Research in Progress
Information Production and Information Sharing: an Analysis of Credit Registries (with Rob Townsend)
Credit registries allow borrowers to build reputation collateral, which is particularly valuable in developing countries, where standard collateral is scarce. This kind of information sharing can discipline borrowers and allow banks to credibly promise not to exploit their good borrowers.
We study the impact of information sharing on the pool of information produced in the first place. We build a two period model with strategic default and moral hazard. There are two lenders and a continuum of borrowers with binary type: low types are always unprofitable, while high types realize a positive investment return with probability increasing in effort exerted. At the beginning of the period, a lender can pay to receive a signal of borrower type, where accuracy is costly. A lender observes the performance of her own borrowers. At the end of period 1, borrower performance and/or any signals gathered are reported truthfully to a credit registry.
Our model yields two key insights. First, in the absence of information sharing, a lender shifts screening costs to good borrowers. Rather than paying an information gathering cost herself, the lender sets interest rates and direct screening levels to induce high types to separate from low types by exerting costly effort. Default is a convincing signal of low type only if high types exert high effort in equilibrium. This in itself dampens high types’ incentives to exert effort. Hence, the information contained in borrower performance is endogenous to lending policy.
Second, information sharing influences the interest rates and screening levels set by lenders in period 1, where borrower performance, signals gathered, or both dimensions must be shared. In general, information sharing causes period 1 interest rates to be higher and informativeness of borrower performance to be lower, so that the quality of information entering the credit registry is low. However, banks use this information anyway, which amplifies the negative effect of low quality information. Empirically identifying this amplification and solving for the socially optimal design of the information sharing system are eventual goals of this work.