I am still playing with the data from the WSJ in my previous post. Letting D2008 and D2004 be the gap between the democrat and republician, and pctinc and pcthouse be the percentage change in income and house prices (all of this by state) I get no significance for pcthouse if I control for D2004 and I do get significance if I only control for whether the state was a bluestate.
The residual maps are really suggestive and I think this makes a great class problem to solve. Start with the data given in the WSJ, then have the students collect the actual D2008 (the number in the WSJ is based on final polls). Then have the students build a model and collect the state by state data that would be better used than those I used.