Let's found an organization to oppose IRV

LAYTON Craig Craig.LAYTON at add.nsw.gov.au
Wed Nov 15 22:34:27 PST 2000

Mike Ossipoff wrote:

>Note that the strategy of major parties crowding the centre seems to be
>equally applicable to Condorcet, Borda & Approval, with Demorep's YES/NO
>system being slightly worse, and *IRV and plurality being slightly better*.

>No doubt the less sincere candidates (which will always include
>the Republocrats) will always crowd the center, no matter what the
>method is.
>But if you're saying that's less of a problem with IRV than with
>Condorcet & Approval, it isn't clear why you believe that.

I don't necessarily think that any of it makes a valid argument (for many
reasons I outlined), but if you are to hold, as some of you do, that
Condorcet and Approval are superior because they elect centrist candidates,
then I would employ the argument that if they do elect centrist candidates,
this is not positive.  You yourself have posited this difference between the

The argument would go thus: In Condorcet, a party/candidate has to win as
many head to head contests as possible.  The closer to the centre you are,
the better chance of winning all contests.  Approval's strategy is similar.
In IRV and plurality, you need to maximise your first preference votes (in
IRV's case high order preferences) in order to win.  To do this, established
parties need to pitch at their traditional supporters as well as the centre
(the parliamentary party is inevetably more centrist than the rank and
file), to avoid bleeding votes to parties to the immediate left & right, and
getting eliminated before the last round.

But I'm exaggerating, the difference is very minor between the methods,
which is why I phrased it 'slightly better', and also why I still support

>When candidates crowd the center, or when there are many candidates,
>IRV does worse at picking sincere CWs in Merrill's simulations.

I'm sure it does.  IRV is terrible at picking sincere CWs in many cases.
Basically, my position is that Approval doesn't have any advantages that
Condorcet doesn't have, but IRV does.  It is much worse than Condorcet, and
I'm not advocating it, but it has enough advantages over Approval (and
approval certainly has advantages over IRV) that I can't see that approval
is so much better to warrant advocating its implementation over IRV, and
possibly compromising Condorcet advocacy in the process.

>By the way, I think you're being a bit of a purist, regarding SU.

I guess I am.  There seems to be an accepted application of SU that is
appropriate to use, even though I might have some misgivings about it.

>Maybe the idea is that the individual differences that you're
>concerned about will tend to average out.

I don't really think that they do.  If you're interested, I could think of
some conceptual examples.  I am concerned that the definition of utility is
not clear when it is used.  People seem to be defining it as preference
satisfaction (in the sense of sincere preference satisfaction, not in the
sense of actual preferences on ballots).  If this is true, then that is
fine, but there sometimes seem to be other connotations.

Measuring utility (my definition) is both complex and vague, and I have only
just begun to think about it in relation to voting theory.  If I sincerely
prefer A over all other candidates, it does not follow that A will maximise
my utility over all other candidates.  To even begin to imagine what kind of
predictive ability would be required to make such a connection is
mind-boggling (a bit Hari Seldon).  The proper approach to utility requires
assumptions and generalisations, starting from very simple ones (eg. having
enough food to eat is utility positive for everyone), and using established
generalisations to make more specific ones (eg. from A, B, C, it follows
that although freedom has no desirability in itself, it seems to be utility
positive to have it as a principle on which society should be based).

Voting theory requires a much narrower scope than I've been able to manage
thus far in thinking about utility, which, for many reasons, is much better
as a broadstroke theory than a precise one.

More information about the Election-Methods mailing list