Energy commodities: A study on model selection for estimating Value-at-Risk

Abstract

Changes in commodity prices can be transmitted directly to the real economy through changes in the marginal cost of production. Therefore, it is extremely important to create some mechanism to protect against these movements in the commodities futures market. Exposure in this market comes along with tail risk, which must be measured and controlled using a risk measure. To help economic agents, this research provides a common statistical specification that can be used to reliably predict the Value‐at‐Risk of four important energy commodities. For this, the predictions of a range of 48 competing models, composed of four heteroskedastic specifications, six conditional distributions, and a Markov chain with up to two regimes, were compared using various statistical tests, and the model with the best average results was preferred.