[EM] STV and weighted positional methods

[Raph Frank]

On Sat, Jan 31, 2009 at 8:30 PM, Kathy Dopp <kathy.dopp at gmail.com> wrote:
> False. It happens whenever the number of candidates is more than the
> number of rankings allowed on a ballot plus the number of seats being
> filled.

Ok, fair enough.  However, IRV is supposed to allow an unlimited
number of ranks.  Fairvote seems to try to apply the name to a huge
range of voting methods, not matter how many ranks are allow.  In
fact, they may have applied it to instant top-2 runoff.

This allows them claim it is in use in lots of places.

Also, if you have 3 ranks, then as long as you give your 3rd rank to
one of the top-2, you should be OK and certainly not any worse off
than with plurality.

>>> Also, if you always rank one of the top-2, then you are likely to be
>>> part of the last round, even if you don't rank everyone.
>
> Unlike top-two runoff or primary/general elections when I am always
> allowed to participate in the final counting round no matter who I
> voted for in the prior election.

Top 2 run-off is also potentially non-monotonic when considered as a whole.

If you vote for a candidate in round 1, it might mean that you end up
electing a worse candidate in round 2, due to eliminating a
compromise.

Each stage is monotonic when taken individually though.

> Whoever made this statement makes false unsupported assumptions about
> my position. While it may be true that plurality is one of the worst
> voting methods available, there is a far worse voting method than
> plurality and that is IRV/STV.

The two methods both tend towards 2 party domination and will likely
give the same results anyway.

I think they are both very poor methods.

>>> I don't think you support plurality in order to maintain the monopoly of current voting machine vendors.
>
> This above statement is hopelessly illogical.

They point I was making was that I don't think you are acting in bad
faith.  I don't think your support of plurality over IRV is due to bad
faith, it is just down to disagreement.

>>> In Ireland, we count PR-STV by hand and there are various checks that can be accomplished.
>
> I believe that in Ireland you also have far fewer issues and election
> contests to vote on for each ballot.  Am I wrong?
>
> Computer scientists have already mathematically proven that counting
> IRV/STV is an exponential problem in computer science. Far far more
> difficult and time-consuming to count accurately than other voting
> methods. I am fairly certain that your assertion about counting time
> is incorrect.

Maybe you are thinking of Meek's method (or one of the really complex
ones like CPO-STV or Schulze-STV)?  They requires a computer.

Standard PR-STV takes at most one pass through the votes per round and
in each round, a candidate is elected or eliminated.

The Irish method is slightly random and is roughly

Initialisation:

1) sort all the ballots into piles based on first choice
2) Count all the piles
3) work out quota

Processing (once per round)

If any candidate has more than the quota in his pile
-- declare that candidate elected.
-- Select ballots equal to the surplus from the ballots for that
candidate at random
-- They are selected using stratified random sampling (based on next
highest ranking)
-- assign those ballots to the other candidates
Otherwise
-- declare lowest candidate eliminated
-- redistribute all the votes in his pile

The first count requires examination of all the ballots, but the later
rounds only require looking at the surplus or the ballots for the
eliminated candidate.  This is a much smaller set of ballots than all
the ballots.

The sub-piles are kept separate so as to allow easy checking during
the recount.  The recount only has to check that each ballot is in a
valid sub-pile and that the total in the sub-pile matches the original
count.