Steve Benen has made a graph of part of the results of a Research 2000 poll:
Now sure, it’s a DKos-sponsored poll, but R2K is a respectable outfit and the sample size is big enough to push the margin of error down to 2%. So even if the absolute numbers are off, the region-to-region comparison ought to be pretty close.
Voters, of course, vote for (and against) candidates, not parties. But still…