[EM] "1 person - 1 vote"

[Elisabeth Varin/Stephane Rouillon]

Mike --

MIKE OSSIPOFF wrote :

> Steph--
>
> I'd said:
>
> >But I wrote those postings because you'd said that Approval violates
> >1-person-1-vote, giving voters unequal amounts of power. You probably
> >don't believe that Plurality violates 1-person-1-vote, and so I showed
> >that actually Plurality gives more disparate voting power than
> >Approval does.

We agree about power given by those methods.
As I said I consider FPTP to respect "1 person - 1 vote" because you
can build a bijection between the voters and an identical weight contribution
to the final results. Approval does not respect this property or criteria.
However FPTP has splitting votes problems Approval has not.
Voting power results from both this aspects.

> You continued:
>
> It gives a slight advantage to voters who would put their approval cut-off
> splitting candidates in two equal groups.
>
> I reply:
>
> >From what I posted on that, Approval can favor one voter over another
> by a factor of 9/5, while Plurality can favor one voter over another
> by a factor of 5. Thus, Plurality can favor one voter over another
> by a factor nearly 3 times greater than Approval can.

These numbers seem really artificial without context.
Please could you post that again?

> You continued:
>
> Approval artificially disfavours extreme
> candidates.
>
> I reply:
>
> That's what you said before: You don't like electing voter median
> candidates. And so when Approval, or Condorcet(wv), or Condorcet(rm)
> elects a CW, it's artificially disfavoring the extreme candidates who
> should have won instead of the CW. I guess I'm not quite sure what
> you mean by "artificially". If you mean "as a result of the voting
> system's properties, then I admit that Approval, Condorcet(wv), and
> Condorcet(rm) artificially disfavor extreme candidates when they
> elect voter median candidates.

No I think, even if I have no proof that Condorcet methods gives more
an equal weight to every voter in the outcome and resolve the vote-splitting
issue by using pairwise comparison. Relative margin, margin and winning -votes
all  fit when all pairwise comparison are expressed. I recommend relative
margins (rm) because it fits that equal weight idea when unexpressed
comparison representing a no opinion (A ? B) occur.

> But what you probably mean is that when Condorcet(rm) elects an
> extreme candidate, in violation of majority rule, it's doing the right
> thing, because it isn't artificially disfavoring extreme candidates.
> But you can't count on rm always doing that. Often even it will elect the
> CW, and therefore fail your standard.

No I do not. I think Winning-votes does a better job in this case.
All I would like to evaluate is, is this gain worth the loss of representation
when there is no unsincere truncation problems (which does happen a lot
more often). I am still trying to evaluate this...

> You continue:
>
> But vote-splitting is by far a greater problem. So Approval is a lot better
> than FPTP even if the latter does not violate "1-person-1-vote".
> End of my logic.
>
> I reply:
>
> But if "1-person-1-vote" means what you said above, then FPTP violates
> 1-person-1-vote about 3 times worse than Approval does.

No see before...

> And if that isn't what you mean by 1-person-1-vote, then exactly what
> do you mean by 1-person-1-vote?
>
> Mike Ossipoff

----
For more information about this list (subscribe, unsubscribe, FAQ, etc), 
please see http://www.eskimo.com/~robla/em