Model Secrecy and Stress Tests

Abstract

Should regulators reveal the models they use to stress-test banks? In our setting, revealing leads to gaming, but secrecy can induce banks to underinvest in socially desirable assets for fear of failing the test. We show that although the regulator can solve this underinvestment problem by making the test easier, some disclosure may still be optimal (e.g., if banks have high appetite for risk or if capital shortfalls are not very costly). Cutoff rules are optimal within monotone disclosure rules, but more generally optimal disclosure is single-peaked. We discuss policy implications and offer applications beyond stress tests.