top of page
Writer's pictureRobin Powell

Active share as a predictor of future performance

Updated: Nov 14







By LARRY SWEDROE


Active share is a measure of how much a fund’s holdings deviate from its benchmark index, and funds with the highest active share tend to have the best relative performance (in order to outperform, you must deviate from the benchmark). Thus, while there’s no doubt that active management underperforms in aggregate and the majority of active funds underperform every year (the percentage that underperform increases with the time horizon studied), if an investor were able to identify the few future winners by using active share as a measure, active management could be the winning strategy.


My February 21, 2022, Advisor Perspectives article examined the evidence from empirical research on the ability of active share to predict future performance of actively managed mutual funds. The evidence demonstrated it is difficult to make the case that active share has any predictive value in terms of future risk-adjusted outperformance of actively managed mutual funds. For example, in their November 2021 study Is Active Share Unattractive?, Morningstar found that despite exhibiting greater risk, high active share funds failed to deliver superior net-of-fee results in any category. Despite this, active share has become an increasingly popular metric in terms of both reporting and evaluation. Given the evidence, there isn’t a logical explanation for the phenomenon other than that high active share is a crucial ingredient for outperformance. Unfortunately, it’s not a guaranteed one.



New evidence

Martijn Cremers, Jon Fulkerson and Timothy Riley, authors of the study Active Share and the Predictability of the Performance of Separate Accounts, published in the First Quarter 2022 issue of the Financial Analysts Journal, examined the relationship between active share and performance for both actively managed mutual funds and separately managed institutional accounts. They explained: “Separate accounts are privately run portfolios that are, in many respects, similar to mutual funds; however, rather than being managed for the benefit of a diffuse group of investors, separate accounts are managed for a single entity and can be personalised for that entity’s specific preferences (e.g., such as tax and ESG considerations).” The authors noted that while separate accounts are less well known, they manage more than twice the assets of mutual funds. Thus, their performance, and the ability of active share to predict future performance, should be of interest.


Cremers, Fulkerson and Riley used two methods to identify a separate account’s benchmark. The first was the self-declared benchmark. The second was to use the benchmark that resulted in the lowest active share (their preferred method). They considered 21 potential benchmarks: the S&P 500, 400 and 600; the Russell 1000, 2000, 3000 and Midcap; and the value and growth components of each of those benchmarks. Their main measure of risk-adjusted performance was the alpha calculated using a seven-factor (CPZ7) model: (1) S&P 500 less the risk-free rate, (2) Russell Midcap less the S&P 500, (3) Russell 2000 less the Russell Midcap, (4) S&P 500 Value less the S&P 500 Growth, (5) Russell Midcap Value less the Russell Midcap Growth, (6) Russell 2000 Value less the Russell 2000 Growth and (7) Carhart momentum factor. Their data sample covered 3,706 separate accounts over the period July 2007-December 2019. Following is a summary of their findings:


  • There was inaccuracy in the self-declared benchmarks — about 21 percent of separate accounts had a difference in holdings between their self-declared benchmark and their minimum active share benchmark that was economically significant (benchmark mismatch greater than 60 percent). That result was similar to mismatches found in mutual funds (24 percent).


  • Separate accounts had an average active share of about 80 percent, comparable to the average active share observed for an equivalent sample of mutual funds.


  • The average net alpha for the separate accounts was -0.76 percent per year, and the net benchmark-adjusted returns ranged from -0.57 to -1.07 percent per year depending on the benchmark —t here is net-of-fee underperformance for the typical separate account.


  • An unconditional sort on active share among separate accounts was insufficient to identify future outperformance. (Note that this is different from the findings of the original study on active share by Martijn Cremers and Antti Petajisto, How Active Is Your Fund Manager?: A New Measure That Predicts Performance, published in the September 2009 issue of The Review of Financial Studies. The authors concluded: “Active Share predicts fund performance: funds with the highest Active Share significantly outperform their benchmarks, both before and after expenses, and they exhibit strong performance persistence.” Since subsequent returns did not bear this out, the authors sought to add other criteria, which should increase your skepticism.)


  • A sort on the top 20 percent of past performance of separate accounts showed no statistically significant evidence of the ability to identify future performance — net alpha of 0.53 percent per year (t-stat = 1.32). Results were similar for mutual funds sorting on past performance.


  • Conditioning on the level of active share, past performance can identify separate accounts that continue to outperform. Among separate accounts, within the top 20 percent of active share, a portfolio of those within the top 20 percent of past performance had a statistically significant (at the 5 percent confidence level) net alpha of 1.38 percent per year (t-stat = 2.11) — there is evidence that positive performance persistence suggests that being a top performer among high active share separate accounts is indicative of managerial skill. Double sorting on mutual funds provided much weaker evidence — the net alpha of that portfolio was only 0.69 percent per year (t-stat = 1.06). (Note that for large institutional investors, separately managed accounts typically have lower fees than do mutual funds and fewer cash flows, resulting in lower trading and market impact costs, helping to explain the difference in performance.)

  • The net alpha for the portfolio of separate accounts in the top quintile of past performance among those with low active share was only -0.14 percent per year (t-stat = -0.28).


  • Negative performance persistence occurred regardless of active share — whether active share was high or low, separate accounts that previously underperformed relative to their active share quintile tended to underperform the market in the future.


  • For a portfolio of separate accounts with strong past performance among those with high active share, there is more future outperformance inside the small-cap style—the portfolio containing the top quintile of past performance within the top quintile of active share had a net CPZ7 alpha of 1.31 percent per year (t-stat = 1.65). However, in the large-cap style there was no evidence of persistent outperformance. The results were weaker among mutual funds.


  • For a portfolio of separate accounts with strong past performance among those with high active share, there is more future outperformance for those using a fundamental analysis approach—the top two quintiles of active share had annualized net alphas of 1.25 percent (t-stat = 2.42) and 1.70 percent (t-stat = 2.22). The equivalent portfolios formed among separate accounts with the other investment approaches did not have positive net alphas.


  • Cash acted as a drag on performance, as higher cash holdings were associated with lower returns, suggesting no evidence of market timing ability. (Note that separately managed accounts can hold less cash because they don’t have to be as concerned about cash outflows from investors, helping to explain some of the difference in performance between them and mutual funds.)


Their findings led Cremers, Fulkerson and Riley to conclude that neither active share nor past performance by itself was an effective means of identifying future outperformers. However, at least for separate accounts, there was some evidence that a double sort on high active share and past performance provided information on future performance. What should investors take away from this new evidence?



Investor takeaways

As noted earlier, Martijn Cremers and Antti Petajisto, authors of the original study on active share, concluded: “Active Share predicts fund performance: funds with the highest Active Share significantly outperform their benchmarks, both before and after expenses, and they exhibit strong performance persistence.” However, that result has not held up in replication studies. And even the new study in which Cremers is one of the co-authors did not find evidence of statistically significant persistence in mutual fund performance sorting by active share, or even double sorting by active share and past performance.


While Cremers, Fulkerson and Riley did find statistically significant (at the 5 percent confidence level) evidence of persistence in performance using the double sort in separately managed accounts, they did not find the same result for mutual funds. In other words, the hypothesis failed a test of robustness of their finding. While not dismissing their finding, another concern investors should have is related to the multiple testing problem—if the first hypothesis fails, keep testing until a statistically significant relationship is found—which it likely will be. (For those interested in learning more about statistical testing, I highly recommend the very readable The Art of Statistics by David Spiegelhalter). Campbell Harvey, Yan Liu and Heqing Zhu, authors of the 2016 study … and the Cross-section of Expected Returns suggested addressing the multiple testing problem by increasing the hurdle for statistical significance from the standard of t-stat = 2 (95 percent confidence level) to a t-stat greater than 3 (99 percent confidence level). If the hurdle were raised to 3, Cremers, Fulkerson and Riley’s results would not have been significant, even for separate account managers. And while their results held for small stocks, they did not hold for large stocks—the alpha was a statistically insignificant -0.28 (t-stat = -0.44) — which make up the vast majority (about 90 percent) of the total market capitalisation.


The bottom line is that for the typical investor who invests in mutual funds, all the evidence suggests that active share does not provide valuable information in terms of future performance. The Cremers, Fulkerson and Riley study provided some evidence that the double sorting on active share and past performance have value. However, the results were not robust to mutual funds, and there was the problem of multiple testing. And finally, their findings only applied to separate account managers. The question is: Did they provide sufficient evidence to convince you?


As you consider the answer, remember that active share has not provided valuable information as to the future performance of mutual funds. In addition, active share combined with past performance will lead to portfolio turnover. That increased turnover can lead to tax inefficiency for taxable accounts.




© The Evidence-Based Investor MMXXIV. All rights reserved. Unauthorised use and/ or duplication of this material without express and written permission is strictly prohibited.


bottom of page