This is our blog. We write about cool experiments, interesting case studies, open talks and technicalities.
Grab the RSS feed


Gavagai on the September 14 Riksdag Elections in Sweden

We at Gavagai have been tracking political opinion in Sweden for a while now. We have some observations we can share with you today, just before the parliamentary election to the Swedish Riksdag. The major contenders in the election is the incumbent cabinet coalition of four liberal and conservative parties (“alliansen”: M, C, FP, KD) and the four opposition parties, three of which are left-leaning (“de rödgröna”: S, MP, V) and have to some extent aligned their political goals and ambitions, and finally one party (SD) with a populist and xenophobic agenda which none of the other parties wish to cooperate with. Formation of a cabinet after the elections may turn out to be a serious challenge since the Riksdag needs to approve the prime minister and it appears likely that neither the governing alliance nor the three left-leaning parties will be able to achieve a majority in the Riksdag.[1]



  • We find that counting mentions or share of voice is not a useful predictor of electoral votes.
  • We believe that the key to understanding public opinion are the issues debated by the electorate, not the parties they mention.
  • We find that issues of interest to the general public do not always translate to political topics.
  • We find that the associations to the ruling cabinet do not significantly indicate more belief in their competence than associations to the opposition.
  • We find that the electorate is not more annoyed with the incumbent government coalition than with the opposition.
  • We find that negative campaigning does not seem to make a lasting impression on political discourse.
  • We predict that MP’s poll figures from the past few years are over-optimistic and that they will receive a significantly lower share of votes in the election.
  • We predict that FI will not make it into the Riksdag.



Share of voice as a measure of political clout

Our measures show that M, the leading party of the incumbent cabinet, gains the most media exposure, followed by S, the largest party, and SD, the most controversial party. The way people discuss politics in social media is thoughtful and comparative – they discuss pros and cons of various parties. While most writers are biased and subjective and make no attempt to be impartial, only a minority of writers profess blind loyalty to some specific party. Writers discuss actions they like and dislike alike by parties they like and dislike alike. A simple tally of media exposure would give an election result no-one expects to come true. More analysis of what is written is necessary to understand the mechanisms of public political discourse.

Some crucially interesting issues: Environment, Equality, and Education

Some political issues have found purchase in net-based debates – notably environment, equality between genders, education, immigration, and profit interests in public services. Other issues have no space in the public debate: pensions, public safety, defence issues, or tariffs and trade.

We systematically track the public opinion on several political issues, both in connection with names of parties and politicians, and without. We find that where taxation clocks in at about 6000 posts per day, education gets somewhat higher numbers, gender equality nets twice that volume, and environmental issues twice again as much. This would seem to be an indication that some issues have  potential to attract voters.

Issue ownership

Political opinion is only partially connected to political parties or politicians. Some issues are strongly connected to a party; others cut across party lines. In the heatmap graph below we display how the amount of discussion for some chosen hot political topics during the last 12 month period and opinion polls for the major political parties correlate[2]. The strongest signal here is that increased discussion on “welfare” topics, mostly to do with health care and elderly care, predicts a boost in figures for KD, V, and S whereas it is predicts lowered figures for MP.

In other measurements we have found how some issues are strongly connected to some party. For instance FI is the main carrier of discussions on gender equality; SD on discourse on immigration and racism; V on profits and welfare providers. This does not always translate to success. A surprising observation is that the vibrant discourse on climate, environment, and conservancy issues does not translate into increased support for MP, V or C, all of which profess to be green: it appears that the general public does not perceive environmental issues to be political, and that MP and C have failed in bringing their signature questions into the political arena[3]. The line graph below shows how the mentions of MP, V, and C co-occur with mentions of environmental questions. The daily frequency, except for occasional peaks coinciding with the EU parliamentary elections in May and the Almedalen political conference in July remain under a hundred mentions per day. This is to be compared with other political topics.

In this graph for the month leading up to the elections, the green line shows MP mentions with environmental issues (green line) and health care issues (blue line). The latter is more or less on par with the first, in spite of the first being the signature issue of the party.

These observations go to show that an issue of great interest of the public (as evidenced by the frequency of mention in general discourse) can fail to become a political topic.

Incumbents and challengers – stature and executive oomph

The incumbent cabinet (“Alliansen”) has been in power for eight years. The contenders (“Rödgröna”) have never governed together. A major strategy both sides have made use of in the election campaign has been to exude competence and executive power and to simultaneously portray opponents as either weary and void of ideas or as fragmented and incompetent.

The following graph shows how the incumbents and the contenders are mentioned in terms which indicate competence and executive capacity. It appears that the incumbent cabinet has not succeeded in projecting more competence than the hopeful opposition parties.

The following graph shows how the incumbent coalition and the opposition are mentioned in terms which indicate tiredness or confusion. It appears that during some periods during the past year, notably right after the elections to the EU parliament in May, the opposition has succeeded in portraying the incumbent cabinet to be tired and dispirited. This effect seems to have worn off during the Summer, however. Negative campaigning seems not to last.


The incumbent cabinet (“Alliansen”) is to some extent a natural focus point for daily frustration. The contenders (“Rödgröna”) use a more aggressive language – one of the parties (V) is formed from a revolutionary tradition and uses verbiage which involves class struggle and confrontation. How does this project into the attitudes and associations visavi Alliansen and Rödgröna? The table below shows that – some peaks of annoyance visavi the incumbents aside – the incumbents and the contenders prompt about the same amount of annoyance and anger among the voters. (One of the later peaks of violent language in connexion with Alliansen, e.g., is about enacting stricter sentencing guidelines for violent criminals rather than anger towards Alliansen itself).

These observations lead us to conclude that negative campaigning seems not to have strong or lasting effects on the associations in general discourse on political actors.

MP and negative publicity

MP is perceived to be a well-meaning party for the urban chattering classes, and its sympathisers are well represented among the media[4]. This has been claimed to lead to less critical reporting about the party and its representatives. We can see some support for this view in our measurements. We have found a significant correlation between amount of negative mentions of MP and lowered support for the party – this is in contrast with how much effect an increased volume negative mentions of other parties has on their opinion polls. This might tentatively be understood to be an effect of MP not being as carefully examined as other parties, thereby explaining why negative news and negative observations with respect to MP have more effect or explain more of a dip in support than for other parties. Nearing election time, media scrutinize and rake through the policies of all the major parties with a much more critical eye than during the parliamentary four year term. We expect this more critical examination to have effects on the vote.

The increased amount of negative publicity for MP will have an effect on their share of the vote. We expect that for MP poll figures from the past few years (ranging around and above 10%) are over-optimistic and that they will receive a significantly lower share of votes in the election.

FI and the 4% threshold

One of the factors which contribute to the uncertainty is whether the newly formed Feministiskt initiativ (FI) party will gain enough support to claim seats in the Riksdag or not[5]. It is to be expected that potential FI representatives in the Riksdag will support a left-leaning cabinet and that most of their voters would have supported some of the left-leaning parties – this makes it strategically important for the party convince potential voters that a vote for FI is not “wasted”.

In previous elections, at times, parties have been able to beat the 4% threshold. This has especially been true for the three EU elections Sweden has held since joining: every election has marked the accession (and subsequent drop) of a new party. Since the institution of a vote threshold a new party has succeeded in entering the Riksdag or the EU seven or eight times, depending on how you tally the results[6]. This has most often coincided with some external event, situation, or idea which has explained the momentum for the acceeding party.

We have been following the rising interest in FI since the beginning of the year. The interest started gaining steam after political manifestations on the International Women’s Day on March 8, and peaked around this Spring’s EU elections.

Frequency of mention for three political parties during the month of Aug 7 – Sep 7 2014. FI: blue; V: green; KD: red

The graph over the latest month shows that FI would seem to have an excellent chance of entering the parliament. However, there are confounders. The major question of interest for FI is gender equality and this is an issue which has major traction in social and editorial media alike. Also, FI has actively promoted its visibility on the net: Gudrun Schyman, party leader, has asked sympathetic voters, including those who will vote for another party, to respond to pollsters with FI rather than other parties, to build up support for them[7].

We find that many of the mentions we see are clear vote declarations: “I voted for FI today!”. These mentions show great mobilization among FI voters, but do not correspond to other parties – they are an indication of fighting spirit rather than broad interest in the party. This illustrates the need of editorial interpretation social media data. No measure is better than its reader! This also illustrates the need for a new party to find a cause to rally around, and to ideally find a cause which established parties have been disregarding or downplaying: environmental issues, personal integrity, immigration, EU skepticism. Without resonance in public opinion, creating media interest in a new political organisation is a near-impossible challenge. FI did this with considerable tenacity and savvy.

The anomalous character of the feed leads us to predict, in spite of its impressive showing in social media, that FI will not gain seats in the Riksdag.

Sep 10, 2014 JiK

[1] Sweden has parliamentary democracy. After an election, the Speaker of the Riksdag commissions some suitable representative of the people to form a cabinet – typically from the party which has been perceived to win the election. The Riksdag signs off on this decision.

[2] The discussion of a topic is measured in a 30-day window and the opinion figures are the weighted figures of various polls as given by Botten Ada, (

[3] On August 16, Prime Minister Fredrik Reinfeldt kicked off the election campaign by giving a well-reported speech on immigration issues and on integration of refugees and migrants. If he would have held a similar speech on environmental and climate issues, we would expect to have seen a very differnet issue map.

[4] A poll among journalists showed that fully 41% of respondents indicated MP to be their preferred party.  Kent Asp ”Journalistkårens partisympatier”

[5] The Riksdag has 349 seats, but to claim seats, a party needs to achieve at least 4% of the vote nationwide or 12% in one electoral district.

[6] 1988-1991 and 1994- Miljöpartiet; 1991- Kristdemokraterna. 1991-1994 Ny Demokrati. 2010- Sverigedemokraterna. In the EU elections slightly different rules apply but the 4% threshold is still there. 2004-2009 Junilistan; 2009-2014 Piratpartiet; 2014- Feministiskt initiativ.

[7] “Fejkad Opinion! Ett upprop om feministisk aktivism”


Category: case studies