Where the Whiter Folk Are

Today I combined some Census data with 2008 election results (thanks Cosma). Though Barack Obama won the vote last fall, he lost the Non-Hispanic white vote. It stands to reason then that the whiter and less Hispanic a county is, the more likely it would be to tilt McCain. I was curious as to geographic variation within this general rule-of-thumb. So I plotted the % who voted for Obama in a county vs. the % who were Non-Hispanic whites (according to the 2000 Census*). I then generated a line of best fit via loess, and used the deviations from the trend to generate a map shaded proportionately. In other words, the bluer a county is the more it voted for Obama above expectation based on the overall relationship of the % white Non-Hispanic within a county and vote for Obama (the converse for red naturally). Again, click the map for the larger image.

I also decided to constrain the data set to those counties which were at minimum 80% Non-Hispanic white. Mostly because the “Black Belt” counties are showing up on the above map.

Finally, a shaded map of the results.

1) Don’t mess with Texas Obama. Despite Obama winning the Hispanic counties in the South, those counties are always underperforming relative to other majority-minority districts. Since some of those counties are 90% Hispanic it probably isn’t just Non-Hispanic white swing in the other direction.

2) Though Obama lost much of the rural North and East, he overperformed when you use the whole nation as a reference point, in particular rural areas in the West and South.

3) Pennsylvania kind of does look like Pittsburgh + Philadelphia, with Alabama in the middle.

4) Obama did well in Greater New England. Less well in the Butternut Region of the southern Midwest settled from the South.

Here are a list of the 200 counties furthest from the trendlines. The first 100 are the most pro-Obama above expectation when the predictor is Non-Hispanic white %. The last 100 the most pro-McCain.

* This is 8 years out of date, but by far the most complete and precise data set until the 2010 Census.