In today’s era of hyperpolarization, the single most accurate predictor of a district’s congressional vote is the presidential lean of the seat. As we previously analyzed here at Split Ticket, 97% of a district’s 2020 congressional vote could be explained simply by its presidential lean that year.
With this in mind, it is likely that 2022 will similarly have a massive correlation between the congressional and past presidential margins. However, a retrospective look at the 2018 cycle shows that other factors like spending and incumbency have an extremely large impact on a district’s vote as well. Moreover, without the presidential race on the same ballot to serve as a regularizer, candidate quality begins to matter more and the potential for significant deviations from baseline partisanship becomes much more likely as a result, causing some results that absolutely nobody would have predicted if assessing races purely by presidential leans.
To analyze the 2018 cycle, we created a Wins-Above-Replacement (WAR) model for the cycle (similar to our 2020 one), which factors in money, incumbency, demographics, and 2016/2012 presidential leans on a district-level basis to calculate how a candidate was expected to perform in the district, given results across the nation. By comparing this to the actual performance, we can get a sense of a candidate’s electoral “value over replacement” in the cycle by margin and get a measure of individual candidate strength. Moreover, by separating out candidate quality from a race, we can also get a better picture of the importance of factors like money and incumbency.
A map of our model is shown below, and a tabular view and interactive map can be found here.
Firstly, the potential for significant overperformances is much greater in a midterm year, showing how candidate quality really does appear to matter more. In 2020, the best-performing Democrat was Brian Higgins, who performed 12.6 points better than expected. In 2018, 11 Democratic candidates performed better than this; in fact, the best performer in either party (Democratic challenger Richard Ojeda, in West Virginia’s 3rd district) actually performed a whopping 27 points above expected, an overperformance that would be unfathomable in a presidential year. The potential for high overperformances is also found on the Republican side, where 13 candidates had double-digit overperformances, as compared to the 5 that did in 2020.
In terms of 2022 implications, we might see candidate quality generally play a bigger role altogether in the upcoming cycle than it did in 2020, which could mean that districts might move significantly from their baseline 2020 partisan lean. For instance, it is easy to envision a universe in which Sharice Davids (D-KS 03) hangs on in a Biden +4 seat while Mike Levin (D-CA 49) loses a Biden +11 seat, simply because Davids’ tends to overperform by a lot more; in fact, her performance-above-expectations was an average of 8 points better than Levin’s across 2018 and 2020. The district’s presidential result, while highly instructive, is not the only thing that matters, and in battleground seats, it is likely that other factors like spending and candidate quality could be enough to tip the seat in either direction.
Our 2018 WAR model’s results also reinforce the importance of moderation. In fact, 14 of the top 15 Democratic overperformers were fairly moderate or median Democrats, which once again fits into the pattern highlighted by Split Ticket with regards to moderation and electoral overperformance. Moderation is strongly correlated with electoral overperformance, and a candidate’s ability to court crossover voters will probably be one of the biggest determining factors in close races in 2022, especially as the presidential race disappears from the ballot in a midterm year and candidates distance themselves from the national party when needed.
Because we purposefully did not control for 2018 statewide race results, regional trends is one thing that this model occasionally struggles with, and results should thus be contextualized appropriately. One may notice a systemic Democratic underperformance in Florida, for example; here, it is less likely that every Florida Democratic candidate was subpar and significantly more likely that Florida became far more unfavorable to Democrats between 2016 and 2018, resulting in almost every Democratic candidate there performing significantly below expectations. The impact of statewide races on House margins is one we would like to explore more in the future; however, for the purposes of consistency, we felt it prudent to exclude it in this iteration.
Moving on to macro-level quantifications, we found that incumbency was worth approximately 4 points on margin in 2018, which was a spike up from the ~3.5% it was worth in 2020. Whether this value declines again in 2022 or rebounds as the presidential race vanishes from the ballot is yet to be seen, and strong cases can be made for either scenario; however, in the latter case, Democrats would likely see their chances of holding the House go up, given their current majority and the state of play across many battleground districts.
In contrast to incumbency, money actually played a comparable role in both 2018 and 2020; our estimates are that $1 million in net spending roughly corresponded to a net point in margin. The money advantage is still highly significant for both sides, and control of the House could well be decided by whether endangered Democratic incumbents like Cindy Axne and Jared Golden swamp their opponents financially, or whether they are matched by their challengers financially.
Without the presence of the presidential race on the ballot, district presidential lean was also less important in 2018 than in 2020; in 2018, approximately 95% of a district’s result could be explained by the presidential result, as compared to 97% in 2020. It remains to be seen as to whether the importance of the presidential result similarly declines from 2020 to 2022; if it does, even the small amount of resulting partisan depolarization may still be significant enough to decide the result in some battleground races.
Presidential lean is highly instructive in determining a district’s final congressional results, but incumbency, spending, and campaign and candidate quality still do matter a great deal, as we’ve highlighted above. With a House map that is shaping up to be quite balanced in overall partisan lean, and with a shrinking number of battleground seats, every point of margin could matter a great deal in determining the party that wins the House in 2022. Understanding the factors at play is critical for parties and candidates seeking to overperform baselines and spring surprises.
(Note: A prior version of this piece posted a version of our model which was run on an older dataset with a few minor errors. This was corrected and the piece updated to reflect the latest model).
I’m a software engineer and a computer scientist (UC Berkeley class of 2019 BA, class of 2020 MS) who has an interest in machine learning, politics, and electoral data. I’m a partner at Split Ticket, handle our Senate races, and make many kinds of electoral models.