[EM] Election districts (was bicameral design poll)

[Sampo Syreeni]

On 2004-01-06, James Gilmour uttered:

>And we here in the UK already know that virtual districts are not worth a
>second thought, at least, not here. [...] Locality and community are very
>important.

I agree, but for a different reason. If we impose democratic rule on a
group of people, we're essentially socialising some set of decisions.
When we pick an election method, we're essentially *defining* "the will of
the people" for this group.

Now, if we only analyse the election method in isolation, we're missing
any game theoretical phenomena which vary based on the underlying demos.
Those phenomena are usually the real worry of proportionality and
representation enthusiasts. For instance, geographic representation thru
districting is essentially an attempt to fix the perverse incentives
created when the demos includes people from far away which have no stake
in local matters.

No districting rule or election method can ever solve all such conflicts
of interest simultaneously, because the latter are several from the
mechanics of the election. Rather the problem is caused by the fact that
the demos was chosen badly -- it fails to include people who are affected
by its decisions, includes people who have no incentive to vote
rationally, or is simply so torn apart by individual interests that no
election method can possibly unearth a collective consensus. Or it might
not even be possible to structure the political surroundings of the
election such that people have an incentive to vote sincerely.

In groups like these, no definition of the "will of the people" works, so
all collective choice can actually fail to unearth people's preferences.
(E.g. if people have reasons separate from the election method to
consistently vote insincerely. Secret ballots are an attempt to alleviate
one such problem.) At the very worst all election methods might deliver
sane, persuasive, profoundly irrational collective preferences. (E.g. the
tragedy of the commons: everybody votes sincerely against the collective
good.) In other words, not all decisions can or should be socialised.

These problems are not really in the realm of social choice theory, but
have to do with wider political science and economics. Talking about them
purely in the context of election methods or the concrete mechanics of
elections just isn't enough. And since the problem isn't in the election
method, trying to fix the method will in this case only make it worse. For
instance, geographical representation will seriously hurt other kinds of
representation (say, along party lines), so it can severely distort
decisions in non-geographically determined questions. In this case the
best fix is clearly to decide local matters locally, not districting.
Even worse, all that virtual districting would do would be to turn the
whole system into a pure plurality one or give a whole new meaning to
gerrymandering, depending on whether people were allowed to choose their
own districts.
-- 
Sampo Syreeni, aka decoy - mailto:decoy at iki.fi, tel:+358-50-5756111
student/math+cs/helsinki university, http://www.iki.fi/~decoy/front
openpgp: 050985C2/025E D175 ABE5 027C 9494 EEB0 E090 8BA9 0509 85C2