Is the monthly stock return reversal effect currently exploitable? In the August 2011 version of their paper entitled “New Evidence on Short-Term Reversals in Monthly Stock Returns: Overreaction or Illiquidity?”, Chris Stivers and Licheng Sun investigate the persistence, size-sensitivity and seasonality of monthly stock return reversal in the context of three competing explanations: (1) investor overreaction to news (exploitable); (2) market illiquidity (perhaps unexploitable); and, (3) large stocks lead small stocks (exploitable). They evaluate simple value-weighted and equal-weighted prior-month loser-minus-winner (LMW) strategies based on a sort of prior-month returns, and five more complex equal-weighted LMW strategies based on double-sorts of prior-month returns and market capitalizations. Using monthly return and market capitalization data for a broad sample of U.S. stocks and 30 industries over the period February 1926 through December 2010, they find that:
- Over the entire sample period, a value-weighted (equal-weighted) LMW portfolio has a gross average monthly return of about 0.95% (2.86%).
- However, this monthly reversal effect dissipates considerably since discovery for all seven LMW strategies. For example, the gross average monthly return for the simple value-weighted (equal-weighted) LMW strategy is 1.27% ( 3.24%) during 1926-1989, compared to -0.02% (1.72%) for 1990-2010.
- The monthly reversal effect is much stronger for small-capitalization and low-priced stocks. For example:
- The gross average monthly return for the equal-weighted LMW strategy applied to the smallest (largest) market capitalization quintile is 3.29% (0.61%) during 1926-1989 and 1.44% (0.25%) during 1990-2010.
- Screening out stocks priced $5 or less depresses the reversal effect by nearly 50% in the full sample and eliminates statistical significance during 1990-2010.
- The monthly reversal effect exhibits strong seasonality, with pronounced profitability in January during both 1926-1989 and 1990-2010. After accounting for this seasonality, the effect largely disappears since 1990. For example:
- The gross average January return for the value-weighted (equal-weighted) LMW strategy is 4.70% (10.2%) during 1926-1989 and 3.73% (9.94%) during 1990-2010.
- The gross average return over the other months for the value-weighted (equal-weighted) LMW strategy is 0.96% (2.62%) during 1926-1989 and -0.36% (0.97%) during 1990-2010.
- There is no evidence that a lag in small-capitalization stock returns relative to large-capitalization stock returns contributes to the monthly reversal effect.
- Applying monthly reversal strategies to industries rather than individual stocks generates negative gross average monthly returns, but January still outperforms other months. (This finding is consistent with the lower performance of a sector momentum strategy with a skip-month compared to one without a skip-month in “Simple Sector ETF Momentum Strategy Performance”.)
In summary, evidence from an array of tests indicates that the monthly stock return reversal effect weakens considerably since its discovery and that January seasonality accounts for it since 1990. Results generally support a belief that the effect derives from difficult-to-exploit market illiquidity rather than systematic investor overreaction.
Cautions regarding findings include:
- Return calculations are gross, not net. The reversal strategy is trading-intensive, with profitability concentrated in stocks with the highest trading frictions. Incorporation of reasonable trading frictions (generally higher for older subperiods and for shorting) would materially reduce calculated returns.
- Calculations assume that investors can execute prior-month sorts just before monthly closes, thereby taking responsive positions at these same closes. This assumption may be problematic for some investors from calculation burden and trade execution perspectives. Delaying trades until the next trading day would systematically miss part of the the turn-of-the-month effect.
- Recent subsamples are short for assessment of seasonal (January) effects.
- Statistical significance tests assume well-behaved return distributions. To the extent that return distributions are wild, these tests lose meaning.
