How much of Congress’s polarization can be attributed to gerrymandering? To get at this, my collaborators, Keith Poole of the University of Georgia and Howard Rosenthal of New York University, and I decided to try our hand at gerrymandering by using our computers to draw our own districts. For each map, we used the characteristics of the districts — partisanship, average income, racial and ethnic composition — to predict how liberal or conservative their representatives might be. But even when we tried our best to create as many heterogeneous and competitive districts as possible, the predicted level of polarization was only slightly below what we observed in the real Congress. So even if there were a radical transformation of how legislative districts are drawn, the effects on polarization would be minimal.
The above quote is from a recent op-ed in the Washington Post by Monkey Cage occasional contributor and Princeton University political scientist Nolan McCarty. The full piece is available here.