I was reading http://bolson.org/dist/
about fair redistricting. He suggests using the distance to the centre of the district as the parameter to optimise. A computer algorithm is then run to try to minimise this parameter while keeping all the districts approx equal in size. I wonder if an additional constraint might be to try to balance the vote in each district so as to reduce the number of safe seats. It would work something like: The votes for each party is examined from the previous election. Are totals released on a polling booth by polling booth basis? In not, then it would have to be based on a survey or something. This gives a geographic spread of support for each of the parties. The constraints that are used for re-districting are then 1) The districts have reasonably similar populations (say +/- 1%) 2) The average distance between each person and the centre of their district is no more than double the minimum determined when ignoring 3) and 4) 3) Consistant with 1) and 2), pick a district configuration that minimises the square of the estimated difference between the top two parties in the district One possible problem with this is that it basically hands all districts to the larger party in the previous election. If a party gets 55% of the vote, then a configuration where they have 55% in every district minimises 3). So, adding 4 4) Randomly order the districts. When calculating 3), the weighting given to each district shall be X% smaller than the weighting given to the district that occurs immediately before it in the ranking. (X could be some value, say 33%) This means that the districts near the start of the list shall be very close to 50/50 while the districts near the end of the list will be more unbalanced/safe seats. The net result is that there would be some safe seats and some not so safe seats. However, they would be selected at random. ---- election-methods mailing list - see http://electorama.com/em for list info
