Journal of Risk Model Validation

Our fall issue contains a number of innovative papers that we hope are of great interest to our readers. A feature of a number of these papers is that they offer new methodologies with attractive empirical properties. Since these are hard to verify at a nonempirical level, I have tended to follow the papers' abstracts in describing the features of each paper. I hope readers will forgive this journalistic lapse.

The first paper in this issue, "Individual and flexible expected shortfall backtesting" by Marcelo Brutti Righi and Paulo Sergio Ceretta, proposes a new expected shortfall backtesting approach. The authors present a Monte Carlo simulation algorithm to determine the significance of the backtest, and they provide an empirical illustration that demonstrates the advantages that their backtests provide. The results suggest that it is not necessary to wait for an entire backtest period in order to prove the prediction that the expected shortfall test is inefficient.

In our second paper, "Toward model value-at-risk: bespoke CDO tranches, a case study", Pierre Cohort, Pierre-Emmanuel Levy ditVehel and Frédéric Patras discuss the importance of model risk. They claim that, in spite of recent advances, unlike market risk and credit risk, model risk still lacks a clear and largely accepted methodology. The authors argue that mainstream research has to be supplemented by new insights and methods and they propose a new method, called model value-at-risk, that hopes to answer questions such as, what is the model-dependent value-at-risk of an investment at one year at a confidence level of 95%?

In the issue's third paper, "Multirating decision model validation: the relevance of the quality of securitization issues", Miguel Á. Peña-Cerezo, Arturo Rodríguez-Castellanos and Francisco J. Ibáñez-Hernández discuss how ratings enable the information asymmetry that exists in the issuer-investor relationship to be reduced, particularly for issues with a high degree of complexity (such as securitizations). However, there may be serious conflicts of interest between the issuer's choice and remuneration of the agency and the credit rating awarded, and this results in the rating having lower quality and information power. The authors propose a model of the number of ratings requested, by analyzing the relevance of the number of ratings to measure the reliability, where multirating is shown to be associated with the quality, size, liquidity and the degree of information asymmetry relating to the issue. This is real-world problem; while rating agencies have been universally criticized, few constructive solutions have been offered. The journal is delighted to publish constructive solutions in this area. Some readers may feel that this is a little tangential to model validation, but my view is that rating agencies, while typically looking at the overall process, should validate any model within their rating process. I am therefore happy to include papers on rating issues in The Journal of Risk Model Validation.

Our final paper, "Expected loss and Impact of Risk: backtesting parameter-based expected loss in a Basel II framework" by Wolfgang Reitgruber, discusses the dependency structure of credit risk parameters: an important factor in determining capital consumption. The author addresses the topic of the lack of availability of an unbiased estimator of risk expense and claims that no established backtesting procedures for expected loss currently exist. He offers a practically oriented, top-down approach to assess the quality of expected loss by backtesting with a properly defined risk measure. He makes the following claim.

The proposed method will deepen the understanding of the practical properties of expected loss, reconcile the expected loss with a clearly defined and observable risk measure and provide a link between upcoming IFRS 9 accounting standards for loan loss provisioning and the regulatory capital requirements under the internal ratings-based approach (IRBA). The method is robust irrespective of whether parameters are simple, expert-based values or highly predictive and perfectly calibrated IRBA-compliant methods, as long as the parameters and default identification procedures are stable.

Readers can decide for themselves whether or not they agree with this statement, but it is clearly of interest to both practitioners and academics.

On a more general note, The Journal of Risk Model Validation would also welcome model validation insights from emerging markets. It is clear that there is much innovation occurring outside the old economies and I would be delighted to include papers from the BRICs or other fast-growing economics.

Steve Satchell
Trinity College, University of Cambridge