Predicting stock performance with mutual fund portfolio disclosure

Research by Russ Wermers

Knowing how the best mutual fund managers are choosing their portfolios can be valuable information for investors, if stocks picked by skilled fund managers really outperform those picked by unskilled managers. But it has been difficult to test this hypothesis, since it is not clear how to use fund holdings to pick stocks.

Russ Wermers, associate professor of finance, with co-authors Tong Yao, University of Arizona, and Jane Zhao, PanAgora Asset Management, developed a statistical model that predicts the future performance of individual stocks based on how heavily they are held or purchased by both successful and unsuccessful fund managers.

Rather than just looking at the results of winning funds, Wermers and his co-authors examined good, average and bad funds to see what highly skilled—and thus very successful—fund managers were buying in common, and what underperforming fund managers were not buying. The model uses a weighted average alpha to determine the outlook for a stock at the beginning of a given year, consisting of the portfolio weight on a stock multiplied by a manager’s past alpha, summed across all managers who held that stock at the beginning of that month.

“You can’t just consider how many winning funds bought the stock, you have to weight how much of it they purchased, and you also have to put some weight on the skill of the manager,” says Wermers. “We considered the performance of every fund manager and use it as the main factor in the weighting of the outlook for a stock.”

The result is a much more accurate prediction of stock returns than any previous model, and significantly better returns result even when adjusted for risk, beating previous methods by as much as 6 percent to 8 percent per year.

“Even controlling for all the risk factors that are already known in finance, this model seems to provide an independent source for stock-picks,” says Wermers. “People have been poring over stock returns for decades now, trying to find a new angle, a new way to make money. We didn’t expect to find something this big. Hedge funds in particular have shown interest in this research because they need an independent source of stock returns, above what is already known by the masses.”

The model uses three stock alpha estimators to extract information about the future alphas of stocks from portfolio data for the cross-section of mutual funds. The net returns and portfolio holdings of actively managed U.S. domestic equity funds from 1980 to 2002 are used to show that that these three stock alpha estimators consistently predict cross-sectional stock returns over the following year. Wermers and his co-authors also developed conditional stock alphas by taking into account stock characteristics and fund characteristics, and found that these characteristics helped to further improve the model. They found that smaller and older funds, and funds with lower expense ratios, higher turnover and higher industry concentration of portfolio holdings were more likely to exhibit persistent skills.

Because mutual funds only disclose portfolio holdings information on a quarterly basis, and because the funds have a 60-day grace period to file their holdings with the SEC, investors interested in using this model to predict stock returns are limited by the time lag in receiving information about what mutual fund managers are holding. Mutual fund disclosures are also staggered, so it isn’t easy to obtain the information—for instance, some funds report their holdings in December, while others report in October. Wermers used a multitude of datasets to create the holdings data and returns data used in the study.

Wermers is interested in using this technology to develop stock signals for European and Asian stocks in the international stock arena. “We think there is a lot of potential in non-U.S. markets because those fund managers are more likely to have inside information, which is perfectly legal in some of those countries,” says Wermers. “We’re trying to find a source of reliable fund holdings in order to examine this data.”

“The Investment Value of Mutual Fund Disclosure” won the Best Paper award at the annual conference of Inquire-Europe, an international consortium of investment companies dedicated to bringing academic research into the public arena. For more information about this research, contact rwermers@rhsmith.umd.edu.