Separating Moral Hazard from Adverse Selection in Automobile Insurance: Longitudinal Evidence from France

This paper uses longitudinal data to perform tests of asymmetric information in the French automobile insurance market for the 1995-1997 period. This market is characterized by the presence of a regulated experience-rating scheme (bonus-malus). We demonstrate that the result of the test depends crucially on how the dynamic process between insurance claims and contract choice is modelled. We apply a Granger causality test controlling for the unobservables. We find evidence of moral hazard which we distinguish from adverse selection using a multivariate dynamic panel data model. Experience rating appears to lead high risk policyholders to choose contracts that involve less coverage over time. These policyholders respond to contract changes by increasing their unobservable efforts to reduce claims.