Monitoring targeting performance when decentralized allocations to the poor are unobserved

How can a central government monitor the performance of a decentralized pov erty program when the incidence of the program's benefits is unobserved at the local level? This article shows that, using a poverty map and the corre sponding spending allocation across geographic areas, one can identify the latent differences in mean program allocations to the poor and the nonpoor. The national measure of targeting performance can also be decomposed into subgroups. An application to an antipoverty program in Argentina is used to assess the program's performance before and after reforms. Increases in fu nding and changes in program design brought large gains to the poor, althou gh Performance differed across Provinces.