[EM] How to combine list and candidate ranking based proportionality?

[Juho]

On May 22, 2010, at 2:15 AM, Raph Frank wrote:

> The ballot layout issues would still be there.  If the voter is to be
> able to rank all candidates, then you need to have each candidate's
> name on the ballot.

Yes, the candidates have to be listed somewhere. Ballots could be  
shorter if there would be separate ballots for each party. I don't  
like this approach very much though since the arrangements become  
somewhat more complex, the voting process becomes somewhat more clumsy  
and the level of privacy may decrease if the ballots must be picked  
somewhere else than from piles within the voting booth. In the  
"neutral ballot" case the candidates could be listed within the voting  
booth and elsewhere on a poster. Voters would simply write the numbers  
of their favourite candidates on the ballot (that could be very small,  
just big enough for e.g. three numbers). In the neutral ballot case  
voters have to draw clear enough numbers. At least in open lists where  
one has to draw only one number that approach seems to work well. It  
could work well for small number of rankings too.

>
> It would still save time, since you could sort the ballots into piles
> for each party.  Also, coalition negotiations could start once the
> party votes were counted and the seats allocated between the parties.
> The PR-STV part could take longer than normal.
>
> Also, the party votes part would be precinct summable, even if the
> PR-STV stage isn't.

If there is only a fixed number of slots (e.g. 3) and not too many  
candidates per party (e.g. 20) summability could still work also at  
"PR-STV level" (=> about 20^3 piles). Not very elegant, and ballots  
could still be marked by voting for some unusual combinations.  
(Another ugly trick would be to count vote A>B>C as one vote to A, one  
inheritance preference from A to B, and one inheritance preference  
from B to C. This would be quite summable but the inheritance rules  
for B would become somewhat distorted since B's votes would be  
inherited always the same way (not taking into account if B already  
has inherited this particular vote or not). One would need also rules  
for cyclic inheritance.)

>
> Another option is the "candidate-list" system where each candidate
> submits a list and you vote for 1 list.

Yes, that would make the method fully summable. This approach is also  
already close to a tree structure based inheritance (more limiting but  
easy to understand at one sight). (The "ugly trick" that I mentioned  
above would in a way allow the voters determine the "candidate- 
list" (or inheritance tree) of each candidate.)

Juho