Long Run Macroeconomic and Sectoral Determinants of Systemic Banking Crises

In a panel comprising 61 countries covering the years 1980-2010 we show that macroeconomic variables such as GDP and real-interest rates lose potency as systemic banking crisis determinants when estimated over a full business cycle and that the choice of panel time-span is of high relevance. Using a shorter panel (1998-2011) involving 75 countries, we show that sectoral variables such as Bank Z-Score, private-credit-to-GDP ratio, bank credit-to-deposit ratio and non-performing loan levels represent an improved model-fit over their macroeconomic-focused counterparts, yielding improved in-sample crisis predictions. Whereas sectoral-centric models may over-estimate the likelihood of systemic banking crises this does not constitute a model weakness if not overlooking embryonic crises is the key objective. Future research is facilitated via the establishment of a control cluster of determinants with both sectoral as well as macroeconomic constituents.