Previous Split Ticket research indicates that a district’s presidential lean is the single greatest predictor of its House result in a presidential year. Polarization has only reinforced this conclusion, causing the total number of “crossover” seats to shrink significantly over the last two decades. (Read more here) Since there is no sign that this current cycle of polarization will cease any time soon, the better analysis question is two-fold:
- Is a district’s presidential lean still a significant predictor of its result if the election cycle is a midterm?
- If so, has the R2 predictive value markedly increased from cycle to cycle in a pattern indicative of the effects of polarization?
By researching the relationship between past midterms and their preceding elections, Split Ticket should be able to provide data-driven insight into how the 2020 presidential race might predict the scale of the GOP’s expected 2022 wave. This is in no sense the exclusive way to predict a midterm result, but it does afford an avenue to understand the basic parameters of midterm cycles as well as the thresholds for pick-ups.
This analysis breaks down each of the midterm elections since 2006 by four metrics. Metrics 1 & 2 deal with the predictive power of a district’s presidential lean on its House result, while metrics 3 & 4 account for average shift calculations and partisanships of flipped seats.
These benchmarks can be used to estimate the framework of President Biden’s upcoming midterm. Before delving into the meat and potatoes of the analysis, let’s break down the reasoning behind the four aforementioned criterion.
Metric 1 – Presidential Lean – The presidential lean of a given district is measured by vote share, a figure that Split Ticket prefers over raw margin because of the impact of third parties. To determine whether the Democratic or Republican presidential candidate’s vote share is used to qualify a given midterm, we look at the party that each off-year cyle favored.
For instance, our 2010 analysis compares John McCain’s 2008 vote share to that of House Republicans because Obama’s first midterm heavily favored the GOP. Following that line of reasoning, 2006 and 2018 can be classified as Democratic years while 2010 and 2014 fall into the Republican column.
To determine R2, we regress (y-var) House share on (x-var) Presidential share. The resulting number shows how much of the variation in House vote share for the victorious party in a given midterm election can be explained by the vote share won by that party’s presidential candidate in the prior national election.
Metric 2 – Under/Overperformance – For each of the four midterms that Split Ticket analyzed, the top five under- & overperforming candidates for the victorious party were determined. These numbers were simply calculated by subtracting presidential vote share from House vote share, ranking the districts by the size of that deviation on a number line scale, and extracting the top five extremes from both categories.
Negative numbers (-) measure House candidates receiving lower vote shares than the previous presidential nominee of their parties, while positive numbers (+) indicate opposite outcomes.
Metric 3 – Average Shift – This calculation is designed to measure the average change in district vote share between presidential and midterm cycles. To get this number, previous presidential share is subtracted from House share. Think of this figure as the average under-/overperformance between different elections.
Metric 4 – Flipped District Partisanship – This number measures the average presidential lean (based on vote share) of all of the seats picked up by a victorious party in a given midterm cycle. It creates a threshold for midterm flips that shows just how strong a given environment was while simultaneously shedding light on the types of districts that changed hands.
Important Notes & Exclusions
No uncontested districts are included among these data. Races without two major party contenders are also excluded because it is not valid to use the vote share of a Republican or Democratic presidential candidate to explain the performance of a third party hopeful. The remaining data parameters are provided below, grouped by cycle for the sake of clarity.
- 2006 (D) Kerry share vs. House D share
- 2010 (R) McCain share vs. House R share
- 2014 (R) Romney share vs. House R share
- 2018 (D) Clinton share vs. House D share
*Since Louisiana uses a jungle primary system, vote share has been combined by party affiliation, regardless of candidate count, in districts where incumbents won reelection without runoffs*
The 2006 R2 of 0.708 might seem high when looked at abstractly, but it is actually a rather low number when one considers the fact that 2018’s R2 was 0.943 – a 23.5 point jump over the course of just 12 years.
Although a strongly positive, linear relationship existed between the 2004 Kerry & 2006 House Democratic vote shares, there was also a predictably-higher level of depolarization fostering greater variance.
In other words, presidential vote share was less predictive of midterm House results in 2006 because there were more crossover district representatives benefitting from split-ticket voting. (i.e., Bush-district Ds and Kerry-district Rs)
Analyzing the top five best and worst Democratic candidates hammers this point home very clearly, suggesting that lower levels of polarization made candidate quality and local brands more important than partisan affiliation in specific circumstances.
Mike Castle is an ideal example. As a moderate Republican and former Governor, he had the perfect qualifications to comfortably win an at-large seat in a state that tracked quickly toward the Democrats during the Clinton Presidency.
2010’s R2 of 0.832 represented a significant increase from 2006 that was directly associated with the rise of polarization in the Obama-era, particularly among rural and working-class white voters. This midterm also saw greater turnover than its predecessor, though that was mostly because Democrats had so much ground to lose after two successive electoral waves.
Additionally, many – but not all – vulnerable Democrats held McCain seats that fundamentally favored Republicans. That’s part of the reason why McCain’s 2008 vote share was a better midterm barometer than Kerry’s 2004 performance.
Looking at the under-/overperformances adds to the polarization painting because no Republican candidate exceeded McCain’s vote share by >30 points, the margin that five aforementioned Democrats had passed Kerry’s with in 2004.*
The second graph also shows that 2006-style ticket-splitting was still a major factor in 2010 despite being diminished. Gene Taylor is a great example of this. Taylor was the leading Democratic overperformer in 2006, receiving 49 percent more of the vote than Kerry did. He did lose in 2010, but his Republican opponent Steven Palazzo received 16 percent less of the vote than McCain.
*Many of these districts already had presidential leans that were too Republican to allow for a GOP overperformance of 30 or 40 points*
In 2014, our R2 was 0.873 – marking only a modest increase from 2010. Although Obama-era polarization was still a major factor by the time the President’s six-year itch rolled around, the foundational trends of the post-2010 realignment had already started to settle in.
Successful GOP gerrymandering during the 2012 redistricting cycle, atrocious voter turnout, and a general lack of competitive races could also be to blame for the slight increase. After holding the House the previous cycle despite losing the national popular vote, the GOP made a net gain of 13 seats to reach a 247-seat watermark majority.
Despite the bleak decline of ticket-splitting and crossover seats between 2006 and 2014, there was still a larger level of variance than there is today. Romney’s 2012 share was on the whole quite predictive of House Republican performances, but it was not instructive everywhere.
In HI-01, former Congressman Charles Djou – a moderate – received almost 20 points more of the vote than Romney in his bid to win back the seat he had lost in 2010. The flip side was WV-02, where Maryland St. Senator Alex Mooney carpetbagged to run for future Senator Shelley Moore-Capito’s open House seat and received 13 points less of the vote than his party’s last presidential nominee.
Trump-era polarization accelerated the underlying trends that had once defined the Obama-era. It is therefore no surprise that 2018’s R2 value clocked in at 0.943 – a full 24 points above the 2006 level. Although Trump’s one and only midterm was successful for House Democrats, it more or less confirmed that the realignment was rapidly progressing.
Democrats were still able to compete in heavily-Republican seats like NY-25 and IA-04 as a result of special circumstances involving controversial incumbents like Chris Collins and Steve King, but the days of winning seats like IN-08 or TX-17 with conservative Democrats like Brad Ellsworth and Chet Edwards had passed into the annals of history.
Even though the 2016 presidential vote had a higher degree of midterm predictive validity than previous cycles did, there was still variation in 2018 due to ticket-splitting. Suburban Republicans like Brian Fitzpatrick (PA-01) and John Katko (NY-24) won close races in Clinton-seats by outrunning fundamentals.
At the same time, Democrats like Collin Peterson (MN-07) and Jared Golden (ME-02) fought and won in districts where Clinton had devastated the national Democratic brand. Others like Ron DiNicola (PA-16) and Richard Ojeda (WV-03) came up short but posted impressive overperformances anyway.
Overall Conclusions – Average Shift
The average shift measures the mean change in district vote share between presidential and midterm cycles for every seat included in the data set. For instance, the average Democratic House candidate in 2006 received 5 points more of the vote than Kerry did in 2004. Over the four midterm years that Split Ticket analyzed, the shift value remained remarkably consistent.
Overall Conclusions – Flipped District Partisanships
What does it mean for 2022?
Assuming the calculations for average total shifts and flipped district partisanships hold up this fall, Republicans should be able to pick up Democratic-held seats in Trump and marginal Biden districts with relative ease. The GOP should be able to compete for reach seats in which Biden received 55-58% of the vote too, though these targets will be harder to hit outright.
Because the GOP’s 213 seat floor is much higher than its position in 2010, only a handful of Democratic-held districts will need to change hands to flip House control. Given the state of the national environment and President Biden’s approvals, it appears increasingly likely that Republicans will make large gains in the nation’s lower chamber this year. The last important thing to remember is that wave years often produce “fluke” flips, pick ups that tend to be temporary. There could be quite a few of those races this fall in unexpected places.
My name is Harrison Lavelle and I am a political analyst studying political science and international studies at the College of New Jersey. As a co-founder and partner at Split Ticket, I coordinate our House coverage. I write about a variety of electoral topics and produce political maps. Besides elections, my hobbies include music, history, language, aviation, and fitness.
Contact me at @HWLavelleMaps or email@example.com