Modeling Bank Panics: Challenges

Our primary finding is that surprisingly small changes in assumptions which determine the amount of net worth available in a bank panic have an important impact on the nature of the equilibria: there may not be a bank panic at all, or there may be several di erent panics of di erent severity. The economic reasons for this sensitivity are clarified by transforming the market economy into a game and studying banker best response functions. To establish robustness to model details, we report similar quantitative results across three di erent model specifications and calibrations. A second, additional result, is displayed in a three-period version of the panic model of Gertler and Kiyotaki (2015). That model naturally suggests the idea that welfare can be improved by imposing a restriction on bank leverage. We compute the Ramsey-optimal leverage restriction, but find that there is an implementation problem: the restriction can be associated with more than one equilibrium, not just the desired one. We discuss one way to address the implementation problem.