[EM] Approval vs. IRV

[James Green-Armytage]

Hi James, and welcome to the list.
>
>I am convinced of the technical superiority of Condorcet over other
>methods. 

	I agree with you there. But keep in mind that "Condorcet" is not a single
voting method but rather refers to any Condorcet-efficient voting method,
the variety of which is potentially infinite. Basically, I support the
beatpath, ranked pairs, and river methods... but I would like to combine
them with a cardinal component, which I won't go into here...

> However, the lack of any real world implementations to point
>to, and the difficulty of explaining the tie-breaker 

	I prefer not to call majority rule cycles "ties". They are logically very
distinct concepts, and calling them ties gives the impression that the
method becomes arbitrary when they occur, which is not true.  

>So I'm down to looking at IRV vs. Approval (Approval being completely
>trivial to explain).

	Okay, but keep in mind that it is possible to do a version of IRV which
allows for equal rankings. Does the wording of your ballot initiative make
it clear that equal rankings are not allowed? Because if there is an
opportunity to do so, I strongly urge you to keep that option open. We
have sometimes called this "equal rankings IRV" (ER-IRV), and there are
two possible ways to count it. If I rank three candidates as tied for
first, then we could give all three candidates 1 point in the first round,
or we could give them all 1/3 in the first round, and bring the fraction
up to 1/2 and 1 as the other two are eliminated. These are the "whole" and
"fractional" versions of ER-IRV.
	ER-IRV(whole) is more of an IRV-approval "hybrid", and does a better job
of reducing the incentive for the compromising-reversal strategy, but
ER-IRV(fractional) is probably more acceptable to the general public, and
less likely to anger people who are very strict in their definition of
"one person, one vote". 
	I'd lean toward saying that ER-IRV(whole) is the better method of the
two, but I expect that ER-IRV(fractional) is a safer bet politically. I
see no excuse for disallowing equal rankings in IRV, when
ER-IRV(fractional) has clear advantages with no drawbacks. The advantage
is that it transforms a lot of the compromising-reversal incentive into
compromising-compression (changing a strict preference into an equality,
rather than reversing it), which will have a less-distorting effect on the
election.
>
>The Center for Voting and Democracy (a group I generally agree with)
>has stated its preference for IRV over Approval.  There are two
>relevant links:

	Yes, most of us are familiar with CVD. I interned there one summer,
actually.
>
>The most compelling argument against IRV in my mind is the empirical
>evidence from Australia.  3rd party candidates are still not viable,
>and voters still vote tactically.  The requirement to rank all the
>candidates also results in some odd side effects (like 'how to vote'
>cards, and the horrific 'donkey vote').

	Yes. Australia's HOR shows us that IRV can maintain stable govenments,
but it does not show us that IRV can help third parties challenge the
status quo.
>
>The most compelling argument against Approval voting from the Science
>mag article is the idea that it will result in non-substantive
>campaigns where candidates try to come across as totally inoffensive
>in order to gain approval from as many voters as possible.  
>
	Actually, the most compelling argument against approval is that it fails
the mutual majority criterion. (If there is a single majority of the
voters who rank every candidate in a set S1 over every candidate outside
S1, then the winner should certainly be a member of S1.) This means that a
candidate can be the first choice of 55%, 60%, even 99% of the voters, and
still lose an approval election. Also, it means that even if a big
majority prefers all of the Democrats to all of the Republicans, a
Republican can still win.
	Of course, this is not really a criticism of approval relative to
plurality, because plurality also fails the mutual majority criterion, but
it is a devastating criticism of approval relative to IRV (which passes
the criterion). This failure also has nasty strategic consequences. See
the link below.
http://fc.antioch.edu/~jarmyta@antioch-college.edu/voting_methods/survey.htm#approval
	I'd say that there is no contest between ER-IRV(whole) and approval; the
former is clearly superior. I would also tend to say that
ER-IRV(fractional) is distinctly better. As for a version of IRV which
doesn't permit equal rankings, I don't know. Probably still a bit better,
but really, I just think it's stupid not to allow equal rankings, so I'd
rather not spend time thinking about it.
>
>It strikes me that this reform will involve a lot of discussions with
>citizens about what "fairness" means in a single-winner election.

	I suggest that the goal of single-winner voting is generally majority
rule. I think that the Condorcet crition (and perhaps generalizations of
the Condorcet criterion, such as the minimal dominant set criterion)
provide the best operational definition of majority rule. IRV's definition
of majority rule is clearly weaker than Condorcet, but a step up from
plurality, etc.
>
my best,
James

P.S. Feel free to visit my voting methods pages at 
http://fc.antioch.edu/~jarmyta@antioch-college.edu/voting.htm