Asymmetric information is a large potential problem for insurance markets, especially for markets insuring health risk. Despite a large theoretical attention over the last 30 years, the empirical evidence of the extent of the problem is still limited. In part this is due to methodological problems. The fundamental problem being that the private information, on which agents' act, is generally not observable to the researcher. This thesis provides empirical evidence on selection and moral hazard, using a private dental insurance natural experiment that solves many of the methodological problems. The initial decision to select into the insurance is analyzed in the first paper. Interestingly, results indicate both adverse selection and its opposite, advantageous selection, in different parts of the sample. These results are captured in a theoretical model where agents differ in their degree of risk aversion. In the second paper the decision to opt out of insurance is related to new information on risk. Results indicate that updating is asymmetric. Agents do not take the insurer's information fully into account, even though it is public. Finally, the moral hazard effect of dental coverage is analyzed in the last paper. The treatment effect of insurance coverage is separated from self-selection, using propensity score matching, IV and difference-in-differences. The results give strong empirical evidence of moral hazard. / Diss. Stockholm : Handelshögsk., 2004 xii s., s. 3-12: sammanfattning, s. 13-104: 3 uppsatser