Julian Dech offered an outstanding analysis in his personal statement, which revealed an in-depth consideration of how to create competitive districts throughout the Commonwealth. His goal was to create a proportional map that reflects the "purple" nature of Pennsylvania's politics. The way that he achieved a competitive balance across the state while limiting county splits was impressive.
The essence of gerrymandering is to skew the distribution of partisan voters among districts. The most commonly noted effect of gerrymandering is to advantage one party over the other, but politicians also draw safe, uncompetitive districts that maintain the partisan balance but are unresponsive to how people vote. Sometimes the goal is the opposite: minimizing the number of safe districts by creating as many competitive districts as possible, so that small voting shifts have large impacts.
My goal was to “gerrymander” a proportional map, where there is an even distribution of districts from solidly conservative to closely competitive to solidly liberal. With such a map, the more votes a party gets overall, the more seats they’ll get, minimizing the efficiency gap.
To meet this goal, I took an aggregate result from all of the statewide races in each precinct, then mapped that as closely as I could to the census tracts used in DistrictBuilder. Pennsylvania’s political geography naturally advantages Republicans because many Democratic voters are already “packed” into liberal strongholds like Philadelphia. So I needed to carve out rural districts with the highest proportion of Republican voters while splitting up places like Philadelphia into multiple districts in order to leave the remaining areas evenly matched.
This map is my second attempt at this method, and this time, I stuck to absolutely minimizing county splits, achieving a mere 15 splits.
Even sticking to a strict split-minimizing rule, I was able to come reasonably close to my goal of a truly proportional map. In keeping areas like the Lehigh and Wyoming Valleys intact, I was unable to get exactly nine districts leaning towards each party. However, by making what would have been a slightly Republican-leaning Harrisburg district more neutral, I was able to come up with four fairly competitive districts. The net result comes close to a party-symmetrical map that would reward the party that got more votes with more seats.
My computed vote margins for each district relative to the state are as follows: