[EM] Changing the Norwegian election system

[Juho Laatu]

On 4.11.2013, at 0.27, Vidar Wahlberg <canidae at exent.net> wrote:

> I've taken on an ambitious endeavour; Attempting to initiate a change of
> the Norwegian parliament election system.
> The chance of achieving anything probably isn't that great, but I do
> have some political connections that have shown interest and wish to
> meet me in order to discuss it further.

Good luck with the effort. The Norwegian system is one of the best, so even if there will be no changes, that’s no problem. Of course one can always make a good system even better.

> 
> This is where you come in. I'd like you to point out any flaws and
> weaknesses in the modifications I suggest.
> Suggestions are welcome, but do note that radical changes, such as
> changing the way we vote and count votes likely won't be well received
> (unfortunately). The idea is to improve the voting system as much as
> possible, while keeping the changes as transparent as possible to the
> voter.
> 
> 
> == Our current election briefly explained ==
> 
> = Representatives per district =
> In total there are 169 parliament representatives. These are allocated
> to 19 districts using Sainte-Laguë based on "population +
> area(km²)*1.8”.

Using area to allocate seats is unusual, but I guess Norwegians are used to that. Strange, but quite ok, maybe even wanted.

> Currently the amount of representatives per district
> ranges from 4 to 19.
> The amount of representatives per district are recalculated every second
> election (every 8th year), based on the population at the start of the
> previous year (district representatives for the 2013 election was based
> on population 1st january 2012).

You could do this also for every election and with fresher statistics, but also the current approach is perfectly fine.

> For more details I created a simple spreadsheet, it's written in
> Norwegian, but it's probably understandable for non-Norwegians. It's the
> first table, the second table I'll address later:
> https://docs.google.com/spreadsheet/ccc?key=0AsOt7wZ7dVTzdFFZLUpRbkp1eHhxand3aWlTT0xqU1E
> 
> = Casting votes =
> We're using a party-list system, each voter have one vote to be cast on
> one party. Each district have their own set of party lists, the voters
> choose their preferred party and may exclude candidates from the ballot.
> To my knowledge no candidates have ever been excluded as 50% of the
> voters needs to exclude the person from the ballot for hir to be
> unelectable. It is already decided that this is to be changed so the
> voters can influence the candidate order in a greater degree, but the
> order of the candidates is a fairly isolated system that I don't intend
> to suggest modifications for this time.

Do you mean that the voters will be able to influence the order of the candidates more in the actual election (i.e. not when forming the lists)? I’m used to an open list system where there is no given preference order of the candidates, so I find that approach appealing. Open lists are also not perfect, but I like the idea that voters will decide who will get the seats instead of the party (or party members before the actual election when forming the list). I think open lists could be improved by using a ranked method within each party (for better party internal proportionality and to make it “safer” to vote also the weak but liked candidates). But if Norway wants to keep the current approach or maybe improve it by giving voters more say on the order, that’s probably fine for Norway too.

> In each district there usually are between 10 and 20 parties to choose
> from, which in this year's election resulted in 8 different parties
> winning parliament representation.
> 
> = Electing the representatives =
> Of the district's representatives, all but one are elected directly by
> the votes in the district using modified Sainte-Laguë (first divisor is
> 1.4, then normal S-L follows). These representatives are called
> "distriktsmandat" in Norwegian.
> The last representative in each district (making 19 in total) is
> reserved for the parties that reached the election threshold (4%), but
> so far have received less seats than they would if all the 169 seats
> were distributed in one district based on the national vote result.
> These 19 representatives are called "utjevningsmandat" in Norwegian.
> A rough explanation on how the last 19 seats are distributed is that the
> first seat goes to the party in the district that was nearest to winning
> a seat. District now got all its representatives and is excluded from
> further seat allocations. If the party reach the amount of seats they
> would've won in a one-district election it too is excluded from further
> seat allocations. This is repeated until all the 19 seats are allocated.
> Once elected, there's no difference between "distriktsmandat" and
> "utjevningsmandat", they are equal in power and influence.
> This page explains very well how leveling seats are distributed, but
> unfortunately it's only available in Norwegian:
> http://www.valginfo.net/index#/utjevningsmandater

I favour full proportionality. This system may also treat different parties and different regions in different ways. I will not dive into the details of possible solutions. I just note that the simplest modification could be to increase the number of utjevningsmandat seats.

> 
> 
> == Some notable problems with the current system ==
> 
> = Election threshold =
> This makes it difficult for new parties to challenge the established
> parties. Arguably worse is that even though we have 7-8 significant
> parties, they tend to form "blocks", and these "blocks" are given a
> great incentive to attack parties near the election threshold in another
> "block" as that will result in several seats being distributed to the
> other parties. In 2009 a party fell just below the election threshold,
> with no election threshold the party would receive 7 seats, but instead
> they only received 2 seats (the election threshold is not absolute, a
> party can still win seats directly, but is excluded from competing for
> leveling seats).

Also here I favour full proportionality. I can’t see how having one or two small (possibly extreme) parties represented in the Stortinget would cause any damage in a stable country like Norway. Better to hear all the voices and not let them radicalize because of not being heard.

> 
> = Different vote weight and overhang seats =
> Votes weight differently across districts, a vote in Finnmark weighs
> significantly more than a vote in Oslo as Finnmark gets more
> representatives than Oslo relative to the population. As there also are
> a limited amount of leveling seats it's not unusual for the larger
> parties to receive overhang seats. In the 3 elections we've had with the
> current system this happened twice, in 2005 and 2009.

Yes, fixes needed. The Finnmark question depends on if people want to give also the "area covered by the voters” some weight. Most countries don’t think this way, but it’s up to Norway if they want to be special in this respect.

> 
> = Arbitrary allocation of leveling seats =
> The way the leveling seats are allocated means that the final seat must
> be given to the district that haven't yet received it's leveling seat.
> In 2005 this resulted in a party winning 1 of 5 seats in a district
> where they had nearly no support (2.2% of the votes). At the same time
> that party had very good support in several other districts, but
> slightly worse than other parties, which happened to have notably
> greater support in the district where the party with 2.2% of the votes
> won a seat.

You can create a more balanced algorithm to implement exact national proportionality, but that would be a bigger change than e.g. just increasing the number of utjevningsmandat seats. All algorithms will have some "rounding errors", but they cn be made smaller.

> 
> = Few representatives in certain districts =
> Some districts have as few as 4 representatives, yet there still are 5-6
> significant parties in the district. A party needs a significant larger
> portion of the votes to win a seat than in districts with more seats.
> This gives the voters an incentive to vote for a party more likely to
> win a seat, so the seat won't go to a third party instead.

Small districts have some benefits that you may want to keep. If so, the problem of small parties not getting any representatives and voters not voting for them even if they support them can be maybe fixed by having good national proportionality. If district A will not get any party X representatives, but neighbouring district B will, then those representatives can (informally) represent also the party X voters of district A. Voters of district A can safely vote for party X if they know that their vote will not be “wasted”, but will lead (with full weight) to election of more candidates of party X in other/neighbouring districts.

> In addition,
> since area is included in the calculation, a large district receives
> more representatives than a small district even though the population is
> identical.
> In the parliament election people tend to be more loyal to the party
> than the candidates, so this is may not be a very significant issue.
> 
> = Wasted votes =
> A vote to a party that didn't win a seat is a wasted vote. With no or a
> very low election threshold, this problem is reduced.

Yes, no threshold, and full national proportionality.

> 
> 
> == Suggestions I believe may be achievable ==
> 
> = Remove or greatly reduce election threshold =
> Norway is a stable democracy, capable of forming a government to both
> sides of the economical axis. With no election threshold we would see
> 2-3 additional parties in the parliament for the last 3 elections making
> up 3-4 seats. With a 1% election threshold only one additional party
> would be represented with 2 seats.
> Such a change would make it more possible for new parties to challenge
> the established parties, there would not be an artifical border that
> gives blocks a strong incentive to attack parties near this border in
> order to "free" multiple seats, and the chance of wasting one's vote by
> voting on smaller parties is reduced.
> This change alone will likely cause more overhang seats.
> 
> = Replace current apportionment with biproportional apportionment =
> This would make all votes weight equally, remove overhang seats and
> distribute seats optimally to where the parties got the most support.

This sounds like a big change. But maybe it is not. Maybe it is a small change in the sense the although the algorithm is very different, from the outside the systen looks pretty much the same to the voters and even to the politicians and candidates.

> This system also cause the election threshold to be absolute, which
> makes it even more important that the election threshold is reduced.
> 
> = Amount of district seats based on degressive proportionality =
> Instead of using Sainte-Laguë based on population and area to calculate
> amount of district seats, a degressive proportionality based purely on
> population may help prevent certain districts from receiving too few
> representatives and ensure that two district with approximately the same
> amount of inhabitants receive the same amount of representatives.
> The second table in the spreadsheet I referenced earlier shows a
> district seat allocation using a degressive proportionality, where
> Sainte-Laguë (divisor = 2s+1) is replaced by a parabolic function, such
> as: divisor = s²+1. The parabolic function likely has to be tuned, this
> only serves as an example on how the increase in population also
> increase the amount of inhabitants per representative. The European
> Union use a degressive proportionality system for their parliament
> members, although they are not using a mathematical function.
> I'm not entirely sure about this modification yet. Without
> biproportional apportionment (or more leveling seats) it will increase
> the differences in vote weight, even with biproportional apportionment
> you'll get more cases of a party winning more seats than a party with
> more votes, and the problem this modification is designed to solve may
> not be a significant problem.
> 
> 
> == Some other suggestions, likely too radical ==
> 
> = Two votes per voter =
> Doubles the effort of voting and counting votes, but serves as a crude
> cumulative voting system which arguably is more expressive than a
> non-preferential single vote systems. No changes to the ballot
> necessary, each voter just cast two ballots. Useful for those who are
> split between two parties, but most people will likely just give two
> votes to the same party.

You could also vote for different parties in different elections, or flip a coin. Maybe a plain and simple system would be better.

I mentioned above the possibility of using a ranked vote, maybe separately within each party. This would solve some problems related to choosing which candidate to vote for, or voting for candidates that one likes but that are not likely to become elected. But I’m not sure how much you want the voters to be able to influence which candidates of certaing party will be elected (and if you want to do this on the ballot or somewhere else, like when deciding the order of candidates on the lists).

> 
> = Specify 2nd preference on ballot =
> Only useful for voters of party that likely won't win a seat. If the
> party doesn't win any seats, the party is excluded from the election,
> votes are transfered to the 2nd preference and a new round is
> calculated.
> Requires modifying the ballot and makes it more difficult both to vote
> and count votes. If election threshold isn't removed or reduced, a
> change like this becomes more important with biproportional
> apportionment (because of the absolute election threshold, which will
> make it even more difficult to challenge established parties).

This would make it easier to vote for some very small parties. This is also already a ranked ballot, and is therefore linked to the possibility of allowing voters to rank candidates (instead of parties). Already two or three slots could do the trick for candidate ranking.

There can be also other approaches, like allowing the parties to form groupings where e.g. a “wasted” vote to a minor left wing party will be counted for the other left wing parties. But parties may be unwilling to show support to each others and form permanent groupings this way.

Note that ability to vote for multiple individuals (=give support to groupings within a party) makes it also possible to vote for "the right wing of a party” or "the northern part of the district” (maybe within party X), thereby making the system more proportional.

That’s all. Just first observations. Hopefully something useful in some of them.

Juho


> 
> 
> -- 
> Regards,
> Vidar Wahlberg
> ----
> Election-Methods mailing list - see http://electorama.com/em for list info