As I wrote earlier this week, Democrats almost certainly need a systematic polling error to win the Senate. By that I mean: They need for the polls to be off everywhere, or at least in certain key clusters of states, to win the Senate. A polling error in just one or two races (say, Beto O’Rourke wins in Texas) probably wouldn’t be enough: Democrats are defending too much territory and have too many problems elsewhere on the map just to get lucky.
That conclusion about the Senate ought to be fairly intuitive, I think. Even if you credit Democrats with wins in all the toss-up races, that wouldn’t be enough — it would only get them to 50 seats. What might be more surprising is that the same conclusion holds for Republicans in the House. They need for there to be a systematic polling error too. If the polls are about right overall but Republicans are hoping to getting lucky in the swing districts, it probably won’t happen — the odds are stacked heavily against them.
The reason it’s counterintuitive is because you can’t really identify 23 districts that are safe bets to flip to Democrats (that’s the number they need to take the House). In the Deluxe version of our model (the one I’ll be focusing on here), only 193 seats are considered to be “solid Democratic” (at least a 95 percent chance of a Democratic victory). If Democrats won only those seats and no others, they’d actually lose two seats from the 195 they control now. Another 15 seats are “likely Democratic,” where Democrats have at least a 75 chance of winning. Win those as well, and Democrats are … still up to a net gain of only 13 seats.
The model then has 34 seats in its three most competitive categories: “lean Democratic” (eight seats), “toss-up” (16 seats) and “lean Republican” (10 seats). If Republicans win 24 of those 34 seats — assuming everything else goes to form — they’ll keep the House.
How hard is that? Because of the possibility of a systematic polling error, it isn’t really that hard at all. If there’s a typical polling error of 2 to 3 percentage points and it works in Republicans’ favor, the House would be a toss-up. We might not even know the winner for several days as everyone waits for additional mail ballots to be returned from California. Thus, the Lite forecast gives Republicans a 2 in 9, or 22 percent, chance of keeping the House based on the possibility of a systematic polling error. Their chances are 18 percent in Deluxe and 15 percent in our Classic version, meanwhile.
That isn’t a great position, but those are real, tangible chances.
Without the possibility of a systematic polling error, however, the GOP’s position is nearly hopeless. If all races were independent from one another and GOP chances in the most competitive races were approximately 50-50, the Republicans would have to do the equivalent of have a coin come up heads at least 24 times in 34 attempts. That’s really hard. The probability of ending up with at least 24 heads in 34 tries is only about 1 in 80.
We can also come up with a more sophisticated version of this coin-flip calculation. Instead of putting the races into different buckets and treating them as 50-50, we can just use the probabilities listed by the Deluxe model in each of the 435 districts — that Republicans have an 83 percent chance of winning New York 24, for example. We can run a Monte Carlo simulation to see how often Republicans wind up with at least 218 seats. The answer is that, assuming that races are independent — again, a bad assumption, but what you get in a world in which there isn’t systematic polling error — Republican chances of holding the House are only about 1 in 1,000.
What if we use the Lite version of our calculations instead? Lite is essentially a “polls-only” forecast; it uses district-level polls in districts where it has enough of them and national (generic congressional ballot) polls and polls of similar districts to make inferences in districts where it doesn’t. If you run a Monte Carlo simulation with our Lite forecast — assuming that each district is independent — Republican chances aren’t much better, about 1 in 700.
But why is it so hard for Republicans to win the House without a systematic polling miss? The short answer is because they’re defending too much territory: The House playing field is exceptionally broad this year, because of Republican retirements, an influx of Democratic cash and other factors. The decisive race won’t necessarily be in a toss-up district; it could very easily be in a “likely Republican” district where a GOP incumbent is caught sleeping at the wheel (perhaps a district where there hasn’t been much polling).
To help illustrate the breadth of the playing field, I’ve sorted every House district based on the projected margin of victory or defeat for the Democratic candidate, according to the Deluxe model, and ordered them from No. 1 (the easiest seat for Democrats to win) to No. 435. Except, I’m only going to show you the districts in the middle of the spectrum; everything from district No. 190 (if Democrats won only 190 districts, they’d lose a net of five House seats) to No. 295 (a net gain of 100 seats).
The 105 seats that will decide the House
Net change to the current distribution of the House if Democrats win that district and every other district with a greater chance of a Democratic victory
If you know these districts pretty well, you can keep scrolling down the list until you tap out and say, “No, that’s too much for me.” For me, you get pretty far down the list before that happens. The 240th district, for example, which would correspond to a Democratic gain of 45 seats, is Georgia 6 — that’s the district where Democrat Jon Ossoff lost in a special election last year but where Lucy McBath has a decent shot because of the potential for higher black turnout. Chris Collins’s district is at No. 248, and Duncan Hunter’s is at 249. Alaska’s at-large district, where Democratic candidate Alyse Galvin actually had a tiny lead in the latest poll, is at No. 256. Steve King’s district is at No. 262. It’s about at that point that we get into upsets that I’d consider really far-fetched, but even then, there are a few interesting races. West Virginia 3, where Democrat Richard Ojeda has sometimes been competitive in polls despite it having voted overwhelmingly for Trump, is No. 280.
Democrats are underdogs in most of these districts individually, but they’re overwhelming favorites to win some of these districts or others like them on the list — unless the polls were wrong all along and were exaggerating the potential for a “blue wave.”
All seat characterizations are as of 10 a.m. Saturday.
All seat characterizations are as of 10 a.m. Saturday.
The figures in the table are as of 11:30 p.m. Friday.