This work, which is posted on the arxiv here, summarizes the work which began with the 2016 Data+ team and was built on by the Quantifying Gerrymandering group. It serves as the basis for Jonathan Mattingly’s testimony in Common Cause v. Rucho.
This work was updated in this subsequent article.
The evolution of the project continued with this second Data+ group. Because of our interaction and collaboration with the Beyond Gerrymandering project, we again concentrated on North Carolina.
A new code base in Java was written. The version was tested carefully for reliability and correctness to a level beyond what had been done before.
Because of its increased efficiency longer runs could now be used.
For the first time in the work the county splitting and not conforming to the VRA were penalized in the score function used to construct the measure. A number of different states were considered in this work.
For this iteration of the project, a new code base was written in Julia. Unfortunately some unidentifiable memory leaks (maybe in the language itself) limited the length of the runs posible.
Some middle bugs in this version of the code were corrected in the next version of the project initiated in the Data+ 2016 project. That work is the basis for Redistricting : Drawing line .
This work posted on the ArXiv here, summarize the initial work done by Christy Vaughn Graves and Jonathan Mattingly between Summer 2013 and Fall 2014. It uses the same basic methodology as the subsequent papers: generate an ensemble of redistricting plans, analysis their properties, and then compare an existing plan of interest to the ensemble of maps.
The central idea being that maps which were outliers in the ensemble were should be seen as unrepresentative.
This analysis did not include the VRA or county splitting terms in the score function as latter works did. The work did look at the typical most african american district in each plan after the plan had been drawn.
We also did not have access to the true length of the boundary between various VTDs. Instead we used an approximation which assumed each district was circular and that the boundary was equally shared between neighbors. This work also used a much smaller number of maps. There were also a small bug in the code which was fixed in subsequent summers.