Jegadeesh and Titman's (2001) "Profitability of Momentum Strategies: An Evaluation of Alternative Explanations" tests whether momentum profits reflect risk or behavioral overreaction. Using CRSP NYSE/AMEX data from 1965 to 1997, past winners outperformed past losers by 1.01% per month in the 1990–1997 out-of-sample period. That nearly matched the 1.11% per month spread from the original 1965–1989 sample. The cumulative momentum profit, however, rose to 11% twelve months after formation before reversing to just .79% by Month 60.
What the Study Found
Past winners outperformed past losers by 1.01% per month in the 1990 to 1997 out-of-sample period, compared with 1.11% per month from 1965 to 1989. Over the full 1965–1997 sample, the winner-minus-loser spread was 1.09% per month with a t-statistic of 5.65. Losers outperformed winners by 2.92% in January during 1965–1989 and by 1.21% in January during 1990–1997. Cumulative momentum profits rose to 11% by Month 12 but declined to just .79% by Month 60 over 1965–1997. Adjusting for cross-sectional dispersion in expected returns still left momentum profits at 1.28% per month, rejecting Conrad and Kaul's (1998) risk-based explanation.
Methodology
The study uses CRSP return data for all NYSE/AMEX common stocks, excluding NASDAQ-listed shares. Stocks priced below $5 at the start of the holding period are excluded from the sample. The sample spans January 1965 to December 1997. It combines the original Jegadeesh and Titman (1993) formation period of 1965–1989 with an out-of-sample test period of 1990–1997. Stocks are ranked into ten decile portfolios based on six-month lagged returns and held for six months using overlapping portfolio formation.
Key Statistics
| Metric | Finding | Context |
|---|---|---|
| Momentum profit (P1-P10), full sample | 1.09% per month (t = 5.65) | Test period: 1965–1997 |
| Momentum profit (P1-P10), out-of-sample | 1.01% per month (t = 4.10) | Test period: 1990–1997 |
| Cumulative profit, Month 12 vs Month 60 | 11% → .79% | Test period: 1965–1997, post-formation months 1–60 |
| Fama-French three-factor alpha (P1-P10) | 1.29% (t = 7.78) | Test period: January 1965–December 1997 |
| WRSS profit, simulation without replacement | 0.00002 (.53% of actual profit) | Test period: 1965–1997, bootstrap without replacement |
| WRSS weight formula | w_it = (1/N)(r_{i,t-1} − r̄_{t-1}) | Defines stock weight in the weighted relative strength strategy |
Why This Matters
The long-horizon reversal pattern is more consistent with delayed overreaction than with a purely risk-based explanation of momentum profits. That complicates the case for treating winner-minus-loser returns as a permanently exploitable trading signal. The findings also weaken arguments that momentum returns are simply compensation for size and value risk exposure. Because evidence of reversal was weaker in the more recent subperiod studied, the authors urge caution before treating the pattern as a stable, exploitable signal.