[EM] margins and winning votes

[James Green-Armytage]

Fan de Condorcet wrote:
>James,
>I wouldn't call myself pro-margins, but since you asked for the views of
>the
>"other side", here's my attempt to give an overview of the arguments:
>Most of the arguments for margins I've seen have been along the lines of,
>"Using winning votes to prevent strategic truncation is misguided, because
>providing a random ranking after a given point has a similar effect."  In
>other words, most of the arguments are negative -- not positive.
>One exception is that many moons ago a few people on this list argued for
>margins (partly) on the grounds that it fared better in simulations that
>measured social utility.
>Here's Blake Cretney's page defending margins or, more accurately,
>criticizing winning votes:
>http://condorcet.org/rp/inc.shtml
>CF


	Thank you very much for the reference. That's a good example of what I
was looking for.
	I did read this page of Blake's, although at present I still find the
winning votes argument to be more compelling. Blake's argument seems to be
on more aesthetic grounds, for example, that it seems ugly for a defeat of
51-49 to be kept while a defeat of 49-2 or 50-2 is dropped. In a way, he
seems to be treating it as somewhat axiomatic that the strength of a
defeat is defined by its margin, in that he simply points out these sorts
of examples and says, hey, look how wide the margin is on the 50-2 defeat,
therefore it's unfair to drop that one instead of the 51-49.
	It seem like this is the positive thrust of his argument for margins and
against winning votes, and the rest of the page is trying to neutralize or
minimize the strength of arguments going in the other direction, without
making any more positive arguments for margins. 
	For example, he does recognize this example

>Group 
>I 49 A -- Actually prefer A > B > C 
>II 11 B > A > C 
>III 40 C > B > A 

as a case where margins allows abuse via strategic truncation, but he
seems to counter the point by saying that the A voters will get the same
effect in winning votes if half of them reverse (A>C>B instead of A>B>C)
and half of them vote sincerely. He's right about this factually, of
course, but I have to go along with Mike O. in saying that it takes a lot
more organization and a lot more deviousness to do this. It is plausible
that voters will truncate against a second favorite to help their
favorite, without anyone telling them to do this. It is also plausible
that voters will order-reverse for the same reason without anyone telling
them to, but I think that it is less probable that it will happen on a
large scale.
	As some people may remember, I am wont to point out that winning votes is
not without some significant possibilities for strategic manipulation, but
on the whole I would still say that strategic manipulation is easier, and
more likely to get out of hand, in margins. 
	Of course, to make a shameless plug, at this point I prefer the weighted
pairwise method to both. (Although you can see that the rules of weighted
pairwise are based much more on winning votes than on margins.)


best,
James