The structural form of our model is as follows:
kS
∆St = α0 + α1∆Bt + α2RS,t + α3 σS,t + α4∆yt + ∑ αi+4∆St-i + εS,t (1)
i=1
kB
∆Bt = β0 + β1∆St + β2RB,t + β3σB,t + β4∆yt + ∑ βi+4∆Bt-i + εB,t (2)
i=1
where the inclusion of lagged endogenous variables has a twofold valence. First,
to be consistent (under the null that the parameters αi and βi are zero for i=1,..,4)
with their stationary nature, the sources of external finance are allowed to evolve
as mean reverting AR stochastic processes. Second, lagged dependent variables
can capture the effects of omitted factors. In fact, it is plausible to think that the
variables included in the model are not the only determinants of the sources of ex-
ternal finance. Moreover, this specification of the model should ensure residuals
not serially correlated. The hypothesis we want to test is that no relationship exists
between issuance of stocks and corporate bonds and financial market volatilities.
There are two potential problems related to this study. The first is that the different
sources of external finance are non stationary. The second is their endogeneity. To
overcome the first problem we first-difference the two series.6 To investigate the
second problem we make use of a system version of the Hausman Test developed
by Revankar and Yoshino (1990). Results suggest that issues of stocks and cor-
reported.
6Both the Augmented Dickie-Fuller and the Philips-Perron tests suggest that when the series
are taken in levels, the null of unit root cannot be rejected at standard significance levels. The null,
however, is strongly rejected when the series are considered in their first differences. These results
are consistent for different specifications of the two tests.