Rational Models of Irrational Behavior

George A. Akerlof; Janet L. Yellen

The American Economic Review, Vol. 77, No. 2, Papers and Proceedings of the Ninety-Ninth
Annual Meeting of the American Economic Association. (May, 1987), pp. 137-142.

Stable URL:
http://links.jstor.org/sici?sici=0002-8282%28198705%2977%3A2%3C137%3ARMOIB%3E2.0.CO%3B2-U

The American Economic Review is currently published by American Economic Association.

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained
prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in
the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/journals/aea.html.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

The JSTOR Archive is a trusted digital repository providing for long-term preservation and access to leading academic
journals and scholarly literature from around the world. The Archive is supported by libraries, scholarly societies, publishers,
and foundations. It is an initiative of JSTOR, a not-for-profit organization with a mission to help the scholarly community take
advantage of advances in technology. For more information regarding JSTOR, please contact support@jstor.org.

http://www.jstor.org
Tue Mar 4 13:08:47 2008
 Rational Models of Irrational Behavior

The General Theory revolutionized macro- behave according to impure assumptions, is

economics with its assertion that classical it not likely that the best models to fulfill the
economics contained a fundamental error. agenda will mirror that behavior?
Keynes argued that a capitalist economy According to standard Keynesian analysis,
could possess equilibria characterized by the key departure from self-interested, maxi-
persistent involuntary unemployment. He mizing behavior is the assumed stickiness of
also showed that aggregate demand would money wages. Keynes argued that because of
play a crucial role in determining output and the importance to workers of relative wages,
employment. Keynes' analysis accorded with they quite typically resist money wage reduc-
common sense and casual observation fifty tions in circumstances where they are willing
years ago. In our opinion, it still does so. to assent to real wage reductions. He recog-
Why then is there a crisis in Keynesian nized that "it is sometimes said that it would
economics? be illogical for labour to resist a reduction of
Keynesian analysis violates the commonly money-wages but not to resist a reduction of
regarded sine qua non of good economic real wages." But, he concluded, "whether
theory-a rnicroeconomic foundation based logical or illogical, experience shows that
on perfectly rational, maximizing behavior. this is how labour in fact behaves" (1935,
In our reading. economists have accorded p. 9).
the assumption of rational, self-interested Keynes' willingness to build a theory which
behavior unwarranted ritual purity, while al- was a pastiche of optimizing behavior and
ternative assumptions-that agents follow sociological/psychological rule-of-thumb be-
rules of thumb, that psychological or socio- havior grounded in the observation of how
logical considerations matter, or that, heaven people appear to behave was probably due
forbid, they act downright irrationally at to his conviction that the central features of
times-have been accorded corresponding his theory in no way hinged on the seem-
ritual impurity. This association between ingly illogical behavior of workers. The Gen-
"impure" assumptions and ritual pollution eral- ~ h e o r ymakes clear that money wage
has had the ill effect of confusing the esthetic stickiness is not in Keynes' opinion the
task of economics-which is to provide clear ultimate cause of involuntary unemploy-
logic for analyzing economic phenomena- ment; indeed, due to the adverse effects of
with the agenda of economics-whch is to falling prices on demand, involuntary unem-
explain the economic events of the real world. ployment might possibly be more severe in
Keynesian theory, with its partial reliance its absence. But most Keynesians accept the
on psychological, sociological, and rule-of- verdict of Patinkin and Pigou " that. in the
thumb behavior to derive departures from presence of a real balance effect, the aggre-
full employment and Pareto optimality, is gate demand curve is not vertical and thus a
the worst casualty of this failure to dissoci- full-employment equilibrium exists. What to
ate esthetic from agenda. If agents really ~ e y n e was
s a minor assumption in a theory
rationalizing business cycles is now interpre-
ted as the key assumption.
During the past two decades Keynesian
theorists- have- struggled to formulate a
*University of California, Berkeley, CA 94720. We "sensible" microeconomic foundation for
are grateful to the Sloan Foundation, Guggenheim
Foundation, and the National Science Foundation for Keynesian economics based on individual-
support under grant SES86-005023, and to Daniel istic optimizing behavior, by relaxing the
Kahneman and Jim Wilcox for helpful comments. assumptions of the perfectly competitive
138 A EA P4 PERS AND PROCEEDTIC GS 114 ). I987

Walrasian model and introducing instead significant deviations from rationality are re-
a dizzying array of market imperfections: quired to rationalize the Keynesian model.
asymmetric information, incomplete con-
tingent claims markets, staggered contracts, I. How Irrational is Keynesian Economics?
transactions costs, imperfect competition,
specific human capital, efficiency wages, etc. Keynes argued that nominal shocks to ag-
Have these efforts been successful? Not en- gregate demand (due, say, to changes in the
tirely. While each of these innovations has money supply)-whether anticipated or un-
enriched economics by modeling important anticipated-can cause prolonged depar-
aspects of reality, the introduction of these tures of the economy from full employment.
imperfections has still not provided a total Such departures simply cannot occur in any
rationalization of Keynesian economics when model that assumes fully rational optimizing
judged according to the rule that the pro- behavior, including rationally formed expec-
posed theory be fully consistent with rational tations, unless nominal prices exhbit inertia.
optimizing behavior and the absence of any The logic is simple. Rational agents should
unexploited gains from trade. In the end, it only care about real magnitudes. If so, any
invariably turns out either that there is an optimizing model with a unique equilibrium
unrealistic assumption or that some clever, will be "money neutral." A cut in the supply
complicated neoclassical contract will elimi- of money should just cause a proportional
nate involuntary unemployment. For exam- reduction in prices and wages with no change
ple, most versions of efficiency wage theory in employment, output, or real wages. This
w h c h are grounded in optimizing behavior conclusion in no way depends on the as-
suffer from the "defect" that there exist con- sumption that markets are perfectly com-
tracts which, while rarely observed, are feasi- petitive, or even that markets clear.
ble in principle (for example, employment Demand-generated business cycles require
bonds, job auctions, tournament contracts) nominal price rigidity. Such rigidity could be
and can eliminate involuntary unemploy- explained by "menu costs" associated with
ment if firms can establish reputations for nominal price changes, or by money illusion.
trustworthy labor relations. In the case of menu costs, one can hardly
This paper begins with the premise that imagine that the objective costs of price
theory whch fits the real world will be based changes are sufficiently large to explain busi-
on assumptions that individuals are not fully ness cycles. The leading argument against
rational. It would be simply unscientific to money illusion as a factor generating busi-
proceed otherwise. For indeed, individuals ness cycles is the implausibility that agents
may actually suffer from money illusion, persistently pass up opportunities for gain.
follow rules of thumb, or give weight to But how large must transactions costs be? Or
considerations of fairness and equity in eco- how foolish are agents whose behavior ex-
nomic matters. Section I1 reviews some evi- hibits money illusion? Are individuals who
dence whch indicates that individuals do change wages and/or prices inertially leav-
behave in such ways, and, furthermore, that ing the proverbial $500 bills on the side-
the sticky money wages assumed in Keynes- walk? Or are they failing to stoop to pick up
ian models are consistent with the known a few pennies?
irrationalities of human behavior. A second
justification for old-style Keynesian models
ear Rational Behavior. According to
previous work by ourselves (1985a,b; 1986)
without full rationality, is that the depar- and others (N. Gregory Mankiw, 1985;
tures from rationality needed to generate Olivier Blanchard and Nobuhro fiyotaki,
Keynesian business cycles are not very great. 1985), many forms of seemingly irrational
Section I argues that rejection of Keynesian behavior may really be "near-rational." By
business cycles on the grounds that fully this, we mean that agents have relatively
rational models must be money neutral is wide latitude for deviating from full optimi-
based on a faulty implicit assumption-that zation without incurring significant losses. In
1'0L. 77 KO. 2 T H E CONTRIBUTION O F K E Y K E S AFTER 50 Y E A R S 139

mathematical terms, t h s is a consequence of sets its price in Bertrand fashion, choosing

the envelope theorem whch states, in effect, the price that maximizes profits, talung rivals'
that the impact of an exogenous shock on a prices as given, and the firms have settled
fully maximizing agent is identical, up to a into a Bertrand equilibrium. If all firms are
first-order of approximation, whether he op- fully rational and the Fed cuts the money
timally changes his decision variable in supply, the new equilibrium will be exactly
response to a shock, or instead responds like the old in real terms. All that happens is
inertially. Stated differently, inertial, or rule- that each firm's nominal price falls in
of-thumb behavior typically imposes losses proportion to the money supply cut. Money
on its practitioners, relative to the rewards is neutral.
from optimizing, which are second-order. Now consider what would happen if some
Thus, slight relaxation of the standards for proportion of firms are nonmaximizers who
"good" model building-so as to tolerate follow an inertial rule in altering prices. To
behavioral assumptions entailing suitably take an extreme case, suppose the nonmaxi-
small losses from nonmaximizing-signifi- rnizers leave their price unchanged following
cantly enlarges the range of behavior to be the cut in the money supply. How much do
considered. For example, it turns out that in they lose? The answer is that their losses are
many contexts, the inertial adjustment of second-order. If the money supply fell by a
nominal wages and prices is near rational. percentage E , the inertial firms will make a
Staggered nominal contracts in the style of pricing error which is proportional to E , but
John Taylor (1979), whch do not quite op- incur losses which are proportional to e2-
timally make use of newly available infor- second-order in terms of the shock to the
mation because they keep nominal prices system. However, the decline in real bal-
constant for two periods, turn out to be ances (and hence output) resulting from the
near-rational as also are the stock adjust- cut in M is first-order-proportional to E .
ment responses assumed to characterize mo- How can this be? Recall that under normal
ney demand, consumption and investment in circumstances, any optimal decision is made
many Keynesian models. It might be thought by just balancing the marginal gains and
that near-rational theories must be close to losses from a change in the decision variable.
fully rational theories; but this intuition is in The price-setting competitors optimally set
fact incorrect. prices by balancing the marginal gain of
A n Example. The logic of the difference greater sales from a reduction in the price
between near rationality and full rationality against the loss due to lower profit on each
can be explained in the context of a simple unit sold. An optimum has not been reached
example. According to t h s example, if firms until the firm is indifferent, to a first-order of
are monopolistically competitive, and a frac- approximation, about its price. The firm's
tion of them change prices inertially in re- profit, as a function of its own price, is
sponse to money supply changes, then sig- almost flat in the neighborhood of an opti-
nificant business cycles result. However, mum. Accordingly, the firm has latitude for
nonmaximizing firms fare only insignifi- making a relatively large error, without
cantly worse than optimizing firms. Consider suffering large losses. An error of size 6
an economy with identical, monopolistically causes a loss proportional to ti2. In this
competitive firms selling differentiated prod- context, inertial price setting is almost cost-
ucts. To make thngs simple, imagine that less even though its macroeconomic impact
output can be produced costlessly. Further, is significant.
assume that each firm's sales depends on its The logic of the preceding example con-
price relative to its rivals and on aggregate cerning prices suggests a similar defense of
demand whlch, again for simplicity, is pro- the standard Keynesian assumption of norni-
portional to real balances-the nominal nal wage rigidity. Nominal wage rigidity is
supply of money divided by the average near-rational if the firms' profits are a con-
price level. Finally, imagine that each firm tinuously differentiable function of the wage
140 A EA PAPERS A N D PROCEEDIA'GS M A Y 1987

they pay. In the efficiency wage paradigm, judgmental mistakes that are due to the use
for example, the "productivity" of workers, of the three heuristics. These judgment er-
broadly defined, is an increasing function of rors are too numerous and too frequently
the real wage they receive. Firms optimally documented in the laboratory with salient
set wages by equating the marginal gains examples in the field to be easily dismissed
from lower wage cost per worker with the as unimportant.
losses from lower productivity per worker; The question arises whether there are cog-
accordingly, inertial wage setting is near- nitive biases that suggest potential reasons
rational. A firm that (nonoptimally) fails to for money wage stickmess. The most natural
cut wages in a recession gets some reward explanation of sticky money wages stems
from its behavior-hgher morale, lower from anchoring. In a typical anchoring ex-
turnover, etc. And whle the behavior may periment, one finds that "irrelevant" ini-
not be strictly optimal, it is almost optimal tial conditions affect outcomes. For example,
since, for a firm whch was initially optimiz- in a classic anchoring experiment, Daniel
ing, the rewards and costs of wage cuts were Kahneman and Amos Tversky (see Bazer-
exactly balanced to start with. man) spun a roulette wheel, and then asked
subjects to estimate the number of African
11. How Rational Are Economic Agents? states with representatives in the United Na-
tions-using the number obtained by spin-
We argued above that nonrational ele- ning the roulette wheel as the initial esti-
ments should be brought into macro models, mate. For those whose initial estimate from
and that, in many models, not much irra- the roulette wheel was 10, the median esti-
tionality is needed to produce business cycles. mate was 25; for those whose initial estimate
In t h s section we shall argue that psy- from the roulette wheel was 65, the median
chology and sociology provide natural ex- estimate was 45!
planations for sticky money wages. The New Could anchoring explain sluggish adjust-
Classical Economics' conclusion, that nomi- ment of money wages? It certainly could if
nal variables are proportional to the ex- last-period's money wage acts as an anchor
pected money supply, is a singularity, in no whlch influences t h s period's wage settle-
way predicted by the judgmental errors and ment. As we shall see presently, in the dis-
concern with equity whch is well docu- cussion of fairness, people's views of fair
mented in psychology and sociology. money wages apparently are anchored in the
Cognitive Biases. Twenty years ago it current money wage.
was widely believed that in most cognitive Fairness. It is hard to believe that pay-
judgments people acted as intuitive scien- ments are not jointly determined by market
tists. However, two decades of work by cog- forces and fairness. Steven Allen, Robert
nitive and social psychologists have un- Clark, and Daniel Sumner (1984) have
earthed a large variety of ways in whch pointed out that some firms have voluntarily
individuals' judgments edubit systematic er- added some indexation to benefits paid to
rors relative to the scientific, objectively ra- already retired employees. While these pay-
tional model. ments undoubtedly had some beneficial effect
People use at least three heuristics whch in enhancing the firms' reputations with cur-
generate biases in their decisions. According rent and prospective employees, a more nat-
to the availability heuristic, they depend more ural explanation for these payments is not
than they should on "salient" information pure profit maximization by the firms, but a
which is easily retrievable from memory. commitment to fair behavior.
According to the representativeness heuristic, In order for fairness to play a role in
they act as if stereotypes are more common determining contract outcomes there must
than they actually are, and in anchoring, be a discrepancy between fair and market
they let their judgments be overly reliant on clearing outcomes. Kahneman, Jack Knetsch,
some initial "anchoring" values. Max Bazer- and k c h a r d Thaler (1986) have provided
man (1986) has enumerated thrteen distinct indisputable evidence of such a divergence in
L'OL 77 KO. 2 THE COhTRIBL~TIONOF KEY'VES AFTER 50 YEARS 141

a recent interview study. In this study they labor services may subjective@ value his or
described a variety of situations to randomly her services by more than the buyer of those
sampled telephone interviewees, and then services. Then, if we take the gap between
asked whether the market-clearing solution the ratio of "reward" relative to investment
was fair. For example, they asked whether it on the part of the buyer and the seller, we
would be fair for a hardware store to charge have a prediction of the extent of resentment
more for snow shovels following a snow- that parties in an exchange will experience.
storm: 82 percent thnk it unfair and 18 Because market-clearing contracts may be
percent think it fair. As a second example, viewed as inequitable, the attempt to impose
they told interviewees that a small photo- market-clearing terms on a transaction can
copying shop has one employee earning $9 easily generate resentment.
per hour. Business continues to be satisfac- The existence of resentment does not, of
tory but a factory in the area has closed and course, imply that economic outcomes can-
unemployment has increased. Would it be not occur which are resented. But in labor
fair to reduce the wage paid to $7 an hour contracts there are good reasons why em-
now paid elsewhere to newly h r e d workers ployers will wish to temper the resentments
with similar talents? Eighty-three percent of their employees and accordingly may offer
think it unfair. contracts that are not market clearing. Quite
Anchoring played a crucial role in most of simply. resentment caused by a sense of
the interview results in t h s study. Most unfair treatment is likely to translate into
questions involved the fair response to a poor performance by workers who can ex-
change from some initial situation. It was, in ercise discretion in the performance of their
general, considered unfair for one party to work. Even if supervision and monitoring
benefit relative to the initial anchoring situa- are feasible at low cost, it may not pay firms
tion while the second party lost. Thus in the to monitor their employees too closely. A
hardware store example. it was unfair for the recent study by Edward Deci, James Con-
store to profit whle customers paid more for nell, and Richard Ryan (1985) has shown
snow shovels; similarly. it was unfair for the that workers who are given more detailed
firm to cut the wages of its existing em- rules and are more closely monitored experi-
ployees. ence less job satisfaction, are less motivated,
Importantly for Keynesian economics the and place more importance on such external
Kahneman-Knetsch-Thaler experiments in- rewards as compensation; in contrast, those
dicated the presence of anchoring based on who are less controlled achieve greater satis-
current money wages. Respondents felt it faction in mastering their jobs. If firms are
fair for a company whose business was bad to take advantage of the self-motivation that
to raise wages by only 7 percent when there comes with on-the-job autonomy, then they
was 12 percent inflation, but unfair to cut must ensure that workers perceive them-
wages by 5 percent if there was no inflation. selves to be fairly treated. The cost can easily
These findings are consistent with the involve paying wages in excess of market
standard sociological theory of fairness clearing. Accordingly, we would expect per-
known as equity theory. (For an excellent ceptions of fairness to play a role in de-
survey see Roger Brown, 1986.) Equity the- termining wage contracts and anchoring to
ory (due to J. Stacy Adams and George cause money wages to be sticky.
Homans) predicts that in exchanges, "out-
comes" (rewards) relative to the "invest- 111. Conclusion
ments" must be equal between the parties to
the transaction. Whle t h s theory is specifi- The bad press that Keynesian theory has
cally derivative from economics, there is an recently received from maximizing, super-
important difference between its predictions rational theory is simply undeserved. The
and those of standard economic theory. In assumptions required to motivate Keynesian
equity theory, the investments and outcomes economics are quite consistent with the be-
are ,SubjectlUe~y measured. The provider of havioral regularities documented by psy-
142 AE.4 PAPERS A N D PROCEEDIh GS 'IfA Y 1987

chologists and sociologists. This motivation Bazerman, Max, Judgment in Munugerral De-
is in no way tortured out of complicated cisionmaking, New York: Wiley & Sons,
assumptions and models. It is highly natural. 1986.
Keynesianism, both as theory and explana- Blanchard, Olivier and Kiyotaki, Nobuhiro, " Mo-
tion of the facts, is alive and well on its nopolistic Competition, Aggregate De-
fiftieth birthday. Happy Birthday, General mand Externalities and Real Effects of
Theory ! Nominal Money," NBER Working Paper
No. 1770, December 1985.
Brown, Roger, Sociul Psychology, the Second
REFERENCES Edition, New York: Free Press, 1986.
Deci, Edward L., Connell, James P. and Ryan,
Akerlof, George and Yellen, Janet, (1985a) "A Richard M., "Self-Determination in a Work
Near-Rational Model of the Business Organization," mimeo, University of
Cycle, with Wage and Price Inertia," Rochester, 1985.
Quarterly Journal of Economics, Supple- Kahneman, Daniel, Knetsch, Jack and Thaler,
ment 1985, 100, 823-38. Richard, "Fairness as a Constraint on Profit
and , (1985b) "Can Small Seeking: Entitlements in the Market,"
Deviations from Rationality Make Signifi- American Economic Reuiew, September
cant Differences to Economic Equilib- 1986, 76, 728-41.
ria?," American Economic Review, Sep- Keynes, J. M., The General Theory of Employ-
tember 1985, 75, 708-20. ment, Interest and Money, London:
and , "The Macroeconomic Harcourt, Brace and World, 1935.
Applications of a Dynamic Envelope The- Mankiw, N. Gregory, "Small Menu Costs and
orem," rnimeo., University of California- Large Business Cycles: A Macroeconomic
Berkeley, 1986. Model of Monopoly," Quarterly Journal of
Allen, Steven G., Clark, Robert L. and Sumner, Economics, May 1985, 100, 529-37.
Daniel A., "Post-Retirement Adjustments Taylor, John, "Staggered Wage Setting in a
of Pension Benefits," NBER Working Macro Model," American Economic Re-
Paper No. 1364, June 1984. view Proceedings, May 1979, 69, 108-13.
http://www.jstor.org

LINKED CITATIONS
- Page 1 of 1 -

You have printed the following article:

Rational Models of Irrational Behavior
George A. Akerlof; Janet L. Yellen
The American Economic Review, Vol. 77, No. 2, Papers and Proceedings of the Ninety-Ninth
Annual Meeting of the American Economic Association. (May, 1987), pp. 137-142.
Stable URL:
http://links.jstor.org/sici?sici=0002-8282%28198705%2977%3A2%3C137%3ARMOIB%3E2.0.CO%3B2-U

This article references the following linked citations. If you are trying to access articles from an
off-campus location, you may be required to first logon via your library web site to access JSTOR. Please
visit your library's website or contact a librarian to learn about options for remote access to JSTOR.

References

A Near-Rational Model of the Business Cycle, With Wage and Price Inertia
George A. Akerlof; Janet L. Yellen
The Quarterly Journal of Economics, Vol. 100, Supplement. (1985), pp. 823-838.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%281985%29100%3C823%3AANMOTB%3E2.0.CO%3B2-N

Can Small Deviations from Rationality Make Significant Differences to Economic Equilibria?
George A. Akerlof; Janet L. Yellen
The American Economic Review, Vol. 75, No. 4. (Sep., 1985), pp. 708-720.
Stable URL:
http://links.jstor.org/sici?sici=0002-8282%28198509%2975%3A4%3C708%3ACSDFRM%3E2.0.CO%3B2-8

Fairness as a Constraint on Profit Seeking: Entitlements in the Market

Daniel Kahneman; Jack L. Knetsch; Richard Thaler
The American Economic Review, Vol. 76, No. 4. (Sep., 1986), pp. 728-741.
Stable URL:
http://links.jstor.org/sici?sici=0002-8282%28198609%2976%3A4%3C728%3AFAACOP%3E2.0.CO%3B2-I

Small Menu Costs and Large Business Cycles: A Macroeconomic Model of Monopoly
N. Gregory Mankiw
The Quarterly Journal of Economics, Vol. 100, No. 2. (May, 1985), pp. 529-537.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28198505%29100%3A2%3C529%3ASMCALB%3E2.0.CO%3B2-O