A Dual Approach to Ambiguity Aversion

In this paper, the assumption of monotonicity of Anscombe and Aumann (1963) is replaced by a weaker assumption of monotonicity with respect to first order stochastic dominance. I derive a representation result where ambiguous distributions of objective beliefs are first aggregated into “equivalent unambiguous beliefs” and then risk preferences are used to compute the utility of these equivalent unambiguous beliefs. Such an approach makes it possible to disentangle ambiguity aversion, related to the treatment of information, and risk aversion, related to the evaluation of the equivalent unambiguous beliefs. An application shows the tractability of the framework and its intuitive appeal.