I added another plot to the output generated by my overnight ideal point scripts: a scatterplot of estimated Senate ideal points against Obama vote share in their state (color coded by party, local linear regression overlays by party, labels for some big residuals).

I suppose I’m surprised by the way that the loess curve for the Democrats flattens out to the left of 50% Obama vote share. It is an interesting picture to stare at and speculate as to why that is the case: party whipping by the Dems, agenda control (distorting our estimates of the ideal points of Dems in ranks, say, 35-60), lack of fit by the one-dimensional model, some combination of all this, something else…

Outliers: Coburn (R-OK), Demint (R-SC) and Ensign (R-NV) among the Reps; Lieberman (really a Dem?, and a positive residual), and two big negative Dem residuals, Brown (D OH) and Harkin (D IA).