The notion of “asymmetric polarization” is commonplace in political discourse today. In particular, some commentators, such as Thomas Mann and Norm Ornstein, have argued that Republicans have moved farther right than Democrats have moved left. Mann has cited as “the strongest evidence for this asymmetry among members of Congress” the DW-Nominate scores of Nolan McCarty, Keith Poole, and Howard Rosenthal. The figure below plots mean DW-Nominate scores for congressional Democrats and congressional Republicans since the middle of the 20th century, using data drawn from here:

The steeper shift of the Republican mean away from the center relative to shift of the Democrat mean away from the center supports Mann’s and Ornstein’s claim of asymmetric congressional polarization.

However, this pattern of asymmetric polarization is not present in other estimates of ideology. Data from Adam Bonica‘s CFscores (“campaign finance” scores), which provide ideology estimates for members of congress based on donations to and from each member, indicate that since 1980 congressional Democrats have moved left slightly more than congressional Republicans have moved right, as shown below.

And estimates developed by Michael Bailey also do not show that Republicans have polarized more than Democrats. If anything, the opposite appears true:
DW-Nominate scores are based on a procedure in which ideology is inferred from the patterns produced by roll call voting: legislators who often cast the same vote as each other are assigned scores close together, and legislators who rarely cast the same vote as each other are assigned scores farther apart. Legislators who disagree with each other most often in roll call voting gravitate toward the poles of the scores, and these legislators are easy to classify as liberal or conservative based on perceptions of their ideologies in the real world.

DW-Nominate scores are uncontroversial as proxies for legislator ideology within a given Congress, but use of DW-Nominate scores to make comparisons across time has drawn questions. DW-Nominate scores are centered over time based on assumptions about cross-time legislator ideological change, but CFscores and Bailey scores attempt to center ideal points across time with non-legislator bridge observations. Bridge observations for CFscores are donors and for Bailey scores are political positions. Examples of such bridges include interest group A donating to legislator B in 1984 and donating to legislator C in 1996 (for CF scores), or legislator D in 1980 and legislator E in 1990 announcing support for the Supreme Court’s decision in case F (for Bailey scores).

It is not that one technique is superior to the others.  But one technique may be more useful for some purposes and the other technique more useful for other purposes. This decision, though, requires careful thought and a detailed understanding of the modeling assumptions of each technique. Although DW-Nominate scores suggest that recent polarization has been driven mainly by Republicans moving right, other estimates present a different pattern, complicating recent accounts and inviting reflection about how polarization is defined and measured.

L.J. Zigerell is an assistant professor of political science at Illinois State University.