Over the last year or so there has been a lot of buzz around the idea known as “Active Share”. Proposed by Cremers and Petajisto, simply put Active Share is a way to measure the degree of deviation from a passive benchmark from a holdings-based perspective. Traditionally the most common way measure the degree of active management is to use tracking error, which is a returns-based way of measuring active management. The authors of Active Share think their way of quantifying active management would be a useful compliment to tracking error but is inherently better. I’d agree with the former but would question the latter half of their thinking.
So what is Active Share? The actual calculation for Active Share is quite simple. It measures the percentage amount your portfolio differs from a passive benchmark. The scale goes from 0% to 100%. A reading of 0% suggests your holdings are identical to the index. An AS score of 100% says you’re completely different than the index without any crossover at all. The actual formula is:
I could see how this might be useful simply to know how different the holdings are from an index, but there are several limitations one should keep in mind when thinking about Active Share.
First and foremost, it is a holdings-based approach, so all of the usual critiques of holdings-bases analysis apply. Plus, there are some new concerns specific to Active Share. The general holdings based concerns are:
1. Timeliness/Availability of. You’re depending upon the manager to provide you with the full list of their holdings. Maybe they don’t want to give you their holdings. Maybe the only holdings they want to provide are three, six, or twelve months out of date. If that’s the case, your AS reading is immediately out of date.
2. Single Point-in-Time Data. Even if the manager wants to give you their data, it will probably be as of quarter end. Who knows what the data looked like on February 17th or July 23rd? It could potentially be a lot different. Whether the portfolio differs due to window-dressing or simply a change in positions is moot: quarterly Active Share readings only give you four snapshots in a 365-day year.
3. Trends. Because Active Share is a single point-in-time datapoint, you need a whole lot of Active Share scores over time to pick up on trends. You don’t really know if the manager has been more or less active over the years unless you calculate the Active Share score a lot of time for a lot of periods.
Those are general issues about any holdings-based method, which lead to the whole idea of looking at returns as an alternative in the first place. That being said, I can foresee additional complications with Active Share when used in the real world. They include:
4. Benchmark Specification. In order to be useful, you need to make sure the right benchmark is being used. The authors of the paper actually did a good job with their research because they ran Active Share scores for 19 different indices and then specified the best fit. That being said, if you’re relying on someone else to provide you with Active Share numbers unless you have the ability to choose the benchmark you think appropriate you would be beholden to their decision.
5. Syncing. This ties in to point #1 above. If the fund is giving you their portfolio holdings as of, say September 30th and it’s now May 31st, you have to go back to whatever the holdings of the index was on September 30th to get a true apples-to-apple comparison. Unless the portfolios are reconciled to the same point-in-time Active Share could be seriously flawed. Cremers and Petajisto acknowledge this in their paper on page 11.
6. Vaugeness. All Active Share will tell you is how different the fund is from the benchmark. If the fund is deviating in terms of holdings, you have no idea just where else that money might be invested. For example, say a fund has an Active Share reading of 60%, meaning 40% of the holdings are different from the benchmark. But what does “different” mean? Are they simply in stocks that happen to be in a different benchmark? Are the holdings in foreign stocks? In cash? Bonds? Gold? Active Share does not specify this at all.
7. Concentration. This is related to the point above. Okay, assume 40% of the holdings are different than the index. That bet could be a 40% position in a single stock or 80 stocks at 0.5% each. Active Share isn’t going to specify, as far as I can tell.
Finally the authors start making some bolder claims like Active Share is actually predictive and can be used to identify good managers going forward. I’m most skeptical of this.
8. Predictive. The authors of Active Share claim that those managers with the highest active share ranking will add the most value. If it were me, I would dial that down a bit. I would say historically speaking, on average, those managers with higher Active Share scores have tended to have more excess return. That’s a big difference.
9. No guarantees. Just because the manager has taken these bets away from the index, there is the very real possibility that these are BAD bets and will blow up. Case in point? The Fairholme fund. At the end of 2010 Fairholme was frequently cited as a great example of awith a high degree of Active Share with persistent outperformance. The worm turned in 2011, as the year-to-date performance of Fairholme through Oct 31st was a stunning -23.4% versus a small gain of 1.3% for the S&P 500.
Whenever we come out with a new metric at Zephyr, we are always cautious in explaining the pros, cons, uses, and shortcomings of that metric. Our thoughts on outside metrics should be no different. Yes, Active Share is an interesting idea and will have its uses. No, it is not the be-all, end-all metric that will answer all your problems. There are limitations to Active Share that should be understood, just like anything else.