Let's found an organization to oppose IRV

MIKE OSSIPOFF nkklrp at hotmail.com
Mon Nov 13 20:36:15 PST 2000

I'd said:

> >There are certainly a number of ways in which Plurality is better than
> >IRV. Participation, IIAC, Consistency, etc.

Blake replied:

>Saying that approval meets IIAC is meaningless, because the method can't
>represent a full ranking of the candidates.  We therefore don't know how
>the voter would have voted with fewer candidates, even assuming

I don't think IIAC is about how people would have voted in a different
election with different candidates. I've heard people refer to that
kind of an IIAC, but I've never heard a precise definition of it.
Nor have I heard a claim that any method could pass such a criterion.

Can you tell me a definition of an IIAC that is about sincere preferences 
and 2 ballotings, with & without a candidate? And then
can you tell me a method that passes it? It seems to me that no
method could pass such a criterion. Being, by itself, unmeetable,
would make Arrow's other criteria unnecessary. Why say "This combination of 
critreia can't be met by any method", when one of
those criteria is one which, itself, can't be met by any method?

So here's what I think is meant by IIAC:

Deleting a loser from the ballots, and then recounting those ballots,
should never change who wins.

[end of definition]

It _is_ meaningful to say that Approval & Plurality meet this simple

Approval meets Arrow's outcome-criteria. IRV doesn't, because it
fails IIAC.

The IRVies always invoke Arrow to try to show that Arrow proved that
we can't have a good voting system, and so we might as well resign
outselves to the worst piece of crap: IRV.

As I was saying before, Arrow can be summed up:

If you want some semblence of majority rule, and if you don't want
the deletion of a loser from the ballots to change the winner when
we recount those ballots, then don't use rank-balloting.

>Participation and Consistency suggest that the method is in some way
>consistent with itself.  Although I don't view these criteria as
>reasonable, someone who did might argue that a methods
>self-inconsistency proves that it isn't giving the ideal answer in all
>cases.  However, I'm not arguing that for IRV, and you're not arguing
>that for approval.

No, I think that when the good rank methods fail Participation, that
can be explained by saying that when you count all of our pairwise
preferences, it shouldn't be surprising if some of them act against
others of them. That's unavoidable for a method that fully counts
all of our pairwise preferences.

Condorcet, because it _does_ fully count those preferences, offers
guarantees that can't be gotten otherwise, and that's more important
than Participation.

But IRV offers nothing in return for Participation. So I _do_ use
Participation against IRV, in the California League of Women Voters
debates, where the only 2 methods they're considering are Approval
and IRV.

> > For me, an important way
> >in which IRV becomes worse than Plurality is when candidates' support
> >tapers gradually away from the voter median point, with the voter
> >median canddiate having the most 1st choice support. That doesn't
> >at all unusual or contived. In fact it sounds typical. And in that
> >ordinary, typical scenario, IRV fails where sincere voting would work
> >fine in Plurality.
>Obviously, if the CW gets a plurality of the votes, I agree that
>plurality is picking the best winner.  However, you seem to be implying
>that a gradual tapering will always result in IRV picking someone
>different.  Is that true?

Well, doesn't it seem so? First the extremes get eliminated, and
transfer inward. Surely, if the taper is at all gradual, candidtates
to the sides of the median CW will accumulate enough votes to
elimiate the CW.

>I wouldn't be so confident in the centre being a plurality winner.  This
>depends on how close the next two most extreme positions are to the
>centre on the ideological spectrum.  If people vote on a left-right
>basis, then a central candidate will on average only get half the number
>of votes between the nextmost left and nextmost right candidates.

True, but it seems to me that it wouldn't be unusual for the
median candidate to get the most votes, except that the middle
tends to be a more crowded place, as you suggest, with more candidates
to divide the 1st choice support up.

Ok, the voter median candidate might not have a plurality, but that
helps show that 1st choice support doesn't mean much, if the popularity
of the median position can make the median candidates get eliminated
all the easier in IRV. That further discredits IRV.

Sure, if Plurality wouldn't elect that median CW
either, that weakens that Plurality vs IRV argument.

>I certainly wouldn't argue that IRV is the best possible.  On the other
>hand, I wouldn't try to organize against the efforts of people trying to
>implement it.

But I bet that would depend on how honest those IRVies are, and
how sleazy their tactics are. Sure, if the IRVies were honest, I
might not oppose them in the way that I do.

>But perhaps you can help me with what I consider to be the biggest
>problem with approval.  Because in approval you must place all
>candidates in one of two conditions, you run into problems if there are
>a large number of reasonable candidates.

Vote for the ones who are above average, if it's 0-info. Otherwise
vote for the ones whose strategic value is positive. (For those who
aren't interested in the mathematical strategy of Approval &
Plurality, it's: Vote for the candidate you'd vote for in Plurality,
and for everyone whom like better than him).

>For example, if parties started running their presidential candidates
>directly in the election, instead of a primary, then a voter has to make
>a choice.  Do I use my ballot to distinguish between candidates from my
>party, and therefore increase the danger of the other party winning, or
>do I vote a strict party slate and get no opportunity to distinguish
>between the candidates of my party.
>When the result comes in it doesn't appear to me to be terribly
>meaningful.  If one party triumphs over another, than can just mean that
>it's supporters were more willing to ignore differences between its
>candidates.  If one candidate wins among the candidates of one party,
>that may just mean that his supporters were more willing to risk loss to
>the other party.  Kind of like a game of "chicken".

Why let the artificial division of party sponsorship become a
complication? Just vote as I described of. If it's 0-info, vote
for all the above-mean candidates. Otherwise vote for all the candidates
with positive strategic value (as calculated at the website

>For this reason, I see IRV as being substantially better than approval
>when there are a large number of winnable candidates.

You mean because a party is a "mutual majority", and IRV's MMC
compliance helps it?But are you sure that you like _everyone_ in
your party better than everyone outside it? If not, then MMC
might not really apply.

Criteria like MMC are usually defined so that either no method meets
them, or so that even Plurality meets them. Here's how I define
an MMC that isn't like that:

If there's a set of voters, comprising a majority of all the voters,
who all prefer every candidate in set S to every candidate not in S,
and if they all vote sincerely, then the winner must come from S.

[end of definition]

That way, some methods pass, but Plurality & Approval don't, which
is probably in keeping with the criterion's intent.

I define sincere voting as follows:

A voter votes sincerely if he doesn't falsify a pairwise preference,
or leave unexpressed a pairwise preference that the voting system
would have allowed him to express in addition to those that he did

[end of definition]

But I always point out that IRV's MMC compliance doesn't count for a
whole lot, because every MMC example is also an IRV badexample, where
IRV fails the defensive strategy criteria.

Using MMC for IRV implies a claim that the mutual majorities will
be common enough to matter, and, if that's true, then you're saying
that IRV will be failing the defensive strategy criteria a lot.
Including criteria that Approval will never fail: WDSC & FBC.

Mike Ossipoff

Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com.

Share information about yourself, create your own public profile at 

More information about the Election-Methods mailing list