Monitoring targeting performance when decentralized allocation to the poor are unobserved

National antipoverty programs often rely heavily on provincial governments. The center targets poor provinces in the hope that they will reach their own poor. Without successful intraprovincial targeting, however, even dramatic redistribution from rich to poor provinces can have little impact on poverty nationally. However, data for assessing performance at provincial level are often far from ideal. Can a centralized government monitor the performance of decentralized social programs in reaching the poor when their benefit incidence is unobserved? The author shows that the poverty map and the corresponding spending allocation across geographic areas allow one to identify the latent differences in mean allocations to the poor versus the nonpoor. The national measure of targeting performance is also subgroup-decomposable. The author uses an application to an antipoverty program in Argentina (Trabajar II) to assess performance in reaching the poor and the measure the relative contributions to the program's performance - before and after reforms - of the center's provincial reallocation and decentralized targeting. Funding and program design changes led to large gains for the poor, although with diverse performance across provinces. Program funding and design choices by the central government can greatly affect the targeting performance of decentralized social programs. The allocation to a province should depend on how successful it is at reaching the poor with the extra resources, rather than how poor it is. Design choices should provide incentives for provincial governments to target resources to the poor . Finding feasible ways to monitor their performance and adjust ventral government's efforts accordingly are then crucial to better outcomes for poor people.