Abstract:

Research on forecasting the price volatility of crude oil futures has focused primarily on foreign markets. Crude oil futures contracts in China are divided into three trading sessions within a single trading day, which differs greatly from those in foreign markets. The segmentation of trading hours may lead to differences in the structure of volatility between the Chinese market and foreign markets. Through the heterogeneous autoregressive realized volatility (HAR-RV) model framework and high-frequency futures contract trading data sampled at a half-second frequency, this paper studies the structural characteristics of price volatility and forecasting issues. It verifies the feasibility of forecasting over smaller time scales from daily to intraday trading session scales in the Chinese market. It is found that Chinese crude oil futures price volatility exhibits interval volatility characteristics, and the inclusion of interval features significantly improves the predictive performance of the model. In addition, it is also found that reducing the forecast horizon facilitates improved stability of the realized volatility series. This paper expands HAR-RV model research to non-stationary time series, and the volatility forecasts can enable investors and risk managers to develop more precise risk management tools tailored to the Chinese crude oil futures market.

Key words: Chinese crude oil futures, heterogeneous autoregressive-realized volatility (HAR-RV) model, time period characteristics, prediction, high-frequency data

摘要：

关于原油期货价格波动率的预测研究主要集中在国外市场，中国原油期货合约在一个交易日内被分为3个交易时段，这与国外市场有着很大的不同。交易时间分段可能使中国市场上波动率的结构与国外存在差异。通过异质自回归已实现波动率（HAR-RV）模型框架与半秒钟采样频率的高频期货合约交易数据，对价格波动率的结构特征及预测问题进行研究。研究验证了中国市场上预测的时间尺度由日缩小到交易时段尺度的可行性，发现了中国原油期货价格波动率有时段波动的特征，且时段特征的加入显著提高模型的预测性能。此外，研究还发现，预测时间尺度的缩小促进已实现波动序列平稳性的改善，发展了非平稳时序下HAR-RV模型研究问题，波动率的预测结果可为投资者和管理者对中国原油期货市场设计出更为精准的风险管理工具。

关键词:

中国原油期货, 异质自回归已实现波动率模型, 时段特征, 预测, 高频数据