3
the center, provided they shift in the direction of their partisan constituencies. To date,
however, the partisan constituencies explanation rests entirely on illustrative arguments.
Our purpose here is, first, to put the partisan constituencies explanation on a firmer
theoretical footing by presenting theoretical results on candidate equilibrium in margin-
maximizing strategies, when voters are moved by partisan loyalties and party identifica-
tion, and may abstain due to alienation from the candidates. Second, we present empirical
tests on Senate elections designed to determine whether the candidates’ observed behavior,
and the electoral consequences of their positioning, is consistent with the partisan constitu-
encies explanation.
Our results suggest that the partisan constituencies explanation has strong theoreti-
cal and empirical foundations. Theoretically, we derive conditions for the existence of a
unique candidate equilibrium, and we show that under these conditions the candidates’
equilibrium positions diverge from the mean voter position, in the direction of their parti-
san constituencies, as the policy beliefs of these constituencies grow more extreme and as
the size of these constituencies increase. Empirically, we find that U.S. Senate candidates’
policy positioning in the period 1988-92 was consistent with our theoretical results and
that, ceterus paribus, candidates who located near the mean positions of their partisan con-
stituencies enhanced their vote margins in general elections compared to candidates who
located near the mean state voter position. These findings are important because several of
them are not predicted by the alternative explanations for candidate divergence summa-
rized above. Our results shed new light on candidate strategies and election outcomes, on
the political implications of higher turnout, and on how we should empirically analyze the
linkages between candidate policies and voting behavior.
2. A Spatial Model with Variable Turnout and Partisan Loyalties
We analyze a two-candidate election involving a Democratic candidate D and a
Republican candidate R, who choose positions along a unidimensional Liberal-
Conservative ideological continuum bounded by the positions [-
α
, +
α
]. The candidates’
positions are denoted by D and R, respectively. Voters’ utilities for the candidates are
based on a simple version of the behavioral researcher’s multivariate voting model, one in
Convention