Do national pollsters...

[ncteechur]

make sure they have an proportional sample from each state? That is to say that a national poll could be considered one that takes a few respondents from a number of states but couldn't they take an overwhelming number of responses from Wyoming or some other red state? Is that the problem we have seen in the Gallup and other polls?

[hlthe2b]

1. They are oversampling Republicans, that's the problem....

Latest Newsweek Sample:

SAMPLE SIZE/MARGIN OF ERROR FOR REGISTERED VOTERS SUBGROUPS:
1,003 Registered voters (plus or minus 4)

391 Republicans (plus or minus 6)
300 Democrats (plus or minus 7)
270 Independents (plus or minus 7)

481 Men (plus or minus 5)
522 Women (plus or minus 5)

99 18-29 (plus or minus 11)
321 30-49 (plus or minus 6)
321 50-64 (plus or minus 6)
236 65+ (plus or minus 7)


http://biz.yahoo.com/prnews/040911/nysa006_1.html

But, also realize that national polling does not reflect electoral college results, since you need state-specific polling for that. So the extrapolation to an electoral result is pretty hard from these national polls. They are a glimpse of national population trends and thats about it...

[AndyTiedye]

2. They Oversample Republicans so they will Match the Diebold Results