June 4, 2014

"Fact, Fiction, and Momentum Investing"

 The AQR posse (Asness, Frazzini, Israel, and Moskowitz) recently issued a working paper that disproves many often-repeated myths about momentum investing, particularly as it applies to individual stocks. The authors back up their reasoning with results from academic papers and publicly available data. Here are the myths they address: 

  •   The momentum anomaly is small and sporadic 
  • ·It works mostly on the short side
  •   It works well only among small stocks
  •   It does not survive trading costs
  •   It does not work for a taxable investor
  •   It is best used as a screen rather than as a regular investment factor
  •   Its returns may not persist
  •   It is too volatile to rely on
  •   Different measures of momentum may give different results
  •   There is no theory or reasonable explanation to support it
 Below is a quick summary of the authors' evidence-based counter arguments:

1)  There is overwhelming evidence from scores of studies showing that momentum returns are stable and robust.

2)  There is little difference in performance between the long and short sides of momentum based on factor model regressions. Based on average returns versus the market, the long side has contributed more to momentum profits.

3)  As for working only among small caps, this is true only if you replace the word "momentum" with the word "value". Two of the authors, Israel & Moskowitz, wrote an important paper last year called "The Role of Shorting, Firm Size, and Time on Market Anomalies" that showed this. That paper was the subject of my post, "Momentum...the Only Practical Anomaly?" Momentum works well across all size stocks.

4)  A study last year by three of these authors called "Trading Costs of Asset Pricing Anomalies" looked at large institutional trades across nineteen developed markets from 1998-2013. They found the trading costs of momentum to be low, despite a higher turnover than from other factors. [1]

5) Several studies show that even though momentum with individual stocks can have 5-10 times the annual turnover of value strategies, momentum actually has a similar tax burden. This is because momentum holds on to winners and sells losers, which avoids short-term gains in favor of long-term ones. Momentum also has a lower dividend exposure than value.

6)   The authors point to papers showing that momentum works better as a factor-based approach than as a screen-based one.

7)  As for momentum's returns disappearing, one can say the same of any anomaly. Abnormal momentum returns have survived though for the past 200+ years. Momentum has held up to considerable out-of-sample validation across time, geography, and asset type. The authors point out, "There is no evidence that momentum has weakened since it has become well-known and once many institutional investors embraced it and trading costs declined."

       8) Relative strength momentum is volatile, but the Sharpe ratio (which includes volatility) of momentum still comes up on top. The authors say, "Who are you calling small and sporadic?" (The authors ignore absolute momentum, which reduces expected volatility and drawdown.)

9)  The authors agree that different measures of momentum can give different results, but they point out that this is true of any strategy. They say that different measures of momentum giving good results are a sign of robustness and not a cause for concern.

10) Momentum can be explained by either risk based or behavioral factors. As long as risks, risk preferences, biases, and/or behaviors do not change, momentum profits should continue unabated as they have for the past 200+ years. (My forthcoming book shows how behavioral biases are part of our DNA and are unlikely to change.)

The authors point out that most of the above myths can be shattered by a quick visit to Kenneth French's data library website. It is refreshing to see the authors, brought up in the Chicago efficient markets tradition, take on the challenge of those who say momentum profits cannot persist (despite plenty of evidence to the contrary) because that would contradict the theory of efficient markets. The authors point out that rejecting data because of a theory (or a one-sided view of the world) can be dangerous. They point to Columbus, Galileo, and the Salem witch trials as examples. Bravo!

The only problem I have with their paper is that the authors, perhaps aware that risk-adjusted momentum profits from individual stocks have been uninspiring over the past thirty years, point out that momentum works best when combined with value because of the low or negative correlations between the two. These low correlations were determined using long/short portfolios of both value and momentum. Few actually invest that way. Long only investing is different. Also, we can see in the Israel & Moskowitz study that value holds up only among the smallest stocks that represent only 10% of total market capitalization, and these are impractical for most investors to hold.

[1] The authors used a small proprietary data set of long/short momentum portfolios. Lesmond et al. and Korajczyk and Sadka found that transaction costs could negate most of the profit from momentum using more focused stock portfolios.