Print
E-MailFrancesco Franzoni is Professor of Finance at USI Lugano. This post is based on a recent paper authored by Professor Franzoni; Marco Di Maggio, Assistant Professor of Business Administration at Harvard Business School; Massimo Massa, Rothschild Chaired Professor of Banking and a Professor of Finance at INSEAD; and Roberto Tubaldi, PhD candidate at USI Lugano & the Swiss Finance Institute.
There is consensus in the theoretical and empirical literature on the fact that short sellers are informed traders. Hence, economic theory suggests that when short sellers interact in the market with uninformed investors the extent to which prices reveal fundamental information (price efficiency) increases. The favorable regulatory environment for short selling in most developed countries reflects these considerations.
However, a more realistic portrait of the market should include multiple groups of informed investors. In this context, the competition among traders affects the incentives for trading and the revelation of information in prices. Specifically, if investors with positive information face the competition of short sellers, they may delay their trades to exploit the decrease in price induced by short sellers. Therefore, when information is diverse and dispersed across different groups of traders, the presence of short sellers may lead to a slow-down of the impounding of positive information.
One may wonder how market participants can actually infer the presence of short sellers. Indeed, several channels contribute to make the market aware of the extent of short selling activity. For example, brokers that intermediate share loans can spread the word to their other clients in order to establish a reputation as valuable sources of information. In addition, data providers publish statistics on short selling activity.
To study empirically informed investors’ reaction to short sellers, we combine short selling information at the stock level from Markit Securities Finance with data on institutional trades from Abel Noser Solutions (ANcerno) from 2002 to 2014.
For our empirical analysis, we select periods in which there is more likelihood of informed trading. To identify such periods, we follow two strategies. First, the period preceding earnings announcements provides a fitting laboratory, as informed investors trade on their private signals before information becomes publicly available. Consequently, part of our analysis is restricted to the period around earnings releases. Second, we identify informed trading by focusing on large trades, which are defined as cumulative volume in one month in the top quartile of the order-flow distribution for a given manager-stock.
Consistent with the conjecture that short selling activity slows down the trades of other informed investors, we find a significant delay in buy trades in the two weeks before the announcement for stocks with higher short interest. In particular, investors reduce the amount of buying by about 22% relative to the mean in the two weeks before the announcement when short selling in the prior-four weeks is one-standard-deviation higher.
Then, we turn to implications of the waiting-game hypothesis for price informativeness. We study the fraction of the cumulative abnormal return in the window between ten days before and one day after the announcement that is realized before the announcement. We find that short selling is related to a decrease in the amount of price discovery before positive earnings announcements. On the other hand, we do not find that short selling increases information impounding ahead of negative news, consistent with a lack of information advantage of short sellers in the period ahead of earnings releases.
To address the issue that short selling is an endogenous variable, we focus on the period of the Reg SHO experiment. In the two years between 2005 and 2007, the SEC suspended short-sale price tests, i.e. the uptick rule, for a randomly selected group of stocks (Pilot stocks). This policy was explicitly designed to provide an exogenous release of short selling constraints for one third of the Russell 3000 universe and assess the effect of short selling on different market outcomes. Using this experiment, we corroborate our prior findings. For example, the information revealed in prices before positive earnings announcements is smaller by a sizeable 20% relative to the mean for Pilot stocks during the experiment. Instead, we do not find any effect on information impounding before negative earnings surprises.
Recent evidence suggests that institutional brokers are a source of information leakage (Barbon, Di Maggio, Franzoni, Landier, 2018; Di Maggio, Franzoni, Kermani, Sommavilla, 2018). Thus, we conjecture that informed investors wishing to reduce information leakage should spread their trades across multiple brokers when short sellers are more present in the market. To test this conjecture, we study informed investors’ hiding behavior in the form of trade breakup across brokers. Consistent with hiding behavior, we find that managers use more brokers in stocks for which there is higher short selling. This result is fully confirmed in the context of the Reg SHO experiment and it is only present for buy trades. We find evidence of trade breakup both before earnings announcements and in the context of large trades. Along with the decrease in trading speed, the evidence of trade breakup provides a channel for the finding that short selling slows down information impounding.
Overall, the evidence suggests that short selling can hamper price discovery by slowing down the trades of investors with positive information. This result is novel in the literature. It does not, however, contradict previous studies reporting a positive effect of short selling on price efficiency. In fact, it is possible that short sellers improve price efficiency over long horizons, especially after the release of public information. At the same time, during concentrated periods when private information is possessed by multiple investors, short selling activity can deter positively-informed investors from impounding their information. It is also possible that when multiple investors privately possess the negative information that is about to be released, short selling does not add much to price informativeness. On net, the effect of short selling in periods with highly dispersed private information can be detrimental for price efficiency.
Accordingly, the regulation of short selling could be made contingent on the extent of information dispersion across investors and on the timing of specific information releases. The practical implementation of this recommendation obviously presents many difficulties, not least the measurement of information dispersion, and is therefore left for future research.
The complete paper is available for download here.
