I've read and read ... party affiliation data is not used to massage the data. the party affiliation numbers in the major polls are a reflection of respondent's attitude and not targets based on previous cycles.
That was my question.
So, that 8% more democrat respondents found in a particular poll is simply that more said so, not that they sought to target 2008 levels or anything like that. The unskew type spin is that the polls are deliberately targeting a certain level, they are not, they are simply reporting what their sample (normalized for race, gender, LV etc) says.
They sift the data by race, gender and such things that they believe to be stable metrics to accurately reflect the LV demographic, but that party affiliation is considered an "attitide" not something they normalize to.Indeed, respondents will give different party affiliations based on what the last questions were, and when during the interview you ask it, but they general won't change their gender or race or age in such a fickle manner.
So, when folks feel good about Dems ... they are willing to identify as such, when they feel poorly about dems, they don't identify as such.
For these reasons they do not normalize to affiliation. So the unskewed folks want the polls to normalize for affiliation (more or less, one calls themselves a dem today, but they really aren't, or, we'll toss them out cause in the end folks will self-identify within a certain range). The point is, how they self-identify today is exactly that, and co-migrates with the candidate choice ... unlike the things they do normalize for (race, gender etc).
In essence pollsters argue that normalizing for affiliation would be like normalizing for the last vote.
53% voted for Obama last time, so we will normalize our data to achieve 53% support for Obama. Certainly, that would be rubbish, and the pollsters find that affiliation and vote go hand in hand, are both plastic, and hence affiliation cannot be normalized to ... affiliation is *part of* what they are measuring in addition to the actual vote and that these readouts comigrate.
Cool. That is what I was hoping. As the race nears conclusion, if the difference narrows, so too will the affiliation, it will happen on its own, not because it is screened that way.
| 