Does adverse selection matter? Evidence from a natural experiment

The empirical evidence of adverse selection in insurance markets is mixed. The problem in assessing the extent of adverse selection is that private information, on which agents act, is generally unobservable to the researcher, which makes it difficult to distinguish adverse selection from moral hazard. Unique micro data, from a dental insurance natural experiment, is here used to provide a direct test of selection. The setup of the insurance makes it possible to observe a proxy for private information. Interestingly, results differ across the distribution of risk, with direct evidence of both adverse selection and advantageous selection. This dual selection can explain the limited empirical evidence for adverse selection in the literature: the two effects may balance out on the aggregate level. The paper also presents a model of insurance choice that can harbor both adverse and advantageous selection, explaining the pattern in the data by differences in the effectiveness of prevention across high and low risks.