Luck vs. Skill in Mutual Fund Alpha Estimates
A long-standing research thread has shown that professionally-managed portfolio returns strongly resemble a random walk about the market average. This is interpreted to mean that professional money managers cannot predictably beat the market.
A new study by Eugene Fama and Kenneth French uses a novel statistical approach to add evidence to that record—but with an important caveat, to be mentioned at the end of this article.
Empirical studies of professionally-managed performance
In most well-constructed academic studies that reach a random walk conclusion, two observations are central:
- Professional investors on average do not outperform the market.
- A professional investor who does beat the market in a series of periods cannot predictably continue to outperform—that is, market-beating performance is random and not persistent.
Both observations would also be made in coin-toss experiments: the average percentage of heads would be 50%, and a series of heads for a particular coin would not predict that the coin’s chance of a head in the future will be greater than 50%.
An adjustment is needed, however, in studies of portfolio returns. One must correct for risk. Otherwise, it will be observed that returns for riskier funds have a tendency to be higher than those for less risky funds—though the variation around that trend will also be higher.
This risk-adjustment results in the notorious alpha (α), the elusive Holy Grail of investing. Alpha is the amount by which a fund beats the market after adjusting for risk.
In most studies, average alpha for professionally-managed funds is indiscernibly different from zero (or negative, if net-of-fee returns are used). Little or no statistical evidence is found that if a particular fund produces a series of positive alphas, its likelihood of continuing to do so is enhanced.
The bootstrap simulation methodology of the Fama-French study
The Fama and French study uses a relatively new methodology to produce corroborating results.
Their data set is equity mutual fund performance from January 1984 through September 2006. They compute alphas by regressing monthly returns not only against market returns but against small stock and value stock returns as well (and in one version, “momentum return”). This uses the Fama-French three-factor model, which the authors published in 1992, and presumably adjusts not only for market risk but also any unique risks related to small or value stocks.
This yields an alpha for each fund. The alphas, not unexpectedly, average about zero when gross returns (before fees) are used, negative when net-of-fee returns are used. Thus, observation 1) above is confirmed.