Democratic challenger in 1994) to 6.27 (a Republican challenger in 1994). If the null
hypothesis is right and national tides act through competitive challengers, the slope of the
line for high-spending challengers should be much steeper than for low-spending
challengers because national tides have a larger impact in those races. On the other hand,
if both tides and competition have strong independent effects, the lines should be parallel
(indicating uniformity of effect), they should have steep slopes (indicating a strong effect
for national tides), and they should be far apart (indicating a strong effect for spending).
[Figure 3 about here]
Figure 3 confirms that the independent effects of spending and the national vote
shift are much larger than any interactive effect between the two. Because the interaction
term between spending and national tides is not zero, we do see some separation at larger
tides: the slope of the line for higher-spending challengers is very slightly steeper than
the slope for lower-spending challengers. But only slightly: the lines are very nearly
parallel, with a difference of 6.0 between the highest values of each line and a difference
of 4.8 between the lowest values. Furthermore, these large and consistent differences
imply that the direct effect of spending is also large. The line for low-spending
challengers never crosses the zero point on the dependent variable, indicating that low-
spending challengers never gain votes, no matter how well their party performs overall.
On the other hand, the national vote shift does an admirable job of rescuing these weak
challengers from themselves: in the best year for their party, these challengers perform
over seven percentage points better than in the worst year. For the highest-spending
challengers, the difference is over eight percent.
Though the interaction between spending and national tides is not the only
interaction in the model, the interactions with district partisanship are just as insubstantial
in size. As expected, the interaction between MARGIN and the national tide is negative:
each one-unit change in the MARGIN variable reduces the size of the national tide’s
effect by 0.01, such that national tides are weaker in districts with lopsidedly partisan
constituencies. To put this effect in some perspective, a change from the 25
th
to the 75
th
percentile of the MARGIN variable reduces the effect of the national tide from 0.70 to
0.61—hardly an enormous change. The district’s partisanship also does little to specify
the effect of the challenger’s spending. This interaction is not only weak (0.005), it is
McGhee
18
Convention