With the election less than a week away, all eyes are on polling and early voting turnout. How much can we trust these numbers, and what do they tell us about the potential results on November 3?
Zachary Donnini, head of data science at VoteHub, joins Tim Dickinson to discuss key House and Senate races, while breaking down what the early voting numbers mean for turning Congress blue. The pair also talk about how prediction markets factor into polling during this midterm election cycle.
Zachary Donnini is the head of data science at VoteHub. He is the author of Data and Divergence on Substack.
The following transcript has been edited for formatting purposes.
Tim Dickinson
Hey, this is Tim Dickinson for The Contrarian. I’m joined today by Zachary Donnini. He’s the Wunderkind Chief of Data Science for VoteHub. Zachary, welcome back to The Contrarian.
Zachary Donnini
Yeah, thanks for having me on again, and very nice of you to say, I appreciate it.
Tim Dickinson
Zachary, everyone, we’re about a week out, everyone’s really trying to sort out how this is gonna go. What’s your view from 10,000 feet about where this election is headed?
Zachary Donnini
Yeah, yeah, so I think that it looks like it’s going to be a blue year, a good year for Democrats. And even with the gerrymandered House districts, it looks like it’s going to be a blue enough year, where Democrats are going to win the popular vote by enough to win the U.S. House. So, Republicans gerrymandered districts in Texas, Florida, among a couple other states, and because of that, Democrats needed to win about 52% of the vote. to win the House, and it looks like they’ll clear that. But the Senate map is on red turf, even redder turf, than the House battle. So that’s Texas, Ohio, Iowa, Alaska. Democrats have to win two out of those four states that President Trump carried by double digits in 2024. So that’s tough, even in a blue year to do. So that’s why the Senate’s so tight.
Tim Dickinson
As you’re looking at, you know, early indicators, you’re bringing in all kinds of data, including real-time sort of early vote metrics. What are you seeing there? What kind of… what is the data whispering to you that maybe the rest of us aren’t getting?
Zachary Donnini
So yeah, the first thing I always like to tell people about early vote is we know who’s voting, not who they’re voting for. So, what we can tell is what different groups’ turnout looks like. So, for example, we have some states in North Carolina, Florida most notably, where we’re getting the party registration of people who are coming out and voting, and what we can do is look at how often they voted in the past, because what a party looks for in an early vote voting period is activating their voters, who don’t always come out, but are coming out in this midterm election. And what we’re seeing is that Democrats, broadly, are activating two sets of voters very, very well, and that’s driving a lot of their gains, and what I would consider to be a good early voting period for Democrats, and it’s young voters and Hispanic voters.
Tim Dickinson
And in recent elections, Democrats have had trouble with that Hispanic vote in particular, and young voters are famously fickle. So what does that… what does that bodes very well for their overall, are the states specific that those turnout metrics are turning… looking good for Democrats?
Zachary Donnini
Yeah, so a lot of the best early voting data we get is from the southeastern part of the country, because early voting data, we care about early voting in states where a higher percent of the votes are going to be cast early. In a state like Iowa, where 70% of the votes are cast on Election Day, maybe I’ll take a dabble and look at some trends, but I don’t want to talk about it too much, because trends change over time, and when 70% of the votes are cast on Election Day. things change. Texas, Florida, Georgia, North Carolina, we’re getting closer to a quarter of the vote on Election Day. So when you’re getting three-quarters of the vote already, by right now, we have about half the vote in, the goose is already cooked to some degree. So in this southeastern portion of the state… of the country. where the data’s the strongest, yeah, we’re seeing that young voters who are turning out, just their party registration are about 5 to 1 Democratic in North Carolina, versus in 2024, they were about even. These are just shifts you don’t see very often. And then Hispanic voters are just turning out at an extremely high rate in Texas, including those who have been historically not very involved in electoral politics. And regarding Hispanic voters, what… 2024 was a very good year for Republicans with Hispanic voters, but in recent polling, especially in South Texas, we’ve seen Hispanic voters shift 20, 30 points back toward Democrats, so seeing these lower propensity Hispanic voters turning out now, we think is a good sign for a candidate like James Talarico.
Tim Dickinson
Interesting, interesting. As we look at the Senate, are there, you know, everyone’s thinking about Michigan, are they thinking about Texas? Some people maybe aren’t thinking too much about Alaska, or… what are the races that, for you, are sort of, like, on a knife’s edge right now? If you were to look at, like, the toss-up contest right now, what are you thinking about?
Zachary Donnini
So, I think the Democratic path to a Senate majority, first of all, you need to hold North Carolina, Georgia, Maine, sorry, North Carolina, Georgia, New Hampshire, and Minnesota. These are must holds, period. You’re losing one of these, you know, you’re in trouble. Then there are 6 more states, Maine, Michigan, which are two that Kamala Harris were competitive, or she won them, but they’re a little tricky this cycle. And then four, Iowa, Alaska, Ohio, Texas, that are in Trump-plus double-digit states. Out of those six races, Democrats need to win four, so you can have two slip-ups, and then it’s over for Democrats. I think one of the most interesting races, and stop me if you’ve heard this before, is the Texas Senate race. Currently, I’m.
Tim Dickinson
No one’s talking about it. about it.
Zachary Donnini
I know not enough people are talking about it. But I, you know, you just have to go for it. This would be the first Democrat to win Texas in 32 years, and I’ve been thinking about it a lot in the last week, because we’ve seen a surge in Black and Hispanic turnout in the state.
Tim Dickinson
Are there other second-tier races that are, you know, Texas has obviously got the big spotlight. Are there other races that give you some personal fascination?
Zachary Donnini
Yeah, I mean, I think Alaska’s another key one with Mary Peltola. So, Peltola did very, very well in the first round. She’s running against not one, but two Dan Sullivans, the second of which Republicans allege is a kind of a plant from the Peltola campaign to siphon off votes, which there’s definitely some evidence for, but I don’t want to weigh in too heavily. But anyways, Alaska’s a state that Trump won by double digits, and Peltola performed really, really well, but she still lost by about 3-4 points in 2024. So she’s hoping that that more blue national environment can boost her to win this November. What’s interesting about Alaska is we might not know who wins for 7 days or 10 days. We’re only going to get about 70-75% of the votes on election night. There’s the added complexity of the ranked-choice voting, which is usually run 10 days after the election, so we could be waiting there for a long time.
Tim Dickinson
So, just to spell that out for people, there could be a scenario where we have the control of the Senate essentially hanging in the balance on this Alaska race that is going to count. Over days and potentially weeks before we really know what the outcome is.
Zachary Donnini
Exactly. So even Maine, we only think we’re gonna get 70-75% of the vote in on election night. But Wednesday and Thursday are big days in Maine. We should know who wins Maine probably by Thursday night at the latest when we go to bed. Alaska, though, Alaska actually takes a break in counting for a week after election night. We usually… we probably won’t get anything until the 10th. But they have a reasonably good idea… good reason for this, which is that they’re a very big state. they have central counting processing facilities, so just the mail-in votes physically, most of which are often being flown in from these towns that even aren’t even connected to the U.S. road network, is one reason it takes so long to count ballots in Alaska.
Tim Dickinson
Yeah, no, but as you can imagine, like, Brooks Brothers riots running rampant in small villages.
Zachary Donnini
Yeah, exactly.
Tim Dickinson
In Alaska, the mind reels. So tell us about how you’re going to be watching election night. Like, what… as you’re… you’re going to be watching all this data flowing in, are there particular House races that are going to be well weathers for you? What are sort of, you know, key indicators that you’re going to be looking for on election night itself?
Zachary Donnini: So it’s 6 p.m. Eastern Time, polls close in Indiana and Kentucky, portions of it, and we start to get our first votes back, but to be honest, those states don’t count especially fast, so by the time we’re getting a lot of useful information, polls have already closed in the Sunshine State in Florida at 7pm. Florida has been a place of a lot of hardships for Democrats across the last four election cycles. It’s been four straight successful cycles for the GOP. But Democrats now have a reason to get their optimism back up about Florida, which is that the early voting period, frankly, just looks much better than I expected for Democrats. Democrats really struggled with turnout in this state in 2022 and 2024. At the Senate level and gubernatorial level, it’s still going to be a red electorate in Florida, but it’s just going to be less red than what a lot of people thought. And if we’re seeing David Jolly, the Democratic nominee for governor and Angie Nixon, the Democratic nominee for Senate, lose by something more like 2 and 7, or 3 and 5, or something like that. What it could do is boost Democrats up in key House races in that 7 to 8 p.m. hour, East Coast time, because Florida counts so quickly. I’ll be watching the number one race I’ll be watching earlier on election night is Florida 22. The 22nd district in Florida is very close to a toss-up in our model right now. There are a couple districts in Florida that are considered more competitive that Democrats really want to win. If Democrats are losing 14 or 25, not a good indicator for Democrats. On the flip side, Republicans really want to be winning in 7, 13, and 27 and 9. If Republicans are losing one of those four districts. Bad indicator for Republicans. 4-22, right in the middle, for Sister Scrollby watching.
Tim Dickinson
I’m not familiar with the numbers offhand. Who is the… who are the candidates there?
Zachary Donnini
So, remember, so Florida actually redrew their congressional map, so they completely kind of manhandled it. So Florida 22 is one of these new districts. It actually stretches across the Everglades. On one side, you have the red retiree-laden Collier County. On the other side, it stretches into Broward County and Palm Beach, and then there you got the Everglades in the middle. So this used to be a Democratic seat. It is now a seat that Trump won by 10. The Democratic nominee, Pia Dandia. against the Republican nominee Casey Ascar in an open district, right? This Trump plus 10 type seat, this is the type of thing where if we’re seeing about a 10-point shift toward Democrats from 2024 to 2026, it’s gonna be highly competitive. On the one hand, it’s only one seat, right? There could be certain specific effects at play here. On the other hand, this is the first kind of toss-up seat that we think we’re going to be able to call on election night. Because of the speed of counting that Florida has.
Tim Dickinson
I see. So, just to recap what I’m hearing, we’re looking for tightness in the Florida statewide ballots, the governor.
Zachary Donnini
That’s what Democrats want to do. Even if Democrats are losing Senate and Governor, it’s Florida. Republicans get hundreds of thousands of people fleeing Ohio for Florida as they reach retiree age and they decide to avoid those cold, cold winters. I know that Sherrod Brown isn’t going to be complaining if they’re padding that margin in Florida for Republicans instead of, you know, casting a deciding vote in Ohio.
Tim Dickinson
So, Florida 22 and tightness there. And there’s been an unusual push by Democrats, by the Progressive Coalition led by Indivisible and No Kings and others. get people voting early. And I wonder, is there a chance that that is creating in itself sort of a blue mirage? Are we sort of picking up on something that people are banking those votes early, and then on election day, we might find that there’s not a lot left in the tank? I wonder what your thoughts are on that.
Zachary Donnini
something we’re always thinking about. So, first of all, 2026 in general, we are seeing what we call democratic front-loading, in the sense that all of these leads Democrats are building in the early vote period will come down, are already coming down. They’re not sustainable. But, if you’re Republicans, it doesn’t matter if they start really high, and then they come down to just normally good for Democrats. And how we can tell how far those early vote leads are gonna come down is by looking at enthusiasm from these lower propensity voters. We call them 0 out of 3s and 1 out of threes, because we’re saying, taking the past three elections, 2020, 2022, 2024, people who don’t vote in very many of those elections. And what’s happening is Republicans are just not getting those 0 out of 3s, 1 out of 3s, even 2 out of 3s out at the rate that Democrats are. So at some point, there are Republican voters that we are pretty, pretty sure that they’re gonna come out and vote. But the issue for Republicans is even getting all of those in a state like North Carolina, Republicans just aren’t bringing out enough of those less consistent voters, while Democrats are. They’re banking and building a lead, not just with the people who come out every single election and vote. But with people who are less habitual voters, and at the end of the day, it’s just a math game, it’s a numbers game, where Republicans just aren’t reaching into the depths of their base as much as Democrats are.
Tim Dickinson
Well, that’s interesting. So you’re saying that Democratic front-loading is not just the go-getter, I’m gonna be out there at no kings rally with my sign. It’s, in fact, some of these lower-propensity voters that are, you know, the kind of folks that, on the other side, Trump has been famous for bringing out, right? Sort of broadening the base.
Zachary Donnini
Like, a lot of people who are watching this right now, you’re probably a more habitual voter, because you’re watching a politics podcast, so you have some interest in it, right? It doesn’t matter when you vote. We don’t care when you vote. We look at your past history, we see you vote every single year, cast your vote early, cast your vote on election day, doesn’t matter, because we think you’re going to cast it eventually. On the flip side, imagine a random woman who’s registered as a Democrat from rural North Carolina. and she voted in 2020 and has no other voting history in the last 8 years, and we see her come out and vote, that’s really good for Democrats, because they brought someone back off the sidelines from 2024. And because she’s registered as a Democrat, and maybe she lives in a precinct that Harris won by 80 points, and her race is Black or something in the California… or the North Carolina state voter file. we see all of these different, you know, pieces of evidence, and we’re like, she’s a really likely Roy Cooper voter, so it’s good that she came out for Democrats.
Tim Dickinson
I see, that’s interesting. Are there other, data points or sort of lines of data that you guys are focused on that maybe wouldn’t be, you know, something that I get from looking at polling averages?
Zachary Donnini
Yeah, so something we look a lot at in early voting is, is age splits. So, and why we do this is because in midterms, getting your young voters out is really hard. It’s really hard. Because young voters, they procrastinate, they vote late, and a lot of these states, then, they don’t come out at all because election day voting, especially in the southeastern portion of the country. Is it too common? Compared to 2024, Democrats are at about 60% of their levels with their registered young voters so far, which is pretty respectable, because it’s a midterm. Fewer people vote in a midterm than a presidential election. Republicans, meanwhile, in these states we’re looking at, they’re at 14%. For every 7 18 to 29 Republican voters who had came out and voted by a week before Election Day in 2024, 6 of them haven’t come out this time, and only one of them has. This isn’t normal. This isn’t the… type of shift we see, and what this means… I’ll give you this. The state of Louisiana, which Trump won by over 20 points, 18 to 29-year-old voters so far in Louisiana are D plus 40. you just don’t see stuff like that very often, and this isn’t a fixed… it’s not like… it’s like, alright, a week from now, it’s all the way down to D plus 5 or something, you know, you know, it’s funnel acid. No, I mean, it’s been getting bluer each day, if anything. Interesting. It’s probably come down some, but just this, this total kind of failure by the current Republican leadership to get any of their young conservative people off the couch and to vote. is developing by the day is surprising to a lot of people, and I’ve just been thinking recently about how the 2028 Republican presidential primaries is… someone’s probably going to be trying to do something about this.
Tim Dickinson
You guys are working with prediction markets, too? I’m curious, it’s sort of a new element in this mix of data that’s coming in. I wonder what, what, what trend lines are you picking up, and are you feeling confident about that as an input?
Zachary Donnini
So what’s interesting about prediction markets, right, is I think a lot of people are always talking about prediction markets or polls. And I don’t know about that idea in the sense the number one thing that’s gonna move prediction markets around is polling data, right? But prediction markets also consider other factors. So, for example, I think something that’s been really useful recently that prediction markets are capable of capturing is how to react to this early voting data. because we really struggle to make models, honestly, that react quantitatively to early vote data, because data availability is so inconsistent at the state and even county level. But something we’ve seen, for instance, on Culture recently is Roy Cooper’s margin has gone from predicted to win by 9 points to 14.5 points in the last, I believe, 10 days or so. Which is… Yeah, I mean, you know, as someone who analyzed a lot of this data, that makes sense to me. I’ve been talking about how Republican turnout has just been putrid among low-propensity voters in North Carolina, but I do think it’s nice to have data points like this that are able to combine a lot of different factors into some, you know, kind of market-based outcome that people are able to converge on, roughly what the conventional wisdom is. That will happen in a race.
Tim Dickinson
you know, predictions, polling can always be off. And so, help us, help us, you know, for folks who are on the blue side of this political spectrum, like, burst the bubble a little bit. What’s like a statistically defensible outcome here that doesn’t go well for Democrats?
Zachary Donnini
Yeah, I mean, I’ll be dead honest, if Democrats lose the U.S. House, they’ll be at a bit of a loss. It would be a rough pill to swallow for the polling industry, and would make Hillary Clinton’s loss in 2016 look like… look like nothing, right? If Democrats lose the House, it’s just a massive, massive miss. So what I’ll give you, for a not-great outcome for Democrats, is we get about a D plus 6.5 national environment, which… As a reminder, 2018 was only a D plus 7.3 national environment, so it wouldn’t be that much worse than 2018, which we still call the blue wave year, but it would… for Democrats for two reasons. One, the gerrymandering in the Senate map means that Democrats would just win a small majority in the U.S. House, maybe similar to what the Republicans have now, with 223 to 225 seats. And then the red Senate map would mean that Democrats with that type of national environment could lose key U.S. Senate states, like Maine and Michigan even, or also just Texas, Ohio, Iowa, Alaska. Losing all six of those races is relatively unlikely, but even if Democrats only win one of them, that would still just be 49 seats. Not only would you not win a majority this cycle, but it would also make it pretty tough for Democrats to win a majority in 2020. 28. So I think that’s kind of the realistic floor for Democrats, where it’s something I can totally see happening. I would probably be able to, you know, explain it and understand it the day after Election Day, but it would definitely be on the negative side for Democrats.
Tim Dickinson
I see. So, so a small… a small taking of the House, and then the sort of red wall holds in the Senate, is sort of the…
Zachary Donnini
And it would be tough for Democrats, because that would be a blue year, right? D plus 6 plus 5 is a blue, blue year, just with how the maps lay out, you don’t go home with very much.
Tim Dickinson
I see, I see. Well, listen to this, it’s always been… I mean, it’s just great to talk to you and get this insight. Do you have any other, big or little thoughts that you want to make sure our listeners and our viewers come away with?
Zachary Donnini
Yeah, I guess the last thing I would say is, is that I think polling is very tricky right now. I don’t poll very much. VoteHub did release two polls last week in Ohio and Texas. But I talk to a lot of people who are pollsters, because we do a lot of forecasting based on polls, and what they say is. It’s really hard to get people to pick up the phone, and even if you’re reaching people who say they are Republicans, or independents who lean Republican, you don’t know that that’s going to be a normal Republican, in the sense, if you’re hitting Republicans, because eventually what pollsters do is they start just kind of texting people who are registered as Republicans, but the 3% of those people or something who respond to your survey, and the 10% of those who say they’re voting for James Tallarico, You know, you can try to wait your way out of it, but just getting a sample of kind of statistically identical to the country as a whole is harder than it’s ever been, because of declining response rates, and specifically, we call this differential non-response, is the lingo, which just means different types of people are more or less likely to respond to your polls. So, pollsters are trying their best, but I think everyone knows that they’re working with just a really tricky situation right now.
Tim Dickinson
Well, Zachary, I appreciate your time, your insight, and hopefully we’ll have you back here soon on The Contrarian.
Zachary Donnini
Thank you very much for having me on, I appreciate it.
Tim Dickinson
Take it easy. Thank you.