The issue I have with analyses like this is that it's binary in some instances - if a district votes conservative 50.1% vs. 49.9% it's given the same weight as one that votes 99.9% vs 0.01%. The color coding skews the underlying dynamics.