Nate at 538 has a great post that compares how Obama is doing among various demographic groups versus John Kerry in 2004. He points out that the one group where Obama is doing worse than Kerry is among self-identified Democrats:
Oh, and what is the one group where Obama fails to outperform Kerry? Democrats — although part of that may simply be that the undecideds aren’t allocated in a pre-election poll whereas they have (necessarily) made up their minds by the time they take an exit poll. Kerry won Democrats 89-11; Obama presently leads by an average of 80-11 over the last three weeks of the Gallup poll.
I take his point about undecideds and would also add that there seems to be some difference in how Gallup (the source of the Obama data) and the 2004 exit polls (the source of the Kerry data) assign independent leaners. Nonetheless, given the near unanimous support for Obama among blacks, who are nearly unanimous in their identification with the Democratic party, these numbers suggest that Obama may be running significantly behind Kerry among white Democrats. Per Nate, I averaged the last three weeks of Gallup data and it shows that Obama is running 60 points ahead (75-15) among white Democrats.
To compare this to Kerry’s performance, I ran the numbers in the 2004 exit poll dataset (copies available from the Roper Center). This shows that Kerry won white Democrats in 2004 by a margin of 74 points (87-13).
Much of this 14 point gap between Obama and Kerry could be the result of undecideds. If the 10 percent of undecided Democrats eventually support Obama, his numbers will look pretty much like Kerry’s. On the other hand, if you look at the Gallup numbers, only about 5 percent of Republicans are undecided at this point. My suspicion is that a chunk (5-10 percent) of white Democrats is cool to Obama and are declaring themselves undecided at this point. Whether they are genuinely undecided or are only saying so as a reflection of the Bradley effect, remains to be seen.