A revealed-preference implication of weighted utility decisions under uncertainty

In a novel formulation of revealed preference analysis, Green and Osband [8] show that for expected-utility maximizers, acts partition the state-simplex into linear polyhedral blocks. The question naturally arises whether this characterization distinguishes expected utility theory from non-expected utility theories. This paper investigates the weighted utility theory of Chew [2] and shows that the corresponding partition is systematically different from the expected utility theory: the boundaries of the partition blocks are quadratic rather than linear. This result contains useful empirical contents.