[EM] IRV inconsistency

[Markus Schulze]

Dear Forest,

you wrote (16 May 2001):
> Markus doesn't consider Approval Condorcet Maximum Approval (ACMA) to be
> a true Condorcet method because it requires more information than the
> other Condorcet methods do, namely the Yes/No approval information or the
> ranking of an (extra) virtual reference candidate MAC (Minimum Acceptable
> Candidate).

Nope! To prove that the consistency criterion and the Condorcet criterion
are incompatible it is not necessary to presume that the winner of the used
election method depends only on the relative rankings of the voters. The
proof is valid for every election method that meets the Condorcet criterion.

******

You wrote (16 May 2001):
> Markus wrote (16 May 2001):
> > Forest wrote (15 May 2001):
> > > As Richard says, IRV ignores and trashes valuable information
> > > willy nilly and still pretends to come up with a top notch
> > > winner. All of IRV's shortcomings can be traced back to this
> > > wasting of information.
> >
> > It is not feasible to say that the one election method uses
> > more information than the other. All election methods use the
> > same information. They only interpret it differently. Election
> > methods only differ in how important they consider which part
> > of this information.
>
> Although many methods have the same available information they
> obviously don't all use it to the same degree, for example random
> candidate doesn't use any of the available information (except who
> the candidates are), and random ballot doesn't use as much as most
> other methods.
>
> IRV looks only at first choices before making an irrevocable
> decision of whom to eliminate first.  After the elimination is
> accomplished all of the comparisons relative to the eliminated
> candidate are irretrievably lost.
>
> This would be of no consequence if IRV both satisfied the IIAC
> and made a careful consideration of whom to eliminate in the first
> round. But it does neither of those.

On the other side, Condorcet methods are criticized very frequently
because the winner depends "only" on the pairwise matrix while
other information is ignored.

Markus Schulze