The Investment Value of Mutual Fund Portfolio Disclosure

This paper uses disclosed mutual fund portfolio holdings to develop stock selection models. Our models aggregate portfolio holdings across mutual funds, weighted by their past performance, to predict future stock returns – an overweighting by successful managers, or an underweighting by unsuccessful managers is considered to be a signal that a stock is currently underpriced. We find that investment strategies based on the models generate returns exceeding seven percent during the following year, adjusted for the size, book-to-market, and momentum characteristics of stocks. This evidence suggests that some managers have superior stock-selection skills, and that these skills strongly persist. Further, returns generated from the models cannot be explained by price pressure due to fund herding or return-chasing behavior of fund flows; and, the models positively predict firms’ future operating performance. Finally, model-forecasted alphas have a low correlation with a large number of quantitative investment signals based on prior-documented market anomalies. Thus, our stock selection signals are unique, and indicate that some fund managers possess private skills to produce fundamental information about stock value.