Early vote data won’t tell you who’s winning
Here’s what it does tell us.
Good morning! It’s Wednesday, October 23, 2024. Election Day is 13 days away.
Except there’s not really a such thing as Election “Day” anymore. Mail-in voting is well underway, and pundits on cable news and social media have already started sifting through available early voting data.
So, before you fall for any premature pronouncements about the early vote trends, let’s talk about what that data can actually tell us — and what it can’t.
The other day, I walked over to my nearby UPS location and cast my ballot in the 2024 election.
Why not the post office? I’m registered to vote in Missouri, and my home state is one of three that requires citizens to get their mail-in ballots notarized. (Mississippi and Oklahoma are the others.) A walk to UPS and five dollars later, and my ballot was graced with the necessary notary stamp; my civic duty was accomplished.
I’m not alone. As of this morning, nearly 23 million Americans have already voted in the 2024 election, according to the University of Florida’s tracker. For context, that’s about 15% of the total number of Americans who cast presidential ballots in 2020.
Across political media, the takes are flowing. According to the left-leaning New York Magazine, the early voting data we have so far “looks good for Harris.” According to the right-leaning Washington Examiner, Republicans boast an “early voting edge.” Forbes says the early voting data shows “bright spots for Democrats,” while the New York Post reports that it signals “bad news for Democrats; ”
So, which is it?
Neither. Trying to read the tea leaves of early voting numbers is a quadrennial exercise — but it’s a futile one. Here’s why:
1. We don’t know the denominators yet. I know a lot of you are skeptical of polls. I understand that. I am too, sometimes. But the nice thing about polls is they represent at least an attempt to model out the overall electorate. Pollsters don’t just take every response they receive, count them up, and click “publish.” Before finalizing their results, they weight the responses of different demographic groups differently, ensuring that each group’s responses are taken into account in accordance with informed predictions of what the final electorate will look like.
Now, their informed prediction could be wrong, of course. NBC News did something helpful thing recently, and shared what their poll results would look like if their demographic weighting was done slightly differently. The first column below represents their best guess at the demographic makeup of the electorate, which ends up with a Harris 48-Trump 48 result (the one they published).
But if the weights were adjusted according to the subgroup sizes in the second column, suddenly the race would be Harris 49-Trump 46. The third column yields Trump 49-Harris 47. Obviously, either of these results are also credible predictions for how the election will end up.
Early voting data, however, doesn’t come with any such predictions about what the final electorate will look like. According to the Georgia Secretary of State’s office, 31.5% of voters in the state so far are white; 24.9% are Black. 41.7% are ages 60-64; 9.6% are ages 18-24. What does that mean? We don’t know, because these numbers are completely in a vacuum — they could end up being similar to the final voting breakdown, or they could look wildly different.
We don’t yet know how many people will vote in the 2024 election, or how different demographic groups will be represented in the electorate. With polls, experts try (maybe incorrectly!) to adjust the numbers they get to match what they think the final electorate will look like. But, with early voting numbers, there’s not even that attempt to control for the makeup of the overall electorate. It’s just raw data, without any sense (or even prediction) of what the final denominators — the final number of voters, or final percentage of voters that will belong to X group — will look like.
2. Partisan affiliation ≠ votes. Some states give more than just data about the race or gender of people who have voted. They also tell you how many members of each party have voted, according to the partisan affiliations voters are registered under. In Nevada, for example, 40% of voters so far have been registered Republicans, according to the secretary of state’s office. About 36% have been registered Democrats.
So, Republicans are winning Nevada, right? Not so fast.
First off, just because someone is registered with a political party doesn’t mean they voted for that party’s candidate. In fact, much of the Harris campaign’s strategy in the closing stretch has been predicated on winning over registered Republicans. Conversely, the Trump campaign is courting young men, many of whom are likely registered Democrats. (I wrote about these trade-offs yesterday.)
Sure, most registered Republicans will vote for Donald Trump and most registered Democrats will vote for Kamala Harris — but we can’t say how many “most” is, at least not with enough specificity to make pronouncements based on early vote data. Especially in a year when both candidates have placed voters on the other side of the aisle at the core of their campaign strategies.
But those crossover votes aren’t even the biggest flaw in the party registration totals. Here’s two more:
2a. They’re often outdated! No, not in the sense that a secretary of state’s office is reporting the wrong numbers. But they might not accurately represent the situation on the ground. Tulsi Gabbard notwithstanding, most voters don’t go back and update their partisan affiliation after first registering to vote — even though their allegiances might have shifted in the intervening decades. In Kentucky, for example, 44% of registered voters are Democrats. But Kamala Harris will not receive 44% of the vote in Kentucky (Biden got 36% there). The state has a lot of ancestral Democrats, who may sometimes vote for a Democratic governor, but that doesn’t mean they are about to send a Democrat to the White House, no matter what their party registration says.
2b. To return to the Nevada numbers, you may notice that 40% + 36% ≠ 100%. (Am I data journalist now?) There are also a huge chunk of Nevada voters (the remaining 24%) who are registered as “Other.” These independents could be voting mostly for Trump, mostly for Harris, or entirely down the middle. We have no clue! But any analysis of the party registration alone will leave out this big chunk of voters, further clouding our visibility into what the early voting data means. I’d also note that young voters are disproportionately likely to identify as independents — so analysis of the party registration data particularly leaves out Gen Z voters, who we know are poised to play a key role in this election
3. Uh, remember 2020? At this point, you may be tempted to push back against my assertion that we’re essentially flying blind whenever we try to sift through early voting data. Aren’t there benchmarks we can compare to? you might want to ask. Can’t we just look at how the numbers match up against the last election?
To which I say: Do you not remember the last election? The 2020 election, you may recall, was held amid the Covid-19 pandemic, when many states relaxed their early voting rules — and many, many more people than usual cast early ballots. Here’s what that spike looked like, per the team at the University of Florida:
Early voting in 2024 (or any other future election, for that matter) will not look like 2020. It was a weird time, with voting behavior that was very different from any other time, and therefore serves as an extremely poor benchmark for voting behavior today. So, yes, we are flying blind, without a recent presidential election to fairly compare the data to.
The other thing you might remember about 2020 is that the early vote was tilted considerably towards Democrats. According to Pew Research Center, 82% of Biden voters cast their ballots before Election Day, compared to 62% of Trump voters.
What is one thing the early vote data is telling us? That will not be the case this year. The party registration data — although not a perfect indicator for presidential choice — does indicate that Republicans have been much more likely to vote early this cycle than in 2020.
The highest-profile example: Donald Trump, who told Fox News yesterday that he’ll be voting early in this election. That’s an incredible turnaround from 2020, when he frequently (and without evidence) trashed early voting as fraudulent.
Republicans have invested tens of millions of dollars in encouraging their voters to vote early — essentially, trying to undo the damage done by Trump’s messaging last cycle — and, so far, it seems to be paying off, with Republicans voting early in much higher numbers than in the past.
That tells us something interesting behaviorally: Republicans are willing to follow Donald Trump, and — because he has switched on the matter — are now much more comfortable with early voting. But it does not tell us something about how the 2024 election will shake out, because we simply don’t know the denominators: we don’t know how many Democrats or Republicans will show up to vote on Election Day, so we don’t know how the early vote numbers fit into the shape of the overall electorate.
I also want to introduce one more point of uncertainty into this conversation: we also don’t know which party benefits from a larger electorate. I’ve written about this before, but it bears repeating: polling evidence and recent off-year election results suggest that it may now be Republicans who benefit when turnout is higher, not Democrats as used to be the case.
On social media, some Democrats have been celebrating stats like those from Georgia, where early voting shattered records in its first week. That’s understandable muscle memory from an earlier era, when Democrats were the low-propensity-voter party — but that no longer seems to be case.
It’s not clear which party stands to benefits if voter turnout is higher this year — meaning not only can the partisan breakdown of the early vote tell us little about the election outcome, but neither can the total number of early voters.
More news to know
Trump’s former chief of staff John Kelly said that the former president “prefers the dictator approach to government.” Also, Trump once asked him why American generals weren’t more like Hitler’s.
More plagiarism allegations surfaced against Harris.
Trump is going on Rogan.
Biden on Trump: “We gotta lock him up.”
Walz on Musk: He’s Trump’s “running mate.”
A Trump transition memo named Aileen Cannon — the judge who dismissed his classified documents case — as a potential attorney general.
Democrats are getting more worried about the Senate.
Russian operatives were behind a fake video smearing Tim Walz.
After decades of refraining from political giving, Bill Gates has donated $50 million to a pro-Harris group.
Georgia poll workers will no longer have to hand-count the state’s ballots. And speaking of Georgia poll workers: two of them are about to own Rudy Giuliani’s luxury condo.
Israel said that it killed Hezbollah’s new leader, less than a month after killing the last one.
The IMF says the U.S. is the fastest growing of the world’s advanced economies.
Could Filipino Americans flip the race? Could Polish Americans? Could Americans living abroad?
Another Trump event was canceled.
The day ahead
Kamala Harris will participate in a CNN town hall in Chester Township, Pennsylvania. Tim Walz will headline a fundraiser in Louisville, Kentucky. Doug Emhoff will hold campaign events in Hallandale Beach and Coral Gables, Florida. Jill Biden will hold a pair of campaign events in Las Vegas, Nevada.
Donald Trump will address a “Believers and Ballots” town hall in Zebulon, Georgia and a Turning Point Action rally in Duluth, Georgia. JD Vance will hold campaign events in Las Vegas and Reno, Nevada.
Green Party nominee Jill Stein and Libertarian Party nominee Chase Oliver will participate in a debate in Los Angeles, California.
Joe Biden has nothing on his public schedule. The House and Senate are on recess. The Supreme Court is off for the week.
Before I go…
Here’s something nice: Residents of the Swedeborg, Missouri — a small community in the middle of the state — were deciding earlier this year who to name an elementary school building after.
Should it be a national figure? A famous Missourian? Instead, they decided to name it the Claudene Wilson Learning Center, after the beloved custodian who has worked at the school for decades.
“These kids get sick, you think they go to the nurse to start with? They go to her,” said Chuck Boren, president of the Swedeborg School Board, which voted unanimously for the new name. “If they had a bad night, they go to her. And she’s there for each and every one of them.”
Hooray for all those kids and the community of Swedeborg, Missouri. We should all be like Claudene. Thanks for sharing this story, Gabe.
Great job on breaking down the early voting numbers; I've been seeing them mentioned recently but wasn’t sure how to interpret them. I find that incredibly helpful.
However, I’m puzzled by how you portray statements like "Trump’s former chief of staff John Kelly said that the former president 'prefers the dictator approach to government,'” and “Trump once asked him why American generals weren’t more like Hitler’s,” as reputable facts.
In contrast to the phrasing "More plagiarism allegations surfaced against Harris".
This inconsistency suggests a bias in how you present different stories. I've noticed it before and didn't say anything but this close to the election makes misinformation a problem.
Its a similar trap that progressive journalists unfortunately fell for with the Hunter Biden laptop story. A great number of journalists joyfully trumpeted that the laptop was misinformation and cited 51 intelligence officials as their sources... and I think its as clear today, as it was then to any unbiased onlooker, that the laptop was not only real but explicitly covered up with lying partisan sources.
So we have stories like the John Kelly one you mentioned that declare Trump in one way or another—always behind closed doors with some disgruntled party who was either fired by Trump or hates his guts present—said he admires Hitler, the Third Reich, white nationalism, and is going to crown himself dictator of the realm. Maybe its true, I don't know and neither does anybody outside of that room. The story is important and deserves to be covered, no doubt, but to state the Kelly story with conviction while the plagiarism accusations against Kamala Harris are downplayed and merely label as "alleged."... when they can be confirmed with a cntrl + f on a wikipedia page! She has been caught in writing plagiarizing.
She was caught using teleprompters when she shouldn't of and dropping out of interviews early. This is the stuff that matters and needs to be stated with conviction. Similar to how JD Vance back pedaled his criticisms of Trump, and I saw that you stated that clearly and with conviction—as you should. This kind of pick and choose coverage turns off independents like myself and half the people I've forwarded this newsletter to. To put it bluntly, it seems like you are intentionally twisting things 13 days ahead of an election and its reminiscent of how the aforementioned Hunter Biden laptop story was covered. I guess I have to resign myself to the fact that everywhere I go to get my political news I am going to encounter spin.
Its not only the above examples but numerous others over the past few weeks but I digress. It's your substack afterall and you can present it how you wish, but I follow you for unbiased journalism and it's disappointing to see this inconsistency.