Estimating Polarization in the United States Congress
Faculty Mentor: Michael Barber, Political Science
I investigated if commonly accepted ideology estimates in Congress
change when restricted to the modern era. I used these estimates to determine if
previous measures of political polarization are accurate.
Polarization is determined by the distribution of ideology in Congress. This
is measured using ideal point estimation. Ideal points are numerical calculations
of ideology allowing for direct comparisons between legislators. Ideal point
estimation was created to combine an empirical explanation to existing theory in
collective decision making. Consequently, ideal point estimation is frequently
used to describe the behavior of legislatures: especially the United States
The popularization of ideal point estimation in American Politics can be
attributed to Keith T. Poole and Howard Rosenthal. Their estimation, DWNOMINATE,
is the most widely used in the study of the United States Congress.
The estimates made by Poole and Rosenthal make legislator comparisons from
1789 to 2014. The problem with using such a large time span is that the party
system has significantly changed over time. The Republicans and Democrats in
our party system today are not directly comparable to the Federalists and
Democratic-Republicans in the 1st Congress. Therefore these estimations may
change according to the time span studied. Because ideal point estimates are
almost exclusively used as a measure of ideology in empirical Political Science,
these potential problems need to be accounted for in our study of Congress.
To accurately estimate polarization in today’s Congress, I selected an
appropriate time period for study. I used data including the roll call votes of the
91th Congress (1969-1970) to the 13th Congress (2013-2014) to calculate new
ideal point estimations. This time period is large enough to perform statistical
procedures and compact enough that the party system is consistent throughout.
Clinton, Jackman, and Rivers created a statistical package in the
programing language R that uses a Bayesian method to calculate ideal points
from the roll call votes in Congress. I used this method for my ideal point
estimation and data collection. I expected my results to show a different story of
polarization in Congress. To test this, I looked for differences between my ideal
point estimates and the estimates made by other scholars. It has been shown
that the modern Congress is increasingly polarized due to the Republican Party. I
expected Congress to be less polarized by restricting the period to the modern
By restricting the period, the estimates were not significantly changed. The
same trends from previous ideal point estimations were comparable. This
reinforces the validity of previous ideal point estimations of congressional role
call voting behavior.