[EM] Suggested name for range voting: free voting

[Anthony O'Neal]

I can't believe I missed such a massive flaw...

Coming up with an approval-range method not vulnerable to vote management is
difficult.

For electing a candidate, we consume an amount of voting power from the
people who had such a candidate marked by V/ N + 1 (the Bischoff quota)
(where V is your definition of voting power, and N is the amount of seats).

But how do we define voting power in a way that is cloneproof?  I'm really
to tired to dwell on that question right now...



>> In effect, your approval for an outcome is just the sum of your
> approval
> >> for each of the individual candidates elected.  However, there is a
> >> limit
> >> to prevent any one vote from becoming to strong.
> >
> >If you reduce the strength of the vote for having multiple candidates
> >approved of it becomes cumulative vote, which is very vulnerable to
> tactical
> >voting.
>
> Sorry, I meant limit the total voting power.  This occurs anyway under
> the system where there are divisors.
>

I see.  Well, it does make sense.



> The approval of outcomes method is probably the only way to overcome
> this.
> >
> > The sequential  method is vulnerable to vote management and introduces
> > tactical voting into it.
>
> The non-sequential method also suffers from tactical voting as I showed
> above
> (unless I made an error).
>

I believe you were right about the tactical issues, which are caused because
highest averages methods are so very blunt.  For small elections, especially
where individuals may choose their own candidates, you must use surplus
votes transfer.
- Show quoted text -


>It's also more vulnerable than a computer total
> >simply because people can just lie about the votes their getting as
> they're
> >hand-counting them.
>
> Ok, we have a fundamental disagreement here.  In a hand count there
> might
> be some small amount of fraud/error.  However, to really rig an
> election,
> you need to get lots of counters involved.  Also, those counters are
> observed.  This makes it easier for there to be a small error but harder
> for their to be a massive error.
>
> A computer has a single point of failure (the program) and cannot be
> readily observed.  Also, the general public doesn't really understand
> computers and those that do are often wary of using computers to do the
> tally.
>
> > For computer methods, the complexity doesn't matter.  It's just as
> easy to
> > make a program that hurts candidates of one party in STV as it is in
> PAV and
> > PRV.  And, actually, the only way to do STV elections without a
> randomness
> > is to use a computer.  The only real alternative to using complex
> methods
> > for proportionality is a party-list, which is undesirable because it
> > completely takes away candidate independence
>
> You cannot do meeks method or some of the more advanced STV-PR by
> hand however, it is possible to do fraction STV-PR by hand (look up
> Gregory method).
>

I know of the Gregory method.  It isn't exactly true to say that the Gregory
method is random, but it doesn't treat every vote the same and isn't as
fair.  Besides, the method is so labour intensive that it is only used for
the senate of Ireland (where there aren't many voters) and Northern Ireland
(where the number of transfers per a ballot is limited).

The problem is the subsequent surplus.  Whenever a vote transfers to a
candidate that's already been elected, most methods just skip over him.
This is, essentially, Hare's method, in that the votes that have been
transferred to and from the already elected candidate are not representative
of that candidate, but representative of the ones that have been transferred
to him instead.  If you try and apply Gregory's method to the subsequent
surplus instead (taking all already consumed votes from the already elected
candidate, adding the votes that have just been transferred to him, and
transferring all votes again at a new fractional value), then some of the
votes may come back to the candidate they were transfered from, and then
transferred to the same candidate again, then transferred back, and so on
and so on, an infinity recursion.

So the pushover strategy becomes very effective in such a case, because you
don't want your vote to be dumped into a basket and never considered again
because one of your favored candidate were elected.

In the case of the subsequent surplus, I'm guessing it could be almost as
effective as Meeks and Warrens method to have votes transferred to an
elected candidate and alter the voting power of the people who voted for him
wit the amount that was just transferred to the candidate, excluding the
votes that would lead to an infinity recursion.  You simply transfer those
votes directly to the next candidate on their list without altering their
fraction.  This is somewhat vulnerable to the pushover strategy, but not
nearly as much as the normal Gregory method.  It is possible to hand count
with this method, but still extremely labour intensive.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20060621/d6c43bfd/attachment-0003.htm>