Soft Clustering for Funds Management Style Analysis: Out-of-Sample Predictability

Abstract

Listen

Translate article

Mutual funds are generally grouped on the basis of their investment objectives. Investors use these groupings, or 'styles', to make investment decisions and compare the performance of fund managers against appropriate indexes. Various authors have criticized the traditional classification approaches because they are made on the basis of funds' stated objectives, not the actual fund styles and are therefore often misclassified. [6] introduced the General Style Classification (GSC) approach as an alternative to the traditional technique which uses statistical (rather than stated) objectives. Results of their study suggest that investment style boundaries are continuous rather than hard. In this work, we classify mutual funds using a soft clustering technique (Fuzzy-C-Means) and compare it with a hard clustering technique (GSC). This comparison demonstrates soft clustering can predict mutual fund performance better out-of-sample.