Andrew Sullivan points out an excellent regional analysis of political and social trends. The authors have divided the country into ten regions at the county level, using thirty-year voting patterns exclusively. They provide sociological and ethnic breakdowns based on those regions, rather than vice-versa. It results in some pretty strange choices, which I'm not too sure about - the northern tier of Eastern Virginian counties are placed in "Southern Lowlands" rather than "Northeast Corridor", and the Mississippi delta counties fall into "Appalachia" instead of the rest of the Delta, (Missouri, Tennessee, Arkansas, and northwards), which were shoved into "Big River". I'll have to ask my Delta people whether they agree with that - I had always thought that the Arkansas and Mississippi sides of the Delta were politically indistinguishable, but I could be wrong. It's a meaty analysis, with a lot of detail and political recommendations for both Republicans and Democrats. It reminds me of Garreau's Nine Nations of North America, except with political instead of sociological emphasis.

Among the other points that the authors make is that if Bush holds the same regions at the same percentages as last time, he'll have a greater majority - he took four of the five regions which have gained most from the last census. One point is that the "Southern Lowlands" and "Big River" regions are the swing votes, with the other eight being mostly committed to one camp or the other - "Upper Coasts" (AKA New England and the northern Pacific Coast), "Great Lakes", "Northeast Corridor" and "El Norte" (AKA the Hispanic southwestern fringe) going Democratic, "Southern Comfort" (AKA Gulf Coast), "Sagebrush", "Appalachia" and the "Farm Belt" going Republican.