Pure Risk

Abstract

We introduce a behavioral notion of domain-specific risk aversion that separates attitudes toward risk from deterministic utility: an agent is more pure risk averse in one domain than in another if, for prizes that are indifferent under certainty, he is more averse to risk in the former domain than in the latter. We develop a model that goes beyond expected utility by allowing risk attitudes to vary across domains, while preserving expected utility within each domain. The domains are subjective and need not be specified in advance; they are identified from choice behavior. We establish uniqueness of the model's parameters and provide an axiomatic characterization.