Introducing PlanScore — Evaluating New Maps

If the Supreme Court specifies a standard for partisan gerrymandering claims this term, it will become very important to be able to evaluate the partisan implications of new district plans. Whether these new plans are remedial (designed after an existing map has been invalidated) or simply enacted after the next Census, their likely partisan consequences will be relevant to their legality.

Until now, only politicians and their hired experts have had the capacity to properly analyze the partisanship of proposed plans. To conduct such an analysis, it was necessary (1) to compile precinct-level election results, demographic data, and incumbency status; (2) to create a precinct-level model of the legislative vote as a function of the other variables; and (3) to use this model to estimate the likely performance of each district in the new map. This work was too laborious and technical to be done by journalists, legislators, or lawyers—let alone ordinary voters.

PlanScore, the new website launched by Ruth Greenwood, Simon Jackman, Eric McGhee, Mike Migurski, and me, tries to fill this data void. Perhaps the site’s most unique feature is an upload page where a user can upload a district plan—and then immediately receive an analysis of the map’s likely efficiency gap, partisan bias, and mean-median difference. The output page also includes sensitivity testing showing the plan’s probable efficiency gap under different electoral scenarios, a visualization of the plan’s districts, and data about the districts’ racial composition.

To provide this information, we did on the back end all of the work that politicians and their experts perform to analyze a new map. In other words, we compiled all of the precinct-level data, constructed the model, and coded the site so that estimates can be generated instantly for any district configuration that a user submits. The model currently does not incorporate incumbency effects, but we plan to add this capability in the very near future. The upload page also currently supports only Pennsylvania, but again we plan to add more states soon (especially ones that might find themselves in remedial proceedings due to ongoing litigation).

As an example of PlanScore’s utility, consider the remedial plan that was recently offered by Pennsylvania’s Republican legislative leaders after the state’s congressional map was struck down. We uploaded the map on the site, and learned the following about it:

  • Its efficiency gap, partisan bias, and mean-median difference are all likely to be very large and pro-Republican, toward the rightward edge of the historical distribution of all three metrics.
  • If Democrats improve substantially on their 2016 performance, the map will grow ever more skewed in a Republican direction. The Republican advantage under the map, in other words, is highly resistant to pro-Democratic shifts in the vote.
  • Compared to their predecessors, the map’s districts are less irregular and more compact. But this aesthetic progress comes without any accompanying gain in partisan fairness.
  • The map creates one black-majority district and one more district with a large enough black population (42% CVAP) to enable African American voters to elect their preferred candidate.

As the Pennsylvania remedial process continues to unfold, we plan to continue to assess proposed plans on PlanScore. And since the upload page is now functional, we invite you to do the same!


Comments are closed.