14 of 36
the mean. Others are defined as mainstream partisans. The results of table 5 clearly show
that moderate partisans behave quite differently from mainstream partisans in this case.
[Table 5]
Furthermore, the pattern of oil PAC contributions is also different for moderate
partisans. As shown in table 6, the oil PACs treated moderate Democrats more like
mainstream Republicans and moderate Republicans more like mainstream Democrats, as
if they knew that moderate partisans would vote differently from their parties on this
issue
15
.
[Table 6]
It is thus reasonable to control the moderate partisans by introducing two new
variables: a dichotomous variable for moderates, which obtains the value of 1 if the
ideology scores lie in the range [-11.15, 14.15], and 0 otherwise; an interaction term for
moderates and party. These two variables represent moderate Democrats and moderate
Republicans respectively. Party now represents mainstream partisans.
[Table 7]
The results are presented in table 7 along with the results from logic regression
without controlling for moderate partisans. The coefficients for the moderate partisans are
significant and have expected signs. Being a moderate Democrat would reduce the odds
of voting for the ban by .146 times and being a moderate Republican would improve the
15
One piece of supporting evidence is that 19 of 29 moderate Democrats belong to the Blue Dog Coalition,
which is a conservative Democrat caucus. The Blue Dog Coalition has 31 members at the time and 18
voted against the ban.
Convention