Polling on this Election: What the Hell Happened
supporter
Posted by mikeh (+302) 5 years ago
Out of the lemons this election handed me (I was with her) I decided to make a little lemonade by making it a teachable moment for my online Stats class.

Three videos - 25 minutes total - and despite not being super slick, I do hope they give insight as to how and why the polls said what they did and the results of the election turned out as they did.

Part I


Part II


Part III


Maybe if this is helpful, my colleague, the New Dad can work with me on something a bit more slick.

[Edited by mikeh (11/9/2016 11:39:27 AM)]
Top
+4
-2
moderator
founder
Posted by David Schott (+17053) 5 years ago
I fixed it for you, Mike. You had some extra [url] [/url] tags in there. They are not needed in conjunction with the [youtube] tag.

- Dave
Top
+2
-1
admin
moderator
founder
Posted by MilesCity.com Webmaster (+10001) 5 years ago
Reply to mikeh (#369109)
Thanks Mike. Enjoyed watching.

Minor point, but technically there's not 50 different elections ... but 51 (with DC), or perhaps more accurately 56 (with the extra NE and ME districts).
Top
+2
-1
supporter
Posted by mikeh (+302) 5 years ago
Thanks for the fix...and the feedback!
Top
+1
Posted by Sam Colt (-138) 5 years ago
This isn't that complicated. We were force feed BS polls for a year plus... as usual. I knew Trump was going to win.
Top
+3
-6
Posted by Oddjob (+194) 5 years ago
This is good stuff mikeh. Thanks for the videos. I know the internet polls are unscientific and too easily manipulated by bots, but wasn't it interesting they all indicated an overwhelming support for Trump. I thought that was just too easily dismissed by the "purists".

It would be interesting to see how many of the major polls in the "too close to call" category fell within their calculated margin of error or confidence interval and think they did OK because of it.. The problem there lies within the fact that almost all of them had Clinton in the lead, which tells me there was something wrong with their basic assumptions or definition of "likely voter". Even so, if the final outcome of vote percentage fell within the margin of error, it was still a major fail.

On my own part, I knew 20 years ago that I was never going to vote for another Clinton and I think a significant number of actual voters were in the same boat. So for us, we were tired of the derision and political argument trolls that have been so pervasive, especially for the last 8 years.

Living in a "swing State" my land line rang constantly. I never answered any of the calls. Never answered any calls on my cell from numbers I did not recognize. The fact of the matter is, that I would have lied anyway if I had been cornered. This is the new reality the pollsters have been facing for the last 4 to 6 election cycles and they either don't know how to fix it or are unwilling to face reality. Another thing is, very few of the pollsters are basically honest in framing their questions and seek a foregone conclusion.

So essentially they are getting the lions share of opinion for their polls from a biased population or they are getting the answers they are interested in seeking. None of them reached the "never Clinton" voters who eventually voted for Trump. (Perhaps with the exception of the LA Times) Apparently there were enough of them for Trump to carry the States he was never expected to win. That's why I believe they were all so far off the mark and their is no solace for any of them for falling within the "margin of error".
Top
+3
-1
Posted by Bob Netherton III (+2773) 5 years ago
Not only all that Oddjob - we don't even have a land line in our household. I'm willing to bet there are many others that are in the same boat. Since Montana isn't a swing state, I'm guessing people here didn't get nearly the calls you did.

Some media outlets were using a "poll of the polls" methodology which I figured would be a safe bet. But if the polling methodologies are flawed in nearly every poll to begin with...

Here in Montana, some polls had Bullock up by 7 points. He barely squeezed pass.
Top
+3
supporter
Posted by mikeh (+302) 5 years ago
Thanks, Bob & Oddjob.

The folks who do polling for a living are looking for the sorts of things you describe - biases - and they are a real problem (in fact you don't even have to have a bias as a person to inadvertantly insert biases into the problem). Most of this seems like "Monday Morning Quarterbacking" - except possibly for one thing. Trump won the Republican Primaries by turning out turned off voters...yet the likely voter models apparently still failed to take some of these disaffected new and returning voters into account sufficiently. They knew about the problem before General Election season but did not find a way to adjust for it accordingly.

A great article on this appeared on FiveThirtyEight last September. I will link to it below.

http://fivethirtyeight.co...-register/
Top