Obama support in NC and IN by race

Time to update my previous work on state-level predictors of Obama support. It turns out that Indiana and North Carolina were strikingly consistent with the trends in racial voting we've seen thus far. Here is a graph of total white and black support for Obama by the state's black population (click for an enlarged version):

The lines are fractional polynomials fit to the data before yesterday's primaries and the shading represents 95% confidence intervals. As you can see, the black vote in Indiana slightly overperformed for Obama but the other results fit almost perfectly with the predicted values.

Update 5/7 3:11 PM: To answer Rob's question in comments, the y-axes represent the share of the total white+black Democratic vote received by Obama from whites and blacks, respectively. For instance, the total white vote (the left y-axis) is calculated as Obama's percentage of the white vote multiplied by the proportion of the white+black electorate that was made up by whites. The total black vote is defined analogously.

The simpler graph below tells a similar story:

Update 5/7 10:38 PM -- Here's an equivalent graph to the top one using just Obama's margin over Hillary in the white vote (corrected after Josh pointed out an error in comments):

Comments

[Rob]

After many re-issues of these charts, it's perhaps a little late in the day for me to be asking this question, but what exactly is this chart showing, i.e., what does "total [white|black] vote for Obama" mean?

[Josh]

I am confused by the third graph. I read your *Description* as meaning that the right y-axis represents the percentage of black voters who supported Obama. But that doesn't seem right, because the graph shows that in most states less than half of black voters supported Obama, which goes against everything I've read.

It could be that the right y-axis is actually the percentage of Obama voters who are black. But that does not seem right either, because the two values for Indiana (~43 % for white, ~15% for black) should come close to summing to 100% (as there are very few Hispanics and Asians in Indiana) but they don't.

Am I missing something?