SBi = amount of new equity as a fraction of new debt.
dumi = dummy variable which is 0 for CBs and 1 for WBs.
The aim here is to explain the cumulative abnormal return on event day +1 from two
variables which are supposed to measure structural differences between individual issues and a
dummy variable that is supposed to measure the effect of WBs versus CBs over and above the
structural differences between the issues.
The null-hypothesis is that β3 = 0. This would mean that shareholders do not perceive CBs
and WBs to be inherently different. Using a Pecking Order framework and the summary of
empirical research on the announcement effects of capital structure changes presented in Dann and
Mikkelson (1984), we can make hypotheses for the other two explanatory variables. According to
the Pecking Order Theory firms prefer to finance new projects with internal capital. If firms have
to raise external capital then they prefer to issue new debt first and they prefer to issue new equity
last. Myers and Majluf (1984) give an explanation for this behavior assuming that management has
more information about the firm’s value than potential investors. One of the conclusions by Myers
and Majluf (1984) is that when managers have superior information, the stock price is expected to
fall when new equity is issued, while it is not expected to fall when new debt is issued5. For our
purposes, this conclusion is most important.
The first variable Snew measures the maximum amount of new equity involved in the issue
in relation to the existing amount of equity. Since the issue of common stock is usually received
with a negative reaction by existing shareholders, it is expected that β1 will be negative. This
would be in line with The Pecking Order Theory.
The second variable SB measures the amount of new equity in relation to the amount of
new debt. Since Dann and Mikkelson (1984) find, in line with the Pecking Order Theory, a
negative relation for the announcement of new equity and a zero relation for the announcement of
new debt, we expect that the coefficient β2 will also be negative.
A last point to mention is that simply using OLS to estimate equation (3) will probably not
be very efficient. Since the abnormal returns are calculated as the residuals from the market model
in (1), they essentially measure the nonsystematic risk of the firm. Since nonsystematic risk will be
different for different firms, the variance of the error term in (1) will not be the same for each i.
The error term in (3) also (in part) reflects this nonsystematic risk. Therefore, this will have as a
consequence that the variance of the error term in (3) is not constant over i either. In other words,
the error terms in (3) are heteroskedastic.
This is conclusion 5 of their paper. See Myers and Majluf (1984, page 220).