Health insurers frequently impose supply-side policies in the form of ‘prior authorization’ to manage healthcare spending. Prior authorization requires providers to fill out paperwork before treatment is eligible for coverage. The stated purpose of these policies is to reduce healthcare spending by encouraging the use of lower-cost treatments of similar quality, and to ensure treatment complies with established guidelines. However, there are concerns that prior authorization may discourage needed care. Using all-payer claims data from Massachusetts in 2009-2013, we estimate the effect of prior authorization on the use of specific drugs in MassHealth, the state Medicaid fee-for-service program. Using difference-in-differences estimation, we compare Medicaid beneficiaries affected by changes in prior authorization requirements to individuals in plans of a major commercial insurer unaffected by these policy changes. We find that prior authorizations lead to large reductions in utilization of drugs that have clear substitutes. These reductions are fully offset by increases in utilization of cheaper but equally effective drugs. However, when clear substitutes are not available, there are reductions in utilization that do not lead to substitution to similar drugs. Prior authorization reduces both high- and low-value use of drugs, suggesting that it is not well targeted.