We are a week out from the midterm elections and something that looked almost impossible a year ago has happened. There is a path for Democrats to win not just the House but the Senate. So a Senate majority for Democrats, to sort of recap this: they need to defend the competitive Senate seats they currently hold, and then they need to flip at least four seats in states currently held by Republicans. A couple of those states had always looked possible, but then came the hard ones — states where Trump had won by double digits. People weren't really talking about Kansas, but now, Kansas. And the remarkable thing about this cycle is that Democrats are polling competitively in all of them. In a handful of those races, they're actually far ahead.
And so, I think the question is, can we trust these polls? What can we really say about where the election stands just a week before the vote? To talk about all this, I want to have Nate Silver on the show. Nate, of course, is, you know, internationally famous elections nerd.
The polling guru Nate Silver, he predicted every state in the last presidential election. Heavy is the head that wears the crown. Silver. Absolutely. These days he edits the Silver Bulletin on Substack, which puts out a ton of great analysis, polling, and election forecasts. He co-hosts the podcast Still Counting. Trump's approval rating is now lower than it was after January 6th when he was...
No, it's, you know, it's as low as it's ever been.
And he's the author of, most recently, On the Edge: The Art of Risking Everything. Nate joins me now.
Nate Silver, welcome back to the show.
Hey, Ezra. How are you, man?
I'm good. I wanted to ask — I didn't follow up when you said this to me a second ago because I want to do it on air. You said to me, looking a little far away in your eyes, that this election feels momentous to you. I've known you a long time. You don't usually say things like that. So, what feels momentous?
If the election goes roughly as the polls say it will go, then in some ways it would mark the end of the Trump era, I think. He'll still be president for two more years. But the minute the election, the votes are counted — might take a while because of Alaska and California and states like that — the 2028 election campaign begins, and if Trump has a bad night, as polls show, then he's a super lame duck, right? You might draw an analogy to Bush after 2006.
And like, part of what people understandably are having trouble believing in the polls — it kind of looks like a different regime, right? Where Kansas is competitive, Texas is competitive, potentially gains for Democrats deep into Republican territory. But maybe we're on the verge of something different, between generational turnover, between Trump being three-quarters of the way through his two constitutionally provided-for terms, and with AI looming too. It's not quite clear to me how big an issue AI is in this election. I think maybe more than people assume, but it could be the biggest issue of 2028. I think you might expect it to be, along with the economy, the biggest issues of 2028.
I was thinking, as you said some of that, about the conversations I had about the election a year ago.
And a year ago the conversation was, Democrats will probably win the House and the Senate. I mean, never say never.
But no. And the reason was that to win back the Senate, Democrats were going to need to win in a number of places where Trump had won by more than 10 points. And given everything we know about partisanship in modern America, that wasn't going to happen. So, you know, when we're talking — it's about two weeks before the election right now, it'll be less by the time this airs — Democrats are at a 70-plus percent chance of winning the Senate in your model. And not just winning the Senate, but possibly winning like beyond that, despair.
Why? What happened?
So, I think a couple of things happened, and some of them have to do with expectations maybe being miscalibrated. You know, after 2024, there was talk of a, I guess, permanent conservative vibe shift. That Democrats were on their back heels and Biden had failed and wokeness.
That was dumb. The idea they'd be permanent.
Yeah. And but if you look at the history of American elections, then the most regular trend in American elections, one of the few things that survives for decades, is there is a backlash against the president's party at the midterm. Maybe a little bit bigger if a president is in his second kind of lame duck term, but you know, we can recite the exceptions on one hand: 1998, after Monica Lewinsky; 2002, after September 11th. That's about it. 2022, Biden had a relatively gentle thumping in the midterm compared to what you might expect based on his approval rating. So, you know, look, in the average midterm the environment shifts by seven and a half points against the president's party versus the previous presidential year. This year it looks more like 11 points.
But why is 11 points important? Well, there are all these states like you said that Trump won by 10, 11, 12, 13 points, right? Ohio, I think he won by 11 points. So if you have an 11-point swing and it's fairly uniform, all of a sudden Ohio becomes a toss-up. What I think wasn't so clear is that Democrats have actually nominated, in most important races, pretty good candidates, and/or lucky enough to face weaker Republican candidates, right? I'm not sure if James Talarico is everybody's cup of tea, but like, you know, he is able to take advantage of Ken Paxton's many sins and so forth, right? In North Carolina, Roy Cooper was a strong recruit from Schumer and Company. You know, Democrats have the strategy now of trying to nominate independents and kind of name-only candidates in states like Nebraska. I think both the candidates, Republican and the Democrat, are strong in Iowa, for example. But, you know, I would say that probably the candidate intangibles, you know, they might move things by five points at most, but you kind of do the math and you're like, okay, well, Texas is a Trump-plus-13 state, the partisan climate moves 11 points toward Democrats, so now it's R-plus-two, and then is Ken Paxton below average enough to turn R-plus-two into D-plus-two? Well, that's the math, right? And that's kind of, that's a logical enough story.
So one question that raises for me is Schumer and his team, the Democratic Senatorial Campaign Committee. It's under Kirsten Gillibrand right now. They looked pretty bad a couple months ago, and they looked pretty bad primarily because of a focus on the fight that they waged and lost in Maine to make Janet Mills the nominee, and instead it was going to be Graham Platner. Graham Platner imploded.
Troy Jackson, who's the Democratic nominee now, you know, is ahead of Susan Collins by a bit. I think people are nervous about that race because Collins tends to outperform her polling, and it's not that, you know, wide of a gap. But if you're looking over the range of states, do they look pretty good? I mean, Turk in Iowa, you know, Peltola in Alaska, Roy Cooper, right? It seems like...
Peltola is a big one. I'm not sure that an average Democrat would be in a toss-up. She's, you know, a slight favorite in that race, actually.
Look, yeah, you have these big fights in Maine and Michigan. I would share Democrats' concern about Maine a little bit. I mean, that's based on nothing but just the fact that Susan Collins beat her polls by like literally like nine points in the last election. There's a whole argument about how persistent polling error is. We tend to assume that it's not, but that's one race where the polls have gotten maybe closer by a point or two. There hasn't been a lot of polling in Maine. Michigan — look, I'm one of those people who says that other things held equal, that a moderate can do better than a progressive, at least in a purple state like Michigan, right? And if you compare the governor's race there, where Jocelyn Benson is ahead by eight points, Elissa Slotkin is ahead by three or three and a half or something, right? Again, that, you know, there is a four or four and a half point candidate quality gap, that is again within that range of what we think candidates, how much they can matter in an environment like this one, right? But if you have enough wind at your back, then you can overcome some of those problems. And by the way, in other ways, Elissa is a good candidate, right? She's very Michigan in certain ways, right? She, in kind of a Zorn-esque way, doesn't present — I mean, she kind of is a bro, right? It's an interesting contrast to like Francesca Hong or somebody, and there are gendered parts of that comparison too, potentially. She's also raised a lot of money, and money is helpful. So, but yeah, I mean, Michigan and Maine — if Democrats were to lose one of them, they'd still have a chance. If they were to lose both, then probably game over for a Democratic Senate flip.
One of the things that has been interesting to me looking at the polls, and people have commented on this — this is not my original observation — is that Democrats seem to be overperforming more in redder states.
So they look better than you would think in a Texas, a Kansas, an Iowa, maybe like a little bit less good in like a Michigan, a Minnesota, you know, it kind of goes on like in the blue states, right? Is there something to this? And if so, what do you make of it?
Well, that's what I think makes Democrats nervous, because it's the red states, especially in the heartland, where the polling overperformance has been concentrated in some of these recent elections. At the same time, the story does make sense on a couple of levels, right? Number one, gas and diesel prices. If you look at where diesel fuel is consumed — not usually by passenger vehicles, but by agriculture, trucking, petroleum, home heating oils, right — it's all kind of the colder Midwestern states where a lot of the battlegrounds are concentrated, as it happens this year. If you believe in kind of a last-in-first-out argument, like what if we are reverting somewhat out of the Trump era? Actually, in 2024, for the first time in a while, you had very mild depolarization. Some of the worst states that Democrats had relative to 2020 were like New York and California. You know, younger voters of color, for example, and younger men in general kind of moved against Democrats. Whereas, you know, like relatively speaking, if you're really looking in the weeds, like Kansas is a state where Harris — I think the PVI, we call it, the gap between that and the national spread declined in Kansas a little bit. So maybe you're having some last-in-first-out where the people that joined the Trump coalition, like Hispanics, working-class whites, maybe just moving a little bit back in the other direction, or maybe just not voting, right? You know, one reason why Trump polls in '24 was overperformance among like young men. The one group you do not want to rely on to turn out in a midterm is usually young men.
So I feel like there's another story you could tell, and I'd be curious how this rates for you, which is that it actually has to do potentially with candidate quality on both sides. So in redder states, Democrats have been, I would say, more rigorously following a view that they need more moderate candidates. So you think of like Adam Hamilton in Kansas. Sherrod Brown is like a longtime populist, and you know very well, but Roy Cooper in North Carolina. John Ov is this like — I mean, I wouldn't say Georgia is a red state anymore, but John Ov is this like relentlessly disciplined figure. Nebraska with Osborn, who's actually running as an independent. I wouldn't say Democrats chose him, but obviously Democrats are aligned, I think, and excited about his candidacy and didn't run a Democrat against him. And as you get more into the more traditionally purple or bluer states, you know, Maine, they have nominated like a more left candidate, you know, Jackson replacing Platner, who is also a more left candidate. Obviously, Abdul El-Sayed in Michigan, the more left candidate won. In Minnesota, the more kind of left candidate won for the Democrats, and it seems like she's going to win, but is performing a little bit worse than people are comfortable with. So, I feel like one way you could squint at this is to say, in redder states, Democrats have been, you know, very purposefully choosing candidates who just seem more independent, right? You know, and that as you go down to where Democrats felt more comfortable and could take more risk, they're taking more risk and trading away some of the possible gains. Like that would be the moderation-wins thesis of like what we're seeing.
Well, and Florida is a state where you had a DSA candidate that isn't on anyone's map because, you know, she's now trailing by six or seven or eight points, which isn't that bad for Florida, but like, but there you see the failure to select a — you know, you need a pretty conservative Democrat probably to win in Florida at this point.
No, I think that's a lot of it. I think a lot of it also, if you look at a state like Minnesota where Peggy Flanagan is ahead, but by three and a half or four points, not by nine or 10, which...
Yeah. You know, the states where you have these Democratic trifectas and you have critiques of Democratic governance is where you tend to see more problems potentially. In Oregon, where you've had a lot of critiques of liberal governance, right? Tina Kotek, the incumbent, I think is only ahead by like one or two points in the polls. You know, Kathy Hochul certainly looks like she'll win re-election, but maybe by a relatively tepid margin. You haven't seen that as much in like Massachusetts, which is a classically well-governed state that's often had a Republican governor in the form of Charlie Baker, right? But you know, I think people are — there's an anti-incumbent mood. People, I think, are making this implicit calculation about who has kind of power over my life, right?
But not against all incumbents, right? Like Ossoff, the incumbent in Georgia, is going to stomp. I mean, from what we can tell in the polls, his challenger. Roy Cooper is the incumbent governor in North Carolina. Yeah. Or, you know, obviously was like, he's going to like win in a walk. So, like there are Democratic incumbents who are doing just fine here. I think Minnesota is the clearest example of this because Tim Walz's approval rating is 17 points underwater or something, to the point where he was going to run for another term and didn't. So, it's kind of like ground zero.
He's had a real fast rise and fall in politics.
The trajectory of losing vice presidential nominees, you know, John Edwards had a rough fall from grace, for example. You know, Paul Ryan no longer in Congress, right? It's pretty, you know, pretty rough.
One more thing for these hypotheses about the heartland and the Midwest. The Economist had some data showing that like the closer that Trump voters live to data centers, the more likely they are to flip and vote Democratic in this election. And in the West in general, you have a lot of data centers, right? Because you want some space, but you don't want to be too far off the grid and too rural. And so that could be a hidden factor, too. There's something I was thinking about while I was like going through and trying to decide which of the candidates seem to me to be overperforming the most. I was thinking about how uncorrelated attention and performance is right now. So it's like if I go through the candidates who are all over my feeds and who are on big podcasts — and I mean obviously some of this is also their choice of how to run — but it's like, Abdul El-Sayed, like omnipresent, right? James Talarico, omnipresent. John Ov, he's not going on many national podcasts at the moment, but his clips are absolutely everywhere. You're not seeing Sherrod Brown, right? You're not seeing very much of Mary Peltola, right? She does not do a lot of national press. You're not seeing a lot of Roy Cooper. But it's really like you look at it and there is almost no connection between who are the candidates who are everywhere, you know, on social media and my feeds, and who are the candidates who are doing practically well or poorly, at least in polls.
Yeah, look, some candidates have different incentives. If you want to run for president in 2028, then you probably want a somewhat higher profile potentially. I think in a case like Talarico or Elissa, they kind of inescapably feel that they're going to get the maximum amount of attention anyway. But look, there's something to be said for kind of keeping a lower profile, too. You know, maybe Ohio is a good example of that. Maybe Iowa a bit. I mean, look, in midterms, there are basically three big questions, right? Will your voters turn out? Will the other side's voters turn out? And then how much persuasion is taking place among, you know, the 15% of the electorate that's possibly swing voters. And you know, if Democrats were to come up short in a couple of races like Iowa or Kansas, they might say, you know what, this race just got like a little bit too much on everybody's radar, and Republicans managed to drag some voters over the finish line. I mean, the persuasion part of this is done. I'd be shocked if Democratic turnout wasn't high by midterm standards. We don't know if Republicans are going to turn out, and that will determine the range between Republicans hold the Senate and keep Democrats to 10 seats in the House and complete 1994-style disaster.
So right now, I mean, we're talking a little bit as if the thing Democrats are hoping will happen will happen.
But we often see polling error in midterms, and maybe we see polling error in general elections too. So, how resilient are Democratic hopes to polling error, and how are you thinking about where the polling error might be? Like when you're building things into the model and thinking about like what you're trying to correct for, you know, you've seen Trump underestimated many a time, seen Susan Collins underestimated many a time. On the other hand, the pollsters know that has happened, right? How do you think about the polling error dimension of this?
Look, the first thing I always tell people is that whenever you're looking at a series of polls in a Senate race or in Senate races plural, right, always ask yourself, what if they're off by three or four points in the same direction in every race? You know, Democrats have a lot of three-point leads in races like Ohio and Texas, more or less, right? If you had a three-point polling error, you would have a very long night where we might be waiting for ranked-choice voting in Alaska. But it's, you know, the House is different. If Democrats were not to win the House, then there would be profound questions about what's going on with the polling and in some ways what's going on with the country, too. I think Democrats have the emotional feeling that when there's a polling error, it is, you know, against them, right? Because it, you know, Hillary Clinton and Joe Biden and just in general, polls actually have repeatedly underestimated Donald Trump. But is that true? And is it particularly true in midterms? Have polling errors been consistently against Democrats?
It has not been true in the two midterms under the Trump era. I guess 2018 was under Trump, and their polls were, you know, one of the most accurate years for polling ever. 2022, similar. In fact, there were a couple of races that were high-profile, like the Fetterman race in Pennsylvania, where polls showed Oz closing the gap after Fetterman's stroke, right? And then Fetterman won by a pretty solid, I think, four or five points or something like that, right? But overall, polls were quite accurate. There are also elections that are held in odd-numbered years, runoffs, things like that. In New Jersey and Virginia last year, the polls were five to 10 points low on Democrats, which is one of the biggest polling errors we've seen in a well-pled, you know, general election. So yeah, the critique I have of some of the people say — they'll say, well, you know, you're undersampling Republicans in these polls, and Democrats are much more enthusiastic and much more likely to respond and trust the pollsters more. Every single person in the industry is aware of that. Look, I think every pollster is aware that polls missed low on Donald Trump in 2016, 2020, and 2024. You know, the three presidential years was kind of the death knell for the gold standard of polling, which is like we're just going to go and survey people and we'll weight to well-known census demographics and we'll just trust the data. The Ann Selzer poll in Iowa — which, bless, I mean, she was literally the highest-rated pollster. She and the New York Times were like, I think, tied for number one or something, right? But now...
So say what happened with Selzer, because maybe some people are emotionally scarred and some people have no idea what we're talking about.
Yeah. What was it — it was the weekend before the election, right? And I had like some dinner plans on Saturday, and then I got like a text from a friend saying, have you heard about this Iowa poll? There's a rumor that like Ann Selzer, the kind of, you know, dame or don of Iowa polling, right, who had predicted Barack Obama's big win, the Iowa caucus in 2008, would have a poll coming out showing Kamala Harris winning, I think, by three points in Iowa. Every other poll of Iowa had been Trump plus nine or something like that, right? And our model was pretty skeptical. You don't usually want to bet against Ann Selzer, but like it kind of opened up this whole world, like what if everyone's just kind of way overcompensating for past elections? And it didn't make a lot of sense, but she had gotten very far by just saying, I'm just going to trust my data. I'm not going to overly massage it, not make too many assumptions about it. And the poll was off by what was it, like 13 or 14 or 15 points.
Trump insanely sued the paper as if like publishing a poll where he was losing was defamation.
Yeah. Yeah, it's never clear what the incentives are supposed to be.
Can I stop on this polling thing for a minute though? Let's get in the weeds methodologically for a minute, because I find whenever we start discussing or I start thinking about how what is happening in polls — to massage demographics, to kind of answer all these problems of people don't pick up the phone or, you know, we're using online sampling and we don't know who will turn out — I feel very disoriented. What do you do? Like what are the pollsters doing?
And I should say, I am not a pollster. I'm a forecaster, right? But that distinction in some ways is becoming blurrier, right?
That's what I mean. They feel like they're forecasting.
I think the upshot of the Ann Selzer thing is that everyone knows that a poll is kind of almost like a mini model with polling-flavored inputs. And I'm not sure what the right metaphor is. But people wanted to make it like it's some broken gauge that like you're — that you know, on your beat-up old car, that when it says 25% of fuel left, you actually have to refill because the gauge is broken, right? Instead, this is more of an iterative process, right? You open a restaurant, the customers complain the food is too spicy, you tamp it down a little bit. Now some of your customers say, no, I wanted the spice back in my food, right? And so, but it's more of an iterative process. And I think pollsters are now aware they'll be judged on their results in the final few weeks at least. And as much as they might say, well, it's just a snapshot of public opinion, there are huge reputational stakes on the line for being right. So I mean, and for being wrong.
So the pollsters have pretty good incentives to correct for these problems potentially. But you know, but sometimes when it — I know it sounds kind of tautological, right? When there are these big wave-change elections or elections that are kind of unexpected, and sometimes pollsters lowball them. In 1994, for example, it looked like a relatively bad year for Clinton that turned into a terrible year, right? You can also look at a year like 2014 as a parallel that I'd be worried about if I were Republicans, because that's a year when the polls were never great for Obama, but you had a really, really big differential turnout where Democrats just weren't that enthusiastic. At the time, Obama was relying on more of a multi-ethnic younger coalition that traditionally does not turn out at the same rates in midterms. And that turned into a nine-seat loss, a total rout for Democrats in the Senate.
Yeah. I mean, this is, I think, what I'm asking you, that we know the midterm electorate is going to be different than the presidential electorate, that some people are going to be more motivated to vote than others. I mean, are they just polling people and saying, do you think you're going to vote? And that's like the big screen just for somebody who does not — like I have read analysis of these, The Upshot a while back, and Nick Cohen did a cool thing where they let you see like how different ways of moving the data around would give you a very different result on the same like underlying poll data. But is there like a kind of conventional thing they're doing to compose a midterm electorate that is going to be really different in 2026 than it was in 2022?
Yeah. So, one difference: if you ask people, are you going to vote? How enthusiastic are you about voting? That produces a very Democratic sample. Other pollsters try to predict whether people will vote based on other types of data, right? And those techniques tend to produce more modest results for Democrats. One thing you can watch in the early voting data, which generally I advise people not to do, but who is winning among like lower-propensity voters, right? People who have voted in no more than one out of the past three elections, counting midterms. Those low-propensity voters appear to be quite Democratic so far. It maybe probably will to some extent. Republicans will cast their vote by election day anyway, and they'll catch up, right? But if not, if Democrats' low-propensity voters turn out and the GOP doesn't, then that's where you might see a polling error that underestimates Democrats.
So if Republicans end up having a better-than-expected night, what do you think will have ended up being the explanation for it?
I mean, look, I think there are two versions of this. Like one is just Democrats lose all the toss-ups, right? Talarico loses by a point. Jackson loses by a point. Maybe Elissa wins by a recount. Sherrod Brown wins, but Turk loses by three points. Like that's very feasible in the kind of thick part of the probability distribution. And then I think, you know, with some questions about the polls, but that's kind of in line. We'll say, okay, partisanship is still really strong. Maybe pollsters and Democrats had their hopes up a little bit too high to think you could win on these, you know, R-plus-12 states. If it would have been R-plus-8, then yeah, but we just kind of reached the point where we ran out of like oxygen climbing up that big red hill, right? If there's a — if there's like a — like let's say the House is kind of close, right? And Democrats only pick up North Carolina in the Senate, right? I mean, then that's kind of hard to contemplate almost, right? Not just not because it's not theoretically possible based on the polls, right? Like, you know, we have these in our model these like fat tails that say sometimes polls can be off by five or six or seven points or something crazy, right? But that would kind of like defy everything that we think we know about how midterms work.
There are, I think, a couple of races that feel to me like they have like a lot of outsized narrative importance. Talarico feels to me like it's got a lot of narrative importance to people. You know, the possibility of Democrats being competitive in Texas, a sort of return of a religious left. Abdul El-Sayed is going to be seen as a very important test of, you know, he's actually not a DSA guy, but sort of a more left populist, you know, unsparing candidate. Like I was watching this video where he just like absolutely annihilates this guy who comes to one of his town halls and is like, will you denounce the Muslim Brotherhood? And I mean, El-Sayed, who's a very, very, very smart guy, leaves this guy like a smoking crater on the floor.
So it feels like there's going to be a lot of attention on how he specifically does.
You know, you've been very attentive to like the factional divisions of the Democratic Party recently. I think you've cut Democrats into what is it, the capital L left, the abundance libs, which I guess is me.
And then Richardsonism, which you've named for Heather Cox Richardson, who's, you know, the Substacker, and that's sort of the big, like, resistance libs, no kings kind of thing. But I'm curious how you think different forms of the election might end up being interpreted and shaping the Democratic Party and its strategies.
So the two big, apart from who wins, right? The two big sub-storylines of November 3rd are, one, how do progressives do relative to moderates? But then is there evidence of, like, an AI backlash from the state-by-state results, right? If people do these ecological analyses afterward, they find, hey, yeah, in states where there was data center construction, Republicans underperformed by three points. I mean, that might be important as well.
Yeah, if I were to lose, then win or lose the Senate, and that's a very bad narrative for progressives. I don't know what happens if, like, you know, let's say that Benson wins by nine points and Elsa wins by four, right? Maybe nerds like me are like, well, he had a below-average wins of a replacement. That proves my thesis, but like, I think that wouldn't really carry the day in the Democratic debate as much.
I think for any race, an important threshold question is, was the race called on election night? Right? That shapes narratives quite a bit. People who aren't nerds tend not to pay much attention to the margin as long as it seems unambiguous and clear.
I mean, look, people wanted to tell this oversupplied narrative of, like, the progressives, like, winning everywhere in the primary, and that wasn't entirely true. You know, Franchesca Hong was ultimately rejected in Wisconsin. I don't know that Jasmine Crockett actually is more progressive than Taller Rico. She's more of a partisan bomb-thrower, and she lost that race. Yeah.
She wasn't a DSA person. That was not her profile in politics.
But like, I think Democratic voters have exercised a fair amount of kind of rational discretion, right? I mean, there's nothing that says that voters have to or necessarily should maximize everything based on electability, right? You're allowed to say, well, I'm trying to calculate some, like, expected value score of what is the chance that this candidate wins times how much good will they do for the issues that I care about, right? Although, I will tell you one thing.
The amount of Democratic infighting really dropped off after these races in Maine, Michigan, Wisconsin. Primaries just create a lot of, I mean, and it's going to go up again when the 2028 primary...
