Journal of Risk Model Validation

Steve Satchell

University of Cambridge

This is the first issue from volume 2 and, following our recently established tradition, we offer the reader an eclectic mix. We are constantly reminded by circumstances around us that it is hard to put boundaries on the subject of risk management and it should follow that it is hard to put boundaries on risk model validation. At this point, it would be easy to lapse into musings about a black swan; I shall resist the impulse.

Editors are both blessed and cursed: they are blessed because they can air their prejudices in a fairly unchecked way and they are cursed because people may actually read the editorials and remind them about what they said in the past. As a case in point, I feel the pain of the then Editor-in-Chief of The Journal of Risk (Volume 9, Issue 1, Fall 2006), who wrote in his editorial that “for several years, volatility and risk premiums in financial markets globally have been shrinking. Brief periods of heightened volatility have not shaken a growing belief that a new, and better, ie, safer, world economic regime has emerged.” In fact, he goes on to warn against too much complacency, but, nevertheless, it does remind us how quickly things can turn around.

Our first paper, by Wong et al, is a valuable contribution on stress-testing credit exposures of retail banks to macroeconomic shocks. This is a difficult problem and one of the reasons why the whole industry has been wrong-footed has been the sudden transition from a benign macro-environment to a rather worrying one. The second paper, by Paolella and Steude, is not (obviously) directly about model validation. However, it is about how to estimate models using estimation procedures that explicitly weight observations of interest, such as down-turns, more highly than the average observation. This is connected very clearly to my earlier discussion and the first paper, since it allows us to build and validate models using more of the bad event information than the good.

The third paper, by Lingo and Winkler, questions the usefulness of one of the fundamental tools of model validation, the accuracy ratio (in fact, it looks at a number of related characteristics). The authors conclude that discriminatory power is of value as part of a validation exercise.

Finally, Dowd looks at the fan charts for the Riksbank inflation density forecast. He illustrates how such a forecasting model can be evaluated in terms of model adequacy. As part of the broader theme of providing macroeconomic scenarios for risk model validation, this paper makes a useful contribution to the risk model validation literature.

All of the above can be thought of in terms of the role of macroeconomics in determining outcomes for the variables that risk models try to forecast. As such, risk model validation includes issues of macroeconomic forecasting.