[EM] More on Gerrymander prevention

Adam Tarr atarr at ecn.purdue.edu
Fri Mar 22 07:55:20 PST 2002

Michael wrote:

>Draw a line between the population centroid and the
>voting centroid (or population median and voting median), and continue until
>you have the number of districts you want. That way, roughly equal voters
>and residents would be in each district.

As you implied before, this could easily end up slicing population centers and 
producing bizzare, sliver-shaped districts.  Also, the population/voter equality 
in this approach is by no means guaranteed... it could in fact end up pretty 
radically skewed just depending on the angle between the two.  Imagine putting 
San Antonio on one side of the line and Houston/Dallas on the other, to pick an 
easy example.

While I stress that I don't think these sorts of simplistic algorithms are 
likely to produce satisfying results, a more reasonable way to partition two 
districts might go something like this:

- start with the geographic centroid of the region.
- draw the minimum-distance line containing this point that bisects the region.
- without changing the angle, shift this line towards the side with greater 
population until the sides are equal in population.

Jurij, I for one would be interested in seeing some details on your automatic 
districting scheme, which I assume is far more sophisticated than this.


More information about the Election-Methods mailing list