New SW method: Weighting with elimination & renormalization

[Steve Eppley]

The following method was proposed by Mike York, a subscriber of a 
maillist which is focussed on developing online voting software.

Mike York claimed it doesn't have the lesser-of-evils dilemma and
isn't subject to tactical manipulation.  I suspect he's wrong on
both claims, but present his method here anyway.

  The ballots 
  -----------
  Each voter rates each candidate using any whole number (0 to N)
  that the voter wishes.  

  Mike York didn't specify this, but it's probably a good idea to
  treat unrated candidates as if they were rated 0.

  The tally
  ---------
  While there is more than one candidate remaining
     Normalize each ballot by dividing each voter's ratings by the 
        maximum rating on his/her ballot.
     For each candidate, sum his/her/its normalized ratings.
     Eliminate the candidate with the smallest sum, and eliminate 
        from each ballot the voter's rating of this candidate. 
  Endwhile

Note that the normalization step is inside the while loop.  If an
eliminated candidate was the top-rated candidate on some ballot,
then that ballot's other ratings will renormalize so the second-
from-the-top gets a full-strength vote.

Example ballot:
  Nader    500
  Clinton    2
  Dole       0

The first time this ballot is normalized, each rating is divided by 
500 and it becomes
  Nader      1
  Clinton    0.004
  Dole       0

Suppose Nader is the first candidate eliminated.  Then this ballot
becomes:
  Clinton    2
  Dole       0
and is renormalized (by dividing each rating by 2) to:
  Clinton    1
  Dole       0

So in this scenario where Nader is eliminated first, the ballot
becomes a full-strength vote for Clinton over Dole. 

--Steve