We construct novel measures of net and gross short covering to examine when short sellers exit positions. We find that idiosyncratic limits to arbitrage, such as adverse stock price movements, volatility, and equity lending fees, are associated with significantly higher position closures. In contrast, we find little evidence that aggregate limits to arbitrage, including VIX, funding liquidity, and market liquidity, affect short covering. Short covering predicts future returns in the wrong direction, but only if it is induced by limits to arbitrage, consistent with the hypothesis that short sellers are forced to exit too early. It is also associated with lower price efficiency, higher future anomaly returns, and better performance of other informed traders. These results show that firm-level limits to arbitrage are important determinants of trading behavior and future returns.

Micro-level geographic proximity is associated with higher returns from short selling, with short trades by institutions near the target headquarters followed by more negative abnormal returns. Proximity matters more for stocks that are small, volatile, and have less analyst coverage, as well as for stocks with low market correlations and inefficient prices. Funds exhibiting larger effect of proximity are smaller and have higher returns and idiosyncratic volatility. The relationship between distance and returns is weaker during the COVID-19 pandemic. Overlapping nearby bars and restaurants between target and short seller matter but not during holidays, suggesting social interactions as a channel.

Short-sale costs eliminate the abnormal returns on asset pricing anomaly portfolios. While many anomalies persist out-of-sample, they cannot profitably be exploited due to stock borrow fees. Using a comprehensive sample of 162 anomalies, the average long-short portfolio return is a significant 0.15% per month before short-sale costs, and the returns are due to the short leg.

However, the average is −0.02% once returns are adjusted for borrow fees. The anomalies are not profitable even before accounting for fees if the high-fee observations, 12% of stock dates, are excluded from the analysis. Thus, short sale costs explain why many anomalies persist.