On Fri, 1 Aug 2003, Jon Johanning wrote:
> But the polling organizations still seem to worry about running into
> more refusers. Maybe it biases the results in some way that can't be
> corrected for, or just because it makes the public *think* the polls are
> less accurate.
>
No, there really is a potential sampling bias problem here. What if the refusers are more likely to (say) support gay marriage? Then the results of any polls on the topic will be skewed, because the refusers didn't participate. The only way high refusal rates would not lead to sampling bias is if the refusers and participants were equivalent, except for the wish to participate in the poll. That's a questionable assumption.
Miles