Originally Posted by alnorth
protip: self-identified party affiliation is very fluid and can change rather quickly.
They aren't usually asking people how they are registered, just what they think of themselves as right now. So, a poll that shows a +10D sample in a state that was +2D 4 years ago isn't necessarily skewed.
This is something I've only accepted recently, so its understandable why people on the right may be confused, but they'll have a rude awakening soon. The left went through this 8 years ago when they openly mocked polls showing Kerry losing because of an "unprecedented" amount of people identifying themselves as Republicans. Turns out, the Republican brand was stronger back then, and people were more likely to tell a pollster they are a republican, and less likely to identify as a democrat. Prior "democrats" started saying they were independent, and prior "independents" started saying they were Republican.
We're going the other way now. I know this happens to some extent because I would have told a pollster I was republican 4 years ago. Now I'm telling them I'm independent. Party identification is not something you can pre-determine and re-weight to, it is something you discover. The only thing pollsters should be weighting is age and race to more accurately reflect likely voters. Hearing someone say "I'm a democrat" merely means they are probably voting for Obama. 4 years ago that person may have said "I'm independent" and struggling to decide between Obama and McCain.
unskewedpolls.com is a very stupid site which will be laughed at shortly.
This is exactly how pollsters do it. With one or two exceptions, all the pollsters weight their polls by demographics
and not by party affiliation.
Which is why the 'skewed polls' argument doesn't make any sense. As it is now, the polls don't cook any party breakdowns into their results at all.