Sweden: election 2010

Note to readers: An excellent discussion is underway in this thread. Thanks to all who are participating.
________

Sweden’s election has produced a parliament with no majority for either main bloc, the incumbent center-right coalition or the Social Democrats. For the first time, an anti-immigrant party, the so-called Sweden Democrats, has won seats.

The BBC compares the Sweden Democrats to their counterparts in Denmark and Norway:

The most important difference between the SD and the others is that, whereas the two Danish and Norwegian parties started out as movements unhappy with everything from crime-fighting to income taxes, the Sweden Democrats have their roots in a racist organisation focused solely on throwing all immigrants and refugees out of the country.

This election also sees the Social Democrats having their worst result since 1921. However, they remain the single largest party, with 113 seats our of 349, compared to 107 for the Moderate Party.

According to preliminary results at the official Swedish elections website, the Social Democrats dropped 4.4 percentage points from the last election down to 30.9%. The Moderate Party gained 3.9 points to stand at an even thirty percent. The Sweden Democrats won 5.7%, an increase of 2.8 over the last election, and giving them 20 seats. The Green Party also scored well, at 7.2%, up 2.0.

Sweden uses proportional representation with a threshold of 4% of national votes or 12% in any given district.

The website indicates that the final count will begin on Monday, September 23. However, on my calendar there is no such date this year.

29 thoughts on “Sweden: election 2010

  1. Well, I don’t know. The ancestor to the Norwegian SD equivalent, Fremskrittspartiet, was at one time the private party of Anders Lange. It’s true, he was angry at crime fighting and taxes, but he was also a warm friend of apartheid South Africa.

    I think the only reason they weren’t actually neonazis at the time (like the SD were) is that Norway and Denmark were invaded during the war. In Norway and Denmark, rabid nationalism and open neo-nazism is just too big of a dissonance.

  2. My website’s Sweden page has detailed 2010 Riksdag election results since last night.

    By the way, I now have election data for Sweden and twenty other countries (including Denmark, Finland, Iceland and Norway) available for downloading in CSV format.

    More on Sweden’s election later.

  3. The seat count which has been calculated by the election authority (val.se) and which is being repeated by the media appears to be wrong. That is, unless they are using a model for predicting how the expat and late incoming absentee vote will turn out. However I doubt this is the case, because the calculation is in line with where the count stood at an earlier stage. All votes will be counted by the end of Wednesday, apparently. Based on raw votes, the present coalition has 174 seats and it is possible that they will get a 175th and thereby an absolute majority.

    Having written that, I took a little time to go back to the 1998, 2002 and 2006 elections, and looked at the differences between the election night count and the final result, as shown on the election site. There was a clear pattern that the more urban parties were somewhat advantaged, and there were indeed clear similarities for each particular party at all three elections. I took the average increase in vote totals for each party and applied it to 2010. For the SD, which has previously been in the ‘other’ pile on election night, I was a little generous and gave it the largest increase that the ‘others’ had had, 2.04% in 2006. For the other ‘others’, I gave no increase, which is roughly in line with 1998 and 2002.

    This is my bold-ish prediction (and may the formatting be with me):

    M 30.1% 107 seats
    C 6.6% 23 seats
    Fp 7.1% 25 seats
    Kd 5.6% 20 seats
    S 30.7% 108 seats
    V 5.6% 20 seats
    Mp 7.3% 26 seats
    SD 5.7% 20 seats
    Oth 1.4%

    A government majority, crisis averted.
    Caveat: the 349th seat looks like it will go to M, but it may go to S. My estimated vote has the final Ste.-Laguë distribution number for M at 8306 and for S at 8300, representing a margin of only 1,261 votes.
    However: The vote evolutions in 1998, 2002 and 2006, taken individually, all yielded the same result.

    I may slightly regret having posted this in a couple of days.

  4. Another caveat is of course the evolution in votes for the SD. If its increase is much higher than estimated, then they may get the last seat.

  5. Another caveat is of course that the SD may get a much larger increase, in which case they may get the last seat. My estimate gave them 20.2935 349ths of the above-hurdle vote. Ceteris paribus, as they say, a 3.15% increase in their vote total from election night would be enough to flip the last seat to them.

  6. The seat totals published on the Swedish Election Authority website are correct – I took the trouble of checking them myself before publishing them on my website – even though they do deviate from a strictly proportional seat allocation; I’ll explain why this is taking place later on.

  7. Yes, thanks Manuel, very sorry about that. As I was just about to post myself, it seems the seat calculation is not wrong. 39 levelling seats have for the first time not been adequate to entirely correct the disproportionalities in the constituency seats. As it stands, the Social Democrats get four more seats than what would be proportional, and the Moderates one, but margins may be tight in some places so the vote evolutions from election night may cause shifts.

    I wish I had gone to a Swedish news site before I wrote all that! Anyway, if my vote predictions come to pass we may have a wrong winner scenario, even if the “wrong winner” in this case happens to be nobody.

  8. Well, let’s hope that the election authority is better at counting and calculating than at copy-editing in English. The Swedish page has the final count starting on Monday, 19 September with the results of all elections expected in ten days (remember that Sweden has local, regional and national elections at the same day, the national result takes priority and we should expect that one on Thursday).

  9. That’s exactly what happened, as I explain here.

    For what it’s worth, on Sunday I did my own projection of Riksdag seats by parties as results came in, using nationwide vote totals until the Election Authority put up seat totals that differed from those I had. However, I quickly surmised some of the parties must have won excess constituency seats, which turned out to be the case (something which, while rare in Sweden, happens quite often in neighboring Norway).

    In any event, this kind of outcome is covered by Chapter 14, Section 5 of Sweden’s Election Act [PDF; in English], which states that:

    “If a party on the distribution of the permanent constituency seats has obtained more seats than are necessary for it to be proportionally represented in the Riksdag, the party and the seats it has obtained shall be ignored when distributing the adjustment seats. This also applies for a party and the seats that the party obtained, if it obtained less than 4 per cent of the votes in Sweden.”

  10. I had a closer look, and using the same “historical method” as above, but with constituency figures, it seems there are at least three seats involving the major parties that are uncertain.

    In Dalarnas län, Mp needs to increase relative to S by 0.413%. Since Mp is the party that usually gains the most after election night, and S is among those that gain the least, this looks very likely. This would free up one seat for the national distribution, and it would likely go to C.

    In Göteborg municipality, Fp needs a relative increase over S of 0.634%. Since Fp is usually only moderately advantaged by the counts after election night, this looks iffy, but it could happen. Nationally, this seat would then probably go to Fp also.

    In Värmlands län, Fp would need a relative increase over S of 0.888% to take the last seat. However, since the local figures do not show that much of a usual advantage, this will probably not happen. If it did, the freed seat would likely go to Kd, with an outside chance for Mp, perhaps V, or even SD if their vote increased by an extraordinary amount.

    If all three seats flipped, M and S would each still have one excess seat. The state of parties would then probably be: M 107, C 23, Fp 25, Kd 20, S 110, V 19, Mp 25, SD 20. There are other seats that may well shift, but these either do not affect the two advantaged parties or involve even more unlikely margins.

    So, there is still a chance for a majority government, but this is not probable. I hereby conclude my short career as a Nate Silver.

    By the way, I agree with the assessment that the Sweden Democrats are of a completely different type than the ‘Progress’ parties in Norway and Denmark. They are slightly more like the Danish People’s Party, but the latter was formed in protest against the traditional buffoonery of the Danish Progress Party* and however right wing it is, its main strategy has been to be a stable partner for governments in return for policy concessions. In Germany, perhaps the DVU is comparable, while the Republicans may be a touch too moderate and the NPD are definitely too extreme.

    * See constant tax evasion charges, strategy shifts, breakaways, and policies like reducing the size of the Folketing by some 80%, or replacing the military with a cannon shooting only a banner saying “we surrender”, in Russian.

  11. Espen, in Dalarnas County the Greens happen to be just 112 votes behind the Left Party, so if they achieved a relative increase of 1.533% over the Social Democrats (and the Left Party) they would take that constituency’s last seat.

    There are two other Social Democrat constituency seats that could switch hands (to the benefit of the Left Party and the Moderates, respectively): one is located in Östergötland County, and the other in Kronoberg County, although both are longshots. As it happens, for my blog posting (see link above), I only considered Social Democratic constituency seats that could be captured by the center-right parties, but the fact is that any extra seat they lost to the Greens or the Left Party would also have an impact.

    By the way, Dr. Lars Alexandersson has just updated his excellent Svensk valmatematik (Swedish election math) page with preliminary 2010 general election results, including detailed Sainte-Laguë quotient tables for every constituency and the entire country; the page is in Swedish only, but the statistics are self-explanatory.

    Finally, I thought the Danish Progress Party wanted to replace the Defense Ministry with a taped “we surrender” recording…

  12. The reason I picked the Greens as the most likely contender in Dalarna is that they usually grow quite a bit faster than the Left between election night and the final result, but you are right that the latter indeed have the higher vote in that district as it stands. So your relative growth percentage is correct for the Greens, while mine mistakenly shows the corresponding number for the Left.

    In Dalarna, the Greens grew by 3.59% after election night in 2002 and by 5.41% in 2006. The Left grew by 2.02% and 3.22%, respectively. The Social Democrats, however, only grew by 1.09% in 2002 and by 1.82% in 2006. These percentages are all a bit higher than those at the national level, but otherwise as expected. By some margin, both the Left and the Greens would overtake the last Social Democrat seat if these patterns reemerged, with the Greens ahead (this must be when I swapped the two numbers before I wrote my last comment).

    In 1998, the increases show the same patterns, but are much smaller, and particularly so in Dalarna, presumably reflecting a much lower level of postal voting. These give the Social Democrats the final seat if applied to the 2010 election night result, but only because the absolute changes are so (unrealistically) small. In any case, we shall see soon whether my assumptions had some merit.

    I think Östergötland may be achievable for the Left if “they” are extremely lucky, but I considered it outside the range of the probable (Kronoberg is even less likely to flip to the Moderates, but all their present seats appear to be safe). I say “they” because the first beneficiaries of the released seats appear not to be the left, but the centre-right. Although, I read in Swedish media that some red-greens may be hoping to lose a few seats on Wednesday since that would save them from having to deal with either the Sweden Democrats or the government other than as opponents.

    Having now googled it, it was a telephone answering machine, but it appears it was official policy only briefly in 1972/73. It may have evolved a bit by the time I read or heard (or remembered) it. Thanks for the link, which I may have first clicked from your excellent website well over a decade ago.

  13. Looking over Dr. Alexandersson’s page, it’s evident that compensatory seats are not given to members-at-large but are distributed to districts. This is based only on where each party has the highest remaining Ste-Laguë quotient, so it’s quite possible for one district to receive more total seats than it deserves. Eg: Västamland and Norrboten are allocated 8 and 9 district seats respectively, but Västamland gets two of the compensatory seats for a total of 10, while the larger Norrboten gets no compensatory seats and so ends up with fewer total seats.

    Has this sort of unequal distribution ever become a political issue in the short-changed districts?

  14. It doesn’t sound like it’s the same. According to Manuel’s site and Wikipedia, each region of Norway has exactly one compensatory seat, so there can be no such unequal distribution of compensatory seats.

    The procedure for actually distributing the compensatory seats between regions in Norway is almost, but not quite, entirely unlike that in Sweden. The only similarity is that a set of quotients are processed in decreasing order; but in Norway the quotients are calculated differently, are compared inter- as well as intra-party, and are subject to the restriction of one seat per region.

  15. The problem with the Norwegian procedure is:
    1) the compensatory seats are awarded one by one
    2) every constituency can only win one compensatory seat

    As a result, the last compensatory seat must fall in the last constituency left open, regardless of the votes of that party in that constituency.
    Manuel calculated it for 2005: the last compensantory seat had to fall in the last empty cell ‘party=Liberals – constituency=Finnmark’; they won the 5th seat there with only 2% of the votes.

    We have the same problematic procedure for some elections in Belgium.

    If you want a fair distribution in both dimensions (the nationwide distribution of seats between parties and between constituencies is fixed), it would be better to use the biproportional apportionment method elaborated by Balinski and Pukelsheim and first used in Zürich.

  16. Bancki, thanks for the pointer to Balinski and Pukelsheim! The chance that the last compensatory seat would be allocated strangely is exactly what I was worried about with the proposed Québec electoral reform. It’s good to know that there exist proposals for dealing with this.

  17. From 1989 to 2001, the procedure used in Norway to distribute party adjustment seats among constituencies was essentially identical to that of Sweden: mandates were allocated at the constituency level on the basis of the largest Sainte-Laguë quotients which had not been used to distribute constituency seats.

    The only significant difference was that for a qualifying party that had won no seats in a given constituency, its first modified Sainte-Laguë quotient i.e. the party vote divided by 1.4 was the figure used to that end, whereas in Sweden the party vote total is used in such cases. In other words, Sweden uses pure Sainte-Laguë (unused) quotients just for that specific procedure.

    However, this system caused the same problems observed in Sweden, namely that some constituencies could be (and were) assigned multiple adjustment seats. In Norway’s 2001 Storting election, Akershus county increased its representation from 12 to 15 seats, while Hordaland County went from 15 to 17. I believe this was one of the reasons why Norway switched to the current one adjustment seat per constituency procedure (in the process increasing the number of such mandates from eight to nineteen). That said, the new system has its own problems, as noted on my website and here.

  18. The counting in Sweden is nearing completion, and the Greens have taken the last seat in Dalarna. The other constituencies I mentioned are nearing completion, and are close. By the way, this is still not the final count, but no new votes would be added after this point.

    About Norway: As long as there were only eight levelling seats, it really was no problem that they tended to cluster in a few urban constituencies, and in fact this helped to mitigate a long-standing bias. The other areas were still guaranteed a favourable and stable share of representation, since the distribution of the 157 regular seats were constitutionally entrenched and largely unaltered since 1953.*

    However, when it was decided to expand the number of levelling seats in order to somewhat increase proportionality, a different system for distributing these seats downward had to be devised.** Representation of various regions is jealously guarded by Norwegian politicians, and it would not have been acceptable to risk as large imbalances as in Sweden. This is also reflected in the new distribution formula for the constituency seats, where clumsily square kilometres are given representation alongside population (I have no objection in principle to some favourable treatment of outlying areas).

    * In 1973, two extra seats were given to Oslo, and three seats to surrounding Akershus, while Bergen and Hordaland were united as they had been for administrative purposes. In 1985, another two seats were given to Akershus. These alterations were late and inadequate responses to long-term population movements.

    ** I agree that the chosen system is horrible in regard to the last seats given out. Better systems could have been created with no added complexity, although I also agree that the unconventional biproportional system is inherently best suited for the task.

  19. Wow, terrible last minute editing of the first paragraph there.

    The media are now reporting that the liberals (Fp) are four votes short in Göteborg, but so far val.se has not been updated to show this.

  20. Espen, on the basis of the figures currently available on val.se, the Social Democrats remain narrowly ahead of the Liberals in the race for the last constituency seat in Göteborgs Municipality, by 8889.67 to 8829.67, with a single polling station (of 286) outstanding. However, the Social Democrats are trailing the Liberals for Värmland County’s last seat by a whisker, 7573.78 to 7577.14, with only two polling stations (of 193) remaining to be tallied. In fact, a distribution of Riksdag seats with the latest figures has the ruling Alliance just one seat short of an overall majority.

    By the way, a third seat may change hands in Stockholm Municipality, but it would switch from the Christian Democrats to the Greens, and would have no impact whatsoever in the overall distribution of Riksdag seats.

  21. All of Göteborg Municipality’s 286 polling stations are now in, and the Social Democrats appear to have just barely held on to the constituency’s last seat, beating the Liberals by
    8949.22 to 8943.00.

    The Social Democrats are also ahead of the Liberals in the race for the last seat in Värmland County, but the margin remains razor-thin – 7601.22 to 7598.57 – with one polling station outstanding.

    In all, the Alliance parties would now be two seats short of an absolute majority.

  22. The last polling station in Värmland County has been tallied, and the Social Democrats have retained the last seat in the constituency over the Liberals, again by the narrowest of margins: 7613.33 to 7608.57.

    It looks like the Alliance will be two seats short of an overall majority.

  23. It looks that way. Very tight results, but still, quite a few votes would have to have been miscounted to change them. I presume this is unlikely. The county administrations will now do a second full count (next week?). The results then established could be appealed further to a central election board, but any ruling there might not be ready until November. The Riksdag first meets in October without regard to any election appeals. This is all according to val.se and svt.se (Swedish public television).

    My “model” seems to have held up pretty well though (apart from that initial unpleasantness). I may do a tiny summing up of that later in the week.

  24. It turns out that even if the liberals had won those two seats, the final national figures now show that the last of those seats would have been captured not by the Christian Democrats, but by the Greens. The next in line after that would have been the Left, narrowly ahead of the slow-growing Christian Democrats. Also, the coalition parties have 1,805 fewer votes than the other parliamentary parties combined, so this is not a “wrong winner” scenario.

    A correction (the explanation at val.se was very confusing): There is no new count next week, the protocols make it clear that the county administrations have approved these counts, thereby authorising the “appointment” of MPs and their deputies by the election authority.

    The national evolution of votes from successive election nights to Wednesdays:

    Year 1998 2002 2006 2010
    Valid 1.40% 1.32% 2.08% 3.28%
    M 2.11% 2.02% 2.68% 3.63%
    C 1.03% 0.90% 1.68% 2.79%
    Fp 2.02% 1.49% 2.45% 3.12%
    Kd 1.36% 1.21% 1.71% 2.77%
    S 0.93% 0.93% 1.42% 2.61%
    V 1.43% 1.57% 2.35% 3.79%
    Mp 2.46% 2.36% 3.78% 5.18%
    Oth -0.48% 1.17% 2.04% 5.43%

    The Sweden Democrats, only counted separately on election night in 2010, grew by only 2.86%. The relative advantage of the centre-right parties fell in 2010, and that of the Left and Greens increased. The overall growth of late incoming votes continued.

    In Dalarna, the Social Democrats increased by 2.35%, the Left by 3.65% and the Greens by 5.99%. The resulting relative gain of 3.56% gave the latter party the last seat.

    In Göteborg, the liberals had a relative gain over the Social Democrats of 0.563%, just below the 0.634% mentioned above.

    In Värmland, they had a relative gain of 0.824% which was just below the needed 0.888%.

    Amazingly, my national vote share estimate above was accurate to the decimal for all the parties, but the seat calculation was still off by a little even when taking constituency seats into account. Barring appeals, the composition of the Riksdag will be as follows (deviation from proportional distribution noted): M 107 (+1), C 23, Fp 24 (-1), Kd 19 (-1), S 112 (+3), V 19 (-1), Mp 25 (-1), SD 20.

  25. Actually, even if the Liberals had obtained the last constituency seats in Göteborg Municipality and Värmland County, and the Left Party captured the last constituency mandate in Östergötland County (the seats in question having been won by the Social Democrats by nineteen, seven and ninety-five votes, respectively), the Alliance would have remained one seat short of a majority: under such an scenario, the nationwide distribution of Riksdag mandates would have assigned the last seat to the Left Party over the Christian Democrats, which narrowly slipped behind in the final count.

    To be certain, the “wrong winner” scenario was avoided, but just barely so. At any rate, I have now updated my site’s Sweden page (see link on the second comment) with definitive 2010 general election results.

  26. Yes – given these national numbers, the only way the government could have retained its majority was if the Social Democrats instead of losing just one constituency seat had lost three or four, at least one of which to the Moderates, or five or more, to anyone. In the former case another “ill-gotten” Moderate seat would have prevented the Greens or the Left from picking it up in the national calculation, and in the latter case the Social Democrats would themselves have been left under-represented.

    Of course, even allowing for extreme luck, such local shifts would surely have altered the national tally so that the resulting government majority would have been well deserved indeed.

  27. I went over the definitive figures once more, and the simplest combination of changes I could come up with that would have given the Alliance a one-seat majority was precisely the one I proposed on my (yet to be updated) blog posting earlier this week (see link above), namely the Liberals gaining the last seats from the Social Democrats in Göteborg Municipality and Värmland County, and the Moderates picking up Kronoberg County’s final seat by outpolling the Social Democrats in that constituency.

    That last seat would have been crucial, because not only it would have been one fewer extra seat for the Social Democrats, but also an additional extra mandate for the Moderates. However, the Social Democrats kept it by 794 votes.

    Incidentally, in this scenario the seat that actually switched hands in Dalarna County would have made no difference whatsoever: it would have remained among one of the “Red-Green” parties one way or the other.

    More importantly, the Alliance parties would have had a majority over the opposition parties represented in the Riksdag despite still having slightly fewer votes than the latter.

  28. Sweden: election 2014 (when is this planting going to sprout?)

    The left of centre parties are 17 seats short of a majority. This leaves the anti-immigration party crowing that they hold the balance of power. The Centre Party — with agrarian roots, strongly pro-immigration, and attracting urban socially liberal voters — holds 21 seats. It lost only one seat to the anti-immigration party, unlike its competition (the Liberal People’s Party) which lost five. It has a previous history of working with the Social Democrats. If they want to boast that they stymied the anti-immigrants, they need to be part of a stable grand centre-left coalition. Right? But it hasn’t happened yet. Playing hard to get???

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s