Stephen J. Brown, NYU Leonard Stern School of Business
William N. Goetzmann, Yale School of Management
The style of fund management is taken the be a key descriptor characterizing differences across managers in the mutual fund industry. We find that the standard stylistic classifications such as "Growth," and "Growth and Income," do a poor job at capturing future cross-sectional return differences. We propose a new empirical approach to the problem of determining mutual fund styles. This approach is simple to apply, yet it captures non-linear patterns of returns that result from virtually all active portfolio management styles. We find that the largest equity fund category, "Growth" typically breaks down into several distinct styles that differ in composition and strategy. Besides categories corresponding to commonly accepted styles such as "Growth," "Growth and Income," "Income," "International," and "Small Cap," we also find evidence of a group of "Timers," as well as a group of "Value" managers.
The classification method we propose groups funds in the space of returns. It is related to clustering methods and to switching regression techniques used in finance and other disciplines. Without using any information about the composition of the fund's portfolio, it identifies fund groupings that are useful predictors of cross-sectional future performance, as well as past behavior. Not only are they superior to common classifications such as "Growth" or "Income", but they also perform better than classification based upon risk measures and analogue portfolios. An out-of-sample test of cross-sectional explanatory value indicates that the return-based stylistic classification works nearly as well as do factor loadings on pre-determined and latent variables.
Download Working Paper?