I read the whole article. Since I teach marketing research, I always find this kind of stuff fascinating. I like his hypothesis that the smaller samples have high hang-up and non-response rates, and non-respondents are more likely to vote Romney.
On the other hand, if a pollster is conducting phone surveys via landlines, that sample is likely to skew more heavily TOWARD Romney as the population of landline owners is older and more stable than the population of voters.
This thing is so close, it really comes down to who shows up to vote, and that is more problematic for OPapaDoc than it is for us. Our people can't wait until November 6, and they are going to have to show some self-control to keep from damaging the ballot, they will be pressing down so hard.