[EM] A better Bucklin flavored FBC satisfying method

[Kevin Venzke]

Forest,

--- "Simmons, Forest " <simmonfo at up.edu> a écrit :
> Put some ideas from MDDA, Bucklin, and MMPO together, and what do you get?
>  
> I'll let Ted Stern name it, but here it is:

I would come up with a name for "collapse ranks until at least one candidate
lacks a majority defeat," say "X," and then call the method "X,MMPO."
  
> Collapse the top ranks as in Bucklin until (according to the collapsed ballots) there is a
> candidate undefeated by a majority.

This means that A=B=C>D>E doesn't become A=B=C=D>E until the fourth round, right?

> If, at that point, there is more than one candidate undefeated by a majority, elect the
> candidate whose maximum pairwise opposition is minimal.

Interesting. I do think that satisfies FBC.
  
> Note that in ordinary Bucklin the ordinal information below the eventual stopping level is
> wasted.  It doesn't matter if four levels or fifty levels are allowed, only the information in
> the top two or three levels ever gets used, since almost all voters will put one of the two
> front runners somewhere in the top two levels.
>  
> This new method remedies that.  Although it collapses ordinal information in the top few ranks
> in order to find a level at which not all candidates are majority defeated, it doesn't waste the
> remaining ordinal information.

However, you get an opposite problem: Information regarding "irrelevant"
candidates (those with majority defeats) can affect who wins. This is why I
don't feel that MAMPO ("Majority Approval, Minimum Pairwise Opposition") is
clearly better than MDDA. (Both satisfy FBC, SFC, and SDSC.)

Also, the MMPO component is huge. MMPO's failure example for the plurality criterion
still works in your method:

1000 A
1 A=C
1 B=C
1000 B

C wins.

Kevin Venzke



	

	
		
___________________________________________________________________________ 
Appel audio GRATUIT partout dans le monde avec le nouveau Yahoo! Messenger 
Téléchargez cette version sur http://fr.messenger.yahoo.com