Forecasting Equity Market Correlation: A Comparison of Alternative Measures

This paper provides empirical evidence to compare the abilities of the forecasting correlations among market returns. The conditional models, which are based on the estimation of the time-varying correlation equations, provide strong forecast accuracy. The unconditional measures of correlation, however, perform poorly in all equity markets. The popular Pearson’s correlation coefficient is inferior to other measures. The robust non-parameter measures of correlation which is insensitive to the outliers are better than Pearson’s correlation. We find that, the measure which calculates by the method of the conditional model with the most parameters provides the best predictive accuracy, while the popular linear correlation coefficient makes poor forecasts in the application of international equity markets.