Skip to main content
Election Update: Democrats Are In Their Best Position Yet To Retake The House

If Labor Day is the traditional inflection point in the midterm campaign — the point when the election becomes something that’s happening right now — then Democrats should feel pretty good about where they stand in their quest to win the U.S. House.

Although it can be a noisy indicator, the generic congressional ballot is showing Democrats in their best position since last winter, with a handful of high-quality polls (including one from our ABC News colleagues) giving them a double-digit advantage over Republicans. Meanwhile, President Trump’s approval rating — as of late Tuesday morning, an average of 40.1 percent of adults approved of his performance according to our calculation, while 54.1 percent disapproved of him — is the worst that it’s been since February.1

As a result, Republicans are in their worst position to date in our U.S. House forecast: The Classic version of our model gives them only a 1 in 5 chance of holding onto the House. Other versions of our model are slightly more optimistic for the GOP: The Deluxe version, which folds in expert ratings on a seat-by-seat basis, puts their chances at 1 in 4, while the Lite version, which uses district-level and generic ballot polls alone to make its forecasts, has them at a 3 in 10 chance. Whichever flavor of the forecast you prefer, the House is a long way from a foregone conclusion — but also a long way from being a “toss-up.”

There are three questions that we ought to ask about this data. First, why have the changes in presidential approval and the generic ballot happened? Second, how likely are they to stick? And third, how much do they matter?

Question 1: Why has Republicans’ position apparently been worsening?

Well, I don’t know. The changes are modest enough that they could be statistical noise. (More on that point below.) But, the fact that both Trump’s approval rating and the generic ballot are moving in the same direction should make us more confident that the trend is real and should suggest that both indicators have something to do with the president.

One clue is that recent polls show an increase in support for special counsel Robert Mueller as well as an increase in the number of voters who want to begin impeachment proceedings. The most obvious explanation, therefore, is that the changes have something to do with the Russia probe and the other investigations surrounding Trump and his inner circle, including the news late last month about Trump’s former attorney, Michael Cohen (who pleaded guilty to illegal campaign contributions), and his former campaign manager, Paul Manafort (who was convicted on eight counts in his tax fraud trial). Before you snicker that none of this matters because the public doesn’t care about “the Russia stuff,” keep in mind that the largest change in Trump’s approval numbers came after he fired former FBI Director James Comey last May, the event that touched off the Mueller probe in the first place.

Another plausible explanation is that voters are tuning into the campaign to a greater degree than they had before — and not liking what they see once they give Trump and Republicans a longer look. The stretch run of the midterms begins on Labor Day, and the generic ballot has historically been a more reliable indicator once you pass it.

There’s just one problem with that explanation. Although we’re past Labor Day now, the polls filtering into our averages were mostly conducted in late August.2 Also, most of the generic ballot polls were conducted among registered voters rather than likely voters. (Pollsters traditionally switch over to likely-voter models in polls conducted after Labor Day.) Republicans usually gain ground when you switch from registered-voter to likely-voter polls, but with signs of higher Democratic enthusiasm this year, that may not be true in this election.3 So although we’re very curious to see what post-Labor Day, likely-voter polls will show about the midterms, we don’t actually have those yet.

Question 2: How likely are these changes to stick?

I have a confession — although it’s one that I’ve made before. The method of calculating the generic ballot that we use on our generic ballot interactive, which currently shows Democrats ahead by 10.8 percentage points, is too aggressive and will usually overestimate swings.4 Our House forecast actually uses a different, slower-moving version of the generic ballot average. In that version, Democrats currently lead the generic ballot by 8.7 percentage points. That’s still pretty good, but the House forecast will need to see Democrats sustain their most recent numbers for a few weeks before it concedes that they’re really up double digits.

Your guess is as good as mine — and probably as good as the model’s — as to whether they’ll do that. Historically, it’s been rare for a party to win the popular vote for the U.S. House by double digits.5 But it wouldn’t be surprising if Democrats wound up with a popular vote margin in the high single digits (i.e. 8 or 9 percentage points) rather than the mid-to-high single digits (i.e. 6 or 7 percentage points). In fact, our model calculates a historical prior based on long-term trends in midterms and presidential approval ratings. That prior figures that Democrats “should” win the popular vote by 8 to 9 percentage points given that midterms are usually rough on the president’s party and that Trump isn’t a popular president.

Put another way, a slight Democratic uptick on the generic ballot to 8 or 9 percentage points6 is arguably bringing the race for Congress more in line with historical norms. That’s one reason to think it could hold. But we’ll need to see more evidence — not only from the generic ballot but also from other indicators — to conclude that Democrats are really ahead by 10 or 12 or 14 points, which would produce a gargantuan wave.

Question 3: How much does all of this matter?

It might seem like we’re parsing awfully fine distinctions — e.g., between an 8-point popular vote margin and a 7-point one. But they matter, because it doesn’t take that much for Democrats to go from House underdogs to potentially taking 40 or more seats.

Here, for example, is the output from a Tuesday morning run of our Classic forecast, showing how a projected margin in the House popular vote translates into potential seat gains for Democrats. If Democrats win the popular vote by “only” 5 to 6 percentage points — still a pretty comfortable margin, but not necessarily enough to make up Republicans’ advantages due to gerrymandering, incumbency and the clustering of Democratic voters in urban districts — they’re only about even-money to win the House. If they win it by 9 to 10 points, by contrast, they’re all but certain to win the House and in fact project to gain about 40 seats!

Democrats have no room for error

How the popular vote translates into House seats for Democrats per FiveThirtyEight’s Classic model as of Sept. 4, 2018

Democratic outcome
Popular Vote Margin Projected Seat Gain Chance of Winning House
14-15 point lead +66 >99%
13-14 +61 >99%
12-13 +56 >99%
11-12 +51 >99%
10-11 +46 >99%
9-10 +41 >99%
8-9 +36 98%
7-8 +32 92%
6-7 +27 78%
5-6 +24 56%
4-5 +20 29%
3-4 +16 11%
2-3 +13 3%
1-2 +10 <1%
0-1 +7 <1%

With that said, there are several reasons for caution. The House popular vote doesn’t actually count for anything, and it’s possible that Democrats (or Republicans) could run up the score in noncompetitive districts. In 2006, for example, Democrats did extremely well in noncompetitive seats, enough to win the popular vote for the House by 8 percentage points, but Republicans did well enough in swing seats to hold their losses to “only” 30 seats — not good, but a lot better than what Democrats experienced in 1994 or 2010, for example.

From a technical standpoint, the popular vote is also challenging to model, as you need to not only project the margin of victory in every district — including a lot of districts where there’s no polling data — but also forecast turnout. As compared with the Classic version of our forecast, which has Democrats performing particularly well in swing seats, both the Lite and Deluxe versions think there’s more risk of Democrats wasting votes in noncompetitive seats.7 So I’d approach our model’s forecasts of the popular vote with a few more grains of salt than its forecasts of seat gains or losses.

Bonus question: Apart from these technical issues, why should Democrats still be worried about their ability to win the House?

You mean, aside from the fact that 1 in 5 chances still happen 1 in 5 times, which is kind of a lot?! Or that the Senate map is still very difficult for Democrats, even if the House looks reasonably favorable to them?

One reason for Democrats to remain nervous — and for Republicans to keep their hopes up — is that we haven’t yet seen the sort of polling at the district-by-district level that would imply a House landslide. In particular, some Republican incumbents are holding their own, such as in a series of three polls of GOP-held congressional districts published late last month by Siena College. Although Democrats will pick up a fair number of open seats as the result of GOP retirements, and a couple more as a result of Pennsylvania’s redistricting, they will need to knock off some GOP incumbents to win the majority — and not just the low-hanging fruit, but Republicans like John Faso in New York’s 19th District, which has historically proven resistant to high-profile Democratic challenges.

These district-level polls, taken in the aggregate, aren’t bad for Democrats, but they imply a popular vote win of “only” 6 or 7 percentage points, with some GOP overperformance among incumbents in swing districts. Those polls make the House look more like a toss-up than the generic ballot or other indicators do.

The caveat to the caveat is that, historically at this point in the election cycle, district polls have had a slight statistical bias toward incumbent candidates. (Or at least they have in House races; we haven’t observed that for other types of elections.) That perhaps reflects name-recognition deficits on the part of challengers, some of whom only recently won their primaries or have not even held them yet. But Democrats have recruited competent candidates in almost every swing district, and they’ve also raised a lot of money, so they should be capable of running vigorous campaigns and perhaps picking up the majority of undecided voters.

To put it another way, Democrats are well-positioned to win the House — but they still have a lot of work to do to turn good prospects into a reality on the ground.

Check out all the polls we’ve been collecting ahead of the 2018 midterms.


  1. Trump’s approval rating among voters (rather than all adults) has also declined, although at 41.6 percent, it’s slightly better than his overall numbers.

  2. You wouldn’t want to conduct polling over the Labor Day holiday anyway.

  3. Our forecast, which adjusts registered-voter polls to a likely-voter basis, figures that likely-voter polls are about as likely to help Democrats as to hurt them as it awaits more evidence.

  4. Long story short, for our generic ballot interactive, we basically adopted the settings from our approval ratings tracker. But approval ratings are a far more stable indicator; as you can see from our interactive, changes in Trump’s approval ratings often hold for weeks or months at a time. Apply those same settings to the generic ballot and you’ll wind up with an overly volatile measure. Generic ballot polls tend to bounce around a lot, and they tend to differ a lot from polling firm to polling firm, so they require more smoothing.

  5. It hasn’t happened since 2008, when Democrats did so, and it hasn’t happened in a midterm year since 1982.

  6. Which is where our House model’s estimate of the popular vote is, vs. the generic ballot’s 11 points.

  7. There’s also a distinction between the generic ballot and the House popular vote, which sort of measure the same thing and sort of don’t, but that’s a discussion for another day. Our House forecast uses a variety of other indicators in addition to the generic ballot in forecasting the House popular vote.

Nate Silver is the founder and editor in chief of FiveThirtyEight.