Do Quality Candidates and Incumbents Still Matter in the Partisan World? Comparing Trends and Relationship Between Candidate Differentials and Congressional Election Outcomes, 1900–2022

Abstract

The increase in elite-level polarization and the changing partisan nature of elections to the U.S. House led scholars to posit that candidate characteristics are minor considerations in determining these election outcomes. However, it is not clear if these trends extend to the U.S. Senate or if candidate considerations have lost the relatively minor predictive power they exhibited during the 2010s, particularly as partisanship continued to rise as a predictor of election outcomes. Using historical data on elections to the U.S. House and Senate from 1900 to the recent 2022 midterm elections, we test whether the incumbency advantage and candidate quality differentials are still salient predictors of congressional elections. We find that the incumbency advantage largely disappeared as a salient component of election outcomes for both chambers as partisanship increasingly shapes these outcomes. By contrast, we find that candidate quality differentials, while waning, still can play a considerable role in shaping congressional election outcomes, particularly in the Senate. We conclude by showing that the declining emergence of quality candidates may have played a pivotal role during the 2022 election cycle by costing Republicans control of the U.S. Senate.

Keywords:

• 2022 U.S. elections
• candidate quality differentials
• congressional elections
• incumbency advantage
• U.S. Senate elections

Acknowledgements

The authors thank the anonymous reviewers, Gary Jacobson, Joel Sievert, and participants of the 2023 Midwest Political Science Association (MPSA) conference for helpful comments on drafts of this article.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

Data are available upon request.

Notes

1 NBC News (8/18/2022): McConnell says Republicans may not win Senate control, citing ‘candidate quality’.

2 Politico (9/1/2022): Defiant Rick Scott explains ‘strategic disagreement’ with McConnell over battle for Senate.

3 To that point, Sen. McConnell unsuccessfully served as NRSC Chair during the 1998 and 2000 election cycles, with Senate Republicans failing to make any gains during the 1998 cycle (i.e., no party gained seats) and losing four seats during the 2000 cycle.

4 We also use these data to match the 2022 congressional districts to their 2020 geographic predecessors and also use these data to identify where House incumbents sought re-election. For example, this procedure allows us to code West Virginia’s 1st congressional district as the geographic successor of West Virginia’s 3rd congressional district since the state went from having three districts to one. There are a total of 94 congressional districts for the 2022 election cycle that had different district numbers than their geographic predecessors.

5 Note that the U.S. Senate election data compiled by Algara (2019) election outcomes from the beginning of the direct-election era in 1914 to the 2016 election cycle was updated to by Algara, Hale, and Struthers (2022) to include results for the 2018 and 2020 election cycles. For county-level U.S. Senate election results covering the same time-period of 1914–2020, see Amlani and Algara (2021).

6 By considering candidate quality differentials as a salient predictor of congressional election outcomes, all forthcoming models only consider election contests contested by both major political parties. Indeed, 10.76% (2,869 cases) of House elections from 1900-2022 and 7.56% (145 cases) of Senate elections from 1914-2022 are uncontested by one of the major two-parties. There are a total of 26,672 House elections and 1,919 Senate elections during our time periods of interest.

7 Jacobson’s (2015) model serves to estimate the incumbency advantage in U.S. House elections and is a modified version of the Gelman and King (1990) model estimating the incumbency advantage in the U.S. House over time.

8 Note that this 2022 value of CP_it, in the context of the U.S. House, takes the form of the two-party margin in the 2020 presidential election calculated under the new district lines used for the redistricting 2022 election cycle. For example, the 2022 and 2020 election year values of CP_itfor New Mexico’s 2nd congressional district is ≈ 6.00 and -12.04, reflecting that in 2020 NM-2 was a Republican-leaning district prior to being redistricted to be a.

Democratic-leaning district by coordination of the Democratic legislature and Democratic Governor. Indeed, New.

Mexico Democrats successfully redrew NM-2 to be such a Democratic-leaning district that the incumbent Republican Rep. Yvette Herrell was unseated during the 2022 election cycle, giving Democrats control of all three congressional districts and—coupled with being represented by two Democratic Senators—a completely Democratic congressional delegation in Washington.

9 Another advantage of using the presidential vote to construct a measure of constituency partisanship is that we hold the candidates constant across all districts since both the Republican and Democratic presidential candidates appear on the ballot in all congressional districts and states. As opposed to constructing the measure from the lagged vote won by the Democratic House or Senate candidate in the preceding election (the approach used by Gelman and King 1990), our measure allows for cross-constituency comparison in partisan preference since each the choice between two presidential candidates does not vary across districts in a given presidential election cycle and the measure is immune from other non-partisan considerations in its construction, such as the influence of a popular local incumbent that sought office in the constituency in the previous lagged election.

10 By contrast with equation 1, this specification in equation 2 omits newly created congressional districts resulting from reapportionment since no party controls a newly created congressional district.

11 For example, during Democratic U.S. Senator Joe Manchin’s 2018 re-election bid in West Virginia, this value of constituency partisanship takes the form of −44.27%, indicating a stark partisan disadvantage for Senate Democrats as they sought to defend a vulnerable incumbent in what has become a very Republican state. By contrast, this value takes the form of 51.39% during Republican U.S. Senator John Barrasso’s 2018 re-election bid in Wyoming, indicating a stark partisan advantage for Senate Republicans defending a seat in perhaps the most Republican-leaning state in the nation.

12 For House elections, this is coded as a trichotomous variable indicating −1 if controlled by the Republicans, 0 if a new seat due to reapportionment, and 1 if controlled by the Democrats. For the U.S. Senate, this variable takes a simple binary, coded 0 for a Republican seat and 1 for a Democratic seat. For the purposes of our analysis and given explicit support by the Democratic Senatorial Campaign Committee (DSCC) in each of their Senate bids, we consider the independent candidacies of Angus King (ME) and Bernie Sanders (VT) as Democratic candidacies. Moreover, we consider the independent challenge of Evan McMullin to Republican U.S. Senator Mike Lee in the 2022 Utah Senate race to be a Democratic candidacy given the explicit endorsement of McMullin by the Utah Democratic Party and the lack of a Democratic candidate in that race.

13 In the appendix, we include a series of robustness checks that take into account uncontested races rather than the subset of congressional elections contested by both major parties reported in the manuscript. These robustness checks including the uncontested cases confirm all forthcoming results reported in the manuscript.

14 Indeed, as the loess regression curves show in Figure 2 Panels A–D, the relationship between candidate-centered differentials are in clear decline if we assess them in the revised Jacobson model or the incumbent party model specifications.

15 As such, 23,143 of our 26,672 House election cases from 1900 to 2022—and 1,488 of our 1,919 Senate election cases from 1914 to 2022—are elections with an incumbent on the general election ballot.

16 In the appendix, we also present the the relationship between constituency partisanship and election outcomes during incumbent re-election bids. These results are very similar to the preceding results estimating the partisan advantage, with constituency partisanship playing an increasing in shaping electoral outcomes.

17 We note that this estimate is taken from the model articulated in equation 3, which estimates the marginal effect of candidate quality differential for each individual election cycle from 1914 to 2022 using Jacobson’s (2015) framework.

18 By seat out-party, we mean that Republicans nominated a quality candidate to contest a Democratic seat in Nevada while Democrats nominated quality candidates to contest Republican seats in Ohio and Wisconsin.

19 This race featured U.S. Rep. Ted Budd (R-NC) being opposed by former statewide-elected Chief Justice of the North Carolina Supreme Court Cheri Beasley (D-NC).

20 In Georgia, Herschel Walker secured the Republican nomination by about a 55% margin over statewide-elected Agricultural Commissioner Gary Black while, in Arizona, Blake Masters performed about 23% better than third-place finisher statewide-elected Attorney General Mark Brnovich en-route to a 12% primary victory.

21 We note that the quality advantage estimate for the U.S. House was far too small in the 2022 midterm election cycle to alter any of the observed outcomes.

22 In these two contests in Arizona and Georgia; Republican voters nominated amateur candidates in the form of venture capitalist Blake Masters and former University of Georgia football star Heisman Trophy winner Herschel Walker to oppose Democratic Senators Mark Kelly and Raphael Warnock, respectively. These amateur challenges proved unsuccessful, with Sen. Kelley defeating Masters by about 5% and Sen. Warnock, in a December runoff, defeating Herschel Walker by close to 3%.

Additional information

Notes on contributors

Carlos Algara

Carlos Algara is the Mary Toepelt Nicolai & George S. Blair Assistant Professor of Politics & Government at Claremont Graduate University and a Faculty Affiliate at the Center of Effective Lawmaking.

Byengseon Bae

Byengseon Bae is a PhD candidate in political science at Claremont Graduate University, where he studies American political thought and development.