The Structure of Credit Markets with New Screening Technologies (with Pablo Kurlat and Maryam Farboodi)

We develop a model of credit market competition with endogenous screening technologies and interest rates, and use it to study two implications of the technological transformation reshaping credit markets. First, lenders deploying similar AI systems and overlapping data may produce increasingly correlated screening errors. We show that correlated mistakes lead ex-ante identical lenders to endogenously specialize across distinct market segments, resulting in a hockey-stick interest rate schedule that echoes the coexistence of traditional banking, fintech lending and private credit, and high-rate indiscriminate lending. Within each segment, lenders charge lower rates and face fewer non-performing loans than absent specialization; yet because credit supply reallocates toward higher-rate segments, the average borrower may end up paying more. Second, technological and regulatory changes affect screening costs unevenly. Big data innovations expand financial inclusion, while broader Open Banking adoption can harm the financially excluded and increase inequality in financial access.