Predicting Pennsylvania

A while back, I pointed out the curvilinear relationship between the percent of African Americans in a state and the vote for Obama. In short, he seems to do well where there are lots of African Americans or almost none.

This patterns seems to be holding up:
In the graph above, I've only used the vote in the primary states. (Incidentally, Obama has earned 12.4 more percentage points, on average, in the caucus states. Also, caucus states tend to be much whiter -- no caucus state is more than 11.5% black.) Note that Texas and Ohio, which are both roughly 11.5% black, appear at the nadir (pun intended) of the curve. That is, we would have expected Obama to narrowly lose those states if we were just using the African American percentage to predict the vote.

So, just for fun, let's use this curve to predict what will happen in the Pennsylvania primary. African Americans comprise 10 percent of the state's population. If we plot that out on the curve, we would expect Clinton to win there, with Obama receiving around 48% of the vote. So this is a nice little do-campaigns-matter experiment. Let's see if all the sound and fury over the next six weeks amounts to much.