Two Tests for Ex Ante Moral Hazard in a Market for Automobile Insurance

Empirically separating the phenomena of moral hazard and adverse selection in insurance markets has occupied researchers in this field for decades. Recently, the potential benefits of using survey data instead of claims data to control for the different dimensions of private information when testing for evidence of asymmetric information have been explored in the insurance literature. This article extends that approach to present two tests for ex ante moral hazard in a market for automobile insurance. In this article we specify (1) a recursive model and (2) an instrumental variables model to address endogeneity with respect to policy selection in cross-sectional road traffic crash (RTC) survey data. We report a statistically significant ex ante moral hazard effect with both models. This result is then subjected to a falsification test, whereby the analysis is repeated in subsamples of at-fault and not-at-fault RTCs. Our antitest produces no evidence of ex ante moral hazard in the sub-sample of not-at-fault RTCs, in which the true moral hazard may reasonably be assumed to be zero, thus supporting the interpretation of the results of our two models. Our extension of the existing literature via these two specifications may have useful analogs in other insurance markets for which survey data are available.