From Michael Harris at Price Action Lab
http://www.priceactionlab.com/Blog/2016/02/explanation-momentum-anomaly/There is massive literature than documents the persistence of the time-series momentum anomaly in most major asset classes. I abduct in this article a simple explanation of this anomaly and I present some evidence that corroborates it.
Summary
The simplest explanation for the persistence of the momentum anomaly is that it has not been arbitraged out yet but it will be in the future
The main reason that the price-series momentum anomaly in equity markets has not been arbitraged out is due to lack of trading discipline
In commodity markets the time-series momentum anomaly has been arbitraged out in the last 10 years due to systematic trend-following programs
Algo trading and robo-advisor domination will cause the momentum anomaly to slowly disappear
Momentum refers to the empirically documented tendency of rising asset prices to rise further and vice versa for falling prices. Numerous academic and practitioner studies have documented this anomalous market tendency. Finance theory argues that once discovered, anomalies should disappear. This assumes that markets are efficient and obey certain equilibrium models, e.g., the CAPM. However, the time-series anomaly has persisted for decades, as the backtests below show. Therefore, either finance models are wrong or the anomaly persists due to some special conditions.
The backtests below are based on monthly S&P 500 data (dividends not included) in the period 01/1900 – 02/2016. No commission is included. The strategy involves buying the index when current close is higher than the close of 12 months ago. This is the simplest form of time-series momentum. Equity is fully invested.
The first backtest is for the period 01/1900 – 12/1937:
There is significant outperformance of buy and hold, as it may be seen from the notes in the above chart. Holding the index in the test period would have resulted in an annualized return of 2.9% and maximum drawdown of nearly 85%. On the other hand, the simple time-series momentum strategy with a 12-month lookback period shows 7.34% annualized return at 50% maximum drawdown. This significant result was noticed in the late 1930s. Of course, this analysis is hypothetical because no one could buy the index but instead a portfolio of securities that tracked the index closely and that involved transaction cost. However, due to the low turnover, the transaction cost would not impact the performance significantly.
Although the annualized return of the momentum strategy in the above period is 30 basis points less than that of buy and hold, maximum drawdown is much lower, resulting in significant outperformance on a risk-adjusted return basis. These results show that the time-series momentum effect has persisted. It can also be shown that the effect is persistent for all lookback periods between 2 and 12.
The goal of the above limited analysis was to review the existence of the time-series momentum anomaly in the US stock market. Much more advanced studies have shown the persistence of this anomaly across many different markets. The puzzling issue is why this anomaly exists in the first place. The abducted hypothesis in this article is that the anomaly exists in markets where there has been no systematic arbitrage in action. In other markets where such action has taken place, the anomaly has almost disappeared. An example is commodities markets due to the systematic programs used by CTAs. It was shown in another article that in the last 10 years, the performance of the CTA groups has deteriorated significantly, as shown in the chart below:
For more details see this article. In the last 7 years, the annualized return of the CTA group is flat versus about 14% annualized for the S&P 500 total return.
CTAs have used systematic strategies for many years in an effort to eliminate emotions from trading and enforce discipline. Because there cannot be a free lunch, the actions to secure proper execution of the strategies actually made them ineffective due to the consistent arbitrage. This is important to understand. Actually, some non-systematic traders may have been able to generate significant outperformance but only due to luck. However, no one should expect to profit from a crowded trade and CTAs have been victims of their own efforts to enforce discipline. Essentially, anyone with market experience and a computer could become a CTA, at least in principle, because the models followed were and are still nowadays in many cases trivial. This is also what is slowly occurring in the equities markets after the popularization of momentum by some authors who fail to recognize the reflexive nature of markets.
By trying to convince people that the momentum anomaly is robust and they should systematically exploit it, authors, practitioners and researchers actually contribute to a slow but effective arbitrage process. As robo-advisors and algo traders offer related strategies, the anomaly will disappear because this is natural consequence and unavoidable: There is no room for everyone to profit from the same strategies and some or even most will have to lose eventually.