Cross-posted at Election Inspection
Yep, you read that title right, Survey USA, probably the best pollster this election season, is probably understating Obama's support in the state of North Carolina. I've been looking at some polling which they've done in southern states, and there is a real trend which has emerged, SUSA has fairly dramatically understated Obama's support there. If you don't believe me, take a look at this stuff:
Poll released 2/2 (Alabama) SUSA: Obama 49 Clinton 47
Actual result: Obama 56 Clinton 42 (Obama outperformed SUSA's expectations by 12 points)
Poll released 1/25 (South Carolina) SUSA: Obama 43, Clinton 30, Edwards 24
Actual result: Obama 55, Clinton 27, Edwards 17 (Obama outperformed SUSA's expectations by 15 points)
SUSA also polled Virginia, and while they were easily the closest to the mark, they still underestimated Obama's support there by 7 points.
It might seem a bit unfair for me to use South Carolina (considering that no one got it right) but in SUSA's case, they weren't really wrong because of their polling method, they were wrong because they screwed up the demographics. SUSA projected that African Americans would make up only 41% of the vote in South Carolina, but in reality, they made up 55% of the vote. The same thing happened with Alabama, SUSA projected that African Americans would make up only 42% of the vote, they made up 51% of the vote.
SUSA wasn't the only pollster to make this mistake, but there were pollsters who did get a closer demographic projection. In South Carolina, for example, PPP was probably the only pollster to correctly guess that African Americans would make up over 50% of the vote in South Carolina. This probably shouldn't come as a surprise, considering that they are based out of North Carolina.
I've been working on a voter turnout model to try and figure out what the turnout will be for North Carolina, specifically what the racial demographics will be. Working out a turnout model, I've determined what percentage of Kerry's support was black and what support was white (for example, in Mississippi, over 75% of Kerry's vote consisted of African Americans, while only 25% consisted of whites). On average, African Americans in southern primaries consisted of 81% of Kerry's African American support in the state, while whites have made up 132% of Kerry's white support in the state. Knowing this, the most likely turnout model will be African Americans making up 37%-42% of the primary, while whites will likely make up 45%-50% of the vote (the remaining will be Latinos, Asians, etc.) SUSA shows African Americans making up only 30% of the vote, while whites making up a massive 65% of the vote. PPP, on the other hand, shows African Americans making up 37% of the vote, while whites make up 58% (which, while still a little off for me, are a lot closer to my turnout model)
For this reason, I would probably say that, despite PPP's lackluster Pennsylvania polling, they are probably the pollster to look at for a better picture than SUSA. This is not to say that SUSA isn't a worthy poll to look at, and they will be especially valuable in Indiana, but with that in mind I don't believe they are accurately portraying the North Carolina race.