Editor's Note: This transcript was automatically transcribed, so mistakes are inevitable. You can contribute by proofreading the transcript or highlighting the mistakes. Sign up to be amongst the first contributors.
I enjoy dolphins and whales, so I thought I'd wear this, save the dolphins, but now I feel bad that I didn't wear my orchestra. We could have been under the McMillion team for a little talk today. Is Mamoulian a word? Sure. Hello and welcome to the 538 Politics podcast, I'm Galen Drake.
I'm Nate Silver, and this is my little talk, perhaps the final little talk of the year, a pre-election.
Well, we'll probably have a model talk once we have all of the results. And we may even have a model talk for the Georgia runoff. But this is definitely our final free November third bottle talk. So it's a big day for days until the election. How's it going at the silver household? You know, I wouldn't say I'm lacking for anxiety, I just think I target my anxiety more precisely.
I got a text message yesterday from your partner asking why we haven't interviewed five Fox on one of our daily podcasts. And I asked him if he had a five Fox costume we could use for this so that we could finally interview five. And the answer was no. So my plan for today was to try to get you to wear a Fox costume, but I failed.
Yeah, I mean, what would five sound like is something that we've talked about before. Do you have an idea in mind that can you do the voice? I can't do that. Hey, it's fine.
Fox Hey, it's fine. FOX We estimate about those polls.
All right, Ferreti. Well, I mean, Halloween is on Saturday. So if you're going to be anything this year, I would think you would be.
I mean of the various things. Right, various things get ruined for me because they're timing with elections, Halloween, fall foliage, the World Series.
Oh, you want to talk about things that get ruined because of the election? My birthday is on Thursday, so if we don't have results by Thursday, I'm going to be out of luck on that one.
But if we do, you can feel less stressed than you have in a long time for the past two years of this election coverage. That's true. So let's talk about the forecast model, which is what we're here for as we sit down to tape this. The forecast currently shows that Biden is favored to win the election with an eighty nine percent chance. That's an all time high that he reached yesterday. Democrats are favored to win the Senate with a seventy seven percent chance, and they're also favored to maintain control of the House with a 98 percent chance.
So those numbers have changed a little bit over the past weeks and months, but they haven't changed all that much. So over the next four days now, how much do you see those numbers changing before Election Day?
I'd be surprised if they changed that much with one pretty important exception we'll talk about in a moment here. So keep in mind that a lot of the polls that we're seeing now are what would be considered final polls. They'll figure into the final real clear politics and fire through the averages. We look them up years later. They're within this window for a while, within twenty one days, the election, they go in our pollster ratings. But like, it's not like I don't think, you know, Quinnipiac pulled Florida, Pennsylvania, Ohio, Iowa.
I don't think they're going to pull those again before Election Day. Maybe they will. That would be a lot of interviewing to do, kind of back to back. Maybe they will be nice, but we're kind of seeing half the final polls now. And we have a pretty clear picture of what the final polls will say. Now, if there were some Last-Minute change in the race, then, OK, maybe you'd missed that in the polls. But there are reasons to think we won't have a big change.
No one being that nothing really in the news cycle has changed. I mean, if anything, the fact that we have this record setting covid cases might be bad for Trump, if anything, but not that much changing the news cycle. Half the country roughly have already voted, if not more. So I wouldn't expect a lot to change. You said that there was one caveat with a couple caveats. One is that if Biden holds his current lead, he still has a little bit more room to grow in our percentage forecast, because even though maybe this time we don't have any reason to think the race is tightening, but it's theoretically possible they could tighten at this point, 2016, the Comey letter was tightening the race.
So if we make it to November 3rd and there hasn't been further tightening, then that will help Biden a little bit further. But there is the other important caveat, which is we have not gotten very many final high quality polls of Pennsylvania right now. Biden leads there by I think once we're taping this five point one or five point two points in our Pennsylvania polling average, if he were to get to six or if that were to fall to four, that would make a decently large difference because Pennsylvania is a tipping point state, a very high percentage of our time in our average, like the final New York Times CNN Upshot poll, which can have a very large sample size apparently in Pennsylvania.
That will make some difference. If that's like a plus nine Biden or a plus three, then that will affect our odds. A decent amount potentially.
Yeah. So that's looking forward to what we could expect between now and November 3rd. But looking back over the past week or so, we have seen some tightening. So can we talk about whether that's been the case? I'm not sure I agree with that. You don't know? From what I've looked at, our forecast has changed a little bit in some of the battleground states. Right. We've seen a little bit of tightening in Florida with Trump's odds of winning that state, increasing a little bit.
We have seen Trump's odds increase slightly in North Carolina and Pennsylvania and then in Georgia and Ohio. It looks like Biden is doing a little bit better. So maybe overall in the national polls, we haven't seen that much tightening. But it seems like there has been some shifting in the battleground states that will decide. The election, well, shifting and tightening are different things, you have some different states bouncing around in different directions, but like I would not say that those clearly directionally favor either Trump or Biden.
OK, so let's unpack that a little bit, because my impression was that Trump was favored more in Georgia and Ohio and now it's moved a little towards Biden. So that looks like tightening. And then Biden was favored a little more in places like Florida and Pennsylvania now and now. That's a little more towards Trump. We don't have to get into so much of a conversation about tightening or shifting or whatever. How do those changes in those battleground states affect how we should be conceiving of the competitiveness of this race?
Well, look, one thing that's interesting is that we've kind of lost a little bit of the hierarchy of importance of states behind Pennsylvania. On the one hand, there's not exactly a clear second best backup plan for Biden. If he loses Pennsylvania, it's probably Arizona. Although Arizona does not have as many electoral votes as some other states, Florida has a ton of electoral votes. If Trump loses it, then he's in deep trouble. But the has been close in Florida, a two point lead for Biden on a four point lead or something.
George actually is polling very close to Florida. North Carolina is right in the middle of that pack, too. You know, on the other hand, like in Wisconsin and Michigan, Biden has leads that look, frankly, somewhat comfortable. And therefore, maybe it is about like Pennsylvania or some other states to replace Pennsylvania. And Biden doesn't have to worry as much about Wisconsin and Michigan. Other hand, you never know. I mean, those states the polls were off last time.
So in our tipping point index, Pennsylvania, where there is no one with a bullet, it's a tipping point state like 40 percent of the time. And there are a bunch of like six or seven states that are grouped together in the kind of 7th through 8th rank.
And so just to clarify, some of the polling there in Pennsylvania right now, as you mentioned, Biden has about a five point lead, whereas in Wisconsin and Michigan, he has closer to a nine point lead. I think actually in both states right now, according to our polling averages, his lead is at eight point six. We kind of clumped all those states together in talking about the upper Midwestern path to a Biden presidency. Is there any sense of what's happened that led to Michigan and Wisconsin breaking away a bit, Pennsylvania staying a bit tighter and being actually closer to states like Arizona and North Carolina?
I mean, Michigan has been polling pretty well for Biden all year. It's kind of like a change state. You know, they can be pretty anti incumbent. It can swing around a lot in a big Democratic year in Michigan in twenty eighteen. The voting laws as a result of that are more favorable to having more people vote now. So Michigan kind of has been pretty steady for Biden. He's also been in fairly good shape in Wisconsin. But like I think it's this covert outbreak.
I mean, it's really bad in Wisconsin right now. And any chance Trump had to close? Well, in Wisconsin, I think is undermined by that very much so.
Considering everything we've just mentioned about these key states, how would you rank them right now in terms of their favorability for Biden or Trump? I mean, our forecast does this to a certain extent, but what are you thinking about? Oh, if this state has gone for Biden or Trump, it seems like the race is over, you know, thinking about starting to process returns on Tuesday night.
I mean, Trump has to kind of run the table. Trump's map would involve winning the following states, most likely Pennsylvania, North Carolina, Georgia, Florida and Arizona. Those are all states where he's currently an underdog in our forecast. And he has to win all five in exactly those five because Wisconsin is further away from the tipping point, then it's unlikely that he could win Wisconsin and not Pennsylvania. Also fewer electoral votes. So he's got to win again, Georgia, North Carolina, Florida, Arizona, Pennsylvania.
If he wins those five, then he is very likely to be a two term president. If he loses any of those five, then he is likely in a lot of trouble.
How much of the forecast right now is relying on strictly polls versus some of the fundamental data that we've talked about throughout the past several months, like the economy or the remaining days until Election Day?
So now it's only three percent of the forecast is based on the prior at this point that will go down to zero by the weekend. So that boosts Trump by it looks like about two tenths of a point. So if nothing changes, then we entirely remove that prior, then Biden will gain two tenths of a point in the average state. Not very much.
And in terms of how much we use within states, just this question of like. So if a state doesn't have much polling, we also have other methods to compute what the polls there say from other states that also we don't need to do very much of now because we can rely purely on the polling in a particular state, because we have enough polling in most swing states.
Now, you mentioned early voting, and that's something that we've received a lot of question. I don't want to dig a little deeper into it, so here are some updated numbers. As of Friday, the number of early ballots cast is already nearing. Eighty five million, and that's more than 60 percent of all of the ballots cast in 2016. We only have party registration data for those voters in some states, but according to the data that we do have, forty seven percent of those ballots have come from registered Democrats, 30 percent have come from Republicans.
Twenty three percent have come from unaffiliated voters. Now, we've repeatedly cautioned people that historically early vote data like that does not provide an indication of who will go on to win the election for many reasons. One, because you may be registered with a party, but you may not be voting for the candidate of that party. And that also Election Day votes count just as much as any early day votes. And so if one party is behind, they can always make up that ground on Election Day.
However, given the record breaking level of early voting that we have seen, should that change our priors at all? The priors being what I just mentioned?
Again, to a first approximation, I think just. Don't pay much attention to early voting numbers because it's reflected in the polls and the polls give you a lot of information in the early voting totals, don't. But those show is voting by party registration in some states that have registration by party, which some states don't. Party registration is not the same thing as your party ID. If I ask you, are you a Democrat or Republican, maybe you're like, well, registered as Republican once 20 years ago to vote in this primary.
But I always vote Democratic now. So you might be registered as a Republican, but ID as a Democrat. Also among people who are voting early, the independents that are voting early, for example, are much more pro Biden than the independents overall. And the Republicans, for that matter, are more pro by not many of them, but more than in the Election Day voting pool. So I think people make too much of this early voting data when you have polls to look at to directly answer these questions of candidate preference.
With that said, I mean, clearly we're not going to have a low turnout election. It would seem like, obviously, some of these votes from Democrats and Republicans are going to people that would have voted on Election Day, but are also a lot of first time voters in this early voting data, substantially more than there were in twenty sixteen.
By all polling indications, both parties, but especially Democrats. In the latest Gallup polling, Democrats are more enthusiastic Americans about voting. Those are pretty reliable indications. If you're going to pull people and say what's your enthusiasm level relative to the past? If you say more enthusiastic, then that's a pretty reliable indication of higher turnout. So we're very likely to have higher turnout in this election. And it's probably not bad news for Democrats. Usually higher turnout helps Democrats.
It's a little bit different now in a world where a lot of lower propensity Trump voters, but at the very least not likely to have a skewed turnout where only Republicans turnout and Democrats don't. And that's one of the conditions Trump might need to win. Like this happened in 2014, for example, where turnout was very low but really, really low among Democrats.
Right. And so Republicans kind of won that midterm based on turnout this year. That's not going to be the way they if they win the election because a lot of swing voters wound up sticking with Trump in the end. Right. Or something went wrong with a lot of male votes being thrown out or the polls would be so far off that like we'd have to do an autopsy. Right.
I don't think low Democratic turnout is the way that Trump will win. There's other ways to win, but probably not that one. I mean, zooming in on Texas in particular, as of this morning, Texas reached one hundred percent of its total 2016 vote in its early vote alone. Early voting in Texas ends today. So, of course, we'll also get more ballots in on November 3rd. How should we process that? I mean, is it the same lesson that you just mentioned from all of early voting in general that just there's a lot of enthusiasm to vote and that it's more of an indication that the polls might be correct?
Or should we think? Well, Texas is historically a low turnout state and we know that a lot of the low propensity voters may be more likely to support Democrats than Republicans. And so in a high turnout year, that specifically seems good for Democrats in Texas.
Would I say what we found is that when you have a big change in turnout, then polls are less accurate. Right. Which makes sense intuitively. If you have a bunch of new people, then maybe your turnout model from the past is going to be inadequate. We don't know that all those new people are Democrats. Remember, you know, people have gotten like a little bit enamored with the Texas purple thing. I mean, Democrats still haven't won like a statewide race there in a long, long time.
So you can have much higher Democratic turnout. And that's why it's a one point race. There are two point race there. That doesn't mean Biden will win. It would just be he might lose by two points instead of 10. Oh, for sure. Clearly, Texas is a place where I think the polling error might be a little bit higher than our model, indicating our model always has like a lot of uncertainty. Right. But I think the range or range might be too narrow in Texas, which would mean maybe you wind up with a Trump landslide after all, or maybe you end up with a Biden win.
It's a hard state to poll. Yeah, I mean, I should say a large part of the purple Texas story up until this point has not relied on turnout as much as persuasion because the highly educated suburban areas around Dallas, Fort Worth and Houston have seen massive swings amongst people who were already regular voters, as opposed to just incorporating lots of new voters into the electorate. So this new data kind of not necessarily contradicts that, but as a different trend than what we saw in twenty eighteen, which was essentially Romney Clinton voters turning out and electing Democrats to the House.
So of course, we will see what happens there on Election Day. I'm sure there will be a lot of coverage of taxes one way or another. Before you have to listen, questions I want to dig into one more hot topic of the election. And it's also something that we got plenty of questions about. So I've asked you about QAI Trump voters before. We've talked about it on this podcast four years ago. Now we're going to talk about it again.
And this time we have to talk about it because Politico just published an article. All about pollsters missing Trump voters and it's called, quote, People are going to be shocked, exclamation point return of the Scheid Trump voter question mark.
And the Politico article speaks with pollsters from Trafalgar Group, which we've talked about on this podcast, and USC Dornsife, which was also talked about on this podcast. Now, in 2016, Trafalgar showed Trump winning an upper Midwestern states that he did go on to win and USC showed Trump winning nationally by three points, which was obviously wrong because Clinton won nationally by two points.
I don't know why the USC guy is giving credit as the man who saw it. I mean, that's so stupid. Like the only poll, the national popular vote. And they were one of the worst polls, the national popular vote. And he does say that in the interview he corrects them. But I mean, that's kind of just utterly bullshit and kind of tells you something about like, look, there are reasons why Trump can win 10 percent chance is kind of high.
Think about how important the election is again.
Look, I don't think it's my job to go and bat down every probably drum theory because there's an outside chance series and are probably really dumb are true. They're probably not true. But they could be they're not that dumb. Right.
Dumb, but not that I'm sure people are wrong sometimes. I don't think there's much evidence for Trump voters, but like the polling error is a thing. Sometimes it's systematic, sometimes it switches back and forth when you're the next. Sometimes it happens the same direction a couple of times in a row. So Trump's got a 10 percent chance. I don't think it's going to be because of QAI voters, as this article necessarily talks about. I mean, if anything, the evidence mounts against this theory.
Let me give you a couple of problems with it. Number one, we have tangible evidence that Democrats are engaged in the form of early voting numbers and in the form of money. So while you wouldn't necessarily usually put a lot of stock in early voting numbers, if you're claiming there's going to be some massive red wave, well, now, A, that wave has to come on Election Day. And, B, it has to be really, really large because we know that a lot of Democrats have already voted so early voting data.
And I wouldn't really have it change my view of the race very much. But it might give you a reason not to think about the side Trump voter stuff. The other reason to be dubious about this theory is that Democratic Party ID, despite the early voting data, is actually not that high in the polls. The pollsters are not saying, oh, Democrats are going to have a plus nine edge nationally and party ID if it was last time like plus five plus six.
In fact, some of them show a smaller edge. The reason that Joe Biden is winning in these polls, because he's doing very well with independents, because he is capturing a lot of people who said they voted for either Trump or Gary Johnson in 2016. When you ask people who do vote for 2016 for Trump, we vote for this year, Joe Biden, not a lot of people, but there are people that say that you're not a shy Trump would if you admit to having voted for him four years ago and say you're voting for Biden now.
So this is more of a persuasion election than a turnout election despite the early voting data. Like I said, pollsters are not projecting a particularly Democratic turnout. They're just saying that like independents are going for Biden by 15 points or something. Also, there are a few more crossover Republicans than crossover Democrats. Not a huge split, right. But Biden is doing a very good job of holding 95 percent of the Democratic vote, whereas if Trump gets ninety one percent of the GOP vote in 92 percent.
Right. That matters a little bit. And you kind of do the math and that kind of all adds up to Biden plus nine points or whatever international polling average.
So I want to read a couple quotes just to get you to react to what their actual theory is here. So USC's pollster, Ari Captain has these questions where he asks about who the respondent is planning on voting for and then asks who is your social circle planning on voting for? And they said they get a 10 point lead nationally for Biden over Trump when asking about personal support. But if you look at the social circle question, Biden gets only five or six points.
So what do we think about that? Well, that's fucking stupid.
You know, I it's stupid is because we have this media narrative for four years about hidden inside Trump support. So people are reflecting the kind of narrative they've heard from every major news outlet for the past four years.
All right. So one other quote in here that I want to get your reaction to the Trafalgar poster size. Quote, We live in a country where people will lie to their accountant, the lie to their doctor, they'll lie to their priest. And we're supposed to believe they shed all of that when they get on the phone with a stranger.
First of all, how to know there also aren't shy Biden voters. If you're like a shy Biden voter in a community where expected you'd support Trump. They're also evidence that some African-American and Hispanic voters are not thrilled about necessarily revealing who they're voting for. If they're on the phone with someone who they perceive to be white, which a lot of interviewers are. So you're saying they're asking voters? I'm saying there's no reason to think there are more shy Trump voters and shy Biden voters.
Also, you can look at other countries, too, where you have candidates that are more explicitly nationalistic than Donald Trump in the next couple years ago about. Every kind of European election with an identifiable right wing nationalist party, there is no tendency over 30 elections for the nationalist right wing politically incorrect party to, I shouldn't say, politically incorrect, you know, often openly racist.
There is no tendency for parties like that to outperform their polls. Look, it's just like quixotic thing, right?
People kind of claim the existence of it, but there's no evidence for it, you know what I mean? And it isn't necessary to explain what happened in twenty sixteen. I mean, we happen in 2016 is that you didn't have enough education waiting that Trump won the undecided voters, which he may not do this year. And even then the national polls were pretty close.
So at some point it's the Loch Ness monster, these Trump voters, you know what I mean?
At some point, like the fact that Trump narrowly beat his polls in 2016 is not sufficient to prove the existence of Trump voters. And by the way, 2016 is not the only election, the election that Trump is on the ballot right in the primaries. Trump did not beat his polls. By and large, you may beat expectations, but he did not beat the polls. In fact, he underperformed relative to Ted Cruz in some states. In twenty eighteen, there were some claims of shy Republican voters.
Trump was not directly on the ballot. Twenty eighteen. The polls were spot on. So, like, you know, I don't know.
I mean, parvenus like our human desire for, like, explanation to tie things in the package. I mean, sometimes the polls are wrong. If they're wrong again this year, we don't know we're going to be wrong in the Biden direction or the Trump direction. If they're wrong in the Trump direction, again, it may be for entirely different reasons than they were wrong in 2016, because it's the reasons people don't anticipate. Because if you anticipate a reason, then you'll probably try to correct for that ahead of time.
Maybe that's do with male voting, for example. I don't know. And on top of that, there's no real evidence that shy Trump voters whereby the polls are wrong in 2016. So I don't know. I mean, I think some pollsters are like they have an intuition or they have a preference or gut or whatever else, some murky ground combination between all of those. Right. And they kind of see confirmation bias for it everywhere. By the way, there is some evidence that, like when you asked people about like what are your attitudes on Black Lives Matter, then there actually might be some Moad effect where if you ask a question anonymously, you may be more likely to say, you know, I actually don't I don't know about these protests.
However, those same polls find you don't find that effect when it comes to Trump. And I think a lot the reason why is like there are lots of reasons why you might vote for Trump. Let's say that you actually want to vote for Trump because you hate immigrants and you think he also hates immigrants. If a poll asked you, why are you voting for Trump, you say, I just want lower taxes. I'm a Republican. You know, I don't like the mainstream media.
They give him too much crap. Oh, I just want continuity. Oh, the economy is improving so you can give, like, a million rationales for voting for someone. It's often hard to tease out the why, but like people usually don't lie to pollsters about who they're voting for. And to the extent they do, there's no particular indication they're going to lie in one direction disproportionately.
All right. So I think I can say that we have comprehensively responded to any questions about that Politico article, and we probably won't be talking about QAI Trump voters again on his podcast, at least before the election. We've addressed that, I think, thoroughly at this point. So let's get to some more specific listener questions. As usual, we got a lot. So I'm going to do my best to get through all that we can. First question blunt one from Jake.
Maybe we can answer this or even just a yes or no. Jake asks, Do national polls have any use at this point? Not much.
National polls are primarily useful for inferring trends. You don't have a lot of recent state data, but in most states, we'll have plenty of polling by the end here.
So national polls are kind of kind of boring. Next question. What does it take for a pollster to be banned from 538 averages that we know or suspect that they faked data or they've committed other gross ethical misconduct?
And do we have any indication that there will be newly banned pollsters after the twenty twenty election, not any of the high profile ones?
I don't think. I mean, we always rethink our rules. And there have been some polls that we think are kind of taking advantage of some of the assumptions we make about how civics and good faith and some stuff like that. That's a different category, though, than a pollster faking data. We do have some issues where we have pollsters that. Have not necessarily disclosed who they're doing a poll for, and that is actually a fairly big problem, we're going to have to introduce a kind of policy interpretation around that, which we'll explain on the site at some point.
But it's not the same as like banning or a degree of severity that would merit banning. Walt, as I know that state results are likely correlated. I assume that polling errors would be as well. And in fact. Well, yes, they are. And we've talked about this a lot after 2016. That's part of why our forecast model gave Trump a better chance of winning the election than our forecast models did. But Walts question is, is there a breakdown of which states are most likely correlated?
And yes, there is. So I guess maybe the question is, which states are most correlated in the forecast, maybe most particularly referring to of the battleground states?
I mean, it's usually the ones that you would think are intuitively correlated, you know, Michigan and Wisconsin and then also Wisconsin, Minnesota, Maine and New Hampshire. We actually have files you can download where we show you the results of forty thousand simulations and you can kind of run the correlation matrix on your own and see which states are more correlated. It does vary a lot like our model assumes that polling error primarily occurs in the form of correlated demographic and geographic errors.
So, oh, Clinton did better or worse than we thought in the upper Midwest. Right. Means a whole bunch of states together, kind of a clustered error. So you'll see those interesting clusters there. There's some interesting behavior in extreme outlier cases. You can even see states that we assume are slightly negatively correlated, like, I think Washington State and Mississippi, because they're so different from one another. But you also see states like Minnesota and Wisconsin that are positively correlated.
For the most part, polling areas are strongly correlated. People like to look at our maps. Right. And they'll see interesting things. And sometimes they have a map where like I'm looking at a map here, we're like Montana, North Dakota and South Dakota and a Biden landslide. I'll go for Biden. That makes sense because we have correlated polling areas and all those groups of states, I think in this map like Kansas and Missouri, also go for Biden.
It makes sense that a cluster of states that are geographically linked, if there's a massive Pullinger and Biden wins by 16 points, it makes sense that you have a correlated group of states that are having similar things happening, behaving together. What about in the South?
I feel like we have more quirky explanations for the way that Florida moves politically versus Georgia versus Texas versus North Carolina, for that matter. I mean, are those Southern states a less correlated than some of the northern states that we talk about?
They should be. I mean, each southern states a little bit different in a way that I think is more true. I think in Florida and Georgia are more different than Wisconsin and Minnesota, for example. You know, I mean, they're two different things here. One is, are the polls moving the same direction, the same states in the south? It's kind of been no. Right. And like Florida, there has been one of the states where Trump has been regaining ground and tightening.
But Georgia has been one of Biden's best states recently. Does that make a lot of sense?
I mean, I'm not sure. I'm not totally sure. Yeah, I think there has been like a lot of late expenditure in Georgia by the Biden campaign, and maybe it's driving the numbers up. But, you know, there's also a fair amount of noise. And because, like, there's a fair amount of pollster hurting, especially late in the year, sometimes you see trends which wants to publish the same numbers, play it safe, and maybe something else is going on behind the scenes.
Actually, after we talked about hurting last week, we got a new question asking regarding hurting. Does the model account for that at all? And maybe put more emphasis on the polls a week or two before the election than the ones right before the election, since they may be hurting so implicitly is you know, we were in lots of tests to say, OK, how do you weight a polling shift at different times of the year?
The model is more aggressive about polling changes at the end than earlier on. But there is just a balance between being too aggressive and not aggressive enough. And like we just ran a lot of simulations to optimize that balance, basically.
But yeah, I do think, like, you know, remember in 2016, there was like a little bit obviously the most important movement was like post comi letter toward Trump. There was a little bit in 2016 of a kind of dead cat bounce for Clinton, where she recovered from two and a half down in national polls to three and a half down to three point eight or whatever it was, which kind of proved to have gone in the wrong direction.
You know, empirically, more often than not, if there is a late shift than it's telling. But, you know, it's easy to recall the times when that wasn't true.
We got a lot of questions that play on this theme generally. And I know you don't necessarily like this question, but I'm going to ask it anyway, which is do you have any thoughts about what kinds of polling errors might exist at this point in the race?
And of course, if you knew them, you would tell the pollsters and they would be able to adjust for them. But just from being a highly data literate, savvy observer of polling in this moment, do you have thoughts on where you're curious about pulling?
Er, you know, I mean, how many non. College educated Hispanic voters, are we getting on the phone in these polls? There have been polls that underestimated Democrats in the Southwest, in the West in recent years. And one issue with polling is that among white voters, higher education levels correlate with being Democratic. So this is kind of the infamous missing white working class Trump supporters among Hispanic voters. It's the opposite, the college educated, well-to-do Hispanics.
That's a pretty big swing vote where as Hispanics who may speak Spanish at home, who are younger without college degrees, may vote Democratic in bigger numbers and they may be harder to reach on the phone. So that could create errors.
Again, as much as I think probably pollsters have corrected their problems, saying you can't be sure about that. You know, a few pollsters still don't wait by education. They were a few problems in twenty eighteen that seemed they were a little 2016, like in other states there weren't. Right. I mean, I put it like this. If you're a pollster and you have Biden with a nice solid lead nationally or in a key state like Wisconsin or whatever.
And there's kind of some indication that maybe you're underestimating Democratic turnout. Do you care that much? So your poll has Biden up seven instead of 11 or something? I mean, maybe not. I don't know. So it could be that this early voting surge, plus various indications of Democratic enthusiasm, are not being picked up enough. I mean, there are some polls there's one poll from like Susquehanna, which has been producing very Republican leaning results that will actually ask, which past elections have you voted in?
And if you haven't voted recently, you're thrown out of the poll. Even if you voted this year, even if you said I already in this year, sorry, you're not even likely voter by Susquehanna. So there's some stuff like that that is probably a bit stupid along this line.
We got a question from Brad who asks, Which would you consider a bigger mess, Trump winning the popular vote or Biden winning the popular vote by 15 points?
So Trump winning would be a bigger mess. It's close, right? But we have Biden forecast to win the popular vote by eight. So we're talking about eight point error if Trump wins versus seven point error if Biden wins by 15.
All right. There you have it, Brad. Let's ask this question from Jenna. She says, Does the model account for different state rules or court rulings regarding mail in ballot deadlines? If not, how large of an effect should we expect from the recent rulings to have on the outcome of the election?
The model does not. I mean, we account for overall ease of voting in a particular state that goes into the prior that our model uses. We are not accounting for the late court rulings. Look, if you start to do the math on this. I'm not sure that it makes as much difference as people are assuming, so let's look at this, for example. According to Dr. Michael McDonald, U.S. Election Project website, so one thing to consider, first of all, although the Democratic vote has been very strong in the early vote so far, the remaining votes are not particularly Democratic because Democrats sent their vote in early or drop them off.
So of return mail ballots. Democrats have a twenty four point edge. They lead 50 to twenty six. A ballot have been returned in registration of outstanding ballots that have not been returned yet.
It's only a thirty seven to twenty six, only a plus 11 that's in party registration. You can imagine various reasons why the outstanding ballots are either more or less Democratic than those numbers indicate. Let's go with the plus one for now. So it is a plus 11 and mail ballots would say that in Pennsylvania in 2016, about one percent of ballots arrived after the deadline. Now, Pennsylvania is a state we actually can't have your ballot right after the deadline.
But let's say that you couldn't anywhere in the country. So take one percent of mail ballots, which is maybe 50 percent of all votes cast. Let's say multiply that by an 11 point edge for Democrats. And it works out to let me do this here 11 percent times, 50 percent times one percent. So it works out to affecting the polls by point six percent, which is not a lot, I mean, it could make a difference in like a Florida style two thousand margin.
I mean, again, obviously some degree of handwaving here, maybe maybe two percent of ballots or three percent get thrown out and it adds up a little bit. But like the fact that Democrats have taken the message and either sent in their ballots early or using Dropbox's for them is pretty important. And people are missing that. I mean, there could even be like some irony where because Republicans procrastinate more and or are not as concerned about the vote not getting counted, that maybe these court rulings actually wind up hurting Republicans.
I don't think that'll happen. But it's within the realm of possibility that, like the mail to trickle in are quite different than the ones that voters sent in two weeks ago.
All right. So hopefully that answers János question and also some of our other listeners questions about how significant or impactful these court rulings will be on who actually wins this election. Next question comes from Sergio and he asks about voting power. The question is, which voters in which states have the most power to sway the election?
Well, if you get all that raw data, I think there's something called the voter power index in the output that we show you. So I'm sure it's states like Pennsylvania, right? It's kind of all the obvious states. It's also states that have a small number of electoral votes, but individual voters have a fair amount of power, like Nevada is a relatively small state where probably the tipping point state is enough that electoral votes, but still bang for the buck wise, you know, Nevada or like the 2nd Congressional District of Nebraska.
It may be true that a voter in Omaha is the most powerful voter in the country right now, maybe Omaha along with Pennsylvania. But that tends to rank pretty high because that one electoral vote there can matter quite a bit.
And New Hampshire as well. Right. Although it's not that close in this election.
But, yeah, New Hampshire is quite pro.
Biden in polls now. So that may not be as much of a swing state as usual.
So congratulations, Omaha.
And I don't know what are people from Omaha, Omaha, O'Mahoney, and how do you say that? Where people from Oklahoma, Hine's, Omaha, Omaha.
OK, congratulations, Omaha. You're the most powerful voters in America. So we got a question about conditional probability.
Essentially, a lot of people have asked, like, hey, if Trump wins Florida, then how does that change our forecast model? And we're not going to go through all of those different probabilities. But I'll just let people know that on the website we have a tool where you can do all of this yourself. Just go to our home page, 538 dotcom, look over on the right and you'll see all of our trackers and interactive. And there you will see a map that you can fool around with and select all different states going for either Biden or Trump.
And then you'll see how our forecast changes. So we'll let you do that on your own. But last couple of questions here as we wrap up. The first one is if Washington, D.C. and Puerto Rico become states, will you need to change the name of your website?
So depending on how they do this, they would probably be more than 538 electoral votes. But now it's got brand equity. Now, folks, we're not going to change the domains where it's probably already taken five forty six dotcom or whatever and would charge an exorbitant amount for it.
So we're we're five thirty eight to stay here. Nor will five Fox change his name.
Sounds good. And last question here. We got a lot of questions about what our plans are for Tuesday night and Tuesday throughout the day. And after we start getting results in, we're going to cover that in our podcast on Monday more thoroughly. But just simply, do we plan on having a forecast that is going to receive the new data that is coming in from the states and combine it with our current forecast? The people would like to know.
No, we will not. So people will have to rely on our analysis and decision desks.
You'll probably hear this 50 times from us on Tuesday night. Do not make inferences until counties or states are fully reported. Just don't do it. Different states, different counties are going to have very different rules for which ballots are processed first. Do not extrapolate. Do not stare at a state that is bluer or redder than you expect and get excited or enthralled or scared by it. Wait until things are fully reported from all different tranches of ballots.
All right, Nate, that's it.
Well, last pre-election little talk, Ailene, her last pre-election model. How does it feel? Like a major marker down.
It doesn't really feel relieving because the election itself is still has to come. I mean, talking on these podcasts is like my therapy, basically. So, yeah, it doesn't really feel relieving. Well, maybe I should charge your your. I guess I do kind of charge you. I got a salary. I don't know something like that. In some roundabout way I'm charging you for therapy, but we're going to have lots of podcasts next week as well.
We'll have a regular podcast on Monday. We're going to sit down at whatever hour we have at least enough results or feel like we need to go to bed on Tuesday night. We'll probably also have a podcast Wednesday. If there are no results, we don't know the winner. We're probably going to be podcasting all next week. Will likely have something out for you in the field also this weekend. So although we're getting down to the wire, there will be no shortage of days to get you through the finish line.
And with that, thank you, Nate. Thank you, Galen. And also to everyone, we talked a lot about Firefox, FOX today. We've got lots of Firefox merchandise over on the 538 store at five thirty eight dotcom. So go check that out. Unfortunately, we do not have that Firefox costume that Nate will be wearing for Halloween, but we'll talk to our colleague Vanessa Diaz and see if she can work on it. Maybe next Halloween.
Would you wear it if we put it on the story? Is it like the dinosaur costume where I can, like, prance around in it? Yeah, you can prance around in whatever costume you want. No one will stop you from prancing as long as we have the costume. OK, yeah, I would wear if Ivy Fox costume for sure. All right.
It's a deal. It is a deal. My name is Garland. Tony Chow is in the virtual control room. Claire Bindaree Curtis is on audio editing. You can get in touch by emailing us at Podcast's at five thirty eight dotcom. You can also, of course, tweeted 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. Tell your friends be like, hey, you want some extra sources of news during the final days of the election?
We're going to have lots of content for you. So tell your friends. Also subscribe on YouTube, as usual. Anyway, thanks for listening and we'll see you soon.