Commodity value-at-risk modeling: comparing RiskMetrics, historic simulation and quantile regression

Abstract

Commodities constitute a nonhomogeneous asset class. Return distributions differ widely across different commodities, both in terms of tail fatness and skewness. These are features that we need to take into account when modeling risk. In this paper, we outline the return characteristics of nineteen different commodity futures during the period 1992-2013. We then evaluate the performance of two standard risk modeling approaches, ie, RiskMetrics and historical simulation, against a quantile regression (QR) approach. Our findings strongly support the conclusion that QR outperforms these standard approaches in predicting value-at-risk for most commodities.

Modeling the return distribution of salmon farming companies: A quantile regression approach

“…This will give an indication of the performance of the VaR estimates, and hence, the robustness of the estimated tails of the return distributions. We are using a standard backtesting procedure (Steen et al, 2015) which assesses the in-sample performance of the VaR-model. In-sample assessments of predictive power in general overstates the out-of-sample predictive power a statistical model and thus caution is required.…”

“…This is the opposite of hypothesis testing for significance in ordinary regression models. See, e.g., Steen et al (2015) for a more detailed presentation.…”

Value-at-risk in the European energy market: a comparison of parametric, historical simulation and quantile regression value-at-risk

“…Previous studies on using QR in various financial markets include, among others, those of Steen et al (2015) and Haugom et al (2016). Steen et al (2015) compare the efficiency of VaR forecasts using twenty different commodities and three different VaR models, among them a QR approach. They find that the QR approach outperforms the HS method as well as the standard RiskMetrics approach.…”

“…They find that the QR approach outperforms the HS method as well as the standard RiskMetrics approach. Haugom et al (2016) use a similar QR approach to that of Steen et al (2015). In their paper, the authors employ a parsimonious QR model to forecast one-day-ahead VaR in both commodity markets and more traditional financial assets.…”

“…We wish to contribute to this literature by providing a study that compares the accuracy of different types of VaR models used to manage risk in the European energy market. Our analysis consists of comparing a RiskMetrics-inspired parametric model, a historical simulation (HS) model and a quantile regression (QR) model, inspired by Steen et al (2015). We have adjusted our HS method using a volatility weighting technique similar to that proposed by Hull and White (1998).…”