The correlation of monthly excess returns for 8 major countries has been studied over the period 1960 to 1990. The unconditional international covariance and correlation matrices were found to be unstable over time. This instability cannot be explained by the time variation in conditional expected returns. A traditional multivariate GARCH model helps to capture some of the evolution in the conditional covariance structure. Information variables such as the dividend yield contain information about future volatility and correlation that is not contained in past returns alone. The proposed GARCH modelling allows the capture of trends in variances but falls short in adequately modelling the observed increase in international correlation. It is found that the correlation increases in periods when the conditional volatility of markets is large.
It has been found that the international correlation of stock markets has increased over the past 30 years and that the correlation increases in turbulent periods. 3 influences on correlation have been demonstrated, exogenous information variables, time trend and market volatility. Unfortunately, these 3 types of variables are correlated.