Using data from the RAND Health Insurance Experiment, the authors examined the effect of mental health insurance on individuals' mental health outcomes, use of mental health services, and economic welfare. Variation in cost sharing induced no significant change over time in mental health status averaged over all beneficiaries. However, there were interactions among plan, initial mental health status, and income. Those with initially good mental health status who were poor had better outcomes under cost sharing than under free care. Those who initially had poor mental health status had a relatively more favorable response to free care, but the research did not determine whether they were absolutely better off under free care. The authors also considered the tradeoff between the loss from overconsumption of mental health care because it is insured and the gain from reducing families' financial risks that stem from unanticipated use of mental health services. The results suggest that for an individual or family, the optimal coinsurance rate for outpatient mental health services is between 25 and 50 percent with a $1,000 ceiling annually on out-of-pocket payments for individuals, or $2,000 per family for all health services (in 1986 dollars) subject to an annual limit of 52 covered outpatient mental health visits.