[EM] Converting the converted: IRV to Condorcet

[Rob Lanphier]

One can have all sorts of fun searching groups.google.com for 
Condorcet.  I found a thread that had veered seriously off-topic on 
alt.philosophy and into the world of voting systems.  The thread started 
here:
http://groups.google.com/groups?selm=29b75758.0304100452.7080fc8%40posting.google.com
(with "Once Instant Runoff Voting (aka Preference Voting) is implemented 
in the US, I'll start voting again.")

...and ended here:
http://groups.google.com/groups?selm=29b75758.0304141120.c17c37d%40posting.google.com
(with "I'm still very much opposed to treating Condorcet as if it were a 
"nearly-perfect" system [...] but it does seem to address the particular 
problem I'm most interested in (spoilers), better than IRV.")

Janet Anderson, who is a particularly vocal IRV advocate, posited that 
perhaps IRV is a good stepping stone to Condorcet.  She hadn't made the 
leap to Condorcet (she has been an advocate of IRV so long that old 
habits die hard), but based on my personal experience, and based on 
anecdotal evidence that I continue to see, I think she may be right.  
Those that would like to see Condorcet implemented may be able to draft 
off of the efforts of the IRV advocates, and as such, we should avoid 
getting too inflamatory with them.

Rob
----
From: Bill Lewis Clark (wclark at eden.rutgers.edu)
Subject: Re: Condorcet vs. IRV (Re: Liberals vs. Conservatives-- general 
questions)
Newsgroups: alt.philosophy
Date: 2003-04-14 12:20:42 PST

robla at robla.net (Rob Lanphier) wrote in message 
news:<bf570e3.0304131155.63a69277 at posting.google.com>...

 > "Condorcet" can be used to describe a method, since Condorcet himself
 > outlined a tiebreaker which is pretty well.

Okay, I'll buy that.

 > Arrow's thereom is an important result, but it leads to a rather
 > stupid fatalism about fixing voting systems.  I'm not sure why you
 > bring it up here...usually, people bring it up when they are arguing
 > that we should just stick with plurality.

My point was simply that there can be no *single* "best" voting
system, so independently of particular criteria, it makes only limited
sense to compare voting systems.  Some systems (like plurality) suffer
from a multitude of problems and fairness criteria violations, and are
rather clearly inferior -- but past a certain point, all voting
systems are pretty much equipotent.

Before reading some of your references, I'd thought IRV to be in that
upper-most class of equipotent voting systems.  However, many
authorities seem to consider it one of the more obviously flawed
systems.  I'll have to do some more of my own research and thinking on
this issue, before I separate the facts from the opinion.

However, my main underlying point still stands -- there are other
systems that are just as good as Condorcet, whether or not IRV happens
to be one of them.

 > Regardless, here's a response to Arrow's thereom:
 > http://www.electionmethods.org/Arrow.html

It's well-written, but I still don't buy it.  It seems rather
disingenuous to argue that the criterion that should be relaxed just
so happens to be the one that Condorcet violates.  Appeals to
intuitions regarding fairness are pointless here, since they just beg
the question.

The claim that "cyclic ambiguities are true ambiguities in the will of
the electorate, and an election method can hardly be faulted for
accurately reflecting them and attempting to resolve them rationally"
is just laughable.

Proponents of *any* voting system could make a similar claim, by
arguing that whatever ambiguities their system exhibited were "true
ambiguities."  One man's "true ambiguities" are another's technical
flaws.

 > IIAC is basically Arrow's statement of the spoiler problem.

Now that's something new to me, and does more to convince me that I
should be supporting Condorcet's over IRV, than any of the other
arguments.  I never realized that IIAC captured the essence of the
spoiler problem so well, but re-reading it now it's very clear to me
that you're right, so thank you for pointing that out to me.

 > IRV elections only help minor parties *when they stay minor*.

That's just fine with me.  I'm against the party system in general, so
anything that dilutes their power is okay in my book.

 > 48% vote for 1. Bush
 > 5% vote for 1. Gore
 > 20% vote for 1. Gore, 2. Nader
 > 27% vote for 1. Nader, 2. Gore
 > Under IRV, Gore would have been eliminated in the first round with
 > only 25% of the vote (20%+5%).  Nader would have then only had 47% of
 > the vote vs. Bush's 48%.

Nice example, but I'm not sure what to make of it.  Granted, on the
one hand Nader spoiled the election for Gore (since Gore beats Bush in
the cycle.)  On the other hand, Nader has stronger support than Gore.

My only real concern with the spoiler problem is that it allows the
dominant parties to "swallow" smaller parties.  Greens become
Democrats, because they're made to feel guilty about costing Gore the
election and thereby putting Bush into office.

However, in this example it's not clear *what* will result.  The
strong support for Nader makes it unlikely that the Democrats will
simply absorb the Greens.  More likely, I'd guess we'd see a party
split within the Democrats, with the smaller group of hard-core Gore
supporters (the 5% group) splitting off from the Democratic voters
more sympathetic to Nader.

That type of party fragmentation is precisely what I'd like, so the
"flaw" in IRV in this case isn't something I'm necessarily interested
in seeing fixed.

 > I started off as an IRV supporter many years ago, so I understand the
 > initial appeal of the system.  Intuitively, it feels like it *should*
 > fix the spoiler problem.  However, after spending some time thinking
 > through spolier scenarios (trying to defend IRV from the criticisms of
 > a Condorcet supporter), I had to  acquiesce that the spoiler problem
 > is better fixed by Condorcet.

Indeed, with a few minor points I'd like to give more thought to (such
as that highlighted in the above example,) you've basically convinced
me, as well.  I'm still very much opposed to treating Condorcet as if
it were a "nearly-perfect" system, or to relaxing IIAC simply for the
sake of making Condorcet pass -- but it does seem to address the
particular problem I'm most interested in (spoilers), better than IRV.

Cheers,

-Bill Lewis Clark