[EM] Preferential Party-List Proportional Representation (PPLPR)

[Juho Laatu]

On 07 Nov 2014, at 00:29, Vidar Wahlberg <canidae at exent.net> wrote:

> Even my distaste for election thresholds I do see a potential problem of
> fragmentation, and too many parties.

I once checked the results of some Finnish parliamentary elections from the point of view of eliminating all rules that cut the small parties out. In Finland there is no threshold, but the fact that the results are counted separately for each multi-member district favours large parties. Also the used D'Hondt method has a similar but smaller effect. I counted the results at national level and allocated seats to parties based on their proportional share of votes. The end result was that there would have been one or two additional small parties with one representative. The total number of representatives is 200. I guess that would cause no other problems in Finland than maybe the need to "waste" some time in the parliament, listening also to the opinions of these parties that anyway do not have the strength to influence decisions much. The benefit would be that this would make the party structure more dynamic by allowing new opinions to get in when there is a need. It may be that the number of small parties would grow a bit higher after the voters realize that also the smallest parties have a chance of winning a seat.

> Your example of accumulating votes
> from the previous election to reduce "lost" votes is intriguing. My own
> thoughts on the subject has been preference IRV-style, meaning that
> votes to a party which doesn't meet the election threshold goes to the
> voter's next preference (repeat process until all remaining parties are
> above the election threshold).

That sounds like a realistic approach (probably more realistic in real life than my accumulated votes).

I note that your proposed method (PPLPR) does also more than that. It is not proportional in the traditional sense that the elected representative body would represent the opinion space of the voters in miniature size. It allocates more power to the centrist or compromise oriented groupings. That is an interesting approach. Typically proportional metods simply give each grouping representatives that most accurately represent the political opinions of that grouping. Many single-winner elections on the other hand try to elect a compromise candidate. Your method extends this ("balancing of the single seat to avoid strong bias in some direction") also to multi-winner elections. That could be called a new category of methods.

> You have one election for an ideology
> (party), and then you have another election for which person within the
> ideologies that should be elected.

I come from an open list background. The party/candidate ranking based approach is valid in open lists too to avoid lost votes to parties that will get no seats. In open lists there is also another need, the need to influence which individual will be elected within your favourite party. And similarly your vote may be partially lost if the candidate that you voted for will not be elected (the party will however get the vote in any case, so it is not completely lost). The traditional approach of electing those candidates that got the highest number of votes within a party also does not respect proportionality within the party. It is quite possible that the votes of some large opinion group within a party spread to multiple candidates and none of them get elected.

The technical approach in this case could be to allow the voters to rank multiple candidates. Within one party this could lead to some STV style system where similar minded candidates inherit support from each others. This makes the method proportional also within the parties, and it is easier for the voters to vote for their favourite candidate even if they feel that this candidate doesn't have good chances of becoming elected this time.

This approach is also related to the "clear opinion space building" role of parties. There can be hundreds of candidates, but the voter need not rank all of them. Even one will do since a vote to a candidate will be inherited by the party anyway. This means that short rankings do not lead to lost votes. And the role of parties remains strong in the sense that they give voters some clear alternative overall directions, and make it clear where each candidate stands (without making excessive studies on that, and cutting out some sweet advertising that is intended to please all possible voter groups (also those that do not actually like the ideology of this party at all)). I also note that existing parties are more likely to accept systems that maintain the party structure than systems that try to get rid of it (just a practical concern).

The description above covered only ranking multiple candidates within one party. If the voter ranks also candidates of another party, such votes could be interpreted so (close to the PPLPR style) that if the favourite party will not get any seats, the vote goes to the next party and thereby will not get lost. (One could transfer also fractional votes (probably only leftovers from the party), but that may be already too tedious when compared to the achieved benefits.) This approach didn't offer additional seats to the compromise parties in it basic form, but it could in principle be extended also in that direction (in PPLPR style).

Juho