Smart-beta indices are constructed to exploit “anomalies” that reward exposure to risk factors beyond what would be expected as “necessary compensation” under the Capital Asset Pricing Model (“CAPM”). Of course, any factor that results in nominal outperformance must be considered on a risk-adjusted basis, since taking on higher risk should engender a greater reward – and investment researchers at S&P Dow Jones Indices think at least some factor “anomalies” aren’t anomalies at all, but just rewards for greater-than-understood risk-taking.
Even still, among the remaining anomalies, the researchers think many are “disappearing,” “statistical,” or “attenuated” – and only a few are truly “persistent.” Writing on behalf of S&P Dow Jones, academic Hamish Preston and S&P Dow Jones Index Investment Strategy professionals Tim Edwards and Craig Lazzara express these views in an October 2015 research paper titled “The Persistence of Smart Beta.”
Disappearing anomalies don’t last. A great example shared by the paper’s authors is the so-called “Weekend Effect” that was popularized by Frank Cross in 1973. Mr. Cross discovered that if investors had bought stocks at their closing prices each Monday and sold them at their closing prices each Friday – avoiding the weekend and the Monday trading session – they would have dramatically outperformed a “buy and hold” strategy from 1950 to the time of his research.
But then, almost immediately after the Weekend Effect became well known, the anomaly didn’t just disappear, it reversed. The Weekend Effect rebounded in 1984, only after another academic research paper called it into question – and then when a paper called “The Reverse Weekend Effect” was published in 2000, the old Weekend Effect returned.
As soon as investors gained knowledge of the Weekend Effect, it reversed. When knowledge of the reversal became widespread, the reversal reversed. Now it’s taken as a given that the Weekend Effect was a coincidence – hence, it was a disappearing anomaly.
Perhaps a better approach is for investors to keep knowledge of anomalies they discover secret – that way they may be less likely to disappear. This is what David Dolos did when he discovered that applying the price movements of the 1720 South Sea Bubble – second only to Tulip Mania in episodes of old-school irrational exuberance – to the Dow Jones Industrial Average inexplicably produced outsized returns. Mr. Dolos never told anyone about his discovery, and he reaped the rewards in anonymity until 2007, when the system broke down.
Why? Well first off, David Dolos didn’t exist. The story is made up, and although the 1720 South Sea Bubble was real, the South Sea Bubble effect was data-mined into existence. As the paper’s authors note, modern computing power can easily produce “false positives” – i.e., anomalies that are purely statistical in nature. In order for an anomaly to be persistent, it must make logical sense.
Momentum is one of the most popular factors. Academic research supports its outperformance, and the concept of momentum stocks – stocks that are going up – outperforming non-momentum stocks makes logical sense. The momentum anomaly is known to anyone who cares to know about it, and yet this knowledge hasn’t caused the anomaly to disappear – instead, it has reinforced it.
The downside is that since investors have become aware of the momentum anomaly, its drawdowns have been bigger. This is what the S&P Dow Jones authors mean by an “attenuated anomaly.”
In 1997, Mark Carhart published a study that showed adding momentum to the famous Fama-French three-factor model boosted returns. This caused more money to flow into momentum stocks, ultimately leading to bigger drawdowns during crashes.
Are there any truly persistent anomalies? The authors say there is at least one: Low volatility. But they conclude with a word of caution: “So far, the investment and attention directed toward low-volatility strategies has not been sufficient to temper their returns or attenuate their risk/return profile.” So far.
As the well-known disclaimer says: “Past performance does not necessarily predict future results.”
For more information, download a pdf copy of the white paper.
Jason Seagraves contributed to this article.