Reference dependent risk preferences and insurance Demand

Abstract

This thesis explores insurance decisions with respect to modest risks. Bernoulli’s expected utility theory is compared to a model where utility depends on both final wealth and changes in wealth relative to some reference point. Optimal insurance is derived within these frameworks. An important result in the expected utility theory is that full insurance is only optimal at actuarially fair premia. I show that when utility also depends on some reference point, full insurance may be optimal at unfair premia and no insurance may be optimal at fair premia. The aim of the study is to investigate which model is best suited to explain preferences over smallscale insurance contracts. The analysis is based on a survey experiment of 904 Norwegian citizens, representative of the general population, in the spring of 2020. The respondents were asked to choose between hypothetical income gambles and menus of small-scale insurance contracts. The results display strong indications that the reference dependent model is superior to expected utility theory when predicting insurance choices. The majority of the sample made insurance choices that can be explained by the reference dependent model, whereas only seven percent made choices that are in line with expected utility theory. There also appears to be high heterogeneity in both the degree of risk aversion and in people’s reference points. Among those respondents whose choices are in line with the reference dependent model, approximately half seem to have full insurance as their reference point and prefer full insurance at unfair premia, whereas the other half seem to have no insurance as their reference point and prefer no insurance at fair premia.