This is worth pulling up and chewing on, I think. Brad writes in the comments:
"I think it's simply an anomaly: when we throw race, education, age, income, gender, and religion into the mix on the right-hand side, we find that we have accounted for the overwhelming bulk of political preferences. That's how Nate Silver was able to do so well in forecasting the 2008 Democratic primary.
But that doesn't work for southern whites: there is something else going on there..."
This is a fair point. Thinking about what shapes the error structure of our models is important. So let's do the same with race. Put region in the right hand side and take race out of it.
We've explained a lot of political preference. But the model doesn't work for blacks, let's say. They are more likely to vote Obama than predicted. In fact let's forget Obama. They are more likely to vote for Democrats than predicted.
I don't have data on me to compare magnitudes, but this is certainly the case regardless. That's why Brad listed it first. It has explanatory power.
And all he's saying is that the South has explanatory power. Sure.
So? What is it that that explanatory power (because it's not really an "anomaly" - it's a correlation. It's not noise - it's information) says to us?
I'm just suggesting it says a lot less salacious things than people are implying. Southerners have for centuries not liked as active or as centralized government. This is not exactly news. After Washington took his tour of New England in 1789 he took a tour of the Southern states. He complained vigorously about how Southerners didn't seem interested in getting together and collectively providing for infrastructure the way New England did. There's always been a conservative, small government streak.
That seems to me to account for the regional explanatory power. You can throw in some more about how it was not exactly a magnet for European immigrants a hundred years ago so new fangled leftist ideas didn't get ingrained in the same way.
In any case it's a twenty point spread. It has explanatory power but I think we can just say "ya - Southerners are more conservative" without alluding to any racial undertones. Because that is where a lot of people are taking it (maybe not Brad) - that this says something substantial about race relations in the South.
If we want to talk in anecdotes there is plenty that the South has to offer in terms of bad race relations. This election cycle, twitter and Facebook have served a lot up on a silver platter. But if we want to explain the twenty point spread, I think it's irresponsible and misleading to turn to that. I know a lot of white Southerners that vote Republican that are not motivated by that. Not all white Republicans are like them. But my strong suspicion is that the vast majority are. There are a lot of omitted variables feeding into regional effects. The people who claim that the racial dog whistle of the Republican party is a substantial one are, in my opinion, quite wrong and don't know much about Southerners.
But it's an omitted variable, so it's tough to arbitrate this.