This study develops two new performance evaluation measures based on a comparison of fund manager portfolio holdings and finds that the performance measures are effective at predicting future mutual fund returns. The measures are based on the following premise: managers who use similar techniques are likely to make similar investment decisions and likely to deliver similar performance in the future. In other words, if two managers own many of the same stocks, it is likely they are using similar techniques and one would want to pay a great deal of attention to the performance of the first manager when evaluating the second. The goal of the performance measures is to judge each individual fund manager’s skill by the extent to which his investment decisions resemble those of other managers who have distinguished past performance records.

            Both holdings based performance measures are constructed in a similar manner. The first measure focuses on the extent to which a manager’s portfolio holdings overlap with the holdings of other managers while the second measure is based on the overlap of changes to portfolio holdings. To calculate the measures, the authors first estimate past manager skill using various risk adjustment procedures including the Carhart model. Then, each stock held by a specific mutual fund is given a quality measure which combines the average skill of all managers who hold that stock in their portfolios and the amount of the stock each manager holds. As an example, if all skilled managers currently hold large percentages of Microsoft stock within their portfolios, then Microsoft would be assigned a high quality measure. Finally, manager performance is calculated as the average quality of all stocks in the manager’s portfolio where each stock contributes to the performance measure according to its weight in the portfolio. The authors calculate this measure on a quarterly basis for all funds whose portfolio holdings data is available. The study runs from the end of the first quarter of 1980 through the second quarter of 2002, and the number of funds analyzed grows from 235 in 1980 to over 1500 funds in 2002. 

            For fund investors, the most important result of the study is that these new measures provide information about future fund returns that is not found in traditional mutual fund measures such as a fund’s alpha. Each quarter, the authors sort funds into deciles based on their holdings based measures. They then track the returns of each decile over the subsequent quarter. They find that the difference between the risk adjusted returns of the top and bottom deciles ranges from 5.9% to 7.4% depending on which measure is used. Combining their measure with a fund’s alpha can lead to even larger risk adjusted returns. A portfolio strategy based on buying funds with the highest alpha and highest holdings based measure while selling funds with the lowest alpha and lowest holdings based measure provides risk adjusted returns as high as 8.5% depending on the model used to measure alpha. The authors also find that the results are similar even when this portfolio strategy is implemented with a one quarter delay after the portfolio disclosure dates. This last result is important given that mutual funds can publicly report their holdings with a lag of up to two months relative to the official disclosure date.

            Overall, the study provides strong evidence that the holdings based measures provide useful information to investors about the future performance of all categories of equity mutual funds.