Short-Term Reversal

Short-term reversal is a pattern in share returns in which the worst performers over a recent short period, usually the past week or month, tend to earn higher returns in the following period than the best performers. A strategy built on it buys recent losers and sells recent winners, holding the positions for days or weeks. It is one of the oldest documented departures from the idea that past prices carry no information about future returns.

The effect runs in the opposite direction to momentum, which appears over horizons of roughly three to twelve months. The two coexist because they operate on different time scales. Short-term reversal is a form of cross-sectional mean reversion and is a building block of many statistical arbitrage strategies.

The effect was documented in two papers published in 1990. Narasimhan Jegadeesh found that United States shares with the lowest returns in one month outperformed those with the highest returns in the next month. Bruce Lehmann reported the same pattern at a weekly horizon. Later work found reversals in many other equity markets.

The finding shaped how momentum is measured. The standard momentum signal uses returns over the past twelve months but skips the most recent month, precisely to avoid contamination by the reversal effect. A separate long-term reversal over horizons of three to five years was reported by Werner De Bondt and Richard Thaler in 1985. A short-term reversal factor for United States shares is published in the data library maintained by Kenneth French.

• Compensation for providing liquidity. When investors sell a share urgently, its price is pushed below fair value to attract buyers. Those who buy, in effect acting as a market maker, are rewarded as the price recovers. On this view the reversal return is payment for supplying liquidity, and it should be larger when markets are stressed, which is what researchers have observed.
• Overreaction. Investors may respond too strongly to news or to price moves themselves, and the excess is later corrected. This account comes from behavioral finance.
• Measurement effects. A closing price recorded at the bid one day and at the ask the next creates an apparent reversal with no change in value. This bounce across the bid-ask spread inflates measured profits, especially in illiquid shares.

Reversal is weaker when the initial price move coincides with genuine news, such as an earnings release, and stronger when it occurs without news, which is consistent with the liquidity explanation.

A standard way to express the strategy, used by Andrew Lo and Craig MacKinlay, gives each share a weight proportional to how far its recent return was from the market average, with the sign reversed:

• $w_i$ is the weight in share $i$; a negative value is a short position.
• $N$ is the number of shares.
• $r_i$ is the share's return over the previous period.
• $r_m$ is the equal-weighted average return of all the shares over that period.

Take four shares whose returns last week were 4, 1, -1 and -4 percent. The average is zero. The weights are $-0.04 / 4 = -0.01$, $-0.0025$, $0.0025$ and $0.01$. The portfolio is short the two winners and long the two losers, with the largest positions in the most extreme movers, and the weights sum to zero, so the long and short sides are equal in value. Practitioners refine this by measuring returns relative to an industry or to a factor model, so that the bet is on the share-specific part of the move.

The strategy trades its whole portfolio every week or month, so turnover is extremely high and each transaction cost matters. Studies have found that the measured profits are concentrated in small, thinly traded shares, where spreads are wide and market impact is large, and that after realistic costs much of the return disappears for an ordinary investor. Among large, liquid shares the effect has weakened considerably since the 1990s as electronic market making has grown.

Firms that do capture it tend to be those with the lowest costs and best execution, which is why the effect is now associated with professional liquidity providers. Short positions also require shares to be borrowed, which can be costly or impossible for exactly the shares the signal selects.

A reversal strategy buys what is falling. If a decline reflects real deterioration, the losers keep losing, and the strategy suffers. Its losses tend to arrive in market crises, when prices trend strongly, liquidity dries up and many participants holding similar positions reduce them together. The events of August 2007, when a number of equity market-neutral funds lost heavily within days, are commonly linked to the unwinding of such positions. The return is therefore better understood as a premium earned for bearing liquidity risk than as a market inefficiency that can be collected without risk.