Abstract This paper examines the choice between using capital and using output subsidies to promote socially desirable production. We exploit a natural experiment in which wind farm developers were unexpectedly given the opportunity to choose between an upfront investment subsidy and an output subsidy to estimate the differential impact of these subsidy margins on project productivity. Using instrumental variable and matching estimators, we find that wind farms choosing the capital subsidy produced 11 to 12 percent less electricity per unit of capacity than wind farms selecting the output subsidy, and that this effect is driven by subsidy incentives rather than selection. We then use these estimates to evaluate the public economics of this policy shift, finding that the introduction of capital subsidies caused the Federal government to spend 18 percent more per unit of output from wind farms than they would have paid under an output subsidy. The potential for increased entry and longer-run distortions is also discussed.