Logically, I agree. But suppose in late October there are say 5-6 incidents of mass shootings and/or bombings perpetrated by Muslims. I’m not so sure that his message of hate and intolerance won’t start selling more than it does.
Not so sure if that is true on election eve. Yes for earlier calls as states and nation polls (and certain individual states) tend to move together. But polls on election eve are not subject to movement. At that point it is like a bunch of 55% picks for a week of football games.
No, but they could be to common bias (e.g. if they’re undercounting a type of voter in one state, they’re probably undercounting in other states).
The way states ultimately vote is correlated. That is, knowing who one state chooses gives us information on how other states will likely choose. We cannot treat each state as an independent football game, where who wins one game gives us no information on will win another game.
Possibly, but different polling companies sharing the same mistake across different states doesn’t seem likely.
It’s still to a great extent the same several polling companies being averaged in each state, so the resulting average could be biased. In part I’m thinking in Likely Voter calculations, as a big narrative is competing turnouts of less-likely voting pools on each side (and of course, we’re not talking about 100% correlation here, just some amount of correlation).
That’s not really how this works, they are not coin flips.
Yes they’re correlated but if they were that darn connected then maybe he shouldn’t put odds on individual states. Hey it’s 51/49 that every swing state goes Obama.
Or, they may be a single coin flip.
Say that there’s a bunch of states, each with a 50% chance to go to either candidate. But, the result is determined by a single coin flip. If it’s heads, they all go to one candidate; if it’s tails, they all go to the other. So in this model a candidate has a 50% chance to win each state, and a 50% chance to win all of the states. But after the coin flip, one candidate will have won all of them. That is how correlation works.
Of course, in the real world, we won’t have 100% correlation between states. But it shows how you can’t treat each state as an independent coin flip.
I was watching the NowCast, which I like because I (wrongly) see it as a leading indicator of the current zeitgeist.
Speaking of the NowCast, she’s hit 70% :).
Thank you @538 Forecast Bot!
Based on a good NH poll, she is up to the following:
Polls-plus ↑ 1.1% (61.2%-38.8%)
Polls-only ↑ 0.9% (63.3%-36.7%)
Now-cast ↑ 2.4% (72.4%-27.6%)
Latest:
Polls-plus ↑ 1.6% (62.4%-37.6%)
Polls-only ↑ 1.6% (65.3%-34.7%)
Now-cast ↑ 0.6% (72.5%-27.5%)
From polls in NH and MI (both HRC +7).
She’s closing on Ohio in the Now-Cast, too. Good news.
To play devil’s advocate and usual caveat about the strengths and weaknesses of the tracker polls made, what do y’all make of the LA Time/USC Tracker moving more Trumpward? :eek:
It does seem to be the outlier and it represent merely a good HRC day 8d ago dropping off but it does fly in the face of what we and most expects are expecting to see come out today and over the next few more days.
Well, PPP has it Clinton+4 when they had it Clinton+5 in late August. Quinnipiac actually had Clinton down by 4 compared to their last poll. Reuters is the only national poll I see where she’s up from the last time they polled. So the LA Times poll isn’t really the outlier.
The state polls are looking better for Clinton but still not seeing too much national poll movement. But it’s still early. We usually look for movement like the first few days after a major event like a debate but sometimes it actually takes a little longer for a bounce to show up.
Latest update from a good Florida poll:
Polls-plus ↑ 1.0% (63.4%-36.5%)
Polls-only ↑ 1.0% (66.3%-33.6%)
Now-cast ↑ 1.3% (73.8%-26.1%)
We are looking for pre and post debate here Addie. Looking at Quinnipiac’s before debate numbers in their last poll does not inform. Comparing to a month ago is of limited value. And you are missing Rasmussen’s move of 4 towards Clinton.
OTOH I am not yet seeing the move you mention in Reuters’ tracker, even as they documented that most viewers felt she “won” the debate. Yes, they agree with your delayed movement hypothesis:
Still, if the national polls do not end up showing an overall movement Clintonward over the next few days then she will not have “won” the debate. Unless one takes a position that “likely” voters answers have not changed much but whether or not they actually vote has. And that becomes a nonfalsifiable position.
The LA Times has been so Trump-leaning as compared to any other source, throuhgout the election, than I find it baffling. 538 gives it a Trump bias of four points or so, and yet *even with that adjustment *it is always 3-6 points more Trumpian than any other pollster.
How much does a debate bounce matter though? Romney got one, but it faded even before his “47%” comments were revealed. Plus with both Clinton and Trump there’s always another negative news story just waiting to come out. Seems like the winner of the election will be whoever is NOT the focus of media scrutiny in the last week of November.
They are using the same group of voters and just polling them over and over. It’s a different approach from what any other pollster is doing so it’s yielding wildly different results. So they’ll either be geniuses and everyone will copy them next time around or they’ll be totally wrong and poll the normal way next cycle.
Seems to me that the LA Times pollster is trying to figure out a way around the recent obstacles to polling that have come up in recent years. If you have like 5000 people who have agreed to let you call them again and again then that certainly makes polling more convenient, but it could just be that Trump voters are more willing to be a part of their unique poll.