[EM] List PR

[Forest Simmons]

Excellent insights into Craig's list PR method!

Also see below for a suggested modification inspired by your insights. 

Forest

On Fri, 15 Jun 2001, Bart Ingles wrote:

> 
> I like it.  It looks as though in the non-partisan case (or one where
> there is exactly one candidate from each party), the system equates to
> cumulative voting.  But at the opposite end, where all candidates belong
> to a single party, it's equivalent to approval voting.
> 
> Or at least a multi-winner form of approval voting.

That is a non-PR multiwinner form of Approval voting.

This fact suggests also using the d'Hondt count within each list to decide
which of the list's candidates get to fill the seats won by the list
(party).

Then if there is only one list, the election will reduce to sequential
PAV, maintaining the PR status of the election.

>  But in that case it
> seems like it would be in the interest of at least some of the party
> members to split the party.

What about allowing parties to submit multiple lists with the party name
on them without actually splitting the party?

If the lists are small enough, then it will be feasible to use full PAV
within each list to pick the candidates that will fill the seats won by
that list. 

If there were only one person on each list, then as you say, the whole
election would just be cumulative voting PR.

How about a rule that says no party can have more than five lists in a
five winner election? 


>  I wonder if there's a corresponding
> incentive to merge when there are too many parties?
> 
> 
> 
> LAYTON Craig wrote:
> > 
> > I've been thinking about ways to improve list proportional representation.
> > I generally like STV for public elections, but it has a number of drawbacks,
> > and is not always viable.
> > 
> > List PR is very simple, and using a d'Hondt count will result in better
> > proportionality than any other method (aside from methods with variable
> > voting power).  The problem is always the method of deciding which
> > candidates from each party get elected.
> > 
> > This is my first thought - it is a free list method, where voters can mark a
> > single box corresponding to a single party.  A regular d'Hondt count is used
> > to determine how many seats each party is awarded.  However, voters have the
> > option of voting "below the line" for candidates, rather than parties.  A
> > voter can vote for as many candidates as they like.  A voter's vote is
> > divided amoung the parties proportional to the number of candidates they
> > vote for in each party.  ie If a voter votes for 2 candidates in party A, 1
> > candidate in party B & 1 candidate in party C, party A gets .5 of a vote,
> > and parties B & C .25 each.  These fractional amounts are added to the above
> > the line (whole) votes to determine how many seats each party gets.  Seats
> > are then awarded to the candidates in each party with the most below the
> > line votes.
> > 
> > So, parties are elected based on plurality/cumulative type voting, and
> > individual candidates using what basically amounts to approval.
> > 
> > It might even be better to avoid the above the line voting totally, and only
> > have voting for candidates, but I've not yet decided on that.  Thoughts?
> 
>