Not a valid point, the shift to Bush was picked up by almost every poll. Zogby is an exception but he weights.
If so, then it is an explanation. The bottom line here is that party affiliation may not be a relevant demographic. When one says Dems were "overpolled" one cannot know if that is correct because no one knows who will turn out. Also, one cannot say "overpolled" unless the results were forced in that regard.
I think you missed the point, I never said dems were overpolled. Here's my opinion on polling. If a voter is registered, he is registerd with a politcal party or as an independent. That should be the second question asked. First is are you registered? Second should be, in what party? Voters can not shift parties at a whim. If they are registered, they're registered. They can and do cross over but that would show up in the polling, no?
There is another consideration. It is not clear to me that phone numbers are on registration lists.
They are not. Phone numbers are dialed randomly which presents a myriad of problems to todays pollsters.
>
I can't, of course. But this is not the only poll that showed this. Gallup showed a shift in partisan mix when Bush took the lead from Kerry. People may not respond to the question from the perspective of how they are registered. They may respond with how they lean that day.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
Not a valid point, the shift to Bush was picked up by almost every poll. Zogby is an exception but he weights.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
I typed that badly. The point was that Gallup's partisan mix shifted when Bush took the lead. Everyone's did. They call randomly and who answers the party question gets their answer recorded.
Look, if the vast majority of variance of a poll's result can be explained by the partisan mix of a poll, then what exactly is being polled? If you call 1000 zip code weighted and randomized phone numbers and you keep getting wide variances in partisan mix -- which produce wide variances in the presidential horse race, what is it we are measuring?
The phrase . . . they weighted their sample with X Dems and Y GOP this time vs last time suggests they did it on purpose. They almost certainly didn't. It would screw up their methodology to constantly vary that mix. They would not be measuring anything at all other than their choice of mix. No one will pay for that result.