- Research shows that highly skilled active managers with high active share, low R2 and a patient investment approach have systematically outperformed by meaningful amounts over time.
- These managers, which academics have termed highly selective, can be identified ex-ante.
- Studies showing active underperformance ignore the degree to which many active portfolios have resembled their benchmarks.
- Highly selective active managers’ performance is strongest during periods of increased volatility in the market.
Active share has been gaining in recognition as a measure of a portfolio’s level of active management. A recent paper combining both academic research and ClearBridge internal analysis on active share found that managers with high active share and long-term conviction (longer holding periods) systematically outperform their benchmarks net of fees.1 These findings are consistent with studies investigating the importance of a portfolio’s R2 (R-squared), which is a statistical metric of active management that measures how much of the variability of the portfolio’s returns can be explained by variability in the benchmark model. We believe that an understanding of both active share and R2 helps to contextualize recent capital flows and performance trends and highlights the scarcity and value of highly selective active management.2
Measurement of active share as a metric of active management is a fairly recent activity. In their seminal 2009 study “How Active Is Your Fund Manager? A New Measure That Predicts Performance,” Martijn Cremers and Antti Petajisto first defined active share as an investment tool that measures the percentage of a portfolio’s stock holdings that differ from its benchmark.3 And a recent paper by Cremers and Ankur Pareek, building on previous research, shows that only strategies with holdings that differ significantly (top quintile of active share) from their benchmarks tend to outperform net of fees. (Results of this research are not susceptible to benchmark manipulation — managers choosing improper benchmarks that may be easier to beat — suggesting that such benchmark “gaming” does not account for the outperformance.4) Of these strategies, those with longer holding periods have shown stronger performance (longer holding periods defined as two or more years, as short-term mispricing tends to be competed away by institutions, while the market’s ability to arbitrage away long-term mispricing is more limited).5 In fact, Cremers and Pareek show that strategies with top-quintile active share and long-duration portfolios have generated five-factor net alpha of 2.05% per year on an equal-weighted basis and five-factor alpha over the benchmark net return of as much as 3.57% per annum in the 1990–2013 period.6 Conversely, funds with low active share have generated negative alpha.
Research also shows that highly selective active managers outperform in both large cap and small cap. A recently published paper by Cremers demonstrates that managers in the top quartile of both active share and fund holding duration generated annualized seven-factor net alpha7 of 1.37% in the large cap category and 1.94% in the small cap category in the 1990–2015 period (Exhibit 1).8 It should be noted that the small cap managers tend to have better performance than the large cap managers in general, as well as higher average active share (of roughly 89%) versus large cap managers (of about 72%).9
Exhibit 1: Large Cap and Small Cap Performance by Active Share and Holding Duration (1990–2015)
Source: Cremers 2017. Performance of large cap and small cap mutual funds shown as seven-factor alphas by active share and fund holding duration (1990–2015).
R2 Confirms Highly Selective Active Management Outperforms
The conclusions generated by this active share analysis reflect the research done on the R2 of mutual fund returns. R2, a more traditional statistical metric of active management, results from a regression analysis comparing the returns of a portfolio to those of a benchmark model; it measures how much of the variability of the portfolio’s returns can be explained by variability in the benchmark model. A high R2 indicates that the benchmark explains most of the portfolio’s returns; in other words, the portfolio’s behavior is not very differentiated from the benchmark. The portfolios of highly selective managers, on the other hand, have low R2, which has been shown to be a determining factor in relative outperformance; in fact, research has shown that portfolios with low R2 historically have generated alpha (calculated using the Fama and French 1993 and Carhart 1997 model (FFC)), net of fees, of 60 basis points annually (Exhibit 2).10 On top of this, when low R2 is combined with skill (defined as top-quintile alpha generation in prior periods), alpha generation rises to 3.8% annually (Exhibit 3). R2, then, much like active share, shows that outperforming managers can be identified ex-ante.
Measures of active management like active share and R2 are important because analysis showing that active managers consistently underperform often ignores the degree to which many active managers’ portfolios generally have resembled their benchmarks. To that point, the median mutual fund’s R2 from 1990 to 2010 was 93%, meaning that, for most portfolios, over 90% of return variability can be replicated by major stock indexes.11 In Cremer’s research, by definition, only 25% of active managers made the cutoff for top-quartile active share.12
This data goes a long way to explaining much of the average active manager’s persistent underperformance. Sharpe observed in his 1991 paper “The Arithmetic of Active Management” that the sum of all money managers equals the market, making it impossible to outperform on average, net of fees. Hence, it is not surprising to find the average “active” manager with a portfolio closely resembling a passive product, and delivering nearly passive returns when adjusted for fees. Indeed, research estimates that only a small percentage of funds (16%) engages in what Petajisto defines as active stock picking (taking “large but diversified positions away from the index”).13
Exhibit 2: Alpha and R2 Levels (Entire Universe)
Alpha measured using FFC model. Source: Amihud and Goyenko 2013.
Exhibit 3: Alpha and R2 Levels (Managers with Highest Prior-Period Alpha)
Alpha measured using FFC model. Source: Amihud and Goyenko 2013.
Levels of Volatility Affect Performance Opportunity
The amount of available alpha in the market (or the amount of opportunity to outperform even for the highly selective managers) is not consistent over time. Researcher Anna Von Reibnitz studied this phenomenon in a paper on the impact of cross-sectional dispersion of returns on mutual fund performance. The paper strongly suggests that the market environment affects manager performance.
Cross-sectional return dispersion measures the difference in performance of stocks within an index or a benchmark. The math states that if all the stocks in the S&P 500 Index rise or fall the exact same amount in a period, generating cross-sectional return dispersion of zero, outperformance in that period will be impossible. Von Reibnitz points out that the market has periods of high and low cross-sectional return dispersion (Exhibit 4). From 1972 to 2013, the active manager universe generated excess returns measured as FFC alpha of 3.06% on an annualized basis during the months ranked in the top 20% of return dispersion. During periods of least return dispersion, the universe of active managers generated FFC alpha of -0.27% based on the same FFC alpha methodology. Note that the highly selective active managers, defined as those in the lowest R2 quintile, generated FFC alpha of over 9% on an annualized basis during the periods of highest cross-sectional return dispersion.
Exhibit 4: Cross-Sectional Return Dispersion is Cyclical
Source: Active vs Passive: How Will the World of Investing Evolve? (part 1 of 2) dated January 31, 2017. © UBS 2018. All rights reserved. Reproduced with permission. As of November 30, 2017.
In short, highly selective active management outperforms when there is dispersion in the market, according to Von Reibnitz’s study. Outside of the top two dispersion quintiles, highly selective managers don’t generate meaningful alpha despite outperforming over the period studied (Exhibits 5 and 6).14
Exhibit 5: Dispersion Provides Opportunity for Highly Selective Active Managers (1972–2013)
Annualized net alpha in the top and bottom quintiles of cross-sectional return dispersion shown by R2 quintiles. Alpha measured using FFC model. Source: Von Reibnitz 2015.
Consequently, it is not surprising that active performance, even for many highly selective managers, has been lackluster over the last several years. Years with generally high cross-sectional return dispersion were 1974–75, 1982, 1988, 1990–91, 1998–02, and 2008–09.15 This helps to explain Cremer’s findings that since 1990, highly selective active, patient managers had the best performance in 1990–2001 and 2007–13.16 Recent years have featured historically low cross-sectional return dispersion (which may be the result of historically low market volatility) and, hence, low opportunity for outperformance.
Exhibit 6: Low R2 Outperforms in Periods of High Return Dispersion (1972–2013)
Annualized net alpha for the highest and lowest quintile R2 shown in each cross-sectional return dispersion quintile. Alpha measured using FFC model. Source: Von Reibnitz 2015.
We believe the history of cross-sectional return dispersion suggests that exactly timing highly selective active managers’ periods of outperformance is difficult. Even the best performers will go through some periods of underperformance, especially when they have high active share. In an analysis done by Vanguard, 98% of outperforming managers have at least four years of underperformance over the 2000–14 period, with 59% underperforming for seven years or more.17
Where Are the Highly Selective Active Managers?
One key message from both active share and R2 research is that highly selective active managers are scarce. The proportion of managers that fit this description decreases even further when patience — longer holding periods — is added as a characteristic. One recent study found that only 1.6% of mutual fund assets are managed with a combination of high active share and patience.18 Index funds did not really become well accepted until the late 1990s, while ETFs did not become widely available until the 2000s; the SPY, which tracks the S&P 500 Index, launched only 25 years ago. We submit that, in the absence of competition for capital from passive products, active managers’ success was at least partially measured by adherence to a specific investment style — value, growth, etc. — which was measured by tracking error. To minimize tracking error, many active managers likely shifted portfolios so that they more closely resembled the benchmark. However, as the prevalence of low-cost passive products increased, investors started to substitute these active managers with ETFs and index funds. This move was likely hastened by low market volatility, which dampened active returns.
Academic research shows that a group of skilled, truly active managers have systematically outperformed their benchmarks by significant amounts. As mentioned, these managers can be identified ex-ante using both active share and R2 methodologies defined in academic research. They share key common characteristics: high active share or low R2 (one must differ from the benchmark to beat it) due to their superior conviction and skill; patience, expressed as long holding periods (one must have the opportunity to take advantage of long-term mispricings); and reasonable fees (high fees detract from performance). Highly selective active managers are especially well positioned to add value today, particularly if volatility and return dispersion begin to re-emerge after years of low volatility.