Dual Momentum: Does it still work? Recent results are not encouraging

Momentum stock trading strategies are premised on the historical observation that an asset’s performance tends to persist into the future. Momentum has been studied extensively by both practitioners and academics alike over the past years and the general consensus is that it is both pervasive across asset classes and time periods. Fama/French (2008) also famously stated that “Momentum is “the center stage anomaly of recent years…an anomaly that is above suspicion…the premier market anomaly.”

Dual Momentum

Dual momentum is a trading strategy that was originally expounded by Gary Antonacci in his original book. Dual momentum combines both relative momentum (cross-sectional comparison against other assets’ performance) and absolute momentum (time series comparison against the asset’s own historical performance.

The investing rules for his Global Equities Momentum (GEM) strategy:

1. Trades only 3 assets:
   • S&P 500 (IVV)
   • MSCI ACWI ex-US (VEU)
   • Bonds (BND)
2. Portfolio is rebalanced monthly to one of the 3 asset classes
3. To determine which asset to invest every month:
   • First, compare the past 12 month performance of IVV against BND (absolute momentum component)
     • If BND > IVV, invest into BND
   • Second, if IVV is higher than BND, compare the past 12 month performance of IVV against VEU (relative momentum component)
     • If IVV > VEU, invest into IVV
     • If VEU > IVV, invest into VEU

The backtested results from Antonacci have been quite impressive. GEM substantially improves annual returns while reducing both drawdowns as well as standard deviation.

Recent results based on my own testing:

As the results from the author’s back test was so promising, I started doing more research into his methodology. However, its recent performance has been terrible. From 2014 till now, the dual momentum GEM strategy has vastly underperformed a simple buy and hold of the S&P500 index.

The results below are based on my own calculations using data from Tiingo and it broadly matches the independent work done by TrendXplorer.

Modified Version

I started changing the lookback period of the original model to see whether I could squeeze out better performance.  I found that I could get the best performance using a lookback period of 2 months

However, if you exclude the 2008 crash, dual momentum with a lookback period of 2 month still underperforms a simple buy and hold strategy by around 100 percentage points. Does it make sense to give up so much upside just to protect yourself against a crash? I don’t think so.

Conclusion

After taking a close look at dual momentum, especially in light of the recent year’s performance, I am not fully convinced on the effectiveness of dual momentum. While absolute momentum does indeed seem to be effective in reducing large drawdowns, the relative momentum component does not do much in allowing us to capture momentum based alpha. In fact, as shown earlier, it actually reduces the performance in recent years.

Most of the research advocating for dual momentum bases its historical performance over close to 40 years. However, while people might find comfort in it working over such a long period, I believe its also important to measure its performance over shorter and more recent periods. This ensures that you can test it over different market regimes. Based on the evidence I have seen, I don’t think dual momentum is all people make it out to be.