How Useful are Regime-Switching Models in Banking Crises Identification?

We employ a regime-switching approach to the identification of banking crises. This approach reduces the arbitrariness in crisis identification by endogenizing the choices of crisis threshold and crisis duration. Using a sample of 47 countries, we show that this approach also subject to several same problems as the common procedures. The method is sample-dependent, tends to invent much more crises, and is less robust to different model specifications. We conclude that a bind application of regime switching model to crisis identification is questionable