Reply #42: Look at This Discussion of Reweighting by a RWer [View All]

[On the Road]

Still looking for a discussion of weighting methodology in exit polls. Here's another hint:

From The Folks Who Brought You President Kerry—The Smoking Hispanic-Share Exit Poll

"Having stayed up all election night, around dawn I noticed that the exit poll data had been quietly rejiggered to show Bush winning narrowly. By Wednesday evening, it had been fiddled with again to give Bush a bigger margin. This doesn't leave me with a lot of faith in the exit poll.

The big difficulty with an exit poll is coming up with a representative sample of polling places. Apparently, the NEP failed to do this.

Most of the exit poll data, as it currently stands, seems not too implausible. It shows Bush's share of the crucial white vote growing from 54 percent to 58 percent, which almost accounts for Bush's three point overall rise by itself. Among blacks, Bush's share grew from 9 to 11 percent, among Asians from 41 to 44 percent, and among American Indians and others from 40 to 41 percent—i.e. it’s trivial.

http://www.vdare.com/sailer/041107_election.htm

Sailer seems to be suggesting that geographic modeling takes the lead and that polling organizations try to back into demographics from the geography and maybe the poll sample. It's worth a look -- Sailer actually has some perceptive things to say on political strategy.