[RD] US 2016 election: Poll watching thread

[dutchfire]

I get the feeling many of you care a lot more about politics than I. (I care a lot about elections, which isn't quite the same as politics.)

https://twitter.com/ForecasterEnten/status/753112599345135616

This thread is not for policy positions, gaffes and quotes. I don't want to hear about e-mail servers, arrested campaign managers or who said what on the sunday morning talk shows. This thread is for talking about the latest polls, for watching the betting markets and the for furiously pressing F5 on the 528 election forecast page.

The direct occasion for this thread, of course, is the Clinton's drop in the 528 models over the last two days, dropping several percentage points in all flavours of the model. Looks like we might be getting a close election after all.

Also, I was wondering if the 2012 (or 2008) forecasts during the election season are still available somewhere. It'd be interested in seeing how many fluctuations there were in the entire season.

 

[GoodEnoughForMe]

2012 had a massive fluctuation right after the first debate (which many said Obama was lackluster at), favoring Romney, but then Obama steadily built back the lead until it was 99%. I can remember that much at least.

 

[dutchfire]

There's a New Hampshire poll out today that has Trump +9. It is literally in the only NH poll in the 538 database that shows Trump ahead. The polling firm, Praecones Analytica, has no website and no internet results except for this poll.

Fake?

 

[Gucumatz]

538*

538 also has pollster ratings here:
http://projects.fivethirtyeight.com/pollster-ratings/

 

[Timsup2nothin]

There's a New Hampshire poll out today that has Trump +9. It is literally in the only NH poll in the 538 database that shows Trump ahead. The polling firm, Praecones Analytica, has no website and no internet results except for this poll.

Fake?

The "paper" that sponsored the poll (NHJournal) is actually a "news source site," meaning that it is a one person operation only slightly removed from a Facebook account. In this case the one person (John DiStaso) is a hack from the Karl Rove stable. Judge accordingly.

And unsurprisingly the "pollster" is not rated or listed by 538.

 

[AdrienIer]

Huh. Rassmussen apparently published a poll with Clinton at +1, even as other poll stations give Trump at +3 or more. Strange.

 

[Bootstoots]

Nate Silver has published an article today explaining why Trump is now near-even with Clinton in his model. The short answer is that their model is fairly aggressive at capturing recent changes and trends in polling data, and a downward trend for Clinton over the preceding two weeks was then added to by Trump's convention bounce, the magnitude of which seems to be around 3-4 points. The RCP average now has Trump+0.9, down slightly from his all-time high of +1.1 yesterday.

It seems likely to me that the picture after both of the convention bounces fade out will be a slight edge for Clinton, of something like 2-4 points, but with substantial oscillations in both directions as this year's unusually intense news cycle continues. The official Bootstoots Odds (or BO for short) of Trump winning the election, derived one part from real data and one part from my posterior, are raised to 30% from 20%.

 

[Kaitzilla]

Based on the vitrol being thrown Trump's way, this election wouldn't surprise me if there were lot of people who told their friends and poll people "no way am I voting for Trump", and then they secretly vote for Trump.
Sort of like a warped Bradley Effect.
https://en.wikipedia.org/wiki/Bradley_effect

0% of the black vote in a few states?
Trump's appeal is aimed mostly at blue collar workers, so it seems improbable that he won't get any African-American blue collar votes.
http://www.nbcnews.com/politics/201...rt-black-voters-ohio-pennsylvania-nbc-n609026

 

[Timsup2nothin]

0% of the black vote in a few states?
Trump's appeal is aimed mostly at blue collar workers, so it seems improbable that he won't get any African-American blue collar votes.

Trump's appeal to blue collar workers is a "back to the good old days" appeal. For whites that means "back when there was less competition for our jobs, no immigrants, no blacks, just us." For blacks that "appeal" means back to the days when an employer could openly toss your application in the trash and say "beat it, _____." Guess what, it isn't very appealing.

 

[Sommerswerd]

Huh. Rassmussen apparently published a poll with Clinton at +1, even as other poll stations give Trump at +3 or more. Strange.

It's not strange once you accept what Rasmussen is. They are nothing but agenda driven propaganda. That is why their results often have no relation to any other polls. If you think of Rasmussen as a legit, but conservative leaning poll, like say FOX News poll, then you will be confused when they don't always lean further right ahead of the trends. When you realize that they just make it up from whole cloth, (using "LV" as a smokescreen for their cherrypicking fabrications), you understand why their results never make any sense.

Just ignore Rasmussen entirely, it makes the analysis so much more clear and reliable. They are pure garbage, to be dismissed out of hand.

That being said. The real, credible polls are showing Trump ahead and I want to attribute this to post convention bump, but I am starting to get nervous again. I have been saying this entire cycle that Trump was the only threat in the Republican field ...

 

[dutchfire]

Today's Suffolk University poll in Pennsylvania is very good for the Democrats (although it is a relatively small sample with a B+ pollster rating from 538).

 

[innonimatu]

Hasn't this Nate Silver lost all credibility yet with his predictions during the last year?

 

[Kozmos]

http://predictwise.com/

Predictwise is 68% for D and 32% for R. I'm curious how accurate this will be.

 

[Camikaze]

Based on the vitrol being thrown Trump's way, this election wouldn't surprise me if there were lot of people who told their friends and poll people "no way am I voting for Trump", and then they secretly vote for Trump.
Sort of like a warped Bradley Effect.
https://en.wikipedia.org/wiki/Bradley_effect

The 538 podcast has dealt with this a couple of times, with the conclusion being that there's no evidence at all that that sort of effect is likely. Trump has thus far outperformed his polls more than he has underperformed them. An equally compelling theory would be that Trump posters are notably vocal, so tend to answer polls at a higher rate than non-Trump supporters.

Hasn't this Nate Silver lost all credibility yet with his predictions during the last year?

If credibility is simply a product of public opinion rather than accurate criticism, then probably yes. The headline has been that he said Trump had a 2% chance of winning the Republican primary at some stage last year (a statement Silver later said was impressionistic, not intended as a mathematical prediction, as the current models are). But then that initial prediction rapidly changed as the actual data did, and he ended up calling almost all the primaries correctly.

In short, there's no actual reason to assume that Silver's models are off, simply because he, like everyone else, didn't think Trump had much of a chance before there was much data to work with.

 

[Arakhor]

The Bradley Effect is what's known as the 'shy Tory' phenomenon in the UK, though presumably with less racist overtones.

 

[rah]

538*

538 also has pollster ratings here:
http://projects.fivethirtyeight.com/pollster-ratings/

Thanks for the link. I'm usually involved in methodology or quality at the respondent level so I sometimes don't see the bigger picture. It's nice to see our company rated at A- with an almost non-existent bias of 0.1 since I know there is no intentional bias.

 

[Sommerswerd]

538*

538 also has pollster ratings here:
http://projects.fivethirtyeight.com/pollster-ratings/

Thanks for this, very informative... I didn't even notice it until Rah quoted it

Thanks for the link. I'm usually involved in methodology or quality at the respondent level so I sometimes don't see the bigger picture. It's nice to see our company rated at A- with an almost non-existent bias of 0.1 since I know there is no intentional bias.

Yeah I would consider anything with at least an A- rating to be legit a B+ so-so, and anything under a B+ to be garbage. I guess you could argue that there's different qualities of garbage, but certainly stuff in the C's and lower are not worth citing as a source for anything. And on that note...

I noticed that out of the hundred (or more) polls listed good ol' Rasmussen was 4th highest in terms of Republican bias, snuggly between two F-graded, banned (by 538) polls, graded in the C's . Another thing that I noticed was that of all the crap polls (by grade), most of them had very few polls analyzed, most had less than 10, and almost all were less than 30.

There were only a few crapsters with a lot of polls analyzed, and far and away the one with the most was Rasmussen. In other words... I'm even more confident that Rasmussen is garbage based on this because of all the craptastic pollsters, they have the largest sample size. They really don't deserve the status they enjoy on RCP... as I have said many times.

EDIT: I just noticed that YouGov (graded B) has a few more polls analyzed than Rasmussen, but I don't think that affects my conclusion that Rasmussen is the by far the worst poll that has top tier media status.

 

[AdrienIer]

Latest Reuter/Ipsos poll doesn't make sense. Clinton vs Trump is 40/35 but Clinton vs Trump vs others is 37/37/5/1. How can Trump have more votes if there are more choices ?
But the Missouri poll (+1 Clinton) is very interesting

 

[Zkribbler]

2012 had a massive fluctuation right after the first debate (which many said Obama was lackluster at), favoring Romney, but then Obama steadily built back the lead until it was 99%. I can remember that much at least.

I remember a few more details. Yes, Obama was lousy in the first debate. Then, in the VP debate, Biden immediately jumped on every wrong statement Ryan made. And there were a lot of them. In the last two Presidential debates, Obama adopted this tactic, and eviscerated Romney.

The electorate is looking for a good guy and a bad guy. When one side is getting caught in falsehood after falsehood, the bad guy quickly becomes apparent.

 

[onejayhawk]

Just ignore Rasmussen entirely, it makes the analysis so much more clear and reliable. They are pure garbage, to be dismissed out of hand.

This is definitely wrong. Where Rasmussen deserves to fall is yet to be established, but it is not the round file.

J