After my post yesterday on proving partisan versus anti-competitive gerrymandering, some readers asked for examples of these theories diverging. In yesterday’s Texas case study, of course, the theories converged. Texas’s 2022 congressional plan was both a partisan and an anti-competitive gerrymander. As shown below, the plan had a more pro-Republican efficiency gap than all 5000 maps in the ALARM ensemble. The plan also had a larger median margin of victory than essentially all those maps (all but four).

But there are many instances in which these analyses tell different stories. In fact, if one imagines a 2 × 2 matrix of partisan effect (yes or no) versus anti-competitive effect (yes or no), congressional plans this cycle can be found in all four quadrants. Consider the category of partisan effect but no anti-competitive effect. As depicted below, New Mexico’s plan has a more pro-Democratic efficiency gap than almost all (around 99%) of the 5000 maps in the ALARM ensemble. Entirely unlike Texas’s 2022 plan, however, New Mexico’s plan has a smaller median margin of victory than the vast majority (roughly 95%) of those maps. Its median margin is about 8% while the ensemble’s median is close to 17%—more than twice the size. This is the signature of a partisan but pro-competitive plan.

What about the opposite scenario—anti-competitive effect but no partisan effect? Per the below chart, Georgia’s congressional plan has a larger median margin of victory than the vast majority (about 98%) of the 5000 maps in the ALARM ensemble. But unlike Texas’s 2022 plan and New Mexico’s plan, Georgia’s plan is quite fair in partisan terms. Its efficiency gap is less than a point away from the ensemble’s median—and on the Democratic side of that midpoint despite the plan’s Republican authorship. These are the traits of an anti-competitive gerrymander that doesn’t skew in either party’s favor.

This leaves the category of no partisan effect and no anti-competitive effect. The below chart indicates that Connecticut’s congressional plan sits in this quadrant. Its efficiency gap is almost identical to the median of the distribution for the ALARM map ensemble (within a quarter point). Its median margin of victory is also somewhat lower than the ensemble’s median (by nearly two points), denoting a modest pro-competitive effect. Connecticut’s plan was designed by Nate Persily and illustrates how both forms of political gerrymandering can be avoided by a good-faith line-drawer.

Returning to Wisconsin, its congressional plan is alleged to be both a partisan and an anti-competitive gerrymander. The above discussion demonstrates that these claims may, but need not, both be true. Partisan and anti-competitive effect sometimes travel together—but other times go their separate ways.