[EM] IRV vs. Condorcet, Approval

[Kevin Venzke]

James,

 --- James Green-Armytage <jarmyta at antioch-college.edu> a écrit : 
> 	As far as the technical side goes, I don't find the IRV / Plurality
> question to be especially interesting either, because I think that it's
> obvious that IRV is better.

Hmm.  Let me put it this way: If I could put any electoral system in place
instantaneously and for free, I would probably implement IRV before Plurality.
But since implementation is not free, and I find IRV's superiority to be so
slight, my mind shuts down and declines to admit its superiority.

> >There is a possibility that low-utility CW's wouldn't win in Approval. 
> >I'm not
> >sure if that is "normative."  (Perhaps someone knows of more reasons.)
>[...]
> 	As far as that goes, your point fits into the category that I was looking
> for. Although I can't say that I find it very convincing. I think that a
> Condorcet winner is a Condorcet winner, and should be elected. 

I'm not going to argue that a low-utility CW shouldn't win.  I'm on the fence on
that issue, and I don't think it would be a big issue in real life.

> I doubt
> that there is a method that is strategy-free enough for public elections
> that will tell us how much utility a candidate has.

I assume you mean "viability" (or "odds"), not "utility."   But Approval doesn't
make any effort to guess the odds for the voter.  It's supposed that the voter
will guess what his own expectation (in terms of utility) when deciding whom to
approve.

> Also I see no
> guarantee that approval will exclude a low utility candidate.

There is no guarantee, but it's more likely than the opposite happening:

> I imagine that you have in mind an example something like this, right?
> 48: A>>>>>B>C
> 3: B>A=C
> 49: C>>>>>B>A
> 
> Despite all of those little carrots, I can't really see how it's justified
> for anyone other than B to win here, assuming that majority rule is what
> you're after. If you give it to A, the plurality winner and IRV winner,
> then all the people who voted A>>>>>B>C will cry hey! And they will want
> to change their vote to A>B>C, or B>A>C, or whatever, depending on what
> weird method you are using that doesn't pick B. And if they are allowed to
> change their vote, then B will win, sure as pudding. If they can't, then
> they will regret their votes.

There are a couple of issues here.  First, if you think the CW should win no
matter what, then there are no axiological grounds on which to prefer Approval.

I assume by "weird method" you mean voter strategy, since this is plain Approval.
But there is not really much debate about which strategies are best.

If the voters have the same information, then although they may regret the way
they voted, they played with the same game as everyone else.  In the scenario 
above, both A and C's supporters prefer to try to elect their candidate than
to "play it safe" and elect the "merely bad" candidate, B.  If either faction
"plays it safe," they elect the CW no matter what the other faction chooses.

So the idea is that Approval may ignore a CW when no one (or not enough people)
feels it's worth it to "play it safe" and compromise on him.

> 
> >> 	However, I am admittedly not totally convinced that Approval is
> >superior
> >> to IRV. (It may well be, but nobody has convinced me of it yet.)

I think you should find it superior to IRV because effective strategy is
simpler.

> >I do believe strategy would be more difficult in [equal-rankings IRV].  A 
> >clueless voter
> >(such as one who who prefers to strictly rank) might be significantly
> >disadvantaged.
> 
> 	I'm not entirely sure that I understand what you are referring to, but
> maybe I do. Strategy is more difficult in equal rankings IRV than
> approval? Hmm, maybe so. But I think that there is a lot of strategic
> thinking that goes into approval. 

If you can estimate the odds of each candidate winning, and also measure your
"preference" for each candidate, you're ready to plug the numbers in.  But chances 
are your gut instinct will agree with the result, and chances are that most
candidates won't have any odds at all (and you'll suspect that fact).

"Approve the candidates that you would rather put in office, than have the 
election at all (if you had the option)."

If you think that may be harder than equal-rankings IRV, I will be forced to
write further on this subject.

> 	Anyway, I would like for an approval fan to really lay out why they think
> approval is so effective. I know that there are some very smart people who
> are approval fans, so I imagine that there must be something to it. But at
> this point, I don't know what that something is, exactly.

I don't think there will be many normative reasons here.  Bart Ingles wrote a
message about his sentiments on the issue; that might be worth finding.  I
do think his arguments were more about practical issues, though.

> >
> >What do you think about MCA?  The odds of electing a majority favorite
> >are 
> >increased.
> 
> I'm sorry to admit that I'm not wholly clear on what MCA is.

The ballot has three ranks (the last being the default).  If any candidate is
placed in the first rank by a majority of voters, then the candidate ranked
first by the most voters is elected.  Otherwise, the candidate ranked in the
first or second rank by the most voters is elected.

The voter can put any number of candidates in each rank.  (Maybe someone feels
that only one candidate should be permitted in the first rank, but that
requirement would complicate zero-info strategy, and create favorite-betrayal
incentive.)

This is basically Approval with a chance at catching a majority favorite.
A practical reason for MCA over Condorcet might be that MCA can be easily
hand-counted.

Kevin Venzke
stepjak at yahoo.fr


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