[EM] Juho's example

[Juho]

On Feb 24, 2007, at 0:46 , Michael Ossipoff wrote:

> The first things that stands out about Juho’s example is the fact  
> that there is no majority preferring B to C. So one could ask in  
> what way it is a problem for C to win instead of B.

First question, which one of the examples are you referring to?  
Actually I already gave 4 examples. Since you refer to C being the  
favourite of only one voter it must be the first example (49:A,  
49:BC, 1:CB).

Only the second one (the "DCR" example) was intended to describe a  
real life practical example. The first one is thus quite theoretical  
but I'll comment in that light.

My first thoughts on why electing C could be seen in a negative light  
are as follows. This example maybe comes from a real life situation  
where B and C are from the same party where the party supporters  
indicate with strength 49-1 that B is better than C. The A party  
supporters don't care.

If one looks at the very basics of margins and wv comparisons it is a  
question of whether 49-1 preference is stronger than 49-48 or if they  
should be seen as having same strength. 49 is not a majority but 49-1  
sounds "unanimous" (except that some voters didn't express their  
opinion or said that the candidates are equal) and 49-48 sounds "tie".

> C is the favorite of only one voter? We don’t claim to be choosing  
> the winner by Plurality. Condorcet chooses compromise winners too,  
> compromise winners who don’t have much 1st place support. Ask any  
> IRVist.

Yes, Condorcet characteristically may elect centrists. I didn't see  
strong "centricism" in C though. A supporters were neutral. B and C  
supporters support each others. Either B or C could be claimed to be  
a centrist. Out of them B has clearly more support.

> B is CW? Why? With the help of that  insignificant number who  
> consider C the best. They (s/he)  made B the CW, and it isn’t  
> unfair if they can unmake that CW-ship if they want to. So not only  
> is B not majority-supported against C, but B isn’t CW independent  
> of the C voter_.

It seems to me that B and C are clones that form one party that has  
majority (narrow one but majority anyway). That majority also makes B  
a Condorcet winner.

I don't think small minorities should have the right to make  
themselves the winners as they please. If C supporters wanted to give  
the victory to A that would be ok. But being able to turn the victory  
to C doesn't look pretty.

> No one here advocates Plurality as the voting system. When you look  
> at that winner in terms of Plurality, it looks awful. But, more  
> fairly, let’s estimate SU, by the sincere Borda scores. Then you’ll  
> find that C has more than half as much estimated SU as B has.

I believe you didn't claim that C would the best using this SU  
estimation method.

> Now, I must admit that I have no idea why those B voters ranked C  
> in 2nd place. But, if we may make a fair guess, we can guess that  
> they (somehow?!) felt that they might need C as compromise, in case  
> B doesn’t win. In other words, they were ok with C winning instead  
> of B. C winning instead of B is not unheard of or unacceptable to  
> those voters. In fact it was not even unplanned or unintended by them.

The votes were sincere, i.e. intended to reflect the sincere  
preferences of the voters. The natural explanation to me is (as in my  
second example) that B and C are from the same party. C is just less  
popular than B there.

B supporters thus sincerely felt that C is better than A. And C  
supporters felt B>A. The problem with winning votes, as I see it, is  
that they ignore the fact that among the otherwise symmetric B and C  
candidates B was clearly more popular (among the voters that  
expressed their opinion about them).

> Along with gross misjudgement about compromise-need on the part of  
> the B  voters, C won by betrayal,. Betrayal of the B voters who  
> were trying to help C. As I always say at this point, I hope that  
> makes you proud of yourself.
>
> Had the B voters, more realistically, not ranked C, that outcome  
> couldn’t have happened.

Ok, now we get to the differences between margins and winning votes.  
Winning votes may in this example require strategic truncation. With  
margins there is no need to do that. As I have said earlier I think  
Condorcet methods lose much of their benefits when we enter from the  
domain of sincere voting to the domain of strategic voting. At least  
in this example winning votes seem to fail earlier than margins.

> To summarize, I don’t find a problem in Juho’s example. If it’s the  
> best bad -example that can be found for wv in comparison to  
> margins, then it amounts to a testimonial and a tribute to wv, and  
> is accepted as such. <smiley>

Do you say that (with winning votes and sincere ballots 49:A, 49:BC,  
1:CB and strategy where 1:CB changes to 1:CA)
1) It was ok that C won (=a relatively good candidate)
2) The ability to change the outcome to C with one strategic vote was ok
3) Requiring B supporters to apply strategic defensive methods is ok
4) Winning votes performed better than margins with these votes

I had some problems with each one of those.

In your earlier mail you promised to post some of your examples that  
demonstrate why margins would be worse than winning votes. Please do  
so. Preferably with real life like ballots rather than theoretical  
ones. I'd like to defend a bit too in addition to attacking :-). In  
the light of this example (and the others that I mailed) I still must  
think that margins outperform winning votes. (although you claimed,  
with smiley, that this amounted to a testimonial and tribute tho wv :-)

Juho

>
> Mike Ossipoff
>
>
> ----
> election-methods mailing list - see http://electorama.com/em for  
> list info


	
	
		
___________________________________________________________ 
All new Yahoo! Mail "The new Interface is stunning in its simplicity and ease of use." - PC Magazine 
http://uk.docs.yahoo.com/nowyoucan.html