Forecasting the volatility of crude oil futures: The role of oil investor attention and its regime switching characteristics under a high-frequency framework

Abstract

The purpose of this article is to investigate whether oil investor attention (OA), measured by Google search volume, contains incremental information content to predict crude oil futures volatility under high-frequency heterogeneous autoregressive (HAR) model specifications. Moreover, to account for possible structural breaks and nonlinearity in the relation between OA and crude oil volatility, this article extends HAR-type models with regime switching considerations. The results of parameter estimation and out-of-sample prediction show that the in-sample and out-of-sample performance of HAR-type and Markov switching (MS)-HAR-type models with OA is significantly better than that of their corresponding HAR-type and MS-HAR-type models without OA. Furthermore, our findings suggest that (i) HAR-type-OA models tend to produce better forecasts for the volatility of the crude oil market at short horizons (1-day) compared to HAR-type, MS-HAR-type and MS-HAR-type-OA models. (ii) MS-HAR-type-OA models have the best forecasting performance at relatively long prediction horizons (1-week and 1-month). Therefore, the result suggests that the OA and regime switching specifications have a significant positive impact on volatility predictions and can be useful for improving the performance of HAR-type models.