This thesis focuses on the identification, propagation, and prediction of banking crises and on the relationship between banking and currency crises. We firstly modify the Von Hagen-Ho index and apply it to a large set of countries, excluding countries with a repressed financial system. The crises identified by our MMPI are more in line with the crises identified by Laeven and Valencia (2010), while the index also gives fewer ‘false alarms’. Then, we investigate the dynamic relationship between currency and banking crises in 94 developing and emerging countries using quarterly data from 1980Q1 to 2010Q4. The results show that in most cases currency crises tend to lead banking crises and vice versa which is robust for using different periods and different samples of countries. We investigate interdependence and spillover effects of financial turbulence across countries during the last decade. Our results suggest that financial turbulence has a significant interdependence effect across countries through the trade and distance channels, while a significant spillover effect through the capital flows channel is also identified. Finally, we compare the performance of the logit model and data mining models in predicting bank failures in the United States during 2002 to 2010. For all three models, the logit model issues more missed failures and false alarms ex post than data mining models, but issues fewer missed failures and false alarms ex ante.