What happened to the Polling?

[Citizen (The) Doctor]

Seriously? What happened to the polls we've seen this cycle? Warner +7? What?

[Ljube]

Orman +12

[○∙◄☻¥tπ[╪AV┼cVê└]

It's easier to answer a poll than to bother to vote when the Democratic party couldn't argue that 1 + 1 = 2 to a math professor.

[Oakvale]

[IceSpear]

Imagine if he picked 2014 to debut instead of 2012. He'd be seen as better than Nate Silver.

[Nichlemn]

Imagine if he picked 2014 to debut instead of 2012. He'd be seen as better than Nate Silver.

There probably wouldn't have been much demand for him, because Republicans looked to have a decent year according to the polls anyway.

[Bandit3 the Worker]

The real question should be "What happened to the election?"

The #1 story on the evening news on every network tomorrow should be "The election: what went wrong?" In every other country in the world, this would be considered a sign that the election was stolen.

[Lief 🗽]

Another problem I think is all this insistence on polling third party/libertarian candidates whatever. In all these races the numbers of libertarians were vastly inflated.

[greenforest32]

Maybe difficulty with turnout? How was turnout compared to previous mid-terms?

[Nichlemn]

The real question should be "What happened to the election?"

The #1 story on the evening news on every network tomorrow should be "The election: what went wrong?" In every other country in the world, this would be considered a sign that the election was stolen.

LOOOOOOL. You think the election was stolen? Oh sorry, just saying that it would be considering stolen anywhere else. Just asking questions?

And you're totally wrong anyway, the US has pretty accurate polling by a worldwide standpoint. Take a look at say, recent Canadian provincial elections. Huge polling errors. No-one says they were stolen.

[Bandit3 the Worker]

Maybe difficulty with turnout? How was turnout compared to previous mid-terms?

Terrible (despite anecdotal statements otherwise).

[Bandit3 the Worker]

The real question should be "What happened to the election?"

The #1 story on the evening news on every network tomorrow should be "The election: what went wrong?" In every other country in the world, this would be considered a sign that the election was stolen.

LOOOOOOL. You think the election was stolen? Oh sorry, just saying that it would be considering stolen anywhere else. Just asking questions?

Some of the polls were off by as much as 20 points - well outside the margin of error. In fact, almost every major poll was off by more than the margin of error in every major election.

All these polls independent of each other showed roughly the same thing, which wasn't nearly as Republican as the actual results.

So the outlier is the election.

[Mehmentum]

Not all the polls were off.  Most of the Alaska polls showing Begich down by about 3 were right in the end.

[Bandit3 the Worker]

Not all the polls were off.  Most of the Alaska polls showing Begich down by about 3 were right in the end.

Probably only because Alaska is geographically so far from the rest of the country.

[rob in cal]

I think the Roberts Orman polling was so far off in part because support for Indys/3rd party candidate often fades on election day, and maybe Orman lost out because he didn't have any official party mechanism out there helping to turnout his voters.

[SPC]

A bit ironic that Alaska was the most accurate state for established pollsters.

[Mr. Reactionary]

In every other country in the world, this would be considered a sign that the election was stolen.

Good think this isn't every other country in the world then, because clearly this election wasn't stolen.

[jd1433]

I think you have to acknowledge the possibility that a lot of polling may not have been that far off when the poll was actually taken, but that the well documented "late break" can be a bigger factor in elections than many people are willing to admit. Even in 1994 when polling was probably it's easiest (the statistical math was developed, populations pretty sedentary, and everybody possessing a land line) and there was still a massive very last minute break in the GOPs favor.


That said the fact that pollsters refused to believe that VA was tightening enough to warrant more polling deserve to be criticized. I mean when "rumors of race tightening" are so prevalent that they actually make it onto RCP's site notes so they can cover their a$$ for the lack of polling just in case it's true... you know you pollsters missed their chance for fame on that one.

[Bandit3 the Worker]

The pollsters will take the wrong lesson from this. Just you watch. I bet they "recalibrate" their methods again, instead of demanding to know why the election was so far off.

[Mr. Morden]

From Nate Silver:

http://fivethirtyeight.com/features/the-polls-were-skewed-toward-democrats/

[The Vorlon]

Seriously? What happened to the polls we've seen this cycle? Warner +7? What?

a SERIOUS answer...

Firstly, it has gotten really, really hard to do good polling.  Less than 10% of people will even agree to do a survey anymore, and this combined with cell phones, caller id, call forwarding, etc had dramatically raised both the difficulty and cost of doing a "good job"

As a result,  there are a stunning number of "junk" polls out there where methodologically they are taking a LOT of shortcuts in order to get the costs of polling down.

In many ways, the media drives this because everybody and their dog wants to have a poll so they have "news" that their talking heads can then talk about, and they want to buy these polls cheap to keep costs down.

The media also likes "exciting" polls where there are big swings from poll to poll.

I remember maybe 15 years ago I was hired to do a series of polls for a local newspaper regarding the mayors race.

The incumbent mayor was fairly hard left candidate who had about 40-42% of the vote, and the remaining 60% or so was distributed maybe 33% to 25% between two other right of center candidates.

Over a period of about 8 weeks my polls gradually showed the 60% on the right consolidating behind one candidate, and my final poll correctly showed about a 5% win for the better of the two center right alternatives.  From a polling perspective I had done a great job.

By contrast, a pollster for the other newspaper in town showed huge swings where on Tuesday X was ahead by 5% but behind by 7% on Saturday.. and the final poll was not even close to correct...

I didn't get the contract to poll the next mayors race because, and I quote "your polls were boring" - cheap and exciting beats quality with the media.

Finally, there is substantial evidence that polls are starting to, as 538 puts it "herd" - where the weaker pollsters with methodological deficiencies adjust there assumptions so their results more or less match the other polls.  As a result, the herd can often be wrong, as they are no longer polling, they are just copying each others homework....

Lastly, directly or indirectly, many of the bad polls use some form of a party id weight that represents little more than a guess.  In this election party ID was 36/36 or so, but I suspect (directly or indirectly) a lot were keyed to the 2012 D+7 electorate.

~~~~~~~~~~~

Hats off the Ann Selzer in Iowa - Her last poll had Ernst at +7, and she was all alone...... and right.

[Bandit3 the Worker]

Oh, it's more than +4.

[Recalcuate]

The real question should be "What happened to the election?"

The #1 story on the evening news on every network tomorrow should be "The election: what went wrong?" In every other country in the world, this would be considered a sign that the election was stolen.

Stolen?

R+7 or even R+8 was reasonable given the polling. R+9 wasn't out of the realm of possibility. North Carolina slightly favored Hagan. The Roberts/Orman race was a flip.

The governor's races had like six races that were basic coin flips.

There wasn't enough reliable polling in the House to really figure out much of anything.

The races that were completely blown were in Maryland and Virginia and to a lesser extent Kansas. Aside of that, it's just error on the fringes of a predetermined result.

A few point bias can make quite a difference, obviously. But stolen? Very unlikely.

[IceSpear]

They were Junk Polls!

[Likely Voter]

maybe there is a bit of the polling companies fighting the last war. Following 2010 GOP wave the 2012 polls were too GOP biased, then when those proved wrong the 2014 were too Dem biased.