the confusion for some seems to be recognizing the variables being measured versus the variables being controlled for.
Currently, when a bunch of folks are surveyed by say FOX, of the likely voters, when asked about party identification, they report at rate rate of Dem +8
Exit polling in most elections over time is Dem +3.
So the right wants this parameter normalized to. The polling organizations say, NO.
They find that party identifcation is part of what they measure, it varies even with the same person in the same interview depending on when the question is asked.
So, they normalize to gender, race, some normalize to income, and so on. The Right wants party affiliation to be normalized to, professional pollsters argue it is part of what is being measured. I can certainly see how a person actually believes that party affiliation should be normalized to, but in the pollsters experience they apparently do not agree.
So the right says the pollsters are "oversampling" dems. They are not, folks are just self-identifying as dems at a higher than normal rate.
Its called brand. These things happen, but they do not normalize to them.