How Do All These Republican Polls Affect The Model?
56:21 Share

You're a podcast listener, and this is a podcast ad. Reach great listeners like yourself with podcast advertising from Lipson Ads. Choose from hundreds of top podcasts offering host endorsements, or run a reproduced ad like this one across thousands of shows to reach your target audience with Lipson Ads. Go to LipsonAds.com now. That's L-I-B-S-Y-N-Ads.com. Do you want to know what I saw on the way walking over to the studio? No. No.

weird al yankovic's dressing room uh-huh i think people who listen to this podcast might be surprised to know how close we are to where the view record well one we're in the same building and two to get here from our offices we have to walk past all of the views dressing rooms who's the biggest celebrity you've seen

Kerry Washington was here last week. That's a pretty good celebrity, right? Yeah. I was in the green room with Constance Wu a couple weeks ago, and she said she follows FiveThirtyEight. That's good. Yeah. Yeah. Isn't that exciting? Yeah. Have you had any recent celebrity sightings? No, I guess not. You're like, I see myself in the mirror every morning, Galen. What are you even talking about?

Hello and welcome to the FiveThirtyEight Politics Podcast. I'm Dillon Druk. I'm Nate Silver. And this is Model Talk. We haven't done that in about a week. The last time we did it was live in Washington, D.C. It was fun, wasn't it? It was a good live show, yeah. Did it change your opinion at all about Washington, D.C.? I feel like my dislike for D.C. has gotten memified more than...

Disproportionately? Disproportionately, yeah. Honestly, it did make me like DC more. We had a good time. The crowd was really fun. We had a good time afterwards. Pretty good food. Pretty good food, yeah. I mean, honestly, no complaints. Wait, is this the headline? Nate Silver goes to DC, has no complaints.

That's a pretty high bar for me. Yeah. I mean, honestly, DC moving on. I'm not like a complainer. Oh, okay. All right. Yeah, yeah, yeah, yeah. Wait, on what scale are you measuring that? I know some of the biggest complaints. I'm not going to name names. I know some of the biggest complainers on the planet. Are you impugning your friends on this podcast? Maybe. Well, I guess Nate Silver's acquaintances, you know who you are. You're apparently even bigger complainers than he is.

I look forward to whatever emails we get from your friends sticking up for themselves. So, Nate, this is Model Talk. It's less than a week away from the election. So let's talk about the model. Okay. According to the Deluxe forecast, Republicans have a 51% chance of winning. A 53. It's now 53. Yeah, it's going up every hour, man. How does this also happen every time? And we even have that Quinnipiac. Well, first of all,

we're getting a lot of polling data and the model is designed to be a little bit more sensitive at the end here. So it's, you know, by the time you're listening to this, it wouldn't shock me if it's 55% GOP kind of every poll today has been pretty good for Republicans. Not every poll, I take that back, but the,

plurality of polling. Okay, so according to the deluxe forecast, Republicans have a 53% chance of winning control of the Senate. The difference may be between 49, 51, and 53 is not a huge difference. However, this is, for the record, the first time since July. Yesterday was the first time since July that Republicans had an advantage in our Senate forecast.

Now, let me make sure that my numbers are right in the House. Republicans have an 84% chance of winning. We've gotten a lot of questions from listeners. A lot of them are along these lines. And it's honestly a question that I have. So not to be annoying, but what's the point of a forecast if it's just going to be like 50-50?

If every forecast we did was 50-50, I would agree it was pointless. But usually they're not. Usually they're 80-20. You're 75-25. And we get a lot of shit if the 25% happens, right? This happens to be one where- Oh, is this just making sure? We're like, well, if it's 50-50, you can't complain either way. Well, I guess that's fair. So are you happy the percentage is at 50-50 right now?

Secretly, yes. Apparently not so secretly. Not so secretly. Although I'm not sure it's going to be. It wouldn't shock me if, I mean, it's been moving toward the GOP every day. We can debate whether the model, why, you know, is that good that it's kind of moving in the same direction every day? Probably not. But like, it wouldn't shock me if it's 60-40 by election day or something. Uh-huh. Okay. So make the case for the information that a 50-50 Senate model gives us that we wouldn't have otherwise. I mean, 50-50 is like not any different than any other number really, right? Uh.

I think there's a cliche that like, I don't know. Say 60, 40, because it seems like they're conveying information. Right. But like, um,

If you're going to a basketball game and I say the Knicks have a 50% chance of winning, that's like useful information. It often isn't 50%. With the Knicks, it's often lower than 50%. I think, though, that people expect that the forecast becomes more confident as we get closer to an election. The Senate forecast shows that there's an 80% chance that the outcome will fall somewhere between Republicans having 54 seats and Democrats having 53 seats.

So that's not super confident as far as these things go. Is that just evidence of how truly close it is and when it's on a knife's edge, all of the competitive races could go in favor of one party or the other? Yeah, I mean, it's not quite like a presidential year where you have the same candidates in the ballot in every state. But certainly there are many years in the Senate where all the toss-up races go the same way and there are –

three races that we classify as toss-ups per se, and then a couple others that are close enough that if Democrats have a bad night, they could lose Arizona and New Hampshire too, for example. If the GOP has a bad night, then who knows about North Carolina, Wisconsin, Ohio. So there's a wide range of outcomes. I know it's like a cliche saying that, but you can't kind of have it both ways, right? You can't criticize...

oh, polling is so inaccurate, right? And then complain when our model reflects that inaccuracy and that uncertainty in some ways. Well, but this isn't just a result of us and the forecast slash model understanding that polling is uncertain and that there oftentimes is error. It's also the result of the polling averages in all of those states being super close, right? The polling average is huge.

Even in Nevada, Warnock leads by one point in Georgia. Fetterman leads by one point in Pennsylvania. At this point, Arizona is a two-point race. I mean, these races are super, super, super close when you look at the polling averages. Put the forecast aside for a second. And so...

I want to try to understand why those polling averages are so close. Is it one, because this is truly just one of the closest races for the Senate we have ever seen? Or is it two, that institutional pollsters are showing an advantage for Democrats and partisan pollsters are showing an advantage for Republicans and those two things are canceling each other out? Oh no. Oh no. Not this partisan pollster bulls**t. I mean, you had to expect it. Not this bulls**t, dude.

You had to expect it. I've never seen so much hopium smoke done. What's the guy's name? Like Simon Rosenberg or something? Someone on Twitter was like, look, I counted up all of the polls in 538's polling average, and a good portion of them are now Republican parties and pollsters, which that is correct. And I'm sure that people who are looking...

at the details enough to see which polls are actually showing up in our averages, we'll notice that this year in particular, there are a lot more partisan pollsters and that those partisan pollsters tend to be Republican and tend to have better numbers for Republicans. So I understand discouraging people from, you know, we support DARE. We don't want people using any kinds of drugs, including hopium. However, let's explain to folks what's going on with a little generosity here.

So first of all, with a little generosity. Yeah, with a little generosity. It's six days before the election. I don't have time for generosity. I always have time for generosity. So first of all, our model knows a few things. It's pretty smart. It's been around the block. It's been around the block. This model has been stomping the pavement for a long time. One thing we have is what's called a house effects adjustment. So if a poll consistently...

skews toward Democrats or Republicans, then the model can adjust for that. And it would understand that a Trafalgar poll showing members odds up by five points is not the same as if Quinnipiac says that or something, right? If it's an explicitly partisan poll, I mean, it conducted for a Republican candidate or organization, then the House of X adjustments even quicker to kick in. So in principle, the fact that you have a

a lot of partisan polls, the model can, for the most part, account for that. I'd also say that there's no reason that Democrats couldn't put out their own polls. To some extent, it reflects some degree of confidence because these pollsters will get a lot of s*** if they're wrong, right? You know, Trafalgar, etc. If Democrats have a good night, that would really hurt their credibility. So they are taking some risk and Democrats, if they believe all the real race numbers are here, they could publish their own polls potentially and actually take that credibility risk. So,

I mean, that's an interesting view of polling, which is that in some ways it's a messaging game. And I think people do at this point look at, OK, are folks publishing polls that advantage one candidate or the other to try to send a message to the electorate? This is something we've contended with before, which is how does having public polling averages or public polling in general shape polling?

the electorate's understanding of a race? Like, is Trafalgar putting out polls that across the board advantage Republicans more than other polls? Whether that's true or not, does it change how voters are going to behave? I mean, no one has a consistent theory about this, right?

Because half the time the sale you're trying to suppress turnout, right? And then say, oh, you made people complacent. So they didn't vote at all if the numbers were good, right? No one has a consistent theory about this. The empirical literature on this frankly sucks, right? It would be a hard thing to test experimentally anyway. So I don't know. It's just hopium. Although again, other people also, okay, you want to get frank here? I mean, it's a problem for real clear politics, right?

Because they don't make any house effects adjustments, and also they seem to take all the dubious GOP polls and reject all the dubious blue-leaning polls, right? So for them it's an issue, but don't lump us together with them. So I think one way of thinking about this is that polls are part of the media landscape, and all political media comes together to shape voters' perceptions of politics.

politics, policy, government, whatever, in some way. But at the same time, what percentage of Americans are paying attention to individual polls of this midterm environment? Like less than 1%? Well, I'm not sure the assertion would be that individual people are so much as it's trying to influence the media narrative. Again, to me, it's not clear kind of

Which way the media narrative, I mean, yeah. I mean, I think people are like wrapped up in their own bubble. And so they're very concerned about perceptions of the campaign and it wouldn't really affect how an ordinary voter would vote either way, right? I do think as like kind of the so-called gold standard polls decline and you're no longer that confident, oh, there's a right way to do polling, it's expensive, but there's a right way to do it. I do think it introduces more subjectivity. And so maybe some of that subjectivity is like partly reflected in

people's beliefs about what's going to happen more than kind of what the data tells them. And so, you know, it's not crazy to think I know how to put it. Like everything else, polling becomes sort of atomized and more partisan in one way or another. It's not crazy to think that like the polling average in some sense is influenced by media discourse about the race.

Which gets to the likely voter model, right? I mean, in order to create a likely voter model, you do have to make some sort of assumptions about who's going to turn out. And this is very much sort of like the art aspect of the art and science of polling. Like what goes into a likely voter model? I mean, if there are a lot of there are as many different likely voter models as there are polls for some people, just if you say you're likely to vote.

look at you that way. Other people look at past turnout and try and say, has this person voted in the past, right? They may ask questions about like, do you know where your polling place is? Have you thought a lot about the election? We can be kind of like a battery of tests, you know, but polls also find a Republican enthusiasm advantage. And so you would expect that also be typical in a midterm year for the out party, opposition party to do well with voter enthusiasm. So like, I mean, it's all like,

It's all quite normal in some sense. You know, I mean, the country is not normal. The consequence of the election may not be normal, but like this isn't that from a polling standpoint, this seems kind of pretty typical. And by the way, it's like not like there's a huge shift. Right. You know, Democrats were ahead by one and a half, two points on the generic ballot. Now it's GOP by one point three. It's not crazy that like.

2% of the country could switch or even maybe not even switching, but like people who are undecided gravitate back toward their natural partisan allegiance. I mean, we're not talking about some like

Okay, so I want to talk about the difference between the forecasts. We've mostly been focused here on the Deluxe forecast.

In the House, the forecasts have mostly converged, which is what I think we've generally seen in past election cycles. However, in the Senate, they have not. So in our deluxe version, as I said, Republicans have a 53% chance of winning. In the light version, Republicans have a 43% chance of winning that chamber. Why the discrepancy? I mean, if you look purely at the polling with no other assumptions or secret sauce, right, um,

purely at the polling, then it still looks slightly better for Democrats, I guess. I mean, they are a tiny bit ahead in the polling average in Georgia. They are a tiny bit ahead in Pennsylvania. If you take the polls at face value, then Democrats have a shot in Ohio, for example, or Wisconsin, North Carolina. So they have a lot of outs, you would say, in poker terminology. So, you know,

Will they remain that way by the time we get to Tuesday? I mean, all these forecasts have been shifting GOP. I don't know. That might wind up more 50-50 anyway. But like,

If you look at the 538 polling averages and simulate the race, which is what the light model is doing, then the balls come up Democratic slightly more often. So David asks, if the only difference between your classic model and the deluxe model is that the deluxe model incorporates experts' ratings and incorporating those ratings moves the model about 10 points towards the GOP, are the experts' ratings as a whole Republican-biased?

So it's a little complicated. We actually look at how experts rate the races relative to one another. Like we don't actually let the experts influence the overall picture. It's just that they have more bearish ratings for Democrats in a number of key Senate races. So the ones I mentioned before, you know,

Ohio, Wisconsin, North Carolina. I think most raters rate those as tilt or lean Republican, right? Whereas these races where Democrats had a small league, like lead like Pennsylvania, Georgia, they're rated as toss-ups. So just on that basis, they kind of go with the more bearish interpretation of those key races. We don't, again, look at like what their overall view is. We think we are better at figuring out the macro environment of

than the Raiders, frankly. But they are good at adding value and saying, okay, I understand that Tim Ryan is close in the polls against J.D. Vance, but I know Ohio, I know it's just not enough Democratic votes out there. And so, you know, they provide a useful opinion in that respect. I have to say, to some extent, it does feel like people who cover these things professionally, like us to some extent, are putting less emphasis on the polls this year. Like if...

I had amnesia and I had no idea what had happened in 2016 or 2020. And you showed me the polling environment today. I would say, okay, Democrats have the advantage. But it seems like a lot of people are looking at the polling environment and everything else and saying, I don't necessarily trust the polls. I think Republicans have the advantage. Are we in some ways in like a post-polling situation?

media landscape again. Like there was a period of time where folks weren't really empirical about this stuff and weren't paying all that much attention to surveys and sampling and stuff like that. You know, in many ways you made the argument we should do this in a more rigorous way. Let's create these averages. Let's get a forecast. And that was really the focus of, you know, elections media for a decade. It feels like are we in like a new era now where folks feel less attached to that stuff? It's worth remembering that like

the era of vibes like sucked, right? And a lot of media coverage still does suck because it is all about narrative and vibes. And the fact that there's more kind of an empirical infusion into how elections are covered, I think is useful. You know, is this shift in the forecast that's happened since September? I mean, how much of that is just vibes? How much of the polls themselves absorb the vibes and reflect the vibes? I don't know.

But what I come back to, again, is that like this is supposed to be a good environment for Republicans. The country is not doing well in certain respects. Democrats are in charge. And even when the country is doing relatively well, the party in power usually gets its ass kicked at the midterms. Right. So, you know, whether it's kind of returning back to gravity or something else, I mean, it's kind of again, if you say let's not look at polling at all.

you would assume the GOP would have a good night on Tuesday. In fact, probably a better night than they probably will have, right? I mean, if you asked me a year ago, what's the environment going to look like at the midterms, I would say it will look like what we just saw, what we're seeing in the November elections this year, which is Glenn Youngkin, the Republican, winning the governorship in Virginia, a state that voted for Biden by 10 points. Republicans coming close in New Jersey. I mean, I do think that

And that's not what we're seeing, to be clear. We're seeing a more competitive race than that. I do think it's true on some level that the incentives are to predict the GOP doing fairly well. Like, in part because people just are still, like, terrified of 2016. And so, like, you just, I don't think, get yourself in trouble by predicting a Republican movement. Are we still living in 2016? Oh, yeah. People have tremendous, in the news media, people have trauma from that. Yeah. I'm using that term literally. Yeah, yeah, yeah. Yeah.

No, I think we're still in an environment where you take more reputational risk by predicting a Democratic win than a Republican win. I think it's true. But A, I don't know if that affects the pollsters. And B, I just think it's the right forecast, right? That like, yes, I mean, Democrats are overperforming in some of these races, but

with inflation and immigration and crime and just the very robust history of the president's party doing badly in the midterms that... Right. It is really interesting. I have so many questions here and we have a lot of listener questions too, but this is an interesting piece of the conversation, so I'm going to pursue it. It's interesting that when you look at aspects of the polling,

you actually do see a story that is good for Republicans, right? When you ask, what's your number one issue? 50% of the country, it depends on the poll, but say it's economy and inflation. And when you ask people, which party do you think would be better at handling this stuff? It's Republicans by a double digit margin. And so when you think like, okay, what are Americans focused on and who do they think will fix it? The answer is Republicans, right?

But when you ask, you know, OK, well, now think about who you want to vote for. Things change. That's understandable. You know, people vote on more than just the economy, more than just on one issue. But it's not as if the polling is only telling one story. The polling seems to be telling us that, like, there are a couple of things going on in Americans' minds right now. Yeah, people, you know, a lot of the kind of soft Biden disapprovers are still going to vote Democratic in theory, right? I mean, the generic ballot gap. Or not show up.

the generic ballot gap is smaller than the president's approval gap, for example. So yeah, I mean, voters in some ways are weighing a lot of issues and being somewhat sophisticated in that treatment. I mean, there's also this conversation about like, you know, well, how could a voter excuse what Herschel Walker did? And it's like, well, if you just want to have policy enacted, maybe it's pretty rational in some sense. You're not having a popularity contest, right? You're

electing someone to represent your views in the Senate, and as long as you think he's going to reliably do that, then who cares? Or John Fetterman. Right. I was going to say, voters on both parties are making this argument. I mean, one relates to a scandal and one relates to health concerns, but increasingly... And look...

we've talked about this before. Voters do see a difference. They see senators as people who will pull an R lever or a D lever and they see governors as people who have to actually administer the responsibilities of a whole state. However, you know, complex to mundane those responsibilities may be. That's why you have a Democratic governor. Delaware,

Pretty easy to be governor of Delaware, don't you think? I feel like you have to deal with a lot of sort of like overseeing corporate policy. Isn't like every company in America headquartered in Delaware? I mean, it has like three counties. Yeah, I don't know. Honestly, who is the governor of Delaware? Do you know who the governor of Delaware is? Yeah. Oh, no. Jack something? Jack something. Watch, it's a woman.

John Carney. That's pretty close. Jack John. That's pretty close. I thought I was thinking of Jack Markle. Is there anything to say about John Carney? John Charles Carney Jr. is an American former football coach and politician serving as the 74th governor of Delaware.

That seems important. Fair enough. Yeah. Okay. So there's a lot going on in Veritas Minds. There's a lot going on in the polling. We did get more data this past week in terms of how to put that all together.

Because the New York Times conducted four House polls. We've gotten very few House polls this year, sadly. In 2018, the New York Times and Santa College did us a huge favor by conducting a lot of House polls, and they were also quite accurate.

So let's talk about this result, even though I know that we don't love to talk about individual polls. So they polled four districts, Kansas three near Kansas City, where the Democrat was up by 14 points, Pennsylvania eight around Scranton, where the Democrat was up by six points, Nevada one, which is suburban Las Vegas, where it was even and New Mexico's second district, which is in southern New Mexico, where the Democrat was up by two points. How would you characterize those results?

Those are very good results for Democrats, at least relative to the polling consensus. Yeah. And they kind of in their article, they kind of said this is proving that the GOP is going to do well in the midterms. And it's like, do you want me to just read their lead? Yeah. Yeah. Is this revised or it's like the original? Yeah. Well, they they revised the headline, but this is the original lead.

Quote, President Biden is unpopular everywhere. Economic concerns are mounting. Abortion rights are popular, but social issues are more often secondary. A new series of House polls by The New York Times and Siena College across four archetypal swing states offer fresh evidence that Republicans are poised to retake Congress this fall as the party dominated among voters who care most about the economy. Yeah, the headline is fake news. I mean, like, because the polls themselves, there may be plenty of reasons to think

the GOP will take over the House and the Senate? This is more evidence of my theory that we're post-polls. Right, but their poll does not provide evidence of it. Oh, for sure. Their poll contradicts that thesis. No, they're literally contradicting their poll. Their poll is good for Democrats, and then they write a lead that says Democrats are about to lose the House. Misinformation. Elon Musk. Elon Musk. If I were on your misinformation board, I would label that article misinformation. No, but the thing is,

We don't know whether or not the lead is accurate, but it's certainly not reflective of the poll that they published. That's right. Yeah. Those polls were like a 9 out of 10 for Democrats. And their Senate polls were pretty good for Democrats, too. So they have a much more bullish—because this is why it's funny. If you're in the New York Times universe, if the only thing you knew was those New York Times House polls and Senate polls—

Then your conclusion would be that, oh, Democrats are going to have a quite good night. Yeah. They're going to win the Senate for sure. Keep the Senate and then the House fighting chance, maybe slight favorites. Right. That's those polls are consistent with a, you know, blue leaning, if everything environment. Those polls look like what Democrats were hoping for back in August at the peak of all these special elections and so forth, where where the winds seem to be blowing in their favor.

And they also polled four Senate races. I'll cite those here. So Arizona, they found a six-point advantage for the Democrat. Pennsylvania, a six-point advantage for the Democrat. Fetterman, Georgia, three-point advantage for Warnock. Nevada was even between Adam Waxall and Catherine Cortez Masto. What should we make of these polls? Because I know that the folks at the time, Nate Cohn and Sienna College, have

made a concerted effort to try to account for the challenges that polling has faced over the past six years or so. Yeah, they have a lot of, I mean, if you read his newsletter, The Tilt, I think it's called, there's interesting kind of back and forth with himself, right, where reasons why the polls could be biased in either direction. They have found there is still some partisan response bias, meaning that Democrats are more likely to respond to polls. They have ways to adjust for that in theory. Other pollsters maybe don't.

They've also, though, had a discussion about like some pollsters are now waiting to the past vote that they'll say, OK, well, Trump won this state by five points in in 2020. So therefore, we know the compensation for the electorate. And if if our poll says that, oh, actually, people voted for Biden.

in 2020, then we'll adjust that based on past vote, right? And that tends to shift things more Republican, but there are some problems with that. People forget who they voted for. People forget who they voted for. And also if there are new, if turnout composition has changed, right? So it's kind of like a little bit of a cheat code that, you know, if the polls have a

anti-GOP bias as they have in 2020, 2016, then you have this little cheat code that fixes it kind of, but it also risks two wrongs making a right and it risks that you kind of overcompensate in the other direction. So it's, you know, I trust, despite their weird headline, you know, I trust the New York Times Sienna poll to like give a thoughtful result that is relatively independent from like

the hive mind, right? We can go post it. And they did, right? And they did. They published a poll that is quite good for Democrats. I'm sure they're worried a little bit because obviously they're polling in 2020 overestimated Democrats by a significant margin. Yeah. No, if Republicans have a good night. But they did it anyway, which is good polling practice. You know, you get a result, you publish it. Yeah. And we can go through pollster by pollster and...

And I can tell you who I trust to herd and who I don't, right? Yeah. Well, maybe after the election we'll do herding Olympics. Herding Olympics. And we'll go through pollster by pollster once we have the results. But to your point about partisan non-response bias, this is the idea that

There are certain parts of the electorate that are more inclined to vote for, in this circumstance, a Republican that just aren't answering their phones or that aren't responding to pollsters. And it doesn't even just break down by Democrat or Republican. It goes beyond that. It's like...

Republicans, because sometimes registered Republicans vote for Democrats and sometimes registered Democrats vote for Republicans. But this is getting at, you know, whether you're a Democrat or Republican, there's something unique about, say, supporting Trump that makes you less inclined to respond to a pollster. This is what they say this time around.

The response patterns by district and state certainly raise the possibility that there's a similar challenge. Where did Democratic response rates outpace those by Republicans the most? Kansas' third district, where Democrats were nearly 70% likelier to respond than Republicans.

I'm pretty sure, this is Nick Cohn writing, I'm pretty sure that's the largest disparity in partisan non-response we've ever encountered. And this district was also the race where we posted by far our most Democratic-leaning result. The incumbent Democrat, Sharice Davids, led by 14 points, a solid 10 percentage points more than I would have guessed before we fielded the poll. Republicans, meanwhile, were likelier to respond to our surveys in Pennsylvania's 8th and New Mexico's 2nd, two districts where the results were closer to my expectations.

Yeah. I mean, the one critique kind of critique I have here is I'm not sure that he's just not looking at random variants. And because he said, like, in his national polls, they hadn't had this issue as much. But, you know, I mean, as Democrats tend to be more kind of news consuming college educated voters, those are people who are more likely to respond to polls. There's some question about, like, whether you can fix that with demographic weighting. There's some question about whether this applies to the same degree in election without Trump on the ballot. Right. And again, I'm not sure.

Pulcers can also do lots of things that kind of maybe fudge the other way a little bit from explicit hurting to waiting to pass vote share, which is probably fine, but not without its risks. Right. I mean, at a certain point, you can try to wait your way out of things. The best sort of practice, though, is to try to reach everyone. Yeah. But that's, I mean, we're nowhere near that, right? I mean, that's where kind of like

New innovations in terms of texting or email or even snail mail potentially might get us closer. I think pollsters are in the process of just trying to figure out how to solve that problem too. One thing pollsters don't do is

Do the people who are harder to reach, are they different than the people who are easier to reach? That might be a good hint. That's a little technique that people could apply. Wait. The answer is yes. Isn't that what partisan non-response bias is? Yeah, but the people you have to work really hard to reach, maybe you should weight them more because they're representative of people that aren't reached. Tricky. Tricky. Tricky.

Okay, so I want to get to listener questions. Before we do that, though, you have two pieces that are coming out on the website this week. One of them is already published as of the time of this recording, and maybe when you're listening to this, the other one has already published. And it's you talking to your imaginary friends. Imaginary? Which, you know, love to see it. Love to see you owning up to your weird behaviors. Okay.

You talking to your imaginary friends from a Grand Central bar, Grand Central Station bar.

And the first one published was you talking to Nathan Redd, your imaginary Republican friend. He's independent. What is the name of your imaginary Democratic friend? Nathaniel Bleu. Nathaniel Bleu. Okay. How pretentious. True to forum. And you literally write out a fake dialogue in the pieces that you're publishing on the website. I've already read the Republican one. Folks should go check it out. It's Nate, I have to say.

it's bringing back the blogger days and I love to see it. Thank you. It was really good. But no, this is kind of, cause like it's not trying to set these people up as straw men. It's saying, what are the arguments that I find most persuasive for the one we published so far is for a good Republican night. And it's kind of what I was saying before, right? That like, look,

The fundamentals of this race should favor the GOP by quite a bit. The polls increasingly match those. But if you ignore the polling, you would expect the GOP to do pretty well. Also, some more minor issues around the margin where in Pennsylvania, I'm not sure the polls fully reflect the impact of that.

debate with Fetterman. I'm not saying it's a total game changer, but if you go from Fetterman plus one in Pennsylvania to Oz plus one in our polling average, that would have a decently large effect on the overall picture of things. So it's trying to engage with these arguments somewhat seriously. And so what is the argument for Democrats overperforming that is going to publish later this week?

I think the best argument for Democrats overperforming is actually the results of these special elections and ballot referenda of various kinds, because that is hard evidence about how people actually voted. And the results for Democrats in the summer were not just OK, but it looked like 2018 or something where it was a very dealing year over performance for Democrats. So the argument is kind of like.

That is somehow real evidence and everything else is in the fog of war. And maybe the incentives in the fog of war are to paint an overly rosy picture for Republicans. I mean, it's more speculative, right? The whole Nathan Redd case is like, keep it simple. It's pretty obvious what's going to happen now. There isn't this disparity between polls and forecasts. Maybe they'll win big Republicans, maybe they'll win small. But like, you know, keep it simple. Whereas Nathaniel Blue's case is much more about like the psychology of the media and, you know,

hurting and things like that, right? Yeah. I mean, it's also worth saying that although historically the outcome is that the out party picks up a lot of seats in the House,

the Senate doesn't always flip back and forth. I mean, in 2018, Democrats had a really good year and the Senate didn't flip. In 2010, Republicans had a really good year and the Senate didn't flip. So like, it's not necessarily the case because the Senate is more complicated. It's not a sort of the pulse of the nation. It's only a third of seats every midterm and blah, blah, blah, blah. Yeah. I mean, so there's also a case that like, hey, candidates actually do matter and that people have gotten infatuated too much with, you know, because you're right. I

But in the Senate, I mean, in the old days, like a Herschel Walker might would probably would not win. I mean, the old days, Georgia would be more red. So he might win. Right. But in the old days, you suffered more of a penalty for candidate quality. And so, you know, you also had more an advantage for incumbency in races like Nevada. So, you know, people, I think, are a little bit selective with their story of what the kind of fundamentals present themselves as. But, you know, if it weren't for

those special elections, I would kind of totally buy the doom case. But like, there's something about how that reflected some real state of the world. And then, and then have things really shifted that much since

August? That smells like some potential hopium for our Democratic listeners. Well, I mean, but it's trying to engage with... I know, I know. We don't have to rehash it. But that seems a better argument to me than like the, hey, man, just trust the early voting numbers. Don't even, you know... Well, okay. Speaking of early voting, let's get to our listener questions.

You're a podcast listener, and this is a podcast ad. Reach great listeners like yourself with podcast advertising from Lipson Ads. Choose from hundreds of top podcasts offering host endorsements, or run a reproduced ad like this one across thousands of shows to reach your target audience with Lipson Ads. Go to LipsonAds.com now. That's L-I-B-S-Y-N-Ads.com.

You're a podcast listener, and this is a podcast ad. Reach great listeners like yourself with podcast advertising from Lipson Ads. Choose from hundreds of top podcasts offering host endorsements, or run a reproduced ad like this one across thousands of shows to reach your target audience with Lipson Ads. Go to LipsonAds.com now. That's L-I-B-S-Y-N-Ads.com.

Okay, so this is truly, I think, our most popular question when we get down to the final weeks of an election. So let's get it out of the way. Brian asks, is there any way to actually use historical early turnout data to model this election's result as they relate to heavy increases in early voting? Because it's a moving target. I mean, we had different parties over the years have emphasized the early vote to greater or lesser degrees. I mean, that's a problem, right? It's not like some regular number that you can just kind of target toward.

It's different in every state, different for every race. Obviously, in 2020, Democrats put a much more emphasis on mail voting and to some extent voting early, but not necessarily voting early in person because they were voting by mail. And so there's just nothing to like anchor to. Next question. Can you ask 5e what his down to the wire prediction on what the tipping point race will be for the House and or Senate? You don't have to ask 5e because you can go if you don't know this.

You can go to the forecast page, and at the bottom of the forecast page, there is a link that says at the very bottom, keep scrolling. Model outputs. Model outputs, and that will kind of actually tell you each state's tipping point number. We've kind of gone for a more ergonomic design where we don't publish everything automatically.

on the page, but you can still find that in the output files, I should say. Yeah. And throughout the cycle, we've seen that Georgia is the likeliest tipping point state in the Senate, then Nevada, then Pennsylvania. But I think because these races are so close right now, this listener is asking us to make a call ourselves. I think if anything, people maybe underestimate

The chance it could be one of those other states. You know, Arizona is getting fairly close. New Hampshire is getting fairly close. I mean, you know, there are things people don't see coming, right? Maybe North Carolina saves Democrats. Maybe J.D. Vance is a flawed enough candidate that he loses somehow. I tend not to think that's likely, but you never know. I mean, you know, you have a bunch of... So what is it, 5E? 5E thinks it's going to be...

Pennsylvania. Five things it's going to be Pennsylvania. All right. Once again, Pennsylvania. Don't keep us waiting until Saturday, Pennsylvania. Just tell us one way or another. Oh, no. And the House, there's like too many districts, but it's going to be North Carolina. Maybe it'll be North Carolina's 13th. Maybe it'll be Iowa's third. Maybe it'll be New York's 22nd. Who knows? We'll just have to wait and see. Okay. Blair asks, saw the Libertarian candidate in Arizona dropped out and endorsed Blake Masters. With not much time left for new polling, how does the model factor this in? It doesn't.

All right. There you go. How do voters usually respond to this? Like, I think this gets covered in a way that just expects voters will say, oh, that's the candidate I was going to vote for. Now they're telling me to vote for this candidate. So I'm going to vote for that candidate. Like, is the impulse of a third party voter to say, I just really don't like these two parties and I would rather vote for a third party? Or is it to say, like, because no one's voting for the third party thinking that that candidate is going to win in the first place?

Yeah, I mean, if you're voting for Libertarian candidates because you kind of probably rejected Blake Masters since Libertarian voters are, well, it depends how you describe it, but kind of conservative-leaning in general. And so their endorsement may not matter that much. But, you know, hey, I mean, it's a close race. And so if it shifts half a point toward Masters, then that's still significant. Raul asks, what was the most surprising race rating in the model and what was the most predictable one? I mean, the fact that Ohio stayed so close in the Senate for—

for so long, I think is interesting, right? The most predictable, the fact that Georgia is coming down to another one point race, I think feels like very par for the course.

On this topic, Kenneth asks a follow-up question here. Specific question about the deluxe models forecast of Texas 15. Every single forecaster has it at least lean R with around half having it likely R. The district went for Trump by three points and the DCCC has completely abandoned the race. Why on earth is it still 53% for Vallejo? Vallejo, Michelle Vallejo is the Democrat running in Texas 15. I mean, is there polling in that race? There has been extremely little polling in that race.

I mean, first of all, it's not just looking at 2020. It's looking at historically, it's been a relatively blue district. It's looking at fundraising. I mean, I don't know. I mean, you know, over the course of 435 races, the model has a pretty good formula for balancing out these different data points. Could you go in and look at

two or three or four races and say, I disagree. Sure, yeah, but we're trying to deal with a whole bunch of races at once. So I think one thing to say here is that our partisan lean metrics, which is part of the fundamentals that the model considers, rely on past elections. And when there is a super fast sort of like realignment or coalitional shift, it's sometimes hard for that to get priced into the partisan lean data quickly. Well, it's not because empirically,

There's some reversion to the mean a lot of the time. There's a decent amount of version of the mean. But I think polls suggest that in South Florida and South Texas, there hasn't been a reversion to the mean amongst Latino voters. We don't know yet. We haven't seen the results. All right. I mean, Trump's not on the ballot, notably, right? I would be surprised if there's a reversion to the mean. I haven't seen any polls that oversample Latino voters, especially. If you go through case by case and trying to test, oh, here's there's going to be reversion here. There's not. You won't. I don't think you're going to do better than like the algorithm.

Okay. All right. I'll, you know, I'll take it. That's why we have these conversations, but I still don't think it's going to revert to the mean in South Texas and South Florida, but you know, we're going to meet back here. We're going to have a model talk post election and one of us is going to be right. Okay.

Okay, so next question. In 2010, you wrote that the rule that, quote, undecided to break towards challengers was false. We have a listener digging back into your 2010 archives. Appreciate it. However, you noted that it had been true back in the 1980s. Has that changed at all since 2010? Do undecided voters break in any empirically predictable pattern?

As far as I can tell, no. I mean, I think it used to be people were reluctant to say they'd vote for somebody who they don't know. And now in a more partisan era...

you know galen druke democrat galen duke republican i don't care about galen druke i just care about the party label by your name you don't care about me i mean very early i mean we turn the model on in the summer basically right if you're looking like a year in advance there might be something to that right um but this is more myth than fact look at the margin and not the i mean you know looking at the share of the vote can be informative too i mean certainly like um

You know, in some of these races, it's less that Democrats have lost ground than the Republicans have gained it, which is what typically happens is fewer people are undecided the closer you get to the election. So it's kind of relevant context. But still, like to make a forecast, the margin is the best number.

Alicia asks, not necessarily a model question, but what should we expect in terms of timing for states being able to report results, networks being able to project results this time around, and which states have updated vote counting procedures regarding mail-in ballots? So I should say, we are going to address this a little more on our Friday podcast with Nathaniel Rakich, but what are you preparing for? I mean, I think you should be prepared that if the election is close, then...

We won't know on Tuesday, right? I'll flip that. If we know on Tuesday, then one side's won pretty clearly, at least in the Senate, right? I guess you could have a case where like Democrats are beating their forecasts and you know the Senate, but the House takes a while to resolve. So one thing we know for sure is that if in Georgia no one is at 50%, then Georgia goes to a runoff. So that's contingency number one.

Number two is Pennsylvania typically is pretty slow to count its votes or at least as issue where they count the mail votes, which tend to be more Democratic afterwards. So that's an issue.

You know, Arizona is a state where the mail vote takes a while to get in sometimes. So are Nevada. So like there's no reason why you're giving me 2020 PTSD right now. Well, there's no reason to. I mean, which states count? Wisconsin counts fast. It didn't 2020. Florida counts very fast. But no, like several of New Hampshire counts fast. But yeah, there are three to four Senate races where where you would not necessarily expect to defend a result on election night. And those happen to be some of the three or four most important Senate races.

Next question. Past election results are useful for establishing a baseline about the political environment, a.k.a. partisan lean. We were just talking about this. How do you adjust for districts that have been gerrymandered? Do you even use a baseline metric? This is an easy question to answer. Yeah. So we can... Are gerrymandered? Yeah. I mean, we still... We just reconstruct new districts using past vote data. That's right. Yeah. And we account for... I mean, there are a whole bunch of procedures in the model for like

How much credit does a candidate get for incumbency if their district has changed a lot? We actually look at how much overlap there is between their current and former district. So if it's like 40% of their former district, they get 40% of the incumbency advantage? I'm not sure it's quite that linear, but that's a general idea. Yeah. Yes. So to answer your question, gerrymandering is taken into account because we just –

take these new districts and say, if they had voted in the 2020 election, how would this have come out? Which is how we have the number for Texas's 15th district. Texas's 15th district did not exist in 2020, but had it existed, Trump would have won that district by three points.

Okay, final serious question, and then we got some funny questions. This is a question we've answered before, but it keeps coming up, so let's answer it again. Alexander asks, is Evan McMullin treated like a Democrat in the model? What would be the probability of a 50 Republican, 49 Democrat, and a McMullin Senate? I mean, I don't think we have McMullin's chances rated particularly highly. That's a race where the fundamentals override polling. That's in some cases been pretty close. Yeah.

So the model gives McMullen to Democrats 75% of the time. He has said he won't caucus with either party. You know, is that to be believed or not? I'm not sure. I'm not even sure how all the procedures work. Like you automatically receive committee assignments apparently in the Senate, whether or not you're caucusing or not. I'm not a senatorial procedure expert. I do think that he's been very openly unhappy with the kind of Trumpified direction of the GOP and

would be a Democratic vote, at least on things like protecting voting rights, nominating a reasonable cabinet picks, confirming them, right? You know, if there's a dispute in 2024, I don't think he would play ball with election deniers and stuff like that. So I think he would kind of be like a Democratic vote on most of the important issues. You know, what he does if it's 50-49,

Let's say Democrats lose a seat in Nevada or something. So it's 50-49 and he could caucus Democrats and then have Chuck Schumer be majority leader. I mean, I don't know what happens there. I don't know if he also negotiates and says, you know, make me majority leader or someone more moderate. He probably won't win. What do we have his chances at?

His chances? 4%. 4%? Yeah. Not great. What's the light model say there? I wonder how, yeah. Light 10%. So even a light forecast, it's a little bit of a long shot. All right, next question. It's been over a month and we still haven't heard Nate's take on this study, which has everything I expect from FiveThirtyEight. Data, maps, restaurants, cultural divides, regional divides. This was made for you. What do you think? Okay.

Have you heard of the study? It's called Places in America with the Most Chain Restaurants. Oh, no, I haven't seen it. You haven't seen it? Wait, wait, wait, don't look. It's from the Washington Post. Can you guess which state has the largest share of chain restaurants? Arizona? No, that's not even in the top 10, interestingly enough. Where is Arizona? Ohio? Upper Midwest? No, no. In fact, there is a famous chain restaurant

named after this state kentucky yeah yes so kentucky is number the share of chain restaurants is 46.3 percent of kentucky's restaurants are chain restaurants you want to guess the state with the least chain restaurants uh maine you are close that is one two three fourth least

Vermont. Yes, you're right. Okay, so Vermont is the least, then Alaska, then Hawaii, then Maine, then New York, then DC, then Montana. And I'll give you the top for the most. So it was Kentucky had the most chain restaurants, then West Virginia, then Alabama, then Indiana, then Kansas, then Arkansas. Speaking of Indiana, do you remember when Mike Pence came to visit New York and

in 2016 i think it was or maybe 2017 and went to eat at the i think it was the olive garden in time square and was like just hear the olive garden in time square eating at the best restaurant in new york city my tons to that yeah again i would be honestly like good trolling behavior i would like trolling behavior i would eat at the good restaurants and like i like all kinds of food so that would make me a good they'll make you a good candidate

Yeah, I would like actually like, I mean, I'd probably like waste too much resources on like having like my best people find like the best barbecue and stuff in random towns instead of like trying to like make sure people showed up at the rally. Don't you feel like you should eat at chain restaurants to relate to voters in Kentucky, West Virginia, Alabama, Indiana, Kansas, Arkansas, Mississippi, Ohio, Oklahoma, Tennessee? I'm not anti-chain. Oh, okay. So you were, okay. What's your favorite chain? Taco Bell. How often do you eat from Taco Bell? Is it, how often do you eat from Taco Bell?

A couple times a month. A couple times a month? Because I live right near a Taco Bell. It's open. It's actually not as easy as it used to be to find restaurants that are open late in New York. Okay, I agree. In fact, I make the claim that COVID isn't truly over until the CVS across the street from my house is actually open for 24 hours a day, like its sign says. Yeah, so I don't live in a neighborhood with a good...

Late night Mexican restaurant. Wait, don't you have tacos in your neighborhood? No, much too far. And it's close. It's kind of closes early anyway. Right. We need more 4 a.m. Taquerias in New York. Yeah. Yeah. Honestly, actually, I will single hand platform to run off, spend enough money there. Right. To like, yeah, make it a viable business.

Okay. Next question is, can you get through a Model Talk episode without saying salient? Wow. I don't think I've said salient. Okay, Sam. I don't think I've said salient once. I mean, now I do. It's a trick. It's a f***ing trick. I actually, I have no idea. We hadn't said salient, but now because he brings it up, I had to say it. You'll have to go back and listen to this episode and tell us if we have said salient or not, because I frankly can't remember.

Next question. This election has been something crazy between the candidates and the polls. What are you looking forward to after the election is over? Going to Florida. The World Poker Tour Hydroc Seminole Poker Tournament. One of the best stops on the circuit. It's in Florida. People are crazy and pretty bad at poker. Is there something about Florida that makes people there bad at poker?

Overconfidence? Bragadocio? Overconfidence. Yeah, YOLO-ness, right, is, you know...

Yeah, like YOLO inspires people to bet too much money at things they're not very good at. Conspicuous consumption, right? And the pros mostly live in Nevada, so you don't have quite as many pros. But no, Florida is a good place for poker. Speaking of, did you see the article in the New York Times that was featuring the new restaurant in Miami from the owners of Carbone? It was like celebrity packed. Rhonda Santos was there. I was kind of surprised to see that Rhonda Santos was there because it was full of, it wasn't very of the people.

Miami is a status-seeking city. I like Miami. But yeah, there's not enough glamorous restaurants relative to the number of- Glamorous people or glamorous-seeking people? Glamour-seeking people, yeah. So you've had a lot of New York restaurants. It's basically like Miami is like the sixth borough, Miami Beach. Literally, they're just random New York restaurant brands that will open up there because it's underserved for some reason.

for that type of restaurant. I'm looking forward to celebrating my birthday. Oh, yeah. Which is on Saturday, but I'm not celebrating it until after the election. So Pennsylvania, count those votes. Count those votes. Georgia, no runoff, please. Is that it? Is that a wrap? I think we're done. Tony, are you coming to my birthday party? Tony Chow is back from Parental Leaf. He is right here with us in the studio. It's so great to see your face. How was Parental Leaf?

Congratulations on the baby. All right, now we'll wrap the show. Thank you, Nate. Thank you, Galen. My name is Galen Druk. Sophia Leibowitz, Kevin Ryder, and Tony Chow are in the control room. Once again, welcome back from paternity leave. You came just in time for the election. I'm sure you're thrilled.

Emily Vanesky is our intern and Chadwick Matlin is our editorial director. You can get in touch by emailing us at podcasts at 538.com. You can also, of course, tweet at us with questions or comments. If you're a fan of the show, leave us a rating or review in the Apple podcast store or tell someone about us. Thanks for listening and we'll see you soon.