Market EfficiencyBehavioral Finance

Market Overreaction: Why Prior Losers Outperform Prior Winners

Summary by Robert Gorak · Published June 10, 2026 · Last reviewed June 10, 2026

Werner F. M. De Bondt and Richard Thaler·1985·Journal of Finance
Sample: between 347 and 1,089 NYSE stocks per replicationData: CRSP Monthly Return FilePeriod: January 1926 – December 1982

Market Overreaction: Why Prior Losers Outperform Prior Winners

Summary by Robert Gorak | Published 2026-06-10 | Last reviewed 2026-06-10


De Bondt and Thaler's (1985) "Does the Stock Market Overreact?" tests whether investors push prices too far past fundamental value. Their dataset is the CRSP Monthly Return File, covering 347–1,089 NYSE stocks between January 1926 and December 1982. Loser portfolios of 35 stocks outperformed the market by 19.6%, and winner portfolios underperformed by about 5.0%, over 36 months. The return spread [ACARL,36 − ACARW,36] = 24.6% (t-statistic: 2.20) across 16 non-overlapping three-year periods.

What the Study Found

Loser portfolios of 35 stocks earned 19.6% above the market over 36 months; winner portfolios earned about 5.0% below. The [ACARL,36 − ACARW,36] = 24.6% (t-statistic: 2.20) spans 16 non-overlapping 3-year test periods. The overreaction effect is asymmetric: losers drive most of the spread, with January excess returns of 8.1% (t=3.21) at month t=1. Loser January returns persist: 5.6% (t=3.07) at month 13 and 4.0% (t=2.76) at month 25. Twelve months into the test period, the return spread is only 5.4% (t=0.77).

Methodology

De Bondt and Thaler (1985) use the CRSP Monthly Return File for all NYSE common stocks from January 1926 to December 1982. Between 347 and 1,089 NYSE stocks participate per replication across 16 non-overlapping 3-year formation periods (January 1930 – December 1977). Winner and loser portfolios are the top and bottom 35 stocks by prior 36-month market-adjusted excess returns. Primary residuals are market-adjusted (u_{jt} = R_{jt} − R_{mt}); the study also tests market model and Sharpe-Lintner CAPM residuals.

Key Statistics

Metric Finding Context
ACARL,36 − ACARW,36 24.6% (t=2.20) 16 three-year periods, 35 stocks, test period end
Loser portfolio ACAR at 36 months 19.6% above market 16 three-year periods, 35 stocks
Winner portfolio ACAR at 36 months about 5.0% below market 16 three-year periods, 35 stocks
January excess return — month t=1 8.1% (t=3.21) Loser portfolio, 16 three-year periods
January excess return — month t=13 5.6% (t=3.07) Loser portfolio, 16 three-year periods
January excess return — month t=25 4.0% (t=2.76) Loser portfolio, 16 three-year periods
Return spread at 12 months 5.4% (t=0.77) Not statistically significant; reversal concentrated in years 2–3
Winner portfolio CAPM beta 1.369 3-year experiment; t-statistic on difference vs. loser: 3.09
Loser portfolio CAPM beta 1.026 3-year experiment; winners are significantly more risky
ACARL,60 − ACARW,60 (5-year formation) 0.319 (t=3.28) 10 five-year periods, 50 stocks

Why This Matters

Loser portfolios carry lower CAPM betas than winners, so risk-adjusted returns understate the true overreaction spread. The reversal concentrates in January of years two and three, not year one, which raises questions beyond tax-loss selling as a full explanation. The results challenge weak-form market efficiency: past return data alone predict future returns without any fundamental information. De Bondt and Thaler trace the pattern to Kahneman and Tversky's representativeness heuristic, connecting laboratory psychology to aggregate market prices.

Frequently Asked Questions

24.6%: that is the cumulative average residual spread between prior losers and winners at 36 months in De Bondt and Thaler (1985). Stocks that previously underperformed subsequently outperform stocks that previously outperformed. The authors attribute this to investors overreacting to news in violation of Bayes' rule.

19.6%: the cumulative market-adjusted excess return for the loser portfolio at 36 months. Winner portfolios earn about 5.0% below the market over the same period. The 24.6% spread (t-statistic: 2.20) spans 16 non-overlapping 3-year periods between January 1933 and December 1980.

1.369 versus 1.026: those are the CAPM betas for winner and loser portfolios in the 3-year experiment (t-statistic on difference: 3.09). Loser portfolios are significantly less risky than winner portfolios by this measure. If CAPM is correct, market-adjusted returns understate the true magnitude of the overreaction effect.

8.1% is the excess return earned by the loser portfolio in January of the first test year (month t=1; t-statistic: 3.21). The January effect persists: 5.6% (t=3.07) at month 13 and 4.0% (t=2.76) at month 25 — as late as five years after portfolio formation.

Source

Werner F. M. De Bondt and Richard Thaler (1985). Does the Stock Market Overreact?. Journal of Finance.

Read the full paper