Transcribe your podcast

Tell me something funny. We need a good cold open. I can't be spontaneously funny like that. I'm not a comedian. We got it.


That's a cold open. We're actually like.


Hello and welcome to the 538 Politics podcast.


I'm Galen Drink.


I am Nate Silver and this is Model Talk.


All right. I don't know if we got it, but we gave it the effort.


You know, it's a rainy morning and we're recording this nothing but darkness on the horizon. So, you know, maybe a little bit lackluster relative to some previous Model Talk intros.


Well, I was going to say we have 18 days until Election Day. How are you doing? It sounds like not great.


No, I'm doing OK. It's like Zeno's paradox for each day. Each step is like twice as long or something. Yeah, I know there's been a lot of the social media and stuff, like a little bit of relitigation of 2016. Always a fun topic. You love to see it. Do you feel like the pressure is on this time around in a way that's any different because people had doubts about the forecast in 2016, whether legitimate or illegitimate.


I think and you can go ahead and I don't know if you listen to, like the Zócalo podcast, like the aggregators. So if you're an aggregator, I think FiveThirtyEight, not only should we not have blamed I think we should have gotten a lot of credit for 2016 for being prescient and for giving Trump a pretty good shot when other people didn't. Well, right.


But anybody who listens to this podcast knows that. So I guess I'm curious, does the pressure feel any different this time around?


Not because whether or not those criticisms are legitimate, we've said that they aren't, but because of the popular perception of forecasting and polling and things like that.


I mean, it's all it's all stupid. All these models. I mean, they used to be a bigger difference is now all these models have Biden at about the same number now at 87 percent. Economists is ninety one percent. One's eighty eight percent. Eighty four percent. They're all saying, OK, look, we're now at the point where you can go back and examine how big is this lead? How off would polls have to be for Trump to win?


Right. It just we're just handicapping services. That's the correct number that Trump is giving him a range somewhere in the range of five percent to twenty percent, depending on different assumptions, given how far behind he is in the polls and given that we have two and a half weeks to go and accounting for the fact that errors are correlated in different states and like we are setting those odds correctly, we do not determine if we're in the one in eight world where he wins that one in eight or one in seven chance.


That's not up to us. It's up to voters. It's up to the candidates, up to the world, right up to the pollsters, I guess, to some extent. But we're setting proper odds. And in a world where people are more rational, then I don't really care if the one in eight comes up or the seven and eight comes up. Now, realistically, will I be in a lot of. Well, the whole polling will be in a lot of probably.


But I don't give a shit because, like, I can't do anything about it. And all I can do is give you an honest answer and give you the best forecast that I can so I can't do anything about it. I'd much rather be on the side where eighty seven percent of the time we'll look at least somewhat smart instead of 70 percent, you know, given that we get shit done, even if, like, we have a narrow favorite who loses, I'd much rather that someone be a heavier favorite.


But beyond that, I can't do anything about it. All right.


You heard it here first, folks. According to 538, Nate Silver has exactly zero to give. If you don't like our forecast, I mean, what can I could I do?


We spend so much time with these models thinking carefully about different sources of uncertainty and making conservative choices. We think not unduly conservative, but understanding that usually when you're back testing a model, you tend to make overly optimistic choices. So when we're kind of making deliberate choices, then we air on the conservative side. Right. Donald Trump is ten and a half points behind in national polls. Probably seven is seven and a half points behind in the tipping point states.


Biden's at fifty to fifty three percent of the vote. There are not many undecided voters left. Twenty million people have already voted as the time we're recording this right. The race has been pretty stable. Trump is trying the same lines that he has for the whole campaign that have him ten points down. Right. There's a pandemic that's killed. How many people now? Two hundred fifteen thousand plus. And it's getting worse again in the US. We're not going on a limb on a limb, but with a model that says Biden is the heavy favorite to me, that Trump still has a one in seven or a one in eight chance or whatever is actually kind of high given all the things that are going wrong for him.


All right.


So I think it's fair to say that how the public views our work is to some extent out of our control at this point. We try to be really diligent. We try to show our work and explain where we're coming from. And we hope that people appreciate that. But let's check in on. Specifically, where our forecasts stand for the presidency, Senate, House, et cetera, and then we'll get into some more specific questions. So as of the time of this recording, which is Friday morning, as you already mentioned, Joe Biden has an 87 percent chance of winning.


Democrats have a 73 percent chance of winning the Senate and a 96 percent chance of keeping the House. I also looked this up before we recorded. It's not listed on the forecast, but Democrats have a combined 70 percent chance of winning all three. So there's a slim chance that Democrats would, I guess, win the Senate, but not the presidency. But in large part, that trifecta hinges on the Senate or vice versa.


So let's start by talking about the Senate, which has the most uncertainty involved in the forecast at this point. And Democrats chances of winning that chamber have gone up by about 10 percentage points in the past 10 days, and they've gone up from about 58 percent when we launched. So what's responsible for the changes that we're seeing in the Senate?


So one thing to keep in mind with all these forecasts is that. Democrats are ahead, and so if time runs off the clock and the situation stays as it was before, then they gained ground because there's less time for Republicans to have something shift that brings the race back toward them. Yeah, with the Senate, it's just a lot of like drip, drip, drip of individual polls and individual states that tend to be fairly good for Democrats. There's been a bunch of polling in North Carolina that tends to show Cal Cunningham, the Democrat.


Still ahead, despite sexting accusations giddy in Maine, the Democrat is up to 63 percent to win. Kind of it's a little bit out of the tossup zone into lean Democratic in Maine. In Georgia, quite a bit of polling there in both those races. Remember, there's a special election race in Georgia that has multiple candidates on the ballot for both parties. What's happened there is that you have two Republicans, Kelly Loeffler and Doug Collins, who are kind of trying to outflank one another and be kind of crazy, far right.


I mean, Kelly Loeffler, like, was endorsed by QAnon supporter the other day and proudly bragged about it, whereas one Democrat, Rafael Warnock, has locked up almost all the Democratic vote. So that's all of a sudden a very competitive race when we play out both the primary and the runoff.


Yeah, on Georgia, we did have several listeners send us questions about that. Folks noticed that in one day Republicans went from a seventy two percent chance of holding that seat in the special election to a 51 percent chance. Was there something specific that happened on that one day?


So it is pretty unusual if you track our models for us to have that sharp a shift. The reason it happened there is that you have all these pollsters that have been conducting polls of Georgia, but they haven't been testing potential runoff match ups. Quinnipiac finally tested Warnock versus Loeffler and Collins one on one, and they have Warnock the Democrat way ahead. So finally getting a runoff polls and we basically we're flying blind made the model more confident that Democrats could potentially win a runoff there.


And since that race is very, very likely to go to a runoff, then that was important to the model when considering the likelihood of the Democrat or Republican winning that runoff election in Georgia.


It's going to happen in January and we're already going to know the results of the election. So does that change the probabilities at all once that happens? Is there a different electorate, say, in a special election runoff than there would be in November?


So currently the Democrats have an edge in the generic ballot. There are six points ahead. And that's kind of the national environment is pro Democratic, four races for Congress by six points. I believe the model assumes that for the runoff that it goes back to a neutral environment or at least an uncertain environment. So it does assume that things are worse for Democrats than they would be if you had that runoff immediately. So it's like Warnock is ahead in the polls of the runoff.


Our models prior is probably that Republicans would win the runoff and it's balancing those two and kind of coming up with 50 50 in Georgia now.


All right. So we were focusing in on the Senate. We've talked about North Carolina and Georgia. Are we seeing any other significant polls coming in, giving us a better picture of how the races are playing out and maybe starting with Alaska, which plenty of people have asked about and has become somewhat of a meme on Twitter?


What's the expression? Don't support Alaska, which I think the election Twitter has been like crowd funding, even polls of Alaska.


I don't know what exactly is so exciting about Alaska. I am sure that it's lively and obviously it's the largest state in the union. So there's lots to love. But what do we know about the actual Senate race there or presidential for that matter?


So we will know more after we finish recording this. A New York Times Siena College Upshot poll will be available in Alaska when you don't have the inside scoop.


I don't have the inside scoop I could ask, but there hasn't been a lot of high quality polling in Alaska. No, when the polling in Alaska is always kind of a mess, you do have some partisan polling suggesting that the race is very close, the Senate race there. I'll be curious to see what that New York Times poll shows. But you to remember in 2008, a similar year for the top of the ticket presidentially, Democrat Mark Begich won a Senate race in Alaska.


And it reminds me just a little bit of this year where outgrossed Democratic candidate in the polls is kind of tied now, although it's mostly partisan polls. Right. Kind of tied now, maybe a couple of points behind Dan Sullivan, the incumbent, I guess the last nonperson poll of Alaska had Sullivan up four. And if the person polls have a tie, then that also would translate into roughly Sullivan plus for Alaskans to see. There's a lot of uncertainty.


And so it's a race to watch with about a twenty five percent chance of winning Upshot's Siena poll and the presidency in Alaska, for that matter. I mean, Biden is doing well, relatively speaking, in a lot of these very northern states, Maine, Minnesota. So we also have Biden with about a twenty five percent chance of winning the presidency in Alaska. So to follow up, we're going to get to more listener questions later about.


We did get a listener question on this, and it's from Dave. He says, Perry wrote an article on the Senate races in Montana, Kansas and Alaska. Right now, the model is showing Dems that thirty two percent, twenty six percent and twenty three percent chance of winning those states. Taken together, that's a sixty one percent chance Dems win at least one of those three Senate seats. Is this the proper way to interpret the model or should we assume they are correlated?


No, because they are correlated. They're not correlated as much as presidential states are. Senate races can operate somewhat independently from one another, but they are correlated for sure. If you look very carefully like our Senate histogram, the graph is a little asymmetric, meaning there is a tail where the polls underestimate Democrats. They win all these lean Republican races. Right, as well as all the toss ups. And then all of a sudden there are fifty five or something Democratic senators.


Right. So Republicans do need to worry about that. But at the same time, if everything holds and Democrats could lose all those tossup races, I mean, Republicans still have a 30 percent chance to win over 20 percent of keeping the Senate. It's not nothing. Yeah.


All right. So I do want to talk about the presidency, but are there any other Senate polls that have caught your eye since, I guess, Monday before we move on? Not really.


I mean, the thing about Senate races, too, is that the polling tends to be pretty stable and our model is pretty conservative about shifting those races around. You know, you get a poll, for example, in Michigan that showed the Republican, John James, only one point behind from Upshot. But then you had other polls, people like, oh, my gosh, is that race a tossup now? But since then, you've had several other polls showing Gary Peters, a Democratic incumbent, ahead by six to nine points, I should say.


So like people don't change their minds in these Senate races as much as, like for the presidency. And they don't change for the presidency much either. But you want to take a pretty long term look at the polls and not expect there to be some type of sharp turning points necessarily.


So when it comes to the presidency last week on Model Talk, you said that the state polls didn't quite reflect the 10 to 11 point race that we were seeing in national polls. Is that still the case or have we seen state polls catch up to the national polling since then?


A little bit. I mean, I'd say maybe there's a point worth of a gap, right, where it's more like nine or 10 points. If you extrapolate from state poll versus 10 or 11 points in national polls, I think some of the higher quality state polls maybe are more in line with the larger leaves for Biden. There's a lot of mediocre Republican leaning robo polls, Trafalgar, etc., that kind of get thrown in the averages. Right. And they tend to sometimes collectively knock Biden down by half a point or a point or something, if there are several of them at any given time.


I don't know. But also 10 or 11 points is pretty gargantuan lead. That's a very big win. So there's some maybe prior by which you'd be skeptical that he's really had by 10 or 11 points. Well, it's big, right? But clearly, Trump's not in great shape. Exactly.


Do we have any better sense at this point of what's responsible for this polling high for Biden last week?


Besides, so many things were happening at once. It's hard to disentangle them. I guess we're a week out from then. Now, or should we expect that the debate is not having any impact, that covid-19 is still having any impact? Of course, covid-19 the pandemic is. But I mean, his personal diagnosis.


I mean, we're outside the range where you call it a debate bounce debates two and a half weeks in the past now. Right. So you would think that if Biden has realized gains from the debate, then maybe now they're semipermanent, especially since Biden's number has come up more than Trump's has come down. You've had some undecideds who have now locked in or decided upon Biden with covid. It's a little hard to know if the president being out of the hospital will help him.


Doesn't seem like it's tightening based on that, but that's a little bit more uncertain. There are a couple of things. Number one is that Joe Biden has a pretty big edge in advertising, especially in swing states. And we think advertising is a little overrated in general. But if Trump is really running out of money, then maybe that helps put some wind in Biden's sails and adds a or so to his numbers and keeps that debate bounce from receding to the mean.


Also, we have a lot of people who already voted. How pollsters handle that is an interesting question. It's possible that pollsters, some people who are deemed to be unlikely voters now have voted. And when if you say you've already voted, then you could count as a likely voter who would have been excluded otherwise. So the early voting could have some effects on the polls where the pollsters are handling that correctly or not. I am not one to say.


I'm going to say I haven't researched enough to know. But, you know, but those are some alternative theories for why Biden is kind of held this lead.


You wrote about this this week. And it's something that some of our listeners noted as well, which is that for the first time on our winding path towards an Electoral College victory snake on the forecast, Georgia shaded blue for the first time, which doesn't mean that. It's now likely to be won by Biden, it's still essentially a toss up, but that for the first time in the forecast, the Democrat Biden was slightly, slightly ahead of Trump there.


Maybe the specific question here we got from Joseph, which is how surprising is that Georgia now seems likelier to go blue than Ohio?


I don't know.


I mean, we have the kind of the same. They both like as close to 50 50 as it gets. I think I would bet on Ohio if I had to pick between one of those two going to Biden. It's a bit more consistent with his strength elsewhere in the Midwest. There are barriers to vote in Georgia. There might actually does try to kind of account for. But but I know I mean, Clinton only lost Georgia by five points in twenty sixteen.


If you only lose by five an election where you won the popular vote by two and then Biden wins popular vote by 10, then you might expect Georgia to shift enough for Biden to win it by a couple of points. So it's not implausible by any means. Which is also a pretty inelastic state mean it doesn't actually swing around that much, in part because you have like a very solid base of Republican conservative evangelical white voters, a very solid Democratic base of black voters, plus Atlanta area multicultural younger voters.


But the GOP base is just slightly larger. But there probably are enough swing voters in the Atlanta suburbs to hand Democrats narrow victories if Democrats are at the top of the range with those voters.


So it's plausible and it's worth saying that Trump won Ohio by eight points in twenty. So actually, Trump did better in Ohio than he did in Georgia. Maybe these are two good states to watch in terms of how Biden and Trump are performing in your standard Rust Belt versus standard Sunbelt state on Election Day. And it's also worth noting that Democrats don't actually have to win either Georgia or Ohio to win the presidency at all. Right.


They do not have to win either state. No. And it's pretty unlikely that either state would be the tipping point. You have to get into some weird maps, right, where like if Biden loses Nevada and Arizona and Florida and North Carolina. Then you have some maps where Biden would need Ohio as part of a strength in the Midwest to overcome losing Nevada and other stuff like that and similar like if Biden got decimated the Midwest and lost Michigan and Wisconsin and Pennsylvania and Minnesota, then you could make up for that if you won Florida and Georgia and North Carolina.


So we're talking about some pretty farfetched possibilities, although keep in mind that, like, you know, to some extent now, if Trump is down to a 13 whatever percent chance of winning, then then weird things would have to happen for Trump to win. You would have to see some big shifts or for the polls to be we offer to win. And if there are big shifts or there's a big polling error, it may not be uniform. And maybe that in certain states something happens to Biden's numbers.


Although I would say, by the way, like probably the state that's like the worst affected by covid right now or one of the states is Wisconsin, which is not good timing for the president.


Yeah. Which was actually a lesson question that we got. So let's dig into a bunch of our listener questions.


We got a question from Sam and it's how have the recent rises in coronavirus cases affected the models covid adjustments? I know that there are not big adjustments based on covid in the forecast, but to what extent are there? How do they change everything? And then also, are we seeing in the polling I got if it's not in the forecast?


So the models doing so, some work in the background to try to figure out if you put an index of covid severity in a regression analysis, does it give you additional power to explain what the polling says? But do we see polling shifts in the immediate aftermath of hotspot in a particular state?


I have not looked at that detail. I mean, the best example might be something like Florida, where Biden was polling very well in June in the midst of their covid peak. And then the race tightened a bit after their declined. So there's probably some anecdotal evidence for it, but they were talking maybe a point or something. I mean, look, I would say the voters are not inclined to vote for President Trump as a result of his Kobad handling.


There's going to be a lot of news and conversations about Rising Kaskowitz, about Lockdown's. We learned this morning that it's very unlikely that a major U.S. company will seek to get FDA approval for emergency use before the election. Pfizer said today we're going to wait until mid-November. Other companies have said that to and or their trials are behind schedule, are on pause temporarily anyway. So the fact that, like we're not in an ebb and Komen case is heading into November 3rd is not favorable for the president.


And it's worth saying we actually have the numbers still. We're aggregating all of the polling on approval of President Trump's response to the coronavirus crisis. And he's out about negative 18 net approval right now. I think the worst he got was negative 20. The best he got was negative 16 over the past several months. So across the board, Americans aren't happy.


The next question we got is a question we got from a lot of people and is also something that the media has been trying to explain recently. Question is, what's behind the GOP's recent new voter registration advantage and what should we make of it for the election?


I think there are a few things. One is that the GOP has been doing more in-person voter registration efforts, door knocking and whatnot that Democrats had not done until recently, too, is that Democrats had a competitive primary. So they registered a lot of voters late last year or earlier this year. Three is that a lot of younger voters tend not to want to affiliate themselves with the party. So the registers independents, but they probably profile is likely Democratic voters.


They might be younger. Hispanics, for example, might tend to register as independent. They probably will not vote for Trump. At least most of them won't. So it's not clear that this GOP advantage in party I.D. necessarily translates one to one into actually having more Trump voters, especially because we know that the pool of independent voters like Biden is winning among independents by like 20 points in some polls. It's a really big edge. And the whole QAI Trump voter theory that polls are I mean, there are various factors that are more and less sophisticated.


But one version is that polls under sample Republicans and therefore they underestimate Trump. In fact, a lot of the high quality polls show Republicans in a good position in party identification, where they're almost tied with Democrats who have closed the gap. And despite that, Trump is losing by 10 points because Biden is winning independents by 20 points. So that really tells a persuasion story. Also, when you ask people who do you vote for in 2016, the upshot, polls do this.


They will find people that say, I voted for Trump in 2016 and this year I'm voting for Joe Biden. You're not a shy Trump voter. If you say I voted for Trump at 16, but not this year, or to say I voted for Gary Johnson in 2016 and voting for Biden this year or Jill Stein even. So, there's all types of evidence that, like Biden, is about persuasion and not turnout. Which to me is actually like a bit more robust to avoid polling error, because turnout can be harder to estimate.


You know what I mean? But if you have evidence of vote switchers, then that's usually pretty solid.


Our next question is more of a process question. And it is, is new polling data manually input into the model or does the system automatically pull data as it's released?


There are a handful of scrapers for some polls, but generally speaking, we we enter stuff in manually which shout out to our colleagues who do a hell of a job.


It's a lot of work. And, you know, we're kind of now the de facto polling collector for the whole industry and we try to be the most comprehensive of anybody. You know, if you're curious, it currently takes about 10 minutes to run. Forty thousand simulations of the presidential model and about 17 minutes to run. Forty thousand simulations of the midterm model or the congressional model. The congressional actually is more code, is more complicated, does more things.


So when you see a new poll pop up in our list of polls at like 1:00 p.m., say, and you'd expect it about 110 or 115, the presidential model would update and then 115 or 120, the congressional model would update.


Next question. While early voting isn't an indicator of final results, will late breaking events like a comi letter have less impact than they would in a normal non pandemic voting environment because so many voters are getting baked in early?


I mean, a bit, right? The question is how much? Because usually people who vote early are weighted toward people who are pretty firm partisans and have kind of decided who they're going to vote on. So let's say that 40 percent of people have already voted. Then a big event happens that shifts things back to Trump. Would that mean only has 60 percent of the impact? I'd say probably not, but maybe we'd only have 80 percent as much impact.


And so it does provide some insulation to bite. Look, there is no world in which it's good for Trump that Biden is making a bunch of votes when he's ahead 10 points. There's no world which that's good for Trump unless there's some huge rejection rate among these ballots. It may not be that bad for him, though. But, you know, I mean, in the event of a commu letter, I mean, maybe because that was Trump won by such a narrow margin in the swing states, maybe accumulator in twenty twenty.


If that comes at a point when 40 or 50 million people have voted, maybe a third of the overall turnout. Right. Maybe that'll be just enough to have saved Clinton in 2016. Maybe.


Katie asks, you've mentioned numerous times that the model assumes tightening in the polls. When does that assumption go away? And the model just uses the polls? Is that only on Election Day or some day prior to that?


I think it's actually the day before the election, the economic weight goes to zero. Let me look out for people kind of where we have this now. It's currently ninety one percent polls and nine percent are prior. So the prior doesn't play a huge role, but it does a little bit. So it means that as of this recording, we have Biden projected to win by eight point three points. So note that that assumes some tightening without the prior, he'd be ahead by nine point one point.


So why nine point one and not ten and a half or whatever he is in national polls? Well, again, that gets back to that state. Pulvers national poll gap. Our forecast of the popular vote is mostly based on state polls, and they're more consistent with like a nine or nine and a half point Biden lead.


We got a question from Aryal who says, What are the percentage point ranges you use for the different levels of certainty on the website? And Ariels referring to the favored, slightly favored, clearly favored demarkation that we assign to certain percent chances, actually have the list in front of me. So I can clarify that for Aryal.


So from fifty to fifty four percent chance, we say it's a toss up from fifty five to sixty nine. We say slightly favored from 70 to eighty nine percent chance we say favored. So that's where we are right now for Joe Biden for clearly favored. The next indication that we have not actually gotten to yet, it's 90 percent to ninety seven percent and the ninety eight or above is very likely to win the election. So we do think we're going to get to either clearly favored or very likely to win the election in the next two and a half weeks.


I mean, so Biden will keep gaining if the polls hold steady in the percentage terms, right. Because it means there's less time for Trump to make up ground. Yeah, I don't know. I mean, in theory and a well calibrated forecast, then the percentage that you have in any given day is also the best prediction of the percentage that you have the next day. So in theory, you would predict that Biden will say eighty seven percent because the race would tighten a little bit.


You know, in practice, I don't know. I mean, in practice we can get into questions of what could do to help facilitate a kind of back. Again, I get back to like ten and a half points is such a big lead relative to even where this campaign has been. Even relative to 2008 with Obama, where he won by seven, 10 after such a big lead that like, I don't know if that's easy to sustain, although who knows, you know, maybe Trump is in a proverbial downward spiral.


I don't know that's plausible, too. But the model that we go with, our focus, our focus is that Biden is going to win the popular vote by eight and a half points and not 10 and a half points. So it does kind of assume the race would tighten a bit.


We've got this question a couple times and it's which party is favored to control the most state delegations in the House. And people are asking this question because if there's a tie in the Electoral College, it goes to the House and then the votes are cast, according to state delegation majority, not based on individual House members. So essentially, the question is, are Democrats or Republicans controlling six plus state delegations? And I should say there's a less than one percent chance of a tie in the Electoral College, according to our forecast right now.


But do we have an answer to this question regardless?


We do so. We currently show a twenty five percent chance that Democrats control the most delegations, a 25 percent chance that Republicans control the most delegations, and a 50 percent chance that there's no majority, that no one is twenty six, it's actually gone up for Democrats. So right now, I think Democrats Republicans control twenty six. Right. So Democrats have to win back some delegations. The easiest way to do that would be to win the Senate races or the House races in Alaska or Montana, which are competitive.


And when there's just one seat. So Democrats have gained in this calculation, it looks like a bit. But keep in mind that like this is conditional in a world where Trump is having a really bad night and loses the popular vote by eight or nine points. Right. And what not if the race tightens, which is a scenario under which the delegations becomes relevant. Right. If you have an Electoral College deadlock and they need the House to break a tie or interpret competing slate of electors, then you have a vote by delegations.


If we have a tie, that probably means that we end up having a much more Republican night than we expected going in. So conditionals near where the delegations matter, the Republicans would probably still have an edge, but it's a reason to watch Alaska, Montana. And if you go to download the data, if you go to the House national toplines file, it now contains data on the projected number of delegations per party.


All right. We just got a couple more questions, I should say. We got, again, hundreds of listener questions. So thanks to everyone who sent them in, we try to get to as many as possible. A lot of these have longer answers. The question is what impact doesn't mean.


Adopting rank choice voting have on the model both for the presidency and for Congress.


So it doesn't much affect. A lot of polls in Maine will use a version where they kind of either simulate the ranked choice themselves or just ask people who they vote for in a two way race. So we use that version of the questions in Maine, where they're testing basically a two way match up instead of with third parties included. We have some code. If you had a case where like it was actually unclear who the top two candidates would be like if an independent, it was a true three way race.


Right. We have some code to handle those things, but it's very, very unlikely that you're going to have a third party actually in the top two in any of the main races. So therefore, we basically just use kind of head to head polls in Maine to simulate the effect of a reallocation of votes. All right.


Sam asks, Have you ever considered trying to model how soon we will know the winner of the election? There are so many variables in that question that I can't even imagine trying to begin forecasting that we have thought about.


Like, you know, there are certain states who are thought to be kind of quick counting states like Florida, Arizona, North Carolina. Right. We thought about like, can you designate a set of quick counting states? And we might we might add something on that. I'm not sure they're going to quite as many bright lines necessarily between fast and slow states as people would want. But we're thinking about trying to have versions of that calculation.


All right. Last question on this question. Do you have a drink of choice for getting through election night through election night itself?


Well, I have to work, so I can't be drinking the booze that I would want.


Well, I feel like it's a getting through election night. It's like coffee or red ball getting to sleep on election night. It's like whiskey. Vodka. Are you a beer guy or wine guy or a liquor guy getting to sleep after a wild election night?


I mean, you usually have a beer guy, but after the election, I'd probably be whiskey or something. Yeah. All right. That's a good note to end on. Thanks, Nate. Thank you, David. And I should also mention big news. We are going to be doing daily podcasts basically from here on out, excluding the weekends for the most part. Maybe if something happens, we'll have a weekend podcast. But essentially from next Monday when we have our regular Monday podcast through Election Day, you're going to be getting podcasts every weekday and then, of course, perhaps some on the weekends.


We'll see. Hopefully not. But get ready for your daily podcast.


Lots of lots of content, daily podcast folks suffered through these final weeks with us.


There we go. My name is Gail and Tony Chao is in the virtual control room. Claire Bita Gary Curtis is on audio editing. You can get in touch with us by emailing us at Podcast's at five thirty eight dotcom. You can also, of course, tweet 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 tell someone about us. Make sure to subscribe on YouTube as well.


As you may know, this episode of Model Talk, along with most of the podcast we do, are also on YouTube. If you would like to watch us from I guess for me, it's my bedroom. For you, it's your library. Talk about all of this fun stuff. Anyway, thanks for listening and see.