Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency
Source: Jegadeesh & Titman (1993) · The Journal of Finance · DOI: 10.1111/j.1540-6261.1993.tb04702.x
TL;DR
Stocks that outperformed over the past 3–12 months continue to outperform over the next 3–12 months. A zero-cost portfolio that buys past winners and shorts past losers earned roughly 1% per month (~12% annualized, t ≈ 3) over 1965–1989 — returns that are not explained by market risk or size. This is the founding paper of the cross-sectional momentum factor.
What anomaly it documents
The paper documents relative-strength momentum: the cross-sectional persistence of returns over intermediate horizons. Rank all stocks by their trailing return; the top performers ("winners") continue to beat the bottom performers ("losers") for the next several months.
Predictor: trailing cumulative return over a formation window of J months (J = 3, 6, 9, 12).
Direction: positive — high past return predicts high future return.
Horizon: the effect is strongest at 3–12 month holding periods. Critically, it reverses beyond ~12–24 months, consistent with the long-term overreaction reversal of De Bondt & Thaler (1985). Momentum is therefore an intermediate-horizon phenomenon sandwiched between short-term (1-month) reversal and long-term reversal.
Why it might exist: the authors are careful not to over-claim, but the pattern is most consistent with underreaction — prices adjust too slowly to firm-specific news. It is not explained by exposure to common risk factors or by serial correlation in systematic risk.
In the OSAP / Chen-Zimmermann taxonomy this corresponds to the Mom6m predictor (6-month formation), classified as 1_clear — among the most robustly replicable anomalies in the database.
How to construct it
The canonical academic construction (the paper's base case plus the now-standard "skip-a-month" refinement):
Sorting variable: cumulative raw return over the formation window. The paper's headline strategy uses J = 6 months. The widely-used academic "UMD" version measures the return from month t−12 to t−2, skipping the most recent month to avoid contamination by 1-month bid-ask-bounce reversal. (Jegadeesh & Titman's base case did not skip; the skip became standard later.)
Universe: NYSE and AMEX common stocks (the original sample; later work adds NASDAQ). Exclude very low-priced stocks (e.g., price < $5) and the smallest micro-caps in practice.
Portfolio formation: at the end of each month, rank stocks into deciles by formation-window return.
Long leg / short leg: long the top decile (winners), short the bottom decile (losers). Zero-cost, dollar-neutral.
Holding period: K months (K = 3, 6, 9, 12), using overlapping portfolios — in any month you hold K cohorts formed in the prior K months and average across them. This is the key implementation detail that smooths the return series.
Weighting: equal-weighted within each leg.
Rebalancing: monthly (1/K of the book turns over each month).
Evidence and replication
Period
Sharpe (approx)
Ann. Return
T-stat
Source
IS (1965–1989), 6×6 strategy
~0.7
~12% (≈0.95%/mo)
~3.07
this paper
IS best case (12-month formation, 3-month hold)
~0.8
~15.7% (≈1.31%/mo)
~3+
this paper
OOS (post-1993)
positive but lower
mid-single-digit to ~8%
—
post-publication studies
OSAP replication (Mom6m)
clear, positive
—
—
Chen & Zimmermann 2022
Key points on robustness:
The original 6-month/6-month winner-minus-loser portfolio earned ≈0.95% per month (t ≈ 3.07), and the strongest 12×3 variant ≈1.31% per month.
The profits are not subsumed by the CAPM or by size — winner and loser betas are similar, so the spread is not beta compensation.
Momentum is one of the more resilient anomalies out of sample. McLean & Pontiff (2016) estimate an average ~58% post-publication decline across anomalies; momentum decays less than the typical factor but is not immune, and it suffers severe, infrequent crashes (see Limitations).
Momentum survives across asset classes and countries (Asness, Moskowitz & Pedersen 2013), strengthening the case that it is not a single-sample artifact.
Why it might work
The debate is genuinely unresolved and both camps have evidence:
Behavioral / underreaction (leading explanation): investors are slow to incorporate information — anchoring, conservatism, slow diffusion of news (Hong & Stein 1999), and disposition-effect selling pressure (Grinblatt & Han 2005). Prices drift toward fair value, producing continuation.
Behavioral / delayed overreaction: Daniel, Hirshleifer & Subrahmanyam (1998) model overconfidence and biased self-attribution generating momentum that later reverses — consistent with the long-horizon reversal Jegadeesh & Titman document.
Risk-based: harder to sustain. Some argue momentum loads on time-varying macro or growth-option risk, but no widely accepted risk story explains the magnitude, and the negative skew/crash profile is the wrong shape for a simple risk premium.
Limits to arbitrage: the short leg concentrates in small, illiquid, hard-to-borrow losers, so the anomaly persists because it is costly to fully arbitrage.
Limitations and risks
Turnover and transaction costs: monthly rebalancing of decile portfolios is turnover-heavy. A meaningful fraction of paper profits is eroded by trading costs, especially in the small-cap short leg. Net-of-cost capacity is the central practical constraint.
Momentum crashes: the strategy has large negative skew and occasional catastrophic drawdowns — most notably in 2009, when beaten-down losers rebounded violently (documented by Daniel & Moskowitz 2016, "Momentum Crashes"). These crashes cluster in post-bear-market rebounds with high market volatility.
Capacity: concentrated in smaller names; the edge shrinks as you scale into large, liquid stocks. Volatility-scaling and crash-management overlays are now standard to make it investable.
Crowding and decay: as a published, heavily-traded factor it is subject to the McLean-Pontiff arbitraging-away effect.
Data/implementation: results are sensitive to the skip-month convention, weighting scheme, and microcap inclusion — easy to overstate with equal-weighted micro-cap-heavy backtests.
Key references
Jegadeesh, N. & Titman, S. (1993) — Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency — Journal of Finance — DOI: 10.1111/j.1540-6261.1993.tb04702.x
De Bondt, W. & Thaler, R. (1985) — Does the Stock Market Overreact? — Journal of Finance — DOI: 10.1111/j.1540-6261.1985.tb05004.x
Hong, H. & Stein, J. (1999) — A Unified Theory of Underreaction, Momentum Trading and Overreaction in Asset Markets — Journal of Finance
Daniel, K., Hirshleifer, D. & Subrahmanyam, A. (1998) — Investor Psychology and Security Market Under- and Overreactions — Journal of Finance
Asness, C., Moskowitz, T. & Pedersen, L. (2013) — Value and Momentum Everywhere — Journal of Finance
Daniel, K. & Moskowitz, T. (2016) — Momentum Crashes — Journal of Financial Economics
Chen, A. & Zimmermann, T. (2022) — Open Source Cross-Sectional Asset Pricing — Critical Finance Review
