Backtesting Analysis. How to Assess the Quality of PD Models in a Retail Banking

Abstract

The paper refers to the probability of default model validation procedure in retail banking. The author presents the idea of backtesting analysis focusing on sensitivity analysis of capital requirements under stress scenarios. The paper addresses statistical methods which can be applied in credit risk management under the backtesting exercise in retail banking. The advantages and drawbacks of specific approaches are discussed. Furthermore, the outcomes of the empirical implementation of selected methods are presented. The author considers the impact of positive asset correlation on various validation approaches, where no correlation is assumed, and proves that the zero-correlation assumptions may result in a more prudent approach. This finding was confirmed by the empirical analysis performed for retail portfolios. The research concerned PD parameters calculated for car and mortgage loans. The backtesting results revealed that PD forecasts created for mortgage portfolios underestimated credit risk during the crisis period which started in 2008. However, car loan portfolio credit risk predictions appeared to be robust.(original abstract)