RISK AVERSION AND EXPECTED UTILITY THEORY: AN EXPERIMENT WITH LARGE AND SMALL STAKES

Authors


  • The editor in charge of this paper was Orazio Attanasio.

  • Acknowledgments: We would like to thank Endemol Italia S.p.A for kindly providing access to recorded episodes of the show Affari Tuoi. In particular, we would like to thank Luca Passerini and Salvatore Spirlí for extensive discussion about the game rules and the selection process of the participants as well as Silvia Brasca for facilitating the data collection. We thank Paul Beaudry, Gorkem Celik, Drew Fudenberg, David Green, Jerry Green, Yoram Halevy, David Laibson, Thomas Lemieux, and Okan Yilankaya for useful comments and suggestions. In addition, Philippe Aghion, Alberto Alesina, and Andrei Shleifer are gratefully acknowledged for numerous discussions. Kareem Carr provided excellent research assistance. We are also grateful to seminar participants at the University of British Columbia, at the 12th Conference on Foundations and Applications of Utility, Risk and Decision Theory, and at Harvard University.

  • E-mail: matildeb@interchange.ubc.ca (Bombardini); ftrebbi@mail.ubc.ca (Trebbi)

Abstract

We employ a novel data set to estimate a structural econometric model of the decisions under risk of players in a game show where lotteries present payoffs in excess of half a million dollars. The decisions under risk of players in the presence of large payoffs allow us to estimate the parameters of the curvature of the von Neumann–Morgenstern utility function—not only locally, as in previous studies in the literature, but also globally. Our estimates of relative risk aversion indicate that a constant relative risk aversion parameter of about 1 captures the average of the sample population. We also find that individuals are practically risk neutral at small stakes and risk averse at large stakes—a necessary condition, according to Rabin’s calibration theorem, for expected utility to provide a unified account of individuals’ attitudes toward risk. Finally, we show that for lotteries characterized by substantial stakes, nonexpected utility theories fit the data equally as well as expected utility theory.

Ancillary