In 2016, one of Hillary Clinton’s mistakes was not campaigning in the wrong states so much as not campaigning in enough states. Of course, it would have been nice if she’d known ahead of time exactly where President Trump was likely to do well — that he really could win in Michigan and Wisconsin, for instance, but not in Virginia and Colorado. But that’s not how polling and election forecasting works in the real world. Polls and forecasts are pretty good instruments, but they’re not all that precise. Nor do they necessarily go wrong in the places you expect them to go wrong. Therefore, campaigns should generally be active in a reasonably broad set of plausibly competitive states and districts. They shouldn’t get too cute by either triaging a race too soon or taking one for granted.
So as you see reports about Republicans or Democrats giving up on campaigning in certain races for the House, you should ask yourself whether they’re about to replicate Clinton’s mistake. The chance the decisive race in the House will come somewhere you’re not expecting is higher than you might think.
Let’s be more specific about that. Below is a table showing House races divided into the seven broad categories we use in our forecast, ranging from “solid D” (at least a 95 percent chance of the Democrat winning) to “solid R” (at least a 95 percent chance of the Republican winning). For each category, I’ve listed the chances that a district within that category will be the tipping-point district, according to FiveThirtyEight’s tipping-point index. The tipping-point index measures the likelihood that a state or congressional district will prove decisive to the overall outcome — in the case of the House, for instance, that it will be the 218th most Democratic or Republican district once all votes are counted on Nov. 6, providing the winning party with at least a 218-217 majority.
The House may be decided in an unexpected place
Projected wins and tipping-point chance by district category, according to the Classic version of FiveThirtyEight’s model as of 5 p.m. on Oct. 15
| Tipping-point chance | Projected wins | ||||
|---|---|---|---|---|---|
| Category | No. of races | Average race in this category | All races in this category, combined | Dem. | Rep. |
| Solid D | 190 | 0.02% | 3.8% | 189.6 | 0.4 |
| Likely D | 16 | 1.1 | 16.9 | 13.7 | 2.3 |
| Lean D | 8 | 1.8 | 14.0 | 5.4 | 2.6 |
| Toss-up | 19 | 1.4 | 27.3 | 9.2 | 9.8 |
| Lean R | 19 | 1.0 | 18.5 | 6.1 | 12.9 |
| Likely R | 48 | 0.4 | 17.5 | 6.7 | 41.3 |
| Solid R | 135 | 0.015 | 2.0 | 0.9 | 134.1 |
In considering the battle for the House, it would be easy to concentrate only on the 19 “toss-up” districts (as of Monday evening)1 — plus the eight that are “lean D.” If Democrats win all those districts, plus all the “likely D” and “solid D” districts, they’ll wind up with 233 seats — a net gain of 38 seats from Republicans and comfortably enough to take the House.
It’s extremely unlikely that everything will work out quite so precisely, however. Individual House races are fairly hard to predict. They don’t get all that much polling. Sometimes the candidates make a big difference, and sometimes they don’t. There will be lots of idiosyncratic and even “surprising” demographic patterns that emerge on Nov. 6, but the whole reason they’ll be surprises is because we won’t know about them ahead of time.
What this means is that Democrats will almost invariably lose some of the toss-ups, perhaps along with a couple of the “lean D” races, even if they’re having a pretty good night2 on Nov. 6. Unless we’re really, really lucky, the races just aren’t going to wind up in the exact order our model lists them or in the exact order that the Democratic Congressional Campaign Committee and National Republican Congressional Committee expect.
So it greatly helps Democrats that they also have a long tail of 19 “lean R” seats and 48 “likely R” seats3 where they also have opportunities to make gains. (Conversely, there aren’t that many “lean D” or “likely D” seats that Democrats need to defend.) These races are long shots individually for Democrats — a “likely R” designation means that the Democratic candidate has only between a 5 percent and 25 percent chance of winning in that district, for instance. But they’re not so unlikely collectively: In fact, it’s all but inevitable that a few of those lottery tickets will come through. On an average election night, according to our simulations, Democrats will win about six of the 19 “lean R” seats, about seven of the 48 “likely R” seats — and, for good measure, about one of the 135 “solid R” seats.4 (That is, it’s likely that there will be at least one total and complete surprise on election night — a race that was on nobody’s radar, including ours.)
When I’ve pointed out the importance of the “lean R” and “likely R” races to other election geeks, the general response has been to agree that these seats might be vulnerable, but to suggest that they’ll only fall to Democrats in the event of a landslide, upping the Democrats’ ceiling but not necessarily helping their median projection.
I think this is probably wrong. I don’t want to say “just trust us,” but we spend a lot of time evaluating uncertainty in the forecast, including the crucial issue of the extent to which races are correlated with one another. Indeed, House races are correlated with one another, but not to the same extent that presidential ones are, where the same two candidates are on the ballot in every state. And this year in particular, competitive House races span a pretty wide array of districts that vary in their partisanship, their demographics and whether there’s an incumbent running or not. They also vary in the extent to which they have reliable polling. Thus, the model thinks there’s a 36 percent chance that the tipping-point district — the pivotal race in a close election — winds up coming from the “lean R” or “likely R” pile. Democrats won’t “need” any of those seats if they win all the toss-ups, but it’s highly unlikely that they’ll win all the toss-ups.
To be fair, some of this is just because there are a lot of “lean R” and “likely R” districts. If you took just one “toss-up” district and just one “likely R” district, the “toss-up” would be about four times as likely to produce the tipping-point district. However, there are about three times as many “likely R” districts as “toss-up” districts, which evens things out a lot.
But that long list of “likely R” races is a tangible risk for Republicans. Democrats have raised a ton of money almost everywhere (contra the media narrative, not just in high-profile elections such as the Texas Senate race) and nominated credible candidates all over the map. Where there are districts with demographics that have shifted more than people expect, or complacent incumbents who don’t do a good job of turning their voters out, or last-minute mini-scandals, Democrats are poised to take advantage of those opportunities. That’s a big part of why they’re 6-in-7 favorites in our House forecast.