Okay! 1 week until EDay in VA, here are my thoughts on the early vote so far. Let's start with some quick caveats: more than 2/3 votes will likely be cast on Eday, and many more early votes are yet to be cast. The TLDR version is that EV is still consistent with a close election.
What you won't find in this analysis - statewide numbers. Our focus is on the battleground House and Senate seats. Statewide comparisons to previous elections are relatively meaningless. Control will be determined across a handful of districts.
What's more, the most important analysis is the comparison to the same point in time in previous cycles. That said, given the significant changes in ballot access in 2020, the only relevant comparisons are '21 and '22, and both are imperfect.
Final caveats: the analysis relies on comparisons of modeled partisanship across '21-'23 at the same point in time in each election (7 days out). VA does not have party reg, so the models are a proxy, but are imperfect. They do not capture candidate support.
At this point, across the battleground Senate and House districts, modeled Dems outnumber modeled GOPs (+13 in the House, +15.9 in the Senate). That said, in '21 at this time those leads were 21.3 and 26.1 in those same districts. The GOP has narrowed the turnout gap in these LDs
Is that cause for panic for Dems in VA? No - the GOP, after functionally abandoning early voting (especially mail) in '20, and carrying into subsequent elections, they have invested heavily in getting their voters out early this time around. So some narrowing was expected.
That said, it is not accurate to say that the GOP is simply "cannibalizing their Eday vote" (not a Halloween reference) by just turning out people who would have voted on Eday anyhow. It is true that a significant majority of early voters (D and R) are regular voters though.
Only about 6 to 8% of early voters in targeted districts are either first time or infrequent voters. So the overwhelming majority of these early votes are cast by fairly high propensity voters.
If we limit our analysis to those early voters who didn't vote at all in '21 (about 9% of all EVs so far), Dems have a modeled advantage of 12% in target HDs and 15% in SDs. That's decent, but in '22 the modeled Dem turnout share advantage among these surge voters was about 20%.
So you get the general idea here, more GOPs are voting early in VA than have in the past two cycles. Most are likely voters, but GOPs have narrowed the turnout gap among the infrequent early voters, so there are signs that both bases are relatively energized.
From a demo perspective, women are accounting for a slightly smaller share of EVs than they were at this point in '21 (0.4 pts). This is a function of more GOPs voting early and isn't especially significant. That said, white women are also accounting for a larger share of the EV.
Voters under age 30 also comprise a slightly smaller share of the EV (0.2 pts), so this isn't substantial, but it is worth keeping in mind that the electorate in VA in '21 was skewed fairly old to begin with.
There are a lot of additional ways you can parse the EV in VA, and I likely will share more in the coming days (we will be updating the data every day). But I'll wrap this thread up with some bigger picture observations.
- GOP investments in early voting are showing
- Dems are still voting early in big numbers
- The EV skews whiter, older, and more male than previous years (see first bullet)
- More than 2/3 votes are yet to be cast in VA
Anticipating questions about the impact of abortion on races in VA: this is my opinion, but I don't believe we've yet seen the full impact of the issue in VA. Recall in 2022, Dem turnout and performance was highest in states where the issue was on the ballot, or in the forefront.
In public polls released earlier this month, independent voters in VA displayed a certain discord: giving Youngkin high approval ratings, but overwhelmingly opposing stricter abortion laws in the state. As the electorate focuses in on his agenda, the GOP will have issues.
I'll add more in the coming days, but as I said in starting this thread, everything I am seeing is consistent with close races in VA, and the work done over the final 7 days will likely determine control of both chambers.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
My election eve reminder - during election day and in the first couple of hours after polls close, the ratio between the appetite for data that tells us how things are going to the predictive value of that data is just massive.
In 2022 people were sharing election day turnout data from a handful of precincts in two states, and drawing inferences across the country, suggesting Dems were doomed. That was wrong.
In 2018, before 830 PM eastern time, a dejected Dem strategist declared on cable news that there would be no blue wave, and that things did not look good at all for Dems (Dems gained 41 seats and won back the House).
A few quick thoughts on NV early vote, since I'm sensing an irrational Dem freakout taking place there.
1) It is still quite early. The total turnout so far is just 25% of the final early vote in 2020, and 38% of the 2022 early vote. We're going to see a lot more votes cast there, and the numbers will move around during that time.
2) We may just be overreacting to a different sequencing of when the votes are cast. In 2020 at this point the Dem turnout % was 3% higher than the GOP turnout %, but by the time early voting ended, the GOP turnout was 5 pts higher. That's a big swing. Compare that to...
A bunch of people have shared this with me, asking what I think. So this will be a mini-thread on that topic.
The tldr version is, if those numbers are accurate (or close to accurate) it is a very good sign for Harris. And they are plausible, but they could also just be wrong.
So theoretically, polling of people who already voted should be more accurate than a likely voter poll, because it is a known universe of voters, so you can quota/weight to the exact right audience, and verify turnout history.
Unfortunately, Marist didn't share sample data for the early voter respondents, so there is no way of validating if they did this. But they're good pollsters, so I would hope they did.
I've gotten a lot of questions about what I think about the NV early vote, so here's a quick thread. From a party share of the EV, there has obviously been a substantial gap closing relative to 2020, where at this point it was +16D, and right now it is +2D.
Of course, while I wouldn't compare raw numbers in a presidential to a midterm, it is worth pointing out that, at this point in 2022 the EV in NV was just +3D, and things turned out fine for Dems then.
Also, let's remember that the final Dem advantage (party reg, not vote) among all EV in NV in 2020 was +3. In 2022 it was +1.5. So Dems started off huge (with lots of mail) in 2020, but things settled back down to a narrow advantage. Meaning, let's keep watching this.
I'm hoping to have the time to write something longer/more detailed on the early vote in the battlegrounds in the next day or two. But for now, a brief thread explaining why I think a lot of the EV analyses are getting it wrong, and how complex this all is this year.
I'm going to use NC as the example, but we could really use any state and find generally the same thing.
Taken at a high level, the early vote data in NC would appear to be very good for the GOP. At this point in 2020 the EV was +22 Dem, and right now it is +1 D. But let's talk about why it's a lot more complicated than that.
This week marks the point where I believe we will have enough early voting data in key states to begin to talk about what we're seeing with more regularity, so I wanted to share some thoughts about what we can and cannot do with this dataset.
1) Early voting data will not tell us who is going to win. Just like a poll won't either. It provides important context. The right way to do this is to use EV data to inform polls and vice versa.
2) Just because one side or the other has more of their people voting early doesn't mean that they are winning that state. This should be obvious, but this gets missed sometimes, it seems.