Forecasting correlation among equity mutual funds

Abstract

Investors often buy multiple funds that are actively managed, specializing within narrowly defined market segments. To successfully implement a strategy of diversification investors must obtain accurate estimates of correlation among mutual fund returns. This paper forecasts mutual fund correlations using eight models that are broadly classified into historical, mean and index. Results indicate that estimate of future correlations from the Multi-Style Index, Dynamic and Fama–French 3-Factor models have the lowest prediction errors. Moreover, the relative ranks of Multi-Style Index and Fama–French 3-Factor models have lower dispersion across different forecasting time periods and in sub-samples of funds belonging to similar or different ‘style’ categories.