What determines output losses after banking crises?

We examine the output costs associated with 150 banking crises using cross country data for years after 1970. Many banking crises do not lead to contractions and most banking crises do not lead to large contractions, a result that holds for developed and developing economies. We examine which variables help to predict output changes after a banking crisis using Bayesian Model Averaging. For developed economies, we find that the output losses are positively related to prior economic conditions such as credit growth. For low-income economies, we find that other factors such as having a stock market and deposit insurance are more important.