Decomposing Automobile Insurance Policy Buying Behavior: Evidence of Adverse Selection

Abstract

This paper examines relationships between voluntary automobile insurance policy choice and risk classification in a heavily regulated market. We investigate the existence of self-selected risk classification based on a unique dynamic data set and test whether incurred claims in one year, a proxy of risk type, is informative about policy choice in the following year. By decomposing policyholders into high and low coverage and high and low deductible groups based on policy choice in the first year, we find evidence of an asymmetric information effect in the second year for low coverage, high deductible groups but not for high coverage, low deductible groups. Our results imply the coexistence of adverse selection and other behavior, including advantageous selection and habit persistence.