How Much of an Anomaly Lives on the Short Side

Almost every published equity factor is reported as a long-short spread: buy the top decile, short the bottom decile. Academic backtests treat the short leg as symmetric to the long leg — sell a stock, buy it back later, collect the spread's return. Real short-selling is not symmetric. To sell a stock short, a broker first has to borrow it from another holder, and that borrow has a price. For a small fraction of stocks, that price is trivial. For the stocks that anomaly research loves to short — small, cheap, distressed, heavily shorted already — that price can be enormous, and the stock may not be reliably available to borrow at all.

Why the short leg is systematically the expensive side

The stocks that anomaly research assigns to the "bottom decile" — overpriced, low-quality, financially distressed — are disproportionately the same stocks that are already hard to borrow, because informed short-sellers have often identified and borrowed against them already, tightening supply. A stock trading "general collateral" (easy to borrow, near-zero fee) costs almost nothing to short. A "hard-to-borrow" stock can cost 5%, 20%, or even over 100% annualized in borrow fees — a cost that eats directly into the very return the short position is supposed to earn.

Worked example

A negative-earnings-quality anomaly reports a backtested long-short return of 9% a year, with 4.5% coming from the long side (buying high-quality names) and 4.5% from the short side (shorting low-quality names). A study of actual borrow rates finds that stocks in the bottom decile average an annualized borrow fee of 6%, well above the roughly 0.3% fee typical for the broad market, because they're disproportionately small, distressed companies that short-sellers have already targeted. Subtracting that 6% fee from the 4.5% short-leg return leaves the short side losing 1.5% a year after borrow costs, even before considering the risk that shares get recalled and the position must be closed at an inopportune time. Total realistic long-short return: roughly $4.5\% - 1.5\% = 3\%$, a third of the reported 9%.

What this means in practice

Any factor strategy that relies heavily on shorting should be evaluated leg by leg, not just as a combined spread, because the long leg alone often survives realistic costs far better than the short leg does. Some funds address this by running the strategy 130/30-style (mostly long, a modest short overlay) or by checking real-time borrow availability before sizing a short position, rather than assuming the backtest's symmetric assumption holds.