For all of us who need to know the Pennsylvania results 30 hours before we will have them, Public Policy Polling (PPP) and Survey USA have given us competing internals. Then we can figure out who is right, or what each candidate needs to do to beat expectations. With internals on vote by gender, race, and region we take the tens of millions of dollars spent and 6 weeks of hard campaigning and boil it down to a spreadsheet. Both pollsters use automated polling rather than live interviewers, but their results are quite different.
I've broken everything down for each of the polls for comparison, and then threw in the Ohio exit polls since Ohio is probably the closest thing we have to an analog of Pennsylvania. Ohio's African-American population is 10.6% versus 9.2% in Pennsylvania, but I'll assume that with Pennsylvania's closed primary, the racial breakdown of the voters works out about the same between the two states.
Women
Survey USA (55% of sample): Clinton 60%, Obama 37%
PPP (58% of sample): Clinton 50%, Obama 41%
Ohio (59% of vote): Clinton 57%, Obama 41%
Men
Survey USA (45% of sample): Obama 53%, Clinton 38%
PPP (42% of sample): Obama 55%, Clinton 34%
Ohio (41% of sample): Clinton 50%, Obama 48%
White Voters
Survey USA (81% of sample): Clinton 58%, Obama 36%
PPP (76% of sample): Clinton 52%, Obama 38%
Ohio (76% of vote): Clinton 64%, Obama 34%
African-American Voters
Survey USA (14% of sample): Obama 87%, Clinton 11%
PPP (18% of sample): Obama 81%, Clinton 12%
Ohio (18% of vote): Obama 87%, Clinton 13%
18-29 Years Old
Survey USA 18-34 years old (21% of the sample): Obama 57%, Clinton 36%
PPP (16% of the sample): Obama 50%, Clinton 39%
Ohio (15% of the vote): Obama 62%, Clinton 35%
30-49 Years Old
Survey USA 35-50 years old (29% of the sample): Clinton 49%, Obama 47%
PPP 30-45 years old (25% of sample): Obama 49%, Clinton 41%
Ohio 30-50 years old (38% of the vote): Clinton 50%, Obama 50%
50-64 Years Old
Survey USA (28% of sample): Clinton 55%, Obama 41%
PPP 45-64 years old (35% of sample) Obama 47%, Clinton 44%
Ohio (32% of sample): Clinton 60%, Obama 37%
65+ Years Old
Survey USA (21% of sample): Clinton 62%, Obama 32%
PPP (24% of sample): Clinton 48%, Obama 42%
Ohio (14% of vote): Clinton 72%, Obama 26%
Southeast PA
Survey USA (43% of sample): Obama 55%, Clinton 41%
PPP (45% of sample): Obama 58%, Clinton 32%
Southwest PA
Survey USA (24% of sample): Clinton 58%, Obama 36%
PPP (26% of sample): Clinton 51%, Obama 38%
Elsewhere in PA
Survey USA (33% of sample): Clinton 56%, Obama 35%
PPP (29% of sample): Clinton 53%, Obama 37%
These two automated polls have some serious differences of opinion on how different demographics will vote, but it seems like Survey USA looks more like Ohio turned out than PPP. PPP tends to think the voters will average a little more female, urban and African-American than Survey USA projects, which would tend to favor Obama a bit. PPP thinks Clinton will only win seniors by 6%, which is pretty hard to believe. Based on Ohio and an assumption that unaffiliated voters who voted in Ohio have a lower percentage of African-Americans than Democrats, I'd say PPP's estimate of 18% African-American looks better than Survey USA's estimate of 14%. PPP's estimate of 58% women in the primary also lines up better with Ohio (and IIRC other primaries as well).
Shifting Survey USA's portion of the African-American vote from 14% to 18% and the women's vote from 55% to 58% but otherwise keeping Survey USA's reasonable vote splits between Clinton and Obama, the final result will be Clinton 52.5%, Obama 47.5%. That would definitely be a pyrrhic victory for Clinton, as her fundraising would dry up after not even meeting expectations in a state she needs to win big. Remember, you heard it here first.