Publication Type: Conference Paper/Unpublished Manuscript
Review Method: Peer Reviewed
Abstract: According to Fenno (1977),
Senators can change their policy positions within their six-year terms
even when their constituencies remain the same. The reason, as Fenno
argues, is that the length of senatorial terms provides for a period,
approximately at the midpoint, during which Senators are reasonably
insulated from electoral pressures due to the distance from elections.
In this paper, we develop a formal model and an empirical test to
explain this spatial movement of Senators’ policy positions across
their six-year terms. The first part of the formal model introduces a
decision-making calculus that Senators employ in evaluating the utility
derived from accepting certain cues. We contend that a key component of
this utility function is Senators’ proximity to election. The second
part of the model is a two-stage game accounting for the fact that
Senators need to win both a primary and a general election. This
results in a constrained optimization problem, in which Senators
maximize their vote share at the primary stage subject to the condition
that they will have to win a majority of the votes in the general
election where they face more moderate constituencies. We test our
formal model empirically by using a hierarchical model. In the first
stage of the model, we rely on Markov Chain Monte Carlo methods to fit
a Bayesian model of ideal point estimation to assess the claim that
there is spatial movement of policy positions. By dividing senatorial
terms into two-year congressional periods, we obtain a natural
experiment that allows us to compare the third of the Senate up for
reelection with the remaining two-thirds to check whether proximity to
election – holding everything else constant – results in greater
movement of policy positions. Our findings, in fact, show that Senators
up for reelection are more likely to change their policy positions than
those that are in the middle or at the beginning of their six-year
term.
In the second stage of the hierarchical model, we estimate a Bayesian
regression model, accounting for the error in the ideal point
estimation in the first stage, to explain the direction of spatial
movement. Specifically, we test the Downsian model according to which
Senators – come election time – should move toward the median of the
constituency critical to their electoral success. Most interestingly,
our findings show that Senators who face quality primary opponents are
more likely to move to relatively extreme policy positions, which are
preferred by their primary constituencies.
The paper offers two main innovations. First, we formalize Fenno’s very
important claim that Senator’s policy positions change across time.
Secondly, we test this model using Bayesian ideal point estimation,
which offers many advantages to previous methods. Perhaps most
importantly, these methods make the important assumption that ideal
points follow a random walk in which Senators’ positions across time
are, at least partially, dependent upon previous positions. Such
assumptions are not possible in constant or independent ideal point
models.