[EM] Single-winner method to PR, part II

[Stephane Rouillon]

Dear John,

> approval and ranked pair are not as described proportional methods,
> but you are right they could be used as inner motor to produce
> proportional methods.
> However, you have to be careful about how you do this.
>
> The key into converting a single-winner method into a motor for
> a proportional multiple-winner method is to make them able to produce
> weights as an output. Why? Because of fairness issues: first a system
> needs to be fair between parties, next between candidates and finally
> between voters. It is one thing to have a single-winner method that
> avoids vote-splitting, does not favor cloning and gives an equal power
> to every voters
> (wheter approval fits or not is not the debate here), it is another
> matter to preserve those
> qualities while
> building your proportional model using such a single-winner method.
>
> For examples, take approval and let`s use it as you said to build a
> proportional model. It is
> obvious that it would introduce a bias toward parties with the same
> program or alike.
> Imagine all ridings with 3 parties at first, each with a 33% support.
> Then one of the party
> (Liberal for example) would split into three others over a detail,
> presenting 3 candidates
> for each riding. Instead of approving only my liberal candidate I would
> approve all three
> candidates. So the final support toward every party would now be
> measured as 20% for
> every party. Clearly the fractional support
> becomes dependent of the number of candidates using such a method...
>
Now about fairness between candidates. Two main problem rise: first some
particular candidate will be compared against which other candidates of
the same party.
All of them or only the ones running in the same riding? If you want to
compare some
candidate with the other of the same party running in other ridings, you
assume that ridings are equivalent in composition (poor/rich,
urban/rural, language, religion,...) so to compare supports as a measure
of the candidates quality. If you don't you will see some fights to run
in the well-known fortress of a party: bias again (this already exist
with FPTP). Some will say ok let's just compare with the candidates of
the same riding, typically STV does that.
Now the comparison is fair, but still the candidacy fight still creates
a problem. In Ireland,
political scientist report huge attempt not only to get a specific
riding, but especially to have bad co-runners.In other words, with STV,
running in the same super-riding than your party leader, you need your
party to elect two seats in order for you havind the second... So your
election is favoured if the other candidates of your party in the same
riding than you are bad, and you are disfavoured if they are all the
good candidates of your party.
So fairness between candidate needs ridings equivalent in composition so
to compare
between ridings. One way of obtaining this is a sampling technic, that
would use equivalent
sample of the electorate instead of geogrphical ridings. In SPPA, I
suggested to use birth date (day, month, modulo of the year) to build
such virtual ridings. Another way is using one single huge riding, but
then the candidate list is humonguous and voters are less interested...

Finally, fairness between voters. This is mainly assured by the
single-method itself.
Wether(?) you think Condorcet favours center-voters or FPTP favours old
parties or Cardinal Rating favours extremist parties, the proportional
system resulting will be as fair as the original single-method. For
example vote-splitting problems of FPTP would end up
into a very fractured representation in the house. The other fairness
issue to ensure is that every vote counts with the same weight, so
ridings (virtual or not) should be made with equal population size...
Again SPPA fits. Finally such a system can elect several MP's in some
riding and none elsewhere. Some voters would perceive this as unfair. I
think they would be rigth in the case of geographical ridings? Do you?

Steph.
PS: If you are interested SPPA on fairvote canada's site deals
with such issues...
http://www.fairvotecanada.org/phpBB/viewtopic.php?topic=8&forum=1&4




-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20030925/3e5d9068/attachment.htm>