October 24, 2012 - 4:22 pm
The sampling is daft. Here’s Time’s write-up, but here are the crosstabs. You’re welcome to read all 49 pages, but here’s the kicker: The sample is D 39, R 30. That’s higher than the 2008 differential.
This poll is at best an outlier, and at worst bogus. CNN just cited it without noting the sampling problem. That’s how the MSM game is played.
Some Gallup guy was on TV the other day and said the way they did it, and he thought the way everyone did it, was to choose your 100 people without (officially) knowing their party affiliation.
You ask for that last, and then report it.
Of course this conflicts with what we all believe, and I thought Dick Morris said, that they FIRST get a party profile from the last election and choose a sample to match.
So – I dunno.
I’ve heard both versions from different pollsters. I believe you’re right, I seem to recall reading an article from Gallup back a few weeks ago that this was the way THEY do it. This brought into focus the revelation from Pew that 91% of the people they call either don’t answer the phone or refuse to cooperate with them. It could be that other organizations do something different.
It depends on the poll. That’s also how Pew does it. But look back at my piece on sampling a few days ago: http://pjmedia.com/blog/polling-theory-when-polling-is-useless/
In that example, I knew I started with close to the same number of red and blue BBs. If I then took a sample and the sample was 60 percent blue and 40 percent red, that could be a perfectly reasonable sample — in fact, we can compute exactly how probable it would be to have a 60/40 sample come out. None the less, 60/40 would be unexpected, and I’d have every reason to start looking for something that could be skewing my sample.
We also know our sample is self-selecting somehow — with ten people out of eleven not cooperating in the poll, we may not know how they’re self-selecting, but we have no reason to think they’re self-selecting purely randomly.
The point is, D+9 may well be the sample they’re getting; the result isn’t necessarily “wrong” in the sense that it’s being purposefully skewed. It should still be looked at with caution.
Now, look at Rasmussen’s poll, which has Ohio tied, and (since PJ bought me a Platinum membership I can see the crosstabs now) has a sample that’s 39/38/23 R/D/O.
That still doesn’t explain all the times they call you, you agree to be polled, they ask you right away what your affiliation is, and then they thank you and hang up. What kind of poll is that? They have to be selecting for a sample because they haven’t asked me anything else, just affiliation.
It could be explained in a poll that has a turnout model — they’d already polled their full quota of Republicans. Of course, that ought to make their turnout model questionable. You don’t happen to have noted the actual polling company, perhaps?
They all know they’re headed for the eastern front
“Real Clear Politics” has almost everything within the statistical margin of error. But, interestingly, the “battleground” states now include: Nevada, Colorado, Iowa, Wisconsin, Michigan, Ohio, Virginia, Pennsylvania, Florida, and New Hampshire. That puts the Electoral College at 206 – Romney, 201 – Obama, and 131 toss-up.
With a 9 point bias, they could only put Obama up by five?
Someone needs to hire a new pollster…
D+3 would be more realistic which would make it R47/O46. If R turnout is better than this or the majority of undecideds break for R then O is toast.
Golly, umeployed Democrats home at 2:00pm for the pollsters and afraid of losing their benefits are up by 5% for the Democrat… hmm.
Average Ohio neighborhood:
Romney-Ryan signs = 26
Obama-Biden signs = 18
Yep, math. Hard. But isn’t that a 30% difference, advantage Romney?
Bottom line: No one’s answering polling calls, but some are posting signs in their yards.
judging from the overall poll’s accuracy in the past (reagan will lose?, didn’t see 2010 coming? – just 2 examples that come to mind) and what i hear now from all around me, i’m of the opinion that they stand between 2 crack houses to take their samples. i can see where some might think it an appropriate, good and likely spot, as that appears to be where some of their candidates got their start.
prediction: intense cheating by dem’s, and still romney wins. then the sore loser will …? not go quietly into that good riddance.