On the (Non-)Uniqueness of Random Non-Expected Utility

Abstract

In random expected utility (Gul and Pesendorfer, 2006), the distribution of preferences is uniquely identified from random choice. This paper investigates whether such identification extends beyond expected utility. We first show that when risk preferences conform to the disappointment aversion model of Gul (1991), the distribution of preferences remains uniquely identified. To assess the scope of this result, we then examine other models of non-expected utility. Within the broader class of betweenness preferences (Dekel, 1986), random utility can be unidentifiable. If preferences are confined to the weighted expected utility class (Chew, 1983), a more nuanced picture emerges: unique identification holds in a three-prize setting but fails with four or more prizes. These findings show that the uniqueness property of random expected utility may persist beyond expected utility, but its persistence critically depends on the class of risk preferences under consideration.