[EM] Tim Hull's PR method

[Warren Smith]

1. Voters vote for up to n candidates - n being either # of open seats or # of candidates
2. Each voter has one vote equally and evenly divided among the candidates they voted for.
3. After doing the first count, eliminate the candidate with the fewest votes.
4. Recount all ballots, dividing votes equally and evenly among *remaining candidates only*
5. Repeat Steps 3 and 4 until there are only as many candidates remaining as there are open seats.
6. The remaining candidates shall be declared elected.

Any comments on this method?
--

WDS: it seems to me this method is not PR in the sense that voters who vote
for single candidate risk having their votes wiped out, and this other problem:

EXAMPLE:
there are 10 Dems & 10 Repubs running for 10 seats.
The voters are 51% Dem and 49% Repub.
Each Repub voter votes for all 10
Repubs (who thus initially each get 0.1 vote).
Each Dem voter voters 100% for just one Dem - Bill Clinton. Really
the Dem voters feel
   Clinton > all other Dems > all Repubs,
but this voting scheme does not allow them to express that feeling, so they
give it all to Clinton.

Result: 9 Dems eliminated, then 1 repub eliminated, then
the 10 winners are Clinton + 9 repubs.

The problem here is that the Dem voters are "pickier" than the Repub voters. 
A lot of PR-design-attempts run onto this kind of reef.

Warren D Smith
http://rangevoting.org