[EM] "Runoff Without Elimination," Condorcet efficiency

[Kevin Venzke]

Back in 1998 or so there was a method known as "Runoff Without Elimination"
or "RWE."  I think it may have been devised by Donald.

I'm not sure if this is a faithful reproduction of the rules, but I have
programmed this:
1. The voter ranks candidates.  (As in Bucklin, voters should decide where to
stop ranking.)
2. Initially count first-place rankings only.
3. While no one has votes from a majority (of ALL voters), AND non-eliminated candidates exist:
   "Eliminate" the non-eliminated candidate with the fewest votes.
   ("Eliminate" has no implied meaning besides what I'll say here.)
   Recount the votes such that every voter approves their highest ranked
non-eliminated candidate, as well as every eliminated candidate preferred to
that non-eliminated candidate.  However, no one approves last-ranked (truncated)
candidates.
4. Elect the candidate with the most votes.

This may not seem interesting, except that it is very similar to IRV (but surely
better), and to "AER" (and doesn't require an approval cutoff).

I did an interesting experiment, I think.  Assuming everyone knows the identity
of the sincere CW before voting, and everyone ranks down to and including the CW,
but no lower, then it seems (based on my simulations) that the CW is always 
elected.

There's the possibility that some factions will truncate the CW to try to beat
him (an experiment I've yet to try), but this finding still seems interesting.


Kevin Venzke
stepjak at yahoo.fr


___________________________________________________________
Do You Yahoo!? -- Une adresse @yahoo.fr gratuite et en français !
Yahoo! Mail : http://fr.mail.yahoo.com