We analyze reductions in bank credit using a natural experiment where unprecedented flooding in Pakistan differentially affected banks that were more exposed to the floods. Using a unique data set that covers the universe of consumer loans in Pakistan and this exogenous shock to bank funding, we find two key results. First, following an increase in their funding costs, banks disproportionately reduce credit to borrowers with little education, little credit history, and seasonal occupations. Second, the credit reduction is not compensated by relatively more lending by less-affected banks. The empirical evidence suggests that a reduction in bank monitoring incentives caused the large relative decreases in lending to these borrowers.

Banks in low-income countries face severe liquidity risk due to volatile deposits, which destabilize their funding, and dysfunctional liquidity markets, which induce expensive interbank and central bank lending. Such liquidity risk dissuades the transformation of short-term deposits into long-term loans and deters long-term investment. To validate this mechanism, we exploit a Sharia-compliant levy in Pakistan, which generates unintended and quasi-experimental variation in liquidity risk, with data from the credit registry and firm imports. We find that banks with a stronger exposure to liquidity risk lower their supply of long-term finance, which reduces the long-term investment of connected firms.