Michael H. Crespin, David W.
Rohde, and Ryan J. Vander Wielen, "Measuring variations in
party unity voting: An assessment of agenda effects,"
Party Politics, 19 (May 2013), 432-457.
[Available at http://ppq.sagepub.com/content/vol19/issue3/
Measures of party divisiveness, commonly referred to as
party unity scores, have been widely used for various
purposes in the congressional literature and elsewhere. In
fact, these party unity scores have even been a motivating
factor in some scholarly dialogues (Brady et al., 1979;
Collie and Brady, 1985; Cox and McCubbins, 1993; Rohde,
1991). At the same time, scholars also recognize the
important influence the agenda has on shaping the roll-call
record (Cox and Poole, 2002; Cox and McCubbins, 2005;
Roberts, 2007). Specifically, certain types of votes are
more inclined to generate conflict than others. Moreover, we
know that there is considerable variation in the content of
the agenda across time (Rhode, 1992).
- Figures and
- Table 1. Expected effect of vote-type on conventional
party unity measure as share increases.
- Figure 1. Percentage party unity by Congress
including adjusted party unity and adjusted party unity
excluding ultra-consensual votes.
- Table 2. Vote-type by Congress.
- Table 3. Party unity by vote-type.
- Table 4. Percentage of roll-calls with majorities of
at least 95 percent.
- Figure 2. Variation in adjusted scores caused by
index Congress selection.
- Table 5. Bivariate regressions.
- Table 6. Replication exercise.
- Appendix A. Vote categorization rubric
- Appendix B. Measurement values
(First paragraph of conclusions) These results have both
substantive and methodological implications for analyses
using measures of aggregate partisanship (and more broadly
for aggregate congressional voting measures generally). As a
general statement, it is imperative that researchers
understand the details of voting data before conducting
empirical analyses of it. When studying voting
longitudinally, for instance, one needs to be sure that the
analysis is comparing 'apples to apples'. This holds for
those examining party unity scores over a long time frame or
other aggregate measures that are derived from roll-call
votes. We have demonstrated that, without adequate controls
the data do not meet this condition due to the sizeable
variation in the roll-call agenda. We show that changes in
the conventional measure of party unity are in no small way
a reflection of changes in the agenda. Increases (decreases)
in traditionally consensual vote-types can and do lead to
decreases (increases) in party unity scores that are wholly
independent of partisan disagreement or cohesion. Inferences
based on many conventional measures may then be spurious.
Instead, for the party unity score -- one such example of a
party divisiveness measure -- we offer a modification that
controls for the variation in the agenda and offers a more
accurate comparison of party unity across Congresses. The
method detailed above can easily be extended to other
measures of aggregate party