Read different news articles, and you’d be forgiven for not being sure whether the House map is expanding or contracting. It was less than two weeks ago, for instance, that The Washington Post claimed the prospect of a Democratic wave had diminished because of “President Trump’s rising approval rating and the polarizing fight over Supreme Court Justice Brett M. Kavanaugh.” Those factors were limiting Democrats’ upside in “conservative and rural districts” where Trump was popular, the article said, probably confining their biggest gains to wealthy suburbs.
But a New York Times article published this weekend told a totally different story, one of a broader House map where, bolstered by their unprecedented fundraising advantage, Democrats had Republicans on the defensive in “more conservative, exurban areas” in addition to the suburbs, as the “field of competition” had grown “well beyond 40 seats.” Recent Republican ad buys in deeply Republican districts such as South Carolina 1 would also seem to support the map-is-broadening hypothesis.
So which case is right? Is the House playing field getting bigger or smaller? The answer is … a bit complicated. In fact, the different versions of FiveThirtyEight’s House model tell somewhat different stories about it.
FiveThirtyEight Senate forecast update for Oct. 30, 2018
The Times’s article is more recent, however, and therefore probably closer to the mark. In early October, in the aftermath of the Kavanaugh hearings, you could make a good case that the map was contracting. But after those stellar Democratic fundraising numbers were reported earlier this month, and after Democrats continued to poll well in generic ballot and district-level polls, the map now appears about as broad as ever, with at least twice as many seats “in play” as in the last midterm, 2014.
As I mentioned, this is one of those times when the different versions of our model — Lite, Classic and Deluxe — show different trajectories. And I think they can help explain why reasonable observers might come to different conclusions about the state of play in the House. But first, let me just show you what they say. The chart below tracks the number of competitive races for each day since Aug. 1 according to the three forecasts, where a competitive race is defined as one in which each party has at least a 5 percent chance of winning.
In the Lite model, which is based on district-level and generic-ballot polling only — and which uses an algorithm we call CANTOR to impute polling in districts that don’t have much of it — the number of competitive districts has gradually been decreasing over time, from about 140 (!) in August to a still-very-high 130 or so today. But that mostly reflects the increasing abundance and availability of district-level polling. Initially, Lite starts with a long list of districts that just might be competitive — races it would place in the “likely Republican” or “likely Democratic” categories (as opposed to “solid Republican” and “solid Democratic”). As polling comes in, some of these districts can be eliminated from being considered competitive, while others get promoted into the more intensely competitive categories (“lean Republican,” “lean Democratic” or “toss-up”).
There’s a big fluctuation in the Lite forecast in late September and early October, however. That reflects a period of comparatively weaker polling for Democrats in both the generic ballot and in district-level polls, which coincides with the apex of the Kavanaugh hearings. However, the Democrats pulled out of that polling slump a couple of weeks ago, with the generic ballot now back to showing an advantage for them of 8.5 percentage points and district polls showing vulnerable Republican incumbents in many different kinds of districts.
Democrats’ ceiling is sensitive to these minor changes in the political environment because there are a glut of districts that are somewhere between 10 and 20 percentage points more Republican-leaning than the country as a whole. Many of these districts were gerrymandered to be “safe” for Republican candidates. And in a medium-sized wave — say, one that featured a Democratic lead on the generic ballot of 6 to 7 percentage points — they probably would be safe. But once Democrats get up to about an 8- or 9-point lead instead, they’re really bumping up against the walls of these gerrymandered districts, and they may encounter incumbents who are underfunded and otherwise not-well-prepared for the challenge. In that sense, The Washington Post’s analysis was right: A relatively minor shift back toward Republicans, especially in conservative areas, could have really limited the Democrats’ upside potential. It was just mistimed; the shift that occurred toward Republicans in early October is no longer really evident in the polls now.1 Instead, Democratic candidates for Congress continue to poll well in some Trump-friendly districts as well as those that voted for Hillary Clinton for president.
Our Classic model incorporates “fundamentals” in addition to the polls — most importantly, including fundraising. And it shows the number of competitive districts generally having increased over time as Democrats’ fundraising numbers get better and better, especially in Republican-leaning areas where you’d ordinarily expect the incumbent to have a heavy fundraising advantage. In particular, there’s a spike in the number of competitive districts in the Classic forecast on or about Oct. 15, when third-quarter fundraising totals were reported to the FEC and incorporated into the model. The Classic model now shows around 110 competitive districts, the vast majority of which are currently held by Republicans. Since Democrats only need to win 23 of these to win the House — potentially plus a few extra to cover a couple of districts of their own they might lose — you can see why it’s become pretty bullish on Democrats’ chances.
Finally, there’s the Deluxe version of the model, which incorporates expert ratings such as those published by the Cook Political Report. That version shows the number of competitive districts as having been steady, at around 100, throughout the election cycle. Our model has its disputes with the expert ratings; in particular, based in part on Democrats’ fundraising prowess, it thinks the set of districts that Cook et al. rate as “likely Republican” are liable to be more competitive than they are ordinarily. But one thing that’s historically very rare for the expert ratings are false negatives — districts they rate as “solid” or “safe” are almost never won by the underdog. Instead, they tend to put any district where an upset is even thinkable into the “likely Republican” or “likely Democratic” categories. In general, these expert raters have shifted districts that were already deemed to be competitive leftward (e.g., from “lean Republican” to “toss-up”) over the course of the election. But they haven’t added to or subtracted much from the list of potentially competitive districts. Thus, the number of districts deemed to be competitive by Deluxe has held pretty steady.
But if the Deluxe forecast is right and “only” 100 districts are competitive, that’s still an awfully high number as compared to recent years. Just 38 House districts would have been classified as competitive by the Deluxe model at the end of the 2016 campaign, for example, or 44 of them in 2014.
|Year||Midterm?||number of Competitive Races|
The number of competitive districts in 2010 was also around 100, by comparison, according to the House model that we published at the time.2 It was easy to see why so many districts were competitive that year, however. Democrats had pushed very deep into red territory after 2006 and 2008 and started out the 2010 midterm with a 256-179 advantage in the House. (By comparison, Republicans have “only” a 240-195 advantage now.) The shift against President Obama and the Democrats was very strong: Democrats went from winning the popular vote for the House by 11 percentage points in 2008 to losing it by 7 points in 2010. And districts back then were slightly less polarized and slightly less gerrymandered (although, there was still plenty of gerrymandering) than they are now. Republicans were reclaiming a lot of naturally red territory.
The 2018 House election is also being fought in red territory — only, it’s Democrats who have an opportunity to make big gains there. And it’s taken a whole assortment of factors to make the map as competitive as it is, somewhat despite the odds:
- The large number of Republican retirements.
- The massive Democratic cash advantage.
- Democrats nominating competent candidates in almost every district.
- The declining incumbency advantage. It used to be that all but the worst incumbents would outperform a hypothetical open-seat race in their districts. Now, that isn’t so clear, as the incumbency advantage is only about half as large as it was a couple of decades ago.
- We’re now eight years removed from when maps were redrawn, and some districts (especially in the Sun Belt) have become more competitive over that period.
- As mentioned, Republican gerrymanders do have a breaking point, and some of the more optimistic polling for Democrats puts them close to that breaking point.
- Finally, redistricting in Pennsylvania and Republican scandals add a few additional competitive seats to the list.
All of this sets up what could be a feast-or-famine evening for Democrats next Tuesday. They have a huge number of opportunities to win Republican seats — around 100 of them! And yet, it would be hard to circle more than about 12 or 15 of these districts that can safely be predicted to wind up in Democrats’ hands. If things go roughly to form nationwide, Democrats almost certainly will get there. But even a modest pro-GOP polling error — if Republicans were to beat their polls by 2 or 3 points across the board — would revert the overall race to being a toss-up. Conversely, even a modest, pro-Democratic polling error could send their number of pickups careening into the 50s, or higher. The wide ranges in our forecast reflect the uncertain conditions on the ground.