Over time the factor exposures inherent in stocks also tend to drift, and that can pose a problem for investors and portfolio managers. SHELDON McFARLAND and LARRY SWEDROE have been looking at the findings of a new study on the issue. So, how serious is factor drift? And what can you do to address it?
For the most part, the arms race came to an end with the collapse of the Soviet Union at the end of the Cold War in 1991. However, by 1992 we were on the precipice of a new conflict — a factor arms race that would result in a proliferation of factors. The seminal Fama-French papers (the 1992 study The Cross-Section of Expected Stock Returns and the 1993 study Common Risk Factors in the Returns on Stocks and Bonds) kicked off an intense interest among academicians and practitioners in the size and value factors, and ultimately in the identification of additional factors, such as profitability/quality, momentum and volatility (low volatility/low beta). At around the same time the seminal Fama-French papers were published, Morningstar (1992) developed their 3-by-3 Style Box methodology to help identify the size and style (value) categorisation of stocks and mutual funds. The Morningstar Style Box framework quickly became a touchstone upon which both investors and practitioners judge diversified portfolios.
A key aspect of factor categorisation in tools such as the style box, however, is the question of factor dispersion. Within each factor box, how widely dispersed is that factor across available stocks and funds? Secondly, if the exposure to each factor is widely dispersed, how stable is that exposure over time, and should that influence the construction of factor-based portfolios?
These questions were addressed by BlackRock’s Keiko Kimura, Katharina Schwaiger, Deepika Sharma and Andrew Ang in their study Factors with Style, published in the April 2021 issue of The Journal of Investing. They investigated the degree to which exposure to five prominent equity market factors (size, style , quality, momentum and volatility) are dispersed across the universe of U.S. stocks and funds, the degree to which those factors drift over time, and possible methods to control that drift in a factor-oriented portfolio.
Factor dispersion
The BlackRock researchers focused primarily on the large-cap universe, using the Large Cap Russell Indexes (Value, Core, Growth) as the representative universe for the three equivalent Morningstar Style Boxes across the Large Cap row. The researchers measured the dispersion within these three style boxes for all five equity factors between the 5th and 95th percentile and found the dispersion to be large for all five factors. The researchers also found that the benchmarks do not generally exhibit factor tilts (other than the tilt that is obvious from their construction, such as Value or Growth). The takeaways for investors are:
There is substantial opportunity to target factor portfolios of varying degrees of factor exposure within each Morningstar Style Box.
Not even all passively managed funds (such as index funds) in the same style box (or asset class) are created equal, as they can provide very different degrees of exposure to the factors that explain the variation in returns.
The researchers also examined whether active fund managers who seek to stay within a given Morningstar Style Box take active factor exposures, and the degree to which the factor exposures they take are dispersed. They found that while active fund managers do take active size and style exposures within the style box into which they are categorised, the dispersion of those exposures is also high — although not as high as the dispersion found among the stocks in the benchmarks themselves. The researchers also found that while the managers took factor-tilted exposures toward size and style, they did not on average tilt on either quality or momentum. A slight average tilt in favor of low-volatility stocks was observed.
Factor drift
In order to test the extent to which equity factors drift, the researchers formed single-factor portfolios and tracked their style exposures over time. The portfolios were formed within the Russell 1000 (Large Cap) universe, with weights initially set using a tracking error optimisation methodology that allowed maximum tracking against the benchmark of 1.5 percent. Limits were also placed on individual security positions (between 0.5 percent to 2.0 percent of the portfolio), sector exposures (+/- 2.0 percent) and beta (0.98 to 1.02). Portfolios were rebalanced back to those initial weights every month. Performance was measured over the period June 2003-December 2020.
The authors then tested the degree to which the factor scores of these single-factor portfolios drifted within the defined bands set by the Morningstar Style Box. Results were mixed. For example, while the value factor portfolio consistently stayed within the Value style box, it fluctuated strongly within the Value box. The momentum factor portfolio exhibited the greatest drift and was prone to drift outside of its original style box location. The quality factor portfolio exhibited the least factor drift, while the size factor portfolio tended to drift between the Large Cap and Midcap style boxes.
Benefits of multi-factor portfolios
The researchers also followed the same process of tracking style exposures over time, but this time they created multi-factor portfolios. These multi-factor portfolios were created to:
Start within a given Morningstar Large-Cap Style Box.
Target an overall ex-ante active risk level of 1.5 percent relative to the Russell 1000 benchmark.
Target a high level of active exposure for each factor.
Have similar weight bands and sector tolerances that were used in the single-factor portfolios.
They noted two key results from their tests on the multi-factor portfolios. First, relative to their categorisation in specific Morningstar Style Boxes, the multi-factor portfolios tended to drift out of their style boxes less frequently than did their single-factor portfolio counterparts. Second, the information ratios of the investment performance of the multi-factor portfolios were higher than that of the single-factor portfolios, establishing the diversification benefit of the pursuit of multiple factors at the same time. In addition, they found that the multi-factor portfolios tended to have lower investment risk, higher returns and less factor drift than the style box the researchers originally examined in their review of active fund managers factor drift.
Conclusion
The proliferation of equity market factors has intensified the portfolio construction process for investment practitioners. And tools such as the Morningstar Style Boxes are believed to help de-escalate that intensity. However, equity market factors that are frequently categorised into qualitative groupings (such as Value, Growth, Large, Small) within the Morningstar Style Boxes can be quite disperse within these groups. Over time the factor exposures inherent in stocks (and by extension active fund managers) also tend to drift across these categorisations, which can greatly frustrate portfolio diversification objectives. The construction of portfolios with multiple factor exposures that risk-constrain the factors with a higher propensity to drift can achieve both diversification benefits over single-factor portfolios and higher risk-adjusted returns than a majority of the active managers who operate in a similar space.
Investors can best address the issue of factor drift raised by BlackRock by selecting funds that not only systematically invest using multiple factors but reconstruct their eligible universes on a more frequent (typically monthly) basis rather than the annual reconstruction methodology used by most index products. Indices that reconstruct annually, such as the Russell and RAFI Fundamental Indices, can experience significant style drift in between reconstruction dates. For example, from 1990 through 2006 the percentage of stocks in the Russell 2000 in June that would leave the index when it reconstituted at the end of the month was 20 percent. For the Russell 2000 Value Index, the figure was 28 percent. The result is that a small-cap index fund based on the Russell 2000 would have seen its exposure to the small-cap risk factor drift lower over the course of the year. For small value funds based on the Russell 2000 Value Index, their exposure to both the small and value premiums would have drifted lower. The drift toward lower exposure to the risk factors results in lower expected returns. To avoid this problem, fund families such as AQR, Bridgeway and Dimensional reconstitute their asset class definitions monthly.