LAUSR

This work aims at contributing to the improvement of the early warning systems of banking crises using a new approach accounting for model uncertainty. We show that a multinomial logit model based on Bayesian model averaging (BMA) is a good strategy to predict banking crisis. To do this, we argue that differences in vulnerability to banking crisis can be largely explained by an asymmetry between financial market evolution and regulation update on a sample of 49 developed and developing countries between 1980 and 2010. When markets are liberalized, competition pushes bankers to take more risks and take advantage of regulatory delays thus increasing crises probabilities. Our empirical evidence supports that crisis probability is higher in country liberalizing their banking system when regulation is not updated. We developed an early warning system for systemic banking crises based on the multinomial logit model. Its main difference to existing prediction models and its contribution to the literature is that it is intended to identify and resolve what is called by Bussiere and Fratzscher [(2006). Towards a new early warning system of financial crises. Journal of International Money and Finance, 25(6), 953–973] as post-crisis bias in binomial models and to develop a new methodology of leading indicators selection based on BMA. Overall, our model predicts all banking crises during our sample period.