01/02/2014 11:12 am ET Updated Mar 02, 2014

Primaries Predict Partisanship, But Only For Democrats

A majority of laws are passed after July in the second year of House terms. Given that many Primary deadlines and elections conclude around June/July, it has become conventional political wisdom that House Republicans will be open to compromising on immigration reform in the last few months before the 2014 midterms (see the full Primary calendar here).

Here's the U.S. map with the 2012 Primary schedules.

The problem is that this conventional wisdom has yet to be calculated. In the absence of statistical analyses, the conventional wisdom for romantic relationships is that opposites attract and birds of a feather flock together -- an inherent contradiction (in case you're wondering, social scientists have found more empirical support for the latter piece of wisdom).

Therefore, to test the belief that Republicans will be less partisan after the primary elections have passed, I used the publicly available DW-NOMINATE dataset on (the same one I utilized to confirm that women are less partisan than men), and coded each 2011-2012 incumbent in the 112th House based on if they faced a Primary challenge (1) or did not face a Primary challenge (2). The data are already coded based on political party.


A univariate ANOVA found a significant main effect of facing a Primary challenge: For Democrats, F(1, 197) = 8.00, p = .005, squared semi-partial = 3.90 pecent. Democrats' partisanship in the 112th Congress was higher among those who faced a Primary (Mean = .43, Standard error = .015) than among those who didn't (M = .375, SE = .011).

There was no significant effect of facing a Primary challenge among Republicans (p = .565).

In other words, Republicans who didn't face a Primary (M = .675, SE = .013) were just as partisan as those did (M = .666, SE = .015).

Women and Men Are Still Different.

There was a significant main effect of political party, F(1, 437) = 328.31, p < .001, squared semi-partial = 42.9 percent, and an interaction between political party and facing a Primary challenge, F(1, 437) = 4.88, p = .028, squared semi-partial = 1.1 percent.

Moreover, this interaction was only significant for women, F(1, 70) = 4.83, p = .031, squared semi-partial = 6.5 percent.

Democratic women's partisanship was higher among those who faced a Primary (M = .477, SE = .024) than among those who didn't (M = .415, SE = .019), whereas Republican women's partisanship was higher among those who didn't face a Primary (M = .663, SE = .028) than among those who did (M = .605, SE = .035).

In short, the interaction simply reinforced the fact that while House Republicans in the 112th Congress were more partisan, only House Democrats' behavior was significantly influenced by a Primary battle.

Limitations and Implications

1. Dear Democrats, don't wait until July to start fighting for immigration reform hoping that the GOP will suddenly become more reasonable. If Boehner won't bring it up for a vote, use the Discharge Petition to make all House members take a vote on immigration reform (if only symbolically). If the Senate minority Republicans want to be regressive by using the filibuster, then the House minority Democrats should be progressive by using the Discharge Petition.

2. This analysis is based on a 2011-2012 Congress that was preparing for a general election and therefore their behavior may have differed from the 2009-2010 Congress which was preparing for a midterm election. Of course, there is no way to qualitatively analyze the effect of the far-right's influence on House GOP members prior to 2011-2012 because the last time they were in power while approaching a midterm was 2005-2006 (and we know how that went). To be clear, it's quantitatively possible, but the GOP of the past few years is qualitatively different.

3. The pairwise comparisons for women revealed a significant effect for Democratic women, but not Republican women. The lack of a significant finding for Republican women when delineating the interaction effect is likely due to there being so few women in the Republican Party. This is also the reason all my Congressional analyses focus on the House and not the Senate. It is difficult to achieve the statistical power necessary to find significant results with only 100 people in the Senate to use as my subjects amd even fewer Republican women in the House to use as subjects.

4. It would be interesting if Poole and Rosenthal at changed their upcoming 2013-2014 dataset so that it would have Representatives' partisanship scores up to their Primary election, and then their partisanship scores for the rest of 2014. That would be the most definitive analysis of this conventional wisdom.