[EM] Another PR method based on ranked ballots

[Stephane Rouillon]

James Gilmour a écrit :

Stephane wrote
> James Gilmour a écrit :
>
> > No.  The largest single defect of single-member districts is that they usually
> > deny representation to half of those who vote.  At best, they can guarantee
> > representation to only half.  The second largest defect of
> single-member districts
> > is that they commonly distort the wishes of the voters in terms of overall
> > representation.  Gerrymandering is a defect of single-member districts, but it
> > comes third on my list.
>
> SPPA (French acronym) solves both lattest problems (totally proportional and no
> circumscription).

I do not understand the term "circumscription".

It is just another word for district or riding. It's latin origin precises that
the riding has to be a connex region. South Liverpool and North Glasgow
could not make a circumscription even if added it would have the right number of voters...

> About the first problem, Montreal simulation done during the Convention
> of electors,
> november 10th, showed direct support using four different methods
> after rallying when it is used:

I do not understand what the term "rallying" means.

rallying (ralliement in french) means when your preferred candidate is eliminated and you join one of the remaining candidates. At each round of an IRV election, voters who preferred the eliminated candidate get transferred to their next choice: they rally behind a new candidate or to a new candidate (sorry my english is limited).

> FPTP (no rallying) 46%
> MMP (no rallying) 38% - but except to quota, proportional
> STV (after rallying) 46% - semi-proportional
> SPPA (after rallying) 69% - integer optimal proportional

What do you mean by "direct support"?  What are the percentages?

"direct support" is the mean or average individual support received by
each elected candidate. Typically in an FPTP election using two districts, this
is how you measure direct support or individual approbation rate:
Imagine the following results:
district 1: A 34%, B 36%, C 30%
district 2: D 53%, E 47%
B and D would get elected with an average individual approbation of 44,5 %

If you combine direct individual approbation rate and the proportionality error
(the difference between votes and seats for every parties) you have a good
idea of the quality of the method.

If the simulation of STV gave a result that can be correctly described as
"semi-propotional" then all I can say is that it was a simulation of a very poor
implementation of STV-PR.  Actual results from real STV-PR public elections show
that STV-PR can be correctly described as "proportional".  What is "optimal" will
depend on the criteria you have selected to optimise and the definitions you have
chosen to use.

The implementation was Irish STV-Droop quota. But the results are just an instance about one election. It should be better in general, I agree.

> Just read and ask...

> http://www.fairvotecanada.org/phpBB/viewtopic.php?topic=8&forum=1&4

I have read through this description but cannot find the answers to the questions
above.

Integer optimal means there is no better integer solution that minimizes the
norm-1 (sum of absolute values of the difference between seats and votes) of the error. You cannot slice representatives... but exact proportionality can be reached using a proxy representation, so elected officials would get a voting weight not equal to 1, but not far.

I am not in favour of any system that allocates seats to parties.  Political
parties already have too much power over the elected representatives.  The balance
of power needs to be redressed in favour of the voters.

The proportionality in your proportional representation is focussed exclusively on
the political parties.  But proportional representation can be, and in my view,
should be, about much more than just PR of political parties.

I do agree with you. This is why I think giving an equal chance to independant candidates is mandatory. My model does that.
Next, what you say is linked to the degree of rigidity of the party line. This
equilibrium is destroyed when party chief can assign dissidents to suicide candidatures at the next election, by shuffling them on another riding or to the bottom of the list. Removing those elements gives back its freedom to the representative.
Finally, I do not think it is realist to believe that politics can work without parties. If I am well informed, in britain, at the parliamentary birth, candidates were independants (like several other places). Nevertheless, whigs and tories regrouped naturally in order to maximize their chance of managing main issues, while accepting to follow the chief on secondary issues. It is human behavior.
I think electoral system designers have to accept parties as one of the actors of the political scene. We only need to balance their power, protecting the representative of the majority who elected him/her from the party leading minority.

If I have understood your description correctly, when no party wins a majority of
seats, you allocate 50% of the seats to the party that wins most votes.  This may
or may not solve your problem of "shaky governments" by providing them with a
"crutch", but it is not proportional representation.  Practical experience of real
PR systems shows that neither coalition government nor minority government need be
shaky.  That is a function of the political culture, not the voting system.

Not exactly. First, the crutch is an option. If the chief of the plurality party thinks (s)he can form a stable coalition without using the "crutch", so be it and go for a full mandate.
The problem arise when their is no stable governement to form (even with FPTP as you just said to Alex, there is no garantee). Then, AND ONLY THEN, it is acceptable to sacrifice proportional representation to gain
governement stability. Please understand that most FPTP defendors are ready to do this sacrifice before knowing if it is necessary.  I just generalize the Malta's correction for majority reversals to be fair while moving from a proportional result to a functional government.  Actually, on some recommandations of M. Blais (Montreal),  made the correction to 50% or 50%-1 seat instead of 50%+1 seat. It garantees a stable bipartite coalition
instead of a 50+1% majority not that stable...

You advocate a system of electoral "ridings" instead of geographically defined
electoral districts.  This would not be acceptable in the UK (and, I suspect, in
many other countries) where locality and geography are considered important in
representation at all levels of government.

I know. It is the same in Quebec. But this is essentially the answer of parliament members, not the answer of most voters. It depends who will decide what will be the next electoral system. Several specialists told me about using a mixed system with my model replacing national lists...
Would you like it? I would prefer STV-Droop personnally.

Your example was for 10 "ridings" and 10 seats.  The UK Parliament has 659 elected
members.  Under your system would this mean there would be 659 electoral "ridings"
and that the votes would be averaged over all 659 ridings to ensure a high degree
of party proportionality?  If you do mean PR to one in 659 (1/659th), I have to
say that this is not necessary and that it is highly undesirable.  Its political
consequences would be disastrous.  That is not speculation on my part - sadly, we
have the we have the evidence from real elections of national legislatures to
confirm the point.

Of course! If you do not accept to analyse the result and to allow the crutch option in case of instability, the result is disastrous. You need the complete set of the tools to make sense.
I think it all goes on to what is an election.
I think an election is a representation exercise. So we should maximize PR and individual approbation. But when the result is not functional, thus trading PR for stability is acceptable.

James

----
Election-methods mailing list - see http://electorama.com/em for list info