This will be a brief diary, intended to encourage Kossacks to keep their eyes open when reading polls.
I have no complaints about demographics (age, race, location) or party ID: Pew weight for the former and not the latter, which is the way good pollsters operate. So, I'm not complaining the poll is "skewed," in the sense that word has acquired in this election.
No, my argument is against accepting the results without question, without looking more closely at the internals and comparing with a politically-aware understanding of the nation and its electorate.
Specifically with regard to the Pew poll (here), without looking at this number in the "Among Whites" section of the table "Patterns of Voter Support, September - October":
18-49, Obama 34, Romney 63.
Which is eye-opening on its own, doubly so when compared with Pew's last poll of the same group:
Obama 47, Romney 48.
So, a 29-point advantage for Romney among whites under 50, and a 28-point shift in his favor since the last Pew poll.
All you smart, reality-based Kossacks out there: anyone buy this? Seriously, anyone think Romney is ahead by 29 points among whites under 50? I sincerely hope the answer is "no," or my faith in you all may be misplaced.
This massive swing, when weighted to the census numbers (a good practice), is what gives Romney his overall lead in the poll. It does not reflect reality. You know this.
This is not to say that Pew is up to something nefarious, or to question their competence. But those margin of error numbers are there for a reason, and sometimes when a poll has an egregiously out of whack number like that 29-point lead, you can assume with considerable confidence that the poll lies toward a more extreme end of that margin of error.
Polls are not holy texts. They should all, good and bad, be examined critically.