The Momentum Premium: What the Data Actually Shows
Momentum is one of the most studied anomalies in finance. The evidence is strong, but so are the failure modes if you implement it naively.
Key Takeaways
01Momentum, or buying recent winners and selling recent losers, has generated a documented premium of roughly 8.3% annualized in U.S. equities since 1927.
02The factor's largest weakness is crash risk during violent reversals, with the most notable example arriving during the 2009 post-crisis rebound.
03Practical implementations try to contain that risk with regime filters, sector constraints, turnover control, and liquidity screens rather than running the raw academic factor.
04Momentum tends to work best as one part of a broader portfolio process alongside value, quality, and risk controls.
In 1993, Narasimhan Jegadeesh and Sheridan Titman published a paper in the Journal of Finance that documented something the efficient market hypothesis struggled to explain. Stocks that had performed well over the prior 3 to 12 months tended to keep outperforming, while recent laggards tended to keep lagging.[1] That result has been examined from every angle since, and it still matters because the signal has remained unusually persistent across time, markets, and asset classes.
Retail investors often hear momentum reduced to a bumper sticker: buy what is going up. That version is incomplete to the point of being dangerous. The real evidence trail is richer, the implementation details matter, and the failure modes are very real.
What Momentum Actually Measures
In quantitative finance, momentum usually means cross-sectional momentum: ranking securities against one another by trailing total return, often over the previous 12 months while skipping the most recent month to avoid short-term reversal noise. It is not the same thing as trend-following, which evaluates whether an asset's own return has stayed above or below zero.
Using the Ken French Data Library, the Fama-French UMD factor delivered about 8.3% annualized in U.S. equities from 1927 through 2023.[2] That number is not a marketing claim. It comes from a public dataset that researchers can inspect and reproduce.
Momentum Factor Performance by Decade, U.S. Equities
Period
Avg. Annual Return
Std. Dev.
Sharpe Ratio
1927–1939
+10.2%
25.1%
0.41
1940–1959
+9.8%
14.3%
0.69
1960–1979
+7.1%
13.8%
0.51
1980–1999
+11.4%
15.2%
0.75
2000–2019
+4.1%
21.7%
0.19
2020–2023
+6.8%
18.4%
0.37
Full Period
+8.3%
17.2%
0.48
Source: Ken French Data Library UMD monthly factor returns, annualized across the full public sample.
Asness, Moskowitz, and Pedersen later showed that momentum appears not just in U.S. stocks, but across international equities, bonds, currencies, and commodity futures.[3] That breadth is one reason the signal remains central to quantitative research.
Why It Persists
The strongest explanation is behavioral. Investors underreact to gradual information, overreact to dramatic stories, and often anchor to stale views long after the evidence has moved. De Bondt and Thaler documented that markets do not absorb information perfectly, especially when narratives evolve over time.[4]
Momentum also feeds on the disposition effect and herding. Investors frequently sell winners too early, hold losers too long, and crowd into established trends once those trends become legible. Those are not temporary software bugs in markets. They are recurring human behaviors.
The Crash Problem
Daniel and Moskowitz showed that momentum can suffer severe drawdowns during sharp reversals, especially when beaten-down losers suddenly rebound and prior winners stall.[5] The most cited example is the first quarter of 2009, when the factor collapsed during the post-crisis snapback.
Momentum's Worst Documented Drawdowns
Period
Peak-to-Trough
Recovery
Trigger
Mar–May 2009
–40.2%
~3 years
Post-GFC reversal
Jul–Aug 1932
–33.7%
~2 years
Post-Depression reversal
Jan–Feb 2001
–25.1%
~14 months
Dot-com rotation
Nov 2020
–15.8%
~6 months
Vaccine rotation
Source: Ken French UMD data with crash analysis conventions aligned to Daniel and Moskowitz (2016).
From Factor to Portfolio Process
The raw academic factor is a starting point, not a finished strategy. Novy-Marx showed that intermediate-horizon momentum captures much of the signal with lower turnover.[6] Professional implementations then add position sizing, breadth checks, sector diversification, and market-regime awareness.
What It Means in Practice
The practical takeaway is not that investors should blindly chase strength. It is that trend persistence is real, measurable, and hard to dismiss. If a portfolio process ignores one of the most persistent cross-sectional effects in finance, that omission deserves explanation.
Momentum is not a secret. It is a well-documented tendency that survives because markets are made of people, and people rarely update as cleanly as models assume. The edge is not in knowing that the effect exists. The edge is in implementing it with discipline when the signal becomes uncomfortable.