Asymmetric Information and Unobserved Heterogeneity in the Accident Insurance

Testing for asymmetric informations has become very important in recent years in order to test theoretical predictions and to outline new directions for research. In this paper we analyse accident insurance, which has not been analysed before in the literature, but covers one of the most important risks faced by individuals in modern society. We show that there is asymmetric information, but the extent depends on the amount of coverage. Moreover, and most importantly, the option of choosing an annual adjustment of the insured sum has strong predictive power both for the occurence of an accident and the chosen coverage, although it should be irrelevant from the point of theory. This results shows new ways to design contracts and variable selection for risk classification. In contrast to previous studies, we also explicitly take into consideration unobserved heterogeneity by applying finite mixture models and so called 'unused' observables. As shown by DeMeza and Webb (2001), the absence of a positive correlation between risk and coverage is still consistent with asymmetric information if unobserved risk anversion influences both the insurance demand and the risk type