[EM] 2-rank and N-rank Condorcet

[Kevin Venzke]

One of the advantages of Approval (or "two-rank Condorcet") is that it
asks the voter to indicate which contests they would prefer to be involved
in.  (They do this by dividing the candidates into two ranks, and only
supporting candidates of one rank against those of the other.)

The idea is that contests be decided by voters who care the most about
them.  For instance, if the Condorcet ("N-rank") rankings would be:

46: A>B>C
10: B>C>A
44: C>B>A

B would win, and there might be concern about that.  But we could avoid
the problem, possibly, with two ranks.  The results might be more like:

46: A>BC
10: BC>A or B>CA
44: C>BA

A or C would win, because most voters would rather express Favorite>B
than B>LeastFavorite.

With two ranks, placing the single "cutoff" has been called "agonizing."
So what if we added one more rank and cutoff?  I suspect that would
produce a good mix of Approval and Condorcet's advantages...  Specifically,
it seems that a compromise could be elected more easily than under
Approval, while a rogue/turkey would have a harder time being elected
than under Condorcet (at least to an extent).  It would also permit
greater expression and less agony than Approval.

I wrote a program where I could in theory test these notions, but I
have some trouble in that I'm not sure what the best zero-info strategy
would be for three-rank Condorcet.  Perhaps you could place the cutoffs
in the largest gaps, or perhaps the candidates should be "plotted" on
three different "points," aiming to minimize the inaccuracy of the
placement relative to the actual distances from the given voter.

And obviously, the more ranks you have, the less information you obtain
about preference priorities.


Kevin Venzke
stepjak at yahoo.fr


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