…the counties where Republican margins grew the largest tended to be predominantly white places in otherwise racially mixed states.
That is political scientist Eric Oliver, in this post at the Freakonomics blog. He is complicating the story of this much-discussed map in the New York Times, which showed the counties that swung toward McCain relative to 2004:
The story, he argues, is not about the composition of these countries, so much as the location of these counties:
Racially isolated whites in Arkansas or Alabama may have been more afraid of voting for Obama not because they are more racist than white voters in Minnesota or Montana, but because they perceive greater racial competition with nearby black populations.
This account dovetails with some recent analysis from Joost van Beek. To get around the ecological fallacy of using county-level attributes to infer individual-level behavior (see Eric’s post for more on this), he looks at the state-level exit polls. Exit polls are not perfect, as he notes—and they force the analyst to look at states rather than countries—but they do have the advantage of providing individual-level data.
van Beek’s map of the white vote in 2008 vs. 2004 looks like this:
The “bigot belt” disappears, being replaced simply with big shifts in a couple Southern states—namely, those states where white populations coexist with large black populations. Charles Franklin’s graph shows something similar:
Oliver’s concluding paragraph is a fitting summation, which also fits with some of my previous posts on the election, which urged against overinterpreting the election’s outcome:
Nevertheless, these results should dispel the idea that with the election of Obama, America has somehow “transcended” race. Undoubtedly, racism is still pervasive in the United States, but where it appears depends a lot on social context.
{ 9 comments }
I’m not sure I buy the premise that it’s no longer bigotry if it’s motivated differently, but going along with that for a second … wouldn’t this rationale predict similar dynamics inside cities with high populations of African-Americans (were such exit polling data available), which is to say, including large, northern cities?
Do I have that right? And does anyone think we would see that, if the data were available?
No Joel, it’s a different context. Whites who coexist with blacks and other ethnic minorities in large urban areas tend to be liberal. If you look at van Beek’s map, you’ll see states have only a few large urban areas – and in those areas, whites broke for Obama. In the rural counties in those states, whites broke overwhelmingly for McCain: 75% >.
It seems like the areas with more blacks, the whites voted for McCain more. Am I reading this right?
It seems like both Franklin’s and van Beek’s analyses simply replace one ecological fallacy with another. Either way you’re talking about correlations between aggregates. The question then becomes: is it fair to assume that all white people in KY and TN are the same?
The county-level map is fascinating, but the ecological fallacy is a problem. Why are heavily white counties in the Tennessee mountains, for instance, voting more pro-GOP this year compared to ’04, while heavily white counties in the NC mountains are voting more pro-Dem in ’04? Also, many of the same rural counties in Tennessee shifting more toward McCain voted (in a majority) for a black senate candidate in 2006.
I glanced at Oliver’s post. I don’t esp. like his use of “racial competition” to denote “perceived competition between blacks and whites for jobs/public goods (under certain specified geographical etc. conditions).” This is at least partly a matter of perceptions, isn’t it? The claim is that the white voters in question *think* they are competing for public goods with blacks, not necessarily that this perception is *correct*. Using “racial competition” as a shorthand for this is, it sees to me, a bit misleading. (It may be standard in the literature, but still…)
Chris, the “heavily white counties in the NC mountains” actually form the only part of the state where a fair number of counties did not vote more pro-Dem than in ’04. Six counties, including the three western-most counties of the state, moved further Republican in the presidential race.
Dividing the state into six constituent parts, the Mountains area turned out to be most resistant to Obama. While Piedmont, the Metrolina area surrounding Charlotte and the Research Triangle showed swings to Obama of 14% and more, the Mountains region swung by a more modest 7.6% — and it would have been 5-6% if not for Buncombe county (Asheville). See: How did North Carolina end up the ultimate toss-up state? Reviewing county data.
I think the fact that Obama did do that well still in the mountains of NC has a lot to do with the intense level of organisation and campaigning in the state, something that was not remotely paralleled across the state border in TN.
But what does the fact mean that despite that unprecedented push, the mountains turned out to be more resistant than the lower-lying areas further east? It’s tempting to speculate about Appalachian specificities. But there’s also just the point that those are the whitest counties in the state. Any move of rural white voters toward McCain, which may well have occurred elsewhere too, was less canceled out there than it was further east by the mobilisation of new black voters by Obama.
To clarify that last bit: I wasn’t saying that rural whites swung to McCain on balance in NC. Just that to the extent that some of them did, and rural whites consequently swung to Obama less than the state did overall, this would naturally show up most clearly in the Appalachian counties anyway – simply because those are the whitest counties of the state.
The depressed swing to Obama in the mountain counties thus does not necessarily need to point to some cultural, Appalachian specificity. Rural whites elsewhere in the state may have swung to Obama in equally more hesitant fashion, and we wouldn’t know it because such a pattern would be canceled out there by the mobilisation of extra black votes.
In short, even if NC was much more Obama-friendly than the states to its South and West, the regional differences within the state are the same as those across the South, with a McCain belt appearing in the Appalachian parts. But the question what that means remains the same too: does that point to some cultural specificity of the mountain counties? Or does it demonstrate a larger pattern among rural Southern whites, which is merely cloaked elsewhere by differing racial demographics?
Oops – looks like the first of the two posts I made yesterday didn’t get through. Hence the mysterious opening sentence starting with “To clarify that last bit”.
In that first post I was pointing out to Chris that actually, the “heavily white counties in the NC mountains” were the least likely in the state to “vote more pro-Dem in ‘04″. Six counties there actually moved toward McCain, including the three western-most counties of the state.
Overall, the mountains area of NC swung to Obama by “just” 7.6%; the least of all seven regions of NC, as the Piedmont Triad, Research Triangle and Metrolina all swung by 14% or more. And without Buncombe county (Asheville), it would have been more like 5%. See here: How did North Carolina end up the ultimate toss-up state? Reviewing county data.
I also said that when looking at the difference between the swing on the NC and the TN side of the state border, a big role must have been played by the intensity of the Obama’s campaign and ground organisation in NC.
And a bunch of other stuff, but I dont remember anymore.
Comments on this entry are closed.