David L Wetzell wrote:
On Fri, Nov 18, 2011 at 7:14 AM, Kristofer Munsterhjelm
<km_el...@lavabit.com <mailto:km_el...@lavabit.com>> wrote:
Loring argues that Plurality councils can swing wildly and deny
representation to people who should be represented, while PR
councils can still be off-center due to kingmaker scenarios, and
that one should therefore pick a center that can break ties while
not giving any voting bloc undue power.
My approach replaces STV with LR Hare, I guess I don't really care
whether rankings get used or not, but I do like having fewer seats with
PR with a Hare Quota, so we can avoid those arbitrary percentage
restrictions. It lets third parties decide who's the party-in-power but
helps the party-in-power get more seats so they can get things done if
they are generally popular, or able to win many of the single-winner
elections.
Personally, I think Sainte-Lague is a more proportional approach than
LR-Hare. In being a divisor method, it can avoid the population-pair
monotonicity problem, and by measures of proportionality, it is usually
about as good as LR-Hare.
Unfortunately, there's no Sainte-Lague (or divisor method) version of
STV yet. The closest I know of is my "Set Webster" method, which
combines a divisor method and the concept of solid coalitions. It is
(most likely) monotone, but not so much better than STV that it's worth
it to switch unless you value monotonicity above all else or are
interested in new methods in and of themselves.
If you want party list, though, use Sainte-Lague (or Webster, same
thing). For the same reason, I think that congressional apportionment
would be better done by Webster's method (as it were, once), than with
the current Huntington-Hill algorithm.
He then proposes to use STV, but shield the CW from losing. The
Condorcet winner represents the center or common consensus position,
while the other winners represent the diversity of opinion among the
people. Because the process is done inside a single method, in the
case the CW is off-center, the proportional representation aspect of
the algorithm will even this out by compensating.
The link you gave me though tends to weigh stronger for the 2nd version,
which is easier to explain to voters by virtue of how it combines two
already existing elections...
I suppose the first version is more like MMP and the second version is
more like parallel voting. That is, the first compensates and the second
just runs in parallel.
You might be able to get something more easily understood yet retaining
some of the compensation part of the first version, by doing something
like this: first elect the single winner/s. Then start STV with the
single winner/s marked as elected (and thus with vote transfers already
done).
The same sort of shielding could be used in any type of multiwinner
system. If it's sequential, you just keep the CW from being
eliminated. If it's combinatorial (like Schulze STV), you only
consider those sets of winners that include the CW.
dlw:If I was gung-ho on getting the CW elected that'd be really great,
but I don't expect great things of the ranking choices of low-info
voters like we have a lot of in the US. Does it matter the ratios? Cuz
I really like 3:1 multi-winner and single-winner.
The ratios matter to some extent, which we can see by considering the
extremes. If you have infinity PR seats to each single-winner seat (i.e.
no single-winner seats at all), then you have a PR council. If you elect
only single-winner seats, then you get a bunch of the group closest to
center, starting at the center and winding its way out (if the method is
good, that is).
So the balancing point depends on how much you value single-winner
balance against PR diversity. You could probably do some calculations to
find out to what degree increasing the single-winner share lowers the
probability of small-party kingmakers getting undue power, but
ultimately, you'd have to make a value judgement.
----
Election-Methods mailing list - see http://electorama.com/em for list info
