The problem with some of these +9% Democrat polls may not be from their failure to weight by party, but their sampling techniques. The phone lists they’re using may be contacting too many Democrat households.
We now know that only 9% of all calls result in interviews.
Assembling a random sample of the entire universe out of a sample that represents only one-in-eleven is virtually impossible.
Most of the polls weight for the standard demographics -- age, race, gender, income, education, etc. But they do not weight for party identification, simply accepting whatever the result is.
Given the weighting by race, the continual over-sampling of Democrats is obviously an artifact of white Democrats being more likely to answer the phone and respond to a pollster than are white Republicans.