Yesterday I pointed out that in the South Carolina Republican primaries this year where Huckabee did well Romney did badly, and vice versa, on a county by county basis. The tendency also seemed in line with the proportion of evangelicals within a county. I did the same analysis with Michigan....
Correlation between candidate proportion and evangelical proportion by county
As you see the correlations are very low (do the coefficient of determination to get a sense of the variance relationships). The only thing that jumps out is that McCain counties tended to not be Romney counties, and vice versa. Probably one can account for this based upon the appeal of Romney's economic message to the eastern part of the state. In any case, South Carolina was quite different. Note:Data file.
Correlation between candidate proportions by county