Journal of Risk

Farid AitSahlia
Warrington College of Business, University of Florida

Modeling comovements while avoiding the use of copulas and scrutinizing the plausibility of stress test scenarios carried out by the US Federal Reserve are among the topics discussed in this issue of The Journal of Risk. They are complemented by a paper on the calibration of stochastic volatility models and one on the estimation of future values-at-risk and initial margins.

“How to build a risk factor model for non-life insurance risk”, our first paper, is by Alessandro Ferriero, who introduces an alternative to copulas as a way to capture loss dependencies in non-life insurance portfolio lines affected by common factors. He relies on the theory of infinitely divisible distributions to support this approach and shows numerically how it can capture nonsymmetric dependencies and multidimensional structures where standard copulas – such as the Clayton, Gumbel and Gaussian – could not.

In the issue’s second paper, “Regularization effect on model calibration”, Mesias Alfeus, Xin-Jiang He and Song-Ping Zhu compare two calibration methods for two stochastic volatility models (the stochastic alpha beta rho (SABR) model and the Heston model) by assessing the regularization effect on out-of-sample pricing accuracy. On the basis of Nasdaq 100 index data, their study shows that regularized calibration is effective only for long-term horizon pricing.

In “Estimating future value-at-risk from value samples, and applications to future initial margin”, the third paper in the issue, Narayan Ganesan and Bernhard Hientzsch compare a variety of methods for estimating future values-at-risk and dynamic initial margins. They highlight in particular the effects of violations of moment constraints and those of additional inner samples, which they address, and they suggest approaches for improvement. They also propose the use of pseudo-inner samples instead of actual inner samples in order to enhance the accuracy and speed of methods such as nested Monte Carlo and Johnson percentile matching.

Our last paper, “Severe but plausible – or not?”, is by Stefan Gavell, Mark Kritzman and Cel Kulasekaran, who rely on the Mahalanobis distance, used to measure statistical unusualness, to determine the plausibility of stress test scenarios devised by the US Federal Reserve in light of the Covid-19 pandemic. Based on standard statistical assumptions, the authors show that these scenarios are practically implausible but offer suggestions for their slight modification to remedy their shortcomings. They further propose an approach to make the distribution of the Mahalanobis distance consistent with empirical evidence, thus avoiding the assumption of normality.