Risk Modelling in Energy Markets: A Value at Risk and Expected Shortfall Approach

Abstract

Value at risk (VaR) and Expected Shortfall (ES) are commonly used risk measures in the financial literature. They have however not been applied to a great extent on energy derivatives. This paper compares the performance of several VaR and ES models for energy commodity futures on some of the world s largest commodity exchanges. In total 14 different VaR models and nine ES models are evaluated; GARCH and GJR-GARCH with normal, student t, GED and skewed student t distributions and EWQR are used to obtain both VaR and ES forecasts. In addition, five CAViaR models are used in the VaR analysis.EWQR is by far the best ES model. It has very good test results for all markets and quantiles considered. The VaR results vary greatly, and there does not appear to be any clear pattern in which some models are better suited for certain markets or commodities. The VaR models with best performance overall are however EWQR, the adaptive CAViaR and GARCH and GJR-GARCH models with student t and skewed student t distributions.