Evaluation of similarity models for expected utility violations

A body of work proposes a decision cost argument to explain expected utility (EU) violations based on pair similarity. These similarity models suggest various measures over the risky pairs that define decision costs and benefits. This paper assesses the empirical modeling success of these similarity measures in explaining risky choice patterns showing EU independence violations. We also compare model fit for these similarity models relative to EU and to a selected generalized EU model. Although the candidate models exhibit some degree of substitutability, our results indicate support for models that use relatively simple measures as instruments for similarity.