Research On Volatility Modeling Of CSI 300 Index Introducing Economic Policy Uncertainty Index

Volatility is an important indicator of financial asset risk measurement,which plays an important role in many financial activities,such as derivatives pricing,portfolio decision-making,calculation of VaR risk management indicators,etc.Therefore,the prediction and measurement of volatility has always been the focus of scholars’ research.On the other hand,with the increasingly complex and changeable macroeconomic environment and the increasing dependence of market economy on the guidance of government economic policies,the market is increasingly unable to regulate the economy independently under the frequent intervention of the "hand of the government".As a result,the uncertainties brought about by changes in economic policies have increased,making macroeconomic fluctuations more intense,which in turn affects the sharp rise and fall of stock prices.Specifically speaking,the domestic stock market is also known as the "policy market",which is greatly influenced by policy factors.This paper attempts to introduce the economic policy uncertainty represented by the economic policy uncertainty index into the Volatility Prediction model,hoping to improve the prediction accuracy of the volatility model,which will have a very important practical significance for practitioners.On the basis of previous scholars’ research,with the help of China Economic Policy Uncertainty Index(CEPU index)compiled by Paul Luk and Huang Yun(2019),this paper attempts to introduce daily CEPU index into HAR family model,and daily and monthly CEPU index into GARCH-MIDAS model respectively,in order to improve the prediction of volatility model.Measuring accuracy.Specifically,using the 5-minute high-frequency trading data of the Shanghai-Shenzhen 300 index from January 1,2014 to December 31,2018,the sample period is divided into estimation period and prediction period according to the ratio of 4:1,and the rolling time window is used to predict the next step.When estimating in the sample,we find that 1)the new HAR family model with CEPU index has improved the goodness of fit compared with the original model,among which the ability to fit short-term daily fluctuations is the least,the ability to fit medium-term weekly fluctuations is the middle,and the ability to fit long-term monthly fluctuations is the greatest.This reflects the effect of macroeconomic policy uncertainty on the goodness of fit.The stock market has a certain lag,and the CEPU index mainly affects the long-term volatility of the stock market.2)GARCH-MIDAS-D is superior to GARCH-MIDAS-M and GARCH-MIDAS models in the estimation period of different delay periods.In addition,the optimal lag time of GARCH-MIDAS-D model is 630 days(21months).In the aspect of out-of-sample prediction,according to the results of MCS test,1)GARCH-MIDAS model with daily CEPU index is generally better than HAR model.2)Only GARCH-MIDAS-D model with 21 months delay can survive the MCS test.This shows that GARCH-MIDAS-D model with 21 months lag can predict the volatility of China’s stock market more accurately,and CEPU index plays a positive role in improving the prediction accuracy of stock market volatility.The disadvantage of this paper is that the sample period is too short to carry out a longer stability test.Next,we plan to estimate and forecast samples for a longer period to judge the stability and accuracy of GARCH-MIDAS-D model with daily CEPU index in different sample periods.