The Half-Life Of Smart Beta

October 23, 2013


The Fate Of These Anomalies
To the extent that these factors—size, style and momentum—can be explained by actual risks, they are likely to persist over time. Those that were actual mispricings will tend to correct, but knock-on effects of market structures might not. As we have seen, the size effect has disappeared since its discovery, and the value effect has diminished, to be found mostly in small-cap stocks that are shunned by institutions. Conversely, the momentum effect has persisted, which suggests that momentum investors are indeed taking on additional risks.

A hot-off-the-presses paper from McLean and Pontiff tested the strength of 72 anomalies before and after they were revealed in academic publications.22 They found that, on average, the risk-adjusted rewards associated with newly discovered cross-sectional characteristics decreased by 35 percent after they became public. They found evidence of higher investor interest (as measured by variance, turnover and dollar volume) in factor portfolios following publication of the factors, and even stronger effects in low idiosyncratic risk, or low-volatility, portfolios, and in portfolios that are easier to arbitrage. This would indicate that the market quickly eliminated mispricings, and that the remaining effects do indeed correspond to real risks.

A little-noticed paper from the Cass Business school authors Hwang and Lu suggests that the Fama-French factors, including momentum, might be the wrong ones altogether.23 They found that “the market portfolio, liquidity and co-skewness explain the stock returns as well as the famous Fama-French three factors with momentum.” This is a fascinating contention. Put together with Asness et al. findings that “Liquidity risk is positively related to value and negatively to momentum, and its importance increases over time, particularly following the liquidity crisis of 1998,”13 this research opens the door to a complete re-evaluation of the real risk drivers behind identified factors or anomalies. What if co-skewness, liquidity and the market portfolio really are all you need?

Co-skewness is the tendency for bad things to happen simultaneously. Skewness is an expression of outlier events—extreme data points that tug a distribution’s mean away from its median. When outlier events happen together, they are more painful than separate, isolated events. Intuitively, we understand why investors would want to avoid co-skewness, and would push down prices of securities that they thought were particularly vulnerable to untimely downdrafts. Co-skewness kills the benefits of diversification. The observation cited earlier that “winner stocks move down more with the market than past loser stocks” is an illustration of co-skewness. Combine this with the remnants of the size effect and value effect, which are found mostly in small, presumably neglected stocks, and you have a decent set of explanations for the origin of risks in the most common factor models.

Low Volatility
Not included in the three- or four-factor models is the low-volatility anomaly. Many index providers and investment strategists are interested in low-volatility strategies. Is low volatility a separate factor, and if so, what risks might underlie it? What might happen to the anomaly now that investors are paying attention to it?

Discovered in the 1970s by Black and Scholes, the low-volatility or low-beta anomaly is the least intuitive of the bunch.24 Contrary to all expectations, stocks with low variance in the range of their price movements, or with low sensitivity to the overall market, have historically generated higher risk-adjusted returns—in the form of Sharpe ratios—than those with high variance and high market sensitivity, in many market environments.

The low-volatility anomaly went largely unnoticed by investors until recent years, when investors developed a strong interest in defensive strategies, and issuers launched a raft of low-volatility indexes and investment products. Early evidence is flowing in. Bornholt, using portfolios formed with low- and high-beta industry returns, found a loss of statistical significance of the low-beta effect after 1993.25

In an efficient market, why would investors not demand greater returns from securities that give them heart attacks than from those that allow them to sleep at night? There are two schools of thought—lottery-seeking behavior; and mismatched incentives. Both look at agency costs.



Find your next ETF

Reset All