Both informal and formal loans matter in agriculture. However, formal lenders provide many more production loans than informal lenders, often at a cost (mostly loan default cost) higher than what they can recover. For example, the Agricultural Development Bank of Pakistan (ADBP), providing about 90% of formal loans in rural areas, incurs high loan default costs. Yet, like other governments, the Government of Pakistan supports the formal scheme on the grounds that lending to agriculture is a high risk activity because of covariate risk. Hence, such policies are often based on a market failure argument. As farm credit schemes are subsidised, policy makers must know if these schemes are worth supporting. Using a recent large household survey data from rural Pakistan (Rural Financial Market Studies or RFMS), we have attempted to estimate the effectiveness of the ADBP as a credit delivery institution. A two-stage method that takes the endogeneity of borrowing into account is used to estimate credit impact. Results reveal that ADBP contributes to household welfare and that its impact is higher for smallholders than for large holders. Nevertheless, large holders receive the bulk of ADBP finance. The ADBP is, thus, not a cost-effective institution in delivering rural finance. Its cost-effectiveness can be improved by reducing its loan default cost and partially by targeting smallholders in agriculture where credit yields better results. © 2003 Elsevier Science B. V. All rights reserved.