CNN "Snap Poll" only polled Southern Whites Over 50. No joke. by @DavidOAtkins

CNN "Snap Poll" only polled Southern Whites Over 50. No joke. (CORRECTED WITH UPDATE)

by David Atkins

Update: It appears that this isn't quite so nefarious, after all. Apparently, every other subsample polled was hit in numbers that were in keeping with statistical sampling, but too few to make reliable projections about that subsample's behavior. Still, as Kevin Drum notes, it's a little odd that the particular subsamples that did have enough responses just happened to be the most conservative demographic samples in the country. But there's no question from this and other polls that most voters seemed to believe that Romney came out the winner.

What's sad is that there is some effort in the "both sides do it" beltway to suggest that progressives who noticed this strange demographic quirk in CNN's poll are somehow equivalent to the conservative "Unskewed" lunatics. As Nate Silver and others have been at great pains to point out, the main reason that the "unskewed" conspiracy mavens are so deluded is that good pollsters use key demographic baselines such as race and age to set their samples. Party ID, by contrast, is fluid. So claiming that a "skewed" party ID sample makes the poll wrong is simple failure to understand the polling process.

By contrast, seeing a subsample where only certain regions, races and ages appear (with "N/A" for everything else) does lead to scientifically valid questions. Questions that CNN does appear to have answered, but questions that were certainly worth asking. Nor do I know of any major progressives who are denying the reality of Romney's recent bounce in the polls. Some of us just believe that it's temporary, and that Romney did himself more harm than good in the end.

But to say that this episode with the CNN poll somehow proves that "both sides do it" when it comes to fact-free delusions is the willful artifice of a holier-than-thou beltway press unwilling to call a spade a spade.

TheSilverMonkey at Daily Kos makes a great catch on that CNN snap poll showing Mitt Romney winning with 67% of undecided viewiers. Check out these curious internals:

Total Men Women White Non-White
----- ----- ----- ----- ---------
Obama 25% 20% 30% 20% N/A
Romney 67% 74% 59% 71% N/A
Neither 3% 2% 4% 3% N/A
Both 5% 4% 6% 5% N/A
No opinion * * * 1% N/A
Sampling Error +/-4.5 +/-6.5 +/-7.0 +/-5.0
18- 35- 50- Under 50 and
Total 34 49 64 65+ 50 Older
----- ----- ----- ----- ----- ----- ------
Obama 25% N/A N/A 26% 22% N/A 24%
Romney 67% N/A N/A 63% 69% N/A 67%
Neither 3% N/A N/A 1% 4% N/A 3%
Both 5% N/A N/A 9% 4% N/A 6%
No opinion * N/A N/A * 1% N/A *
Sampling Error +/-4.5 +/-8.0 +/-7.0 +/-5.5
Under $50K No Attended
Total $50K or more College College
----- ----- ------- ------- --------
Obama 25% 37% 20% N/A 23%
Romney 67% 54% 72% N/A 67%
Neither 3% 2% 4% N/A 3%
Both 5% 7% 4% N/A 6%
No opinion * * * N/A 1%
Sampling Error +/-4.5 +/-8.5 +/-6.5 +/-5.5

Demor Indep- Repub- Lib- Mod- Conser-
Total endent lican eral erate vative
----- ----- ------ ------ ----- ----- -------
Obama 25% 51% 17% 3% N/A 34% 6%
Romney 67% 35% 75% 94% N/A 55% 91%
Neither 3% 4% 3% 1% N/A 3% 1%
Both 5% 9% 4% 2% N/A 8% 1%
No opinion * 1% 1% * N/A * 1%
Sampling Error +/-4.5 +/-8.0 +/-8.5 +/-8.0 +/-8.5 +/-7.0
North Mid- Sub-
Total east west South West Urban urban Rural
----- ----- ----- ----- ----- ----- ----- -----
Obama 25% N/A N/A 22% N/A 34% 20% N/A
Romney 67% N/A N/A 71% N/A 57% 72% N/A
Neither 3% N/A N/A 3% N/A 3% 3% N/A
Both 5% N/A N/A 4% N/A 6% 5% N/A
No opinion * N/A N/A 1% N/A * 1% N/A
Sampling Error +/-4.5 +/-8.5 +/-8.5 +/-7.0

It's hard to read in this format, but if you follow the link and read carefully, something shocking is apparent. The poll sample was

Seriously. All older white Southerners, mostly men. And Romney's favorability rating among the respondents prior to the debate was twelve points higher than the national average.

Now, it may well be that the audience of debate watchers skews more conservative than the general public or even the likely voting universe. But it's not entirely older, Southern whites. That's just malpractice in the first degree.


.