Research On The Volatility Of China's Stock Market Based On Bayesian CV-DC Model

In China's financial market,stocks are a kind of financial products with strong participation,high liquidity and high risk.The theory of the average return on the stock market returns that the long-term forecast variance is smaller than the shortterm forecast variance.This theoretical point of view suggests that long-term investors should hold more stocks than bonds than short-term investors.However,the theory ignores the process of continuously learning and changing the predictive parameters of returns after new data is added,that is,the variability of the predictive parameters.Therefore,this paper attempts to establish a time-varying parameter model to predict stock returns and forecast variance from the perspective of investors,and then studies the relationship between long-term forecast variance and short-term forecast variance when the parameters are time-varying.This is of great significance to the allocation of long-term investors' asset portfolios and the selection of asset portfolios in the long run.At present,the domestic GARCH and stochastic volatility categories are the mainstream of volatility modeling,but the parameters of the yield model are often constant within the sample interval.The term structure characteristics of forecasted returns,variances and forecasted volatility.This article theoretically provides ideas for the study of forecasted returns and volatility and in reality provides guidance for investors in portfolio management.In this paper,the 155-month excess returns and dividend data of the CSI 300 Index are selected as the explanatory variables and explanatory variables(predictors)of the yield prediction model.The MCMC algorithm is used to estimate the CV(Constant Variance)model and time-varying Parameter CV-DC(Constant Variance-Drifting Coe ffi cient)model.In order to further illustrate the advantages of MCMC algorithm in estimating parameters,this paper also compares and studies two other non-time-varying parameter models of rolling regression estimation.Based on the estimated parameters,this paper performs out-of-sample forecasted returns and variances on the data to obtain term structures with different forecast variances under the CV and CV-DC models.Finally,this paper selects 155 monthly data from the same data interval of the Shanghai Stock Exchange 50 for robustness test,which verifies the rationality and correctness of the relevant conclusions of this paper.The main conclusions obtained in this paper are:(1)According to the MSE and MAE indicators,it is found that when the MCMC algorithm is used to estimate the parameters,the effect of out-of-sample forecasting returns after adding the timevarying parameter model is the best and the most accurate.(2)The parameters estimated by the MCMC algorithm are used to calculate the forward variance of the forward K period using different models.It is found that the forecast variance before adding the time-varying parameter model is lower than the short-term variance in the long run,which is in line with the characteristics of the traditional mean recovery.However,after adding the time-varying parameter model,the forecast variance in the long-term investment level is significantly larger than the short-term variance.(3)Through the robustness test,similar forecast term structure is obtained with different data,which shows that the conclusion of this paper has certain robustness.