Poll-Defying Pattern Predicts Sanders Victory

[GreatGazoo]

Bernie Sanders' historic upset in Tuesday night's Michigan Primary, which the nation's most influential polling aggregator estimated he would lose by 22 points, may prove to be the turning point in this election. That's because Sanders' two point victory in Michigan follows just a week or less after three other huge, but barely unnoticed poll-defying victories by Sanders in Colorado, Minnesota, and Kansas. Taken together with these other states, the Michigan upset is not, as America's foremost poll analyst Nate Silver claimed, a freak event not witnessed since the New Hampshire primary of 1984, but part of a new pattern of poll-defying results that will, if they continue, carry Bernie Sanders into the White House.

In the March 1st Super Tuesday contest, Sanders won Colorado by 19 points. The most recent poll there had shown him winning by only 6% of the vote, resulting in a poll-to-reality discrepancy of 13 points. In the under-reported Kansas contest on March 5, Bernie won by 35 percent, instead of losing by 10 percent as predicted, a poll-to-reality discrepancy of 45 percent.

The compelling question that eight days of election results in Michigan, Kansas, Colorado and Minnesota raises is how accurate are all the other recent polls showing Clinton victories on the March 15th Super Tuesday sequel? If Bernie surpasses the polls in these states by as much as he just did in Michigan, he stands to score historic upsets in the important delegate-rich states of Ohio and even North Carolina.

If Sanders does nearly as well as the 35 percent average poll-to-reality discrepancy of the four state pattern described above, Bernie may even win Illinois and Florida next week. Should that happen, it will be Bernie, not Hillary, who will have become "inevitable."

http://www.huffingtonpost.com/jonathan-greenberg/polldefying-pattern-predi_b_9434118.html

[MFM008]

1. i dont believe anyone polls

his, hers, NBC, none of them. No one has ever polled me.

[GreatGazoo]

4. there is a science to polling and they weight the people they DO poll

to try and match the general demographics of the larger population.

The fallout from 538's "99% chance of Clinton win in MI" has been a total re-examination of the way polls are done and weighted. Fascinating stuff for those of us who do multivariate analysis.

Wired points at cheaper polls reliance on landlines and outdated data modeling.
http://www.wired.com/2016/03/sanders-michigan-win-shows-pollsters-bernie-blindspot/

[Stevepol]

8. One Big Problem in polling!

Ever since the voting machines began to be used everywhere and to give "tilted" results due to malicious programming introduced into the machines in some way (who knows?), the polling procedures have been altered to "fit" the results, which are assumed to be accurate of course. This adds to the problem with polling. This is on top of the bigger problem: that everything that used to be science is now being twisted to support and maintain the status quo, which means the whims and wishes of the moneyed class.

[PATRICK]

9. One question

The internal polling of the best campaigns is both secret and superior to public polling- which came about to let the public(the ones polled) get as much knowledge as the candidates. Now it is so secret no one talks about internals at all except to lie and discourage one's opponent. Parties have access to probable lists and organization GOTV histories that may be better than phone pollers with more general lists?

Also it is apparent that some candidates have broad support and name recognition and machine loyalty but in voters' passion it can be both weak and vulnerable. Expectations can be routed, the curse of the known frontrunner. Primaries depend on passion which the party had planned on excising totally to make it cheap and safe to do all the real campaigning in the fall. A very stupid and dangerous and unnecessary error. How long can you keep shallow, passionless support amorphous and expect it to solidify in the fall into something other than a brick?

[LiberalElite]

2. Kicking nt

[brooklynite]

3. Simple question: how many MORE delegates will Sanders get compared to Clinton?

[JonLeibowitz]

5. Nobody knows. The future is unknowable, as you (hopefully) are aware.

[brooklynite]

10. How interesting that you didn't offer this weighty opinion to the prediction in the OP

[Major Hogwash]

6. If Sanders wins in North Carolina, it is all but over.

He's already ahead in Ohio, Illinois, and Missouri.
And that would mean that Florida would be the only state that Hillary would win on Super Tuesday 2.

Not inevitable, anymore!!!

[Betty Karlson]

7. Thanks for the voice of reason.

Stunning analysis.

[GreatGazoo]

11. a 35% variance from polls is sign of how real this revolution is

data models expect past results to predict future results but many of those models are being re-examined right now.

The other part of this is INTERNAL polling that the campaigns use to portion out their ad buys and decide which states are in play. Perhaps after the election we will hear more about whether or not internal polls were any better than the public polls.

Also points to how wide the gap between those who rely on old media and those who don't is. And that old media is struggling to connect with a younger audience. Trump is all about TV and fear while Sanders is building from a base that has newer media preferences. Trump may yet turn out to be the "enough rope" that the old media "hangs itself" with.

Fascinating stuff and today will be more of it!