Shorting in Broad Daylight: Short Sales and Venue Choice
In a guest post featuring University of North Carolina research, we consider where investors are choosing to short, and why it matters.
Off-exchange trading in so-called “dark pools” has been around since the 1980s, but these venues have become an increasingly common place for institutional equity managers to transact. Yet, in the fallout from the financial crisis, dark pools have been criticized in the popular press as being vulnerable to conflicts of interest and predatory trading by high-frequency traders (for example, in the 2014 book Flash Boys by Michael Lewis). However, very little rigorous research has examined how informed traders use dark pools versus traditional exchange platforms. For example, when a trader has time sensitive information and needs to execute very quickly, would they be better off trading on an exchange or with a dark pool? If dark pools are the preferred venue, the claim that dark pools provide a safer place for traders to make uninformed (e.g., liquidity) trades may need to be revisited.
First, what is a dark pool? While they come in many specific flavors, all dark pools share the common characteristics of being off-exchange platforms where orders are not publicly observable to other market participants. Orders are matched in order to save on bid-ask spread costs and to potentially reduce price impact. However, if no match within the dark pool exists, a trade won’t take place. Intuitively, dark pools can reduce the costs associated with information asymmetries that would occur on public exchanges, but introduce more execution risk and potential opportunity costs of non-execution.
The biggest challenge in determining where informed traders should transact is simply identifying informed traders in the data. In a recent paper “Shorting in Broad Daylight: Short Sales and Venue Choice” co-authored with Mehrdad Samadi (SMU Cox School of Business) and Jonathan Sokobin (FINRA), we utilize the fact that the literature has established short sellers as well-informed investors whose transactions tend to out-perform the market. We then examine venue choice of short sellers using data from individual stock exchanges and from the FINRA Trade Reporting Facilities (TRF) which includes dark pool trades.
The premise of our analysis is that exchanges and dark pools present tradeoffs for an investor seeking to make a large trade in a timely manner. Exchange orders can be submitted to execute immediately even if the resulting execution prices fall outside of the prevailing National Best Bid and Offer (NBBO). In contrast, dark pools are required to match orders at prices within the NBBO. So very explicitly, dark pools can lead to lower bid-ask spread costs, but if no match within the NBBO exists, then the trader loses the potentially very valuable time spent seeing if there is a match.
In our analysis we consider three main questions in order to understand how short sellers weigh these trade-offs: First, in which venue type is short sale activity more prevalent? Second, where are short sales more informative about future returns? Finally, where do short sellers trade surrounding corporate news releases associated with short-lived information when execution risk is acute?
We begin by looking where short sale activity is more prevalent. Figure 1 presents statistics showing short sale volume divided by total volume, by venue during 2012-2014. Exchange short sales represent 45.7% of a stock’s exchange trading volume on average. Short sales executed in dark pools represent just 37.0% of a stock’s dark pool trading volume on average. This difference is statistically significant at the 99% confidence level, so it seems that overall short sellers prefer exchanges to dark pools. But these results don’t tell us if short sales perform better on exchanges versus dark pools.
As noted already, the literature finds that heavily shorted stocks underperform lightly shorted stocks, capturing the essence of long-short, market neutral, and statistical arbitrage strategies and indicating that short sellers are good at relative valuation. To examine performance in our data we take a traditional calendar-time portfolio approach to measure differences in future returns by venue. Stocks are sorted daily into quintiles based on the previous day’s short sale ratio and portfolios are held for 20 days. Figure 2 plots cumulative raw returns of portfolios representing the difference between the most heavily and lightly shorted stocks.
The results show that shorting on both exchanges and dark pools is associated with informed trading (i.e., positive returns). However, exchange short sales are significantly more informative than dark pool short sales. Heavily shorted stocks in dark pools underperform corresponding lightly shorted stocks by 0.53% on a risk-adjusted basis over the subsequent 20 trading days (or 6.41% annualized). This difference is more pronounced on exchanges: heavily shorted stocks underperform lightly shorted stocks by 0.89% (10.70% annualized). Furthermore, the difference-in-differences between dark pools and exchanges is 4.29% and statistically significant.
When we further dig into the results in Figure 2, we find that the difference in return predictability is even larger on corporate news release days when short sellers are likely to have a larger information advantage. This is an important difference that suggests timeliness of information and the related execution risk are important determinants of trading venue. To examine this issue more closely we examine short sale venue choice around corporate news releases in five-minute increments. We split up short sales into exchange short sale, midpoint dark pool short sales (yMP ), and non-midpoint dark pool short sales (yNMP ). We then estimate a statistical model which examines the cross-relations of these sales and also the relationship between scheduled and unscheduled news events (the specific model is called a panel vector autoregressive model with exogenous variables or VARX). Results are presented in Figure 3.
The coefficients in the bottom two rows of Panel A tell us that scheduled news forecasts more short selling activity in the dark pools and overall. In contrast, unscheduled news only forecasts more short selling activity on exchanges. In Panel B, we show how short sale shares in dark pools respond to scheduled and unscheduled negative news events. We find a significant decrease in midpoint and non-midpoint dark pool short sale shares (relative to their average short sale share in the raw five-minute intervals) during the five-minute interval of unscheduled news. These results suggests that short sellers prefer to trade on exchanges even more when short-lived information makes execution risk more severe.
Overall these results suggest that informed traders (in this case short sellers) prefer to trade on exchanges and benefit from the lower execution risk, despite the more visible nature of their activity. In many ways this confirms the common anecdotal evidence that dark pools are better for investors when they have uninformed trading needs (e.g., liquidity).
This article summarizes a research paper originally published by the Kenan Institute of Private Enterprise and SMU Cox School of Business. The authors of the paper were Adam V. Reed (UNC Kenan-Flagler Business School), Mehrdad Samadi (SMU Cox School of Business), and Jonathan S. Sokobin (FINRA). View the original paper here.