Actually, most pollsters do not weight their results based on Party ID. They weight to meet certain demographics (usually to make sure the demographics roughly match the area being polled), but do not weight to meet a certain Party ID number - they just report the Party ID of their samples, as collected. The problem is that, for whatever reason, those samples this year have massively, and consistently, over-stated the number of Dems (and understated the number of Republicans). Whether that's because conservatives are less likely to talk to media pollsters, more likely to be out of the house (y'know, at work), or for some other reason, the samples seem to be off. It's a sampling problem, not an artificial weighing problem.
Here’s Morris explaining his take on the mis-weighting of the polls: http://www.dickmorris.com/prediction-dick-morris-tv-special-election-alert/.