According to herding theories, individuals follow the crowd to appear talented and to learn from others. In the context of investment management, differences in ability or information quality can drive differences in manager herding tendencies. Specifically, less-skilled managers may tend to follow the trades of the institutional crowd to enhance the market’s perception of their ability, whereas high-skilled managers may choose to go against the crowd. The authors look at the link between mutual fund herding behavior and performance.

Differences in herding tendency across mutual funds are shown to have strong predictive power for the cross-sectional pattern of returns, with top-decile “herding” funds underperforming bottom-decile “anti-herding” funds by an average of 2.28% a year, both before and after expenses. On a risk-adjusted basis, this annual average performance gap ranges from 1.68% to 2.52%, depending on the risk loadings used to measure alpha. This performance difference is also shown to persist over periods of up to two years.

The study finds that fund managers with anti-herding tendencies generally are those that demonstrate other types of investment skill (e.g., stock selection), indicating that manager skill is the driver of a fund’s anti-herding tendency. Overall, the authors show that herding tendency is a useful predictor of fund performance, even after controlling for other predictive measures, such as active share and past performance.

The authors’ sample consists of all actively managed US equity funds from 1990 to 2009 with net assets of at least $5 million, using monthly fund return data from the CRSP mutual fund database and fund stock holdings data from the Thomson Reuters Mutual Fund Holdings database. The authors use 2,255 distinct mutual funds in their sample.

The authors devise a measure of fund-level herding by correlating a fund’s trades in a given quarter with the trading decisions of all institutional investors in the prior quarter. In other words, herding funds have a greater tendency to buy the same stocks as the crowd did in the recent past. This measures captures a fund manager’s tendency to imitate the trading decisions of the crowd. When measuring herding tendency, the authors control for stock characteristics that represent common mutual fund investing styles (e.g., size, value, and momentum). The authors find substantial heterogeneity in herding tendency across funds.

With the fund-level measure of herding tendency, the authors show that differences in herding tendency across funds help predict subsequent cross-sectional fund performance.

The authors then conduct further tests to see whether herding tendency can be explained by manager skill. They find the following:

• Anti-herding funds make better returns on stocks they hold that are not included in the set of stocks traded by the institutional crowd—that is, they demonstrate better stock selection skill than herding funds.
• Anti-herding funds outperform herding funds by a larger margin when there is greater potential for managers to demonstrate skill, such as periods of high idiosyncratic risk—that is, skillful managers take advantage of opportunities.
• The performance gap between herding and anti-herding funds persists over periods of up to two years after fund herding measurement—that is, the performance advantage of skillful managers does not result from chance.
• Trades by anti-herding funds anticipate the trades of the institutional crowd, which is not the case for herding funds—that is, skillful managers provide useful signals to the crowd.

The authors do a good job of explaining how individual fund performance can be affected by the individual or collective behavior of fund managers and have provided an additional facet to the components of portfolio performance attribution. A future study could look at how this fund herding behavior might affect considerations among the investing public on the active versus passive debate. It would also be interesting to see what findings could come from studying herding behaviors among algorithmic programs that follow the same market triggers.