As ttujoe remarked yesterday, crosstabs can reveal major differences in polling samples. For example, the latest PA poll from Rasmussen (subscription) has Clinton winning 21% of the Black vote, which as I'll discuss is about 10% too high. Thus the 5 point gap (C 49, O 44) in the poll might be too large.
Similarly, the latest poll from Mason Dixon (PDF) shows that Catholics make up 41% of the survey sample (though only 29% of the population), whereas Protestants are just 31% of the sample (but fully 50% of the state). Even if Catholics are disproportionately Democratic in PA, the MD sample with its huge Catholic shift appears to be out of whack. The crosstabs show that Clinton wins twice as many Catholic votes as Obama, and the reverse is true of Protestants. I notice also that unlike most recent polls, MD gives Obama a much higher unfavorability rating than Clinton. So it could be that the sample needs correcting, in which case Obama might trail Clinton by less than the 5% difference MD shows (C 48, O 43).
One of the most interesting points of comparison among competing polls are the crosstabs for race. Mason Dixon has fairly typical numbers for Black voters (C 10, O 83, undecided 7). SUSA presses undecideds hard, and its numbers are revealing: C 11, O 87, undecided 2. In other words, it looks as if nearly all undecideds break for Obama and he will pick up fully 89% of the Black vote.
That suggests Obama will get the large margins in Philly and suburbs that he needs to be competetive statewide. As BooMan points out, the largest increases in Democratic registration in the Philadelphia metro area are in heavily Black neighborhoods. In fact, throughout the state it seems to me that the big increases in Dem numbers came primarily from Obama supporters, particularly in areas where he's more competitive. So for example in western PA, such as Clinton stronghold Westmoreland County, registration generally was pretty flat. In the west it was only Pittsburgh, where Obama is more popular, that saw major increases comparable to those in the southeast. There were over 316,000 newly registered Democrats in the state (7% of Democrats are newlyl registered), and from the patterns of new registrations I surmise that Obama will get a 4 to 5% bump from that; his net gain will be about 100,000 votes (mainly in Philly and suburbs, which have 43 of the 103 delegates awarded by Congressional district).
Incidentally, pollsters appear to be undersampling the newly registered voters. For example, in this Franklin & Marshall poll new voters constitute only 4% of the sample. It's another indication that polls may be slightly underassessing Obama's support. [Correction: I misread the F&M poll. Newly registered voters comprise fully 10% of the sample, not 4%.]
And what do crosstabs indicate about White voters (82% of the population)? Rasmussen has C 55%, O 37%, undecided 8%. Mason Dixon has C 58, O 33, undecided 8, "other" 1. Most other polls put Clinton's share of the White vote in the mid to high 50s, and that tracks pretty closely her level of support in western PA and the "T" which are heavily white regions. SUSA presses the undecideds and comes up with: C 58, O 36, undecided 2, "other" 5. In other words, the undecided are not undecided about Clinton; she gains almost nothing, whereas a few move to Obama and the rest prefer "other". That suggests to me that Clinton's support in her strongholds will top out just about 59-60%, not enough perhaps to give her the overwhelming victory that she needs in PA.
The math for delegates awarded by Congressional district is unfavorable to Clinton to begin with. Only 6 out of 12 of her stronghold districts have an odd number of delegates. If Clinton does not win more than 60% of the vote there, she won't gain more than 6 delegates against Obama outside of southeast PA. That gives Obama a very good chance of winning a slight majority of those 103 delegates awarded by Congressional district, even if as expected he loses the popular vote. The final tally of delegates will include another 55 at-large delegates awarded in proportion to the statewide popular vote.
And for what it's worth, I'm guessing that Obama will finish within 4% of Clinton in Pennsylvania.
Comments are closed on this story.