Simple question, simple answer. Use lists between parties (or other
groupings) and candidate ranking within them. Open lists try to
implement proportionality within the lists in one quite primitive way.
Use of candidate ranking within the parties allows us to offer also
proper party internal proportionality.
Candidate ranking based methods are not very good in elections with
numerous (e.g. 100+) candidates. Combined list + candidate ranking
based methods could limit candidate ranking to candidates of one party
only. That would keep the counting rules simple and keep the filling
of the ballot easy enough (not too tedious).
In environments where the political structure is party based the
dividing lines between parties are often quite strict. Elected
representatives tend to work with and for the party. In this kind of
environment clear dividing lines between parties may be what one wants
to have (to give a clear picture to the voters) and there is maybe not
much need for the ability to cast ballots that rank the most liked
candidates from multiple parties. This approach also reduces party led
vote management.
Within the parties there may be interest _not_ to display the dividing
lines to avoid too clear fragmentation of the party to competing
subgroups. For this reason candidate ranking might be a good option
within the parties.
Since the assumption is that voters rank only candidates of one party,
ballots that will be exhausted (e.g. in a STV process) can be
inherited by the party. This means that short or bullet votes are not
a big problem. This further makes this approach more viable in
environments with numerous candidates. One could also accept limiting
the maximum number of ranked candidates. The ballot could contain e.g.
only three slots without causing any major problems to in the method.
And in any case this would be better than the one slot only approach
that open lists implement. Note that this approach could be used with
ballots that contain e.g. three boxes where voter may write the
numbers of her three favourite candidates (of her favourite party). Do
you think this kind of limitations (three slots + inheritance to the
party) work well enough in practical elections?
The key idea of methods like this is to implement proportionality both
between and within parties in environments that have multiple
candidates (and maybe also multiple representatives and parties). This
approach can be seen to be an improvement of open lists to offer
better proportionality within the parties. This approach can be seen
to be an improvement of candidate ranking based methods in the
direction of allowing them to be used in elections that have a high
number of candidates (while limiting the ability to rank candidates
from several parties). The size of the ballots could be kept
manageable by using party specific ballots that list all the
candidates of that party, or neutral (party and election (year)
independent) ballots that do not list the candidates but allow the
user to identify the most preferred candidates.
A simple implementation first counts the number of votes of each party
and allocates the seats to them (as in party list based methods), and
then uses e.g. STV to distribute the allocated seats independently
within each party.
Juho
P.S. Also a tree based structure (maybe to group the parties) could be
used instead of a flat list based structure, although this approach
can also be used instead of a (party internal) tree structure. One
could also combine the tree and candidate ranking based approaches
within the parties so that voters would be allowed to rank subgroups.
A vote could be e.g. "Candidate_A > Candidate_B > Subgroup_Z". One
more approach would be to allow ranking of candidates of multiple
parties and use the default inheritance to one party only if all
ranked candidates of a ballot are from the same party. This would be a
candidate ranking based method with automatic or indicated inheritance
rules (to a party / subgroup / tree branch) where practical. I'll skip
the details of possible counting rules for now.
----
Election-Methods mailing list - see http://electorama.com/em for list info
