Dynamic Connectedness Among Economic Policy Uncertainty,Investor Attention And Stock Volatility

In order to investigate the dynamic connectedness among the four variables,which are economic policy uncertainty(EPU),good and bad volatility,and investor attention,we propose a four-variable TVP-VAR-SV model with mixture innovation(MI-VAR)and select the optimal model by model comparison method.Then we explore the characteristics of the risk spillover effects among the four variables by the DY index proposed by Diebold and Yilmaz(2009,2012)and the Wald-type statistics based on the posterior simulation.Firstly,based on the MI-VAR model proposed by Koop,Leon-Gonzalez,and Strachan(2009),this paper introduces three parameters that can determine whether the VAR coefficients,the variance of the disturbing term,and the contemporaneous relationship of the perturbation terms are time-varying at each moment,respectively,in order to estimate when,where,and how the parameters change.The method of model comparison is then applied to select the optimal model,and the parameters are found to have different evolution characteristics,demonstrating the superiority of using a mixture innovation process to describe the evolution of the parameters.Second,in order to better explore the asymmetry of volatility spillover effects,this paper decomposes realized volatility into positive and negative realized semivariance volatilities(good volatility and bad volatility).Moreover,we use the number of posts containing a specific security name in several stock forums integrated by Uqer.com and the EPU index proposed by Huang & Luk(2020)as a proxy variable for investor attention and EPU.After estimating the parameters using Gibbs’ algorithm,this paper calculates the spillover index using variance decomposition and constructs the Wald test statistic using simulation sample from posterior distribution to investigate the time-varying characteristics of the variables and asymmetric spillover effects,respectively.Accordingly,the main conclusions of this paper are as follows:(1)The results of the model comparison show the probability of time-variation of the VAR coefficients and the variance of the disturbance terms is close to 100%,while the contemporaneous correlations of the disturbance terms are relatively stable.(2)EPU index and investor attention are net recipients,i.e.,the risk spillover effect to other variables is smaller than the risk spillover effect from other variables to them.And both good and bad volatility are net contributors,which are sources of risk contagion in all times.(3)From posterior-based Wald-type tests,we find that there are significant two-way asymmetric spillovers between good,bad volatility and EPU,as well as between good,bad volatility and investor attention.The significant asymmetric spillover effect of good and bad volatility on EPU occurs more frequently than the asymmetric spill-in effect,in which the bad volatility has a bigger effect on EPU.(4)Good and bad volatility have an asymmetric risk spillover effect on investor attention,and there is a stronger spillover effect between good volatility and investor attention,except the time of two stock market crashes,which can be served as evidence of the “ostrich effect” inthe stock market.