As we have discussed before, one of the major problems for the first formal asset pricing model developed by financial economists, the capital asset pricing model (CAPM), was that it predicts a positive relation between risk and return. But empirical studies have found the actual relation to be flat, or even negative.
Over the past five decades, the most “defensive” stocks have furnished higher returns than the most “aggressive” stocks. In addition, defensive strategies (at least those based on volatility) have delivered significant Fama-French three-factor and four-factor alphas.
The superior performance of low-volatility stocks (as well as closely related low-beta stocks) was initially documented in the academic literature in the 1970s, before even the size and value premiums were “discovered.” The low-volatility anomaly has been found to exist in equity markets around the globe, not only for stocks, but for bonds. In other words, it has been pervasive.
One of the CAPM’s assumptions is that there are no constraints on either leverage or short-selling. In reality, however, many investors are constrained against employing leverage (by their charters) or have an aversion to its use. The same is true of short-selling, and the borrowing costs for some hard-to-borrow stocks can be high. Such limits to arbitrage prevent arbitrageurs from correcting the pricing mistake.
Another assumption made by the CAPM is that markets have no frictions, meaning there are neither transaction costs nor taxes. Of course, in the real world, there are costs. The evidence shows that the most mispriced stocks are the ones with the highest costs of shorting.
The explanation for the low-volatility anomaly, then, is that, faced with constraints and frictions, investors seeking to increase their returns elect to tilt their portfolios toward high-beta securities to garner more of the equity risk premium. This extra demand for high-beta securities, and reduced demand for low-beta securities, may explain the anomaly of a flat or even inverted relationship between risk and expected return relative to the CAPM’s predictions.
Some recent papers (Robert Novy-Marx’s 2016 study, “Understanding Defensive Equity,” and Eugene Fama and Kenneth French’s 2015 study, “Dissecting Anomalies with a Five-Factor Model”) argue that the low-volatility and low-beta anomalies are well-explained by asset pricing models that include the newer factors of profitability and investment (in addition to market beta, size and value).
For example, Fama and French write in their paper that when using their five-factor model, the “returns of low volatility stocks behave like those of firms that are profitable but conservative in terms of investment, whereas the returns of high volatility stocks behave like those of firms that are relatively unprofitable but nevertheless invest aggressively.”
They add that positive exposure to RMW (the profitability factor, or robust minus weak) and CMA (the investment factor, or conservative minus aggressive) also go a long way toward capturing the average returns of low-volatility stocks, whether volatility is measured by total returns or residuals from the Fama-French three-factor model.
David Blitz and Milan Vidojevic contribute to the literature on the low-volatility anomaly with their July 2016 paper, “The Profitability of Low Volatility,” which covers the period July 1963 through December 2015.
While they did not question the empirical results of the Fama and French and Novy-Marx papers, the authors do argue that direct evidence of a linear, positive relationship between market beta and returns, which is assumed in the aforementioned models, is still lacking.
They write: “We are unable to construct high-beta portfolios with high returns and low-beta portfolios with low returns by controlling for factors such as profitability, while it should be possible to do so if the low-beta anomaly is fully explained by such factors.”
They also found “more pronounced mispricing for volatility than for beta. This suggests that the low-volatility anomaly is stronger than the low-beta anomaly, and, given that the two are closely related, that the low-volatility anomaly is the dominant phenomenon.”
Blitz and Vidojevic noted that they did include momentum as one of the control factors in their analyses because it is widely recognized as an important driver of stock returns in the cross section of returns.
In addition, they found that the results held across all size groups. The authors concluded that “exposure to market beta in the cross-section is not rewarded with significantly higher returns, regardless of whether one controls for the additional factors proposed by Fama and French.”
They then go on to add: “These results imply that the relation between risk and return in the cross-section is flat instead of positive. We also find that the mispricing is even more pronounced for volatility than for beta.”
Finally, Blitz and Vidojevic end with the following caveat: “The results in this paper represent just one attempt at obtaining a positive risk-return relation by controlling for the factors that supposedly explain the low-risk anomaly. The fact that this attempt is unsuccessful does not rule out that portfolios constructed in a different manner do exhibit a clear positive risk-return relation consistent with the predictions of the Fama and French and Novy-Marx models. For instance, the market betas or factor exposures used in this paper might not be appropriate, and it is possible that a different methodology would lead to different conclusions. But as long as the data indicates that portfolios with higher risk do not generate higher returns, it is premature to conclude that the low-risk anomaly has been resolved.”
Before you draw any conclusions, however, let’s look at some of the other evidence.
Xi Li, Rodney Sullivan and Luis Garcia-Feijoo, authors of the 2014 study, “The Limits to Arbitrage and the Low-Volatility Anomaly,” found that the excess return associated with forming the long-low-volatility/short-high-volatility portfolios are basically present only in the first month following formation, and that they are largely subsumed by the high transaction costs associated with low-liquidity stocks (such as low-priced/high-volatility stocks).
They also found that the anomalous returns within value-weighted portfolios are largely eliminated when omitting low-priced (less than $5) stocks, and were not at all present within equal-weighted portfolios.
In fact, the average price of stocks in the highest-volatility quintile was just over $7, indicating that many, if not most, would be considered “penny stocks.” And finally, they found that the low-risk effect has been noticeably weaker since 1990—new regulations were passed in that year aimed at reducing any fraud associated with trading penny stocks.
(I would add that many of the high-beta stocks simply disappeared after the dot-com crash, and the number of such stocks on public U.S. exchanges has shrunk dramatically since then.) The authors concluded: “Our findings cast some doubt on the practical profitability of a low risk trading strategy.”
The 2016 study by Bradford Jordan and Timothy Riley, “The Long and Short of the Vol Anomaly,” which covered the period July 1991 through December 2012, was motivated by previous research that has shown that both high-volatility stocks and stocks with high short interest exhibit poor risk-adjusted future performance.
(While the “conventional wisdom” holds that high short interest is a bullish signal because it predicts future buying from short covering, the reality is that stocks with high short interest perform poorly, on average.) However, to date, no one had studied the two together.
The authors found that while, on average, stocks with high prior-period volatility underperformed those with low prior-period volatility, the comparison is misleading because, among high-volatility stocks, those with low short interest experience extraordinary positive returns. On the other hand, high-volatility stocks with high short interest experience equally extraordinary negative returns.
The bottom line is that high volatility on its own is not an indicator of poor future returns. In fact, the authors found that for the period July 1991 through December 2012, stocks with high volatility and low short interest would have outperformed the CRSP value-weighted index by 9 percentage points a year! They also found that a portfolio long on high volatility and high short interest stocks had a four-factor alpha of -9% a year.
When Volatility Outperforms
Another important finding was that high-volatility/low-short-interest stocks outperform the market in turbulent times, such as the dot-com crash and the recent financial crisis. During the dot-com bubble, an equally weighted high-volatility/low-short-interest portfolio posted an annualized compound return fully 3.5 percentage points greater than that of the CRSP value-weighted index. During the financial crisis, that same gap was 13.3 percentage points.
Importantly, Jordan and Riley’s findings have implications for long-only strategies, as buying high-volatility stocks with low short interest avoids the high costs of shorting strategies and limits to arbitrage.
While they did find that high-volatility stocks with low short interest are less liquid than the average stock—execution costs could prevent investors from fully realizing the returns—the authors found no significant difference in performance between stocks in the group with high or low liquidity.
And low-short-interest stocks are typically larger stocks with higher trading volumes. High-volatility/high-short-interest stocks have an average size of $559 million, compared with about $2 billion for high-volatility/low-short-interest stocks. Thus, long-only investors can use screens to eliminate stocks with low liquidity, or they can use patient-trading strategies to minimize turnover costs.
The authors reached the following conclusion: “Based on the evidence in this study, the current ‘low vol’ investing fad has little or no real foundation.”
There is yet another issue we need to cover. The research shows that low-volatility strategies have exposure to the term factor—a factor Blitz and Vidojevic did not consider.
The fact that low-volatility strategies have exposure to term risk (the duration factor) should not be a surprise. Generally speaking, low-volatility/low-beta stocks are more “bond like.” They are typically large stocks, the stocks of profitable and dividend-paying firms, and the stocks of firms with mediocre growth opportunities. In other words, they are stocks with the characteristics of safety as opposed to risk and opportunity. Thus, they show higher correlations with long-term bond returns.
Ronnie Shah, author of the 2011 study “Understanding Low Volatility Strategies: Minimum Variance,” found that over the period 1963 through June 2010, the low-beta strategy had exposure to term risk. The “loading factor” (degree of exposure) was a statistically significant .09 (with a t-statistic of 2.6).
As further evidence, Tzee-man Chow, Jason Hsu, Li-lan Kuo and Feifei Li, authors of the 2013 study “A Study of Low Volatility Portfolio Construction Methods,” found a correlation of 0.2 between the BAB (betting against beta) factor and the duration factor.
Given their positive exposure to term risk, low-volatility stocks have benefited from the cyclical bull market in bonds that we have been experiencing since 1982. That rally cannot be repeated now, with interest rates at historic lows.
And finally, we need to consider if low-volatility strategies have become overgrazed.
Have Low-Volatility Strategies Become Overgrazed?
As with so many well-known anomalies and factors, the problem of potential overgrazing exists. Findings regarding the premium, combined with the bear market caused by the financial crisis of 2008-2009, led to a dramatic increase in the popularity of low-volatility strategies. As one example, as of July 2016, the iShares Edge MSCI Minimum Volatility USA ETF (USMV) had assets exceeding $15 billion.
We’ll examine how cash flows can change expected returns to factors by comparing the valuation metrics of USMV to those of the iShares Russell 1000 ETF (IWB), which is a market-oriented fund, the iShares Russell 1000 Value ETF (IWD), and the DFA Large Cap Value Fund (DFLVX). (Full disclosure: My firm, Buckingham, recommends DFA funds in the construction of client portfolios.) The table below is based on Morningstar data as of July 15, 2016.
It’s clear that cash inflows have raised the valuations of defensive (low-volatility/low-beta) stocks, dramatically reducing their once-significant exposure to the value premium to zero or negative, lowering expected returns. Specifically, as low-volatility stocks have been bid up in price, low-volatility portfolios have lost their value characteristics, in turn reducing the forward-looking returns. In other words, while low volatility still may predict low volatility, it may no longer result in higher returns than high volatility.
While it may not yet be resolved whether the low-volatility and low-beta anomalies can be well explained by exposures to other well-known factors, the popularity of the strategy certainly has changed the valuation metrics of low-volatility stocks. At the very least, this should raise a flag of caution for investors who have been enticed by the historical data. Forewarned is forearmed.
Larry Swedroe is the director of research for The BAM Alliance, a community of more than 140 independent registered investment advisors throughout the country.