harder to measure that P/D ratios forecast stock returns. Consider
since it depends on horizon and timing issues. the conditional version of equation 6
Still, for horizons of a year or so, 0.2 is a pretty
generous number. γ 1 or 2 is standard; γ = 10 Et ( r ) − rt f
is a very generous value. Putting this all together, 7) = γ σt ( ∆c) corr t ( ∆c, r),
σ t (r )
10 × 0.01 × 0.2 = 0.02 rather than 0.5. At a
20 percent standard deviation, a 0.02 Sharpe
ratio implies an average excess return for where Et, σt , corrt represent conditional moments.
stocks of 0.02 × 20 = 0.4% (40 basis points) I showed above that the P/D ratio gives a strong
rather than 8 percent. signal about mean returns, Et(r). It does not
This devastating calculation is the celebrated however give much information about the
“equity premium puzzle” of Mehra and Prescott standard deviation of returns. Figure 5 does
suggest a slight increase in return standard
16 ECONOMIC PERSPECTIVES
intermediaries 48 percent to keep their money. bet to avoid taking it. For example, in the low-
This just does not happen. er right-hand corner, the family with γ = 250
This observation can also be phrased as would rather pay $9,889 for sure than take a
a conceptual experiment, suitable for thinking 50 percent chance of a $10,000 loss. This predic-
about one’s own preferences or for survey tion is surely unreasonable, and has led most
evidence on others’ preferences. For example, authors to rule out risk aversion coefficients
what does it take to convince someone to skip over ten. Survey evidence for this kind of bet
a vacation? Take a family with $50,000 per year also finds low risk aversion, certainly below γ = 5
income, consumption equal to income, which (Barsky, Kimball, Juster, and Shapiro, 1997), and
spends $2,500 (5 percent) on an annual vaca- even negative risk aversion if the survey is taken
tion. If interest rates are good enough, though, in Las Vegas.
the family can be persuaded to skip this year’s Yet the results for small bets are not so unrea-
vacation and go on two vacations next year. sonable. The family might reasonably pay 5 cents
What interest rate does it take to persuade the to 25 cents to avoid a $10 bet. We are all risk
family to do this? The answer is ($52,500/ neutral for small enough bets. For small bets,
$47,500)γ – 1. For γ = 250 that is an interest
rate of 3 × 1011. For γ = 50, we still need an amount willing to pay to avoid bet size of bet
interest rate of 14,800 percent. I think most of ≈γ .
size of bet consumption
us would defer the vacation for somewhat
lower interest rates.
The standard use of low values for γ in Thus, for any γ, the amount one is willing
macroeconomics is also important for delivering to pay is an arbitrarily small fraction of the bet
realistic quantity dynamics in macroeconomic for small enough bets. For this reason, it is
models, including relative variances of invest- easy to cook numbers of conceptual experi-
ment and output, and for delivering reasonable ments like table 6 by varying the size of the bet
speeds of adjustment to shocks. and the presumed wealth of the family. Signifi-
cantly, I only used local curvature above; γ
Risk aversion represented the derivative γ = –Cu″(C)/u′(C).
Economists have also shied away from high In asking how much the family would pay to
curvature γ on the basis that people do not avoid a $10,000 bet, we are asking for the
seem that risk averse. After examining the response to a very, very non-local event.
argument, I conclude that there are fewer solid The main lesson of conceptual experiments
reasons to object to high risk aversion than to and laboratory and survey evidence of simple
object to high aversion to intertemporal substi- bets is that people’s answers to such questions
tution via the consumption-interest rate relation- routinely violate expected utility. This observa-
ships I examined above. tion lowers the value of this source of evidence
as a measurement of risk aversion. As a similar
Surveys and thought experiments cautionary note, Barsky et al. report that whether
Since Sharpe ratios are high for many an individual partakes in a wide variety of
assets, much analysis of risk aversion comes risky activities correlates poorly with the level
from simple thought experiments
rather than data. For example, how
much would a family pay per year TABLE 6
to avoid a bet that led with equal Amount family would pay to avoid an even bet
probability to a $y increase or de-
Risk aversion γ
crease in annual consumption for
the rest of their lives? Table 6 Bet 2 10 50 100 250
presents some calculations of how
$10 $0.00 $0.01 $0.05 $0.10 $0.25
much our family with $50,000 per
100 0.20 1.00 4.99 9.94 24
year of income and consumption
1,000 20 99 435 665 863
would pay to avoid various bets of
10,000 2,000 6,921 9,430 9,718 9,889
this form.8 For bets that are reason-
ably large relative to wealth, high γ Notes: I assume the family has a constant $50,000 per year
consumption and an even chance of winning or losing the
means that families are willing to indicated net, per year.
pay almost the entire amount of the
18 ECONOMIC PERSPECTIVES
New utility functions and state variables bad a representation of risk aversion. Maybe
If changing the parameter γ in u′(C) = C-γ a modification of preferences can disentangle
does not work, perhaps we need to change the the two attitudes.
functional form. Changing the form of u(C)
is not a promising avenue. As I have stressed State separability and leisure
by using a continuous time derivation, only the With the latter end in mind, Epstein and Zin
derivatives of u(C) really matter; hence quite (1989) started an avalanche of academic research
similar results are achieved with other functional on utility functions that relax state-separability.
forms. A more promising avenue is to consider The expectation E in the utility function of equa-
other arguments of the utility function, or tion 3 sums over states of nature, for example
nonseparabilities.
Perhaps how people feel about eating more U = prob(rain) × u(C if it rains) + prob(shine) ×
today is affected not just by how much they are u(C if it shines).
already eating, but by other things, such as how
much they ate yesterday or how much they “Separability” means that one adds utility
worked today. Then, the covariance of stock across states, so the marginal utility of con-
returns with these other variables will also sumption in one state is unaffected by what
determine the equity premium. Fundamentally, happens in another state. But perhaps the mar-
consumers use assets to smooth marginal utility. ginal utility of a little more consumption in the
Perhaps today’s marginal utility is related to sunny state of the world is affected by the level
more than just today’s consumption. of consumption in the rainy state of the world.
Such a modification is a fundamental change Epstein and Zin and Hansen, Sargent, and
in how we view stock market risk. For example, Tallarini (1997) propose recursive utility func-
perhaps more leisure raises the marginal utility tions of the form
of consumption. Stocks are then feared because
they pay off badly in recessions when employ-
ment is lower and leisure is higher, not because Ut = Ct1− γ + e − ρ f Et f −1 (Ut +1 ) .
consumers are particularly averse to the risk
that stocks decline per se. Formally, our funda- If f(x) = x, this expression reduces to power
mental equation 6 is derived from utility. These utility functions are not state-
separable, and do conveniently distinguish risk
Et (r ) − rt f = covt ( ∆uc , r ), aversion from intertemporal substitution among
other modifications. However, this research is
and substituting uc = ∂u(c)/∂c. If I substitute uc only starting to pay off in terms of plausible
= ∂u(c,x)/∂c instead, then uc will depend on models that explain the facts (Campbell, 1996,
other variables x as well as c, and is an example) so I will not review it here.
Perhaps leisure is the most natural extra
ucx variable to add to a utility function. It is not
E(r ) − r f = γ cov( ∆c, r ) + cov( x , r ). clear a priori whether more leisure enhances
uc
the marginal utility of consumption (why bother
buying a boat if you are at the office all day
Since the first covariance does not account for and cannot use it?) or vice versa (if you have to
much premium, we will have to rely heavily on work all day, it is more important to come
the latter term to explain the premium. home to a really nice big TV), but we can let
There is a practical aim to generalizing the the data speak on this matter. However, explicit
utility function as well. As illustrated in the versions of this approach have not been very
last section, one parameter γ did two things successful to date. (Eichenbaum, Hansen, and
with power utility: It controlled how much Singleton, 1988, for example). On the other
people are willing to substitute consumption hand, recent research has found that adding
over time (consumption and interest rates) and labor income growth as an extra ad-hoc factor
it controlled their attitudes toward risk. The can be useful in explaining the cross section of
choice γ = 50 to 250 was clearly a crazy repre- average stock returns (Jagannathan and Wang,
sentation of how people feel about consump- 1996; Reyfman, 1997). Though not motivated
tion variation over time, but perhaps not so
20 ECONOMIC PERSPECTIVES
in the artificial economy and subject them to Habit models with low risk aversion
the standard statistical analysis reviewed above. The individuals in the Campbell-Cochrane
The artificial data replicate the equity premium model are highly risk averse. They would respond
(0.5 Sharpe ratio). The ability to forecast returns to surveys about bets on wealth much as the
from the P/D ratio and the P/D variance decom- γ = 50 column of table 6. The model does not
position are both quite like the actual data. The give rise to a high equity premium with low
standard deviation of returns rises a bit when risk aversion; it merely disentangles risk aver-
prices decline, but less than the rise in mean sion and intertemporal substitution so that a
returns, so a low P/D ratio forecasts a higher high risk aversion economy can be consistent
Sharpe ratio. Artificial data from the model with low and constant interest rates, and it
also replicate much of the low observed correla- generates the predictability of stock returns.
tion between consumption growth and returns, Constantinides (1990) and Boldrin,
and the CAPM and ad-hoc multifactor models Christiano, and Fisher (1995) explore habit
perform better than the power utility consump- persistence models that can generate a large
tion-based model in the artificial data. equity premium without large risk aversion.
The model also provides a good account That is, they create artificial economies in
of P/D fluctuations over the last century, based which consumers simultaneously shy away
entirely on the history of consumption. Howev- from stocks with a very attractive Sharpe ratio
er, it does not account for the currently high of 0.5, yet would happily take bets with much
P/D ratio. This is because the model generates lower rewards.
a high P/D ratio when consumption is very Suppose a consumer wins a bet or enjoys
high relative to habit and, therefore, risk aver- a high stock return. Normally, the consumer
sion is low. Measured consumption has been would instantly raise consumption to match
increasing unusually slowly in the 1990s. the new higher wealth level. But consumption
Like other models that explain the equity is addictive in these models: Too much current
premium and return predictability, this one consumption will raise the future habit level
does so by fundamentally changing the story and blunt the enjoyment of future consumption.
of why consumers are afraid of holding stocks. Therefore, the consumer increases consumption
From equation 9, the marginal utility of con- slowly and predictably after the increase in
sumption is proportional to wealth. Similarly, the consumer would borrow
to slowly decrease consumption after a decline
−η in wealth, avoiding the pain of a sudden loss at
uc = Ct− η
FG C − X IJ
t t
. the cost of lower long-term consumption.
H C K t The fact that the consumer will choose to
spread out the consumption response to wealth
shocks means that the consumer is not averse
Thus, consumers dislike low consumption
to wealth bets. If consumption responds little
as before, but they are also afraid of recessions,
to a wealth shock, then marginal utility of
times when consumption, whatever its level, is
consumption, uc(C), also responds little, as does
low relative to the recent past as described by
the marginal value of wealth, VW (W,⋅) = uc. Risk
habits. Consumers are afraid of holding stocks
aversion to wealth bets is measured by the
not because they fear the wealth or consumption
change of marginal utility when wealth changes
volatility per se, but because bad stock returns
(∂ln VW /∂lnW = –WVWW /VW).
tend to happen in recessions, times of a recent
The argument is correct, but shows the
belt-tightening.
problem with these models. The change in
This model fulfills a decade-long search
consumption in response to wealth is not elimi-
kicked off by Mehra and Prescott (1985). It is
nated, it is simply deferred. Thus, these models
a complete-markets, frictionless economy that
have trouble with long-run behavior of con-
replicates not only the equity premium but also
sumption and asset returns.
the predictability of returns, the nearly constant
If consumption growth is considered inde-
interest rate, and the near-random walk behavior
pendent over time (formally, an endowment
of consumption.
economy), which is a good approximation to
the data, the model must feature strong interest
22 ECONOMIC PERSPECTIVES
Heterogeneous agents and the much more variable purchases of durable
idiosyncratic risks goods, such as cars and houses.
In the above discussion, I did not recog- More fundamentally, the addition of idio-
nize any difference between people. Everyone syncratic risk lowers the correlation between
is different, so why bother looking at represen- consumption growth and returns, which lowers
tative agent or complete market models? While the predicted Sharpe ratio. Idiosyncratic risk is,
making an assumption such as “all people are by its nature, idiosyncratic. If it happened to
identical” seems obviously foolish, it is not everyone, it would be aggregate risk. Idiosyn-
foolish to hope that we can use aggregate cratic risk cannot therefore be correlated with
behavior to make sense of aggregate data, the stock market, since the stock market return
without explicitly taking account of the differ- is the same for everyone.
ences between people. While differences are For a quantitative example, suppose that
there, one hopes they are not relevant to the individual consumption of family i, ∆c i, is deter-
basic story. However, seeing the difficulties mined by aggregate consumption, ∆c a, and idio-
that representative agent models face, perhaps syncratic shocks (such as losing your job), ε i,
it is time to see if the (aggregate) equity premium
does in fact surface from differences between
∆c i = ∆c a + ε i .
people rather than common behavior.
The empirical hurdle For the risk ε i to average to zero across people,
Idiosyncratic risk explanations face a big we must have E(ε i ) = 0 and E(ε i∆ca) = E(εir)
empirical challenge. Look again at the basic = 0. Then, the standard deviation of individual
Sharpe ratio equation 6, consumption growth does increase with the
size of idiosyncratic risk,
E( r ) − r f
= γ σ(∆c) corr ( ∆c, r ).
σ( r ) σ 2 (∆c i ) = σ 2 ( ∆c a ) + σ 2 (ε i ).
This relationship should hold for every (any) But the correlation between individual con-
consumer or household. At first sight, thinking sumption growth and aggregate returns de-
about individuals seems promising. After all, clines in exact proportion as the standard devi-
individual consumption is certainly more vari- ation σ(∆ci) rises,
able than aggregate consumption at 1 percent
per year, so we can raise σ(∆c). However, this E (r ) − r f cov ( ∆c a + ε i , r ) cov( ∆c a , r )
argument fails quantitatively. First, it is incon- =γ =γ .
σ( r ) σ( r ) σ( r )
ceivable that we can raise σ(∆c) enough to
account for the equity premium. For example,
even if individual consumption has a standard Therefore, the equity premium is completely
deviation of 10 percent per year, and maintain- unaffected by idiosyncratic risk.
ing a generous limit γ < 10, we still predict a The theoretical hurdles
Sharpe ratio no more than 10 × 0.1 × 0.2 = 0.2. The theoretical challenge to idiosyncratic
To explain the 0.5 Sharpe ratio with risk aver- risk explanations is even more severe. We can
sion γ = 10, we have to believe that individual easily construct models in which consumers
consumption growth has a 25 percent per year are given lots of idiosyncratic income risk.
standard deviation; for a more traditional γ = 2.5, But it is very hard to keep consumers from
we need 100 percent per year standard devia- insuring themselves against those risks, pro-
tion. Even 10 percent per year is a huge standard ducing a very steady consumption stream and
deviation of consumption growth. Remember, a low equity premium.
we are considering the risky or uncertain part of Start by handing out income to consum-
consumption growth. Predictable increases or ers; call it “labor income” and make it risky
decreases in consumption due to age and life- by adding a chance of being fired. Left to
cycle effects, expected raises, and so on do not their own devices, consumers would come
count. We are also thinking of the flow of up with unemployment insurance to share
consumption (nondurable goods, services) not this risk, so we have to close down or limit
markets for labor income insurance. Then,
24 ECONOMIC PERSPECTIVES
where m t is a strictly positive discount factor13
F C I=η y −1y
10) ln G it 2 that prices all assets under consideration. That
H C JK
i ,t −12
it t t ,
is, mt satisfies
L F C I R OP LM
δit = δ it −1 exp ηit yt −
1 2
yt . OP
11) y = σ Mln G it N 2 Q
t
MN H C JK PQ
it −1
t
26 ECONOMIC PERSPECTIVES
them? So far, I have tried to find representations If these segmented market views of the
of people’s preferences or circumstances, or a past equity premium are correct, they suggest
description of macroeconomic risk, in which that the future equity premium will be much
stocks aren’t that attractive after all. Then the lower. Transaction costs are declining through
high Sharpe ratio is a compensation for risk. financial deregulation and innovation. The
Instead, we could argue that stocks really explosion in tax-deferred pension plans and
are attractive, but a variety of market frictions no-load mutual funds means more and more
keep people from buying them. This approach people own stocks, spreading risks more evenly,
yields some important insights. First of all, driving up prices, and driving down prospective
stock ownership has been quite concentrated. returns. Equation 6 will hold much better for the
The vast majority of American households have average consumer in the future. One would expect
not directly owned any stock or mutual funds. to see a lower equity premium. One would also
One might ask whether the consumption of expect consumption that is more volatile and
people who do own stock lines up with stock more closely correlated with the stock market,
returns. Mankiw and Zeldes (1991) find that which will result in a fundamental change in the
stockholders do have consumption that is more nature and politics of business cycles.
volatile and more correlated with stock returns
Technology and investment
than non-stockholders. But it is still not volatile
and correlated enough to satisfy the right-hand So far, I have tried to rationalize stock
side of equation 6 with low risk aversion. returns from the consumer’s point of view:
Heaton and Lucas (1996b) look at individual Does it make sense that consumers should not
asset and income data. They find that the richest have tried to buy more stocks, driving stock
households, who own most of the stocks, also returns down toward bond returns? I can ask
get most of their income from proprietary the same questions for the firm: Do firms’
business income. This income is likely to be investment decisions line up with stock prices
more correlated with the stock market than is as they should?
individual labor income. Furthermore, they The relative prices of apples and oranges
find that among rich households, those with are basically set by technology, the relative
more proprietary income hold fewer stocks number of apples versus oranges that can be
in their portfolios. This paints an interesting grown on the same acre of land. We do not
picture of the equity premium: In the past need a deep understanding of consumers’ de-
most stocks were held by rich people, and sires to figure out what the price should be. If
most rich people were proprietors whose other technology is (close to) linear, it will determine
income (and consumption) was quite volatile relative prices, while preferences will deter-
and covaried strongly with the market. This is mine quantities. Does this argument work for
a hard crowd to sell stocks to, so they have stocks?
required a high risk premium. The Campbell- Again, there is a standard model that has
Cochrane and Constantinides-Duffie models served well to describe quantities in growth,
specify that stock market risk is spread as macroeconomics, and international economics.
evenly as possible through the population, The standard model consists of a production
whereas if the risk is shared among a small function by which output, Y, is made from
group of people, higher rewards will have to capital, K, and labor, L, perhaps with some
be offered to offset that risk. uncertainty, θ, together with an accumulation
These views are still not sorted out quanti- equation by which investment I turns into new
tatively. We don’t know why rich stockholders capital in the future. In equations, together
don’t buy even more stocks, given low risk with the most common functional forms,
aversion and the tyrannical logic of equation 6.
We don’t know why only rich people held 15) Yt = f ( Kt , Lt , θt ) = θt Ktα L1t −α
stocks in the first place: The literature shows Kt +1 = (1 − δ) Kt + It
that even quite high transactions costs and
borrowing constraints should not be enough Yt = Ct + It .
to deter people with low risk aversion from
holding stocks.
28 ECONOMIC PERSPECTIVES
well. The longer-term variation in P/D is not that surrounds the size of the risk aversion
mirrored in investment: This simple model coefficient, γ. From equation 19 and the fact
does not explain why investment stayed robust that investment growth has standard deviation
in the late 1970s despite dismal stock prices. of about 10 percent, a ≈ 2 is needed to ratio-
However, the recent surge in the market is nalize the roughly 20 percent standard devia-
matched by a surge in investment. tion of stock returns. With I/Y ≈ 15% and
This kind of model has been subject to an Y/K ≈ 33%, and hence I/K ≈ 1/20, a value a ≈ 2
enormous formal empirical effort, which pretty means that adjustment costs relative to output are
much confirms the figure. First, the model is a I I 2 1
= F I
consistent with a good deal of the cyclical varia- 2 K Y 2 20 H K x 15% = 0.75%, which does
tion in investment and stock returns, both fore- not seem unreasonable. However, estimates of
casts and ex-post. (See, for example, Cochrane, a based on regressions, Euler equations, or
1991c.) It does not do well with longer-term other techniques often result in much higher
trends in the P/D ratio. Second, early tests values, implying that implausibly large frac-
relating investment to interest rates that imposed tions of output are lost to adjustment costs.
a constant risk premium did not work (Abel, This model does not yet satisfy the goal of
1983). The model only works at all if one rec- determining the equity premium by technological
ognizes that most variation in the cost of capi- considerations alone. Current specifications of
tal comes from time-varying expected stock technology allow firms to transform resources
returns with relatively constant interest rates. over time but not across states of nature. If the
Third, the model in equation 18 taken literally firm’s own stock is undervalued, it can issue
allows no residual. If prices deviate one iota more and invest. However, if the interest rate is
from the right-hand side of equation 18, then low, there is not much one can say about what
the model is statistically rejected—we can say the firm should do without thinking about the
with perfect certainty that it is not a literal descrip- price of residual risk, and hence a preference
tion of the data-generating mechanism. There is approach to the equity premium. Technically,
a residual in actual data of course, and the resid- the marginal rate of transformation across
ual can be correlated with other variables such states of nature is undefined.
as cash flow that suggest the presence of finan-
cial frictions (Fazzari, Hubbard, and Peterson, Implications of the recent surge in investment
and stock prices
1988). Finally, the size of the adjustment cost, a,
The association between stock returns and
is the subject of the same kind of controversy
investment in figure 7 verifies that at least one
connection between stock re-
turns and the real economy
FIGURE 7 works in some respects as it
Value weighted portfolio P/D and investment should. This argues against the
view that stock market swings
are due entirely to waves of
Investment-output irrational optimism and pessi-
ratio mism. It also verifies that the
flow of money into the stock
market does at least partially
correspond to new real assets
and not just price increases on
existing assets.
In particular, the surge in
stock prices since 1990 has been
Price/dividend accompanied by a surge in invest-
ratio ment. If expected stock returns
and the cost of capital are low,
’62 ’69 ’76 ’83 ’90 ’97 then investment should be high.
Notes: Investment-output ratio and P/D of value weighted NYSE. Statistically, high investment-
Investment = gross fixed investment, output = gross domestic output.
Series are stretched to fit on the same graph. output or investment-capital
30 ECONOMIC PERSPECTIVES
in this model; linearizations near the steady In the economic models that generate the
state and dynamics with the shocks turned off equity premium, every investor is exactly happy
give dramatically wrong predictions about the to hold his or her share of the market portfolio,
model’s behavior. The costs of business cycles no more and no less. The point of the models is
are orders of magnitude larger than in standard that the superficial attractiveness of stocks is
models. In the Constantinides-Duffie model, balanced by a well-described source of risk, so
one has to explicitly keep track of micro- that people are just willing to hold them. Simi-
economic heterogeneity in order to say anything larly, the time variation in the equity premium
about aggregates. does not necessarily mean one should attempt
The new models are also a drastic revision to time the market, buying more stocks at times
to finance. We are used to thinking of aversion of high expected returns and vice versa. Every
to wealth risk, as in the CAPM, as a good start- investor in the Campbell-Cochrane model, for
ing place or first-order approximation. But this example, holds exactly the same portfolio all
view cannot hold. To justify the equity premium, the time, while buy and sell signals come and
people must be primarily averse to holding go. In the peak of a boom they are not feeling
stocks because of their exposure to some other very risk averse, and put their money in the
state variable or risk factor, such as recessions market despite its low expected returns. In the
or changes in the investment opportunity set. bottom of a bust, they feel very risk averse, but
To believe in the equity premium, one has to the high expected returns are just enough to
believe that these stories are sensible. keep their money in the market.
Finally, every quantitatively successful To rationalize active portfolio strategies,
current story for the equity premium still requires such as pulling out of the market at times of
astonishingly high risk aversion. The alternative, high price ratios, you have to ask, who is there
of course, is that the long-run equity premium who is going to be more in the market than
is much smaller than the average postwar 8 per- average now? And, what else are you going to
cent excess return. The standard model was do with the money?
right after all, and historically high U.S. stock More formally, it is easy to crank out
returns were largely due to luck or some other portfolio advice, solutions to optimal portfolio
transient phenomenon. problems given objectives like the utility func-
Faced with the great difficulty economic tion in equation 3. Assuming low risk aversion,
theory still has in digesting the equity premium, and no labor income or other reason for time-
I think the wise observer shades down the varying risk exposure or risk aversion, solutions
estimate of the future equity premium even typically suggest large portfolio shares in equi-
more than suggested by the statistical uncer- ties and a strong market timing approach, some-
tainty documented above. times highly leveraged and sometimes (now)
even short. (See Barberis, 1997; Brandt, 1997.)
Portfolio implications
If everyone followed this advice, however, the
In sum, the long-term average stock return
equity premium and the predictable variation in
may well be lower than the postwar 8 percent
expected returns would disappear. Everyone
average over bonds, and currently high prices
trying to buy stocks would simply drive up the
are a likely signal of unusually low expected
prices; everyone trying to time the market
returns. It is tempting to take a sell recommen-
would stabilize prices. Thus, the majority of
dation from this conclusion. There is one very
investors must be solving a different problem,
important caution to such a recommendation.
deciding on their portfolios with different
On average, everyone has to hold the market
considerations in mind, so that they are always
portfolio. The average person does not change
just willing to hold the outstanding stocks and
his or her portfolio at all. For every individual
bonds at current prices. Before going against
who keeps money out of stocks, someone else
this crowd, it is wise to understand why the
must have a very long position in stocks. Prices
crowd seems headed in a different direction.
adjust until this is the case. Thus, one should
Here, a good macroeconomic model of
only hold less stocks than the average person if
stock market risk could be extremely useful.
one is different from everyone else in some
The models describe why average consumers
crucial way. It is not enough to be bearish, one
are so afraid of stocks and why that fear changes
must be more bearish than everyone else.
over time. Then, individuals in circumstances
APPENDIX: DERIVATIONS
lim
j →∞
F∏ R I P
GH
j
JK D −1
t+k
t+ j
. z
u ′(Ct ) Pt = Et e u ′( Ct + s ) Dt + s ds +
k =1 t+ j Et e − ρk u ′(Ct + k ) Pt + k .
∞
pt − dt = ∑ ρ j (∆dt + j − rt + j ) +
0 = Et F dP I + D dt + E F dΛ I + E LM dΛ dP OP. t
j =1 H PK P H ΛK NΛ P Q t t
t
lim ρ ( pt + j − dt + j ) .
j
j →∞
32 ECONOMIC PERSPECTIVES
1
Applying this basic condition to a risk-
free asset,
∑ δ j u(C − x ) = 2 ∑ δ j u(C + y) +
j j
1
f L dΛ O
r dt = − E M P = ρdt − E M
L du′(C) OP = ∑
2 j
δ j u(C − y) .
t
N QΛ
t
N u′(C) Q t
x = C − LM (C + y ) OP
1 1− γ 1 1− γ
+ (C − y )1− γ
0 = Et F dP I + D dt − r dt = − E LM dΛ dP OP.
f t N2 2 Q .
H PK P t
NΛ P Q t
t
Et
LM dΛ dP OP = Cu′′(C) E F dC dP I .
t
wealth risk by
N Λ P Q u ′( C ) H C P K
t
t
1
x
= 1−
1 LM F
1+
y I 1− γ
1
+ 1−
yF I OP
1− γ 1− γ
.
Finally, using the symbols C 2 MN H
C K 2 CH K PQ
dP D
r= + dt, r f = r f dt, This equation is the basis for the calcula-
P P tions in table 5.
−Cu′′(C) dC For small risks, we can approximate
γ= , ∆c =
u ′(C) C
1
we have equation 5, u(C − x ) = u(C + y) + u(C − y)
2
1
Et (r ) − r f = − γ covt ∆c, r = − u ′ (C ) x ≈ u′′(C) y 2
2
a f
− γ σ t ∆c σ t (r )ρt ( ∆c, r ) . x
≈
−Cu ′′(C ) y F I 2
=γ F yI 2
C u ′( C ) C H K H CK
I drop the t subscript in the text where it 2
is not important to keep track of the difference
x
≈ γ
y F I
between conditional and unconditional moments. C C H K
γF I .
x y
Risk aversion calculations ≈
What is the amount x that a consumer is
y H CK
willing to pay every period to avoid a bet that
either increases consumption by y every period
or decreases it by the same amount? The an-
swer is found from the condition
34 ECONOMIC PERSPECTIVES
LM 1 γ y + 1 γ y + ln R OP
1 = Et −1 exp −ρ + 2 2 2
t +1
We can sometimes construct such discount factors by
N 2 t
2
t
Q picking parameters a, b in mt = max [a + bRt, e−ρ (Ct /Ct–1)−γ ]
L 1 OP to satisfy equation 14. However, neither this construction
1 = Et −1
F 2 IJ F ln 1 + ρI + ln R
exp M −ρ + γ ( γ + 1)G
MN 2 H γ (γ + 1) K GH R JK
t
t +1
PQ
nor a discount factor satisfying equation 13 is guaranteed
to exist for a given set of assets. The restriction in equation
1 = Et −11. 13 is a tighter form of the familiar restriction that mt ≥ 0 is
equivalent to the absence of arbitrage in the assets under
consideration. Presumably, this restriction is what rules
13
There is a possibility that the square root term in equa- out markets for individual labor income risks in the
tions 11 and 12 might be negative. Constantinides and model. The example m = 1/R that I use is a positive dis-
Duffie rule out this possibility by assuming that the count factor that prices a single asset return 1 = E(R–1R),
discount factor m satisfies but does not necessarily satisfy the restriction in equation
13. For high R, we can have very negative ln1/R. This
Ct is why the lines in figure 6 run into the horizontal axis
13) ln mt ≥ −ρ − γ ln
Ct −1 at high R.
REFERENCES
Abel, Andrew B., “Optimal investment under Black, Fischer, “Studies of stock price volatility
uncertainty,” American Economic Review, Vol. changes,” Proceedings of the 1976 Meetings of the
73, No. 1, March 1983, pp. 228–233. Business and Economic Statistics Section, Ameri-
can Statistical Association, 1976, pp. 177–181.
, “Asset prices under habit formation
and catching up with the Joneses,” American Boldrin, Michele, Lawrence J. Christiano, and
Economic Review Papers and Proceedings, Vol. Jonas D.M. Fisher, “Asset pricing lessons for
80, No. 2, May 1990, pp. 38–42. modeling business cycles,” NBER, working
paper, No. 5262, 1995.
Abel, Andrew B., and Janice C. Eberly, “Opti-
mal investment with costly reversibility,” Review Brandt, Michael, “Estimating portfolio and
of Economic Studies, Vol. 63, 1996, pp. 581–594. consumption choice: A conditional Euler equa-
tions approach,” University of Chicago, Ph.D.
Aiyagari, S. Rao, and Mark Gertler, “Asset dissertation, 1997.
returns with transactions costs and undiversifi-
able individual risk,” Journal of Monetary Eco- Brown, Stephen J., William N. Goetzmann,
nomics, Vol. 27, No. 3, June 1991, pp. 311–331. and Stephen A. Ross, “Survival,” Journal of
Finance, Vol. 50, No. 3, 1995, pp. 853–873
Barberis, Nicholas, “Investing for the long run
when returns are predictable,” University of Campbell, John Y., “Stock returns and the term
Chicago, Center for Research on Security Prices, structure,” Journal of Financial Economics, Vol.
working paper, No. 439, 1997. 18, 1987, pp. 373–399.
Barsky, Robert B., Miles S. Kimball, F. Thomas , “Understanding risk and return,”
Juster, and Matthew D. Shapiro, “Preference Journal of Political Economy, Vol. 104, April
parameters and behavioral heterogeneity: An 1996, pp. 298–345.
experimental approach in the health and retire-
ment survey,” National Bureau of Economic Campbell, John Y., and John H. Cochrane, “By
Research, working paper, No. 5213, 1997. force of habit: A consumption-based explanation
of aggregate stock market behavior,” University of
Bernardo, Antonio, and Olivier Ledoit, “Gain, Chicago, working paper, 1997. [Revision of Na-
loss and asset pricing,” University of California tional Bureau of Economic Research, working
at Los Angeles, Anderson Graduate School of paper, No. 4995, dated January 1, 1995.]
Management, working paper, No. 16-96, 1996.
Constantinides, George M., and Darrell Duffie, Ferson, Wayne E., and George Constantinides,
“Asset pricing with heterogeneous consumers,” “Habit persistence and durability in aggregate
Journal of Political Economy, Vol. 104, April consumption: Empirical tests,” Journal of Financial
1996, pp. 219–240. Economics, Vol. 29, October 1991, pp. 199–240.
Craine, Roger, “Rational bubbles: A test,” Jour- Friend, I., and M. Blume, “The demand for
nal of Economic Dynamics and Control, Vol. 17, risky assets,” American Economic Review, Vol.
1993, pp. 829–846. 55, 1975, pp. 900–922.
Deaton, Angus S., Understanding Consumption, Hansen, Lars Peter, and Ravi Jagannathan,
New York: Oxford University Press, 1992. “Restrictions on intertemporal marginal rates of
substitution implied by asset returns,” Journal of
Deaton, Angus, and Christina Paxson, Political Economy, Vol. 99, 1991, pp. 225–262.
“Intertemporal choice and inequality,” Journal
of Political Economy, Vol. 102, June 1994, Hansen, Lars Peter, Thomas J. Sargent, and
pp. 437–467. Thomas D. Tallarini, “Robust permanent in-
come and pricing” University of Chicago, manu-
Duesenberry, James S., Income, Saving, and the script, 1997.
Theory of Consumer Behavior, Cambridge, MA:
Harvard University Press, 1949.
36 ECONOMIC PERSPECTIVES
Heaton, John C., “An empirical investigation of Deepening the equity premium puzzle,” Journal
asset pricing with temporally dependent prefer- of Monetary Economics, Vol. 34, December
ence specifications,” Econometrica, Vol. 63, 1994, pp. 325–341.
May 1995, pp. 681–717.
Mankiw, N. Gregory, “The equity premium and
Heaton, John, and Deborah J. Lucas, “Evaluat- the concentration of aggregate shocks,” Journal
ing the effects of incomplete markets on risk of Financial Economics, Vol. 17, September
sharing and asset pricing,” Journal of Political 1986, pp. 211–219.
Economy, Vol. 104, June 1996a, pp. 443–487.
Mankiw, N. Gregory, and Stephen P. Zeldes,
, “Saving behavior and portfolio “The consumption of stockholders and non-
choice: Which risks matter?,” Northwestern stockholders,” Journal of Financial Economics,
University, working paper, 1996b. Vol. 29, March 1991, pp. 97–112.
Hodrick, Robert, “Dividend yields and expected Marshall, David, and Kent Daniel, “The equity
stock returns: Alternative procedures for infer- premium puzzle and the risk-free rate puzzle at
ence and measurement,” Review of Financial long horizons,” Macroeconomic Dynamics, Vol.
Studies, Vol. 5, No. 3, 1992, pp. 357–386. 1, No. 1, 1997 forthcoming, pp. 452–484.
Jagannathan, Ravi, and Z. Wang, “The condi- Mehra, Rajnish, and Edward Prescott, “The
tional CAPM and the cross-section of expected equity premium puzzle,” Journal of Monetary
returns,” Journal of Finance, Vol. 51, March Economics, Vol. 15, March 1985, pp. 145–161.
1996, pp. 3–53. [Updated in Staff Report, No. 208,
Federal Reserve Bank of Minneapolis, 1997.] Merton, Robert C., “An intertemporal capital
asset pricing model,” Econometrica, Vol. 41,
Jermann, Urban, “Asset pricing in production September 1973, pp. 867–887.
economies,” Journal of Monetary Economics,
1997 forthcoming. Poterba, James, and Lawrence J. Summers,
“Mean reversion in stock prices,” Journal of Finan-
Kandel, Shmuel, and Robert F. Stambaugh, cial Economics, Vol. 22, October 1988, pp. 27–59.
“Expectations and volatility of consumption and
asset returns,” Review of Financial Studies, Vol. Reyfman, Alexander, “Labor market risk and
3, No. 2, 1990, pp. 207–232. expected asset returns,” University of Chicago,
Ph.D. dissertation, 1997, Ch. 3.
, “Asset returns and intertemporal
preferences,” Journal of Monetary Economics, Rietz, Thomas A., “The equity risk premium: A
Vol. 27, February 1991, pp. 39–71. solution,” Journal of Monetary Economics, Vol.
22, July 1988, pp. 117–131.
Kocherlakota, Narayanna R., “On the ‘discount’
factor in growth economies,” Journal of Monetary Shiller, Robert J., “Consumption, asset markets,
Economics, Vol. 25, January 1990, pp. 43–47. and macroeconomic fluctuations,” Carnegie
Rochester Conference Series on Public Policy,
, “The equity premium: It’s still a Vol. 17, Autumn 1982, pp. 203–238.
puzzle,” Journal of Economic Literature, Vol.
34, March 1996, pp. 42–71. Telmer, Chris I., “Asset pricing puzzles and
incomplete markets,” Journal of Finance, Vol.
Lamont, Owen, “Earnings and Expected Re- 48, December 1993, pp. 1803–1832.
turns,” National Bureau of Economic Research,
working paper, No. 5671, 1996. Weil, Philippe, “The equity premium puzzle and
the risk-free rate puzzle,” Journal of Monetary
Lucas, Deborah J., “Asset pricing with undiver- Economics, Vol. 24, November 1989, pp. 401–421.
sifiable income risk and short sales constraints: