The strong Democratic showing compared with Mr. Chen’s simulations doesn’t necessarily indicate that the map is a Democratic gerrymander. For one, the simulations aren’t perfect. And they aren’t necessarily representative of realistic partisan-blind maps. To take a concrete example: The simulations often split the city of Pittsburgh, something few human map-drawers would choose to do given the requirement to avoid unnecessarily splitting municipalities.
Perhaps more important, the remedial map still slightly favors the Republicans with respect to the statewide popular vote.
In the average 2016 contest on the new map, Democrats would have carried an average of 8.4 districts (out of 18), even though Democrats won the statewide popular vote in the average contest. The median congressional district favored the Republicans by a point in the average 2016 contest.
Over all, the new court-ordered map comes very close to achieving partisan symmetry in an evenly divided state.
The seeming contradiction between the analysis based on partisan symmetry and one based on simulated nonpartisan congressional districts gets at the heart of what may be the next big debate in gerrymandering: whether nonpartisan maps should strive for partisan symmetry, or whether they should try to avoid political considerations altogether.
The question is important because both methods of analysis are routinely employed to identify Republican gerrymanders.