Nonexpected utility and coherence

The descriptive power of expected utility has been challenged by behavioral evidence showing that people deviate systematically from the expected utility paradigm. Since the end of the 70's several alternatives to the classical expected utility paradigm have been proposed in order to accommodate these deviations. This thesis examines from a theoretical and experimental point of view the rank-dependent models, currently the most influential alternative to expected utility. The rank-dependent models propose a natural extension of expected utility by permitting nonlinear weighting of probabilities through decision weights. This thesis investigates the intuition underlying the rank-dependent models, the extension of de Finetti's coherence principle to the rank-dependent case, and the experimental elicitation of the decision weights. In addition, an alternative generalization of expected utility, the gambling effect model, is presented.