Matador
Well-known member
So back to the original post: I had some more time to read through the presented paper, and it really is an interesting approach.
It starts off with some definitions of mathematical fairness, which interesting enough is *not* a partitioning of a plurality of districts so that both Democrats and Republicans are evenly represented. Instead they define a set of all possible partitions of a district that conform to certain rules: among them 'compactness', 'continuity', and 'equal population'. The set of possible partitions of the districts of the average state is staggering: the average number calculated was over 1080, which is commensurate with the number of atoms in the observable universe (!).
Instead of equal representation, the authors propose using a supercomputer cluster to test a large number of random samplings of a set of districts within a state, and calculate what the vote shares would be using those partitions. The average number tested was around 100 million, and the idea is that they want to converge on the average number of representatives split between the two parties. The current NC split (based on the GOP drawn map) is 3 democratic districts, and 10 Republican, despite the fact that on a state-wide basis, the vote share of Republicans is approximately 53% versus 47% for the Democrats (this is estimated, since the 9th district didn't have votes counted).
What they found was that of the 100 million tested district partitions, 2 standard deviations of seat shares yielded a mean 61% district share for the GOP and 39% for the Democrats, meaning that of the 13 districts in NC, in the average case 8 should be GOP seats and 5 should be Democratic, with a certainty of about 90%. Obviously this is far from the current situation. The average number of wasted votes was less than 4%, which matches the efficiency gap calculations proposed before the Supreme Court several years ago.
The really interesting result is that some argue that the seat share should be proportional to the vote share, which in NC is roughly evenly split. However due to the population and their proximity to the big cities (and the relatively sparse population areas between Ashville and Charlotte ), the only way to get equal district splits between the GOP and the Dems is to violate compactness rules in favor of the Dems (essentially, it would be to partially gerrymander in favor of the Dems). Hence an equal 7-6 split really isn't possible in NC, however the most probable fair split is 8 GOP and 5 Dem.
It starts off with some definitions of mathematical fairness, which interesting enough is *not* a partitioning of a plurality of districts so that both Democrats and Republicans are evenly represented. Instead they define a set of all possible partitions of a district that conform to certain rules: among them 'compactness', 'continuity', and 'equal population'. The set of possible partitions of the districts of the average state is staggering: the average number calculated was over 1080, which is commensurate with the number of atoms in the observable universe (!).
Instead of equal representation, the authors propose using a supercomputer cluster to test a large number of random samplings of a set of districts within a state, and calculate what the vote shares would be using those partitions. The average number tested was around 100 million, and the idea is that they want to converge on the average number of representatives split between the two parties. The current NC split (based on the GOP drawn map) is 3 democratic districts, and 10 Republican, despite the fact that on a state-wide basis, the vote share of Republicans is approximately 53% versus 47% for the Democrats (this is estimated, since the 9th district didn't have votes counted).
What they found was that of the 100 million tested district partitions, 2 standard deviations of seat shares yielded a mean 61% district share for the GOP and 39% for the Democrats, meaning that of the 13 districts in NC, in the average case 8 should be GOP seats and 5 should be Democratic, with a certainty of about 90%. Obviously this is far from the current situation. The average number of wasted votes was less than 4%, which matches the efficiency gap calculations proposed before the Supreme Court several years ago.
The really interesting result is that some argue that the seat share should be proportional to the vote share, which in NC is roughly evenly split. However due to the population and their proximity to the big cities (and the relatively sparse population areas between Ashville and Charlotte ), the only way to get equal district splits between the GOP and the Dems is to violate compactness rules in favor of the Dems (essentially, it would be to partially gerrymander in favor of the Dems). Hence an equal 7-6 split really isn't possible in NC, however the most probable fair split is 8 GOP and 5 Dem.
