The rise of passive investment strategies may play a role in the market’s increasing fragility.
A more detailed version of this paper will be published in Financial Analysts Journal in 2012.
Index investing has experienced tremendous growth over the past two decades. This phenomenon is undoubtedly due, in part, to the appeal of index mutual funds and exchangetraded funds; they generally offer lowcost, comprehensive, diversified exposure to various market segments. Indexing’s appeal is compounded by the challenges of sourcing skilled active managers, which, of course, is a zero-sum—even negative sum-game after costs.
The popularity of index funds and ETFs, however, comes at the cost of “trading commonality,” or “basket trading,” across the market. Many stocks within an index being traded are simultaneously bought and sold. As a consequence, the stocks in an index tend to move together throughout the trading day.
At the same time, there has been a steady increase and convergence of U.S. equity betas across size and styles since 1997. Exhibit 1 plots the time series of equal-weighted cross-sectional beta estimates for U.S. stocks in four major equity size and style groups. We will return to this figure in detail later as A more detailed version of this paper will be published in Financial Analysts Journal in 2012. Index investing has experienced tremendous growth over the past two decades. This phenomenon is undoubtedly due, in part, to the appeal of index mutual funds and exchangetraded funds; they generally offer lowcost, comprehensive, diversified exposure to various market segments. Indexing’s appeal is compounded by the challenges of sourcing skilled active managers, which, of course, is a zero-sum-even negative sum-game after costs.
The popularity of index funds and ETFs,
however, comes at the cost of “trading
commonality,” or “basket trading,” across the
market. Many stocks within an index being
traded are simultaneously bought and sold. As
a consequence, the stocks in an index tend to
move together throughout the trading day.
At the same time, there has been a steady increase and convergence of U.S. equity betas across size and styles since 1997. Exhibit 1 plots the time series of equal-weighted cross-sectional beta estimates for U.S. stocks in four major equity size and style groups. We will return to this figure in detail later as we study how this rise in systematic risks connects with the rise in popularity of passively managed index funds and the associated increased trading commonality. Through a battery of tests, we examine the impact of trading commonality related to passive investing and how this affects systematic market risk.
Wurgler (2010)1 reviewed the consequences of index investing and suggests that index investing is distorting both stock prices and risk/return trade-offs. This, in turn, may distort a host of things, including corporate investment and financing decisions, investor portfolio allocation decisions, and assessments of fund manager skill. These effects could intensify if index-linked investing continues to gain popularity.
Growth in Index Trading
During the past three decades, there has been substantial growth in institutional investing. The average portion of a firm’s equity shares held by institutions has grown from 24% in 1980 to 44% in 2000 and to 70% in 2010. The influence of institutional investing has been steadily growing and has become a dominant force in today’s stock market activity.
As the role of institutional investors has increased, the amount of assets invested in index mutual funds and ETFs has also risen rapidly. Exhibit 2 compares total U.S. equity assets under management for active mutual funds versus passively managed funds from February 1993 to September 2010. We measure passively managed investments as assets invested in index mutual funds plus ETFs.2 As of October 2010, total assets in equity mutual funds and ETFs reached $3.5 trillion. Of that, $2.3 trillion was actively managed, with the remaining $1.2 trillion being passively managed. The passively managed assets are split up almost evenly, with $619 billion in equity index mutual funds and $605 billion in equity ETFs. As Exhibit 2 shows, passively managed funds represent only about one third of all fund assets, but the average annual growth rate for passive assets during the past 17 years has been about twice that of actively managed assets (26% per year for passive versus 13% per year for active).
Measuring Trading Commonality
The accelerating growth of passively managed assets places increased demands on indexrelated trading. This index-trading activity, in turn, precipitates higher trading commonality in the cross-section of stocks that make up each index. These basket orders are sometimes spread over a few days or weeks in an effort to minimize the price impact. Buying (or selling) in one period will likely be followed by more buying (or selling) in the next period as the index manager attempts to mitigate market impact by spreading out the required order flow over time. Furthermore, index funds with a similar focus create similar volume changes for many stocks to the extent that their fund holdings overlap. It follows that index-related trading must, at the margin, reduce the cross-sectional dispersion of trading-volume changes.
Understanding the market impact of indexrelated trading requires a focus on how such trading affects a change in the volume of stock trading from period to period, rather than in the absolute level of trading volume. We, therefore, focus on examining the impact of the cross-sectional dispersion of trading volume changes on overall market activity. To accomplish this, we first measure the logarithmic change in trading volume from one period to the next, calculated as:
We then calculate the cross-sectional dispersion of volume change as the standard deviation of VCt as measured across all stocks for each time period (t).
Exhibit 3 shows the relationship between the cross-sectional dispersion of trading volume changes and the growth in passive equity fund assets, calculated as the percentage of total passive assets relative to the total U.S. stock market capitalization. (We refer to this ratio as the “passive market share.”) The passive market share is linearly interpolated from 1979 to 1993 by reasonably assuming that the percentage of passive assets was zero percent in 1976.3
Our novel results from Exhibit 3 show that the dispersion of cross-sectional volume changes has two distinct regimes: a flat regime from 1979 to 1996 and a persistent declining regime from 1997 to 2010. Further, the second regime declines at an almost constant rate. Exhibit 3 also shows that decline in cross-sectional volume change over time is nearly a perfect inverse of the growth in passive assets. This finding appears highly consistent with our thesis that increased index trading volume drives higher return covariance among an index’s stocks. The systematic decrease in changes in stock volume dispersion over time is consistent empirically and intuitively with the rise in index trading during the same time period.
In our study, we use the growth of passive assets as the proxy of index trading. However, index trading can come from a broad range of investment vehicles. For example, high-frequency traders may seek to capture any pricing spread differential between ETFs and index funds. Such common, systematic trading across a basket of stocks would likely contribute to lower dispersion of changes in trading volume.
Measuring Pairwise Correlations
As we mentioned earlier, trading commonality among index funds and ETFs tends to result in the prices and volumes of constituent stocks moving in the same direction during any given period. Therefore, we would expect to see an increase over time in the average pairwise correlations among both stock prices and trading volumes, especially after the late 1990s (the period associated with rapid growth in passively managed assets). In Exhibit 4, we plot the average of all stock-by-stock correlations for the NYSE/ AMEX/Nasdaq universe of stocks. The figure shows that the equal-weighted average pairwise correlation for both daily price return and trading volume changes increases rather dramatically after 1997. The correlation for each stock is measured on 26 weeks of daily returns and the corresponding volume changes and then averaged across all stocks.
As with Exhibit 3, Exhibit 4 also shows that the average pairwise correlations for both price return and change in volume have two distinct regimes: a relatively flat period (from 1980 to 1996) and a sloped period (from 1997 to 2010). Changes in prices and volumes each became meaningfully more pairwise-correlated after 1996. Consistent with our thesis, we suggest that this regime change was in large part a result of increased index trading, which resulted in a significant increase in average pairwise correlations among stock prices and changes in volume.
One may argue that the sloped period experienced unusual market turbulence -namely the 2000 technology, media, and telecommunications crash and the 2008 financial crisis—and that these effects might then be responsible for the increased average correlation observed. While correlations across securities and across markets do tend to increase during crashes (in what is known as asymmetrical correlation), there are only about four years (2000–2002, 2007–2009) of the severe downside markets in this period, while the other 10 years experienced more-normal market conditions. So, our findings on the shift in average pairwise correlation suggest meaningful persistent increase from 1997 to 2010, regardless of the status of the market.
Impact on Systematic Risk and
How does this tie into the steady increase and convergence of U.S. equity betas across size and style? The answer to our question lies, in part, with links to trading commonality driven by index trading.
In Exhibit 4, we showed that average pairwise correlations among stocks have increased since 1997. From this, we can reasonably infer that the rise in correlations has simultaneously yielded a rise in average betas.
To estimate our betas for Exhibit 1, we use 26 weeks of equal-weighted daily returns and then sort the stocks by size or book/market ratio (for our two style groups). The equalweighted average beta for small stocks is then measured as the average beta for stocks whose market cap is below the 50th percentile of the universe. Likewise, the equal-weighted average beta for growth stocks is the average beta for those stocks whose book/ market ratio is below the 50th percentile of the universe.
Strikingly, the observed differences in betas have dramatically narrowed over the past 10 years, as betas for all size and style categories have converged. This convergence comes from the rise of small- and valuestock betas to the same level of large and growth stocks during the first half of the decade, beginning in 2000. During the second half of the decade, the average betas across all four size and style categories together began a steady rise, exceeded one, and have remained elevated ever since.
This increase and convergence of betas suggests that diversification benefits during the second period (1997–2010) were reduced for all types of portfolios (small, large, growth, or value), a situation that remained (at least) through the end of our study period. This unambiguously means that there is increased risk of unexpected negative events for investors.
To further demonstrate the reduced diversification benefits since 1997 for both large-cap and small-cap stocks, we examine the excess return volatility for portfolios of large- and small-cap stocks while varying the number of stocks in the respective portfolios. For both the largest- and smallest-cap stock quintiles, we constructed 26-week, equal-weighted portfolios containing a different number (five to 50) of randomly selected stocks (without replication), similar to the empirical methodology of Kamara, Lou, and Sadka (2010)4. Using daily returns, we calculated the annual excess return volatility of each portfolio relative to the market, defined as the difference between the portfolio’s standard deviation of return and the standard deviation of return of a value-weighted portfolio made up of all the stocks in the sample. To examine changes over time, we again subdivided our sample into two periods, 1979–1996 and 1997–2010. For each period, we calculated the average annual excess volatility for each portfolio.
Exhibit 5 shows that for both large and small portfolios, diversification benefits are dramatically diminished in the second period (1997–2010). In other words, an investor who wishes to maintain the same excess return volatility level after 1997 would need to meaningfully increase the number of stocks he or she holds, for portfolios of both large- and small-cap stocks.
Passively managed index funds and ETFs have seen accelerating growth of their assets in recent decades, and passive funds now hold more than half of the level of assets held by actively managed mutual funds. The increased level of trading associated with passive investing, however, comes with important consequences for investors. It means an increased trading commonality among index constituents through the interactions of market participants. Such trading commonality then gives way to a rise in systematic fluctuations in overall demand. This, in turn, leads to a fundamental impact on the overall market and investors’ portfolios.
The rise in systematic risk emanates, in part, from an increase in trading commonality across time and across stocks associated with growth in passive investing. Though perhaps not the only explanation for rising systematic risk, our results provide strong evidence that the observed increase in trading commonality since 1997 has indeed led to lower cross-sectional dispersion of volume changes and, therefore, greater systematic risk since then.
As evidence, we report that both pairwise and cross-correlations between return volatility and volume volatility have significantly increased since 1997. Furthermore, we show that the diversification benefits of equity investing have decreased for all styles of stock portfolios (small, large, growth, or value). Altogether, our results suggest that the U.S. equity market has become more fragile over recent decades.
1 Wurgler, Jeffrey. 2010. “On the Economic Consequences of Index-Linked Investing.” Forthcoming in Challenges to Business in the Twenty-First Century: The Way Forward, W.T. Allen, R. Khurana, J. Lorsch, G. Rosenfeld, Eds. American Academy of Arts and Sciences. 2 Most ETFs—although not all—have passive mandates and track an index, such as the S&P 500 or the Russell 3000. 3 The Vanguard 500 VFINX, founded in 1976, was the first index mutual fund. 4 Avraham Kamara, Xiaoxia Lou, and Ronnie Sadka. 2010. “Has the U.S. Stock Market Become More Vulnerable Over Time?” Financial Analysts Journal, Vol. 66, No. 1: 41–52.