Discussion for article #229469
Fascinating. Playing smarter, not meaner.
I have no desire to succumb to “may-it-be-so” unskewing, but I think there are good odds that blacks, Latino/as, the young, and women (especially) are being significantly undersampled in the '14 mid-terms polls. I hope I’m right.
1 Like
Votes the size of cantaloupes!
6 Likes
Generally speaking, for some reason, the polls tend to be a 1 or 2 points off for Dems in Colorado.
Even if that holds true, however, this race is pretty much like every other race…extremely tight.
I expect him to win by about 4 points. Nonwhites are always under polled, even when it’s them
1 Like
Absolutely. I’ve asked so many times but I’ll ask again…How are pollsters building likely voter models? Are they simply relying on 2010 demographics or is it some mix of 2012 and 2010? Are these polls almost exclusively land lines or is it a good mix of cellphones and land lines? Young and minority voters are far more likely than older, white folks to have cell phone only homes.
1 Like
This poll may seem like it would be left-leaning, but note that the sample is 40% Republican as compared to 33% Democrat. This is probably a more Republican-leaning electorate than what will show up on Election Day. Expect this to be a close one, but Udall has a good shot.
1 Like
Men. Smells alot like unskewing to me. We’ll see what happens, but my gut is telling me Udall failed.
2 Likes
I’ve got pol fatigue, and I’ve got poll fatigue.
Also, Cory Gardner is a preening, self-satisfied con man of the lowest order.
1 Like
Some will screen you by asking you if you intend to vote, how strong that intention is, and sometimes if you’ve already voted.
Others, apparently, use a predetermined demographic model. These seem to be the models that fuck the goat (See: Gallup, circa 2012.)
1 Like
Democrats have been under-estimated in polling—win or lose—by about 2 or 3 percentage points since the 2010 midterms.
Smells like LV modeling.
“What is unique about the poll, however, is that it intentionally over-samples Hispanic voters and then re-weights the findings to reflect their share of the anticipated electorate on Election Day. “The sample was weighted to ensure a proportional demographic representation of the likely 2014 electorate,” the pollster explained.”
This is what I have been saying for a long time now, that minorities are notoriously under polled in most of these polls. That this poll specifically states what it did to try and find what the true statistics look like, by making it an aim to poll Hispanics, which most other polls never do, proves that most polling is based on outdated demographic models. As someone else also pointed out, they had a 7 points higher in Republican respondents than Dems, and yet this poll still found that with this model of polling, that Udall is up by 1 point.