Closet indexing continues to flourish in mutual fund land (and bring down average performance)

Retail Investing 29 Nov 2010

The stated aim of most hedge funds is to produce alpha.  In a sense, hedge funds are refined expressions of active management.  But hedge funds don’t have a monopoly on alpha-generation.  Although mutual funds deliver large portions of market beta, some mutual funds are more active than others.  So it can be useful to examine the differences between active and passive mutual funds in order to better understand the differences between (prominently active) hedge funds and (relatively passive) mutual funds.

Some mutual funds allege to be active, but their performance seems to track equity indices pretty closely.  So are these managers B.S.-ing their clients or are they really making active bets?  It turns out they can make a large number of active bets on individual securities, but if those active bets are randomly distributed across all stocks, they can cancel each other out and leave investors with an obfuscated index fund.  Conversely, a manager might make very few active bets on individual securities but have them all occur in, say, two specific sectors.  If the manager overweighs every tech name in their portfolio, the aggregate dollar value of those active bets can be modest, but the effect on the overall portfolio can still be significant.  In other words, all of their active bets are pulling in the same direction (the tech “factor”).

You might call the first manager a “stock-picker” and the second manager a “factor-bettor.”  But which is better?  Which provides higher, more consistent returns?

A study by Yale’s Antti Petajisto of US equity mutual funds released in October provides some insight here.  If you were an AAA reader back in August 2006, you may remember this post about Petajisto’s earlier work with Yale colleague Martijn Cremers.  At that time, we recommended that if you read only one academic paper in 2006, that ought to be the one.  The new paper makes several refinements to its earlier cousin.  Most importantly, the new paper includes data up to the end of 2009 while  the previous paper used data up to the end of 2003.

Petajisto uses the concept of “active share” to determine the extent to which a fund is a stock-picking fund.  Readers of Cremers and Petajisto’s paper will recall that this refers to the percentage of the fund’s security weightings that differ from the weightings of those in the index.

Naturally, there is a positive relationship between active share and the resulting tracking error of the fund, that is the deviation from the index returns.  The greater the active share, the greater the divergence from the index.  The table below from the paper makes this point by showing the active share and resulting tracking error of 1380 funds in the study.  When you look across tracking error buckets from low (0-2%) to high (14%+), you see a monotonically rising mode to each distribution of active share. The lowest tracking error was logged mainly by funds with a 0-10% active share, ie. index funds.  The highest tracking error (14%+) was logged mainly by funds with an 80-100% active share.

Curiously, many highly active funds seem to produce markedly lower tracking errors; these are the ubiquitous “closet indexers”  and they only pretend to be active.  They even act like they’re active by taking active bets.  But at the end of the day, they deny the result: an approximation of the index at a much higher cost than the real thing.  Or, as Petajisto says,

“This is of course the opposite of what investors are paying active managers to do, since investors can always buy a cheap index fund if they want to reduce volatility relative to the index. […] A closet indexer charges active management fees on all the assets in the mutual fund, even when some of the assets are simply invested in the benchmark index. If a fund has an Active Share of 33%, this means that fund-level annual expenses of 1.5% amount to 4.5% as a fraction of the active positions of the fund.”

This can be interpreted as a loose reference to the thesis of Ross Miller’s seminal Journal of Investment Management article on the true cost of active management.

Petajisto also channels Miller when he uses Fidelity’s Magellan fund as an example of how active share can change depending on who is managing the fund.  He laments that Magellan’s active share “plunged” soon after Robert Stansky took over management in 1996.  The result, reports Petajisto is that “performance suffered during this closet indexing period.”  Although, he says, it was “not disastrous performance. And just as a note, Petajisto is charitable compared to Miller, who analyzed the effective fee for active management of Magellan under Stansky and wrote a paper on it called Stansky’s Monster: A Critical Analysis of Fidelity Magellan’s ‘Frankenfund’.

The proportion of funds that can be described as “closet indexers” (active share 0-60%) rose dramatically in the 1980’s and 1990’s.  But did that trend continue?  The chart below from the paper shows that active management actually staged a modest comeback when markets bottomed-out earlier this decade.  But it rose again to all time highs by the end of last year.

So which kind of fund performs better?  As you might guess, it’s the stock pickers. The chart below, based on data in Table 6 of the paper, shows the benchmark-adjusted returns of each type of fund (high, medium and low active share, plus a couple of variants):

We wondered if stock pickers or concentrated funds were also more consistent than closest indexers since their stock picking skills shouldn’t really wax and wane with the markets.

It turns out that concentrated funds showed what Petajisto calls “remarkable persistence” in their returns.  In other words, last year’s winners are likely to be winners again.  Conversely, he found that only the closet indexers didn’t show any return persistence.  This seems to fly in the face of the efficient markets hypothesis.

In the end, Petajisto’s conclusion reads like the alpha/beta separation philosophy of this website:

“When selecting mutual funds, they should go with only the most active stock pickers, or combine those funds with inexpensive index funds; in other words, they should pick from the two extremes of Active Share, but not invest in any funds in the middle.”

It’s not a huge stretch to move beyond stock picking mutual funds and include long/short equity hedge funds in this analysis.  In Cremers and Petajisto’s earlier paper, they noted:

“…the Active Share of a hedge fund can significantly exceed 100% due to its leverage and net short positions in individual stocks.”

So this analysis suggests that long/short strategies, whether in a mutual fund or hedge fund context, may indeed be a superior form of investing.

Be Sociable, Share!


  1. Luke
    November 30, 2010 at 6:00 pm

    Interesting article and not unsurprising, but there are some flaws in it…or at least the way it is presented. (Caveat that I haven’t actually read the paper yet.) First, what is the benchmark against which the tracking errors and active share are calculated? If it is the S&P 500 across the board, then that is just plain foolishness and ignores the realities of the mutual fund world. Most of the 1,380 funds are likely not benchmarked to the S&P 500. Should a small cap fund be measured against the S&P? I hope most here realize the answer is no. So, only if the active share and tracking error are measured against the proper index are this data even remotely reliable.

    Second, the bar chart is very interesting. If I’m reading it correctly, funds with high active share AND high tracking error underperform on average. Hmmm. But the broad category of high active share outperforms. Does this mean high active share and medium/low tracking error does that much better? It seems to indicate this is true, but someone can correct me. Is it that inconceivable that highly active managers might use good risk management systems to lower their vol vs. the index? (i.e., lower their tracking error). After all, TE is obviously not a good predictor of alpha in and of itself. If I was a highly active manager, and I wanted to get on 401k platforms, etc, where the big money comes in for mutual funds, then absolutely I would do what I could to maintain my level of “activeness” and lower my tracking error, thereby increasing my Information Ratio. IR is what every CIO and fund consultant out there wants, and it’s been shown to be a good tool. Just because a fund has high active share and “only” a 6% tracking error (which isn’t low, by the way) does NOT mean it’s a closet indexer…just that they use good risk management to keep the vol vs. the index at a reasonable level.

    Anyway, enough for me. Am I missing something here?

  2. Daniel Plainview
    December 1, 2010 at 11:25 am

    Luke in the paper he discusses benchmarking and the funds are not all bench marked to the S&P 500.
    The only way an active manager can add value is by deviating from her benchmark index
    The way I understand it is tracking error measure systematic risk the volatility of the fund not explained by movements in the fund’s benchmark. Active share measures the percentage of the fund that differs from the benchmark; so you can have a fund with high active share and low tracking error( The two measurers are positively correlated) you get a concentrated fund if it takes both active stock selection and has high tracking error. In the paper how active management shows up depends on one key question: do the active positions have exposure to systematic risk? The data supports stock picker’s performance over tactical allocators.
    The last comment has me confused, as defined in the paper closet indexers have lower active share and lower TE, plus the closet indexes charge fee as if they are active managers. An active manager has high active share and low TE.

  3. Luke
    December 1, 2010 at 12:11 pm


    Thanks for the reply. Tracking error is the standard deviation of the fund’s excess returns versus the index. It is usually measured on either a monthly or quarterly basis. So, if a very active stock picker (high active share) has consistently outperformed the index over the time period measured, the tracking error would be quite low even though the fund isn’t anywhere close to a closet index. So you just have to be careful with tracking error. Many people equate lower tracking error with lower potential for alpha. And while that may be true on an average basis, many funds have shown the ability to create very nice alpha versus the appropriate index and do so with a low/moderate tracking error because of the risk management approach that they take.

    The last comment has to do with the specific comment in the article above: “Curiously, many highly active funds seem to produce markedly lower tracking errors; these are the ubiquitous “closet indexers” and they only pretend to be active.” Closet indexers, to your point, are those with low active share and low TE. High active share and closet indexer are mutually exclusive in my mind. If the fund manager uses risk management to keep the tracking error from being too high, but yet the active share is still very high, is that a closet index? Not a chance. Just beacuse the R^2 might be higher doesn’t really mean anything except your volatility against that index is lower, which for most people, isn’t a bad thing if you are still getting the stock picking.

Leave A Reply

← Investors in event-driven strategies told to look for manager with extensive hat collection Insider Traders: Rogues or Whistleblowers? →