How costly is the misallocation of production that we might expect to result from distortions such as market power, incomplete contracts, taxes, regulations, or corruption? This paper develops new tools for the study of misallocation that place minimal assumptions on firms’ underlying technologies and behavior. We show how features of the distribution of marginal products can be identified from exogenous variation in firms’ input use, and how these features can be used both to test for misallocation and to quantify the welfare losses that it causes. We then consider an application in which thousands of firms experience demand shocks derived from a lottery-based assignment of public procurement contracts for construction services in Ecuador. Using administrative tax data about these firms, we reject the null of efficiency but estimate that the welfare losses resulting from misallocation are only 1.6% relative to the first-best. Standard parametric assumptions applied to the same setting would suggest losses that are at least an order of magnitude larger.
