Banking Crisis Prediction with Differenced Relative Credit

Abstract

Indicators based on the ratio of credit to GDP have been found to be highly useful predictors of banking crises. Differences in this ratio seem a highly promising early warning indicator. We test a large number of slightly different versions of the differenced credit-to-GDP ratio with data on euro area members. The optimal time interval of the difference is about two years. Using the moving average of GDP over several years rather than the latest annual data is shown to have little impact on forecasting performance. The proposed indicator demonstrates particular promise at relatively short forecasting horizons (2–3 years).