On Thu, 21 Oct 2004, Doug Henwood wrote:
> The AAPORites beef with him include his tendencies to weight respondents
> according to his feelings about turnout among various demographic groups,
> and his practice of weighting by party ID.
Speaking of which -- has anyone come up with a reasonable explanation why so many polls are consistently hugely oversampling Republicans?
If those oversamping are all wrong -- which I still think is the case -- then the weighting method the mainstreamer have been holding against Zogby begins to look more and more legitimate. The basic principle used against him is that we know from trend charts that registration changes over time, and if you specify it, you'll miss that, and your sample won't reflect changing reality. But since the same trend charts show that it doesn't change this much this fact, in now seems that same principle argues now in favor of his method over whatever unknown factor is going wrong with the mainstream polls -- to the point where Ruy Teixeira, a stand-up and diligent guy, has been more and more incorporating it into his standard repetoire.
I thought the more damning indictment against Zogby was that he called up to half his sample during the day, and counted non-responders as part of N. Those two practices seem to make nonsense of the whole statistical procedure. But maybe he only does that sometimes, like during the primaries, when time is of the essence?
Michael