(Promoted from the diaries by kos. I don't buy the final prediction, but this is a good look at why the numbers vary from poll to poll. I was about to write up a similar post, but now I don't have to. )
In the last couple of days, a slew of PA polls have come out, with ranges from a 13 point Clinton lead to a 3 point Obama lead. Follow me below the jump for some interesting things that come out of the crosstabs.
For this diary, I'm focusing on the five polls for which some level of crosstabs are available (I have the Rasmussen subscription to see their crosstabs):
I was waiting to get the SurveyUSA crosstabs before publishing the diary, but I'm impatient- I did use some data from the WCAU news story on that poll.
I couldn't care less about anything Zogby or ARG say.
Here are some things that strike out at me:
The Suffolk regional demographics
Obviously, the Suffolk poll is the most favorable to Clinton out of this group. There is one big thing that stands out in this poll:
Pittsburgh/Southwest PA: 38% of all polled for Dem primary
Philadelphia/Southeast PA: 38% of all polled for Dem primary
Rest of state (the "T" area): 24% of all polled for Dem primary
Suffolk is the only polling firm approximating that SW PA (including the Allegheny area) is an equal voting bloc to Philly & SE PA. Other firms have the numbers at:
45% Philly/SE, 27% Pittsburgh/SW (Mason-Dixon)
45% Philly/SE, 26% Pittsburgh/SW (PPP)
46% Philly/SE, 23% Pittsburgh/SW (SurveyUSA April 15)
Obviously, with Philly/SE being Obama's strongest area, these regional demographics are a big deal. Suffolk has him up 57-40 in the SE region (higher than any other polling firm other than PPP- we'll get there next), but a huge disadvantage in Pittsburgh/SW (63-29 Clinton- also higher than any other pollster) means a large disadvantage in the state.
Let me put this on record, using both other polling data and historical/demographic data- there is no way Obama loses Pittsburgh+SW PA by 34 points (Quinnipiac has Pittsburgh proper at Clinton 50-44 and the rest of Southwest PA at Clinton 68-26; I think that means about a 20-25 point loss in Pitts+SW PA. Let me also add: there is no way that SW PA is equal to SE PA in total number of votes. Philly/SE has far more registered voters and every other pollster has that region as just under half of total voters. This makes sense to me.
In conclusion, Suffolk is way off- maybe not on the final number, but definitely on the way they got there.
The PPP "Outlier"
PPP, while a reputable pollster this primary cycle, seems way off on their poll this time showing Obama up 3. I've heard talk that the poll is an outlier; not exactly correct since an outlier is that 1/20 polls that is just way off, and PPP has been consistently showing Obama stronger than others. Regardless, there is not another pollster within 8 points of this projection.
Strangely though, their data on most of the state is roughly similar to that of other pollsters. The big difference: Philly/SE PA.
In Philly/SE PA, the pollsters show:
Suffolk: Obama 57, Clinton 40
Mason-Dixon: Obama 49, Clinton 44
SurveyUSA: Obama up 14 (not sure on the specs until we get crosstabs)
Quinnipiac: Obama 54-41 in Philly, 50-46 in the rest of SE PA- lets call that Obama 52-43
PPP: Obama 58, Clinton 32
There's your difference. And it sets up a baseline for what would have to happen for Obama to win PA- he would need a 20+ point gap in Southeast PA and Philly, or a 15-20 point gap with a massive turnout to make that area over 50 % of the total vote. Probably not going to happen.
One interesting thing about the PPP poll- it does show undecideds breaking 60-40 Clinton, which is about what I and others expected.
Rasmussen & the A-A vote
One oddity about the Rasmussen poll- it shows Clinton with 21% of the African-American vote. If that happens, I'll never post an analysis again- just not going to happen. Every other pollster has it at 81-84 percent Obama, 10-12 percent Clinton, with about 7 percent undecided. I think it'll end up like 86-14 Obama.
There are some other oddities here and there. Clinton's lead among women varies from 8 points (Rasmussen) to 23 points (SUSA). Obama leads or is tied among men in every poll but Suffolk (I think the regional analysis shows why he's down with men in Suffolk), though it ranges from a tie (Rasmussen) to 21 points up (PPP). White voters are fairly static in the polls (ranging from a 14-21 point Clinton lead). The non-Pittsburgh/Philly parts of the state seem fairly close as well- about a 15 point Clinton lead.
My gutcheck prediction
If Philly/SE comes in at about a 12 point Obama lead (far below PPP, in between M&D, SUSA and Suffolk, and the rest of the state (including Pittsburgh) comes in at about a 15 point Clinton lead (pretty close to what everyone other than Suffolk says), all of which I believe:
Clinton 51, Obama 49
That's my pick, and I'm sticking to it until misled by exit polls tomorrow.
Update by kos: Sorry to intrude into this post, but SUSA has released its crosstabs, and their regional breakdown is:
43% Philly/SE -- 41C, 55O
24% Pittsburgh/SW -- 58C, 36O
4% Northwest -- 61C, 36O
7% West Center -- 59C, 21O
10% South Center -- 49C, 43O
11% Northeast -- 60C, 37O