If you look at a map of the Democratic voter registration surge (by zip code) in the five county Philadelphia region, an important pattern emerges. It will help if you know the region and the city. In Philly, registration is through the roof in University City and the rest of West Philly. This is Obama’s strongest area (District 2, Chaka Fattah) in the state. Registration is also extremely high in the North (the blackest area of Philly) and in the Center City, Society Hill, Queens Village area (District 2, Bob Brady) that is socially liberal and filled with urban professionals and artists (the Creative Class). By contrast, registration is low in the Northeast, Fishtown, and deep South Philly (white ethnics, most opposed to Obama).
The same pattern holds in Bucks County, where heavily black Morrisville saw a huge spike, in Montgomery County, where heavily black Norristown and Conshohocken saw a huge spike, in Delaware County, where heavily black Chester and Darby saw a huge spike, and Chester Country, where heavily black Coatsville and West Chester saw a huge spike. Overall, the heaviest registration has come from college towns/areas, black neighborhoods, and the Creative Class neighborhoods of Philadelphia.
This portends a very high differential turnout of Obama voters to Clinton voters in the five county area. It also portends an historically unprecedented differential turnout of the black/creative class vote over the white ethnic vote.
I’m not a polling expert, but I do know that pollsters are not free just to pull turnout models out of thin air. If the percentage of the Democratic vote that was black in 2004-2006 was say 17%, the pollsters are not going to postulate that it will be 25% this year. Likewise, if the Philly region represented say 33% of the statewide vote in recent election, then the pollsters will not stray too far from that precedent.
If you are polling the whole state using just 700-1000 voters, you have to get the regional turnout numbers right, you have to get the differential race, age, and gender numbers right. If you don’t, your poll is going to be off.
On the eve of the South Carolina primary the polls predicted the following result:
Obama: 43.1%
Clinton: 28.5%
Edwards: 17.0%
The actual result?
Obama: 55%
Clinton: 27%
Edwards: 18.0%
Two out of three ain’t bad. They nailed the vote for Clinton and Edwards, but they dramatically underestimated Obama’s vote. Looking at the voter registration numbers in the Philly area, I think there is a real chance that the pollsters are about to be proven wrong about this race.
Here is one sign that backs up my suspicions.
A recent Keystone Poll of Pennsylvania voters, conducted by Franklin and Marshall College, found that 62 percent of Democrats who registered statewide within the last three months planned to vote for Obama.
They might be weighting their polls using anachronistic turnout models. If so, look for Obama to dramatically outperform the polls, just as he did in South Carolina.
