Forecasting the oil futures price volatility: Large jumps and small jumps

Abstract

Macro news drives jumps, however, a jump does not seem to improve the predictability of the simple heterogeneous autoregressive realized volatility model (HAR-RV) in the oil futures market. This paper provides a new insight and seeks to investigate whether truncated jumps can help improve the forecasting ability compared to that achieved using the HAR-RV model and its various extensions with jumps. Our results provide strong evidence that the models incorporating both large and small jumps gain a significantly superior forecasting ability. Specifically, including small jumps in a high-frequency model significantly improves the forecast accuracy at the 1-day forecasting horizon, while including both large and small jumps can achieve a higher forecast accuracy at the weekly and monthly horizons. These findings reveal that considering the decomposed jumps with a certain threshold can increase the forecast accuracy of the corresponding model.