Journal of Risk Model Validation

Steve Satchell

University of Cambridge

It is interesting that the big risk around as I write this editorial is sovereign risk, yet none of the four excellent papers in this issue address sovereign risk modeling or the validation of sovereign risk models. Before readers conclude that this is evidence of irrelevance, I note that there seems to be an historical sequence in risk model validation; namely, validation often occurs as a consequence of model prediction failure. That is, by analogy to medicine, we cure disease rather than prevent it. Research on validation typically follows some evidence or perception of model failure. However, without making too strong a delineation at this stage, some of the papers that follow might be seen to have a forward-looking aspect to them.

The fourth volume of the journal commences with a paper by Skoglund, Erdman and Chen on the performance of value-at-risk (VaR) during the financial crisis. There has been a great deal of anecdotal talk concerning VaR and its failures in the crisis, so it is interesting to see some proper analysis on the topic. This is a topic where risk validation has a great deal to add. For example, there has recently been a focus on the length of the look-back period; the importance of the look-back period of VaR is well discussed elsewhere, such as in the Basel Committee’s risk-based capital proposals (issued in January and July 2009). Further validation work could be done on backtesting the length of the look-back period and the parameters of VaR could also be investigated.

In a similar vein, our second paper looks at the validation of loss given default (LGD) in the context of retail portfolios; the authors of this paper are Hlawatsch and Reichling. The validation of LGD is very much part of the internal ratings based approach for retail loan portfolios and is a key component of Basel II. As the authors point out, there is a shortage of models in this area and they provide a number of examples.

The third paper, by Huschens, Lehmann and Tillich, examines a portfolio tranche structure and calculates the hitting and wipe-out probabilities of each tranche. The underlying model structure is that of a Vasicek-distributed portfolio loss. The key feature of such a model is a common latent risk factor. Whilst this may seem overly simplistic, it does lead to tractable analysis and the use of a tranche structure gives it some extra flexibility. Indeed, the calculation of hitting probabilities becomes complicated very quickly as we try to add structure to our models, so the above analysis is of great value in establishing “benchmark” calculations.

The final paper, by Cespedes, Herrero, Rosen and Saunders, presents an analysis of exposure at default (EAD) – a key component of counterparty risk. The authors provide calculations of important parameters of EAD, such as wrong way risk, the alpha multiplier and counterparty credit capital. The calculations are done within the regulatory rules of Basel II internal ratings-based methodology. I would hope that this paper, indeed all four papers, would be of immediate value to practitioners. I have been delighted on my occasional forays into industry to find groups of lenders quoting from papers in our journal.

Editors are cursed by a requirement that they should be minor prophets. One feels a responsibility to let the readers know where the subject is heading or should be heading. My feeble pronouncement is that we will be getting papers on the validation of sovereign risk models before too long.