Except that hand and machine counts are not sampling the same populations within the state. You would expect left-leaning local governments to resist Diebold machines and Republican-controlled local governments to be early adopters. So there you have a highly nonhomogeneous socioeconomic variable that varies across the state and 1) is plausibly a major predictor of the machine/hand distribution; 2) being a political variable, is plausibly correlated with the Clinton/Obama swing.

We have met the enemy, and he is us — Pogo