Comment Re:Cell phone people are different (Score 1) 206
Another important point is that modern telephone polls have response rates in the neighborhood of 30%. In the same way that those with cell phones are different than those without, people who are (a) at home when the poll takes place and (b) agree to participate are different than those who are not or do not.
In surveys done for academic research, this kind of selection bias is taken very seriously. Political pollsters either need to model the selection process, or they need to be more honest about the accuracy of their surveys. The +/- 3% rule only applies if the sample was drawn independently, an assumption that is violated in the presence of selective non-response.
In surveys done for academic research, this kind of selection bias is taken very seriously. Political pollsters either need to model the selection process, or they need to be more honest about the accuracy of their surveys. The +/- 3% rule only applies if the sample was drawn independently, an assumption that is violated in the presence of selective non-response.
