20
impact of rival expenditures during the year-based period, but shaped more through their
own expenditures.
Finally, Model III shows the combined effects of campaign-period spending and
year-based expenditures in the same model. Own fund agent, local party and support
group expenditures all demonstrate the expected statistically significant and positive
effect on DPJ vote shares. Likewise, rival campaign-period spending is the only variable
with a statistically significant and negative correlation with DPJ votes, which suggests
that rival effects may not be entirely insignificant for the DPJ.
Aside from the effects of the expenditures on DPJ vote shares, I turn to the results
for the control variables. In terms of candidate characteristics, the measures for
incumbent status and the previous number of terms politicians were elected to office have
a statistically significant and positive effect on DPJ vote shares. Moreover, the number of
support groups registered to DPJ members appears to have no systematic impact on the
percentage of votes they receive, which parallels the results for LDP in Table 2.
19
For district characteristics, the measure for the percentage of the district that is
densely inhabited has a negative and statistically significant impact on DPJ vote shares,
which does not differ from the LDP results. As for the level of competition faced by DPJ
candidates, there is a statistically significant and negative effect on vote shares in each
model, which stands in contrast to the meaningfully positive effect for the LDP.
20
Finally,
DPJ vote shares are negatively affected by districts with greater numbers of candidates in
the race.
19
This result is likely related to the fact that only half of DPJ members have any registered support group
beyond their local party branch and fund agent.
20
Unlike the LDP, greater differences between the DPJ candidates and the frontrunner or second-place
finisher appear to negatively contribute to their vote shares.
Convention