Multivariate market association and its extremes

This paper analyzes the global market association with the cross-sectional dispersion measure proposed by [Solnik, B., Roulet, J., 2000. Dispersion as cross-sectional correlation. Financial Analysts Journal, 56 (1), 54–61]. It is demonstrated with a simple asset allocation problem that the dispersion measure increases investors’ welfare under certain conditions. From a policy makers’ perspective, the measure can be used to detect herding behavior and assess the evolution and characteristics of global market association. An empirical analysis of eleven developed stock market indices shows that the global market association has increased in recent years and that there are asymmetric effects of jointly positive and negative shocks. In addition, the market association is not always increasing in crisis periods. A comparison of the dispersion measure with estimates of a multivariate GARCH model indicates that correlation breakdowns exist but cannot be identified with the correlation coefficient. Quantile regression estimates further reveal that the volatility of market association decreased in recent years.