The Short of It: Why Anomaly Profits Come from the Short Side

The Short of It: Why Anomaly Profits Come from the Short Side

Of the 11 well-documented stock market anomalies examined by Stambaugh, Yu, and Yuan in their 2012 study, every single one produced larger abnormal returns on the short side than on the long side. The pattern was not marginal. Across anomalies ranging from accruals to momentum to financial distress, the short leg delivered monthly alpha that was, on average, three to five times the magnitude of the long-leg alpha. This observation upends a comfortable assumption embedded in much of the factor investing industry: that buying high-quality, cheap, or high-momentum stocks captures the bulk of the anomaly premium.

It does not. The premium lives disproportionately on the other side of the trade, in the stocks that are overpriced and should be sold short. And for most investors, that side is effectively inaccessible.

The Theoretical Foundation: Why Overpricing Persists

The intellectual roots of this asymmetry trace to Miller (1977), who proposed a deceptively simple idea: when investors disagree about a stock's value and short selling is constrained, the market price will reflect the views of the most optimistic participants. Pessimists who would normally sell the stock short are sidelined by borrowing costs, regulatory restrictions, or institutional mandates that prohibit short positions. The result is a price that sits above the consensus valuation, tilted toward the optimistic tail of the opinion distribution.

This creates a structural asymmetry in how mispricings are corrected. Underpriced stocks can be purchased by anyone with capital. The barrier to entry is minimal: open a brokerage account, click buy. Overpriced stocks, by contrast, can only be corrected by short sellers, a much smaller population of market participants who face a gauntlet of frictions. D'Avolio (2002) documented these frictions in granular detail using proprietary data from a large institutional lender. He found that while the vast majority of US stocks (roughly 80%) can be borrowed at negligible cost (around 25 basis points annualized), the distribution has a long right tail. Approximately 9% of stocks face loan fees exceeding 1% annualized, and the most constrained names, often small, volatile, heavily shorted firms, face borrowing costs above 10%.

These are precisely the stocks that populate the short legs of anomaly portfolios. The stocks ranked worst by accruals quality, profitability, or momentum tend to be smaller, more volatile, and more heavily shorted. They are exactly the names where short-selling constraints bind most tightly.

Shleifer and Vishny (1997) formalized this insight into the broader "limits of arbitrage" framework: even when sophisticated investors identify mispricing, they may be unable or unwilling to correct it. Short positions carry unlimited downside risk, require margin maintenance, and can be forcibly closed through short squeezes or share recalls. These risks deter arbitrage capital from flowing toward overpriced securities, allowing the mispricing to persist.

The Empirical Evidence: 11 Anomalies Decomposed

Stambaugh, Yu, and Yuan (2012) tested this theoretical prediction by decomposing the returns of 11 prominent anomalies into their long and short components. The anomalies span a wide range of return predictors: financial distress (Campbell, Hilscher, and Szilagyi), O-score (Ohlson), net stock issuance (Ritter), composite equity issuance (Daniel and Titman), total accruals (Sloan), net operating assets (Hirshleifer et al.), momentum (Jegadeesh and Titman), gross profitability (Novy-Marx), asset growth (Cooper, Gulen, and Schill), return on assets (Fama and French), and investment-to-assets (Titman, Wei, and Xie).

For each anomaly, the authors sorted stocks into decile portfolios based on the relevant signal, then calculated four-factor (Carhart) alphas for the long leg (top decile) and short leg (bottom decile) separately.

The results were striking. In 10 of 11 anomalies, the short-leg alpha was statistically significant. By contrast, the long-leg alpha was significant in only 6 of 11. The average monthly short-leg alpha across all 11 anomalies was roughly 0.40%, compared to approximately 0.13% for the long leg.

The distress and O-score anomalies showed the most dramatic asymmetry. Stocks of financially troubled firms, which theory suggests should earn higher returns as compensation for distress risk, actually performed far worse than predicted. The short leg (most distressed firms) generated monthly alpha of over 0.50%, consistent with severe overpricing driven by speculative demand from retail investors and constrained shorting.

The Sentiment Channel: When Overpricing Gets Worse

The paper's second major contribution was linking this asymmetry to investor sentiment. Using the Baker and Wurgler (2006) sentiment index, a composite measure derived from six market-based proxies including closed-end fund discounts, IPO volume, and equity share in new issuance, Stambaugh et al. divided their sample into high-sentiment and low-sentiment months.

The prediction was precise: if anomaly profits derive primarily from overpricing, and if overpricing is amplified when optimistic investors dominate the market, then anomaly short legs should perform better following high-sentiment periods. The long legs, which represent underpriced stocks, should be less affected by sentiment because there are no structural barriers to buying underpriced securities.

The data confirmed this prediction with unusual clarity. Following months of high sentiment, the average short-leg alpha across the 11 anomalies roughly doubled compared to low-sentiment periods. The long legs showed no statistically meaningful variation across sentiment regimes.

This finding has a testable corollary that Stambaugh, Yu, and Yuan (2015) explored in a follow-up paper: "arbitrage asymmetry." They argued that many asset pricing puzzles, including the idiosyncratic volatility anomaly (the finding that high-volatility stocks underperform), can be explained by the same mechanism. When arbitrage is more difficult on the short side, overpricing dominates underpricing, and signals that predict both overpricing and underpricing will appear to predict returns primarily through the short leg.

Implications for Factor Investing

These findings create an uncomfortable reality for the growing factor ETF industry. Most factor products are long-only: they tilt a portfolio toward stocks that rank well on value, momentum, quality, or low volatility, but they do not short the stocks that rank poorly. According to Stambaugh et al.'s analysis, this means they are capturing the smaller portion of the anomaly premium.

Consider a concrete example. A long-only value factor ETF buys the cheapest quintile of stocks by book-to-market ratio, overweighting them relative to the market index. The monthly alpha of this long leg is approximately 0.15%. A long-short value strategy that also shorts the most expensive quintile captures an additional 0.33% per month from the short leg. The long-only investor receives roughly one-third of the full anomaly spread.

This does not mean long-only factor investing is worthless. A 0.15% monthly alpha, compounded over years, remains economically meaningful. But it does mean that the academic evidence on anomaly returns overstates what is achievable for the vast majority of investors. When a researcher reports that a value or momentum strategy delivers 80 basis points per month, a long-only investor should mentally discount that figure by 60-80%.

The low-volatility anomaly is particularly affected by this dynamic. The observation that low-volatility stocks outperform high-volatility stocks is driven almost entirely by the poor performance of the most volatile stocks on the short side. A long-only low-volatility strategy captures a modest premium by holding boring stocks; it misses the much larger premium available from shorting lottery-like speculative names.

Practical Barriers to Capturing the Short Side

Even for institutional investors who can short sell, capturing the full short-leg alpha faces significant headwinds.

Nagel (2005) documented that stocks with low institutional ownership, which tend to populate anomaly short legs, are particularly expensive to borrow and subject to recall risk. When a lender needs their shares returned, the short seller must cover the position regardless of whether the trade has reached its intended horizon.

The costs compound in several ways. First, the direct borrowing fee reduces gross alpha. For hard-to-borrow stocks, this fee can consume a substantial share of the expected return. Second, the market impact of covering positions in thinly traded names can be severe. Third, short squeezes can force covering at the worst possible time, transforming a statistically profitable strategy into a realized loss.

These frictions help explain why anomaly short-leg alpha persists: it is compensation for the real costs and risks of short selling. In economic terms, the alpha is not "free money" but rather a premium paid to those willing to bear the operational, financial, and career risks of maintaining short positions in overpriced securities.

Connecting to Behavioral Biases

The Stambaugh et al. framework connects directly to behavioral finance. The overpricing that drives anomaly short-leg returns arises from well-documented cognitive biases: overconfidence leads investors to overweight their private information, representativeness causes them to extrapolate recent performance into the future, and the disposition effect makes them reluctant to sell losers.

During high-sentiment periods, these biases operate with reduced resistance. Optimistic retail investors flood into speculative stocks, pushing prices above fundamental value. Normally, short sellers would provide a corrective force. But when borrowing is expensive or impossible, there is no effective mechanism to push prices back toward fair value until the overpricing unwinds on its own, often through disappointing earnings, restructuring, or eventual delisting.

This creates a predictable cycle. Sentiment rises, overpricing increases, and the gap between fundamental value and market price widens for the most speculative names. When sentiment reverses, these stocks experience outsized declines as the accumulated overpricing corrects. Anomaly strategies that are short these names during the correction phase capture the largest returns.

What Survives After Costs

A natural question is whether anomaly short-leg alpha survives after accounting for all implementation costs. The answer depends on the anomaly, the investor, and the market regime.

For the most constrained stocks (the smallest, most volatile, hardest to borrow), the answer is often no. Borrowing costs, market impact, and recall risk can exceed the gross alpha. This is consistent with the limits-of-arbitrage theory: the alpha persists precisely because it is expensive to capture.

For moderately constrained stocks, the picture is more nuanced. Institutional short sellers with prime brokerage relationships, access to locate desks, and patient capital can capture a meaningful portion of the gross alpha after costs. The key is selectivity: rather than shorting the full bottom decile (many of which are untradeable), focusing on the subset of short-leg stocks that are liquid enough to trade efficiently.

The long-short factor strategies offered by some hedge funds and liquid alternatives attempt to navigate this tradeoff. They sacrifice some theoretical short-leg alpha (by excluding the hardest-to-borrow names) in exchange for practical implementability. The net alpha is smaller than the academic estimates but still meaningful.

The Asymmetry in Perspective

Stambaugh, Yu, and Yuan's insight that anomaly profits concentrate on the short side has become one of the most cited findings in modern empirical asset pricing. It explains why so many anomalies that look impressive in backtests deliver disappointing returns in live portfolios: the portion of the anomaly that most investors can access, the long leg, is the weaker component.

For individual investors evaluating factor-based strategies, the key takeaway is calibration. The returns documented in academic research represent the theoretical ceiling of a long-short strategy implemented without friction. A long-only factor tilt captures a fraction of this, and that fraction, while still positive, is meaningfully smaller than headlines suggest. Understanding this gap is essential for setting realistic expectations about factor investing returns.

The overpricing that drives anomaly profits is not a market flaw waiting to be eliminated. It is a structural consequence of the asymmetry between buying and short selling. As long as short-selling constraints exist, and they show no signs of disappearing, the most profitable anomaly opportunities will remain concentrated on the side of the market that most investors cannot reach.

References