[EM] Competitive Districting Rule

[Juho Laatu]

--- raphfrk at netscape.net wrote:

> Juho <juho4880 at yahoo.co.uk> wrote:
> >
> > > I don't think an electorate like ours would
> tolerate
> > > your proposed re-allocation of some of the
> > > single-district winners to achieve party
> > > party balance.
> >
> > Few remaining comments on this topic:
> >
> > As said, I think it makes sense to look at these
> matters from the 
> point of
> > view
> > of unavoidable rounding errors. In the method I
> presented (which is 
> not one o
> > my favourite methods) we achieved extreme locality
> (single member 
> districts)
> > and
> > very good PR (party based).
> 
> I wonder if a non-party system could be achieved. 
> For example, each 
> candidate
> could be required to rank all (or at least alot of)
> the other 
> candidates.
> Eliminated candidates would then tranfer their vote
> to other parties.

Interesting scenario. The "proximity" of different
candidates could be measured this way. Maybe
candidates would list e.g. their favourite in every
other district. There are different methods to make
the calculations (and I can't collect my thoughts and
name my favourite here :-). 

There are few basic sources of party structure and
proximity related information.
1) preferences given by a candidate
2) preferences given by a voter
3) traditional party structure
4) a more detailed / hierarchical party structure
(party coalitions and party internal subdivisions)
I mention these since any of these could be used in
producing PR, and since I find also the fourth point
worth a thought (don't remember such EMs discussed on
this mailing list). The two first ones build the
relationships algorithmically based on the opinions of
many individuals. The latter two rely on an agreed
structure. The agreed structure is more rigid, which
may be seen as a limitation or as a positive feature
(helps voters in their decision making, and opens up
the game of party internal policy making to the
voters). Note that the granularity of the hierarchical
structure can go down to groupings with two members.

Also combinations of different methods are possible
but that may make the voting system more complex. One
could get also some quite interesting and complex
scenarios by allowing voters to vote anyone, not just
candidates of their own district (geographical PR
could still be guaranteed). I have also nothing
against offering voters and candidates the possibility
of setting preferences that include parties/groupings,
not just individual candidates (except that the voting
methods gets quite complicated and we might often get
practically equally good results with some simpler
mechanisms).

> 
> Hmm, I have a recollection of making a suggestion
> like that before 
> (maybe to
> this list)
> 
> Btw, if someone gets more than 50% in a district are
> they gaurenteed a 
> seat ?

It is possible that some big party gets say 55-60% of
the votes in every district, and they may get a seat
only in districts where they have a 57-60% share of
the votes but not in districts where they have only
55-56% of the votes. The result is fair in the sense
that the best candidates of this powerful party got
the seats and the _relatively_ weak ones were left
without one.

> 
> I guess that an independent who gets 51% of a
> district isn't really 
> entitled
> to a seat.

I tend to think in terms of share of the total votes.
If there are n seats, then 1/n of the votes could
guarantee a seat. Additional rules that limit the
possibilities of small groupings to get seats are
possible but it may be better to see them as a
deviation from the main rule. Independent candidates
are problematic since they don't benefit of votes
given to similar candidates in other districts.
Creating a party or other grouping of the "x type"
independents would be a good idea if they have some
common "x type" targets. 51% of the votes of one
district could well be rounded up to one seat (1/n =~
100% of the votes of one district). At least from
scientific point of view methods that favour large
parties should explicitly mention this in their
definition.

> 
> > One can implement PR nationally (like in the
> method that I described) 
> or
> > independently within each district. National levwl
> PR calculations are
> > typically
> > needed if one wants to achieve maximum PR. I.e. to
> cover also to the 
> smallest
> > parties/groupings that will get only about 1/n of
> the votes (where n 
> is the
> > number of representatives to be elected). PR may
> thus be with or 
> without
> > cut-off. STV based methods don't normally use the
> concept of 
> parties/grouping
> > and therefore counting the votes of e.g. all
> animal protection 
> oriented
> > candidates together (nation wide) to achieve one
> seat is not 
> possible. (and a
> > a
> > result a cut-off (which is one type of a rounding
> error) (or maybe 
> intentiona
> > in some cases) is introduced at some level)
> 
> Also, under your system, some constituency somewhere
> would have to end 
> up
> being represented by a party a party that got hardly
> any votes.

Yes. A party that gets 1/n of all the votes could be
entitled to one seat. They could get evenly about 1/n
of the votes in each district. If n=100, then the one
candidate that will be elected would have only 1% of
the votes in her/his district (probably a bit higher
though in this particular district). What is still
fair in this case is that this particular district
obviously had only some weak large party candidates.

(We could for example have three parties with support
50-49-1, and same number of seats (out of the 100)
would be given to each party. In one district the
votes could go 48-47-5 and both "48" and "47"
candidates would be left without a seat since they got
too poor results (for candidates of such strong
parties). It is possible that there was another
district with vote distribution 50-44-6 where the "50"
candidate got elected. Maybe electing this particular
small party candidate was the best solution to the
problem of allocating the one PR based (=well
justified) seat to them and minimum damage was done.)

> 
> It suffers from the spoiler problem.  A 2nd
> candidate from a party that
> is likely to win could reduce the plurality for that
> party in the
> constituency and force it to be one of the
> constituencies that are 
> switched.
> It would have no effect on the national vote as they
> would both be 
> counted
> as part of the party.

Yes but the method could be changed to something more
appropriate if we so want. If parties are expected to
set several candidates, we could e.g. change rules 3
and 4a as follows.

3) Find the strongest PARTY (MEASURED AS THE
PERCENTAGE OF VOTES ITS CANDIDATES GOT WITHIN THE
DISTRICT) among all the districts that have not
elected their representative yet
4a) Elect THE MOST VOTED CANDIDATE OF THIS PARTY if
his/her party doesn't yet have as many seats as
decided at point 2

I mentioned also hierarchical party structure above.
With such a structure one could determine the
proximity hierarchically so that one would e.g. first
find the district where the coalition of left wing
parties is strongest, and then (after making some
other "stronger" decisions in other districts first)
within this coalition decide which party gets the
seat, and then which grouping within the party, and
finally which candidate was strongest.

> 
> 
> Btw, what are people's views on rescaling of votes
> for a constituency 
> before
> working out the totals to be used for the PR
> calculation.
> 
> For example if the turnout was 50% in a district
> each vote would could 
> for 2
> when doing the national vote and if turnout was 40%,
> each vote would 
> count for
> 2.5 in the national vote.  This means that if people
> don't vote their
> district still has the same effect on the national
> vote.  Also, it
> reduces the benefit of ballot stuffing.
> 
> One party could get 60% in half the constituencies
> and 40% in
> the others, then its average would be 50%.  However,
> if the 
> constituencies
> were had different turnouts, they could end up with
> less than 50%.
> 
> In effect, going to full PR reduces the power of
> constituencies with low
> turnout.  Although, another way of looking at it is
> the PR increases 
> the incentive to
> vote.

Yes, this is quite interesting. These questions must
be answered although their impact may not always be
that big.

On the other hand balancing the votes (by counting
also the sleeping voters in) is a nice thing but on
the other hand I have some sympathies towards methods
that give more power to those who express their
opinions in the elections. In a way this is a very
concrete way of telling voters that it pays off to
vote (maybe better than making voting mandatory or
telling people that voting is their duty).

One could also consider multi-member districts where
the number of seats per district depends on the number
of active voters. (One could however alternatively
balance the scale e.g. so that kids without voting
rights could be counted in even if those adults that
didn't vote would not be counted (maybe the same
percentage of kids as there were active adults).)

Note also that the opinions of those who voted in some
district and those who didn't may not correlate
strongly (e.g. right wing voted in some district and
left wing slept, and the other way around in some
other district). This would speak in favour of not
letting the votes of the persons who voted to
determine the opinion of the others.

(In the case of dynamically determined number of
multi-member district seats above the home area of
those who voted and those who didn't would correlate
very strongly but for other reasons we might still let
the sleepers sleep and not get representatives :-). I
think both philosophies, one where those who are
interested will have representatives and one where all
citizens will have representatives (to guarantee
geographic PR) irrespective if they are active or not
are possible.)


Good comments, let's work forward.

BR, Juho

> 
> 



	
	
		
___________________________________________________________ 
All new Yahoo! Mail "The new Interface is stunning in its simplicity and ease of use." - PC Magazine 
http://uk.docs.yahoo.com/nowyoucan.html