Access to and use of health services are concerns in poor countries. If implemented correctly, health insurance may help solve these concerns. Due to selection and omitted variable bias, however, it is difficult to determine whether joining an insurance scheme improves medical care–seeking behaviors. This paper uses representative data for the whole country of Ghana and an instrumental variable approach to estimate the causal impact on healthcare use of participating in Ghana’s National Health Insurance Scheme. Idiosyncratic variations in membership rules at the district level provide exogenous variation in enrollment. The instrument is the existence of nonstandard verification methods to allow enrollment of children. Using the 2008 Ghana Demographic and Health Survey and a census of all district insurance offices, this paper finds that insurance membership increases the probability of (1) seeking higher-quality (but no greater quantity of) maternal services and (2) parents’ becoming more active users of child curative care. Instrumental variable estimates are larger than ordinary least squares ones, indicating that “compliers” have much higher returns to being insured than the average participant. Results are robust to several validity checks; this paper shows that the instrument is indeed idiosyncratic and proves that government officials did not establish less-cumbersome membership rules in districts with worse initial indicators.