[EM] [ApprovalVoting] Self-Gerrymandering, Fairness, Extremism, & Electability

Kristofer Munsterhjelm km_elmet at t-online.de
Wed Jan 27 13:48:10 PST 2016


On 01/20/2016 11:12 PM, Jeffry R. Fisher wrote:
> On 01/15/2016 03:31 AM, Walabio at MacOSX.COM [ApprovalVoting] wrote:
>>
>> On Jan 15, 2016, at 9:46 AM, Kristofer Munsterhjelm
>> <km_elmet at t-online.de> wrote:
>>
>>> Suppose you have a districting system that is supposed to follow the
>>> borders of natural communities, and that is supposed to be fair. Now
>>> suppose that cities vote predominantly for X, and rural areas
>>> predominantly for Y. Then there's no way of squaring the circle: if you
>>> use a method that is fair, the method will clip through communities
>>> because the communities are so lumpy. On the other hand, if you use a
>>> method that follows communities, e.g. the borders make sense from the
>>> point of view of a group, then you get safe seats. At least one has to
>>> go.
> 
> Good analysis. That's why proxies (transferable votes) are attractive:
> Voters are freed from geographic districts. Isn't geography an archaic
> basis for voting in the Internet age anyway?

If we take as an ideal of democracy that every person should have a say
in a decision proportional to how he would be affected by it, then there
would still be a place for geography because local events have a greater
impact on the people who live there than do global events, all else equal.

If you start at the extreme of geographical representation by using
single-winner districts, you can probably get a significant improvement
in political representation by sacrificing only a small amount of
geographical quality, e.g. by going from single-winner to multiwinner
districts. A similar observation applies to party list: biproportional
representation can improve national political proportionality at only a
slight cost of local political proportionality.

At some point, the trade-off is no longer worth it. I have no idea
exactly how large the ideal district would be, though; it probably
depends on factors like economies of scale and the preferences of the
people.

In the absence of such knowledge, it probably makes sense to consider
location to be another property that should be given some degree of
proportional representation. If political PR is of the shape that "if
10% of the people prefer this group of candidates to everybody else,
then 10% of the candidates should be elected", then geographical PR
would be similar to "if 10% of the voters live in the vicinity of a
certain city, then 10% of the representatives should live in the
vicinity of that city too". The trade-off is a matter of how strict the
proportionality on either should be, when tightening the proportionality
on one will degrade the proportionality on the other.


More information about the Election-Methods mailing list