314 B.E. Eckbo et al.
The intuition behind these predictions forARis as follows. Starting with the first line
(uninsured rights), firms with highkprefer to issue using the relatively low-cost unin-
sured rights method. Since there is no inspection, there is also no information conveyed
by the issue decision, thusARur=0. Firms with medium and lowkvalues prefer qual-
ity inspection (Figure 3). Thus, issuers of uninsured rights with medium or lowkhave
necessarily been rejected twice by the inspection, soARur<0.
Second, in the line for the standby rights method, medium-kissuers prefer standbys,
creating a positive market reaction (ARsr>0) due to the positive inspection result.
Low-kissuers prefer private placement (Figure 3). Thus, low-kissuers that issue using
standbys have been rejected by the private placement inspection before accepted by the
standby underwriter inspection. From the market’s point of view, these two inspection
results cancel out, so there is no new information andARsr=0. Similarly, in the line for
the private placement method, medium-kissuers that use private placement have first
been rejected by the standby underwriter, thusARpp=0. Low-kissuers prefer private
placement (Figure 3), so the successful inspection result impliesARpp>0.
We now turn to a summary of the international evidence on SEO announcement re-
turns, and then draw inferences about the theoretical predictions above.
4.4. Evidence on issue announcement returns
Abnormal returns are typically measured over the two-day window [−1, 0] ending with
the public announcement date (day 0), or over the three-day window [−1,+1]. Abnor-
mal return to issuerion daytis typically defined using a simple market model:
γit≡rit−E(rit)=rit−(αi+βirmt), (2)
whereritis the daily stock return in excess of the risk-free rate,rmtis the daily excess
return on the value-weighted CRSP market return, andαandβare estimated during
some pre-event period. For event windows containing multiple periods, the cumulative
abnormal return is found by adding daily abnormal returns. With the market model
estimation, it is important not to “contaminate” the estimate ofαwith the well-known
average stock price runup over the year prior to the typical U.S. stock issue. If this runup
is treated as “normal” then the estimate ofαwill be overstated, resulting in a downward
bias in the estimated abnormal returnγ. One solution to this problem is to estimate the
market model parameters using post-issue stock returns.
Some studies estimateγdirectly by means of a conditional market model,
rit=αi+βirmt+γidt+it, (3)
whereditis a dummy variable that takes on a value of 1 during the event window and
zero otherwise, anditis the regression error term. If the event dummydttakes on a
value of one overωdays in the event window, then the cumulative abnormal return over
the event window isωγi.^33
(^33) SeeThompson (1985, 1995)for details of this event-study approach.