Preference Types, Limited Attention and Welfare in Property Insurance Markets

This paper builds a procedure to partially identify and estimate a mixture model of decision making under risk, where households behave either according to standard expected utility theory or according to the dual theory of choice under risk (Yaari, 1987). The model allows for unobserved heterogeneity in households' attitudes towards risk and attention levels within each preference type. It is estimated on novel data on households' deductible choices across three lines of property insurance coverage, and the estimation results are used to measure the welfare implications of policy interventions in these markets. Specifically, a restructuring of the auto coverage lines offered to households is considered, where two lines of coverage currently offered as separate products are combined into a single product. Although prima facie such an intervention might appear welfare reducing by restricting the choice set of households, average household welfare may increase due to non-expected utility behavior of a subset of households in the sample. The proposed procedure leverages advances in random set theory to overcome identification problems due to heterogeneity in households' attention levels. It is computationally tractable and informative about the underlying preference structure, the shares of various preference types and attention levels, and the welfare effects of the intervention of interest.