Andrea Frazzini, Jacques Friedman and Lukasz Pomorski question whether there is any evidence to show that active share predicts active managers’ relative performance
The investment community is abuzz with ‘active share’. Many asset managers tout their active share, leading investment consultants emphasise it as a key ingredient for manager selection, and some institutional investors have gone so far as to embed a high active-share requirement in their investment guidelines.
Active share does seem to offer a simple solution to a complex problem: an easy-to-calculate, intuitive statistic that promises to predict investment performance. Unfortunately, active share is no panacea. The simple truth is the statistic has no backing from economic theory and, as we will show, empirical results for its manager performance prediction are weak.
Active share is designed to gauge the degree of active management in a portfolio. It takes the weights the portfolio allocates to various securities, and if they are far away from the benchmark index weights (high active share), the manager is thought to be more active.
According to its proponents, active share captures more than just how far from the benchmark you are; it also predicts fund performance. Two papers that developed the measure, published in 2009 and 2013 by Martijn Cremers and Antti Petajisto, and the white papers that followed, suggest investors can improve their performance by over two percentage points per year simply by selling their low-active-share funds and investing in high-active-share funds. A new AQR working paper, ‘Deactivating Active Share’, evaluates this bold claim.
Our concerns are twofold. First, there is little theory to suggest that active share might predict future returns. Of course, a portfolio with the same weights as a benchmark (zero active share) cannot outperform. But there is no reason why deviations from the benchmark should be positive rather than negative for relative returns. Simply put, you’re not more likely to be right just because you have a high conviction.
With no economic theory behind active share, all support for the measure rests on one piece of empirical evidence: the analysis of US equity mutual funds carried out in the two papers mentioned above. Herein lies our second concern – the robustness of that evidence. We evaluate it using the same underlying dataset, and the same sample period, as the seminal active share studies, and offer a series of observations we think are informative about active share’s usefulness for investors.
First, the active share of a fund depends strongly on the market capitalisation of its benchmark. Figure 1 illustrates this point by presenting the average active share of funds grouped by their benchmark. The graph makes it clear that a sort on active share is also a sort on the market cap of the benchmark. For example, large-cap funds (left-hand side of the figure) have lower active share than all-cap funds; all-cap lower than mid-cap (middle); mid-cap lower than small-cap (right-hand side). Investors who rely on active share should sell their large-cap funds and invest the proceeds in small-cap funds.
Of course, few would look at all US equity funds at the same time, choose a high-active-share fund and accept whichever benchmark falls out of the analysis. Most investors do the opposite. They start with a preferred benchmark, and then evaluate funds within that benchmark. Figure 2 compares apples to apples by pitting high-active-share funds versus low-active-share funds within the same benchmark and measuring the difference in their performance (technically, their annualised alpha to the standard model that controls for the market exposure, and value, size, and momentum tilts).
After this simple step, any evidence of active share predictability disappears. Within individual benchmarks, active share is as likely to predict better as it is to predict worse performance; it predicts outperformance in eight out of 17 indices and underperformance in the remaining nine. In only two benchmarks is the relationship statistically significant: positive in one, negative in the other. Within each market-cap category, we find benchmarks where active share predicts positive performance and benchmarks in which it does the reverse (for example, outperformance within the S&P 500, but underperformance within the Russell 1000).
“Within individual bench-marks, active share is as likely to predict better as it is to predict worse performance”
Overall, we find no support for the idea that active share predicts performance or that investors would be better off in high-active-share funds. We do agree with active share proponents on one important point: fees matter. Being close to a benchmark may not imply poor performance, but it cannot justify high fees. In general, fees should be commensurate with the active risk funds take. There are many ways to measure the degree of ‘activity’ other than active share, including predicted and realised tracking error or other concentration measures. A prudent investor should use multiple measures to determine if a manager is taking risks commensurate with fees.
Andrea Frazzini and Jacques Friedman are principals and Lukasz Pomorski is a vice-president at AQR Capital Management