As I write this, Trump no longer has a majority of the vote and that is not going to change. As the last few votes trickle in the margin between him and Harris continues to get smaller. From the latest data from the Cook Political Report, the margin stands right now at just over 2.5 million votes (out of 154 million), which makes for a 1.65 point advantage for Trump. This probably represents 99% of the vote, which means there is probably only a million votes left to count - thus making it impossible for Harris to win the popular vote (although for a few days last week it looked like it could happen).
Source: Cook Political Report
I have taken my sweet time writing a reaction to the election results of November 5, 2024. In part, that is because I wanted to give plenty of time for the data to come in. Also in part, I was exhausted upon returning from my campaign work. I still am tired, but it’s back to work whether I like it or not.
The most important thing we can take from the election results is confirmation of how disingenuous the claims of election fraud have been from Republicans and their supporters. As late as mid-day on November 5th, Trump was screaming online about “massive” voter fraud in Philadelphia. Once the results started coming in, you haven’t heard one word about fraud anywhere - except from a few sore loser Republicans who no one - even on their side - is taking seriously. There never was any significant voter or election fraud in this country. Not in 2016, not in 2020, and not in 2024. But each year, Trump screamed about he was being scammed out of a victory, only to shut up about it the two times he won. When he lost, he tried to overthrow the Republic.
The polling averages were accurate this year, nationally and in the battleground states. Now, being accurate for a poll is not the same thing as getting the outcome exactly right. All of the results were within a standard 3-4 point polling miss (call it the margin of error, if you prefer, but that’s not exactly what it is when it comes to the averages). It’s an example of why polling in close races does not help us understand who will win.
The results in an election are precise; the victor can win by one vote out of millions cast. Polling is never so precise; this is why it is better used for understanding how people are thinking about issues than it is who they will vote for. Getting the issues right does not need to be precise; if you want to use polling for the horserace, it does (but can’t be). In a close race such as this one, all we were ever likely to know from the polling itself is that the race would be close. In that sense, the polling was accurate no matter who won. If the results went the same way nationally and in the battleground states for Harris, they would be been accurate as well.
But what about recalled-vote weighting? Wasn’t that supposed to show that Trump’s polling numbers were artificially inflated. Not exactly. In a normal political environment, it was likely those numbers would have been inflated, but this technique alone did not demonstrate that the numbers were artificially inflated. But it did suggest they might be, and when you consider the other indicators we had about the election it appeared it might be the case - and I definitely thought so (and I was wrong).
However, the New York Times/Siena polling did not use recalled-vote weighting and they were pretty much on the money. They caught that Florida was not going to be close for Harris or the Democratic Senate nominee. They still saw the battleground states and the national race close, just as the polls that used recalled-vote weighting. That was something that puzzled me at the time, and it still does. Why - in most places, but certainly not Florida - did the use of this technique not make a noticeable difference in the estimates? Hopefully, Nate Cohn will figure it out for us (if he hasn’t already).
All indications pointed to the polls being wrong this year and that non-polling indicators pointed to Harris and Democrats as winners. The reverse happened. Why? Let’s briefly look at two things today: late-breaking undecideds and early vote data.
For starters - and I cannot stress this enough - we never know what the outcome of an election will be until people actually cast their votes and they are counted. No matter how rigorous we get with our methods, voting will always be a choice made by fickle humans at one point in time. It is not physics.
But do not misread that as meaning we cannot know anything. In fact, when an election is not close we can tell from polling and indicators that this is the case. In 2020 - despite what a number of pundits thought - Biden was clearly going to win the election. People think that was in doubt because so many thought he was going to win in a double-digit landslide instead of by the 4.5 point margin he did. I’ve noted before that this had more to do with focusing on the margin in the polling that the topside number Biden had. The polls largely got Biden’s number right. I am not sure why so many people thought it was possible that Biden could get 52% of the vote and win by ten points in a two-person race. It was always much more likely to be 52-47 than 55-45, although those numbers are off. I think some people are so bad with math that people thought Biden would get 60% of the vote if he won by 10 points. That’s not possible either in a two-way race.
What pollsters think they missed were “hidden” Trump voters (I use this term because I think it’s insanely inaccurate to call any Trump supporter “shy”), but at least in some cases what they were missing were undecideds who broke to Trump. And that happened again this year. A New York Times analysis shows that Trump won the undecideds who broke late. But how do we square that against the reports we had from the Harris campaign that their internal polling showed undecideds breaking their way?
Let’s dispense with the house effect argument up front. Yes, maybe there was a house effect. Harris campaign officials should have incorporated that into their thinking before they made such a public demonstration about it (nothing was leaked, it was all intentional). Did they? I don’t know. It’s unlikely we will ever see the polling data they were looking at (other than maybe the topline results). What might have happened is that Harris was leading among undecideds, but in the end most of her’s didn’t vote. That would be consistent with something we know happened: Democratic turnout was way down.
Two things happened in Democratic cities that helped Trump win the electoral and the popular vote. Democratic turnout was down, which helped Trump capture the Blue Wall states, and Democratic share of the vote was lower, which helped Trump win the popular vote. If Harris had won the same number of votes that Biden did in California, New York, New Jersey, and Massachusetts she likely would have won the popular vote. However, she still would have lost the electoral vote. Democratic turnout was way down in Detroit, leading to Trump’s narrow victory there. In Philadelphia, there were at least 36,000 fewer Democrats voting than in 2020. Overall, Harris would have won the Pennsylvania if she got Biden’s vote total. Wisconsin has turned out to be the closest state, and the drop in Democratic turnout surely made a difference in the outcome. But even if turnout alone did not win the Blue Wall states for Trump, he saw shifts in support in those states (similar to that in most of the country) made it a sure thing.
The drop in support among Democratic states like New York and New Jersey is troubling, but it was something we saw in the polling. It is why I argued that it was possible, maybe even likely, that Harris would lose the popular vote but win the electoral vote. And she almost did. At the moment it looks like a shift in 200,000 votes across the Blue Wall states would have won Harris the election, but not the popular vote. Imagine all the talk of voter fraud then from the Right.
What about the early vote? Well, here is a good example of how we cannot always intuit the way voters will behave by looking at party designation and demographics. Women and Black voters gave less of their support to Harris than I assumed. It appears Harris won 52% of women, which is in line with what Biden won. Considering the impact of Dobbs and gender politics more generally, it appeared she would increase that to perhaps 54% or higher. She did not. Likewise, about 86% of Black voters supported Harris. That’s surely a large number, but less than the 90% that supported Biden. The fact that so many Black voters have to vote Democratic for the party’s candidate to win shows you captive white voters - particularly white men - are to Trump. It would be unfair to blame Harris’ loss on women or Black voters, but she simply could not win without more support from those groups.
It appears that Harris lost the election because (1) Democrats did not turnout to vote as they did in 2020, but Trump voters did and (2) among voters there was a shift in vote towards Trump nearly everywhere. The polling actually did capture this. The other indicators did not. This is what happened. Why it happened is for another discussion.