On Apr 28, 2010, at 6:37 PM, Jameson Quinn wrote:
2010/4/28 Raph Frank <[email protected]>
On Wed, Apr 28, 2010 at 4:05 PM, Juho <[email protected]> wrote:
> Do you mean that voters would concentrate on the first rankings and
> strongest candidates? The used method should be such that this
kind of
> behaviour will not be rational.
Yes. If the order of election matters, then your first rank is
effectively for the president's position .. and it is a plurality
election.
Minor note: I proposed using order-of-election for vice president,
not for president.
How about this: Elect the council with STV. Elect the president from
the council with Condorcet. Elect a two-member subset of that
council with PR-STV. Any members of that two-member council who
aren't the president are vice presidents.
It gives a variable number of vice presidents. However, it seems
like a very fair all-around system, and needs no innovative new
methods.
If one uses the same votes in all three elections or in the latter two
then the result could be quite proportional and quite free of
strategic incentives. This method doesn't have the burden of keeping
the president included in the elected "P+VPs" set (that is an
"innovative new method"). But as a result the number of VPs may vary.
If the president is not included in the VP set then the president is
probably a compromise candidate from a small grouping. That causes
some distortion in proportionality of the P+VPs set, but on the other
hand I understood that there is also a strong interest to elect a
centrist president and therefore this solution may be preferred to
full proportionality. (Also the method where the president was forced
to be included in the (fixed size) P+VPs set has this property.) We
may thus not want full proportionality in the P+VPs set if we can find
a good president "outside of the few leading groupings".
Or, if you elected a 3-member subset, I suspect it would be very
rare that the president was not in that subset. If she wasn't, and
if 3 VPs were too many, you could then repeat the STV to choose two
of those 3, or let the board elect 2, or let the president pick 2,
or eliminate the Condorcet loser among those 3.
We are now sliding back to the world of "innovative new methods". I
think none of the solutions is perfect (the first one is maybe the
best of them). But if one wants an exact number of VPs then something
must be done to reduce their number by one (or add by one). One more
approach would be to use STV to pick either two of all the candidates
depending on if the president is included in the set of three or not
(one needs however an additional rule on what to do in the rare case
that the president is included in the two but not in the three).
(I still like my RBV method, and would still be willing to code it
open-source if the Czech greens are interested. But I understand if
they want something more proven.)
I didn't form yet any strong opinions on the RBV method. Is
monotonicity the target that makes you like it more than STV?
Juho
Jameson Quinn
----
Election-Methods mailing list - see http://electorama.com/em for list info
