Credit Risk Analysis and Security Design

Abstract

We consider the security design problem of a lender who can assess the borrower’s project prior to making an accept or reject decision. The lender’s subjective assessment is represented by a private signal. Unless the lender extracts the full surplus from the project, her cutoff signal above which she is willing to accept the project is inefficiently high, i.e., the lender is too conservative. The unique optimal security is standard debt. Debt maximizes the lender’s payoff from financing bad–i.e., low-signal–projects, thus implementing a lower cutoff signal than other securities. While the lender could, in principle, make the loan terms indirectly contingent on the signal by choosing a security from a prespecified menu, such ex-post fine-tuning is generally not optimal. Rather, it is optimal to either grant credit at standardized terms or not at all. Our model suggests a natural segmentation among lenders, whereby inside (i.e., local or relationship) lenders attract low-NPV borrowers while arm’s-length lenders attract high-NPV borrowers.