Rational Overoptimism (and Other Biases) Forthcoming in the American Economic Review

Rational agents with differing priors tend to be overoptimistic about their chances of success. In particular, an agent who tries to choose the action that is most likely to succeed, is more likely to choose an action of which he overestimated, rather than underestimated, the likelihood of success. After studying the comparative statics of this mechanism, I show that it also causes agents to attribute failure to exogenous factors but success to their own choice of action, to disproportionately believe that they will outperform others, to overestimate the precision of their estimates, and to overestimate their control over the outcome. (JEL A12, B49, C70, D81, D84) Human inference and estimation is subject to systematic biases. The evidence shows, for example, that people are overoptimistic about future life events (Neil D. Weinstein 1980, 1982, 1984 and the reviews therein), that they tend to attribute success to their own actions but failure to external factors (Miron Zuckerman 1979, Susan Fiske and Shelley Taylor 1991, Roy F. Baumeister 1998, Thomas S. Duval and Paul J. Silvia 2002), that they overestimate their contribution in joint projects (Michael Ross and Fiore Sicoly 1979), that they overestimate the precision of their estimates (Stuart Oskamp 1965), and that they also overestimate the degree to which they have control over an outcome (Ellen J. Langer 1975). The economics and psychology literature has suggested various explanations for these phenomena, which I discuss in more detail below. The purpose of this paper is to propose an alternative mechanism that is structurally different from the explanations forwarded by the existing literature. The basic idea is as follows. If agents sometimes overestimate and sometimes underestimate the probability of success of an action (relative to other agents) and try to select the action with the highest probability of success, then they are more likely to select actions of which they overestimated the probability of success. They will therefore tend to be overoptimistic (relative to these other agents) about the likelihood of success of the actions they undertake. This mechanism is similar ∗Sloan School of Management, MIT, Cambridge MA 02142 (e-mail:[email protected]). John Roberts’ suggestions and support made an enormous difference to me. Bengt Holmstrom’s remarks also had a big impact on this and other work. I got extensive suggestions from Mark Satterthwaite, Ján Zábojńık, and the anonymous referees, and very useful comments from Dan Ariely, Bill Barnett, Max Bazerman, Bob Gibbons, Thomas Hellmann, Ed Lazear, Jon Levin, Sendhil Mullainathan, Paul Oyer, Drazen Prelec, Emily Pronin, Garth Saloner, Antoinette Schoar, Joel Sobel, Muhamet Yildiz, and the participants in the Stanford Strategy Conference, the Harvard-MIT Theory Seminar, the MIT Organizational Economics Lunch, and the SITE conference. Financial support from the Interuniversitair Centrum voor Managementwetenschappen (ICM/CIM), Stanford Graduate School of Business, and MIT-Sloan are gratefully acknowledged.