2012/6/7 Juho Laatu <[email protected]> > On 7.6.2012, at 5.21, Michael Ossipoff wrote: > > Sainte-Lague isn't the only PR formula that is unbiased with respect to > party-size, but it's the only unbiased formula that doesn't share the > avoidable errors of STV and Largest Remainder. > > > Largest Reminder has some paradoxes but I wouldn't call them errors. > Usually those properties can also not be exploited as strategies. The > possibly surprising seat allocations in the Alabama paradox can be said to > be fair and not problematic. > > Also, I like emphasis on party platforms instead of personalities and > hairdo, etc. > > But I recognize that many would like to vote for individuals, even in a PR > election. Of course that can be done in open list systems, such as those in > Belgium, Luxembourg and Switzerland. The best of both worlds: Optimal > proportionality and opportunity to support individual candidates. > > > Some people support the idea of reducing the role and power of the parties > to the minimum. They may like STV since in it the party stucture is not > binding but just possible background information. You refer to party > platforms as an alternative to electing representatives based on > personalities and hairdo. I symphatize also that since it is also good if > the political system is stable, simple to the voters and also binding to > the elected representatives. In the party / grouping based approach regular > voters will in some sense know better than what they will get than in a > system that is based only on the smiling faces and smooth talk of the > candidates. I think in an ideal system we need a good balance between these > needs and different directions of interest - ability to influence on which > individuals will be elected, and having a clear political map and > directions available to the voters. > > Open lists typically have the problem that within the party there are no > guarantees that different wings of the party will bet the correct > proportional number of seats. The methods may approximate this to some > extent, but we could do better too. It is for example possible to combine > open lists and STV by allowing voters to rank candidates within the party > list. (Abiliy to mix and rank candidates of different parties is lost, but > this may not be a big problem.) One step more party/candidate oriented (but > more informative approach to the voters) is to use candidate given > preference orders. >
PAL representation blends this. Within a party, candidates start out with their explicit number of votes, but as candidates are eliminated bottom-up, votes are shared by that candidate's in-party approvals. Thus there is a mechanism for candidates to ostracize other candidates from their own party, for factional or other reasons, but mostly it's an open-list system. Also the apportionment of PAL is interesting. It starts out with a Hare quota, and fills as many seats as possible. When that runs out, it retroactively changes that to a droop quota, and goes through appointing the same seats in the same order. Thus, the final leftover droop quota includes leftover [Hare minus Droop] slices from all the Hare quotas originally present. This last seat is therefore likely to go to a centrist party (small or large); that is, a party that comes high in many candidate's preference orders outside their own parties. > Yet another possibility is to use a tree structure to divide the party > into smaller subgroups. Trees are very informative and already quite > binding to the representatives. Some parties may find them even too > explicit since they may emphasize fragmentation within the party and they > may give the voters too much (from party leaders' point of view) power on > what policy the representatives will drive during the next term (no chance > to change opinion in line with what the party leaders say if one was e.g. > the candidate of the pro-nuclear-power grouping). > > I'd > thought that Finland had open list, but Juho says that they don't. > > > Finland uses open lists. The seat number of each district is calculated > before the election based on population and Largest Reminder. Within each > district (of different size) the seats are allocated to the parties using > D'Hondt. Within the parties candidates with highest number of votes will > get the seats. > http://en.wikipedia.org/wiki/Finnish_parliamentary_election,_2011 > > Juho > > > > > > > ---- > Election-Methods mailing list - see http://electorama.com/em for list info > >
---- Election-Methods mailing list - see http://electorama.com/em for list info
