SEC’s Regulation SHO-Pilot Program on NYSE: A Problematic Natural Experiment
Kevin M. Zhao Department of Economics and Finance Middle Tennessee State University
Abstract
Examining the private information content of short sales surrounding the implementation of the SEC’s Regulation SHO-pilot program, this paper finds evidence that short sellers were on good behavior, not trading in pilot stocks based their private information about forthcoming analyst downgrades due to the concern of “heightened scrutiny” during the program. Results here indicate that the SHO-pilot program, which generates the SHO data, may not be a clean experiment. Although pilot and control samples are generally matched based on firm characteristics, there is a propounding difference between two samples as far as the informed short sellers’ behavior concerned, making any comparison between the pilot and control samples problematic. This also casts doubts on whether the removal of the uptick rule by the SEC in July 2007 is a proper decision.
Key Words: The SHO-pilot Program, Uptick Rule, Short Selling, Analyst Recommendation, Informativeness of Short Selling
JEL Category: G12; G14
1. Introduction
The uptick rule on the New York Stock Exchange (NYSE) prohibits short selling when stock price has already trended down. The impact of the uptick rule on stock market efficiency has been controversial since its establishment during 1930s. As a typical short selling constraint, the uptick rule tends to have negative effect on stock price efficiency (Miller, 1977; Diamond and Verrecchia, 1987; Jones and Lamont, 2002; Deither et al, 2002; Nagel, 2005; Hong and Stein, 2003). To systematically examining the effectiveness of the uptick rule, U.S. Security Exchange Commission (SEC) implemented on May 2, 2005 a Regulation SHO-pilot program, which suspends the uptick rule for a pre-chosen set of “pilot” stocks, but remain effective for “control’ stocks. The SEC designed the Regulation SHO-pilot program as a natural experiment to facilitate the study of the effect of uptick rule on variety of aspects of market quality in a controlled environment. Several studies using the SHO data generated by the Reg SHO-pilot program shows that the uptick rule has moderate negative effect on market efficiency (SEC, 2007; Diether et al., 2009; Alexander and Peterson, 2008; Boehmer and Wu, 2008; and Zhao, 2009). The SEC’s Office of Economic Analysis (2007) uses SHO data during the period from January to October 2005 to compare pilot and control stocks along a number of dimensions, and find that uptick rules constitute an economically relevant constraint on short selling. Suspending uptick rules for pilot stocks affects the mechanics of short selling, order routing decisions, displayed depth, and intraday volatility, but does not have a deleterious impact on market quality or liquidity. Similarly, Diether, Lee, and Werner (2009) examine a SHO dataset for a shorter period from February to July 2005. They find that the suspension of the uptick rule for NYSE pilot stocks is
associated with wider spreads, more symmetric trading patterns, and higher volatility, while there is no significant effect on market quality for NASDAQ pilot stocks, suggesting that lifting the bid-price test rule for NASDAQ stocks may not improve stock price efficiency. Further, utilizing a two month SHO data from April to May 2005, Alexander and Peterson (2008) find that lifting the uptick rule for NYSE pilot stocks is associated with a decrease in liquidity, but similar measures of market efficiency as control stocks, and that lifting the NASDAQ bid test rule has little impact on nearly all market price efficiency measures. Nevertheless, these studies implicitly assume that the Regulation SHO-pilot program, which generates the SHO data, was a clean natural experiment, in which no systemic difference between the pilot sample and the control sample exists. Indeed, previous studies using the SHO data show that the pilot and control samples are well-matched in terms of firm characteristics, such as trading volume, firm size, and book-to-market ratios. However, the controlled environment created by the Regulation SHO-pilot program would have been compromised, if “traders might behave differently if a rule were permanently and completely removed than if it is only temporarily or incompletely removed. Moreover, it is possible that traders with manipulative intentions might be on good behavior if they believe that heightened scrutiny during the Pilot increases their chances of getting caught” (SEC Office of Economic Analysis, 2007). This paper addresses this issue by examining consistency of short sellers’ trading behavior surrounding the SEC’s implementation of the pilot program. I hypothesize that short sellers were aware of heightened scrutiny in pilot stocks during the pilot program, and intentionally suspended their private information driven short selling to avoid being caught by the SEC. This is the regulatory concern hypothesis. To test the hypothesis, I compare private information
driven short selling behavior in both pilot and control stocks before, during, and after the Regulation SHO-pilot program. The empirical testing results in this paper show that prior to the Regulation SHO-pilot program short sellers consistently capitalize on their private information associated with forthcoming recommendation changes in both pilot and control stocks. However, during the Regulation SHO-pilot program, such privation information driven short selling disappeared in pilot stocks. After SEC permanently removed the uptick rule in July 2007, the private information driven short selling were restored in pilot stocks. The results here is consistent with the regulatory concern hypothesis and provide clear evidence that short sellers were on good behavior, not trading in pilot stocks based their private information about forthcoming analyst downgrades due to the concern of “heightened scrutiny” during the Regulation SHO- pilot program. Several robust tests are performed to show that results are not driven by systematic differences between the pilot and control stocks, such as availability of exchange-traded-options, firm size, the market-to-book ratio. I also adopt alternative specification of time interval prior to recommendation changes and different periods when the pilot program was in place. Additional results are consistent with the initial findings. The alternative explanations of the results could be that the removal of uptick rule restrictions make the market more efficient. I name this as the shorting efficiency hypothesis. When short selling restrictions are relaxed, the short sellers’ privately held information are impounded into prices at a higher speed and with a greater magnitude, thus spoils the market surprises and reduces the incentives for informed short seller to take advantage of their privately held information associated with forthcoming analyst recommendation downgrades. Additional test however does not provide support to this hypothesis.
This paper contributes to the literature in following ways. First, the main finding in this paper suggest that informed short sellers were on good behavior when they believe that heightened scrutiny during the pilot program increases their chances of getting caught. As far as informed short sellers’ behavior is concerned, the pilot and control sample are not matched, suggesting the controlled environment created by the Regulation SHO-pilot program was contaminated. Therefore, using SHO data to compare pilot and control samples without taking account of the systemic difference in short sellers’ informed trading behavior would be problematic and may lead to biased conclusions in previous studies. This also raises the concern whether the removal of uptick rule restrictions by the SEC in July 2007 is a proper decision. This is our first and the most important contribution. Second, this paper complements the literature on informativeness of short selling by linking short sales to private information in which an analyst downgrade is soon to occur. Existing literature show that short selling is informative by examining the relationship between short selling activities and subsequent risk adjusted returns without linking the short selling activities directly to private information held by short sellers. For example, Diether, Lee, and Werner (2008) conclude that short selling are informative based on the negative relation between increasing short sales and subsequent future returns. Also, recent studies have used earnings announcements as a source of private information. For example, Christophe, Ferri, and Angel (2004) investigated short selling before earnings announcements and found that abnormal short selling significantly related to post-announcement stock returns. Moreover, Reed (2007) used quarterly earnings announcements as information events, where earnings are taken to be private information until they are publicly announced, to test the hypothesis that short sale constraints reduce the speed at which prices adjust to private information. These studies ignore the legal risk
involved in insider trading associated with firm’s earnings information. Further, given the same earnings information, investors may formulate different evaluations on stock prices. Using analyst recommendations as information events is a more appropriate choice because trading on forthcoming analyst recommendations imposes less legal risk than trading on forthcoming earnings information. Also, it allows for different opinions given the same earnings information. Nevertheless, this study has several limitations. First, only the impact of short selling on a specific information event is examined: analyst recommendations. There may be other events for which the different information driven short selling patterns associated with the suspension of the uptick rule can be discovered, such as earnings release, Mergers and Acquisition announcements, etc. Second, the impact of altered informed short sellers’ behavior on many market quality measures for uptick rule unrestricted stocks has yet been examined. This important extension would provide the regulators valuable information regarding short selling regulation. Third, this paper only examine short sellers’ private information driven short selling around the implementation of the Pilot program on NYSE. Similar study focuses on NASDAQ would generate fruitful insights. The rest of this paper proceeds as follows. Section 2 provides background information about the SEC’s Regulation SHO-pilot program. Section 3 constructs samples and describes data. Section 4 describes the regression methodology, presents results, conducts robustness tests, and evaluates the shorting efficiency hypothesis. Section 5 concludes.
2. Background Information
2.1. Uptick rule on NYSE
Short selling in the U.S. have been subject to the tick test rule since 1938 according to the Rule 10a-1 under the Securities Exchange Act of 1934. The tick test rule prohibits short selling on downtick for stocks. Short sales are only allowed at or higher than the last traded prices, where the last traded price is higher the most recent trade. This is commonly referred as the uptick rule on NYSE. Since NASDAQ was not recognized as an exchange until August 1, 2006, stocks traded on the NASDAQ were not subject to the tick test rule prior to August 2006. However, NASD (National Association of Securities Dealers) introduced a NASD Rule 3350, which imposes so call “bid test rule” on NASDAQ listing stocks. The rule only allow short sales to occur at prices one penny above the bid price if the bid is a downtick from the previous bid. Due to the different market microstructure and between NYSE and NASDAQ and the difference between the uptick rule on NYSE and the bid test rule on NASDAQ, this paper focuses on the uptick rule on NYSE and leave the bid test rule on NASDAQ for future study.
2.2.SEC’s SHO-pilot Program
In July 2014, the Securities and Exchange Commission (SEC) issued an order to establish a pilot program, which exempts a third of the stock in the Russell 3000 index from either uptick rule or bid test rule restrictions. The SEC implemented the Pilot Program between May 2, 2006 and April 28, 2006, and extended the program until July 2007, when the SEC removed the all price test rules permanently. The SEC picked pilot stocks from the Russell 3000 membership list by ranking on average daily dollar trading volume for the previous year and selecting every third
stock as a pilot stock with the remaining stocks are specified as control stocks. This process should eliminate any systematic difference between pilot and control stocks. The purpose of the SHO-pilot program is to generate empirical data to help assess whether short sale price test rules should be removed. Suspending the uptick rule for pilot stocks during the Pilot program creates a controlled environment where researchers can empirically examine the impact of the uptick rule on many dimension of market quality by comparing trading behavior of the pilot stock to that of the control stocks. Any discernable effect caused by the short sale price test rules would appear as differences between the pilot stocks and control stocks. Figure 1 show the timeline of the Pilot program. Although the pilot program went into effect on May 2005, stockbrokers and dealers are required by to report tick-by-tick short sale data to the SEC as early as January 2, 2005. Between January 2, 2005 and May 2, 2005, both pilot and control stocks are subject to the same the uptick rule on NYSE. I specify this as the pre-SHO period in the paper. Beginning May 2, 2005, the pilot program went into effect. The program lasted for a total of 22 months and ended July 2007. I specify this time-period as the SHO-pilot program implementation period. To facilitate comparison, I chose a 4-month period from January to April 2006 as the during-SHO period. On July 6, 2007, SEC announced to remove all short sale price test rules of all exchange-traded securities. After July 2007, both pilot and control stocks are no longer subject to the uptick rule. I chose a 4-month period from July to October 2007 and specify this period as the post-SHO period. [Insert Figure 1 here]
3. Sample and Data
3.1 Sample Regulation SHO established a set of “pilot” stocks and a matched set of “control” stocks on the NYSE and NASDAQ for which transactions based short sale data starting January 2, 2005 would be collected and eventually made available to the public. Starting May 2, 2005, uptick rule restrictions were suspended from the pilot stocks and the transactions data continued to be collected until July 6, 2007 when the SEC voted to permanently revoke uptick rule restrictions for all exchange-traded securities. Pilot stocks were picked by the SEC from the Russell 3000 membership list by ranking on average daily dollar trading volume for the previous year and selecting every third stock as a pilot stock with the remaining stocks are specified as control stocks. This process should eliminate any systematic difference between pilot and control stocks. To further eliminate the potential effect of index inclusion or index exclusion on stock returns, we require that sample stocks were members of the Russell 3000 index after the June 2004 reconstitution and remained in the Russell 3000 member list after the June 2005 reconstitution. Further, we exclude stocks that were added to the index due to IPOs during the period June 2004 through the end of 2005, as well as stocks that were eliminated during the same period due to mergers, bankruptcies, and ticker changes. Excluding stocks on the pilot stock list as of May 2, 2005 results in the control sample, which is roughly as twice as large as the pilot sample. Previous literature reveals that the impact of uptick rule suspension on market quality is different for NYSE-listed stocks and for NASDAQ-listed stocks. Therefore, we perform separate analyses of NYSE and NASDAQ-listed stock in order to accommodate the different natures of the uptick test constraint on the two exchanges. As the results of these screens, our samples consist of 464 pilot stocks and 826
control stocks listed on the NYSE and 404 pilot stocks and 778 control stocks listed on the NASDAQ. To facilitate the comparison of the private information content of short sales for pilot and control stocks, we construct four samples during the post-SHO period from May to December of 2005: pilot sample on the NYSE, control sample on the NYSE, pilot sample on NASDAQ, and control sample on NASDAQ. If the suspension of the uptick rules contributes to different information driven short sales for the pilot and control samples, then the control and pilot samples should have similar patterns before the Pilot Program, when pilot stocks and control stocks have the identical treatment on the uptick rule restrictions. Thus, we also construct four more samples as above, but for the pre-SHO period from January through the end of April of 2005, a period featured by the same uptick rule restrictions for pilot and control stocks.
3.2 Data
3.2.1. Regulation SHO-pilot Program Data To comply with the Regulation SHO, market dealers and brokers were required to report tick-by-tick short sales transactions to the SEC during a two and half year time-period from January 2005 to June 2007. The resulting data is Regulation SHO data. I obtain NYSE Regulation SHO data from NYSE TAQ dataset for a period from January 2005 to June 2007. After June 2007, market dealers and brokers on NYSE are no longer required to report tick-bytick short sale transactions. However, NYSE retained such data after June 2007. To compare short selling between the during-SHO period and the post-SHO period, I obtain intra-day short selling data from NYSE for a period from July to October 2007. Merging the Regulation SHO-
pilot data with pilot list and control stock list in pre-, during-, and post-SHO periods yields six samples for subsequent analysis. Table 1 compares daily short sales activities for the pilot and control samples during three time periods. The short sale ratio, the size of short orders, the number of short orders, and the short sale volume are reported in Table 1 to measure the daily short sale activities. Table 1 shows that during both the pre- and post-SHO period, there is no statistically significant difference between the pilot and control samples in all measures of short sale activities. It also shows that short sales constitute about 21% of total trading volume on the NYSE consistent with previously documented increasing short sales levels in recent years. For the during-SHO period, however, the short sale ratio for pilot stocks is significantly greater than for control stocks by approximately 2%, indicating that the suspension of uptick rules in pilot stocks stimulates short sale activities. Both short sale volume and the number of short sales are significantly greater in the pilot sample than in the control sample. In addition, the size of short sale orders is significantly smaller in pilot stocks than in control stocks.
(Insert Table 1 about here)
3.2.2 Analyst Recommendation Data Buy-side brokerage firm analyst recommendations change information is obtained from the First Call historical dataset. I assign to recommendations the values of one for “buy”; two for “overweight”; three for “neutral”; four for “underweight”; five for “sell”; and zero for the initiation of recommendation. To measure different types of recommendations, I define recommendation changes as the difference between the previous recommendation and the
current recommendation for the same brokerage firm. For the initiation of a recommendation, I specify the recommendation change as the deviation of the current recommendation from the neutral recommendation with the rating value of 3 . Finally, I classify a recommendation change as a downgrade if the value of the change is negative and a recommendation change as a nondowngrade if the change is either zero or positive. To avoid having to identify intra-day price response to recommendation changes, and to mitigate the problems with identifying a specific time a recommendation change becomes public knowledge, I focus on recommendations that are released outside of regular trading hours, either before 9:30am or after 4:00pm. Further, to ensure that the analyst recommendations represent significant information events we restrict our sample to those recommendations issued by the top 20 U.S. brokerage research departments, as designated by Institutional Investor, that distribute their recommendations through the First Call. I also restrict our sample to those have no previous recommendation changes within five trading days before the current recommendation change date. In addition, we excludes firms with a quarterly earnings announcements fall in the window of five trading days prior to the recommendation change date to protect our samples from the impacts of earnings announcements. Table 2 documents features of analyst recommendation changes in our samples. Panel A of Table 2 presents the number of downgrades and non-downgrades for both pilot and control samples for three sample periods. Panel B details the magnitude of recommendation changes for pilot and control stocks on the monthly basis. It shows that there is no significant difference on the magnitude of recommendation changes between the pilot and control samples. (Insert Table 2 about here)
4. Empirical Tests 4.1 The Regression Model In this section, I adopt a regression approach to examine whether short sellers engage in informed trading prior to the release of analyst recommendations by testing the relationship between abnormal short selling in the days prior to analyst recommendations and the information content of such short selling. In the regression model, the dependent variable, abnormal short selling, ABSS(-5,-1), is defined as the average daily short sale volume during the five days preceding the analyst recommendations divided by the average daily shorting volume during the 4-month sample period, all minus 1. I implicitly assume that the average daily short selling ratio in the sample period is a fair representation of the firm’s typical daily level of short selling. More specifically, a stock’s average daily abnormal short selling during the five days prior to the release of analyst recommendation, ABSS (-5,-1), is measured as ABSS (5,1)
AverageSS (5,1) 1 AverageSS
(1)
Where the average SS (-5,-1) is the average daily shorting volume during the five days prior to the analyst recommendation, and the average SS is the average daily shorting volume during the sample period. Use of a five-day pre-recommendation interval is a reasonable choice when both the short-term pattern of short selling and the limited market timing capability of informed short sellers are concerned. On the one hand, short selling unrelated to the recommendations may be falsely incorporated into the analysis if a longer time-period than five days is adopted. On the other hand, it is unlikely that short selling related to the recommendations occurs only on one or two days prior to it, since informed short sellers may have strong incentives to engage in a number of smaller trades over a relatively longer period instead of a single large trade concentrated at one time. As a robustness check, I also conduct regressions based on a three-day
and a one-day pre-recommendation interval and results are consistent with the choice of five-day interval. The following OLS regression model that is similar to the one used by Christophe, Ferri, and Angel (2004) is adopted to test whether abnormal short selling is associated with short sellers’ private information regarding the upcoming analyst recommendations. ABSS (5,1) 0 1 INFO 2 RET (5,1) 3 AVOL(5,1) ,
(2)
where INFO is the two-day stock return from the closing prices of day -1 to +1, RET(-5,-1) is the stock return from the closing price of day -5 to -1, and AVOL(-5,-1) is the abnormal trading volume during the period from day -5 to day -1, measured as the average daily trading volume for five days prior to the analyst recommendation divided by the average daily trading volume for the non-recommendation period, all minus 1. A two-day stock return after the recommendation changes is chosen here as the proxy for information content associated with analyst recommendation changes because the market reaction to the recommendation reveals whether the recommendation contains a surprise. Hence, a negative two-day return following the recommendations indicates that the market views the analyst recommendation as a negative surprise, and a positive one implies a positive surprise. A statistically significant and negative coefficient on INFO would reveal that the abnormal short sales increase prior to a downgrade recommendation, and that short sellers are capitalizing their private information associated with forthcoming analyst recommendation changes. A nonnegative and insignificant coefficient would reject the hypothesis that private information about the upcoming downgrade recommendation drives short selling prior to the release of the recommendation.
It is possible that upward or downward changes in the stock price affect the level of short selling in the days leading up to the release of analyst recommendations. For instance, a prerecommendation increase in stock price might induce “contrarians” to short sell “over-valued” stocks. Thus, RET(-5,-1) is included in the regression model to control for “contrarians” short selling behavior. To account for potential contemporaneous correlation between abnormal short selling and trading volume, VOL(-5,-1) is included as the control variable to account for the possibility that increased short selling is caused by increased trading volume. VOL(-5,-1) is measured as the percentage difference between the average daily volume in the 5 day interval and the average daily volume during the sample period.
4.2 Regression Results In this section, the OLS model specified in Equation (2) is estimated separately for pilot and control stocks over time periods before (pre-SHO), during (during-SHO), and after (post-SHO) the suspension of uptick rule restrictions. Results are presented in Table 3. (Insert Table 3 about here)
In the pre-SHO period, pilot and control stocks face the same uptick rule restrictions. Left two columns (columns 1 and 2) in table 3 report the results of the estimation of Equation (2) during the pre-SHO period. For both pilot and control samples, the coefficient on INFO is significantly negative, suggesting that higher levels of abnormal short selling are related to a larger negative surprise, and lower levels of abnormal short selling are related to a larger positive surprise. Both tipping and prediction hypotheses developed in Christophe, Ferri, and Heieh
(2008) can be used to explain the negative relation between pre-recommendation abnormal short selling and recommendation related market surprises. While tipping hypothesis states that informed short sellers front-run analyst recommendation changes based on their private held information about the upcoming recommendations, prediction hypothesis argues that short sellers predict analyst recommendation changes based on their skillful analysis on publicly available information. Following Christophe et al. (2008) that provides evidence supporting the tipping hypothesis, the results here suggest that informed short sellers were capitalizing their analyst recommendation related private information during the pre-SHO period in both pilot and control stocks.The coefficients on RET(-5,-1) are significantly positive. This is consistent with the observation that short sellers frequently act as contrarians, selling into an increasing market. It also shows that when markets are uptick rule restricted it is difficult to short sell when the market is declining. The coefficients on AVOL(-5,-1) are significantly positive as well. Since the dependent variable is already normalized for trading volume, the significance of this coefficient means that high trading volume induces an increase in short selling activity. In addition, the sings, magnitudes, and significance levels of the coefficients for the control samples are very similar to the coefficients for the pilot sample during the pre-SHO period. In the during-SHO period, pilot stocks do not face uptick rule restrictions while control stocks do. If the uptick rule is a binding constraint on short selling activity, then we would expect short selling of pilot stocks to differ from that of control stocks in the during-SHO period. Since control stocks are subject to the same uptick rule restrictions over both pre- and during-SHO periods, we would expect that short selling behavior in the during-SHO period to be similar to that in the pre-SHO period. Columns 3 and 4 present regression results of the estimation of Equation (2) for the during-SHO period. All of the coefficients on the control sample regressions
during-SHO are of the same sign, significance, and similar magnitude as those on the pre-SHO regressions. The coefficients on INFO remain negative and significant indicating that in the during-SHO period, short sellers of control stocks managed to short more in anticipation of a larger downgrade surprise. However, the coefficient on INFO for the during-SHO pilot samples is no longer significantly negative, indicating that once the uptick rule is suspended in pilot stocks, abnormal short selling no longer relate to the recommendation announcement surprise. In other words, informativeness of short selling in pilot stocks disappeared when the uptick rule was suspended in the during-SHO period. This is a surprising result given that the removal of uptick rule restrictions would make it easier for short sellers to enter trades to exploit any private information, strengthening the relation between ABSS(-5,-1) and INFO. This result could arise from two different mechanisms. First, the removal of uptick rule restrictions might make the market more efficient, impounding the short sellers’ privately held information into prices at a higher speed and with a greater magnitude, thus spoils the market surprises and reduces the incentives for informed short seller to take advantage of their privately held information associated with forthcoming recommendation changes. I define this as the shorting efficiency hypothesis. Alternatively, informed short sellers might have chosen not to trade on their private information in pilot stocks in order to avoid potential regulatory or market scrutiny; with plenty of non-pilot program stocks available, why risk trading in pilot program stocks? The SEC’s analysis has mentioned this possibility, “…it is possible that traders might behave differently if a rule were permanently and completely removed than if it is only temporarily or incompletely removed (during the SHO-pilot program)…”. Therefore, it is possible that informed short sellers are on good behavior if they believe that heightened scrutiny during the SHO-pilot program
increases their chances of being caught. I define this as the regulatory concern hypothesis. The reason behind the non-significant coefficient on INFO for during-SHO pilot stocks has regulatory implications; if informed short sellers were avoiding pilot stocks in the during-SHO period, then the SEC’s conclusion that the removal of uptick rule restrictions improves market efficiency without an increased incidence of predatory short selling behavior may be unfounded. If, on the other hand, short selling more efficiently imparted negative information into stock prices for pilot stocks during the post-SHO period, then the SEC’s conclusions were likely correct. In the post-SHO period, the uptick rule is permanently removed for both pilot and control stocks. Regarding the informativeness of short selling prior to analyst recommendations, there are three potential scenarios to consider. First, if the removal of the uptick rule would make it easier for short sellers to enter trades to exploit any private information, then we would observe a stronger relation between ABSS(-5,-1) and INFO for both pilot and control stocks. Second, if the removal of the uptick rule makes the market more efficient and spoils the market surprises, then we would expect a weak relation between ABSS(-5,-1) and INFO for both pilot and control stocks. Third, if the regulatory concern motivates short sellers to curb their information driven short selling in pilot stocks during the SHO-pilot program, then after SEC concludes the SHO program short sellers would resume their information driven short selling pilot stocks, which are no longer subject to SEC’s close monitoring. Column 5 and 6 in Table 3 reports regression results during the post-SHO period. It shows that coefficients on INFO in both pilot and control samples are significantly negative, suggesting a similar pattern of information driven short selling as observed in the pre-SHO period. Notably, coefficients on RET(-5,-1) are no longer
significant for both pilot and control samples, suggesting that short sellers act as contrarians less frequently when it is easier to short sell in a declining market.
4.3 Robustness Checks
To ascertain that the different private information driven short selling behavior between the pilot and control samples in the during-SHO period are not caused by statistical flukes or by other systematic differences between the pilot and control samples, I conduct several additional tests to demonstrate robustness. First, the regression has thus far ignored the possibility of alternative investment vehicles for an informed investor to trade on negative private information. Chen and Singal (2003) and Senchark and Starks(1993) suggest that the level of short interest is affected by the availability of exchange traded options. Options provide investors with a direct alternative to capitalizing their negative private information when short sales are restricted. The lack of a reliable relationship between abnormal short selling and recommendation downgrade may reflect a greater use of options by sophisticated investors for pilot stocks in the sample period. To account for the impact of exchange-traded-options on the analysis, I obtain option availability data from CBOE (Chicago Board of Option Exchange), and re-estimate Equation (2) for both pilot and control stocks with and with no options in the during-SHO period. Results reported in Table 4 shows that, for both pilot stocks with options and with no options, the coefficients of β1 are consistently not distinct from zero, suggesting that option availability is not a factor that induces short sellers to bypass the information-driven trading opportunities preceding analyst recommendations. ( Insert Table 4 about here )
Second, previous studies addressing the relationship between short selling activities and the subsequent stock returns suggest the necessity of controlling for firm size and book-to-market ratio. For example, Asquith, Pathak, and Ritter (2005) find a small firm effect associated with short selling and stock return relationship. Moreover, Chirstophe et al. (2004) have highlighted the “torpedo” effect, which states that even slight negative information can lead growth stock to suffer a large drop in price.1 When the recommendation is negative, short sellers are more likely to target the growth stocks with low book-to-market ratios. It is possible that the different price response patterns in the pilot and control samples are driven by different compositions of value stocks and growth stocks in both samples. Therefore, the implication here is that the absence of a reliable relationship between short selling and subsequent stock returns following the information event may be biased by a value stock dominated pilot stock sample. Therefore, we run the pilot sample OLS regressions for three sub-samples characterized by the ranks of firm size and book- to-market ratio. Table 5 presents the pilot sample OLS regression results when firm size and book–to-market ratios are controlled for. It shows that the coefficients of β1 are far from being significant across different size categories. Also, the estimated value of coefficients on β2 decline when the firm size decreases, indicating that passive short sellers are more likely to aim at larger firms than small firms. This makes sense because large firms have more shares available for shorting than small firms. In addition, it also shows that the growth stock sample has a larger positive coefficient of β2 than the value stock sample for pilot stocks. More specifically, for one percentage point increase in abnormal short sales, growth stocks experience a 2.39% price increase while value stocks experience a 1.73% price increase over a two-day period (0,+1), representing a reverse “torpedo” effect for pilot stocks. In sum, the results in Table
1
Skinner and Sloan (2002)
5 reinforce the view that informed short sellers forego the opportunity to trade on their private information associated with pilot stocks. The results also suggest that passive short sellers are more likely to target large stocks and growth stocks when stock prices are in an upward trend, a typical contrarians’ short selling strategy. (Insert Table 5 about here) Thirdly, in the previous analyses, a five-day pre-recommendation interval was chosen arbitrarily. The potential shortcoming of choosing such a long period is that when the uptick rules are relaxed, informed short sellers may adopt a less risky strategy that concentrates on short selling in a shorter period prior to the downgrade. Thus a five-day period may dilute the significance of shorting activities prior to the downgrade. In this section, regressions are run with two alternative specifications for pilot stocks on the NYSE during the sample period from May to December 2005. The first one uses a three-day period prior to the recommendation in the regression, and the second one adopts a one-day period. ABSHO(3,1) 0 1 RET (0,1) 2 RET (3,1) 3 ABVOL (3,1) , (3) ABSHO(1) 0 1 RET (0,1) 2 RET (1) 3 ABVOL (1) ,
(4)
The regression results are presented in Table 6. It shows the results with a three-day period prior to the recommendation and provides no evidence that short sellers concentrate their shorting in a three-day period before the recommendations for pilot stocks. It also rules out the possibility that short sellers focus their shorting just on the day before the recommendation. This additional robustness test confirms the results that informed short sellers tend not to trade on their private information in pilot stocks. (Insert Table 6 about here)
Finally, we run regressions based on another during-SHO period, from May to December in 2005, an eight-month period immediately after SEC’s implementation of the suspension of the uptick rule pilot stocks. The results detailed in Table 7 are consistent with what we find in Table 4. (Insert Table 7 about here)
4.4. The Shorting Efficiency Hypothesis vs. Regulatory Concern Hypothesis
It is a surprise to discover that informativeness of short selling in pilot sample disappeared when uptick rules were suspended during the post-SHO period. This result could arise from at least two different mechanisms. First, the removal of uptick rule restrictions might make the market more efficient, impounding the short sellers’ privately held information into prices at a higher speed and with a greater magnitude, thus spoils the market surprises and reduces the incentives for informed short seller to take advantage of their privately held information associated with forthcoming recommendation changes. We define this as the shorting efficiency hypothesis. Second, informed short sellers might have chosen not to trade on their private information in pilot stocks in order to avoid potential regulatory or market scrutiny; with plenty of non-pilot program stocks available, why risk trading in pilot program stocks? In fact, the SEC’s analysis has mentioned this possibility, “…it is possible that traders might behave differently if a rule were permanently and completely removed than if it is only temporarily or incompletely removed (during the pilot program)…”. Therefore, it is possible that informed short sellers might be on good behavior if they believe that heightened scrutiny during the Pilot increases their chances of getting caught. We define this as the regulatory concern hypothesis. The reason behind the non-significant
coefficient on INFO for post-SHO pilot stocks has regulatory implications; if informed short sellers were avoiding pilot stocks during the post-SHO period, then the SEC’s conclusion that the removal of uptick rule restrictions improves market efficiency without an increased incidence of predatory short selling behavior may be unfounded. If, on the other hand, short selling more efficiently imparted negative information into stock prices for pilot stocks during the post-SHO period, then the SEC’s conclusions were likely correct. Under the shorting efficiency hypothesis, we would expect abnormal short sales prior to the downgrades have greater power to drive down simultaneous returns significantly in pilot stocks than in control stocks during the post-SHO period. Also, since private information has been impounded into pilot stock prices before the recommendation change announcement date, we would expect the market response to downgrade recommendation would be much smaller in pilot stocks than in control stocks. To evaluate the shorting efficiency hypothesis, we conduct two separate tests. First, we begin with studying the relationship between pre-downgrade abnormal short sales and simultaneous 5-day pre-recommendation stock returns by estimating the following equation: RET(-5,-1)=a + b1* ABSS(-5,-1)*Pilot Dummy + b2* ABSS(-5,-1)* Control Dummy (5) If short sales in pilot stocks become more efficient in driving down stock prices, then we would expect a larger significant negative coefficient of b1 than b2. Table 8 presents the results. (Insert Table 6 about here) It shows that the signs on the coefficients on both ABSS(-5,-1)*Pilot Dummy and ABSS(-5,-1)*Control Dummy are either positive or zero, rejecting the shorting efficiency hypothesis. Under the regulatory concern hypothesis, if informed short sellers choose to leave the market, then the remaining short sellers would be relatively uninformed; shorting more in an up
market since they tend to be contrarians. Results in Table 4 provide some supports for this hypothesis. Comparing the coefficients of RET(-5,-1) for the pilot and control samples in the post-SHO period, I find that this coefficient is significantly greater in pilot stocks than in control stocks, suggesting that abnormal short sales prior to recommendation changes are more likely to be induced by “contrarian” short selling in pilot stocks than in control stocks. The direct way to test the regulatory concern hypothesis is to examine whether informed short sellers resume their private information driven short selling behavior after June 2007, when the uptick rule restrictions be permanently removed by the SEC and Regulation SHO mandated short selling data collection be ceased. Although the collection of short sale data is no longer mandatory after June 2007, the NYSE Data Service continues to collect the data. I obtain a fourmonth short sale data for a period from July to October 2007 from the NYSE Data Service. Equation (2) was estimated for a four-month period from July to October 2007. Regression results presented in Columns 5 and 6 in Table 4 shows that coefficients on INFO now are significantly negative for both pilot and control stocks, suggesting that abnormal short selling in pilot stocks has been restored its informativeness after the SEC removed uptick rule restrictions permanently.
5. Conclusions
This paper examines short sales prior to analyst recommendation changes for pilot stocks that are exempt from uptick rule restrictions and for control stocks with binding uptick rules around the introduction of the pilot program. I find evidence that informativeness of short sales has been disappeared in pilot stocks during the pilot program in 2005. In particular, informed
short sellers were capitalizing on their private information associated with upcoming recommendation changes in the control sample. However, such behavior seems to disappear in the pilot sample when uptick rules were suspended during the pilot program. The results are robust when accommodating a number of potentially confounding factors, such as exchangetraded-option, firm size, book-to-market ratio, different time interval prior to recommendation changes, and different calendar periods. This is a surprising result because it would be easier for short sellers to conduct such information-driven trading when the uptick rules were suspended for pilot stocks. The shorting efficiency hypothesis and the regulatory concern hypothesis are developed to provide explanations for this surprising result. The shorting efficiency hypothesis states that the removal of uptick rules improve the efficiency with which informed short sellers’ private information are impounded into stock prices. Prior to the recommendation downgrades, stock prices might have already been driven down so significantly that it provides no incentives for informed short sellers to short on their private information, and the market surprises of downgrades might have been spoiled by uptick rule unrestricted short selling. The regulatory concern hypothesis argue that during the pilot program, informed short sellers were aware of the regulatory scrutiny from both the SEC and the public, and were very cautious in trading on their private information associated with forthcoming analyst recommendation downgrades. Additional tests aiming at evaluating these two hypotheses show that the regulatory concern is more consistent with the data. I do not find evidence that market surprises were spoiled in pilot stocks when we examine the intra-day market response to downgrades. Using short sale data for a period from July to October 2007, I find that informativeness of short selling has been restored
in pilot stocks since June of 2007 when the SEC universally removed uptick rules and ceased mandated short sale transactions reporting. It is safe to conclude that the Regulation SHO-pilot program, which generates the SHO data, may not be a clean natural experiment. Although pilot and control samples are generally matched based on many firm characteristics, there is a propounding difference between two samples as far as the informed short sellers’ behavior concerned, making any comparison between the pilot and control samples problematic ones. This further casts doubts on whether the removal of uptick rule restrictions by the SEC in July 2007 is a proper decision. I propose reexamining the impact of uptick rules on many aspects of market qualities by considering the systematic difference in informed short sellers’ behavior between the pilot and control samples in the future.
References Alexander, Gordon J. and Mark A. Peterson (2008), The Effect of Price Tests on Trader Behavior and Market Quality: An Analysis of Reg SHO, Journal of Financial Markets, Vol. 11, No. 1, 84-111
Asquith, Paul, Parag A. Pathak, and Jay R. Ritter (2004) Short Interest, Institutional Ownership, and Stock Returns, Journal of Financial Economics, Vol. 78, No. 2, 243-276
Boehmer, Ekkehard, Charles M. Jones, and XiaoYan Zhang (2008), Which Shorts are Informed? Journal of Finance, Vol. 63, No.2, 491-527 Boehmer, Ekkehart, and Julie J. Wu (2012), Short Selling and the Informational Efficiency of Prices, Review of Financial Studies, Vol. 26, Nov.2, 287-322
Christophe, Stephen, Michael G. Ferri, and James J. Angel (2004), Short-Selling Prior to Earnings Announcements, The Journal of Finance, Vol LIX, No.4, August 2004
Danielsen, Bartley R. and Sorin M. Sorescu (2001), Why do Option Introductions Depress Stock Prices? An Empirical Study of Diminishing Short Sale Constraints, Journal of Financial and Quantitative Analysis 36, 451-484
Diamond, Douglas W, and Robert E. Verrecchia (1987), Constraints on Short-Selling and Asset Price Adjustment to Private Information, Journal of Financial Economics, 18 (1987) 277-311 Diether, Karl B., Christopher J. Malloy, and Anna Scherbina (2002), Difference of Opinion and the Cross Section of Stock Returns, The Journal of Finance, Vol. LVII, No. 5, October 2002 Diether, Karl, Kuan Hui Lee, and Ingrid M. Werner (2008), Can Short-Seller Predict Returns? Daily Evidence, The Review of Financial Studies, Vol. 22, No. 2, 575-607 Diether, Karl, Kuan Hui Lee, and Ingrid M. Werner (2009), It’s SHO time: Short-Sale Price tests and Market Quality, Journal of Finance, Vol. 64 No. 1, 37-74
Hong, Harrison and Jeremy C. Stein (2003), Differences of Opinion, Short Sales Constraints, and Market Crash, The Review of Financial Studies, Summer 2003 Vol.16, No.2 487-525
Jarrow, Robert (1980), Heterogeneous Expectations, Restrictions on Short Sales, and Equilibrium Asset Prices, Journal of Finance, Vol. 35, No.5 (Dec. 1980) 1105-1113
Jones, Charles M. and Owen A. Lamont (2002), Short Sales Constraints and Stock Returns, Journal of Financial Economics 66, 207-239 Miller, Edward M.(1977), Risk, Uncertainty, and Divergence of Opinion, Journal of Finance, 32
Nagel, Stefan (2005), Short Sales, Institutional Investors and the Cross-Section of Stock Returns, Journal of Financial Economics 78 (2005) 277-309
Office of Economic Analysis of the SEC (2007), Economic Analysis of the Short Sale Price Restrictions under the Regulation SHO Pilot, The SEC Working Paper Reed, Adam V. (2007), Costly Short Selling and Stock Price Adjustment to Earnings Announcements, Working Paper, University of North Carolina Senchack, A. J., Jr. and Laura Starks (1993), Short-Sale Restrictions and Market Reaction to Short Interest Announcements, Journal of Financial and Quantitative Analysis, Vol. 28, No. 2, June 1993 Skinner, Douglas J., and Richard G. Sloan (2002), Earnings Surprises, Growth Expectations, and Stock Returns or don’t let an earnings torpedo sink your portfolio, Review of Accounting Studies, 7, 289-312
Figure 1 Time line of SEC Regulation SHO-Pilot Program
The Regulation SHO Pilot program went in effect on May 2, 2005. The uptick rule is temporarily relaxed for pilot stocks, but not for control stocks. The pilot program was initially set to end in May 2006, but was extended by the SEC until July 2007.
Pre-SHO Period
During-SHO Period
October-07
On July 2, 2007, the uptick rule is permanently terminated by the SEC for both pilot and control stocks.
September-07
August-07
July-07
May-07
June-07
April-07
March-07
January-07
February-07
December-06
October-06
November-06
September-06
July-06
August-06
May-06
June-06
April-06
March-06
February-06
January-06
December-05
October-05
November-05
September-05
July-05
August-05
May-05
June-05
April-05
March-05
February-05
Short selling data collection begins January 2005. Both pilot and control stocks are subject to the same uptick rule restrictions.
Post-SHO Period
Table 1 Descriptive Summary of Short Selling Activities This table presents daily short selling activities for both pilot and control stocks for Pre-SHO, during-SHO, and postSHO periods. Short sale ratio equals short sale volume divided by the trading volume. Short order size is the number of shares in each shorting order. The number of short orders is the number of short sale transactions for each stock. Short sale volume is total number of shares shorted for each stock on each given day, and is measured in thousand shares. The differences of shorting measures between the pilot and control samples are measured as Difference, followed by t-stat. *** is significant at 1% level. Pilot Mean (1)
Stderr
Control Mean (2)
Stderr
Difference (3)=(1)-(2)
t-stat
Short Sale Ratio
21.2%
0.0028
21.74%
0.0019
-0.55%
-1.57
Size of Short Orders (shares)
437.6
10.5
446
5.3
-8.4
-0.71
Number of Short Orders
459
14.2
447
10
13
0.73
Short Sale Volume (1,000 shares)
201
1.6
199.2
1.34
1.8
0.86
Short Sale Ratio
23%
0.0025
21%
0.0017
2%***
6.83
Size of Short Orders (shares)
337
9.7
552
3.4
-215***
-20.9
Number of Short Orders
623
17
366
8.8
257***
13.44
Short Sale Volume (1,000 shares)
210
1.4
202
1.01
8.08***
4.67
24.2%
0.003
23.91%
0.002
0.29%
0.83
Size of Short Orders (shares)
352
10.7
360
8.9
-16
-0.57
Number of Short Orders
709
12
695
13.1
11
0.79
Short Sale Volume (1,000 shares)
250
1.4
250
1.01
-0.63
-0.37
Pre-SHO period (January to April 2005)
During-SHO period (January to April 2006)
Post-SHO period (July to October 2007) Short Sale Ratio
Table 2 Descriptive Summary of Analyst Recommendation Changes Table 2 presents the number of downgrades and non-downgrades for pilot and control stocks by month in panel A, and magnitude of recommendation changes for both pilot and control samples in panel B. Analyst recommendation changes are from First Call. Observations are limited to those made by top 20 brokerage firms ranked by Institutional investors.
Panel A Number of Recommendation Changes by Month Month January, 2005 February, 2005 March, 2005 April, 2005 January, 2006 February, 2006 March, 2006 April, 2006 July, 2007 August, 2007 September, 2007 October, 2007
Pilot Downgrades 42 42 36 24 40 55 38 26 37 28 35 44
Sample Non-downgrades 95 70 82 77 101 91 75 75 78 75 80 103
Total 137 112 118 101 141 146 113 101 115 103 115 147
Control Downgrades 83 86 59 59 86 85 72 61 73 80 67 95
Sample Non-downgrades 181 153 163 141 196 195 130 143 129 143 198 196
Total 264 239 222 200 282 280 202 204 202 223 265 291
Panel B Magnitude of Recommendation Changes by Month Month Jan 2005 Feb 2005 Mar 2005 Apr 2005 Jan 2006 Feb 2006 Mar 2006 April 2006 July 2007 Aug 2007 Sep 2007 Oct 2007
Pilot (1) -1.39 -1.48 -1.36 -1.63 -1.32 -1.47 -1.39 -1.47 -1.46 -1.47 -1.34 -1.39
Downgrades Control Difference (2) (3)=(1)-(2) -1.48 0.10 -1.43 -0.05 -1.46 0.10 -1.46 -0.17 -1.27 -0.05 -1.38 -0.10 -1.43 0.05 -1.46 -0.01 -1.50 0.04 -1.38 -0.10 -1.46 0.12 -1.43 0.05
t-stat 1.06 -0.49 0.93 -0.98 -0.44 -0.92 0.39 -0.07 0.30 -0.92 1.27 0.39
Pilot (4) 1.11 1.11 1.05 1.04 0.85 0.79 1.16 0.98 0.93 0.79 0.82 1.16
Non-Downgrades Control Difference (5) (6)=(4)-(5) 0.96 0.14 1.18 -0.06 0.96 0.09 0.91 0.12 0.83 0.02 0.82 -0.04 1.05 0.10 0.93 0.05 1.06 -0.13 0.82 -0.04 0.92 -0.11 1.05 0.10
t-stat 1.54 -0.55 0.85 1.19 0.16 -0.36 1.07 0.39 -1.18 -0.36 -1.05 1.07
Table 3 Informativeness of Short Sales Prior to Analyst Recommendation Changes
ABSS (5,1) 0 1 INFO 2 RET (5,1) 3 AVOL(5,1) , This table presents the results of OLS regressions based on the above equation. The dependent variable ABSS(-5,-1) is the abnormal short sales prior to the analyst recommendations, measured as the average daily short selling in the five-day prerecommendation period divided by the average daily short selling in the non-recommendation period, all minus 1. INFO, the private information held by informed short sellers prior to the recommendation change is the stock’s 2-day percentage return following the recommendations. The control variable RET(-5,-1) represents the stock’s percentage return measured form the closing price on day -5 through end of day -1. The control variable AVOL(-5,-1) is the stock’s abnormal volume in a five-day pre-recommendation period, measured as the average daily volume in the pre-recommendation period divided by the average daily volume in the non-recommendation period, all minus 1. White’s (1980) robust standard errors are presented in parentheses below the coefficients. *,**, and *** are statistically significant at the 10%, 5%, and 1% level, respectively.
Pre-Pilot
During-Pilot
Post-Pilot
Pilot
Control
Pilot
Control
Pilot
Control
0.0002
0.014
-0.001
0.012**
0.030***
0.035***
(0.013)
(0.010)
(0.014)
(0.008)
(0.009)
(0.006)
-0.822***
-0.488**
-0.206
-0.376***
-1.004***
-0.939***
(0.365)
(0.249)
(0.373)
(0.101)
1.919***
2.174***
1.769***
1.632***
(0.189) -0.011
(0.138) 0.066
(0.383)
(0.294)
(0.253)
(0.402)
(0.151)
(0.102)
0.826***
0.882***
0.790***
0.783***
0.906***
0.889***
(0.029)
(0.030)
(0.019)
(0.021)
(0.020)
(0.013)
Adj. R square
0.56
0.69
0.54
0.59
0.61
0.59
# of observations
468
925
501
968
480
981
Intercept INFO RET (-5,-1) AVOL(-5,-1)
TABLE 4 CONTROLLING FOR EXCHANGE TRADED OPTIONS IN DURING-SHO PERIOD The sample period is from January to April 2006. This table details the results of OLS regression and was fitted to subsamples determined by availability of exchange traded option at the time of analyst recommendations. Robust standard errors are in parentheses below the coefficients. *,**, and *** are statistically significant at the 10%, 5%, and 1% level, respectively.
Stocks with options
Intercept
INFO RET(-5,-1)
AVOL (-5,-1)
Adj R # of obs
Stocks with no options
Pilot
Control
Pilot
Control
0.002
0.014
0.011
0.038**
(0.012)
(0.009)
(0.022)
(0.018)
-0.101
-0.182
-2.277
-0.544**
(0.757)
(0.123)
(2.201)
(0.287)
-0.211
0.242
3.952*
0.27
(0.917)
(0.369)
(2.429)
(0.639)
0.764***
0.764***
0.563***
0.807***
(0.048)
(0.044)
(0.091)
(0.048)
0.613
0.579
0.416
0.558
177
416
290
516
Table 5 Controlling for Firm Size and market to book ratio for Pilot Stocks The sample period is from January to April 2006. This table reports the results of OLS regression for pilot stocks controlling for firm size and Market-to-book ratio. *,**, and *** are statistically significant at the 10%, 5%, and 1% level,
respectively.
Intercept
INFO
RET (-5,-1)
AVOL (-5,-1)
Adj R square
Large firms
Median firms
Small firms
Growth firms
Median firms
Value firms
-0.004
-0.031
0.028
0.000
-0.026
0.016
(0.016)
(0.019)
(0.025)
(0.020)
(0.017)
(0.023)
0.171
0.049
0.245
0.253
-0.481
0.315
(0.679)
(1.027)
(0.658)
(0.843)
(0.617)
(0.759)
2.340***
2.310***
0.929
2.394***
1.972***
1.703**
(0.593)
(0.638)
(0.731)
(0.817)
(0.484)
(0.957)
0.719***
0.574***
0.653***
0.623***
0.810***
0.629***
(0.070)
(0.105)
(0.115)
(0.117)
(0.059)
(0.110)
0.708
0.463
0.447
0.486
0.721
0.461
Table 6 Shorter Periods prior to Analyst Recommendations This table details the results of regression by adopting two shorter time periods for ABSHO. ABSHO (-3,-1) is the abnormal short selling during the 3-day period prior to the recommendation changes and ABSHO(-1) is the abnormal short selling the day prior to the recommendation changes. *,**, and *** are statistically significant at the 10%, 5%, and 1% level,
respectively.
ABSHO(3,1) 0 1 RET (0,1) 2 RET (3,1) 3 ABVOL (3,1) , (3) (4) ABSHO(1) 0 1 RET (0,1) 2 RET (1) 3 ABVOL (1) ,
Dependent variables Intercept
ABSS(-3,-1) 0.016
Dependent variables Intercept
(0.015) INFO (RET (0,+1))
0.065
2.303**
INFO (RET (0,+1))
0.549***
RET(-1)
0.45
3.286 (4.463)
AVOL(-1)
(0.075) Adj R square
0.387 (1.08)
(0.917) AVOL(-3,-1)
0.132 (0.034)
(0.603) RET(-3,-1)
ABSS(-1)
0.421*** (0.099)
Adj R square
0.38
Table 7 Alternative during-SHO period (May to December 2005) The sample period is from May to December 2005. This table details the results of OLS regression based on equation (2). Robust standard errors are in parentheses below the coefficients. *,**, and *** are statistically significant at the 10%, 5%, and 1%
level, respectively.
Intercept
Pilot
Control
-0.002
0.0142**
(0.014)
(0.008)
0.1055
-0.410**
(0.348)
(0.120)
1.861***
1.051***
(0.280)
(0.309)
0.660***
0.758***
(0.022)
(0.034)
Adj. R square
0.53
0.58
# of observations
804
1964
INFO RET (-5,-1) AVOL(-5,-1)
Table 8 Shorting Efficiency Hypothesis Test This table report OLS regression results based on the following equation. RET(-5,-1)= Intercept + b1* ABSS(-5,-1)* Pilot Dummy + b2* ABSS(-5,-1)*Control Dummy Diff is the difference between the coefficient of ABSHO(-5,-1)*Pilot Dummy and the coefficient of ABSHO(-5,1)*Control Dummy, followed by the t-stats. ***, **, and * are significant at 1%, 5%, and 10% level respectively.
Panel A. The Full Sample Variable Intercept DuringSHO
Pre-SHO
Estimate 0.005***
Stderr 0.001
t-stat 4.14
abshopilot abshocontrol
0.011*** 0.027***
0.004 0.002
2.82 10.84
Intercept abshopilot abshocontrol
-0.002 0.015*** 0.015***
0.001 0.005 0.003
-1.15 2.72 5.19
R2
Adj
Diff *
t-stat
0.042
-0.016***
-3.64
-0.0001
-0.02
0.024
Panel B. The Subsample of Downgrades
DuringSHO
Pre-SHO
Variable Intercept
Estimate 0.006
Stderr 0.003
t-stat 2.16
abshopilot abshocontrol Intercept abshopilot abshocontrol
0.014 0.048*** -0.002 0.006 0.007
0.008 0.006 0.003 0.009 0.005
1.67 8.34 -0.67 0.61 1.37
2
Adj R 0.076
Diff *
t-stat
-0.035***
-3.50
-0.002
-0.15
0.005
