Accounting for Uncertainty in Decision Weights for Experimental Elicitation of Risk Preferences

Abstract

In the generalized expected utility framework, the multiplicative relationship between preferences and beliefs complicates the identification of risk preferences. In experimental or field settings, the respondent's decision weight must be known with certainty to confidently infer accurate attitudes towards risk. Factors such as probability weighting or the influence of past experiences may result in an individual applying a decision weight that differs from the probability used to infer risk preferences. We propose a novel Monte-Carlo based method for expressing uncertainty in the individual's decision weight as uncertainty in their inferred risk aversion coefficient. We implement this procedure on experimentally elicited risk preferences that were obtained via a mail survey and show that this procedure improves model fit when the risk aversion coefficient is used as a determinant of behavior in a reduced form model of insurance demand.

Keywords: Risk Preferences, Field Experiment

JEL Classification: C9, C83, D81