Do generalized expected utility theories capture persisting properties of individual decision makers?

Abstract

Expected utility theory and two generalizations of it (rank-dependent utility theory and cumulative prospect theory), which have been proposed to describe individuals' risky choice behavior, have been examined. A characteristic feature of these models is subjective weighting of outcomes and probabilities. This study investigates whether individually determined weighting functions (utility functions and weighting functions for probabilities) capture individual traits of decision makers. To verify this, temporal constancy of individual subjective weighting functions and predictability of individuals' future decisions were considered. Sixty-four subjects had to choose between the same 150 gamble pairs in three separate sessions, with an interval of one week between sessions. Descriptive analyses showed considerable temporal inconsistency in the individuals' decision behavior. It was, therefore, not surprising that the analysis of temporal constancy of individual weighting functions yielded frequent rejections of the constancy assumption. An additional analysis was conducted, predicting decision behavior in later sessions using individual weighting functions estimated from earlier sessions. This analysis takes the inconsistency into account explicitly and suggests that the models' weighting functions, indeed, seemed to capture persisting properties of the decision makers. While the assumption of subjective weighting of outcomes (as in expected utility theory) proved very important for predicting future decisions, the additional assumption of subjective weighting of probabilities (as in rank-dependent utility theory or cumulative prospect theory) improved predictability only to a minor degree.