A macro-prudential perspective posits a link between bank fundamentals and the likelihood of banking crises. We articulate this link by developing a dynamic model with bank runs as coordination failures. Introducing a global game, the model endogenizes a bank run probability as a function of bank fundamentals including leverage.

The model generates pro-cyclical leverage and shows that credit growth tends to precede banking crises, replicating the empirical ending of Schularick and Taylor (2012). Countercyclical leverage restrictions can improve social welfare by reducing the crisis probability despite dampened economic activities in normal times.