[EM] STV - the transferrable part is OK (fair), the sequential round elimination is not

[Raph Frank]

On Sat, Oct 31, 2009 at 2:29 PM, Kathy Dopp <kathy.dopp at gmail.com> wrote:
> Rather than reply individually to the three response to my former
> post, I'll just make some observations:
>
> 1. It seems like the pro-IRV/STV group has begun to dominate this list,

I am pro-PR-STV but against IRV.

As with all election methods, it is a trade-off.  The benefits of
PR-STV outweigh the disadvantages.  It gives max control to the voters
while giving reasonable PR.  The more seats elected the better.  With
small constituencies, it isn't so great.

I guess my thoughts would be that PR is better than a single seat
method, and PR-STV is better than a party list system.

> 2. the assumption that "Later-no-harm" is a desirable feature of a
> voting method is very odd. I would claim that the opposite is true, in
> agreement with Abd ul Lomax. Later-no-harm is a feature that prevents
> a voting method from finding majority-favored compromise candidates
> and ensures that IRV/STV tends to find candidates supported by either
> extreme leftist or rightist groups

I agree with this too.  The biggest weakness of PR-STV is that is
collapses to IRV in the single seat case.

I think that it might be worth looking at the elimination ordering to
help with this.

For example, you could have an approval vote held at the same time and
eliminate the least approved remaining candidate, if no candidate is
elected.  This would collapse to (roughly) approval followed by
(instant) top-2 runoff in the single seat case.

The key point would be to use a different method for determining who
is eliminated than is used to determine who is elected.  The preserves
the proportionality of the method while allowing it perform better in
the single seat case.

The problem is that PR-STV is already reasonably complex and most
proposed changes make it even more complex.

There are methods like CPO-STV and Schulze-STV which are similar to
PR-STV.  Both of these methods are condorect compliant in the single
seat case (and so presumably break later-no-harm).  However, that are
so complex, that they would require a computer to perform the tally.

> 3. STV does *not* achieve proportional representation at all unless
> there is no vote splitting and just the right number of candidates run
> who support each group's interests. I.e. the success of methods like
> STV to achieve proportional representation rest in the unlikely
> assumption that just the right proportion of candidates run (or more
> precisely an equal proportion of candidates run) in proportion to the
> number of voters in each separate group.  This is just simple
> mathematical fact.

Generally it does achieve reasonable proportional representation.
Parties might get less than proportional in one constituency and more
than proportionality in another, due to randomness.

However, the smaller the constituencies the bigger the "seat bonus"
given to larger parties.

Again, the more seats per constituency, the better, as that gives
better proportionality and makes it easier for smaller parties to get
seats.

> 5. It always amazes me how irrationally the supporters of IRV/STV
> support a nonmonotonic system that creates more problems than it
> solves when there are clearly better alternatives available that
> actually solve more problems than they create.

I think the issue is that you look at PR-STV and IRV and refuse to see
any difference.

Many people (including many on this list) feel that IRV is a bad method.

However, PR-STV has some advantages over other PR methods.  That is
why people refuse to dismiss it out of hand.