Polls Missing the Call

[SGOS]

Missed the presidential prediction.  Missed the Senate, too.  My question is, considering their past accuracy, why did they get it wrong?  Polls are just numbers.  A honest poll would be nothing more than a tabulation of people polled.

Did they poll the wrong people?  That's my first guess.  Second guess was that the sample was not big enough.  Out of millions of voters, polls usually sample about one thousand, which could easily not be a representative sample.  All my life I've questioned sizes of poll samples when used to make predictions of large populations.

Did an unexpected demographic decide to behave in a different manner and either not show up, or show up, to vote in unexpected numbers?  I wonder about this. 

Of course the polls showed a close race just before the election.  They might still be close, just not close enough to have called it correctly. 

[Sal1981]

That's basically the failure of extrapolation of definitive numbers to a whole set.

Unless it's a true distribution, it will be skewed and become an incorrect prediction in polls, say.

Hell, this election reminds me of a statistician going to an NRA rally and asking how many guns per capita they own (or any other trite collection of like-minded individuals).

[AllPurposeAtheist]

I didn't trust polling from day one.  I kept telling people polling is bullshit, but nobody believed me. They kept telling me Nate and his genius followers couldn't be wrong, but I knew that the polling was flawed from day one.

[Cavebear]

As more voters live online, polling gets less certain.

[SGOS]

That's basically the failure of extrapolation of definitive numbers to a whole set.

Unless it's a true distribution, it will be skewed and become an incorrect prediction in polls, say.

Hell, this election reminds me of a statistician going to an NRA rally and asking how many guns per capita they own (or any other trite collection of like-minded individuals).

Pollsters are supposed to know about unintentional skewing.  I guess once in a while, the usual methods aren't going to work, because eventually you're bound to extrapolate from an unrepresentative sample, even when you follow the rules.  The results of this election also got the reporters wondering about the Comey influence.  Before, the election, the general opinion seemed to be, people have already decided, it won't make a difference.  Personally, I think it had to hurt.  Yes a lot of people were decided, but not everyone decided well enough to withstand a jolt of that magnitude, and that subgroup may have been unrepresented in the polls taken before Comey.

[SGOS]

As more voters live online, polling gets less certain.

I've heard there is a problem with polling via the net, but I don't know specifically how the net skews results.  Sometimes the explanations are surprising when they are articulated.

[SGOS]

I didn't trust polling from day one.  I kept telling people polling is bullshit, but nobody believed me. They kept telling me Nate and his genius followers couldn't be wrong, but I knew that the polling was flawed from day one.

While errors in polls can always be accidently introduced, I have been more surprised by how right they are most of the time.  Actually, I'm more surprised by that than surprised by last nights snafu, which I supposed was destined to happen some time or another.  I think bullshit is a bit too strong of a criticism, because that implies the error was intentional, common in propaganda polls, but not usually by Reuters and the like.  I never heard of this Nate Silver guy until the other night.  I'm not sure how he established himself, but his name is certainly not as recognizable as Pew or even USA Today.

[Baruch]

I didn't trust polling from day one.  I kept telling people polling is bullshit, but nobody believed me. They kept telling me Nate and his genius followers couldn't be wrong, but I knew that the polling was flawed from day one.

I believed you ;-)

[Cavebear]

While errors in polls can always be accidently introduced, I have been more surprised by how right they are most of the time.  Actually, I'm more surprised by that than surprised by last nights snafu, which I supposed was destined to happen some time or another.  I think bullshit is a bit too strong of a criticism, because that implies the error was intentional, common in propaganda polls, but not usually by Reuters and the like.  I never heard of this Nate Silver guy until the other night.  I'm not sure how he established himself, but his name is certainly not as recognizable as Pew or even USA Today.

I will never trust polls again. 

[SGOS]

I will never trust polls again. 

I think we just took them too far.  I was infatuated with their track record, and gave them more credit than they deserved.  The commentators I watched were like that last night, too.  It's true, you can't completely trust them, but you can trust them some.  To give credit to the better pollsters, these guys are making the best predictions they can in what are almost always very close popular votes.  They attempt to parse out something meaningful from very small disparities.  Then factor in the effect of the electoral college, and how close they manage to come still seems impressive to me.

[drunkenshoe]

I believed you ;-)

He is serious about that. APA never believes in any kind of polls, lol.

[Blackleaf]

Undoomed had an interesting perspective in this. He's a Trump supporter, but I still think his explanation for the downfall of the democratic party is valid. One of the things he thinks led to Trump winning was this: While the Republican party traditionally represented crazy Christian nutjobs, the Democratic party ran on identity politics. The Left Wing media portrayed their opposition as racists, sexists, homophobes, idiots, etc. As a result, the Democrats began to be perceived as nutjobs as well. It may also be that many Trump supporters were simply too ashamed to admit their position when polled, which again is related to identity politics and a fear of being labelled a racist, sexist, homophobe.

https://www.youtube.com/watch?v=JBDnp-R75P8

[Unbeliever]

My question is, considering their past accuracy, why did they get it wrong?  Polls are just numbers.  A honest poll would be nothing more than a tabulation of people polled.

The mainstream media must have close elections, or their viewers will be completely uninterested in watching the coverage, so the advertisers get no bump in sales. The Republicans must have a close election, because only then can a bit of illicit jiggling win the thing.

Therefore, close elections, polls be damned.

[SGOS]

NPR just interviewed a pollster for Edison Research.  He said  he doesn't know any pollsters that know what went wrong, but everyone is trying to figure it out.  His bet is the area of turnout.  People answer the questions, perhaps sincerely in pre election polls, but this doesn't guarantee that they actually turnout to vote.  It's a simple enough explanation.

So the next question would logically be why did the demographics of turnout change from the intent to the actual turn out?  I'm going to paraphrase here a good bit, because it was a radio broadcast.  The Edison Research guy said there was a striking difference between Trump vs  Clinton supporters over the state of the economy.  Those that thought it was OK to great were going to vote for Clinton.  Those who thought it was a disaster were going to support Trump.  The next part of his explanation sounds key to the whole thing; Those thinking the economy was a disaster numbered in the mid 60% range.  My added interpretation (and I think the pollster's) is this constitutes a group in dire straights that is highly motivated for change, and highly motivated to actually vote.  Trump simply resonated with the most disenfranchised wanting more change.

This makes sense to me.  Clinton being OK and all sounded kind of stale to me, a product of the roaring 80s, telling people everything is good.  Even if people think it's good, what motivates them to vote?  It has to be what they perceive as the status quo, and people just don't get excited about the status quo.

This was Obama's strength.  With the end of the Bush era, hope for change became overwhelming for voters.  Things were in dire straits with wounded hopes in Iraq, an international banking crisis.  There was lots to be done.  Lots to be restored.  Lots of opportunity to change direction.  Stated openly or not, that is what people perceived in Obama, At least enough people to vote him into the White House.  Even then, Hillary was not imparting that message as articulately, I thought.

[Poison Tree]

The nature of the electoral college is helping to amplify how much the polls appear to be off. As 538 points out, if 1 out of every 100 Trump voters had voted Clinton instead then Clinton would have won 307 electoral votes, the polls would have nailed the popular vote and correctly called every state except North Carolina. While higher then expected turnout among white voters without a college degree looks like the early front runner for what the polls missed, it could simply be that 1% of Trump voters switched away from Clinton too late for the polls to notice.