[EM] Condorcet-Approval hybrid method

[Russ Paielli]

Forest Simmons simmonfo-at-up.edu |EMlist| wrote:
> Russ asked about what we used to call "Approval Completed Condorcet."
> 
> The legendary Demorep was an avid proponent of several variations of 
> this idea, one of which he christened ACMA for Approval, Condorcet, 
> Maximum Approval, a three step method:
> 
> Step 1: Approval: first eliminate all candidates with more disapproval 
> than approval.
> 
> Step 2: Condorcet: elect the Condorcet Winner among the remaining 
> candidates if there is one.
> 
> Step 3: Maximum Approval: in the case of no CW in step 2, elect the 
> candidate with maximum approval.

The first step is arbitrary and I would eliminate it.

I would start by simply choosing the CW if one exists, or paring the 
field down to the Smith set otherwise. Then I would eliminate the 
candidate with the lowest approval and repeat.

I thought of this yesterday while I was working out, and I thought I had 
come up with something big. Then I searched the EM archives and 
discovered that Kevin Venzke had mentioned it way back in 2003.

Oh, by the way, I would *not* allow equal rankings. Why not? I just 
don't like them. They strike me as an unnecessary complication and 
little more than a way to game the system.

> Like I say, Demorep and others came up with many variations on this, 
> including using truncation for the approval cutoff.  But note advocating 
> this truncation approach for an approval cutoff is tantamount to 
> thumbing one's nose at the later-no-harm criterion that Kevin is trying 
> valiantly to rescue.

I don't follow you here. I would require the voter to rank the approved 
candidates and simply not rank the unapproved candidates. That's simple 
and intuitive. Creating an approval cutoff separate from the rank list 
would be highly undesirable, I think, in terms of public acceptance. If 
it's absolutely needed for strategic concerns, then I just don't think 
the method will ever fly. Ditto for equal rankings.

> My idea was to use standard ordinal ranking ballots with a kind of 
> artificial candidate called MAC for Minimum Acceptable Candidate, or 
> NOTB for None of the Below, or some other catchy name.  The voters 
> indicate their approval cutoff by ranking all of their approved 
> candidates above this artificial candidate, and all others below it. 
> Approval scores can then be found in the column of the pairwise matrix 
> corresponding to this artificial candidate.
> 
> More recently I suggest adding a suitable lottery to the list of 
> candidates.  This lottery serves as a kind of approval cutoff.  One can 
> then implement Demorep's ACMA with a ready answer to the frequent 
> question, "What if no candidate gets more than 50% approval in step one?"

As I said earlier, I think step 1 is arbitrary and unnecessary.

--Russ