[EM] Methods of elimination in quota preferential STV

[Craig Carey]

Using quotas to increase the loss of voters power in transfers away
from eliminated candidates ought allow a reduction in the loss of power
when transferring votes away from winners.


Some rationales/motives for eliminating with quotas:

* Putting quotas for losers into STV can allows the method to be more
   compliant with monotonicity/P1. Monotonicity is imposed and leads to
   a minimal change then quotas result.

* Using quotas to reject losers would reduce the 100% transfer values
   that STV will use when shifting votes away from candidates being
   eliminated. That allows:

     (a) votes held with eliminated losers to have less influence. This
         great power allocated to votes leaving eliminated losers seems to
         get complained about. It looks like it got complained about just
         below: "eliminated candidates' preferences ... the result seems
         to become increasingly arbitrary". I recently showed in my mailing
         list that for the papers (a:AB, b:B, c:C), the transfer value T,
         for the losing paper can be T=1 when a=1/4, but it is T=0 when
         a=1/3.

    and also,
     (b) votes transferred away from winners, can have more power. The
         design of STV may be an impediment that prevents people from
         figuring out how to transfer from winning parcels with less loss
         of power.

         The transfer values need to be kept under 1 or else my P4 rule
         is not satisfied. P4 fails the Approval Vote.

         This reduced loss would not be evident from a considering of the
         1 winner 3 candidate Alternative Vote (there is no 2nd winner).




At 15:03 04.10.00 +1000 Wednesday, LAYTON Craig wrote:
 >Yep, but I was considering the basis for excluding candidates in the first
 >place.  If, for a moment, you accept the system of electing by quotas and
 >distributing surplus votes at fractional values (an assumption I might be
 >less inclined to make after I've looked more closely at Craig Carey's
 >arguments), the basis for eliminating candidates when you get to a round of
 >voting in which there is no candidate with a quota is still something that
 >doesn't seem very clear.  In fact, once the eliminated candidates'
 >preferences start affecting the count, the result seems to become
 >increasingly arbitrary.
 >
 >>Hi Layton,
((Mr Layton, or is there a missing comma?.))

 >>
 >>where Nc is the number of candidates remaining that have not been elected
 >>or excluded, one can conduct a "sub" STV election of Nc-1 candidates. The
 >>candidate left over can be excluded.

David's method could be worse than STV. No tests are around to say whether
the method was improved....