Journal of Risk

Robustness is a running theme across the four papers in this issue of The Journal of Risk. Specifically, the variability in the credit valuation adjustment (CVA) used in many over-the-counter markets, the combined effects of interacting macroeconomic variables on regulatory stress testing, the accuracy of quantile estimates obtained via Cornish–Fisher approximation, and a method of robust risk-preference elicitation are addressed herein.

The great recession of 2008–9 illustrated the need for better estimates of CVA that account for counterparty risk. In our first paper, “Counterparty risk: credit valuation adjustment variability and value-at-risk”, Miche`le Breton and Oussama Marzouk use a recursive approach to determine the CVA’s distribution, and this leads to an efficient and robust estimation of it. As a result, the authors show that the ad-hoc assumptions commonly adopted in practice result in significantly inaccurate value-at-risk (VaR) estimates of CVA.

At present, supervisory stress tests involve scenarios for macroeconomic variables in such a way that their mutual effects are ignored. In “A generic stress testing framework with related economic shocks and possible regulatory intervention”, the second paper in this issue, Dror Parnes and Michael Jacobs Jr. propose a model that accounts for the interaction between the different factors used for regulatory stress testing. This approach is flexible enough that it can also help identify the best timings for regulatory interventions within this testing, such as preventative measures or emergency rescue actions, instead of viewing these as subsequent to stress testing.

VaR and conditional VaR have, despite their known shortcomings, become standard measures of risk used by both regulators and industry professionals. Related quantile estimates can be efficiently approximated via the Cornish–Fisher expansion. However, the nature of this approximation and the shape of the associated Cornish–Fisher distribution are poorly understood. In “Making Cornish–Fisher fit for risk measurement”, this issue’s third paper, John D. Lamb, Maura E. Monville and Kai-Hong Tee show how to correctly determine such a distribution given the first four moments, highlighting its unimodal form in most cases and its smoothness properties relative to portfolio weights. The authors also demonstrate how this distribution can be extended to higher dimensions and illustrate how it helps to correct underestimated risk.

Our final paper, “Measuring latent risk preferences: minimizing measurement biases” by Gosse A. G. Alserda, addresses a long-standing issue, namely that of eliciting risk preferences from individuals. Based on pension data, the author shows how the composite score of four common methods contains more information on risk attitude, and is more robust, than a factor-weighted or an unweighted combination of their separate results.

Farid AitSahlia
Warrington College of Business, University of Florida