The polls should also already be accounting for enthusiasm if they are likely voter polls.
Pennsylvania isn’t going to Trump. PA is a tease for Republicans most years. If RCP’s no toss ups map is reliable, then Trump’s best path to victory is to win CO, VA, WI, or NH. I think PA is a state he’d only win if he already won those other four.
VA is WAY down the list. According to 538, Trump’s best opportunities to steal away a state from Clinton are: CO @ 37.5%, NH @ 36.9%, or PA @ 31.5%. MI is in there at 30.9%, WI is at 27%, and VA is at 23.7%.
Virginia has only gone Democratic in the last two elections. PA has been Democratic for a lot longer.
Which is wonderful historical background, but completely ignores the polling in 2016.
That was my point, that PA teases the GOP every year and tricks them into pouring resources in.
Not according to every model I can seem to fine. Clinton is around 2% higher nationally, and Colorado she’s ahead by 1%. Plus, even with Colorado, Trump needs a little more to win.
But really, what is 1-2 percent in polls? It’s well within the sampling error, and more to the point, the momentum has been shifting to Trump. I don’t take satisfaction in this, mind you – just pointing it out.
I think Kaine will protect his him turf for Hillary. Colorado, OTOH, is much more in play and maybe Trump can steal another state.
Near-panic time, though it’s just due to a single poll – the Selzer (538’s highest rated pollster) national poll is Trump +2, and all by itself it brought down the polls-only from about 58-59% to 51.5%. I didn’t know a single poll could do that!
Yeah, that’s a bit of a butt-clencher.
It’s not the single poll. There were a bunch of other polls added to the model this morning.
And the Selzer poll isn’t getting much more weight than the new Quinnipiac poll (Clinton +3). I suspect it’s that CNN state poll that’s doing the real damage.
It was the single poll (based on my memory of the last few hours) – I checked back after the earlier update this morning, and it was still at ~58%, and then the only update that brought it down to 51.5% was the Selzer poll and a good VA poll (HRC +6).
I question your assumption that the model is so perfectly automated and instantaneous that we can determine which polls have affected it by looking at half-hour time samples.
We know enough about how the model works to know that a single national poll, which is not being weighted more than counter-balancing national polls and which did not move the national margin by itself very much, shouldn’t have moved the model that far. We also know that the CNN state polls and 3 other national polls trending Trump, which you report as not having moved the model at all, should indeed have moved it down. Knowing all that suggests that the model just didn’t update as quickly as you’d expect.
Or maybe this is too much Kremlinology. 
We can look at the “update” tab, which I do each time it changes (if I see it). Unless my memory is totally off, some time around an hour and a half ago, I checked, and it was at about 58%. The most recent polls in the “updates” tab were those CNN polls (and a couple of others), which didn’t look great for Clinton, but weren’t terrible.
Then I checked again about 40 minutes ago or so, and it was all the way down to 51.5%, with the only updates being a good VA poll and the Selzer Trump +2 poll. Crucially, Selzer is an A+ rated pollster – I think the only A+ pollster.
So yes, unless my memory is failing me, or the “updates” records new polls that haven’t actually been incorporated into the model (which I’m pretty sure specifically contradicts the explanations provided earlier by Nate and his team), then this is what occurred.
It’s possible my memory is off, but I’m about as certain as one can be without writing this stuff down that this is the case.
That is the assumption I’m questioning, at least in the context of this particular 28 minutes. I don’t think there’s any way the earlier update didn’t move the model at all, given its content (which included, again, all three other national polls trending Trump which we know are collectively weighted far more heavily than the Selzer poll).
The alternative is that the model is a bad model, since knowing the things we do know, it should not have shifted not at all in response to the first update (with lots of data largely pointing in one direction) and quite substantially in response to the second (with very little and contradictory data).
I’m not sure if those are weighted more heavily. Also, it wasn’t all bad – Quinnipiac had HRC +3 adjusted, and LA Times and RKM were a tie and HRC +2 adjusted. Selzer was Trump +2 (both reported and adjusted). That Selzer “adjusted” (which, IIRC, are the numbers that go into the model) is a lot worse than the other adjusted numbers. Plus, it’s A+ and the others are not, which should mean it’s weighted more heavily.
I don’t think the 1st update was as negative as you’re portraying, and I’m not sure if you’re right that those other polls are collectively weight significantly more (or more at all) than the Selzer.
Clinton - 45.1 Trump - 54.9 Yikes!
iiandyiiii, the weighting of the polls is stated in the state and national polling pages. You can see for yourself what I’m saying about weights. I think you’re also omitting trends, which are very important.
Well, you can all take solace in the fact that, despite the race being essentially a coin flip at the moment, the fallout from the debate is likely to overwhelm whatever polls from this morning had to say.
Solace and/or grim acceptance.