Modeling and forecasting the oil volatility index

This paper models and forecasts the crude oil ETF volatility index (OVX). Themotivation lies on the evidence that the OVX has been used in the last years as an important alternative measure to track and analyze the volatility of future oil prices. The analysis of the OVX suggests that it presents similar features to those of the daily market volatility index. The main characteristic is the long range dependence that is modeled either by autoregressive fractional integrated moving averaging (ARFIMA) models or by heterogeneous autoregressive (HAR) specifications. Regarding the latter family of models, we first propose extensions of the HAR model that are based on the net and scale measures of oil prices changes. The aim is to improve the HAR model by including predictors that better capture the impact of oil price changes on the economy. Second, we test the forecasting performance of the new proposals and benchmarks with the model confidence set (MCS) and the Generalized-AutoContouR (G-ACR) tests interms of point forecasts and density forecasting, respectively. Our main findings are as follows: the new asymmetric proposals have superior predictive ability than the heterogeneous autoregressive leverage (HARL) model under two known loss functions. Regarding density forecasting, the best model is the one that includes the scale measureas a proxy of oil price changes and considers a flexible distribution for the errors.