Empirical tests of weighted utility theory

Abstract

Cumulative empirical evidence indicates that expected utility theory fails as a general representation of actual choice behavior under uncertainty. In particular, tests that involve a problem structure introduced by Allais (1953) show that individuals often violate expected utility's linearity assumption (i.e., strong independence axiom). In recent decades, many theories of choice under uncertainty have been proposed as alternatives to expected utility. This paper describes and tests weighted utility theory, a generalization of expected utility in which the independence axiom is weakened such that nonlinear choice behavior, including the Allais paradox, can be accounted for. The tests use a generalization of Allais' problem structure which demands consistency over a greater number of choice situations. The results indicate the superior predictive ability of a particular nonlinear variation of weighted utility referred to as the "light" hypothesis.