Ending Political Gerrymandering with HPC

New project offers a means for determining when a legislative map is unfair.

Researchers Wendy K. Tam Cho and Yan Y. Liu have pioneered a new technique for using high-performance computers to evaluate whether a legislative map shows signs of gerrymandering. Gerrymandering is the practice of drawing legislative district lines in a way that favors a particular candidate or party. US courts have expressed displeasure with the practice, but courts have traditionally lacked an objective standard for determining whether the lines are reasonable or unfair.

The only way to evaluate whether a map is as fair as possible is to compare it with other possible maps of the district, and the number of possible alternatives is theoretically infinite. Cho and Liu used a mathematical model to reduce the number of alternative maps to a finite number then put the NCSA Blue Waters supercompter to work on drawing the large sample of maps needed for a meaningful comparison. According to Cho, “We drew 800 million maps … no one has drawn more than 10,000, and usually the number is about 1000.”

The proposed solution won first place at the Common Cause 2016 First Amendment Gerrymandering Standard Writing Competition.