A buzzword in the investment community these days is active share, a specific way to measure how different a portfolio is from its benchmark. On its own, such a measure wouldn’t attract much attention; what’s gotten people excited is the claim that higher active share predicts higher excess returns.
So does it? No, as we show in a new AQR white paper, “Deactivating Active Share.” That shouldn’t be too surprising. Just because a portfolio claims to have “high conviction” and thus looks a lot different than its benchmark doesn’t mean it should perform better than its benchmark. The idea that success requires high conviction may seem intuitive and appealing, but there’s just no clear economic reason or theory to link “different” to “better.” 1 1 Close It’s even more difficult to explain why active share should predict performance while other measures of “activity” such as tracking error do not! Here, there may be some confusion about “probability” and “amount” of outperformance. That is, being “different” with higher active share or higher tracking error may lead to different returns but not necessarily higher returns. And that’s exactly what we find in the data: it’s just as likely for active share to be associated with outperformance as underperformance.
So where did the active share results come from? We use the same data as the original studies on the topic and find it wasn’t a fund’s active share driving the results; it was the market cap of its benchmark. Once you control for the apples-to-oranges problem of different funds having different benchmarks, active share’s usefulness as a predictor of alpha goes away.
However, proponents of active share make a good point when it comes to fees — specifically, that it’s a rotten deal for investors to pay “active-like fees” for a fund that is essentially passive. But even here, we'd argue (ex ante) tracking error is a better way to evaluate how "active" a fund is. Tracking error is imperfect as well, and by no means a panacea, but unlike active share it recognizes correlations between stocks, industries, etc. Moreover, tracking error is more comparable across benchmarks (an active share of 70% may be large if your benchmark is the S&P 500, but small if your benchmark is the Russell 2000) and asset classes (active share can be meaningless for long/short, levered strategies). As we’ve argued before, fees should be in line with the active risk a manager takes, and tracking error is a more consistent and useful way to measure that.