The Dobbs decision, consistent Democratic overperformance in special elections, Trump’s underperformance in Republican primaries, the Harris Campaign’s incredible ground operation with millions of volunteers, Trump’s shocking behavior and ugliness especially these last few weeks, Independents and newly-registered voters (unpolled!) overwhelmingly preferring Harris – IMHO, pollsters have failed to sufficiently take these factors into account.
Moreover, bad-faith pollsters have been releasing an unprecedented number of questionable, with the clear aim of impacting the polling averages such as 538. To an astonishing degree, they have succeeded in manipulating the media election narrative. (Likewise, Polymarket has been a case study in narrative manipulation.)
During this2024 election, especially, basing predictions on the 538 average, means trying to predict on the basis of skewed, highly-manipulated data. PEC should, in my opinion, have restricted itself to an average of just the high-quality independent polls.
Interestingly, the average of the highest-quality independent polls HAS NOT CHANGED.
It’s worth paying attention to one notable pollster exception, one who doesn’t do manipulative weighting or make a whole bunch of hidden assumptions, is Ann Selzer. There are good reasons why she is perhaps the most respected and revered pollster in America.
Here are my predictions, going out on a limb and clearly at odds with the consensus:
PRESIDENCY: Kamala Harris wins at least six of the swing states: MI, WI, PA, NV, GA and NC. Stunningly, she also wins FL and IA. Texas will be surprisingly close. It’s possible that Trump wins AZ.
SENATE: Democratic winners include Gallego AZ, Alsobrooks MD, Slotkin MI, Rosen NV, Casey PA. Baldwin holds in WI, as does Sherrod Brown in OH and Tester in MT. Allred defeats Cruz in TX, and Mucarsel-Powell unseats Scott in FL. Osborn loses a close race against Fischer in NE.
Democrats with caucusing Independents expand control to 52–48.
HOUSE: 226–209, Democratic control. I would not be surprise to see Dems reach 230.
Based on what I see in the special-election data, I think a realistic upper-bound estimate for Harris performance would be the +2.5 point estimate I made. Call it the Hopium Bound. It includes Arizona, North Carolina, and Georgia, and adds up to Harris 319 EV, Trump 218 EV.
I think we're too polarized for things to go any further. Of course I could be wrong!
That could well be. But I think the massive gender gap, the Selzer poll, Harris’ huge advantage in the polls of those who had already cast and Early Vote (far too little emphasis on those polls!), likewise the gaps in specialized large-sample polls of specific demographics (younger voters; Blacks; Latino/Hispanic voters) paint a very different picture.
Add to that the Florida poll of people who had voted early. It showed Harris leading with +1, despite the fact that Rs had a 10+ point advantage in voters!
In other words, despite the polarization, there is a huge swathe of Republican voters (say, almost half of Haley’s voters) who are voting for Harris rather than Trump, because the cannot abide the Fascist. Likewise, there have been repeated indications of Independents in most swing states breaking heavily for Harris.
In other words: The political polarization that focuses on Party is camouflaging rather than clarifying the true underlying dynamics of what is happening.
And pollsters clearly have not adjusted to the post-Dobbs reality, to the referenda results, and to the almost-unbroken series of Democratic overperfomances in the special elections.
(I realize full well that I am almost alone in believing that Harris can win Iowa or Florida. That said, 319 EV would be an excellent result – and well beyond "The Margin of Steal"; hard to see how SCOTUS and others courts, or disloyal electors, can mess with that!)
Kudos to you for taking the special elections into account!
Florida has already been called for Trump, so your predictions are starting to be wrong. I'm all in for the "Hopium" result myself, but you've gone overboard.
i was curious about the switch of wisconsin, dark blue in the unadjusted poll map to light blue in the tilt by 1.5 points to harris map. if the results in the tilted towards harris map would generate better results for her, why did the color of the map seem to indicate a worse outcome for her in wisconsin?
Some of the Vote Maximizer recommendations for close House races that I donated to are coming home. E.g., Riley has just been declared winner in NY-19.
A couple in CA remain very close, though only about 50% of votes counted.
At the moment, in 11 districts Democrats are in the lead to flip Republican districts, and 5 Republicans are in the lead to flip Democratic districts. If this holds, a net gain of 6 seats would be enough to give Democrats control of the House.
In these districts, the average Vote Maximizer per-voter power is 87 out of a possible 100. So it appears that Vote Maximizer was successful in identifying important districts.
I also note that of these 16 districts, 6 were drawn by an independent citizens' commission, 4 were drawn by a court (and 3 of those in New York by an Electoral Innovation Lab affiliate), and 2 used ranked-choice voting.
"Renowned Pollster J. Ann Selzer said Tuesday she would be reviewing her data to determine why a Des Moines Register/Mediacom Iowa Poll released just days before the election produced results so far out of line with former President Donald Trump's resounding victory." [Des Moines Register at 11:40 pm CT]
Re: ".The House of Representatives has little net overall bias, thanks to reductions in gerrymandering in the last decade (three cheers for anti-gerrymandering reformers!). Therefore we can take the national vote as a measure of who will control the House in 2025." I thought that https://www.pnas.org/doi/10.1073/pnas.2217322120 showed that there's about ten-seat (or 1% vote) advantage for the GOP in the House due to geography advantage, even if there was no net intentional gerrymandering.
They did not show that. They made specific assumptions that went into their simulations. For example, one can draw random maps without regard to race, or one can draw districts that comply with the Voting Rights Act.
Current election returns confirm my analysis. The average margin is R+1.1% and the median seat is R+1.5%. That is a 0.4-point advantage for Republicans. Based on specific results so far, they have a 1-2 seat advantage. That may change slightly, but basically my estimate is correct.
I note that at present, the House appears headed for a likely outcome of 221 R, 214 D seats. Maybe as few as 219 R, 217 D.
Sam, here's the relevant sentence from the Kenny paper I was referring to: "To win a majority in the US House of Representatives under the enacted plan, Democrats need more than 51.1% of the national two-party popular vote, just 0.14 percentage points more than under the nonpartisan baseline. "
That paper modified their SMC algorithm in a few states when needed to draw enough minority districts to comply with the Voting Rights Act. I could imagine that the high uncertainty in Kenny's votes-to-seats model might lead to an erroneous calculation of the national bias. But their district-drawing method looks sound to me.
Ohio Issue One (would have established anti-gerrymandering commission) failed. Ballot language from Republican Sec'y of State, approved by R dominated Ohio Supreme Court, was grossly misleading to voters, too many of whom were deceived into thinking "no" was the anti-gerrymandering choice. Mother Jones, among others, has the grisly details.
Some thoughts...
The Dobbs decision, consistent Democratic overperformance in special elections, Trump’s underperformance in Republican primaries, the Harris Campaign’s incredible ground operation with millions of volunteers, Trump’s shocking behavior and ugliness especially these last few weeks, Independents and newly-registered voters (unpolled!) overwhelmingly preferring Harris – IMHO, pollsters have failed to sufficiently take these factors into account.
Moreover, bad-faith pollsters have been releasing an unprecedented number of questionable, with the clear aim of impacting the polling averages such as 538. To an astonishing degree, they have succeeded in manipulating the media election narrative. (Likewise, Polymarket has been a case study in narrative manipulation.)
During this2024 election, especially, basing predictions on the 538 average, means trying to predict on the basis of skewed, highly-manipulated data. PEC should, in my opinion, have restricted itself to an average of just the high-quality independent polls.
Interestingly, the average of the highest-quality independent polls HAS NOT CHANGED.
It’s worth paying attention to one notable pollster exception, one who doesn’t do manipulative weighting or make a whole bunch of hidden assumptions, is Ann Selzer. There are good reasons why she is perhaps the most respected and revered pollster in America.
A prediction
Here are my predictions, going out on a limb and clearly at odds with the consensus:
PRESIDENCY: Kamala Harris wins at least six of the swing states: MI, WI, PA, NV, GA and NC. Stunningly, she also wins FL and IA. Texas will be surprisingly close. It’s possible that Trump wins AZ.
Harris wins 355–183 with AZ, 344–194 without it.
https://www.270towin.com/maps/zRBxY
SENATE: Democratic winners include Gallego AZ, Alsobrooks MD, Slotkin MI, Rosen NV, Casey PA. Baldwin holds in WI, as does Sherrod Brown in OH and Tester in MT. Allred defeats Cruz in TX, and Mucarsel-Powell unseats Scott in FL. Osborn loses a close race against Fischer in NE.
Democrats with caucusing Independents expand control to 52–48.
HOUSE: 226–209, Democratic control. I would not be surprise to see Dems reach 230.
Based on what I see in the special-election data, I think a realistic upper-bound estimate for Harris performance would be the +2.5 point estimate I made. Call it the Hopium Bound. It includes Arizona, North Carolina, and Georgia, and adds up to Harris 319 EV, Trump 218 EV.
I think we're too polarized for things to go any further. Of course I could be wrong!
That could well be. But I think the massive gender gap, the Selzer poll, Harris’ huge advantage in the polls of those who had already cast and Early Vote (far too little emphasis on those polls!), likewise the gaps in specialized large-sample polls of specific demographics (younger voters; Blacks; Latino/Hispanic voters) paint a very different picture.
Add to that the Florida poll of people who had voted early. It showed Harris leading with +1, despite the fact that Rs had a 10+ point advantage in voters!
In other words, despite the polarization, there is a huge swathe of Republican voters (say, almost half of Haley’s voters) who are voting for Harris rather than Trump, because the cannot abide the Fascist. Likewise, there have been repeated indications of Independents in most swing states breaking heavily for Harris.
In other words: The political polarization that focuses on Party is camouflaging rather than clarifying the true underlying dynamics of what is happening.
And pollsters clearly have not adjusted to the post-Dobbs reality, to the referenda results, and to the almost-unbroken series of Democratic overperfomances in the special elections.
(I realize full well that I am almost alone in believing that Harris can win Iowa or Florida. That said, 319 EV would be an excellent result – and well beyond "The Margin of Steal"; hard to see how SCOTUS and others courts, or disloyal electors, can mess with that!)
Kudos to you for taking the special elections into account!
Florida has already been called for Trump, so your predictions are starting to be wrong. I'm all in for the "Hopium" result myself, but you've gone overboard.
I deny that I made a prediction for President.
Not yours! Arctic Stones's!
Yes, indeed. I was very, very wrong.
i was curious about the switch of wisconsin, dark blue in the unadjusted poll map to light blue in the tilt by 1.5 points to harris map. if the results in the tilted towards harris map would generate better results for her, why did the color of the map seem to indicate a worse outcome for her in wisconsin?
Oh yeah, I should probably fix that. Don’t pay any attention to those little color differences!
Generally the New York Times models are doing well, from a technical standpoint. Here is the projection for Iowa, which Trump is headed for winning by 12 points. That is pretty different from the Selzer poll, which had shown Harris up by 3. https://www.nytimes.com/interactive/2024/11/05/us/elections/results-iowa-president.html#forecast
Some of the Vote Maximizer recommendations for close House races that I donated to are coming home. E.g., Riley has just been declared winner in NY-19.
A couple in CA remain very close, though only about 50% of votes counted.
Thank you for using the information!
At the moment, in 11 districts Democrats are in the lead to flip Republican districts, and 5 Republicans are in the lead to flip Democratic districts. If this holds, a net gain of 6 seats would be enough to give Democrats control of the House.
In these districts, the average Vote Maximizer per-voter power is 87 out of a possible 100. So it appears that Vote Maximizer was successful in identifying important districts.
I also note that of these 16 districts, 6 were drawn by an independent citizens' commission, 4 were drawn by a court (and 3 of those in New York by an Electoral Innovation Lab affiliate), and 2 used ranked-choice voting.
Wednesday morning, 10:00am: the net gain by Democrats might not hold up. Current flips are now 6 D's, 5 R's, a net gain of 1 seat. https://www.nytimes.com/interactive/2024/11/05/us/elections/results-house.html
"Renowned Pollster J. Ann Selzer said Tuesday she would be reviewing her data to determine why a Des Moines Register/Mediacom Iowa Poll released just days before the election produced results so far out of line with former President Donald Trump's resounding victory." [Des Moines Register at 11:40 pm CT]
Re: ".The House of Representatives has little net overall bias, thanks to reductions in gerrymandering in the last decade (three cheers for anti-gerrymandering reformers!). Therefore we can take the national vote as a measure of who will control the House in 2025." I thought that https://www.pnas.org/doi/10.1073/pnas.2217322120 showed that there's about ten-seat (or 1% vote) advantage for the GOP in the House due to geography advantage, even if there was no net intentional gerrymandering.
Geoff, thanks for the link. An interesting study.
They did not show that. They made specific assumptions that went into their simulations. For example, one can draw random maps without regard to race, or one can draw districts that comply with the Voting Rights Act.
To repeat my statement: in our pre-election calculations there is approximately 0% bias in the House of Representatives, meaning that a majority national vote would produce a majority caucus most of the time, for either party. See https://www.theatlantic.com/ideas/archive/2022/11/midterm-house-elections-2022-gerrymandering-new-york/672145/
Current election returns confirm my analysis. The average margin is R+1.1% and the median seat is R+1.5%. That is a 0.4-point advantage for Republicans. Based on specific results so far, they have a 1-2 seat advantage. That may change slightly, but basically my estimate is correct.
I note that at present, the House appears headed for a likely outcome of 221 R, 214 D seats. Maybe as few as 219 R, 217 D.
Sam, here's the relevant sentence from the Kenny paper I was referring to: "To win a majority in the US House of Representatives under the enacted plan, Democrats need more than 51.1% of the national two-party popular vote, just 0.14 percentage points more than under the nonpartisan baseline. "
That paper modified their SMC algorithm in a few states when needed to draw enough minority districts to comply with the Voting Rights Act. I could imagine that the high uncertainty in Kenny's votes-to-seats model might lead to an erroneous calculation of the national bias. But their district-drawing method looks sound to me.
Ohio Issue One (would have established anti-gerrymandering commission) failed. Ballot language from Republican Sec'y of State, approved by R dominated Ohio Supreme Court, was grossly misleading to voters, too many of whom were deceived into thinking "no" was the anti-gerrymandering choice. Mother Jones, among others, has the grisly details.