[EM] Re: What respected IRVists are saying about Condorcet, Approval, etc

[James Green-Armytage]

I had some free time this afternoon, and for some reason I decided to
reply to that Jim Lindsay pdf you posted. I live in California, so I might
as well try to get in touch with some of these Cali-IRV people to see what
they're about. 

When exactly did Lindsay post this, and under what subject heading? I'll
reply on that list, when I find out. 

In the meantime, I'll include the text of my reply so far below. Most of
this is old hat for folks here, although I hope that there are one or two
interesting points.

my best,
James
________________


***********************************
“Majority vote wins” in IRV and runoffs: “YES”.
***********************************

	This depends on what you mean by “majority vote”, and whom you consider
to be a majority winner. Consider the following example:

35: Right > Center > Left
33: Left > Center > Right
16: Center > Right > Left
16: Center > Left > Right

	Right wins using IRV; Center wins using any Condorcet method. Is Right a
majority winner? Well, he does have a majority win over Left. However,
does he have a majority win over Center? No. 
	Actually, Center has a 65-35 majority win over Right, and a 67-33
majority win over Center. Hence, I argue that the only real majority
winner is Center. What definition of majority rule allows Right to be
considered the majority winner? I suggest that it is not a sufficiently
strong definition. (The same critique applies to the two round runoff.)

***********************************
 “It is a very strong system if a tiebreaking procedure is used that
discourages bullet voting. (Unfortunately ties would probably be common in
this system, and good Condorcet tiebreaking procedures are very complex.)”
***********************************

	First of all, you seem to be confusing “ties” with “majority rule cycles”
A majority rule cycle (or just a “cycle”) is a situation where A pairwise
beats B, B pairwise beats C, and C pairwise beats A. This is often
mistakenly referred to as a “tie”.
	An appropriate usage for the word “tie” within a pairwise tally procedure
is when there is a pairwise tie. Here is an example of a pairwise tie:

49: Bush > Gore > Nader
49: Gore > Bush > Nader
1: Nader > Bush > Gore
1: Nader > Gore > Bush

	There is a 50-50 pairwise tie between Gore and Bush. Rest assured that
pairwise ties among viable candidates are astronomically improbable in
public elections, just as precise ties in plurality and IRV are
astronomically improbable.
	Now on to majority rule cycles. Here is an example of a majority rule
cycle:

33: A > B > C
32: B > C > A
35: C > A > B

A beats B 68-32. 
B beats C 65-35. 
C beats A 67-33.

	This is a weird result, not because the pairwise voting system is weird,
but because the electorate is weird. If everyone who likes A best likes B
second best, why does everyone who like B best like A the least? And so on.
	If cycles exist in the preferences of the electorate, this is not the
tally method’s fault. Rather, it is the result of a strange asymmetry in
the voters’ preferences among the candidates, or perhaps the result of
strategic voting.
	Anyway, majority rule cycles don’t cause any particular problem for
Condorcet tally methods, and there are plenty of cycle-breaking rules that
are not “very complex”. 
	Here’s one: Knock out the candidate who has the largest majority defeat
against them, and then re-tally. (Repeat until someone is undefeated.) In
our example above, you would eliminate B, who has a 68% defeat. This would
leave you with A and C. C beats A, so C would be the winner. This is not
really much more difficult than IRV.
	Another simple method is to drop the weakest defeat in the cycle (change
it into a pairwise tie) until there is an unbeaten candidate. In our
example above, you would change the 65-35 B over C defeat into a tie,
leaving C undefeated. So C wins. Again, not very complex.
	By the way, if you measure the strength of a defeat by size of the
winning majority, rather than the margin between the winning and losing
sides, this does indeed discourage bullet voting.

***********************************
 “Condorcet tends to elect compromise candidates, so it might be well
utilized for a “healer” type executive position. Let’s say that a group
has had a lot of internal fighting. Electing the President of this group
via Condorcet might be good choice in this situation.
Condorcet would be a perfectly reasonable choice for electing a mayor of a
city, too.”
***********************************

	I agree with you here, and I believe that the US government on all levels
is strongly in need of “healer” executives.
	In that vein, you write that IRV is “good at electing compromise
candidate”. Unfortunately, I think that IRV is significantly less
effective than Condorcet at doing this. My Right/Center/Left example above
seems like a pretty good illustration of this. I suggest that only
Condorcet methods should be fully trusted to elect compromise candidates.
I also suggest once again that we in this nation are sorely in need of
political compromise. The people badly need to establish common ground and
common cause, and I argue that only pairwise tally methods do an excellent
job at revealing majority-supported compromises.

***********************************
“Eliminates spoiler effect” IRV: “YES”. 
***********************************

	How do you define the spoiler effect, exactly? I suggest that, although
it is quite a bit weaker, IRV still retains more of a spoiler effect than
Condorcet methods. Take this example:

45: Bush > Gore > Nader
6: Gore > Bush > Nader
20: Gore > Nader > Bush
29: Nader > Gore > Bush

	Bush wins using IRV. However, if Nader had not run, Gore would have
beaten Bush, 55-45. This situation gives Nader voters an incentive to rank
Gore equal to or above Nader, in order to assure that Nader is eliminated
first. While its short-term effect may be to give the victory to the
Condorcet winner, this kind of strategic voting could prove to be a
significant barrier to third-party candidates actually winning major
public office.
	In Condorcet, Gore wins all of his pairwise comparison, and wins the
election. Hence, there is no incentive against Nader>Gore>Bush voters
ranking Nader over Gore.

***********************************
“Encourages ‘honest’ voting; discourages ‘strategic’ voting” Plurality:
“no”; Approval: “YES”
***********************************

	I don’t understand this. In my opinion, approval requires just about as
much strategy as plurality. Should you approve your top 3 candidates? Your
top 5? Only your favorite? I suggest that strategic concerns will usually
play a role in the answer to this question.

***********************************
 “Requires enthusiastic support to win” Condorcet: “no”; Approval: “YES”
***********************************

	Are you sure that you didn’t get these two mixed up? What makes you think
that Condorcet doesn’t require enthusiastic support to win? What makes you
think that approval does? 
	Approval voters cannot differentiate between candidates whom they
strongly support and candidates whom they weakly support. Condorcet voters
can make that distinction by ranking one above the other. 
	Condorcet will usually pick a compromise candidate. If that candidate has
mostly lukewarm support, that’s too bad, but it’s still better than a
candidate whom a majority is hostile towards. However, I see no reason to
assume that all compromise candidates will necessarily have unenthusiastic
supporters. I suggest that, with the advent of Condorcet, we will see new
majority-supported policies come into the political fore, but were not
enacted previously because the two major parties were too timid/cautious
to incorporate them in their platforms. This will be something to be
enthusiastic about. I suggest that Condorcet elections will include both
rather dull centrists (bland, avoid tough issues, Joe Lieberman, etc.) and
interesting centrists (intelligent people of obvious good will who can
recognize good and sensible solutions to important problems).
	Condorcet minimizes the spoiler effect, and thus allows the most viable
choices of any single winner system, and thus the end result is more
likely to be a winner whom voters are really enthusiastic about. 
	When there are only two viable candidates, there is an inhuman amount of
pressure on those candidates; a single slip-up could put millions of
dollars in campaign funds to waste, and result in the election of one’s
ideological polar opposite. Hence these two candidates are extremely
cautious about speaking honestly or trying new ideas, which seriously
limits voter enthusiasm. This is a big problem in plurality, and I argue
that it would still be a problem under IRV, because most IRV elections are
likely to boil down to two viable candidates in the last round, usually
the two candidates who were identified as viable before the election. 
	In Condorcet, however, the number of viable candidates is not
artificially limited by the tally mechanism; it is only limited by the
availability of campaign resources and the attention span of voters. This
will provide a much stronger opportunity for unfettered, uncorrupted,
creative, inspiring candidates to rise to the top.