[EM] IRV's "majority winner". What if we let the people choose?

Bart Ingles bartman at netgate.net
Wed May 19 00:59:01 PDT 2004


Adam H Tarr wrote:
>
> >James Gilmour wrote:
> >> Now consider:
> >> 49  A<C<B
> >> 48  B<C<A
> >>  3  C<B<A
> >> IRV winner = B;  CW winner = C.
> >> I doubt very much whether most electors would accept C as the "winner"
> >> if this were an election for State Governor, much less for a directly
> >> elected President of the USA.  If anyone has evidence to the contrary I'd
> >> like very much to see it.
> >
> Bart Ingles replied:
> >
> >Whether C has widespread acceptability depends almost entirely on
> >information which is not captured in ranked ballots.  At the extremes, C
> >may enjoy either unanimous popularity, or near total rejection.
> 
> [AT]
> A hardcore Condorcet supporter would say that the distinction is irrelevant,
> since either way C wins all contests pairwise.  To put it another way, C would
> win any approval election where the voters have perfect information.

[BI]
If the voters had perfect information, C would win the IRV election as
well (assuming the voters were willing to use obvious strategy).  So the
"perfect info" scenario doesn't provide a strong argument for or against
any particular method.


> >Approval voting is able to distinguish between these extreme cases with
> >ease.
> 
> Approval only accurately distinguished between these cases when all the voters
> use the same utility-based cutoff.  If some voters have different utility
> cutoffs than others, or if some of them actually look at the polls before
> voting, then the ability of approval to distinguish between the two extremes
> becomes a bit muddied.

I'm assuming that in the two extreme cases, the utilities are high or
low enough to be above or below any reasonable cutoff.  I thought this
was especially relevant to the question being raised, to the effect that
voters might not accept the outcome of a Condorcet election.

In less extreme cases-- where voters are fairly ambivalent about their
middle choices relative to the other two-- of course the approval
outcome is less clear-cut.  But in that case I wouldn't expect the
voters to be lined up with pitchforks in order to overturn the results
of the election (either approval or Condorcet).


> >Ranked methods can only do so to the extent that they encourage
> >insincere strategy.
> 
> What strategy would that be?  The sincere votes here are already a Nash
> Equilibrium.  The only thing the outer factions can do is throw the election to
> the other outer faction.  This is a pretty strategy-free example in Condorcet
> voting.

By 'ranked methods' I wasn't necessarily assuming Condorcet here. 
Although I still believe that a prisoner's dilemma outcome is possible
if C is disliked enough, and if neither of A and B has a clear advantage
over the other.  But even if you don't buy that possiblity, my original
point remains-- if there is no insincere strategy, then a ranked method
can't distinguish between the two extreme cases that fit the ranked
example above.



More information about the Election-Methods mailing list