On Aug 13, 2011, at 11:31 PM, Greg Nisbet wrote:
On Sat, Aug 13, 2011 at 6:21 PM, Dave Ketchum <[email protected]
> wrote:
Glad to see thinking, though we part company on some details.
On Aug 13, 2011, at 5:25 PM, Greg Nisbet wrote:
All current forms of party list proportional representation have
each voter cast a vote for a single party. I say this is inadequate
since a small party can be eliminated and hence denied any
representation (this is particularly relevant if the legislature has
a threshold). However, votes for a party that doesn't have
sufficient support to win any seats in the legislature are simply
wasted. Thus I propose an alternative method.
That some party may get zero seats, that does NOT make their attempt
a pure waste:
. If they are growing, they are on the way - and a warning to
other parties that their apparent goals deserve more attention -
perhaps to be honored by those who do get seats.
Under this system, we would in fact see greater support for small
parties since it is less of a gamble. Even IF my first choice
(probably a niche party) does not get a seat, my vote will be
eventually transferred to a party that *does* have a seat. This
means that I'm more likely to support my first choice to begin with.
(This isn't fool proof though in the original formulation ...
ranking other parties at all increases their weight which helps them
compete against my preferred niche party, I don't think this is a
huge vulnerability though and it can be solved by allowing greater
flexibility in rankings).
I read this as following the IRV approach that requires going back and
rereading ballots to do such transfers. MANY of us se this as failing
too often. We argue for the Condorcet approach that reads ballots,
ONE time, into an N*N matrix for analysis. Since parts of a district
such as precincts can be read into matrices, then to be summed
together, there is more opportunity for encouraging, and checking on,
quality of counting.
Looking closer, winners do not have to be first choice - they simply
need to be ranked above enough of their competition.
I would base the voting and counting on the ranking we do in
Condorcet for single seats - same N*N matrix and whoever would be CW
be first elected, with next the one who would be CW if the first CW
was excluded.
. If the above could elect too many from any one party, exclude
remaining candidates from that party on reaching the limit.
. Note that the N*N matrix has value that does not often get
mentioned - it is worth studying as to pairs of candidates, besides
its base value of deciding the election.
I'm sure I don't have to remind you a Condorcet Winner does not
always exist. I don't completely understand your description of your
method. How does it work with parties?
Condorcet methods accept that three or more candidates may be a cycle
rather than one being a CW - and have to accept responsibility for
deciding which cycle member shall be, effectively, CW.
Seemed simple to treat parties, rather than persons, as candidates. I
thought of parties being allowed to fill more than one seat and, for
this, wanting to have multiple candidates such as G1, G2, and G3.
Even with this, seems like voters would want to identify the person
holding a seat even if the seat's existence was identified with the
party.
Each voter votes for as many parties as they wish in a defined
order. My vote might be democrat>green>libertarian>republican or
something like that.
Anyway, first we calculate each party's "weight". Weight is
calculated simply by counting the number of times the party appears
on a voter's ballot in any position (this should be reminiscent of
approval voting). Each party also has a status "hopeful", "elected",
or "disqualified".
Next, pick your favorite allocation method. D'Hondt, Sainte-Laguë,
Largest Remainder, anything else you can think of, with or without a
threshold.
We then use this allocation method to determine each party's mandate
if everyone voted for their first preference. If every hopeful party
has at least one seat, then all the hopeful parties are declared
elected. If at least one hopeful party has no seats at all, the
party with the lowest weight is disqualified, its votes are
redistributed, and the allocation is done again with the new list of
hopeful parties.
I see "first preference" and think of avoiding IRV's problems -
which the above ranking attends to.
I am assuming candidates identified with their parties, and parties
getting seats via their candidates getting seats. Thus, once all
the seats get filled, remaining parties - due to their lack of
strong candidates - get no seats.
My system does not have voters voting for candidates at all. In
fact, candidates needn't even exist (theoretically of course) for my
method to be well-defined. Instead people simply vote for parties,
with parties that can't get any seats dropped from the lowest weight
first. Making the system more candidate-centric could be done, but
my algorithm (or class of algorithms) is supposed to be a minimal,
easily analyzable change from non-preferential party list methods.
This method has some advantages over traditional systems. People
would not be motivated to betray their favorite party for fear that
it will lack enough support to win any seats in the legislature and
hence their vote would be wasted. This method can also be slightly
modified into a cardinal method, with a voter's first choice being
defined as the highest rated party on their ballot remaining and
weight being calculated by the arithmetic mean of a party's rating à
la Range Voting. This class of voting method is probably compatible
with MMP, but I haven't yet worked out the details of how that would
work.
----
Election-Methods mailing list - see http://electorama.com/em for list info
