Prediction Of Price Volatility In Crude Oil Futures Markets

Energy is an indispensable material foundation of modern society and occupies a important strategic position in the economic development of countries in the world today.Among them,crude oil is the most important and most representative nonrenewable energy source.With the gradual financialization of crude oil,investors and national policy makers are paying more and more attention to fluctuations in the price of crude oil derivatives.Because the crude oil futures market and the spot market respond to information synchronously,it is also of great practical significance to study the price fluctuations of the crude oil futures market.Crude oil futures prices fluctuate greatly and have time-varying feature.If futures prices fluctuate significantly,this will not only affect the investment decisions of financial traders and the production plans of manufacturers,but even affect some government policy formulation and the global economic development.At the same time,the volatility of crude oil futures markets has become increasingly linked to asset allocation,asset pricing,risk measurement and management.Therefore,analyzing the volatility characteristics of the international crude oil futures market has very important guiding significance for market risk management and control.In the empirical research in this paper,in order to improve the prediction accuracy of the price volatility of the crude oil futures market,this paper will focus on time-varying volatility and combination strategies.First,in this paper,we will consider how to improve the accuracy of volatility prediction of the crude oil futures market in a time-varying environment.Based on the five-minute high-frequency data of the crude oil futures market price WTI and the HAR family model,this paper uses a combination strategy to construct five constant combination models and a time-varying parameter combination model.Through empirical research inside and outside the sample,the results show that when forecasting the future crude oil futures price volatility,the combination forecast methods can help improve the prediction accuracy of the model.The combination model is more adaptable,which can not only reduce the instability of a single model,but also reduce the influence of external factors on a single model at different time nodes.Among the constant combination models,the shrinking combination model performs best.The time-varying parameter combination method is used to dynamically combine various high-frequency volatility models at different points in time,which shows the feasibility and superiority of the dynamic model averaging method in predicting future crude oil futures price volatility.Second,we will consider the various risks in the financial market,the structural change may reduce the long memory of price fluctuations in the crude oil futures market,which will have a certain impact on the volatility forecast.However,the structural changes often also contain a lot of current market information,which helps us to identify potential structural inflection points and then find the more potential influencing factors of crude oil futures price fluctuations in a specific period.In this paper,we use the sub-sample regressions to find that excluding the impact of structural changes caused by external factors,we mainly rely on the continuous sample path deteriorating part in the short-term forecast,and when the price changes are large,we should consider the effects of leverage effects.This verifies the importance of considering structural changes and demonstrates the necessity of adopting dynamic prediction models.Finally,from the perspective of the volatility state transition,Finally,I combined the HAR model with the Markov transfer mechanism.The dynamic prediction model can capture the characteristics of multiple volatility states.Finally,it is concluded that the MS-LHAR-RV-CJ model is the best model with the highest prediction accuracy and the shrinking combination prediction model was the optimal combination prediction model.