Abstract
Throughout American history some members of Congress are beset with the unfortunate predicament of facing a fellow incumbent in their bids for reelection. One culprit is responsible for these atypical contests: redistricting. Using district and sub-district level data, this research note provides the first systematic coverage of all dueling incumbent general election U.S. House races from 1843 to 2018. We chronicle and analyze when we expect to see dueling incumbent races, the ability of parties to target out-party incumbents, and the electoral value of previously represented constituents for incumbents in reconfigured districts. Although incumbent duels are uncommon, they comprise a substantial number of incumbent general election defeats in contests following redistricting.
In this research note we provide the first comprehensive examination of the 73 dueling incumbent (incumbent vs. incumbent) general election contests in the U.S. House of Representatives from 1843 to 2018. 1 Although these types of races are infrequent, since 1952 they account for 22% of incumbent defeats in general elections held immediately after a decennial redistricting. 2 In this exploratory and primarily descriptive account we consider: (1) the determinants of the emergence of dueling incumbent races such as redistricting, reapportionment, and partisan targeting, (2) the role of partisan mapmakers in eliminating out-party incumbents when redistricting occurs, and (3) evaluate the support incumbents garner among the voters they retain following a redistricting.
Dueling Incumbent Races, 1843–2018
Dueling incumbent races are seldom mentioned in extant literature on congressional elections
3
, and the few previous accounts characterized their existence as a “failure” of redistricting strategies that come with (by definition) a high rate of incumbent defeat (e.g., Bullock, 1975; Bullock, 2010). We expect dueling incumbent races to primarily manifest in two instances, as described in Figure 1. First, redistricting is a causal factor for observing incumbent versus incumbent contests. Changing district lines forces incumbents to decide where to run for reelection and all 73 of our dueling incumbent contests take place in the confines of redrawn congressional boundaries. Dueling incumbent determinants and expectations.
Next, we expect to see more duels in states that lose a district, an event we call electoral retrenchment. When this occurs, the number of incumbents is greater than the number of districts so we should observe a duel if all incumbents decide to run for reelection. Our data support this hypothesis with 63% of our cases in states experiencing electoral retrenchment. The rest of the dueling incumbent matchups happened when states either gained a seat (10%) or retained the same number of seats following reapportionment (27%). We also expect to observe more duels when one party draws district lines since they can “target” the out-party by drawing boundaries unfavorable to an incumbent of that party. Our data show that 68% of duels occur when one party controls redistricting. The remaining contests occur in comparatively less partisan settings, including when district boundaries are reconfigured by the courts, commissions, or via a divided state government (e.g., different parties in control of the legislative chambers and/or the governorship).
Figure 2 presents all 73 contests in our dataset. Twenty-eight different states had at least one dueling incumbent contest. Populous states consistently losing districts have the most duels: Ohio with 10, followed by Pennsylvania (8), New York (5), and then several states with 4. Forty-one dueling incumbent races occurred in states losing at least one seat through reapportionment. Another five contests operated under the condition of electoral retrenchment because an at-large seat (or seats) was subsequently converted to a single-member district in a multi-district state, which then materialized into a dueling incumbent race. Twenty dueling incumbent contests took place in states where the number of seats remained constant. In only seven cases did a dueling incumbent contest occur in a state adding at least one seat. Historical overview of dueling incumbent general elections, 1843–2018.
The numerical rarity of incumbent duels offers face validity for the expectation that these contests are the most competitive, given two incumbents decided to run for the same seat. In other words, from the perspective of these incumbents they would appear to believe they have a reasonable possibility of winning, although delusions of success (Maisel, 1982) surely play a role in specific contests that greatly favor one incumbent’s party over the other. Nevertheless, with a modified and supplemented version of Gary Jacobson’s House dataset for the post-World War II era (1946 forward), we can compare the House vote margin in two-party contested races with: (1) an incumbent facing a challenger, (2) an open seat, and (3) a dueling incumbent matchup. Importantly, these data are limited to only House contests where redistricting took place 4 , and in the election cycles including a dueling incumbent election (1952, 1962, 1966, 1968, 1970, 1972, 1982, 1992, 2002, 2004, 2012, and 2018). In short, this is our best attempt at an apples-to-apples comparison of competitiveness across these three categories of elections.
Dueling Incumbent Races are the Most Competitive Based on Vote Margin.
Note: Number of cases by contest category are in parentheses. Data originally from Jacobson but supplemented by the authors. Elections only include those in which redistricting took place and in cycles with a dueling incumbent race, which occurred from 1952 to 2018. We have omitted the four dueling incumbent contests that took place among members of the same party, but their inclusion hardly alters the data displayed in Table 1.
Party-Motivated Considerations
To be sure, it is not typically the case that if one party controls redistricting it will seek to remove an out-party incumbent by altering congressional boundaries to foster a duel favoring a member of the in-party. Indeed, pitting two incumbents of the same party against each other at the primary stage is a foolproof means to eliminate at least one member of the opposing party. Nevertheless, there are several instances where it is evident that the party controlling redistricting aims to defeat an out-party representative in the general contest by pairing that member against a party-aligned incumbent. 5 Thus, the question we consider is what do the win/loss numbers for incumbent duels look like in those scenarios where one party redraws the congressional districts?
Redistricting Plan Type by Party of Winner.
The “% Duels of Total” column is included as a baseline and was calculated by dividing the number of duels that took place following standard redistricting years (e.g., 1990, 2000, 2010) by the total number of plans in that category that were classified as “major redistricting” in standard redistricting years.
†The 1882 Mississippi race (District 2) was decided in the House, so it is not included here in the win total for Democratic-drawn plans.
Partisan Success in Targeting Out-Party Incumbents, 1882–2012.
Note: The 1882 Mississippi race (District 2) was decided in the House, so it is not included here in the total for Democratic-drawn plans.
In Table 3 a “success” is defined when a greater share of the in-party’s incumbents run (compared to their seat share prior to the election with the new maps) versus the out-party’s incumbents after the in-party has executed a partisan plan. 7 Conversely, a “failure” would be when a greater share of out-party incumbents run for office, despite the in-party being in control of redistricting. A “push” occurs when the same number of incumbents from both parties run, despite the state losing a seat. Many of these “push” contests are dueling incumbent races. Just looking at the overall “success” rate, the ability of the in-party to dissuade out-party incumbents from running looks questionable at best, with only a 25% success rate for Democrats and a 34% success rate for Republicans. If parties are viewed as successful when achieving a push or a pure success, then both parties had a considerable margin of success, with Democrats achieving a push or pure success in 64% of races and Republicans in 72% of contests.
Race-by-Race Analysis
In this section we provide a descriptive analysis of cases with sub-district data. We have data on two-party and same-party contests in multi-district states, and two-party contests in at-large states. Except for Iowa in 1942, all data are from the Baker (1962) “Reapportionment Revolution” era (Cox & Katz, 2002) and later. Sub-district data allow us to measure the population and partisanship of the old district incumbents bring with them to the new district as well as a similar measure for areas that are new to both incumbents. Areas previously represented by Democratic incumbents are labeled “Retained D” and we use “Retained R” for areas previously represented by Republican incumbents. The area new to both incumbents is labeled “Redrawn.”
Descriptive Statistics for Select Dueling Incumbent Elections.
Note: LA race features two Republicans at the runoff stage.
Democratic Vote/Wallace Vote.
Non-weighted precinct level.
Of course, just looking at constituents does not account for their partisanship. In the Ohio 2012 race, most voters in the Retained D and Retained R parts of District 16 were under 50% Democratic, giving the Republican a presumed advantage in these areas. Also, the Redrawn portion of Ohio 16 was slightly Democratic (51.3%) but was only 37.6% of the population. In Texas 19, the partisan targeting was highly effective as the new district was overwhelmingly Republican in all three areas. Texas 32 was likewise successfully targeted by the GOP, as the Democrat brought in a partisan stronghold (65.9% Democratic), but it was only a relatively small percentage of the new district (20.4%). The other partisan maps also appear effective on this measure of the percentage of partisans. Considering these results along with election outcomes data presented earlier, it appears mapmakers were able to draw plans furthering their intentions of defeating the out-party. Nonetheless, several of the targeted incumbent candidates overcame these redistricting-induced electoral deficiencies.
Individual Incumbent Performance
In this penultimate section, we estimate a series of regression models to determine how incumbents perform in dueling incumbent contests. Our dependent variable is the Democratic proportion (two-party) of the House vote at the smallest level of voting geography available (county, precinct, etc.). 9 The key dichotomous independent variables in the models are: Retained D, and where applicable, Redrawn. These two variables are compared against the omitted category, which is Retained R. These variables are coded 1 to correspond to the applicable area of the new district. Thus, we are interested in seeing if there is a significant difference in the House vote cast by Retained D constituents and Redrawn constituents versus Retained R constituents. The expectation is that Retained D is positive and significant since the Democratic incumbent should perform better among their “old” constituents versus those of the Republican incumbent. 10 The Redrawn variable is a wild card. All things constant, it is not certain how this segment of the electorate behaves since they have not been represented by either of the incumbents (Desposato & Petrocik, 2003).
We include controls for the Democratic (two-party) proportion of the most recent presidential vote and in some cases, we can control for race by including the proportion white. Generally, the higher proportion white should be negatively related to the Democratic House vote as more non-whites/minorities affiliate with the Democratic Party. In most cases we also run regressions weighted by population since there can be significant variation in the populations of the geographic units, especially counties. This model is tweaked for a co-partisan dueling incumbent contest in Louisiana (LA 3 in a 2012 Republican runoff). In this case, the dependent variable is the vote share for Republican Charles W. Boustany who wins the race. The key variable is Retained Boustany precincts, leaving his opponent Jeff Landry’s Retained precincts, as the omitted comparison category. It is expected that Boustany will garner a higher share of the vote among his old constituents versus Landry’s constituents.
Finally, our at-large dueling incumbent contests are slightly different than most of the types discussed above. Like Louisiana District 3 in 2012, both Montana in 1992 and South Dakota in 1982 have no Redrawn variable. In collapsing two congressional districts into a statewide race, roughly half the constituents are Retained D with the other half Retained R (Rush, 2000). Ultimately, the incumbent who performs markedly better among their retained constituents or alternatively, not as poorly among their opponent’s retained constituents, will be the victor.
Figure 3 provides a visual summary of the key coefficients (and 95% confidence intervals) from our models with full results available in the appendix.
11
We find the effect for Retained D is positive, and in only two cases (NC in 1968 and IA in 1972) does the coefficient fail to reach significance at conventional levels. In terms of effect size, incumbents do between 2% (PA 2018) and 37% (LA 2012) better among their retained constituents even after controlling for partisanship measured as the district presidential vote. The Louisiana case is the only one where two co-partisan incumbents ran against each other, so we are cautious about drawing inferences from this one type. If we ignore this race, the next largest coefficient is from the Pennsylvania 2002 race, which registers a 25% advantage.
In the three presidential elections in which the redrawn coefficient is statistically significant, it is positive, indicating the Democrat did better in these dueling incumbent matchups (IA 2 in 1992, TX 19 and 32 in 2004, and OH 16 in 2012). In the two midterms where the redrawn coefficient registers significance, it aligns with the partisan direction of short-term conditions (Petrocik & Desposato, 2004), favoring the Democrat in Pennsylvania 17 in 2018 (a Democratic cycle) and the Republican in Iowa 2 in 1942 (a Republican cycle).
Conclusion
Given the strategic behavior of members of Congress (Jacobson & Kernell, 1983), the historical occurrence of general election dueling incumbent contests is a rarity. Nonetheless, due to redistricting and especially under conditions of electoral retrenchment (seat loss) and unified partisan control of mapmaking, since 1843, incumbents have vied against each other more than 70 times. In this note we have documented these cases at the district level and taken another step by examining the extent to which incumbents’ retention of their old constituents aids their reelection prospects in several more recent contests with the analysis of sub-district data (e.g., county and precinct).
In nearly every case we find that incumbents do significantly better among the constituents they represented prior to redistricting vis-à-vis the voters belonging to their incumbent adversary. Nonetheless, there is considerable variation in the extent of this electoral advantage, and it appears to have declined in more recent years of nationalized elections (Jacobson, 2015). But retained constituents are just one important feature of these House races. In many instances there also exists a nontrivial segment of the district population comprised of redrawn voters new to both incumbents. 12 Depending on the partisan bent of these voters, and the direction of the partisan tide (Goedert, 2017), redrawn voters can be the pivotal group in determining the outcome. Additionally, we addressed the ability of majority parties to target the out-party. Our evidence suggests they draw maps that on paper advantage their party in terms of retaining prior population and drawing in co-partisans (based on the district-level presidential vote). However, the election results tell a more balanced story as the winners are split relatively evenly between both in- and out-party incumbents. Perhaps this outcome is not entirely surprising because, all things constant, some incumbents are more formidable candidates than others, and incumbent duels are by far the most competitive House contests (see Table 1).
To our knowledge, though brief, this study is the most comprehensive in evaluating dueling incumbent House contests and yet we have merely scratched the surface regarding this novel line of research. For instance, future analyses could examine dueling incumbent primaries. For partisan mapmakers, these types of races might be the best way to target the out-party because at least one incumbent is guaranteed to lose reelection. More complete coverage of the downstream effects of congressional redistricting will further enhance our understanding of the electoral dynamics shaping these vastly understudied but clearly consequential contests.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Notes
Appendix
Analysis of At-Large Dueling Incumbent Elections
| South Dakota 1982 | Montana 1992 | |
|---|---|---|
| Democratic pres. Vote (%) | .911 (.069)* | .963 (.061)* |
| White (%) | −.070 (.031)* | −.010 (.048) |
| Retained D | .036 (.013)* | .082 (.012)* |
| Constant | .255 (.042)* | −.040 (.062) |
| N | 66 | 56 |
| Adjusted R2 | .853 | .900 |
Note: The dependent variable is the Democratic (two-party) percentage of the House vote. Data are at the county level. The omitted comparison category for Retained D is Retained R.
*p < .05 (two-tailed).