Because 2012 had the “skewed polls” meme, I think that made a lot of people and institutions skeptical about not believing polls this time around, and we see the result. But even then, I can’t imagine that the Democratic internal polling predicted exactly this in exactly this way.
So what happens to political polling in the future, especially how the electorate reacts to it?
I just compared the final RCP numbers with the results and it appears that the polls predicted Hillary’s share pretty well. They just massively under-predicted Trump and practically all the undecided moved to him. I think part of it is that many of the undecided were really Shy Trump voters as some had speculated and secondly that he attracted some genuinely new voters who were difficult to reach in states like Wisconsin and Michigan which were not heavily polled anyway.
I don’t this invalidates the idea of polling altogether but it does suggest that polling is much harder for truly unique candidates like Trump and when there is a relatively large number of undecided voters. I suspect that if Trump is the candidate next time around, polls will be more accurate since he will be much more of a known quantity.
The media certainly seems to have believed the polls. The DNC certainly seems to have believed the polls. Hillary certainly seems to have believed the polls.
The only people who didn’t believe the polls where the majority of the voters. The voters who simply don’t care what the polls say.
BREXIT passed. The poll numbers were wrong. Trump was elected POTUS. The poll numbers were wrong. The poll numbers have proven to be useless. What’s the point of even having political polling that appears more biased than factual?
That is not what happened.
Brexit was more like Trump’s victory in the primaries, where the Experts talked until they were blue in the face about how the distinctly possible thing was not even possible.
I was going to start a thread on this myself. It may be that the polls get it right next time and people start trusting them again, but until then they are going to really have to wonder. And if the next time is wrong, then I wonder if there will even be money to pay for media polls any more.
The pollsters messed up because polling is flawed for this purpose. Many people screen their calls and don’t answer, or they answer and lie (as I’ve done on several occasions). Sure, pollsters try to adjust for this, but it really is a “garbage in, garbage out” flawed process.
The pollsters can go to hell, because they are the fuel for presidential contests being covered as horse races.
Polls as recent as two days ago had Clinton ahead of Trump by 5 points in PA, but Trump won. Exit polling is a little better, but still depends on people telling the truth of who they voted for. I’ve never been approached at a polling place, but there is no way I’m telling some pollster schmuck who I voted for.
I also noted there was talk among some here that the press were showing a narrowing race because ‘the media have an interest in peddling a narrow horse-race’; while that may be true, you can’t simply discount without looking at the stats, and also, why is that a bad thing? A close-race should, ideally, boost turnout.
I think the electorate is so volatile lately that pollsters don’t know how to model likely voters.
My guess is that the campaigns had internal polling showing Michigan and Pennsylvania in play, though, given how they made a point of spending time there at the end.
Nate Silver was right, and Sam Wang was wrong.
I mean, literally you could not have come up with a more stunning proof of Silver’s model being right and Wang’s being hopelessly wrong, and their respective arguments - Silver’s, that there was great uncertainty because of the high number of undecided/third party votes, versus Wang’s, that everything was settled long ago and there was no uncertainly and Silver was crazy - were, respectively, entirely correct and stupidly wrong.
What this means for pollsters? Not a lot. The polls were very close and the results mostly within the MOE.
This needs and will be repeated many times. Silver quantified the uncertainty well. No one else did.
The New York Times had Trump at 33% chance yesterday.
This 538 map is a great example. It has my county at +25 blue, Clinton got less votes here than Obama did - she got less than 10% of the vote to Obama’s ~12%. So not only was it not +25 blue, it was actually a slight bit redder.
They weren’t just wrong, they were on the wrong planet.
Trump also got less than Romney, but the Libertarians had a gain of basically the Romney/Trump difference.
I questioned the predictions of that map in a couple of posts here, I wrote off the disparity to Republicans hating the nominee - not some new found Hillary love. Turns out my gut feeling was right all along, there was no way the county could have a +25 blue shift.
If it’s any consolation, I lost $100 on the election. I just couldn’t fathom Trump winning, I guess I put too much faith in polls.
Again, 538 did not deal in absolutes. They dealt in probabilities, unlike the folks saying Clinton would win for sure or close to for sure. They have been saying for some time Trump could win due to a polling error that failed to account for where undecided voters would go. And they were right.
It’s akin to two people going like this when Cleveland was up 3-2 in the World Series going back to Cleveland:
SAMUEL: This is it. The Indians will win. I’m more than 99 percent sure. I’ll eat a bug if they don’t.
NATHAN: That is not how baseball works. You can’t assume that because Cleveland is winning to the point that that has to continue. The Cubs are a really good team and winning two games in a row is not that unlikely. I mean, I think the Indians will probably win. The odds are in their favour. But, man, it’s not that unlikely the Cubs will beat them up.
SAMUEL: Impossible. You’re a hobbyist. You aren’t good at math. Cleveland will win, it’s better than 99 percent.
NATHAN: Well, I’ve got this model I’ve designed based on how baseball works, and, you know, I think Cleveland will win, but the Cubs have like a 30 percent chance. If you look at previous World Series where one team was up 3-2, you’ll see…
SAMUEL: Nope. No chance at all.
NATHAN: Well, I’m a Cleveland fan and I hope you’re right, but the Cubs have a chance. Don’t bet anything against them you really would not want to lose.
Then the Cubs won. Clearly, Nathan was right. He said Cleveland was likely to win, *but his position that there was great uncertainty was correct. *He understands baseball, and he understands probability. Samuel, well, doesn’t. I look at the map you posted and maybe 538’s numbers are wrong about your county, but look at the color on that map. Where is Trump’s upside? The very places he surprised everyone the most; the Rust Belt.
The future of polls lies with non-profit polling stations lol
Exactly how are pollsters doing polls nowadays? One thing is we have had a big decline in landline usage and a switch to cell phones. To what extent are cell phone numbers available to pollsters?
We don’t have the final national popular number, but right now RCP shows 0.2% popular ‘victory’ for Clinton vs. 3.2% final lead in the polls. That’s less than the error in 2012, where final RCP avg showed Obama up by 0.7% and he won by 3.9% points. It wasn’t as shocking then because it didn’t mis-predict the EC outcome. However in the few elections before that in ‘the RCP era’ that untweaked average was much closer to the final outcome.
So like many such things the commentariat will probably go overboard obsessing about poll error but it was pretty bad. And it was particularly bad in ‘rust belt older white voter’ states. ‘Shy Trump’ was effectively true, despite not having showed up net in the primaries. But as I mentioned prior to the election (though I’m not pretending I ‘predicted’ 
) it was plausible for Shy T to show up in the general after not doing so (much, net) in the primaries because Trump’s GE vote was from a 4 or more times larger and different group of voters than his primary voters, more of whom perhaps had reason to be reluctant to admit (even perhaps to themselves) they’d finally vote for him, or reluctant to participate in polls.
Over at the LA Times they are crowing about how their constant outlier tracking poll predicted the Trump wave.
Though with it looking like Trump lost the popular vote, they kind of missed the mark too.
Well the first thing to do is to figure out just why the polls seem to be pointing one way but the results were the opposite. I have no idea. Anyone?
Clinton won the popular vote by a very small number - that’s exactly what the polls said?
Polls are no predictor of elections. But they create a nice, albeit bogus, media narrative
The media believed their own push-polling.
Well, no one has ever accused journalists of being especially smart.