Such a great contrast... 1 candidate refuses to behave like he's ahead (he is) and the other refuses to accept he's behind (he is) *Resistible force meets movable object, or something.
Something i've been wondering is, how do you explain the number of Dem Senate candidates firmly ahead in Swing States where Trump is also firmly ahead? It seems bizarre and unlikely to me that there would be a large between Jaclyn Rosen voters and Trump voters or in Arizona Gallego voters and Trump voters. But I feel that's sort of what you'd need to square these numbers?
In the same vein something i've wondered is you've mentioned Nate that Trump significantly over-performed his polls in 2016 and 2020 (which obviously is true). Which makes me wonder is there any chance that pollsters tried to update their methodology to account for this and are being overly friendly toward Trump?
I really have no idea and don't know enough about polling methodology to say whether that's possible or impossible, but am curious if that could be realistic.
> Which makes me wonder is there any chance that pollsters tried to update their methodology to account for this and are being overly friendly toward Trump?
This has definitely happened after each election, it's just difficult to judge the effects. After 2016 pollsters claimed the issue was not resampling based on education level. But then they did that, and 2020 still happened. So my impression is that after 2020 they couldn't figure out a good way to address the issue and just decided to put their thumb on the scale towards Trump (and Trump alone). But again, impossible to predict how effective it will be.
Yeah I don’t buy this either. Those people who go in and vote for a dem senator to protect abortion rights are going to also throw a vote to Biden. I don’t buy Trump taking AZ, but I’ve been incorrect in every single political prediction I’ve ever made (not joking)
Yea what does a Ruben Gallego/Donald Trump voter even look like? Hard to believe more than like 1% of the electorate fits that mold. And equally hard to believe that someone is passionate enough about the D senate candidate but also so anti-Biden they'd throw a vote away for RFK.
I'm not even necessarily saying that polls are overstating Trump's support (though them updating methodology feels like a better explainer to me?), they could easily swing the other way. I just can't really believe that these results can be accurate in terms of D Senate polling vs Presidential polling; something is getting missed for one side.
Suggest that the reason for “the relatively civil and sophisticated discussion in our comments section” is your subscribers are mostly data wonks, in the first instance, and partisans, secondarily. Makes for a great exchange of questions and ideas - happy to be here.
One group that is consistently underrepresented in discussions about presidential debates is the "the debates are bad actually" camp. Boston College political scientists has written some good posts over the years and one of my favorite is about how debates are a good example of how the press thinks they should be the ones picking presidents (with some great points about how the nominations process has changed over time): http://www.honestgraft.com/2019/03/in-fox-debate-flap-press-defends-its.html
At least the NYT has a headline that mentions replacing him. It's really not an option on the table in D.C. I asked DC folks at a Senator's event recently if they felt Biden would step down if Trump had a double-digit lead and they said no chance. The media reaction to your model was fascinating. It basically just "Nate Silver, who has been trashed for his incorrect predictions" to dismiss the numbers. I am convinced we are now in a place where the left is just going to remain in denial through November. They don't like the 65% so it must be your fault. I mean it's pretty sad that reporters can't even point to, "hey, here's a reason to have hope" instead of "Nate Silver, bad and wrong". I did not waste any of my evening watching the debate but I am looking forward to your take on it. I will give it five minutes of my life at WSOP but not one hour.
To his credit, Nate has been pretty vocal for many months now that the Democrats would be much better off if Biden stepped down and they had an open convention. I wonder what he thinks the odds of that happening are now? Maybe that's gone from 4% to 33%?
Well that wasn't the game changer that Biden was looking for. Not sure how much any debate will move the needle, but this one was just horrific for the Democrats.
I have a question. How do you know whether your model works or not? This is not a snarky question, but one I have been wrestling with for a very long time. When you create a betting model the test is whether you make money over the betting period, year after year. You can do that because there are many outcomes. Another way to do it is to calibrate the model when estimating a win percentage. So if you say a team should win 70% of the time and it wins 40% of the time your model is off.
With an election model, though, especially a national, presidential election model, you have one outcome every four years. It's hard to get a significant answer from that. How do you check?
What’s the chance of the early debate mostly being a Biden audition for superdelegates and activists rather than a performance for the general electorate? Seems like we were past the point of a Biden dropout, but maybe they felt the need to create an out?
Only Biden himself knows what he's thinking, but the worst place for Democrats to be is far enough behind that they're likely to lose, but not so far behind that they can persuade Biden to drop out. I wonder how close we are to that 'worst place' spot. It's hard to tell because the first part of it is a lot easier to know than the second.
Depends on the reason. Decided after the debate that this country needs fresh ideas and a younger voice. These seniors of 8 decades should just bow out.
I figure that part of the reason they wanted a debate this early is that if Biden completely nosedives (falls down on stage, goes catatonic for 10 seconds, etc.) there's still the Hail Mary play of an open convention. I don't think either of these things are particularly likely to happen (like <10%), but it's a form of catastrophic insurance.
More likely, if Biden puts in merely a poor (but not catastrophic) performance, the early debates give it time to fade from the public's mind before November.
If Biden puts in a poor performance that's kind of the worst case scenario that Zach outlined above. The question I have is whether or not a second debate happens if Biden doesn't do well.
My initial gut reaction is that Biden's too proud and hates Trump too deeply to shrink away from a second even if the first goes poorly, but who knows.
Hard sell to pull out of the second debate unilaterally. Ultimately Biden can probably do whatever he wants though, because the people who have jobs in his administration are vested in the nominee being him rather than another Democrat, so it's a classic problem of incentives (the net benefit to Democrats of a change is [probably] positive, but the people closest to the situation are the ones hurt most by it, so it doesn't happen). The only person I've ever thought might be able to talk Biden into dropping out would be Barack Obama, but it seems very unlikely he would even try.
Yes, if you're a Biden apparatchik if Trump wins you're out of a job in January. If Biden withdraws you're unemployed even sooner. Not much incentive to try to get a stronger candidate.
I was listening to your podcast discussion on homeowners insurance. I think you're missing a big reason insurance companies are leaving entire states and not just selling to the low risk areas. Look up how FAIR plans are funded. State FAIR plans act as backstop homeowners insurance for high risk homes that are uninsurable on the voluntary market. FAIR plans are funded by private insurance companies in proportion to their share of the voluntary market, so insurance companies are liable for high risk areas if they do business in low risk areas. That's why it makes sense to leave states like California and Florida entirely.
I will do the 'isolation chamber ' debate watch and commit to post my take as a reply to this comment before reading anyone else's take. So sharing at least one independent data point on here!
Straightforwardly bad for Biden. He looks old, sounds old, doesn't speak fluently, and just doesn't seem to have what it takes to counter Trump's well-honed line of bullshit.
Based on current polling averages, there is zero gap between the popular vote and electoral college. Trump is up 1.5 pts in the national average, and Trump is up in PA, the current tipping point state, by the same margin.
However, in the forecast, Biden is expected to win the popular vote by 0.1 pt, but Trump is expected to win the PA vote by 1.2 pts, a 1.3-pt gap between the popular vote and electoral college. I think I understand the logic behind why the forecast for Biden nationally is better than the current national polling averages, but can you discuss why that wouldn't be the case for the state forecasts and polling average?
This seems like a crucial question, as the model currently has Biden losing the electoral college in nearly 1/3 of the instances where he wins the popular vote (~32% of the time)
"Based on current polling averages, there is zero gap between the popular vote and electoral college. Trump is up 1.5 pts in the national average, and Trump is up in PA, the current tipping point state, by the same margin."
There's no reason those two numbers should be correlated in terms of magnitude, i.e. it's pure coincidence.
Well, I think that's the question. Obviously these are small changes so it could be noise, but maybe for some reason Biden is holding up in those states compared to others. I'm not sure, but given the wider national vote/EC gaps in 2016 and 2020, thought it was noteworthy
1) From one perspective Biden is holding up better in the swing states than the national average. But from the perspective of past polling (2016 and 2020) this is the first time that Trump has had any sort of lead in WI/MI/PA. So I would be cautious.
2) The simplest explanation I can muster is that the race is genuinely close in those states compared to the rest of the country (and the rest of the swing states).
I would guess that it's because Biden is already overperforming in PA/WI/MI relative to his national polls, and the model is guessing that the numbers in those heavily polled states are more reliable, so it's adjusting Biden up in other places, rather than adjusting him down in those three states. Probably what you're thinking too?
Yeah, could be something like that, or maybe it has something to do with the mix of polls in those states vs nationally, or some weight being placed on a prior expectation of how those states perform vs national (I really should spend some more time reading more about the model methodology)
It's possible the electoral college disadvantage for Democrats has lessened or even completely disappeared this cycle. It could even have shifted in the other direction, if Trump wins by more than expected and runs up the score in noncompetitive states. It's just that the gap only matters in terms of how fair or unfair the public thinks our system is. If Biden loses any of PA, MI, or WI, he loses the election regardless of what the national popular vote is.
Never say never. Not in this world. I wouldn't be surprised if his wife really put the pressure on him to walk away. The sticking point is getting Kamala to step aside. She's pretty much unelectable. Mayor Pete would be an excellent candidate. Only major name with a net favorable. I think I saw 32-30. Of got him at 9 cents for the nomination on PredictIt. I can dream.
The big thing with spin is this. I was in that position for several years. What you don't EVER do is lie outright. It has to be the best possible take on the facts as they exist. Every lawyer knows how to do it and lawyers who actually lie are in trouble. What you say must be technically true. Jen Psaki, much as I disliked her, was conscious of that line; she crossed it sometimes but never grossly. Karine Jean-Pierre just goes out and says whatever she's told to say, even when it absolutely defies all reality.
Probably the best was Ari Fleischer, who served Bush II in the first few years.
Such a great contrast... 1 candidate refuses to behave like he's ahead (he is) and the other refuses to accept he's behind (he is) *Resistible force meets movable object, or something.
Sustain the friendly vibes? That's no fun. Lets get toxic AF
Math is for nerds. Haha!
Your at the mean for being mean on x but two standard deviation away from being nice here.
hah!
I got toxic AF with your mum tonight mate.
As a new subscriber and longtime fan of your work, just want to say this is a fantastic Substack. Thanks for sticking with it after the 538 fallout.
Something i've been wondering is, how do you explain the number of Dem Senate candidates firmly ahead in Swing States where Trump is also firmly ahead? It seems bizarre and unlikely to me that there would be a large between Jaclyn Rosen voters and Trump voters or in Arizona Gallego voters and Trump voters. But I feel that's sort of what you'd need to square these numbers?
In the same vein something i've wondered is you've mentioned Nate that Trump significantly over-performed his polls in 2016 and 2020 (which obviously is true). Which makes me wonder is there any chance that pollsters tried to update their methodology to account for this and are being overly friendly toward Trump?
I really have no idea and don't know enough about polling methodology to say whether that's possible or impossible, but am curious if that could be realistic.
Yes, I’ve thought the same. Here’s a vote for a piece trying to explain what’s going on with this discrepancy.
I can't speak broadly, but in Ohio there genuinely are a significant number of Trump voters who will be voting for Sherrod Brown.
> Which makes me wonder is there any chance that pollsters tried to update their methodology to account for this and are being overly friendly toward Trump?
This has definitely happened after each election, it's just difficult to judge the effects. After 2016 pollsters claimed the issue was not resampling based on education level. But then they did that, and 2020 still happened. So my impression is that after 2020 they couldn't figure out a good way to address the issue and just decided to put their thumb on the scale towards Trump (and Trump alone). But again, impossible to predict how effective it will be.
Yeah I don’t buy this either. Those people who go in and vote for a dem senator to protect abortion rights are going to also throw a vote to Biden. I don’t buy Trump taking AZ, but I’ve been incorrect in every single political prediction I’ve ever made (not joking)
Yea what does a Ruben Gallego/Donald Trump voter even look like? Hard to believe more than like 1% of the electorate fits that mold. And equally hard to believe that someone is passionate enough about the D senate candidate but also so anti-Biden they'd throw a vote away for RFK.
I'm not even necessarily saying that polls are overstating Trump's support (though them updating methodology feels like a better explainer to me?), they could easily swing the other way. I just can't really believe that these results can be accurate in terms of D Senate polling vs Presidential polling; something is getting missed for one side.
Suggest that the reason for “the relatively civil and sophisticated discussion in our comments section” is your subscribers are mostly data wonks, in the first instance, and partisans, secondarily. Makes for a great exchange of questions and ideas - happy to be here.
One group that is consistently underrepresented in discussions about presidential debates is the "the debates are bad actually" camp. Boston College political scientists has written some good posts over the years and one of my favorite is about how debates are a good example of how the press thinks they should be the ones picking presidents (with some great points about how the nominations process has changed over time): http://www.honestgraft.com/2019/03/in-fox-debate-flap-press-defends-its.html
Biden should have held more formal press conferences but alas too old Joe is not up to it.
Those things are often a waste of time to though "MR PRESIDENT WHY DO SOME PEOPLE SAY YOURE SENILE!?!?!?!" etc
At least the NYT has a headline that mentions replacing him. It's really not an option on the table in D.C. I asked DC folks at a Senator's event recently if they felt Biden would step down if Trump had a double-digit lead and they said no chance. The media reaction to your model was fascinating. It basically just "Nate Silver, who has been trashed for his incorrect predictions" to dismiss the numbers. I am convinced we are now in a place where the left is just going to remain in denial through November. They don't like the 65% so it must be your fault. I mean it's pretty sad that reporters can't even point to, "hey, here's a reason to have hope" instead of "Nate Silver, bad and wrong". I did not waste any of my evening watching the debate but I am looking forward to your take on it. I will give it five minutes of my life at WSOP but not one hour.
To his credit, Nate has been pretty vocal for many months now that the Democrats would be much better off if Biden stepped down and they had an open convention. I wonder what he thinks the odds of that happening are now? Maybe that's gone from 4% to 33%?
Well that wasn't the game changer that Biden was looking for. Not sure how much any debate will move the needle, but this one was just horrific for the Democrats.
I have a question. How do you know whether your model works or not? This is not a snarky question, but one I have been wrestling with for a very long time. When you create a betting model the test is whether you make money over the betting period, year after year. You can do that because there are many outcomes. Another way to do it is to calibrate the model when estimating a win percentage. So if you say a team should win 70% of the time and it wins 40% of the time your model is off.
With an election model, though, especially a national, presidential election model, you have one outcome every four years. It's hard to get a significant answer from that. How do you check?
https://projects.fivethirtyeight.com/checking-our-work/
Not updated since Nate left, obviously, but this is the go to source for these questions.
Having state-level outcomes forecast helps.
What’s the chance of the early debate mostly being a Biden audition for superdelegates and activists rather than a performance for the general electorate? Seems like we were past the point of a Biden dropout, but maybe they felt the need to create an out?
Only Biden himself knows what he's thinking, but the worst place for Democrats to be is far enough behind that they're likely to lose, but not so far behind that they can persuade Biden to drop out. I wonder how close we are to that 'worst place' spot. It's hard to tell because the first part of it is a lot easier to know than the second.
The other problem is that if Biden was going to drop out he should have done it last year.
Yes, definitely, but in the absence of a time machine the last year of the campaign is a sunk cost. Not that humans make decisions rationally.
Depends on the reason. Decided after the debate that this country needs fresh ideas and a younger voice. These seniors of 8 decades should just bow out.
I figure that part of the reason they wanted a debate this early is that if Biden completely nosedives (falls down on stage, goes catatonic for 10 seconds, etc.) there's still the Hail Mary play of an open convention. I don't think either of these things are particularly likely to happen (like <10%), but it's a form of catastrophic insurance.
More likely, if Biden puts in merely a poor (but not catastrophic) performance, the early debates give it time to fade from the public's mind before November.
If Biden puts in a poor performance that's kind of the worst case scenario that Zach outlined above. The question I have is whether or not a second debate happens if Biden doesn't do well.
My initial gut reaction is that Biden's too proud and hates Trump too deeply to shrink away from a second even if the first goes poorly, but who knows.
Hard sell to pull out of the second debate unilaterally. Ultimately Biden can probably do whatever he wants though, because the people who have jobs in his administration are vested in the nominee being him rather than another Democrat, so it's a classic problem of incentives (the net benefit to Democrats of a change is [probably] positive, but the people closest to the situation are the ones hurt most by it, so it doesn't happen). The only person I've ever thought might be able to talk Biden into dropping out would be Barack Obama, but it seems very unlikely he would even try.
Yes, if you're a Biden apparatchik if Trump wins you're out of a job in January. If Biden withdraws you're unemployed even sooner. Not much incentive to try to get a stronger candidate.
I was listening to your podcast discussion on homeowners insurance. I think you're missing a big reason insurance companies are leaving entire states and not just selling to the low risk areas. Look up how FAIR plans are funded. State FAIR plans act as backstop homeowners insurance for high risk homes that are uninsurable on the voluntary market. FAIR plans are funded by private insurance companies in proportion to their share of the voluntary market, so insurance companies are liable for high risk areas if they do business in low risk areas. That's why it makes sense to leave states like California and Florida entirely.
I will do the 'isolation chamber ' debate watch and commit to post my take as a reply to this comment before reading anyone else's take. So sharing at least one independent data point on here!
Straightforwardly bad for Biden. He looks old, sounds old, doesn't speak fluently, and just doesn't seem to have what it takes to counter Trump's well-honed line of bullshit.
Based on current polling averages, there is zero gap between the popular vote and electoral college. Trump is up 1.5 pts in the national average, and Trump is up in PA, the current tipping point state, by the same margin.
However, in the forecast, Biden is expected to win the popular vote by 0.1 pt, but Trump is expected to win the PA vote by 1.2 pts, a 1.3-pt gap between the popular vote and electoral college. I think I understand the logic behind why the forecast for Biden nationally is better than the current national polling averages, but can you discuss why that wouldn't be the case for the state forecasts and polling average?
This seems like a crucial question, as the model currently has Biden losing the electoral college in nearly 1/3 of the instances where he wins the popular vote (~32% of the time)
"Based on current polling averages, there is zero gap between the popular vote and electoral college. Trump is up 1.5 pts in the national average, and Trump is up in PA, the current tipping point state, by the same margin."
There's no reason those two numbers should be correlated in terms of magnitude, i.e. it's pure coincidence.
Well, I think that's the question. Obviously these are small changes so it could be noise, but maybe for some reason Biden is holding up in those states compared to others. I'm not sure, but given the wider national vote/EC gaps in 2016 and 2020, thought it was noteworthy
1) From one perspective Biden is holding up better in the swing states than the national average. But from the perspective of past polling (2016 and 2020) this is the first time that Trump has had any sort of lead in WI/MI/PA. So I would be cautious.
2) The simplest explanation I can muster is that the race is genuinely close in those states compared to the rest of the country (and the rest of the swing states).
I would guess that it's because Biden is already overperforming in PA/WI/MI relative to his national polls, and the model is guessing that the numbers in those heavily polled states are more reliable, so it's adjusting Biden up in other places, rather than adjusting him down in those three states. Probably what you're thinking too?
Yeah, could be something like that, or maybe it has something to do with the mix of polls in those states vs nationally, or some weight being placed on a prior expectation of how those states perform vs national (I really should spend some more time reading more about the model methodology)
It's possible the electoral college disadvantage for Democrats has lessened or even completely disappeared this cycle. It could even have shifted in the other direction, if Trump wins by more than expected and runs up the score in noncompetitive states. It's just that the gap only matters in terms of how fair or unfair the public thinks our system is. If Biden loses any of PA, MI, or WI, he loses the election regardless of what the national popular vote is.
I think you referred to this. A June debate could show the party what a liability Biden is and still give them time to run someone else.
Biden is not going to step down though.
Never say never. Not in this world. I wouldn't be surprised if his wife really put the pressure on him to walk away. The sticking point is getting Kamala to step aside. She's pretty much unelectable. Mayor Pete would be an excellent candidate. Only major name with a net favorable. I think I saw 32-30. Of got him at 9 cents for the nomination on PredictIt. I can dream.
Yeah this comment of mine did not age well. It seems very possible now!
The big thing with spin is this. I was in that position for several years. What you don't EVER do is lie outright. It has to be the best possible take on the facts as they exist. Every lawyer knows how to do it and lawyers who actually lie are in trouble. What you say must be technically true. Jen Psaki, much as I disliked her, was conscious of that line; she crossed it sometimes but never grossly. Karine Jean-Pierre just goes out and says whatever she's told to say, even when it absolutely defies all reality.
Probably the best was Ari Fleischer, who served Bush II in the first few years.