Retail Short Selling And Stock Prices via Columbia Business School
Eric K. Kelley and Paul C. Tetlock
This study tests asset pricing theories that feature short selling using a large database of retail trading. We find that retail short selling negatively predicts firms’ monthly stock returns and news tone, even controlling for overall short selling. Predictability from retail shorting is strongest in stocks with low analyst and media coverage, high idiosyncratic volatility, and high turnover; it does not depend on short sales constraints. Retail buying positively predicts returns in similar types of stocks. These results are consistent with the theory that retail short selling informs market prices, but are inconsistent with alternative theories in which retail short selling is a proxy for sentiment or attention.
Retail Short Selling And Stock Prices – Introduction
Short selling now accounts for roughly one quarter of all trading in the US stock market (Diether, Lee, and Werner (2009)), suggesting shorting is a key determinant of stock prices. While numerous studies find that short interest and short selling negatively predict stock returns, far fewer systematically test competing theories of why this is the case.1 Understanding the economic forces driving the relation between short selling and prices is important because recent models highlight vastly different mechanisms, ranging from informed trading to attention-based mispricing. Such mechanisms determine how markets incorporate information, the role of traders’ potentially incorrect beliefs, and the implications of short sale constraints.
We test competing asset pricing theories by exploiting unique data on retail investor behavior, including short selling. Because retail investors differ widely in their trading skill and access to information, our data likely includes trades motivated by perceived information and those motivated by genuine information.2 Our data can help us distinguish these motives if short sellers are more likely than other retail traders to be informed, as argued in Boehmer, Jones, and Zhang (2008). In addition, retail short selling may be easier to interpret than institutional short selling, which depends on portfolio managers’ beliefs and conflicts of interest as well as their clients’ beliefs and liquidity needs (Lamont and Stein (2004)).
We propose a parsimonious model of stock prices that features three primitive investor types: rational arbitrageurs without short sale constraints; and two types of agents with belief biases, one of which is subject to short sale constraints. Retail investors consist of an unknown mixture of the unconstrained arbitrageurs and the biased and constrained agents. The model nests three leading joint hypotheses about retail investor behavior and stock pricing. First, retail short sellers could be uniquely informed and predict decreases in firm value. Second, retail short sellers could act on pessimistic investor sentiment, which could cause stock underpricing and positively predict stock returns. Third, retail short selling could simply reflect attention from traders whose opinions differ. Because differences in investors’ opinions and attention can combine with short sales constraints to cause overpricing (e.g., as in Miller (1977)), retail short selling could negatively predict returns even if these traders are not well-informed.
We analyze the comparative statics of the above information, sentiment, and attention hypotheses and distinguish among models using empirical return predictability. Our main empirical result is that retail short selling robustly negatively predicts returns in a wide range of stocks—all except the top NYSE size quintile—with annualized magnitudes of 5% to 10%. Controlling for retail trading activity, overall short selling, and changes in mutual fund ownership breadth does not materially reduce predictability from retail shorting. Predictability decreases with firm size, analyst coverage, and media coverage; it increases with idiosyncratic volatility and turnover; and it does not depend on proxies for short sales constraints. Furthermore, retail buying is a positive predictor of stock returns, especially in small stocks.
The evidence that return predictability from retail shorting is negative in nearly all subsamples and horizons contradicts the sentiment hypothesis. Reconciliation with the attention hypothesis is also difficult. The strongest evidence against the attention hypothesis is the positive return predictability from retail buying, particularly in small stocks in which retail short selling negatively predicts returns. Moreover, short sales constraints do not interact with predictability as required by the attention hypothesis.
See full PDF below.