Selection of Value at Risk Models for Energy Commodities
In this paper we investigate different VaR forecasts for daily energy commodities returns using GARCH, EGARCH, GJR-GARCH, Generalized Autoregressive Score (GAS) and the Conditional Autoregressive Value at Risk (CAViaR) models. We further develop a Dynamic Quantile Regression (DQR) one where the parameters evolve over time following a first order stochastic process. The models considered are selected employing the Model Confidence Set procedure of Hansen et al. (2011) which provides a superior set of models by testing the null hypothesis of Equal Predictive Ability. Successively information coming from each model is pooled together using a weighted average approach. The empirical analysis is conducted on seven energy commodities. The results show that the quantile approach i.e. the CAViaR and the DQR outperform all the others for all the series considered and that, generally, VaR aggregation yields better results.