I'm not loving the polls right now - seems to be a bit of a drip-drip-drip inching to the right across multiple modes. FiveThirtyEight's forecast will probably be at 50/50 within a day or two.
That said, I go look at the details and I see extremely, nearly implausibly, R-favorable electorates. Take Ipsos/Reuters, most recent poll linked b…
I'm not loving the polls right now - seems to be a bit of a drip-drip-drip inching to the right across multiple modes. FiveThirtyEight's forecast will probably be at 50/50 within a day or two.
That said, I go look at the details and I see extremely, nearly implausibly, R-favorable electorates. Take Ipsos/Reuters, most recent poll linked below. It has Harris +3 among likely voters. But she is +91 among Ds, -82 among Rs, and +9 among Is. How is this possible? It's possible if your sample of likely voters is somewhere between R+3 and R+4 (least-squares regression indicates about R+3.3).
Similarly, that Fox News poll that had Trump +2 nationwide has Harris +85 among Ds, -87 among Rs, and +9 among Is. Again, least-squares regression indicates an R+3.1 composition of the likely electorate, and indeed their writeup says it's R+3.
This seems a little hard for me to buy. I don't know why you'd assume it's a 2014 electorate. If it is, I agree we're in deep trouble, but it just doesn't seem likely. It feels to me like 2020, where partisan samples favored the out party by a lot (D+5, D+6, etc.) and we ended up with a D+1 electorate instead. If Harris out-performs the polls by 3-4 points, like Trump did in 2020, this might be why.
Furthermore, a number of other polls that have moved right seem to mostly be moving right because of sample. Emerson two weeks ago had a D+1.6 sample with Harris +1.4. Today they have a D+0.1 sample with Harris +0.3.
Many bad pollsters have taken Steve Bannon’s advice to heart: "Flood the zone with shit." All too many pollsters are really not in the business of conscientious polling, but instead have found an extremely cost-effective way to impact the media narrative. There’s now a massive flood of bad polls, mostly of swing states. A clear objective is to impact the polling averages and create the illusion that Trump is winning.
Unfortunately, all the bad-faith polling also creates a very fertile ground for the guaranteed claims of "Election fraud!" and "Stolen election!" after Trump loses.
If it were just the red-wavey pollsters, the Emersons of the world, I'd agree that this is what's going on. But Ipsos/Reuters aren't partisan, and in the past Fox hasn't been partisan in its polling either (though that R+3, plus the release timed to ambush Harris live, makes me wonder). I think there's something fundamentally weird right now. I just don't quite know what it is. Maybe it's that the electorate really will be R+3, but I'm suspicious.
There may well be something going on, such as some Republicans "returning home". And I share your impression that Fox commissions high-quality polls. But if you look at the makeup of the recent wave of swing-state polls, it’s hard to avoid the conclusion that much of the narrative is manufactured.
It’s worth noting that bad pollsters did exactly the same in 2022, creating the illusion of a "Red Wave" that never materialized. In fact, a chorus of pundits were mocking Simon Rosenberg, Tom Bonier, Michael McDonald, Jennifer Rubin and a few others that were insisting: "There may be a Red Wave coming, but so far we see no signs of that."
Nate Silver went so far as to accuse Simon Rosenberg of "Smoking Hopium!" That’s how the Hopium Chronicles were born. :)
Suggestion: Look very closely at the assumptions built into each good-faith pollster’s Likely Voter model. My impression is that they have tweaked their model to account for Trump’s overperforming the polls in 2016 and again in 2020 – but that they have failed to adjust for the post-Dobbs reality. Consider:
– Democrats overperformed in the 2022 Midterm Elections
– Democrats have overperformed in every Special Election, post-Dobbs
– Trump underperformed the polls in almost every single Republican primary.
I would hope they already had tweaked their model to account for that but recent tweaks could be the cause. I think there is also a very good chance that they’re not accounting for (potentially) new Harris voters. The difference between her and Biden is so stark, that along with Dobbs I fully expect there to be a significant number of new women voters, young voters, and voters of color. There is data from new voter registrations to support that, but if pollsters are modeling based on the 2020 electorate those people will not be included.
I have to also consider that may be bias or wishful thinking, but there is enough evidence, both solid and anecdotal, that I would be more surprised by the polls underestimating Harris by a few points than Trump. It’s been stated on here multiple times that Trump has a hard ceiling of 46-47%. The only way he gets to 49-51% which is the number in a couple of polls today is if Dem enthusiasm craters. There is absolutely zero indication that will happen, and indeed every sign points to the opposite.
These folks voted for Haley long after the nomination was decided; they won't vote Trump and will walk over hot coals barefoot to defeat him at the polls(and I think you are correct in your post above about normal partisan %)
Trump's "hard ceiling" of 47% is an article of faith, leaning far too heavily into past performance and ignoring the significant coalitional dynamism that's been playing out in the last decade.
I usually think of floors and ceilings in terms of absolute numbers of voters, not shares of the electorate. If turnout collapses for one party, the other party can break through its supposed ceiling of the vote share even if its own turnout isn't that special.
That’s a fair point, but I don’t believe for a second that the majority of American’s support him. As sacman701 states below, him getting a higher percentage than that is heavily dependent on turnout collapsing on the Democratic side. I’ll concede that the 47% ceiling is based on slim to limited evidence, but that compares pretty well to the complete lack of evidence that he has majority support.
what makes it a poor argument, too, is easy: 47% of what electorate? If GOP turnout craters he’s not getting 47. If D turnout craters he probably beats that number. It’s entirely relative
It could be that pollsters are now getting actual voting data from early/mail in voting and its showing an R leaning year.
I often think - I basically live in a bubble with very little true interaction with how a good chunk of the country thinks.
But it could very well be that at the end of the day - the price of milk and rent and gas is just too high for the infrequent voter to ignore. Or that alot of the country is just too racist and misogynist. Or just too anti-California liberal - maybe those "Kamala saying in her own words how proud she is that she approved money for murderers to get trans surgery" ads are working.
Or that the softness in enthusiasm among infrequent Dem voters that we saw clearly when Biden was the candidate never really went away and now that the Harris announcement/Walz announcement/DNC/debate bounces have receded, we are in not so great shape.
Or just enough of all of the above for Trump to win. We will have to see.
All I can say is - if Trump wins and the GOP takes the house and senate it will prove that basically nothing matters in national politics but vibes. Not fundraising or organization or candidate quality or even ideology.
That's exactly how Glenn Youngkin got elected Governor here in Virginia. He managed to get father of five Terry McAuliffe perceived as "anti parent" and exit polls indicated that voters thought Youngkin, not McAuliffe was the "moderate." Not to mention the hard feelings that McAuliffe even ran in the race to begin with - Virginia is the one state in the USA where Governors are not permitted to serve consecutive terms - NEVER went away. Many "progressives" and African American voters wanted an African American nominee for Governor because "our turn" and felt that McAuliffe's run violated the "spirit of the law" I described above. I knew McAuliffe was in trouble when he let Youngkin beat him to the punch about repealing the grocery tax (Virginia is a VERY anti tax state) and when there were Democrats claiming that the law I described above meant that Governors could only serve one term. In short, vibes absolutely matter and we do ourselves no favors by pretending otherwise.
I would say perceived ideology matters. Not actual ideology. Reagan was VERY right wing but because he looked good on television, he wasn't seen as such.
He can switch on almost anything *that isn't what his voters actually care most about,* which is bigotry in support of a rigorously enforced social order. They're extremely ideological - they just don't actually care very much about traditionally Republican-coded policies of low taxes and pro-business policies. They care about people behaving the "correct way" and obeying their given place in the hierarchy.
Of course, because that's exactly who Trump is at his core, he won't be switching.
The oddest thing about it is they’re modeling these Republican leaning electorates while simultaneously showing Democratic enthusiasm to be fairly high.
I wonder if it's just that they didn't change their models after Biden left the race? Which, I can't totally blame them if that's the case. Like, if you think people didn't like Biden because they were unhappy about the economy and other issues, then you should keep your models favoring a conservative electorate. But if it was just a lack of enthusiasm that has flipped, you'll get the wrong electorate. It sure doesn't seem like Dems are suffering from a lack of enthusiasm.
I like the idea. Nice thought. Here's another variation: what if they changed them, but changed them during the sugar high of the replacement/DNC when response bias was favoring Ds, so they had to deflate the D share a bit? Would produce similar results.
Most of the polls results seem to be based on modeling the electorate, and I’m glad they’re publishing their numbers no matter what they are. Modeling an electorate in this era is difficult.
I agree that an R+3 electorate seems… unlikely, as of right now. Maybe I’m wrong! But that Emerson example you give seems to show that all that slide is almost 1-1 correlated with their assumed electorate model and that leads me to my prior that the race remains static and narrow, as it has more or less since Labor Day
This is a big reason why I think Harris will outperform the polls. This country obviously isn't R+3 (there have been more Ds than Rs in America for most of my life [which is why Rs have only won the popular vote once during it]), and there's no reason to believe the electorate will have more Rs than Ds. I'm expecting a roughly D+1 electorate, about the same as 2020, and that the results will reflect that (and therefore Harris will outperform any polls that use an R+3 electorate).
Are they weighting by recalled 2020 vote? If so, that will likely skew the weighted sample towards Republicans, since people tend to misremember voting for the winning candidate (Biden).
Some of them are, yes. Ipsos/Reuters is, for example. I just looked at the most recent YouGov/Economist, where they weight on it, and I see the same pattern (Harris +93 among Ds, -82 among Rs, +3 among Is, but only +3 overall, meaning a roughly R+2 sample).
Good catch. Of course self-reported party ID is fluid, but nothing has happened over the last 4 years that would cause a big shift toward the GOP. The ABC exit poll for the 2022 midterm showed an R+3 electorate. This cycle, we have pundit ratings and campaigns' spending patterns both suggesting an environment much closer to 2020 (when exit polls showed a D+1 electorate) than to 2022. At worst, I would expect an R+1 electorate, and my best guess would be dead even but with Dems doing slightly better among indies than in the past because young voters tend to register indie.
And now that I'm noticing this pattern, I'm seeing it everywhere. YouGov, Harris +3, R+2 sample. Scott Rasmussen (RMG), Harris +1, R+3 sample (last week was Harris +3 in an R+1 sample).
I'm not loving the polls right now - seems to be a bit of a drip-drip-drip inching to the right across multiple modes. FiveThirtyEight's forecast will probably be at 50/50 within a day or two.
That said, I go look at the details and I see extremely, nearly implausibly, R-favorable electorates. Take Ipsos/Reuters, most recent poll linked below. It has Harris +3 among likely voters. But she is +91 among Ds, -82 among Rs, and +9 among Is. How is this possible? It's possible if your sample of likely voters is somewhere between R+3 and R+4 (least-squares regression indicates about R+3.3).
Similarly, that Fox News poll that had Trump +2 nationwide has Harris +85 among Ds, -87 among Rs, and +9 among Is. Again, least-squares regression indicates an R+3.1 composition of the likely electorate, and indeed their writeup says it's R+3.
This seems a little hard for me to buy. I don't know why you'd assume it's a 2014 electorate. If it is, I agree we're in deep trouble, but it just doesn't seem likely. It feels to me like 2020, where partisan samples favored the out party by a lot (D+5, D+6, etc.) and we ended up with a D+1 electorate instead. If Harris out-performs the polls by 3-4 points, like Trump did in 2020, this might be why.
Furthermore, a number of other polls that have moved right seem to mostly be moving right because of sample. Emerson two weeks ago had a D+1.6 sample with Harris +1.4. Today they have a D+0.1 sample with Harris +0.3.
(Ipsos link: https://www.ipsos.com/sites/default/files/Reuters%20Ipsos%20Core%20Political%2010%2016%202024%20PDF.pdf.
Fox link:
https://static.foxnews.com/foxnews.com/content/uploads/2024/10/fox_october-11-14-2024_national_cross-tabs_october-16-release.pdf)
Many bad pollsters have taken Steve Bannon’s advice to heart: "Flood the zone with shit." All too many pollsters are really not in the business of conscientious polling, but instead have found an extremely cost-effective way to impact the media narrative. There’s now a massive flood of bad polls, mostly of swing states. A clear objective is to impact the polling averages and create the illusion that Trump is winning.
Unfortunately, all the bad-faith polling also creates a very fertile ground for the guaranteed claims of "Election fraud!" and "Stolen election!" after Trump loses.
If it were just the red-wavey pollsters, the Emersons of the world, I'd agree that this is what's going on. But Ipsos/Reuters aren't partisan, and in the past Fox hasn't been partisan in its polling either (though that R+3, plus the release timed to ambush Harris live, makes me wonder). I think there's something fundamentally weird right now. I just don't quite know what it is. Maybe it's that the electorate really will be R+3, but I'm suspicious.
There may well be something going on, such as some Republicans "returning home". And I share your impression that Fox commissions high-quality polls. But if you look at the makeup of the recent wave of swing-state polls, it’s hard to avoid the conclusion that much of the narrative is manufactured.
It’s worth noting that bad pollsters did exactly the same in 2022, creating the illusion of a "Red Wave" that never materialized. In fact, a chorus of pundits were mocking Simon Rosenberg, Tom Bonier, Michael McDonald, Jennifer Rubin and a few others that were insisting: "There may be a Red Wave coming, but so far we see no signs of that."
Nate Silver went so far as to accuse Simon Rosenberg of "Smoking Hopium!" That’s how the Hopium Chronicles were born. :)
https://www.hopiumchronicles.com/
Suggestion: Look very closely at the assumptions built into each good-faith pollster’s Likely Voter model. My impression is that they have tweaked their model to account for Trump’s overperforming the polls in 2016 and again in 2020 – but that they have failed to adjust for the post-Dobbs reality. Consider:
– Democrats overperformed in the 2022 Midterm Elections
– Democrats have overperformed in every Special Election, post-Dobbs
– Trump underperformed the polls in almost every single Republican primary.
I’d rather they tweak their model to account for their misses in 2016/20 than the other way around, to be sure
I would hope they already had tweaked their model to account for that but recent tweaks could be the cause. I think there is also a very good chance that they’re not accounting for (potentially) new Harris voters. The difference between her and Biden is so stark, that along with Dobbs I fully expect there to be a significant number of new women voters, young voters, and voters of color. There is data from new voter registrations to support that, but if pollsters are modeling based on the 2020 electorate those people will not be included.
I have to also consider that may be bias or wishful thinking, but there is enough evidence, both solid and anecdotal, that I would be more surprised by the polls underestimating Harris by a few points than Trump. It’s been stated on here multiple times that Trump has a hard ceiling of 46-47%. The only way he gets to 49-51% which is the number in a couple of polls today is if Dem enthusiasm craters. There is absolutely zero indication that will happen, and indeed every sign points to the opposite.
" I think there is also a very good chance that they’re not accounting for (potentially) new Harris voters."
A modest example:
A New York Times/Siena College/Philadelphia Inquirer poll found Harris winning 12% of Republicans in Pennsylvania.
https://nitter.poast.org/RpsAgainstTrump/status/1847108612839076264#m
Parties typically usually only win about 4-5% of other-partisans, no?
These folks voted for Haley long after the nomination was decided; they won't vote Trump and will walk over hot coals barefoot to defeat him at the polls(and I think you are correct in your post above about normal partisan %)
Trump's "hard ceiling" of 47% is an article of faith, leaning far too heavily into past performance and ignoring the significant coalitional dynamism that's been playing out in the last decade.
I usually think of floors and ceilings in terms of absolute numbers of voters, not shares of the electorate. If turnout collapses for one party, the other party can break through its supposed ceiling of the vote share even if its own turnout isn't that special.
That’s a fair point, but I don’t believe for a second that the majority of American’s support him. As sacman701 states below, him getting a higher percentage than that is heavily dependent on turnout collapsing on the Democratic side. I’ll concede that the 47% ceiling is based on slim to limited evidence, but that compares pretty well to the complete lack of evidence that he has majority support.
what makes it a poor argument, too, is easy: 47% of what electorate? If GOP turnout craters he’s not getting 47. If D turnout craters he probably beats that number. It’s entirely relative
It could be that pollsters are now getting actual voting data from early/mail in voting and its showing an R leaning year.
I often think - I basically live in a bubble with very little true interaction with how a good chunk of the country thinks.
But it could very well be that at the end of the day - the price of milk and rent and gas is just too high for the infrequent voter to ignore. Or that alot of the country is just too racist and misogynist. Or just too anti-California liberal - maybe those "Kamala saying in her own words how proud she is that she approved money for murderers to get trans surgery" ads are working.
Or that the softness in enthusiasm among infrequent Dem voters that we saw clearly when Biden was the candidate never really went away and now that the Harris announcement/Walz announcement/DNC/debate bounces have receded, we are in not so great shape.
Or just enough of all of the above for Trump to win. We will have to see.
All I can say is - if Trump wins and the GOP takes the house and senate it will prove that basically nothing matters in national politics but vibes. Not fundraising or organization or candidate quality or even ideology.
What matters is information ecosystems. Which create vibes.
That's exactly how Glenn Youngkin got elected Governor here in Virginia. He managed to get father of five Terry McAuliffe perceived as "anti parent" and exit polls indicated that voters thought Youngkin, not McAuliffe was the "moderate." Not to mention the hard feelings that McAuliffe even ran in the race to begin with - Virginia is the one state in the USA where Governors are not permitted to serve consecutive terms - NEVER went away. Many "progressives" and African American voters wanted an African American nominee for Governor because "our turn" and felt that McAuliffe's run violated the "spirit of the law" I described above. I knew McAuliffe was in trouble when he let Youngkin beat him to the punch about repealing the grocery tax (Virginia is a VERY anti tax state) and when there were Democrats claiming that the law I described above meant that Governors could only serve one term. In short, vibes absolutely matter and we do ourselves no favors by pretending otherwise.
I would say perceived ideology matters. Not actual ideology. Reagan was VERY right wing but because he looked good on television, he wasn't seen as such.
Trump is almost anti-ideological, and if he switches on almost anything his voters will come along with him.
If he switched on choice, they wouldn't. The Republican Party remains hardline socially conservative.
He can switch on almost anything *that isn't what his voters actually care most about,* which is bigotry in support of a rigorously enforced social order. They're extremely ideological - they just don't actually care very much about traditionally Republican-coded policies of low taxes and pro-business policies. They care about people behaving the "correct way" and obeying their given place in the hierarchy.
Of course, because that's exactly who Trump is at his core, he won't be switching.
The oddest thing about it is they’re modeling these Republican leaning electorates while simultaneously showing Democratic enthusiasm to be fairly high.
I wonder if it's just that they didn't change their models after Biden left the race? Which, I can't totally blame them if that's the case. Like, if you think people didn't like Biden because they were unhappy about the economy and other issues, then you should keep your models favoring a conservative electorate. But if it was just a lack of enthusiasm that has flipped, you'll get the wrong electorate. It sure doesn't seem like Dems are suffering from a lack of enthusiasm.
That’s actually not a terrible theory
I like the idea. Nice thought. Here's another variation: what if they changed them, but changed them during the sugar high of the replacement/DNC when response bias was favoring Ds, so they had to deflate the D share a bit? Would produce similar results.
Most of the polls results seem to be based on modeling the electorate, and I’m glad they’re publishing their numbers no matter what they are. Modeling an electorate in this era is difficult.
I agree that an R+3 electorate seems… unlikely, as of right now. Maybe I’m wrong! But that Emerson example you give seems to show that all that slide is almost 1-1 correlated with their assumed electorate model and that leads me to my prior that the race remains static and narrow, as it has more or less since Labor Day
This is a big reason why I think Harris will outperform the polls. This country obviously isn't R+3 (there have been more Ds than Rs in America for most of my life [which is why Rs have only won the popular vote once during it]), and there's no reason to believe the electorate will have more Rs than Ds. I'm expecting a roughly D+1 electorate, about the same as 2020, and that the results will reflect that (and therefore Harris will outperform any polls that use an R+3 electorate).
Are they weighting by recalled 2020 vote? If so, that will likely skew the weighted sample towards Republicans, since people tend to misremember voting for the winning candidate (Biden).
Some of them are, yes. Ipsos/Reuters is, for example. I just looked at the most recent YouGov/Economist, where they weight on it, and I see the same pattern (Harris +93 among Ds, -82 among Rs, +3 among Is, but only +3 overall, meaning a roughly R+2 sample).
Doesn't explain Fox, but, well, Fox.
Good catch. Of course self-reported party ID is fluid, but nothing has happened over the last 4 years that would cause a big shift toward the GOP. The ABC exit poll for the 2022 midterm showed an R+3 electorate. This cycle, we have pundit ratings and campaigns' spending patterns both suggesting an environment much closer to 2020 (when exit polls showed a D+1 electorate) than to 2022. At worst, I would expect an R+1 electorate, and my best guess would be dead even but with Dems doing slightly better among indies than in the past because young voters tend to register indie.
Very thoughtful comment.
Thanks, that's very kind of you!
And now that I'm noticing this pattern, I'm seeing it everywhere. YouGov, Harris +3, R+2 sample. Scott Rasmussen (RMG), Harris +1, R+3 sample (last week was Harris +3 in an R+1 sample).