Coordinated failure? A cross-country bank failure prediction model

This paper empirically investigates the causes of bank failures in Japan and Indonesia. Using logistic regression analysis of financial ratios, we explore the usefulness of domestic bank failure prediction models with a cross-country model that allows for cross-correlation of the error terms. Our results suggest that loans, both as a ratio to total assets, deposits and in some cases the ratio of non-performing loans, are the most significant predictors of bank failure in both Japan and Indonesia. Regulatory capital ratios, on the contrary, do not seem to be significant indicators of failure. In addition to the domestic models, we explore the usefulness of a cross-country model of bank failure prediction and find that this model outperforms the domestic models on several diagnostic tests.