The Final Pre-Midterm Model Talk
49:10 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.

That's a loud computer. It is, yeah. Is that because that's where the model lives? The model is safely protected on a server. You don't carry the model with you? No. Where is the server? I don't know. Chris knows where it is. You don't know where the server is? No. Like, your baby's just somewhere on a server, and you don't even know where it is? I mean, there's a hard copy here, to be fair. So it is right here? It's a lot of places. The model lives in the metaverse. Does the model live in our minds?

I hope not. It's up here, man. It's up here. I mean, these days it might. Speaking of which, hello and welcome to the FiveThirtyEight Politics Podcast. I'm Galen Druke. I'm Nate Silver. And this is Model Talk. This is our last pre-2022 midterm Model Talk.

We freeze the forecast at midnight tonight. At the stroke of midnight. At the stroke of midnight. So we're still about. So sketchy. I'm not going to name the pollster. The sketchy pollsters who like release all the polls the day before and they all perfectly match the 538 polling average. Your time is up at midnight tonight.

Ooh, you're accusing some pollsters of hurting at the last minute. Oh my God. Oh my God. This one pollster hurts every year. Who is it? They're a non-American based pollster. I'm just going to subtweet them. A non-American based pollster? Can you at least tell us what country they're based in? I think...

I think they're Canadian. You think they're Canadian? Yeah. Who is it? Just tell us who it is. I'm not going to say who. I'm not going to say who. Really playing court with that information. Okay, listeners, if you can think of a maybe Canadian pollster that tends to herd towards 538's polling averages on the last day before an election, let us know. Anyway, we are freezing the forecast at midnight, which is 12 hours away from now. It is about noon on Monday.

So we don't have final, final numbers, but more or less, this is probably what the final forecast will look like. Republicans have a 55% chance of winning the Senate and 83% chance of winning the House. The governor's races are toss-ups in Wisconsin, Nevada, and Oregon. Arizona leans Republican, Kansas leans Democratic, and every other governor race is either likely or solidly for one party.

I only get to ask you this question one last time before election day. And I know it's your favorite question. So how would you characterize the race at this point, Nate? Too close to call. No, it's – according to the headline on our interactive, it's a dead heat.

Although, as someone who cares about small percentage differences, you'd rather have a 55% chance of something than a 45% chance. So therefore, it tilts ever so slightly toward Republicans. But it's important to remember that there's a wide range of outcomes. Well, you do realize that Democrats are still above 50% in the light and classic models. So they haven't fully converged on Republicans having the slightest of advantages. How do you feel about that? Yeah, but Deluxe is the better forecast. Deluxe is the better forecast? Yeah.

But I have a feeling that if Democrats barely hold on, you'll be like, well, the white forecast had Democrats up all along. It's a very close race, right? There's not really going to be any forecasting knowledge imparted by either party winning the Senate. If Democrats won the House, then that would be interesting. Yeah. Yeah. I mean, so how would you characterize the House?

I would characterize the House as, was it 83-17? I mean, you have a clear favorite in Republicans, but, you know, one in six chance, roughly. What's that Russian roulette? It's not that crazy. I mean, which is certainly not a game I would play. So, yeah. Take those out seriously. It's not that crazy for Democrats to win the House. It would require roughly like a three-point polling error, but that's a so-called like normal-sized polling error. People are just so, people are so f***ing stupid in politics. Yeah.

They're so stupid they can only remember like what happened two years ago. There have been elections in the past like 1998 or 2012 where Democrats beat their polls. It can happen. It happened in some of like the special elections and so forth just this year. People are so dumb though. They can only think about one thing at a time. It's certainly possible for Democrats to beat their polls and in that case then the house becomes competitive. Yes. Although look.

I think that if Republicans were to beat their polls again, that would feel more urgent because it would seem like election after election. We've talked about this plenty on this podcast. It would feel more urgent. I mean, it's a big night for the polling industry. And it would be in many ways more urgent, right? It's a big night for the polling industry. It is. Let's acknowledge that. We don't have to be cynical or coy. This is a big election for polls. Yeah.

Do you remember the last time one of our forecasts showed a 50-50 proposition like it does in the Senate right now? Wasn't it like 2020 in the Senate or something? No. When? It was 2016, the Senate. Oh, 2016 in the Senate. Mm-hmm. Okay.

Which is, I mean, kind of ironic. I don't know if that's the correct use of ironic because it was exactly six years ago. So it's the exact same seats that are up. Oh, yeah. This cycle. Do you know how that ended up turning out? That 50-50 proposition? Republicans won, right? Republicans won. It was 52 to 48. Yeah. Democrats flipped Illinois and New Hampshire. Yeah. Republicans held Arizona and Georgia.

And yeah, how much has changed since then? In the past six years? I mean, Arizona and Georgia have become more blue. Well, thanks for that analysis. Yeah. Okay. Hashtag hot takes. Okay. So we will see what happens this time. I think our forecast suggests that people should be prepared for any result in the Senate ranging from Republicans winning 54 seats to Democrats winning 52 seats. That's the 80% range. Yeah. I mean, there's kind of no like...

Either one party kind of wins big or you're kind of in for a fairly long evening, right? There's like no like – Week potentially or month. Week or month that there's a runoff in Georgia. And let's not neglect the fact that like increasingly there's rhetoric from the Republican side primarily about like whether we should respect the election results. They don't turn out the way that we think. And so those scenarios are on the table too. A close election –

And with the country being what it is right now and a lot of election deniers, Republicans on the ballot, is quite a scary proposition, frankly. Yeah. So let's talk specifically about how close this is. There are currently eight Senate races that are separated by five points or fewer. In our polling averages, Pennsylvania is eight.

even. In Georgia and Nevada, the Republican leads by one point. In Arizona and New Hampshire, the Democrat leads by two. In Wisconsin, Republican Ron Johnson leads by three. In North Carolina, Republican Ted Budd leads by four. And Republican J.D. Vance leads by five in Ohio. The average polling error in Senate races over the past 25 years has been about five points. Should we take that to mean that really anything could happen tomorrow?

Well, let me make one point. The five-point polling error is on a individual poll. Polling averages in theory are less error-prone because they average out eccentricities and outliers, which is important in this election because you have a pretty big divergence between the traditional pollsters that show a pretty good night for Democrats and the more modern pollsters, how to put it, who generally show a good night for Republicans. And so what's the average polling error on an average?

Well, in individual race, it's still fairly high. But less than five? Less than five. By definition, like an average is more accurate than an individual poll. To your point about institutional pollsters showing decent polls for Democrats and the sort of more modern, more partisan, and in this case, more Republican pollsters showing good polls for Republicans. Does that mean that if the polls are spot on, like the averages are spot on, that actually...

everyone got it wrong because the institutional pollsters underestimated Republicans and the Republicans overestimated. It means 538 got it right. Well, yeah. It means averaging is good. I mean, sure, but it would suggest that there's still trouble in the industry. No, I mean, you can't. Okay, well, now that's kind of what I mean. Like if the polling averages get it right, then you can't complain too much. But look, I mean, this is model talk. We're looking under the hood here. If there are two sets of polls,

And one is significantly Democratic-leaning and one is significantly Republican-leaning. And the average gets it, right? This is not that big a difference, right? No, I mean, in some ways, a pluralistic view of polling is a decent way to go. If there's no gold standard anymore, right, then people try different things. And it's actually not so bad. It's not bad if you...

You know, let's say that Democrats have a fairly good night. Let's say the polls are mostly right. Democrats do a little bit better than expected. Lose the House. Hold the Senate. You know, I mean, like, yeah, Trafalgar might look dumb, but they had a good year in 2020 and 2016. So that's OK. Right. It's OK. You don't have to nail every election. OK. Fair point. How do the methods of the two types of polls differ? One method is that you try to reach some universe of voters as though you are

Trying to do pollsters the traditional way? I mean, where are you asking going? So, like, if we end up with results that hew closer to the institutional pollsters, what does that mean about what are the best practices? And if we get results that hew closer to the partisan Republican pollsters, what does that mean about the best practices? I mean, the traditional best practice is that you use either random digit dial, meaning you literally just dial working phone numbers, or you use a list of registered voters and randomly select from those. So...

The premise is that you have an equal chance of contacting each person. You wait for known demographic divergences relative to your sample or relative to the electorate and then maybe apply a likely voter screen. There's a little bit of art in that and then go from there, right? That's a traditional way to do polling.

The non-traditional way is no one knows what the f*** they're doing because they're extremely non-transparent. It's probably you contact voters however you can, right, and then do a lot of massaging and waiting behind the scenes to the point where, you know, when does a poll begin and someone's opinion begin? I don't know. When does a poll end and someone's opinion begin? I don't know. But no, I mean, these groups are not very transparent. And over the long run, non-transparency is associated with higher polling error. But it hasn't been recently, so who knows?

Speaking of the likely voter model, we have expected that as we get closer to Election Day, that likely voters would probably hew more Republican than registered voters because...

As you suggested, pollsters have to make some determination about who's actually going to turn out to vote in a midterm election like this. Somewhere around 50% of eligible voters will turn out. So a lot of people who may have opinions on the election one way or another aren't actually going to register at the polls on election day. What is the difference right now? What do we see between registered voters and actual likely voters? So this has been kind of all over the place. And the example I give is in the Marquette University Law School poll of Wisconsin. They're a pretty good pollster.

In their previous poll, they had the race tied among Senate race, tied among registered voters, but Ron Johnson, the Republican, leading by six points among likely voters. In their most recent poll, they have Johnson ahead three among registered voters, but only two among likely voters, being that Mandela Barnes actually gained a point from the likely voters screen. So no one has any idea, really. There were a couple of polls this weekend from ABC News, our employers, and NBC News that had very little difference between likely and registered voters. If that's the case, if there's not...

much of a turnout gap, if any. Then Democrats are in probably a pretty strong position. Most of these polls assume that there'll be some GOP enthusiasm advantage. If there's not, the Democrats are going to feel pretty comfortable about holding the Senate and might even be viable in the House. I'm not sure if I count on that. Usually the Republicans have a traditional turnout advantage. Also, usually the opposition party is more fired up. But we're in a different world now where like all these college educated voters that Democrats are now, I was going to say, in charge of,

There's a big educational split in who votes for which party. And traditionally, people who attended college are much more likely to vote in midterm elections. So maybe that's shifted. Maybe Democrats will have a turnout neutral or edge even, and then those upside scenarios, Democrats are entirely viable. Is there anything that's happened over the past month that would suggest that Democrats' enthusiasm has increased? I mean, it seems like if you look at the polling and if you look at the likely versus registered polling results, that like...

that like something has happened in the past month that has amped up Democrats more to vote. I mean, you can tell some bullshit vibes narrative, right? We're like... Well, but that's based on the data. I mean, I don't know what's behind it. Well, they're not enough. So one thing that ABC and NBC did, which is good, is they actually publish the likely voter and registered results both so you can compare them. Most of the time, we have no idea, right? If you have a candidate four points ahead on the likely voter screen, is it because...

She was tied among registered voters, but the turnout enthusiasm helps her, right? Is it, you know, why does that occur or the opposite? Most pollsters are not very transparent about that. And so there could be a lot of fingers on the scale on the likely voter screen potentially. I mean, it's not crazy to think that like,

The Dobbs decision, Roe v. Wade being overturned, energized some Democrats early on. Then the GOP kind of got their low-hanging fruit excited about the election with crime and immigration and inflation and whatnot, right? And now there are some Democrats who are hearing maybe the Biden campaign

on the ballot message are like, this Republican's unacceptable to me. I have to vote. The election seems fairly close, which it is in the polls, right, in the Senate anyway. I mean, it's somewhat viable, I guess. There's also the issue of like non-response bias in polls, which is, you know, maybe people were excited about the vibes or down on the vibes and then affected whether they responded to polls or not. But I mean, it's a plausible story. It's just like,

not that robust because we don't get enough polls that actually make this head-to-head comparison between likely and registered voters, which we should. Any pollster publishing a likely voter number should also publish its registered voter number.

Well, it's interesting the story that you tell about Democrats being enthusiastic to vote after the Dobbs decision and there just being a backlash to that ultimately unpopular ruling from the Supreme Court. Then inflation, immigration, crime coming to the fore. Throughout this whole process, there have been, I think, what we've come to call gas price determinists who have followed gas prices throughout the year as some indicator of the most apparent, obvious, in-your-face situation

measure of inflation. And it looks like gas prices have held pretty steady over the past several weeks. Right now, the average gas price in America is $3.80. It hit a high average of $5 in the middle of June.

How are you sort of processing? I mean, I have an electric car, so I wouldn't know about the price of gas. No, I'm just kidding. I don't actually. I was like, Nate, don't bluff. You don't have a license. I know that for a fact because I've driven you around. I have a car, though. My partner and a car. Okay, fair enough. It's not an electric car. How do you know that? I did not know that. But I would think given that you don't have a license, you would be less aware of gas prices.

I am not a gas price determinist. I mean, I think it's important. I think it's probably a coincidence that it's tracked so closely with like the generic ballot and other measures. It's very visible, obviously. You literally see it when you drive on the street, what the gas prices are. I think voters are a little bit more complex than that. You can't will everything down to gas prices. But yeah, I mean, the dumbest thing, I mean, there are still now and then these Democrats who are like the media. Inflation is just a creation of the media. Like if you say that,

don't ever comment. I don't like to do this, right? Because people can be wrong about stuff and I have my share of terrible takes. If you think that like concern about inflation is just a media thing, then you are one of the dumbest people. I think that's like faded significantly. There's still, there's still, there's still, there's still some of that, dude. And I think,

The most obvious example is that all throughout the summer when abortion was top of mind and inflation wasn't being covered that heavily in the polls, inflation never budged from being the most important issue. You know what it is when you're paying more for gas prices and groceries. Yeah.

Well, right. Which is why, I mean, we talked to economists earlier on in the year who were saying during a recession, people are concerned about losing their jobs. A quite high, very high unemployment rate is 10% in the United States. And that may cause a lot of people to fear losing their jobs, but it ultimately doesn't affect everybody. Inflation affects absolutely everyone. For sure. And so that can be a sort of... Although, I mean, some people can benefit, you know, if you own a home in

In some parts of the country, housing values have gone up a lot, right? But you don't necessarily notice that so much, right? You're not like, oh my gosh, you know, I'm not getting like an extra paycheck. It's just that if I were to choose to sell my house at some point, it's worth more. But like it's not an immediate dividend to you. So we've been engaging in looking at the data and trying to figure out what the narrative of this election is, why voters are behaving the way that they're behaving, how they think about the two parties in the current era of politics. What data are you going to be paying the most attention to in order to understand why

why this election turned out the way it did. The votes, Galen. I'm going to be looking to see who votes for which party. But look, the narrative building is going to start at like 7 p.m. tomorrow. So I do mostly buy, even though it's kind of cope, right? I do mostly buy the take that like this

Ought to be a pretty bad election for Democrats and the fact that they're competitive is a little surprising in some ways. You have 8 point whatever percent inflation. You have Biden with a 41 percent approval rating. Maybe that means that actually the polls are wrong. We're going to see a big red wave. It's very possible, by the way. But if it doesn't occur, if you have this ambiguous outcome or a Democratic Senate, then like that's that's a decently impressive outcome.

result and suggest that all the GOP shenanigans and extremism does have some effect, at least for some of the candidates who are nominating these key Senate races that did matter at the moment. What's in the bucket of shenanigans and extremism? Is that abortion, January 6th, election denialism? Well, abortion's not, we shouldn't lump all those. I mean, January 6th and

election denialism are in a different and, to me, more severe bucket. Well, sure. But it seems like we had elections last November that took place after January 6th, in which Republicans did very well. Glenn Youngkin, the Republican, won the governor's race in Virginia. Republicans came pretty close in New Jersey. So we saw a post-January 6th election where Republicans did well. Now we're dealing with a post-Dobbs election. So in terms of parsing which matters more— It's weird how little, almost—

That 2021 elections, it's weird how little they seem to have to do with the 2022 elections, right? I mean, I think the GOP had in some ways some like low hanging fruit from like COVID restrictions and like education policy and whatever critical race theory means to people, whatever that bucket is, right? And like those things have not been really –

top of mind at all in this election. In some ways, it's been much more substantive, frankly, than 2021. So I don't know what there is to learn from 2020. Well, my suggestion slash question is that then it maybe is all about Dobbs. Like, what changed between then and now?

Dobbs is the very least would seem to neutralize this GOP offensive on like, quote unquote, cultural issues. You had a lot of mass shootings this year. Gun control is not an issue. We talk about that much. You also have much worse inflation. I mean, you've had Biden accomplish much more in 2022 than 2021. You have a war in Ukraine. There's a lot of stuff that's happened this year. It's been a long year. Big year for happenings. And it's not over yet. We still have an election. OK, let's talk about the House a little bit before we move on to listener questions.

It's the day before Election Day. The forecast suggests that Republicans will win the House popular vote by four points. Our generic ballot polling average, which is the average of polls that ask, would you rather see Republicans or Democrats control Congress?

Shows Republicans ahead by one point. We have talked about this before. I know that there are more districts where Republicans are running unopposed than where Democrats are running unopposed. So that will help boost Republicans' House popular vote margin. HPV, House popular vote? Yes. You had to do it. You had to do it. How old are we? Are we like 16 years old? That's the humor level of this podcast. Yeah.

Specifically model talk. Okay, so that being said, what gives? Because I think we expected some convergence of the generic ballot average with where the forecast thought the House popular vote would end up. But there's still a three-point gap, and the election is tomorrow. So let's talk about a few things. The generic ballot average you see on 538.com is not actually the same one. No, it's not. It's...

It's whatever pollsters give you. We prefer likely voters to give you both options, but it's mostly likely voter with some registered voter mixed in. The actual number our model uses, we adjust the registered voter polls to a likely voter basis. Well, then why don't we make the one that our model uses the one that the public can see? You know what, Galen? I can't tell you that.

You're like, it's the day before election. It's too late for these arguments. I can't tell you that. But so we have the generic ballot average are we show you on the site is like one point for the GOP. What is it right now? Yeah. 1.2. Yeah. The one Armada uses is more like 1.8. So closer to two points. So that's part of it. But then it still thinks there's another two points on top of that. So one and a half points or so is because of these missing Democratic candidates. Right. In like 23 districts, there's no Democrat in the ballot. There's also no Republican in the ballot in like 12 districts. But like that actually has like

an effect on the size of like one and a half points. And there's some other stuff. We don't actually use

only the generic ballot. In a district, if you have lots of district polls, we use those polls instead. There's other factors, incumbency and so forth. And so when you forecast each district, including districts where there's no Democrat on the ballot or no Republican, it comes out to about a four-point prediction for a GOP popular vote win. If it weren't for the missing candidates, it would be more like GOP plus two and a half. So that's kind of our view of the national environment is that in a neutral district, partisan lean of zero, no incumbent,

two identical candidates, both sides-ism in the best version of that world, right? We think the Republican would win by two and a half points based on the national environment, so-called. So more than our generic ballot average shows, but less than our popular vote estimate shows. Okay, here's a question.

What is a Republican overperformance or a Republican underperformance? Would you say our baseline is four points or two and a half points? Our forecast of the House popular vote is that the GOP will win by four points. Part of that is because they actually have candidates in the ballot and more districts than Democrats do. But we don't account for that. So what's an overperformance and an underperformance?

Well, I mean, if they win the House popular vote by more than four points, it'll be an overperformance relative to their model. Yeah. Okay, so...

That's interesting because I think if they were to win it by four points folks would say like oh, that's decent for Republicans That's pretty good for Republicans. That's pretty Typical what is it like give it seven or eight point shift? They lost the House popular vote by like three points two years ago. Mm-hmm So what's typical shift? I used to know this if there's like a seven or eight point shift, which is fairly average Then you get to like four or five But again, it's a little bit inflated by the fact that you have these missing Democratic candidates 50-state strategy

No more. Democrats just give up sometimes. So if they get their four points, that's like not ultimately that good using history as a guide. What was it in 2014 U.S. House election? That was the last really good Republican year in the House. And in that year, they won the House popular vote by six points. I do that right? Yeah. 51 minus 45 is six. You know, I'm tired. I can do that. All right. So final question before we get to our listener questions.

Did we have an October surprise this year? And is Trump our October surprise?

The surprise, Galen, is that there was no surprise. Is Trump getting back in the news talking about his potential 2024 announcement? I guess that's not very surprising. But at this point, is there much to say about how he is shaping the final stretch of this election? I mean, times when Trump's been more present in the news cycle have not been good for Republicans. So it's not helpful, I don't think. But I'm not sure that people are being swayed by—

By Trump or by Elon's endorsement of Republicans this morning. I'm not sure that's going to shape things too much. You know, I missed that. Yeah. What did he do? He's like, we need balance in this country. So vote Republican this year. I am sure that that will have a really important influence on American voters. Okay. Let's get to some of 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.

All right. Our first question comes from Jacob Rabashkin, who's actually a big fan of the podcast.

been a contributor on our live blogs. Wow. Someone actually qualified and professional asking us questions. Rude. We got some really good listener questions. We get great listener questions on being asked. How would the models forecast change if you took out all of the House polling but left in Senate and Governor polling?

To get all the House polling, the left and Senate and governor polling? I have no idea. That's too good a question. All right. It's too good a question. All right, Jacob. Sorry, we can't answer your question. Well, it's just, is the House polling better for Republicans or worse for Republicans? I think it's been worse. Well, the Times polling was worse, but I guess... I have to be honest. I have been so focused on the Senate this year because it's so much more close and competitive that I...

don't have the encyclopedic knowledge of the House polling landscape. Our apologies, Jacob. Actually, this might be something that you can figure out yourself if you look into our averages. There you go. Oh, breaking news. Breaking news.

Republicans now have a 56% chance of winning the Senate. That means they are slightly favored instead of a dead heat, according to 5E Fox. Ooh, breaking news right here on Model Talk. Yeah. We often break news on Model Talk. We do. We really do. Okay. Ariel asks, are there any plausible outcomes of these elections that will cause you to make fundamental changes to the model for next cycle? You're like, uh, but that breaks my brain. Next cycle? Next cycle. I mean, if the polling's really shitty, then maybe just turn the model off.

Right. I'm not going to battle with impossibly terrible polling. I mean, you could make a model, right? But I'd just turn this off probably. Yeah. Go back to vibes only. I mean, you can still short the vibes.

You can still short the vibes. Yeah. I mean, you'd make a polling average and say, well, this is at least what the polling shows, but. Make a polling. I mean, yeah. But if it becomes like impossible to like know if anyone has any idea what they're doing. That's, you know, that's kind of letting the haters win, Nate. If you turn the forecast, all the, all the people have been saying these forecasts are garbage in garbage out.

Early man, I have so many other things I could do that I enjoy more than the f***ing election model. Okay, we'll get existential like this after the election actually happens and we see whether or not we all have to find different jobs. Next question. Henry asks, is there any evidence of hurting in the polls? Of course, you called someone out earlier in this podcast, but apart from the unnamed possibly Canadian pollster.

Yeah, go to like the 538 poll scroll, polling page, and like— The poll scroll? I like that. I dare you to find like one remotely surprising result in that. Any poll released in the past 24 hours, no one's remotely been— I mean, you know, Trafalgar has Raphael Warnock losing in Georgia. Shocker. Canadian polling company matches the 538 averages exactly in every state. Like there's not a single interesting result there. So you think there's quite a bit of hurting going on? Of course. A lot. And that's what the incentives dictate.

Should that make our alarm bells go off? Because usually, on average, hurting makes averages less accurate. I mean, there hasn't been, like, a ton of movement in the last five or six days. But, like, basically, I should check and see this. I wonder if you, like, in presidential primaries, this is the exception because their races move so fast you need the last minute polling. I do kind of wonder if, like, looking at the polling averages on Friday is better than Tuesday morning sometimes because it's all ass covering from Friday through Friday.

Tuesday. You never learn anything that anyone would remotely actually put themselves on the line. Oh, that's interesting. What to you is the most interesting polling result you've seen in the past week? I do think those polls showing that very small likely register voter gap, ABC and NBC were interesting, right? Oh, so the most interesting poll from the last week is our employers poll? Our employers poll. Nice name. Correct. Nice. Gotcha. Company man right here. Yeah. Okay. Next question is from Jacob.

Is there an impact to the model based on differences in voter engagement on state-level races? Like a House Democrat in Ohio might have a better shot than normal due to the high-profile Senate race.

whereas a competitive New York district doesn't have that boost. No, we don't account for that. And that would probably be a smart thing to account for. So because Tim Ryan, the Democrat in the Senate race in Ohio, is doing significantly better than you would expect a Democrat to do, that other down-ballot Democrats may end up doing better. It's probably true, but we don't account for that. So, yeah.

Okay, next question. This is such a classic question. Blair asks, like Nate, I'm a big sports fan. And for sports, momentum is key. How does momentum factor into political races? So in theory, the model is supposed to be designed in such a way that there's no autocorrelation. And what that means is that

The forecast today doesn't tell you anything about the forecast tomorrow, right? Is that what we call a random walk? A random walk. If you go from 60% to 55% GOP chance, the best forecast for what the forecast will say tomorrow is 55%. You can't derive anything from the trajectory of the forecast so far. Sometimes you'll see in the midterm forecast where our model is pretty conservative about updating for new information. So maybe you have some sense based on the generic ballot. Things are shifting fast.

but you don't see an individual races until later. Conversely, our presidential forecast is designed to be more aggressive and make inferences about how races shift from one race to another. But in general, I don't think MoMTN tells you that much. I mean, clearly the GOP has closed the race pretty well, but I think that should be accounted for by the fact that they're now

at 56% in the Senate when they were at 29% September 17th. So there's been momentum, but I think it's pretty well priced into our forecast. Well, okay. So here's a good follow-up question in that case from Henry, which is, is it troubling that the deluxe forecast moved consistently towards Democrats from July to September, then consistently towards Republicans from September to November? A little bit. That's not great. That doesn't seem very random walkie and you'd want it to be more random walkie. So that's not... Wait, what are some potential reasons that it isn't random walkie?

Such a nice little curve here. So one reason could be that it's not well designed and that it should update more aggressively, right? Although weirdly that would have caused a bigger bounce back and forth. Maybe it was too optimistic about Democrats after Dobbs and should have clung to like the priors a bit longer.

But it could also have to do with like the vibes that pollsters are hurting more and more in an environment with less and less gold standard polling and therefore that they're kind of copying off one another and that this is partly artificial. I don't know. No, that's great though. It's not a very natural pattern for a random walk. So something's a little amiss. Would you change something? If you're seeing a lot of pollster hurting back and forth, then that kind of I think violates certain assumptions of the model and it probably in some sense reflects public sentiment but is not ideal necessarily. Yeah.

I mean, it's also like not crazy to think that there are I mean, there are benign explanations, right? Democrats legitimately did have like a lot of good news cycles in a row in the summer and some more problematic ones after that. That's not crazy to think. Or maybe it's a sense of voters like we talked about before. Maybe Democrats got energized earlier because of Dobbs and GOP came later on. And a lot of these races you see actually not Democrats losing ground, but GOP gaining votes from undecided. So it's not it's not crazy.

But it's not ideal. If this pattern happened in every election, then I'd say something's wrong with the way that we're forecasting. Okay. Virginia Contrarian asks, we've gotten questions from this contrarian from Virginia plenty of times throughout this cycle. How do you know it's from Virginia? From Twitter. Nate has talked about the theory that polling error in 2016 and 2020 was tied to Trump turning out low propensity voters. This person says through his celebrity, but there are various reasons that it could have happened.

Is there any evidence that celebrities do better in elections or tend to beat their polls? And could this logic apply to Walker and Oz? That's interesting. We'll see. That's a good question. I like that question. I have no answer, though.

Okay, next question. Pre-Trump slash COVID, it seems Republicans always won the early vote. Afterwards, that flipped to Democrats. Who is largely expected to win the early vote and Election Day vote this time around? Or should it be looked at as neutral or even an unknown? And does this even matter?

I mean, I guess I assume that Democrats tend to do more early mail voting and that the JIP begins on Election Day. But, you know, I get so turned off by like conversations about the early vote because they're almost always just partisan wish casting that like I don't.

haven't looked at it as much as I should, but it will matter. I mean, we still have like, are going to have in a lot of states like a red shift or a blue shift where certain batches of ballots are counted first. I assume again, that mail ballots are going to be somewhat democratic. They didn't used to be, but I assume there's some hangover effect from, from COVID still. So I think this is an important place to say that, um,

A lot of the stuff that we talked about in 2020 still applies in 2022, which is don't jump to early conclusions on election night based on the results that you're seeing in places like Arizona, Nevada, all of the states that we're going to be watching closely. It's still kind of is a little bit of a,

So we also got some questions this time around from the folks watching along on YouTube.

on YouTube, shout out to everyone on YouTube. We're in the studio, we've got all this nice lighting, especially like we even have extra lights today set up for election day. - Extra bright and shiny. - I know, so we look extra good, thanks Tony. So here's one of the questions from the folks on YouTube. How does your confidence in your election model compare to your confidence in the various sports models FiveThirtyEight provides? Have you learned anything from the non-political modeling that's informed the model in past years?

I mean, sports in some ways is easier in the sense that you just have a lot more data. It's more data rich environment. And so you can kind of be more like, oh, this is just how things work, which is less true for politics. You know, we also have more team members, to be honest, devoted to politics and elections. So things get updated a little bit faster.

If there's some pole, they'll get entered very fast usually. If there's some injury in the NBA model, then although we're now trying to automate some of this, it might take a while to enter in. But you know that New York offers a plethora of sports betting options? I've learned a little bit from that. I've learned a lot about sports betting. What have you learned? Any hot tips? Okay, you have to bet early. The best lines are lines that are posted early.

Right at the beginning when it opens and not after everyone else has gotten to make the market move. So then you're spending it's like one random dude who sets the lines and not the entire sports betting community. Be aggressive about shopping for different lines. If there are eight different like legal sports books open in New York, you're going to get a better price somewhere than others. And that can actually take you from.

negative expected value to neutral or positive specialize hot tips specialize if you have a particular type of sport or bet that you follow then focus on that that gives you maybe a finding chance to be better than the consensus if you just kind of dabble in things that's less likely yeah thank you for all of that yeah it's not going to help me but i hope it helps someone all right next question desi asks did debates affect any poll for senate or governor this year

I guess we've been wondering, do the debates matter? Will they matter? I think the conventional wisdom is that they matter more in primaries because there's not quite as much of a hurdle to get over to switch your candidate. We had a Pennsylvania close from – I mean at one point Federman was up 10.7 points on September 10th and now it's tied. So clearly that made some – well, actually I take that back because like this – A lot of the closure happened. You can't actually point to like – yeah. It looked like a fairly steady closure since mid-October and not –

Because of the debate necessarily? I remember on the day of the debate, it had closed to a three-point gap in the polling average. It's still a pretty big shift. From three to even? Yeah. Although we don't know. Is it momentum? PA is very vibes-y. It's very vibes-y. You mean the polling is vibes-y? Very vibes-y polling, yeah. So you think it's less accurate? People are just going with...

what they feel like might be the result as opposed to what the numbers are really telling them. I mean, there's this whole narrative about pollsters flooding the zone, which I think is mostly dubious, but it's a pretty zone-floody mix of polls in Pennsylvania. Which is to say a lot of partisan pollsters putting their cell on the scale for Republicans. You know, I don't think putting—I mean, they might be right. They've been some of the most accurate pollsters in the past couple election cycles, so they might very well be right, but like—

but Pennsylvania feels like a lot of vibes. A lot of vibes. Okay, so on that topic, in a sense, we get a lot of listener questions along the lines of,

So in this case, for example, Chester says, I've been talking door knocking to voters for the last few months in deep red parts of Ohio, and it seems like Tim Ryan is quite popular. What are his chances? Could the model be off by five points? So this is a genre of question. I'm going to broaden it out. Are there any areas where you actually think a candidate is being unpopular?

underestimated or overestimated in the forecast right now, the day before Election Day. We've talked about this before. I know Georgia is one that you've said maybe Raphael Warnock has a better shot than the forecast suggests, but anything else?

And I'll name a couple of the other names that came up from our YouTube watchers. They were Franken in Iowa, so the Democrat in Iowa, or O'Day in Colorado, who's the Republican in Colorado. I mean, in general, I am anti-upsets in those races. And also I'm like anti-Lee Zeldin upset in New York. I mean, the partisan leanings of these states are often pretty strong. Colorado is a blue state at this point.

New York is a very blue state. Ohio is a red state. That's not where I'd be looking for upsets. I think Armada does a good job of calibrating those different estimates and there's whatever statistical chance there is. But, you know, voters sometimes come home. I mean, in New York, I think I tweeted this or mastodoned it. I'm just kidding. I tried to get it in a mastodon. I couldn't. You know, this has the feeling of a race in New York that if there had been no hype about it, like the New Jersey governor's race last year,

It maybe gets close because the GOP voters are more excited here in New York. We don't have competitive House races. We don't have a mayoral race. There's not much version to vote in New York. But now you see 73 billion New York Post headlines about Lee Zeldin's comeback and then you show up to vote. And that probably keeps Hochul in the governor's mansion. You're saying that the New York Post is keeping Cavi Hochul in the governor's mansion? That's a hot take.

I know what you're saying. It's hard for – and Lee Sheldon is like a Republican-y Republican, right? He's not some like George Pataki moderate. Right. I mean most significantly on Trump and January 6th issues. Yeah. I mean he's been a little bit ambiguous but like he's not –

Throughout Trump's presidency. Best buds. And on January 6th on Fox News from the Capitol, you know, coming up with excuses and stuff like that. Yeah. Although I think our interactive does not officially classify him as he's classified as like election denial curious by our interactive, I think. Yeah. I mean, yes, sure. Yeah. But he is not someone that has a traditional crossover appeal and like it's a very blue state. So the fact that this has been so publicized as an upset in some sense makes it less likely, I think.

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, last serious question, then we'll get in a few jokey questions before we wrap. We got plenty of a version of this question, which is, is there a particular race or state, district, et cetera, that you're going to be watching on? So you're going to be on ABC News. Folks should tune into ABC News. Folks should also tune into our live blog at FiveThirtyEight.com. They should also tune into our YouTube channel. We're going to be recording videos there.

All night, we're also going to have a podcast out at the end of the night. So tune into all of that. But you are going to be on ABC News responsible for telling the nation what is happening as it happens. What are you going to be paying most attention to in that hot seat doing your duty for the nation, Mr. Silver? I'll give a little bit of a...

boring answer, which is New Hampshire of all things. And the reason why is that New Hampshire tends to count its vote pretty quickly. It's a pretty uniform state. There's not like super blue and red patches of New Hampshire. So New Hampshire is a place where we can get

a relatively early read on things. There's two House seats there. There's a Senate race. The governance race is not supposed to be competitive. It's also pretty swingy. It's pretty swingy. If Maggie Hassan lost, then that looks very bad for Democrats. That's a two-point race right now in the polling average. And Democrats kind of helped boost Don Bullock, who his opponent in the primary was a more establishment candidate. So that could be a real, what's the expression that I'm looking for?

f*** up if Democrats ended up boosting Donald Bolduc and then he actually won against Maggie Hassan? Yeah, no, I mean, that would be not really that much of an upset according to our model, but that would make it much harder for Democrats to retain the Senate. That'd be bad news for them for sure. Okay, so here's some jokey questions. Michael asks, would you rather have, one, die and go to hell where you have to analyze an election cycle that gets closer and closer to election day but never happens?

Or two, the model be off by 15 points. Two, because I could just quit. Yeah. Obviously, one is literally like rewrite Dante's Inferno because that is a unique and horrible way of experiencing eternity. A hundred percent. Okay. Next question is, how are you preparing for the stress and lack of sleep awaiting both of you on election night?

I mean, the good thing about election night is that not much really happens on election day. You don't have to be at the office until, you know, late afternoon, early evening. I already voted. We can vote early now in New York. So I'm just going to kind of sleep in tomorrow probably. Well, you know what day actually sucks? Wednesday. Wednesday is the worst day because you've had –

Maybe two to three hours. We'll see what happens. Unlike Tuesday where you're not really covering the election results all day, you're covering election results all day on Wednesday and on much less sleep. So I don't know. Maybe a lot of coffee and, you know, nicotine maybe, booze. I don't know. You know, who knows? Who knows? Who knows? Final question here. Oh, breaking news. Oh, no. It's now a dead heat for the Senate again. It's going back to 55-45. You know what? F*** this s***. I'm out of here.

Okay, final question. How much PTO is 5e taking after Tuesday? Well, like I said, 5e is not off duty until well into the week. I mean, I think foxes have like a different conception of like work-life balance. Because they're always working. I don't know. I mean... Or are they always not working? They are just much happier existentially than humans are. So like 5e is always doing what 5e wants to do. Show me the receipts. Where's the empirical evidence that foxes are happier than humans? Just look at them. 5e is right there. Okay.

He's just doing a Rubik's Cube on the counter. Yeah. I don't know. I feel like foxes are always working, though, because, like, you're always trying to find food and shelter and, like, worry about your safety. Foxes are very Nietzschean, right? They want to, like, achieve success and righteousness through work.

Yeah. So are you saying that 5E is not taking any PTO? 5E will do it. 5E is preparing for Trump to potentially announce his 2024 candidacy on November 14th. So 5E can't take any PTO. That's not 5E's department. 5E doesn't have to, you know. I mean, 5E's got to make a Republican presidential primary forecast. We have a cool presidential primary model. 5E's hard work is done. Yeah. It's to show up on the forecast and...

say fun, comforting stuff. And yeah. What is 5E currently saying on the forecast? Each dot represents a potential electoral outcome according to our model. That's some text there. Hover over. Members need context. Read our election updates for more.

That's very un-Fivey-like. I feel like those words were put into Fivey's mouth. Members need context. No, they don't. Here's another one. Since the 2010 election, Republicans have held the majority of governorships. They currently hold 28 seats, while Democrats hold 22. That's interesting. Mm-hmm. Fivey's giving us historical context. Thank you, Fivey.

Oh, there's a, congrats, you made it. Everything is fine, Fabio says. Oh, wow. I didn't even know. I hadn't even seen that one. Oh, wait. Scroll all the way down to the bottom. Oh, wow. I didn't know. That's a great little Easter egg that even I wasn't aware of.

And then there's an upside down smiley face. And 5E's drinking a coffee in a room that appears to be on fire. That is... Wow. Did somebody break into the forecast? Maybe. I love that. And change 5E? Okay. What's your final thought, Nate? Anything you want to get on the record before Election Day, you've got to get on the record now. Okay. So when I pre-register...

One, it's a big night for pollsters. Two, I've never had like less intuition for what would happen. I want to pre-register that take. And three, you know, I don't know if I say enough, all this shit that's happening with like election deniers on the ballot and more political violence, it's pretty f***ing scary, right? I'm pretty worried about the future of this country. So I just wanted to pre-register that take. All right. Well, there you have it. The final pre-2022 election.

midterms model talk. We're going to leave it there. Thank you, Nate. Thank you, Galen. My name is Galen Druk. Tony Chow and Emily Vanezki are in the control room 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 any questions or comments. If you're a fan of the show, leave us a rating or review in the Apple podcast store or just tell someone about us. Thanks for listening and we'll see you soon. ...