"Behavioral economics" replaces strong rationality assumptions used in economic modeling with assumptions that are consistent with evidence from psychology , while maintaining an emphasis on mathematical structure and explanation of naturally-occurring (field) data. Of course, for decades social scientists have criticized economic models for assuming too much rationality, and economists defend the models as useful approximations. In behavioral economics we believe more realistic assumptions will make for better approximations. The only interesting question is how to incorporate the psychology into economics. Herbert Simon, who coined the term "bounded rationality" in the 1950s, thought theories of individuals in economics should resemble theories in cognitive psychology, which specify algorithms or detailed mechanisms by which decisions are reached. Economists never took up this suggestion with any vigor, perhaps because Simon's suggestion came just as economists were finding ways to characterize economic decisions and equilbria in unusually elegant mathematical terms. The elegant mathematics left no room for messier cognitive theories.

Indeed, the "literary" tradition in economics before about 1930 - due to Adam Smith, Keynes, Marshall, Fisher and others - is full of psychological insights which came to be neglected as the core ideas were mathematized by later economists to fit together neatly. Smith, for example, is famous in economics only for The Wealth of Nations, in which he suggests that people get their dinner "not from the benevolence of the butcher, the brewer, or the baker", but "from their regard to their own interest". However, Smith wrote an earlier book, The Theory of Moral Sentiments, all about the ways in which people care about others (see V. L. Smith, 1998). Why is the latter book virtually unknown, and the first so famous? Perhaps because the race to prove Pareto-optimality of competitive equilibria was so greatly simplified by assuming people's utilities depend only their own allocations. Having proved that, perhaps it is time to ask how economic analysis is changed by incorporating the insights in Smith's book on moral sentiments.

In the 1970s, cognitive psychologists began studying judgment and economic decision making. These studies took a different approach from the one Simon suggested. They took expected-utility maximization and Bayesian probability judgments as benchmarks, and used conformity or deviation from these benchmarks as a way to theorize about cognitive mechanisms. Important psychology of this sort was done by Ward Edwards in the 1950s, and later by Amos Tversky, Daniel Kahneman, Baruch Fischhoff, Paul Slovic, and many others. Because the output of this research often consisted of psychological principles or constructs that could be expressed in simple formal terms, this sort of psychology provided a way to model bounded rationality which is more like standard economics than the more radical departure that Simon had in mind. Much of behavioral economics consists of trying to incorporate this kind of psychology into economics.

A good example of how the cognitive psychology improves economic predictions is the "prospect theory" which Kahneman and Tversky proposed as an alternative to expected utility theory. The central principle in prospect theory is that people adapt to hedonic sensations, and therefore, utilities are determined by gains and losses from some reference point, rather than by overall wealth. Many studies suggest behavior toward losses and gains is different in two ways: Losses are disliked about twice as much as equal-sized gains ("loss-aversion"), and people often seek risk in the domain of losses when they can "break even" (i.e., reach the reference point), while they avoid risk in the domain of gains (the "reflection effect"). In addition, in expected utility theory, attitudes toward risk are expressed solely by curvature of the utility function. In prospect theory (and many other alternative theories), risk attitudes are also influence by nonlinear weighting of probabilities-for example, a person could buy a lottery ticket, even if her utility function for money outcomes is concave, if she overweights the small chance of winning. Indeed, the hypothesis that small probabilities are given too much weight (which is backed by many experiments) can explain why people with concave utility for gains would love high-skewness lotteries with a tiny chance of winning, and also explains why people who gamble over losses would nonetheless buy insurance against small chances of disastrous losses.

It is crucial to note that prospect theory is not an ad hoc customization of standard theory designed to fit a few experimental data. Nonlinear weighting of probabilities, differential sensitivity to gains and losses, and reflection can all be justified by more basic psychophysical principles which characterize a wide range of human behavior. For example, nonlinear weights result if people are unable to discriminate among probabilities equally well (e.g., going from a zero chance of winning a lottery to a .0001 chance seems like a bigger leap than going from .0001 to .0002). Seen this way, expected utility effectively assumes that people discriminate among probabilities equally well, throughout the range from impossible to certain, which is highly implausible. In addition, prospect theory has an axiomatic underpinning, and it has been compared with many other alternative theories (and with expected utility) using sophisticated econometric tests on thousands of experimental choices (see Camerer, 1995). And prospect theory has proved useful, or at least inspiring -- the 1979 paper by Kahneman and Tversky is the most widely-cited publication in Econometrica.

A common concern among the economists who are skeptical about behavioral economics (who are increasingly few in number) is that the ideas are too informal and fragmented to serve as a basis for economic theory. We take this concern seriously because the goal of behavioral economics is indeed to find parsimonious principles which explain field data. But we think more research will prove this pessimistic prediction wrong. In fact, recent research has already produced theories which are candidates to replace standard theory-while maintaining formal structure and reasonable parsimony-in seven crucial areas: (1) Utility maximization could be replaced by theories of reference-dependent preference (in which preferences exist, but are sensitive to current consumption or another reference point) and by theories of preference "construction"; (2) Expected utility theory can be replaced by prospect theory; (3) Subjective expected utility theory (in which "personal" probabilities are expressed by judgments, rather than derived from objective evidence) can be replaced by theories with non-additive probability; (4) Discounted utility can be replaced by "hyperbolic discounting", in which very short-term discount rates are much higher than future discount rates, reflecting a temporary impatience or impulsiveness; (5) Bayesian updating could be replaced by "support theory" or by formalizations of cognitive heuristics like availability (easily retrievable information is overweighted) and representativeness (hypotheses which are well-represented by evidence are thought to be likely); (6) theories of self-interest can be replaced by theories of "social preference" (e.g., Rabin, 1993); (7) and theories of equilibrium behavior can be replaced by (or perhaps justified by) theories of adaptive learning (e.g., Camerer and Ho, 1999).

These seven tools are arguably the most important ones in the economists' toolbox for modeling individuals. In each case, alternative theories have been proposed, mostly with good backing from a wide range of experimental data. Some of these theories are more formal and well-developed than others: Hyperbolic discounting and prospect theory are best-established, while alternative to Bayesian updating and theories of preference construction will take a lot more work. Nonetheless, it seems like only a matter of time before these tools prove useful in explaining and predicting field phenomena, and finding their way into economics books. For example, prospect theory has already proved capable of explaining ten different phenomena discovered in field data, from stock market pricing anomalies to downward-sloping labor supply and asymmetric price elasticity. An optimistic long-term prediction is that we will look back, decades from now, and regard assumptions like exponential discounting, self-interest, or even equilibrium as special examples of more general theories, which are convenient for some kinds of modeling, much as Cobb-Douglas production or homothetic preferences are special examples of more general functions which we often assume for simplicity.

Only a few economists dared to call themselves behavioral economists. The first generation was led by Richard Thaler, who has collaborated with Kahneman on many projects and has worked on a remarkable range of mainstream economic problems, including consumer reactions to price changes, savings-consumption behavior, and stock market. (Kahneman, along with Eldar Shafir and Drazen Prelec, are important among psychologists for their contributions to behavioral economics.) A generation or so behind Thaler, and clearly influenced by him, are myself, Linda Babcock, Catherine Eckel, George Loewenstein, and Matthew Rabin. The young turks include David Laibson, Terry Odean, and Sendhil Mullinaithan. I've defined behavioral economists as those who take direct inspiration for theorizing from psychology. Others who have questioned rationality and suggested new approaches, usually drawing less directly from psychology, include George Akerlof, Bob Frank, and Bob Shiller. Many, many other economists have also done work which fits under the broad definition above, by being sensitive to psychological realism (and data) when choosing assumptions. Prominent among them are many women economists, including Sheryl Ball, Rachel Croson, Elizabeth Hoffman, Rachel Kranton, Annamaria Lusardi, and Lise Vesterlund. Indeed, so much interesting new research could be placed under this heading that someday soon the term "behavioral economics" will no longer be a useful label. That's precisely the goal! The point is not to truly create a separate approach or field but, instead, to impose more psychological discipline on economic theorizing, which relied so much on assumptions of unbounded computation, willpower, and self-interest for a long time as economists struggled to get the mathematical underpinnings down pat. © Colin Camerer

Top >