Aurora Borealis
2 April 2021 Financial Analysts Journal

Identifying Hedge Fund Skill by Using Peer Cohorts (Summary)

  1. Phil Davis

This is a summary of “Identifying Hedge Fund Skill by Using Peer Cohorts” by David Forsberg, David R. Gallagher, and Geoffrey J. Warren, published in the Second Quarter 2021 issue of the Financial Analysts Journal.

Listen to an audio version of this summary.

In analyzing hedge funds, “cohort” models formed by cluster analysis are more effective assessors and predictors of skill and performance than the established factor models.

What’s the Investment Issue?

Factor models are widely used by investment professionals to analyze fund performance. But they are less useful for finding the best-performing hedge funds given that hedge funds use such a wide range of strategies and styles and provide limited information about them.

A “cohort” model separates hedge funds into cohorts based on the clustering of their returns, which gets around the problems of lack of information and the complexity of strategies. The correlation of returns should identify peer groups of funds that have similar strategies and are likely to be exposed to the same factors.

The authors show how to identify groups of hedge funds that use similar strategies and how to assess the performance of each fund within its cohort, or peer group.

How Do the Authors Tackle the Issue?

The cohorts are created using cluster analysis of returns reported by 4,469 hedge funds in the Hedge Fund Research and eVestment databases.

The performance of each individual hedge fund within a cohort is then compared with the cohort as a whole, which distinguishes returns related to manager-specific skill from returns generated by the strategy followed by the cohort. By implication, the cohort model also separates manager skill from the factor exposures of the cohort.

The authors use performance analysis to compare the alpha in each cohort with estimates of alpha implied by the seven-factor model of William Fung and David Hsieh (“Hedge Fund Benchmarks: A Risk-Based Approach,” Financial Analysts Journal, 2004), which is widely used in analysis of hedge funds. This process allows the authors to see how well the cohort model can predict out-of-sample returns relative to the seven-factor model.

The authors also form portfolios of hedge funds with the greatest cohort alpha from the 15 largest cohorts and evaluate the performance of these portfolios. This process is aimed at designing a high-performing fund of hedge funds.

What Are the Findings?

The cohort model does a considerably better job than the seven-factor model in identifying and categorizing the underlying drivers of hedge fund returns. The cohort model explains a higher percentage of hedge fund returns (based on R2) than the seven-factor model for 99% of the periods of the study and for more than 90% of the funds.

Under the cohort model, hedge funds that outperform their cohort over the previous two years continue to outperform for the following three years. In contrast, alpha using the seven-factor model is both less sizable and less persistent, continuing for less than one year.

By forming portfolios of hedge funds with the greatest alpha in each cohort, there is a significant alpha gain compared with performance both from the seven-factor model and from lower-performing funds from the same cohorts. This increase in alpha ranges from 10 bps to 24 bps a month, or about 1.2%–2.9% a year.

The authors further find that by allocating to hedge funds not initially assigned to a cohort but then assigned to their closest cohort, the cohort model is better than the seven-factor model at identifying skilled managers.

What Are the Implications for Investors and Investment Managers?

A cohort-based approach is better at finding skilled managers and explaining performance than traditional approaches. The outperformance is observed not just in sample but persists strongly out of sample too.

The process of forming peer cohorts facilitates insights into the strategy groupings within the hedge fund universe — in particular, which fund belongs to which group. These insights can help optimize strategy exposures within fund-of-funds portfolios, helping managers better diversify their holdings and pick the best funds.

The cohort approach also has the advantage of being relatively straightforward to apply, with cohorts formed using simple return correlations and cohort alpha being estimated by the usual factor regression. 

We’re using cookies, but you can turn them off in Privacy Settings.  Otherwise, you are agreeing to our use of cookies.  Accepting cookies does not mean that we are collecting personal data. Learn more in our Privacy Policy.