Backtesting, stress testing and sensitivity analysis

Rafael Cavestany

So far, we have described an end-to-end statistical process of estimating capital requirements for operational risk. In this chapter, we will look at the backtesting and stress testing of such models, to provide a quality control and validation of completeness of the operational risk capital model. The backtesting represents an ex post validation of the accuracy of the modelling, and compares the new experienced operational risk losses with those predicted by the models used during the capital estimation. On the other hand, stress testing estimates the potential losses of adverse operational risk scenarios and serves as an additional capital adequateness validation.

OPERATIONAL RISK BACKTESTING

Backtesting is a necessary analysis in any risk estimation, and one that provides us with an ex post evaluation of the precision of the calculations. Backtesting of operational risk capital estimations involves the challenge of backtesting risk estimations performed for a one-year time horizon and a high confidence interval of generally 99.9%. This implies that more than 1,000 yearly observations of total operational losses would be required for the direct backtesting of operational