Princeton Election Consortium

Innovations in democracy since 2004

Outcome: Biden 306 EV (D+1.2% from toss-up), Senate 50 D (D+1.0%)
Nov 3 polls: Biden 342 EV (D+5.3%), Senate 50-55 D (D+3.9%), House control D+4.6%
Moneyball states: President AZ NE-2 NV, Senate MT ME AK, Legislatures KS TX NC

Counting, conflict, and consequences

November 4th, 2020, 10:03am by Sam Wang

Based on traffic stats, the great majority of you arrived in the last few days. We’re living out some likely events I suggested in September. First, mail-in votes are different this year. They are taking a while to count in Wisconsin, Michigan, and Pennsylvania. Second, given the Trump Administration’s attitude to the law, it matters who is in charge of election administration, especially Pennsylvania.

It seems like a full vote count (for example, if mail ballots were not delayed; hi, Louis DeJoy!) would lead to WI/MI/PA for Biden, GA a toss-up, and 291 to 307 electoral votes for Biden in the end (or one less, if Trump wins Maine’s 2nd Congressional District).

Three things are on my mind today: (a) voting rules matter a lot for the Presidency – and for the Senate, in Georgia and Maine; (b) polarization took over in Senate races, and (c) electoral reform took a major hit last night.

Also, there is no avoiding the question of the large polling error. It’s hard to tell until all the votes are counted, but the error seems to be at least as large as the 2016 error of 3 percentage points. That puts the outcome in line with the PEC Trump+3% map, which shows a mode of 309 electoral votes. In Florida and Texas the polling error looks larger, 4-6 percentage points. That is huge.

I don’t know why this would be. One possibility is that Hispanics are assimilating with other whites in self-identification. If they vote more like their white neighbors in both Florida and in New York, then they vote differently in those places, and would need to be sampled accordingly. However, I have to say it is starting to get tedious figuring out which polling error was made in each successive election.

Anyway, you see why I listed all the results in the banner in terms of a Meta-Margin, which among other things defines how large a polling error would have to be to create a perfect toss-up. I thought 5.3% was pretty large, but if the polling error were Florida-sized everywhere then there’d be even more ‘splaining to do.

A huge practical outcome for 2021 is the Senate. Montana, North Carolina, and Iowa went Republican. It is the same pattern as 2016: every state went the same way for the Presidency and the Senate. Polarization wins again. That gets to 48 Democratic seats. It might still come out at 50 seats, the bottom of the range mentioned before. But it depends a lot on unusual voting rules in Georgia and Maine.

Note: Vote-counting is not done yet in North Carolina until November 12th. There are processes such as vote-curing taking place.

In Georgia, if a candidate does not initially get to 50% of the vote, there is a runoff election on January 5, 2021 – the day before the new Congress comes in. The open seat goes to such a runoff between Reverend Raphael Warnock (D) vs. Senator Kelly Loeffler (R-inc.). Warnock leads Loeffler in head-to-head surveys by 4-14 points. (Note that the polling error in Georgia appears to be pretty small this year.) However, turnout may well drop off in the Because of that, and because Loeffler has time to consolidate the Republican vote, Warnock still faces trouble. That’s a key  difference between a top-two primary and a ranked-choice/instant-runoff system.

(But if you’re not registered to vote yet, you can do so online!)

The other Georgia Senate race may also go to a runoff, though there is still a chance that Senator David Perdue (R) will make it over the 50% threshold. We should find out soon. The runoff eliminates a Libertarian candidate, which is bad for Jon Ossoff (D)’s chances.

In Maine, Senator Susan Collins (R-inc.) leads Sara Gideon (D) so far. However, she may fall below the 50% threshold. In that case, they do have ranked-choice voting. (Here at the Electoral Innovation Lab, we worked on a recent court case addressing the constitutionality of ranked-choice voting in Maine.) If independent progressive Lisa Savage gets more vote share than the difference between Collins and Gideon, then Gideon still has a chance to win – if enough of Savage’s supporters pick her as second or third choice.

All of this goes to show how critically important the rules of democracy are. We’re forced to live with the Electoral College, where a convincing popular win can be overwhelmed by the peculiarities of swing states. And the population distortions of the Senate, as well state-by-state rules, have big consequences.

Whatever happens in these races, the picture for electoral reform in the next few years just got a lot worse. National reform has to go through the Senate, which currently needs 60 votes to move legislation. Democrats probably needed 55 votes to lower that threshold, though who knows, maybe they can make a more modest rules change. Such a change would be needed to pass a new Voting Rights Act. In addition, if they can’t even get to 50 votes, then I have no doubt that Biden’s judicial appointments face an extremely difficult road in 2021.

Finally, there is the question of 2022. Democrats seem likely to lose seats then, putting them below 50 seats. So at most, they have two years to get anything done.



I haven’t even gotten to the subject of state legislatures and ballot initiatives. There, my hardworking team here at Princeton was deeply involved. Because of that, and because of its importance, I will write about that story in a separate post. The picture is not all bad…but it’s very much a blue-state, red-state story.

Tags: 2020 Election · President · Redistricting · Senate

55 Comments so far ↓

  • Amitabh Lath

    Public polls were way off (+17 in Wisconsin?). However I suspect the Biden/Harris team had access to higher quality polls. I wondered why they were spending so much time in Midwestern states and then Georgia?

    According to public polls they should have been in NC, FL… (although Biden did go to TX).

    • Sam Wang

      Polls were off, though maybe did you mean 7 points rather than 17 points?

      I am curious about what they knew. I would say that they played a strategy that was both cautious and aggressive at the same time, and the cautious part turned out to be a good idea.

    • Pechmerle

      Yes, both cautious and aggressive. they didn’t make Hilary’s mistake of taking WI, MI, and PA for granted. Of course the Biden campaign had that cautionary clearly in view as they made resource allocation decisions.

  • Rick

    I think he’s talking about that outlier ABC/WaPost poll that had Biden +17 in late Oct.

  • A. Netliner

    Three possible explanations for the election results:

    1. Fundamental flaws in U.S. polling models, with the exception of Rasmussen and Trafalgar. Needs to be studied and more detail embodied in state-level polls. An additional key question: how is the presidential popular vote stacking up versus Biden’s lead in the polls? The 2016 polls called the popular vote. What about 2020?

    2. Voter suppression. Needs to be examined, even though turnout was exceptionally high. Poll responses are not suppressed. Votes might be.

    3. Hacking. A continuing deviation between polling data and vote totals could suggest deliberate manipulation of the results. We hate to even raise this possibility in the United States, but it warrants a hard look. What do election security experts think?

    • Sam Wang

      In regard to #2, I am really interested in the subject of mail delays. John Kruzel of The Hill is reporting a large number of undelivered mail ballots, including 27% of ballots from south Florida, where Biden is said to have massively underperformed.

    • ArcticStones

      Sam, I have been reading John Kruzel’s Twitter feed, but I searched in vain for an article about this on The Hill.

      Kruzel’s story about the undelivered mail-in ballots ought to be their banner headline!

    • ArcticStones

      PS. I saw your follow-up Tweet with its caveat, and read the Vice article to which you linked. I sincerely hope the explanation is as described therein.

    • Herb Stoller

      “Poll responses are not suppressed.” I think you are wrong here. And this is at the heart of the polling errors. As has been pointed out elsewhere, the Trump supporters are self-suppressed when it comes to responding to polls but not when it comes to voting.

  • ArcticStones

    Three thoughts:

    1) When all is said and done, one of the biggest stories will be how Trump lackey Postmaster General Louis DeJoy prevented timely deliver of a huge number of mail-in ballots.

    I would not be surprised if a careful analysis reveals that these essentially-stolen ballots – if delivered in a timely manner – would have changed the outcome in several states that went for Trump (I am looking at you, Florida), as well as widened Biden’s margin of victory other states.

    2) Perhaps an election result with a fair count of every vote – including every mail-in ballot sent – would have been seen as an unequivocal rejection of Trumpism as well as Trump.

    Perhaps these “stolen” ballots are also allowing Mitch McConnell to keep his Senate majority?

    3) A lot of observers are touting the impressive turnout. I beg to differ. The national turnout appears to have been approx. 66.9 percent. This means that one-third of Americans who have the right to vote didn’t even bother casting a vote, never bothered to make their voice heard.

    That’s tragic.

    • Nff

      Hi Arctic Stones,

      I’m not sure it’s so much that a lot of Americans can’t bother to vote as that in a lot of places it’s freaking difficult to even register!

      Trust me, I speak from experience. Somedays we’ll have a virtual beer & I’ll tell you how 3 different clerks in a PA election office tried to tell me a US passport wasn’t valid ID; that I needed a PA divers license; and that license had to be more than year old with no tickets or points.

      Which was all BS. I’m a rich white person, BTW, whose address is in the hoity-toity part of town. I have no idea how someone without my white & class privilege could overcome that.


    • Herb Stoller

      A couple of observations:
      1) it is a federal crime to tamper with the mail as you suggested was done. It is inconceivable that one individual could accomplish this much misdirection by themselves, it would take a small group. To think a small group would be able maintain the conspiracy at the risk of prison terms is beyond belief.
      2) y personal experience with the FLA Board of Elections is very positive. When we called asking for a mail-in ballot they followed up immediately and we had the ballot within days. They notified my spouse that her application was rejected and it turned out someone in the office had misspelled our last name (a common occurrence) They apologized and send out the new ballot immediately. It was also possible for us to track their mailing of the ballot online, it’s reception and logging in!
      Your conspiracy theory about the Post Office eating the ‘missing’ ballots has about as much basis as Trump claiming victory.

    • ArcticStones

      Hi Herb, just a point of clarification. I am in no way suggesting a USPS “conspiracy”.

      What I am saying is that the changes introduced by Postmaster General DeJoy achieved their goal: slowing down the post. (The measures have been detailed and the slow-down has been documented in numerous articles, so I won’t do so here.)

      No USPS conspiracy is required. But DeJoy’s orders severely damaged the efficiency of the US Postal Service. That is beyond dispute.

      DeJoy’s changes caused mail-in ballots to not be delivered before the various state deadlines – effectively “stealing” those votes.

      We do not, however, know how many ballots were slowed down, where, and the election impact of those ballots not being counted.

  • Dave kliman

    Is there any way to statistically analyze the data to discover if the sabotaged USPS “lost” a large number of ballots?

    It just seems too odd that every single poll would be so far off in the same exact direction, but Occam’s razor seems to imply that a bit of skullduggery—evidence of that we have already seen, was responsible for the discrepancy.

    It seems if that’s so, it wasn’t enough to turn the presidential election, but definitely many Dow ballot races and initiatives surely have been affected.

    • Herb Stoller

      Why in the world would you think that civil service postal employees would want to sabotage the Biden campaign? I am sure some are pro-Biden and some pro-Trump, it that environment (and I worked in the Post Office) it would be impossible to get a group together to sabotage anything. Much more likely simple human error if any ballots did not get delivered properly. You need to sharpen your Occam’s razor.

  • NCDem

    Off topic, but where is there a link that is tabulating votes for the presidential contest in Maine’s 2nd CD. (Not the congressional contest, but the presidential one). I can see that it is not “called,” but I can see no place where actual vote numbers for president are tabulated.

  • 538_Refugee

    Are we anymore confident of the self reported post vote breakdown of the demographics than we are of the pre vote numbers? Was Trump premature in ejecting the guy that ‘squandered’ large amounts for online influence? These unaccounted for Trump voters are accounted for somewhere.

  • Tracy Lightcap

    If I had to come up with a quick analysis of these results I’d fall back on the old polsci adage: it is hard to beat an incumbent president. And, sure enough, it is, although it looks like it will happen this time.
    Given that, it might be a good idea to include a “fudge factor” for incumbency in the models next presidential year. Abramowitz does this in his “time for change” model. There’s no need to get too sophisticated about that; Silver worked something like it into the 538 models this time. (Though those did not work as well as he had, no doubt, hoped.)
    Well, just a suggestion. Now we wait.

    • Sam Wang

      That’s a good suggestion. Trump’s approval was at 44.5%, below the 46-47% threshold I cited a few months ago. But not that far below.

      Similarly, the House generic poll was only 1.5 points above the 3-point threshold I mention in that article.

    • Herb Stoller

      Trump was not only the incumbent, we have never seen a President so subvert so many facets of government to his reelection as Trump has, from the AG’s Office, to the Education Dept, to the DNI, to the State department, to the use of the White Houst itself, etc. etc.

    • Stephen Li

      @Pechmerle, thank you for your reply.
      With your reply, I will risk over-staying my welcome by pointing out another example of progressives going for all or nothing. My pet peeve on Bernie Sanders was that he was insisting for Medicare for All. Why not look for every opportunity to expand it? California is passing Proposition 22 to allow gig contractors working for Uber and Lyft not be classified as employees. Bernie should now work on a legislation to allow gig contractors to register for Medicare. Medicare already have worked out an actuarial table of premiums by age, it can easily be expanded from 65 back to 18. Gig contractors under 65 signing up will have to pay the monthly premiums of course. Many union negotiations often get mired in the issue of health benefits as a final sticking point. Bernie can also work to expand Medicare to cover Postal Union employees for example. Then Automobile Union members. Then Teachers Union members.

  • Stephen Li

    I think it’s wrong to focus on the undelivered ballots when the Democrats in charge in WI MI PA NC GA did not start with counting mail-in ballots yesterday. Just absolute incompetence!

    • Sam Wang

      No, that was required by law. They’ve been trying for months to get the law changed. See the post I linked to.

    • Franklin

      FYI, Democratic governors in Wisconsin, Michigan, and Pennsylvania sought legislation which would authorize advance counting of mail-in ballots. It was Republican controlled state legislatures in each state which refused to pass such legislation.

      North Carolina did undertake advance counting of early ballots. And as for Georgia, Democrats are not “in charge” of any branch of the state government.

      So much for blaming the Democrats.

    • Stephen Li

      I do not understand why most progressives are automatically “all or nothing”. Off-topic: like Bernie insisting on Medicare for All, instead of looking for opportunities to expand it: like moving Postal Union members into it, and letting Uber/Lyft drivers join. Later Teachers Unions.
      Back to the topic, don’t understand why I say “count mail-in ballots first” you guys automatically talk about advanced counting. The guy running the show could have said all the mail-in ballot boxes are all ready, so they would be opened first so counting could start right on time at 8pm without waiting for ballot boxes from voting locations.

    • Stephen Li

      I wonder why most progressives automatically adopt the “all or nothing” attitude? When I mentioned counting the mail-in ballots first the responses were automatically but most states don’t allow advanced counting. Counting first doesn’t imply advanced counting. If the law provides for counting to stop at 8pm on Election Day, then the Registrar could have all the received mail-in or drop-off ballots ready in counting centers so that they get counted first before paper ballots collected on Election Day. Or if we don’t want the vote percentages to swing too much, then the Registrar can assign 2/3 of the staff to count mail-in ballots, and 1/3 of the staff to count Election Day ballots.

    • Pechmerle

      @Stephen Li,
      I think your idea makes an enormous amount of sense. I agree with you that election officials would easily have the discretion to count in this order, and it would have taken away the pernicious “red mirage” created by counting election day votes first.

  • KZ

    On polling, I can’t shake the thought that Republicans are systematically lying to pollsters, maybe just democratic leaning pollsters. This seems inconceivable, but the memory of the bloc voting for Hugo award — “sad puppies” — keeps coming back.

  • Anthony Lamantia

    Hi Sam! I hope you are well—all of the current uncertainty must be stressful in every direction.

    What I wonder about is whether the polling errors reflect the unique status of Donald Trump as a polarizing figure in US politics. It’s very possible that one just can’t get a completely accurate measurement of opinion/voter probabilities because of the specific nature of Mr. Trump and the reactions of his supporters and detractors. If pollsters are seen in the context of that polarization, then the outcome would be exactly what we’ve seen in 2016 and 2020. Was there as much error in 2018 when Mr. Trump was not on the ballot?

  • Dave Kliman

    Let me bring this article to peoples’ attention, if you all haven’t seen it yet. This shows the top of the iceberg. It makes perfect sense.

  • Andy

    Why on earth would Sara Gideon concede Maine when holding Collins to under 50% trigger the ranked choice vote? I guess I’m missing something, but I can say definitively that no Republican would concede with 75% of the vote counted.

  • Amitabh Lath

    So if not for Covid-19 and 250k dead, we are looking at a 2nd term for Trump?

    • Sam Wang

      I actually don’t think the pandemic made a difference. His approval rating has been so flat that it seemed immovable. The pandemic is not a detectable event in that long plateau.

    • Amitabh Lath

      So are polls basically useless? (WTF Wisconsin?)

      Last time it was “oh wait, we didn’t separate non-college, we will do that and all will be good”. This time it will be some other sub-tranche that messed up.

      It feels like epicycles upon epicycles trying to prop up Aristotelian cosmology. We need a Kepler.

  • Jeffrey Biever

    If Biden ends up with 270 EV, must we worry that a rogue elector casts his vote with Trump?

    • Dan Martin

      Not likely — electors are generally pulled from among the party faithful…

  • Jon P

    Finally, there is the question of 2022. Democrats seem likely to lose seats then, putting them below 50 seats. So at most, they have two years to get anything done.

    How do you figure that? The 2022 elections are for those who won in the red wave of 2016. Of the 34 seats up for grabs, only 13* are held by Dems. California, Connecticut, Hawaii, Illinois, Maryland, New York, Oregon, Vermont, Washington all seem quite solid. That leaves Kelly (AZ), Bennet (CO), Masto (NV), Hassan (NH) as potentially vulnerable, and I don’t think any of those would be underdogs.
    On the other side, Burr (NC) and Toomey (PA) have already said they’re not running again, and Murkowski (AK), Blunt (MO), and Johnson (WI) seem potentially vulnerable, to say nothing of Rubio.

    * – plus Warnock if he wins in January.

  • ArcticStones

    A rather worrisome potential development:

    The Justice Department told federal prosecutors today that the law allowed the government to send armed officers to ballot counting locations, the New York Times reports.

    The information was relayed via email, according to three sources who spoke with the Times:

    A law prohibits the stationing of armed federal officers at polls on Election Day. But a top official told prosecutors that the department interpreted the statute to mean that they could send armed federal officers to polling stations and locations where ballots were being counted anytime after that.

    The statute “does not prevent armed federal law enforcement persons from responding to, investigate, or prevent federal crimes at closed polling places or at other locations where votes are being counted,” the official, Richard P. Donoghue, told prosecutors in an email that he sent around 1:30 a.m. on Wednesday.

    • Bob

      What are you worried about happening at these vote counting centers? Are they going to force the vote counters to stop? For what reason? They would actually need a reason to forcibly stop the votes from being counted. I don’t subscribe to these fears of pro Trump government law enforcement officials using the power of their job to help Donald Trump steal the election.

  • Dave Kliman

    I wonder if 24 million requested absentee ballots not being returned is a normal thing.

  • Joseph Poirier

    As others have posted above, I believe there is a fairly significant numbers of Republican leaning voters who are lying to pollsters purposely, as a sort of anti-elite massive prank on the polling industry. They are not “shy Trump voters” in the sense that they don’t want to announce their support of Trump to other live pollsters, they are “prank Trump voters” that want to see the mainstream media (of which the polling industry and polling aggregators are a part) trolled and humiliated.

    If you follow Trump-leaning pundit and cartoonist Scott Adams, he have even encouraged his 500,000 followers and 50,000 daily Periscope viewers to do exactly that – lie to pollsters to prank the industry.

    If only 5-10% of Republican voters choose to prank pollsters or are indeed true shy Trump voters, that is enough to swing things significantly.

    How do pollsters guard against determined liars?

  • ArcticStones

    Herd immunity has finally been achieved!
    (headline at Vox com)

    Donald Trump is trying to
    steal the 2020 election.
    America is ignoring him.

  • Christopher Madden

    I follow and respect Sam Wang very much but I’m am less than ignorant regarding all this statistical crunching. That said, it seems like everyone is doing contortions to avoid the idea that the generally non-audited electronic vote was hacked. It feels like our Democracy has had its throat cut (almost fatally) by Occam’s Razor and everyone’s trying to find the guy with the gun.

  • Christopher Madden

    I follow and respect Sam Wang very much but I am less than ignorant regarding all this statistical crunching. That said, it seems like everyone is doing contortions to avoid the idea that the generally non-audited electronic vote was hacked. It feels like our Democracy has had its throat cut (almost fatally) by Occam’s Razor and everyone’s trying to find the guy with the gun.

  • LK

    I’m sure I’m not the only one who ignored emails from Senate campaigns that stated close polling and instead came here to see which races were (supposedly) really close/competitive and then made lots of donations via ActBlue. And then ultimately these Democrats lost by a lot more than was suggested by the polls. Again. So ‘poll aggregation’ is only as good as the polls, and it seems the polls were way off. So my conclusion is that this is all pointless…

    • Pechmerle

      You make an excellent point, that “moneyball” resource allocation doesn’t work if the polling data on where the best allocation is is inaccurate. The polling industry needs to figure this out; I don’t know how they do it.
      As Sam has noted, polling was pretty accurate in 2018 (the midterms). It’s only when Trump is on the ballot that things have seemed to not work right. At least he can’t be on the ballot in 2022.

    • 538_Refugee

      At the risk of sounding ghoulish, if Trump has violated laws for which there are still legal charges possible, they need to be pursued. The people that ended up being most negatively vocal about Nixon were his defenders that got burned. The mask needs to be ripped off and and a stake driven through the spot his heart was supposed to have been in or this zombie will haunt us for years to come. Even now the GOP is cowering in terror of him as he pretty much teeters on the brink of treason.

      This simply must be put to rest in decisive terms.

  • Fred

    I think a huge number of unreturned mail ballots is pretty plausible. Mail voting emerged early on as a sensible response to the pandemic, and lots of people latched on. Then, Trump & co started telegraphing an apparent intention to hinder mail voting. So, it’s easy to imagine someone requesting a mail ballot, but then changing their mind and voting in person. I was considering it, although I did end up voting by mail.

    Then there’s Nevada, where for the first time every registered voter got sent a ballot. Of course, they’re not going to be an appreciable fraction of 24mil, but you get the idea.

  • Charles Greenberg

    Let’s talk about the GA runoffs. Looks like we’re getting two of them. People can still register up through Dec 7 to be able to vote in the runoffs. What are the best organizations to donate to to boost D registration and voting?

    • Pechmerle

      Charles, an absolutely critical point. These two seats in Georgia are our last, best chances at a Dem. majority in the Senate (counting V.P. Kamala Harris). Everything possible that can be done must be done to affect these results. I shudder to think how much money is going to be spent in these two runoffs.
      I think the super-critical thing (which I can’t do anything about) is door-knocking in Dem. – leaning precincts in GA.

  • ArcticStones

    Pennsylvania: As we know, Biden took the lead this morning, and he has since expanded it to 16,300 votes. That is 0.25 %. And there are many votes yet to be counted.

    Anyone care to shed light on whether Biden can further increase that lead to a full 1 %, thus eliminating the need for a recount?

    Nevada: Do we know why the counting slowed to a crawl, especially in Clark County? Seems it’s taking forever to count that last 14–16 % of the votes. (I have been following Jon Ralston of the Nevada Independent, but have yet to see a good explanation. In fact, he himself seems rather frustrated with the pace.)

    Senate: With regards to the Senate, it seems to me we should wait until the full results from Alaska are in.

  • Sam Wang

    Martin wrote: Polls have become a tool to AFFECT elections. There is no such thing as “the polls”. It is overwhelmingly the pro Democrat polls that were biased.

Leave a Comment