From Uniform Expected Utility to Uniform Rank-Dependent Utility: An experimental study

Abstract

This paper presents an experimental investigation of the Uniform Expected Utility (UEU) criterion, a model for ranking sets of uncertain outcomes. We verified whether the two behavioral axioms characterizing UEU, i.e., Averaging and Restricted Independence, are satisfied in a pairwise choice experiment with monetary gains. Our results show that neither of these axioms holds in general. Averaging in particular, appears to be violated on a large scale. On the basis of the current study and a previous one, we can conclude that none of the models for set ranking that have been axiomatically characterized so far is able to model observed choices between sets of possible outcomes in a satisfactory fashion. In this paper we therefore lay out the foundations for a new descriptive model for set ranking: the Uniform Rank-Dependent Utility (URDU) criterion.

Rank Dependent Weighted Average Utility Models for Decision Making Under Ignorance or Objective Ambiguity

“…It seems unlikely that such a news would increase (even weakly) the attractiveness of the treatment from the patient's point of view. Vridags and Marchant (2015) have shown that the very fact of ruling out both forms of extremism from the part of a RDWAU decision maker-a requirement referred to as dominance in (Barberà, Bossert, and Pattanaik, 2004)-leads to a significant restrictions of the rank dependent weights used by him or her when evaluating decisions with different numbers of outcomes. Specifically, the following Proposition is proved in Vridags and Marchant (2015) (Proposition 1) (with mild changes in the notation and definition).…”

“…Hence, a pessimistic patient who gives more weights to the samples where the treatment performs poorly to those where it performs well could very well choose the third treatment over the second even though he has chosen the first treatment over the second. In the only instance we know where the averaging axiom has been tested in an experimental context (Vridags and Marchant, 2015), it has been rejected by a vast majority of subjects.…”

“…The criteria analyzed can be viewed as variants of the rank-dependent expected utility family originally proposed by Quiggin (1982) (see also Quiggin (1993)) that are suitable to the considered finite set theoretic framework. These criteria were hinted at in Vridags and Marchant (2015) where they were referred to as Uniform Rank Dependent Utility criteria because they are equivalent to a RDU model applied to a uniform probability distribution. However, in this paper, we use the more explicit name of Rank Dependent Weighted Average Utility (RDWAU) to designate these criteria.…”