Poor Poll Sampling

[badrose]

http://www.zerohedge.com/news/2016-10-23/new-podesta-email-exposes-dem-playbook-rigging-polls-through-oversamples

[badrose]

Emails revealed by Wikileaks show how Democratic operatives planned to encourage “oversamples for polling” in order to “maximize what we get out of our media polling.” In other words, sample more Democrats than Republicans in order to make people believe that Hillary’s lead is far greater than the reality of a tight race.

I can see how that could backfire.

[Doc]

I'm normally suspecious of polls but I actually believe the ones in this elections are fraudulent as the article implies. I'm putting zero faith in the polls. Typically repbulicans are under-represented, or at least reported and there is no reason to believe this year is any different, and every reason to believe it will be exaggerated in this election.

[KSRBEvans]

Many of us got hung up on "skewed" and "unskewed" polls in 2012, but the polls generally turned out to correctly predict Obama's victory. A lot of the talk about "oversampling" seems to be similar to that.

This article by Fox Digital Politics Editor Chris Stirewalt explains why the talk about oversampling doesn't mean what many think it means:

http://www.foxnews.com/politics/2016/10/24/word-about-polling.html

It's long but a good read.

[Catonahottinroof]

The polls in Kentucky had Jack Conway being our next elected governor. How did that turn out?

[KeithKSR]

Many of us got hung up on "skewed" and "unskewed" polls in 2012, but the polls generally turned out to correctly predict Obama's victory. A lot of the talk about "oversampling" seems to be similar to that.

This article by Fox Digital Politics Editor Chris Stirewalt explains why the talk about oversampling doesn't mean what many think it means:

http://www.foxnews.com/politics/2016/10/24/word-about-polling.html

It's long but a good read.

It isn't oversampling that yields the skewed differences, it is the overforecasting of the potential turnout. The number of newly registered or switched party Republicans this cycle is greater than the number of newly registered or switched party Democrats. Using 2012 election turnout to forecast the 2016 election is fools gold, and perhaps doubly so. Hillary isn't going to get the turnout Obama did, he had people voting for him who never voted before and had a lot more charisma than Hillary.

The TIPP/IBD poll, which has the best prediction record over the last three presidential cycles and has this race a virtual tie: http://www.investors.com/politics/trump-clinton-in-dead-heat-as-race-hits-final-two-week-stretch-ibdtipp-poll/

They forecast the Obama-Romney race within 0.1%, which is very good.