PA PrimD: Public Policy Polling: Obama Leads PA by 4%

[Verily]

PPP was right on in WI, and wrong nearly every place else.  I'm not counting on it.

Texas: C+6, actual C+4
Ohio: C+9, actual C+10

There's a reason people have said they've been good. They have.

[Verily]

Going back otherwise...

NY: C+12, actual C+17
GA: O+10, actual O+35
TN: C+11, actual C+13
SC: O+20, actual O+29

One problem they did have for some of the Feb 5 states was that Edwards was still included in their polls (in NY and TN). And they were bad at dealing with the black vote in the South, apparently. Then again, all of their Feb 5 polls were conducted almost a week before the election.

[Verily]

Anyway, comparing internals. SUSA in red, PPP in blue

Race
White: 81 | 76
Black: 14 | 18

I'll give this one to PPP; as Alcon pointed out, 22% of Kerry voters in PA in 2004 were black.

Gender
Male: 45 | 42
Female: 55 | 58

I give this one to PPP, reflecting exit polls in other states. But paradoxically this would imply that Obama does better with more women involved, obviously not true.

Age
Unfortunately, they broke down age groups differently, so these will be listed separately.

SUSA
18-34: 21
35-49: 29
50-64: 28
65+: 21

PPP
18-29: 16
30-45: 25
46-65: 35
65+: 24

Ironically, the one area where they broke down the numbers the same way, the 65+ range, PPP had more voters. So, again, Clinton's most favorable demographic was actually larger with PPP than with SUSA!

Region
Northeast: 11 | 9
Southeast: 43 | 45
South: 10 | 10
West: 7 + 4 | 10
Southwest: 24 | 26

SUSA divided West and Northwest in their numbers; PPP did not. PPP is marginally more regionally favorable to Obama, but not in a way that could explain more than a fraction of a percent of difference. As both are so close to each other, I'll call it a tie.

[Verily]

In conclusion to the above, the chief difference between SUSA and PPP appears to be actually finding different results, not using different base demographics.

For example, PPP finds a six-point margin among elderly voters (seven with leaners); SUSA has Clinton ahead by thirty among those older than 65. But such numbers are unreliable as the margin of error is quite large, although the MoE is smaller on PPP's crosstabs because of a very large sample size.

I would be inclined to put my dollar with SUSA simply because they agree with the conventional wisdom, but PPP being accurate would not be a complete surprise. (Of course, they could also split the difference with a virtually tied result.)

[Verily]

PPP actually hasn't had much of a swing in their polls, either, certainly not anything I'd call wild, save for their C+26 poll at the beginning of the PA campaign to O+2 thereafter.

[Verily]

The new ARG?

No, not at all. PPP is still really good, they just had a very bad day. At least, that's all I can explain.