The Polls Look Fantastic for Democrats. Should We Believe Them?

Still Counting

The hosts analyze recent midterm polling that suggests a significant Democratic advantage heading into the election. They evaluate the relia

Key takeaways

• Recent polling shows Democrats leading on the generic ballot by as much as 10-12 points in some models.
• General election polls are historically more accurate than primaries because they rely on stronger partisan baselines and predictable turnout.

Main topics

• Midterm polling accuracy and methodology
• Economic impact on voter behavior

Notable quotes

Every gas sign in America is an ad for Democrats.

Conclusion

While the polls show a strong Democratic outlook, the true test lies i

Transcript preview

Speaker 3 (0:00) And in that election, the polls were pretty good. And like underestimated, for example, Beto O'Rourke in Texas. You Speaker 1 (0:09) always say his name with a little like je ne sais quoi. What do you mean? Can you say Tallarico? Speaker 2 (0:16) How would you say Tallarico? James Tallarico? I don't know. Speaker 2 (0:24) Welcome to Still Counting. I'm Galen Drake. I'm Claire Malone. Speaker 3 (0:28) I'm Nate Silver. Speaker 2 (0:29) On today's show, the closer we get to Election Day, the better the polling gets for Democrats. But are the polls to be believed? We're Speaker 1 (0:37) going Speaker 2 (0:37) to dig into the latest polling and ask what errors we might expect. Then, say we wake up on November 4th and Republicans have kept Democratic Senate candidates at bay everywhere but North Carolina and control of the House is still unknown. How would we explain that scenario? Okay, now say we wake up on November 4th and Democrats have swept the Senate battlegrounds and even won in Kansas. How would we make sense of that? We're going to get into the bull and bear cases for Democrats this November. Speaker 1 (1:09) I love it. It's funny, I used to babysit for this family. And when the kid was like three and people would ask him what his dad did, he said he worked with bears and bulls. Speaker 2 (1:17) because he was like Speaker 1 (1:18) some Speaker 2 (1:18) kind Speaker 1 (1:18) of banker. It was very funny. Anyway, that's what I think. Or it could Speaker 3 (1:21) be a Chicago sports writer. That's true. There you go. Good Speaker 1 (1:25) Chicago plug. Okay. But one thing that we want to do is just sort of level set a little bit and talk about the current state of the midterms, what's been happening in all these Senate races this week. What's the big news? So I'm just going to run down a few quick little items that have kind of caught our eyes. First of all, Kamala Harris. Heard of her? Wait, Kamala Speaker 2 (1:47) who? That's right. Speaker 1 (1:49) Campaigned with Abdul El-Sayed in Michigan, which is notable because not a lot of other Senate candidates want her to Speaker 2 (1:56) show up at their Speaker 1 (1:57) races. And in fact, some have preemptively disinvited her. In Maine, there is a ProPublica piece that came out. Speaker 3 (2:04) Publica or Publica? Speaker 1 (2:06) Publica, as my years of Latin class would tell me. Speaker 3 (2:10) I guess Speaker 1 (2:10) you're right. It would be Publica, like Publius Maxus. Anyway. ProPublica reported that there was a stalled federal investigation that had been circling Maine Senator Susan Collins. And now that's all the news in Maine is whether or not Susan Collins' broader circle might have been actually doing some sort of shady financial things. Speaker 2 (2:34) Diesel Speaker 1 (2:35) prices hit an all-time high of $6.51 per gallon this week. That's not a Senate candidate, but that's a factor. Speaker 3 (2:42) That matters more than the first two. I mean, the Susan Collins things could matter, but gas prices are causing a lot of the Republican problems, I think. Speaker 1 (2:51) And they're making it into Senate ads and things like that, like the Speaker 3 (2:55) actual... Gas prices are their own ad, right? You drive down the highway and you literally see an ad saying shit's expensive. That's a bad ad for Republicans. Speaker 1 (3:04) It is. Well, just one more item on this list of what's been happening. Louisiana Senator John Kennedy campaigned in Iowa, Nebraska, and Kansas, and he told reporters that those states should be, quote, fat guy layups for Republicans. Sorry, Speaker 2 (3:18) what's a fat guy layup? Isn't that hard? Isn't a fat guy layup, like, not easy? Speaker 1 (3:22) I think it's when you stand under the board and don't have to move. And he said, if you told me six months ago that I was going to have to be spending my time in those three states, I wouldn't have believed you. You got to give him points for quotability and just being Speaker 2 (3:36) honest. Yeah. Speaker 1 (3:38) We're Speaker 2 (3:39) going Speaker 1 (3:39) to talk more about the Republicans' midterm struggles later in the episode. But now I kind of want you both to give me what's your 30-second pitch for the state of the midterms? Galen, Speaker 2 (3:48) go. Okay. It is... post-Labor Day, obviously. But the way that that has manifested for Democrats is it's median voter time. You see Mary Paltola in Alaska being like, we love oil, drill, baby, drill. We are going to get the price of diesel down. Like, she is talking about opening up wildlife reserves for drilling in Alaska. And then you see James Tallarico, for example, in Texas being like, there are two genders. Let me explain why Charlie Kirk agrees with me that God is non-binary and so does the Pope. He also came out and said that there should not be gender-affirming surgery for minors in the state, and that trans girls should not participate in girls' sports. So pivoting to the median voter pretty aggressively in those two places, and really across the country more broadly, trying to play as aggressively as they can to get across the line in these harder states. Then for Republicans, the natural thing for them to do is to run away from Trump, right? And you've seen it a little bit, you know, in South Florida, you've seen a little bit from Susan Collins on the issue of... you know, Canadian tariffs and the like. But it's just a much more complicated relationship on the Republican side. Like, each candidate on the Democratic side can choose their own adventure, right? If Abdul El-Sayed wants to align himself with Kamala Harris, he probably wouldn't advise it. But that's his choice. On the Republican side, Trump has been hugging the Republican Party close for so long, it's hard to break loose at this point. Speaker 1 (5:08) Okay. Speaker 2 (5:09) That's good. Speaker 1 (5:10) I will note that you are a Paltola head. You love... It's a fascinating family freedom. Yes. You think it's... Have you been to Speaker 2 (5:17) Alaska? No. I haven't either. We should do an Alaska cricket show. Live show in Alaska. Anchorage, here we come. Speaker 1 (5:23) Yeah, I like it. Okay, Nate, same thing. Let's say a minute and 30 seconds state of the midterms right now. Speaker 3 (5:30) Yeah, I mean, Galen got like at least two minutes. Speaker 1 (5:32) Yeah, that's fine. Everything's inflation here. Speaker 3 (5:36) I'm going to be a little bit more macro zoomed out. Recent polling suggests that possibly... The bottom is falling out from Republicans and Trump. Democrats are ahead now by almost nine points on the likely voter version of the generic ballot, meaning when you account for enthusiasm, differentials, you're routinely seeing some of these polls print numbers like plus 10 or plus 12. That would be like for the biggest shock since 2010. We can go back and define kind of what the biggest shifts have been in recent midterms. You know, we had a poll in Iowa this week where Josh Turek was up by eight points. We have polling in Kansas and Nebraska showing, if you believe the polls, you probably should take them with a little bit of a grain of salt, showing those states as near toss-ups. It's hard to find any polls that don't show James Taylor go ahead. And Trump's approval is as low as it's ever been, at least in the second term. Speaker 1 (6:36) Yeah. I actually just want to give some of the latest in Senate race polling because we've gotten... a lot of stuff in the last week. So let me just run it down. Nate already mentioned this, but in Iowa, Democrat Josh Turek is up eight points on Republican Ashley Hinson in a Marist poll. In Texas, James Talrico is up by six points, according to Marist. In Michigan, Abdul El-Sayed is up seven points against Mike Rogers in a Suffolk poll. In Kansas, Democrat Adam Hamilton is leading incumbent Roger Marshall by two points in a recent poll by Emerson. And that's just to name a couple of- Speaker 3 (7:08) It's a little, it's- It's a slightly cherry-picked list. Oh, 100%. But not that cherry-picked. Speaker 1 (7:13) Yeah, Speaker 2 (7:14) I Speaker 1 (7:14) mean, we should say, right, for people who are not familiar with our ethos, you know, you're supposed to take the, you know, you're supposed to look at the polls in totality, not just take individual ones. But we're kind of gearing up to a point of, there's a lot of recent polling. How should people sort of be smartly thinking through this? No, Speaker 3 (7:32) but usually you see, one funny thing about this cycle is that usually you see, They're spammy Republican pollsters, just like they're spammy Democratic pollsters. This year, the kind of spammy Republican pollsters are kind of... sitting out. Like Rasmussen Reports actually has pretty good results for Democrats, for example. Republican campaigns are not releasing a lot of internal polls. You hear anecdotal accounts of their internal polls suggesting a very bad environment too. So it's oddly, the polls are oddly in consensus, right? It's just kind of a question of like, do you believe some of these bigger numbers for Democrats? Speaker 1 (8:10) Okay, that's a good setup, Nate, because I want to say. Before we jump into the actual politics of everything, we want to take a step back and look at the polls themselves and how believable they actually are. So, Galen and Nate, you guys are kind of in the thick of day-to-day poll-y stuff, data-y stuff. You know, in the past five or six years, I've done a lot more sort of media reporting and conventional politics reporting. I'm now sort of in the thick of reporting on, like, American medical skepticism and the Maha movement. I kind of wanted to use today to also just like, you know, ask you guys questions about broadly polling and data and the state of it. So... To just track back to the primaries, there were some very notable polling errors in Wisconsin's governor's race and Michigan Senate primary. They were both off by double digits. Are the general election polls generally higher quality? And I think also we should tell people, like, what do we mean when we say a higher quality poll? Speaker 3 (9:08) They're generally more accurate. Why are they more accurate? Because you have much stronger priors or baselines when it comes to general election voting, right? So first of all, maybe only 15 % of the votes are in play, so it's much less fuck-up-able, whereas in the primary electorate, every contest is kind of sui generis, and you can fuck up that entire 100 % of the electorate, right? Also, in general elections, there's a lot of persistence based on partisanship. So if you shift the overall climate from R plus 2, which is what it was in 2024, to D plus 8, right? A relatively uniform 10-point swing would give you a fairly decent guess about how an individual state or district would poll. That tends to bound. Like, none of this is pure polling anymore, right? It's not people are randomly dialing numbers on the phone and everybody answers. It's all modeled. And to have more structure, better assumptions, like basically the polls are like little mini models. That's Speaker 2 (10:10) the Speaker 3 (10:10) way to think about it. Speaker 2 (10:11) I think we can be quite specific about this, which is when folks look at what happened in Michigan during the primary, we're talking about a statewide primary for which, historically speaking, the average error in those polls is about 12 percentage points. Which is crazy. Which is... It's Speaker 3 (10:29) not Speaker 2 (10:29) the margin. That's Speaker 3 (10:30) not Speaker 2 (10:30) the average. Speaker 3 (10:30) That's the average. Now, Speaker 2 (10:32) if you cast forward to statewide general election polling, historically, the average error... is more like four to five percentage points. So just like, even if you take away the specifics of what's going on this cycle, and I should also say that's statewide races in, you know, in the general election. If you want to talk specifically about the popular vote nationally in a midterm, the average error is three points. So those average errors are significantly smaller than what we see in primaries. The reason for that, Nate laid it out a little bit, is because in a place like Michigan, you have no fucking clue who's even going to show up in that election, right? Because independents and Republicans can vote in a very competitive Democratic primary if they want to, especially when there's no competition on the Republican side whatsoever. And so trying to model turnout in that environment is pretty hard. Now, we are currently in a high turnout environment. We can expect that probably over 50 % of eligible voters are going to turn out in this midterm. And when that happens, everything gets compressed, as Nate said, closer to the baseline. Now, we still don't know, like, are people who are really upset with Trump gonna, who are maybe more usual Republican voters, but are really upset with the current circumstances, will they turn out? Those are tricky decisions. We have to talk about Latino voters who may be regular voters and that they turn out in every election, but they've been switching a lot back and forth because they don't have. like a close affiliation with one party or the other, there are still plenty of places for the polls to fuck up in the general election. But just historically speaking, we're not talking nearly about the bans that we were talking about in the primaries. Speaker 3 (12:09) Let me give you a pedantic correction, Galen, because it is kind of important. Probably won't have 50 % of the voting eligible population show up. That has only happened recently in 2018. Speaker 3 (12:21) I think. Ooh. Speaker 2 (12:22) Wait, was it not 2018? Speaker 3 (12:24) Oh, excuse me. I just misspoke. I Speaker 2 (12:27) don't know where Speaker 3 (12:27) the Speaker 2 (12:27) fuck. Okay. Keep that Speaker 1 (12:28) in. Speaker 2 (12:28) Keep that in. Keep that in. I'm incorrect. Let the record show. Let the record show. Speaker 3 (12:33) Only, I somehow zoomed back 20 years. Only in 2018 did, recently, did 50 % of the eligible population turn out in a midterm. The Speaker 1 (12:43) first, and we should say, the first midterms where Trump was president, right? So the first kind of big. I think, nonetheless, you Speaker 2 (12:50) might be right, Nate, because I would imagine that Democratic turnout is similar to 2018, but I don't know if Republican turnout, you know, it's really tricky to try to guess at this stuff. But Speaker 3 (12:59) that's kind of one of the tail risks for Republicans, I guess we're preempting our next segment partly, is that voters just like, I don't give a fuck. I voted for Trump and Republicans so many times, right? They seem to be out of ideas. Gas is really expensive. The Democrat in my district is a wacko, but I don't give a fuck. I'm just going to watch, like, you know, football, whatever that night instead of voting. Speaker 1 (13:25) Still Counting is brought to you by Incogni. Have you ever weirded out by how much of your internet presence and data is available? Speaker 3 (13:33) Yeah, it's kind of crazy. Everything you need to know about me is in the cloud somewhere. That seems probably like a stupid risk for me to take, Claire, isn't it? Speaker 1 (13:41) Well, let's talk about something that doesn't get enough attention. Corporations profiting off your personal data without your consent. Speaker 2 (13:47) Boo. Bad. Bad use of consent. Speaker 1 (13:50) That's right. Data brokers, companies you've never heard of, are quietly collecting your home address, phone number, even your family member's names, and selling that information to anyone willing to pay. No opt-in, no transparency, just pure surveillance capitalism. And that exposure is exactly what scammers and bad actors are using to target people for identity theft and fraud. Speaker 3 (14:10) I want to sell my own data and profit from it. Speaker 1 (14:12) That's surveillance socialism, Nate. Okay. That's where Incogni comes in. They act as your personal privacy advocate, filing legal removal requests and getting your information taken down across more than 420 plus harmful data broker sites. Automatically and on a recurring basis, since brokers try to relist your data every 30 to 60 days. Those Speaker 2 (14:33) persistent scoundrels. I mean, the devil works hard, but data brokers work harder. Speaker 1 (14:38) That's right. Incogni is independently verified by Deloitte. You can track everything from their iOS or Android app. And if you're on the unlimited plan. The exposure scanner even hunts down your info on news sites, blogs, and forums with real human teams handling those manual takedowns for you. Speaker 3 (14:54) Didn't Deloitte used to be Deloitte and Touche? What happened to Touche? Where'd Speaker 2 (14:59) Touche go? Touche is still at the club. They left her behind. Speaker 1 (15:02) Oh, Touche. And the best part, Incogni Speaker 2 (15:07) offers Speaker 1 (15:07) a full 30-day money-back guarantee. Reclaiming your privacy shouldn't be a luxury. They can't harm you if they can't find you. Use code COUNTING at incogni.com and get 60 % off an annual plan. That's incogni.com slash counting. I-N-C-O-G-N-I dot com slash counting. Code COUNTING. Speaker 1 (15:33) Can I ask a question? So just to go back to the idea of what do we mean when we say higher quality polling? And when we're taking, you know, like, for instance, when we're looking at primary polls and you're saying you're basically building models, in some ways, I guess that means, right, that the person has to take into account the news on the ground and see who is most affected by that. Speaker 2 (15:52) What Speaker 1 (15:52) is it? What Speaker 2 (15:53) are Speaker 1 (15:53) the what makes a good in this era where fewer people pick up the phone, where there are more Internet polls? What makes a person a good pollster? Speaker 3 (16:02) They have to have some sense for how reliably they're reaching different segments of the electorate. Who are they oversampling? Who are they undersampling? Like, can I