Fundamental to this error is a clear misinterpretation of what polling firms are doing.
A few polling firms normalize for affiliation. The good ones do not. Affiliation is an attitude that can change even over the course of a single interview. The right seems to think affiliation is like gender, race, and so on ... something one normalizes to. It is not.
from the CBS article:
"The state polls weren't oversampling Democrats and undersampling Republicans - there just weren't as many Republicans this time because they were calling themselves independents."
And this is what competent pollsters said from the beginning. They were not going out and selecting a set with 7% Dem "oversampling". They sampled a set based on age, race, gender and so on ... and of those people 7% more at that moment said "Democrat". So the right would change their data to for their projections of what "turnout" would be (that is what peoples attitude would be on exit polling as again, affiliation is an attitude).
He/she who normilizes to affiliation may as well normalize the vote. "We think Romney is going to get 53%, so we will normalize to that".
It is a fundamentally broken premise. Again "The state polls weren't oversampling Democrats and undersampling Republicans - there just weren't as many Republicans this time because they were calling themselves independents." One cannot normalize to what a respondent says as a function of current mood ... that is what you are seeking to determine, current mood, to normalize against that is simply to assert without evidence that their mood will change.
I brought this up a couple months ago and researched it. The right kept saying the polls were somehow deliberately calibrating to a 2008 Dem %. They were not. They said they were not. They were simply reporting what the respondents said on that matter. They calibrated to race, gender, age, LV and so on ... not attitude.
| 