2024 Presidential Election -- new thread for the final week

What will be the outcome of the 2024 Presidential Election

  • 269-269 Tie - I am someone who likes maximum chaos

    Votes: 3 3.5%

  • Harris wins 270-285 EVs - She basically wins by one or two states

    Votes: 14 16.3%

  • Harris wins 285-305 EVs - a close race, but she largely sweeps the swing states

    Votes: 36 41.9%

  • Harris wins 306+ EVs - Ann Selzer was right!! It ain't actually that close

    Votes: 21 24.4%

  • Trump wins 270-285 EVs - The race where it doesn't get called until next week

    Votes: 3 3.5%

  • Trump wins 285-305 EVs - The 2016 sequel

    Votes: 6 7.0%

  • Trump wins 306+ EVs - The MAGA Mandate, Harris concedes by midnight

    Votes: 3 3.5%
  • Total voters
    86
  • Poll closed .
Status
Not open for further replies.
House P said:
PA is likely to take longer than election night, while AZ and NV could take days.
Click to expand...

Just for reference, in 2022 in Arizona the AG race was extremely close (final margin was 280 votes, roughly 0.01%) and after the mandatory recount, some county-level shenanigans, and a bit of litigation the winner was finally declared on December 29. So, uh, let's hope this thing doesn't hinge on Arizona.
 
OldPhiKap said:
Hillary had a bigger polling lead in the month leading up to the election than Harris does, even with Harris having Dobbs to help her. That may account for some of it to the extent the models are similar.

And of course, Hillary got 3 million more votes than Trump did. But not in the right places. While it is possible that Harris takes 5 or even 6 of the swing states, it is also possible that Trump does the same.
Click to expand...
Yep. The new fivethirtyeight model (now that Silver was let go by ABC and took his model with him) relies mainly on the polling data at this point, as does Silver's model. Clinton had a lead in the polls for most of the swing states in 2016. It's much closer in the polls than it was in 2016. The polls were unwittingly underrepresenting Trump voters in 2016.

Of course, pollsters have changed their polling methodology since 2020. And in addition, there has been a dramatic increase in partisan pollsters since 2016. So it's hard to know if they've accurately captured Trump voters, or overrepresented them, or still underrepresented them. But the fivethirtyeight model (and Nate Silver's model) think Trump's chances are better in 2024 than in 2016 because the polls say his chances are better in 2024 than they did in 2016.
 
acdevil said:
Some of the outputs don’t pass the smell test for me.

Trump wins the popular vote 33% of the time? I’d happily lay money on that at 1:2.

I have a feeling the model has been bamboozled by the partisan polls flooding the zone.
Click to expand...
Yeah. Trump's not winning the popular vote.

Electoral college is a different thing all together.
 
CDu said:
Yep. The new fivethirtyeight model (now that Silver was let go by ABC and took his model with him) relies mainly on the polling data at this point, as does Silver's model. Clinton had a lead in the polls for most of the swing states in 2016. It's much closer in the polls than it was in 2016. The polls were unwittingly underrepresenting Trump voters in 2016.

Of course, pollsters have changed their polling methodology since 2020. And in addition, there has been a dramatic increase in partisan pollsters since 2016. So it's hard to know if they've accurately captured Trump voters, or overrepresented them, or still underrepresented them. But the fivethirtyeight model (and Nate Silver's model) think Trump's chances are better in 2024 than in 2016 because the polls say his chances are better in 2024 than they did in 2016.
Click to expand...
Makes sense, but is an indictment of the model. And since these models playing both sides of the outcome 538 and Silver can say they got it right, and keep their credibility intact. It reminds of market experts who tell you their "base" case for the S&P 500 and then tell you that is their base case, which they warn you could be wrong if their assumptions prove incorrect. Six months later after their base case was way off, they are still making predictions on TV or in the media. If these people were that good, they would sell their information to the highest bidder and keep it quiet.
 
No matter what happens this will be the strangest election for me. I will never have been sadder to win a bet, and conversely, I will have never have been more happy to lose a bet, in my life.
 
tendev said:
Makes sense, but is an indictment of the model. And since these models playing both sides of the outcome 538 and Silver can say they got it right, and keep their credibility intact. It reminds of market experts who tell you their "base" case for the S&P 500 and then tell you that is their base case, which they warn you could be wrong if their assumptions prove incorrect. Six months later after their base case was way off, they are still making predictions on TV or in the media. If these people were that good, they would sell their information to the highest bidder and keep it quiet.
Click to expand...
Yes, these models aren't terribly useful in predicting the winner in a close race. They can only tell you that it's close. And that's what they are saying - they are trying to be very clear that they are predicting the winner. I guess that's a lesson learned after 2016.

And they aren't at all useful if the polling data aren't sound (as we saw in 2016, where there appears to have been a systematic error in the polling).
 
tendev said:
Makes sense, but is an indictment of the model. And since these models playing both sides of the outcome 538 and Silver can say they got it right, and keep their credibility intact. It reminds of market experts who tell you their "base" case for the S&P 500 and then tell you that is their base case, which they warn you could be wrong if their assumptions prove incorrect. Six months later after their base case was way off, they are still making predictions on TV or in the media. If these people were that good, they would sell their information to the highest bidder and keep it quiet.
Click to expand...
You can validly evaluate these prediction gurus and it’s not that hard. You just can’t do it on the basis of a single race. To evaluate Nate take all of his predictions: House races, Senate races, Presidential races, etc. and put them in bins. One bin has races where the leading candidate was said to be a 90-99% fave, then a bin for 80-89% faves, …., 59-50.1% faves. Then look at the results. Did approx 95% of the first bin win? 85% of the second bin? And so on. If most of the bins prove accurate then Nate’s prediction model is valid. If not, then it isn’t. So there is accountability for these prognosticators - if you want to do the work.

Of course none of this will detect if Trump is a special “unpredictable” case. That is unknowable unless, god forbid, he runs several more times!
 
The more I understand about the Anne Selzer poll - and her history of putting out surprise polls - the more confident I am that tomorrow will be a huge day for Harris.

The Selzer poll had Biden -18 in June, Harris -4 in September, and Harris +3 over the weekend.

She is not trying to model off past elections at all to predict results. She is doing polling the old-fashioned way of talking to 1,000+ Iowans to find 800 who have either voted or will definitely vote and then tweaking the weighting of each of the 800 to match the Iowa census. Essentially she's not trying to forecast turnout - she's letting the voters tell her who is turning out. I don't think any of the major pollsters are taking this approach which would seem to explain why she's such an outlier.

It's been highly accurate the past 4 election cycles and foretold a surprise Trump victory in 2016 and a surprise very close election in 2020.

There is no logical reason her poll would be off this time around and even if she misses wildly by like 6 points and Iowa is Trump +3, that should still predict a big national win for Harris.

And the fact that Repub pollsters Emerson and IinsiderAdv just came out with big Trump leading polls in Iowa shows that the pollsters are cooking the data unless you believe Selzer's methodology is suddenly broken after being very precise for many years.

The best critique of the Selzer poll in my mind is not that it's off but instead that Iowa is not representative this cycle because its recent 6-week abortion ban has overly energized women in Iowa in a way we won't see in the other battleground states. Ok, maybe. But the 7-point move toward Harris from Sept to now in Selzer's poll has all happened after the July abortion ban.

Maybe this will be the cycle that Selzer's poll is wildly inaccurate, but I believe the other explanation is WAY more plausible. Modeling 2024 election turnout off 2020 election turnout is mistakenly skewing the polls in Trump's favor.
 
Last edited:
Status
Not open for further replies.
Back
Top