Market risk management in commodity market: an application with the value at risk method

Aim of this study is to determine the appropriate distribution and models for VaR forecasted by the Variance-Covariance Approach used for quantifying market risk in commodity markets, and also to assess of the models' success in forecasting market risk by backtesting. In this study the series of daily returns on gold, silver, copper, aluminum, Brent and West Texas Intermediate (WTI) oil, natural gas, wheat and soybean are used for the period January 2003-November 2013. VaR is calculated by the Variance-Covariance method with the symmetric and asymmetric generalized autoregressive conditional heteroskedasticity (GARCH) models based on normal distribution, student-t distribution and generalized error distribution (GED). Analysis results related with metal commodities suggest that the models based on student-t distribution are more succesful and have more accurate predictions of VaR for gold and silver, while the models based on normal distribution for copper, and the models based on both normal and GED distributions for aluminium seem more successful. For energy commodities, normal distribution models indicate better performance, whereas models based on student-t and GED distributions tend to overestimate the real market risk. For agricultural commodities, like in the case of metals, student-t and GED distributions seem to make more reliable predictions. Furthermore, the performance of asymmetric GARCH models appears to be better than symmetric models in commodity markets. This finding shows that positive and negative information do not influence commodity prices similarly. This feature should be taken into account during predictions.