- Détails du magazine
- Format
- Magazine
- eISSN
- 2003-184X
- Première parution
- 30 May 2019
- Périodicité
- 1 fois par an
- Langues
- Anglais
The Swedish national election on 9 September 2018 has been called one of the most controversial campaigns in Swedish history (Wixe & Ek, 2018). It was the first parliamentary election after the 2015 “migrant crisis” in Europe, when Sweden took in among the highest numbers of refugees relative to population (Traub, 2016), and polls predicted record results for the far-right nativist-populist Sweden Democrats (SD) in 2018. Although SD's nearly 18 per cent vote share was less than anticipated, it was still striking for a party that only a decade ago claimed a small fringe of the Swedish electorate (Statistics Sweden, 2018) and established SD as one of the country's major parties (Oscarsson & Strömbäck, 2019: 325). In international coverage, the election was taken as a sign of the growing strength of nationalist politics – they could take hold “even in Sweden!” As
Media research on Sweden's election has largely focused on the domestic public sphere, with particular attention to the growing strength of right-wing online actors and alternative media (Larsson, 2020; Sandberg & Ihlebæk, 2019; Schroeder, 2020). Some research on the Swedish election and social media has discussed international interest in the election. Researchers at the LSE Institute of Global Affairs released a report in October 2018 titled
With this as a starting point, we examine how foreign media events construct, and are constructed by, transnational Twitter audiences. Although foreign events might lack the monopolising force that media sociologists Daniel Dayan and Elihu Katz (1992) originally described in their concept of media events, we consider Hepp and Couldry's (2010: 10) contention that these events may nevertheless be experienced as “thickenings” – thickenings of attention and meaning, varying in strength in different territories, and produced by both mass media (press, television, and radio) and by online networks. Rather than examine an overtly global event, we choose as our case study a national election, a classic example of a media event intended for a geographically bound polity. This allows us to better examine how globalised digital media lead media events to, in the words of Dayan and Katz (1992: 15) “create their own constituencies”. We adapt Tomlinson's (1999) concept of deterritorialisation to the online world, using it to describe the way global constituencies interpret, remix, and transfer national politics to new cultural contexts.
Empirically, we use a corpus of English-language tweets collected in real time during the 2018 election campaign and its aftermath. Drawing on both media events theory and the literature on political networks on social media, we pose the following questions: 1) What transnational networks formed on Twitter around the Swedish national election? 2) What themes did these networks spread? and 3) What was the role of the news media in creating moments of thickening among Twitter users? Together, these questions contribute to a better understanding of how audiences deterritorialise politics and political events on global social media platforms. In the following section, we explore previous research on media events, global media events, and transnational communication on social media.
The notion of a media-constituted “public” has long been a central concept in cultural, political, and media studies. In Anderson's (1983)
In the age of social media and Big Data, computational methodologies have taken understanding of these events to new levels. Researchers have found that conventions, debates, and election-night returns are periods of high Twitter traffic, as audiences live-tweet about the candidates and root for their party (Larsson & Moe, 2016; Robertson et al., 2019), forming networks through interactions. In recent years, researchers have contributed other models of media events that focus more on the role of audiences in shaping the event, such as conflictual media events (Hepp & Couldry, 2010; Mortensen, 2015) and media event chains (Sreberny, 2016).
Among the main threads in these developments is a challenge to Dayan and Katz's implication that media events are top-down phenomena, where audiences accept messages as intended, or even experience the event as one audience. Particularly in the case of elections and other contest-style phenomena, media events may put more emphasis on partisanship and societal division than on unity. In contrast to a mass-media based
Revisiting his own concept in 2010, Dayan described a kind of meta-contest that now takes place within media events: “In the
Written before the arrival of social media platforms, John Tomlinson's
The idea of media events moving beyond the nation-state was always embedded in the concept. Dayan and Katz (1992: 16) hinted at the possibility, writing that broadcasting allows media events to talk “over and around conventional political geography”. Taking up this thread, Hallin and Mancini (1992) examined television coverage of Cold War–era US–Soviet summits. The scholars observed that journalists used “we” in the broadcasts not to refer to inhabitants of their home country, but to describe humanity as a whole – contributing, the scholars argued, to a sense of global community. More recent studies have examined the unifying nature of other international events as experienced through globalised media, including ritual international competitions like the Olympics (Roche, 2002) and Eurovision (Kyriakidou et al., 2018); ceremonies like the D-Day anniversary (Robertson, 2010); and conflicts like the Danish cartoon controversy (Eide et al., 2008).
One of the persistent findings of this research is that “global” media events do not have the totalising, world-encompassing force implied by the name. Even seemingly identically witnessed events like the Olympics or live commemorations – or perhaps we should say especially these events – are subject to “domestication” through coverage by national media outlets (Clausen, 2004; Eide et al., 2008; Frandsen, 2003; Robertson, 2010). Sreberny (2016: 3500) observes that global media events are typically not experienced simultaneously at all, but through a “deterritorialized assemblage of contemporary event chains”, thanks in part to the way digital media allow events to be experienced on the user's schedule. However, scholars argue a sense of immediacy remains, and that furthermore (as Tomlinson suggested), global media platforms help establish transnational reference points among geographically dispersed audiences. Volkmer and Deffner (2010: 218), drawing on interviews across nine countries, argue that a transnational “eventsphere” is emerging, in which events are increasingly viewed in terms of their global meaning.
Reviewing the literature on global media events, we find that the concept generally has been applied to either 1) planned mega-events, or 2) unplanned conflictual events that gain international attention. Mega-events, as defined by Roche (2002), are large-scale events in the original Dayan and Katz mould, but organised internationally, such as the Olympics, Eurovision, and world summits. Conflictual events, like the Charlie Hebdo killings or the Danish cartoon controversy, are generally unanticipated local occurrences that have resonance beyond their original geographic context and are deemed global. What is not clear from these two strands of literature is how, given a digital media environment that facilitates rapid deterritorialisation and an emerging transnational eventsphere,
Thus, we seek to offer a third perspective, maintaining Dayan and Katz's concept of a media event as anticipated, but exploring the creation of unanticipated audiences. Drawing on Tomlinson's (1999) concept of deterritorialisation, we examine an ostensibly national media event from the contest genre: a national election campaign. Unlike conflictual events, these are planned rituals with established rules familiar across many countries. Yet, unlike global mega-events, elections are not designed for a global audience. Rather, these events are witnessed ad hoc by international observers through globally available platforms. And such events arguably have real or perceived stakes for audiences elsewhere, as they incorporate them into their own phenomenological worlds (Tomlinson, 1999).
Additionally, this helps fill a gap in the understanding of elections. Despite wide intercultural familiarity with the patterns of election campaigns, the increased digitisation of the public sphere, and the tendency of these events to cast the audience in a participatory role, little research has so far addressed the globalisation of national elections (for a couple of exceptions, see Cheng & Chen, 2016; Sevin & Uzunoğlu, 2017). Research in this vein more often focuses on government-backed “foreign interference” campaigns (Blackwill & Gordon, 2018; Bradshaw & Howard, 2018; Colliver et al., 2018) and “cyber warfare” (Singer & Brooking, 2018). Our approach instead brings in the citizen-oriented tradition found more often in social movement research. As we discuss in the next section, this approach views digital technology as giving users new ways of participating in the global flow of information.
Much of the scholarship on transnational communication on social media comes out of social movement research, starting with the pivotal Arab Spring protests and uprisings. While popular imagination has painted these events as birthed by a global Twittersphere, scholars argue the relay of foreign movements from one cultural context to another is much more complex. Foreign events do not transfer unchanged from one cultural context to another; rather, they have resonance because audiences reshape foreign issues, transposing them or plugging them into other geographies and power struggles (della Porta & Diani, 2006). Lim (2018: 112) has developed the concept of portability to capture the way events on the ground are distilled into archetypes and symbols that transcend the local context and “evoke shared emotion” in far-away others.
Such a process can be seen in globally visible movements like The Arab Spring, Occupy Wall Street, Black Lives Matter, and #metoo; each are rooted in unique events, but became portable on Twitter through universalised enemies and simplified themes of democratic freedom, wealth, race, and gender. In part due to such movements, Twitter is viewed by scholars as an important platform for connecting audiences and offering them a “front row seat” to events (Jackson et al., 2020). More than traditional news media, Twitter allows users to not only witness an event, but to witness others’ witnessing of it. Empirical research also demonstrates that Twitter tends to facilitate cross-border networks more than Facebook (Ghemawat, 2016), and though Twitter contains many domestic language communities, English is by far the most used language: Scholars suggest it has become a kind of lingua franca of transnational networks (Hänska & Bauchowitz, 2019; Mocanu et al., 2013).
Due in part to the cosmopolitan associations with global polyglot audiences, much research on transnational social movements has focused on left-wing causes. However, scholarship has been shifting as a result of the recent rise of openly right-wing nationalist discourse in many countries (Bieber, 2018; Bob, 2012). Davey and Ebner (2017: 23) found in an analysis of the German hashtag #Merkelmussweg [#Merkel has to go] that 38 per cent of the geotagged tweets came from outside Germany. Grumke (2013: 50) suggests the shared enemies of globalism and global elites provide an ideological unity among what he calls an “international of nationalists”. While empirical research does not point to widespread online coordination among established groups and parties on social media, there is evidence that nativist messages – and especially anti-immigration and anti-Islam messages – have transnational portability in online networks (Caiani & Kröll, 2015; Froio & Ganesh, 2018).
Tweets were collected between 10 August and 28 September 2018 from Twitter's real-time Streaming API using the DMI-TCAT (Digital Methods Initiative Twitter Capture and Analysis Toolset) (Borra & Rieder, 2014). This time period begins a month before election day (9 September) and runs until three weeks after, during which a post-election “crisis” over government formation unfolded. The keywords used were “election”, “elections”; “vote”, “voter”, “voters”; “party”, and “parties” – paired with “Swedish” and “Sweden”. Any combination of these words (regardless of letter case) anywhere in the tweet (including as a hashtag) qualified the tweet for collection by the TCAT. Using keywords to collect data from the Streaming API can run into a rate limit that Twitter imposes on free data collection (Sloan & Quan-Haase, 2017). This rate limit caps data to around 1 per cent of the entire Twitter stream; queries that would otherwise return more tweets therefore miss an unknown number of tweets. We therefore chose focused keywords and do not believe the data was subject to this limit, based on the number of tweets collected and the rate limit alert feature of the DMI-TCAT. As part of the data processing, we also checked for signs of bot activity using tweet metadata (Bovet & Makse, 2019) and the tool Botometer (Yang et al., 2019), which suggested automated accounts had minimal impact (~5% of users).
In total, our English-language keywords resulted in a combined dataset of 198,635 unique tweets, sent by 91,797 users (an average of 2.2 tweets per user). Almost all of the tweets were in English (98%), followed by Swedish (0.6%), and French, Romanian, and undetermined languages (each with 0.2%).
In addition to the English-language tweets, we also collected tweets containing #svpol, #val18, and #val2018 (abbreviations for “Swedish politics” and “Election 2018”, also used by Colliver et al., 2018; Petrarca et al., 2019). This returned 221,686 tweets in Swedish. This Swedish-language data serves as a point of reference when examining temporal patterns in the English-language data and understanding moments of thickening.
As noted by Bruns and Burgess (2011), Twitter is not a single social network, but a series of smaller subnetworks, around which users group based on shared interests. According to social network theory, information flows faster within subnetworks than across subnetworks (Himelboim et al., 2017). These subnetworks, also called communities and clusters, can be detected algorithmically by identifying groups of nodes (users) who have more ties (tweets) with each other than with other users (Blondel et al., 2008). We use Blondel and colleagues’ modularity class algorithm in the network analysis software Gephi (run at 2.0 resolution) to identify the largest subnetworks in the full English-language data collection. We then qualitatively characterise these subnetworks based on the most prominent users (those who received the most retweets and @mentions).
Einspänner and colleagues (2014: 103) have suggested using an “iterative and cyclical process” to create coding schemes for Twitter data. In that vein, we developed the coding scheme 1) deductively, based on known topics in the Swedish election (Petrarca et al., 2019) and themes from the literature on media events, transnational communication, and social movements (particularly Froio & Ganesh, 2018; Jost et al., 2018; Lim, 2018; Volkmer & Deffner, 2010), and 2) inductively, based on themes we discovered in the data.
The coding scheme was developed as follows. We began with a list of expected themes. We then conducted what Neuendorf (2002: 103) calls a “qualitative scrutiny of a representative subset”, where the two authors independently examined a random selection of the data. Some of the deductively established themes were also prevalent in the tweets, such as immigration, nationalism, and horserace aspects, as well as users comparing Sweden to their own local or national context. However, we discovered that many themes we expected to find (e.g., economy, education, and women) were scarce or nearly absent from the sample. In addition, we wrote down repeated themes we did not expect, such as “speculations of election fraud” and “critique of sensationalist coverage”.
Based on our initial review, we then drafted a coding scheme that combined the inductively and deductively developed themes. This coding scheme went through three rounds of testing. In each round, the two authors independently hand-coded a sample of the same 100 tweets and an intercoder reliability test was run to see if we had the same findings. After each round, themes were added, removed, changed, or refined. The final coding scheme and intercoder reliability scores for each theme are listed in Table 1.
Coding scheme
| 1 | NationalistRise | The tweet puts emphasis on the success or expected success of SD in the election. | .820 | 1,256 |
| 2 | Horserace | The tweet puts emphasis on updates of who is winning and losing, including poll results, voter turnout, results of the election, and updates on government formation. | .837 | 1,422 |
| 3 | Violence | The tweet puts emphasis on reports of violence, threats of violence, rape, terrorism, or other violent crime. | .784 | 498 |
| 4 | HistoricUpheaval | The tweet puts emphasis on the historic nature of the election or the permanent mark it will leave on Sweden | .678 | 373 |
| 5 | Migration | The tweet puts emphasis on immigration policy, (im)migrants, refugees, Islam (as implicit to migration in Sweden), or multiculturalism. | .801 | 1,431 |
| 6 | DebateDistortion | The tweet puts emphasis on external factors: Russian or other foreign interference, fake news, or platforms manipulating content. | .949 | 321 |
| 7 | ElectoralFailure | The tweet puts emphasis on internal factors: voter fraud, public corruption, unfair treatment of parties, unfair voting rules, and other institutional failures that would impact the results. | .959 | 415 |
| 8 | GlobalPolitics | The tweet puts emphasis on a relationship between the Swedish election and politics in other places (e.g., Europe, the UK, the EU, the West, the world). | .795 | 665 |
| 9 | WelfareState | The tweet puts emphasis on Sweden's welfare state, including taxes and welfare benefits. | .887 | 61 |
| 10 | UtopiaDystopia | The tweet puts emphasis on Sweden as a model leftist, progressive, socialist, or social democratic country. This may be in a positive or negative light. | .660 | 188 |
| 11 | Counternarrative | The tweet puts emphasis on the idea that the media or dominant narrative sensationalises, exaggerates, or ignores some aspect of the election. | .818 | 609 |
| 12 | Environment | The tweet puts emphasis on climate change, wildfires, or other environmental issue. | 1.00 | 32 |
| 13 | Racism | The tweet puts emphasis on racism in Swedish politics, including referring to a party as Nazi or having Nazi roots. Note that this does not refer to tweets that express racist views themselves. | .764 | 336 |
| 14 | Rooting | The tweet puts emphasis on personal support for a political “team”, including encouraging voter turnout (before the election) or expressing celebration or disappointment about the result (after the election). | .752 | 467 |
| 15 | Financial | The tweet puts emphasis on the election's impact or potential impact on global markets, investments, the SEK, etc. | .830 | 35 |
| 16 | Other | Emphasis of tweet not captured by the above categories. This includes tweets that are not about the election at all, are apolitical jokes about politicians, are not in English, or are unintelligible. | .700 | 259 |
All themes are based on manifest content (Neuendorf, 2002); however, we allow tweets to be labelled with as many themes as apply, since we found, as Sandberg and Ihlebæk (2019: 435) did, that themes often overlap. We also coded for thematic content that was more a mode of expression, such as rooting for a winner (Rooting) or comparing Sweden to other countries (GlobalPolitics), while others are more traditional election issues (Violence, WelfareState, Environment). This is another reason it was important to allow codes to overlap. The analysis of these overlappings became essential to understanding the data, as we will describe in the “Findings” and “Discussion” sections.
Two random samples from the full English-language dataset were hand-coded. The first is a population random sample – that is, it samples all tweets in the collection. We term this the PopRand sample (
Our themes analysis focuses on the coded PopRand sample (
Using retweets and mentions in the data, a network graph was made using the network analysis program Gephi. Running the modularity algorithm in Gephi (resolution = 2.0), we find 1,715 “communities”, or subnetworks (Blondel et al., 2008). Within these, 93 per cent of users are found in the six largest subnetworks; the remaining users are part of small subnetworks and dyads unconnected to the main network. After examining the largest subnetworks, we further collapse the users into three main subnetworks for analysis (see Figure 1). These subnetworks are summarised as follows: European and American far-right media and commentators (45% of users) in the upper left of Figure 1; mainstream international media and centre-left commentators (35% of users) in the upper right; and a British-specific subnetwork (13% of users) in the lower part. This last subnetwork was largely non-existent until a few days before the election; 90 per cent of the tweets by this network were sent 7–12 September.
Figure 1
We allowed each tweet to contain multiple codes, and the content analysis demonstrates it was common for tweets to contain more than one theme: 55 per cent contained one theme; 31 per cent contained two; 9 per cent contained three; 5 per cent contained four; and 1 per cent contained five. Of those containing only one code, about a quarter (27%) exhibited the Horserace theme. Horserace was also one of the most common themes overall (25% of the sample), along with Migration (28%) and NationalistRise (25%). This is followed by GlobalPolitics, Counternarrative, and Violence. Environment, Financial, and WelfareState were marginal themes.
Additionally, from the sample of non-retweets (non-RT) we see that top themes in original tweets were also Horserace, NationalistRise, and Migration. However, comparing the non-RT sample to the population sample (PopRand), we can see that certain themes received more amplification – in particular, Migration, Violence, and especially Counternarrative. That means that relative to the number of original tweets emphasising these themes, their overall representation in the volume of tweets was disproportional. Figure 2 compares the portion of each theme in the two samples, showing which themes were amplified.
Figure 2
To understand the connections between themes, we performed a calculation of co-occurrence through a multimodal network analysis. We use the PopRand sample for this analysis since it includes retweets and therefore better represents the content circulating on Twitter at the time. In the network shown in Figure 3, each tweet is a node connected to the themes it contains. (The Other category was for tweets not connected to any of the themes, so these tweets have been excluded from the network graph.) The themes are sized according to how many connections they have – that is, how many tweets contained the theme. As expected, we see that Migration, Horserace, and NationalistRise are the largest nodes.
Importantly, the visualisation also provides information on the relationship between themes: Tweets that contain multiple themes will draw those themes toward each other in the network. In Figure 3, we see that NationalistRise is centrally located, reflecting connections to a variety of different themes. Migration is also fairly centrally located. We also see that Racism and GlobalPolitics are closely aligned with NationalistRise, while Violence is more closely aligned with Migration. Financial, ElectoralFailure, and DebateDistortion appear on the periphery, reflecting that these themes were less often combined with other themes when they appeared, although ElectoralFailure and DebateDistortion are sometimes connected to each other, as might be expected given the content. Where connections are found within the main graph, DebateDistortion has more connections with Racism and GlobalPolitics, while ElectoralFailure has more in common with Violence and Migration, and to some degree Rooting (as will be discussed later, the connection between ElectoralFailure and Rooting appears to be disappointment that the user's favoured party did not win and the suspicion that fraud is to blame). We also see that major campaign issues like Environment and WelfareState – while not heavily emphasised in the tweets – do have connections to other themes. Environment has connections to Rooting, and WelfareState is closely tied to the central themes of NationalistRise and Migration. (These relationships are also confirmed numerically using the Jaccard Index, a statistical measure of overlap ranging from 0 to 1, where 1 is perfect overlap; see the Appendix.)
Figure 3
We can also use the network data (see Figure 1) to identify the relationships between the themes and the different subnetworks discussing the election. Figure 4 colourises the themes co-occurrence network according to the three main subnetworks identified previously.
Figure 4
In this graph, we see that certain subnetworks emphasised certain themes and theme combinations. In particular, Migration, Violence, and ElectoralFailure were mainly themes expressed by the right-wing subnetwork, while DebateDistortion was more varied across the network. Counternarrative was largely expressed by the British network (as we will discuss later, this was largely due to critiques of the BBC's reporting on the election). Tweets in the Environment, Racism, and Financial themes are generally originating from the mainstream media subnetwork. The subnetwork oriented around the media also emphasises the rise of the Sweden Democrats and their stance on immigration, leading to the dominance of NationalistRise, Migration, and Horserace, although these themes, along with GlobalPolitics, are also prominent throughout the network. We will explore the differences between the subnetworks further in the “Discussion” section.
Across all subnetworks, retweets and “via”-tweets dominate the data (81% of the collection); about one-third of all tweets include links to URLs outside Twitter. We also see from the network map (see Figure 1) that many of the most retweeted and @mentioned accounts were mainstream news outlets, such as Reuters, NBC, CNN, and
To better understand the role of news media, we investigate the timeline of tweeting. Figure 5 compares the Swedish language data with the English language dataset. As expected, the major peak in tweeting in both timelines is around election night. However, the fluctuations in tweeting in the global Twittersphere do not match up with the Swedish sphere prior to election day. The Swedish-language tweeting spikes modestly during the televised debates (marked with grey lines in Figure 5), a moment of “thickening” predicted by previous literature on Twitter use during national elections. The English-language tweets do not follow this pattern; for example, there is little increase in tweeting immediately after the first televised debate on 14 August. Instead, we see a spike a few days later when NBC published a story about the election headlined “Far-right Sweden Democrats hope to topple century of socialism”, a story that was retweeted 589 times that day, according to our dataset. Another spike occurs on 7 September, the Friday before the Sunday election. This is when France24,
Figure 5
Right-wing accounts also played a role in circulating mainstream stories. For example, a major source of links to the NBC story on 17 August was a tweet sent by
Moving along the global timeline, one of the larger swells in attention came on 28–29 August, when right-wing YouTube personalities Paul Joseph Watson (@PrisonPlanet) and Alex Jones sent several tweets about YouTube removing right-wing Swedish content, which they framed as an attempt by global tech giant Google to try to influence the election (DebateDistortion). Here, we also see direct connections between the global right wing and right-wing alternative news outlets in Sweden. In one of his tweets about the removal of YouTube content, Paul Joseph Watson links to an article in the Swedish language outlet
Twitter commentators on other parts of the political spectrum were also influential. An uptick in tweeting on 3–4 September is partially tied to retweets of threads that same day by popular academics and Twitter personalities Yascha Mounk and Matt Goodwin (@GoodwinMJ). These received many more tweets than a
We have sought to better understand the way that contemporary media events “create their own constituencies” (Dayan & Katz, 1992: 15) on global digital platforms and how those constituencies in turn “try to affirm their own control” of the event (della Porta & Diani, 2006: 75). Here, we further discuss this competition for control, bringing in examples of tweets that represent how themes intersected in the global network.
Two central themes united the overall network. First, we see that transnational publics engaged closely with the Horserace dimension of the election, seen in the high volume of tweets emphasising polls, culminating in a huge wave of interest on election night and the following day. Even though global users were not in tune with the Swedish-language live televised debates, the familiar pattern of the election
The second important uniting theme was more specific to the case at hand. The anticipated success of nationalist politics, coded as NationalistRise, was a central theme across all subnetworks. We anticipated this to some degree based on reading the international news coverage. We also discovered when making the coding scheme that common election issues – health programmes, the economy, education – were not prominent in our data. However, we did not anticipate the degree to which NationalistRise would emerge as the central theme that connects nearly every other theme. Essentially, it is what Hepp and Couldry (2010: 11) called the thematic core.
However, this theme is interpreted differently by different subnetworks, as we can see in the other themes through which NationalistRise is refracted. In the mainstream subnetwork, NationalistRise often connected to Racism and UtopiaDystopia, as exemplified by the highly shared It's looking like the Swedish election will put Nazis in power. So the nightmare continues. #Brexit. There's an election in Sweden this weekend and from what I can see it's going to be as much of a catastrophe as the American election.
For the mainstream subnetwork, the Swedish election was largely understood through the Racism and NationalistRise themes, which were emphasised in periodic news coverage that prompted spikes in tweeting in the subnetwork. Sweden became a shorthand or “narrative crux” (Colliver et al., 2018) for those seeking to make a point about the enormity of these threats, since they exist “even in Sweden”.
In the right-wing subnetwork, NationalistRise was likewise an anchoring theme. However, here it was tethered to Horserace, Migration, and Violence (and the latter two were almost always connected to each other). This popular tweet by Fox News host Laura Ingraham is an example of how all four themes could come together in the right-wing subnetwork:
Swedish election: Main blocs neck and neck, lose seats...as nationalists gain. Under-reported story--ALL major parties moved to right on immigration/refugee issue.
The above tweet also reflects how the familiar ritual of the election cycle could help create an avenue for Dayan's (2010) Trojan horses. The event is not only a contest between different Swedish parties, but between different 1 more month t’ill Sweden's elections. Sweden, you have a chance to be that prosperous nation you once were. Please vote for Sweden Democrats and get rid of the illegals. #MakeSwedenGreatAgain
This tweet is also an example of the way people used Twitter to express a team spirit, integral to the contest genre. Indeed, the right-wing subnetwork was more likely than the other subnetworks to send Rooting tweets (appearing in 12% of their tweets). As Eide and colleagues (2008) found in the Danish Cartoon controversy, we observed during our analysis that users in this subnetwork ported the Swedish election to a global sphere by emphasising “civilizational” themes – discussing barbarism, survival, and the West. We have selected the following two tweets as examples of this occurring in the overlap of Migration and Rooting:
#Sweden's election is this coming Sunday! #svpol Swedes need to step up and vote out the current Government that is allowing this barbarism to occur in their country! Good luck to Sweden today. We Europeans have to remember: we can either let liberalism pursue demographic replacement and islamisation, or we can choose to SAVE our heritage and our people!
Importantly, Rooting tweets also position the user more as a witness or participant than a spectator (Dayan, 2010). Although not in the cosmopolitan spirit that Hallin and Mancini (1992) and Robertson (2010) sought, this sense of collective witnessing may be why the right-wing subnetwork was especially active in the data – possessing, as Grumke (2013) has described, a kind of cosmopolitan spirit with nationalist ideology.
Dayan and Katz (1992: 46) theorised that of all the genres of media events, contests were most open to a kind of meta-debate prior to their occurrence, focused on the rules participants would follow. Our findings point to such a debate happening transnationally about the Swedish election as well, through the themes of ElectoralFailure and DebateDistortion.
In Figure 4, we can see that these themes are largely discussed separately from other topics. Additionally, these themes were addressed differently by subnetworks. DebateDistortion tweets by the mainstream subnetwork focused on issues well-known in other countries: potential foreign interference and efforts to combat misinformation. As previously noted, the right-wing subnetwork also used the theme DebateDistortion, but with the focus on YouTube parent-company Google potentially interfering in the Swedish public sphere by removing content by the far-right nationalist party Alternative for Sweden. This story was responsible for a peak in tweeting around 28 August, and it was also retweeted by parts of the mainstream subnetwork, resulting in a modest increase in tweeting in that subnetwork as well.
However, the more important theme in the far-right subnetwork was Elector-alFailure, characterised by tweets about alleged voter fraud and elite corruption that altered the results. After SD did not perform as well as polls predicted, this theme became even more prominent. An example of an ElectoralFailure tweet from this subnetwork dated after the election came from a Swedish user writing in English: “I believe we just had our first fully fake election”. Another user in Germany connected ElectoralFailure to GlobalPolitics: “There was voter fraud in the Sweden election. No doubt we’ll see the same in Germany's next election”. Later, the increase in tweet volume on 23 September (see Figure 5) is largely due to a series of tweets by Peter Imanuelsen (@PeterSweden7), in which he describes “900 reports of election fraud” and suggests the election was so riddled with problems that a new election is needed. Imanuelsen was retweeted over 2,000 times that day. In other words, the integrity of Sweden's voting system appears to be an important aspect of the election's portability to a global sphere. As one user wrote in reply to @PeterSweden7, “So basically Sweden is now run by the American Democratic party?”
Our analysis points to social media users’ continued reliance on the traditional foreign press for understanding the event and thickening attention to the Swedish election. Outside of the right-wing subnetwork, the most shared URLs were to stories from Reuters, the BBC, Bloomberg News, NBC, and
This counter-response, captured by the code Counternarrative, is best seen in the British subnetwork, which formed later than the others, largely in reaction to the BBC's coverage of the election. The Counternarrative theme appears in 59 per cent of tweets from users in this subnetwork. One of the most retweeted was by Labour politician Andrew Adonis:
BBC reporting of Sweden's election sensationalist – a narrative of Brexit-style far right takeover. Only problem – the result of the election. The far right came 3rd, with 17%. The leading party was – wait for it – the Social Democrats, which BBC had on the verge of extinction.
Counternarrative, Horserace, and GlobalPolitics overlap in this tweet as Adonis leverages the Swedish election to critique domestic media. Other tweets call out the BBC for not being as tough on the British right wing as they are on Sweden's right-wing politicians, accounting for a high overlap of GlobalPolitics and NationalistRise in this subnetwork.
To a lesser degree, Counternarrative also appears in the mainstream subnetwork (9% of tweets). However, one of the most retweeted users in the data was Christian Christensen (@ChrChristensen), a journalism professor at the University of Stockholm, who frequently tweeted about problems with the international media's coverage of the Horserace aspect of the election and focus on SD:
To international outlets covering the elections in Sweden: Don’t give media oxygen to ANYONE pushing the “collapsing Sweden” narrative. It's a childish, nihilistic, bigoted message that just deflects from real politics and real issues.
As mentioned, such Counternarrative messages were disproportionately amplified by users (see Figure 2). Thus, while NationalistRise may be the thematic core of the discussions of Sweden's election, we also see Counternarrative as its counter-weight. This finding is in line with Volkmer and Deffner's (2010: 226) argument that among transnational audiences, particularly those online, “the role of media powerfully defining this center is being renegotiated”. The real global threat that the Swedish election represents, according to the Counternarrative tweets, is not the rise of nationalism. Rather, it is the obsession with spectacle in the global media (Kellner, 2010) and the dominance of a single narrative that smooths out and simplifies the local realities of a distinct national event.
The 2018 Swedish election was not just a Swedish event. Our goal in this article has been to understand media events in an age when communication by ordinary people is less restricted by traditional national boundaries. We examined how a classic form of media event – a national election – becomes deterritorialised in a globalised information environment characterised by digital networks. As advocated by Hepp and Couldry (2010: 12), we have sought to investigate how social actors use media events for “constructing reality in specific, maybe conflicting ways [and] to establish certain discursive positions and to maintain those actors’ power”.
In response to the first research question on the networks that formed, we found three distinct subnetworks, characterised as international media and centre-left or left commentators; British commentators and politicians; and international right-wing media and commentators (the largest network in terms of both users and tweets). Our second research question asked what themes global Twitter networks emphasised to make sense of the election. Here, we find that the subnetworks were largely united in framing the election as a fight over nationalism, but they deployed that theme differently and through various existing global and local debates. Finally, regarding our third research question on the role of news media, we find that the traditional foreign press was still a critical starting point for understanding the event, and it contributed to moments of collective thickening. However, individual users often acted as the gatekeepers, even for stories from major news outlets, and also promoted alternative or counter-readings of the event. On the political right in particular, users acted as political activists in the process, contextualising the election through a civilizational worldview. There are also indications of a much larger commitment to following Swedish politics on the right and connections between English- and Swedish-language alternative media.
The findings demonstrate that as public spheres move online, it is likely that national elections – not only global “mega-events” like American presidential elections – become part of discussions in transnational Twitter networks. Here, we have demonstrated how a small Scandinavian country's election became a symbol for threats facing the entire Western world, and that gaining ownership of the narrative about Sweden became part of the fight to define politics far beyond Sweden's borders. This analysis had limitations, however. Using manual coding meant we had to take samples of the larger dataset. Computational text analysis or linguistic approaches may be able to capture more dimensions of the way global audiences interact with events. Likewise, a qualitative discourse analysis could provide more insight into the interpretation of foreign events by social media users. Another question we did not address in this study is the degree to which global interpretations become part of the national election discourse – or are in effect,
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Coding scheme
| 1 | NationalistRise | The tweet puts emphasis on the success or expected success of SD in the election. | .820 | 1,256 |
| 2 | Horserace | The tweet puts emphasis on updates of who is winning and losing, including poll results, voter turnout, results of the election, and updates on government formation. | .837 | 1,422 |
| 3 | Violence | The tweet puts emphasis on reports of violence, threats of violence, rape, terrorism, or other violent crime. | .784 | 498 |
| 4 | HistoricUpheaval | The tweet puts emphasis on the historic nature of the election or the permanent mark it will leave on Sweden | .678 | 373 |
| 5 | Migration | The tweet puts emphasis on immigration policy, (im)migrants, refugees, Islam (as implicit to migration in Sweden), or multiculturalism. | .801 | 1,431 |
| 6 | DebateDistortion | The tweet puts emphasis on external factors: Russian or other foreign interference, fake news, or platforms manipulating content. | .949 | 321 |
| 7 | ElectoralFailure | The tweet puts emphasis on internal factors: voter fraud, public corruption, unfair treatment of parties, unfair voting rules, and other institutional failures that would impact the results. | .959 | 415 |
| 8 | GlobalPolitics | The tweet puts emphasis on a relationship between the Swedish election and politics in other places (e.g., Europe, the UK, the EU, the West, the world). | .795 | 665 |
| 9 | WelfareState | The tweet puts emphasis on Sweden's welfare state, including taxes and welfare benefits. | .887 | 61 |
| 10 | UtopiaDystopia | The tweet puts emphasis on Sweden as a model leftist, progressive, socialist, or social democratic country. This may be in a positive or negative light. | .660 | 188 |
| 11 | Counternarrative | The tweet puts emphasis on the idea that the media or dominant narrative sensationalises, exaggerates, or ignores some aspect of the election. | .818 | 609 |
| 12 | Environment | The tweet puts emphasis on climate change, wildfires, or other environmental issue. | 1.00 | 32 |
| 13 | Racism | The tweet puts emphasis on racism in Swedish politics, including referring to a party as Nazi or having Nazi roots. Note that this does not refer to tweets that express racist views themselves. | .764 | 336 |
| 14 | Rooting | The tweet puts emphasis on personal support for a political “team”, including encouraging voter turnout (before the election) or expressing celebration or disappointment about the result (after the election). | .752 | 467 |
| 15 | Financial | The tweet puts emphasis on the election's impact or potential impact on global markets, investments, the SEK, etc. | .830 | 35 |
| 16 | Other | Emphasis of tweet not captured by the above categories. This includes tweets that are not about the election at all, are apolitical jokes about politicians, are not in English, or are unintelligible. | .700 | 259 |
Jaccard Index – measure of overlap between themes (PopRand sample)
| – | 609 | 321 | 415 | 32 | 35 | 665 | 373 | 1,422 | 1,431 | 1,256 | 259 | 336 | 467 | 188 | 498 | 61 | |
| 609 | – | 0.004 | 0.005 | 0.000 | 0.002 | 0.111 | 0.002 | 0.099 | 0.066 | 0.059 | 0.000 | 0.051 | 0.013 | 0.005 | 0.020 | 0.009 | |
| 321 | 0.004 | – | 0.041 | 0.000 | 0.000 | 0.038 | 0.001 | 0.000 | 0.005 | 0.005 | 0.000 | 0.011 | 0.003 | 0.004 | 0.006 | 0.000 | |
| 415 | 0.005 | 0.041 | – | 0.000 | 0.000 | 0.006 | 0.003 | 0.002 | 0.018 | 0.006 | 0.000 | 0.000 | 0.008 | 0.007 | 0.025 | 0.000 | |
| 32 | 0.000 | 0.000 | 0.000 | – | 0.000 | 0.006 | 0.000 | 0.000 | 0.007 | 0.000 | 0.000 | 0.003 | 0.014 | 0.005 | 0.004 | 0.069 | |
| 35 | 0.002 | 0.000 | 0.000 | 0.000 | – | 0.004 | 0.000 | 0.003 | 0.001 | 0.005 | 0.000 | 0.000 | 0.000 | 0.005 | 0.000 | 0.021 | |
| 665 | 0.111 | 0.038 | 0.006 | 0.006 | 0.004 | – | 0.023 | 0.027 | 0.084 | 0.091 | 0.000 | 0.051 | 0.068 | 0.017 | 0.007 | 0.013 | |
| 373 | 0.002 | 0.001 | 0.003 | 0.000 | 0.000 | 0.023 | – | 0.044 | 0.107 | 0.191 | 0.000 | 0.086 | 0.065 | 0.214 | 0.093 | 0.021 | |
| 1,422 | 0.099 | 0.000 | 0.002 | 0.000 | 0.003 | 0.027 | 0.044 | – | 0.097 | 0.148 | 0.000 | 0.016 | 0.020 | 0.004 | 0.015 | 0.001 | |
| 1,431 | 0.066 | 0.005 | 0.018 | 0.007 | 0.001 | 0.084 | 0.107 | 0.097 | – | 0.000 | 0.026 | 0.109 | 0.067 | 0.019 | |||
| 1,256 | 0.059 | 0.005 | 0.006 | 0.000 | 0.005 | 0.091 | 0.191 | 0.148 | – | 0.000 | 0.118 | 0.039 | 0.113 | 0.074 | 0.020 | ||
| 259 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | – | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | |
| 336 | 0.051 | 0.011 | 0.000 | 0.003 | 0.000 | 0.051 | 0.086 | 0.016 | 0.026 | 0.118 | 0.000 | – | 0.009 | 0.076 | 0.008 | 0.000 | |
| 467 | 0.013 | 0.003 | 0.008 | 0.014 | 0.000 | 0.068 | 0.065 | 0.020 | 0.109 | 0.039 | 0.000 | 0.009 | – | 0.008 | 0.094 | 0.004 | |
| 188 | 0.005 | 0.004 | 0.007 | 0.005 | 0.005 | 0.017 | 0.214 | 0.004 | 0.067 | 0.113 | 0.000 | 0.076 | 0.008 | – | 0.112 | 0.020 | |
| 498 | 0.020 | 0.006 | 0.025 | 0.004 | 0.000 | 0.007 | 0.093 | 0.015 | 0.074 | 0.000 | 0.008 | 0.094 | 0.112 | – | 0.015 | ||
| 61 | 0.009 | 0.000 | 0.000 | 0.069 | 0.021 | 0.013 | 0.021 | 0.001 | 0.019 | 0.020 | 0.000 | 0.000 | 0.004 | 0.020 | 0.015 | – |