Selective Sampling

October 4, 2012 3:14 PM

Daily Kos has reported on the statistical sampling behind the CNN poll that described Mitt Romney winning the first presidential debate. Interestingly, the poll seems to have consisted entirely of only white, college educated Southerners over the age of 49, who self identified as "moderate" or "not liberal". [PDF, see page 8]
posted by dobie (3 comments total)

Talking Points Memo looked into this:

Late Update: We’ve dug in a little deeper on this, and there seems to be a straightforward non-nefarious explanation. It appears that where the subgroups within CNN’s representative sample of 430 voters were too small to yield statistically valid conclusions about the subgroups themselves, CNN declined to publish those results simply because they were not reliable on their own. We’re seeking confirmation from CNN, but this is the most likely explanation — one that is valid. —dk

Later Update: CNN provided us with the internals of the poll, and the demographics of the poll respondents are very much line with normal standards for randomized sampling. —dk

posted by Kattullus at 3:18 PM on October 4, 2012 [1 favorite]

Romney didn't win it, Obama lost it. Some games score points, others score disappoints.

I just commented on this in another thread, which is too large, so I'll repost it here:

Obama lost the debate in the first five minutes. The moderator actually stopped the exchange and asked him to comment on Romney's "trickle down" comment. It was a chance to call Romney out on the right wing theory it represented, from tax breaks to the rich, and then question whether Romney knew what it meant or not. He avoided the issue entirely and acted confused instead. I wonder now if Obama knows what it meant.

Another major zinger Obama let slide was the "borrowing money from China test." I could hardly believe it when I heard it, like those idiots who claim they don't want their tax dollars going somewhere they don't like, boo hoo. The point is that any borrowed money is potentially from China, even for the military and tax breaks for the rich.

Then there's the Romney gaffe of referring to the poor as "...your poor..." indicating Obama's poor, as if he alone represents them. He could have slam dunked that one as a 47% sentiment on national television.

Major debates can't let mistakes like that just slide by, or people will wonder.
posted by Brian B. at 3:21 PM on October 4, 2012

It appears that where the subgroups within CNN’s representative sample of 430 voters were too small to yield statistically valid conclusions about the subgroups themselves, CNN declined to publish those results simply because they were not reliable on their own.

I can't parse this. Does this mean that the poll included so few non-white non-conservatives that they omitted the non-white non-conservative results? Yet still published stories implying that it was a fair random sample of the country?

And does the 'later update' imply that this is a normal thing?
posted by ook at 3:22 PM on October 4, 2012