Predicting Systemic Banking Crises: An Early Warning System Based on Panel Logit Approach

This paper presents the determinants of systemic banking crises and develops an early warning system using a dynamic panel logit model based on data from 42 countries between 1980 and 2021. Empirical findings reveal that the credit gap is one of the strongest leading indicators of banking crises. It was also found that economic growth and trade openness have significant effects on the probability of a crisis. Moreover, the model’s predictive performance was evaluated through a classification matrix and ROC curve analysis. The results indicate that the model could correctly identify about 78.0% of the systemic banking crises. The AUC value of 0.9488 obtained from the ROC analysis demonstrates that the model possesses a very high predictive capacity in distinguishing between crisis and non-crisis periods. Simultaneously, the low false alarm rate suggests that the model is able to limit unnecessary crisis signals while maintaining a satisfactory level of crisis detection. The evidence indicates that the model provides useful signals of potential crisis periods but should be used as a supportive policy instrument rather than a stand-alone decision-making tool.