We estimate the effects of a large program of public investment subsidies to private firms in Italy. Investment projects presented by firms were ranked by numerical scores reflecting both objective measures of project quality and the discretion of local politicians, and they were funded according to such ranking until exhaustion of available funds. We estimate that subsidies increased investment of marginal firms near the cutoff by 39 percent, and employment by 17 percent over a 6-year period. Building on recent advancements in the econometrics of regression discontinuity designs, we characterize treatment effect hetero-geneity and cost-effectiveness of subsidies across inframarginal firms. Smaller firms exhibit higher employment growth upon receiving the subsidy, but larger firms generate more jobs at a lower cost, and younger firms exhibit do better than older firms. Using the estimated distribution of treatment effects, we compute the effects of the policy under different allo-cation criteria. Eliminating political discretion would reduce the cost per job by 10 percent, and adopting an optimal criterion based on a set of observable firm characteristics would reduce the cost by over one third relative to the actual policy.