Lending with Hard and Soft Information

Abstract

What is the effect of credit scoring on loan-officers banking? I construct a model in which digital lenders compete with traditional banks. Digital lenders have access to more precise credit scores (economies of scale in credit scoring), but traditional banks can employ information gathered by loan officers at a lower cost (diseconomies of scale in the acquisition and transmission of soft information). I find that digital lenders cream-skim the best firms in the market and displace traditional banks from a lending segment where traditional banks are more efficient, leading to an inefficiently low amount of soft information acquisition that translates into higher rates of default. I also show the conditions under which this cream-skimming inefficiency more than offsets the cost-reduction gains introduced by credit scores. A per-loan Pigouvian tax on digital lenders combined with a revenue-neutral lump-sum rebate eliminates the inefficiency. The paper provides a set of empirical predictions that match the received empirical evidence and new testable hypotheses.