Failure of the Credit Organisation

The question addressed in this thesis is whether it is possible to construct a model based on the financial and macroeconomic variables. That model has to accurately predict future defaulters as well as healthy banks based on the specific criteria. Classical econometric methods fail due to the number of assumptions when the outcome of the event is of binary nature (default or live). That is why this study employs logit model. Model will be constructed using data from 2004 to 2011 and then tested on periods of different length (half year; year; two years) of 2012-2013. The main findings of this research are: half annual periods are optimal, inclusion of macroeconomic variables improves predictive power, and absolute value of H1 coefficient doesn’t matter by itself, the highest predictive power obtained (measured by Specificity estimate) was 67.74%.