presence of different regimes in volatility (Forbes and Rigobon (1999)). This paper makes a step
forward and suggests to analyze time varying conditional correlation between international stock
markets by utilizing the recent methodology by Engle (2002), a multivariate GARCH dynamic
conditional correlation analysis (DCC-GARCH).
A DCC-GARCH class of models encompasses the parsimony of univariate GARCH
models of individual assets volatility with a GARCH-like time varying correlations. The
estimation of the DCC-GARCH model is a two-step procedure. In the first step, univariate
GARCH models are estimated for each time series, in the second step transformed residuals from
the first stage are used to obtain conditional correlation estimator. The model assumes that returns
from the k series are multivariate normally distributed with zero mean value and covariance
matrix Ht:
r,∖F,-ι~ N (0, Ht ) (5)
Ht ≡DtRtDt, (6)
where Dt is a kxk matrix of time varying standard deviations from univariate GARCH
models with h^ on the ith diagonal, following a univariate GARCH model. The proposed
dynamic correlation structure is:
Rt = (Qt*)-1Qt(Qt*)-1, (7)
where Qt* is a diagonal matrix composed of the of the square root of the diagonal elements
of the Qt and Qt follows a GARCH type of process:
MNM N
Qt = (1 — ∑αm -∑βn)Q + ^αm(ε'tεt) + ∑βnQt-n, (8)
m =1 n=1 m=1 n=1
where Q is an unconditional covariance matrix of the standardized residuals from the first-
stage estimation.
16