Anatomy of international banking crises at the onset of the Great Recession

Abstract

The paper examines a wide range of potential predictors of 25 international banking crises that broke out in 2007–2011 on the basis of cross–sectional logit models and the BCT (binary classification tree) algorithm, a novel technique in assessing the causes of banking crises. The major determinants of the crises arise from an excessive credit depth (measured as private credit to GDP ratio) and illiquidity of the banking sector (credits to deposits ratio). The implementation of explicit deposit insurance schemes is also a pro-crisis factor due to the moral hazard effect they tend to cause. On the contrary, higher values of remittance inflows to GDP decrease the susceptibility to banking crises. These findings are robust under both methodologies. Lower bank concentration, bigger values of cost to income ratios as well as a higher level of economic liberalization and inflation make countries more vulnerable to banking crises, as derived from the logit analysis. The pre-crisis credit depth, credits to deposits ratio, inflation, financial openness and net interest margin are also significant predictors of the crisis costs proxied by the peak ratio of non-performing loans (NPL) relative to gross loans, the increase in public debt to GDP ratio and real output losses.