Final Day of the Campaign: What the Latest is Telling Us

November 4, 2024

There’s definitely a lot of enthusiasm and momentum, but we’re still the underdogs here … and feel like we’re going to keep doing every ounce of work here to make sure we’re getting our vote out and we’re persuading anyone that is open to hearing from us.

Harris campaign official, quoted by Axios

This past weekend - the last of the 2024 presidential campaign - over 90,000 volunteers knocked on 3 million doors across the country for Harris. From what I heard from organizers in Philadelphia is that fully a third of those doors were in Pennsylvania. The Harris campaign is closing strong, and the Trump campaign seems to be spiraling into chaos - or at least Trump himself is. This suggests what the campaigns and candidates themselves think is happening: Harris is winning.

There is no way to know for sure yet, of course. Votes will not be counted until Tuesday nights and in some places it may take days to finish counting. But there are things that suggest to us how an election might turn out. We have discussed some of those things already in a couple of recent posts. Since then, there are more signs that things are looking favorable for the Democratic ticket.

Politico reports that one MRP model (Focaldata) shows Harris winning. MRP (which means “multilevel regression and post-stratification”) might be the future of election polling in the United States. It already is in Europe, where it has accurately projected the results in UK and French elections. If you want to understand it better, this is a good place to start. In a nutshell, MRP uses statistical techniques on very large datasets to disaggregate characteristics of voters, analyze how they impact voting behavior, and then make projections on electoral outcomes in smaller jurisdictions.¹ Then, it incorporates polling data into the analysis. In the graphic below, the combination of MRP and polling create the projections of Harris wins in states such as Michigan and Nevada. Focaldata sees Trump winning close ones in Georgia and North Carolina.

Source: Politico

I think the polling industry is going to have a very difficult time after Tuesday. It’s not because the quality pollster are not good at what they do, although I think we will see they have missed projecting the electorate in their modeling. It’s that there has been an amazing amount of shit polls that have been produced to make it look like Trump is either competitive or winning. And the reaction of the polling aggregators and pundits has been to “throw it in the average,” as if this will just water down poor or corrupt results all on its own. As we can see, the reaction of the shit pollsters is to produce even more. AtlasIntel gave us poll after poll showing Trump winning everywhere over the past week. At the very least, there was no way they were doing this much polling and so quickly. It’s obviously bullshit. Throwing it in the average just makes the average underestimate Harris’ support. But if we look at individual polls, we see a lot of good news for Harris. At least a lot more than the media narrative suggests.

One of the more underreported things about this election is how close Iowa has been in polling. Harris has been within single digits of Trump in the past few months. However, Ann Selzer – considering to be one of the best pollsters in the country, and has an amazingly successful track record in Iowa – has a new poll out this weekend for the Des Moines Register showing Harris winning Iowa 47-44. In September, Selzer had Trump up +4. Trump beat Biden by eight points in 2020 and beat Clinton by almost ten points in 2016. Before that, Iowa was a fairly reliable Democratic state, although Bush won by less than a point in 2004 (and lost by a similar margin in 2000).

What Selzer does differently from a lot of pollsters is she gets a cross-section of all Iowa residents, then she asks if they have voted or tell her they are definitely voting. That is her likely voter screen. She does not weight the data according to what she thinks the electorate will look like, but from what the data reveals to her is probably what the electorate will look like based on self-reporting from respondents. She has had amazing success with this simple technique. She calls things like recalled-vote weighting “polling backwards” and her method “polling forwards” because she doesn’t care who voted or not before only who says they will definitely vote this year. (She does not include probably-will-votes as likely voters.) Here’s a fascinating conversation with Selzer about the poll and her methods:

What is really interesting about Selzer’s poll is that her methodology is much better at seeing shifts in the electorate from one cycle to another than almost all other pollsters use. If the electorate is going to heave bigger shares of women, young folks, and Black voters this year, recalled-vote weighting will not see it. But Selzer’s method should. Now, there are very few Black voters in Iowa so this poll won’t help us much on that. However, Selzer notes that women – particularly senior women – are driving the electorate towards one more favorable to Harris. That is why she is winning there in this poll. Not exclusively, but in large part. If this shift is a national one, then Republicans – particularly Trump – are in for a very bad night tomorrow.

The Harris campaign claims that its internal polling shows late deciding voters breaking to the vice president in the past week. The cause of this appears to be related to Trump’s Nazi Rally at Madison Square Garden. About 75 million voters have cast their ballots already. This might be 50% of the total vote. But unlike similar returns in 2020, more Republicans have voted early and more Democrats will vote on Election Day. There has been a big turnout early among women and Black voters nationally. Women make up 54% of the national share of the early vote, and Black voters are over 22%. The party demographics do not favor Democrats as strongly as one might think from these numbers. Some think that an early surge of Republican voters were Harris votes because those voters were mostly women and from areas where Haley performed well in the primaries. There is also some indication that senior men are breaking to Harris as well as they are repulsed by Trump’s fascist rhetoric. That may be wishful thinking, but we do know that seniors were a surprisingly strong voting bloc for Biden.

There is something else going on that is entirely qualitative and could be wrong: the projection of confidence from the behavior of the candidates and the campaigns. In the last weekend, Harris stopped mentioning Trump in her speeches and focused on a message of unity. Trump, on the other hand, seemed to have some kind of psychotic snap on the podium. While Trump’s campaign has claimed they are going to win in a landslide, that sounds like nonsensical faux-boasting designed to make an extremely insecure candidate feel better while also priming his supporters to believe the election was stolen from them when he loses. It’s no secret he’s planning to declare victory before the votes are counted like he did in 2020 in an effort to make it appear that the election is being stolen. A winning candidate does not do things like this. A winning candidate can wait to declare victory until it is certain. While this election could be very close, there is no indication anywhere that Trump might win in a landslide. Is it possible he wins a big electoral vote majority by barley eking out popular vote wins across all the battleground states? Yeah, that’s possible. That’s not a landslide, though.

You might be thinking: wait! We thought a lot of these things in 2016, too. Clinton was in great shape, she was confident, Trump thought he was going to lose, everything was moving Clinton’s way, etc. That election is going to be a source of trauma for years to come. However, a lot of what we think we remember about that election is wrong. I have said this before, but it bears repeating: the reason people thought Clinton was going to win was because of the election forecasts from Princeton Election Consortium, FiveThirtyEight, and others. In fact, a lot of the things that this year hint at a Harris victory were not present in 2016. Late movement was away from her, not towards her. Polling - what there was of it - in battleground states was moving away from her, not towards her. What we were hearing from folks on the ground in places like Wisconsin and Michigan was that Trump was surging, not Clinton. We just didn’t think it was possible that more people would support Trump than Clinton - and that was correct, by the way. Let’s not forget that. Clinton won the popular vote.

There seems to be little indication that we will see the loser of the popular vote win the Electoral College this year, but it is possible - especially if the polling is correct and so many states are on a razor’s edge that could go either way. But this year, it appears as likely that Harris could win while losing the popular vote as the reverse. In either case, the likelihood of this happening appears much less than it did months ago (and it still was not likely then).

Here are some things to watch tomorrow night as the returns come in

North Carolina will probably be called by about 10pm ET on Tuesday. They count votes fast there; one reason is that, unlike Michigan, election workers are allowed to prepare mail ballots for counting ahead of Election Day so that they can just be tabulated  on that day. If North Carolina is called for Harris, the election is over and she has won. There is no plausible route to victory for Trump if he loses the state. It is, of course, possible - but supremely implausible. If the state looks like it is going for Harris, I would not be surprised if Trump declares victory early and before the results are in there.

It will take some time to count the votes in Pennsylvania because mail ballots cannot legally be prepared for counting until the polls open on Tuesday. However, I am hearing that election workers think it will be possible to call the election by about 2 or 3 am on Wednesday. This is probably because they have gotten better at the preparation and counting process; it’s not because there are fewer ballots to count. On Tuesday night, the returns we will see coming in will be Election Day votes. If by 10pm or so, Harris has a lead in Election Day returns, she is probably going to win Pennsylvania. She almost certainly has a comfortable lead in the mail votes.

Florida counts votes fast as well. Considering how the state has been trending, Trump should win by a comfortable margin here, maybe even as much as ten points. If Florida is close, Republicans might be in for a long night.

Senate races: Aside from the the several states we’ve been hearing about for the past few months, keep an eye on Florida, Texas, and Nebraska. A really good night for Democrats could mean Republican losses in one or more of those states. The Texas race will likely be close no matter what. Florida and Nebraska could be close or the Republican could win comfortably even on a night that Harris wins and Democrats otherwise do well. Republicans and the media have been acting like Tester had already lost his reelection in Montana, but the polling - what little there has been of it - has been close, his Republican opponent has been dealing with self-inflicted wounds (pun intended), and Tester has a history of winning by tiny margins there. I think there is a good chance Tester survives, but it will not be a shock if he loses.

One last thing for today: election certification

You will likely hear a lot about certification of elections in the coming days and weeks. Trump’s plan if he loses the election is to get his allies to deny certification in enough states to throw the election to the House of Representatives. Trump allies have been demanding that election workers investigate their bullshit allegations of fraud before they can certify results. In fact, the Georgia Board of Election has overstepped its authority and made rules to this effect (which a court has thrown out).

Certification has nothing to do with whether improper votes were counted or not. Sorry, folks, but this is not a controversial statement. However, it looks like one which is why low-information Republican supporters both believe it and are using it as a rationale to reject certification if they have no fully investigated fraud. Here’s what certification is: a process that validates that the number of votes counted was accurate. That’s it. Nothing more or less. And that is why it is a ministerial, rather than a discretionary, action. The question of whether fraud has occurred is for the judicial system, not appointed (or elected) local election workers. If there was no question of illegality in the casting of the vote at the time of the count, and the votes were counted, certification validates that the tabulation of the count is accurate. Aside from hoping to throw the race to the House should Trump lose, I think one reason behind this strategy is that the courts would not accept their arguments about fraud without any evidence in 2020.

Well, that’s it for now. We are out of time. Get out and vote! And then get others out to vote! Tomorrow’s the big day.

1. MRP does a very good job of understanding unusual things that impact voters broadly, but it cannot really deal with a highly localized unusual event – like a huge plant shutdown in a Congressional district. Nevertheless, it works fairly well and it seems to be a better method than our current polling techniques.