A [FiveThirtyEight]( email
[Election Update](
Wednesday, September 30, 2020
By [Nate Silver](
President Trumpâs quest to win a second term is not in good shape. He entered Tuesday nightâs debate with roughly a [7- or 8-point deficit in national polls]( putting him [further behind at this stage of the race]( than any other candidate since Bob Dole in 1996.
If we look at potential tipping-point states, the race is a bit closer, but not that much closer. After a couple of strong polls for Joe Biden earlier this week in Pennsylvania — the state thatâs currently [most likely to decide the election]( — Trump [now trails there by 5 to 6 points](. Heâs down by about 7 points in [Michigan]( and [Wisconsin]( meanwhile. Those states, along with [Minnesota]( [Maine]( and [New Hampshire]( — where Biden has also polled strongly lately — suggest that Biden is winning back some of the Obama-Trump white working-class voters who flocked to Trump four years ago. Indeed, Biden is as close to winning [South Carolina]( or [Alaska]( as Trump is to winning Michigan and Wisconsin, based on recent polls of those states.
At a time when Trump desperately needed a boost, the debate probably didnât help him either — it may have hurt him. Every scientific poll weâve seen had Trump losing the debate, some by [narrow margins]( and some by [wide ones](.
That includes the [poll FiveThirtyEight conducted with Ipsos]( which surveyed the same group of voters before and after the debate. While the poll didnât show a massive swing — most voters stuck to their initial preferences — more voters did rate Bidenâs performance favorably, and Biden gained ground relative to Trump based on the number of voters who said they were certain to vote for him, roughly tantamount to a [3-point swing toward Biden in head-to-head polls](.
Now, Iâm not predicting this will happen, but if Bidenâs national lead were to expand to 9 or 10 points, which is consistent with the sorts of polling bounces weâve seen in the past for candidates who were perceived to win debates â especially [challengers debating an incumbent for the first time]( — Trumpâs situation could become quite desperate.
To be clear, none of this means that Trumpâs chances are kaput. As of this writing, our forecast still gives him around a [21 percent chance of winning the Electoral College](. Thatâs not great, but itâs a lot better than zero.
But itâs possible Trumpâs chances may decline further after post-debate polling begins to roll into our forecast. Furthermore, the mere passage of time helps Biden in our model, because every day that Trump doesnât gain ground is a day when his fate becomes slightly more sealed. ([Lots of people have already voted!]( Case in point: In an election held today — Trump has no more time to make up ground — his chances would be 9 percent, not 21 percent, according to our forecast.
Then again, there are some possibilities that our model doesnât account for, and they have become more pertinent after Trump has [repeatedly refused to commit to a peaceful transfer of power]( and [declined to commit to respecting the election results](. As we [wrote when launching the forecast](
We assume that there are reasonable efforts to allow eligible citizens to vote and to count all legal ballots, and that electors are awarded to the popular-vote winner in each state. The model also does not account for the possibility of extraconstitutional shenanigans by Trump or by anyone else, such as trying to prevent mail ballots from being counted.
---------------------------------------------------------------
Letâs back up for a second. This is FiveThirtyEightâs fourth presidential election campaign. And in the previous three, there was at least some question about who was ahead in the stretch run of the race. John McCain, for instance, [briefly pulled ahead of Barack Obama following the 2008 Republican convention]( and Obama didnât really solidify his lead until early October. In 2012, national polls were [very tight between Obama and Mitt Romney]( following the first presidential debate, and remained fairly tight thereafter (although Obama always maintained an Electoral College edge). And people forget how close the 2016 race was for stretches of the campaign; [it was not such a huge upset](. In fact, Hillary Clinton led by only [1.4 points in our national polling average]( heading into the first debate that year.
But there isnât any of that ambiguity this time. Since we launched our general election polling averages on June 18, Biden has never led by less than 6.6 points nationally. Literally only one national poll — a Rasmussen Reports [poll]( that put Trump ahead by less than a full percentage point — has shown Trump leading by any margin during that period. Itâs been an exceptionally stable race.
But, amazingly, that hasnât really shaken peopleâs confidence in [Trumpâs ability to win](. In our own poll with Ipsos, we found respondents thought Biden and Trump had roughly equally likely chances of winning. And maybe that boils down to three perpetual sources of anxiety I hear in conversation with liberal friends or liberal readers:
- Trump could win the Electoral College despite losing the popular vote by a wide margin.
- There could be a large polling error in Trumpâs favor.
- Trump could somehow steal the election.
All three are legitimate sources of concern for Biden backers. The first two are relatively easy to quantify, however. Indeed, the whole purpose of a model like FiveThirtyEightâs presidential forecast is to answer questions like those. The third one, however, is harder to get a handle on, so letâs talk about No. 1 and 2 first..
The Electoral College could still help Trump, but it only goes so far
The possibility of an Electoral College, popular vote split [remains a point in Trumpâs favor](. In fact, thereâs an 11 percent chance that Trump wins the Electoral College but not the popular vote in our forecast (but less than a 1 percent chance the other way around). At the same time, Bidenâs strength in the Upper Midwest relative to Clintonâs — at least, if polls are correct there — potentially mitigates this disadvantage to some extent. The table below shows Bidenâs probability of winning the Electoral College given various popular vote margins, according to our forecast as of Wednesday afternoon. And as you can see, Biden is only truly safe to win the Electoral College once he has a popular vote margin of 5 points or more! But, heâs a fairly heavy favorite with a 3- to 5-point margin, and has roughly break-even odds with a 2- to 3-point margin.
Bidenâs favored, if he wins the popular vote by +2 to +3 points
Chances of Biden winning the Electoral College under different popular vote scenarios, according to the FiveThirtyEight presidential forecast, as of Sept. 30
POPULAR VOTE MARGIN scenarios Bidenâs chances
of winning the ELECTORAL COLLEGE
Biden +6 to Biden +7 >99%
Biden +5 to Biden +6 98
Biden +4 to Biden +5 93
Biden +3 to Biden +4 77
Biden +2 to Biden +3 54
Biden +1 to Biden +2 29
TIE to Biden +1 11
Trump +1 to TIE 3
Trump +2 to Trump +1 <1
So, for practical purposes, you can take Bidenâs lead in national polls and subtract 2 or 2.5 points from it to infer his margin in tipping-point states. In other words, if heâs ahead by around 7.5 points in national polls, thatâs more like the equivalent of a 5-point lead in the Electoral College. Thatâs still a reasonably large advantage; empirically, itâs [not that easy to overcome a 5-point deficit at this stage of the race](.
A big polling error could help Trump ⦠or Biden
One of the misconceptions I hear about FiveThirtyEightâs forecast is that âit assumes that polls are right.â Actually, in some sense the whole purpose of the forecast is to estimate the chance that the polls are wrong. In 2016, the polls did show Clinton ahead, but between tight margins in tipping-point states and the large number of undecided voters, there was a [fairly high probability]( — around 30 percent, according to our forecast — that Trump was going to win anyway.
So while a polling error is possible — indeed, our forecast assumes thereâs likely additional error this year [because of an uptick in mail voting]( — it would still take a bigger error than in 2016 for Trump to win.
Assume that current polls hold until Election Day, and subtract 3 points from Bidenâs margin in every state (roughly the average error in swing state polls in 2016) … Biden still wins Pennsylvania, Michigan and Wisconsin fairly comfortably, and therefore, the Electoral College; heâd also be a slight favorite in Arizona. And as our friends at the Upshot have calculated, even if you had a polling error [of the exact same magnitude in the exact same states as in 2016]( Biden would still win, albeit narrowly.
Of course, nothing intrinsically rules out a larger polling error. We had one in 1948 — when [Dewey didnât defeat Truman, after all]( — and in 1980, when Ronald Reagan won in an epic landslide instead of the narrow margin that polls predicted.
But thereâs no guarantee such an error would favor Trump. Historically, the direction of polling bias has [not been predictable from cycle to cycle]( the same polls that underestimated Trump in 2016 tended to underestimate Obama and Democrats in 2012, for instance. If anything, to the extent there are polling errors, they [sometimes come in the opposite direction of what the conventional wisdom expects](.
I want to spend more time on this topic in the coming days, so I wonât go on at too much length here. But for now, know that a 7-point Biden lead on Election Day could, indeed, turn into a 2-point Biden popular vote win where Trump narrowly wins the Electoral College.
As I wrote earlier in the piece, our forecast gives Trump about a 9 percent chance of winning an election held today despite his current deficit in polls â not bad when youâre 7 points down! But itâs about equally likely that a 7-point Biden lead could translate into a 12-point Biden win, in which heâd not only carry states like Georgia and Texas, but would also have a shot in South Carolina, Alaska and Montana.
Trumpâs comments on respecting the election outcome are deeply worrisome, but itâs hard to estimate his chances of overturning the result
Hoo, boy. At some point Iâm going to have to write a column about this too, I suppose. As I said at the outset, our forecast assumes that the election is free and fair — at least to the extent that past elections that we used to train the model were free and fair. (Throughout American history, there has always been plenty of voter suppression and [voter disenfranchisement](
But for now, let me advance a few propositions:
- Even a small probability that the U.S. could become a failed or manifestly undemocratic state is worth taking seriously.
- There are a [wide range of things that Trump could attempt to do]( many of which would be quite damaging to the country, but they are not necessarily equally likely to succeed.
- Trumpâs actions are much more likely to actually change the result of the election if the outcome is close, and right now, the most likely scenario is that Biden wins by a not-so-close margin.
Beyond that, itâs hard to estimate the probability that Trump could steal the election to any degree of precision. It requires, at a minimum, some knowledge of the probabilities in a free and fair election plus some knowledge of election law and how many votes could realistically come under dispute plus some theory of the institutional incentives of the Supreme Court and various other courts plus some opinions on how Congress might interpret the Constitution in the event of a disputed election. Maybe a panel of experts could get together and try to put together some reasonable bounds on the probability of various scenarios, but I donât know that any individual could — certainly not me.
After Trumpâs actions over the past few weeks, though, I wonder if thereâs some tradeoff between Trumpâs chances of winning legitimately and his willingness to engage in authoritarian rhetoric and behavior, even if it probably wouldnât succeed at stealing the election. Itâs not like this is coming entirely out of left field; Trump [also said in 2016 that he wouldnât necessarily respect the election results](. But his recent statements have come at a moment of increasing peril for his campaign. Itâs hard to know for sure, but I think Trumpâs comments might be more tempered if he were 2 points ahead in Wisconsin instead of 7 points down.
Itâs not easy to see which cards Trump has left to play or which contingencies could work in his favor enough for him to win — other than if the polls have been wrong all along.
Consider that Trumpâs convention produced, at best, a [very meager bounce in his favor](. His attempt to pivot the campaign to a âlaw and orderâ theme fell [completely flat in polls of the upper Midwest](. Heâs thrown the kitchen sink at Biden and [not really been able to pull down Bidenâs favorables](. His hopes that weâd turn the corner on COVID-19 before the election are diminishing after [cases have begun to rise again in many states](. His campaign, somehow, is [struggling to hold on to enough cash]( to run ads in the places it most needs to run them. The New York Times and other news organizations are likely to [continue publishing damaging stories on his taxes and personal finances from now until the election](. And now heâs seemingly lost the first debate.
If Trump intuits that heâs unlikely to win legitimately — itâs not hard to imagine him escalating his anti-democratic rhetoric and behavior. Itâs also not hard to imagine this rhetoric further eroding his position in polls. Itâs [highly unpopular in focus groups]( (yes, take those with a huge grain of salt) and Trumpâs polling over the past several days has [been particularly bad]( (although thereâs been a lot of other news, too).
So we could be headed for a vicious cycle where Trump increasingly gives up on trying to persuade or turn out voters and voters increasingly give up on him. But from a polling standpoint, this is one of the clearer elections to diagnose: Biden isnât home-free, but heâs in a strong position. Nonetheless, the outlook for whatâs actually in store for America has rarely been more cloudy.
Check out our latest [2020 election forecasts](.
[Latest Forecast](
---------------------------------------------------------------
Share
[Facebook](
[FiveThirtyEight]
[View in browser](
[ABC News](
[Unsubscribe](
Our mailing address: FiveThirtyEight,
47 West 66th Street, New York, NY 10023.