[EM] summary answers

James Green-Armytage jarmyta at antioch-college.edu
Mon Apr 4 02:33:40 PDT 2005


James G-A replying to Curt Siffert...
>
>I am curious about the answers to the following questions regarding 
>Condorcet.  Forgive me if these are simplistic, I am kind of looking 
>for a summary of what the consensus is here:

	I don't know if you can get a consensus, but you can certainly have my
opinion.
>
>1) Are there cases where you would consider a candidate outside the 
>Schwartz set to be the proper winner?
>2) Are there cases where you would consider a candidate outside the 
>Smith set to be the proper winner?

	For public elections, no and no. Actually, just one no, really, since the
two sets are practically identical given large electorates. In general,
when majority rule is the goal, my answer to your question is "no".
>
>I'm curious about that in terms of criteria/strategy, but I'm also 
>curious about it in a larger way - because I think I remember Mike 
>saying that if we could be assured of sincere ballots, he'd prefer the 
>Borda winner - even if there were a different Condorcet Winner.  

	I rarely deal with the question of guaranteed sincere ballot scenarios.
But even if we could guarantee sincere ballots, I'm sure that we could
come up with a better method than Borda. Maybe something with cardinal
ratings...??
>
>3) In a two candidate race, if 51% mildly preferred A to B, and 49% 
>passionately preferred B to A, who should win?

	If enough of the A voters are willing to abstain out of respect for the B
voters' preference, B. If not, A. I have also suggested a method where
voters have the option voluntarily decrease the weight of a given
preference in response to a situation like this.
	If you're not satisfied with this, then majority rule is not your best
friend. I know that Nicolaus Tideman did a great deal of work on majority
rule (e.g. ranked pairs), but in the process he became aware of its
limitations, and has done other work on secession rights and the
"demand-revealing process", which are ways of de-centralizing majority
rule. I have my doubts about the demand-revealing process, but it's hard
for me to argue with secession rights...
	Anyway, it's all interesting food for thought.
>
>I am also curious how "common" multi-member Schwartz (or Smith) sets 
>are.  

	Nobody really knows. Various people have done statistical models,
computer simulations, examinations of ranked ballots used for other
methods, etc., but there is no single consensus answer. Personally, I
believe that sincere majority rule cycles will be rare (1 election in 10
or 15, perhaps?) when there are less than a dozen or so candidates.
>
>I tend to carry around the definition in my head that Condorcet methods 
>are "perfect" up through the identification of the Schwartz Set.  For 
>public polling purposes, I would probably just communicate a 
>multi-member Schwartz Set as a tie.
>
	Certainly not a tie, but an ambiguous result, as far as the choice
between Schwartz set members is concerned. 

my best,
James




More information about the Election-Methods mailing list