[EM] Theoretical Gerrymandering Solution

[matt at tidalwave.net]

> This sounds somewhat promising. Can you cite any sources on the
> mathematically unsolvable nature of this problem, or expand on that a
> little bit?

This is a technical topic but optimization problems of this type are
computationally NP hard to solve.

Doing this would be radical departure from current practice because even
small shifts in population probably would result in very different
districts.  I don't think drastic changes to the districts every ten years
is bad.  Term limits are popular and in my opinion this a better way to
implement a type of term limit.  It doesn't prevent incumbents from being
candidates, so in that sense it is less strict then term limits, but it
also doesn't assure a seat for a particular party, and in that sense is
more effective than term limits in making elections more competitive.

Also, constraints can be placed on optimization models so that other
goals, in addition to small perimeter sums, such as minority
representation, could be specified, although there is a risk that piling
on constraints could make the problem infeasible.  A better way to assure
racial/ethnic/religious/political minorities can get candidates elected is
to use some type of proportional representation system. For example,
California could use a system like that used in Peoria, Ill. for municipal
elections. Instead of electing 40 state senators from 40 districts, voters
in 10 districts could elect four senators each.  Candidates who win at
least a quarter of the vote would earn a seat (it is more complicated then
this since it will often be the case that four candidates do not each
receive at least a quarter of the votes but that is another topic). These
districts would be far more likely to be bipartisan, even electing some
urban Republicans and rural Democrats.