A Bayesian Approach to Uncertainty Aversion

Abstract

The Ellsberg paradox demonstrates that peoples belief over uncertain events might not be representable by subjective probability. We relate this paradox to other commonly observed anomalies, such as a rejection of the backward induction prediction in the one-shot Ultimatum Game. We argue that the pattern common to these observations is that the behavior is governed by rational rules. These rules have evolved and are optimal within the repeated and concurrent environments that people usually encounter. When an individual relies on these rules to analyze one-shot or single circumstances, paradoxes emerge. We show that when a risk averse individual has a Bayesian prior and uses a rule which is optimal for simultaneous and positively correlated ambiguous risks to evaluate a single vague circumstance, his behavior will exhibit uncertainty aversion. Thus, the behavior predicted by Ellsberg may be explained within the Bayesian expected utility paradigm.