Profitability of Momentum Strategies: An Evaluation of Alternative Explanations
Source: Jegadeesh, N. & Titman, S. (2001) · Journal of Finance 56(2), 699–720 (NBER WP 7159) · DOI: 10.1111/0022-1082.00342
TL;DR
Revisits momentum eight years after Jegadeesh & Titman (1993), using 1990–1997 as a genuine out-of-sample window. Momentum profits persisted at nearly the original magnitude (~1%/mo), ruling out data snooping. Post-formation, momentum portfolio returns reverse, but only in years 4–5 (no significant reversal in years 2–3); the cumulative return in months 13–60 is negative. This pattern supports behavioral delayed-overreaction models and sharply rejects the Conrad–Kaul claim that momentum is merely cross-sectional dispersion in unconditional expected returns.
What anomaly it documents
Predictor: prior 3–12 month return (the paper centers on 6-month ranking, 6-month holding).
Direction: past winners continue to outperform past losers over the next 6 months.
Shape: short-horizon continuation followed by long-horizon (years 4–5) reversal — the mispricing-then-correction signature.
Each month rank stocks into deciles on past 6-month return (a $5 price screen is applied; results are sensitive to including low-priced stocks in January).
Long P10 (winners), short P1 (losers); equal-weight; hold 6 months with overlapping monthly cohorts.
For the reversal test, track cumulative portfolio returns over the 60 months following formation.
Evidence and replication
Period / horizon
Result
t-stat
Source
In-sample 1965–1989, P10−P1, 6×6
1.11%/mo
(orig. 3.07)
this paper, confirming JT 1993
Out-of-sample 1990–1997, P10−P1
1.01%/mo
—
this paper
Full 1965–1997, P10−P1
similar magnitude
4.61
this paper
Post-formation years 2–3
no significant reversal
—
this paper
Post-formation years 4–5 (months 13–60 cumulative)
significant reversal, negative
—
this paper
The near-identical OOS magnitude (1.01% vs 1.11%) is striking given that contemporaneous anomalies decayed: the FF size factor was 0.53%/mo (t=2.34) in 1965–1981 but −0.09% (t=−0.37) in 1982–1997, and the HML factor 0.51%/mo (t=2.61) in 1965–1989 fell to 0.19% (t=0.76) in 1990–1997.
January seasonality (winners beat losers in all months except January) also recurs out of sample.
Why it might work
Behavioral: investors underreact to information initially, then overreact, with the overreaction eventually reversing (Barberis-Shleifer-Vishny 1998, Daniel-Hirshleifer-Subrahmanyam 1998, Hong-Stein 1999). The years 4–5 reversal is the predicted correction.
Against risk: the negative long-run post-formation returns are hard to square with momentum being compensation for risk, and the analysis rejects Conrad-Kaul's unconditional-dispersion explanation (which predicts continued positive post-holding returns).
Limitations and risks
The behavioral support is "tempered with caution": reversal timing (years 4–5, not 2–3) and its strength vary across subperiods (notably weaker outside 1965–1981).
Momentum carries crash risk (Daniel-Moskowitz 2016) and high turnover/transaction costs; OOS here is only the 1990s, and later decades show further decay.
Results are sensitive to the $5 price screen and January treatment of low-priced stocks.
Key references
Jegadeesh, N. & Titman, S. (1993) — Returns to Buying Winners and Selling Losers — Journal of Finance
Conrad, J. & Kaul, G. (1998) — An Anatomy of Trading Strategies — Review of Financial Studies
Daniel, K. & Moskowitz, T. (2016) — Momentum Crashes — Journal of Financial Economics
Provenance: verified/generated from the paper's full text.
