I want to address one "unskewing" point -- % Dem vs % Rep
Statistical analyst in real life, do this kind of work for a living, blah, blah, blah.
Anyway, during the primaries I did a few primers on polling, and one of the issues I touched on was setting population parameters. These are the factors that pollsters weigh their polls by and can include things like race, gender, socioeconomic status, etc.
One of the things a good pollster never does is weigh by party self-identification (Not the party the respondent is registered with, but the party he or she most closely identifies with). The reason for this is very simple--forcing a sample to match percent weights for parties forces results into uniformity. In other words, you wouldn't need to actually poll anyone, because the weight you set to each party would tell you what result you're going to get.
Instead, pollsters simply ask what party the person identifies with and records that result. When someone asks a pollster a question like, "How could you have X% for Democrats?" The answer they will inevitably be given back is, "Because that's how many people self-identified as Democrats when we asked." In the case of CNN, they ended up with more people self-identifying as Republicans than Democrats.
CNN may very likely be an outlier on the LV side, and it is the belief of most other major pollsters and analysts that it is, in fact, an outlier. But the party ID results is more a reflection of random selection for a sample than "skewing" a poll to be anti-Hillary.
= new reply since forum marked as read
and not be "skewing" a poll to be anti-Hillary.
