1. A Psychological Perspective on Economics
By D ANIEL KAHNEMAN*
My rst exposure to the psychological as- of trust and reciprocation have begun to appear
sumptions of economics was in a report that (Kevin McCabe et al., 2001), and they con rm
Bruno Frey wrote on that subject in the early the signi cance of these games as social situa-
1970’s. Its rst or second sentence stated that tions, in which behavior is determined to a
the agent of economic theory is rational and substantial extent by motives other than pro t.
sel sh, and that his tastes do not change. I found A considerable amount of evidence, drawn
this list quite startling, because I had been pro- from two-person games and from public-goods
fessionally trained as a psychologist not to experiments, suggests that many people, at least
believe a word of it. The gap between the as- in the Western culture, start out trusting and
sumptions of our disciplines appeared very benevolent and reciprocate both good and bad
large indeed. behaviors. This propensity for reciprocity has
Has the gap been narrowed in the intervening been studied both empirically and theoretically
30 years? A search through some introductory (Ernst Fehr and Simon Gachter, 2000). Many
textbooks in economics indicates that if there people also have a propensity to punish, even at
has been any change, it has not yet ltered down some costs to themselves, misbehaviors of one
to that level: the same assumptions are still in stranger toward another stranger. An important
place as the cornerstones of economic analysis. theoretical discovery is that the presence of a
However, a behavioral approach to economics suf cient number of individuals with these mo-
has emerged in which the assumptions are not tives in a population will turn individuals who
held sacrosanct. In the following I comment do not have the same motives into apparent
selectively on the developments with regard to cooperators (Fehr et al., 2002).
the three assumptions, on both sides of the The experimental and theoretical studies of
disciplinary divide. sel shness that some economists have con-
ducted represent a general advance for social
I. Sel shness science. They also represent a signi cant move
in economics, beyond the model of economic
The clearest progress has occurred in correct- agents that Amartya Sen (1977) famously la-
ing and elaborating the assumption of sel sh- beled “rational fools.” Some of the agents in
ness, and the progress has come entirely from Fehr’s models are “opportunistic with guile”
developments in economics, where the inven- (Oliver Williamson, 1985), but their behavior is
tion of the ultimatum game (Werner Guth et al., strongly constrained by the fact that they are
1982) had a great impact. Experiments con- compelled to interact with people who care
ducted in low-income countries by investigators about being treated fairly and are willing to do
armed with dollars con rmed conclusively that something about it.
quite a few people will forgo a substantial sum
for the sole bene t of denying a larger sum to an II. Rationality
anonymous stranger who has treated them un-
generously (Lisa Cameron, 1999). Other evi- No one ever seriously believed that all people
dence indicates that offers that would be have rational beliefs and make rational deci-
rejected if they came from a person will be sions all the time. The assumption of rationality
accepted if they are generated by a computer. is generally understood to be an approximation,
Brain-imaging studies of people playing games which is made in the belief (or hope) that de-
partures from rationality are rare when the
stakes are signi cant, or that they will disappear
* Department of Psychology, Princeton University, under the discipline of the market. This belief is
Princeton, NJ 08544. not shared by everyone: some economists have
162

2. VOL. 93 NO. 2 VIEWS OF ECONOMICS FROM NEIGHBORING SOCIAL SCIENCES 163
questioned both the idea that small deviations legitimized and encouraged the development of
from rationality do not matter (e.g., George economic theories that model departures from
Akerlof and Janet Yellen, 1985) and the idea economic rationality in speci c contexts. There
that arbitrageurs will drive irrationality out of have been quite a few of those, including the
the marketplace (Andrei Shleifer, 2000). Their following:
position, if accepted, increases the relevance of
nonrational behavior in economic analysis. (i) a stock market in which all traders
The standard of rationality in economics was, believe they are above average (Ter-
and remains, the maximization of subjective rance Odean, 1998);
expected utility—a combination of von Neumann- (ii) a stock market in which traders are
Morgenstern preferences and a Bayesian belief myopic and loss-averse (Shlomo Ben-
structure. There have been important challenges artzi and Richard Thaler, 1995);
to this de nition of rationality. Both Maurice (iii) a market in which traders are too
Allais (1953) and Daniel Ellsberg (1961) dem- quick to jump to conclusions (Matthew
onstrated preferences that violate expected- Rabin, 2003);
utility theory but have considerable normative (iv) models in which discounting is quasi-
appeal. A rich literature has developed in at- hyperbolic (David Laibson, 1997);
tempts to formulate a theory of rational choice (v) models in which self-control is an
that will legitimize the Allais and Ellsberg pat- acknowledged problem (Thaler and
terns of preferences. Herbert Simon (1955) Hersh Shefrin, 1981).
introduced the concepts of satis cing and
bounded rationality, which can be interpreted as But the rationality model continues to provide
de ning a realistic normative standard for an the basic framework even for these models, in
organism with a nite mind. which the agents are “fully rational, except
In the mid-1980’s Amos Tversky and I artic- for ...” some particular deviation that explains a
ulated a direct challenge to the rationality family of anomalies.
assumption itself, based on experimental dem-
onstrations in which preferences were affected III. Unchanging Tastes and the Carriers
predictably by the framing of decision prob- of Utility
lems, or by the procedure used to elicit prefer-
ences (Tversky and Kahneman, 1986). We Economists are thoroughly habituated to the
argued that the demonstrated susceptibility of sight of indifference maps, but for someone
people to framing effects violates a fundamental who has been trained as a psychologist they can
assumption of invariance, which has also been be a source of puzzlement. It took me a long
labeled extensionality (Kenneth J. Arrow, 1982) time to realize that the representation looked
and consequentialism (Peter J. Hammond, 1989). odd because I kept looking for an indication of
Unlike the paradoxes of expected-utility theory, the individual’s current position in the map.
violations of invariance cannot be defended as There is no such indication, of course, because
normative. Furthermore, these violations are not this parameter is supposed to be irrelevant: pref-
restricted to the laboratory. The labeling of erences for nal states of endowment are as-
taxes is an obvious example of framing (Ed J. sumed to be stable over variations of current
McCaffery, 1994). The power of default options endowment. This assumption, called reference-
is another. Brigitte C. Madrian and Dennis F. independence by Tversky and Kahneman (1991)
Shea (2001) reported that the enrollment rate in is the interpretation of unchanging tastes with
401(k) plans is close to 100 percent when en- which I am concerned here. As I will show
rollment is automatic, but if action is required to below, reference-independence can also be
enroll, only about half the employees will join viewed as an aspect of rationality.
the plan within their rst year of employment. The assumption of reference-independence
The cost of the activity is hardly suf cient to (or, equivalently, the idea that nal states of
rationalize this behavior. endowment are the carriers of utility) has a long
The various questions that have been raised history in the theoretical analysis of decision-
about the rationality assumption appear to have making under risk. Modern decision theory

3. 164 AEA PAPERS AND PROCEEDINGS MAY 2003
traces its origins to the famous St. Petersburg a theory of choice that completely neglects
essay in which Daniel Bernoulli (1738) formu- the short-term emotions associated with gains
lated the original version of expected-utility and losses is bound to be psychologically
theory. Bernoulli’s decision-maker values - unrealistic.
nancial outcomes as states of wealth and orders Prospect theory (Kahneman and Tversky,
options by the expected utility of these states. 1979) was offered as a descriptive model of
The model incorporates an assumption of xed risky choice in which the carriers of utility are
tastes, because the utility of states of wealth not states of wealth, but gains and losses rela-
does not depend on current endowment. This tive to a neutral reference point.1 The most
assumption has been retained in all subsequent distinctive predictions of the theory arise from a
versions of expected-utility theory. property of preferences called loss-aversion:
An important article by Matthew Rabin the response to losses is consistently much more
(2000; see also Rabin and Thaler, 2001) showed intense than the response to corresponding
that attitudes to wealth cannot explain the levels gains, with a sharp kink in the value function
of risk aversion observed when the stakes are at the reference point. Unlike a reasonable
low. Rabin developed a method that permits Bernoulli agent, a loss-averse decision-maker
inferences of the following kind (Rabin and will always reject a single 50-50 bet to lose
Thaler, 2001 p. 222): “Suppose, for instance, we $100 or win $105. Estimates of the coef cient
know that a risk-averse person turns down of loss aversion (the ratio of slopes in the neg-
50-50 lose-$100-or-gain-$105 bets for any life- ative and positive domains) converge on a value
time wealth level less than (say) $350,000, but of about 2 (Tversky and Kahneman, 1992).
know nothing about his or her utility function for The idea of loss-aversion was rst extended
wealth levels above $350,000,except that it is not to riskless choice by Thaler (1980), who used it
convex. Then we know that from an initial to explain the endowment effect—the now well-
wealth level of $340,000 the person will turn documented discrepancy between willingness-
down a 50-50 bet of losing $4,000 and gaining to-pay and willingness-to-accept for the same
$635,670.” Most people will reject the small good. Other implications were explored by
bet, and most will nd it absurd to reject the Kahneman et al. (1991) and by Tversky and
large bet. Attitudes to wealth cannot explain Kahneman (1991), and in several sources
these preferences. collected by Kahneman and Tversky (2000).
If Bernoulli’s formulation is so transparently Reference-dependence and loss-aversion are
incorrect as a descriptive model, why has it both involved in the sharp distinction that most
been retained for so long? The answer may well people draw between opportunity costs and
be that the assignment of utility to wealth is an losses. Among many other phenomena, the rel-
aspect of rationality. The following thought ex- ative neglect of opportunity costs explains
periment illustrates the point. target-income behavior by New York cab driv-
ers, who stop work earlier on rainy days, when
Two persons get their monthly report their opportunities are best (Colin Camerer et
from a broker: al., 1997). Loss-aversion contributes to sticki-
A is told that her wealth went from ness in markets, because loss-averse agents
4M to 3M; are much less prone to exchanges than nal-
B is told that her wealth went from states agents. In experiments in which some
1M to 1.1M.
Which of the two individuals has more randomly chosen participants were endowed
reason to be satis ed with her nancial with a consumption good and allowed to trade
situation? it, the volume of trade was about half of the
Who is happier today? amount expected from standard economic the-
ory (Kahneman et al., 1990). Loss-aversion for
In Bernoulli’s analysis only the rst of
these questions is relevant, and only the long-
term consequences matter. The wealth frame 1
Habit-formation models incorporate similar ideas, in a
ts a standard of mature reasonableness. But format that many economists will nd more congenial.

4. VOL. 93 NO. 2 VIEWS OF ECONOMICS FROM NEIGHBORING SOCIAL SCIENCES 165
the consumption good was involved, because sodes of consumption. Loewenstein and Daniel
exchanges of money tokens in the same institu- Adler (1995) showed that participants in an
tion conformed quite precisely to the standard experiment underestimated the price that they
analysis. Not all exchanges involve loss- would demand to part from an object, if they
aversion. For example, there is little reluctance were asked the question before actually taking
to trade a ve-dollar bill for ve singles, and possession of it.
aversion to giving up goods is unlikely to The evidence of grave de ciencies in taste
affect the merchant who exchanges a pair of prediction appears to pose a signi cant chal-
shoes for cash (Tversky and Kahneman, 1991). lenge to many applications of the rational-agent
But the drying up of sales in real-estate mar- model. In particular, it is dif cult to reconcile
kets that have experienced declining prices this evidence with the extraordinary feats of
illustrates an unwillingness to accept losses hedonic prediction that are assumed in theories
relative to an existing reference price (David that assume the rationality of the choice to be-
Genesove and Christopher Mayer, 2001). The come addicted (Gary Becker and Kevin Mur-
boundary conditions for loss-aversion are yet phy, 1988). Perhaps more than any other, the
to be delineated with precision (Ian Bateman rational-addiction model highlights the large
et al., 1997). gap that persists between the criteria of reason-
The rules of fairness that embody a regard for ableness that are applied to views of human
loss-aversion also induce stickiness in markets. motivation in the disciplines of economics and
For example, cutting the wage of an employee psychology.
is considered unfair even when the employee An increased willingness of economists to
could easily be replaced at a lower pay. In consider subjective data is a salient develop-
general, imposing losses on others is considered ment of recent years. The many applications of
unfair under conditions where failing to share measures of happiness in economic research
gains is entirely acceptable (Kahneman et al., reviewed by Bruno Frey and Alois Stutzer
1986). The asymmetry between losses and fore- (2002) included their own studies of the effects
gone gains is recognized in many aspects of of democratic institutions, as well as research
the law (David Cohen and Jack L. Knetsch, by others on the effects of in ation and unem-
1992). ployment (Alberto Alesina et al., 2001). An
Adaptation and the consequent shift in the interest in the experienced utility of outcomes
reference point that separates gains from losses (Kahneman et al., 1997) is a natural side-effect
have been interpreted here as a common form of of the willingness to consider agents who are
taste change. The ability of a decision-maker to less than fully rational. If these agents do not
anticipate such changes in tastes is an essential necessarily maximize the quality of their out-
but often neglected aspect of rationality (James comes, then choice is no longer the sole relevant
March, 1978). Reviewing the relevant literature, measure of utility. This idea, if accepted, could
George Loewenstein and David Schkade (1999) have implications for many domains of eco-
concluded that people generally underestimate nomic analysis.
the extent of hedonic adaptation to new states.
Hedonic and affective forecasts are susceptible IV. Will the Gap Close Further?
to a substantial duration bias (Daniel Gilbert et
al., 1998). Assistant professors, for example, Much has happened in the conversation be-
greatly overestimate the effects of a tenure de- tween economics and psychology over the last
cision on their happiness a year later (Gilbert 25 years. The church of economics has admitted
and Wilson, 2000). Failures of hedonic predic- and even rewarded some scholars who would
tion are even common in the short term. The have been considered heretics in earlier periods,
participants in a study reported by Kahneman and conventional economic analysis is now be-
and Jackie Snell (1992) showed little ability to ing done with assumptions that are often much
anticipate how their enjoyment of a piece of more psychologically plausible than was true in
music or a helping of their favorite ice cream the past. However, the analytical methodology
would change over a period of eight daily epi- of economics is stable, and it will inevitably

5. 166 AEA PAPERS AND PROCEEDINGS MAY 2003
constrain the rapprochement between the disci- Political Economy, August 1988, 96(4), pp.
plines. Whether or not psychologists nd them 675–700.
odd and overly simple, the standard assump- Benartzi, Shlomo and Thaler, Richard H. “Myo-
tions about the economic agent are in economic pic Loss Aversion and the Equity Premium
theory for a reason: they allow for tractable Puzzle.” Quarterly Journal of Economics,
analysis. The constraint of tractability can be February 1995, 110(1), pp. 75–92.
satis ed with somewhat more complex models, Bernoulli, Daniel. “Exposition of a New Theory
but the number of parameters that can be added on the Measurement of Risk.” Econometrica,
is small. One consequence is that the models of January 1954, 22(1), pp. 23–36 [original
behavioral economics cannot stray too far from work published 1738].
the original set of assumptions. Another con- Camerer, Colin F.; Babcock, Linda; Loewenstein,
sequence is that theoretical innovations in George and Thaler, Richard H. “Labor Supply
behavioral economics may be destined to be of New York City Cabdrivers: One Day at a
noncumulative: when a new model is developed Time.” Quarterly Journal of Economics,
to account for an anomaly of the basic theory, May 1997, 112(2), pp. 407–42.
the parameters that were modi ed in earlier Cameron, Lisa A. “Raising the Stakes in the
models will often be restored to their original Ultimatum Game: Experimental Evidence
settings. Thus, it now appears likely that the gap from Indonesia.” Economic Inquiry, January
between the views in the two disciplines has 1999, 37(1), pp. 47–59.
been permanently narrowed, but there are no Cohen, David and Knetsch, Jack L. “Judicial
immediate prospects of economics and psychol- Choice and Disparities between Measures of
ogy sharing a common theory of human Economic Value.” Osgoode Hall Law Re-
behavior. view, 1992, 30, pp. 737–70.
Ellsberg, Daniel. “Risk, Ambiguity and the
REFERENCES Savage Axioms.” Quarterly Journal of
Economics, November 1961, 75(4), pp.
Akerlof, George A. and Yellen, Janet L. “Can 643– 69.
Small Deviations from Rationality Make Sig- Fehr, Ernst; Fischbacher, Urs and Gachter, Si-
ni cant Differences to Economic Equilib- mon. “Strong Reciprocity, Human Coopera-
ria?” American Economic Review, September tion, and the Enforcement of Social Norms.”
1985, 75(4), pp. 708 –20. Human Nature, January 2002, 13(1), pp.
Alesina, Alberto; Di Tella, Rafael and MacCul- 1–25.
loch, Robert. “Inequality and Happiness: Fehr, Ernst and Gachter, Simon. “Fairness and
Are Europeans and Americans Different?” Reciprocity.” Journal of Economic Perspec-
National Bureau of Economic Research tives, Summer 2000, 14(3), pp. 159– 81.
(Cambridge, MA) Working Paper No. Frey, Bruno S. and Stutzer, Alois. “What Can
8198, 2001. Economists Learn from Happiness Re-
Allais, Maurice. “Rational Man’s Behavior in search?” Journal of Economic Literature,
the Presence of Risk: Critique of the Postu- June 2002, 40(2), pp. 402–35.
lates and Axioms of the American School.” Genesove, David and Mayer, Christopher. “Loss
Econometrica, October 1953, 21(4), pp. 503– Aversion and Seller Behavior: Evidence from
46. the Housing Market.” Quarterly Journal of
Arrow, Kenneth J. “Risk Perception in Psychol- Economics, November 2001, 116(4), pp.
ogy and Economics.” Economic Inquiry, Jan- 1233–60.
uary 1982, 20(1), pp. 1–9. Gilbert, Daniel T.; Pinel, Elizabeth C.; Wilson,
Bateman, Ian; Munro, Alistair; Rhodes, Bruce; Timothy D.; Blumberg, Stephen J. and Wheat-
Starmer, Chris and Sugden, Robert. “A Test of ley, Thalia P. “Immune Neglect: A Source of
the Theory of Reference-Dependent Prefer- Durability Bias in Affective Forecasting.”
ences.” Quarterly Journal of Economics, Journal of Personality and Social Psychol-
May 1997, 112(2), pp. 479 –505. ogy, September 1998, 75(3), pp. 617–38.
Becker, Gary S. and Murphy, Kevin M. “A Gilbert, Daniel T. and Wilson, Timothy D.
Theory of Rational Addiction.” Journal of “Miswanting: Some Problems in the Fore-

6. VOL. 93 NO. 2 VIEWS OF ECONOMICS FROM NEIGHBORING SOCIAL SCIENCES 167
casting of Future Affective States,” in J. For- Madrian, Brigitte C. and Shea, Dennis F. “The
gas, ed., Feeling and thinking: The role of Power of Suggestion: Inertia in 401(k) Par-
affect in social cognition. New York: Cam- ticipation and Savings Behavior.” Quarterly
bridge University Press, 2000, pp. 178–97. Journal of Economics, November 2001,
Guth, Werner; Schmittberger, R. and Schwarz, B. 116(4), pp. 1149– 87.
“An Experimental Analysis of Ultimatum March, James G. “Bounded Rationality, Ambi-
Bargaining.” Journal of Economic Behavior guity, and the Engineering of Choice.” Bell
and Organization, December 1982, 3(4), pp. Journal of Economics, Autumn 1978, 9(2),
367–88. pp. 587–608.
Hammond, Peter J. “Consistent Plans, Conse- McCabe, Kevin; Houser, Daniel; Ryan, Lee;
quentialism, and Expected Utility.” Econo- Smith, Vernon and Trouard, Theodore. “Func-
metrica, November 1989, 57(6), pp. 1445– tional Imaging Study of Cooperation in Two-
49. Person Reciprocal Exchange.” Proceedings
Kahneman, Daniel; Knetsch, Jack and Thaler, of the National Academy of Sciences (USA)
Richard. “Fairness as a Constraint on Pro t 25 September 2001, 98(20), pp. 11832–35.
Seeking: Entitlements in the Market.” Amer- McCaffery, Ed J. “Cognitive Theory and Tax.”
ican Economic Review, September 1986, UCLA Law Review, September 1994, 41(7),
76(4), pp. 728 – 41. pp. 1861–1947.
. “Experimental Tests of the Endow- Odean, Terrance. “Are Investors Reluctant to
ment Effect and the Coase Theorem.” Jour- Realize Their Losses?” Journal of Finance,
nal of Political Economy, December 1990, October 1998, 53(5), pp. 1775–98.
98(6), pp. 1325–48. Rabin, Matthew. “Risk Aversion and Expected
. “The Endowment Effect, Loss Aver- Utility Theory: A Calibration Theorem.”
sion, and Status Quo Bias: Anomalies.” Jour- Econometrica, September 2000, 68(5), pp.
nal of Economic Perspectives, Winter 1991, 1281–92.
5(1), pp. 193–206. . “Inferences by Believers in the Law of
Kahneman, Daniel and Snell, Jackie. “Predicting Small Numbers.” Quarterly Journal of Eco-
a Changing Taste.” Journal of Behavioral nomics, 2003 (forthcoming).
Decision Making, July 1992, 5(3), pp. 187– Rabin, Matthew and Thaler, Richard. “Anoma-
200. lies—Risk Aversion.” Journal of Economic
Kahneman, Daniel and Tversky, Amos. “Prospect Perspectives, Winter 2001, 15(1), pp. 219 –
Theory: An Analysis of Decisions under 32.
Risk.” Econometrica, March 1979, 47(2), pp. Sen, Amartya. “Internal Consistency of Choice.”
313–27. Econometrica, May 1993, 61(3), pp. 495–521.
, eds. Choices, values and frames. New Shleifer, Andrei. Clarendon Lectures: Inef cient
York: Cambridge University Press, 2000. markets. Oxford, U.K.: Oxford University
Kahneman, Daniel; Wakker, Peter P. and Sarin, Press, 2000.
Rakesh. “Back to Bentham? Explorations of Simon, Herbert. “A Behavioral Model of Ratio-
Experienced Utility.” Quarterly Journal of nal Choice.” Quarterly Journal of Econom-
Economics, May 1997, 112(2), pp. 375–405. ics, February 1955, 69(1), pp. 99 –118.
Laibson, David. “Golden Eggs and Hyperbolic Thaler, Richard. “Toward a Positive Theory of
Discounting.” Quarterly Journal of Econom- Consumer Choice.” Journal of Economic Be-
ics, May 1997, 112(2), pp. 443–77. havior and Organization, March 1980, 1(1),
Loewenstein, George and Adler, Daniel. “A Bias pp. 36 – 60.
in the Prediction of Tastes.” Economic Jour- Thaler, Richard and Shefrin, Hersh M. “An Eco-
nal, July 1995, 105(431), pp. 929 –37. nomic Theory of Self Control.” Journal of
Loewenstein, George and Schkade, David. “Wouldn’t Political Economy, April 1981, 89(2), pp.
It Be Nice? Predicting Future Feelings,” in D. 392–406.
Kahneman, E. Diener, and N. Schwartz, eds., Tversky, Amos and Kahneman, Daniel. “Rational
Well-being: The foundations of hedonic psy- Choice and the Framing of Decisions.” Jour-
chology. New York: Russell Sage Founda- nal of Business, October 1986, 59(4), Part 2,
tion, 1999, pp. 85–105. pp. S251–78.

7. 168 AEA PAPERS AND PROCEEDINGS MAY 2003
. “Loss Aversion in Riskless Choice: A nal of Risk and Uncertainty, October 1992,
Reference-Dependent Model.” Quarterly 5(4), pp. 297–323.
Journal of Economics, November 1991, Williamson, Oliver E. The economic institu-
106(4), pp. 1039– 61. tions of capitalism: Firms, markets, rela-
. “Advances in Prospect Theory: Cumu- tional contracting. New York: Free Press,
lative Representation of Uncertainty.” Jour- 1985.