[EM] A pairwise elimination satisfying SFC,SDSC

[Blake Cretney]

On Fri, 6 Apr 2001 15:39:52 -0700 (PDT)
Forest Simmons <fsimmons at pcc.edu> wrote:

> What we need is a method that will appeal to IRV psychology and
still not
> haphazardly eliminate excellent candidates in the early stages of
the
> simulated runoff.

I wonder if that is the best approach.  For one thing, the method that
best appeals to IRV psychology is undoubtedly IRV.  If you want to
convince IRV advocates, you have to modify their psychology.  That is,
you have to show that some of their assumptions are wrong.  But if you
succeed in convincing them of that, it isn't clear that they will
still want a method as close to IRV as possible.  

Of course, I might be wrong, but it seems that these compromise
methods are always proposed by Condorcet advocates as a way of
placating IRV advocates.  I don't see anyone arguing that they
personally want a method as close to IRV as possible, while still
meeting the Condorcet Criterion.

Some of the reasons that IRV is advocated, which are not simple errors
in reasoning are the following:

1 Less of certain kinds of strategy.  It is quite debatable what kinds
of strategy people are likely to use, so it can be argued that IRV
prevents those kinds of strategies which are most likely.
2 Simplicity.  Just how simple IRV is compared to other methods is
debatable, but most compromise type methods add extra rules to IRV,
and are therefore clearly more complicated.
3 Inertia.  IRV is already used in some countries.  Large groups have
been set up to advocate this reform.  It's hard to change this goal
now.
4 Vote splitting/clone independence.  One of the chief advantages of
IRV is its lack of vote-splitting.  Although some Condorcet criterion
methods, like Ranked Pairs and Schulze, don't have this problem, many
do.

The point is, that the compromise methods all lose the first three
advantages, and most lose the fourth.  So, an IRV advocate is not
going to give up IRV, unless given a good reason.  And once they
understand why IRV should be abandoned, they won't necessarily want a
similar method.

There's perhaps an even more serious problem.  Many IRV advocates have
never considered counting a ranked ballot any other way.  This gives a
sense that IRV is the "right" answer.  Not merely an arbitrary
procedure with some desirable qualities.  Unfortunately, an IRV
advocate who might consider Condorcet's criterion important, is
presented with a seemingly endless number of seemingly arbitrary
completion methods.

Until Ranked Pairs was invented in 1987, all Condorcet criterion
methods violated either monotonicity, clone independence, or weren't
defined for all cases.  So, it's only been fairly recently that we've
had a method that seems plausible (doesn't violate monotonicity), and
doesn't have a particular disadvantage over IRV (lack of independence
of clones).

However, Markus has suggested another method with those qualities, and
people quite frequently suggest methods that don't (like the recently
proposed methods on this list).  The proliferation of methods was
particularly enhanced when some people on this list misinterpreted
Condorcet as specifying "defeat-support" instead of margins for
determining majority strength.  They set about advocating that, which
close to doubled the number of possible methods.

So, to come to my point, although I like discussing the various
suggested Condorcet completion methods, and don't want to discourage
debate, I think that advocating more methods isn't the way to go. 
Instead, it makes sense for people to search for a single standard
Condorcet completion method.  Hopefully, over time, some partial
consensus will develop.

---------------------------------------------------------------------
Blake Cretney   http://www.fortunecity.com/meltingpot/harrow/124/path

Ranked Pairs gives the ranking of the options that always reflects 
the majority preference between any two options, except in order to
reflect majority preferences with greater margins. 
(B. T. Zavist & T. Tideman, "Complete independence  of clones in the 
ranked pairs rule", Social choice and welfare, vol 6, 167-173, 1989)