Skip to main content
Menu
Election Update: Here Are 25 Districts Where Our Model And Other Experts Disagree

Welcome to our Election Update for Thursday, Aug. 23!

As of 1 p.m., the “Classic” version of FiveThirtyEight’s House forecast gives Democrats a 5 in 7 chance (72.4 percent) of flipping the chamber. That’s down a tad from Wednesday morning, when Democrats had a 3 in 4 chance. The average gain for Democrats is now 33 seats, as opposed to 34 — so we really are talking a small difference here.

On Wednesday, we took a broad look at how our forecast stacks up against the three OGs in the field of election prognosticating: the Cook Political Report, Inside Elections and Sabato’s Crystal Ball. On the whole, our views match up very well, with only a few exceptions. Today, we’d like to dive into those exceptions.



The individual districts that FiveThirtyEight and election gurus disagree on actually reveal a lot about our different approaches to election forecasting. Here are the 25 districts where our Classic model’s probabilities of a Democratic win differed the most from the probabilities implied by Cook’s, Inside Elections’s and Crystal Ball’s ratings (we’re showing the experts’ ratings in the table below, rather than the probabilities we inferred from them, because the handicappers don’t really endorse converting their ratings into exact win probabilities)1:

Where our model differs from the handicappers’ ratings

As of 1 p.m. on Aug. 23, 2018

District Democratic Chance of Winning in 538 Classic Model Cook Political Report Inside Elections Sabato’s Crystal Ball
1 TX-32 11% Toss-up Toss-up Lean R
2 WV-03 6 Lean R Lean R Toss-up
3 VA-02 6 Lean R Lean R Toss-up
4 PA-07 96 Lean D Tilt D Lean D
5 PA-17 97 Lean D Lean D Lean D
6 TX-23 70 Lean R Tilt R Toss-up
7 FL-27 96 Lean D Lean D Lean D
8 CA-25 73 Toss-up Tilt R Toss-up
9 CO-03 37 Likely R Solid R Likely R
10 IL-06 24 Toss-up Toss-up Toss-up
11 AZ-02 90 Lean D Tilt D Lean D
12 MN-02 74 Toss-up Toss-up Toss-up
13 NE-02 56 Lean R Lean R Lean R
14 CA-10 70 Toss-up Tilt R Toss-up
15 IA-01 72 Toss-up Toss-up Toss-up
16 MI-07 37 Likely R Likely R Likely R
17 VA-05 47 Lean R Likely R Lean R
18 NY-22 70 Toss-up Toss-up Toss-up
19 KS-03 27 Toss-up Tilt R Toss-up
20 IA-03 65 Toss-up Tilt R Toss-up
21 NY-24 34 Likely R Likely R Likely R
22 FL-07 98 Likely D Lean D Likely D
23 PA-01 22 Lean R Tilt R Toss-up
24 VA-07 28 Toss-up Tilt R Toss-up
25 GA-06 4 Lean R Likely R Likely R

Topping the list is Texas’s 32nd District, where handicappers seem more bullish on the Democrat than our model does. They may see this district as akin to Texas’s 7th District — both are affluent suburban areas unexpectedly won by Hillary Clinton four years after Mitt Romney carried them by solid margins — and therefore rate them similarly. By contrast, our model thinks the two are very different. Specifically, our “fundamentals” variable — things like how much money the candidates have raised, past election results and the incumbent’s ideology — is heavily Republican-leaning for the 32nd District and slightly Democratic-leaning for the 7th. Why the discrepancy? Perhaps the candidates’ fundraising: The 32nd District’s incumbent, Pete Sessions, has raised $2.8 million to his Democratic opponent’s $1.9 million. In the 7th, it’s the Democrat who has outraised Republican Rep. John Culberson.

Second on the list is West Virginia’s 3rd District, where, yet again, the experts give the Democrat, Richard Ojeda, a much better chance than our model does. The 3rd is extremely conservative on the presidential level (Trump won it 73-23), and its (departing) Republican representative won re-election by 44 points in 2016, leading our model to view its “fundamentals” as overwhelmingly Republican. But the handicappers have a good argument too: This coal-country district is ancestrally Democratic and still willing to back a certain brand of West Virginia Democrat. (Democratic Sen. Joe Manchin, for example, carried it by 34 percentage points in his 2012 campaign.) Ojeda represents an even Trumpier (78-19) district in the state Senate and led in the only poll of the race.

A pair of Keystone State districts make the top five on the table; the experts view both Pennsylvania’s 7th and 17th districts as competitive, while our Classic model says they’re all but sewn up for Democrats. In the 7th District, that’s because the only poll of the race, taken before the primary in May, gave the eventual Democratic nominee an 11-point lead — and that was before it was revealed that Republican Marty Nothstein was the subject of a sexual-harassment claim at his workplace, a scandal that significantly discounts his chances in our model. But the handicappers might point out that the poll is nearly four months old and the district has a roughly even partisan lean.

And the 17th District is 2018’s only incumbent-vs.-incumbent general election, as Republican Rep. Keith Rothfus and Democratic Rep. Conor Lamb were thrown together after court-ordered redistricting. Other prognosticators are probably wary of giving a significant edge to one of these two obviously battle-tested candidates, but we’re not: Lamb proved his ability to win on tough turf and raise massive amounts of dough in March’s special election for Pennsylvania’s old 18th District; plus, a Monmouth poll last month gave him a double-digit lead.

I’ll spare you going through all these districts. (We’ll soon have district-specific pages that walk through the model’s “thinking” in each and every race.) That said, we’d love to hear from you on Twitter or in the comments if you have thoughts on whether we or the race raters will wind up correct. For the record, we are fully expecting to win some and lose some. The model’s systematic approach has a lot of advantages when looking at the House as a whole, but it can miss district-specific idiosyncrasies that the handicappers are smart about incorporating.

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

Footnotes

  1. Here’s how we calculated the probabilities: We assigned each race-rating category a probability based on the midpoint of FiveThirtyEight’s ranges for each category (you can see those ranges just above the map here). “Solid Democratic” races have a 97.5 percent chance of going blue; “likely Democratic,” 85 percent; “lean Democratic,” 67.5 percent; “toss-up,” 50 percent; “lean Republican,” 32.5 percent; “likely Republican,” 15 percent; “solid Republican,” 2.5 percent. (Inside Elections’s “tilt Democratic” and “tilt Republican” categories are between a “lean” and a “toss-up” race, so we assigned them probabilities of 60 percent and 40 percent, respectively.) We then took the average probability implied by the three forecasters and compared it with our Classic model’s probability of a Democratic win. The districts below are the 25 with the biggest gaps between our probabilities and those implied by the expert ratings.

Nathaniel Rakich is FiveThirtyEight’s elections analyst.

Comments