Elections Why Were The Polls Wrong Yet Again?

[Natural Order]

Polls, lol. Are you cans even awake?

 

[White Whale]

Donald Trump just created five new swing states for 2028

Exit polls suggest that Republicans could take more states from Democrats in 2028 after Trump flipped 6 of the 2024 swing states.

www.newsweek.com

Donald Trump Just Created Five New Swing States for 2028

Five states could potentially become swing states in 2028, according to election exit poll data.

Exit polls suggest that Republicans could take more states from Democrats in 2028, including Minnesota, New Hampshire, New Jersey, New Mexico and Virginia. In each of those states, the Democrats' lead was narrowed to a margin of 3 to 6 points, down from leads of up to 16 points in the 2020 election.

The small leads for Democrats mean all five states could be in play for Republicans in the 2028 elections, in addition to the seven swing states from this election.

The polls were so bad Donald Trump may have created five more swing states.

Pollsters overlook many people.

 

[Madmick]

Out of dozens and dozens of late race polls, a couple were moderately off.

On the whole they were quite accurate.

The latest polls before the election showed an extremely close, within the margin of error race, and it was an extremely close, within the margin of error vote.

Nate Silver had his final projection at:

You can't get any more uncertain than that.

Certainly the specific ones that were off need to investigate and see if there's something that can be done better in the future.

But the narrative some people want to sell of "don't believe the polls" is pure BS.

That's the electoral simulation. That isn't the projection for the votes themselves. Derp.

 

[Gamer2k4]

When you demonize people for backing a particular candidate or party, they're going to be a lot less likely to tell you that's who they're voting for.

 

[Andy Capp]

It's probably wishful thinking on my part but I'd like to open an honest discussion about the nature of polls, whether or not they have provided utility in prior elections and why they've seemed to be off for almost a decade.

I understand that we're ultimately looking at a situation in which one of only two predictions are going to be correct. But the Iowa Selzer poll, amidst almost every other that came out in the weeks and months leading up to the election, demonstrate that something is seriously wrong with either the way in which the data is being gathered, or the reception to the prompts for that data.

I myself will not get rustled if you just come in to tell me how stupid I am. Maybe I am stupid. That's the reason I try to learn from different people, and it's the reason I'm making this thread. I'd ask that you be respectful to each other however. If possible, let's not let this devolve into a shit-slinging contest.

The fact is that presidential polls have, up until last month, never underrepresented a party three times in a row. Something is seriously off. I realize that there's probably no one single element, but a number of contributing factors. I'd like to examine them.

I'm going to go out on a limb and say it's pretty silly to simplify it down to "they're liberals so they favor liberals". I don't think that every single major news network aside from Fox would just lie about data. Harris' own campaign does admit in this interview that their internal polling never had a great outlook, but that was not common knowledge until after the election:choo


So what's going on? Everyone who has ever said "I don't care about polls" has, at one point or another, utilized polls whenever it suits them. But unless people figure out what went wrong and how it went so wrong, they'll become less and less useful.

Conversely the betting odds had it precisely correct for the most part. Harris had a short lived surge but electionbettingodds.com had it 312 for Trump almost the entire time. Some of the prominent people who are by now in hiding for a bit insisted this was due to a "mystery bettor" who placed two large bids for Trump, essentially thumbing the scales. But I think it was about a million maybe, which represented less than 1% of the pool.

Anyway I'm curious to get feedback on what could have caused the polls to get it wrong three times in a row. Some have said that Trump supporters are less likely to engage with analysts, or less likely to admit they voted for Trump. I can see this, but I have a hard time believing this is the sole element.

I think the polls don't capture people choosing to just stay home come election day and that appears to be where all the apparent support for Democrats fizzled. It's surely not the only reason but I believe that one of the most significant is they way Harris was "selected" as the candidate. The Democrats failed to adequately paint the picture of necessary expediency and the need for a united party against Trump and MAGA---they should have attacked them more instead of the joy thing, I think, sadly--and instead avoided that out of fear of undermining Harris instead of helping.

Hindsight is 20-20, as they say, but I asserted this (the bold text) right after Biden stepped down. I believe a lot of people were turned off by the apparent side-stepping of discussion of her "selection". I use quotes because with the time they had they really had no other option and people should have been discussing that in the media instead to let people develop and understanding of the alternatives, such as they were.

Having said that, the misinformation campaign was very strong and I think a lot of Dems who weren't sure what to believe got swept up in it.

To boot, I think there are big flaws in the way polls are being conducted and reported on. Consider that woman in Iowa who is getting shat on for her outlier poll.

When reporting polls, they (whatever media source) almost never correctly state the true margin of error. It should be +/- some number followed by some frequency, usually 19 times out of 20 or 95% of the time. Usually, there's no margin of error stated in the reporting, and even when there is, it almost never mentions the latter bit. Anyway, in the case of that pollster, it seems like that one poll was just one of the 5% but she's somehow been shamed right out of her profession. Sigh.

Note the liberal (lol) used of "I believe" and "I think". I can be convinced otherwise if new information/data comes to light.

 

[jeremyemilio]

I haven't seen the EC margin separation that asserts that claim. I know urban center turnout was lower than 2020 which really impacted kamala in swing states.

You got the swing state comparisons handy?

Yeah. CNN has them at:

https://www.cnn.com/election/2020/results/president and https://www.cnn.com/election/2024/results/president

Here's what it looks like:

Arizona - Biden 2020 = 1,672,143 ; Trump 2024 = 1,770,242
Georgia - Biden 2020 = 2,473,633 ; Trump 2024 = 2,663,117 (Kamala actually got more votes in Georgia than Biden in 2020)
Michigan - Biden 2020 = 2,804,040 ; Trump 2024 = 2,816,636
Nevada - Biden 2020 = 703,486 ; Trump 2024 = 751,205 (again, Kamala got more votes in Nevada than Biden in 2020)
Pennsylvania - Biden 2020 = 3,459,923 ; Trump 2024 = 3,543,579
Wisconsin - Biden 2020 = 1,630,866 ; Trump 2024 = 1,697,626 (and Kamala outperformed Biden 2020 here, too)

If there's any argument to be made about voter turnout that is backed up by numbers, it's probably that Trump's pre-election hedge play to discredit mail in ballots as a means of contesting the election if he lost may have been what cost him the election in 2020, suppressing his own vote with his own conspiracy BS. The guy's a buffoon. But he won this election decisively, and it wasn't because Kamala's voters didn't turn out.

 

[jeremyemilio]

I think the polls don't capture people choosing to just stay home come election day and that appears to be where all the apparent support for Democrats fizzled. It's surely not the only reason but I believe that one of the most significant is they way Harris was "selected" as the candidate. The Democrats failed to adequately paint the picture of necessary expediency and the need for a united party against Trump and MAGA---they should have attacked them more instead of the joy thing, I think, sadly--and instead avoided that out of fear of undermining Harris instead of helping.

Hindsight is 20-20, as they say, but I asserted this (the bold text) right after Biden stepped down. I believe a lot of people were turned off by the apparent side-stepping of discussion of her "selection". I use quotes because with the time they had they really had no other option and people should have been discussing that in the media instead to let people develop and understanding of the alternatives, such as they were.

Having said that, the misinformation campaign was very strong and I think a lot of Dems who weren't sure what to believe got swept up in it.

To boot, I think there are big flaws in the way polls are being conducted and reported on. Consider that woman in Iowa who is getting shat on for her outlier poll.

When reporting polls, they (whatever media source) almost never correctly state the true margin of error. It should be +/- some number followed by some frequency, usually 19 times out of 20 or 95% of the time. Usually, there's no margin of error stated in the reporting, and even when there is, it almost never mentions the latter bit. Anyway, in the case of that pollster, it seems like that one poll was just one of the 5% but she's somehow been shamed right out of her profession. Sigh.

Note the liberal (lol) used of "I believe" and "I think". I can be convinced otherwise if new information/data comes to light.

See my post just above ^^^.

Trump's 2024 state by state vote totals beat Biden's 2020 totals in each of the seven swing states. Kamala actually got more votes than Biden in three of the seven swing states and still lost all seven where Biden won all seven. It's just demonstrably false that Kamala lost because the democratic vote "fizzled" and stayed home.

 

[JudoThrowFiasco]

Yeah. CNN has them at:

https://www.cnn.com/election/2020/results/president and https://www.cnn.com/election/2024/results/president

Here's what it looks like:

Arizona - Biden 2020 = 1,672,143 ; Trump 2024 = 1,770,242
Georgia - Biden 2020 = 2,473,633 ; Trump 2024 = 2,663,117 (Kamala actually got more votes in Georgia than Biden in 2020)
Michigan - Biden 2020 = 2,804,040 ; Trump 2024 = 2,816,636
Nevada - Biden 2020 = 703,486 ; Trump 2024 = 751,205 (again, Kamala got more votes in Nevada than Biden in 2020)
Pennsylvania - Biden 2020 = 3,459,923 ; Trump 2024 = 3,543,579
Wisconsin - Biden 2020 = 1,630,866 ; Trump 2024 = 1,697,626 (and Kamala outperformed Biden 2020 here, too)

If there's any argument to be made about voter turnout that is backed up by numbers, it's probably that Trump's pre-election hedge play to discredit mail in ballots as a means of contesting the election if he lost may have been what cost him the election in 2024, suppressing his own vote with his own conspiracy BS. The guy's a buffoon. But he won this election decisively, and it wasn't because Kamala's voters didn't turn out.

Ahhh, I see what your initial point is now. Yeah, it was a clear cut blowout. Dems may cling to the notion the PV margin was tight but Donald/ GOP swept every major contest the candidates focused on. It was Donald in 4 games despite if people want to point out the score was 2-1 in each game

 

[Andy Capp]

See my post just above ^^^.

Trump's 2024 state by state vote totals beat Biden's 2020 totals in each of the seven swing states. Kamala actually got more votes than Biden in three of the seven swing states and still lost all seven where Biden won all seven. It's just demonstrably false that Kamala lost because the democratic vote "fizzled" and stayed home.

Ok. Good to know.

 

[VulcanNervPinch]

That's the electoral simulation. That isn't the projection for the votes themselves. Derp.

The simulation is driven by the EC votes, which is driven by the state popular votes.
In a close race, getting the state votes significantly wrong will impact the simulated winner.
Derp.

 

[Err]

LOL, wtf are you talking about?

They were worse than ever before. The final popular vote predictions from the major pollsters projected Harris to win by anywhere from 0.6-1.8 pts. She got stomped by 1.6 points. Them being off by 2.2-3.4 points might be within the realm of predictable error, statistically, but it's awful. It's just a terrible prediction. That's nearly triple the error from Clinton vs. Trump in 2016.

We're not talking about a County or a State. This is the whole enchilada. There are thousands of polls contributing to it, and a greater pool of votes to stabilize the estimate.

It's a question worth asking because misses favoring Democrats are becoming so consistent it raises an eyebrow.

The writing was on the wall for anyone that wanted to look at the polls objectively, especially when weighed against his performance in 2020 and 2016.