Ideological Primary Competition and Congressional Behavior

Abstract

Primary challenges on an incumbent’s more extreme flank have become increasingly common in congressional elections. We explore the consequences on the legislative behavior of successful incumbents following these types of challenges. We propose that the effect of extreme ideological primary challenges is conditioned on whether the member belongs to the majority or the minority party. We test our expectations on primary elections from 2000 to 2012 and corresponding changes in voting behavior in the next Congress. We find that incumbents in the majority party who have defeated an extreme ideological primary challenge are less likely to support their party, especially on key votes. We expect this reaction is because they fear appearing too moderate and thereby encouraging another extreme ideological primary challenge. We find that incumbents in the minority party who face an extreme ideological challenge do not change their general voting behavior, but are more likely to vote with their party on key votes.

Acknowledgments

We thank William Patton and Maura McDonald for their research assistance. We also received valuable feedback and suggestions from Jason Roberts, the American Politics Research Group at University of North Carolina, participants at several conferences, and the anonymous reviewers.

Notes

1 The word ‘ideological’ is used in line with other work on congressional primaries that defines a challenge from the incumbents’ extreme flank as an “ideological one.” This classification is based on the issue positions/talking points of the candidates running and is explained in greater detail later in text.

2 We do find members of the minority who faced an ideological challenge are more likely to support their party on key votes in the Congress following the ideological challenge. This finding is discussed in greater detail in following text.

3 One possible explanation for the discrepancy between the conclusions of Jacobson (2012) and Sides et al. (2018) is their reliance on self-reported participation versus validated vote choice.

4 Looking at the openness of primary elections, rather than the competitiveness of the primary, McGhee et al. (2014) conclude that there is no effect on the ideological extremism of state legislators.

5 Of course, incumbents do not face primary challenges for only ideological reasons. Other types of primary challenges may also result in a member of Congress modifying voting behavior. Ideology, however, is an increasingly common reason that may motivate a primary challenge, and we believe that there is a strong theoretical rationale for expecting a member of Congress to react to an ideological challenge. In addition, our exploration of all incumbents who faced a primary challenge in addition to the inclusion of a variable capturing the competitiveness of the primary helps to parse out how members of Congress respond to a variety of primary challenges.

6 Kanthak and Morton (2001) contend that the relationship is more nuanced, noting the importance of distinguishing between how easy it is for voters to adopt or switch party affiliations at the polls, and suggesting that there may be differences between the parties as well.

7 Boatright (2014) has some of this information and utilizes it in his analysis, but he only examines changes in mean NOMINATE scores for incumbents who faced primary challenges and does not conduct a multivariate analysis that also controls for competitiveness. In addition, there are many primaries in his data for which the reason for the challenge is listed as “unknown.”

8 As noted previously, from a formal modeling perspective, this act is not strictly rational because the member is voting against a bill which, strictly speaking, moves a policy closer to his or her ideal point. However, we argue that the signal sent by casting such a vote garners a member electoral benefits which outweigh the benefits the bill would bring. One such electoral benefit is the deterrence of a future, ideologically extreme challenger in a primary election.

9 We chose to use percent of the time that the member votes with the Minority Party Leader (opposed to Majority Party Leader) as it better illustrates our theoretical expectation of majority party members exhibiting their ideological extremism by voting against their party. For robustness, we fit models using agreement with the Majority Party Leader as well. The results do not change significantly. The roll call dependent variable includes procedural votes, but only if the minority leader cast a vote. We also fit the model just using key votes and this model is reported later. All roll call data were accessed at http://voteview.com/Party_Unity.htm

10 More important, our dependent variable is immune to changes in partisan control of Congress because it is created by utilizing the voting history of the Minority Party Leader in the Congress following the primary challenge and then utilizing that Leader’s voting history in the previous Congress and comparing it to the member being analyzed. Thus, when there is a change in party control, the voting record is compared to whoever the Minority Party Leader is in the current Congress and that same Leader’s voting record in the previous Congress, even if previously he or she was the Majority Party Leader or not a leader at all.

11 As described following, key votes are based on identification by Congressional Quarterly.

12 These categories include old age (i.e., the incumbent is old and should retire), scandal, a prominent national issue, or a focus on ideological position.

13 We also fit our models with a variable indicating that the incumbent won the primary, but captured less than 90% of the vote share, as Burden (2004) has done. Although a primary where an incumbent secured 89% of the primary vote may not be considered highly competitive, this measure represents incumbents who faced non-trivial competition, or as Burden (2004) says, “more than token opposition.” With this change in operationalization, our results do not change significantly or substantively.

14 In-party spending is the natural log of the amount spent in 2010 constant dollars.

15 Both models include Republican and Democratic incumbents who faced a primary challenge. The distinction between the models is that in the second model, if the challenger was classified as a Tea Party candidate, this is considered an ideological primary challenge. Table A in the Appendix provides models that look only at Democrats, to further demonstrate that the results are not driven by or confined to the Tea Party and/or the Republican Party.

16 Given that no Tea Party primary challenge occurred against a Democratic incumbent, we fit a single Democrats-only model. The full model and results can be seen in the Appendix.

17 There are two things to note about the key vote model. The first is that the N decreases. This is because we only fit the model for members who voted on all the key votes in a Congress. The second point to note is that we exclude any key votes where the majority party was rolled at final passage. In these cases, a vote with the minority party would be suggestive of a moderate position and therefore they are excluded.

18 We also fit this model for Democrats only. The results for this model can also be found in Table A in the Appendix.

19 As noted previously, we removed votes on which the majority party was rolled from this model.

20 When considering ideological challenges that exclude Tea Party challengers, we observe a 6.8% increase.