Dual approaches to the analysis of risk aversion

We present a dual formulation of choice under uncertainty based on a few simple assumptions about preferences. It is shown that the additive separability restriction on preferences, key to expected-utility theory, can be dropped with little loss of analytic power for a broad class of choice problems. Dual risk premiums are characterized, and it is shown that placing various invariance restrictions on them leads naturally to generalizations of the concepts of CARA, CRRA, and LRT familiar from expected-utility theory. Each of these generalizations conforms to a notion of homotheticity