[EM] Gerrymandering and competitive districts

[Raph Frank]

Was doing a google search on "redistricting competitive districts" and
hit this book

http://www.routledgepolitics.com/books/Redistricting-and-Representation-isbn9780415964524

They argue that competitive districts are actually a bad thing.

>From reading the summary their theory seems to be that in non
competitive districts, a greater proportion of the voters will have
voted for their Representative, so happiness is improved.

This seems the poltical equivalent of Keynes' economic theory, it
gives politicians a plausible reason to do what they want to do
anyway.  The question is, are homogeneous or heterogeneous districts
better.

With plurality, the answer is almost certainly No.

However, I wonder if all districts were arranged so that 80% of the
voters in each district supported one or other party, would that party
run more than 1 candidate, or would you still have a situation where
the party only runs 1 candidate?

With plurality, if a district has more than 2/3 of the population from
the same party, then a party can run 2 candidates in complete safety,
as if each party supporter votes for one or other of them, their
preferred candidate will get more than 1/3.  The non-party members
still get to influence which of the 2 party members wins.  It is kinda
like top-2 runoff without the first round.

Likewise, under Range/Scorevoting, the party could run lots of
candidates and the voters would be pretty sure that no matter how they
voted, someone from their party would win.  The reduces the Burr
dilemma (if it is a problem at all).

The main point being that with a competitive election method which has
no spoiler effect, voters can remove incumbents no matter how the
districts are gerrymandered.  This may dull the incentive for
politicians to gerrymander.  At the moment, there is both a personal
benefit and a party benefit.

If heterogeneous districts were acceptable, the method could assign
the number of districts to each party based on a party popularilty
vote.  For example,

- break State into small election areas
- each voter ranks the parties
- least popular party is excluded until all parties have won a
majority in each areas to make up at least 1 seat
-- votes are gives to highest ranked party
- assign seats/districts to remaing parties using d'Hondt
- create election areas so as to maximise support of the party in its
districts according to party votes
-- Other rules would also apply

This splits the districts between the parties using PR and results in
heterogeneous districts.  Tiny parties which don't have concentrated
support wouldn't be assigned any districts though.  This is true for
districting methods based on geography.