The divide between the House outlook and the Senate outlook continues to widen. Democrats’ chances of winning a majority remain at or near their all-time highs in our House forecast — ranging between 78 percent (7 in 9) and 85 percent (6 in 7) in the various versions of our model. But they’re at their lowest point yet in the Senate. All three versions of our forecast give them only about 1 in 7 shot (about 15 percent) of taking over the Senate from Republicans.
This is normally the point at which you might expect us to give you a throat-clearing “well, actually” about how 1 in 7 chances happen all the time. Indeed, they do. One in seven days of the week is a Thursday. None of us woke up this morning screaming “Oh my gosh, I just can’t believe it’s a Thursday!” And nobody should really be that surprised if Democrats win the Senate next week, or if Republicans keep the House.
At the same time, Republicans have a fairly clear advantage in the Senate (as Democrats do in the House) — clearer than the edge Hillary Clinton had before the 2016 election, when President Trump had roughly a 3 in 10 chance to win the Electoral College. In 2016, a normal-sized polling error (if it worked in Trump’s favor) was probably going to be enough to give him a victory in the Electoral College. And that’s exactly what happened: The polls weren’t great in 2016, but they were about as accurate as they have been on average since 1972. Because the race was close and because Clinton was underperforming in the Electoral College, a small and routine but systematic polling error was enough to give Trump the win.
The difference this year is that a normal-sized polling error in Democrats’ direction would merely make the race for the Senate close. (Likewise, a normal-sized polling error in the GOP direction would make the House close, but Republicans would still have to fight it out on a district-by-district basis.) A sports analogy, for those so inclined: In 2016, Trump was doing the equivalent of driving for the game-winning touchdown with the odds somewhat but not overwhelmingly against him. If enough undecided voters in the Midwest broke toward him, he was going to win the Electoral College. In the Senate this year, by contrast, it’s more like Democrats are driving for the game-tying touchdown; they still have to win in overtime even if they score.
By a systematic polling error, I mean one that occurs in a correlated way across every race, or in certain groups of races — not merely errors that happen on a one-off basis. Our models account for the possibility of several different types of systematic errors, but in this article, I’m going to focus on the simplest type of systematic error, which is a uniform swing that applies to every race. In certain simulations, for example, our model will randomly simulate a 4-percentage-point uniform swing toward Republicans, in which it adds 4 points to the Republican margin in every state and district. From there, it proceeds to consider the other types of error and uncertainty.
What’s clear is that Democrats will very probably need some type of systematic polling error to win the Senate: They’ve fallen too far behind in too many races to have much of a shot at winning just by getting lucky on a case-by-case basis.
Problem No. 1 for Democrats is that they don’t have a clear path to a majority. Even if they were to win all of the “toss-up” Senate races, plus all of the races in which they’re favored, they’d wind up stuck at 50 seats, with Republicans also having 50 seats and Vice President Mike Pence casting the tie-breaking vote to preserve a GOP majority. Instead, Democrats will have to win at least one of the races that our model currently deems as “lean Republican” or “likely Republican” — namely North Dakota, Texas, Tennessee or (less plausibly) the Mississippi special election.
But in certain ways, the no-clear-path problem is overrated. Democrats aren’t favored in any of these races individually, but if you put them together collectively, they have a decent shot at winning at least one. According to the Deluxe version of our model, Democrats have a 23 percent chance to win North Dakota, a 20 percent chance to win Tennessee, an 18 percent chance to win Texas, and a 12 percent chance to win the Mississippi special election.15 Assume for a moment — this is a bad assumption but it’s useful as an illustration — that each Senate race is independent from the next one.16 Under that assumption, there’s a 55 percent chance that Democrats win at least one of those four races.
But here’s Problem No. 2: Even if Democrats manage to pull off a big upset somewhere, they also have to win a lot of other contests in which they’re somewhere between modest favorites and mild underdogs. The most problematic of these races for Democrats are Missouri, Indiana, Nevada, Arizona and Florida. There’s also a smaller but non-trivial chance of a loss in Montana, West Virginia, New Jersey and the Minnesota special election. In fact, if you assume that each Senate race is independent, Democrats’ overall chances of winding up with a majority is only 4 percent or 5 percent, according to our numbers. There are a lot of scenarios in which Beto O’Rourke beats Ted Cruz in Texas but Claire McCaskill loses her race in Missouri, or Dean Heller outlasts Jacky Rosen in Nevada — or Bob Menendez somehow blows it in New Jersey.
It’s the fact that Senate races aren’t independent from one another that gives Democrats a somewhat better chance.
Consider, for example, what would happen if there were a 2.5 percentage point uniform swing in Democrats’ favor. Why 2.5 percentage points? Because according to our model — which is based on how often systemic polling errors have occurred in congressional races since 1990 — 2.5 percentage points represents about one standard deviation’s worth of uniform swing on election night.17 About one-sixth of the time, Democrats will beat their polls by at least one standard deviation (or 2.5 points) in the average competitive Senate race. Another one-sixth of the time, Republicans will beat their polls by at least 2.5 points in the average Senate race. The remaining two-thirds of the time, we’ll wind up somewhere in between. Here’s what the eight most important Senate races would look like with a 2.5-point uniform swing in Democrats’ favor, according to the Deluxe version of our forecast:
Democrats can compete in the Senate … if there’s a uniform polling error in their favor
Forecasted vote margins in eight key Senate races, according to FiveThirtyEight’s “Deluxe” model vs. with a hypothetical uniform swing of 2.5 points toward all Democrats, as of Nov. 1 at 2 a.m.
2.5 point Dem. swing
Democrats must win six of these eight races to win the Senate, assuming less-competitive races go as predicted.
Keep in mind that these projections are not meant to be deterministic. They just recalibrate the numbers in Democrats’ favor based on the hypothetical of a uniform swing in their direction. This is a good scenario for Democrats, but not a best-case scenario: They’d still have to win each race on a case-by-case basis, just under more favorable conditions than we originally expected. In fact, to take the Senate, Democrats would need to win six of the eight races on the list above.18
How easy would that be? With a polling error in their favor, Democrats would still be underdogs in Tennessee, North Dakota and Texas — but lesser underdogs than before, down by about 3 percentage points in each race. They’d be favored in the other five key races, although not by much except in Florida and Indiana. Overall, the Senate would be a lot more competitive than under the status quo, although you’d still rather be in Republicans’ position.
Of course, Democrats aren’t necessarily limited to a 2.5-percentage-point polling error. About 1 election in 40, there will be a two standard-deviation error in their favor, or 5 percentage points’ worth of uniform swing. Even then, Tennessee, North Dakota and Texas would only be toss-ups, but the overall odds would be stacked pretty heavily toward Democrats. There are also various other types of regional or demographic polling errors that could help Democrats. If they outperformed their polls in rural areas, that would be helpful to their chances in the Senate, for example.
If the polls are right, Democrats will lose the Senate. But our model’s job is to consider all these different ways in which polls and other data might be wrong, based on how often the polls have been wrong in the past, and by how much. There’s about a 1 in 7 chance that the polls are wrong enough — and wrong enough in just the right ways — that Democrats win the Senate instead of Republicans.