Predicting European Bank Stress Tests: Survival of the Fittest

This paper develops an early warning system to predict whether European banks pass stress tests conducted by regulatory agencies. All banks participating in stress tests in 2010, 2011 and 2014 are examined. Using an AdaBoost classifier ensemble approach with financial ratio and macroeconomic variables, we are able to identify over 98 percent of failing and passing banks in the training subsample and predict about 90 percent of banks in the test validation sample. Additional analyses of predictor importance and robustness compared to other competing model approaches are conducted. Based on our findings, we conclude that ensemble methods can help identify the underlying risk dimensions of individual banks.