[EM] Automatic gerrymandering

[Michael A. Rouse]

I generally look at gerrymandering as a problem that is to be avoided at 
all costs, but I thought I would have a little fun and come up with an 
"automatic gerrymandering" method that might yield bizarre 
representative districts but would still be relatively fair. Here's what 
I came up with:

1. Count the total votes cast in an election and use the Droop Quota to 
calculate the minimum number of voters to be represented by a single 
legislator.
2. Determine the winners of each precinct.
3. If a contiguous group of precincts have the same winner, and the 
number of voters in the precincts is equal to or greater than the 
cutoff, the winning candidate becomes a representative.
4. If  a winning candidate has "surplus precincts" above the cutoff , 
the surplus becomes available for the closest candidate on down in each 
precinct as long as they do not eliminate an existing district by 
changing the winner, making the district non-contiguous, or dropping the 
number of voters below the Droop Quota. In other words, if a candidate 
won 100 precincts of equal size but only needed 90 to win, 10 precincts 
are available for redistribution to the second-place candidates. 
Unallocated precincts then become available to the third-place 
candidates, and so on.
5. If a group of unallocated precincts  is surrounded by one or more 
completed districts, then they are added to the neighboring completed 
district with the highest-scoring candidate as long as they do not alter 
his election. His surplus precincts then become available to other 
candidates (you are basically adding precincts on one side and making 
them available on the other).
6. If there are still representatives needing to be picked, unallocated 
precincts are combined into "super-precincts" until there is a winner. 
(I'm not sure what method of combination would be best, since there 
might be multiple ways of adding precincts together which would yield 
different representatives).

Of course, the downside to this would be the amount of calculation 
necessary to determine an election -- candidates who win outright are 
easy to pick, but choosing a winner among the other candidates is much 
more difficult -- but at least with this version of gerrymandering it's 
the voters in each precinct that makes the decision, not the 
politicians. You might also want to make a rule that unallocated 
precincts are first given to candidates who have won a sufficient number 
of non-contiguous precincts, and/or candidates with the highest total 
number of votes overall.

I'd be interested in seeing other ideas or refinements that others might 
come up with, To be honest I still prefer anti-gerrymandering districts 
(a vector quantization/Voronoi cell method would be great), but this 
method might yield interesting results in states like California and 
Texas. :)

Michael Rouse
mrouse1 at mrouse.com