Last month, FairVote released its projections for the November 2018 U.S. House elections that will take place nearly two years from now. If every current incumbent (excluding the five members of the 115th Congress who have already vacated their seats) were to seek re-election, we can confidently project that at least 368 of them, 205 Republicans and 163 Democrats would win (Figure 1).
We can make full projections, that is, projections for all 435 congressional seats, grounded in assumptions of different kind of election years. For example, assuming an election year in which Americans are evenly split between the two parties, we project wins for 190 Democrats and 245 Republicans, which would represent a net loss of four seats for House Democrats.
If voters instead swing in favor of Democrats as strongly as they did in 2006, with an overall advantage among voters of 54% to the Republicans' 46%, then we project that Democrats would win 208 seats, whereas in 2006 that vote distribution delivered 256 seats to Democrats. Conversely, if voters swing to the GOP by the same 54-46% margin, as they did in 2010, we project Republicans would win 250 seats, compared with 242 in 2010.
In 2014, we published high-confidence predictions for 361 2016 U.S. House races, 202 which we forecasted Republicans would win, and 159 which we forecasted Democrats would win. Those predictions were 100% correct. In 2016, our full projections were 99.5% accurate in terms of the net makeup of the House of Representatives (we projected 192 seats for Democrats, 194 Democrats won) and 97% accurate in terms of forecasting individual outcomes (only 7 Democrats and 5 Republicans won races we projected them to lose). Our full 2018 projections, based on the same methodology, project that 429 incumbents will win re-election
Here are key facts from our 2018 analysis:
Projections Reveal Startling Lack of Competition and Structural Bias.
The 368 districts that are truly safe for their incumbents represent more than 84 percent of all House seats.
Only 24 districts are labelled as true "toss-ups," representing just five percent of House seats.
Of the Republicans' 245 seats, we project 205 are virtually certain to be retained, just 13 short of an absolute majority. Democrats start with 163 near certain winners.
If no incumbents ran, all seats were open seats, and voters nationally split 50-50 in their preference for Republicans or Democrats, we would expect the resulting House to consist of 237 Republicans and 198 Democrats. Representatives from both parties benefit from incumbency, but Republicans, who have more incumbents, stand to benefit significantly more. Additionally, our data shows that Republican incumbents benefitted from an incumbency bump about 1 point larger than Democrats, which along with national two-party congressional preference explains how they were able to win in 23 districts that Clinton carried. However, while accounting for incumbency explains why Democrats are unlikely to take 8 additional seats, the incumbency bump fails to explain the vast majority of Republican structural advantage.
Even if in 2018 as many as 55% of U.S. voters preferred a Democratic House to a Republican one, Republicans would likely retain control of the body. Based on our projections Democrats would need to be preferred to Republicans by 55.5% to obtain a narrow majority in the House. The incumbency bump plays a role here; without it Democrats would be projected to win the House with 52.8% of the vote. This suggests that a continued decline for the incumbency bump (2014 was the lowest year for this metric since 1992, with only a minor recovery in 2016) could present an opening for Democrats.
Why is Democratic control of the House so unlikely, even in the case of a major wave year for Democrats? Gerrymandering, campaign spending and incumbency advantage play a role, but the biggest cause is well understood: Republicans are distributed in a more geographically advantageous way than Democrats for single-winner geographic districts. Democratic voters are mostly situated in large cities with small geographic footprints. As a result, Democratic candidates win their 163 safe districts by much higher margins than they need to, while Republican candidates win suburban and rural districts by lesser margins. The average safe Democratic district was 68% Democrat, 18 points from equal distribution, while the average safe Republican district was only 64% percent Republican. While this may seem like a small difference the additional 4% of Democratic votes that are functionally wasted could have made a huge difference in the 28 districts with between 46% and 50% Democratic partisanship.
For all our projections and detailed information for each district, please download our Monopoly Politics 2018 spreadsheet, which will be continually updated as House membership changes over the next two years. As described more fully below, the spreadsheet allows you to simulate projections if voters nationally favor Republicans or Democrats and to simulate outcomes with all incumbents seeking re-election or if every seat were open.
Summary of Monopoly Politics 2018 Projections:
FairVote first released its Monopoly Politics projections in 1997. Monopoly Politics 2016, the most recent installment, is a comprehensive report analyzing the effects of an increasingly polarized electorate coupled with a House of Representatives elected exclusively from single-winner districts. That report contained projections for 361 of the 435 house seats, using a straightforward metric that did not rely on polling data, campaign expenditures, or any of the other typical information used for predicting races. Instead, it projected which party's candidate would win a seat based on how voters in that district had voted in the 2012 presidential election and 2014 and 2010 congressional elections, along with a measure of incumbency strengthFirst released in the November 2014, just weeks after the November 2014 election, and updated as seats became open, our report was correct in 100% of high-confidence projections, and 97% of other projections.
Understanding Our Projections:
Because these projections rely only on past election performance of the incumbent and the presidential candidates, it becomes possible to project 2018 elections immediately following publication of the 2016 election results. Consequently, we can now project the results of 369 congressional contests with full confidence that these projections are near certain so long as incumbents choose to seek re-election. We make projections for the remaining 66 seats as well. Due either to even partisanship or partisan volatility these seats are more susceptible to partisan waves and other exogenous factors and so are therefore riskier to project.
As you view the full spreadsheet, keep in mind that by default it assumes that nationally voters will be split 50-50 between those who prefer a Democratic House and those who prefer a Republican House. However, you can change the "National Party Preference (Dem.%)" by altering the number in cell E9 of the spreadsheet, and the projections will automatically updated to reflect that national split, rather than the default 50-50 split.
Additionally, the spreadsheet by default assumes that every incumbent will seek re-election, but it also provides the projection for each seat if the incumbent does not seek re-election. As candidates announce that they will not seek re-election, the table will be updated, so that the final projections will reflect the reality of which incumbents are running and which are not.
Note also that there are two unusual circumstances that may change some projections. First, although there is no regular redistricting until 2021, at least one state (North Carolina) is likely to redraw district lines under court order. In addition, if any incumbents leave office mid-term and are replaced in special elections, then their districts will have an incumbent, but the incumbent will not have the same incumbency metric that the former member had. We will update the spreadsheet as needed any time this occurs.
Mid-term elections typically feature significantly more partisan volatility than presidential elections, due to an“enthusiasm” gap” that spans multiple demographics. Older people, more educated people, and people who identify with the out-of-power party are all likelier to vote in a midterm year than the young, the less educated, and members of the party in power. These relationships will significantly impact Democrats' prospects for retaking the House in contradictory and overlapping ways.
We first released our full projections for the first time in our 2016 Monopoly Politics report, these were 100% accurate, likely because 2016 wound up being very close to a 50-50 partisanship election. Considering that our model is based on a 50-50 partisanship year, we are unlikely to reach the same level of accuracy as our 2016 predictions given the inherent volatility of a midterm election, particularly one in which the president's party controls all levels of government. As of February 2017, lists compiled by both parties' congressional committees include many of the districts we identify as true toss ups (and therefore do not include among our high-confidence projections), but also a few districts our model sees as safe. This suggests that Democrats are banking on a Democratic wave year, as we modeled above, while Republicans believe that structural advantages in their favor can continue to erode support for congressional democrats.
Still, it is extremely unlikely that our high confidence projections will fail given that they have remained mostly accurate through wave elections such as 2006 and 2010. We expect to at least match our 99.7% accuracy from our 2014 projections in these new projections for Election Day 2018.
FairVote proposes reforms aimed at combating problems with our congressional elections, such as ranked choice voting and fair representation voting, which can be found on our website. For more information on partisan skew and noncompetitive districts check out our upcoming Monopoly Politics 2018 report, and our Monopoly Politics 2016 report.