Decentralized Targeting of an Agricultural Credit Program in India: Political versus Private Intermediaries

Résumé

Decentralized implementation of development programs is typically motivated to increase accountability and use of local information by service providers. Most developing countries have delegated implementation to elected local governments. However this has created problems arising from elite capture and clientelism of local governments. In this paper we conduct a field experiment which compares delegation of an agricultural credit program in West Bengal, India to an intermediary agent appointed by local governments (GRAIL), to one where the intermediary is selected randomly from private traders operating in the local area (TRAIL). Both agents are formally empowered to select recipients of subsidized loans, and are incentivized based on commissions linked to loan repayments. Both programs raise agricultural output by approximately 25 percent, but only TRAIL is successful in raising borrower incomes. The relative success of TRAIL owes partly to selection of more productive borrowers. GRAIIL selection is distorted by political motives of clientelism and cronyism. We also find evidence that informal roles of monitoring and assistance played by the TRAIL agent augmented the performance of selected borrowers. Differences in behavior between private (TRAIL) and political (GRAIL) agent can be explained by differences in their respective information/expertise and incentives. The results in the paper suggest the possible value of selecting private intermediaries to implement development programs, accompanied by suitable incentive schemes. (joint work with Pushkar Maitra, Sandip Mitra and Sujata Visaria)