[EM] Retiring from voting systems, quitting EM

[Kevin Venzke]

Hi Daniel,

What I see in the example is that in the WV election, the addition of one voter moved

the win from a candidate with a lot of first preferences to one who had had none. But
that can't be the intention exactly, because if one cares about first preferences then margins
isn't a good method either.

I guess (though it's not crystal clear to me) that you aren't saying you have a problem with

the 101-voter scenario's outcome per se, only how it can be reached and how it changes
via the addition of one single ballot.

For this to be a general claim of margins over WV we would probably run simulations
aiming to show that margins has greater stability of its result, I suppose. I can see why
that finding might be possible. But instability doesn't seem to have inherent victims, which
would make me more tolerant of it.

I like to imagine how this would play out if ABCD were four options being considered by
an assembly. If the larger blocs are unwilling to vote for anyone but their favorite, then 
indeed the 101st voter will have to flip a coin to break the A-C tie, if he felt like it. But 
otherwise, how can the A and C factions avoid trying to form a coalition with the B>D 
voter? The alternative is to lose entirely.

(As far as why to bring an assembly into it: Because members of an assembly typically
have good information about how many votes are prepared to go in which ways, and
consequently they aren't likely to regret voting for/against some proposal for the reason
that they didn't realize another option was viable. Thus trying to imitate the process
of a deliberating assembly, when designing a method, should help reduce voter regrets,
i.e. strategy incentives.)

Kevin



________________________________
 De : Daniel Bishop <danbishop04 at gmail.com>
À : em <election-methods at electorama.com> 
Envoyé le : Mercredi 23 juillet 2014 21h58
Objet : Re: [EM] Retiring from voting systems, quitting EM
 

> 
>WV is not without its disadvantages.  Consider, for example (thanks to
>Juho Laatu), the election
>
>
>50: A>B
>50: C>D
>
>
>This could happen in an alternate US history in which the 12th Amendment
>introduced Condorcet voting to the Electoral College instead of
>separating the presidential and VP ballots.  A and C would be the
>presidential candidates, and B and D their respective running mates.
>
>
>Four of the six pairwise contests (A-C, A-D, B-C, B-D) are 50-50 ties,
>and the other two (A>B, C>D) are 50-0 landslides.  This is unsurprising
>considering that *nobody* voted B>A or D>C.
>
>
>But consider what happens if just one B>D vote is added.  This creates
>an A>B>C>D>A Condorcet cycle.  I shall assume the Tideman method will be
>used to break it, but I believe that Schulze gives the same results.
>
>
>Under WV, the co-strongest defeats are B>C, B>D, and D>A, with 51 votes
>each.  They do not contradict each other, so we can lock all three of
>them.  This reduces the 24 possible orderings of candidates to three
>(B>C>D>A, B>D>A>C, B>D>C>A), all of which have B as the winner.  IOW, a
>single elector can turn a tied presidential election into a victory for
>one of the *vice*-presidential candidates!  How does this make any
>sense?!
>
>
>Under Margins, however, the strongest victories are A>B and C>D (with a
>net of 49 votes each).  After that point, it's ambiguous whether A or C
>wins, but that's what we'd expect.  B would need another 50 votes in
>order to win the election.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20140725/41ac1508/attachment-0001.htm>