Short Interest as a Signal
How much of a stock's float is sold short is one of the better-documented predictors of future returns — heavily shorted stocks underperform on average — but the same crowded short position that predicts underperformance can also fuel the sharpest short squeezes, which is why the signal is treated with more caution than its backtest alone suggests.
Prerequisites: How Short Selling Works
Sophisticated investors are the ones who mostly bear the cost and risk of short-selling — paying borrow fees, facing theoretically unlimited loss if the stock rises — so when a large fraction of a stock's shares outstanding are sold short, that's a concentrated bet from informed money that the stock is overpriced. Short interest, expressed as days-to-cover (shares short divided by average daily volume) or as a percentage of float, has been shown repeatedly to predict below-average returns for the most heavily shorted names.
The signal and its two-sided risk
Asquith, Pathak, and Ritter (2005) found that stocks in the highest decile of short interest as a percentage of shares outstanding underperformed the broad market by several percentage points annually, and the effect was strongest for stocks that were both heavily shorted and had low institutional ownership — the combination proxying for stocks where informed short-sellers had identified a problem that a thin ownership base hadn't yet priced in.
Worked example. Stock A has 35% of its float sold short with 8 days-to-cover, versus a typical stock in its sector at 3–4% short interest. A strategy flagging the top decile of short interest as a short candidate would short Stock A, expecting the historical underperformance pattern to play out — perhaps 3–5% of relative underperformance over the following two quarters based on decile-level studies. But that same 35% short interest and 8-day cover means any sharp piece of good news (a beat-and-raise earnings print, a buyout rumor) forces short-sellers to buy back stock into a market with limited natural sellers, and the stock can gap 20–40% in a single session as shorts race to cover — a short squeeze that erases months of the strategy's expected edge in one day. This asymmetry is exactly why heavily shorted names carry fat-tailed upside risk that a smooth backtest of the average decile return doesn't show.
High short interest predicts below-average returns on average, but it is also a direct measure of squeeze risk — the same crowding that creates the edge creates the tail. Days-to-cover, not just percent-of-float, tells you how violent a squeeze could be.
What this means in practice, and what erodes it
Desks running this signal size positions smaller than the raw decile spread would justify, precisely to survive the tail, and many pair it with a borrow-cost filter: stocks that are "hard to borrow" (expensive or unavailable to short) often signal the crowded-short thesis is already known and priced, muting the forward edge while keeping the squeeze risk. The 2021 meme-stock episodes (GameStop, AMC) are the extreme public case of what a crowded, high-days-to-cover short position can do to a book that sized the position as if backtest averages were guaranteed outcomes.
Short interest data is typically reported with a lag (in the US, twice-monthly, several days stale by the time it's published) and by the time a stock's short interest is visibly extreme in the data, the crowding it reflects may already be old news to the specialized short-sellers who built the position early — you're often seeing the signal well after its freshest value has decayed.
Related concepts
Practice in interviews
Further reading
- Asquith, Pathak & Ritter, Short Interest, Institutional Ownership, and Stock Returns (2005)