[EM] RE: [Condorcet] Can we come to consensus?
simmonfo at up.edu
Wed Sep 7 14:18:53 PDT 2005
Jeff wrote ...
Now that we have presented a few favorites and examined them, I hope we all
at least understand each other, even if we do not yet agree. It now looks
like we have the following:
* DMC has one ardent supporter and several contributors who think it
* Shulze (WV) is widely acceptable, already in use, and hated by none.
* Smith//Approval is widely acceptable, but it has a few detractors who
Shulze and Smith//Approval have been around the longest, so they are more widely known than DMC.
But let's compare them head-to-head with DMC.
DMC is identical to Shulze when defeat strength is measured by winning approval (instead of winning votes). The only disadvantage of DMC in this comparison is that it requires an approval cutoff. In all other ways, including ease of description, DMC is as good or better than Shulze.
In particular, because Shulze lacks approval information it cannot take it into account. Because of this DMC tends to pick a better winner than Shulze under zero-info sincere voting.
On the other hand, "winning approval" is a more robust measure of defeat strength than "winning votes." Because of this fact, Shulze is at least as vulnerable to strategical manipulation as DMC is. To understand this better see the example I posted this morning to both the Condorcet and the EM list. [It hasn't shown up yet on the Condorcet list.]
As in Smith//Approval the DMC winner always comes from the Smith set. When the two methods choose different members of the Smith set, DMC chooses the pairwise winner of the two, while Smith//Approval chooses the one with the higher approval. So Smith//Approval favors approval information over pairwise information compared to DMC.
I would rather see plain old garden variety Approval or Range(-1,0,1) than Smith//Approval, which seems like a cobbled together kluge in comparison with DMC.
Also, as has been pointed out, the Smith//Approval winner X can lose pairwise to the Smith//Approval runner up Y. In other words it can happen that in a runoff between the winner X and the candidate Y who would have won had the winner dropped out, this candidate Y would win.
DMC is not afflicted by this problem.
The perception that Shulze is more widely acceptable than DMC may be an illusion due to the fact that members who have not weighed in on this issue consider the 15 reasons to be so definitive as to make the conclusion obvious.
In my opinion the best Condorcet proposal is DMC. I suspect that a majority of members who have read the fifteen reasons would agree.
However, if there is no simple ballot option for the voters who don't want to rank the candidates, i.e. for those voters that would rather delegate that task to their favorite, then in that case Approval is a better public proposal.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Election-Methods