Financial crises in developing countries and early warning systems: Markov switching modelling

The large number of financial crises in developing countries that erupted in the recent years and the consequences of these crises have led to an increased interest in the subject of designing early warning systems. In this study we propose Markov switching model, also known as the regime switching model, as an alternative approach to build early warning systems for financial crises of a series of Latin American (i.e., Argentina, Brazil, Mexico), South Eastern Asian (i.e., Korea, Malaysia, Thailand) and Eastern and Southern European (i.e., Bulgaria, Russia, Turkey, Ukraine) countries during the period 1990:M1-2006:M12. For this aim, we provide a survey of the theoretical and empirical literature on currency crises in order to assess empirical definition of ?crisis? and the initial set of potential indicators. Then, we present a detailed review of standard and alternative methodologies for early warning systems. The primary result of our analysis is that the Markov-switching models are adequate to build reliable early warning systems of financial crises. The results of our models perform well in anticipating and giving proper warning that crisis is going to occur. On the basis of this analysis, the most useful early warning indicator seems to be the deviations from the real effective exchange rate. The Banking Sector Fragility Index and LIBOR (3-months London Interbank Offer Rate) are the other two important indicators distinguishing crises from the non-crises periods. In view of the results, the indicators, which are important in determining crisis likelihoods for one country, are unimportant for another country due to heterogeneity across developing countries.