Let's start the dissection - polls and their seeming inaccuracy

[Tigers!]

British General election, Brexit, Trump, no agreement with FARC in Columbia.
It seems at present that the pollsters can not take atrick.
Why is that?
I can quickly think of 3 reasons
1. People being polled are telling the;pollsters one way but voting another. Why is that? I suspect that is because the polled already perceive what the pollster wishes and to avoid being seen as bogan/red-neck/reactionary/conservative - choose the epithet you most wish. The TV series' Ýes Minister' has an excellent 2 minutes of how pollsters could distort results. Must see if i can find it. https://www.youtube.com/watch?v=G0ZZJXw4MTA
2. The pollsters do not believe or accept what the results are saying so they 'modify them' to be in more accord with what they think the results should say.
3. The pollsters are not getting broad enough samples.

All it takes is 2-3% of polled to not be truthful and the margin of error explodes in their faces.

What say you?

 

[Jayjay]

Polls are always normalized based on results and demographics from previous elections and hence trail 2-4 years behind the times.

Also, when polls predict an overwhelming victory for one side as it did with Hillary and "Bremain", its supporters (or those on the fence) get complacent and don't show up to vote.

 

[zorq]

Option 4.... The vote is rigged. Electronic voting machine software is stealing votes.

I'm not advocating this one, but it would also explain the polling errors.

 

[Tom Sawyer]

Option 4.... The vote is rigged. Electronic voting machine software is stealing votes.

I'm not advocating this one, but it would also explain the polling errors.

If there were isolated discrepancies, yes. They didn't create a fake country wide movement, though.

 

[credoconsolans]

Option 4.... The vote is rigged. Electronic voting machine software is stealing votes.

I'm not advocating this one, but it would also explain the polling errors.

If there were isolated discrepancies, yes. They didn't create a fake country wide movement, though.

If Hillary had won, you certainly would have heard Trump and his deplorables baying about the 'rigged' election. But since he won, the election was as clean as Ivory soap.

 

[Tigers!]

Option 4.... The vote is rigged. Electronic voting machine software is stealing votes.

I'm not advocating this one, but it would also explain the polling errors.

Of course you are advocating it, otherwise why mention it?

 

[Keith&Co.]

Option 4.... The vote is rigged. Electronic voting machine software is stealing votes.

I'm not advocating this one, but it would also explain the polling errors.

Of course you are advocating it, otherwise why mention it?

Well, that's just silly. Imagining ways something COULD have happened is not the same as determining this is what DID happen.

 

[Jimmy Higgins]

Option 4.... The vote is rigged. Electronic voting machine software is stealing votes.

I'm not advocating this one, but it would also explain the polling errors.

Exit polls weren't too bad. Though there did seem to be some discrepancies. Though, the exit polling was why I thought VA was easy. I am still confused by PA.

 

[Tom Sawyer]

If there were isolated discrepancies, yes. They didn't create a fake country wide movement, though.

If Hillary had won, you certainly would have heard Trump and his deplorables baying about the 'rigged' election. But since he won, the election was as clean as Ivory soap.

Ya, they would have said it, but they would have been wrong. When the call is that the guy who caught Brady's touchdown pass against the Giants stepped out of bounds, everybody in New England can see clearly how it was a fair catch overturned by blind refs and everyone in New York will praise the refs on making what was clearly an excellent ruling on a close play.

The fact that they would have been screaming bloody murder if things had broken a couple of percentage points the other way doesn't mean that it's therefore ok to scream bloody murder. This election wasn't rigged, it's just that Trump's models of what was going on in the country were better than the models most of the pollsters used.

 

[Jolly_Penguin]

Maybe it is just that it was not politically correct to support Trump and people were too hesitant to admit that they did, until doing so anonymously in the voting booth.

 

[skepticalbip]

All polls didn't turn out to be such nonsense. The poll that was scoffed at by most media and Democrats was almost dead on for this election and the previous election - The LA Times daily tracking poll.

Polls are generally weighted to "correct" for what the pollster believes may skew the data. These "corrections" are their personal bias even though they make valiant hand waving arguments to justify why the weighting must be there. Apparently in this election the general pollster bias turned out to be about five to ten percent in favor of Clinton.

 

[TV and credit cards]

There may be some polled who did not want to be associated with the picture painted by the media as one of the "uneducated" and/or allying themselves with such an ill-mannered individual. But their desire for change, desprately seeking any change won them over.
Perhaps there is a demographic of people who do not answer polls that are closely aligned with the fed up majority seeking change.
Aren't the majority of polls conducted by cold calling folks? There is a demographic of those who do not answer unrecognized numbers or quickly hang up on perceived bullshit calls.
I asked about this demographic here a month or so back but got no reply. Are these people accounted for?

We will get change. God knows what kind.

 

[zorq]

All polls didn't turn out to be such nonsense. The poll that was scoffed at by most media and Democrats was almost dead on for this election and the previous election - The LA Times daily tracking poll.

Polls are generally weighted to "correct" for what the pollster believes may skew the data. These "corrections" are their personal bias even though they make valiant hand waving arguments to justify why the weighting must be there. Apparently in this election the general pollster bias turned out to be about five to ten percent in favor of Clinton.

That is a bit generous. The final prediction for the LA Times had Clinton winning with 352 EVs
http://www.latimes.com/nation/polit...lectoral-map-of-the-1478473458-htmlstory.html

And if you look at their final daybreak poll they had Trump beating Clinton by 3 points. But the popular vote looks to be almost dead even but favoring Clinton. This is still good compared to other competing poll weightings but they were still barely at the threshold of accurate. http://www.latimes.com/politics/la-na-pol-usc-latimes-poll-20161108-story.html
[Edit ]
I do want to mention that that last link does support what Jolly Penguin and TVACC just said.

But Trump voters were notably less comfortable about telling a telephone pollster about their vote. Voters who backed a third-party candidate were even less comfortable responding to a poll. Women who said they backed Trump were particularly less likely to say they would be comfortable talking to a pollster about their vote.

These Trump and 3rd party voters were likely embarassed for strangers on the telephone to know their preferred choice.

 

[maxparrish]

All polls didn't turn out to be such nonsense. The poll that was scoffed at by most media and Democrats was almost dead on for this election and the previous election - The LA Times daily tracking poll.

Polls are generally weighted to "correct" for what the pollster believes may skew the data. These "corrections" are their personal bias even though they make valiant hand waving arguments to justify why the weighting must be there. Apparently in this election the general pollster bias turned out to be about five to ten percent in favor of Clinton.

That is a bit generous. The final prediction for the LA Times had Clinton winning with 352 EVs
http://www.latimes.com/nation/polit...lectoral-map-of-the-1478473458-htmlstory.html

And if you look at their final daybreak poll they had Trump beating Clinton by 3 points. But the popular vote looks to be almost dead even but favoring Clinton. This is still good compared to other competing poll weightings but they were still barely at the threshold of accurate. http://www.latimes.com/politics/la-na-pol-usc-latimes-poll-20161108-story.html

True. However a few clarifications:

First, the LA Times electoral prediction was made by IGNORING their tracking poll. Clearly they did not believe it.

Second, the LA tracking poll was actually as bad or worse as many other polls. While the LA Times poll has deservedly become a far more credible poll, they are not close to being correct.

According to Nate Silver's list of polls, in the poll period ending November 7th the LA Times poll had Trump ahead of Hillary by 2 percent in the popular vote. But as of now, Hillary is fractionally (several tenths of a point) ahead of Trump in the popular vote. And Projections are that after all the votes are counted that she will end up 1.2 to 1.3 percent ahead of Trump.

If so, then LA Times missed it by 3.2 to 3.3 percent in favor of Trump. HOWEVER, the polling average of other polls were in favor of Hillary was 3.5 to 4 points...much closer to her projected margin of 1.2 points. THEREFORE, the average poll was twice as accurate the LA Times Tracking poll.

As Nate Silver has pointed out, take the polling average and add 6 points to Trump and THAT is the reading the LA Times Tracking Poll.

So which pollsters were the winner? Google Consumer Survey and IBD/TIPP. Both of these polls surveyed through November 7th, and projected a 2 percent lead for Hillary Clinton.

PS - THE REAL STORY is to be found in the several states where pollsters missed it, including Republican party pollsters. Penn., Michigan, and Wisconsin battleground polls were way off. It would be far more interesting to find out IF ANY battleground state pollster got it right.

PPS - I forgot, Rassmussen was the other clear winner. They also projected (as of Nov 6th data) that Hillary led by 2 percent.

 

[Sarpedon]

I suppose that probably the same people who view journalists as the enemy also may consider pollsters to be the enemy.

I think calling this 'embarassment' is playing into our own sense of superiority. If we learn anything from this election, it is that we should examine our own flaws.

 

[Tom Sawyer]

I suppose that probably the same people who view journalists as the enemy also may consider pollsters to be the enemy.

I think calling this 'embarassment' is playing into our own sense of superiority. If we learn anything from this election, it is that we should examine our own flaws.

That's true, but it doesn't stop this from being a major embarrassment for your country. Speaking for the rest of the world, we're laughing our asses off at you morons right now. Then we're shuddering in terror as we realize that it could stop being funny really fast.

 

[zorq]

I suppose that probably the same people who view journalists as the enemy also may consider pollsters to be the enemy.

I think calling this 'embarassment' is playing into our own sense of superiority. If we learn anything from this election, it is that we should examine our own flaws.

You are right. Completely. I stand corrected. But it still can't be completely discounted. Because otherwise, what is the difference between a web survey conducted by a newspaper and a phone survey conducted by some other journalism organisation?

 

[Jolly_Penguin]

We're also building a wall on our southern border to keep American's out. It'll take some time though. It's a big wall.

 

[Jayjay]

I wonder if an entirely different methodology would help: instead of conducting random polls, choose certain representative sample of people as a focus group and follow how their opinions change during the entire election cycle. So for example every three months (and maybe more often closer to the election) ask them same questions about who they'd vote or if they are happy with their last vote or their representatives or if they think they are more likely to vote than they were last time. Of course same questions could be asked retroactively but then people are more likely to lie, and who can remember what they thought four years ago anyway?

Surely this can't be such a revolutionary idea that it wouldn't have been tried before?

 

[Tigers!]

We're also building a wall on our southern border to keep American's out. It'll take some time though. It's a big wall.

Just play Bieber music loud and continuous and Celine Dion

- - - Updated - - -

Maybe it is just that it was not politically correct to support Trump and people were too hesitant to admit that they did, until doing so anonymously in the voting booth.

That I believe is a very important point. That was suggested in the last UK general election and Brexit.