A note on concave utility functions - Rice （关于凹的效用函数大米）.pdf

A note on concave utility functions1 Martin Monti Simon Grant Daniel Osherson Princeton University Rice University Princeton University April 1, 2004 1Contact information: M. Monti, Dept. of Psychology, Green Hall, Princeton Univer- sity, Princeton NJ 08544. Fax: 609-258-1113. Electronic mail: mmonti@princeton.edu, sgrant@rice.edu, osherson@princeton.edu. Abstract The classical theory of preference among monetary bets represents people as expected utility maximizers with nondecreasing concave utility functions. Critics of this account often rely on assumptions about preferences over wide ranges of total wealth. We derive a prediction of the theory that bears on bets at any fixed level of wealth, and test the prediction behaviorally. Our results are discrepant with the classical account. Competing theories are also examined in light of our data. JEL classification: D81, C91. keywords: gambling, risk aversion, concave utility function, expected utility, prospect theory A note on concave utility functions 1 An influential theory of preferences among bets represents people as expected utility maximizers with nondecreasing concave utility functions. In what follows, we shall call anyone who behaves this way a classical agent. The theory that people behave towards bets as if they were classical agents has been the subject of intense discussion, with alternative hypotheses prompted by experimental findings at variance with the classical account.1 A new kind of objection has recently been formulated by Rabin (2000a,b; Rabin Thaler, 2001). Let (g, p, ) denote the bet yielding gain $g with probability p and loss $ with probability 1 − p . Rabin deduces predictions of the