Nate Silver kind of looks like a doofus for showing Harris winning all those swing states but not being favored. He’s taking heat for his “convention bounce remover” … let’s see where it’s at in 2 weeks,
If accuracy is important after an audit bush did beat gore by a couple 100 votes. Honestly Florida was a functional tie. Sometimes when you flip a coin it lands on its edge not often but it can happen.
Betting odds fluctuate even more than the polls and at this point in the election cycle they're even less meaningful. As an example on September 7 in the 2016 election cycle the betting odds had Hillary Clinton over Donald Trump by 71% to 29% margin. Although the odds may outperform the polls within the last couple of days prior to the election at this point in the cycle they're virtually meaningless. In fact even up to election day in 2016 they were way off. Although the national polls had Hillary Clinton winning the swing state polls actually had her ahead by such a small margin that most of those sampled close to the election were within the margin of error meaning they were statistical dead heats. Betting odds from 2016.
Ah, I don’t think he should be looking like a dufus for letting his model run the way he programmed it. He gave his reasons for the adjustment. And I don’t think we should be too hard on models that diverge from the consensus. These outliers can provide additional information to the population.
I think I'd prefer that the model be based on the best current data and not an arbitrary adjustment. Silver talked a lot of crap about 538 for sticking to their model and their "adjustments", showing Biden winning, even when Biden was clearly losing. Now Silver is showing Trump winning, when he is clearly losing, but has an explaination ... Here's a blast from the past... Why I don't buy 538's new election model - by Nate Silver
To be fair, I don’t think this adjustment is any more arbitrary than any other decision being made by any of the model builders. The magnitude and duration of the adjustment is no doubt trained on historical tendencies, just as is done with the other parameters of the model. It of course could be that Kamala’s convention bounce will not resemble that of the average candidate of the past, but this same risk exists for all the other model assumptions. The real question is whether the probability is more accurate with or without the adjustment. Rational arguments can be produced on both sides of this question, and the adjustment cannot be tested empirically for reasons outlined in the link I posted above, so my best advice would be to split the difference.
Is Silver's model making an adjustment for polling bias? I've not seen a model yet that is. Silver is probably a little low on his odds of Trump winning, because while he is making his adjustments, he is taking the polling data at face value. Does a poll that shows Kamala leading Trump by 3 in Wisconsin mean she's really ahead by 3 in Wisconsin? The last two elections would tell us there's likely a 5-point bias in that poll. I think it is a little unrealistic to believe all of these media pollsters are suddenly going to dial in Wisconsin and Michigan perfectly after missing so badly in '16 and '20.
Thankfully the last couple of elections have shown us that they've undercounted Democrat turnout so Harris' lead may be larger than what shown today in the polls
I wonder how dick Cheney (can anybody sane call him anything but a true Republican) endorsing Kamala will impact rational supporters of trump.
Jesus himself could endorse Trump and his supporters would start talking about how they really like Judas better anyway.
I believe he weights different polls differently based on their past accuracy but does not do a bias adjustment. Such an adjustment would be a risky venture, as there is no guarantee that past bias will translate into a future one. And others pointed out, the forecasts seemed to have done a fine job in 2022, so a bias adjustment then would have made the forecasts less accurate.
Harris is dropping in the polls as expected, that's probably why she did the CNN interview, which was heavily edited What CNN didn't show you:
