As predicted, Obama got drubbed in West Virginia, so it's time to update my series on the predictors of his state-level support. If we put the exit poll data in context, we can see that he did much worse among whites in West Virginia than we would expect based in states with similar black populations:
However, he did about as poorly as we would expect among whites given the proportion of state residents with college degrees, though West Virginia has fewer college graduates than any previous primary state in the sample:
[Both graphs include linear fits with 95% confidence intervals that are predicted based on previous primary results.]
Update 5/15 10:07 AM: Josh Marshall argues Obama's struggles with downscale white voters are the result of the cultural and racial history of Appalachia, the location of most counties that he's lost by wide margins (in purple):
I've been telling you for months that this analysis is ahistorical nonsense. Everybody knew that Clinton would win big in WV based on previous primary results. If I didn't point that out here, I certainly did elsewhere. % of black people in a state is not driving white voting behavior. Correlation is not causation here. Kentucky will be the same story - 7% African-American, but Clinton will win way more than the 58% or so of the white vote predicted by this model. And again, based on previous primary results and minimal knowledge of political geography, this is easy to predict.
% of whites with college degrees, on the other hand, does seem to have some explanatory power, though things are complicated by states where Obama is probably a better cultural fit also being better educated. This makes sense in light of exit polls.
Posted by: ikl | May 15, 2008 at 12:48 AM