Research Looks At False Discoveries in Mutual Fund Performance
New research from Maryland Smith finance professor Russell Wermers rebuts a critique of another paper he and two co-authors published in 2010 that introduced a new way to evaluate mutual fund performance.
Wermers, chair of Smith’s finance department and director of its Center for Financial Policy, worked with Laurent Barras of McGill University and Oliver Scaillet of the University of Geneva on the original paper and this latest research, published in the Journal of Finance, a leading scholarly journal.
In their original work, the researchers apply a new econometric approach – the False Discovery Rate – to the field of mutual fund performance. They use the approach to estimate the proportions of zero and nonzero alpha funds in the population and form portfolios of funds that differ in their ability to generate future positive alphas.
A 2019 paper by other researchers conducts a critical evaluation of the 2010 research, challenging Wermers’ and his co-authors on their simulation analysis based on their choice of parameter values.
The False Discovery Rate has increasingly been used to evaluate mutual fund performance, as well as other areas of finance, because it’s simple and fast, write Wermers, Barras and Scaillet: “Simply put, it amounts to estimating a simple average based on t-statistics. This contrasts with alternative approaches that impose much more structure to improve estimation performance.”
In this paper, Wermers and his co-authors replicate their critics’ results, then explore an issue the critics raise about bias by using different parameter values and updating the sample period.
With these new tweaks, the researchers show the performance of the False Discovery Rate (FDR) improves dramatically, and that the critique misses some important real-world parameters. These results, along with those of the critical research, indicate that using the FDR in finance should be done with careful evaluation of the underlying data-generating process, especially with small samples.
“The results represent good news for the academic community, which typically favors statistical methods that are simple and fast—two major advantages of the FDR approach,” write the researchers.
Read the full research article, “Reassessing False Discoveries in Mutual Fund Performance: Skill, Luck, or Lack of Power? A Reply,” in the Journal of Finance.