Is it all demographics?

As some of my previous posts about the Indiana and Pennsylvania primaries suggested, a lot of voter behavior in primaries is easily predictable from previous primaries. If you know the demographic composition of a state and how those demographic groups have been voting in previous primaries, there just aren't that many surprises.

Now Josh Marshall has posted an excellent short essay showing how Clinton's blowout victory in West Virginia tonight -- along with her big wins in western Pennsylvania and eastern Ohio -- can be explained by Appalachia. This region, extending from upstate New York to northern Alabama, is largely populated by the descendants of Scots-Irish immigrants who were as vehemently anti-slave as they were anti-slavery. Today, the Appalachian region tends to be poorer, older, less educated, and whiter than surrounding areas, and for demographic and cultural reasons is highly likely to prefer Clinton to Obama.

I was discussing this with a colleague, who pointed out to me that the Scots-Irish have spread out from Appalachia over time into places like the Ozarks of Missouri and southeastern Colorado. Check out the caucus night results in Colorado:Although Obama won the state 2-1, Clinton won in the southeastern counties, where the Scots-Irish settled.

This strikes me as an under-reported phenomenon in elections. The media focus a lot on momentum and other campaign effects. After Pennsylvania, Obama was expected to not do well in Indiana because Clinton had momentum and because Obama had a spate of tough press associated with Jeremiah Wright. When he did better than expected, it was assumed he did so because he's "tough" or had somehow blunted the attacks. Or just maybe it was because Indiana is not part of Appalachia.

Just how much of these state-to-state variations (or, really, county-to-county variations) can be explained by the ethnic compositions of the local populations? If we measure this stuff right, is there any sort of campaign effect left?