[RD] US 2016 election: Poll watching thread

[metalhead]

If by deliberately you mean it is part of the usual methodology, yes. Clinton was never up by nine, much less stable at that level. Five is quite sufficient.

J

You're only looking at one source. That's not a smart thing to do. Bloomberg currently has their polling average at 8.3, because they exclude the nonsense polls like Rasmussen from their average. They focus on high quality, in-person phone pollsters who publish at least some of their data as any good pollster ought to do.

More importantly, the national polls are understating the margins among Latino voters, and by a significant margin. According to Bloomberg, the current average margin among Hispanic voters is Hillary +29. However, Latino Decisions' polling puts the national margin among Latino voters at Clinton +59 in their most recent polling, with her getting 74% of the vote.. They also carried out polls focused on four swing states. There they got similar results - in 3 of the states, Trump was under 20%. The margins were +52 in Arizona, +40 in Florida, +55 in Nevada, and +50 in North Carolina. The swing state polls show slightly lower margins, but you can peg it at +50, and peg Hillary's share of the Latino vote nationwide at around 75%.

So, don't just trust a polling average to be 100% accurate. The difference between Clinton getting 65% of the Latino vote and 75% of the Latino vote, is a difference of over a full percentage point in the nationwide popular vote margin. The polling is off by at least that much.

 

[El_Machinae]

I notice Rasmussen is an outlier. Do we know why? Like, actually know?

 

[metalhead]

Nope, they put all of their details behind a paywall.

 

[El_Machinae]

But it must have been looked at. All my science publications, cept one, are behind a paywall too

 

[metalhead]

I'm not aware of anyone actually looking into it. I think they've been written off in the polling nerd community as a rank partisan polling outfit, and nobody even gives them enough credence to check their methodology.

 

[El_Machinae]

Oh, it looks like a methodology artifact. That's fair enough. It's why we look at averages. It'll happen

 

[AdrienIer]

And what about the IBD/TIPP tracking polls ? They're also consistently skewed towards Trump

 

[metalhead]

Well, the suspicion is that they don't really have a set methodology, but rather they cook it each time they publish a poll to get a desired outcome. This is quite easy to do if you understand what the electorate is going to be. All pollsters do this to some extent, but while pollsters typically "cook" their results in such a way to match their own biases about what the electorate will look like and which groups will turn out and stay home, Rasmussen does it for editorial purposes to make Republicans look good.

It's all in how you weight your results; if you take your polling data and weight it so your projected electorate is 80% white and 70% without a college degree - a 2000-ish electorate - you're going to get very different and far more pro-Trump results than if you project your electorate to be 70% white and 60% without a college degree, which is in line with current demographic trends. You can use the former weighting if you want, and amazingly some pollsters were up until a month or so ago, but you ought to have a good reason for it.

At this point, there is no good reason for it, so if that is how a pollster is turning their polling data into a Clinton +1 result, it is highly suspect. But many pollsters don't publish their weighting or projections for the electorate that is going to show up, and its possible Rasmussen doesn't even do so behind their paywall.

 

[El_Machinae]

538 gives IBD an A- on their methodology. It's better than a few names you would assume to be credible

 

[Sommerswerd]

538 gives IBD an A- on their methodology. It's better than a few names you would assume to be credible

LA Times is so transparent about their methodology that the flaw in their result is much better known and much easier to account for. I haven't read anything about IBD, but if 538 gives them an A- then I would assume that at least their methodology is sound and like LA Times, there is some nuanced flaw that is making them an outlier. A- is not perfect afterall so there is some flaw in there, but with such a high grade I would expect the flaw to be something isolated/specific. So if you can figure out what the nuance is, you can figure out how to use their poll to get useful info.

metalhead is correct however that Trashmussen is just garbage water... to be rinsed out of the can, poured into the sewer drain, and otherwise disregarded.

 

[metalhead]

IBD brags on their polling page, right in the lede, about how they are the most accurate pollster. That right there gives me pause about their credibility, 538 grades notwithstanding.

 

[Leoreth]

In case a state election is going to be challenged, what happens when we run into a Bush v Gore situation and the Supreme Court needs to decide? It's currently dead-lockable. Do these cases go to the SC directly or do Federal Courts decide first, so that there is a decision to fall back on?

Even though I don't think the election will be close enough for a close state to matter, I'm still worried about a state called for Clinton being contested by Trump without there being an institution to confirm the call.

 

[tetley]

Right now I'm seeing a little pro-Trump drift. The interestingly-timed Obamacare hike went out on the news, plus it looks like Hillary and Obama both are setting their sights on congressional races, while Trump is going after presidential battleground states. 7.7 million votes already cast.

 

[metalhead]

In case a state election is going to be challenged, what happens when we run into a Bush v Gore situation and the Supreme Court needs to decide? It's currently dead-lockable. Do these cases go to the SC directly or do Federal Courts decide first, so that there is a decision to fall back on?

Even though I don't think the election will be close enough for a close state to matter, I'm still worried about a state called for Clinton being contested by Trump without there being an institution to confirm the call.

A 4-4 decision leaves the lower court decision in place. Interestingly, in 2000 that would have meant the Florida Supreme Court decision to keep the recount going would have remained in place, and the outcome may have been different.

 

[Camikaze]

A saving grace might be that you wouldn't think Roberts or Kennedy would really be that keen on Trump. This election probably splits the court along lines which aren't purely partisan.

 

[metalhead]

I have to imagine that all of the justices save perhaps Thomas are aghast at the idea of Trump as president.

 

[Gori the Grey]

In its own way, that would be fun to see. It'd be fun to see every element of the "system" "rigged against him": those corrupt Clintonians, Congresspeople from his own party, * , the fourth estate--all the way up to the Supreme Court.


*

Spoiler :

the electorate, of course

 

[tetley]

Early voting indicators seem to be yielding a mixed bag for Clinton. Which is all she needs--for Trump to win, he has to win them all, and then some.

 

[tetley]

Whoa. 538 just gave Democrats a big bump in the Senate. They're favored 66% to take control. Hillary's advantage had steadily eroded from 87% to 84%, and that just bumped back up to 86%. Which is a bigger deal if you think it in reverse: Trump's chances slowly went up from 13% to 16%, then bounced back to 14%.

 

[AdrienIer]

It's been looking good for the democrats in the senate for the past month. CO WI and IL are almost certain, and they need 3 out of : NV, MO, IN and PA where they're narrowly winning, NH where it's mostly a tie and in NC where they're narrowly losing.