We conduct a field experiment in India comparing two approaches to appointing a local commission agent to select eligible smallholder farmers for a subsidized credit program: a private trader in TRAIL, versus a political appointee in GRAIL. Although both schemes had similar loan take-up and repayments and similar treatment impacts on borrowing and farm output, only TRAIL raised farm profits significantly. This cannot be explained by greater connectedness between TRAIL agents and farmers, or differential patterns of borrower selection.