STV for party candidate lists?

[Herman Beun]

[ Contains pretty long introduction you might want to ]
[ skip, but please read questions at the end.         ]

Hello world,

I am a member of a more or less liberal-democratic party (cf. the
lib-dems in the UK) in the Netherlands called D66. Democratic reform
is one of this party's main goals. Though a member of the government
coalition, D66 is not a very big party: currently 9.3% of the seats
in parliament (PR, party lists), however, it was bigger during the
last parliamentary term (16% of the seats). This important loss is
one of the reasons why I think the time is ripe for a democratic
reform of this democratic reform party ;-) and I'd like to hear this
list's opinion about an idea I have.

As I said, the Netherlands have a PR party list election system.
Voters have only one vote, which _can_ be given to a specific
candidate, but is only counted as such if the candidate receives more
than a certain quota of votes. As a result, 99% or so of the MPs owes
his/her place in parliament to his/her place on the party list,
instead of to preferential votes. So, the order in which candidates
appear on the lists is extremely important. 

Now, most Dutch parties have closed committees to make up their list
of candidates for parliamentary and other elections. D66 doesn't:
this party lets the ordinary members make up the list by means of an
election by mail on a one (wo-)man, one vote basis. The system used
is Borda: D66 members make a preference list with maximum N names,
and candidates receive N points for each first place on a preference
list, N-1 points for each second place, etc. The final candidate list
as it appears in the parliamentary elections has the candidates
sorted according to the number of Borda points they received in the
internal election.

The Borda system has two main disadvantages:
- it is extremely centrist; councils chosen according to Borda show a
much narrower range of opinions than the voters do. As a result of
this centralising tendency, the best strategy to get elected under a
Borda system is "say nothing and oppose no-one". This is exactly the
behaviour we see in internal candidate elections, and, as a probable
result, also among D66 representatives elected in councils and
parliaments. Another result of the centralising effect is that
minority views and new (little known) candidates have no chance of
being represented, even if they are supported by a considerable
proportion of the voters.
- it is sensitive to manipulation or 'insincere voting': voters do
not make lists of their real preferences, but anticipate on other
voters' behaviour by placing candidates they suspect have few chances
higher than they would have otherwise. The non-linear (feed-forward)
effect this introduces makes elections less representative than they
should be.

An alternative preferential system like Single Transferable Vote (the
system used for the Australian senate and the Irish parliament) would
do much better than Borda in both these respects. However, STV
normally does not produce a ranking order of the candidates; after
all, in principle each candidate is finally elected with a number of
votes that is exactly equal to the election quota. That is not a
problem when you elect a council with a fixed number of seats, but it
is when you use it to draw up a candidate list for a general
election: then you _don't_ know the number of seats your party is
going to occupy in advance.

Doing an STV count to select the candidates, followed by ordering
them according to the Borda system is not an option: Since some 2/3
or so of the candidates finally appears on the candidate list (but
only a small fraction of those finally ends up in parliament of
course), the result would still be very much Borda and very little
STV. No, I need an internal voting system that provides candidate
lists which _always_ give an STV composition, no matter if the top 3,
the top 14 or the top 24 of the list is eventually elected in
parliament.

I have been thinking about this problem, doing a few simulations with
Robert Loring's election simulation program 'PolSim', and my
preliminary conclusion is that probably the best way to solve this is
by using the order in which STV eliminates one chanceless candidate
in each round. In this way, each candidate is subjected to the STV
shuffling of votes as long as possible: that's why the (at first
sight more obvious) order in which STV assigns seats in each round is
not the right choice. 

My preliminary conclusions then: Electing party lists of candidates
by STV can be done by performing an STV count for a (seemingly) one
seat election. The order in which candidates are eliminated (one per
round) can be used to rank the candidates on the list: the first
candidate eliminated comes last on the list, the second candidate
eliminated comes second last, etc.

My question (finally!): Do you agree with this reasoning? Is there
anyone on this list who knows if this system is used by any political
party in the world? Experiences? Or is this completely weird? If so,
what flaws are there in my reasoning? Are there any alternative
preferential vote counting systems solving my problem equally well
which are used by political parties elsewhere in the world?

Regards,

---------------------------------------------------------------------
Herman Beun                                                    Arnhem
http://home.worldonline.nl/~chbeun/                        Gelderland
CHBeun at worldonline.nl                                       Nederland
                                                                   EU
**** Representative democracy is a contradiction in 4 year terms ****
---------------------------------------------------------------------