Judis & Bowers on Research 2000 polling
by Jerome Armstrong, Mon Sep 22, 2008 at 11:42:28 AM EDT
Chris Bowers asks:
Having watched how Research 2000 has polled in the past, the answer is probably that the demographics of those polled generally mirror the population; with quotas assigned to reflect the voter registration.
It's a pretty sketchy likely-voter methodology. One that would inherently favor Democrats, as Judis notes, and one that doesn't reflect the reality of the partisan make-up of those that vote, as Bowers notes, and more.
To believe the Research2000 prediction for 2008, you have to believe that latinos & blacks will vote in equal numbers to their populations as whites do, that Democrats will outnumber Republicans by 9% and Independents/Refused will be nearly 40% of the voting day population, and to assume that older people (60+) are going to be less of the voting population (even though the trend says they will be more).
It seems absurd to take it at face value. But this isn't something that Markos thought up, he's just paying for it. It's entirely the methodology of Research 2000. I'm not sure exactly how I'd re-weight the poll either; given how off it is across the board, it's probably not salvageable. Well, the MOE is 3% so I'll guess that it favors Obama by that much.