atarr at purdue.edu
Tue Feb 11 20:35:46 PST 2003
matt matt wrote:
>My perspective is that the ranked pairs and beatpath methods describe an
>algorithm for determining contest outcomes independently from the details
>regarding how equal ranking is counted or ballots are completed or ties
>are broken or defeats are measured and the like.
"how equal ranking is counted" and how "defeats are measured" are, indeed,
more closely related to margins vs. winning votes than they are to ranked
pairs or beatpath. How ballots are completed is, as I interpret the
phrase, essentially what ranked pairs and beatpath are all about. But I
think this is a useless semantic debate, so let's move on.
>I don't understand your claim that A>B=C>>D makes approval cutoff based
>ballot completion into A>B=C>E=F>D irrelevant.
I had no idea that's what you meant by "approval cutoff" - that the unvoted
candidates are placed equally ranked ON the "approval cutoff". Interesting
idea. Nevertheless, it is mostly irrelevant to the margins vs. winning
votes debate. This idea basically amounts to a shorthand way for voters to
fill out a ballot. Voters are allowed to explicitly vote tied rankings, so
there's no ballot you can make this way that a voter couldn't make with a
You should probably give this approach another name, since this is not what
other people think of when you say "approval-completed Condorcet" or
What most people on this list mean by approval completed Condorcet is an
ordering where all candidates above a specified rank are "approved". With
approval completed Condorcet, you don't need to worry about whether you're
using winning votes or margins, because you never actually measure a
defeat's strength. If there isn't a Condorcet winner you take the
candidate with the highest approval count (or in some approaches, the
highest approval count within the Smith set). So defeat strength is
>Do you consider margins vs winning votes irrelevant to approval-completed
>Codorcet because approval cutoff based ballot completion makes more sense
>with equally ranked getting 1/2 vote each?
It's certainly not irrelevant in your flavor of "approval-completed
Condorcet". I'd argue that it definitely does NOT make more sense to give
equally ranked candidates 1/2 vote each, since half votes give you margins
results in stead of (potentially) winning votes results.
>So the only reason to keep margins as a measure of defeat is to
>accommodate someone who wants margins and unvoted ranked last with 0 votes
>each for equally ranked?
That's not what margins does. The effect of margins is the same as the
"1/2 vote for each equal candidate" that you have advocated. If you really
want that, then you want margins. But I think you should sit down with
some examples and test it first. The effect of doing this is that voting
your full preferences can hurt you in many common situations. It's
extremely unlikely that expressing your full preferences can ever hurt you
in winning votes.
>I will look at the 14 variants. I would appreciate further explanation of
>the qualification "the same ballot input as the winning vote
>method". What ballot input is permissable when margins is the measure of
>defeat that is impermissable for winning votes?
All I meant by that is the ballots have to be capable of conveying the same
information; you can't take away expressivity from the winning votes method
and expect it to be able to do the same job. So it wasn't much of a
qualification at all.
I think a lot of the confusion in this discussion was just each of us
having a completely different idea of what "approval completion"
meant. Just as much my fault as yours.
For more information about this list (subscribe, unsubscribe, FAQ, etc),
please see http://www.eskimo.com/~robla/em
More information about the Election-Methods