The Modern Business of Election Polling

[GGT]

With all the Trump bluster about polls being "rigged", this might be a good time to delve into polling methods, and when survey data is considered 'scientific'.

http://www.nytimes.com/2015/06/21/op...ling.html?_r=1

http://cstl-cla.semo.edu/rdrenka/Renka_papers/polls.htm

http://video.cnbc.com/gallery/?video=3000560769

http://www.businessinsider.com/nate-...crisis-2015-11

http://www.newyorker.com/magazine/20...he-new-machine



"The polls we actually see are advertising for something else".

[Loki]

Have you found a new technical bogeyman to fear?

[GGT]

I'm just pointing out the American obsession with polls, and how often they're cited incorrectly. There are so many polls released every day, too many IMO, that people aren't paying attention to the semantics or methodology. All too often survey results of 'likely voters' (or registered Republicans or Democrats) are attributed to 'Americans'. As in, over 50% of Americans believe the US is on the wrong track, for example.

And the media reports every poll as if they're all the same, which they're not. Most are land line phone surveys (because cell phones can't be robo-dialed by law) which misses the growing number of people who only have cell phones. And internet surveys have their own flaws, especially since their sample groups aren't random. Beyond that, I'm not sure all this polling doing our political process any good. It's already a circus, more like a reality TV show, focused on personalities and he said/she said gossip, entertainment for short term attention spans.....instead of in-depth policy debates.

[ImAnOgre]

Have you found a new technical bogeyman to fear?

I understand (from working in the industry,our company owns a TV station or two, a few newspapers,and various other online publications) that most polls only call those with landlines. As many people have dropped home phone service and rely on cells only, it's pretty easy to see how ever poll can be incorrect. Gigo would apply here.

[Loki]

I understand (from working in the industry,our company owns a TV station or two, a few newspapers,and various other online publications) that most polls only call those with landlines. As many people have dropped home phone service and rely on cells only, it's pretty easy to see how ever poll can be incorrect. Gigo would apply here.

No, most credible polling agencies call cell phones (the ones that do manual calling, instead of having a computer do it). This gets a much lower response rate, which means the people who do answer their cell phone get weighed more heavily. This could cause some problems, but it's one that polling agencies constantly try to fix.

[wiggin]

Loki - While I agree with your basic point, I do think that really understanding polling requires a certain degree of sophistication - there are many different methodologies for sampling/weighting/etc. and it does mean that polls may be biased in one way or another (though not necessarily due to a partisan bent, but rather methodological issues underpinning polling). This is particularly true if we suspect that a given election will call into question some of pollsters assumptions that they use in building their models. It's by no means obvious that polls are an accurate representation of reality, though I agree that they are continually trying to improve their predictive nature.

[Loki]

My point isn't that polls are perfect (far from it). It's that they have important uses and shouldn't be dismissed out of hand by the Luddites of the world.

[GGT]

My point isn't that polls are perfect (far from it). It's that they have important uses and shouldn't be dismissed out of hand by the Luddites of the world.

And today the pollsters are wiping egg off their faces, congratulating a President Elect Trump. Questioning the sampling, methodology, and assumptions of polling doesn't make someone a Luddite -- but relying on them too much is foolish.

[Aimless]

Will be interesting to see what the turnout was in various age groups.

[Hazir]

I think we can safely say that trusting the polls is liable to make a fool of you.

[GGT]

For Loki: http://www.aapor.org/Publications-Me...on-Pollin.aspx


A quick search shows the entire polling industry, and news agencies, are asking the same questions I did.