2024 Pre-Election Political Polls | POLL - Trump would have had 7 point lead over Biden

  • Thread starter Thread starter nycfan
  • Start date Start date
  • Replies: 6K
  • Views: 183K
  • Politics 
Enigma94 said:
So does the model provide a statistical chance that Harris wins the popular vote? I understand virtual certainty isn't precise language. It just seems overwhelmingly likely that Harris wins the popular vote, so any model that gives her a 17% chance of winning it while losing the EC would naturally mean it's unlikely she loses the EC. But then he has Trump as the favorite to win the EC. Maybe I'm overestimating the chance Harris wins the PV?
Click to expand...
If the model gives a 17% of Harris winning the popular vote (PV) but losing the EC and you know that Trump is predicted to win 52% of the time and Kamala 48%, you can do the rest of the math...

Kamala win PV and EC: 48%
Kamala win PV but lose EC: 17%
Trump win PV but lose EC: 0% (Admittedly, this is an assumption, but it seems a safe one. If there are situations where Trump wins the PV but loses the EC, they'd have to be statistically very small)
Trump win PV and EC: 35%

So, Harris would be projected to win the PV 65% of the time.
 
Enigma94 said:
So does the model provide a statistical chance that Harris wins the popular vote? I understand virtual certainty isn't precise language. It just seems overwhelmingly likely that Harris wins the popular vote, so any model that gives her a 17% chance of winning it while losing the EC would naturally mean it's unlikely she loses the EC. But then he has Trump as the favorite to win the EC. Maybe I'm overestimating the chance Harris wins the PV?
Click to expand...
You are. But again, this stuff is counter-intuitive, and I'm not sure it makes that much sense if you think about it enough.

I don't have the numbers in front of me because it's paywalled, but I'm guessing that the model has Kamala about 60% to win PV, maybe a little higher (there is a chance that Trump could win PV and lose election, though it's unlikely). But an unknown % of Trump's 40% might just be from temporal uncertainty. That is, something could happen between now and election day. Maybe Kamala dies. Maybe it comes out that the bear RFK dumped in Central Park was actually killed by Kamala after she had sex with it. Or maybe the economy crashes.

Accounting for those contingencies is sort of different than accounting for the possibilities that the polls are wrong, or that random chance produces certain outcomes. It's more like this, I think: [we don't know that much before Labor Day and thus all estimates should be discounted] * [the polls might be wrong or the votes in the tossups all go Trump's way]. Nate's statistical models multiplies those terms together to spit out a single percentage, but I think they are analytically distinct concepts and that's what confuses people.
 
Thanks to Snoop and Super for the analysis. It doesn't seem very helpful to use a model that assumes a 60-65% chance that Harris wins the popular vote. I understand there are externalities that can occur and must be accounted for in a statistical model, but I just can't give any credence to a model with that low a chance.
 
aGDevil2k said:
When Texas starts to turn blue (and it will happen), that's when the National Popular Vote compact will pick up steam. the GOP will be FUBAR without NY, Texas and California, so then they will want to act to go to the popular vote
Click to expand...
They would lose more with the popular vote than they do now.
 
Enigma94 said:
Thanks to Snoop and Super for the analysis. It doesn't seem very helpful to use a model that assumes a 60-65% chance that Harris wins the popular vote. I understand there are externalities that can occur and must be accounted for in a statistical model, but I just can't give any credence to a model with that low a chance.
Click to expand...
Well, you are assuming that she will win the popular vote. And maybe she will. But the whole point of using a model is to correct for biases in your assumptions. It might seem like Kamala is a slam dunk to win the popular vote . . . but if the polls are saying that the race is within 3 points nationally, and the polls are correct within their MOEs, then 65% would be a decent estimate of the probability. It might seem wack to you, but that's probably because it would seem wack that Harris was only up by 3 in the polls.

Again, I think a big part of the problem is the nature of the probabilistic estimate. Maybe it will help to use a concept from finance: risk versus uncertainty. Risk is considered to be variance that you can measure. Uncertainty is variance that you can't. For instance, a BB team down by 8 with 2 min to go has (IIRC) about a 10% chance of winning if they have the ball. So the winning team has a 10% risk of losing. Uncertainty would refer to the possibility that, say, someone slips a roofie into the winning team's gatorade and suddenly the players all get super groggy. That possibility is, we hope, extremely remote. But if you add up the all the remote possibilities, they could amount to something significant.

To build on this example, I remember watching the ESPN gamecast of a UNC-Duke game in 2020, the year the NCAA tourney got cancelled. The year we were awful because of injury and underperformance. I don't remember which game (I really don't remember that season much at all), but in one of them, UNC was up big on Duke in the second half. Like 20 points maybe. The gamecast said that UNC's chances to win were 99.5%. I told my wife that a UNC victory was actually a coin flip, because what the gamecast didn't know was 1) the ability of that team to self-destruct; and 2) the refs would often intervene for Duke in the waning minutes of a close game. Those factors would fall under the rubric of uncertainty. We don't know how to quantify the chance that the players would lose their collective minds, but it must have been significant because we did in fact lose.

So again, the 65% estimate includes both risk and uncertainty. It's counter-intuitive.
 
I hate issue based polling like that because such a big part of it is just allegiance to the individual. If they listed the actual policies, specifically with Russia and Ukraine, it wouldn’t be close. But Trump cultists will always select Trump if he’s an option.
 
The challenge right now is the pollsters are (rightly) scared of undercounting the Trump vote...again.

But there is something happening with registrations (ridiculously hard to model in polls because these arent typical likely voters) and small $$ donations. Could it all be missed? Are they missing the "something going on" like they missed in 2016? In 2016 we kept hearing something was happening....and it really was with the MAGAs. Maybe what we all feel happening over the last month (also hard to model when you have never had a 3.5 month campaign for president) is really confounding the pollsters?

That's very possible. That's one thing Bouzy (and Wasserman for that matter) are talking about that you dont see so much from Nate SIlver
 
HeelInTheOzarks said:
Laughing at Nate’s state by state probabilities basically saying Kamala wins and then still putting Trump to win on the top line.
Click to expand...
Look, I'm not a big Nate fan, but they aren't "his" state-by-state probabilities. He built a model. He puts the inputs into the model and reports what comes out. The model is almost assuredly better than your eyeball test.

I suspect that you're not taking account of correlation. Which would be almost impossible for you to do. Which is why we build models.
 
superrific said:
Look, I'm not a big Nate fan, but they aren't "his" state-by-state probabilities. He built a model. He puts the inputs into the model and reports what comes out. The model is almost assuredly better than your eyeball test.

I suspect that you're not taking account of correlation. Which would be almost impossible for you to do. Which is why we build models.
Click to expand...
100% true. The challenge Nate has, and why he's "lost" so many people that respected him, is that he is the classic narcissist that won't admit when he is deficient. I see him try to do it more but it's usually wrapped in wonky over-explanation. Never "my model wasn't build for a 3.5 month race, so it's going to be all over the place and we will see how it pans out"
 
You must log in or register to reply here.
Back
Top