David Atkins passes along a strange find this morning. CNN's instapoll of the debate had Romney winning by 67-25. But if you look at the crosstabs, it appears that their sample was entirely white, entirely over 50, and entirely from the South. Maybe there's a good explanation for this, but it seems a bit odd, no?
UPDATE: TPM says, "CNN provided us with the internals of the poll, and the demographics of the poll respondents are very much in line with normal standards for randomized sampling." OK. Though I'm still not sure how you account for such odd crosstabs.