9
Validation of Credit Portfolio Models
Portfolio credit risk models produce a probability distribution for portfolio credit losses (and gains, if it is a mark-to-market model). To validate the quality of a given model, we can examine whether observed losses are consistent with the model's predictions.
Some people argue that portfolio models are difficult or even impossible to validate empirically. Usually, such an opinion is justified by a comparison to market risk models. Market risk models produce loss forecasts for a portfolio (which might be the trading book of a bank) as well, but the underlying horizon is much shorter – often, it is restricted to a single day. A standard validation procedure is to check the frequency with which actual losses exceeded the Value at Risk (VaR). In a market risk setting, risk managers usually examine the 99% VaR, which is the loss that is predicted not to be exceeded with a probability 99%. Over one year containing roughly 250 trading days, the expected number of exceedances of the 99% VaR is 250 × (1 – 0.99) = 2.5, provided that the VaR forecasts are correct. When we observe the number of exceedances differing significantly from the expected number, we can conclude that the predictions were incorrect. Significance can be assessed with a simple binomial test.
Obviously, such a test is not very useful for the validation of credit portfolio models, which mostly have a one-year horizon. We would have to wait 250 years until we gain as many observations ...
Get Credit Risk Modeling Using Excel and VBA with DVD now with the O’Reilly learning platform.
O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.