Financial research has uncovered many relationships between investment factors and security returns. Given that popularity is a curse in investing, the growing popularity of factor investing has led to worries that factors have become overvalued, posing risks to investors in these strategies.
For investors, an important question is whether the past relationship between factors and returns will continue after the research has been published and factor investing becomes popular. Said another way, if everyone knows about it, should we expect the premium to continue outside of the sample period?
In the introduction of our new book, “Your Complete Guide to Factor-Based Investing: The Way Smart Money Invests Today,” my coauthor Andrew Berkin and I provide five criteria that a factor should be required to meet before the premium can be expected to continue: The factor should be persistent, pervasive, robust, investable and have logical, risk-based and/or behavioral-based explanations.
Does Popularity Destroy Persistence?
Chapters 1 through 7 of the book provide evidence and explanations for why we believe the premiums for each factor addressed (beta, size, value, momentum, profitability/quality, term and carry) should be expected to continue.
However, this conclusion says nothing about the size of the premiums, provoking the question: Does the publication of research impact the future size of premiums? The question is important on two fronts.
First, if anomalies are the result of behavioral errors—or even investor preferences—and publication draws the attention of sophisticated investors, it is possible that post-publication arbitrage would cause the premiums to disappear. Investors seeking to capture the identified premiums could quickly move prices in a manner that reduces the return spread between assets with high and low factor exposure.
However, as we explain in the book, limits to arbitrage (such as aversion to shorting and its high cost) can prevent arbitrageurs from correcting pricing mistakes. And the research shows this tends to be the case when mispricing exists in less liquid stocks, where trading costs are high.
Second, even if the premium is fully explained by economic risks, as more cash flows into funds acting to capture the premium, the size of the premium will be affected. At first, publication will trigger inflows of capital, which drives prices higher and thus generates higher returns. However, these higher returns are temporary because subsequent future returns will be lower.
With these questions in mind, we’ll turn to the academic literature. Paul Calluzzo, Fabio Moneta and Selim Topaloglu contribute to our understanding of how markets work and become more efficient over time (the adaptive markets hypothesis) with their 2017 study “When Anomalies Are Publicized Broadly, Do Institutions Trade Accordingly?”
They hypothesized: “Institutions can act as arbitrageurs and correct anomaly mispricing, but they need to know about the anomaly and have the incentives to act on the information to fulfill this role.”
How Institutions Affect Anomalies
To test their hypothesis, Calluzzo and colleagues studied the trading behavior of institutional investors in 14 well-documented anomalies: net stock issues, composite equity issues, accruals, net operating assets, asset growth, investment-to-assets, distress, momentum, gross profitability, return on assets, book-to-market, Ohlson O-score, post-earnings announcement drift and capital investment.
To determine if investors exploit these anomalies and help bring stock prices closer to efficient levels, Calluzzo and colleagues built portfolios that consisted of long stocks with positive expected returns, and short stocks with negative expected returns. Their study covered the period January 1982 through June 2014.
Following is a summary of their findings:
- Trading with the anomaly was profitable during the original in-sample period. The alpha of the equally weighted (across each of the anomalies) portfolio was 1.54% per quarter.
- Raw returns in the period after publication decay to an average of 1.05%—a 32% relative reduction. Using the Fama–French three-factor model, there were reductions in nine of the 14 anomalies.
- During the in-sample, prior-to-publication period, institutional investors did not take advantage of stock return anomalies.
- In the post-publication period, institutions traded to exploit the anomalies.
- There is a significant negative relationship between institutional trading and future anomaly returns in the ex-post portfolio. Institutional trading after anomaly publication is related to the post-publication decay in anomaly returns.
Calluzzo and colleagues concluded: “Institutional trading and anomaly publication are integral to the arbitrage process which helps bring prices to a more efficient level.” Their findings demonstrate the important role that both academic research and hedge funds (through their role as arbitrageurs) play in making markets more efficient.
These findings are consistent with those of R. David McLean and Jeffrey Pontiff, authors of the study “Does Academic Research Destroy Stock Return Predictability?” McLean and Pontiff re-examined 97 factors published in tier-one academic journals and were able to replicate the reported results for 85 of them. That the remaining 12 factors were no longer significant may be due to a variety of reasons, such as incomplete details in the original paper or changes in databases.
They also found that, following publication, the average factor’s return decays by about 32% (note the agreement of that figure with the one in the Calluzzo, Moneta and Topaloglu paper), and returns do not decay to zero but remain positive.
In addition, McLean and Pontiff found that factor-based portfolios containing stocks that are more costly to arbitrage decline less post-publication. This is consistent with the idea that costs limit arbitrage and protect mispricing.
We can draw two conclusions from the research. First, anomalies can persist even when they become well-known. As McLean and Pontiff remark: “We can reject the hypothesis that return predictability disappears entirely, and we can also reject the hypothesis that post-publication return predictability does not change.”
Second, research appears to lead to increased cash flows from investors seeking to gain exposure to the premiums, which in turn lead to lower future realized returns. However, where logical, risk-based explanations exist, premiums should never disappear.
For example, no one expects the market-beta premium to disappear even though it has been well-known for decades. However, investors should not automatically assume that future premiums will be as large as the historical record.
Heiko Jacobs and Sebastian Muller provide us with evidence from international markets with their 2016 paper, “Anomalies Across the Globe: Once Public, No Longer Existent?” Their study covered the pre- and post-publication return predictability of 138 anomalies in 39 stock markets that account, on average, for almost 60% of global equity market capitalization and more than 70% of global gross domestic product. The data covered the period January 1981 to December 2013.
While their findings for the United States were similar to those of McLean and Pontiff, showing declining premiums, none of the 38 international markets yielded a significant post-publication decline in anomaly returns.
In fact, Jacobs and Muller found that returns to anomalies in international markets actually increased. Equally weighted (value-weighted) monthly returns increased from 34 (28) basis points between 1981 and 1990 to 56 (40) basis points between 2001 and 2013.
Following is a summary of their findings:
- Averaged over the sample period, long/short anomaly returns in various subsets of international markets turn out to be similar in magnitude to the estimates for the U.S. market.
- Many anomalies tend to be a global phenomenon and thus are unlikely to be driven mainly by data mining.
- In almost every country, equally weighted portfolios generate larger returns than value-weighted portfolios. This is consistent with the notion that both mispricing and limits to arbitrage tend to be stronger for smaller stocks.
- For the majority of countries, pooled long/short returns are positive and statistically significant at the 1% level.
Jacobs and Muller concluded that, while their findings point to a strong negative time trend and increased post-publication arbitrage trading in the United States, they did not find reliable evidence for an arbitrage-driven decrease in anomaly profitability in international markets. In addition, the authors found that differences in standard arbitrage costs “seem to explain at best a fraction of the large differences in post-publication.”
They did add that they explored “several possible mechanisms behind the surprisingly large differences between the return dynamics in the U.S. and international markets” but were “unable to fully explain the results” and their findings were “consistent with the idea that sophisticated investors learn about mispricing from academic studies, but then focus mainly on the U.S. market.” They conclude with the following: “Our results may thus be interpreted as a puzzle that calls for further theoretical and empirical investigation.”
Are Factors Expensive?
Clifford Asness, Swati Chandra, Antti Ilmanen and Ronen Israel contribute to the literature on factor investing with their March 2017 study “Contrarian Factor Timing is Deceptively Difficult.”
The paper will appear in a forthcoming edition of the Journal of Portfolio Management. It addresses two issues: Just how rich, if at all, have style premia factors (often referred to as smart beta) become? And should investors hold off on investing in rich factors until they cheapen?
Following is a summary of their findings:
- While value spreads for some well-known styles are expensive relative to history, as a group they are not, and none are at bubblelike (or the opposite) levels.
- When comparing the impact of value timing (can dynamic allocations improve the performance of a diversified multistyle portfolio?), they found “lackluster results—strategic diversification turns out to be a tough benchmark to beat.”
- Despite the proliferation of factor and “smart beta” investing, the spreads remain historically reasonable and exhibit a pattern of frequent mean reversion, not steady richening, in response to growing investor demand.
Value Timing Of Factors
Asness, Chandra, Ilmanen and Israel noted that value timing of factors, because it is buying what is relatively cheap, is correlated to the standard value factor as it adds further value exposure to a portfolio.
They explain: “If a multi-style portfolio already includes value at optimally diversified levels, value timing the styles may increase value exposure to levels that undermine diversification, leading to weaker performance, particularly in a risk-adjusted sense. For many investors, the original intention of a multi-style allocation is to balance risk across multiple sources of return and capitalize on the power of diversification. Value timing a multi-style allocation may work against that very purpose by effectively increasing the allocation to value.”
They add: “Portfolio math tells us that returns add linearly while risk adds quadratically. Hence, at larger tilts, the increase in risk from timing may be proportionately larger than any increase in return, resulting in lower risk-adjusted returns.”
The authors also examined whether timing would add value if it were done only at extreme levels—when spreads passed a certain threshold. They found: “The timed strategy Sharpe ratio improves as we increase the threshold, but it is a very modest improvement. The timed strategy Sharpe ratio barely exceeds the Sharpe ratio of the non-timed. In fact, increasing the threshold further leads to a slight drop in the Sharpe ratio.”
These findings add further support to investors building portfolios that are strategically (as opposed to tactically) diversified across each of the factors that show persistence of their premiums; low correlation to other factors; are pervasive around the globe and across asset classes; have intuitive reasons to believe the premiums should persist (whether behavioral- or risk-based); and are implementable (survive transactions costs).
The Importance Of Factor Diversification
Chapter 9, Implementing a Diversified Factor Portfolio, in “Your Complete Guide to Factor-Based Investing,” demonstrates that a portfolio that is diversified across factors has been more efficient than any of the individual factors, producing dramatically higher Sharpe ratios. The following table from the book covers the period 1927 to 2015.
For each factor, it shows the mean premium, volatility and the Sharpe ratio. It also provides the same information for three portfolios. Portfolio 1 (P1) is allocated 25% to each of four factors: market beta, size, value and momentum. Portfolio 2 (P2) is allocated 20% to each of the same four factors and 20% to the profitability factor. Portfolio 3 (P3) is allocated the same way as P2, substituting the quality factor for the profitability factor.
You can also see the benefits of diversifying across factors in the following table from the book, which shows the odds of underperformance over various time horizons. It too covers the period 1927 to 2015.
As you can see, no matter the time horizon, the odds of underperformance are lower for each of the three portfolios than for any of the individual factors.
The research makes clear that, while publication and ensuing popularity has, on average, reduced the size of premiums by about one-third, the premiums can disappear entirely for factors that are pure anomalies (behavioral mispricings or perhaps the result of data mining activities).
However, for the factors that have passed the tests of persistence, pervasiveness, robustness, intuitiveness and implementability, the premiums don’t disappear, and in some cases, may not have even shrunk.
Given that there aren’t any crystal balls that can tell which factors will deliver premiums in the future, the prudent strategy is to strategically diversify across factors that meet your criteria for investment and then stay the course, rebalancing but avoiding tactical adjustments based on value spreads.
Larry Swedroe is the director of research for The BAM Alliance, a community of more than 140 independent registered investment advisors throughout the country.