Nate Silver on what the data says about the midterm elections

Hello, still counting fans. This is Tommy Vitor from Podsave America. We are less than two weeks out from the midterms. So I decided to January 6th this channel and force one of your hosts, Nate Silver, to let me pepper him with questions about the latest polling and how things look. Nate, great to see you. You too, Tommy. Listen, if you're not subscribed to Still Counting on YouTube, please subscribe. We have people literally making up fake polls out there and spreading it through the internet and when you subscribe to Still Counting, you help them get quality information into the YouTube ecosystem. And so just please do it. Also, it's free. You should have to pay for this. This is good stuff. Nate, should we let Flip her out of her tank? Flip her. I guess this is still kind of count. People are aware that Flipers name her house. Oh, wow. We got Reald Off and Reald Off and Sound Effects. Well, I went to YouTube, but I might have found a fake one that sounded better to me than the real one. It's like when they use fake food and commercial and shit, but we love Flip her here. What's that? What were you saying about Flip her? No, look, it's been a big year for bottlenose dolphins. Actually, I don't know what else they've done apart from Flipers, but like in general, you know, always a good year for bottlenose. The Miami dolphins are not good, but no, no, they got rid of McDaniel, so maybe that'll help them. All right, Nate, let's talk about the Senate. What is Flip her telling you big picture about the Democrats' chances of taking back the Senate? Help my anxiety here. Well, I want to give people anxiety in some sense, because we're not in the type of election where you can say with much confidence what's going to happen in the Senate until November 4th or maybe November 5th or 6th, depending on how long Alaska and other states take into account. But we can see Democrats are favored. They have been at around 75% in Flip her for a couple of weeks now. Other models in prediction markets and things like that are kind of catching up to that number. If you look at Polymarket Calchi, it's around 70%. Or so, the way I kind of think about it is Democrats kind of have two ways to win the Senate, right? First of all, if just the overall map is really blue, like you had New York Times, Sienna University poll this week that had a D+1 generic ballot, right? The average is like D+9. Anything at like 9.9.5, 10, and certainly above, there's just too much pressure even in these red states for Republicans to probably hold on to all these races, right? The number of people should keep in mind is, what if there's an 11 or 12 point swing? In 2024, President Trump won Ohio, Texas, Alaska, Iowa, Florida. All these states buy around 11 or 12 points, right? And that's when he won the national popular vote by two points. If we're in a D+10, that's a 12 point swing. And so even without considering regional stuff or candidates, those states become toss-ups, and people can't kind of wrap their head around why the map would look like when it has such a blue tint, I suppose. So that's one way, right? If Democrats beat the points for the generic ballot, then statistically, it's very likely they'd win, we cannot say exactly which combination, right? If it's more like a D+7 or D+8, then you're counting on candidate quality advantages for Democrats, right? You know, personally, I think the two races that if I were Democrat, I would feel safest about, not safest about least anxious about, well, one North Carolina, right? After that, probably Ohio, because there's proof of concept with Sherrod Brown winning several times statewide, including in 2018, in a midterm environment, he was in an incumbent technically. Now people still think he's an incumbent, some of them, but he won in a similar environment in 2018, and Mary Phatola has not become a satir before, but she has also won like statewide races. You know, Talarico in Texas, I mean, I don't think I'm breaking news by saying that like whoever wins Texas is probably whatever, 80 or 90% likely to win the Senate. There's been maybe a point or so of tightening there in the polls, if you look at prediction markets, though, Talarico's odds have gone up. I think some people are reading early voting tea leaves and seeming to think it's stronger in democratic areas and/or among infrequent Democratic voters. I always urge a lot of caution about that, right? But you have some polls showing Talarico up for some showing a tie. Maybe there's a little bit of hurting. No one seems to want to step to it far out of line in Texas, right? But they get to like a a plus two and like, you know, again, Ken Paxson is one of the most scandal-plague candidates in, you know, left of George Santos, or right of George Santos, I guess, right? And like it's kind of, it's not that hard, you know, when Beto O'Rourke lost my two and a half points in 2018 against a candidate who might have been the most charming guy in the world, but Ted Cruz isn't scandal-plague, right? And then you take a guy with Ted Cruz's charm and six million scandals. It's like, it's not that hard to see why Democrats win by a point, instead of lose by two. Yeah, well, I want to ask you a little more about Texas, because Fox News has a couple of polls out this week that, you know, they have main tied at 49.49. They have Texas tied at, I think, 49.49. In both cases, the Republican there is doing a little better than the kind of flipper-pulling average. It is sacrilegious to make too much of a single poll on this channel, and I shall not do that ever. But I heard a few people speculate that maybe Fox has a likely voter screen that might be different from what we're seeing elsewhere. Have you heard this theory, like, is this the kind of thing you're talking about when you talk about hurting? I think I've maybe contributed to this theory, right? It's both Fox and Marist also has polls that are using a likely voter, or have you likely voter screen that is more red than the registered voter version of the polls? Look, that's one of the more subjective elements of polling. One of the differences here is that there are different ways to do likely voter models, right? If you just do a based on who says they're going to vote and who is enthusiastic, you have a huge democratic lead among likely voters, right? I think in the Times poll, it goes from like 11 to like 14 points. If you go by past voting history, then particularly in the more red states, then that fades a bit, right? Maybe it's a plus eight or something. I think maybe some of these pollsters, given that they have some choice, are hurting a little bit toward a tie. We saw a lot of this in 2024 when a remarkable number of polls show to tie, right? Trump or Harris up by exactly one point. This is a little bit different, right? You definitely still do see some polls showing Calarico winning. You don't see very many polls showing Pakistan winning. Maybe I'll buy a point here and there. I kind of trust the process a little bit at the same time. In this era where people are very self-conscious of getting things wrong, particularly being high on Democrats, and they have gotten things wrong, so it's a well-placed anxiety. I sometimes wonder if the polls three weeks before the election are more informative than the polls in this kind of like three-week window where silver, both in other sites, grade polls are based on their performance. There's a lot of scrutiny, you know? But yeah, I don't know, ties. They wish just banned ties, right? You have to show someone up by one point. Yeah, I like that rule. By the way, trust the process if you're from Philly means you're going to shit the bed for a decade, and then you trade for LeBron and Jalen Brown. So I don't know how we apply that here. Let's talk about polling error. I think you or Galen said this week that the average error in Senate polls is five points. So the average difference between the final results and the polling average in Senate races is five points. That's, you know, that's a big difference. You guys have also talked about how there could be a systemic polling error this cycle. It could swing for Democrats. It could swing to Republicans, right? It could be nationwide or not. I want to read for you a take I saw on Twitter today and kind of get your reaction. This is from Dan Rosenheck, who I believe is a data scientist at the Economist. He said, "Every credible forecaster said you can't predict direction of polling error. This may be true nationally, but is demonstrably false for states districts. In any given year, errors are most favorable to Dems in blue states, and Republicans in red ones, most consistent pattern I've ever seen. What's your reaction to that? You know, I've seen a lot of patterns in my day. You know, the sun coming up as I would say, a more consistent pattern than systematic polling error. So models like Flipper are usually not purely polling averages. They blend polls with what we call fundamentals and kind of fundamentals usually start with national polls, the generic ballot, and then the presidential lean of a state, right? So I'm repeating myself a bit, right? But like in states like Texas, the fundamentals don't necessarily show a Paxton lead because you were in a D plus nine or nine and a half environment in Texas at R plus 10 or 11 state, right? And so there's any candidate quality advantage at all for Telereco, right? We actually have a scandal variable in our model that says like a scandal it's probably worth a few points, right? And so like there isn't this gap between polls and fundamentals as you would think. Maybe in Kansas a little bit, even in Kansas though, you know, we have the fundamentals at R plus four. And Adam Hamilton is ahead by a point in the polling average. That's a five point gap. That's high-ish, but within the range of, I mean, looking at all these governance rates, you have like, you know, between El Syed and Joslyn Bettson, you have a five point gap between, you know, Hosa and Telereco, you have like a four or five point gap, right? And so like, you know, a five point gap versus fundamentals to me is within the realm of plausible, you know, I have a little bit more trouble believing South Carolina, for example, but like the other thing too that's funny is like, you know, one of the best polls or Democrats I thought I saw this week was in Nebraska. Pete Ricketts put out a candidate showing, I got the right Ricketts, right, showing himself up three points over Dan Osborne in Nebraska, right? The fact that Republican internal polls, and by the way, the typical bias in internal polls is three or four points, like that's consistent with the toss up in a nonpartisan poll, you know, Nebraska is a race that flippers been skeptical about. The fact that Republican internal polls show their candidate only by three points in Nebraska, which is quite a bit redder than Texas, like that doesn't seem good for Republicans. And their posture has been very defensive, kind of abandoning the house, depending on which articles you read, you know, be concerned about South Carolina. And I mean, again, that's kind of where, you know, beyond Kansas is when I start to get a little bit more skeptical, all the time. It's kind of like quasi-independent thing going for him. But I don't think I've ever seen worse vibes from a party than I have from the GOP just based on the public internal polls, the rumored internal polls, and like where the resources are being spent. Yeah, I mean, I guess the argument I could make for South Carolina is maybe if you're a South Carolina Republican and you're as pissed off as people are in the rest of the country, you're not used to your vote matter mattering in Senate campaigns. You don't have a very exciting candidate, so maybe you just stay home or just like don't think that there's actually a contest. And that's different from the mentality of a Republican in the state like Iowa, for example, where they do feel like there's a contest because it's hotly contested. I still struggle to imagine Democrats winning in South Carolina, but like that's kind of like the fake intellectual case I can make for it. No, look, you need to have a total collapse in a Republican turnout. And maybe you can't, I mean, look, you can even go to things like, wasn't in Kansas where there was like an abortion referendum, right? But like in cases like that where turnout is low and Democrats are really fired up and Republicans aren't, then kind of funny things can happen. I mean, you know, it's not one of the more remembered midterms, but 2014 is a year when you had quite low midterm turnout, especially among irregular Obama voters. And we'd already looked like a fairly bad year for Democrats in the Senate turn into like a nine seat route, Republicans beat their polls by a couple of points because I think in that year, Obama had beaten his polls in 2012 and at the time the concern was polls are low balling Democrats and missing these younger voters and Hispanic voters, et cetera, right? And younger voters don't really turn out that much in the midterms, right? One of the bigger splits is that, you know, notoriously kind of like 20-something and 30-something men in particular shifted bigly to Trump in 2024, right? The biggest turnout differentials are like those guys are not turning out for Roger Marshall or Ken Paxton, probably, right? Trump has this iconic status. I mean, when I remember, we also had like the assassination attempt against Trump in 2024, you know, Trump has his iconic status, you know, I don't want to get into some act to beat like whether he counts as like cool or what, right? But like, but he's got a certain charisma that Ken Paxton doesn't have and like, you know, so now I wonder if like having been burned by missing these low propensity Trump voters in three presidential elections in a row, if if posters are kind of like spotting them credit for for an extra field goal that that doesn't work when when the kicker left the stadium already. Where's the last sportsman to force here? I like it. I like I look I'm here for them. I don't get to do them on my other shows. Um, you have a piece up today about that sort of ticket splitting among these in among voters in these states where there's contested governor and senate races. So as folks probably know, there's states like Texas, Michigan, Iowa have targeted senate races, but also big governor's races happening that maybe don't get as much national attention, but are still very important. There's a delta between the vote share that the two seem to be getting in a lot of states. Like what dynamics are you seeing in ticket splitters? And like when I talk to my friends back in Iowa, uh, who are Democrats, they express anxiety to me about, you know, everyone feels really good about Rob Sandt and the idea that enough Republicans or independence could pull the lever for Rob Sandt, they wonder if Republicans will be willing to pull the lever for two Democrats. Maybe that's a bridge too far, whether it's Josh Turek or anybody else. I don't know. Yeah, you know, and I would say subjectively, um, Iowa is one of the races I'd feel less good about for Democrats in the senate because they are, I think, both candidates are relatively strong and normal, right? So you can't cite this kind of candid quality gap as much like you would in a Ohio or a Texas or in Alaska of that of that quad of four states. But yeah, look, I think, um, I think these splits reinforce kind of common sensical notions about candidate quality, right? The fact that Talleriko is doing better than he knows, I think reflects maybe more Ken Paxton than Greg Abbott, right? Um, everybody thinks that Rob Sandt is a really bright prospect of a gubernatorial candidate. He's explicitly disavowed Biden and Harris and things like that. Um, in Michigan, uh, you know, you kind of have this four point if the polls are right, progressive penalty relative to Jocelyn Benson, that is kind of consistent with what political science models would say. Someone like me who would say, yeah, probably, um, El Sayed, especially someone who's never actually won a general election before. Um, yeah, probably about four points is the effect that you might expect. But if Michigan's a D plus eight or D plus nine state in this environment, then it's a good year to have a four point penalty, right? And then I'll have an opportunity in office if you were to win to, to define himself differently potentially. But I think, I think the all makes sense. I think people should, you know, with that said, in the median state, where there's both a governor in a Senate race, we're projecting like a five point difference between them, which is much smaller than it used to be, right? I mean, you have cases like Vermont where Phil Scott will probably be elected, although there's contradictory polling there. And Democrats will win the at large house seat by whenever 26 gazillion points and things like that, right? New Hampshire seems very likely to split potentially. Um, you know, the one race where, uh, Democrats might come up a little bit short. That's interesting is Florida potentially, you know, jolly might be a relatively good candidate, but like, um, but just the Democratic Party has taken so many Ls there in voter motivation, although even in Florida, the early vote is typically very, very red. It's a state where you have a lot of older voters who traditionally voted earlier by mail and it's more purple this year. So, you know, even in states where I don't expect Democrats to do well, you still may have signs of, of a relative shift. And in Florida, there are quite a few competitive house races that, you know, Rhonda Santis built a risky math that was designed for like a presidential type year where Republicans win Florida these days, right? Not designed for like a R plus three year as much. That would make things a little bit more tenuous for Republicans. Yeah, not designed for a Brent crude at 112 year. Um, oh no. And like, and I mean, this is like maybe the October surprise. I've done, this is not like the odd lots podcast, but I've done a fair amount of research into like, you know, gas prices. And so, um, gas prices, it turns out, I just learned us this morning, right? There are different formulations of gasoline that are used in the summer and the winter, basically. The winter one is a little bit cheaper. And so typically gas prices fade at this time of year. And they had been fading a little bit. However, this renewed shack and oil prices, like the AAA price for gas nationally went up, yesterday for the first time in a couple of weeks, it went up by 20 cents in Michigan. It seems like a bad state for that to happen if your block Rogers diesel has remained very high. So, you know, and there have been nervousness about like an AI bubble over the past couple of weeks too. And so like, you know, if you get into a final 10 days with, um, with renewed economic concerns, then, I mean, Republicans kind of need, right? For the polls to be wrong, they can be, a, they can be, you have been wrong all along, right? The other polls, wake polls can kind of be technically right. In practice wrong is if all the undecideds kind of go one way, right? Especially in red states where probably the undecided voters in Kansas are probably mostly Trump voters, right? If they don't turn out, um, and if you're 10 days out and you're still undecided, you might just not turn out, then especially when there's not a presidential race. Sometimes in presidential years you get dragged over the finish line because like a voter's like, well, I'm going to vote for Trump and I guess I'll vote for Gold Roger Marshall, right? But you're just gonna, you're just gonna go chill at home and go to Applebee's or something instead, right? If you're not inspired by any of those races on the ballot, very much. And so that would be a fear and have if I were Republicans. Yeah, me too. Um, let's push past the midterms for a minute. So I've been told you're a big fan of our Secretary of War, Pete Hegseth. The folks watching probably know that Hegseth is a big reason why the war in Iran is going so well. There's a very funny article this week about his team breaking Eisenhower's globe, which is, you know, like resistance, chum, but it's still just kind of a hilarious metaphor for him smashing up the world. Um, Democrats in Congress are talking quite openly about impeaching Hegseth if they win. Uh, I cringe. I win some a little bit when I hear Democrats talk about impeaching Trump, because I think we've run that play twice now, voters kind of hated it. I don't think that's going to change. But how do you think about a possible impeachment of like a cabinet level official, like someone like Haguezeth? Well, you wonder if Trump might kick him out before they have a chance to impeach him potentially. I mean, look, if, if, if, if not for the decision to go to war in Iran, then I don't think the Senate is lean D, right? I think the Senate is a toss up in that environment, both because, you know, but again, the double whammy of like directly violating a campaign promise on no more stupid wars and on lowering gas prices. And then a little bit with the upstein stuff too, right? It just kind of starts to create a narrative where it's like Trump has broken his promises. And like the thing I think the most important data point that people miss is if you look at his approval rating, it's whatever a 37% you think, well, at least that 37% is ironclad. It's not half of that is weak approval instead of strong approval. You know, the number of true Trump super fans has declined significantly. And, you know, it's been a very linear decline in Trump approval. I think he said a new low this morning, Silver Bolton, like negative 24. Because people kind of go through this like permission structure of going from big Trump fan to lukewarm Trump fan to lukewarm Trump skeptic, right? I don't think the lukewarm Trump skeptics are going to vote in the selection. You know what I mean? I think they're going to sit out. And then, but the strong Trump skeptics are going to vote in droves. And it's like half the country by itself now. Yeah. Yeah. It's a lot of folks. You guys on this week's episode talked a lot about an organization called Wedgwood polls that was putting out just like fake survey data. Wedgwood, they claim that this was some sort of social experiment. My opinion is, I think similar to yours that that is bullshit. I suspect we'll learn there's a financial motive whether it was gambling on Calche or whether it was, you know, crowdfunding money to pay for the polls and just kind of pocketing it. You made the point though in that conversation that if we know about a handful of these fraudsters, there's probably, you know, two, three, four times more that we don't know about. I think a problem, do you think this is? Like, is it really fixable when you look, you guys take pains at deciding who goes into your model, but there's a lot of polling accounts that we all follow. I don't even know who's running them half the time that just share stuff without thinking twice. I know I've done it. So how do you think about this issue? Yeah. Look, I mean, it's kind of common sense. If you see one ant in your kitchen, and you see a second ant, they probably are not exactly two ants in your kitchen, right? So I have complicated thoughts, right? You know, number one, fake pollsters have always been a thing. I helped expose one called strategic vision back in the day in 2009. Research 2000 was another, they prominently pulled for state newspapers for daily coasts and other places and at least substantial parts of their polling were, were, were falsified. You know, what makes it harder now is a couple of things. When back in the day, it was expensive to conduct a poll. And so you could have strong suspicion that someone was fake based on saying, if they said we have done the equivalent of like several hundred thousands of dollars worth of telephone polls, right? And yet they have no sign of a business model of any kind or any kind of public or, you know, profit generating mechanism, right? That becomes implausible. Now you can have these polls where you pay poll fish or something 900 bucks or a thousand bucks to do a, a survey or a quote unquote survey. And like, that's part of the problem, right? I think the barriers to entry became a little bit too low. You know, I think part of the problem too is like, I tried to, well, we make an exception with the New York Times, CNN College Bowl, but like you said at the top, like we try to avoid giving undue attention to just one poll. But, you know, there is vetting in the industry, the New York Times. We rely primarily on the New York Times' data feed. And they've caught a lot of, you know, preemptively screened out a lot of suspicious pollsters, right? This way made it through and the problem of someone's going to like make a fake poll, they may also be smart or charismatic enough to like lie if you call them on the phone or forge other documentation. I mean, it is a pretty difficult problem, right? Like a lot of crime is deterred by criminals being lazy or stupid, right? If criminals are smart, then, then that takes a lot more kind of resources and things like that. Yeah, and as you mentioned, the incentive with like prediction markets where the thing wedge would did is they pulled, I'm using the scare quotes, if you're listening, an audio version. They pulled some races where there weren't a lot of polling. And if you're the first firm, like they were the first firm to poll, quote unquote, "main" with Jackson instead of Graham Platner. That's the kind of poll that might move a prediction market quite a bit. So it's circumstantial evidence, right? But the fact that, you know, they were the first non-partisan poll of Mississippi, for example, one of the first of like the Nebraska governor's race. But that might be cases where, hey, you're the first poll, a good juristic reason. Well, we know anything about this race. Now, we have some poll, it's kind of a shitty poll, wedge would, but like some information is better than none and that my move markets. And so, you know, that that worries me. So I think the industry is like doing a better job than wedge would get some credit for of like policing this stuff. But like if the amount of crime is increasing faster the amount of police, then you can have problems. Yeah, and also look, look, I maybe I wouldn't be a good cop, but I like come to my day-to-day life with the presumption that someone is not entirely fabricating the thing they're telling me. You know, it's like I don't know that like we should certainly do some vetting of polling and, you know, the the outlets we go to, but it's going to be hard if people are just fabricating data anyway. I think it's complicated. Last question I want to ask you is just about prediction markets and politics because I find it to be kind of a fascinating new input because anytime someone has money behind a prediction, right, you assume that they care about the outcome. Maybe they have an information edge, right? There's some reason they're getting involved in this market. But there's a hobbyist. But also, you know, it does seem like a lot of folks are just kind of looking at public polls, vetting off their gut feeling, maybe vetting off some wedge-wood polls. So, like, how are you thinking about the value of that data? I mean, look, I consult for Polymarkets. I should disclose that. I'm an advisor to Polymarkets. I think there's a lot to be said for like putting something on the line, right? Like, look, I'm somebody who, we build sports models in Silver Bulletin and I bet based on the sports model, I think it gives them more credibility if I'm like actually seeing how they're doing in the real market. But like prediction markets are this weird in-between thing where they're kind of in-between the models and some kind of vibes, zeitgeist kind of thing, right? You know, sometimes there are stupid narratives that, like, I don't expect to affect American politics that much, like Elon tweeting about riots in France that I expect not voters not to care about very much at all, right? And they kind of, you know, I think more traditional platforms are libertarian or kind of conservative leaning. But at the same time, they, like, if you look at the races where prediction markets are lower on Democrats than flipper is, it states where you've seen polling errors in the past like Maine and Iowa. And that might be pretty reasonable potentially. At the same time, we have seen this uptick in Democrats' chances at prediction markets in the last few days. You know, I think that might be some combination of entrails from early voting. It might be the notion that, like, the election is just coming to a close now, right? Maybe people thought there'd be some, like, massive October surprise. Well, people are already voting in a lot of these states, right? It's kind of too late for October surprise. And, you know, the kind of known thing that's priced in now is that we have all this pressure from, like, crude oil prices that might send gas, another 10 or 15 cents higher by election day, potentially. So, but, like, but yeah, so far this year, the models have led instead of lagged the prediction markets. Not true, not true every year, but they've been playing catch up, I think, a little bit. Yeah, the only October surprise or attempt at one I've seen is something about the sexual predilections of a now deceased senator. So I don't know that anyone is really delivering. Shocking, shocking revelations about Lindsey Graham for me. The least breaking news I've ever seen in my life. If you've been a sentient, I'm a human being. No, I had a, I was on a flight of the day and the wife, I didn't work. And then I landed and like, uh, and myself, I was having trouble getting a signal. I'm like, Oh, shit, man, maybe it's just the AI agents. Finally, an agent swarm took down the whole internet, right? Like that would, that would have worries me a little bit when I'm disconnected from the internet from more than a couple hours at a time. It does seem like every company in the world was hacked in the last week by some AI agents. So yeah, more fun to come. Well, listen, if you are still listening to this and you've not subscribed to still counting by now, for the love of God, please click the little button, it's free right there. Nate, thanks for doing this. I feel, you know, my anxiety is sort of steady state, but I appreciate you talking me through all the context here. Keep it, keep yourself in a medium state of anxiety. That's my advice for life in general. Good 10 cups of coffee a day. That'll do it. [BLANK_AUDIO]