Time-varying Risk Aversion And Crude Oil Futures Price Volatility

Crude oil is the most actively traded commodity in the world and has important strategic significance.Crude oil futures are the best tool to manage the risk of crude oil prices,and their price volatility reflect investors’ forward-looking views on the uncertainty of the crude oil market,which not only affects investors’ trading decisions,but also affects the policy-making of government departments.At the same time,the price volatility of crude oil futures is highly time-varying,which also has a significant impact on the global economy.Therefore,how to fit and predict the price volatility of crude oil futures market has become a research hotspot of scholars.Existing studies have shown that the volatility of crude oil futures market is asymmetric and highly persistent,and ignoring these features will reduce the accuracy of forecasting crude oil futures price volatilities.Some studies have shown that the high-order moment of the crude oil futures yield distribution is time-varying,and ignoring the time-varying highorder moment of the yield distribution will bring certain financial risks.In addition,investors are an important part of the crude oil futures market,their aversion to risk will have an important impact on the price volatility of crude oil futures.In view of this,the asymmetry of crude oil futures volatility should be considered when modeling the volatility of crude oil futures.and high persistence characteristics,the time-varying high-order moment characteristics of crude oil futures yield distribution,and the impact of investor risk aversion on crude oil futures price volatilities,in order to improve the fitting and forecasting performance of volatility,it has important practical significance for reducing the risk of crude oil futures market.Considering that volatility is asymmetric and highly persistent,returns are characterized by time-varying higher-order moments,this paper applies the GJRGARCH framework,which is most commonly used in modeling volatility,to capture the high persistence of volatility using the mixing frequency(MIDAS)structure,and the third-order moments and fourth-order moments of asset returns are time-varying in the model,and a new GJR-GARCH-MIDAS-SK model is constructed,which can simultaneously characterize the time-varying higher-order moments of asset returns and capture the high persistence of volatility,and examine the predictive role of both on crude oil futures price volatility.Then,based on the GJR-GARCH-MIDAS-SK model,a measure of investors’ risk aversion(RA)is incorporated into the model to examine the impact and predictive role of investors’ risk aversion on crude oil futures price volatility.This paper selects the daily closing price of Brent crude oil futures and WTI crude oil futures as research samples to empirically compare the newly proposed GJRGARCH-MIDAS-SK and GJR-GARCH-MIDAS-SK-RA in this paper for in-sample data fitting,out-of-sample volatility forecasting performance and comparative benchmark models(GJR-GARCH,GJR-GARCH-SK and GJR-GARCH-MIDAS models).The main research conclusions are as follows:(1)By comparing the AIC values and log-likelihood values of the GJR-GARCH-MIDAS-SK model and the three comparative benchmark models,it is found that the GJR-GARCH-MIDAS-SK model that integrates the characteristics of high volatility persistence and high-order timevarying moments of returns has better full-sample fitting performance than the model that considers one of these characteristics alone.(2)The full-sample estimation results of the GJR-GARCH-MIDAS-SK and GJR-GARCH-MIDAS-SK-RA models indicate that investors’ risk aversion has a significant positive effect on crude oil futures price volatility,and its inclusion in the models can improve the full-sample fit performance of the models on crude oil futures price volatility.(3)Out-of-sample forecasting evaluation results based on loss function value comparison,DM test and MCS test show that the GJR-GARCH-MIDAS-SK model has better out-of-sample forecasting than the comparative benchmark model,and the GJR-GARCH-MIDAS-SK model with the introduction of time-varying risk aversion indicators has the best forecasting performance.(4)The robustness tests of different forecast intervals,different forecast periods and different forecast evaluation indicators demonstrate the robustness of conclusions(1)-(3).