[EM] an example of BTR-IRV (the name as now been changed to STV-ME)

[Anthony O'Neal]

Anthony O'Neal wrote:
    > (This is the election I pulled out of the Wikipedia article for
    CPO-STV.
    > I just didn't feel like making up an election where the results from
    > CPO-STV and STV differ right now. If you want to see how
    > the results for the CPO-STV and STV results were arrived upon, then
    > go to the article)
    >
    >(PS, does anyone know what the BTR part of BTR-IRV means?
    > Honestly, I can't figure it out, but that's what Warren on
    Rangevoting.com
    > calls it. I'm thinking of changing the name to Majority Elimination
    by IRV,
    > or ME-IRV, and ME-STV, but if BTR makes more sense...)It is probably
"Bottom Two Runoff" or something similar.

I guess I really need to change the name then, because it's not always a
runoff between the bottom two, but a runoff between the bottom N + 1 (where
N is the number of seats in the STV election that are currently needing to
be filled).  I'm thinking Single Transferrable Vote with elimination by
majorities, although this also doesn't really fit, because it's actually
elimination by the majorities of the Droop Quota.

    > Now for this election, using the Hagenbach-Bischoff quota of
    votes/seats + 1,
    > the amount needed to get elected is 25 votes. So Andrea and Carter
    are
    > immediately declared elected, as their amount of votes exceeds the
    quota.

    Why not use the Droop quota, that is much fairer?

    Otherwise, "vote management" is a viable strategy. If a party has 44%
    of the vote
    and can manage to split the vote evenly between two candidates 22%
    each, then they
    both get elected. If they don't bother with vote management, the more
    popular
    candidate will only be able to transfer votes above 25%. The second
    candidate
    will only get 19% of the vote which isn't enough for a 2nd seat. The
    system of
    election should not encourage tactical voting if possible, people
    should be
    encouraged to vote honestly.

I believe you are confused.  You are talking about the Hare quota, which is
a rather idealistic attempt to represent all voters.  The problem is that
the minimum amount a person needs to get elected isn't votes/seats, it's
votes/seats + 1, and people can exploit this by making their total vote for
each candidate Votes/seats + 1 instead of having the method take away
votes/seats.

There is a difference of one vote between the quota I'm using and the droop
quota.  The droop quota is (votes/seats + 1), and the quota this election
uses is votes/seats.  I would be surprised is either quota ever actually
produced different results in real election, with such a small difference.
But the quota I'm using is actually an attempt to completely get rid of the
vote management strategy, because droop technically wastes 1 more vote than
they really need to get elected.  The quota I'm using, however, may elect
more people than there are seats.  The reason I use it is because that's how
the election in the Wiki article used it, and it was one of those rare
elections that would be changed around if I had changed the quota (because
Adrea wouldn't have gotten elected in the first round, and sicne no one else
placed her on their ballots she wouldn't get elected at all), and I didn't
feel like messing with the vote totals to find one that would work with
droop whenever the difference was so miniscule.

However, in recommending quotas I'd probably go with Droop rather than
Hagenbach-Bischoff, because in the extraordinarily unlikely event where the
quotas would give different results in real elections, having a person
elected because more votes were wasted than needed is quite a bit less
catostrophic than having more people elected than there are seats.




This brings me to another point, though.  Why even transfer surplus votes?
Why not simply elect the person with the most votes, then transfer all of
that persons voters votes to the second person on their list at a value of
V/M * 2 + 1 (where V is the value of their ballot, and M is the number of
candidates that person has had elected)?  The difference between the surplus
transfer methods and this method is basically the difference between largest
remainder method and highest averages method , and the highest averages
methods are certainly more precise.  I don't believe it would have very many
more chances for strategic voting then the system of transfer of surplus
votes, it may even have less.  You may believe it would be a good idea for a
vote to not vote for a person who is likely to win already, but this is
already a problem under the surplus transfer method.  I don't believe any
group could take advantage of it for strategic purposes, as the groups
collective vote would always be reduced by how many candidates they could
get elected.

It is also much, much simpler than the only two STV counting methods that
are completely fair, Meek and Warren.  Actually, STV-ME could be counted by
hand using my proposed method.

    > I believe that the second method, not counting already consumed
    voters,
    > is the proper method, but this may just be bias since I'm trying to
    paint my
    > method out to be simialar to CPO-STV. I have presented both, so if
    any
    > of you know more about the subject feel free to point me in the right
    direction.

    That seems fair. It is unreasonable that people who have already
    elected a candidate (or will elect a candidate) should have a say in
    who gets
    eliminated.

    What about something like multiple stages

    Stage 1: standard PR-STV is used to select potential winners.

    Stage 2+: BTR-STV is run to select a (potentially) different set of
    potential
    winners

    Ballots which are held by potential winners from the previous stage or
    candidates
    from the current stage who have been deemed elected shall not be
    considered
    when determing which of the bottom two should be eliminated (unless
    they are
    held by one of the bottom two)

    Keep going until 3 stages in sequence result in the same set of
    potential winners
    and deem those elected

    After a cutoff number of stages, either select the winners of the last
    stage or
    the winners of the second last stage as winners by a "coin toss".

    I wonder if this diminishes the problem that BTR violates later no
    harm. Your
    lower ranked choices will not be considered unless your current
    preference
    is unlikely to get elected.

    Any Condorcet method violates later-no-harm, but I only believe they do
so where there is no clear Condorcet winner or the voters can push the
election into not having one.  It would be difficult for a voter to
ascertain whenever such a situation exists, and whether it would be an
effective strategy to lie.  Not as difficult as with IRV, but honestly, I
think literally less than 1% of the population will be able to effectively
take advantage of such flaws in a BTR-IRV or STV-ME election.  BTR-IRV
doesnt' violate later-no-harm by much, and where it does it isn't going to
cause enough truncation to results in something like the Bucklin methods
problem, where it was basically stupid to even put a second choice unless
you knew your guy absolutely wasn't going to win.

Then again, Condorcet would be absolutely detrimental to elections if
burying were implemented widely (and stupidely).  If it were, then I may
support a change to IRV.  The reason I support Condorcet is the type of
consensus candidates it elects, but it is more manipulable than IRV, and if
the people start manipulating it to a wide degree it may produce worse
candidates than IRV.  Then again, it probably will never produce worse
winners the pluarality, which practically enforces tactical voting.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20060608/83b84b81/attachment.htm>