Either that or ABC has Nate by the balls so the numbers are “correct”. For all we know, if 538 wasn’t owned by ESPN, the marker could be so much different.
Wang’s model appears to have no bias influenced by the media, and his indicators are showing a split senate. Tim Kaine is about to have a lot of action in two years.
I don’t have numbers or cites at my fingertips because this is just from reading constant Twitter commentary but, yes, the number of polls has gone dramatically since 2012. Part of that is difficulty in polling in general and part is because of aggregation. Putting out a public poll used to be a good way to get your firm’s name out there and in the news but, with the obsession with aggregation, these days your number just gets tossed into the blender at best and, at worst, fails to meet the aggregated average and gets mocked for being an outlier and your reputation suffers. So places are either restricting themselves primarily to private polling for organizations or else backing off of election polling in general.
Last I saw, the people at 538 were denying that it was a factor and Public Policy Polling tweeted that, yup, it was part of the reason. Obviously 538 has a vested interest in denying that their business model is hurting election polling in general.
To be fair, Kerry also didn’t perform especially well throughout the same region:
Minnesota: +3.5
Wisconsin: +0.4 (!!)
Michigan: +3.4
Pennsylvania: +2.2
Also, New Hampshire (+1.3) was closer than you might guess based on Obama’s success. Kerry, of course, lost but my point is just that these “blue states” still need to be worked at.
OK, but Kerry lost by 2.5% overall. So compared to Kerry’s nationwide (negative) margin, it looks more like:
Minnesota: +6.0
Wisconsin: +2.9
Michigan: +5.9
Pennsylvania: +4.7
Compared to Obama’s 3.86% margin in 2012, he had:
MN +3.8
WI +3.1
MI + 5.6
PA +1.5
Both did better in these four states than they did nationwide, but Kerry more so than Obama.
I wouldn’t be surprised if Hillary’s margin in at least one of these four states turns out to be smaller than her nationwide margin.
Now consider Obama 2012 in these states:
NC -5.9
AZ -12.9
TX -19.6
Hillary’s gonna do WAY better than that.
I don’t think that trying to adjust Kerry’s state performance against his national percentage is especially meaningful. Kerry wasn’t especially popular, including in the upper Midwest. Clinton is more popular including in the upper Midwest. She’s doing (based on current polling) better in the upper Midwest than Kerry did.
I think you guys are reading too much into the pretty picture.
Clinton’s chances on 538 have barely changed at all. They’ve gone from like 48.6% to 51.1% and numbers like that. As it happens their map has no color for “no favorite” but the prediction is not substantially different.
I’m not sure what you are looking at, but I’ve been seeing her chances on 538 increase today almost by the minute. Right now over 68%.
I believe RickJay was referring to 538’s predictions for either FL or NC.
Must be. The movement may have been slight, but just since yesterday in both FL and NC, 538’s gone from Trump being a slight favorite to Clinton being a slight favorite. Ditto NV, only more pronounced there - Hillary’s up to a 54.5% chance of winning in NV.
Meanwhile, nationally, at 7am, Hillary’s win chances were 63.6%, and now they’re up to 68.3%. And her lead in the 538 polling average is up to 3.5%.
There is really little difference in the numbers you guys and throwing around. Her chances on 538 were about two in three a few days ago, and are about two in three now. The difference between 63.6 and 68.3 is noise.
The question, really, is whether you believe the Silver model or the Wang model or something in between. THERE, there’s a difference.
I’m bullish on Hillary in Florida. Not so much in North Carolina, though it could be in play 4 years from now as demographics overwhelm even the most aggressive vote suppression efforts. But vote suppression will win North Carolina for republicans in 2016.
If I had to bet right now I’d say Hillary takes it 307 - 231.
Re: early voting, I basically understand Silver’s desire not to go against the history here, however, it’s possible that the massive surge in Hispanic voters may force him (and us) to rethink those assumptions in the future. If nothing else, the surge in Latino voters clearly shows that they are even more determined to be a political force to be reckoned with. However, it’s entirely possible that Trump’s empire strikes back at the ballot box tomorrow. When in doubt, I go with polling data and history, both of which look somewhat safe for Hillary tomorrow, provided there’s still a sense of urgency among her voters.
This is awesome:
Why is that awesome?
If you look at the individual polls, Clinton is consistently polling 4-7 points above Trump in Michigan. I don’t really see it as a 25% chance of a Trump victory, which as he says, depends on Dems sleeping and Repubs showing up in droves.
Nate Silver’s coin toss analogy is apt in that Nate Silver fought the whole ‘horse race’ thing early on and others got attention (and clicks) by playing it, so now he’s using ‘coin toss’ a lot more. Being caught short by Trump winning the nomination seems to have turned him into the Unskewed Polls guy of 2016.
He’s had Minnesota listed as variously a battleground or… I dunno, forgive me I’m at work and I’m forgetting the words - but much less than the solid lock it is for Hillary Clinton. The latest 4 polls show her up by 5,8,9 and 10 points.
The problem with the coin flip analogy is, it’s not a coin flip we’re dealing with. We’re dealing with real humans, with real emotions about this race and about politics in general. And we’re dealing with two candidates, one of whom has laid out a real complex operation that is dedicated to mining votes, and another campaign that, while highly efficient in terms of votes per dollars spent, will ultimately probably fall short in getting all of the voters it needs. The debates, and the results of those debates, were in my view a harbinger of things to come. Clinton’s discipline and focus on the task of winning will trump the Donald’s ability to attract attention of the masses. Trump needed to cross over more and build a broader constituency, and he couldn’t quite do that. He could come surprisingly close in Michigan and Pennsylvania – that I can believe. But I think Clinton’s going to grind out the victory in the end.
Interestingly, Nowcast and Polls-Plus are both 70%ish for Clinton. It’s just Polls-Only that’s still in the 60s (albeit 69.4%). Will all three will converge tomorrow?
Hot damn, though, am I glad to see that 70% again!
They are three independent Monte Carlo simulations, so even going by the same variables, you’d expect them to differ just by randomness. But for the last run tomorrow, he might just go with a single run since they’ll all be using the same data.
[QUOTE=Nate Silver]
We usually run at least 20,000 simulations for each version of our model each day. That’s a lot, but it still produces a small amount of sampling error. You shouldn’t worry too much when win probabilities change by less than a percentage point.
[/QUOTE]
Interesting.
PEC started off early on stating that roughly C+4 +/-3 would be where this election would end up. End of May he unveiled the state snapshot and at that point has it at meta-margin C+4.2 with the EV at 336/202. His forecast has remained relatively stable. Today’s meta-margin is C+2.7 with the EV of 312/ 226.
Silver no one here needs to be told has during that time been all over the place, now adjusting back to the middle on election eve, flipping FL, NV, and NC all light Blue and giving an average prediction of 299.7/227.5 and a no tossups prediction of 323/215 (a major bump over 1 d). Popular vote has not flipped as much, from +7 to virtually tied and now C+3.5.
RCP 4-way is now C+3.2. RCP no tossups map is C 272/266.
Weird aint it how Wang’s early prediction of how polls run and where they’d end up turned out pretty dang on the money and that after lots of more and less uncertainty Silver’s election eve predictions have come round to what wang was saying all along, even if the residual anything can happen factor is still as fairly high as Wang’s is fairly low.
FWIW I’ve made a sheet with PEC’s and 538’s state by state margins and probabilities. I’ll leave to statisticians in days to follow to come up with Brier scores and the like but if I can on election day I will try to keep up some with reporting what each of them predicted for states as they get called and with what degree of confidence.
I appreciate your eyes on the numbers, DSeid. Maybe you can put up your spreadsheet when it’s all done so the number/statistics geeks here can have at it?
There’s nothing weird about it. Silver, like most others, was pretty confident (80-90%) until the news cycle started moving against Clinton, and that movement started before the Comey announcement. It actually started with Obamacare’s rate hike. Silver goes where the data takes him. He’s not wrong.