Agents' insurance behavior: a non-additive decision model approach.

Summary The objective of this thesis is to study the insurance behavior of agents whose preferences in an uncertain world are represented by non-additive models that generalize the utility expectation model and allow the separation of the attitude towards wealth from that towards uncertainty. This representation of preferences, together with the assumption that agents do not know precisely the probability distribution of their losses, makes the modeling of insurance behavior more realistic and allows a better understanding of the choices observed in the markets. The first chapter of the thesis is devoted to the presentation of choice models in uncertain probabilistic and non-probabilistic conditions. It is followed by the study of the demand for insurance in the classical framework of a single risk, first when the agent is able to locate the probability distribution of his losses within a set of given distributions and his preferences are represented by the Jaffray model, then when he is in a general framework of non-probabilized uncertainty and his preferences are represented by the utility expectation à la Choquet. In the third chapter, we extend the scope of the study by considering the demand for insurance in the presence of an additional, uninsurable risk. The impact of this additional risk on the demand for insurance is studied in the case where the two risks are comonotone and anti-comonotone. We then focus on the supply behavior of an insurer in an anti-selection model, when the information of the insureds on their loss probabilities is imprecise. The last chapter of the thesis is devoted to an empirical study of the determinants of the purchase of life insurance contracts, with the objective of testing the impact of agents' beliefs on their actual choice of insurance contracts.