How much do our neighbors really know? The limits of community-based targeting

Abstract

Social assistance programs in developing countries often rely on local community members to identify potential beneficiaries. As community members may observe neighbors’ welfare, their reports may capture transitory shocks better than the proxies typically observable by a centralized policy implementer. To test this, we conduct a lab-style experiment in Central Java, in which participants rank other community members’ welfare, using benchmarks that vary in sensitivity to transitory shocks, and target small cash transfers. We find little evidence that community-held welfare information better reflects transitory shocks and find that targeting decisions mostly depend on perceived differences in overall wealth.