[EM] Proportional Representation via Approval Voting (fwd)

[LAYTON Craig]

>But only if your utility formula is accurate for a given election.  It
>is not enough to say that it is accurate on average, since the unranked
>method may adapt better to variations in individual elections, yielding
>a higher average utility.
>
>In fact, given STV's slim 3.5% lead using the Borda assumptions, I think
>it highly probable that an unranked method would win under more
>realistic models (just as single-winner approval consistently beats out
>IRV).

It was a weak example, but the thing that I was most interested in was that
the results of the two systems do seem to vary significantly (in terms of
who gets elected).

Yes, IRV is a flawed voting system, but I don't think it is safe to assume
that because it doesn't work well for single winner elections, it shouldn't
be applied to multi-winner elections.  The main problem with IRV is that it
doesn't examine a voter's full set of preferences, so significant
information may be ignored.  STV goes through almost all of the majority of
voters preferences.  Like the utility calculations in Proportional Approval,
if a voter's first choice (A) is elected, it isn't considered terribly
important that he slightly prefers B over C (and the allocation of only a
fraction of his vote reflects this).  He's already secured representation.
The number of preferences that are ignored is very small in comparison to
IRV, and the problems much less, well, problematic.  Certainly, the
eliminations are always a little questionable, but they cause much more of a
problem when there are truncated votes and inadequate count rules (like the
rule that states if there are 3 uneliminated candidates, and 3 seats
remaining, they are all declared elected).

>I have serious doubts about the proxy system.  Would the variable votes
>carry over into committee assignments?  Would an individual's larger or
>smaller voting power affect the chances of even getting a committee
>assignment?  Should an representative's time on the floor be in
>proportion to voting power, etc.?  How about fundraising limits and
>operating budgets?  It all sounds like a big mess to me.  Much better to
>minimize inequity before the candidates are elected.  I think the quest
>for proportionality can be taken to excess anyway -- it should be
>balanced with other things, such as utility & practicality, etc.

Fundraising? Well, you need to fix that anyway.

I'm not sure how the US committee system works, but I think that members of
parliament should probably be equal in all things except in actually passing
motions and legislation in the house (I would generally support a unicameral
system for any form of PR).  Whether or not the parties want to carry
variable voting power over into the party room is up to them.

I don't think that is messy at all.

The issue is the size of electorates.  STV (and PAV) only work well in
relatively small electorates, with a limited number of candidates running.
As a result, proportionality suffers significantly.  The proxy system party
mitigates this, and the accompanying criticism that minor parties with a
non-localised support base are unfairly disadvantaged by electorate-based
electoral systems.

It is much more difficult to do utility calculations with a proxy system,
but it seems to me that utility would be improved, as more voters get higher
choices elected (the number of wasted votes, or votes helping to elect
candidates much further down the ballot paper is really quite small).  

So, you get improved political participation, as there is a higher chance
your vote will count and you're happier with the outcome, as there is a very
good chance that one of your highest choices will be elected.  It's also
fairly proportional, and much easier, cheaper and faster to count than other
PR systems.  The only real problem is the eliminations (and differences
between electorates, but that can't be helped).  There may be ways to fix up
the elimination process, like looking for Condorcet losers, but I haven't
investigated this yet.