In this paper, a fully choice-based theory of disappointment is developed. It encompasses, as particular cases, EU theory, Gul's theory of disappointment (1991) and the models of Loomes and Sugden (1986). According to the new theory, the risk premium of a random prospect is the sum of two premiums: a concavity premium that is nothing but the usual Arrow-Pratt premium and a second premium that may be identified to expected disappointment. The corresponding representing functional belongs to the class of lottery-dependent utility models (Becker and Sarin 1987) since disappointment is the deficit between the utility of the realized outcome and its expected value. However, unlike the lottery-dependent approach, the theory is choice-based and its axioms are experimentally testable.
