Banking Crises as Rare Events

This paper argues that the estimations coming from the vast literature on the determinants of the banking crises' occurrence probability could be significant biased. Bias is associated with the simple fact the banking crises are strange and infrequent events. In fact, once these biases are corrected, the probabilistic models, commonly used to prevent the occurrence of a banking crises' episode, face substantially increased its predictive power. This is relevant in the early warning systems' design since huge cost, associate with the banking crises' occurrence, could be reduced if it is opportunely identified. (Audio: Department of Economic Publications)