Forecasting crude oil futures volatility with extreme-value information and dynamic jumps

Abstract

In this paper, we propose the realized EGARCH model with jumps (hereafter REGARCH-Jump model) to model and forecast the crude oil futures volatility. A key feature of the proposed REGARCH-Jump model is its ability to account for the extreme-value information as well as time-varying jump intensity. We apply the REGARCH-Jump model to the Brent crude oil futures price data. Our empirical results provide evidence of the presence of time-varying jumps in the crude oil futures market. More importantly, we show that our proposed REGARCH-Jump model outperforms the GARCH, EGARCH, HAR, and REGARCH models in terms of both empirical return fit and out-of-sample volatility forecast. Moreover, the superior forecast performance of the REGARCH-Jump model is robust to alternative out-of-sample forecast windows. Finally, a Value at Risk (VaR) analysis demonstrates the economic value of the improved volatility forecasts from the REGARCH-Jump model. In summary, our findings highlight the importance of accommodating the extreme-value information and jump dynamics in forecasting the volatility of crude oil futures prices.