Energy Market Risk Prediction Based On Mixed Frequency Model:High Frequency Information Or Low Frequency Information?

As the world’s largest oil importer and the second largest oil consumer country,China is highly dependent on international crude oil imports for its oil consumption.Therefore,the change of international crude oil price will affect the steady development of our economy,and how to accurately predict the risk of crude oil futures has become a topic of extensive research.Crude oil prices are influenced by two factors:fundamental information and intraday high-frequency information,but it is difficult to capture such information when using the traditional model to predict the risk of crude oil futures market.Therefore,taking Brent futures and WTI futures as our research object,this paper studies the following questions based on the framework of two types of mixed sampling models:Compared with the daily closing price,whether low-frequency fundamental information or intraday high-frequency price information can help predict the risk of the crude oil futures market,and which variable is the most helpful to improve the accuracy of risk prediction.Firstly,different risk prediction models are constructed.Based on the framework of GARCH-MIDAS model,this paper considers the fundamental factors including crude oil supply,demand and inventory,and constructs one type of risk prediction model.Besides,based on the framework of Realized GARCH model,this paper considers different realized volatility indicators to build another type of risk prediction model.Secondly,according to above models,the ES predicted values of two crude oil futures markets from May 3,2021 to May 31,2022 are calculated,and the prediction accuracy of above models is tested by back testing.Finally,according to the empirical results,this paper draws the following conclusions.First,the negative impact information in intra-day high frequency price information is the most helpful to improve the prediction accuracy of ES in crude oil futures market,and the RS~—model has the highest prediction accuracy among all models.Second,not all mixed information can help predict the ES of the crude oil futures market.Among the two types of mixed sampling models,there are some specific models whose prediction performance is weaker than that of the traditional GARCH class model.Third,in intraday high frequency price information,different sampling frequencies will affect the accuracy of prediction.In conclusion,by comparing the performance of different mixed sampling models in predicting ES indicator,this paper enriches the research of GARCH-MIDAS model and the Realized GARCH model in the field of risk prediction,and provides a reference for the information to be used in the crude oil futures market for risk prediction.