A while back, we debuted a Wins-Above-Replacement model for the US Senate that tried to assess candidate quality through answering the following question for each race: Assuming everything else (money, national environment, incumbency, etc) was held constant, what would the expected outcome be if the the matchup was a generic Republican vs a generic Democrat?
Today, we’ll try to extend this model to the US House of Representatives, with the help of data from Noah Wyhof-Rudnick, Daily Kos Elections, DRA, and the US Census Bureau. In our model, we use a multilinear regression that controls for the Presidential lean of the districts over the last three cycles (2012, 2016, 2020), the raw-dollar spending delta between candidates, incumbency, and the demographics of the districts. Our results are displayed on the map below, and an interactive version with a full, detailed table is available here.
Before delving into the macro-level takeaways, it’s worth touching on some individual performance highlights. Firstly, Republican candidate quality and downballot overperformance delivered them up to 7 districts that Democrats were expected to have won in hindsight. John Katko (NY-24), David Valadao (CA-21), Young Kim (CA-39), Maria Elvira Salazar (FL-27), Carlos Gimenez (FL-26), Ashley Hinson (IA-01), and Mike Garcia (CA-25) all won seats that, based on fundamentals, were favored to be won by the Democratic candidate instead. Meanwhile, the only Democratic candidate who overperformed enough to change the outcome of a race was Jared Golden (ME-02), who overperformed by nearly 7 points even as Trump carried his district comfortably.
The strongest Democratic incumbent in a competitive race was Collin Peterson (MN-07), who overperformed fundamentals by a whopping 10 points, followed closely by Ben McAdams (UT-04), who overperformed the fundamentals by 9. On the Republican side, the strongest two candidates in competitive seats were David Valadao (CA-21) and John Katko (NY-24), both of whom overperformed by over 13 points, enabling them to narrowly win seats that should have been comfortably Democratic. The weakest electoral candidate on either side was likely Ilhan Omar, who won a Biden +63 seat by less than 40 points, underperforming fundamentals by just over 17 points
Analyzing the map as a whole, it is worth noting that the single most predictive factor, as previously discussed at Split Ticket, is the district’s 2020 presidential lean. That said, looking at the results, a few things pop out as striking. Firstly, a few Democrats that overperformed Biden in margin, such as Cindy Axne, are found to have underperformed relative to fundamentals, while some that underperformed him, such as Conor Lamb, are listed as overperformers. The reason for this is that presidential lean, while increasingly predictive, is not the only factor that matters in elections, and in many close districts, several other elements, such as incumbency and spending, can tip the scales of expectations in the opposite direction.
This brings us to a larger takeaway from the model: the importance of campaign spending. In the wake of 2020, this factor was often mocked as overrated, especially as Democrats experienced a somewhat disappointing cycle relative to polling and fundraising expectations. However, as covered by Noah Rudnick and as further reinforced by our modeled findings, the notion that money does not matter is quantifiably false — spending was one of the biggest deciding factors in several closely contested seats, and our explanatory model found that a point in margin gain corresponded to roughly $1,000,000 in net spending advantage, on average (though there is variance depending on the media market and spending strategy). Democrats will probably rely heavily on the increased spending power of their base and campaign apparatus to keep their chances at holding the House afloat in 2022, where the value of a dollar will be critical to their (still non-negligible) chances of maintaining control of the chamber in 2022.
Finally, a model like this allows us to assign a rough point value on incumbency, which remains important (if declining in power). Our estimate is that on average, incumbency provided a ~1.7 point boost in a candidate’s vote share (or roughly three-and-a-half points in margin), which is a slight decline from Gary Jacobsen’s 2018 estimate of ~2 points in vote share. But this is still a non-trivial benefit, and if Democrats didn’t have this going for them, Republicans likely would have won the House in the last cycle itself. The number of retirements in this cycle blunts a good deal of the advantage that Democrats would have enjoyed on this front; however, Republican hopes of winning the majority will still run through flipping seats like the Trump +0.1 district held by Cindy Axne in Iowa. In places like these, incumbency will come as a welcome boost to Democratic House members, who will likely be swimming against the tide in trying to stave off Republican gains.
Our model helps us approximate the impacts of incumbency and fundraising, and the lessons learned may be instructive for 2022. While the environment in this coming cycle will likely be significantly to the right of the one in 2020, a combination of high spending, candidate quality, and incumbency benefits could see Democrats stave off some losses in important seats. This will likely not be enough to enable them to retain control of the House in 2022, but it would still make the task of winning the chamber in 2024 substantially easier.
Given their electoral fortunes of late, this is probably a trade several of them would gladly take.
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.