19
the electorate, incumbency, the candidate’s experience (i.e. whether the candidate had pre-
viously held elective office), campaign spending, whether the candidate had a difficult
primary election, and region. The descriptions and codings of these variables are given in
Appendix 4. Thus our aggregate-level specification was
Democratic vote margin
= b
1
+ b
2
[Democrat’s proximity advantage] + b
3
[Democrat’s partisan proximity advantage]
+ b
4
(State partisan composition) + b
5
(Democrat’s incumbency advantage)
+ b
6
(Democrat’s experience advantage) + b
7
(Democrat’s spending advantage)
+ b
8
(Democrat’s primary advantage) +b
9
(South) . (4)
Table 2, column1 reports the estimated regression coefficients for the analysis of
the 95 Senate elections held between 1988-92, for which information was available. The
partisan proximity advantage coefficient is positive and statistically significant at the .05
level. Furthermore, the magnitude of the coefficient suggests that the electoral gains of
partisan proximity are substantively significant: the estimated value, 10.33, suggests that if
40% of the state’s eligible voters identify with the candidate’s party, then this candidate’s
vote share increases by approximately four percentage points for each additional unit that
the candidate approaches to her partisan constituency’s mean position, along the 1-7 ideo-
logical scale. By contrast, the proximity advantage coefficient is not statistically signifi-
cant, and in fact has the wrong sign.
22
These estimates support the Election Outcomes Hy-
pothesis, that candidates’ vote margins increases with their proximities to the mean policy
position of their partisan constituencies, even when controlling for their proximities to the
mean voter position.
23
22
The difference between the partisan proximity advantage coefficient estimate and the proximity advantage
coefficient estimate is statistically significant at the .01 level, providing strong evidence that candidates de-
rive electoral benefits from diverging from the mean state voter position, in the direction of their partisan
constituencies.
23
We note that we estimated the parameters of an additional election outcomes model in which voters were
specified as evaluating the candidates’ ideological positions according to the directional model (Rabinowitz
and Macdonald, 1990), which Lacy and Paolino (1999, 2001) have argued may capture U.S. voters’ behavior
Convention