Study On Dynamic Correlation And Spillover Effects Between Financial Markets Based On SCC GARCH Model

Since the1980s, along with the development of modern communication technology,the liberalization of capital movement and the integration of global economy, theco-movement between different financial markets has become more and more remarkable,especially during the financial crisis erupted in recent years. To identify and quantify thesimultaneity between different financial markets is of theoretical value and realisticmeaning.This paper firstly gives an overview of mutual spillover effects, dynamic conditionalcorrelation and summarizes the common findings among related empirical research. Somewidely used multivariate GARCH models are introduced in the following section. Somerecent literature is focused on identifying simultaneity in multiple equation system throughautoregressive conditional heteroscedasticity applying the multivariate GARCH modelssuch as VECH models, BEKK models, CCC models and DCC models. DCC-MVGARCH,proposed by Engle （2002） is most commonly used owing to its advantage in parameterestimating, calculating and explaining. However, as same as many other multivariateGARCH models, DCC model simply assumes the structural innovations are uncorrelatedand requires the two-step method to estimate, in which case all the co-movement betweenthe endogenous variables needs to be exclusively explained by volatility spillover effectsrather than the dynamic correlation between different markets. Different from thislimitation, this paper allows for instantaneous covariances. Under the assumption ofstructural constant/dynamic conditional correlation （SCCC/SDCC）, the dynamics ofinstantaneous covariances is still identifiable. The methodology is applied to the dailyreturn of S&P500, HSI and Shanghai Composite indexes between January,1987and July,2011. The innovation of this paper is that we introduce conditional correlations to thestructural innovations, turning them to endogenous variables rather than imposing prioriparameters restriction. The empirical finding in the paper also indicates remarkableexistence of conditional correlations, which verifies our model specification.