Title: Investigating the contagion effect with using Bayesian Copula-GARCH models
Authors: Justyna Mokrzycka - Cracow University of Economics (Poland) [presenting]
Abstract: The aim is to present a concept of identifying the contagion effect consisting in comparing a posteriori probabilities of two Bayesian dynamic Copula-GARCH models with t Student or normal copula. In case of a higher a posteriori probability of model with modified dynamics and a positive value of an additional parameter, these results are proposed to be interpreted as a confirmation of contagion effect on the financial markets. The contagion effect is understood in a narrow sense, as an increase in linkages between financial markets. Results for simulated and empirical data are presented. Using Copula-GARCH models enables a determination of asymmetric structures of dependencies between financial markets, and moreover, in separation from a dynamics of volatility in individual markets. Student's conditional marginal distributions may have a different number of degrees of freedom, the dependence may be asymmetric, i.e. coefficients of upper and lower tail dependencies may have different values. A formal choice of a proper structure of dependence in Copula-GARCH models is possible by applying Bayesian inference.