There is a growing body of research noting that return correlations for individual securities have been increasing. A recent paper by Sullivan and Xiong  titled “How Index Trading Increases Market Vulnerability” not only documents this occurrence, but also cites a potential culprit: the increasing popularity of index funds. Since index funds tend to be value-weighted—and therefore trade the same securities in the same relative portion—as index funds gain more assets, more and more securities are being traded in the same way at the same time, regardless of the under-lying attributes of the stocks themselves.
The increasing “commonality” across individual securities doesn’t appear to bode well for active managers, who by definition seek to add value through individual security selection. This paper will provide insight as to how the level of “active” management has been changing in actively managed mutual funds over the last 21 years by reviewing the historical relationship between gross returns and a benchmark based on each fund’s respective Morningstar category. As one might expect, given the increase in individual security correlations, the average mutual fund correlation to its benchmark has increased over the test period, suggesting that active managers are in fact becoming less active.
Here Come The Index Funds
Index investing has exploded over the last two decades, growing at roughly twice the rate of active investments. Of households that owned mutual funds, 31 percent owned at least one index mutual fund in 2010. The Investment Company Institute estimates that 37 percent of all index fund assets were invested in S&P 500 index funds, while 32 percent were tracking some other domestic equity index. About 40 percent of the new money that flowed into index funds was invested in funds indexed to bond indexes, while one-third was directed toward funds indexed to global and international stock indexes, and one-quarter went to funds indexed to domestic stock indexes.
While equity index assets are only 14.5 percent of mutual fund assets, Sullivan and Xiong estimate indexes represent roughly one-third of total fund assets today when factoring in ETF assets. The first ETF, the SPDR S&P 500 ETF (NYSE Arca: SPY), was introduced in January 1993. Significant growth in ETF assets really didn’t start until 2000, though, when there were roughly $66 billion in assets and 100 options; those numbers have grown to more than $1 trillion in assets with more than 1,400 ETFs available. ETF trading has grown from virtually nil in 2000 to now accounting for roughly 30 percent of total dollar trade volume and about 20 percent of total share volume.1
Just as indexing has changed the nature of stock ownership, so too has the rise of institutional investors. The average fraction of a firm’s equity shares held by institutions has grown from 24 percent in 1980 to 44 percent in 2000, and reached 70 percent in 2010.2 In a world where all institutional investors are trading according to their own respective beliefs, this may not be a problem; however, with the rise in indexing, more stocks are being held by institutions that seek to replicate the return of a given index and minimize tracking error.
Since the vast majority of indexes are value-weighted, they tend to hold the same stocks in the same relative portions. Therefore, when an investor buys (or sells) an index that holds one of these securities, the security is bought (sold) in conjunction with the other securities that make up the index, regardless of the relative attractiveness of the stock itself. This creates an increase in “commonality” among stocks, especially those in the more popular “baskets,” such as those in the S&P 500.
Impact On Active Managers
The most obvious impact on actively managed portfolios from increasing individual security correlations would be higher levels of market correlations (i.e., a decrease in the “active” portion of the portfolio). Although this might seem intuitive, it may not necessarily be the case, if (for example) the portfolio manager was trying to maintain some level of tracking error against his or her respective size and style benchmark. If the portfolio manager were to recognize that correlations among individual securities were increasing, he or she could decide to hold fewer stocks or tilt the portfolio more toward certain sectors.
If a portfolio does not maintain a constant level of tracking error (on average), the portfolio effectively becomes either more active or passive through time. If the portfolio is becoming less “active” and charging the same fee, it becomes increasingly unlikely that the portfolio manager will outperform his or her benchmark. This is because, in the absence of any active management skill (which should cancel out in the aggregate, regardless), the active manager should be expected to underperform the appropriately selected benchmark by the total fees of the portfolio.3 Therefore, in order to outperform the benchmark, the portfolio manager will need to take on active risk, thereby deviating from the benchmark.
If individual securities are, in the aggregate, exhibiting less idiosyncratic risk and the portfolio manager does not change the risk profile through some other means, the probability of having a return more similar to the benchmark increases. This in turn increases the probability the portfolio manager will underperform the benchmark by the total amount of fees.
In order to determine whether or not active managers are becoming more or less “active,” an analysis was performed. For the analysis, all data were obtained from Morningstar Direct. The mutual funds included in the analysis are those categorized by Morningstar as domestic equity mutual funds from January 1991 to December 2011. Funds are included regardless of when they exit or leave the test set, and since the available test population is updated monthly, survivorship bias is not a concern.
In order to be included, the mutual fund must be in one of the nine domestic equity style boxes. Also, to ensure fund “purity,” only those funds with the same Morningstar Category and Style Index are included for a given test period. The oldest share class for each fund is used, and any fund classified as an enhanced index, index fund, or fund of funds is excluded from the analysis. These screens limited the number of funds to 2,059 over the entire test period.
Each fund is compared with its respective Russell Index, across value, blend and growth, using the Russell 1000, Russell Mid Cap and Russell 2000 for large cap, midcap and small cap, respectively. The analysis is conducted on a rolling monthly basis and the available test set is determined for that respective historical rolling period. The analysis uses a 12-month rolling historical period for all calculations. All returns are gross returns; i.e., they do not include the impact of investment management fees. The purpose of the analysis is not to opine on the great active versus passive debate, but rather to determine how the nature of active management has changed over time. Note, however, that since expense ratios are relatively constant, they would have little effect on the statistics calculated for this paper (primarily correlation and standard deviation).
While the average correlations are determined at the individual Morningstar category level, the category results are aggregated into a single value for each rolling test period based on the weighted average number of funds available for the given test period by category. This approach overweights styles that typically receive more assets (e.g., large cap versus small cap). However, although only the aggregate results are presented, the overall results are virtually identical across the individual categories.
Past research by Sharpe  noted that style and size explain approximately 80 to 90 percent of mutual fund returns, while stock selection explains only 10 to 20 percent. The results of this analysis confirm these general findings, with the average correlation of the respective fund to its Russell index based on its category being 0.93 (and the average correlation 0.94). That translates into a coefficient of determination (R²) of 86 percent, which is right in the middle of Sharpe’s original estimate. However, the correlation has not been constant through time, as exhibited in Figure 1, which includes the average rolling annual correlation to the respective category index.
Since 1991, the average correlation for actively managed mutual funds has been increasing. As of December 2011, it was at its approximate highest level in history, with the average fund having a correlation of 0.982 to its respective category index. Viewed differently, a correlation of 0.982 means that 98.2 percent of the return of a given active manager can be described entirely by the underlying benchmark index. This suggests the active manager is only adding roughly one-thirtieth of the total deviation in returns, but in many cases charging 10 times or more than what a comparable passive strategy costs. Note that the t-statistic associated with slope is 11.73, suggesting an incredibly high level of statistical significance.
Another way to view the changing “active” exposure of mutual funds through time is the standard deviation of the correlations. This metric captures the dispersion of all active managers through time. It is possible that while the average is increasing, there could also be a greater level of dispersion among portfolio managers. Unfortunately, as demonstrated in Figure 2, the average rolling category correlation standard deviations have also been decreasing. This suggests that, on average, an increasing number of actively managed mutual funds are clustering more and more tightly around their respective category benchmarks. The t-statistic associated with slope is -8.84, suggesting an incredibly high level of statistical significance.
The final test to determine the direction of the “active” portion of actively managed mutual funds is based on the average tracking error of the fund versus its respective category index (Figure 3). For this test, other than two noticeable spikes, the clear trend has been a decreasing level of tracking error through time. Again, this suggests active managers are in fact less “active” than they used to be. The t-statistic associated with slope is -3.13, suggesting a relatively high level of statistical significance.
It is impossible to pinpoint an exact reason why actively managed mutual funds are becoming less “active,” but the evidence would certainly suggest this is the case. One theory this author believes could be a driving force behind this change is increased movement to index investing. As more investors “index” their portfolios, more and more trading volume is based not on some technical analysis about the future earnings of a given company (for example), but instead whether or not a given security is included in a particular index (the S&P 500 in particular) and to what extent. Alternatively, it could be that “style purity” is becoming increasingly important for benchmarking purposes. Regardless of the reason, active funds have clearly become less active through time.
These findings present an interesting environment for active management. First, as more money flows to index funds, less money must be flowing to active strategies. Although active as well as passive investments can be “winners” from a positive flows perspective, the relative gain of either category definitely comes at the cost of the other. Second, if higher levels of flows into passive funds do create an environment that makes it more difficult for active management to outperform, and if an increasing amount of flows go to passive/index strategies, the ability of an active manager to outperform is likely to become increasingly hampered. Finally, for an efficient market to exist, there must be some active management. It is beyond the scope of this article to venture a guess as to where this point is, but we don’t appear to be there yet. It will certainly be interesting to see what happens if we do ever get there.
Investment Company Fact Book. 2011. http://www.ici.org/pdf/2011_factbook.pdf
Sharpe, William. 1992. “Asset Allocation: Management Style and Performance Measurement.” Journal of Portfolio Management, vol. 18 No. 2: 7-19
Sias, R.W., L. Starks and S. Titman. 2006. “Changes in Institutional Ownership and Stock Returns: Assessment and Methodology.” Journal of Business, vol. 79 No. 6:2869-2910.
Sullivan, R. and Xiong, J. X. 2012. “How Index Trading Increases Market Vulnerability.” Financial Analysts Journal. Forthcoming, available at SSRN: http://ssrn.
1 Sullivan and Xiong 
2 Sias, Starks and Titman ; Sullivan and Xiong 
3 Both explicit management fees and implicit fees that result from trading, such as the bid/ask spread and commissions