Cross-sectional momentum
Momentum compares assets against each other. Over horizons of roughly 3 to 12 months, assets that outperformed their peers have tended to keep outperforming for a while.
Drag across the chart to read values
The textbook construction
- Take a universe, for example the 500 largest stocks.
- Rank each by its total return over the past 12 months, skipping the most recent month (the "12−1" signal). The skip avoids the short-term reversal effect.
- Hold the top decile, optionally short the bottom decile.
- Rebalance monthly.
The effect was documented by Jegadeesh and Titman (1993) and has since been found in international equities, bonds, currencies and commodities.
Momentum crashes
The strategy suffers sharp losses when markets rebound violently after a bear market, because the short side is full of beaten-down stocks that rally hardest. March to May 2009 is the standard example. Daniel and Moskowitz (2016) show that scaling exposure down when recent volatility is high reduces this risk.
Dual momentum
Gary Antonacci's dual momentum combines relative momentum (which asset did best) with absolute momentum (did it beat cash). If even the best asset underperformed treasury bills over the past year, the rule moves to bonds or cash.
Practical costs
Monthly turnover in a decile portfolio is high, so trading costs and taxes eat a meaningful share of the paper return. Many implementations rebalance quarterly or use buffers to reduce turnover.
Practice
Three quick tasks. Pick an answer to check it, or open the answer below each task.
Why does the 12-1 signal skip the most recent month?
Show answer
To avoid the short-term reversal effect. One-month returns tend to reverse, which would contaminate the signal.
When do momentum strategies tend to crash?
Show answer
In sharp rebounds after bear markets. Beaten-down losers on the short side rally hardest, as in spring 2009.
Dual momentum moves to bonds when...
Show answer
Even the best asset underperformed cash. Absolute momentum fails, so the rule steps aside.
Your score: 0 / 0