[EM] Re: question/comments re DMC

[Dave Ketchum]

On Mon, 29 Aug 2005 11:01:21 -0700 Araucaria Araucana wrote:

> On 28 Aug 2005 at 17:19 UTC-0700, Warren Smith wrote:
> 
>>--(also true of range)
>>I'm a bit worried here.  Heitzig was telling me DMC could be done with
>>	equality rankings
>>like A>B=C>D=>F too.  However...  in the plain Condorcet world, permitting equalities
>>*and* using winning votes makes a big difference and you need to do *Both*
>>to reap the benefits.  So I would like to know how DMC handles that, in view
>>of the previous remark 9 that DMC eliminated  the margins/winnign
>>votes debate.
>>
> 
> There are those who maintain that when candidates are equal-ranked,
> each should get half a vote.  As far as I can tell, all those who
> support this position are marginal-votes proponents.


SOMEBODY does not understand - what I see:
      With margins the difference in counts between A and B matters - and 
adding half a vote to each has no effect on the margin.
      I argue for adding the half votes with AV - which I back and where 
they do make a difference.

> 
> If approval cutoff is implemented using an extra fictional
> Not-Approved candidate, with votes "for X against Not-Approved"
> counted as approval, then the half-vote-split equal-rank tabulation
> artificially inflates X's approval rating.  X might then avoid being
> eliminated (in DMC) by a truly higher approved candidate.
> 
> In my winning-votes-biased opinion, equal ranking should be considered
> as an abstention in the contest, tabulated as *no* vote for either
> candidate.  As it happens, this Does The Right Thing on an approval
> cutoff ballot, no extra gimmicks required.


What I say above fits with my belief that two votes for A=B should have 
the same effect as one vote each for A>B and B>A (neither of these show 
voter intent to abstain.  On the other hand, truncation is deliberate 
abstention).

> 
> Furthermore, ER-half-vote-split loses information.  Once the totals
> are accumulated, you can no longer tell how many equal-rank votes were
> cast.  But that information is still available directly from the
> pairwise matrix if you use ER-equals-abstention:
> 
>     (X=Y) = total - (X>Y + Y>X)
> 
> where
> 
>     (X=Y): total number of equal ranks
>     total: total number of ballots
>     X>Y  : total number of ballots ranking X over Y
>     Y>X  : total number of ballots ranking Y over X
> 
> Shouldn't a voting method try to avoid entropy, rather than increasing
> it?  :-)


I get lost in this last paragraph.  I need an entry in the array for A>B, 
and one for B>A.  With truncation these do not add up to total ballots. 
Do you want a half an array to count explicitly A=B.  I do not see it as 
valuable enough to keep around, and simply fold into A>B and B>A.

> 
> Q

-- 
  davek at clarityconnect.com    people.clarityconnect.com/webpages3/davek
  Dave Ketchum   108 Halstead Ave, Owego, NY  13827-1708   607-687-5026
            Do to no one what you would not want done to you.
                  If you want peace, work for justice.