[EM]

[LAYTON Craig]

Hugo Harth wrote:

>Open question : Looking for a fair order of n candidates.
>
>Many of the methods in this EM-forum discuss finding a winner out of
>n candidates and as a by-product produce an order of these candidates.
>
>Question : If the focus switches from finding a winner to finding an
>order (and with an intermediate  situation where the order of the
>first candidates is more important), are the best methods for finding a 
>single winner still the best methods or is there better ?
>
>Rationale behind this question :
>Political parties have to propose an ordered list at elections.
>Here in my country [Belgium], candidates in the top places are more likely 
>to be elected because besides votes in favor for candidate(s), votes in 
>favor of a whole list are possible.
>Candidates on the upper part of the list tend to profit more from this.
>Hence, top places are much desired.
>Last few years, the last place (often occupied by a retiring politician)
>is also much desired, people tend to sympathise with the "list-pusher".

Candidates desiring to come last is a curious anomoly.  You would have to
work it out of the equation: the person who comes fourth (or third or sixth
- depending on just how desirable it is) gets the last spot instead of the
fourth spot, and the person who comes fifth gets the fourth spot etc.

>I want therefore to broaden my question.
>Is it possible to design a mechanism or a game that leads to a good
>(that is : generally accepted and peaceable) solution.
>There must be people around who have thought about the same problem.
>
>A few thoughts :
>
>- The committee that sets up the list may contain candidates and 
>non-candidates.
>Would that be a point to grant non-candidates more votes or power?
>It suggests also that such a mechanism should not be easily manipulable.

No.  I think that's a little messy.  If all the candidates are on the
committee, their self-serving votes will cancel each other out.  If only
some of the candidates are on the committee, I think there is a distinct
problem with the committee.

>- One might argue that some candidates deserve a better chance to be
elected
>  (perhaps by giving them more voting power).
>Some reasons :
>= expected good result because of
>  a) good result at previous elections
>  b) well known personality (TV , sports, ...)
>= merit and excellence
>  a) long standing membership
>  b) prestige, "voice that carries weight" , ...

I would think that the merit of the candidates should be decided by the
actual vote, not beforehand.  The committee should be expected to take all
these things into account when voting on the order.  After all, that's what
the committee is for.

This is what I would propose; firstly, hold some sort of proportional
representation method to choose the people who's names will actually appear
on the list.  STV or proportional approval.  If you have definite factions
and many voters you could even do list PR.

Once you have selected the candidates, you should hold a separate ballot for
finding the order.  Simple Approval would probably be the best method for
this.  You simply order the candidates, with the most approved candidate
going on the top of the list etc.

Borda would actually be much better for ordering than for single winner.
With the list of candidates fixed, the worst problem (the teaming problem
Blake identified) is mitigated significantly.  There is also significantly
less incentive for reversing a preference, and as there is no "winner" as
such, the majoritarian criteria doesn't really apply.

Borda could be used on the STV numbered ballots (altered to ignore
eliminated candidates), so that only one vote needed to be held.