Click to edit Master text styles
Second level
Third level
Fourth level
Fifth level
Specifically talk about rule 105 that prohibited traders from covering short sales made within 5 days of the offering with shares obtained in the offering.
transactions level short sale data, aggregated to the daily level allows to examine changes in daily short selling activity throughout the SEO period in detail.
This table reports the results of statistical tests of abnormal trading activity during the pre announcement period (H1A)
Mean announcement day abnormal return -2.3% and is statistically significant at the 1% level consistent with the prior literature. (red box)
No evidence that either short selling or volume immediately before the announcement is greater than benchmark periods. (blue box)
The mean values of ABSS (abnormal SS) and ABVOL (abnormal trading volume) are insignificant and the median values are significantly negative. So there is some evidence that the decrease in SS is disproportionately smaller than the total trading volume.
The mean pre-announcement value of ABRELSS (Abnormal relative SS) is significant but the median value of the same is insignificant (green box)
These results suggest that although there may be relatively higher SS in a subset of stocks, over all these results do not agree with the hypothesis that SS will be relatively high before the announcement date.
To test this Henry and Koski regress abnormal short selling in the 5 trading days prior to the announcement date on abnormal returns on the announcement date, abnormal returns in the 5 trading days prior to the announcement date and abnormal trading volume in the 5 trading days prior to the announcement date.
If abnormal short selling and abnormal returns are negatively correlated then the coefficient on abnormal returns on the trading day should be negative.
Although the coefficient is negative it is not statistically significant.
As a robustness check they rerun the regression with relative short selling on AD-5 to AD-1 and abnormal relative SS on AD-5 to AD-1 as dependent variables and get the same result.
SS in the period prior to the announcement date is not negatively related to announcement returns.
These hypotheses are conditional on the fact that short sellers are especially skilled at processing information contained in an SEO announcement.
Both SS and total volume are abnormally high on the announcement date but neither the mean nor the median value of abnormal relative SS is significant which implies that there does not appear to be a disproportionate increase in SS.
Results of regressing abnormal announcement date SS on contemporaneous returns: 2 of the 3 coefficients are positive and none are significantly negative hence Henry and Koski’s results do not support this hypothesis.
So no evidence of informed short selling before or on the announcement date.
Issue discount: % difference between the SEO offering price and the closing price on the day before the offering.
To test this hypothesis the authors run regressions of SEO issue discounts on various firm and offering characteristics and include measures of short selling intensity prior to the issue date—the explanatory variables are designed to capture various characteristics of the firm or offering that are known to capture discount rate. These regression are similar to Corwin’s 2003 work.
The coefficients on the control variables in the regressions are generally consistent with previous research
CARNeg: the cumulative abnormal return over the 5 days preceding the offering if it is negative is seen to have a significantly negative relation with issue discount. This implies that firms with more negative abnormal returns in the 5 days prior to the issuance have larger underpricing—the authors suggest that these variables could be capturing some of the effects of short selling and its influence on pre offer prices.
The key variable they use to measure short selling intensity ABRELSS. The coefficient of ABRELSS[ID-10, ID-1] is highly significant. Under the 119 Gerard and Nanda model this is consistent with higher levels of manipulative short selling before SEO issue dates being associated with larger issue discounts.
Similar results hold with ABSS[ID-10,ID-1] and RELSS[ID-10,ID-1].
The coefficient on RELSS in the benchmark period is negative and significant in model 5.
In their 1993 work Gerard and Nanda point out that restrictions on SS can be costly to issuers if they constrain informed short sales. This implies that firms that normally have higher level of relative short selling have smaller issue discounts which is consistent with the literature that short sellers usually increase price efficiency.
But firms with higher relative short selling prior to the SEO have larger issue discounts, supporting the view that short selling before the issue date is making prices less informative.
To be able to compare their results with prior literature the authors collect monthly short interest data for their sample of SEOs and run analyses similar to those carried out by Safieddine and Wilhelm. They find that consistent with prior research there is no evidence of manipulative short selling. However they continue to find that the coefficients on daily short selling to be consistently positive and significant.
They find no relationship between monthly short interest rate and issue discount they do find a consistent relationship between daily short selling and issue discount.
The lack of evidence of manipulative trading in prior research is attributed to the lack of power in the short interest data.
The primary conclusion of this section ia that abnormally high levels of SS are related to larger issue discounts.
To test this hypothesis the authors regress post issue performance on various measures of pre issue short selling intensity.
They regress post issue returns measured over one day after issue to 5 days after issue on pre issue SS measured over 1 to 5 days prior to issue, they also include a shelf dummy to distinguish between shelf offering and non-shelf offering and well as an interaction of this dummy with SS measures.
The results of this regression are insignificant at the 5% level. These insignificantly positive coefficients are consistent with either informed short selling or a lack of power in tests.
So the authors don’t really take a call on this hypothesis.
I didn’t include table 7 in the presentation because I thought this might give a better picture but if you look at table 7 in the paper you will see that all the measures of SS intensity increases monotonically as the issue date nears with an abnormal spike in SS activity one day before the issuance. In this figure look at ABRELSS. This suggests that Rule 105 does not effectively prevent short selling prior to SEO issue dates.
To determine whether rule 105 affects informed or manipulative trading strategies the authors separate ABRELSS[ID-10, ID-1] into 2 subsets based on the number of days prior to the issue date. (Model 6 in Table 4)
The results in this model show that there is a significant relationship between pre-issue SS and issue discount for non-shelf offerings both, when constrained and unconstrained by rule 105. They also regress SEO issue discount on various control measures and the daily level of ABRELSS for each of the 7 days prior to the issuance. They find that the relationship between pre-issue SS and the issue discount is the strongest on day -6. The coefficients on ABRELSS 6 and 7 days prior to the issuance is positive and significant. These results are consistent with SS by manipulative traders who adhere to rule 105 and transact before the deadline.
The Coefficient on ABRELSS 2 days before the issuance is also significantly positive.
Rule 105 does not appear to be effective in restricting all manipulative SS right before issuance.