Politics As (Un)Usual?

POLITICS AS (UN)USUAL?

RELATIONAL, FUNCTIONAL, AND DISCURSIVE DIMENSIONS OF POLITICAL COMMUNICATION ON SOCIAL MEDIA

Abstract

This dissertation examines the opportunities provided by social media for previously disadvantaged political actors and perspectives. Against the broader normalization theoretical framework, which argues that new technologies normalizes into, rather than challenges, existing power structures, the empirical studies test previously overlooked aspects of how politicians use social media and how these platforms shape political communication. Given the characteristics of the medium itself, the opportunity structure of social media is expected to form differently depending on political actor and vary by platform. Two types of political actors are examined more closely in the dissertation, politicians of populist parties and female candidates. The empirical studies support the idea that social media provides several opportunities for populist parties to strengthen their position relative to other parties. Populist parties’ evoke engagement and succeed in politicizing their followers’ feelings of indignation on social media. Through social media, they also maintain a strong adversarial position towards mainstream media. Female candidates’ on the other hand are more likely to utilize interactive opportunities of social media and they view social media as more important than male candidates do. Furthermore, the results show the importance of distinguishing between different social-media platforms; both in terms of how they are used by political actors and in terms of discursive opportunities they provide. Parties’ perceived competence over issues by voters were largely mediated through discussions on Twitter rather than forming an alternative issue agenda, indicating that the platform architecture of Twitter is more conducive to a normalized pattern of communication. News content shared on Facebook on the other hand, indicate a substantial influence of right-wing alternative media on this platform, pointing to a reinforcement of previously marginalized voices on Facebook.

Keywords: political communication, social media, parties, populism, political campaigning

Sammendrag

Denne avhandlingen utforsker mulighetene sosiale medier skaper for aktører og perspektiver som har hatt vansker med å slippe til på egne premisser tidligere, særlig populister og kvinnelige politikere. Avhandlingen består av flere empiriske studier som undersøker hvordan partier og kandidater anvender sosiale medier og hvordan politisk kommunikasjon utspiller seg på disse plattformene. Dette gjøres i lys av normaliseringsteori, hvis antakelse er at nye medier integreres i eksisterende maktstrukturer framfor å utfordre dem. Forskningsfunnene diskuteres også ut ifra mulighetsstrukturer og digital plattformarkitektur for å forstå akkurat hvordan politikk endres når den utspilles på sosiale medier. Resultatene indikerer at sosiale medier har bidratt til et brudd fra tidligere normaliseringsmønstre, både for konkurranse mellom partier og innad i partier. Mulighetene sosiale medier skaper varierer tydelig avhengig av aktør og politisk kontekst. Mine studier gir støtte til antakelsen om at sosiale medier styrker populisters posisjon sammenliknet med andre partier. Populistiske partier skaper større engasjement og lykkes med å politisere følgernes indignasjon, samt opprettholde sin posisjon som motstander til tradisjonelle medier. Når det gjelder kvinnelige politikere så er de både mer tilbøyelig til å se sosiale medier viktigere enn mannlige kandidater og til å anvende dem interaktivt. Sett under ett viser studiene også at det er tydelige variasjoner fra mellom plattformer når det gjelder diskursive muligheter og anvendelse. Borgernes oppfatning av partienes sakseierskap samsvarte med hvordan disse partiene og sakene ble diskutert på Twitter. Plattformens arkitektur normaliserer til en viss grad politisk kommunikasjon i tråd med eksisterende maktstrukturer. I kontrast skaper den mer anvendte plattformen Facebook diskursive muligheter som bryter med etablerte maktstrukturer. Dette demonstreres av hvor hyppig nyhetssaker fra høyreorienterte, alternative medier deles.

Nøkkelord: Politisk kommunikasjon, sosiale medier, partier, populisme, politiske kampanjer.

List of tables and figures in the cover chapter

Table 1. Overview over articles and research questions addressed ………...11

Table 2. Overview of data material methods ………………………………..53

Figure 1. Theoretical framework…………………………………………….22

Contents

Acknowledgements

1 Introduction 1 1.1. A tale of two sides: How is social media reshaping politics? 5 1.2 Research problem 8 2 Theoretical perspectives and previous research 14 2.1 Ruling the void: Social media and representative democracy 14 2.2 Theoretical framework 22 2.3 The (social) medium is the message 23 2.4 Politics as usual? 30 2.5 Summary 43 3 Research approach and methods 45 3.1 The political context and generalizability across platforms 47 3.2 Data and methods 49 3.3 Ethical considerations 57 4 Summary of articles 58 5 Concluding discussion 62 References 69

Articles I- V

List of articles:

1 Sandberg, L. A. C., & Öhberg, P. (2017). The role of gender in online campaigning: Swedish candidates’ motives and use of social media during the European election 2014. Journal of Information Technology & Politics, 14(4), 314-333.

2 Jacobs, K., Sandberg, L. A. C., & Spierings, N. (2020). Twitter and Facebook: Populists’ double-barreled gun? New Media and Society, 22(4), 611-633.

3 Sandberg, L. A. C., & Jacobs, K., & Spierings, N. Populist legislators on the people’s platform: Facebook adoption and affective reactions in Austria, the Netherlands, and Sweden. Manuscript submitted for publication.

4 Sandberg, L. A. C. Socially mediated issue ownership. Manuscript accepted for publication in Communications: The European Journal of Communication.

5 Sandberg, L. A. C., & Ihlebæk, K. A. (2019). Start sharing the news: Exploring the link between right-wing alternative media and social media during the Swedish 2018 Election. Statsvetenskaplig Tidskrift, 121(3), 421-440.

To Marina and Raimondas

‘And he said, “The Age-of-muscular-force is dead. The Age-of-nervous-force has killed him with the knife he holds in his hand; and silently and invisibly he has crept up to the woman, and with that knife of Mechanical Invention he has cut the band that bound the burden to her back. The Inevitable Necessity is broken. She might rise now.” And I saw that she still lay motionless on the sand, with her eyes open and her neck stretched out. And she seemed to look for something on the far-off border of the desert that never came. And I wondered if she were awake or asleep. And as I looked her body quivered, and a light came into her eyes like when a sunbeam breaks into a dark room.’ Three Dreams in a Desert by Olive Schreiner (1855–1920)

Acknowledgments

I dedicate this thesis to my dear friends Marina and Raimondas, whose support and friendship helped me see this dissertation through. Thank you!

Before I started writing this dissertation, I worked together with Professor Marie Demker as a research assistant at the Department of Political Science in Gothenburg. Marie’s intellect and passion for research inspired me to pursue a research path in academia. She also continuously supported me in my efforts to acquire a Ph.D. position. Many colleagues from the department also encouraged and supported me to pursue an academic career and I would like to direct a special thanks to Professor Jon Pierre in this regard. During that time I also got to collaborate with Professor Ulf Bjereld. Our work became the foundation of what I later came to study in this dissertation. Ulf later became my second supervisor and I am truly grateful for his guidance. He has a very clear sense of what needs to be done and often helped me to structure a course of action. When I started my Ph.D. position at the Department of Media and Communication in Oslo, I met my main supervisor Professor Eli Skogerbø for the first time. I was asked to request a supervisor before my arrival and although I had never met Eli, she never gave me any reason to doubt my choice. I am very thankful for all her thoughtful support and supervision. Associate Professor Jon Polk who later in the process became my third supervisor had long functioned as an informal supervisor and mentor. Thank you for all the crucial insights you have provided and your guidance throughout this process.

During my Ph.D. period, I have visited three different departments. I would therefore like to thank the Faculty of Humanities in Oslo for their generous support to make this possible. During my first guest research stay, I returned to Gothenburg and the Department of Political Science there. Together with Patrik Öhberg I conducted research that would later be published in Journal of Information Technology & Politics. I learned a lot from Patrik. He is someone who gets things done efficiently and we had great fun working on this project together.

I also spent a sublime time in Nijmegen at the Radboud University’s Department of Sociology working together with Kristof Jacobs and Niels Spierings, two of the most inspiring researchers within my research field. In a very smooth and effortless collaboration, we managed to produce two research articles in a relatively short amount of time, out of which one got published in New Media and Society. Kristof and Niels contributed a great deal to my academic growth. I will always be very thankful for their generous guidance. I would also like to thank everyone at the Department of Sociology for the good times and for taking such good care of me.

During the last six months of my Ph. D. period, I visited The Digital Social Science Core Facility (DIGSSCORE) in Bergen. I would like to thank Professor Elisabeth Ivarsflaten and my colleagues here for welcoming me to Bergen and providing a new academic home. A special thanks to Ingrid Ovidia who made my last months as a PhD candidate truly enjoyable.

I am really thankful to Erik Knudsen at DIGSSCORE, who stepped in as my opponent during the obligatory “end- seminar” in the last minute. Erik’s constructive feedback was exceptionally helpful and inspiring (in a short amount of time he managed to thoroughly read my draft and all articles, during the seminar he had also prepared a 34 pages long power point presentation with his comments). His last advice that I will pass on is that- A thesis is never truly finished, it is eventually abandoned and the best theses’ are the ones that are submitted.

One of the best perks of being a doctoral candidate is all the great colleagues and fantastic researchers you get to meet during summer schools and conferences. I can’t thank you all in name here since so many of you have inspired me and contributed to improving my work. However, I want to direct a special thanks to my computer science colleagues Shaza Jaradat and Nima Dokoohaki for all their help and collaboration. I must also thank Steffen Krüger, my friend and colleague at the Department of Media and Communication (Oslo) who made my time there enjoyable – and surprisingly - physically challenging. Steffen and I played some intense table tennis matches, and despite giving it everything I had he usually beat me quite soundly.

I also want to thank all my dear friends and my parents, Laine and Per. Lastly, thank you Lars Erik for all your invaluable support through our time together and to our daughter Leonor whose love, energy and humor helped me push through and overcome whatever obstacle faced me along the way.

1. Introduction

To ignore social media is to turn a blind eye to the reality of politics today. Social media platforms structure political information flows and the way citizens and politicians interact and engage with one another. Such new communication technology has the potential to alter the organizational landscape of democracies in that certain types of parties and political actors can more readily adapt the technology and benefit from it (Ward & Gibson, 2008). To increase our understanding of how social media influence contemporary politics and political communication processes, this dissertation examines whether and how social media form opportunities for “previously disadvantaged” political actors and perspectives.

In order to address the potential differences in how political actors use social media, the dissertation builds on two competing theories that suggest opposing patterns of effects on party competition and political communication due to new communication technology - normalization and equalization (Jacobs & Spierings, 2016; Ward & Gibson, 2009). The equalization hypotheses emphasize the potential of new technology to create more even opportunities for political candidates, removing barriers that favor some actors over others (Barber, 2001). These barriers can be understood as systemic structures that in most instances create advantages for already established or privileged actors such as the party or media system. Technological innovations are in this understanding leveling the playing field, making it easier for previously disadvantaged political actors such as minor party candidates or marginalized groups to gain power (Gibson & McAllister, 2015). The normalization hypotheses, in contrast, posits that new technology eventually will become incorporated into existing political structures (Margolis & Resnick, 2000), that is, they will become normalized. Applied to political parties, normalization essentially implies that large traditional parties will come to predominate as they do in other media. The two conflicting assumptions pertaining to the normalization/equalization framework thus concern whether new technological developments, such as social media, shift political power or primarily reflect existing power relations. However, new technology might not necessarily facilitate changes in the uniform manner that the normalization/equalization theoretical framework suggests, as the opportunities for different actors that can be described as previously disadvantaged might not form in a consistent way. As a result, we lack an understanding of how elite power relations in politics change with the coming of social media.

This dissertation addresses that research gap. It sheds further light on how politics changes when taking place on social media by incorporating perspectives on social media opportunity structures and digital platform architectures. Overall, it is argued that how one political actor communicates on one platform does not translate to another (due to specific platform architectures), and if one type of party makes successful use of social media it does not mean that it is a one-size-fits-all model (i.e., opportunity structure forms differently).

Delving down, this dissertation examines the specific dimensions of the overarching normalization/equalization framework, which consists of three dimensions; a relational, functional, and discursive dimension (Schweitzer, 2011). I argue that it is crucial to study all three dimensions to understand the wider political implications of social media. The relational dimension concerns how new technologies either normalize into existing structures of party competition or how they exert an equalizing effect on these structures. Previous research often examines this dimension by comparing less resourceful candidates from minor parties to those of bigger parties (Jacobs & Spierings, 2016). Within the relational dimension we can distinguish amid relations between (inter) parties and within (intra) parties. Social media use has implications for both and several studies in this dissertation examine both inter- and intraparty aspects of online campaigning. The functional dimension concerns the actual use of social media in various ways and political actors can be assumed to approach these functionalities differently (Schweitzer, 2011). Moving away from merely assessing indicators such as adoption and activity levels, the dissertation also seeks to establish to what extent social media is utilized differently by previously disadvantaged actors. Finally, the discursive dimension implies the communicative behavior on social media and is addressed by looking at the extent to which communication on these platforms resembles a normalized discursive pattern.

By addressing all three dimensions of normalization/equalization, this dissertation makes a comprehensive contribution to research at the intersection of technology and politics. Crucially, I argue that it is necessary to link potential political implications of social media to the specific features distinguishing social media from traditional media and other web-based environments, as well as recognizing that individual social media platforms themselves differ in their architecture. By doing so, this dissertation offers a nuanced understanding of social media’s role in contemporary politics, rather than pointing to social media as either transforming political power in one specific direction or having no effects on politics, as the equalization/ normalization perspectives implies. Furthermore, the results show that it is necessary to

2 distinguish between different previously disadvantaged actors, as social media presents them with distinct opportunities.

Specifically, the dissertation focuses on two sets of actors that are dissimilar in many respects – populist parties and female candidates – but have in common that they have been defined as “disadvantaged” in terms of absent or negative traditional media coverage (Jacobs & Spierings, 2016). These are certainly not all actors that can be defined as previously disadvantaged in relation to traditional media. However, they represent two important groups of actors – those that challenge the liberal democratic regime, in the case of populist parties (e.g., Galston 2018; Norris & Inglehart, 2019), and those that favor the development of democratic principles, in the case of female candidates (e.g., Wang et al. 2017). The choice of these two political groups out of the wider spectrum of actors who could be defined as disadvantaged therefore helps examine how the two opposite forces in democratic politics interact with social media use.

Taken together, the empirical studies conducted in this dissertation have identified factors that moderate or mediate key relationships within the normalization framework (such as electoral context, party structure and party ideology). Theoretically, several main components of social media opportunity structures have been linked to the actors examined and to communication patterns on these platforms. The overall results of the empirical studies suggest that social media disrupts normalized patterns of party competition, both between and within parties. Politicians of populist parties adopt social media platforms to a lesser extent but are more active when they do so. More crucially, they have a distinct way of utilizing social media by seizing the opportunity to evoke emotional reactions on Facebook and openly criticizing the news media on Twitter. This points to an equalizing potential and possibilities to strengthen their position in the electoral system. The party leaders of populist parties are particularly active, suggesting that social media has implications for intra-party dynamics in strengthening the role of the party leadership rather than weakening it. On the contrary, this result points to a normalized pattern of intra-party dynamics as these parties are distinguished by the strong party leader. Social media use therefore shows little potential of altering the power balance within these parties.

This dissertation tests the role of gender in the European parliament election as it is likely that second-order elections generate more added value of social media use. The results show that women candidates valued social media platforms as campaign tools more than male candidates did and they were more active and interactive during noncampaign periods. Apart from gender differences, these results highlight the importance of distinguishing between campaign vs non- 3 campaign periods when studying politicians’ social media use. Furthermore, it underlines the importance of electoral context, as the candidates running for office in the European election are lesser known and less present in the media. In this situation utilizing social media therefore has added value. Considering this, social media offers greater campaigning opportunities to candidates in general, as well as providing an (intraparty) equalizing potential for the female candidates.

Another article in the dissertation examines to what extent parties’ perceived issue competences (i.e. issue ownership), as measured in representative surveys, can explain variation in how parties are addressed together with political issues on Twitter. The result suggests that a socially mediated issue ownership dimension forms on Twitter and, to a large extent, it correlates with issue ownership as measured in surveys. Since traditional news media are key both in forming issue preferences (agenda setting) and strongly influence political discussions on this platform, it is likely that news media influence the consistency with which parties are discussed together with the issues they are perceived to own by the voters. The discursive issue ownership dimension on Twitter thus points to a normalized discursive pattern, as this aggregated and mediated issue agenda did not deviate from voters’ perceptions.

The last article of the dissertation also examines the discursive dimension of normalization by looking at whether and how Facebook reinforces the position of marginal actors as new agenda- setters in a high-choice media environment. More specifically, the influence of alternative far- right media sites on Facebook is examined in terms of shares and engagement compared to traditional news media. The result points to an increased visibility of content from right-wing alternative media sites on this platform. Many of the sites included in the analysis can be classified as far-right, and thus Facebook potentially reinforces the position of these previously marginal actors and controversial views. This indicates that Facebook provides such actors with unique discursive opportunities to influence the wider public agenda. In contrast to Twitter, content that engaged and became visible on Facebook therefore signifies a break from normalized discursive patterns.

In sum, the different empirical studies in this dissertation suggest opposing patterns of potential effects of social media, pointing towards both normalization and equalization trends. In order to understand why this is the case, the dissertation suggests that it is important to recognize that opportunity structures of digital and social media form more closely in relation to the specific types of actors. In relation to other parties, specific opportunities form for populist parties. 4

Within these parties, it is the party leader that might strengthen his or her position vis-à-vis the other party representatives. Other opportunities seem to manifest for lesser known female candidates in a second-order electoral context compared to their male counterparts. Social media use might trigger a shared underlying mechanism when compensating for a previously disadvantaged position, but it also seems to matter whether political actors belong to a historically marginalized group or whether they are fringe or in an oppositional position. By incorporating perspectives on opportunity structures and digital architectures, the interaction and dynamic between social media platforms and political actors or communication patterns can be addressed more comprehensively. In terms of communication patterns, Twitter discussions resembled a mainstream agenda whereas far-right partisan news gained visibility on Facebook, further pointing to strengthening (previously) marginal far-right perspectives. This is suggested, to either be related to the fact that social media consist of multiple online public spheres forming a layered structural environment (Ceron & Splendore, 2019), or that the digital platform architectures provide different discursive opportunities.

This chapter is structured as follows. First, in more general remarks, I discuss how social media directly and indirectly contribute to changing the dynamics of politics. This is followed by a presentation of the specific research problem and research question that is addressed in the remainder of the dissertation.

1.1 A tale of two sides: How is social media reshaping politics?

New technological advances repeatedly raise questions of whether these innovations will lead to a new social and political order (Chadwick, 2013). In many ways, social media platforms have transformed the structural conditions of communication, but to what extent and how this transformation has impacted politics is more difficult to evaluate. One reason for this is the problems of analyzing a medium that is evolving more rapidly than the pace of the scholarship. Social media platforms also differ in their digital architecture, so, for example, while politicians use Facebook in one way, they adopt a different communication strategy on Twitter (Bossetta, 2018; Stier, Bleier, Lietz & Strohmaier, 2018), which makes establishing causal mechanisms and general trends more difficult. In the early days of social media and the Internet, research findings supported a politics as usual scenario, particularly in established democracies (Chadwick, 2006; Gainous & Wagner, 2013; Norris & Curtice, 2008; Bimber & Copeland,

2013). Some of these findings are driven by the limitations of studying a medium that has evolved drastically in only a few years.

A crucial aspect of how social media transform the structural conditions of communication is that the previously dominant actors, such as traditional broadcasters, have lost part of their hold on controlling the spread of content and ideas. They are no longer gatekeepers in the same way, as content spreads outside of their control. Parties and candidates can in this way shape and dictate their own content, bypassing mass-media filters. Politicians can to some extent reduce their dependence on traditional intermediaries, such as journalists by communicating directly with citizens, keeping control of the content, distribution, and timing of their messages. At the same time, social media increasingly influence journalists’ news judgement and they more often rely on social media as a source (McGregor & Molyneux, 2018). By not speaking directly to journalists and allowing them to ask questions, politicians can take full control over their message as it appears in traditional media content. Journalists frequently refer to tweets in their news reporting (often the case with Donald Trump in the U.S.). Another example of this changed dynamic is the former Swedish Prime Minister, Carl Bildt, who, accused of corruption, referred journalists to Twitter instead of giving interviews.1 In this manner, politicians can shape relationships with media in a way that gives them more control over their message. Donald Trump has even suggested that if it were not for Twitter, he would not be president.2 In addition, social media allows politicians to criticize news coverage and content publicly through their own channels, something that Trump regularly does on Twitter by referring to the media as “fake news.”3

The network effects of social media might amplify the transmission of a political message through social connections and make social media a valuable part of an election campaign. Typically, strong partisans are likely to receive information from parties suggesting that digital activism is mostly a supplementary channel for members’ input (Copeland & Römmele, 2014; Gibson, Greffet, & Cantijoch, 2017), but the dissemination of such content to a wider audience points to a two-step flow of information that increases the likelihood of parties’ ability to reach out (Norris & Curtice, 2008). User-generated content supporting a party or a candidate is

1 https://www.aftonbladet.se/nyheter/a/XwzzRm/sa-har-svarade-bildt--pa-twitter 2 Trump made this statement in a FOX News Channel interview: https://www.realclearpolitics.com/video/2017/03/15/trump_i_wouldnt_be_here_if_it_wasnt_for_twitter_i_h ave_my_own_form_of_media.html 3 https://www.washingtonpost.com/opinions/global-opinions/trump-is-spreading-his-fake-news-rhetoric- around-the-world-thats-dangerous/2019/11/19/a7b0a4c6-0af5-11ea-97ac-a7ccc8dd1ebc_story.html 6 another way to bypass traditional mass media, and the authenticity of user-generated content might have a powerful effect. However, it also can involve risks if, in the end, the message is not one for which the candidate or a party would wish. Examples of using such content for purposes of sabotage include the circulation of a fake feminist party program in Sweden, that received a lot of negative attention before the party managed to take control over the situation.4

Parties may also use social media data to analyze various measures of their efforts and increase the power of analytics. The 2012 campaign by President Obama pioneered new ways of electioneering in its use of innovative data analytics. In particular, the campaign’s use of social media to facilitate efforts to seek out and mobilize undecided voters contributed to its success (Katz, Barris, & Jain, 2013). One advantage of using social media analytics is the ability to target specific voters with tailored messages. On the other hand, this use of social media data is less transparent, and it opens up possibilities for negative campaigning. In recent years, the power to shape and even manipulate voter choices trough social media have become a great concern. Digital data and analytics introduce new factors, such as foreign influence and partisan news, as potential drivers of voter decision-making.

In a high-choice media environment, concerns arise about social media contributing to polarization, fragmentation, selective exposure, information asymmetries, hate speech, and manipulation (Van Aelst et al., 2017). Leading to an erosion of the common public sphere, this development might create a considerable challenge for democracy. It has even been argued that no company has contributed more to the global collapse of basic tenets of deliberation and democracy than Facebook (Vaidhyanathan, 2018). If, and how social media influence the functioning or well-being of established democracies is an ongoing debate. There are competing views on the importance and direction of this influence, i.e., to what extent and how digital and social media have transformed politics and political life as we know them. A main divergent line is whether reliance on digital technology has fundamentally reshaped power relationships in politics, or if the technology is merely reflecting those already in existence. This is part of the research problem addressed in this dissertation and will be discussed more thoroughly below.

4 https://www.metro.se/artikel/borgerlig-debatt%C3%B6r-bakom-fejkat-valmanifest-fr%C3%A5n-fi-xr 7

1.2 Research problem

The overarching research questions this dissertation address are whether social media creates an advantage for previously disadvantaged actors and perspectives and if so, how? The disadvantaged position essentially reflects actors marginalized relative to the established media or the party system at large (Gibson & McAllister, 2015). However, much research is still unclear about why social media would have an impact on politics, and on which political power relations social media would have such an impact (Jacobs & Spierings, 2016). The equalization perspective is not precise over which specific organizations will benefit from social media and the literature on equalization trends uses the resource argument quite inconsistently (Jacobs & Spierings, 2016). Social media is cheap and easy to use; still, it requires resources and time to maintain an active presence. Similarly, the opportunity to bypass traditional media is central within equalization theory, although inconsistent with the opportunity to target traditional media through social media. Therefore, theorizing how the unique social media attributes influence politics or whether politician’s actual usage fits these theoretically derived expectations, to avoid resorting to either techno-optimism or pessimism, is crucial. For that reason, this dissertation links the equalization perspective to the opportunity structures of social media.

Online opportunity structures have been defined as the “factors inherent to the online media system” (Engesser, Fawzi, & Larsson, 2017, p. 1280). The concept of opportunity structures is used similarly in this dissertation to distinguish the unique opportunities that form through social media platforms, what can be understood as the inherent effects of the technology itself (Bimber, 1999). Following the understanding that each medium embodies a bias, an inherent predisposition to communication possibilities or constraints (Innis, 2008), this will both influence how different political actors use a medium (to which I refer as opportunity structure) and condition what type of content the platform is likely to generate (Bossetta, Dutceac Segesten, & Trenz, 2017). Therefore, it is important also to distinguish between different social media relative to the potential opportunities it provides for political communication. What has been defined as the digital architectures of social media shape how these platforms communicate politics (Bossetta, 2018). This concept includes the openness of a network, its functionality, filtering, and datafication. In the empirical studies, the focus lies on Facebook and Twitter, the most widely used social media platforms in politics to this point. Since the communicative opportunities form differently depending on the digital architecture, it is likely that politicians adopt and use Facebook and Twitter for different purposes.

A major advantage with the normalization/equalization framework is that it allows (or necessitates), the inclusion of both structural elements of communication technology as well as actor-centered perspectives. Applied to party competition, this theoretical framework allows one to formulate testable hypotheses on why some parties or candidates might benefit from new technologies over others. Critique raised against this framework involve the focus on over- simplified two-dimensional approaches to equalization versus normalization and it has been argued that such “one size fits all” explanations should be rejected (Ward & Gibson, 2009). This overarching theoretical framework therefore needs to be complemented with how political actors interact with the specific opportunities shaped by social media. This dissertation further argues that social media opportunity structures do not exist in isolation but form against the political context and in relation to political actors.

Ward and Gibson reason that “systemic and technological opportunity structures provide the broad political and technological parameters within which political organizations operate” (Ward & Gibson, 2008, p.35). Within national boundaries these structures may alter the extent and how different political actors use technology. As I understand the previously disadvantaged position, it forms in relation to systemic structures such as the party system or media system. Both the traditional media and party system typically favors already established political actors. In many western European countries, mainstream news media even is or have been owned by parties or publishers loyal to them, usually conservative or social democratic (Ward & Gibson, 2008). Social media provide opportunities to sidetrack these structures, which allows other political actors to level the playing field. In this understanding, a disadvantaged position vis-à- vis traditional media or party system is moreover not limited to political parties, it can also include other actors challenging the establishment, such as the broader far right movement.

As discussed above, a crucial shortcoming of the literature on equalization is the lack of making a distinction between previously disadvantaged actors. For example, Ward and Gibson (2018), argue that, outsider, oppositional, or fringe organizations are more likely to benefit disproportionately from the rise of new communication technology, and in doing so they pose more of a challenge to the mainstream political establishment. However, previous literature also supports the ability of historically marginalized groups such as women candidates to appear as a function of equalization (Jacobs & Spierings, 2016). Both perspectives challenge the “politics as usual” scenario but opportunity structures of social media are not likely to form in the same way for all actors that can be considered previously disadvantaged such as women, minor or fringe party candidates. 9

Importantly, equalizing effects of technology would have very different consequences for democracy depending on the type of previously disadvantaged actor. The rise of populism can be seen as a result of weak party systems and challenging to representative democracy (Mair, 2013), whereas female representation is central to the idea of political representation. Historical political subordination of women makes women representation in democratic societies crucial for legitimacy in political decision making (Mansbridge, 1999). Subordination of women or lack of women in representational bodies, will be a democratic issue in all liberal democracies that that build on principles of individual liberty, equal justice, and equality of opportunities (Held, 2006, p.88). As this dissertation examines both female candidates and populist MP’s as a function of equalization, this does not imply that they are equally disadvantaged or similarly disadvantaged in a normative sense. On the contrary, they differ in their disadvantaged position, and that is why comparing the two cases in their supposedly disadvantaged role according to the equalization theory in relation to social media opportunity structures gives meaning.

Research questions and sub-studies

The first three articles (see table 1 below), center on the political actor in relation to social media opportunity structures. In Article 1, the focus lies on women candidates running for election and how their motives for using social media in their campaign efforts and actual use differ compared with their male counterparts. Article 2 and 3 focus on politicians belonging to populist parties and their use of social media. More specifically, how populist Members of Parliament (MP) make use of social media, compared with MPs of other parties, and how this might be linked to social media opportunity structures. Theoretically and empirically the focus in these articles is linked to social media opportunity structures and concerns both the relational and the functional dimensions of normalization.

Article 4 and 5, concern a discursive dimension of normalization, examining platform specific (structural) communication patterns linked to potential discursive opportunities. The capacity to set the agenda is a fundamental resource in gaining political influence. With established actors losing their gatekeeping function, social media opens for previously marginalized voices. The extent to which the discursive patterns on social media reflect conventional patterns in opinion formation and to what extent social media enables previously marginal views to gain visibility, are examined in these two articles. Article 4 in the dissertation more specifically examines the extent to which social media mediates mainstream opinions, by examining how

Twitter discussions linking parties with political issues correspond with voters’ perceptions of issue ownership. Compared to prevailing public opinion on party-issue ownership, the potential deviations regarding what issues are addressed with what parties on Twitter are measured to test whether issue ownership theory can explain discursive variations on Twitter. To what extent a mediated issue ownership dimension corresponds to voter’s perceptions of ownership and the relationship between these two processes, relate to opportunities or constraints that arise depending on platform and its user base. Article 5 similarly examines communication patterns but on Facebook. By comparing news-article sharing practices around content from right-wing alternative media and traditional news media, it examines the extent to which this platform enables previously marginal actors and perspectives to gain visibility.

Table 1. Overview over articles and research questions addressed

Article Title Research questions number Article 1 The role of gender in online (1) Do women candidates consider social media to be a campaigning: Swedish candidate’s more important campaign tool than do male motives and use of social media during candidates?

the European election 2014 (2) Are female politicians more likely to use Twitter to enhance the political messages of their party or their personal agendas?

(3) Will women politicians behave differently than male politicians on Twitter by being more active and communicating more interactively?

Article 2 Twitter and Facebook: Populists’ (1) Do politicians from populist parties differ from double-barreled gun? nonpopulist ones in terms of naming and shaming on Twitter and anger activation on Facebook?

Article 3 Populist legislators on the People’s (1) How do the attributes and architecture of Facebook, Platform: Facebook adoption and in theory, link to the key ideological and organizational affective reactions in Austria, the characteristics of populist parties?

Netherlands and, Sweden (2) Are populists more prone to use Facebook pages?

(3) Do populists post more emotional eliciting messages on public pages?

Article 4 Socially Mediated Issue Ownership What similarities and differences emerge in voter perception of parties’ issue ownership compared with a mediated issue-ownership dimension on Twitter?

Article 5 Start Sharing the News: Exploring the (1) To what extent did right-wing alternative news Link between Right-wing Alternative media manage to gain visibility and create engagement Media and Social Media during the on Facebook during the election campaign in Sweden, in comparison to established news media? Swedish 2018 Election

(2) What type of political topics evoked most engagement?

Both article 4 and 5, more strongly relate to the structural features of communication on social media. By extension, this impact actor’s capacity for action, in that certain actors are either restricted or enabled by the structural element of communication.

Structure of the remaining introductory chapter

The extent to which new technologies are exploited and the ways in which they are used are arguably shaped by the political systems within which the technologies are immersed. In order to understand how new technology interacts with the functioning of democracy, it is important to link research to existing literatures or place it within current political and social contexts, something than internet studies in general neglect (Ward & Gibson, 2009). I will therefore begin the first section in the theoretical chapter by relating social media to existing trends within representative democracy as this forms the relevant political context to the empirical studies. Building on the literature on representative democracy and contemporary challenges to political representation in established democracies, I will discuss the role of social media in these processes and situate the empirical studies within a unifying framework of representative democracy.

Moreover, any explanation as to why social media would reshape power relationships in Western democracies would be insufficient without connecting theoretical perspectives to the functioning of social media. In the following chapter, I continue with discussing the characteristics of social media platforms and how political communication on these platforms can be understood. I proceed with a description of Facebook and Twitter, their similarities and differences, as these platforms are examined in the empirical studies. This discussion forms the basis for the next part of the theoretical chapter, where I elaborate on how opportunity structures of social media form for the previously disadvantaged actors linked to the normalization/equalization framework. Results from previous studies on parties’ and candidates’ adoption and use of social media, as well as what this dissertation examines as a discursive dimension of normalization are discussed in relation to opportunity structures of social media. The dissertation uses a combination of various methods and different types of data, which will be addressed and discussed in subchapter 3. This is followed by a summary of the individual studies and in the last chapter I will address the main conclusions and discuss these.

2. Theoretical perspectives and previous research

The general logic of equalization and normalization is based on the rise of digital technology brought about by the Internet, but that logic is not sufficiently elaborated relative to the unique characteristics of social media (Gibson & McAllister, 2015; Jacobs & Spierings, 2016). Prior to the emergence of social media, politics on the Internet has mainly been politics as usual (Margolis & Resnick, 2000). New communication technology, on the contrary, appears to be hastening the alleged decline of traditional representative democracy and accelerating some of the more long-term observable trends such as individualization and disaggregation (Ward & Gibson, 2008). In the first part of this chapter, social media is situated within broader trends of representative democracy, drawing on the literature of political representation. These trends have direct bearing on the empirical studies, such as the growth of populism and an increased importance of issue competition. The aim of this discussion is to place the empirical studies within their current political context and to provide a unifying mantle to the potential implications for democracy at large. The following sections of this chapter proceed from the specific theoretical framework of the dissertation that builds on social media opportunity structures and the equalization/ normalization perspective.

2.1 Ruling the void: Social media and representative democracy

To understand and explain political developments, especially related to the growth of social media platforms, it is important to recognize that democracy and political communication are closely intertwined. It has been argued that any theory with regards to political phenomena is half-blind if it does not include the media, and vice versa regarding theories about media (Strömbäck, Ørsten, & Aalberg, 2008, p. 11). Contemporary political communication in advanced democracies reflects the disconnection of publics from institutions of press and politics due to partisan dealignment trends and growing social divides (Bennett & Pfetsch, 2018). However, what this implies for political communication remains largely unanswered. While plenty of research studies take interest in the rise of populism and populist communication, few links this growth to the role of declining institutional legitimacy and its relation to disruptive and disconnected public spheres (Bennett & Pfetsch, 2018). In order to address how social media form opportunities for previously disadvantaged political actors and

14 perspectives, it is therefore important to situate social media within these broader trends and contemporary challenges to representative democracy.

Furthermore, contemporary political communication processes challenge theoretical approaches in the field. Many traditional theoretical concepts assume functionalities of political communication and functionalist relations among core actors that no longer hold true (Bennett & Pfetsch, 2018). Fluid and transitory hybrid media systems, transnational flows of communication, and complex political networks created by digital and social media requires that we rethink some classical theoretical premises (Bennett & Pfetsch, 2018). Concepts such as agenda setting, and issue ownership are rooted in understandings of media and politics that in many ways have lost their fundamental logic today. Issue ownership, the idea that parties will promote issues on which they are perceived as competent by the voters, was formed as a concept at a time where traditional media was central in opinion forming processes and in conveying party’s policy positions to citizens (Bélanger & Meguid, 2008; Petrocik. 1996). To update the field of research, the complex two-way interactions with mainstream news media and other political influences of information flows from social media and digital networks, needs to be included.

A long asserted trend in Western societies are the weakened ties between parties and citizens, that can be observed in a decline in party membership, more volatile voting behavior, weakened party identification and decreased trust in the system of party representation (Katz & Mair, 2018; Rohrschneider & Whitefield, 2012; Rokkan, 1970; Selle & Svåsand, 1991). Simultaneously, there has been a corresponding rise of new forms of political organization: single issue campaigns, new social movements, and radical direct-action protest (Ward & Gibson, 2008). The liberal democracy model that characterizes most Western democracies, requires that political parties function as a link between citizens and decision makers by aggregating and articulating voters’ interests, and engagement of members in party organizations (Bjereld, Blombäck, Demker & Sandberg, 2018). A weakened link between parties and citizens, will from this perspective pose a challenge to representative democracy. In, Ruling the void, Peter Mair goes as far as claiming that the age of party democracy has passed. Parties are becoming so disconnected from the wider society that their competition lacks meaning, and they are therefore no longer capable of sustaining democracy in its present form (Mair, 2013 p.1). On the one hand, parties are failing in their capacity to engage ordinary citizens, which primarily results in decreased party identification and declined party

15 membership. On the other hand, political leaders direct their ambitions towards external public institutions. Parties are used as steppingstones to other offices and positions resulting in a process of mutual withdrawal (Mair, 2013).

According to Mair and others, this process has had two notable concomitants. The widening gap between rulers and ruled facilitate a populist challenge that is now a feature in many advanced European democracies. Populists’ share among other things, a hostility to what is seen as the national political class (Mair, 2013 p. 19; Mudde, 2004). It becomes easier to mobilize voters positioning the interests of a ruling class and media elites against the people’s concerns when parties are weak. The other closely related trend is according to Mair, the growing acceptability and legitimation of non-political or depoliticized modes of decision- making. The EU is a clear manifestation of this tendency and when power moves from elected representatives to the EU and other international organs, of which accountability is more difficult to exercise; politics becomes de-politicized (Mair, 2013 p. 19). When power becomes more diffuse or becomes vested in supranational organizations, it becomes more difficult for citizens to hold national elected representatives to account (Follesdal, & Hix, 2006). At the same time, many of the concerns facing society today are global such as the environment, migration and global economy which makes them difficult to solve nationally. However, democratic legitimacy is bound to the nation state and its borders.

Both globalization and the fundamental changes in communicational technology pose challenges to the principle of territory as the basis of political power. A fragmentation of the public sphere and an outflow of political power through increased dependencies, global capital movement and supranational collaboration, risk draining power from political institutions (Bjereld et al., 2018). As a result, the question of which representative role the parties should fill arises. If the institutions that govern political processes detach from political attitudes and behaviors, this leads to a political vacuum and frustration among citizens, which forms a breeding ground for populist movements that put the people against the elites (Bjereld et al., 2018). At the same time, the growth of populist and radical right parties in many countries is supported by so-called “alternative media” and digital networks that bypass routines of conventional journalism and attacks traditional media as elitists (Bennett & Pfetsch, 2018). In the light of this, it is imperative to address to what extent social media reinforces the right-wing populist message.

When parties become less relevant for citizens political engagement and identification, citizens either abandon politics or retreat into more specialized and ad hoc forms of representation. Trends supporting this notion are increased electoral volatility, a lack of stable commitment which also make voters decide later what party to vote for, and membership decline (Mair 2013, p. 16). In individualized times, fewer citizens seek out collective movements formed around ideologies that claim to cover all societal issues. Communication technology facilitate forms of networks and engagement around single issues instead of committing to a specific ideology or package deal that is entailed when joining a party (Bjereld et al., 2018). If forms of political engagement particular to social media platforms, such as increased focus on single issues, contribute to a more individualized form of political engagement, this hampers political parties’ legitimacy to represent citizens. Lack of identification with parties and declining trust makes it difficult for parties to represent a social base or collective identity. A more individualized society in itself is not a democratic problem but if parties become representatives of the opinions of citizens and not the interests of groups’, this leads to decreased levels of trust in parties over time (Bjereld et al., 2018). Political parties are in this way crucial in their capacity to represent different communities within a nation and have the democratic legitimacy to do so.

In principle, social media opens up new ways that parties can strengthen their linkages to the electorate. In Sweden, there are recent examples of how new and previously disadvantaged parties have managed to harvest latent forms of political engagement through digital and social media into membership, party activity and votes. Both the Pirate party and the Feminist party were very successful in this respect (Blombäck & Sandberg, 2018), but the engagement around both parties turned out to be relatively shorted lived which likely reflects the nature of digital activism in our time. Things might have turned out differently for these parties if they had managed to get seats in the national parliament (both parties however achieved to get seats in the European Parliament). Rather than strengthening the bond between parties and citizens, parties tailored toward niche interests and special issues is a symptom of a dealignment trends (Mair, 2013). Others argue that the proliferation of e.g. environment and animal-rights organizations should be seen as a significant counter-trend to the decline of established parties (Ward & Gibson, 2008; Kriesi et al., 1995). Overall levels of participation in Western societies are not necessarily declining, but citizens now seen more willing to support single-issue campaigns and engage in unconventional forms of protest activity, rather than join broad-based catch-all parties (Ward & Gibson, 2008). Some parties are clearly better equipped at meeting these new demands and more successful at harvest latent forms of political engagement,

17 especially through new communication technology. This could instead point towards a realignment trend, rather than a crisis of political organizations (Dalton, 2018). Whether or not there is a uniform trend towards citizen disengagement, it is becoming clear that political engagement is taking place in other forms (Ward & Gibson, 2008). Moreover, although parties continue to be the main structuring element of politics (Fisher, 2014), their role in the process of representation has become critical because the lack of trust in the political elite which reinforces processes of political disillusionment (Smith, 2014).

The notion of the mass party model has also been challenged by the individualization of participation within organizations. Trends of centralization and professionalization of campaigning within parties are replacing traditional local campaigning activities of party members. These traditional forms are becoming superseded by national campaigning, particularly through the media (Ward & Gibson, 2008). Social media places the individual at the center and political information flows are adopted at an individual level, which arguably support trends of increased individualization. In the same way, the individual politician uses a personal platform on social media, which may make politicians less dependent on the party organization for making a political career. It would seem like these processes already shape the U.S political system. Both Barack Obama and Donald Trump can be used as examples of the media gaining a larger influence in nomination processes, and both were outsiders in relation to their respective party establishment (Jungherr et al., 2019). Their active and unique (although in a very different respect) use of social media during their campaign periods also distinguish them from traditional electioneering.

Not only can social media be said to contribute to reinforcing individualization trends within representative democracy. The largest social media platforms cross over boundaries of nations. Several recent protest movements are international and mobilize supporters mainly through social media. A recent example of how the international social media engagement can impact politics over national borders is when South- Korean pop fans (K-pop) allegedly helped foil Donald Trump’s return to the campaign trail by encourage people to register to attend Trump's event and not attend, through the social media platforms TikTok and Twitter5.

To sum up, trends of globalization and individualization challenge political representation through parties and obscures accountability. Although not clear-cut, social media and

5 https://edition.cnn.com/2020/06/22/asia/k-pop-fandom-activism-intl-hnk/index.html 18 disconnected public spheres would seem to contribute to reinforcing these trends and might in this regard further weaken the links between parties and citizens. Single- issue campaigns and populistic parties, are in this respect a sign of dealignment rather than realignment. Others have argued that political parties are not in crisis but in flux. As Norris (2002) puts it, they are evolving rather than dying. It has also been argued that new communication technology may be used to modernize representative democracy on a consumerist model (Ward & Gibson). In such a model, citizens are viewed more as consumers of public services and the focus is on value and efficiency and providing individuals with increased choice through access to information (De Vries & Hobolt, 2020; Ward & Gibson, 2008). Although the extent of systematic developments will be shaped by different opportunity structures depending on national context, Ward and Gibson (2008) suggest that we may be moving towards a more fragmented and more contested democratic model.

Changes brought by digitalization and the arrival of social media have, regardless of whether political representation through parties are declining or transforming, added a further layer to debates about the current political system, role of political parties and its organizational infrastructure (Ward & Gibson, 2008). Thus far, the chapter has situated social media within larger development trends of globalization and individualization, and their potential impact on political representation. What implication dealignment trends have on political agenda-setting processes and how social media potentially influence these will be elaborated on below.

Party competition and agenda-setting

Political attention matters. In our time, information is abundant while attention, at the same time, perhaps is the scarcest of all political resources (Jones, 2001). Influencing the political agenda, is crucial for political parties in a representative democracy. When citizens are more likely to operate on the basis of short-term considerations and influences, media plays a more important role in the agenda setting processes (Green-Pedersen, 2019; Mair, 2013). Today, inter-media agenda setting dynamics also make social media important in opinion shaping processes (McGregor and Molyneux, 2018). Due to the speed of communication and rapid spread of information, social media has become one of the fastest ways to shape opinion.

Unlike traditional news media, social media is a participatory medium. However, this shall not be understood as participation in a classical understanding, when conventional politics belonged to the citizens, according to Mair. Instead conventional politics has become part of an

19 external world which people watch from the outside (Mair, 2013 p. 43). This form of audience democracy is stronger when parties are weak, and weaker when parties are strong. Mair points out that “strong parties are difficult to sustain when politics turns into a spectator sport, and that it should turn into a spectator sport is hardly surprising given the fading of the real differences that divided the parties in the first place” (Mair, 2013 p. 44). Traditional media have tended to focus more on politics as a game rather than the content of politics, a focus on spectacle rather than ideas (Gurevitch, Coleman, & Blumler, 2009). Social media similarly seems to reinforce the insubstantial aspects of politics over the content of policy (Trilling, 2015). Similarly, negativity and attacks on political opponents seems to generate most engagement on social media (Lee & Xu, 2018). Social media might therefore more resemble a spectator sport than a medium for strengthening political participation, especially in forms of party engagement.

When the ideological differences between parties evaporate and the role of party members diminishes, campaign periods become more crucial and more is at stake for the parties trying to win new voters at each election while making efforts to keep the old. This also implies that issue-based voting increases and that parties’ issue competition grows in importance (Green- Pedersen, 2007). If trust in parties’ ability to handle the issues on the top of the political agenda becomes more important than representing different voter bases or the social bases in society (Knutsen, 2007), the parties themselves can be expected to react to or follow opinion rather than forming opinion. Mair argues that when voters become more “particularized” and the vote has become more free-floating, so have the parties themselves as a reaction (Mair, 2013). Consequently, the ideological differences might become even less clear which in a downward spiral would hamper trust, party affiliation and identification even more.

As a complement to input from party members, it has become more and more common for parties to use sophisticated polls and focus groups to get input on political processes in society (Bjereld et al., 2018 p.19). Replacing the importance of members with such working methods is easier the larger and financially stronger parties are. As party membership decline (Van Biezen, Mair & Poguntke, 2012), and parties’ face considerations of electoral success, they are encouraged to take cues from the electorate at large. This implies that the voice of ordinary voters can be seen as least as relevant to the party as active party members and the views of focus groups count more than those of conference delegates (Mair, 2013 p. 84). Input from social media similarly work as a cue for the parties (Barberá et al. 2018). This development might bias parties’ issue standpoints as social media do not reflect public opinion nor the views of party members and reactions can often appear as much more negative than they are. As

20 discussed further in article 4, Socially mediated issue ownership, this input can easily backfire if treated as a complement to membership response on party issue strategy.

As social media opens new ways to influence the issue agenda and creates new discursive opportunities, raising issues left out of the mainstream media scope becomes a possibility. Given that socio-economic issues determine party preferences to a lesser extent today and the overall decline in class voting (Vestin, 2019), it could be expected that new media puts more emphasis on nontraditional left-right issues. However, to what extent social media amplify discussions around newer salient issues such as environment and sustainable development or immigration is not evident. During Swedish party leader debates, political discussions on Twitter did not amplify these newer issues that have become salient to the electorate at large (Sandberg et al., 2019). Other studies on the contrary show a significant influence of particular right-wing discourse on Twitter, to a large extent driven by political bots during the Swedish 2018 election (Fernquist, Kaati & Schroeder, 2018).

To summarize, the disrupted relations among media, publics, and democratic institutions constitutes a new age of political communication (Bennett & Pfetsch, 2018). According to the discussion above, the “crisis” of representative democracy can be argued to consist of several trends that together represent a challenge for political parties’ ability to represent citizens (Bjereld, et al., 2018 p.16). Used as an umbrella term for a larger variety of factors and challenges that face representative democracy today, it provides a unifying mantle for the potential impact of social media on representative democracy.

To what extent new communication technology and digital forms of engagement contribute to weakening traditional organizations and organizational forms is strongly linked to challenges facing political representation in contemporary liberal democracies. If social media allow fringe causes a louder voice or streamline organizational hierarchies in European political systems, this points to an equalizing potential that would reinforce trends pertaining to a hollowing of western democracy (Mair, 2013). If marginal or populist parties strengthen their position in the electoral system, this has radically different consequences for democracy than potential consequences related to the internal party organization (such as personalization and professionalization trends). However, both outcomes can be related to party system change.

2.2 Theoretical framework

In Figure 1 below, the theoretical framework of this dissertation is summarized. To begin, I argue the importance of political context, as it determines who can be regarded as previously disadvantaged. The equalization perspective emphasizes a disruption of the status quo, which makes the framework sensitive to the political context. Ward and Gibson argue that both “Systemic and technological opportunity structures provide the broad political and technological parameters within which political organizations operate” (2008, p.35). Primarily two systemic structures are important here. The political environment i.e. the basic political system shapes the use of new communication technology such as federalism, party system, electoral system etc. (Ward & Gibson, 2008). In addition, I consider how the electoral context may also shape how politicians use social media. The media environment at large is also regarded as a systemic structure that affect how parties and political candidates use social media (Ward & Gibson, 2008). As I will discuss later in this chapter, a broadly defined disadvantaged position applies in particular to actors with an oppositional or marginalized position vis-à-vis mainstream media.

Figure 1. Theoretical framework

Next, the argument is made that opportunity structures do not shape the same way for all actors who could be defined as previously disadvantaged, as the equalization perspective suggests. Within the equalization framework, both female candidates and populist actors have been considered as disadvantaged. To understand how social media potentially benefits some actors over others, it is central to include actor centered explanations and the functionality of communication technology. Only then can we understand the interaction between actor and social media opportunity structures (actor x platform architecture). This chapter subsequently explores what more specific factors could shape parties’ and candidates’ incentives and

22 behavior. Drawing on the literature, I argue that party and candidate centered motives for which social media is being used should be understood in relation to opportunities that form differently depending on the actor and on the architecture of different social media platforms. Thus, actor types interact with the platform architecture differently. It is also important to assess the different dimensions of normalization/equalization that are interrelated but not always pointed in the same direction. For example, populist members of parliament might be less prone to adopt social media that relate to party structural traits, but when they do, they distinguish themselves in many important ways, which, in turn, might be linked to opportunities relating to the functional and discursive dimensions of normalization/equalization. These aspects will be further discussed later in this chapter. First, the combined as well as platform specific features of social media will be discussed, to distinguish these platforms from other means of communication.

2.3 The (social) medium is the message?

A symbiotic relationship exists between message distribution and the form of the medium which influence how a message is perceived. Innis argues that each medium embodies a bias in terms of the organization and control of information, which ultimately forms a bias of communication (Innis, 2008). In a similar vein, it has been argued that a medium not only affects society by the content it holds but also by the characteristics of the medium itself— known by the phrase “the medium is the message” (McLuhan, 1994). A long these lines, it is important to recognize that there is a reciprocal relationship between communication technology and politics, as well as constant interaction in shaping certain outcomes. As the media form and content to some extent are interrelated, it is important to ask how specific communication technologies operate and what form of communication they facilitate. Much of the research on the political implications of social media has disregarded the technologies as such, and rarely is any distinction made between the different platforms (Bossetta, 2018). Each social media platform has a unique design, with different technical protocols and usage analytics that make up algorithms for what content becomes visible, ultimately shaping different communication patterns. The unique combining, as well as the different features of social media platforms, should therefore be considered in the light of their potential implications. This section sets out to describe the distinctive features of social media that underpin the formation of opportunity structures for political actors and political discourse.

Before the concept of social media platforms gained prominence, social media, such as Facebook and Twitter, were often referred to as social-networking sites, emphasizing the connections between users in the network (Boyd & Ellison, 2007). Helmond (2015) argues that a shift from network sites to social media platforms occurred when these sites offered Application Programming Interfaces (APIs). Following this definition, a social media platform should provide an interface that allows for programmability. APIs enable social media platforms to expand into the rest of the Web, while external web and app data can be made platform ready (Helmond, 2015).

Like one-to-many web communication platforms such as webpages, social media facilitates communications that are direct and personal (Jacobs & Spierings, 2016). Instead of reaching a general audience through such means as television advertising, social media offers a direct form of communication with no traditional gatekeepers (such as editorial media), making messages appear unmediated. More than the web, social media holds the potential for targeting specific groups based on interest rather than geography (Jacobs & Spierings, 2016). This opportunity to mobilize niche groups more effectively might be more appealing to nonmainstream parties in theory. Similarly, the possibility of establishing a close and direct connection to followers, without external or journalistic interference, might be especially important for actors in a marginalized position, vis-à-vis mainstream media. Parties and actors with less chance of being covered by mainstream media (or those being covered in unfavorable terms) might therefore benefit more from this opportunity.

Social media is also cheap and easy to use. The low threshold is more important for actors who gain little visibility in traditional media or have few resources (Jacobs & Spierings, 2016). A quote from the Italian populist far-right leader, Matteo Salvini, made on Facebook illustrates this: “Who will win? Salvini will” “The others have newspapers, television, banks and corporate cash . . . we have you, we have the network . . . so long as it remains free.”6 For political actors, it is not enough to only have an account though; it takes time and effort to maintain. The potential gains of using social media might diminish in light of the difficulties of using them properly (Gibson & McAllister, 2015). Additionally, political actors identify risks of bad reputation and negative media attention as disadvantages of using social media (Kalsnes, 2016). These risks can also vary depending on the actor; some can be punished harder by the

6 https://www.theguardian.com/world/2018/dec/17/populist-social-media-playbook-who-is-best-facebook- twitter-instagram-matteo-salvini-narendra-modi 24 electorate for something they expressed on social media in the heat of the moment. For certain politicians, other negative aspects of being active on social media may emerge, such as online misogyny and harassment directed at female politicians (Krook & Sanín, 2019).

The interactive features that characterize social media platforms facilitate a reciprocal relationship. Within networks, interaction may also occur around a politician—for example, on her or his Facebook page—which can make political supporters into active agents in communication and community building. The interactive feature of social media makes it possible for politicians to monitor the response on social media, a cheap and efficient way to analyze what political messages are engaging. Therefore, the interaction with and around posts can function as popularity cues (Porten-Cheé et al., 2018), signaling to a politician or a party how the message is perceived in real time. On most social media platforms, this would also increase the visibility of a post, causing it to travel beyond the closest group of followers. Interactivity has often been measured as politicians’ responsiveness to communicating directly with users (see e.g., Jungherr, 2016). An interaction effect is arguably present in a reciprocal relationship on social media, in which all kinds of feedback can be considered interactive, regardless of the direct response rate.

Unique characters of social media are also that political messages diffuse across networks and societal strata, through mechanisms such as likes, shares, and retweets (Jacobs & Spierings, 2016). This means that content can spread fast and increase in volume by the network of users, which may compel political actors to communicate messages that they think hold high potential for evoking attention (Klinger, 2013). In turn, content that becomes prominent on social media is often picked up by traditional media and reaches a broader segment of the population. With no, or a limited, gatekeeper function, it is more likely that a certain type of content disseminates on social media—for example, issues or angles left uncovered by editorial media. Diffusion of messages through networks can make some issues seem bigger than they are, which become problematic if social media is taken as an expression of popular opinions. Another negative aspect of message diffusion on social media is the possibility of manipulating engagement, which can be used to amplify illiberal voices. By making proponents online seem more numerous than they are, the mainstream media coverage can be influenced (Tucker, Theocharis, Roberts & Barberá, 2017).

Direct, personal communication as well as interactive features, low threshold, speed, and volume characterizes social media platforms (Jacobs & Spierings, 2016). In transforming means of communication, social media opens new ways to communicate, interact, or take part in information. This is not to say that political communication as a whole might change fundamentally. However, to understand social media’s potential to reshape power balances in politics and ways of communicating politics, it is important to discuss how different social media characteristics interact with how political actors use them. A comparison can be made with the process of mediatization, defined as a dynamic process in which media and its format, terms, and content influence the political sphere. The more mediatized politics is, the bigger the impact media has on it (Strömbäck, 2008, p. 387). Similarly, the more importance journalists and politicians give social media, the bigger the impact will be on the political processes.

In a hybrid media system where old and new media logics intertwine and complement each other, actors who understand these dynamics and have the capacity to influence information flows between different forms of media increasingly define politics (Chadwick, 2013). Social media platforms have been argued to operate with a distinctly different logic from that of traditional mass media while overlapping with it (Klinger & Svensson, 2015). The literature has defined this logic differently, as a social media logic (van Dijck & Poell, 2013) or a networked media logic (Klinger & Svensson, 2015, 2016). The latter has been defined as the rules/format of communication on social media platforms, leading to different ways of producing content, distributing information, and using media (Klinger and Svensson, 2015). A more recently introduced concept, similar to that of a logic, is the digital architecture of social media (Bossetta, 2018). Social media’s digital architectures include network structure (how connections interact, including privacy settings and search functions), functionality (how content is mediated, accessed, and distributed across platforms), algorithmic filtering (the selection, sequence, and visibility of posts) and datafication (data generated from usage). Separately and together, these aspects influence the network formation, the strength of ties among users, and the type of content likely to be generated on a platform (Bossetta, 2018; Bossetta, Dutceac Segesten, & Trenz, 2017). This concept is argued to better conceptualize political information and communication flows across several platforms, as the digital architecture on one platform shapes how politics is communicated (Bossetta, 2018). Thus, following this conceptualization, it is important and useful to also distinguish between platforms.

This dissertation views the unique characteristics of social media that shape interactions and certain communication practices as forming the basis of possible opportunity structures, the “factors inherent to the online media system” (Engesser, Fawzi, & Larsson, 2017, p. 1280). Engesser, Fawzi, and Larsson compare this concept to the political opportunity structures that have been used in social-movement literature to refer to characteristics of a political system that positively affect a given phenomenon (Kitschelt, 1986; Kriesi et al., 1995). Opportunity structures may also refer to the dimensions of political competition. For example challenger parties are at a disadvantage on the traditional left-right dimension, therefore, in addition to ideological commitments, there are strategic incentives for these parties to mobilize politically on other issues such as immigration or the EU (De Vries & Hobolt, 2020). In social media studies, the term “affordances” has been used as a similar concept, referring to the properties of communication a platform enables (Boyd & Ellison, 2007). However, this concept lacks an agreed-upon definition and is rarely used in studies examining politicians’ social media use (Bossetta, 2018). As the term “opportunity structure” has become more established in work focusing on political actors and political mobilization, it is the term used here.

Compared with the concept of digital architecture that underscores how the properties of the technology directly influence a certain use (Bossetta, 2018), opportunity structures can instead be argued to approach certain outcomes the technology, such as social media, enables or constrains. Furthermore, the opportunity structures arguably form distinct possibilities or constraints for different political actors. Some opportunities will also be more closely linked to the specific platform architectures, such as reaching journalists on Twitter and a broader voter segment on Facebook (an argument that is elaborated on in article 2). Thus, how one actor communicates on one platform should not translate to another (due to specific architectures), and one type of party making successful use of a social media platform does not mean that it is a one-size-fits-all model (i.e., opportunity structures might form differently).

Here, opportunity structures are thought of as a way to understand how social media potentially benefits one type of actor over another. However, politicians and parties are important agents in this framework, the reason that these actors and their potential motives must be analyzed in relation to social media opportunity structures. Actor-specific opportunities that might arise will therefore be elaborated on further in this chapter, linked to the normalization perspective, with a specific focus on the disadvantaged position. Before this discussion, the potentially different opportunities linked to specific platforms will be addressed. For example, on Twitter, the

27 traditional media and journalist are easier for politicians to target, whereas a broader segment of the population is on Facebook. These aspects are discussed further below.

Different social media architecture: Platform politics?

The number of available platforms and channels offer new campaign possibilities for parties and candidates, influencing the way campaigns are conducted (Kreiss, 2012; Stromer-Galley, 2014). Studies have shown that campaigns do employ different strategies, depending on the platform (Kang et al., 2018; Stier, Bleier, Lietz, & Strohmaier, 2018). For example, Facebook is mostly used for one-way communication by politicians, whereas Twitter is used in a more interactive way (Jacobs & Spierings, 2016). Twitter messages are limited to a small number of characters, suggesting that its architecture aims at evoking discussions through interactions, whereas Facebook is more of a personal platform. Predictably on Facebook, the platform rewards emotional appeals, as the Facebook architecture encourages personal and emotional expressions in the form of emoji’s.

Since Facebook and Twitter are the main social media platforms on which the empirical studies focus on, some further defining features of the architecture are worth highlighting, in terms of both similarities and differences between the two platforms. All social media platforms support interactions among users. On Facebook, these are friends and on Twitter, followers. Twitter has a more unidirectional and open network structure than Facebook, where befriending is the main connection between users. It requires a confirmed relationship between both parties, whereas Twitter encourages a broader following and using hashtags to share content to a wider audience. Furthermore, users share content on Facebook, whereas, on Twitter, content from another account is retweeted. Both platforms allow users to establish connections with parties and politicians. Additionally, on Facebook, users can maintain public pages that have a different interface and set of tools, compared to the average account. For instance, they allow politicians to analyze and monitor reactions. Regarding functionality and distribution of content across platforms, Facebook maintains a centralized broadcast feed, the so-called “News feed,” in which the algorithmic filtering decides what content becomes visible to users. On Facebook the feed is more heavily filtered than on Twitter that has a chronological filter (Bossetta, 2018). Both platforms allow parties and politicians to pay for content promotion by the platform.

Besides differences in architecture, a fundamental difference between Facebook and Twitter is the number of users. To date, a much smaller percentage of citizens use Twitter compared to

28 the massive and broad user base of Facebook (Bossetta, 2018). Facebook is less an elite medium (Larsson & Moe, 2014) and would be the platform on which politicians interact to try to mobilize ordinary citizens. How politicians perceive different social media reflects these differences. On Twitter, the primary audience is perceived to be media and journalists, whereas Facebook is used to reach a broader and bigger segment of the population (Kreiss et al., 2018). Research also shows that journalists are particularly active in following and sourcing politicians’ statements from Twitter (Kreiss, 2016). This may give lesser-known candidates or populist politicians’ incentives to try to increase media coverage or influence the journalistic framing of news. For mainstream parties, this opportunity might not reach its full potential, but rather be regarded as just another channel on which to broadcast political statements.

Citizens also differ in the way they use social media, for example, as a source of political information. Citizens with stronger populist attitudes are more likely to use Facebook as their source of political information, while nonpopulist citizens use Twitter for information purposes (Schulz, 2019). To summarize, the architecture and audiences on different social media platforms form different patterns of political communication and a different set of opportunities depending on type of political actor.

2.4 Politics as usual?

According to the normalization perspective, new technological advances, such as social media, will be normalized into existing power structures. Hence, differences in political influence and resources persist, despite new technical possibilities. In its most basic meaning, normalization theory states that actors who already have the most influence will be the ones most likely to benefit from technological advances (Southern, 2015). The general normalization framework of whether and how new web technologies influence the power balance between parties’ date back to when the Internet had started to expand, over 20 years ago. Nonetheless, it is still the most comprehensive framework for understanding the overall transformative power of new technologies on democracy. In the first formulation of the normalization theory as applied to the Internet, the power relations between major and minor political actors were predicted to remain unaltered (Resnick, 1998). Prior to the rise of social media, research in large part confirmed a relational normalization of Internet use and supported functional as well as discursive normalization on the Internet. Persistent gaps between major and minor political actors, top-down and controlled information flows, prevailing practices of negative campaigning, personalization or strategic news, in all of the above-mentioned dimensions, have pointed to a normalized pattern of party competition and political communication on Internet (Schweitzer, 2011).

Social media innovations, on the contrary, have been argued to refute trends of normalization (Southern, 2015). In comparison, these platforms are decentralized, nonhierarchical, and built on multilateral interactivity. The networked information flows and user-produced content, which allow for vertical interactions and a horizontal change in technology adoption, have been argued to empower new or previously disadvantaged parties without the same history of the traditional organization (Kalnes, 2009). Prior to the introduction of social media, political actors still needed access to resources and existing organizations, such as the news media, or were dependent to a greater extent on the party organization. It has been suggested that above all, social media closes “the technological gap between insiders and outsiders,” as the political elite has lost its part of technological advantage (Mounk, 2018, p. 146). The literature often refers to this perspective as equalization, implying that new technologies, such as social media, will level the playing field in redistributing the power balance in favor of previously disadvantaged parties and candidates (Gibson & McAllister, 2015).

The extent to which smaller parties can overcome the disadvantages they face in the offline- media environment is the most widely examined indicator of equalization (Jacobs & Spierings, 2016). However, the disadvantaged position is often not elaborated further than this (e.g., Gibson & McAllister, 2015). Gibson & Ward (2000), argues that both organizational capacities and incentives must be considered in understanding adaptation to new media, introducing nonmaterial factors such as ideology, target audience, and the role of “outsider” or “opposition”. Although the theoretical links between social media and a disadvantaged or outsider position have been addressed, the “outsider” position also lacks a precise definition in relation to the normalization/equalization perspective (e.g., Gibson & Ward, 2000; Strandberg, 2009). Later research also emphasizes the importance of the role of political actors as outsiders (Gerbaudo, 2019; Jungherr, Schroeder, & Stier, 2019; Schroeder, 2018). However, these studies have not been situated within the normalization perspective and therefore, they do not specifically compare different types of parties.

Jungherr, Schroeder, & Stier (2019) argues that digital media has led to a shift in the opportunity structures in politics, by allowing outsiders to “route around established institutions and to become entrenched and powerful political actors without their support” (p. 1). The argument is similar to that of the classical equalization perspective but it emphasizes more strongly the positional role of outsiders as challengers vis-à-vis the political status quo. Most importantly, digital media have weakened the ability of established political institutions and the media “to structure political discourse and thereby limit the degree to which outsiders were able to challenge the political status quo” (Jungherr et al., 2019, p. 2). Therefore, challengers or outsiders can sustain an adversarial stance toward legacy media and democratic institutions, even as these actors become established.

Jungherr et al. (2019) define outsiders as “actors or groups outside the established party systems. Often, they motivate their bid for political power by openly contesting or rejecting the legitimacy of established political institutions, actors, or media gatekeepers” (p. 3). In their understanding, an outsider candidate refers to both Barack Obama and Donald Trump, as they were outsiders to their respective party establishment and challenged their party in pursuit of the presidential nomination (Jungherr et al., 2019). Barr (2009) makes an analytic distinction between political outsiders, anti-establishment politics, and populism. According to his argument, the notion of outsiders connotes a political actor’s relationship to the system of competitive parties and is not necessarily associated with anti-establishment politics or populism. Nevertheless, he points out that outsiders and mavericks (i.e., insiders who do not 31 act in a way that maintains the status quo) can credibly offer anti-establishment appeals and present themselves as agents of change (Barr, 2009). The positional role as an outsider and the opportunity to sustain an adversarial stance toward legacy media is arguably more important for populist actors using anti-establishment rhetoric than for any political candidate acting as an outsider in relation to the party establishment. Therefore, I refer to the term “previously disadvantaged” when discussing the opportunity structures of social media. However, the structures that enable (previously) disadvantaged parties or candidates might yield slightly different opportunities. In this dissertation, I argue that the opportunities that arise for so-called outsiders, such as candidates of populist parties, differ from those of traditionally marginalized groups, such as women or ethnic minorities, while the opportunity structure takes the same form in terms of opportunities to both reach visibility and sidestep traditional media. Whereas populist parties can maintain their adversarial stance and oppositional role while still becoming established in the political system, the position of marginalized groups is more closely linked to gaining visibility and reaching potential voters, outside the traditional media that either neglect or misrepresent these groups. Thus, at times, social media might seem to benefit different political camps. Yet, I argue that in spite of these apparently different consequences, the different types of previously disadvantaged actors trigger a shared underlying mechanism leading to potential strengthening of their position (i.e., equalization).

The following section is structured as follows. First, the relational dimension of normalization/equalization is discussed in terms of inter- and intraparty aspects, focusing on populist parties and female candidates. Second, previous research on how these actors use social media is addressed, relating to a functional dimension, followed by a discussion of the discursive dimension addressed in the empirical studies.

Relational dimension

In general terms, before the emergence of social media, the Internet did not fulfill expectations for equalizing the potential to improve conditions of democratic competition (Margolis & Resnick, 2000). If anything, larger parties benefited from establishing a presence online (Gibson, Lusoli, & Ward, 2008; Margolis, Resnick, & Wolfe, 1999; Margolis, Resnick, & Levy, 2003; Schweitzer, 2011). When social media came to complement webpages and other digital tools, earlier studies did indicate a shift in the power balance, in favor of minor-party candidates (e.g., Gibson & McAllister, 2011, 2015; Gibson, Römmele, & Williamson, 2014; Lassen & 32

Brown, 2011; Southern, 2015). It allowed smaller parties with less resources to reach a wider audience, bypassing traditional media and thereby overcoming the disadvantages they face in the offline media environment (Gibson & McAllister, 2015; Vergeer & Hermans, 2013). As social media became more established and the ability to reach a larger share of potential voters increased, the initial advantage for smaller parties seemed to have vanished (Evans, Cordova, & Sipole, 2014). It has been argued that recognizing different development phases is important, as the logic behind parties and political candidates’ adoption and use of digital technologies changes (Gibson & McAllister, 2014; Jacobs & Spierings, 2016). Following this understanding, smaller parties’ incentives to experiment with innovative tools created an initial advantage, whereas the traditional parties initially had a mindset less compatible with the values of social media (Jackson & Lilleker, 2009). When larger parties saw the added value of a social media presence, it led to normalizing interparty relations again. Larger parties have more resources in terms of funding and personnel, which gradually gave them a greater advantage. Traditional small parties that were not early adopters had a harder time catching up and have remained behind, thus normalizing their position as a smaller party. On the other hand, early adopters have remained relatively strong, supporting evidence for social media equalizing party competition. However, the remaining variances might best be described as “change within a continuity” (Jacobs & Spierings, 2016, p. 5). Different development phases have shed light on the relationship development between minor versus major parties, and although party size has been the most widely explored factor, several other aspects are relevant to consider.

In terms of party characteristics, organizational traits, resources, internal structure, and ideological outlook have been argued to affect the likelihood of parties’ openness to implementing new techniques (Gibson & Römmele, 2001). Prior to the introduction of social media, ideological positions had a great impact on the uptake and usage of political blogs. Party affiliation and ideological positions toward individualism and collectivism shaped blogging, making it “a strongly ideologically situated practice of political communication” (Åström & Karlsson, 2013, p. 434). Similarly, ideology is important in explaining how different parties utilize social media and the different ways they do so. In particular, postmaterialist parties, such as green parties, have been argued as distinguishing themselves in this regard (Gibson & McAllister, 2011; Jacobs and Spierings, 2015; Southern & Lee, 2018). Few studies analyzing politicians’ use of social media have included populism as an independent variable (Engesser et al., 2017a). Despite this lack of empirical studies, research suggests that social media is particularly well suited to populist actors and populist communication (Engesser et al., 2017a;

Gerbaudo, 2018; Krämer, 2017), addressed below in terms of party characteristics linked to online opportunity structures. Given the current body of knowledge on the topic, it is not always possible to distinguish the unique opportunities of social media compared to other online media nor to distinguish between the unique opportunities relating to each platform’s architecture. In terms of actor-specific opportunities, in practice, they often overlap.

Populist parties and social media

Regarding party structure, new-technology adoption has been suggested as more advantageous for parties that have a top-down internal structure and a more hierarchical culture, typical of populist radical-right parties (Gibson & Römmele, 2001). Although the populist radical-right tends to gain a large and active follower base, these politicians have been found less likely to adopt social media (Jacobs & Spierings, 2018). In Article 3, we find that those MPs of populist parties that do adopt social media, are more active than MPs of other parties. In particular, the party leaders make active use of social media, which (we argue) relates to the centralized organization of these parties, discussed later on. Furthermore, the few existing studies on populists’ use of social media focus on Twitter, even though Facebook can be regarded as more important for populist communication (Engesser et al., 2017a; Jacobs & Spierings, 2018). The important need to distinguish between different social media platforms is addressed below in terms of a functional dimension.

In most cases, populist parties are regarded as outsiders with more limited funding than other parties and restricted in their offline opportunities (Jacobs & Spierings, 2018; Strandberg, 2008). In addition, they often have a problematic relationship with traditional media and are covered in an unfavorably way (for them) (Esser, Stępińska & Hopmann, 2016). However, it is only recently that studies have started to address the relationship between populist political actors and social media.

Recent work has identified several opportunities that might make social media particularly well- suited channels of populist communication (Ernst et al., 2019b; Engesser, Fawzi & Larsson 2017). Ideologically, populists oppose themselves to traditional elites (anti-elitism) and align themselves directly with the people (people-centrism), claiming to represent the people’s will and restoring popular sovereignty (Jacobs & Spierings, 2018; Rooduijn, 2014). Advantages of social media for populist actors have been argued to involve circumventing mainstream media

34 to disseminate their messages unfiltered by direct access to “the people” they claim to represent, and anti-elitism in terms of social media’s low threshold (Krämer, 2017; Ernst et al., 2017). Social media allows them to bypass the traditional media, which they see as either ignoring or distorting their message. Populists can use social media to criticize the political elite (Van Kessel & Castelein, 2016) and endorse values whose expression would be restricted in editorial mainstream media or changed by “media elites” (Klinger & Svensson, 2015). Messages can be disseminated on social media without journalists altering the message or adding their own negative evaluations. Furthermore, the speed and volume of social media can push the media to report quickly about populists’ posts, letting them remain in a raw and unaltered form (Jacobs & Spierings, 2018), ultimately providing populists with yet another channel to disseminate their message unfiltered.

The recent growth of political actors challenging the legitimacy of mass media and the political establishment has often somewhat paradoxically developed them into important actors in the political systems they attack. This could be explained by the opportunity to route around institutions that structure political discourse, such as traditional media, and maintain positions that previously would have been filtered out. Jungherr, Schroeder & Stier, (2019) argue that “digital media change the political opportunity structure for outsiders allowing them to maintain strong antisystem challenges while still becoming influential in political systems” (p. 2). Therefore, not only is the opportunity to bypass traditional media important for these actors; social media also allows populist parties to maintain their oppositional role by criticizing the mainstream media as distorted and unfair. Social media simultaneously functions as alternative media to reach citizens that share a distrust of traditional media. The mass-networking capabilities of social media have been argued to provide means for disaffected individuals to express themselves and form partisan communities. This makes social media a suitable channel for “mass politics” and appeals to “the people,” typical of populism (Gerbaudo, 2018). In these communities, belonging and recognition among otherwise scattered groups creates a feeling of protected spaces and perceived anonymity, which allows for communicating hostility, using an uncivil tone, and cultivating an in-group identity that is particularly important for populist online mobilization (Hameleers, 2018b). Because social media can be used to create protected spaces and the feeling of community, one-sided and anecdotal evidence can be accumulated in large quantities and made accessible to followers (Ernst et al., 2019b). Allowing members of these communities to interact with other likeminded members of this in-group and respond to politicians’ populist messages, social media arguably empowers the “silenced majority”

(Bracciale & Martella, 2017; Hameleers, 2018a). Populist and marginal actors can take greater advantage of this and profit from the feeling of community and selective exposure to one-sided information, to mobilize their supporters and coordinate political actions (Krämer, 2017; Ernst et al., 2019b).

To summarize, the ideological core elements of populism—people-centrism, anti-elitism, and the claim to represent the will of the people—have all been argued to interact with the opportunity structures of social media in a favorably way. In particular, the direct connections to followers and avoiding the gatekeeping of traditional media give populists more freedom to spread their messages (Engesser et al., 2017a). They may also maintain their oppositional role, even as they become a part of the establishment. Similarly, the in-group identity formation, in contrast to exclusionist constructions of “out-groups” as credible scapegoats (Aalberg et al., 2017; Hameleers & de Vreese, 2018), interact positively with the mass-networking capabilities of social media.

Populist messages are characterized by assigning blame to elites in an emotionalized way (Hameleers, Bos, & de Vreese, 2017). Similarly, populists’ anti-elitism has been linked to a more straightforward communication style (Bos & Brants, 2014). In terms of a populist communication style, simplification, emotionalization, and negativity (which can be applied to a variety of political actors, not only populist parties) have been argued to fit the attention economy on social media (Engesser, Fawzi & Larsson, 2017). Not surprisingly, it is politicians belonging to a populist party who are most likely to communicate in a populist way on social media (Ernst et al., 2019a). On social media, the competition over visibility is particularly intense due to the abundance of information; therefore, this type of content may evoke more reactions and is another feature of social media opportunity structure that works favorably for populist actors.

Intraparty relations, shifting power balances within parties To a greater extent, the equalization-normalization perspective has focused on interparty relations and how minor, major, or fringe parties benefit from technological innovations in communication. The strong focus on interparty relations might have led to underestimating the impact of social media as the transformative potential of social media extends beyond interparty relationships (Jacobs & Spierings, 2016). Social media more than before, opens for a targeted and personalized form of communication which also has implications for intraparty

36 competition, as individual politicians can shape their own messages vis a vis the party and strengthen their position, at the expense of the parties’ loss of control (Hansen & Kosiara- Pedersen, 2014; Spierings & Jacobs, 2014).

Social media therefore presents an opportunity to level the playing field not only for previously disadvantaged parties but also for previously disadvantaged individual politicians. The individual characteristics of candidates are of importance, as many advantages of social media reside at the individual level (Margolis & Resnick, 2000; Vergeer, Hermans, & Sams, 2013). In general, it is the most well-known politicians who receive substantial amounts of traditional media coverage that also have a larger followership on social media (Keller & Kleinen-von Königslöw, 2018). On the other hand, social media use may trigger intraparty competition and intraparty equalization for the candidates that manage to utilize social media platforms efficiently, a scenario also supported in previous work (Hansen & Kosiara-Pedersen, 2014; Kruikemeier, 2014; Spierings & Jacobs, 2014). Low-ranked or unknown politicians have been argued to benefit from social media use since visibility in mainstream media is limited (Spierings & Jacobs, 2014). Less has been theorized about the opportunities for traditionally underrepresented groups or politically marginalized groups. Without gatekeepers from the older traditional arenas, social media provides opportunities for marginalized political groups to tailor their representative strategies. The nonmainstream profiling, a possibility for building large, personalized, and unmediated networks, and tailor messages to specific target groups unbound by geographic distances, has been argued to benefit politically marginalized and underrepresented groups, such as women and ethnic minorities (Jacobs & Spierings, 2016).

Some studies show gender differences in social media adoption and use, indicating that being a woman candidate is of importance (Evans, Cordova & Sipole, 2014; Jackson, & Lilleker, 2011), while others do not (Karlsen & Enjolras, 2016). In Article 1, we show that women candidates running in the European election place a higher value on using social media as a campaign tool. One of the reasons for different studies coming to different conclusions may be that the electoral context shapes the use of social media. This seems likely, as the opportunities social media offers are closely linked to specific types of political arena. For unknown candidates, in elections such as the European election, the visibility in mainstream media is very limited, changing the dynamics of social media use, compared to national elections.

Social media has different implications for intraparty competition within populist parties compared to a potential strengthening of female representation within parties. Individual politicians with access to their own campaign tools, as well as to direct communication channels with journalists on social media, might undermine the centralization of the party and make individual candidates more independent of the party leadership. This poses a threat to centralized parties, such as populist radical-right parties (Mudde, 2007; Van Kessel, 2015), as it can empower backbenchers. A populist leader is the central actor and embodies the people, and the party headquarters tightly controls the party (Jacobs & Spierings, 2018). Furthermore, the personalized communication on social media might be particularly favorable for populist party leaders. Right-wing populism is highly personalized, in the sense that it is centered on the person of the leader (Krämer, 2017), and successful populist parties rely on charismatic leadership. Previous research confirms that social media posts by populists indeed are focused on the populist leader (Grill, 2016). Similarly, the authentic, unmanaged use and direct communication (Rooduijn, 2014) are features that might be most beneficial for populist leaders. As we argue in Article 3, this might constrain the activity of MPs of populist parties on social media and Facebook might in this respect be a particularly well-suited platform for populist leaders.

Functional dimension

While the relational aspects of normalization often focus on parties’ and candidates’ uptake and activity on online platforms, the functional dimension emphasizes forms of use. More specifically, this dimension of normalization deals with the extent to which political actors realize the media-specific opportunities in online communication (Schweitzer, 2011). Previous research for example looks at the extent to which campaign activities on social media reinforce professionalization trends. According to Schweitzer (2011), incentives for political actors to adopt a more professional use of social media can be driven by efforts to increase their competitiveness, generate attention (also in the traditional mass media), be perceived as modern, and meet users’ rising expectations of a sophisticated presence online. On the other hand, the media-specific opportunities might be neglected for lack of resources and time, or due to more strategic considerations, such as loss of control over political messages and image- building (Schweitzer, 2011). Important here is that the extent to which parties’ and candidates seek to utilize these functionalities or not will differ depending on party type and candidate characteristics.

Previous research finds that political candidates from less established parties use social media differently. They are more interactive and share more personal information, whereas established and bigger parties employ communication strategies more similar to webpages (traditional top- down). In particular, green party candidates adopt more interactive use of social media, thus subverting normalization to a significant degree (Southern & Lee, 2019). Populist politicians on the other hand, are less likely to interact with their followers (Jacobs & Spierings, 2018; Tromble, 2016). Their overall communication strategy on social media follows a top-down approach, by not involving users and by a lack of responsiveness (Krämer, 2017). Populists also follow few accounts, but the ones they do follow are more likely to get retweeted on Twitter, indicating that they are primarily other party affiliates (Jacobs & Spierings, 2018). The functional aspects of populists’ use of social media are examined in two of the empirical studies, both testing previously overlooked aspects. Article 2 deals with the question of what populists can achieve by using different social media platforms. The results show that populists use Twitter to criticize the media to a greater extent than other parties, whereas Facebook might be used to activate anger. This indicates that populists do utilize the media-specific opportunities of social media platforms, and, therefore, it is important to look beyond interactivity as a measure of functionality or opportunity. Previous research does not adequately consider the role of reciprocal interactivity, in terms of popularity cues and feedback loops on social media. On Facebook, in particular, the emotion-eliciting appeals of populist messages should be addressed as an important advantage for these actors. Article 3 further examines this aspect.

Research findings on women candidates’ use of social media are more inconsistent, but it appears that they have motives to act differently on social media platforms (Evans, Cordova, & Sipole, 2014; McGregor, Lawrence, & Cardona, 2017), at least in the U.S. context. However, little is known about how the actual motives and rationales behind use differ between candidates. In Article 1, we therefore examine whether female candidates state other motives for using social media and compare this to their actual use, as well as their evaluations of using social media for different purposes. In functional terms, we specifically examine whether social media is used for personal or party-centered purposes.

Discursive dimension

The discursive dimension of the normalization/equalization perspective address whether typical communication practices offline are transferred to online platforms. Research on this dimension has focused on explaining parties’ discursive behavior on social media and whether party representatives use different issue strategies in different channels (e.g., Benoit et al. 2011; Druckman et al. 2009). On social media, the candidate’s campaign agenda is not simply “politics as usual,” as this agenda does not respond to the media agenda or the public agenda (Van Dalen, Fazekas, Klemmensen, & Hansen, 2015). The discursive dimension of normalization has in these previous studies been used to explain the behavior of political actors. In the dissertation, this dimension is addressed by analyzing the structure of political communication on social media platforms. To what extent aggregated text data on Twitter indicate a potential break in established patterns of issue competition and issue ownership formation, is studied as one aspect of the discursive dimension. Issue competition reflects a vital aspect of political communication and is an essential component of representative democracy; still, this competition is largely unexplored in the communication practices on social media. Therefore, Article 4 examines the extent to which parties on Twitter are mentioned together with the same issues on which the electorate perceives them to have the best policies.

The discursive dimension of normalization is also addressed in Article 5, examining the influence of new agenda-setters in the social media environment and whether social media increases the visibility of controversial political views. Social media has been argued to provide highly effective, quick, and cheap means of information diffusion to put challengers’ agenda before wider audiences (Benkler, Faris, & Roberts, 2018). Governments and media companies previously had a stronger function as gatekeepers and could set the standard for acceptable political discourse—for example, by not publishing racist content or deceitful information (Mounk, 2018). From an equalization perspective, it could be argued that social media platforms allow for controversial and populist messages to thrive, given the lack of traditional gatekeepers. This idea is not a new one; when the Internet became increasingly used as a political communication medium, it was thought to redistribute power away from established players, and out toward previously unheard and marginalized voices in society (Castells, 1996; Dahlberg, 2001). In most cases, this was understood to strengthen democratic discourse. Concerns about the Internet not achieving its full potential were often directed toward powerful actors, such as the state. For example, Rheingold (1994) asked if censorship and

40 commercialization of the Internet would lead to the loss of the greatest resource ever for community-building and free expression of ideas. However, some scholars of that period suggested that instead, the Internet would redistribute power in a populist direction (Bimber, 1998). Later research saw the Internet as offering considerable benefits for far-right actors due to the anonymity it provides to users, in both accessing and exchanging information. (Bratten, 2005; Caiani & Parenti, 2009). Some of the arguments as to why marginal or extreme actors might be more successful in utilizing possibilities brought about by the Internet are similar to those regarding social media. However, the interplay of online mass media and social media makes it increasingly difficult to analytically distinguish between discursive opportunities in the online environment. In particular news sharing practices capture this interplay, so-called “false news” seems to spread faster and wider through social media than other news (Zhuravskaya, Petrova, & Enikolopov, 2019). It is therefore important to address how social media and in particular how specific platforms might reinforce the position of extreme or marginal actors.

The interest-bound and like-minded peer networks on social media are one important opportunity structure and function of discursive normalization. The interactivity and anonymity (perceived or not) have been argued to benefit marginal actors and views in particular (Krämer, 2017). Messages are not mediated, edited, interpreted, nor filtered by professional media actors, which allows for voicing controversial opinions. Subsequently, a discursive opportunity structure forms on social media platforms. The growing pluralization of the online-media environment and the impact of social media content sharing have been suggested as imperative for understanding how fringe political actors avoid both overt and implicit barriers to popularizing extreme positions (Littler & Feldman, 2017). Therefore, social media might function as propitious spaces for populist and counterpublic communication, understood as discursive opportunities (Koopmans & Olzak, 2004), brought about by social media and the high-choice media environment. To what extent social media deny nuance in favor of extreme opinions, is conditional on their digital architecture. Particularly on Facebook, certain type of messages drive visibility through engagement that the algorithms reward and content that evoke engagement is arguably often more emotionally loaded or polarizing. The Twitter algorithm prioritizes recent content and messages are displayed in reverse chronological order. Bossetta, Dutceac Segesten & Trenz (2017) suggest that the ingrained architectural features of a specific platform have direct implications for the types of communication and political information that

41 flow across it. Therefore, it can be expected that these discursive opportunities vary by platform, to some extent, discussed further in the final chapter.

To conclude this chapter, the importance of context is addressed as most of the implications discussed are conditional, dependent on political context or more present during particular electoral contexts. After that, a short summary of this theoretical chapter will follow.

Context dependency- Equalized for whom and when?

Regarding technological advances as empowering insiders or outsiders in more general terms, social media can function as a tool for giving voice to those traditional media previously excluded from political discussion. Tucker, Theocharis, Roberts, & Barberá (2017) argue that the same tools that in authoritarian regimes can help a prodemocratic opposition to connect, coordinate and organize, also can be used to oppress, censor, and manipulate information, to try to silence other voices. Employing bots and trolls to change the online conversation is particular to the digital age (Tucker, Theocharis, Roberts & Barberá, 2017). In democratic societies, other actors are excluded from mainstream media. As discussed previously, these actors can be less established parties and candidates or marginalized groups. In democracies, these can also include illiberal groups opposed to liberal democracy. These groups can take advantage of the same features of social media that prodemocracy forces in autocratic regimes do, using the ability to find like-minded people, irrespective of geographical proximity, and use social media as a means of organizing (Tucker, Theocharis, Roberts, & Barberá, 2017). Thus, the effect of social media is highly conditional on the political context. In Europe, the Internet and social media are argued as contributing to the electoral success of populism, whereas in immature democracies and semi-autocratic regimes, they reduce support for the ruling parties (Zhuravskaya, Petrova, & Enikolopov, 2019). Above all, social media seems to interfere with the status quo and since the political context determines who challenges the ruling elite or is marginalized, the effect of technology can have both negative and positive democratizing potential.

From a theoretical and empirical point of view, the context also matters for who can be regarded as disadvantaged within nation-states. In national contexts, the different political arenas matter. Theories of social media use and adoption have mostly been applied to first-order, national elections (Gibson & McAllister, 2014), and the importance of electoral context has been

42 overlooked (Jacobs & Spierings, 2016). To understand in what ways social media has impacted politics, it is important to recognize that the political arena and not only the political system condition the potentially transformative power. The media cover general elections extensively, and citizens are better informed of parties’ policies and candidates. The added value of social media is thus relatively limited, and national elections might have a moderating effect (Jacobs & Spierings, 2016). Considering the characteristics of social media, it is more likely that second-order elections, such as supranational ones, generate more added value for social media use. Given that these arenas attract much less attention from traditional media, and that many of the candidates are lesser known, social media use might yield greater impact.

2.5 Summary

This chapter started by placing social media within larger trends in representative democracy and perspectives on contemporary political representation in Western societies. After this discussion, the communication specific to social media, along with how the features of these platforms are distinguished from other web platforms, were addressed. The characteristics of social media that positively affect a given phenomenon, is referred to as opportunity structures of social media. What can be understood as the architectures of social media platforms were also argued to shape different communication opportunities, using Twitter and Facebook as examples. Next, this chapter discussed the overall theoretical framework of normalization and its three-dimensional components, as the dissertation includes articles that each address one or several of these. In relational terms, the central question is whether social media normalizes patterns of party competition or leads to equalization for previously disadvantaged parties or candidates. The functional dimension of the normalization/equalization framework emphasizes forms of use, in terms of media-specific features and the extent to which political actors realize them. For example, if interactive features are utilized or if parties (despite this opportunity) focus on traditional top-down communication. Additionally, the discursive dimension of normalization emphasizes the extent to which typical communication practices offline are transferred to the Internet.

This chapter linked the normalization/equalization perspective to the opportunity structures of social media by focusing on two types of actors regarded as previously disadvantaged. The main argument as to why social media would reinforce the position of these actors is the opportunity to route around traditional media and establish direct contact with potential voters. Prior to social media, it was difficult to reach out without media visibility. Digital tools, such

43 as webpages that are structured in a more hierarchical way with top-down information flows, did not offer enough for actors disadvantaged in or by the media. Interactivity and network effects of social media have more leverage potential, making it more likely that political communication on these platforms yields a larger impact. The social media opportunity structures were argued to form differently for previously disadvantaged actors, as some opportunities are actor specific. In particular, as recent research has argued, opportunities directly linked to populist actors and populist communication overlap to some extent with the identified opportunities for disadvantaged actors following the equalization perspective; in other ways, they are specific to the populist ideology and communication style.

Situated within the discursive dimension of normalization, the chapter also addressed aspects pertaining to social media opportunities for controversial views and previously marginal actors to spread their message and increase their visibility. If so, this would point to a more equalized discursive pattern. Similarly, the extent to which political discussions on social media tend to give priority to some issues and parties over others could indicate a disrupted normalized discursive pattern. For example, if discussions give more weight to highly polarized and emotionally loaded issues such as immigration and the parties associated with these.

3. Research approach and methods

The dissertation consists of five articles, each testing hypotheses derived from the broader normalization/equalization framework. Focusing on what the literature has defined as previously disadvantaged actors (see e.g. Jacobs & Spierings, 2016), the overall research question addresses how the opportunity structures of social media interacts with this alleged position. The main motivation for comparing populist parties with female candidates stems from the assumption that social media provides different opportunities for these actors while their use of social media at the same time might point to a shared underlying mechanism of new technology levelling the playing field. To what extent social media strengthens the position of women in representative bodies or a populist opposition have distinctly different effects on democratic quality. In this respect the two types of actors can be described as each other’s opposites. For instance, if social media contributes to strengthening female or ethnic minority representation, this equalizing effect can be understood as normatively good within the framework of representative democracy inasmuch as it reduces inequality. In contrast, populism and populists could be understood as a challenge to minority rights. It should be noted, however, that populist parties pose different challenges to minority rights depending on whether they are left-wing or right-wing populist parties (Huber & Schimpf, 2017). In this dissertation, populism is examined and understood as a thin ideology (Mudde, 2004), and therefore it is also likely that opportunity structures of social media to some extent forms differently depending on whether populism is combined with a left or right- wing host ideology. In this dissertation, I focus on the core element of populism and as such I do not make any distinction between left- and right-wing populism. Some elements of social media are however likely to be more beneficial for right- wing populists or far-right mobilization as discussed previously.

By testing a disadvantaged position that includes both right-wing populists at one side of the spectra and women representing a historically disadvantaged group at the other, we can address both the shared underlying dynamics and opportunities potentially unique for a particular type of a disadvantaged actor. As I discussed in the theory section, politicians belonging to ethnic minority groups have also been considered disadvantaged in terms of a marginalized position in representative democracy. Following the theoretical argument of this dissertation, the assumption about the effect of social media on their political communication patters would be similar to the potential equalization effects for female politicians. Studying communication of politicians with an ethnic minority background, however, is beyond the scope of this

45 dissertation. It would also be a more complicated procedure to compile this type of background information on candidates as it is not publicly available information. The fact that the actors examined in this dissertation are disadvantaged in a different sense has consequences for our understanding of equalization. It is therefore important to examine how social media opportunity structures can shape both similarly and differently depending on the actor type.

Apart from examining questions pertaining to a relational dimension of normalization/equalization, I also address the functional and discursive dimensions. For the sake of analytical clarity, these three dimensions were discussed separately in the theoretical chapter, but in terms of social media opportunity structures, they are interwoven. For example, the value that female candidates place on campaigning and reaching voters through social media is in theory related to the interactive functions and the discursive opportunities to raise issues left out of mainstream media. Similarly, the prominence of populist parties on social media can be understood in light of all three dimensions in conjunction. In particular, the discursive opportunities of social media platforms and opportunities to openly attack the traditional media is interrelated for these actors. Below, I focus on the core aspect of how each article contributes to our understanding of different dimensions of the overarching theoretical framework of normalization.

In Article 1, the relational (interparty) dimension of normalization is addressed by focusing on the extent to which social media is seen as a more favorable campaign tool among women candidates. It also examines a functional dimension in terms of the candidate’s motives (party centered or personalized strategy) and actual use of Twitter (party centered or personalized strategy). Article 2 compares populist members of parliament to those of other parties and examines the relational dimension of normalization. It also addresses the functional dimension in terms of populist parties’ use of social media, condemning the media on Twitter and evoking anger on Facebook, thus, showing the importance of distinguishing between platforms as they create somewhat different opportunities. Article 3 similarly examines both relational and functional dimensions of populists’ use of social media. In relational terms, my co-authors and I do not only examine the intraparty dynamics, but also the dynamics within the party, as we argue that the party structure of populist parties has consequences for their MP’s use of social media. The article also examines activity levels of populist parties and the emotional component of their messages, as a part of the functional dimension.

Articles 4 and 5 address the discursive dimension of normalization. Article 4 focuses on to what extent voters’ issue-ownership perceptions are mediated on Twitter which relates to the opportunity to put forward an alternative political agenda or not. Article 5 has a similar focus, but instead of examining an aggregated form of political discussion it examines the opportunity for controversial and marginal perspectives to gain visibility on Facebook. Although the two studies address different questions, they both provide distinct insights about the discursive dimension as it is likely that communication varies between the platforms due to the unique digital architecture and user base.

3.1 The political context and generalizability across platforms

Within established democracies, the political systems are likely to influence social media’s role in politics (Ward & Gibson, 2008). Most of the previous work on the political implications of social media addresses the situation in majoritarian political systems, such as the U.S. and Great Britain (Jacobs & Spierings, 2016). Research results on politicians’ social media use in these countries are likely to differ from those obtained in pluralistic multiparty systems. In majoritarian contexts, the gap between large and small parties is much wider, and smaller parties, as well as candidates from these parties, struggle more than in other political systems. Furthermore, the individual politician in majoritarian systems is more central, which is likely to influence communication through social media (Vergeer, Hermans, & Sams, 2013). The degree of personalization within the political system can also be expected to shape how social media is used. The same holds for other aspects, such as the media system and campaign- finance regulations.

This dissertation focuses mostly on Swedish politics and arguments made as to why the mechanisms behind social media would reinforce the position of some actors and perspectives should be considered taking into account the Swedish type of political system. The dissertation also includes two comparative case studies that address the dynamics between similar parliamentary systems, namely, those of Austria, the Netherlands, and Sweden. Strong political parties and a party-centered form of representative democracy characterize all three countries. In all three countries, the media system also follows a democratic corporatist model (Hallin & Mancini, 2004). Parties in different countries might additionally experience different development cycles that set the boundaries for their behavior, why it is an advantage that the politicians in these countries started to appropriate social media around the same time.

Comparing forms of use across these countries adds a valuable perspective to our understanding of the social media effects on political communication since there is a lack of comparative studies in this field, with most studies covering one country and one election at a time (Gibson, Römmele, & Williamson, 2014). The comparison of Austria, the Netherlands, and Sweden follows the logic of a most similar system design, which gives an opportunity to examine whether the causal relationship of interest can be observed in different contexts with similar characteristics and assessing whether the obtained conclusions can be generalizable to similar countries. The dissertation as a whole is heavily reliant on Swedish data but the comparative studies helps me to alleviate some of these concerns by showing that there is a pattern to populists’ use of social media that is different from other politicians, which could not have been done in absence of comparative research. Three of the articles in the dissertation, focus solely on Sweden, and each of them discusses the relevance of using Sweden as a case in more detail (article 1, 4 and 5). Focusing on a single case helps conduct a more thorough examination of the mechanisms at play and in this way the dissertation lays the ground for future research using a more comparative approach.

Social media users are not representative of the general population (Mellon & Prosser, 2017), which is why results as a rule cannot be generalized outside social media platforms. Moreover, user demographics and behaviors also differ between platforms. Therefore, when stretching research outside of a single platform it is important to consider the influence of the platform architecture. By studying political communication across platforms, we can better understand what types of patterns are universal and which are specific to a particular platform (Bode & Vraga, 2018). In particular, article 3 addresses this question by comparing platform specific communication patterns. Article 1 also asks about the importance of various social media platforms to the candidates, as well as other web-related platforms. Besides asking candidates about social media use in general and perceptions of the importance of various social media and web platforms, the empirical studies focus on behavior tied to Facebook and Twitter, as these are the most widely used social media platforms by political actors and for discussing politics. However, this might change over time as new platforms emerge.

A limitation with analyzing social media data at large is that platforms may not only shift in their overall popularity but also in who uses them and why. In addition, they might disappear altogether. Therefore, we have to exert caution when generalizing patterns both between platforms and from one platform to the external world. For example, if Facebook facilitates a

48 certain behavior that distinguishes some parties over others, it is not given that this pattern would endure if Facebook would vanish. This volatile nature of social media data will be discussed more below, along with other opportunities and limitations of using these data.

3.2 Data and methods

The growing social and political relevance of social media calls for researching these platforms in their own right. Most of the data collected in this dissertation therefore consist of data collected from social media platforms. Mainly, these data consist of text but also reactions such as shares, emoji’s or mentions, and are analyzed through a mixed methods approach of both quantitative and qualitative methods.

Social media use produces a vast amount of data. Collecting, storing and analyzing these data transform the social science discipline and perhaps our understanding of the political world as big data might be “even ameliorating some of the most important, but previously intractable, problems that affect human societies” (King 2014 , 166). Big data refers to new technological innovations such as machine learning that allows researchers to gather either new types of data such as social media data or vast quantities of traditional data with less expense (Clark & Golder, 2015). Social media data presents new opportunities for the study of politics and society, providing digitized and instantaneous data on processes and phenomena that were previously difficult to study empirically. This calls for rethinking methods traditionally used in social sciences. Conventional sampling and content analysis methods are out of tune with the forms and volumes of “big data” in hybrid information networks (Bennett & Pfetsch, 2018). However, social media also poses novel challenges for researchers seeking to abide by the standards of transparency, generalizability and replicability that underlie the empirically orientated social and political research (Sudulich et al., 2014).

I will return to a discussion of generalizability, validity, and reliability of the empirical material. Prior to that I will briefly discuss some key characteristics of online political information that shape and constrain any study of politics on social media. First, as noted above, social media generate large volumes of data; second, the heterogeneous character of social media data and third, the versatile nature of social media. Social media has introduced a previously unimaginable levels of automated or ‘natural’ data production, which occurs as a byproduct of the interactions that take place online (Sudulich et al., 2014). Digital trace data are continuously collected and provide reflections on major events as they unfold, unlike e.g. surveys (Jungherr,

2015). These data are also non-reactive, not produced by interactions between researchers and those who are studied that may allow avoidance of classical issues such as social desirability bias or other types of interview effects (Salganik, 2019). Another rare feature of digital trace data is that it captures social relationships such as networks of politicians. Although data appear “naturally”, it is not “natural” in a sense that behavior or digital trace reflects what people (at large) choose to share. This creates a form of bias in social media research. On the other hand, what people choose to share is what creates social media, and it accurately represents its nature.

The volume of data requires different methods to scrape and analyze information from social media. Twitter data collected in this dissertation rely on semi- supervised machine learning approaches to collect and structure the textual data. Some of the other studies, have been reliant on online program interfaces such as Netvizz, to retain data and making data queries. Accessing data from private social media platforms come with some trade-offs. For example, it is difficult to assess whether the data you can access are representative, although research indicates that samples show high similarity to the “full” data in terms of volume and topics (Hino & Fahey, 2019). Another issue with access is that data are often inaccessible. While Twitter provides publicly available data (not unfettered access, however), the majority of data generated on Facebook are private. At the moment of writing this even open data has been restricted and cannot be scraped.

The second feature I will focus on is the heterogeneous character and diverse nature of social media data (Sudulich et al., 2014). Social media data do not only comprise textual information or information about networks. Data can be transmitted in several other formats such as graphic or audio. Data contain various information about responses and dissemination as well as e.g. (URLs) links to other sources. Therefore, while this aspect of online political data presents opportunities to generate data that were previously unavailable to political and social science, it also holds difficult challenges and trade-offs.

The changeability of social media is a third characteristic. Content generated on social media might not be traced back (retrieved) over a long period of time and once something is not accounted for at the beginning of the data collection process, it is most likely that it will be irretrievable afterwards (Sudulich et al., 2014). In addition, the structure of the platforms is not constant – some activities and types of interaction are altered or discontinued, while others are introduced (Sudulich et al., 2014). As discussed above, the functionality of social media

50 platforms might vary over time and the extent to which this variation impacts political behavior on a platform will affect the results obtained when replicating previous studies. Comparing behavioral patterns between parties and candidates is one way to deal with this issue, but it will not solve the issue of replicability with new data over time.

Social media can be described as an environment where ‘everything flows’ at a previously unprecedented pace, with content and connections constantly being created, amended and deleted (Sudulich et al., 2014). There is thus an emerging disjuncture between social media’s permanence as an element of social and political life and the large-scale, diverse and highly ephemeral nature of its content (Sudulich et al., 2014). This has implications for both validity and reliability. The versatile nature of social media calls for making sure that data are stored appropriately and in a way that allows for reproducibility. Portions of data collected from social media platforms might be generated by ‘bots’ created to systematically distort data which poses another challenge to researchers. I discuss this issue in more detail in article 5. Employing strategies for parsing e.g. by only collecting data relevant to the topic of interest and distinguishing such data from non-relevant ‘noise’ when analyzing the data, have been suggested remedies (Sudulich et al., 2014).

What is important for all studies on political aspects of social media, is that it is often not accurate to generalize patterns outside social media platforms and the same goes for generalizing between platforms, as I discussed previously. Social media users, as previously mentioned, are not representative for the population at large. One way to address this uncertainty is to combine data sources. In several of the articles I do precisely this, by e.g. comparing behavior between Twitter and Facebook. Politicians official pages on Facebook or Twitter profiles are examples of additional sources that provide new publicly accessible and self-defined communication, which might yield insights to their overall campaign behavior but is likely to be influenced by the platform architecture.

To alleviate some of the problems that arise when analyzing a single platform or with single case studies, I utilize a broad array of data sources, such as social media and survey data, and combine them in several of the articles. For example, by not only measuring politicians’ self- reported motives of using social media but connecting this to their actual behavior on social media, can we say something more profound about potential differences in this case between male and female candidates.

Many of the empirical studies in the dissertation combine statistical analyses with text and content analyses. Article 4 employs semi-supervised automated methods in combination with manual classification techniques. Benoit and Laver (2006) discuss the differences between human coding and computerized coding in terms of reliability versus validity when studying election manifestos. Both techniques have benefits and limitations. Computerized techniques can claim perfect reliability while human coding provides a higher level of validity, as human coders with an awareness of the context can make inferences about content that is difficult for machine learning. In article 4, I combine techniques that allow for controlling the semi- automated output.

Table 2 below present the scope, material, and methods of each article. These are further discussed in more detail below. In Article 1, The role of gender in online campaigning: Swedish candidates’ motives and use of social media during the European election 2014, my co-author and I formulated questions that were incorporated in the Comparative Candidates Study in Sweden about the candidates’ motives for using social media and whether these were related to a more personalized or a more party-centered campaign strategy. We also formulated questions asking them to evaluate their own use of social media. Additionally, questions that were standard items in the survey were used, such as the assessment of important campaign activities. As there are known weaknesses with survey data, such as response biases, we also wanted to analyze the candidates’ actual use of social media, in this case, Twitter. Data on the candidates’ use of Twitter was collected during three time periods to capture a potential campaign effect of candidates’ use. We analyzed the differences in actual use and activity on Twitter by collecting all tweets from the candidates with a Twitter account and the use of retweets, mentions, hashtags, and links. To distinguish between different forms of use, we created two aggregated variables that we named “interaction” (retweets and mentions) and “information” (use of links and hashtags). In the models, we controlled for a number of variables that previous research had found important in some or most cases (i.e., age, incumbency, placement on the party list, political party, and seniority). The methodological strengths of the paper are that we take several periods into account, as most research focuses on the campaign period, and we compare motives for social media use with the actual use, unlike the majority of previous studies. Limitations to our design include that we only test candidate behavior on Twitter, and thus, we cannot say if the same pattern is likely to occur on other platforms, such as Facebook.

Table 2. Overview of data material methods

Article Title Scope Data material Methods 1 The role of gender in online Swedish The Swedish Factor analysis campaigning: Swedish candidates’ candidates candidate survey (principal- motives and use of social media running in the and behavioral component), linear during the European election 2014 2014 European data on the regression (OLS), election. candidates’ use of and negative Twitter during binomial three time regression. periods. 2 Twitter and Facebook: Populists’ Members of Data from official Multilevel double-barreled gun? parliament in Facebook pages regression models, Sweden, the of MPs and content analysis Netherlands, and Twitter data from Austria their open accounts. 3 Populist legislators on the People’s Members of Data from official Multilevel Platform: Facebook adoption and parliament in Facebook pages regression models, affective reactions in Austria, the Sweden, the of MPs included logit and linear Netherlands, and Sweden Netherlands, and in the study. regression Austria 4 Socially Mediated Issue Political issues Survey data Linguistic lexically Ownership and party (Swedish based approach, mentions National Election Pearson correlation Study) and Twitter data 5 Start Sharing the News: Exploring Traditional media Facebook data Content analysis, the Link between Right-Wing and right-wing covering links Pearson correlation Alternative Media and Social media outlets in shared (URLs) Media during the Swedish 2018 Sweden and link and interactions Election sharing on around these Facebook posts.

In Articles 2 and 3, Twitter and Facebook: Populists’ double-barreled gun?, and Populist legislators on the People’s Platform: Facebook adoption and affective reactions in Austria, the Netherlands, and Sweden, we examine populist members of parliament’s use of social media, relying on data from both Twitter and Facebook. Data was collected by scraping Twitter (Article 2) and Facebook data (Articles 2 and 3) for all the Members of Parliament (MPs) of Austria, the Netherlands, and Sweden. We collected 10,355 Facebook posts from the 287 MPs who had an active Facebook Page and 9,852 tweets for 475 MPs who were active on Twitter, for one month. The specific period was chosen to represent a comparable period, as no electoral campaigns were held at that time in any of the three countries.

Dependent variables in Article 2 are mentions of media accounts on Twitter and posts activating anger on Facebook. We first examined whether populist MPs use Twitter in order to publicly “name and shame” the media. Of the tweets collected from the MPs, we focused on those

53 containing a @-mention, to establish whether the mentioned account was a media account. We then examined if populist MPs named (@-mentioned) the media more often than other MPs. The dependent variable in our (multilevel) negative binomial regression model was the number of times an MP @-mentioned a media account. Our regression models include controls for several other important characteristics of MPs that other studies have shown to matter: gender, age, seniority (operationalized as days in parliament), and parliamentary leadership. We then carried out a content analysis of the tweets mentioning a media account and coded them as either “shaming” the media or not. The tweets that we considered to include shaming show the media source or journalist addressed in one (or more) of the following ways: (a) reproaching, (b) accusing, or (c) insulting. Simply voicing disagreement with the media account were not considered as shaming. All 3,232 tweets from both populist and nonpopulists that @-mentioned (“naming”) a media account were hand-coded by two different coders per country (the authors and three assistants).

To establish whether populists’ Facebook posts generate more angry reactions than those of other MPs, we measured the proportion of angry reactions relative to the total number of reactions (i.e., like, sad, love, anger, haha, or wow) on the post level. We use multilevel regression models and nested the messages (level 1) on the MPs page (level 2), not only because the nested structure of the data, but also because of the variability between the Facebook Pages, in terms of the number of followers. We complemented our multilevel analysis with a content analysis to assess whether the posts generating anger could be linked to the content. A subsample of the post evoking the angriest reactions and a random sample of the remaining posts were coded by the same coders as described above, using a coding scheme in which posts (a) included an active play on a sense of injustice leading to perceived unwanted and undeserved outcomes and/or (b) invoked a cultural backlash targeting the political elite. For examples of coded tweets and Facebook posts, see Article 2.

In Article 3, we zoomed in on the Facebook data to analyze all emotions, in the form of emojis, elicited by the MP’s posts. We also examined the uptake of Facebook profiles and pages as well as activity, to detect differences not only between populist MPs and others but also between the party leaders. Regarding adoption, we focused on whether the MPs have no account, a simple Facebook profile, or a more professional Facebook page. As this first set of dependent variables is dichotomous, we used logit regression models. With regard to activity, we included the number of posts as separate dependent variables and, given the distribution of the number of

54 posts, we used a linear regression model. The emotional reactions were measured as the relative number of emojis expressed as the “like,” “angry,” “love,” “haha,” “wow,” or “sad” emoji. As the proportion might also be a function of the attention a post has received overall, we controlled for the total number of reactions, in addition to the control variables described above, in our multilevel linear regression model.

The data collection from Facebook was done using Netvizz, a software used for collecting data from public Facebook groups and pages. The data contains no information about users nor user identification numbers. Since we collected the data, Facebook has imposed restrictions on access, and therefore Netvizz no longer exists. This obviously hampers any attempts to replicate the study in the future and, given the current situation, getting data access from Facebook is challenging. Replications using social media data are generally problematic in any event since the social media user interfaces change. For example, for a long time on Facebook, the “like” button was the only reaction one could give. In 2016, Facebook officially introduced an extended set of reactions for users, five predefined emotions, including "Love," "Haha," "Wow," "Sad," or "Angry." On Twitter, there was a restriction to 140 characters, which doubled in 2017. It is difficult to estimate how these major changes in the user interface impact results from previous studies.

Article 4 relied on survey data and data scraped from Twitter. The Swedish National Election Survey (2014) was used to examine issue ownership as perceived by the Swedish electorate. The question used in the survey was phrased as follows: “Are there, in your opinion, any party or parties with a good policy on [issue area]?” The analyzed issue areas consisted of nine different policy domains—namely, education, social welfare, the labor market, immigration/integration, taxes, health care, pensions, the economy, and the environment. The issue areas were compared to the discussion of these issues in relation to mentioning a Swedish party on Twitter.

The Twitter data were gathered from February 2014 to September 2014, using the Twitter Streaming API. There will always be a degree of uncertainty to Twitter’s sampled API service; however, little is known about how the mode of data collection biases the data that Twitter allows researchers to gather without additional costs (Jungherr, 2016). These limitations should be acknowledged, although it is difficult to estimate if and to what extent this ultimately affects the reliability of the data.

All tweets that mentioned one of the Swedish parties represented in the parliament were included in the data set. A dictionary-based automated lexical approach was used to structure data (i.e., the textual content of individual posts) from Twitter, using stemming and lemmatization libraries specific to the Swedish language. With this technique, the text is parsed into separate words, and inflectional endings are removed. This approach made it possible to manually code the content of the preprocessed words in the whole subsample and match these to the issue categories from the survey, using a predeveloped coding scheme of phrases and words relating to each political issue. The Swedish National Election Survey developed the coding scheme to code issue areas in open-ended survey questions. To my knowledge, this approach has not been used previously to classify political discussions on social media. However, it has a great advantage since a large corpus of textual data still can be analyzed more or less qualitatively, depending on the level of analysis. For example, a manual classification has advantages over choosing a predefined set of keywords that would limit the scope of analysis beforehand. A weakness with this approach is that it does not estimate positive or negative sentiment in the context of party mentions and political issues. For the aim of this article, this is less relevant since the focus lay on measuring how Twitter discussions mediate an issue ownership dimension, and this does not necessarily imply that there would be a positive association.

In Article 5, Start Sharing the News: Exploring the Link between Right-Wing Alternative Media and Social Media during the Swedish 2018 Election, we took an exploratory approach to analyzing sharing and engagement around content from right-wing alternative-media outlets on Facebook, compared to content from traditional news media. Data was collected by The Swedish public service broadcaster (SVT) and consisted of shared links (URLs) in open Facebook groups and engagement around these on Facebook during the three weeks prior to the Swedish national election and one week after the election on September 9, 2018. We analyzed the content in links shared from nationwide traditional news media and sites that we classified as right-wing alternative media, to compare issue engagement between the two sources on Facebook. The content analyses were done by journalists at SVT, and a random sample of links was coded by one of the authors to establish intercoder reliability. In addition, the content in the most engaging links originating from the alternative sites was examined qualitatively to reveal what type of messages and frames lead to high engagement. Our results indicate that Facebook enables visibility and the potential impact of messages from right-wing media. However, we cannot know if this engagement is driven by automatic or semi-automatic

56 accounts in attempts to manipulate the political agenda. Despite this uncertainty, our results suggest that Facebook provides an important platform for spreading controversial views and messages from previously marginal actors.

3.3 Ethical considerations

Addressing some ethical questions related to the gathering and analyzing of the social media data is in order. Data collection from social media took place from only publicly accessible data; on Twitter, only data from open accounts were collected. Collecting data on public figures, such as politicians, we assumed that they use social media to send messages to a public audience and in a public domain. Data from Twitter on politicians’ use was only collected on verified MPs or candidates, so no members of the general public are part of these data sets. On Facebook, we only collected data from elected politicians verified public pages, and the responses to politicians’ posts on Facebook are anonymized when accessed through the API. In Article 4, a larger Twitter dataset was used, based on public Twitter posts; however, it did not contain any user information, only text in tweets matching the search criteria.

4. Summary of the empirical studies

The role of gender in online campaigning: Swedish candidates’ motives and use of social media during the European election 2014

Article 1 investigates political candidates’ motives and use of social media platforms for campaign purposes from a gender perspective. How traditional media covers female candidates is argued to increase the importance of a medium in which a candidate can have full control over her message and sculpting her image. Social media also provides opportunities to establish direct contact with potential voters and form communities around specific issues. Sweden is used as a case, and since this context is relatively gender-equal, the new campaign possibilities that social media might provide for women candidates are seen as a function of equalization, whereas a lack of observed difference would be seen as a normalized pattern, suggesting that men and women are campaigning under relatively similar terms. The role of gender is tested using survey data on Swedish candidates and behavioral data from Twitter. Previous research into candidates’ use of social media has mostly examined campaign behavior. In this study, we examine the use of Twitter during three periods: before, during, and after the 2014 European election, to establish whether the campaign period has an impact on the relationships we test. Our findings show that female candidates tend to value social media to a greater extent, they differ in their stated motives of using social media, are more active and interactive. We also find that the election campaign affects activity level and how male and female candidates interact with their followers (male competitors step up their game, and the differences even out). This has implications for our understanding of normalization, as the campaign period leads to a normalized pattern of campaigning between the genders. This dynamic might thus be similar in the context of a national election, and the patterns we find are likely to be linked to the supra-national electoral context. Our result suggests that both electoral context and campaign versus noncampaign periods are important to address at the intraparty level.

In the Swedish context of equal representation in parliament but remaining gender differences in political power, the expectations we can derive from the normalization/equalization framework are not straightforward. Social media provides more equal conditions for women candidates but the potential effects of this use can still resemble a normalized pattern. However, in terms of social media opportunity structures, the results support the notion that women

58 candidates in their motives and use of social media realize the equalizing potential of social media.

Twitter and Facebook: Populists’ double-barreled gun?

In Article 2, we link populism to the specific architectures of Twitter and Facebook and examine which opportunities each of them offers to populists. We argue that social media can be a powerful weapon in the hands of populists, more specifically a double-barreled gun with two different targets. On Twitter, populists have the opportunity to target journalists and publicly “name and shame” them to increase media coverage. On the other hand, Facebook might be used to mobilize broad audiences directly by activating anger. We found that both strategies occur, though activating anger was more clearly linked to populists than naming and shaming practices on Twitter (naming the media in a reproaching, accusing, or insulting way). There were notable differences between populists and nonpopulists, as the latter hardly ever used naming and shaming and triggered anger less frequently.

The results illustrate what populists can achieve by using different social media platforms, and we confirm that differences in user base and architecture matter empirically. Therefore, distinguishing between the ideological character of populists’ compared to other party families, as well as distinguishing between different social media architectures are important since the architecture and user base of these platforms provide different opportunities, depending on the actor. As normalization theory stresses that the opportunity to use social media lies in bypassing traditional news media, our results suggest that actors can use social media instead to deliberately try to influence the news-media agenda. Twitter, in particular, should more closely be considered as part of the hybrid media system in which the media-specific opportunities overlap.

Populist legislators on the People’s Platform: Facebook adoption and affective reactions in Austria, the Netherlands, and Sweden

While social media is often said to benefit populist actors and reinforce the populist message, empirical studies to the contrary suggest that populist politicians are less likely to use social media. In this article, we explored this puzzling finding and extended the scope of analysis to Facebook, as most previous work focused on Twitter. We found that populist party leaders are most likely to use Facebook, whereas few of these parties’ elected parliamentarians are active in their professional role. Although populist ideology from a theoretical point of view goes well

59 together with the opportunity structures social media provides, it seems like the party structure of these parties makes a more personal and widely adopted use of social media within the party less favorable. The populist politicians that do use Facebook in their role as elected politicians, however, are more active compared to other MPs. Particularly the party leaders of populist parties seem to make active use of Facebook. In the second part of the article, we examine the role of emotions in evoking reactions on Facebook, as an opportunity that might be specifically linked to populist parties. As a part of a reciprocal relationship in populist mass-self- communication, the role of emotions has not been examined on social media, although it is present in the theoretical definitions of populism and populist communication. We find that populists overall get more emotional responses on Facebook in terms of emojis, particularly anger and laughter, compared to those of other parties. Therefore, we argue that the populist leaders might be beneficiaries of politicizing these communities' feelings of indignation while strengthening the bond with followers by humor, something that we interpret as often taking the form of irony or cynicism.

Populist parties might strengthen their position in relation to other parties through social media, which would indicate a more equalized pattern of party competition. Given potential interparty effects, the results from this article suggest that populist party leaders strengthen their position which on the contrary displays a normalized pattern since these parties are characterized by their strong leaders.

Socially Mediated Issue Ownership

This article examines to what extent the concept of issue ownership can explain variations in terms of how parties are addressed together with issues on Twitter. Survey measurements of issue ownership - the perceptions voters hold about parties’ relative competence to handle specific issues - was compared to how these issue-ownership perceptions mediate through political discussions on Twitter. In a hybrid media system, the Twitter agenda has been observed to strongly follow the traditional news media agenda when it comes to which topics are covered. Therefore, it is expected that Twitter discussions are structured in ways that can be anticipated, and thus the issue-ownership dimension that is relatively stable in Sweden is expected to strongly influence what issues are addressed with what parties in Twitter discussions. The results show a significant degree of issue consistency in issues with which parties are associated on Twitter and how representative surveys measure issue ownership. Some issue-ownership deviations also emerged, however, which could be interpreted as short-

60 term changes in parties’ issue competition. The results of this study imply that Twitter discussions follow a normalized discursive pattern in terms of few issue-ownership deviations, indicating that Twitter discussions resemble politics as usual. This result is perhaps not so surprising, given that Twitter has been argued to resemble an elite medium and, in particular political discussions around the parties might be carried out by citizens with high political interest, opinion-makers, politicians, and journalists. It is therefore uncertain to what extent the pattern found here holds true for the broader Twitter sphere.

Start Sharing the News: Exploring the Link Between Right-wing Alternative Media and Social Media during the Swedish 2018 Election

Article 5 examines sharing practices around content originating from right-wing alternative media on Facebook during the Swedish 2018 general election. We compared the number of links originating from these sites with traditional news media, examining what political issues were covered and how they differed, in terms of triggering shares and engagement. We also examined the content in the most engaging links to further establish the appeal of this type of content on Facebook. Compared to traditional news-media content, the engagement around links shared from right-wing alternative media was almost as high, and, of all the links included in our sample, 28% originated from right-wing media. This indicates that content originating from such sites had notable visibility on social media in comparison to traditional news-media content. In terms of political topics covered, immigration and integration-related issues dominated the covered content shared from right-wing alternative media. Content in links that evoked the highest engagement on Facebook similarly revolved around immigration, crime, and the Sweden Democrats. This content was framed in a highly partisan way.

A majority of sites that we included in our analyses are on the far right, with connections to extremist groups. The sizable amount of content shared from such sites is remarkable. It indicates that Facebook contributes to giving voice to contested views of these controversial news sites and a platform to broadening their reach. The growth of right-wing and far-right sites in Sweden and elsewhere made possible through social media distribution cannot be ruled out. This article also demonstrates the importance of examining social media as a part of a hybrid media system where various online media intertwine and present an opportunity structure for previously marginal perspectives to gain visibility. It is possible that in the future, Facebook will take on a much more active role as gatekeepers for the type of content allowed to be shared on the platform, which will then most likely dampen the influence of a far-right agenda.

5. Concluding discussion

Situated within a theoretical framework that attempts to explain the impact communication technologies has on politics in its broader sense, this dissertation has tested whether different dimensions of the normalization/equalization paradigm hold true in this age of social media. According to the general normalization thesis, we should find that social media replicates existing power differences and mainly benefits the already dominant political actors. According to the equalization perspective, on the contrary, we should find that social media contributes to leveling the playing field for previously disadvantaged political actors. To test these arguments, this dissertation has focused on actors that (arguably) were disadvantaged at the time when traditional editorial media functioned as stronger gatekeepers, moderating political discourse.

The overarching questions addressed in this dissertation were if social media create opportunities for “previously disadvantaged” actors? And, if so, how? In short, my answer to the first question is yes. Social media does create opportunities for previously disadvantaged actors. The way in which these opportunities present themselves, however, differs depending on the specific actor and social media platform. This is the answer to the question of how. That insight about the variations across actor and platform type is crucial to keep in mind when attempting to assess the specific implications the current online environment has on democracy. These short summary answers deserve to be unpacked. This is done in the subsequent paragraphs before I return to the pivotal issue of what implications my findings indicate the current online environment has on democracy.

From a normalization/equalization perspective the position of being previously disadvantaged is central to the idea of social media changing or maintaining power relationships in politics. However, based on the empirical studies, the conclusion can be drawn that opportunity structures do not form in the same way for these actors. Instead the two - that is actor and opportunity structure - are more closely connected. For populist actors, social media provides a way to mobilize supporters efficiently and maintain a hostile stance towards elites despite already having become influential. As for women candidates, we found that they place higher value on social media, are more active and interactive. This translates into an equalizing potential. Social media might therefore equalize the position of both type of actors vis-à-vis others, but this occurs in different ways and with qualitatively speaking different outcomes. The underlying assumption as to why this might come about is nevertheless the same; traditional

62 gatekeepers are losing ground. Not only does this enable certain types of political actors. It also creates discursive opportunities. On Twitter, the opportunities which otherwise might have become manifest are likely constrained by facets of the platform architecture in combination with the strong presence of traditional elites, at least in the Swedish context. Facebook, on the other hand was found to reinforce the position of previously marginal actors and perspectives through increased visibility of partisan news.

In terms of theoretical contributions, the dissertation adds to our understanding of why normalizing or equalizing trends occur, by demonstrating that it is imperative to assess the opportunity structures of social media platforms and the way these opportunity structures form for different actors. While some generic change linked to social media can be expected to take place among parties, social media do not hold the same potential for all candidates or parties. Previous research primarily forms expectations about the potential of minor parties or unknown candidates, based on their resources or lack thereof. This dissertation extends that argument to actors that have been overlooked within the normalization framework and links the equalizing potential of the platforms to actor types, showing that this link is more important than have been previously asserted. In an attempt to bridge the gap between actor- and structure-centric explanations, I also connected normalization/equalization as a three-dimensional theoretical framework to social media opportunity structures. Building the link between actor-centered motives and behavior on the one hand and the properties of social media platforms on the other hand, provides a more comprehensive picture of the ways that social media can impact representative democracy.

This work has also furthered our understanding of how gender matters within the (relational) equalization framework. In terms of theoretical contributions, this article also shows how the electoral context and the electoral cycle can moderate otherwise present effects, which has implications for our understanding of normalization/equalization. For female candidates, opportunities can be linked to taking better advantage of the interactive features of social media, as we saw in Article 1. In something of a contrast, populist actors have been found to be the least interactive. Populists may instead benefit from the reciprocal activity on social media platforms and Facebook in particular, where they can activate and harness the core emotional traits of the populist message.

Article 2 looked at and uncovered the extent to which populists use Twitter for expressing hostility toward established media and Facebook to activate anger among followers. Detailing

63 these linkages between opportunity structures of social media and populist parties add to our understanding of how emotional appeals increase engagement on Facebook. It also tells us that social media is more than a way to bypass traditional media for disadvantaged actors. It can, and is, also used to specifically target the traditional media and journalists. Article 3 introduces a distinction between populist ideology and populist party structure, which, we argue, explains the discrepancies in adoption and use among party representatives. For although populist ideology seems to go together well with social media opportunities, social media also pose a risk to these parties. It is therefore primarily the party leader or leadership that makes active use of social media, as we observe in our study. Furthermore, by comparing populist behavior on two platforms, not only do we provide insight into the relationship between populism and social media. We also empirically link populism to two specific but different social media opportunities. On Twitter, populists actively name and shame the media, whereas Facebook is used to activate anger.

Articles 4 and 5 add to our understanding of the discursive opportunities social media provides. Greater inclusion in public discourse through online platforms has often been seen as something positive for democracy since opportunities increase for marginalized voices to be heard. The rise of right-wing extreme, anti-feminism counterpublics challenge this conception and can be said to instead reproduce privilege (Holm, 2019). Both understandings of counterpublics include a challenge to status quo and the prevailing majority discourse. Social media can in this way function as a counterweight to the dominant discourse by offering discursive opportunities to raise issues left out of the mainstream agenda. In this dissertation, news-sharing practices on Facebook indicate that marginal and controversial views succeed in gaining visibility, whereas on Twitter, discussions resemble a normalized discursive pattern that if anything reinforced a mainstream media agenda by mainly mediating popular perceptions of issue ownership. At least in the cases examined here, Twitter discussions resemble more closely politics as usual, whereas the right-wing agenda were prominent in news- sharing practices on Facebook. Clearly, these two studies did not examine the same phenomenon across the two platforms, but, in terms of platform architecture and an equalizing potential, the two platforms might indeed offer different discursive opportunities. It is important to remember that single social media platforms are not a unitary environment; they consist of multiple online public spheres forming a layered structural environment (Ceron & Splendore, 2019). For instance, those discussing political parties on Twitter (Article 4) might be distinct from common users. Similarly, on

Facebook, the overall pattern of sharing and engagement around content on individual pages and in closed groups might look different than news sharing practices in open Facebook groups.

I now turn to the empirical contributions. First, gendered aspects of social media use were tested thoroughly in terms of motives, evaluation, and actual use. A key finding was that female candidates utilized social media more actively and interactively than their male peers. Second, populist parties were found to use Twitter and Facebook in a different way than other parties. On Facebook, the populist party leaders are particularly active and evoke an emotional response; they do not only activate anger but also expressions of humor or cynicism, as well as appreciation. The dissertation also furthers our understanding of Twitter’s role in opinion- formation processes, showing that perceptions of issue ownership strongly structure variation in a mediated issue ownership dimension on Twitter. Lastly, examining the sharing of partisan right-wing media content on Facebook demonstrated a connection that previously was only hypothesized. As much of the research in this field has been limited to one platform and one type of data, the overall important empirical contribution of this dissertation is the combination of data types and the cross-platform approach.

As all the articles relate to the overall theoretical framework of normalization/ equalization, some important caveats relating to this perspective should be addressed. To ascertain that party competition has been equalized, a strong presence on the web or social media should also be linked to an electoral advantage (Gibson McAllister, 2015). Similarly, the normalization thesis implies that offline ways of conducting political campaigning will override any style of electioneering specific to online channels (Lilleker & Jackson, 2011). To put it differently, social media should not matter for explaining relationships found outside of these channels from a normalization perspective. In contrast, it should matter from an equalization perspective. In this dissertation, these effects are not tested directly but merely hinted at. However, by identifying several important opportunities that social media offers, the results of the empirical studies indicate a change in the dynamics of politics, if not its effects. This dissertation therefore offers more of an entry point rather than a final answer to the supposed effects of social media. Future research should therefore continue the pursuit of establishing causal links between presence on and use of social media to effects measurable outside these platforms.

That perceptions of politics (traditionally) mediated through the press shape public opinion is known. In a hybrid media system, this mediation occurs on a multitude of platforms and through digital social networks, making mediation effects more difficult to deduce. How politics is

65 being mediated through social media is equally likely to influence both journalists and politicians’ perceptions of where the majority of the public stands on an issue, perceived majorities (Entman & Herbst, 2001). For example, news reporting might not affect the actual sentiments of individuals, but it does shape the perceived majority of opinions. Therefore, it is crucial to examine how these platforms mediate politics in their own right. That is not to say that social media offers a “true” image of reality. On the contrary, social media have so far turned out to be easily manipulated. For example, the party leader of the Sweden Democrats lost a quarter of his followers on Twitter after the company deleted thousands of fake accounts in 2018.7

Importantly, all major social media platforms are privately owned profit-driven companies. These technology companies are active agents in shaping political processes and political communication of electoral campaigns (Kreiss & McGregor, 2018). Any effects of social media on politics and how politics is mediated will also depend on their terms and business models.

Will social media deepen the void in Western democracy?

In the beginning of the theoretical chapter I discussed how individualization and globalization have contributed to weakened links between parties and citizens. This political vacuum has partly been filled by new anti-political-establishment parties in Western Europe. In majoritarian electoral systems, where entry of new parties is difficult, we have instead seen the rise of new types of political leaders. Both Trump and Obama are examples of outsider candidates and both have appropriated social media in distinct ways. Barack Obama became known as the first social media president in a positive sense (Katz, Barris & Jain, 2013). Unlike Obama, Trump has become known for misogyny, racism, and ignorance (Jones, 2018). They can in this regard be seen as each other’s opposites, but they do share the underlying trait of coming from “outside” the party establishment.

As of now it might seem as if social media facilitates right-wing populistic mobilization, but it is probably too early to rule out a counter movement to the left. What does not seem as plausible at the moment is that traditional parties increase popular support through social media above

7 https://omni.se/akesson-tappade-flest-nar-twitter-rensade-fejkkonton/a/kajrVB

66 and beyond what they otherwise would have. In the understanding of Mair (2013) and others, parties represent more than a ticket at the ballot box, and for democracy to stay healthy it is required that parties remain a strong link function by aggregating and articulating voters’ interests, and through the engagement of members in party organizations. What is central to arguments about the quality or performance of representative political systems “are the functions of collective political organizations” (Ward & Gibson 2008, p.26). Populism, whether left or right, challenges the idea of democratic representation and liberal democracy. In particular, what have been defined as authoritarian populism (Norris & Inglehart, 2019). Currently, social media seems to consolidate the populist message, which makes the prospect of social media improving representative democracy rather dim. Besides, for social media to contribute to reinforcing the link between citizens and parties, it requires more than just evoking attention and engagement on Facebook or Twitter.

Previously, I also discussed whether individual politicians might become more important at the expense of the party through connecting directly with voters on social media or by using social media to gain visibility in traditional media. At least in the Swedish context, however, we do not see any signs of increased personalization trends through social media. Even though female candidates valued social media to a greater extent than male candidates in the European election, they showed stronger motives for a party-centered campaign style.

It is important to recognize both the structural elements of communication and the actor driven motives - it is at the intersection at which actor meets communication technology that opportunity structures form. Traditionally, mass media such as TV and radio have been the medium for established parties. With the Internet came an additional channel for spreading information but it never altered the balance in favor of less established political actors (Margolis & Resnick, 2000). Through social media, information is spread through networks of users which change the dynamic and the speed by which political information travels. Social media is likely to strengthen the position of previously disadvantaged actors and these actors are likely to appropriate social media differently, differently from established actors but also differently from other actors that can be described as previously disadvantaged. This is why opportunity structures of social media should not only be understood in a uniform manner but in relation to a political actor and a specific platform. In this way, potential negative or positive implications for democracy can be addressed without landing on a dichotomous conclusion arguing that these developments are either (purely) inherently good or bad.

Normalization and the maintenance of the status quo has predominantly been portrayed in negative terms before. In some ways this is the whole premise of the “hollowing out” thesis by Mair (2013) and others. Namely that the maintenance of the status quo has led to ossification and subsequently the erosion of representative democracy. In contrast, equalization has been seen as a positive. Clearly, the reality on the ground belies this. This dissertation has laid bare a striking paradox – equalization can both benefit and potentially harm liberal democracy depending on which actors gain ground. It is therefore not given when politics as unusual is preferable to politics as usual, or vice versa.

6. Reference list

Aalberg, T., Esser, F., Reinemann, C., Strömbäck, J., & de Vreese. C. H. (2017). Populist

political communication in Europe. London: Routledge.

Barber, B.R. (2001). The uncertainty of digital politics. Harvard International Review, Vol.

23(1): 42-47.

Barr, R. R. (2009). Populists, outsiders and anti-establishment politics. Party Politics, 15(1),

29-48.

Bélanger, É, & Bonnie M. (2008). Issue salience, issue ownership, and issue-based vote

choice. Electoral Studies 27(3), 477-491.

Benkler, Y., Faris, R., & Roberts, H. (2018). Network propaganda: Manipulation,

disinformation, and radicalization in American politics. New York, NY: Oxford

University Press.

Benoit, K., & Laver, M. (2006) Party Policy in Modern Democracies. London: Routledge.

Benoit, W. L., Glantz, M. J., Phillips, A. L., Rill, L. A., Davis, C. B., Henson, J. R., &

Sudbrock, L. A. (2011). Staying "on message": Consistency in content of presidential

primary campaign messages across media. American Behavioral Scientist, 55(4),457-468.

Bimber, B. (1998). The internet and political transformation: Populism, community, and

accelerated pluralism. Polity, 31(1), 133-160.

Bimber, B. (1999). The Internet and citizen communication with government: Does the

medium matter? Political Communication, 16(4), 409-429.

Bimber, B., & Copeland, L. (2013). Digital media and traditional political participation over

time in the US. Journal of Information Technology & Politics, 10(2), 125-137.

Blombäck, S., & Sandberg, L. (2018). Sociala medier har blivit en självklarhet. I Nord, L.,

Grusell, M., Bolin, N. & Falasca, K. (red.) Snabbtänkt - reflektioner från valet 2018 av

ledande forskare. Sundsvall: Mittuniversitetet, DEMICOM.

Bode, L., & Dalrymple, K. E. (2016). Politics in 140 characters or less: Campaign

communication, network interaction, and political participation on Twitter. Journal of

Political Marketing, 15(4), 311-332.

Bode, L., & Vraga, E. K. (2018). Studying politics across media. Political

Communication, 35(1), 1-7.

Bos, L., & Brants, K. (2014). Populist rhetoric in politics and media: A longitudinal study of

the Netherlands. European Journal of Communication, 29(6), 703-719.

Bossetta, M. (2018). The digital architectures of social media: Comparing political

campaigning on Facebook, Twitter, Instagram, and Snapchat in the 2016 US election.

Journalism & Mass Communication Quarterly, 95(2), 471-496.

Bossetta, M., Dutceac Segesten, A., & Trenz, H. J. (2017). Engaging with European politics

through Twitter and Facebook: Participation beyond the national? In M. Barisione & A.

Michailidou (Eds.), Social media and European politics: Rethinking power and

legitimacy in the digital era (pp. 53-76). London, England: Palgrave Macmillan.

Boyd, D. M., & Ellison, N. B. (2007). Social network sites: Definition, history, and

scholarship. Journal of Computer‐Mediated Communication, 13(1), 210-230.

Bracciale, R., & Martella, A. (2017). Define the populist political communication style: The

case of Italian political leaders on Twitter. Information, Communication & Society, 20(9),

1310-1329.

Bratten, L. C. (2005). Online zealotry: la France du peuple virtuel. New Media &

Society, 7(4), 517-532.

Caiani, M., & Parenti, L. (2009). The dark side of the web: Italian right-wing extremist groups

and the Internet. South European Society and Politics, 14(3), 273-294.

Castells, M. (1996). The information age: Economy, society and culture. Oxford, UK: Wiley

Blackwell.

Ceron, A., & Splendore, S. (2019). ‘Cheap Talk’? Second screening and the irrelevance of TV

political debates. Journalism, 20(8), 1108–1123.

Chadwick, A. (2006). Internet politics: States, citizens, and new communication technologies.

Oxford University Press, USA.

Chadwick, A. (2013). The hybrid media system: Politics and power. Oxford, UK: Oxford

University Press.

Clark, W. R., & Golder, M. (2015). Big data, causal inference, and formal theory:

Contradictory trends in political science?. PS, Political Science & Politics, 48(1), 65.

Copeland, L., & Römmele, A. (2014). Beyond the base? Political parties, citizen activists, and

digital media use in the 2009 German federal election campaign. Journal of Information

Technology & Politics, 11(2), 169-185.

Dahlberg, L. (2001). The Internet and democratic discourse: Exploring the prospects of online

deliberative forums extending the public sphere. Information, Communication &

Society, 4(4), 615-633.

Dalton, R. J. (2018). Political Realignment: Economics, Culture, and Electoral Change.

Oxford University Press.

De Vries, C., & Hobolt, S. (2020). Political entrepreneurs: the rise of challenger parties in

Europe. Princeton University Press.

Dolezal, M. (2015). Online campaigning by Austrian political candidates: Determinants of

using personal websites, Facebook, and Twitter. Policy & Internet, 7(1), 103-119.

Druckman, J. N., Kifer, M. J., & Parkin, M. (2009). Campaign communications in U.S.

congressional elections. American Political Science Review, 103(3), 343-366.

Engesser, S., Ernst, N., Esser, F., & Büchel, F. (2017a). Populism and social media: How

politicians spread a fragmented ideology. Information, Communication & Society, 20(8),

1109-1126.

Engesser, S., Fawzi, N., & Larsson, A.O. (2017b). Populist online communication:

Introduction to the special issue. Information, Communication & Society, 20(9), 1279-

1292.

Entman, R. M., & Herbst, S. (2001). Reframing public opinion as we have known it. In

Bennett, W.L., & Entman, R.M. (Eds.), Mediated politics: Communication in the future

of democracy (pp. 203-225). New York: Cambridge University Press.

Ernst, N., Blassnig, S., Engesser, S., Büchel, F., & Esser, F. (2019a). Populists prefer social

media over talk shows: An analysis of populist messages and stylistic elements across six

countries. Social Media + Society, 5(1), doi: 2056305118823358.

Ernst, N., Esser, F., Blassnig, S., & Engesser, S. (2019b). Favorable opportunity structures for

populist communication: Comparing different types of politicians and issues in social

media, television and the press. The International Journal of Press/Politics, 24(2), 165-

188.

Ernst, N., Engesser, S., Büchel, F., Blassnig, S., & Esser, F. (2017). Extreme parties and

populism: An analysis of Facebook and Twitter across six countries. Information,

Communication & Society, 20(9), 1347-1364.

Esser, F., Stępińska, A., & Hopmann, D. (2016). Populism and the media: Cross-national

findings and perspectives. In Aaalberg, T., Esser, F., Reinemann, C., Strömbäck, J., & De

Vreese, C. H. (Eds.), Populist political communication in Europe (pp. 365-380). New

York, NY: Routledge

Esser, F., & Strömbäck, J. (2014). Mediatization of politics: Towards a theoretical framework.

In F. Esser and J. Strömbäck (eds.) Mediatization of politics: Understanding the

transformation of Western democracies (pp. 3-28). London: Palgrave Macmillan.

Evans, H. K., Cordova, V., & Sipole, S. (2014). Twitter style: An analysis of how House

candidates used Twitter in their 2012 campaigns. PS: Political Science & Politics, 47(2),

454-462.

Fernquist, J., Kaati, L., & Schroeder, R. (2018). Political Bots and the Swedish General

Election. 2018 IEEE International Conference on Intelligence and Security Informatics

(ISI), Miami, FL, 2018, pp. 124-129, doi: 10.1109/ISI.2018.8587347.

Fisher, J., Fieldhouse, E., & Cutts, D. (2014). Members are not the only fruit: Volunteer

activity in British political parties at the 2010 general election. The British Journal of

Politics and International Relations, 16(1), 75-95.

Follesdal, A., & Hix, S. (2006). Why there is a democratic deficit in the EU: A response to

Majone and Moravcsik. JCMS: Journal of Common Market Studies, 44(3), 533-562.

Galston, W. (2018). The populist challenge to liberal democracy. Journal of Democracy

29(2):5-19.

Gainous, J., & Wagner, K. M. (2013). Tweeting to power: The social media revolution in

American politics. Oxford, UK: Oxford University Press.

Gerbaudo, P. (2018). Social media and populism: An elective affinity? Media, Culture &

Society, 40(5), 745-753.

Gerbaudo, P. (2019). The digital party: Political organisation and online democracy. London,

England: Pluto Press.

Gibson, R., Greffet, F., & Cantijoch, M. (2017). Friend or foe? Digital technologies and the

changing nature of party membership. Political Communication, 34(1), 89-111.

Gibson, R., Lusoli, W., & Ward, S. (2008). Nationalizing and normalizing the local? A

comparative analysis of online candidate campaigning in Australia and Britain. Journal

of Information Technology & Politics, 4(4), 15-30.

Gibson, R., & McAllister, I. (2015). Normalising or equalising party competition? Assessing

the impact of the web on election campaigning. Political Studies, 63(3), 529-547.

Gibson, R., & Römmele, A. (2001). Changing campaign communications: A party-centered

theory of professionalized campaigning. Harvard International Journal of Press/Politics,

6(4), 31-43.

Gibson, R., Römmele, A., & Williamson, A. (Eds.). (2014.) Chasing the digital wave.

international perspectives on the growth of online campaigning. Journal of Information

Technology and Politics, 11(2), 123-129.

Gibson, R., & Ward, S. (2000). An outsider's medium? The European elections and UK party

competition on the Internet. British Elections & Parties Review, 10(1), 173-191.

Grill, C. (2016). How anti-European, populist parties campaigned in the 2014 EP election. In

C. Holtz-Bacha (Ed.), Europawahlkampf 2014. Internationale studien zur Rolle der

Medien (pp. 75–96). Wiesbaden: Springer VS.

Gurevitch, M., Coleman, S., & Blumler, J. (2009). Political Communication—Old and New

Media Relationships. The Annals of the American Academy of Political and Social

Science, 625, 164-181. Retrieved July 10, 2020, from www.jstor.org/stable/40375913

Hallin, D. C., & Mancini, P. (2004). Americanization, globalization, and

secularization. Comparing political communication, 25-44.

Hameleers, M. (2018a). Augmenting polarization via social media? A comparative analysis of

Trump’s and Wilders’ online populist communication and the electorate’s interpretations

surrounding the elections. Acta Politica, doi:10.1057/s41269-018-0119-8.

Hameleers, M. (2018b). A typology of populism: Toward a revised theoretical framework on

the sender side and receiver side of communication. International Journal of

Communication, 12, 20.

Hameleers, M., Bos, L., & de Vreese, C. H. (2017). “They did it”: The effects of

emotionalized blame attribution in populist communication. Communication

Research, 44(6), 870-900.

Hameleers, M., & de Vreese, C. (2018) To whom are “the people” opposed? Conceptualizing

and measuring citizens’ populist attitudes as a multidimensional construct, Journal of

Elections, Public Opinion and Parties, DOI: 10.1080/17457289.2018.1532434

Hansen, K. M., & Kosiara-Pedersen, K. (2014). Cyber-campaigning in Denmark: Application

and effects of candidate campaigning. Journal of Information Technology & Politics,

11(2), 206-219.

Held, D. (2006). Models of Democracy. Third Edition, Stanford University Press.

Helmond, A. (2015). The platformization of the web: Making web data platform ready. Social

Media+ Society, 1(2), doi:10.1177/2056305115603080.

Hino, A., & Fahey, R. A. (2019). Representing the Twittersphere: Archiving a representative

sample of Twitter data under resource constraints. International journal of information

management, 48, 175-184.

Holm, M. (2019). The Rise of Online Counterpublics?: The Limits of Inclusion in a Digital

Age (Doctoral dissertation, Department of Government).

Huber, R. A., & Schimpf, C. H. (2017). On the distinct effects of left-wing and right-wing

populism on democratic quality. Politics and Governance, 5(4), 146-165.

Innis, H. A. (2008). The bias of communication. Toronto, Ontario, Canada: University of

Toronto Press.

Jackson, N. A., & Lilleker, D. G. (2011). Microblogging, constituency service and impression

management: UK MPs and the use of Twitter. Journal of Legislative Studies, 17(1), 86-

105.

Jackson, N. & Lilleker, D. (2009). Building an architecture of participation? Political parties

and Web 2.0 in Britain. Journal of Information Technology & Politics, 6(3-4): 232-250.

Jacobs, K., & Spierings, N. (2016). Social media, parties, and political inequalities. New

York: Palgrave Macmillan.

Jacobs, K., & Spierings, N. (2018). A populist paradise? Examining populists’ Twitter

adoption and use. Information, Communication & Society, 22(12), 1681-1696.

Jones, B. D. (2001). Politics and the architecture of choice: Bounded rationality and

governance. University of Chicago Press.

Jones, K. (2018). ‘A man who revels in his own ignorance, racism and misogyny’:

Identifiable referents trump indefinite grammar. Functional Linguistics, 5(1), 11.

Jungherr, A. (2015). Analyzing political communication with digital trace data. Cham,

Switzerland: Springer.

Jungherr, A. (2016). Twitter use in election campaigns: A systematic literature review.

Journal of Information Technology & Politics, 13(1), 72-91.

Jungherr, A., Schroeder, R., & Stier, S. (2019). Digital Media and the Surge of Political

Outsiders: Explaining the Success of Political Challengers in the United States, Germany,

and China. Social Media+ Society, 5(3), DOI: 2056305119875439.

Kalnes, Ø. (2009). Norwegian parties and Web 2.0. Journal of Information Technology &

Politics, 6(3-4), 251-266.

Kalsnes, B. (2016). The social media paradox explained: comparing political parties’

Facebook strategy versus practice. Social Media+ Society, 2(2),

doi:10.1177/2056305116644616.

Kang, T., Fowler, E. F., Franz, M. M., & Ridout, T. N. (2018). Issue Consistency? Comparing

Television Advertising, Tweets, and E-mail in the 2014 Senate Campaigns. Political

Communication, 35(1), 32-49.

Karlsen, R., & Enjolras, B. (2016). Styles of social media campaigning and influence in a

hybrid political communication system: Linking candidate survey data with Twitter data.

The International Journal of Press/Politics, 21(3), 338-357.

Katz, J., Barris, M., & Jain, A. (2013). The social media president: Barack Obama and the

politics of digital engagement. New York: Palgrave Macmillan.

Katz, R. S., & Mair, P. (2018). Democracy and the cartelization of political parties. Oxford

University Press.

Karpf, D. (2012) Social Science Research Methods in Internet Time. Information,

Communication & Society, 15 (5), 639–661.

Keller, T. R., & Kleinen-von Königslöw, K. (2018). Followers, spread the message!

Predicting the success of Swiss politicians on Facebook and Twitter. Social Media+

Society, 4(1), 1-11.

King, G. (2014). Restructuring the Social Sciences: Reflections from Harvard’s Institute for

Quantitative Social Science. PS: Political Science and Politics, 47 (1): 165 –72.

Kitschelt, H. P. (1986). Political opportunity structures and political protest: Anti-nuclear

movements in four democracies. British journal of political science, 57-85.

Klinger, U. (2013). Mastering the art of social media: Swiss parties, the 2011 national election

and digital challenges. Information, Communication & Society, 16(5), 717–736.

doi: 10.1080/1369118X.2013.782329

Klinger, U., & Svensson, J. (2015). The emergence of network media logic in political

communication. A theoretical approach. New Media Society, 17(8),1241–1257.

Klinger, U., & Svensson, J. (2016). Network media logic. Some conceptual clarification. In:

Bruns, A., Enli, G., Skogerbø, E., Larsson, A.O., Christensen, C. (eds) Routledge

companion to social media and politics, (pp. 23-38). London: Routledge.

Knutsen, O. (2007). ‘The Decline of Social Class.’ In R. J. Dalton and H.D. Klingemann eds.,

The Oxford Handbook of Political Behavior. Oxford: Oxford University Press, pp. 457–

480.

Koopmans, R., & Olzak, S. (2004). Discursive opportunities and the evolution of right-wing

violence in Germany. American Journal of Sociology, 110(1), 198-230.

Kreiss, D. (2012). Taking our country back: The crafting of networked politics from Howard

Dean to Barack Obama. New York: Oxford University Press.

Kreiss, D. (2016). Seizing the moment: The presidential campaigns’ use of Twitter during the

2012 electoral cycle. New Media & Society, 18(8), 1473-1490.

Kreiss, D., & McGregor, S. C. (2018). Technology firms shape political communication: The

work of Microsoft, Facebook, Twitter, and Google with campaigns during the 2016 US

presidential cycle. Political Communication, 35(2), 155-177.

Kriesi, H., Koopmans, R., Duyvendak, J. W., & Giugni, M. G. (1995). The politics of new

social movements in Western Europe. A Comparative Analysis. London: University of

Minnesota.

Krook, M. L., & Sanín, J. R. (2019). The cost of doing politics? Analyzing violence and

harassment against female politicians. Perspectives on Politics, 1-16,

doi:10.1017/S1537592719001397 .

Kruikemeier, S. (2014). How political candidates use Twitter and the impact on votes.

Computers in Human Behavior, 34, 131-139.

Krämer, B. (2017). Populist online practices: The function of the Internet in right-wing

populism. Information, Communication & Society, 20(9), 1293-1309.

Larsson, A. O., & Moe, H. (2014). Triumph of the underdogs? Comparing Twitter use by

political actors during two Norwegian election campaigns. Sage Open, 4(4),

DOI:10.1177/2158244014559015.

Lassen, D. S., & Brown, A. R. (2011). Twitter: The electoral connection? Social Science

Computer Review, 29(4), 419-436.

Lee, J., & Xu, W. (2018). The more attacks, the more retweets: Trump’s and Clinton’s agenda

setting on Twitter. Public Relations Review, 44(2), 201-213.

Littler, M., & Feldman, M. (2017). Social media and the cordon sanitaire. Journal of

Language and Politics, 16(4), 510-522.

Mansbridge, J. (1999). Should blacks represent blacks and women represent women? A

contingent" yes". The Journal of politics 61(3): 628-657.

Margolis, M., & Resnick, D. (2000). Politics as usual: The cyberspace “revolution”. London:

Sage.

Margolis, M., Resnick, D., & Levy, J. (2003). Major parties dominate, minor parties struggle:

US elections and the Internet. In: Gibson, R, Nixon, P, Ward, S (eds.) Political Parties

and the Internet: Net Gain? London: Routledge, pp. 53–69

Margolis, M., Resnick, D., & Wolfe, J. (1999). Party competition on the Internet: Minor

versus major parties in the UK and USA. Harvard International Journal of

Press/Politics, 4(4), 24-47.

McGregor, S. C., Lawrence, R. G., & Cardona, A. (2017). Personalization, gender, and social

media: gubernatorial candidates’ social media strategies. Information, Communication &

Society, 20(2), 264-283.

McGregor, S. C., & Molyneux, L. (2018). Twitter’s influence on news judgment: An

experiment among journalists. Journalism, 21(5):597-613.

McLuhan, M. (1994). Understanding media: The extensions of man. Cambridge, MA: MIT

press.

Mellon, J., & Prosser, C. (2017). Twitter and Facebook are not representative of the general

population: Political attitudes and demographics of British social media users. Research

& Politics, 4(3), doi: 2053168017720008.

Mounk, Y. (2018). The people vs. democracy: Why our freedom is in danger and how to save

it. Cambridge, MA: Harvard University Press.

Mudde, C. (2004). The populist zeitgeist. Government and opposition, 39(4), 542-563.

Mudde, C. (2007). Populist radical right parties in Europe. Cambridge: Cambridge

University Press.

Norris, P. (2002). Democratic phoenix: Reinventing political activism. Cambridge University

Press.

Norris, P., & Curtice, J. (2008). Getting the message out: A two-step model of the role of the

Internet in campaign communication flows during the 2005 British general election.

Journal of Information Technology & Politics, 4(4), 3-13.

Norris, P., & Inglehart, R. (2019). Cultural backlash: Trump, Brexit, and authoritarian

populism. Cambridge University Press.

Pappas, T. S. (2016), Are populist leaders “charismatic”? The evidence from Europe.

Constellations, 23, 378-390.

Petrocik, John R. "Issue ownership in presidential elections, with a 1980 case study."

American journal of political science (1996): 825-850.

Porten-Cheé, P., Haßler, J., Jost, P., Eilders, C., & Maurer, M. (2018). Popularity cues in

online media: Theoretical and methodological perspectives. SCM Studies in

Communication and Media, 7(2), 208-230.

Resnick, D. (1998) Politics on the internet: The normalization of cyberspace. In: C. Toulouse

and T. W. Luke (eds.) The Politics of Cyberspace (pp. 48-68). New York: Routledge.

Rheingold, H. (1994). The virtual community: Finding connection in a computerized world.

London: Minerva.

Rokkan, S. (1970). Citizens, Elections, Parties. Oslo: Universitetforlaget.

Rohrschneider, R., & Whitefield, S. (2012). The strain of representation: How parties

represent diverse voters in Western and Eastern Europe. Oxford University Press.

Rooduijn, M. (2014). The nucleus of populism: In search of the lowest common

denominator. Government and Opposition, 49, 573-599.

Römmele, A. (2003). Political parties, party communication and new information and

communication technologies. Party Politics, 9(1), 7-20.

Salganik, M. J. (2019). Bit by bit: Social research in the digital age. Princeton University

Press.

Sandberg, L., Bjereld, U., Bunyik, K., Forsberg, M., & Johansson, R. (2019). Issue Salience

on Twitter During Swedish Party Leaders’ Debates. Nordicom Review, 40(2), 49-61.

doi: https://doi.org/10.2478/nor-2019-0033

Schroeder, R. (2018). Social theory after the Internet: Media, technology and globalization.

London, England: UCL Press.

Schulz, A. (2019). Where populist citizens get the news: An investigation of news audience

polarization along populist attitudes in 11 countries. Communication Monographs, 86(1),

88-111.

Schweitzer, E. J. (2011). Normalization 2.0: A longitudinal analysis of German online

campaigns in the national elections 2002–9. European Journal of Communication, 26(4),

310-327.

Selle, P. & Svåsand, L. (1991). Membership in Political Parties and the Problems of Decline

of Parties. Comparative Political Studies. 4, 459-478.

Smith M. (2014). A Crisis of Political Parties. In: Richards D., Smith M., Hay C. (eds)

Institutional Crisis in 21st-Century Britain. Understanding Governance Series. Palgrave

Macmillan, London.

Southern, R. (2015). Is web 2.0 providing a voice for outsiders? A comparison of personal

web site and social media use by candidates at the 2010 UK general election. Journal of

Information Technology & Politics, 12(1), 1-17.

Southern, R. & Lee, B. (2019): Politics as usual? Assessing the extent and content of

candidate-level online campaigning at the 2015 UK general election, Journal of

Elections, Public Opinion and Parties, 29(2), 179-198.

Spierings, N., & Jacobs, K. (2014). Getting personal? The impact of social media on

preferential voting. Political Behavior, 36(1), 215-234.

Stier, S., Bleier, A., Lietz, H., & Strohmaier, M. (2018). Election campaigning on social

media: Politicians, audiences, and the mediation of political communication on Facebook

and Twitter. Political Communication, 35(1), 50-74.

Strandberg, K. (2009). Online campaigning: An opening for the outsiders? An analysis of

Finnish parliamentary candidates’ websites in the 2003 election campaign. New Media &

Society, 11(5), 835-854.

Stromer-Galley, J. (2014). “Political Discussion and Deliberation Online.” In Kenski, K., &

Hall Jamieson, K. (Eds.), The Oxford Handbook of Political Communication. New

York: Oxford University Press. doi:10.1093/oxfordhb/9780199793471.013.015.

Strömbäck, J. (2008). Four phases of mediatization: An analysis of the mediatization of

politics. The International Journal of Press/Politics, 13(3), 228-246.

Strömbäck, J., Ørsten, M., & Aalberg, T. (2008) Communicating politics: political

communication in the nordic countries. Gothenburg: Nordicom.

Tromble, R. (2016). The great leveler? Comparing politician-citizen Twitter engagement in

three Western democracies. European Political Science, 1, 1-17.

Tucker, J. A., Theocharis, Y., Roberts, M. E., & Barberá, P. (2017). From liberation to

turmoil: Social media and democracy. Journal of Democracy, 28(4), 46-59.

Van Aelst, P., Strömbäck, J., Aalberg, T., Esser, F., de Vreese, C., Matthes, J., ... Stanyer, J.

(2017). Political communication in a high-choice media environment: A challenge for

democracy? Annals of the International Communication Association, 41(1), 3-27.

Van Biezen, I., Mair, P., & Poguntke, T. (2012). Going, going,... gone? The decline of party

membership in contemporary Europe. European journal of political research, 51(1), 24-

56.

Van Dalen, A., Fazekas, Z., Klemmensen, R., & Hansen, K. M. (2015). Policy considerations

on Facebook: Agendas, coherence, and communication patterns in the 2011 Danish

parliamentary elections. Journal of Information Technology & Politics, 12(3), 303-324.

Van Dijck, J., & Poell, T. (2013). Understanding social media logic. Media and

communication, 1(1), 2-14.

Van Kessel, S. (2015). Populist Parties in Europe: Agents of Discontent? Houndsmills:

Palgrave Macmillan

Van Kessel, S., & Castelein, R. (2016). Shifting the blame. Populist politicians’ use of Twitter

as a tool of opposition. Journal of contemporary European research, 12(2), 594–614.

Vaidhyanathan, S. (2018). Antisocial media: How Facebook disconnects us and undermines

democracy. Oxford University Press.

Vergeer, M., & Hermans, L. (2013). Campaigning on Twitter: Microblogging and online

social networking as campaign tools in the 2010 general elections in the Netherlands.

Journal of computer-mediated communication, 18(4), 399-419.

Vestin, E. (2019). The Decline of Class Voting in Sweden 1968–2014: Reconsiderations,

Explanations and the Role of the New Middle Class. Göteborg Studies in Politics,

161, http://hdl.handle.net/2077/60085

Wang, Y., et al. (2017). Women's rights in democratic transitions: A global sequence analysis,

1900–2012. European Journal of Political Research. 56(4): 735-756.

Ward, S., & Gibson, R. (2008). European political organizations and the internet:

mobilization, participation, and change. In Routledge handbook of Internet politics (pp.

41-55). Routledge.

Zhuravskaya, E., Petrova, M., & Enikolopov, R., (2020). Political Effects of the Internet and

Social Media (2020). Annual Review of Economics, Vol. 12, pp. 415-438.

Åström, J & Karlsson, M. (2013). Blogging in the shadow of parties: Exploring ideological

differences in online campaigning, Political Communication, 30(3), 434-455.

JOURNAL OF INFORMATION TECHNOLOGY & POLITICS 2017, VOL. 14, NO. 4, 314–333 https://doi.org/10.1080/19331681.2017.1369918

The role of gender in online campaigning: Swedish candidates’ motives and use of social media during the European election 2014 Linn A. C Sandberg and Patrik Öhberg

ABSTRACT KEYWORDS This study addresses politician’s motives and use of social media platforms for campaigning Election campaign; purposes from a gender perspective. The role of gender is tested using survey data on Swedish European election; gender; candidates and behavioral data from Twitter before, during, and after the 2014 European election. social media; Twitter Our findings show gender-related patterns in how candidates assess the usefulness of social media as a campaign tool. Female candidates tend to value social media to a greater extent and evaluate social media use more positively than male candidates. We also find that the election campaign affects how male and female candidates interact with their followers.

Introduction of online campaign tools. We therefore know very little about the contours of gendered differences, Social media can provide candidates with better con- both in terms of motives and actual use. Moreover, trol over their self-images and public narratives. As a most of the studies concerning online campaigning result, candidates have altered their campaign tactics have been conducted in Anglo-American contexts to avoid the filtered interpretation of campaign mes- (Hermans & Vergeer, 2012; Jungherr, 2014), where sages released through traditional media outlets the political systems are more candidate-oriented (Gainous & Wagner, 2013). Even as social media is and female politicians are quite few. increasingly integrated into election campaigning, In this study, we test the relationship between the question of what it signifies for politicians to be gender and candidate approaches to social media active online remains largely unanswered. This is campaigning by exploring a party-oriented system especially true for female candidates. with a strong presence of female politicians, namely Because individual candidates can influence pub- Sweden. Women have been present in politics and lic perception through strategic use of social media, parliament for a long time in Sweden, and gender- such media might be a particularly useful tool with equal political representation is the norm rather than which women candidates can counteract masculine an unachievable ideal. Exploring gendered campaign communication norms in politics (Meeks, 2013). strategies and practices in this relatively gender- The possibility of direct communication and inter- equal political setting allows us to test whether action with voters and followers may also be parti- there is a difference between the motives behind cularly compelling to women who wish to avoid the and use of social media platforms for campaigning images given to them by traditional media, which is purposes by male and female candidates, even in a generally less favorable to female candidates (Heith, seemingly gender-equal context. Moreover, Sweden 2003;Jalalzai,2006;Kahn,1996). In this regard, provides an interesting case for examining candi- social media could provide a means to strengthen dates’ use of social media in general, since Swedish women candidates by presenting new opportunities parties and candidates stand out as early adopters to reach out, mobilize voters, and target particular and the most extensive users of Web technologies groups, such as young women. There are, however, (Vergeer, Hermans, & Cunha, 2013). very few studies that explicitly explore gender-based Additionally, most studies concerning the factors differences in social media use, rather than simply driving Web campaigning do not account for indi- including gender as a control variable in the uptake vidual attitudes and subjective assessments of the

CONTACT Linn A. C Sandberg [email protected] Gaustadalléen 21 Forskningsparken, OSLO 0349, Norway. © 2017 Taylor & Francis JOURNAL OF INFORMATION TECHNOLOGY & POLITICS 315 value of online platforms as campaign tools. research has shown that engagement with voters Previous research findings have also confirmed a online also leads to positive evaluations of politi- discrepancy between the intention to use Web cam- cians (e.g., Grant, Moon, & Grant, 2010; Utz, 2009; paigning and the actual adoption (Marcinkowski & Vergeer & Hermans, 2013). Politicians who use a Metag, 2014; Segaard & Nielsen, 2013). By combin- more interactive communication strategy can ing survey data with data regarding how candidates additionally obtain a larger number of followers actually use social media, we expected to gain a (Lilleker & Koc-Michalska, 2013), which in turn deeper insight into the factors that drive the use can redistribute content. The effects of social of social media for election campaigning. When media interactivity in political campaigns, such as considering types of activity on Twitter, we also exposure to comments, can conversely shape both used three time periods to capture the candidates’ positive and negative perceptions of candidates behavior during noncampaign periods, whereas (Brewer et al., 2015), suggesting that it is impor- most studies have concentrated only on the cam- tant for politicians to control the content on their paign period surrounding elections when assessing social media profiles. the online behavior of politicians (Larsson, 2015a). Because it gives control over messages and stra- The paper will proceed as follows. First, we tegies, social media may work particularly well for discuss candidates’ use of social media during elec- female candidates, who are often unelected or have tion campaigns and the possible impact of this use. junior status and, therefore, are less likely to be Second, we discuss research concerning gendered constrained by having to conform to the party campaigning practices, emphasizing the online message (European Parliament, 2013). For female environment. Thereafter, we discuss Sweden as a candidates in particular, social media can also case for studying gender and online campaign allow the contradiction of pertinent gender stereo- practices together with data and design. Finally, types, such as negative appraisals of issue compe- we discuss our results and their implications. tencies, in voter evaluations (Dolan, 2010). The potential benefits will be discussed further in the next section, along with previous research con- Social media and personalized campaigning cerning gender and campaigning, as well as pre- The mediatization of politics and the arrival of vious studies on candidates’ use of social media. social media coincide with a development toward individualized campaigning, in which a more per- Gender representation as a function of the sonal communication style has become advanta- equalization perspective geous (see Bimber & Davis, 2003; Strömbäck, 2008; Vergeer & Hermans, 2013). The characteris- The arrival and amplified use of social media as an tics of social media provide individual politicians interactive, open, and transparent platform has with new opportunities for personal campaigning raised questions of its democratizing potential. A and political communication. The individual politi- main issue to be addressed is whether new tech- cian can easily adopt social media as a campaign nologies can improve democratic processes by tool and the opportunities for more personalized allowing interaction among candidates, parties, campaigning through social media can be expected and voters, or by organizing election campaigns to increase even within party-centered systems (Enli in less costly and more direct ways (Gainous & &Skogerbø,2013;Karlsen,2011), such as Sweden. Wagner, 2013). According to the equalization the- In addition to bypassing media filters and jour- ory (Barber, 2001), the use of the Internet helps nalistic frames, advantages for individual politi- remove barriers that favor some groups and indi- cians include opportunities to make statements viduals in the electorate, and in this view, online about different events, add interpretations, or platforms can provide a means by which people make comments (Gainous & Wagner, 2013). and politicians gain more opportunities for inter- Furthermore, politicians might have other advan- action (Barber, 2003; Hagen & Mayer, 2000). In tages in addition to accessibility through social contrast, research has shown that use of the media for individual campaigning; experimental Internet and new technology is being normalized 316 L. A. C. SANDBERG AND P. ÖHBERG into the current electoral paradigm to be used as a similar way, be an indication that social media upsets part of the conventional political structure traditional campaign dynamics or breaks traditional (Bimber & Davis, 2003; Hindman, 2008; Ward, patterns of gender representation. Thus, gender repre- Gibson, & Lusoli, 2003). Normalization theory sentation can also be thought of as a function of states that actors who already have the most influ- equalization. Flowing from the same line of reasoning, ence will be most likely to benefit from new online differences in Internet-based campaigning that show campaign methods (Margolis & Resnick, 2000). male advantages and overrepresentation would sub- Parties that have more resources would, in this sequently indicate a normalized pattern. Perceived view, afford more professional campaigns, and male bias during elections would, from this perspec- social media would be used as just another chan- tive, result in women candidates using the platform nel, causing no disruption to traditional patterns over which they have the most control. If women are in off-line politics (Lilleker et al., 2011). perceived as being portrayed disadvantageously by Overall, the literature suggests that different editorial media, this would also speak in favor of functions of normalization apply for patterns of using alternative campaign strategies. In a relatively Web site adoption as compared to social media, gender-equal context such as Sweden, equalization in which offers some evidence to refute normaliza- terms of new campaign possibilities for women can- tion (Southern, 2015). Minor parties, for instance, didates can be seen as a way for women to avoid appear to be more effective in gaining votes gendered perceptions that continue to persist. A lack through social media than through the adoption of difference in social media use would, in this con- of personal Web sites (Gibson & McAllister, 2015). text, suggest that men and women are campaigning It is possible that given time, the differences in under relatively similar terms. social media use compared to other Web platforms Research on Web-based campaigning and will even out. It could be argued, however, that candidates’ use of social media has, to a very specific attributes of social media would be more limited extent, examined the role or function of beneficial for minor parties or lesser known can- gender. The literature concerning gender and didates. One difference, and perhaps advantage, representation has likewise not examined would be the interactive attributes of social whetherandhowgenderhasanimpacton media and the possibilities of actively building a online practices and strategies. This article pro- personal platform through which candidates could vides a first attempt to test the perseverance for more efficiently reach out to voters and media, as gendered campaign strategies online by study- opposed to Web pages and similar platforms. ing a context where the conditions for cam- Although the definition of normalization as a paigning are relatively equal between male and concept tends to lack specificity (Southern, 2015), female candidates. Below, we will develop our party size, campaign budget, and incumbency discussion of Sweden as a case, in the context often serve as measures to indicate whether poten- of the European elections. Prior to that, the tially weaker parties or candidates can gain favor literature on gender and campaigning will be through Web-based campaigning (Vaccari, 2013). discussed, as well as literature concerning can- Lusoli (2005) suggested that the influences of didates’ useofonlineplatformsandsocial Internet use on women candidates’ chances of media in particular. electoral success, or lack thereof, should also be thought of as functions of normalization. The same logic applies to social media use by under- Gender and campaign practices represented or politically marginalized groups, There remain substantial gender differences in such as women and ethnic minorities (Jacobs & modern democracies pertaining to representation, Spierings, 2016). although women generally perform as well as their As the term “equalization” in this context often male counterparts in similar types of electoral refers to social media giving opportunities to minor races (Darcy, Welch, & Clark, 1994). Recent parties or lesser known candidates, the new opportu- research has found little evidence to support pre- nities it could provide to female candidates would, in a vious claims that voters employ gender stereotypes JOURNAL OF INFORMATION TECHNOLOGY & POLITICS 317 when they evaluate women candidates (Dolan & Larsson & Kalsnes, 2014; Lassen & Brown, 2011; Lynch, 2016). Furthermore, no financial disadvan- Strandberg, 2013; Vergeer, Hermans, & Sams, tages appear to constrain women candidates in 2011). Existing research indicates that younger particular (Hogan, 2007; Werner, 1997). A great candidates belonging to minor parties and coming issue involved in increasing the positions of from challenger positions are more likely to use women to elected office remains the significant social media. lack of female candidates. In general, gender Some studies conducted in the early years of affects the decisions of women and men to run online campaigning using Web sites indicated for office based on their personal perceptions of that gender would generate similar campaign capability. In the United States, for instance, approaches online (e.g., Carlson, 2007), whereas women are more likely to doubt their abilities to others did not. Women candidates have, for engage in the campaign process (Fox & Lawless, instance, been found to be more likely to use 2011). Although research has shown that women blogs in their campaign and also to utilize a are equally supported in elections when they run, broader range of blogging practices, such as inter- this is not necessarily because they are treated active features (Carlson, Djupsund, & Strandberg, equitably by the electorate. Instead, it could 2014). The role of gender in relation to uptake of depend on their political quality; that is, women social media in particular has proven to be impor- who do run must be even better candidates to be tant in some studies. Gender had an impact in a equally competitive with male contenders (Fulton, study by Jackson and Lilleker (2011), which found 2012; Pearson & McGhee, 2013). that female British MPs were overrepresented on Gender differences related to campaign behavior, Twitter, compared to their numbers in Parliament. in particular, have largely focused on the different A study of Twitter use by the House candidates in issues addressed during the campaign. In previous the 2012 campaign also found that women were studies, candidates’ genders were, to some extent, more likely to use Twitter (Evans, Cordova, & related to the issues on which they focused (e.g., Sipole, 2014). In contrast, gender has proven to Dabelko & Herrnson, 1997; Larson, 2001). Later be a statistically nonsignificant factor for online research does not, however, offer any support that presence and adoption by candidates running in female candidates would focus on different issues the European Parliament elections or by the mem- than would male candidates, or that a particular set bers of European Parliament (Larsson, 2015a; of gender-stereotyped issues, such as social issues, Vergeer et al., 2013). exists (Dolan, 2005). There appear to be few gender Few studies look beyond activity level to variation differences persistent in overall campaign behavior in actual use, but there are some exceptions. Research and the use of varying campaigning tools. For exam- conducted in the United States shows that female ple, similarities tended to outweigh the differences politicians use Twitter more intensely than do male between male and female candidates in the way they politicians, have significantly more followers, and dif- are presented in television advertisements (Sapiro, fer in their manners of using social media, in terms of Walsh, Strach, & Hennings, 2011). Much less is content of messages and issues covered. Women were known, however, concerning the extent to which more likely to use “attack tweets,” andusedcampaign male and female candidates use different communi- issues or mobilization tweets more often. Male candi- cation strategies during their campaign or differ in dates were slightly more likely to use personal tweets. motives for using those strategies. Nonetheless, it is Overall, the findings suggest that women use Twitter during the campaign that candidates can choose how more actively as a campaign tool (Evans et al., 2014). to present themselves and their political agenda, as In contrast to these findings, Just and Crigler (2014) well as how to communicate with voters, which later reportedalackofgenderdifferenceinrelationto forms the basis of their representation. some of the above measures. Candidates tended to For the overall uptake of online campaigning by be as active, focus on similar topics, and use similar candidates, important factors include party size, tone. They did, however, find significant gender dif- competitiveness, challenger versus incumbent ferences in the social aspects of social media use, in the position, and age (Gulati & Williams, 2013; way that women connected more positively with 318 L. A. C. SANDBERG AND P. ÖHBERG followers. In an analysis of state legislators on Twitter, less dependent on the party and the campaign orga- party and gender effects interacted; specifically, non- nization. This can be especially rewarding for less- Republican women mentioned family somewhat established (female) candidates. The European (sec- more often than did non-Republican men, but within ond-order) election may therefore provide more the group of Republican legislators, women were variation in personal campaign strategies across par- much less likely to talk about family (Cook, 2016). ties and candidates than do (first-order) national Although research concerning the effect of can- elections, where individual candidates’ campaigns didate gender is inconclusive, gender does appear are more dependent on party organization. Lesser to be an important variable in politicians’ use of known candidates would also struggle more to gain social media platforms. It would be premature to traditional media attention. declare that gender is irrelevant in newer forms of online campaigning, but it is also a demanding Sweden as a case task to study becausw social media, such as Twitter, is a moving target, and observed patterns In this study, we see Sweden as a critical case in one election could look very different in the because female politicians are nearly as prevalent next. Furthermore, the different national electoral as male politicians in representative assemblies. systems matter for representatives’ social media For more than 20 years, female politicians have use in Europe (Obholzer & Daniel, 2016). Below, made up at least 40% of all three elected levels of we will discuss the European elections as a parti- representative assemblies (local, regional, and cular case for studying online campaigning, as national). In government, there has been equal compared to national elections. representation of men and women since 1994 (Krook, 2006). The importance of gender-related issues is also salient in the Swedish debate European parliament elections (Wängnerud, 2015). In a comparative European Previous research controlling for gender when perspective, Sweden has an equal distribution of measuring uptake and level of activity on social representatives in the European Parliament (out of media has mainly looked at politicians who were 20 seats, 11 went to female legislators). Moreover, already holding office. The outcome may be dif- in the 2014 election, a pronounced feminist party, ferent for challenger candidates, particularly in Feminist Initiative, made it into the European European elections, in which fewer established Parliament with one seat. candidates run compared to national elections. Sweden has a fairly typical party-centered Women are in general more successful in the European political structure, in which the members European election compared to national elections, of local and national assemblies are loyal party repre- because national political parties view these elec- sentatives (Gilljam, Karlsson, & Sundell, 2010). tions as having less at stake and thus are willing to Sweden uses a list system, whereby candidates are nominate more women or other, less traditional placed on the list according to party selectors, and candidates (Freedman, 2002; Harrison, 2005). the higher they are on the list, the greater their The logic of political competition varies across chances of being elected (Pierre, 2015). The candi- national and European elections, with the latter dates do have an incentive to cultivate personal votes, often referred to as second-order national elections however, as a candidate who receives 5% of the party’s (Schmitt, 2009). Although the positive effects of this votes can “break” the list. Nonetheless, this semi-open second-order setting for small parties are well list system does not change much. For example, in the known, the implications for independent candidates 2014 election, only one candidate became an MP due have rarely been discussed (for an exception, see to personal votes (Berg & Oscarsson, 2015). The poli- Ehin & Solvak, 2012).Thereareanumberofreasons, tical parties are therefore central actors in the success however, why independent or nonincumbent candi- of an aspiring candidate. We know from previous dates may benefit from running in European research that in strong party systems, the party leader- Parliament elections. For example, in this electoral ship stands a better chance of ensuring that their context, the individual candidates’ campaigns are representatives share the party’s preferences than JOURNAL OF INFORMATION TECHNOLOGY & POLITICS 319 they do in more candidate-oriented systems (Müller, (3) Will women politicians behave differently 2000). The leeway for individual candidates to culti- than male politicians on Twitter by being vate their own relationships with voters is therefore more active and communicating more more restricted. interactively? Taken together, the high numbers of women politicians in the Swedish system and their activ- Data material and methods ities within a strong party system, in which This study used the Swedish candidate survey of individual candidates’ actions are centered on candidates running for the 2014 European elec- party interests, make gendered findings less tion. The survey contained questions regarding plausible. Nonetheless, gendered perceptions the importance of different campaign activities, can be expected to be preserved even when motives for Twitter use, and evaluation of the representation is equal in numbers. For instance, candidates’ use of Twitter. Previous research has womeninthelegislaturetendtospeaklesson found two main styles of social media campaign- “harder” policyissuesthandomeninSweden ing: a party-centered and an individualized style (Bäck, Debus, & Müller, 2014). Furthermore, a (Karlsen & Enjolras, 2016). Female politicians in recent study on Swedish municipal politicians general are more concerned about representing the revealed that gender inequalities continue to party than representing themselves, when com- obstruct women from political power, due to pared to men (Scully & Farrell, 2003). In the the so-called “glass ceiling effect” (Folke & survey, we therefore asked questions concerning Rickne, 2016). The gender-related patterns we the dimension of a personal versus party-centered may discover in this study should be relevant communication style. The questions on motives in contexts where female politicians are less also covered other aspects of uses for political common and where individual politicians have purposes, such as whether the primary audience more pronounced roles to play during the elec- consisted of media, politicians, or voters. In the tion campaign. Women should, according to this survey, we asked how important Twitter was in argument,beevenmoremotivatedtousesocial terms of the following: communicating the politi- media in other modern democracies with less cian’s individual daily activities, sharing informa- female representation. tion about the party’s daily activities, displaying a Studying gender-based differences in campaign more personal side, contacting voters personally, behavior in Sweden can improve our understand- highlighting personal views, highlighting the ing of whether increased gender equality in repre- party’s position, debating with political opponents, sentation cultivates new behavior, to the extent criticizing political opponents, reaching as many that male and female politicians behave fairly simi- voters as possible, and making an impact in tradi- larly, or if gendered campaigning is something tional media (see questionnaire in Appendix 4). more fundamental, and typical male and female In terms of gender and online campaigning, behavior among political candidates continues to social media can potentially offer both opportu- exist even in gender-equal contexts. Against the nities and challenges for female candidates. backdrop of a development toward personalized Because women are more likely to be exposed to campaigning and the interactive features brought the risk of online harassment and trolling forward by social media, we address the following (Megarry, 2014), this could, for instance, hamper three questions: the use of social media platforms for female candidates. We therefore also asked the candidates (1) Do women candidates consider social media why they did not use social media, and if the risk to be a more important campaign tool than of harassment would be one reason not to do so. do male candidates? To analyze the candidates’ actual Twitter use, (2) Are female politicians more likely to use we collected data on Twitter activity during three Twitter to enhance the political messages time periods for candidates with Twitter accounts. of their party or their personal agendas? These time periods included the campaign period 320 L. A. C. SANDBERG AND P. ÖHBERG

(April–May 2014), a pre-campaign period elections of 2014, party belonging was important. (February 2013) and a period after the election There was a higher likelihood that candidates belong- (January–March 2015). From the Twitter ing to mainstream parties would use Twitter as accounts, we collected the number of tweets sent, opposed to more extreme party candidates, on an number of @ mentions, retweets, number of tweets economic left–right scale (Nulty, Theocharis, Popa, with hashtags (#), and number of tweets contain- Parnet, & Benoit, 2016).Themoreextremecandidates ing links. are on the Green-Alternative-Libertarian vs The questionnaire, a Web survey, was sent out Traditional-Authoritarian-Nationalist (GAL/TAN) directly after the 2014 election to all 315 candi- dimension, however, the more often they use dates from the main political parties. The response Twitter (Obholzer & Daniel, 2016). In Sweden, the rate was 46% (see Appendix 3 for response rate per Feminist Initiative party and the Sweden Democrats party, age group, and gender in the survey). In the are on the opposite ends of the GAL/TAN scale. Based models, we controlled for age, incumbency, place- on a study of the parties’ official social media ment on the party list, political party, and seniority accounts, the Feminist Initiative party was among (whether or not they had been members of the the most active, whereas the Sweden Democrats, on national parliament). These variables had all the contrary, were among the least active (Larsson, proved to be important in various previous stu- 2015b). For this reason, we also controlled for party dies, although their importance was not entirely affiliation. consistent throughout the different contexts. For social media uptake and use, age was the one Dimensions in campaign activities factor that stood out as being most significant (Larsson, 2015a; Lassen & Brown, 2011; Vergeer The empirical section begins with a description of & Hermans, 2013). Electoral vulnerability had no how the candidates assessed the importance of social effect on uptake according to Lassen and Brown media platforms, such as Facebook and Twitter, in (2011), and Strandberg (2013) reported incum- relation to other campaign activities. The survey cov- bency to be positively correlated with social ered 18 campaign activities, and the candidates media use. In many other cases, challenger candi- ranked these from unimportant to very important dates employed social media to a larger extent on a 5-point, Likert-type scale. Respondents ranked (Druckman, Kifer, & Parkin, 2007; Gibson, Facebook as the third most important activity; 84% of Lusoli, & Ward, 2008; Jackson & Lilleker, 2009). the candidates used it, 67% of which indicated that it According to a later study of the European 2014 was of importance to their campaign (Figure 1). election by Obholzer and Daniel (2016), MEPs at Respondents ranked Twitter as the sixth most impor- risk of losing their seats are more active on tant, with 65% of candidates using it, 38% of which Twitter. Moreover, previous studies have shown reported it to be important. Other Web-related activ- that seniority has an impact on activity on ities, such as individual blogs and Web pages, were Twitter. Senior MPs are in general more active seen as less important compared to social media users (Jackson & Lilleker, 2011). channels. All in all, although social media platforms Party size is sometimes used as a measure of how were not the most important campaign aspects, resources influence uptake of social media; the larger Facebook and Twitter were within the upper third the party, the more resources are available. For exam- of the most important campaign activities. ple, Lilleker and Koc-Michalska (2013)foundthat To study the relationship between the importance MEPs from minor national parties perform better of social media and gender, we employed a factor for different categories of online campaigning. In analysis (principal-component). The results yielded addition to party size, the ideological nature of the five dimensions of campaign activities (see Appendix parties must be considered if we want to explain 1). Web campaign (i.e., Twitter, Facebook, blogs, and uptake and use of social media. Party ideology was Web pages) emerged as a distinct factor (Cronbach’s unrelated to adoption of Twitter as a campaign tool alpha = 0.73). These activities were all related to the according to Vergeer and Hermans (2013); however, candidates’ online presences. Other factors were in a more recent study of candidates in the EU related to what we termed either traditional JOURNAL OF INFORMATION TECHNOLOGY & POLITICS 321

Figure 1. Mean importance of campaign activities to candidates in the 2014 European election. Note. The scale runs from 1 (unimportant) to 5 (very important). Standard deviation within parentheses. campaign activities (spreading the party’scampaign womeninlessequalcontextsarereluctanttostep material, meeting with party groups, contacting forward as candidates; in other words, female candi- media, and making public speeches; Cronbach’s datesassessthatanelectoralcampaigndemands alpha = 0.70), outreaching activities (calling voters, more of them than it does of men. visiting workplaces and organizations, or communi- Among the individual control variables, we can- cating through individual e-mail lists or SMS; not see that there exists the same type of consistent Cronbach’s alpha = 0.67), personal activities (visiting pattern as there was between male and female can- door-to-door, putting up individual election posters, didates, with the exception of age. Younger candi- or sending letters to voters; Cronbach’salpha=0.65) dates see Web campaigning as more important than and media activities (releasing individual advertise- do their older colleagues, whereas the opposite is ments in newspapers, radio, and television channels; true when it comes to traditional campaigning, cinemas; or individual election pamphlets; where older candidates stress its importance. It is Cronbach’s alpha = 0.67). worth mentioning that Web campaigning is the When analyzing these dimensions and their rela- only thing that the incumbents assessed as more tionships to gender, we found that women candidates important than did the other candidates, which is overall tended to place greater value on all campaign in line with Strandberg’s results. Strandberg (2013) activities than did male candidates (Table 1). Out of suggested that a reason for this could be that journal- five different campaign categories, female candidates ists’ use of social media as a news source would favor assessed three as significantly more important than more established candidates. did men. Among these were activities on the Web If we study the importance of party affiliation (mean for female candidates was 0.47, mean for men and the candidates’ assessments, it is evident that was 0.36).1 It is of interest that female candidates Sweden Democrat candidates assessed all cam- assessed more activities as important. One of the paign activities as less important than did candi- reasons why female candidates see more activities dates from other parties. We cannot offer a solid as important may be related to the fact that women explanation for this finding. The Sweden think that more activities are necessary to pursue a Democrats had their best election results ever, so political career (Fox & Lawless, 2011). This tendency it would be natural to assume that they would be to stress that more things are important, even in a very optimistic about the usefulness of their activ- gender-equal context, could partly explain why ities, but that is not the case. It remains plausible 322 L. A. C. SANDBERG AND P. ÖHBERG

Table 1. Assessment of the Importance of Campaign Activities by Candidates in the 2014 European Election. Web campaign Traditional Outreach Personal Media Sex (women) 0.11** 0.08* −0.06 0.09** 0.07 (0.05) (0.04) (0.13) (0.04) (0.04) Placement on the party list (1–43) −0.00 −0.00*** −0.01 −0.00 −0.00 (0.00) (0.00) (0.01) (0.00) (0.00) Age (22–83) −0.00** 0.00** 0.00 −0.00 −0.00 (0.00) (0.00) (0.00) (0.00) (0.00) Incumbent 0.22** −0.03 −0.03 0.12 0.02 (0.11) (0.11) (0.31) (0.09) (0.10) Experience as MP −0.02 −0.08 −0.04 −0.04 −0.07 (0.07) (0.06) (0.20) (0.06) (0.07) Party (Social Democrats are ref. category) Center Party 0.03 −0.05 −0.39*** −0.44*** −0.08 (0.09) (0.08) (0.25) (0.07) (0.08) Christian Democrats 0.03 −0.05 −0.33 −0.43*** −0.13 (0.09) (0.09) (0.25) (0.07) (0.08) Conservative Party −0.01 0.02 −0.21 −0.06 0.00 (0.10) (0.09) (0.30) (0.09) (0.09) Feminist Party −0.01 −0.30** −0.81* −0.66*** −0.35*** (0.15) (0.13) (0.41) (0.12) (0.13) Green Party 0.03 −0.12 −0.43* −0.51*** −0.22*** (0.08) (0.08) (0.24) (0.07) (0.08) Left Party −0.18** −0.06 −0.86*** −0.53*** −0.36*** (0.09) (0.08) (0.25) (0.07) (0.08) Liberal Party 0.12 −0.03 −0.26 −0.39*** 0.05 (0.10) (0.09) (0.28) (0.08) (0.09) Pirate Party −0.02 −0.21** −0.87*** −0.53*** −0.30*** (0.11) (0.10) (0.30) (0.09) (0.10) Sweden Democrats −0.38*** −0.38*** −1.14*** −0.55*** −0.36*** (0.11) (0.11) (0.33) (0.10) (0.11) Intercept 0.57*** 0.63*** 1.18*** 0.65*** 0.39*** (0.10) (0.10) (0.30) (0.09) (0.09) R2 0.24 0.26 0.18 0.49 0.29 N 142 142 137 143 144 Notes. Linear regression (OLS). All dependent variables run between 0–1. Standard errors in parentheses. ***p < 0.01. **p < 0.05. *p < 0.1.

that the attitude stems from their status as an collapse the four variables into one, there is a sig- antiestablishment party, and that the candidates nificant gender difference). The only other signifi- feel stigmatized by the editorial mainstream cant gender difference was found for Facebook, media and the parliament’s current cordon sani- which was in fact even more important for women; taire around the party. This appears to also affect 76% of the female candidates saw Facebook as rather how they view social media, something we will or very important, whereas 58% of male candidates discuss further below. felt the same. In looking at each of the activities related to social For the control variables, list placement had no media and Web-based campaigning, there was a effect on any campaign activity. Age was significant significant gender difference for Twitter, where for Twitter use (as supported by previous research); women assessed Twitter as more important than older candidates were less optimistic about its use- did men (Table 2.). Nearly half of the women fulness. Incumbents assessed Twitter as more valu- (49%) assessed Twitter as “rather important” or able than did the other candidates. Among the party “very important,” whereas among the men, the cor- variables, the two most interesting significant and responding number was only 28%. We can also see nonsignificant results were from the Sweden the same pattern emerging as in Table 1. The ten- Democrats and the Feminist Initiative. During the dency for female candidates to assess all activities as election campaign, the Sweden Democrat candidates more important than did men is present (if we and their activities on social media were scrutinized JOURNAL OF INFORMATION TECHNOLOGY & POLITICS 323 by the media. Some of these candidates even had to Table 2. Assessment of the Importance of Web Campaign resign as a result of this public scrutiny. It is therefore Activities by Candidates in the 2014 European Election. Home interesting to note that candidates from the Sweden Twitter Facebook Blog page Democrats were slightly more skeptical about the Sex (women) 0.14** 0.11** 0.08 0.09 usefulness of social media. In contrast, the Feminist (0.06) (0.05) (0.06) (0.06) Placement on the party list −0.00 −0.00 0.00 −0.00 Initiative candidates based their campaigns to a large (1–43) extent on social media (Blombäck & De Fine Licht, (0.00) (0.00) (0.00) (0.00) 2015) and were the most active on various social Age (22–83) −0.01*** −0.00 −0.00 −0.00 (0.00) (0.00) (0.00) (0.00) media platforms. The Feminist Initiative party also Incumbent 0.42*** −0.05 0.26 0.28* had more followers than any of the other parties, but (0.15) (0.13) (0.16) (0.16) Experience as MP 0.04 −0.16* −0.07 0.06 Feminist Initiative respondents did not assess (0.10) (0.08) (0.10) (0.10) Twitter and Facebook as crucial tools for their suc- Party (Social Democrats are cess. The results here do not support previous ref. category) Center Party 0.04 −0.19* 0.14 0.09 research in which candidates from more extreme (0.12) (0.11) (0.13) (0.12) parties on the GAL/TAN scale used social media Christian Democrats −0.05 −0.05 0.17 0.02 (0.12) (0.10) (0.13) (0.12) more extensively. Conservative Party 0.06 −0.16 0.07 −0.01 It has been theorized that social media not (0.14) (0.12) (0.15) (0.15) only allow disadvantaged parties and candidates Feminist Party 0.10 −0.03 −0.19 0.07 (0.20) (0.18) (0.21) (0.21) to bypass traditional media, but may also serve Green Party −0.04 −0.21** 0.21 0.16 populist parties as they strive to appeal to “com- (0.13) (0.10) (0.12) (0.12) − − − mon people” by offering direct contact opportu- Left Party 0.15 0.40*** 0.02 0.20 (0.12) (0.10) (0.12) (0.12) nities for this aim (Jacobs & Spierings, 2016). Liberal Party 0.15 −0.05 0.29* 0.21 Conversely, populist radical right parties such (0.13) (0.12) (0.14) (0.14) Pirate Party −0.08 −0.35** 0.26* 0.09 as the Sweden Democrats are often organized (0.14) (0.13) (0.15) (0.15) around strong leaders who personify the party Sweden Democrats −0.36** −0.79*** −0.11 −0.23 and seek to control the party as a whole. As (0.16) (0.14) (0.17) (0.17) Intercept 0.81*** 0.97 0.20 0.30** these parties are more centralized, this control (0.14) (0.12) (0.15) (0.14) decreases as a result of social media (Jacobs & R2 0.25 0.32 0.12 0.17 N Spierings, 2016). The less motivated candidates 144 148 147 147 Notes. Linear regression (OLS). All dependent variables run between in the Sweden Democrats could therefore, in 0–1. Standard errors in parentheses. addition to being affected by tough media cover- ***p < 0.01. **p < 0.05. *p < 0.1. age, also be the result of a more centralized control over candidate behavior.2 In summary, the results shown in Table 1 and the purposes of (a) displaying a relatable personal Table 2 help us to answer our first question, sug- side of themselves, (b) getting in personal contact gesting that women candidates consider social with voters and (c) making their personal views media to be a more important campaign tool visible (Cronbach’s alpha = 0.85). To measure than do male candidates. There is a tendency, Twitter’s usefulness to party-centered campaign however, for women candidates in general to approaches, the candidates were asked about the assess all campaign activities as more important. extent to which their Twitter usage was important for (a) spreading information about the party’sdaily activities and (b) making the party’s views visible Motives for Twitter use and evaluation (Cronbach’s alpha = 0.81). The different sets of ques- In the survey, we asked whether the candidates tions were merged into party-centered and personal- prioritized using Twitter for individual or party-cen- centered categories. In the questionnaire, the scale tered purposes. To measure Twitter’s usefulness for runs from 1 “to a very small extent” to 7 “to a very individual campaigning practices, we constructed an large extent”; in the analyses, however, the items are index based on three questions that asked the candi- merged into two indices that range between 0 “to a dates to assess the importance of using Twitter for very small extent” and 1 “to a very large extent.” 324 L. A. C. SANDBERG AND P. ÖHBERG

Table 3. Assessment of the Importance of Twitter for Personal When women candidates evaluated their party- or Party-Centered Purposes and Evaluation of the Use of Twitter centered uses of Twitter, they were also more posi- for These Purposes in the Candidates’ Campaign. Twitter motivation Twitter evaluation tive than male candidates (Table 3). The results Personal Party Personal Party indicate that women candidates use Twitter as a Sex (women) 0.12** 0.18*** 0.05 0.13* professional campaign tool more often than do (0.06) (0.06) (0.06) (0.07) men, and these findings strengthen the picture that Placement on the party list −0.00 −0.00 −0.00 0.00 (1–43) emerged above: women candidates value Twitter as (0.00) (0.00) (0.00) (0.00) more important in their campaign efforts than do Age (22–83) −0.00 0.00 −0.00 −0.00 male candidates, as a way to reach out to voters with (0.00) (0.00) (0.00) (0.00) Incumbent 0.10 −0.11 0.20 0.10 the party message (mean for women was 0.53, mean (0.13) (0.14) (0.13) (0.15) for men was 0.40). In their evaluation of how well Experience as MP −0.08 −0.19* 0.07 −0.03 (0.13) (0.10) (0.10) (0.11) they performed with their personal usage of Twitter, Party (Social Democrats are however, women were not significantly more satis- ref. category) fied than were men (mean for women was 0.45, Center Party 0.18 0.02 0.16 −0.03 (0.12) (0.13) (0.13) (0.13) mean for men was 0.40). Because the number of Christian Democrats 0.14 0.15 0.16 0.06 respondents was limited, it is difficult to further (0.12) (0.13) (0.13) (0.14) Conservative Party 0.07 0.01 0.03 −0.18 investigate the difference between how male and (0.12) (0.13) (0.13) (0.14) female candidates evaluate their success in bringing Feminist Party −0.01 0.02 0.14 0.20 forward their personalized campaign. We can only (0.18) (0.20) (0.19) (0.20) Green Party 0.02 −0.02 0.04 −0.21 observe that the difference between male and female (0.11) (0.12) (0.12) (0.12) candidates’ motivations for using Twitter was 0.12, Left Party −0.28** 0.05 −0.19 −0.22 (0.11) (0.12) (0.13) (0.13) whereas when they assessed their success of forward- Liberal Party 0.12 −0.03 −0.04 −0.19 ing their personalized campaign, the difference (0.12) (0.13) (0.13) (0.14) remained at 0.5. This could be another indication Pirate Party −0.16 −0.03 −0.19 −0.20 (0.13) (0.15) (0.14) (0.15) that women put more pressure on themselves on the Sweden Democrats −0.06 0.01 0.08 −0.04 campaign trail. (0.21) (0.23) (0.22) (0.24) To answer our second question of whether or Intercept 0.58*** 0.54*** 0.52*** 0.53*** (0.14) (0.15) (0.15) (0.16) not female politicians are more likely to use Observations 92 91 88 88 Twitter to enhance the political messages of their R-squared 0.28 0.19 0.24 0.22 party or their personal agenda, we have seen that Notes. All dependent variables run between 0–1. Standard errors in parentheses. female candidates actually work harder than men ***p < 0.01. **p < 0.05. *p < 0.1. to promote both themselves and their parties. We understand this tendency to be an indication of gendered campaigning. There was, in fact, a significant difference between The presence of incivility can have hampering male and female candidates in terms of motivations effects for online political discourse (Gervais, forusingTwittertoexposethemselvesascandidates 2015), yet it was not an issue reported by the (Table 3).Themeanforwomencandidateswas0.63 candidates in our survey. When we asked the and 0.51 for men. At the same time, more female than respondents who did not use Twitter about their male candidates stated that they used Twitter to draw reasons for refraining from doing so, only one attentiontotheirparties(meanforwomenwas0.68, male and one female candidate stated that they mean for men was 0.50). This result is again in line do not use Twitter because they “do not want to with what we found earlier. Women candidates stress be exposed to personal harassment.” more aspects as being important in their campaign To evaluate the candidates’ actual usage of than do men, and as a consequence, Twitter is more Twitter (see Table 4), we collected tweets from all important for women, both personally and also for the Swedish candidates active on Twitter during the their party-oriented purpose. The control variables 2014 election campaign (between March 25 and did not offer much variation. In fact, gender was the May 25). To detect any special campaigning effects, only consistent pattern for the party-centered model. we also collected tweets from the candidates one JOURNAL OF INFORMATION TECHNOLOGY & POLITICS 325

Table 4. Activity on Twitter During and After the Election Campaign (NBREG). 2014 2014 2014 2015 2015 2015 Number of Twitter Twitter Number of Twitter Twitter tweets interactions information tweets interactions information Sex (women) 0.15 0.13 0.22 0.47** 0.46** 0.26 (0.18) (0.21) (0.19) (0.20) (0.23) (0.21) Placement on the party list (1–43) −0.02* −0.02* −0.01* −0.02* −0.02 −0.01 (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Age (22–83) −0.02** −0.02** −0.01* −0.03*** −0.03*** −0.02** (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Incumbent 0.98** 1.03** 1.19*** 0.96** 1.06 1.23 (0.38) (0.44) (0.40) (0.47) (0.54) (0.47) Experience as MP 0.60** 0.82*** 0.37 0.35 0.42 0.30 (0.26) (0.31) (0.27) (0.29) (0.33) (0.29) Party (Social Democrats are ref. category) Center Party 0.40 0.24 0.25 1.39*** 1.52** 0.99* (0.36) (0.43) (0.36) (0.45) (0.51) (0.47) Christian Democrats 0.08 −0.17 0.12 0.64 0.74 0.32 (0.35) (0.39) (0.35) (0.45) (0.51) (0.45) Conservative Party 0.25 0.08 0.34 −0.05 −0.21 −0.02 (0.36) (0.42) (0.36) (0.44) (0.50) (0.45) Feminist Party 0.97* 0.65 0.11 −0.71 −0.77 −0.86 (0.55) (0.59) (0.55) (0.68) (0.78) (0.68) Green Party 0.49 0.47 0.50 0.35 0.52 0.21 (0.33) (0.39) (0.34) (0.43) (0.48) (0.43) Left Party 0.59 0.47 0.49 1.06** 1.25* 0.81* (0.37) (0.44) (0.38) (0.47) (0.53) (0.47) Liberal Party 0.21 0.17 0.27 0.86* 1.22** 0.58 (0.34) (0.39) (0.35) (0.44) (0.50) (0.44) Pirate Party 0.77* 0.55 0.81** 0.83 1.12* 0.64 (0.41) (0.47) (0.41) (0.51) (0.57) (0.52) Sweden Democrats −0.57 −0.20 −0.39 −0.18 0.08 −0.42 (0.49) (0.56) (0.51) (0.57) (0.64) (0.58) Intercept 5.62*** 5.11*** 5.25*** 5.75*** 5.10*** 5.25*** (0.44) (0.51) (0.45) (0.52) (0.58) (0.53) Pseudo R2 0.01 0,01 0.01 0.02 0.02 0.01 N 196 196 196 195 195 195 Note. Standard error in parentheses. ***p < 0.01. **p < 0.05. *p < 0.1. year before the election (February 2013) and again disaggregate links, hashtags, retweets, and @ after the election (January to March 2015). In order replies. Examining the contents of tweets and a to explore differences in actual use and activity possible gendered communication style would also between male and female candidates, we collected be of interest. In this article, we restricted our scope the number of tweets and the use of retweets, men- to a variable for the total number of tweets that tions, hashtags, and links. Both retweeting and candidates sent out (in addition to forms of use). mentioning can indicate a more interactive style Because our measurements consist of count data of use, whereas the use of links and hashtags indi- with the existence of overdispersion, we used nega- cates intention to contribute to a debate on an issue tive binomial regression in our models.3 at hand or provide further information on topics There were no gender differences for either such as campaign issues. To separate the different numbers of tweets or Twitter activity during the uses of Twitter, we created two variables: Twitter election campaign (Table 4). Although there was a interactions (consisting of retweets and mentions) tendency for women to be more active, the differ- and Twitter information (consisting of links and ence was not significant. There were, however, hashtags). Future studies into communication prac- some significant effects in the model. Older candi- tices, however, may benefit by taking a more gran- dates were less active on Twitter, tweeting, inter- ular approach to different uses based on acting, and sharing information to a lesser extent 326 L. A. C. SANDBERG AND P. ÖHBERG than did younger candidates. The same tendency Conclusions was evident among candidates who were lower In this study, we examined the importance of ranked on the party list. We correspondingly gender in how political candidates use social observed that incumbents were more active on media for campaigning purposes. Using survey Twitter in comparison to the other candidates; data on Swedish candidates and behavioral data here, it also appears that seniority matters. from Twitter during the 2014 European election, Candidates who had been members of the national we found gender-related patterns in the level of parliament used Twitter to a higher extent. It is value candidates placed on social media as a cam- difficult to speculate why, but one explanation paign tool. Women candidates tended to evaluate could be related to party loyalty, as when the social media as more important and more useful election was over and a year had passed, the than did male candidates, particularly in terms of more senior candidates were no longer more party-centered campaign approaches. During the active on Twitter. election campaign, however, women candidates The effects of election campaigning were further did not use Twitter more or differently than evident for other candidates’ activities a year later. men. The logic of election campaigning makes Female candidates were more active on Twitter and candidates act in more predictable ways. The high- also had more interactions with their followers. The est-ranked candidates and the incumbents were difference between male and female candidates who generally the most active; however, we observed shared Twitter information was nearly significant, that more senior candidates became more active as well. It appears that gender is relevant, but not during the election campaign, but lost their inter- when it is election time. The competition between est after the election was over. Between the elec- the candidates makes gender less important during tions, we could also see gender patterns for how the election campaign than it is before or after the men and women interacted using Twitter. These campaign. When the competition is over, however, patterns have implications for the interpretation of women candidates are more interested than men in so-called permanent campaigning (see Farrell, communicating with their followers. This result was 1996) and the logic of online campaigning. not driven by the intensity during the election During the intense campaign period, the level of campaign. In the 2014 election campaign, the can- activity rose as can be expected, but from these didates tweeted nearly 38,500 times, and in the theoretical perspectives, the rationale for social three first months of 2015, they sent more than media usage appeared to yield unexpected differ- 40,500 tweets. To further validate the results, we ences between elections. A distinct normalized or looked at the candidates’ activities a year before the equalized pattern in online campaigning appears election (February 2013). Here, we found the same difficult to argue for and as this study shows, this pattern: women candidates were more active than especially holds true when considering the period men on Twitter (see Appendix 2). These results between elections. It would appear that the more underscore the relevance of studying what politi- experienced and routine candidates value social cians are doing between elections, not only what media to a greater extent, judging by their activity they are doing during campaigns (Esaiasson & and interactivity during the intense campaign per- Narud, 2013). iod. This indicates a normalized pattern emerging Among the other variables, we observed that on social media; however, the pattern was not higher ranked candidates used Twitter more consistent between elections, where we addition- often even when the election was over. Similarly, ally found that female candidates were more pro- younger candidates tweeted more and had more minent in their use. interactions than did their older colleagues. The normalization/equalization dichotomy pro- Incumbents were also more active. There were no vides a limited understanding of the patterns of larger discrepancies between the parties, however, social media use in this case. Indicators of equal- other than candidates from the Center party being ization, such as gender, offer inadequate explana- more active than the other party candidates after tory power in a relatively equal setting. In a the election. JOURNAL OF INFORMATION TECHNOLOGY & POLITICS 327 context of equal representation in parliament but therefore also study what the candidates are doing with existing gender differences in de facto politi- between elections. This is perhaps especially true for cal power, it is not unambiguous how a pattern of social media, and such research is necessary to more extensive social media use by women politi- further assess whether social media is used as a cians should best be understood. It could be professional tool for a more permanent online pre- argued that social media in this case provides sence or merely as a temporary additional platform more equal conditions but the results from this during electoral campaigns. use can still resemble a normalized pattern. It It has been suggested that, in their effort to sidestep would be valuable if future references to this rather traditional media, female candidates aim to project an dated theory provided more elaborate attempts to image of a more open, personal, and interactive poli- explain why social media represents a breaking tician (Carlson et al., 2014). The extent to which this point or continuation of previous conditions. couldbelinkedtoatypicalfemaleconversationalstyle This could be done by, for example, considering online may be relevant to study, given that some different electoral settings, and campaign vs. non- previous research indicates dissimilar gendered beha- campaign periods. As this article shows, the con- vior on social media and that women in general tend cept of normalization appears to hold true when to use technology differently than men (e.g., more looking at a period of intense campaigning, sociably; Cunha, Magno, Goncalves, Cambraia, & whereas the regular day-to-day basis of political Almeida, 2014; Lasorsa, 2012;Tufekci,2008;Walton communication on social media platforms indi- &Rice,2013). Given the results presented in this cates a break from this pattern. article, gender should be studied in greater detail in Research has shown that women and men are relation to the political use of social media platforms. similar in their general campaign styles (Bystrom & Sweden has one of the highest voter turnouts in Kaid, 2002), as well as in their overall online activity European elections and the gender distribution of in the recent European elections. Our results suggest candidates is one of the most equal; nevertheless, we that gender does play a role in social media cam- can detect gendered campaigning. These findings paigns, which calls for further comparative research should therefore be expected to be even more promi- in different electoral contexts and with varying nent in other contexts. In contrast to American and degrees of descriptive representation. This study British majority election and candidate-centered sys- provides insight into the evolution of campaign tems, Sweden’s pluralistic, party-centered parliamen- practices in a context with a relatively gender-equal tary system can provide different implications for representation, where candidates adopt and adapt to individual online campaigning practices. Research new media environments. In this regard, social has for example suggested that being online has a media is seen as an important platform for individual significant effect on intra-party competition (Hansen candidates, and its importance can be expected to & Kosiara-Pedersen, 2014). In contrast, the relation- increase over time. According to Lilleker, Tenscher, ship between gender and social media practices in this and Štětka (2015), indications of development case may not be as strong in the context of national toward a “hypermedia” campaign style can be elections, because the European election possibly pre- found in the integration and involvement of several sents other opportunities for less established candi- platforms as professional campaigning strategies in dates, such as younger female contenders. Europe. As the European campaign structure has Political candidates use of social media can developed toward campaign professionalism, this possibly yield additional preferential votes pattern can help to explain the emphasis put on (Spierings & Jacobs, 2014). A relevant issue to social media campaigning. Separating these practices address in future research could therefore be the as strategic choices in processes of professionaliza- extent to which a candidate’s individual charac- tion is necessary (Tenscher, 2013). In line with this teristics, such as gender, make that candidate reasoning, our results could also indicate that a more more or less likely to benefit from using social strategic use of social media entails a less active media as a campaign strategy. If women candi- presence outside the campaign period. To better dates succeed in establishing themselves online understand election campaigns online, we should and using social media to connect to voters— 328 L. A. C. SANDBERG AND P. ÖHBERG especially female voters, who are more likely to Career Ambition in Representative Democracy” was pub- assess female politicians positively (Kosiara- lished by Kansas University Press in 2017. Pedersen & Hansen, 2015)—this could in time strengthen the status of younger women candidates in the electoral system. References

Bäck, H., Debus, M., & Müller, J. (2014). Who takes the Notes parliamentary floor? The role of gender in speech-making 1. All the means in the article were calculated with the in the Swedish Riksdag. Political Research Quarterly, 67, – margins command in Stata 14.1, while holding other 504 518. doi:10.1177/1065912914525861 factors constant at their average values. Barber, B. (2001). The uncertainty of digital politics. Harvard – 2. In the party’s ethical guidelines from 2011, there are International Review, 23(1), 42 47. instructions for social media use that state that it is Barber, B. (2003). Strong democracy: Participatory politics for important to be honest and sincere, and treat people a new age. Berkeley: University of California Press. the same way on the Internet as in real life. Among Berg, L., & Oscarsson, H. (Eds). (2015). 20 år med personval other things, the guidelines state that candidates [20 years of preferential voting]. Valforskningsprogrammets should: not use social media when in an upset state rapportserie. Statsvetenskapliga institutionen, Göteborgs of mind, always think about the political conse- universitet. Göteborg. quences when publishing, avoid internal discussions Bimber, B., & Davis, R. (2003). Campaigning online: The – (the guidelines say this is best handled within the Internet in U.S. Elections. 387 397. New York, NY: party), and respect the law (pp. 5–6). These state- Oxford University Press. ments give a clue as to some of the trouble the Blombäck, S., & De Fine Licht, J. (2015). Feministiskt initi- party has had with their representatives on social ativs sympatisörer [The sympathizers of Feminist initia- media platforms. tive]. In A. Bergström, B. Johansson, H. Oscarsson, & M. 3. Because alpha was significantly different from zero, Oskarson (Eds.), Fragment. Göteborgs universitet: SOM- we preferred negative binomial regression over institutet. Göteborg. Poisson regression. Brewer, P. R., Habegger, M., Harrington, R., Hoffman, L. H., Jones, P. E., & Lambe, J. L. (2015). Interactivity between candidates and citizens on a social networking site: Effects Acknowledgments on perceptions and vote intentions. Journal of Experimental Political Science. doi:10.1017/XPS.2014.29 We are in great debt to Hampus Brynolf who helped us Bystrom, D., & Kaid, L. (2002). Are women candidates trans- collect all the Twitter data and made the study possible. In forming campaign communication: A comparison of addition we would like to thank Jonathan Polk at the Political advertising videostyles in the 1990s. In C. S. Rosenthal Science Department, University of Gothenburg, for invalu- (Ed.), Women transforming congress. Norman, OK: able comments and advice, as well as the seminar participants University of Oklahoma Press. from the Institute for Social Research, Oslo, who gave helpful Carlson, T. (2007). It’s a man’s world? Journal of Political feedback on the paper draft. Marketing, 6(1), 41–67. doi:10.1300/J199v06n01_03 Carlson, T., Djupsund, G., & Strandberg, K. (2014). Taking risks in social media campaigning: The early adoption of Notes on contributors blogging by candidates. Scandinavian Political Studies, 37, 21–40. doi:10.1111/scps.2014.37.issue-1 Linn A.C. Sandberg is a PhD candidate at the department of Cook, J. M. (2016). Gender, party, and presentation of Media and Communication at the University of Oslo. family in the social media profiles of 10 state legislatures. Sandberg is conducting research on how digitalization and Social Media + Society, 2(2), 1–11. doi:10.1177/ the growth of social media affect representative democracy 2056305116646394 and the democratic values that are traditionally viewed as Cunha, E., Magno, G., Goncalves, M. A., Cambraia, C., & providing legitimacy to political decisions. Almeida, V. (2014). He votes or she votes? Female and Patrik Öhberg is an associate professor in political science at male discursive strategies in Twitter political hashtags. Plos the University of Gothenburg. Öhberg has a wide research One, 9(1). doi:10.1371/journal.pone.0087041 interest in theories and practices of representative democracy. Dabelko, K. L. C., & Herrnson, P. S. (1997). Women’s and Specific interests include the perceptions and behaviors of men’ s campaigns for the U.S. house of representatives. political and administrative elites. Öhberg’s work is published Political Research Quarterly, 50(1), 121–135. in journals such as British Journal of Political Science, Darcy, R., Welch, S., & Clark, J. (1994). Women, elections, Comparative Political Studies, and West European Politics. and representation (2nd ed.). Lincoln: University of His latest book “Ambitious Politicians. The Implications of Nebraska Press. JOURNAL OF INFORMATION TECHNOLOGY & POLITICS 329

Dolan, K. (2005). Do women candidates play to gender Gervais, B. T. (2015). Incivility online: Affective and beha- stereotypes? Do men candidates play to women? vioral reactions to uncivil political posts in a web-based Candidate sex and issues priorities on campaign websites. experiment. Journal of Information Technology & Politics, Political Research Quarterly, 58(1), 31–44. doi:10.1177/ 12(2). doi:10.1080/19331681.2014.997416 106591290505800103 Gibson, R., Lusoli, W., & Ward, S. (2008). The Australian Dolan, K. (2010). The impact of gender stereotyped evalua- public and politics on-line: Reinforcing or reinventing tions on support for women candidates. Political Behavior, representation? Australasian Political Studies Association, 32, 1. doi:10.1007/s11109-009-9090-4 43(1), 111–131. Dolan, K., & Lynch, T. (2016). The impact of gender stereo- Gibson, R., & McAllister, I. (2015). Normalising or equalising types on voting for women candidates by level and type of party competition? Assessing the impact of the web on office. Politics & Gender, 12(3), 573–595. doi:10.1017/ election campaigning. Political Studies, 63(3). doi:10.1111/ S1743923X16000246 1467-9248.12107 Druckman, J., Kifer, M., & Parkin, M. (2007). The technolo- Gilljam, M., Karlsson, D., & Sundell, A. (2010). Politik på gical development of congressional candidate web sites: hemmaplan. Tiotusen fullmäktigeledamöter tycker om How and why candidates use web innovations. Social demokrati [Politics on the Home Turf: Ten-thousand Science Computer Review, 25(4), 425–442. doi:10.1177/ Councilors Thinking About Politics and Democracy]. 0894439307305623 Stockholm, Sweden: SKL Kommentus. Ehin, P., & Solvak, M. (2012). Party voters gone astray: Grant, W. J., Moon, B., & Grant, J. B. (2010). Digital dialo- Explaining independent candidate success in the 2009 gue? Australian politicians’ use of the social network tool European elections in Estonia. Journal of Elections, Public Twitter. Australian Journal of Political Science, 45(4), 579– Opinion & Parties, 22(3), 269–291. doi:10.1080/ 604. doi:10.1080/10361146.2010.517176 17457289.2012.682579 Gulati, G. J., & Williams, C. B. (2013). Social media and Enli, G., & Skogerbø, E. (2013). Personalized campaigns in campaign 2012: Developments and trends for Facebook party-centred politics. Information, Communication and adoption. Social Science Computer Review, 31(5), 577– Society, 16(5), 757–774. doi:10.1080/1369118X.2013.782330 588. doi:10.1177/0894439313489258 Esaiasson, P., & Narud, H. (2013). Between-election democ- Hagen, M., & Mayer, W. (2000). The modern politics of racy: An introductory note. In P. Esaiasson & H. M. Narud presidential selection: How changing the rules really did (Eds.), Between-election Democracy: The representative change the game. In W. G. Mayer (Ed.), In pursuit of the relationship after Election Day (pp. 1–14). Colchester, White House 2000: How we choose our presidential nomi- England: ECPR Press. nees (pp. 1–55). New York, NY: Chatham House European Parliament. (2013). Women in decision-making: The Publishers. role of the new media for increased political participation. Hansen, K. M., & Kosiara-Pedersen, K. (2014). Cyber-cam- Directorate General for Internal Policies, Policy Department paigning in Denmark: Application and effects of candidate C: Citizens’ Rights and Constitutional Affairs. Brussels. campaigning. Journal of Information Technology & Politics, Evans, H. K., Cordova, V., & Sipole, S. (2014). Twitter style: An 11(2), 206–219. doi:10.1080/19331681.2014.895476 analysis of how House candidates used Twitter in their 2012 Harrison, L. (2005). Electoral systems and party strategies in campaigns. PS: Political Science and Politics, 47,454–462. Britain: Women candidates in the 1999 and 2004 Farrell, D. (1996). Campaign strategies and tactics. In L. European parliamentary elections. Representation, 41(2), Leduc, R. G. Niemi, & P. Norris (Eds.), Comparing 85–95. doi:10.1080/00344890508523293 Democracies (pp. 160–183). London, England: Sage. Heith, D. (2003). The lipstick watch: Media coverage, gender, Folke, O., & Rickne, J. (2016). The glass ceiling in politics: and presidential campaigns. In R. P. Watson & A. Gordon Formalization and empirical tests. Comparative Political (Eds.), Anticipating Madam President (pp. 123–130). Studies, 49(5), 567–599. doi:10.1177/0010414015621073 Boulder, CO: Lynne Rienner. Fox, R. L., & Lawless, J. L. (2011). Gendered perceptions and Hermans, L., & Vergeer, M. (2012). Personalization in political candidacies: A central barrier to women’s equality E-campaigning: A cross-national comparison of personali- in electoral politics. American Journal of Political Science, zation strategies used on candidate websites of 17 countries 55,59–73. doi:10.1111/ajps.2011.55.issue-1 in EP Elections 2009. New Media and Society, 15,72–92. Freedman, J. (2002). Women in the European parliament. doi:10.1177/1461444812457333 Parliamentary Affairs, 55(1), 179–188. doi:10.1093/parlij/ Hindman, M. (2008). The myth of digital democracy. 55.1.179 Princeton, NJ: Princeton University Press. Fulton, S. A. (2012). Running backwards and in high heels: Hogan, R. E. (2007). The effects of candidate gender on campaign The gendered quality gap and incumbent electoral success. spending in state legislative elections. Social Science Quarterly, Political Research Quarterly, 65(2), 303–314. doi:10.1177/ 88, 1092–1105. doi:10.1111/ssqu.2007.88.issue-5 1065912911401419 Jackson, N. A., & Lilleker, D. G. (2009). Building an archi- Gainous, J., & Wagner, K. (2013). Tweeting to power: The tecture of participation? Political parties and web 2.0 in social media revolution in American Politics. New York, Britain. Journal of Information Technology & Politics, 6(3), NY: Oxford University Press. 232–250. doi:10.1080/19331680903028438 330 L. A. C. SANDBERG AND P. ÖHBERG

Jackson, N. A., & Lilleker, D. G. (2011). Microblogging, Lassen, D., & Brown, A. (2011). Twitter: The electoral con- constituency service and impression management: UK nection? Social Science Computer Review, 29(4), 419–436. MPs and the use of Twitter. The Journal of Legislative doi:10.1177/0894439310382749 Studies, 17(1), 86–105. doi:10.1080/13572334.2011. Lilleker, D. G., & Koc-Michalska, K. (2013). Online political 545181 communication strategies: MEPs, E-representation, and Jacobs, K., & Spierings, N. (2016). Social media, parties, and self-representation. Journal of Information Technology & political inequalities. New York, NY: Palgrave MacMillan. Politics, 10(2), 190–207. doi:10.1080/19331681.2012.758071 Jalalzai, F. (2006). Women candidates and the media: Lilleker, D. G., Koc-Michalska, K., Schweitzer, E. J., Jacunski, 1992–2000 elections. Politics and Policy, 34(3). M., Jackson, N., & Vedel, T. (2011). Informing, engaging, doi:10.1111/j.1747-1346.2006.00030.x mobilizing or interacting: Searching for a European model Jungherr, A. (2014). Twitter in politics: A comprehensive of web campaigning. European Journal of Communication, literature review. Retrieved from http://ssrn.com/abstract= 26(3), 195–213. doi:10.1177/0267323111416182 2402443, 10.2139/ssrn.2402443. Lilleker, D. G., Tenscher, J., & Štětka, V. (2015). Towards Just, M., & Crigler, A. (2014). Gender and self-presentation in hypermedia campaigning? Perceptions of new media’s social media: An analysis of the 10 most competitive 2012 importance for campaigning by party strategists in com- U.S. Senate races (APSA 2014 Annual Meeting Paper). parative perspective. Information, Communication and Retrieved from http://ssrn.com/abstract=2453741. Society, 18, 747–765. doi:10.1080/1369118X.2014.993679 Kahn, K. (1996). The political consequences of being a woman: Lusoli, W. (2005). The Internet and the European parliament How stereotypes influence the conduct and consequences of elections: Theoretical perspectives, empirical investigations political campaigns. New York, NY: Columbia University and proposals for research. Information Polity, 10(3), 153–163. Press. Marcinkowski, F., & Metag, J. (2014). Why do candidates use Karlsen, R. (2011). A platform for individualized campaign- online media in constituency campaigning? An application ing? Social media and parliamentary candidates in the of the theory of planned behavior. Journal of Information 2009 Norwegian election campaign. Policy and the Technology and Politics, 11, 151–168. doi:10.1080/ Internet, 3. doi:10.2202/1944-2866.1137 19331681.2014.895690 Karlsen, R., & Enjolras, B. (2016). Styles of social media Margolis, M., & Resnick, D. (2000). Politics as usual: The campaigning and influence in a hybrid political commu- cyberspace revolution. London, England: Sage. nication system: Linking candidate survey data with Meeks, L. (2013). A woman’s place: Gender politics and Twitter data. The International Journal of Press/Politics, Twitter in the 2012 elections (Unpublished doctoral disser- 21(3), 338–357. doi:10.1177/1940161216645335 tation). University of Washington, Seattle. Kosiara-Pedersen, K., & Hansen, K. M. (2015). Gender dif- Megarry, J. (2014). Online incivility or sexual harassment? ferences in assessments of party leaders. Scandinavian Conceptualizing women’s experiences in the digital age. Political Studies, 38,26–48. doi:10.1111/scps.2015.38. Women’s Studies International Forum, 47(1), 46–55. issue-1 doi:10.1016/j.wsif.2014.07.012 Krook, M. L. (2006). Reforming representation: The diffusion Müller, W. C. (2000). Political parties in parliamentary of candidate gender quotas worldwide. Politics & Gender, 2 democracies: Making delegation and accountability work. (3), 303–327. doi:10.1017/S1743923X06060107 European Journal of Political Research, 37, 309–333. Larson, S. G. (2001). Running as women? A comparison of doi:10.1111/1475-6765.00515 female and male Pennsylvania assembly candidates’ cam- Nulty, P., Theocharis, Y., Popa, S. A., Parnet, O., & Benoit, K. paign brochures. Women and Politics, 22, 107–124. (2016). Social media and political communication in the doi:10.1300/J014v22n02_04 2014 elections to the European Parliament. Electoral Larsson, A. O. (2015a). The EU Parliament on Twitter— Studies, 44, 429–444. doi:10.1016/j.electstud.2016.04.014 Assessing the permanent online practices of parliamentar- Obholzer, L., & Daniel, W. T. (2016). An online electoral ians. Journal of Information Technology and Politics, 12(2), connection? How electoral systems condition representa- 149–166. doi:10.1080/19331681.2014.994158 tives’ social media use. European Union Politics, 17(3), Larsson, A. O. (2015b). Going viral? Comparing parties on 387–407. doi:10.1177/1465116516630149 social media during the 2014 Swedish election. Pearson, K., & McGhee, E. (2013). What it takes to win: Convergence: The International Journal of Research into Questioning “gender neutral” outcomes in US House elec- New Media Technologies, 23(2), 117–131. tions. Politics & Gender, 9(4), 439. doi:10.1017/ Larsson, A. O., & Kalsnes, B. (2014). “Of course we are on S1743923X13000433 Facebook”: Use and non-use of social media among Pierre, J. (2015). The Oxford Handbook of Swedish Politics. Swedish and Norwegian politicians. European Journal of New York, NY: Oxford: Oxford University Press. Communication, 29(6), 653–667. doi:10.1177/ Sapiro, V., Walsh, K. C., Strach, P., & Hennings, V. (2011). 0267323114531383 Gender, context, and television advertising: A comprehen- Lasorsa, D. (2012). Transparency and other journalistic sive analysis of 2000 and 2002 House races. Political norms on Twitter: The role of gender. Journalism Studies, Research Quarterly, 64(1), 107–119. doi:10.1177/ 13(3), 402–417. doi:10.1080/1461670X.2012.657909 1065912909343583 JOURNAL OF INFORMATION TECHNOLOGY & POLITICS 331

Schmitt, H. (2009). European parliament elections after east- Utz, S. (2009). The (potential) benefits of campaigning via social ern enlargement. Journal of European Integration, 31(5), network sites. Journal of Computer-Mediated Communication, 525–535. doi:10.1080/07036330903290629 14, 221–243. doi:10.1111/jcmc.2009.14.issue-2 Scully, R., & Farrell, D. (2003). MEPs as representatives: Vaccari, C. (2013). A tale of two e-parties: Candidate websites Individual and institutional roles. JCMS: Journal of in the 2008 US presidential primaries. Party Politics, 19(1), Common Market Studies, 41(2), 269–288. 19–40. doi:10.1177/1354068810391287 Segaard, S. B., & Nielsen, J. A. (2013). Local election blogs: Vergeer, M., & Hermans, L. (2013). Campaigning on Twitter: Networking among the political elite. Information Polity Microblogging and online social networking as campaign 18, 4(October), 299–313. tools in the 2010 general elections in the Netherlands. Southern, R. (2015). Is Web 2.0 providing a voice for out- Journal of Computer-Mediated Communication, 18(4), siders? A comparison of personal website and social media 399–419. doi:10.1111/jcc4.2013.18.issue-4 use by candidates at the 2010 UK general election. Journal Vergeer, M., Hermans, L., & Cunha, C. (2013). Web cam- of Information Technology and Politics, 12(1). doi:10.1080/ paigning in the 2009 European parliament elections: A 19331681.2014.972603 cross-national comparative analysis. New Media and Spierings, N., & Jacobs, K. (2014). Getting personal? The Society, 15(1), 128–148. doi:10.1177/1461444812457337 impact of social media on preferential voting. Political Vergeer, M., Hermans, L., & Sams, S. (2011). Is the voter only Behavior, 36(1), 215–234. doi:10.1007/s11109-013-9228-2 a tweet away? Micro blogging during the 2009 European Strandberg, K. (2013). A social media revolution or just a case parliament election campaign in the Netherlands. First of history repeating itself?—The use of social media in the Monday, 16(8). doi:10.5210/fm.v16i8.3540 2011 Finnish parliamentary elections. New Media and Walton, S., & Rice, R. (2013). Mediated disclosure on Twitter: Society, 15(85). doi:10.1177/1461444812470612 The roles of gender and identity in boundary imperme- Strömbäck, J. (2008). Four phases of mediatization: An ana- ability, valence, disclosure, and stage. Computers in lysis of the mediatization of politics. The International Human Behavior, 29(4), 1465–1474. doi:10.1016/j. Journal of Press/Politics, 13(3), 228–246. doi:10.1177/ chb.2013.01.033 1940161208319097 Wängnerud, L. (2015). The principles of gender-sensitive par- Tenscher, J. (2013). First- and second-order campaigning: liaments. New York, NY: Routledge. Evidence from Germany. European Journal of Ward, S., Gibson, R., & Lusoli, W. (2003). Online participation Communication, 28(3), 241–258. doi:10.1177/ and mobilization in Britain: Hype, hope and reality. 0267323113477633 Parliamentary Affairs, 56(4), 652–688. doi:10.1093/pa/gsg108 Tufekci, Z. (2008). Grooming, gossip, Facebook and Werner, B. (1997). Financing the campaigns of women can- Myspace. Information, Communication and Society, 11(4), didates and their opponents: Evidence from three states, 544–564. doi:10.1080/13691180801999050 1982–1990. Women & Politics, 19,81–97. 332 L. A. C. SANDBERG AND P. ÖHBERG

Appendix 1

Table A1. Campaign Activities. Factor Analysis (Principal Component). Traditional Web campaign Outreach Personal Media Meetings with party group 0.80 Contact with media 0.72 Spread the party’s campaign material 0.70 Public speeches 0.59 Blog 0.83 Individual Web page 0.81 Twitter 0.59 Facebook 0.56 Calling voters 0.73 SMS messages 0.72 Individual e-mail list to inform voters and supporters 0.59 Visit workplaces and organizations 0.57 Door knocking 0.75 Letters to voters 0.72 Individual election poster 0.70 Individual commercial in radio, TV, cinema 0.75 Individual ads in newspaper 0.62 Individual election pamphlet or similar print 0.62 Eigenvalue (initial) 4.45 2.34 1.55 1.52 1.11

Appendix 2

Table A2. Gender and Activity on Twitter, February 2013 (NBREG). 2013 2013 2013 Number of tweets Twitter interactions Twitter information Sex (women) 0.91*** 1.02*** 0.83*** (0.31) (0.36) (0.31) Placement on the party list (1–43) −0.02 −0.01 −0.02 (0.02) (0.02) (0.02) Age (22–83) −0.02** −0.03*** −0.02* (0.01) (0.01) (0.01) Incumbent −0.11 −0.11 −0.12 (0.62) (0.75) (0.62) Experience as MP −0.90** −0.73 −0.98** (0.45) (0.54) (0.45) Party (Social Democrats are ref. category) Center Party 0.90* 0.93 0.89* (0.54) (0.64) (0.53) Christian Democrats −0.12 −0.39 0.15 (0.53) (0.62) (0.54) Conservative Party 1.11* 0.67 1.38** (0.65) (0.77) (0.64) Feminist Party −0.42 −0.36 −0.43 (1.04) (1.23) (1.04) Green Party 0.18 0.15 0.20 (0.49) (0.59) (0.50) Left Party 0.60 0.43 0.76 (0.57) (0.69) (0.57) Liberal Party 0.74 0.90 0.64 (0.56) (0.67) (0.56) Pirate Party 0.74 0.77 0.75 (0.64) (0.76) (0.63) Sweden Democrats 1.73** 2.20*** 1.06 (0.70) (0.82) (0.72) Intercept 4.34*** 3.69*** 3.71*** (0.62) (0.71) (0.64) Pseudo R-squared 0.02 0,03 0.02 N 101 101 101 Note. Standard error in parentheses. ***p < 0.01. **p < 0.05. *p < 0.1. JOURNAL OF INFORMATION TECHNOLOGY & POLITICS 333

Table A3. Response Rate per Party (%). (K) Personal spots on radio, TV, movie houses Political party Response rate (L) Personal flyers or other campaign material (give-aways) Center Party 47 (M) Personal Web site Christian Democrats 47 (N) Mailing list to inform supporters and voters about my Conservative Party 30 campaign Feminist Party 44 (O) Own blog Green Party 60 Left Party 51 (P) Facebook Liberal Party 29 (Q) SMS Pirate Party 45 (R) Twitter Social Democrats 53 (S) Other, please specify: Sweden Democrats 35

Response options: Appendix 3. Descriptive Statistics (1) little important (2) somewhat important Table A4. Response Rate per Age Groups (%). (3) very important Age groups Statistics Sweden The survey (4) most important 18–29 17 16 30–49 46 41 50–64 30 30 Importance of Twitter 65+ 8 13 “ ” Note. The statistics are taken from Statistic Sweden’sWebpage: How important was Twitter for you when it comes to: http://www.scb.se/sv_/Hitta-statistik/Statistik-efter-amne/ Demokrati/Allmanna-val/Europaparlamentsval-nominerade-och- (A) Telling people about your daily activities as a politician valda/#c_undefined (B) Disseminating information of the party’s daily activities (C) Being able to show a more personal side of yourself (D) Getting into personal contact with voters Table A5. Response Rate and Gender (%). (E) Visualizing your personal views Gender Statistics Sweden The survey (F) Highlighting the party’s positions Women 47 51 (G) Debating with political opponents Note. The statistics are taken from Statistic Sweden’sWebpage: (H) Criticizing political opponents http://www.scb.se/sv_/Hitta-statistik/Statistik-efter-amne/ (I) Reaching out to as many voters as possible Demokrati/Allmanna-val/Europaparlamentsval-nominerade-och- valda/#c_undefined (J) Having an impact on traditional media

Response options: Appendix 4 Questionnaire 1—To a very low extent, 7—To a very high extent Importance of Campaign Activities

“Were any of the following activities part of your campaign? Evaluations of Twitter And if yes, how important were they? “To what extent have you been successful through Twitter at: Items:”

(A) Telling people about your daily activities as a politician (A) Door-knocking, canvassing (B) Disseminating information of the party’s daily activities (B) Distributing party campaign material (C) Being able to show a more personal side of yourself (C) Calling up voters on the phone (D) Getting into personal contact with voters (D) Visiting businesses and social organizations (E) Visualizing your personal views (E) Meetings with party elites/members and/or party (F) Highlighting the party’s positions groups (G) Debating with political opponents (F) Media activities (interviews, press releases) (H) Criticizing political opponents (G) Public speeches and rallies (I) Reaching out to as many voters as possible (H) Personal campaign posters (J) Having an impact on traditional media (I) Direct mailing (J) Personal newspaper ads 1—To a very low extent, 7—To a very high extent NMS0010.1177/1461444819893991new media & societyJacobs et al. 893991research-article2020

Special Issue Article

new media & society 2020, Vol. 22(4) 611–633 Twitter and Facebook: © The Author(s) 2020 Article reuse guidelines: Populists’ double- sagepub.com/journals-permissions https://doi.org/10.1177/1461444819893991DOI: 10.1177/1461444819893991 barreled gun? journals.sagepub.com/home/nms

Kristof Jacobs Radboud University, The Netherlands

Linn Sandberg University of Oslo, Norway

Niels Spierings Radboud University, The Netherlands

Abstract Social media are said to be a core driver of populists’ current success. Yet, our knowledge of how populist politicians use social media is limited. We argue that they can use Twitter and Facebook, politically the most important platforms, as a “double- barreled gun,” each serving a different target. Based on the architecture of the platforms and the populist ideology, we expect that Twitter is used to name and shame journalists publicly, Facebook to activate anger among citizens. Both types of use are examined by studying the Members of Parliament (MPs) of Austria, The Netherlands, and Sweden. We collected 9852 tweets for the 475 MPs on Twitter and 10,355 Facebook posts from the 287 MPs with a Facebook Page. Using negative binomial regression and content analyses, we find that populists seem eager to activate anger. They are not more likely to @-mention media accounts, but “shame” them roughly three times more often.

Keywords Facebook, hybrid media, political communication, populism, Twitter

Corresponding author: Kristof Jacobs, Institute for Management Research (IMR), Radboud University of Nijmegen, Elinor Ostrom building 2.361, Nijmegen 6500HK, The Netherlands. Email: [email protected] 612 new media & society 22(4)

Introduction We live in a “populist era” (Mudde, 2016: 28). It is often claimed that online media are an important factor contributing to this development (e.g. Groshek and Koc-Michalska, 2017; Mudde, 2016: 28–29). Despite this alleged importance, the study of populism and online media is still in its infancy. Indeed, in their seminal paper about how populism manifests itself on social media, Engesser et al. (2017a) note that “[n]umerous scholars have analyzed politicians’ use of Facebook and Twitter but they either have ignored the concept of populism or alluded to it only briefly” (p. 1110). Moreover, and second, even though Facebook has a far broader user base, the focus of the existing studies is on Twitter (e.g. Jacobs and Spierings, 2018; Spierings et al., 2018; Tromble, 2018a, 2018b; Waisbord and Amado, 2017; Wells et al., 2016). And the few studies that do examine Facebook, highlight a third limitation of the existing literature: they only examine to what extent all types of political actors use a populist communication style (Engesser et al., 2017b; Ernst et al., 2017; Zulianello et al., 2018). These studies are important to get an insight into what populist online communication looks like, and into the degree to which populists contribute to and use this type of communication. However, they tell us little about how populist politicians specifically use social media (e.g. to engage with ordinary citizens, to pressure journalists, and to share information about kindred news outlets). Overall, despite the alleged importance of online media for populists, our knowledge of the actual use of social media by populist political actors is surprisingly limited. This matters because it hampers our overall understanding of how online media may contribute to the rise of populism. To examine this subject, a first requisite, in line with Engesser et al.’s (2017a) call, is a more systematic theorization of the linkage between populism and social media use. In this study, we set out to provide such a theoretical understanding by going back to the core definitions of populism. We then link populism to the specific architectures of Twitter and Facebook and examine which opportunities each of them offers to populists. Our main argument is that social media can be a powerful weapon in the hands of populists. To paraphrase the popular trope, using social media, populists bring their gun to a knife-fight. Specifically, they seem to bring a double-barreled gun, whereby Twitter and Facebook can each serve a different target. Twitter can be used to target journalists and publicly name and shame them to increase media coverage. Facebook, on the other hand, can be used to mobilize broad audiences directly by activating anger. The central question of this study is therefore, Do politicians from populist parties differ from non-populist ones in terms of naming and shaming on Twitter and anger activation on Facebook? The remainder of this study is structured as follows. In the next paragraph, we discuss the overall concept of populism and link it to the broader online opportunity structure of social media (cf. Engesser et al., 2017a). Afterwards, we discuss how Twitter can be used by populists for public naming and shaming and how Facebook can be used to activate anger. In the “Methods” section, we detail how we collected our data and discuss the method and operationalization. Our empirical analysis is based on a unique dataset covering Austria, The Netherlands, and Sweden. While it is unclear how long Facebook will Jacobs et al. 613 still allow the scraping of such data, we were able to collect both Twitter and Facebook data for all the Members of Parliament (MPs) of Austria, The Netherlands, and Sweden via the Netvizz application for Facebook and the Twitter API. We collected 9852 tweets for 475 MPs who were active on Twitter and 10,355 Facebook posts from the 287 MPs who had an active Facebook Page. To these data, we apply negative binomial multilevel regression analyses and complement this with content analyses to examine to what extent our theoretical expectations hold. Finally, we discuss the implications and generalizability of our findings.

Populism, Twitter, and Facebook Populism and online opportunity structures In line with most recent research on populist political communication, we conceive of populism “as an ideology that is communicated discursively” (De Vreese et al. 2018: 425). Within this conceptualization, there are two main approaches: an “actor-centered” and a “communication-centered” approach (Stanyer et al., 2017: 353; see also De Vreese et al., 2018: 428). The first approach focuses on a priori defined populist political actors, while the second focuses on how populism manifests itself without an a priori classification of which political actors are populist. Specifically, the actor-centered approach starts from populism as a (thin-centered) ideology “that considers society to be ultimately separated into two homogeneous and antagonistic groups, ‘the pure people’ versus ‘the corrupt elite’, and which argues that politics should be an expression of the volonté géné- rale (general will) of the people” (Mudde, 2004: 543). Based on this definition, political actors are classified as populist. Afterwards, for example, communication strategies of populists are compared with those of non-populists. The communication-centered approach starts from the “key characteristics of populist communication and then analyzes the extent to which different actors make use of these, without an a priori classification of who is populist or not” (De Vreese et al., 2018: 428). Specifically, this approach considers populism a communication style that frequently refers to the people, positions itself on their side, attacks elites, and does so in a way that is simplistic, emotional, and negative (Engesser et al., 2017a: 1282; Stanyer et al., 2017: 359). Clearly, the two approaches are not mutually exclusive and refer to the same core (i.e. a Manichean worldview based on people-centrism and anti-elitism).1 The choice between the two approaches thus mainly depends on the research question one is interested in. When one wishes to study how a priori classified populist political actors behave (and how they differ from non-populists, for example, in whether they use Twitter for naming and shaming), like we do, then the actor-centered approach is best suited. Several studies have suggested that social media may be particularly useful for populist political actors. Specifically, social media are linked to online opportunity structures,2 the “factors inherent to the online media system” (Engesser et al., 2017a: 1280).3 For instance, social media are unmediated and allow populist political actors to speak directly to “the people” without “elitist” journalists being able to filter and/or alter the populists’ raw message (Jacobs and Spierings, 2016). Interactivity is another important online opportunity structure. Social media offer a means of “demonstrating that the 614 new media & society 22(4) politician and party are ‘of the people’ and not the political establishment” by allowing to engage with members of the public (Tromble, 2018b: 229). Moreover, the mobilization potential of social media via posts that go viral, is another feature of social media that fits very well with the way populist political actors conduct politics (Engesser et al., 2017a: 1286). These online opportunity structures offer opportunities (i.e. they have the potential to be used by politicians), which already indicates that not necessarily all of them are actually used. This has been particularly the case with interactivity. While one can expect that “members of populist parties may be particularly likely to use social media to engage the public” (Tromble, 2018b: 696), in practice studies found them to engage less, both in general (Jacobs and Spierings, 2018; Tromble, 2018b; Waisbord and Amado, 2017) and specifically with ordinary citizens (Spierings et al., 2018).4 The unmediated and mobilization opportunity structures are examined less in studies of populists’ social media use and need more attention. When studying this potential of social media more in detail, it is crucial to avoid lumping together different platforms, such as Twitter and Facebook (Bossetta, 2018: 474). After all, different platforms all have their own architectures in terms of the openness of the network, functionality, algorithmic filtering, and “datafication.” These characteristics in turn influence the specific opportunities that a platform can offer political actors. Moreover, the user base of a platform also greatly determines who political actors can reach and target. In this study, we theoretically and empirically zoom in on the two platforms that are currently the most used and relevant to political actors, Twitter and Facebook (Bossetta, 2018; Jungherr, 2016). Twitter is an “elite” medium: it has a user base that is not at all representative of the broader population (Jungherr, 2016: 78) as it mostly consists of opinion leaders, such as journalists (Jacobs and Spierings, 2016: 162). Twitter’s architecture allows to easily target and publicly tag these journalists (Bossetta, 2018: 488). Facebook, on the other hand, has a very broad user base and a highly sophisticated matching service to ensure that “the right users” see their posts (Bossetta, 2018: 489). This makes Facebook an excellent platform for broad mobilization (Bond et al., 2012). In sum, Twitter may be especially useful for populists to target journalists, while Facebook is useful to target potential voters. Below, we develop each of the two consecutively.

Populism and Twitter: naming and shaming The dominant view about how populists use Twitter in relation to traditional media is best illustrated by a famous cartoon by Dana Summers. The cartoon pictures of Donald Trump throwing a ball with the label “social media” on it over the heads of a group of journalists, directly into the lap of an ordinary citizen (Summers, 2016). The cartoon suggests that populists like Trump use social media primarily to bypass the media gatekeepers and directly speak to the people. This view is dominant in most of the contemporary populism literature (Engesser et al., 2017b; Krämer, 2017; Van Kessel and Castelein, 2016). While the bypass function may hold well for Facebook, Twitter especially is often used by political actors as a “propaganda tool to communicate in real time with [. . .] Jacobs et al. 615 political journalists” and as a means “to influence coverage in older media” (Chadwick, 2013: 196). In other words, as Chadwick (2013) has pinpointed, our current media system is a hybrid one whereby the boundaries between the different media types are blurred. Twitter can function as a new way to get into the old media, a new door into an old house so to speak. Returning to the cartoon-example, this argument is well-illustrated by the Dutch populist radical right politician Geert Wilders who retweeted the aforemen- tioned cartoon, somewhat ironically, to celebrate the fact that he got into numerous mainstream media programs thanks to “one tweet” (Wilders, 6 February 2017).5 As Krämer (2017) summarizes, populists can use social media to put journalists “under pressure to cover them more intensively and favorably” (p. 1303). But how would they do that? And why would Twitter be suited to do so? Here we propose a mechanism that fits very well with the Twitter architecture: public naming and shaming. Specifically, the Twitter architecture allows populist politicians to @-mention journalists directly (“naming”) and attack or bully them (“shaming”).6 In addition, because Twitter has an open network structure (Bossetta, 2018: 479), such naming and shaming allows politicians to accuse journalists publicly and visibly, questioning their ethics and reputation via rhetorical entrapment (“you claim to be neutral but you are biased!”; cf. Katzenstein, 2013). This strategy on Twitter has two advantages: it draws the attention of the journalist by directly tagging him or her, and it does so in a way that is visible for everybody, even people who do not follow the politician. Both of these advantages make it harder for journalists to ignore the attack. Naming and shaming is different from shaming without naming: tweeting negatively about the press without @-mentioning them directly. This latter strategy is different in target and in goal. While the former is targeted at journalists (as it explicitly draws their attention) and aims to improve and/or increase media coverage, the latter is targeted at the populists’ followers, does not explicitly draw the attention of the journalist him/herself and aims to reduce the credibility of the “lying” press in general. While the “shaming without naming” strategy can be used on many different platforms, such as Facebook, blogs, or even alternative online media, the Twitter architecture makes that platform especially well-suited for the “naming and shaming” strategy. So far, empirical research into the matter is fairly limited. Two seminal studies exist, and while they do not distinguish between the two strategies, they still allow indirect insight into how much Twitter is used for naming and shaming. Both find that populists do attack the media but it is unclear whether this is naming and shaming or just shaming without naming. In their study about four populist and four non-populist Latin American presidents, Waisbord and Amado (2017) do find that populist presidents attack the press far more than the non-populists. Similarly, Van Kessel and Castelein (2016) find that the Dutch populist radical right party leader Geert Wilders attacks journalists when he gets “bad press,” though it is unclear whether he @-mentions those journalists in his tweets and whether he does so more often than non-populist Dutch MPs (p. 604). In this study, we will examine whether Twitter is used by populists for naming and shaming. We do so for all of the populist politicians, not just the leaders. After all, it may well be that the task of pressuring and attacking the press falls to backbenchers, as to minimize the potential costs of attacking journalists for the party leader. 616 new media & society 22(4)

Summarizing all of this, one can thus expect that

Hypothesis 1. Populist political actors are more likely to name and shame media accounts on Twitter than non-populist ones.

Populism and Facebook: activating anger Contrary to Twitter, Facebook is more of a mass medium and has a far broader (and more representative) user base (Bossetta, 2018: 481). As a result, it can be useful for the “activation of the citizenry” (Larsson, 2015: 463). This can be online and offline mobilization. In its simplest form, mobilization is the activation of users by triggering engagement that can spill over to the offline world. Empirical research has suggested that Facebook has the potential to mobilize in terms of donating, turning up at rallies, and ultimately a higher voter turnout (Bronstein, 2013). But how would populists mobilize on Facebook? Empirical research in the field of traditional cyber-campaigning suggests that mobilization is mainly triggered using emotional appeals (Brader, 2005), which holds even more for mobilization on Facebook (Bronstein, 2013). As Engesser et al. (2017b) highlight, populists “often rely on emotions,” specifically “anger, fear, and resentment” (p. 1285). Indeed, several different emotions can be activated. However, research by Rico et al. (2017) shows that populist citizens are primarily driven by anger. The same holds for populist politicians. The anger is inherent to their ideology, which is rooted in a Manichean worldview that considers the elites morally wrong and corrupt and wants to restore the power to the “pure” people. Such an interpretation of the world is both “integrative and divisive: not only does it attempt to unite an angry and silent majority, but it also tries to mobilize this majority against a defined enemy (e.g. the establishment)” (Mudde and Kaltwasser, 2017: 11). As populism scholars have shown recently, citizens’ populism and anger typically “lie dormant and require activation” to create successful mobilization (Hawkins et al. 2018: 4). Such activation can occur due to changes in the context (e.g. uncovering corruption scan- dals), but populist political actors themselves can also contribute to the activation. How would they do this in practice? In her seminal work, Wahl-Jorgensen (2018) shows that populist politicians, and specifically Donald Trump, try to activate anger-based appealing to a sense of injustice (Wahl-Jorgensen, 2018: 768). Related, activation can occur via pro- voking a backlash against pervasive cultural changes “eroding the cultural norms one knew since childhood”—in essence a cultural backlash (Inglehart and Norris, 2017: 446). By doing so, populist political actors create and/or amplify a feeling that society is in crisis and “inject urgency and importance to their message” (Mudde and Kaltwasser, 2017: 106). The architecture of Facebook is well-suited to activate anger among citizens. Since 2016, Facebook allows its users to engage with posts by reacting via an anger emoji instead of the classic thumb up like. Populists can thus strategically use Facebook to trigger and harvest a sense of anger. Moreover, the Facebook datafication enables “extremely sophisticated” targeting to activate specific types of users and offers complex, real-time analytics to monitor the reaction of users and fine-tune these efforts (Bossetta, 2018: 488). Using such tools, populists can micro-target those audiences who are the most Jacobs et al. 617 susceptible to messages appealing to a sense of injustice and/or a cultural backlash (cf. Vandeputte, 2018). In this study, we will examine whether populists indeed activate anger on Facebook. Just like in our Twitter analysis, we will do so for all of the populist politicians, not just the leaders, and compare them with non-populists. Although a number of studies examine the degree to which parties employ a populist communication style on Facebook (Engesser et al., 2017b; Ernst et al., 2017; Zulianello et al., 2018), to the best of our knowledge, no empirical study has examined if populists indeed activate anger on Facebook. Based on the theoretical argumentation outlined in this section, we expect that

Hypothesis 2. Populist political actors are more likely to generate anger on Facebook than non-populist ones.

Methods Overall approach Our overall methodological approach combines statistical analyses with content analyses. The statistical analyses provide insights into the prevalence of (elements of) the two strategies, whereas examining the content more closely helps to properly assess what populists tweet and post. Furthermore, to create greater leverage, we apply this mixed strategy to three comparable countries: Austria, The Netherlands, and Sweden. This selection follows the logic of a most similar systems design. The three countries have similar political systems (proportional list electoral system; parliamentary system) and media systems (democratic corporatist model; cf. Hallin and Mancini, 2004). Moreover, they all have a high internet penetration rate (Internet World Stats, 2018) and in all three, there is a significant presence of populist parties.7 Our main focus is thus on such contexts, but we will reflect on the larger significance of our results in the “Conclusion” section, also with respect to, for instance, majoritarian political systems. Furthermore, our case selection strategy focusing on most similar systems does not mean that the three selected countries are exactly the same. Especially, their media system and historical experience with populist parties show differences below the surface. Regarding the former, in a more recent classification, Brüggemann et al. (2014) point out that the Netherlands is more liberal and open than the other two, while Austria shows more political parallelism and Sweden a higher degree of journalistic professionalism. Especially, the restrictiveness of the offline media system affects how political actors behave online (Strandberg, 2008). Of the three, the Swedish system is the most restric- tive to populists.8 Hence, one might expect that the Swedish populist MPs are more likely to opt for naming and shaming to increase their media coverage and use anger activation to mobilize their supporters. Regarding the historical experience with populist parties, Sweden is also different from the two other countries. Despite the fact that the Sweden Democrats (SD; populist radical right party, cf. Rooduijn, 2018) is now equally large as its counterparts in Austria and the Netherlands, the party only entered the 618 new media & society 22(4) parliament for the first time in 2010 and is therefore less “normalized.” Again, this could entail that the SD may need to recur to social media to draw attention. In the “Results” section, we will highlight whether the general patterns among MPs holds across countries.

Populist and non-populist MPs We collected data from the 690 MPs, whereby we included both populist and non-populist MPs and both party leaders and backbenchers. Of these MPs, 475 have an active Twitter account and 287 a Facebook Page. The classification of which political actors were populist follows the dominant definition and classification in the literature (cf. Mudde, 2016; Rooduijn, 2018). Specifically, in the Netherlands, the Party for Freedom (PVV), Forum for Democracy (FvD), and the Socialist Party (SP); in Sweden, the SD; in Austria, the Freedom Party of Austria (FPÖ) is considered populist. We compare the MPs of populist parties with those of the non-populist parties. As prior research has shown that postmaterialist parties also tend to stand out (e.g. Gibson and McAllister, 2015; Jacobs and Spierings, 2016), in addition, we distinguish MPs of these parties in order not to overestimate the differences between populists and non-populists. For the Netherlands, GreenLeft (GL), the Party for the Animals (PvdD) and Democrats 66 (D66) are classified as postmaterialists, in Sweden, the Environment party is classified as such (MP; cf. Bolin, 2016; Jacobs and Spierings, 2018). The Austrian green postmaterialist party lost all its seats in the 2017 general election, but Liste Pilz is by some considered to belong to the postmaterialist party family, as it is a spinoff from the Green party (Bodlos and Plescia, 2018). However, others consider it to be a populist party (cf. Buzogány and Scherhaufer, 2018; Eberl et al., 2017b). In this light, we have run our models classifying the eight MPs of Liste Pilz as postmaterialist, and rerun them classifying Pilz as populist. This did not lead to substantially different conclusions and when and where results devi- ated it is mentioned in the text. Our regression models include controls for several other important characteristics of MPs that have shown to matter in other studies: gender, age, seniority (operationalized as days in parliament), and parliamentary leadership. Political leaders are more likely to adopt social media early, as are younger MPs, women (especially the more personal Facebook), and more senior politicians (see Jacobs and Spierings, 2016, 2018; Tromble, 2018b). We also add technical controls about MPs social media behavior (see below). Finally, to control for country differences, we include the countries at a separate level in our multilevel models, which in practice acts as the equivalent of fixed effects models (we do not include variables at the country level). The differences between the groups of populist and non-populist MPs as presented in the “Results” section are thus corrected for compositional differences regarding the characteristics discussed above (see the “Results” section for the main descriptive statistics).

Twitter Data collection. For all the Austrian, Dutch, and Swedish MPs on Twitter, we collected the data for June 2018. June represents a recent comparable period across the three Jacobs et al. 619 countries as it was neither a summer break nor an official campaign month in any of the three. Concretely, we collected information on all the accounts that were actively @-mentioned by the MP during that month, as well as the messages in which these @-mentions occurred. In total, we collected 9852 tweets including an @-mention (excluding retweets). These data were analyzed following two main steps.

Measuring and analyzing “naming.” First, to establish whether the @-mentioned account was a media account, we followed the procedure developed by Spierings et al. (2018). We applied an automated search in the accounts’ screen names and self-given descrip- tions using identifier words, such as “journalist” (in German, Dutch, and Swedish), “political reporter” (Spierings et al., 2018) as well as the labels of important news venues for each country (e.g. NPO, RTL, Der Standard, Riksmedia, Sveriges Radio).9 This identified 24.8% of the 9852 tweets as @-mentioning media accounts.10 Next, we examine whether populist politicians @-mention (“name”) media more often than their non-populist counterparts. The dependent variable in this analysis is the number of times an MP account @-mentioned a media account. Concretely, for each of the MPs, we summed all media account @-mentions (i.e. tweets including an @-mention, but exclude retweets [which also include the handle of the retweeted account]). The analyses of these data are controlled for the degree to which MPs @-mention in general. As the distribution of the number of media @-mentions is clearly skewed, with a high number of zeroes, we estimated (multilevel) negative binomial regression models. We carried out several robustness tests to examine whether different operationalizations of the dependent variable would change our results, but the main results stayed the same.11

Measuring and analyzing “shaming.” After determining the “naming,” we proceed to the “shaming” part of our analysis and carry out a content analysis of tweets @-mentioning the media accounts. We hand-coded all 3232 tweets from both populist and non-populists that @-mentioned (“naming”) a media account in June. All tweets were coded by two different coders per country (the authors and three assistants); each coder was either native speaker or proficient in the language of the tweets. The same holds for the authors enabling them to check the coding by the assistants. We considered tweets to include “shaming” (1) when the media source or journalist was addressed in one (or more) of the following ways: (a) reproaching, (b) accusing, (c) insulting. If none of these types of shaming were present, the naming tweet was coded as (0). This means simply voicing disagreement with the media account is not shaming; it must be an attack on the media account itself. Also, a tweet including an @-mention of a media account and an insult (e.g. an insult directed at another politician, whereby the tweeting MPs includes a media account via an @-mention to draw attention to his or her tweets) is in itself not sufficient to receive a (1) code. The insult should be directed at the media account. All tweets were manually coded. In contrast to other studies, we coded both the content of populist and non-populist actors (cf. Van Kessel and Castelein, 2016). To illustrate a shaming tweet: when a populist responded to a journalist with the tweet “Fake! This is only mentioned in the provisional (!) protocol.”12 This is considered shaming and coded as (1). To be clear: this shaming logic applied to all MPs, so if, for instance, an MP of the—non-populist—Swedish Liberals accuses a media source to produce fake 620 new media & society 22(4) news, this is also considered “shaming.” Moreover, if a political leader insults a certain media source in a televised debate and another MP tweets a supportive message including and supporting this insult and @-mentioning the media source, this is also coded (1). Alternately, when an MP replied to a journalist writing and tweeting about the need for the public broadcast to earn its own money with the following tweet “I do agree with you. Though one could also imagine that the advertisement money the public broadcast collects could be spent to help investigation journalists at other media,” it was considered naming, but not shaming and thus coded (0). In the “Results” section, we present more examples. The coders first coded a subsample of 15 tweets that were identified by the authors as difficult to classify. Only in one case, did this lead to discrepancies, after which the coding scheme was adjusted (to result in the final coding scheme as described above). The actual coding of the whole corpus of tweets had a strong inter-coder reliability, the coding resulted in the following Krippendorff’s alpha: .8648 for the Swedish MPs, .8467 for the Dutch ones, and .9369 for the Austrian ones.

Facebook Data collection. Regarding the Facebook analyses, we scraped data from the Facebook Pages of the 342 MPs who had a Facebook Page in the time period ranging from 1 November 2017 till 1 February 2018. Again, the period is chosen to represent a comparable and rather recent period: in none of the countries, electoral campaigns were held at that time. For all the MPs, we collected their Facebook handle and per handle (i.e. account), we automatically scraped data about the activity of the Page via the app Net- vizz. We obtained posts of the 287 MPs who were active in that time period (the others did not post anything). We could collect all the posts during this time period since no Page posted more than the maximum 999 posts that Netvizz app can process.13 In total, we collected information about 10,355 Facebook posts. Again, these data were analyzed first in regression models, after which we zoomed in on the content more specifically.

Measuring and analyzing “angry reactions.” First, we examine whether populists’ Facebook posts generate more angry reactions than those of other MPs. We nest the messages (level 1) in the MPs (level 2).14 Our dependent variable is operationalized as the message-level reactions to each message that is Anger proportional to the total number of “reactions” for that post (i.e. the combined number of like, sad, love, anger, haha, or wow). We use multilevel regression models because not only do they capture the nested structure of the data (preventing deflated errors), but also because of the variability between the Facebook Pages, such as the different characteristics of followers (Heiss et al., 2018) and differences between posts.

Measuring and analyzing “activating anger.”. While the regression analyses outline above, will show whether the followers of populist MPs post more anger reactions, this does not directly mean that these MP actively try to evoke this anger. It could also be that the followers of Facebook Pages of populist MPs may be angrier in general. This is not unlikely given that populist citizens tend to be angrier on average (Rico et al., 2017). Hence, and Jacobs et al. 621 second, we complement our multilevel analysis with a content analysis of a subsample of the more than 10,000 posts,15 to assess whether the posts generating anger actually activated anger. Below, we discuss how we selected the subsample, which is following by the operationalization or coding scheme for coding this sample. To start with, we selected all posts that generated the most anger: posts that score two standard deviations above the total corpus average on the absolute number of anger reaction as well as two standard deviations above the average proportion of reactions that is anger. These 111 Facebook posts provide us with a large enough sample to see whether posts receiving a high degree of anger reactions actually include activation thereof (more on the coding below). Analyzing the posts that are the most responded to with an anger reaction, however, does not tell us about how successful MPs are in activating anger. It only includes hits, not misses. MPs might try to activate anger in many more posts, but fail in doing so. To assess this, we also hand-coded a random sample16 of the remaining posts, which received fewer anger reactions. As the variation in anger is larger among that subset, we decided to draw a sample twice the size of the “anger posts.” The 111 “anger” and 222 “bench-mark” posts were coded using the same coding scheme discussed below, and by the same coders as described for the tweet coding (again two coders per country, again first using a subsample). Krippendorff’s alpha was .8648 for the Swedish MPs, .9196 for the Dutch ones, and .8899 for the Austrian ones. Posts were considered to “activate anger” in line with the theoretical considerations discussed in the “Populism and Facebook: Activating anger” section. This means that it was not sufficient for the MP to just give information, present plans, or an opinion (that lead to anger), but the post needs to (a) include an active play on a sense of injustice leading to perceived unwanted and undeserved outcomes and/or (b) invoke a cultural backlash targeting the political elite (which is implicitly or explicitly blamed for societal changes). To illustrate our coding, when a political actor posted the following, it was coded as (1) (activating anger): “The government ignores the ordinary Dutchman. But there are billions of Euros for foreign investors and for free health care for immigrants, who already get priority for social housing. #riseup” (Wilders, 22 January 2018; emphasis added). The italicized parts clearly show that the MP plays into an injustice at the cost of the audience. Conversely, when an MP, for instance, posted the following, it was coded (0): “Why does the whole issue with the A50 highway take so long? Question for the Minister of Infrastructure regarding the postponement of widening the ‘highway of the dead’.”17 This message does address an issue MPs are dissatisfied about and strong words are used, but not in a way to activate anger; the message is not about injustice or backlash, but a signal that the MP considers it serious and is on top of it.

Results In this section, we discuss to what extent the two strategies—naming and shaming and anger activation—occur in practice. First, we will shortly discuss the descriptive statistics (Table 1). Next, we discuss each platform (and strategy) consecutively. For each, we start by discussing the prevalence of the strategy and then turn to the regression results (Table 2) to see whether there is a systematic difference between populist and 622 new media & society 22(4)

Table 1. Descriptive statistics of core variables (unit of analysis = MP).

Min.–Max. All MPs MPs using Twitter MPs using Facebook

% or Mean (SD) % or Mean (SD) % or Mean (SD) Media @-mentions 0–211 6.81 (21.96) Average % angry 0–15.88 1.50 (2.45) reactions Party family Populist 0–1 20.1 17.7 13.2 Postmaterialist 0–1 10.9 14.1 11.9 Another party 0–1 69.0 68.2 74.9 Leader 0–1 3.8 4.8 8.0 Seniority 0–25.61 6.32 (5.46) 6.11 (5.14) 5.47 (5.00) Age 24–96 48.05 (11.15) 46.54 (10.87) 46.05 (10.65) Female MPs 0, 1 40.0 40.2 39.0 Country Austria 0, 1 26.4 14.1 34.5 The Netherlands 0, 1 23.0 32.4 29.6 Sweden 0, 1 50.6 53.5 35.9 N 690 690 475 287

SD: standard deviation. non-populist MPs. Finally, we discuss the content analysis, where we provide examples of how the strategies play out and are linked to populism.

Descriptive statistics Table 1 summarizes the main descriptive statistics discussed earlier (for the sake of clarity, we aggregated them here at the level of MPs). First of all, the numbers show that more MPs are using Twitter than Facebook Pages. This is partly due to the division over the three countries: Facebook Pages are somewhat less popular among Swedish MPs, and more so among the Austrians, who in turn use Twitter less. Table 1 also shows that in both the Twitter and Facebook analysis, we have a lower proportion of populist MPs than there are in parliament: populist MPs are on average less likely to have an account/Page (which is in line with other research, cf. Dolezal, 2015; Spierings and Jacobs, 2018). Furthermore, the table shows notable variation in the degree of media @-mentions on Twitter and the average proportion of anger reactions on Facebook. This will be discussed in more detail in the “Naming and shaming on Twitter” and “Activating anger on Facebook” sections.

Naming and shaming on Twitter The raw data show that of the 475 MPs who have an active Twitter account, 52% did not @-mention any media account, 25% @-mentioned media accounts only one to five times, Jacobs et al. 623

Table 2. Results of the Twitter and Facebook analyses (multilevel negative binomial).

Model 1. Twitter media @-mentions Model 2. Facebook anger

B-coefficient B-coefficient Populist −.074 2.137*** Postmaterialist .043 .369 Another party Ref. Ref. leader .243 1.364** Seniority −.007 .003 Age −.011* .004 Female MPs −.017 −.040 Total monthly @-mentions .016*** − No. of reactions to − .000* FB post (count) Intercept 1.043** −1.571** Negative binomial .921 13.782 MP level 1.000 2.777*** Post level 1.000 BIC 1570.613 58,101.081 N 475 MPsa 287 MPsa 10,355 posts

MP: Member of Parliament. The difference in the number of MPs is due to the fact that more MPs have an active Twitter account than a Facebook Page. aWe included country at the highest level to control for between-country variation. Obviously, because of the low N, we do not include any independent variables at the country level: in practice, it functions as a fixed effects model. The country-level variance for was .105 in model 1 and .111 in model 2. Both were not significant. *p = .05; **p < .01; ***p < .001.

15% 6–20 times. Of the remaining 8%, some MPs were very active in @-mentioning media accounts. Five MPs even @-mentioned media accounts no less than 130–211 times. Whether these differences relate to populism or other characteristics of MPs are shown in Table 2. The negative binomial model shows that populists do not @-mention media accounts more often than non-populist MPs. They seem equally active in seeking the attention of media accounts. As Chadwick (2013) has already noted: mainstream parties also contact media accounts. Our (presented and additional) models confirm this across the three countries.18 Actually, the only factor that seems to make a difference is age: older MPs seem to @-mention media less. While there is no difference between populist and non-populist MPs in terms of how often they “name” media accounts, it could still be that there is a difference between the two in the way they tweet to these media accounts. As highlighted in the theoretical section, one can expect populist MPs to be more prone to shaming. 624 new media & society 22(4)

Table 3. Content analysis of tweets naming a media account.

Populist MP Non-populist MP Shaming 60 (10.6%) 82 (3.1%) No shaming 507 (89.4%) 2583 (96.9%) Total 567 (100%) 2665 (100%)

MP: Member of Parliament. A Mann–Whitney U test revealed that the difference between the two was statistically significant (p < .001; Mann–Whitney U = 698,824.500).

Naming and shaming clearly occurs. For instance, the Austrian MP populist Christian Höbart (FPO)—responsible for nine of the “naming and shaming” tweets— criticized Thomas Mayer, a journalist for the quality newspaper Der Standard, for being inconsistent (after the journalist called migration worries scaremongering; Hobart, 1 June 2018):

Herr Mayer, scaremongering like in 2015–2016? Yesterday you found the migration wave of 15/16 problematic, but now you feel migration worries are scaremongering? Hmm . . .

Similarly, the Dutch Populist MP Fleur Agema (PVV) accused the current affairs program EenVandaag of being one-sided and acting as a propaganda machine of the government. She posted the following tweet while also citing the original media tweet:

This is what you get when you make propaganda items for @minvws [the Ministry of Health Care, Welfare and Sports] without hearing the other side @EenVandaag. He [the minister] obscures the fact that the government only pays for half the costs. The other half will be paid for by budget cuts in the local health sector. (Agema, 1 June 2018)

Similarly, an MP from the SD, Martin Kinnunen, accused a journalist from the newspaper Dagens Nyheter of scaremongering and being subjective:

Isn’t it strange that you are trying to nazify someone who wants to strengthen the Jewish minority in Sweden? The entire text is hysterical. (Kinnunen, 24 June 2018)

Table 3 provides a systematic comparison between populists and non-populists in terms of shaming in the tweets naming media accounts. Although our analysis shows that shaming was not pervasive among the naming tweets, MPs from populist parties did it far more often than non-populist MPs.19 Indeed, populists did so more than three times as often as non-populists. It must be noted though that neither populist or non-populist MPs engage often in naming and shaming activities: only a fairly limited proportion of the tweets naming media accounts can be characterized as shaming.20 With this caveat in mind, we can cau- tiously say that our hypothesis 1 is corroborated: populists do name and shame media accounts more often than non-populists. Jacobs et al. 625

Activating anger on Facebook The descriptive statistics of the Facebook data (Table 1) show that though there is quite some variation between individual posts, the percentage of anger reactions is relatively limited. Most users still simply click the like button (which is also in line with other research, e.g. Eberl et al., 2017a). At the same time, some MPs have an average anger response of 15%, and 105 messages (1%) got more than 50% anger reactions.21 Zooming in on systematic differences in the relative number of anger reactions, there are notable differences between posts from populist and those from non-populist MPs. On average, posts by populist politicians have 4.14% angry reactions, whereas their non-populist counterparts’ posts only have 1.09% of such reactions. This considerable difference holds in the multivariate analysis (see Table 2). The single most important, and robust,22 factor affecting the proportion of anger reactions to a Facebook post is whether it comes from a populist MP, and this holds across countries.23 Moreover, additional models show that this anger does not stand alone, but that post receiving relatively more anger are on average shared and commented on more.24 This holds for populist and non-populist MPs, whereby the former receives more comments and shares on their posts to start with, leading to less additional value of anger garnering posts,25 although it should be remembered that they garner a lot more anger with their posts (see Table 2) and this makes posts more likely to be shared and commented on nevertheless. Across parties, garnering anger seems an effective strategy for gaining attention. Next to populism, the only other significant independent variable is being a leader. Ceteris paribus, for party leaders, the average relative number of angry reactions is also higher, though only about by half as much as the populism effect size. While our regression analysis does not give us insight into why this is the case, it could be that party leaders are generally the most polarizing figures of a party as they are in the limelight most often. Moreover, they are more likely to attract followers who do not support their party but want to see what leaders from other parties post. Future studies could dig deeper into this and examine the dynamics at play. The next question is whether this anger results from populist MPs actively seeking such angry reactions. Do they activate the anger? Several posts were actually very explicit about playing into anger, and many others were less explicit but clearly played into it as well. For instance, two posts by the Austrian FPO–MP Mario Kunasek even included the anger icon (cf. Figures 1–2). These activated anger by evoking a cultural backlash against changes to the identity of the country. Other posts activated anger by appealing to a sense of injustice, such as this post by the leader of the Dutch left-wing populist SP, Lilian Marijnissen (22 January 2018):

A 1,000 billion dollars. That is what the 500 richest persons earned extra last year. 2.7 billion dollars a day. But at the same time, 815 million people are still starving. Every day. How much longer will we put up with this? Support us: https://doemee.sp.nl/lilian.

Finally, some of the posts, such as this one by the Dutch MP Geert Wilders (13 January 2018), combined a sense of injustice with the cultural backlash: 626 new media & society 22(4)

Figure 1. Crispy French fries. Text says “EU prohibits our crispy French fries. Idiocy” (Kunasek, 24 November 2017).

Figure 2. Third gender. Text says “Does Austria need a third gender? Never! Scandal!” (Kunasek, 10 November 2017). Jacobs et al. 627

Figure 3. Anger activation in and reactions on (non-)populist MPs’ Facebook posts.

What sadness and what a big disgrace. Asylum seekers can bring women and children to the Netherlands in huge numbers by means of family reunification. But two 94-year-old Dutch people who have been together for 72 years are separated from each other. The Dutch are discriminated against!

It was Wilders’ post with the highest number of angry emojis (N = 3635). The post clearly pits two groups against each other, the in-group (Dutch people) and the out-group (asylum seekers), who are disrupting the country. At the same time, it explicitly refers to a sense of injustice: “they” can unite with their families but “our own people” are separated from each other, implicitly blaming the government/ political elite. But is there a difference between populist and non-populist MPs and how effective is their anger activation? To examine this, we carried out a content analysis of the 111 posts with the most anger reactions and a random sample of all other posts (double that number: N = 222). The outcomes are shown in Figure 3, based on which several comparisons can be made and conclusion drawn. To begin with, the posts receiving a lot of anger overwhelmingly belonged, and statistically significantly more often,26 to populist MPs: a whopping 73% of the angering messages was posted by a populist MP (the bars dark segments), while just 20% of the total number of posts we collected came from populist MPs (as is also reflected in the right bare in Figure 3). Moreover, for the posts receiving most anger, it was clearly found that almost all were also referring to a sense of injustice and/or cultural backlash. They did not simply elicit anger, but were activating anger. This holds for both the populist and non-populist MPs’ posts, although somewhat less so for the non-populists, who activated anger statistically significantly less27 (cf. the relative size of the solid to the zig-zag striped segment in the left bar). 628 new media & society 22(4)

Finally, the right bar clearly shows that it is not the case that almost all messages on Facebook—regardless of whether they are responded to with anger—include a sense of injustice and/or cultural backlash to activate anger. In the posts leading to less anger, more than 90% did not include such an activation. Noteworthy is that, comparing the two bars, the populist posts with anger activation (the solid segments) are, statistically significantly,28 more successful than their non-populist counterparts (the lighter, solid segments): of the posts activating anger of populists, 95% lead to anger (i.e. is in the left bar), whereas this is 63% for non-populists. In sum, while the overall number of anger reactions to MPs’ posts on Facebook Pages is modest, those of populist MPs generate significantly more anger reactions than those of non-populist MPs. While followers of Facebook Pages from populist MPs are likely to be angrier than those of non-populist ones (cf. Rico et al., 2017), it is clear that populist MPs also use Facebook to activate that anger by playing onto a sense of injustice and/ or evoking a cultural backlash. Hypothesis 2 is thus corroborated.

Conclusion In this study, we set out to examine differences in the use of Twitter and Facebook between populist and non-populist MPs in Austria, the Netherlands, and Sweden. Thereby, we argued that it is important to focus on what sets the populist ideology apart from other party families and the platforms architectures. Moreover, we distinguish between different platforms because the architecture and user base of these platforms provides different opportunities. Specifically, we theorized that Twitter can be used by populist MPs to publicly tag and pressure journalists (“naming and shaming”) to improve and increase media coverage. Facebook on, the other hand, can be used to mobilize citizens by activating anger. Our main research question was “Do politicians from populist parties differ from non-populist ones in terms of naming and shaming on Twitter and anger activation on Facebook?” This question can be answered positively. We found that both strategies occur, though activating anger more so than naming in shaming. Furthermore, we found that there are notable differences between populists and non- populists, as the latter hardly ever used naming and shaming and triggered anger less frequently. Our results have several implications for studies on populism and social media. First, it shows that it is indeed important not to lump different platforms together under the header “social media”: the differences in user base and architecture matter empirically (cf. Bossetta, 2018). Second, while most research about populists on Twitter analyzes the platform from the starting point that it offers a means to bypass traditional media, it seems fruitful to consider social media, and Twitter in particular, as part of the hybrid media system where actors sometimes deliberatively seek spillovers from social media to traditional media (and vice versa; cf. Chadwick, 2013). Empirically, we analyzed three countries with similar political and media systems. While it seems plausible that our results can be generalized to other countries with similar political and media systems, especially because we found the findings are consistent across the three countries included in the analysis, it remains to be seen whether they can also be generalized to majoritarian political systems and other media systems. Especially Jacobs et al. 629 in majoritarian settings, populists may resort to general anti-media critiques. In such a setting, naming and shaming may be more prevalent. The anger activation is something we expect to be more consistent across different types of political and media systems as it fits with the ideological core of populism (i.e. a Manichean worldview based on anger and anti-elitist feelings). In addition to testing our mechanisms in other types of countries, it is fruitful to also examine other, newer platforms such as Instagram and Snapchat. So far, these other platforms are fairly peripheral in political campaigners’ strategies (Bossetta, 2018), but they may become more central in the future. Furthermore, it needs to be stressed that our study focused on two distinct strategies (naming and shaming and anger activation), but it is likely that populist political actors are different on other fronts too, such as other types of emotional appeals or the spreading of specific types of content (e.g. retweeting tweets from ordinary citizens or sharing disinformation). Moreover, while anger is the emotion that fits best with populist citizens (cf. Rico et al., 2017) and populist politicians’ Manichean worldview (cf. Mudde and Kaltwasser, 2017), it is not the only emotion that can be activated. Examining other emotions such as fear or resentment, or even more positive emotions such as admiration (for the leader) is a promising venue for future research. Finally, we wish to emphasize that our analysis covered “peace-time” tweets and posts. It seems likely that during election campaigns populist MPs, and may be also non-populist ones, are more actively naming and shaming journalists. Similarly, because anger is such a powerful mobilization tool (Rico et al., 2017), it seems likely that anger activation will be used more often during election campaigns.

Funding The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs Kristof Jacobs https://orcid.org/0000-0002-4485-4927 Linn Sandberg https://orcid.org/0000-0002-9440-7084

Notes 1. It should, therefore, not come as a surprise that empirical studies find those parties who are considered to adhere to a populist ideology are also the ones the most likely to use a populist communication style (Engesser et al., 2017b; Waisbord and Amado, 2017; Zulianello et al., 2018). 2. This argumentation is in line with the general argument in social media studies that different social media platforms have different affordances (Boyd and Ellison, 2007; Faraj and Azad, 2012; Majchrzak et al., 2013). In the field of populism and social media studies, the term “online opportunity structures” is used (cf. Engesser et al., 2017a), but this essentially con- veys the same message: architecture matters. 3. For a detailed discussion of the relationship between populism and online opportunity structures, see: Engesser et al. (2017a) and Jacobs and Spierings (2018). 4. Although Tromble (2018b: 234) does find that Tea Party Republicans have more conversa- tions with ordinary citizens than non-Tea Party Republicans, the caveat being that the Tea party was not a party, but a movement. 630 new media & society 22(4)

5. He finished the tweet with the hashtag #ilovetwitter (Wilders, 6 February 2017). 6. Populists can also employ a strategy of “naming, but not shaming” (i.e. retweeting or tweeting about positive news coverage when it fits the populist narrative). 7. One difference is that the populist radical right FPÖ is in government in Austria. However, as research has shown, populists in government tend to act as if they are one foot in, one foot out and still wish to distance themselves from the “elites” (Albertazzi and McDonnell, 2005). Donald Trump is a good example of this: despite him being in office, he still uses both naming and shaming of journalists and anger mobilization. 8. The Netherlands has the most open system, and the Austrian political parallelism coupled with the fact that the populist radical right FPÖ is in government makes that system more open to populists as well. 9. The coding script is available via the authors. 10. The number of media accounts is similar to the one reported by Spierings et al. (2018), who use a different sample of data. 11. For instance, some MPs had extreme scores on the dependent variable. These high numbers can be due to intense debates on Twitter between the same accounts. Some MPs might end up in such debates quite often. Therefore, we tested the robustness of our analyses by rerunning the models truncating the dependent variable at different levels: with a maximum of 80, 50, and 20. Each of these models returned similar results and lead to exactly the same conclusion as the ones presented in the “Results” section. 12. Tweet by the Austrian populist MP Dagmar Belakowitsch (12 June 2018). 13. One can only collect data on Facebook Pages, as Facebook profiles are not public. Facebook’s Pages provide an open network structure where users can either like or follow a page to get updates and comment or respond on posts. For the individual politician, it is more convenient to manage communication on pages compared with profiles and it also gives the politician tools to monitor and track response in addition to being able to select more functions, including verification of account and statistical measures to follow responses and users. 14. We also add the countries as controls to filter out between-country variance (level 3), but add no explanatory variables at this level (cf. fixed effects models). 15. Given the large number of posts and the need for hand-coding, we could simply not code all messages. 16. An SPSS random sampling syntax procedure. 17. The post was by the Dutch Christian Democratic MP Van Helvert CDA (1 November 2017). 18. This result holds up in models coding PILZ as populist, without the control for general number of @-mentions, and in the three country-disaggregated models. All more detailed results can be obtained from the authors. 19. The shaming was certainly not restricted to leaders. Among non-populists, 3.1% of the naming tweets was shaming for non-leaders and 3.4% for leaders. (For populist, to few leader tweets named to draw a strong conclusion on leaders, but the results are similar.) 20. These results are in line with what Waisbord and Amado (2017) and Van Kessel and Castelein (2016) find. 21. The average number of emoji engagement of these posts was 447; these were not post with only two angry emojis and one like. 22. Considering PILZ as populist changes the B-coefficient by only .003. 23. Analyzing the three countries separately shows a positive, and the strongest effect, for populism in each. 24. The models are the same as the one presented in Table 2, but then including the relative anger as independent variable and taking the absolute count of comment and shares per post, respectively. The more detailed results can be obtained from the authors. Jacobs et al. 631

25. This was shown by including interaction terms between anger and populism in the models discussed in the previous note: both interaction terms were negative (i.e. weakening) and statistically significant. More detailed results can be obtained from the authors. 26. A Mann–Whitney U test for populist versus non-populist on likelihood to be among the most anger-reacted-to posts: 8,466,787.50 (p < .001). 27. A Mann–Whitney U test for populist versus non-populist on likelihood of posts to be coded anger activating: 6660.00 (p < .001). 28. A Mann–Whitney U test for activating versus non-activating on likelihood to be among the most anger-reacted-to posts: 2492.50 (p < .001).

References Albertazzi D and McDonnell D (2005) The Lega Nord in the second Berlusconi government: in a league of its own. West European Politics 28(5): 952–972. Bodlos A and Plescia C (2018) The 2017 Austrian snap election: a shift rightward. West European Politics 41(6): 1354–1363. Bolin N (2016) Green parties in Finland and Sweden: successful cases of the North? In: Van Haute E (ed.) Green Parties in Europe. Milton, ON, Canada: Routledge. Bond R, Fariss C, Jones J, et al. (2012) A 61-million-person experiment in social influence and political mobilization. Nature 489(7415): 295–302. Bossetta M (2018) The digital architectures of social media: comparing political campaigning on Facebook, Twitter, Instagram, and Snapchat in the 2016 US Election. Journalism & Mass Communication Quarterly 95(2): 471–496. Boyd D and Ellison NB (2007) Social network sites: definition, history, and scholarship. Journal of Computer-Mediated Communication 13(1): 210–230. Brader T (2005) Striking a responsive chord: how political ads motivate and persuade voters by appealing to emotions. American Journal of Political Science 49(2): 388–405. Bronstein J (2013) Like me! Analyzing the 2012 presidential candidates’ Facebook pages. Online Information Review 37(2): 173–192. Brüggemann M, Engesser S, Buchei F, et al. (2014) Hallin and Mancini revisited: Four empirical types of western media systems. Journal of Communication 64(6): 1037–1065. Buzogány A and Scherhaufer P (2018) Austrian Greens: from pyrrhic presidential victory to parliamentary exit. Environmental Politics 27(3): 566–571. Chadwick A (2013) The Hybrid Media System: Politics and Power. Oxford: Oxford University Press. De Vreese C, Esser F, Aalberg T, et al. (2018) Populism as an expression of political communication content and style: a new perspective. The International Journal of Press/politics 23(4): 423–438. Dolezal M (2015) Online campaigning by Austrian political candidates: determinants of using personal websites, Facebook, and Twitter. Policy & Internet 7(1): 103–119. Eberl JM, Tolochko P, Greussing E, et al. (2017a) Emotional reactions on Austrian parties’ Facebook pages during the 2017 Austrian Parliamentary Election. Computational communication science lab report, University of Vienna, Vienna, October. Eberl J-M, Zeglovits E and Sickinger H (2017b) Austria election preview: Sebastian Kurz and the rise of the Austrian “anti-party.” Available at: http://eprints.lse.ac.uk/85414/1/europp- blog-2017-10-04-austria-election-preview-sebastian-kurz-and-the.pdf (accessed 15 October 2018). Engesser S, Ernst N, Esser F, et al. (2017a) Populism and social media: how politicians spread a fragmented ideology. Information, Communication & Society 20(8): 1109–1126. 632 new media & society 22(4)

Engesser S, Fawzi N and Larsson A (2017b) Populist online communication: introduction to the special issue. Information, Communication & Society 20: 1279–1292. Ernst N, Engesser S, Büchel F, et al. (2017) Extreme parties and populism: an analysis of Facebook and Twitter across six countries. Information, Communication & Society 20(9): 1347–1364. Faraj S and Azad B (2012) The materiality of technology: an affordance perspective. In: Leonardi PM, Nardi B and Kallinikos J (eds) Materiality and Organizing: Social Interaction in a Technological World. New York: Oxford University Press, pp. 237–258. Gibson R and McAllister I (2015) Normalising or equalising party competition? Assessing the impact of the web on election campaigning. Political Studies 63(3): 529–547. Groshek J and Koc-Michalska K (2017) Helping populism win? Social media use, filter bub- bles, and support for populist presidential candidates in the 2016 US election campaign. Information, Communication & Society 20(9): 1389–1407. Hallin D and Mancini P (2004) Comparing Media Systems: Three Models of Media and Politics. Cambridge: Cambridge University Press. Hawkins K, Kaltwasser C and Andreadis I (2018) The activation of populist attitudes. Government and Opposition 55(2): 283–307. DOI: 10.1017/gov.2018.23. Heiss R, Schmuck D and Matthes J (2018) What drives interaction in political actors’ Facebook posts? Profile and content predictors of user engagement and political actors’ reactions. Information, Communication & Society 22: 1497–1513. Inglehart R and Norris P (2017) Trump and the populist authoritarian parties: the silent revolution in reverse. Perspectives on Politics 15(2): 443–454. Internet World Stats (2018) Internet in Europe Stats. Available at: https://www.internetworldstats. com/stats4.htm#europe (accessed 15 October 2018). Jacobs K and Spierings N (2016) Social Media, Parties, and Political Inequalities. New York: Palgrave MacMillan. Jacobs K and Spierings N (2018) A populist paradise? Examining populists’ Twitter adoption and use. Information, Communication & Society 22: 1681–1696. Jungherr A (2016) Twitter use in election campaigns: a systematic literature review. Journal of Information Technology & Politics 13(1): 72–91. Katzenstein S (2013) Reverse-rhetorical entrapment: naming and shaming as a two-way street. Vanderbilt Journal of Transnational Law 46: 1079–1099. Krämer B (2017) Populist online practices: the function of the Internet in right-wing populism. Information, Communication & Society 20(9): 1293–1309. Larsson A (2015) Pandering, protesting, engaging. Norwegian party leaders on Facebook during the 2013 “Short campaign.” Information, Communication & Society 18(4): 459–473. Majchrzak A, Faraj S, Kane GC, et al. (2013) The contradictory influence of social media affordances on online communal knowledge sharing. Journal of Computer-mediated Communication 19(1): 38–55. Mudde C (2004) The populist Zeitgeist. Government and Opposition 39(4): 541–563. Mudde C (2016) Europe’s populist surge: a long time in the making. Foreign Affairs 95: 25–30. Mudde C and Kaltwasser C (2017) Populism: A Very Short Introduction. New York: Oxford University Press. Rico G, Guinjoan M and Anduiza E (2017) The emotional underpinnings of populism: how anger and fear affect populist attitudes. Swiss Political Science Review 23(4): 444–461. Rooduijn M (2018) What unites the voter bases of populist parties? Comparing the electorates of 15 populist parties. European Political Science Review 10(3): 351–368. Spierings N and Jacobs K (2018) Political parties and social media campaigning. Acta Politica 54: 145–173. Jacobs et al. 633

Spierings N, Jacobs K and Linders N (2018) Keeping an eye on the people: who has access to MPs on Twitter? Social Science Computer Review 37: 160–177. Stanyer J, Salgado S and Ströbäck J (2017) Populist actors as communicators or political actors as populist communicators: cross-national findings and perspectives. In: Aalberg T, Esser F, Reinemann C, et al. (eds) Populist Political Communication in Europe. New York: Routledge. Strandberg K (2008) Online electoral competition in different settings: a comparative meta- analysis of the research on party websites and online electoral competition. Party Politics 14(2): 223–244. Summers D (2016) Social media. Available at: https://www.washingtontimes.com/cartoons/toon- ing-president-trump/social-media/ (accessed 15 October 2018). Tromble R (2018a) Thanks for (actually) responding! How citizen demand shapes politicians’ interactive practices on Twitter. New Media & Society 20(2): 676–697. Tromble R (2018b) The great leveler? comparing citizen —politician Twitter engagement across three Western democracies. European Political Science 17: 223–239. Van Kessel S and Castelein R (2016) Shifting the blame. Populist politicians’ use of Twitter as a tool of opposition. Journal of Contemporary European Research 12(2): 594–614. Vandeputte B (2018) Bent u fan van N-VA en leest u Het Laatste Nieuws? Dan wil Vlaams Belang u bereiken op Facebook. Available at: https://www.vrt.be/vrtnws/nl/2018/10/05/bent-u-fan- van-n-va-en-leest-u-het-laatste-nieuws-dan-wil-vlaam/ (accessed: 15 October 2018). Wahl-Jorgensen K (2018) Media coverage of shifting emotional regimes: Donald Trump’s angry populism. Media, Culture & Society 40(5): 766–778. Waisbord S and Amado A (2017) Populist communication by digital means: presidential Twitter in Latin America. Information, Communication & Society 20(9): 1330–1346. Wells C, Shah D, Pevehouse J, et al. (2016) How Trump drove coverage to the nomination: hybrid media campaigning. Political Communication 33(4): 669–676. Zulianello M, Albertini A and Ceccobelli D (2018) A populist Zeitgeist? The communication strategies of Western and Latin American political leaders on Facebook. The International Journal of Press/Politics 23(4): 439–457.

Author biographies Kristof Jacobs is associate professor in Comparative Politics at Political Science Department of the IMR, Radboud University. His research focuses on the challenges to democracy (e.g., populism) and how political actors solve these (e.g., by using social media or implementing institutional reforms). He has published, among others, in the Information, Communication & Society, Electoral Studies and Political Behavior and the European Journal of Political Research. Linn Sandberg is PhD candidate at the Department of Media and Communication at University of Oslo. Sandberg’s research interest mainly include party politics and social media. Sandberg’s work has previously been published in journals such as Journal of Information Technology and Politics. Niels Spierings is assistant professor in Sociology at the Department of Sociology and Radboud Social Cultural Research, Radboud University. His research interests include political inequalities; the social implications of technological developments, such as social media; the political parties- voters connection, and the rise of populism in European politics. On these issues he has published, among others, in the Information, Communication & Society, Patterns of Prejudice, Political Behavior and West European Politics.

Populist legislators on the People’s Platform: Facebook adoption and

affective reactions in Austria, the Netherlands, and Sweden

The argument goes that social media reinforces the rise of populism due to populists’ actors emotionally charged language. Whether this benefits populist politicians, however, depends on their social media adoption and whether their language actually elicits more emotional responses. This is a study of those two core elements of populist politicians’ presence on Facebook. It looks at 682 MPs Facebook uptake in the Netherlands, Sweden and Austria and the emoji responses to 10,355 of their posts. Firstly, we argue that populist parties’ centralized structure enforces discipline, which makes Facebook use by their MPs more problematic, with the exception of their leadership. Our study shows that populist party leaders use Facebook prolifically to reach out to their community, whereas relatively few of populist parties other elected parliamentarians use Facebook. Second, we argue that Facebook posts of populist MPs’ will activate feelings of indignation, leading to responses that are more emotional. Empirically this is borne out. Messages posted by populist actors receive more emojis than ordinary likes compared to other parties. ‘Love’ and particularly ‘anger’ and ‘haha’ stand out. An exploration of messages content shows that ‘haha’ also reflects sarcastic laughs ridiculing political opponents, paving the way for anger.

Keywords: Populism, Social media, Facebook, Emotions

Introduction

Social media provides political actors the freedom to articulate their ideology and spread their messages directly to citizens, which is said to benefit populists particularly (e.g. Engesser,

Ernst, Esser & Büchel, 2017; Ernst, Engesser, Büchel, Blassnig & Esser 2017). Part of this logic is empirically corroborated in recent empirical studies showing that populists’ communication more strongly includes emotionally loaded language (e.g., Engesser, Fawzi,

Larsson, 2017; Heiss & Matthes, 2019; Wirz 2018). However, to understand if and how populists do indeed benefit from social media, the other elements in that larger chain of reasoning needs to be studied. In this paper we therefore theorize and provide empirically founded answers to two adjacent core questions which have remained largely unanswered: does (a) the adoption of Facebook and (b) the extent to which legislators’ posts elicit emotional responses differ between populist and non-populist legislators?

To advance our theoretical and empirical understanding of populist legislators’ use of and impact on social media, we focus on Facebook. Until recently most studies have focused on Twitter, which is used by relatively few ordinary citizens. In contrast, Facebook can be described as ‘the people’s platform’. As it is both widely used and facilitates direct communication with potential voters, it is a crucial arena in the study of populism and social media (Jacobs & Spierings, 2016).

Theoretically, we propose an analytical framework that combines the attributes and architecture of Facebook with the key organizational and ideological characteristics of populist parties. The organizational aspect implies that previous findings showing that political candidates from populist parties are less likely to make use of Twitter (Dolezal,

2015; Jacobs & Spierings, 2018) holds for Facebook too. However, for Facebook specifically, our framework adds that the ideological focus on the people in populism aligns with

Facebook being the “people’s platform”. Consequently, populist legislators might be expected

to be more prone to adopt Facebook than non-populist legislators. Additionally, we theorize that the distinction between leaders and other legislators matters more for populist parties than non-populist parties. Second, we argue that given the architecture of Facebook and the ideology and emotional communicational style of populist MPs (e.g., Engesser, Fawzi,

Larsson, 2017; Heiss & Matthes, 2019; Wirz 2018), their messages will garner more emotional responses than messages of other MPs, in particular the anger, love and haha emoji-responses available to Facebook users.

Empirically, we investigate Facebook adoption and emoji reactions to posts of all 682

Austrian, Dutch and Swedish MPs. All three countries have prominent populist parties and rather similar electoral systems. We gathered data from politicians Facebook pages over the course of three months, covering all posts and responses in the form of likes and emoji reactions. This data set allows us to analyze whether populists are more prone to use

Facebook pages and do so more actively compared to non-populists. By collecting the emoji responses, we can uniquely study whether populists’ posts elicit more emotional reactions on public pages, and what type of emotions they seem to activate.

Below, we first provide a brief summary of the literature to explain the core concepts and current knowledge gaps, after which we move on to theorize the connection between populism and Facebook activity more concretely. In the results section we first address the relative presence of populist MP’s and party leaders, and then the emoji responses to posts.

The quantitative analyses are illustrated and explored further through looking at the contents of several of the populist MP’s posts.

Literature and concepts: Populism and Social media

Populism

Populism as a concept has been revitalized by the recent rise of populist radical-right parties in Western Europe (Rooduijn, 2015). Most scholars studying populism agree that populism can be defined as a set of ideas, whether expressed as a discourse or style (de Vreese et al.,

2018; Jagers & Walgrave, 2007), that represent a (thin) ideology (Mudde, 2004): populism in a minimalistic understanding says something about the Manichean relationship between the

(corrupt) ‘elite’ and ‘the people’ (p. 543). It is people-centered and considers society to be divided between these two groups. According to populist parties, politics should be conducted in the general will of the people, which is anti-elitist by (their) definition. These core elements of populism, as defined by Mudde, can be connected with other ideological standpoints such as nativism, socialism or authoritarianism (Kriesi, 2014, Mudde, 2004). Most populist parties combine populism with a supplementary ideology that is either radical left (e.g. Syriza and

Podemos) or radical right (e.g. the Party for Freedom, the Freedom party of Austria or the

Sweden Democrats).

Regarding populist parties, strong party leaders and a more direct communication between party leadership and party supporters are general traits that fit them particularly.

Features such as charismatic leadership and direct communication between the leader and ‘the people’ are common among populists and although they do not the defining characteristics of populism, they certainly are more likely to go together with populist actors (Mudde, 2004, p.

545).

Populists use of social media

While populism can refer to populist actors and to actors communicating in populist ways, this study treats populism primarily through an actor-centered perspective: we focus on populist MPs and the responses to their posts, compared to (those of) MPs of non-populist parties. As such our focus is not directly on the overall degree of populist communication

among for instance party representatives – the communication-centered perspective (cf.

Engesser et al., 2017a; Ernst et al., 2019; Zulianello et al., 2018), which has been shown to be more present among populist MPs (Ernst et al., 2019), but we will show how this translates to differences along party lines.

Despite the seemingly advantageous traits of social media for populist politicians that have been put forward in previous work and that we will address below, research on whether populist politicians are more present than others on social media and how they use it is scarce.

Indeed, when studying the uptake of online campaign tools, only few studies include populism as an independent variable (Engesser et al., 2017b; Spierings & Jacobs, 2018).

These studies suggest that populist actors are less prone to use social media. Moreover, the few studies that look more closely at how populists use social media show they are less interactive and respond less to Twitter @-mentions (Tromble, 2018; Jacobs & Spierings,

2018). Populists’ behavior on Facebook has hardly been examined systematically, however, so it is unclear whether these results point to a broader pattern.

Social media architecture

Political communication on social media is mediated by the given platform’s digital architecture, which shapes user behavior and enables or constrains a certain type of communication style (Bossetta, 2018). Additionally, the users of social media platforms differ. Due to these factors, politicians and campaigners use Twitter and Facebook for different purposes depending on what audience they want to reach (Kreiss, Lawrence &

McGregor, 2018; Stier, Bleier, Lietz & Strohmaier, 2018). On Twitter, the primary audience is perceived to be media and journalists, whereas Facebook reaches a broader and bigger segment of the population (Kreiss et al., 2018). It has therefore been suggested that populism manifests itself more strongly on Facebook than on Twitter (Ernst et al., 2017). On Facebook,

politicians can reach out directly to a more closed group of followers without journalistic interference. This communication is more under the radar and is more likely to reach more potential voters directly than on Twitter.

Altogether, the literature suggests that populism matters for social media use and that results on Twitter might not translate to Facebook. This necessitates further empirical studies and theorization of the linkages between populist MPs social media uptake and reception comparatively to other politicians.

Theorizing populists and communication on social media

Considering the architecture of Facebook and populist ideology and party structure, we will now first theorize how the adoption of the platform might differ between legislators from populist and non-populist parties and then move on to the question regarding the emotions they might elicit.

Adoption

Recent work has identified several opportunity structures of social media that are compatible with populist ideology. First, the direct access to an audience circumventing mainstream media, a close connection to ‘the people’, and possibilities of personalized communication have been argued to be interlinked with core architectural aspects of social media that fit a populist ideology. Particularly, the ideology’s people-centrism and anti-elitism have been linked to these architectural aspects (Engesser et al., 2017b; Ernst et al., 2017; Jacobs &

Spierings, 2018; Spierings & Jacobs, 2018). Given their ideology, one can expect that populist parties use social media to avoid traditional editorial media (as these can alter, frame or subvert the populists’ raw message) and use the possibilities of engaging directly with ‘the people’ (Jacobs & Spierings, 2018). While already having become influential in political

systems, social media provide populists an added opportunity to maintain their position as strong anti-system challengers (Jungherr, Schroeder & Stier, 2019). Therefore, populist might not only use social media to bypass traditional media, they may also actively use them as platforms to maintain their oppositional role and criticizing “the elite”.

Social media enables disaffected individuals to express themselves and form partisan communities. This mass networking opportunity has been suggested to be a particularly suitable channel for populists’ “mass politics” and appeals to “the people” (Gerbaudo, 2018).

In allowing members of these communities to interact with other likeminded members of their in-group and respond to politicians’ populist messages, social media is therefore argued to empower the ‘silenced majority’ (Bracciale & Martella, 2017; Hameleers, 2018b). These communities might further the cultivation of their in-group identity and communicating hostility toward others in self-perceived anonymity and protected spaces (Hameleers, 2018a).

Facebook might in this regard be a particularly useful platform.

Turning our attention to party structure, another element argued to distinguish populist parties from others (Mudde, 2004; Spierings & Jacobs, 2018), we suggest that populist MPs utilization of social media might be comparatively modest to due to the potential threat this poses to the leadership. As social media allows individual politicians to build an unmediated relationship with voters, this can be seen as a clear disadvantage for the top echelon of populist parties due to their high degree of centralization, empowering backbenchers to the detriment of leadership control (Spierings & Jacobs, 2018).

Populist parties concentrate power in the leadership and this power concentration requires party discipline. The struggle for coherence is therefore particularly persistent in these parties (Heinisch & Mazzoleni, 2016: 239–240), while the exposure of internal conflicts are likely to go viral on social media (Vergeer & Hermans, 2013: 401). The centralized nature of populist parties may thus induce them to curtail back-benching members of the party from

speaking up on social media platforms or select and reward withholding back-benchers, as the leadership of these parties is rather intolerant of dissenting opinions (Taggart, 2000: 100–

103). In Sweden, for example, the populist radical right party is the only party that warns their members for potential negative consequences of social media and urges them to post carefully

(Sandberg & Öhberg, 2017). It is therefore the highly centralized party structure, not populist ideology as such, that indicates a relatively low uptake and activity on social media among

MPs of populist parties (Jacobs & Spierings, 2018).

Lastly, however, it should be noted that a charismatic leader is often regarded as a prominent feature of populism or at least a facilitating factor (Kriesi, 2014, p. 363; Mudde,

2004: 545). The very centralized nature of populist parties with dominant party leaders who make direct claims on behalf of ‘the people’ leads to altogether different assumptions than for their general MPs. For populist leaders, the ideological and structural aspects are aligned. We can therefore expect that populist party leaders take control over the social media communication and do so more than in other parties.

Summarizing, we can now formulate three hypotheses:

H1ideology: Populist MPs adopt and use Facebook more than MPs of other parties.

H1party structure: Populist MPs adopt and use Facebook less than MPs of other parties.

H2: Populist party leaders adopt and use Facebook pages more than party leaders of

other parties.

Eliciting responses: emotions

Starting from the observation populist use more emotionally loaded language, we raised two questions to further unravel the impact thereof. Above we addressed whether populists are more present on Facebook; here we address whether their post also elicit more emotional

responses than non-populist MPs on Facebook do, an aspect for which systematic empirical assessments are lacking. Our combined architecture-parties perspective mainly draws attention to the supply side eliciting such emotional responses.

In a user-driven media environment where users can act both as sender and as receiver of populist ideas - a populist mass self-communication (Hameleers, 2018a) - politicians and users simultaneously post messages on social media whereby supply and demand side influences each other. Consequently, the modes of communication are reciprocal and the type of intentional or unintentional response a post evokes partly reflects the content and tonality of posts on politicians’ pages.

In this environment, what type of messages get posted can be shaped by the more or less immediate response a post gets, which is comprised in the architecture of most social media. More particularly, on Facebook, users can react (emotionally) to posts by either liking, loving, laughing, showing surprise, crying or getting angry. The emotional interaction possible on Facebook we see as an additional aspect of populist mass self-communication. It is in this respect that, although acknowledging that emoji responses do not directly measure the content of a post, do reflect how populist MPs use Facebook in emotional terms, whether it conscious or not on their part.

Turning to theorizing the link between emotions and populism, the emotionalized style of blaming the elite has been argued to explain populism’s persuasiveness (Hameleers, Bos, & de Vreese, 2017). This ‘emotion eliciting effect’ of populists’ communication has been suggested in recent work but not examined sufficiently (Wirz, 2018). Populists’ communication is nonetheless commonly assumed to be inherently more appealing to strong emotions given its Manichean nature, and therefore has the potential of being more persuasive than non-populists’ communication. Accordingly, studies using experiments show that populist appeals elicit stronger emotions than non-populist appeals, that these emotions

mediate the persuasiveness of the appeals, and that populist appeals are more persuasive when they are explicitly emotional (Hameleers et al., 2017; Wirz, 2018). However, it remains untested whether messages from populist MPs evoke more emotions overall and in real-life politics.

Based on the experimental studies and work regarding the populist posts’ content we thus expect more emotional responses to populist MPs’ post. Facebook’s architecture allows such emotional responses with a simple click. One can like a post or select one of six more specific emojis: Like, Love, Haha, Wow, Sad and Angry. The ‘like’ option is the standard setting and for a long time the only response option on Facebook, thus it will be considered the baseline or default response item. Only if people take more effort can they react to a message with a more distinct emoji. Considering all this, we translate the theoretical logic to the following hypothesis.

H3: Reactions to populist MPs’ post include more emojis compared to likes

than the post of MPs of other parties do.

Specifying elicited emotions: anger, pride and sarcasm

Beyond this general expectation, it is likely that particular emotions are more dominant among the responses. It has not only been argued that ‘populism relies, above all else, on an emotional appeal’ but that it particularly plays on a variety of emotions: anger, outrage, disgust, a sense of betrayal, and a sense of loyalty (Fieschi & Heywood, 2004). A fundamental theme in populist ideology and rhetoric is pitting the conflict between the out- groups (e.g. elites, foreigners) against the people whose well-being is threatened.

Populist MPs are said to convey messages in a manner that is more direct and more strident, appealing to gut feelings rather than logical analysis. Populism defines itself in part by accepting this emotional, non-intellectual characterization, arguably as it helps to remain

on the outside of mainstream politics or at least depict that it is the case (Fieschi & Heywood,

2004). The populist actors’ statements to defend the ‘pure people’s’ virtues and interests against the out-groups are also argued to reflect emotions of pride that providee a sense of in- group identity and identification with the populist leader (Wirz, 2018). Thus, conflictive messages referring to threats faced by the people and blaming the elite or other out-groups for bad situations elicit emotions of anger, while appeals demonstrating populist actors’ or leaders’ engagement for the people will elicit the positive emotions of loyalty or pride

(Engesser, Fawzi & Larsson, 2017b).

A slightly different but still related emotional aspect is the role of humor in attacks on the elites (MacMillan, 2017), or more precisely a manifestation of cynical irony which has been linked to the populist leader (Milburn, 2019). Populists’ use of irony can be a way to make sincere statements of political belief while avoiding accusations of either hypocrisy or naivety (Milburn, 2019), and social media might just be the perfect arena for such statements.

Irony and cynicism play on both negativity and positivity. One basic function of irony is communicating something opposite to what is being said, projecting both humor and anger at the same time. Populists are said to use cynical irony as a way to attack the elite or political opponents. Apart from purely humorous traits, it can therefore be expected that ironic criticism or sarcasm also trigger negative emotions of anger, disgust and contempt (Thompson et al., 2016).

In sum, we expect that the demonstrated stronger emotional loaded language of populists’ communication translate as follows: positive messages will evoke a sense of loyalty or pride and negative messages (including irony and sarcasm) will evoke anger or gloating.

Turning to Facebook’s design, it is crucial to realize that Facebook allows different emotions to be displayed, as said: Like, Love, Haha, Wow, Sad and Angry. Positive emotions, like pride, are primarily tied to the party leader that presents him- or herself as being an embodiment of

the people and by promising to take action to face the current threats. This might not be directly represented as emoji on Facebook, but it translates well and steers people to expressions of ‘love’ as a reaction. Therefore, we can particularly expect ‘love’ to stand out in the positive sense, reflecting pride, loyalty and support for the party leader. Similarly, negative elicited emotions are easily translated in selecting the anger emoji in the Facebook architecture, both supportive and dismissive of the message. Considering sarcasm, we expect that populist elicits more haha emojis, which are most likely to be supportive to the MPs post.

Altogether, we thus formulate three expectations.

H4: Facebook page post of MPs belonging to populist parties receive a higher proportion

of angry emoji’s compared to posts of MPs of other parties.

H5: Facebook page post of MPs belonging to populist parties receive a higher proportion

of haha emoji’s compared to posts of MPs of other parties.

H6: Facebook page post of MPs belonging to populist parties receive a higher proportion

of love emoji’s compared to posts of MPs of other parties.

Data and methods

Case selection: Austria, the Netherlands and Sweden

All three cases, namely Austria, the Netherlands and Sweden, are middle-sized European countries with proportional representation and similar electoral systems, and they all have populist parties. These three cases largely function as prototypical cases of European proportional democracy systems. Findings that are stable across these three countries can therefore be expected to be visible in other countries with similar political systems that have, or will see, a rise in populist parties.

Facebook data

We collected information on all MPs from the three countries, as well as whether they had a

Facebook account or a Facebook page. We then collected all posts and emoji reactions per post from the MPs with an active Facebook page. Facebook’s pages provide an open public network structure allowing users to like or follow a page to get updates and react to posts.

Moreover, contrary to a standard Facebook profile, a page gives the politician tools to monitor and track responses in addition to being able to select more functions including verification of account and statistical measures to follow responses and users.

Data was collected for the time period ranging from 1 November 2017 until 1

February 2018. This period was chosen because no electoral campaigns were held at that time in any of the three countries. For all MPs with a page we collected their Facebook handle (i.e. account) and per handle we automatically collected data on the activity of the Page via

Netvizz. The maximum number of post the application can process is 999 per page (Rieder,

2013), but this number was never reach in that period, indicating that we collected all relevant

Facebook posts, totalling 10,355. Of these posts we know how many likes as well as other reaction emoji’s the post received.

Dependent variables and models

Regarding adoption, we focus on whether the MPs have a simple Facebook profile or not and whether they have a more professional Facebook page or not. As these two dependent variables are dichotomous, we use logit regression models. As part of the question whether populist MPs use Facebook to the same degree, we also include the number of posts as a separate dependent variable for those MPs with a Facebook page. The number of posts is distributed approximately normally, so we can use the easiest to interpret type of regression models: OLS.

Regarding the elicited emotions, we zoom in on the emotional reactions to posts measured as the relative number of emoji’s expressed as either ‘like’, ‘angry’, ‘love’, ‘haha’, ‘wow’ or

‘sad’ emoji. In other words, we took the number of a specific reaction and divided that by the total number of emojis on a post. As the proportion might also be a function of the attention a post has gotten overall, we control these models for the total number of reactions too (see below). The emoji data are analyzed with multilevel linear regression models, which captures the nested structure of the data and thus the inter-profile variability. In these analyses, the individual posts are the units of analysis. Table 1 presents the descriptive statistics on our dependent variables.

Explanatory variable: populist party family

The main focus of this study is to examine if and how populist politicians differ from other parties. The MPs are classified based on the party they represent in parliament. The parties are classified as populist in accordance with the common definition of these parties in the literature (e.g. Jacobs & Spierings, 2018; Rooduijn, 2018). In the Netherlands; the Party for

Freedom (PVV), Forum for Democracy (FvD)i and the Socialist Party (SP) are considered to be populist; in Austria, The Freedom Party of Austria (FPÖ) is considered to be populist; and in Sweden, we consider the Sweden Democrats (SD) to be populist.

Control variables

The existing literature on adoption has shown that in many respects politicians of so-called

‘post-materialist’ parties differ from those of the mainstream and smaller non-populist parties, often being more prone to use social media and do so professionally (e.g., Gibson &

McAllister, 2015; Spierings & Jacobs, 2018). Including the post-materialists in the reference

group might then lead to bias our results in favor of the hypotheses. Therefore, we distinguish between MPs of post-materialist parties and all other non-populist parties. Following the common definition in the literature (Bolin, 2016), GroenLinks (GL), the Party for the Animals

(PvdD) and Democrats 66 (D66) are considered to be post-materialist in the Netherlands, List

PILZ in Austria,ii and the Green party (MP) in Sweden.

Additionally, each model is controlled for the seniority of politicians (measured as days in parliament), as well as age and sex (see e.g. Xenos et al., 2017). Younger MPs are often early adopters of social media platforms and the same holds for those with less experience in politics (Larsson & Kalsnes, 2014). For the models on the emojis we also include the total number of reactions to a post, as the number of a certain type of emoji might be a function of the number of followers, views and other elements captured by the total number of reactions on a post. Lastly, variations between countries is accounted for by including countries as a separate level. With only three observations we can obviously not include country-level explanatory factors, but including a random intercept at the country- level does filter out country-level differences – this approach is equivalent to fixed effects models.

Table 1. Descriptive statistics

Mean Std Min- Max Adoption FB Page (active) 0.42 0.495 0,1 FB Profile 0.82 0.388 0,1 Number of posts 36.0 39.2 0- 217 Emotional response (#) Anger 8.25 69.7 0- 3 645 Love 10.26 52.0 0- 1 585 Haha 2.70 19.9 0- 661 Wow 1.15 6.8 0- 202 Sad 3.46 31.6 0- 1 625 Likes 49.54 205.0 0- 4 109

Results

There is quite some diversity in the adoption of open Facebook pages (ranging from 40% in

Sweden to 65% in the Netherlands). Table 2 presents the adoption results and clearly shows that populist MPs use Facebook pages and profiles less than other MPs. The differences are statistically significant and most strongly for Facebook pages in particular. While there are differences in uptake between the countries (see fixed-effects in Table 2), additional models show that the core negative relationship between belonging to a populist party and Facebook uptake only differ slightly between the countries. In all three cases, representing a populist party shows a significant negative effect (see Appendix).

Table 2. Uptake of Facebook pages, profiles (logistic regression) and number of posts

(OLS)

Model 1 Model 2 Model 3 FB Page FB Profile Number of posts B coefficient B coefficient B coefficient Populists -1.239 *** -0.763 ** 18.215 ** Post-materialist -0.251 -0.339 -11.962 Female MPs -0.090 0.100 -0.615 Age -0.032 *** -0.018 0.216 Seniority -0.012 -0.072 *** -0.195 Austria 1.257 *** -2.542 *** 13.239 * Netherlands 1.229 *** -1.757 *** -2.259 Intercept 0.955 * 4.446 *** 22.394 *

*p < 0.05 **p < 0.01 ***p < 0.001

Model 3 in Table 2 additionally shows that if MPs of populist parties do have Facebook pages, they are more active on the other hand. They have a higher number of posts than other

MPs, even controlling for gender, age, seniority and country. The mean number of posts for

MPs belonging to populist party was 109 compared to 32 for post-materialist party representatives and 74 for other party representatives. In sum, populist MPs use Facebook

pages to a lesser extent but when they do use them, they are significantly more active. These results can indicate that populist MPs are overall dissimilar to other MPs and do post more, but given the lower uptake it can also indicate a selection effect whereby among populist parties only the most vocal or professional MPs have a Facebook page. Of other parties, the less vocal and professional MPs are active with a Facebook page too. Zooming in on the party leaders will shed more light on this issue and taking the result discussed below into account suggests Hypothesis 1b (on party structure) gets more support than H1a (on ideology).

Populist party leaders

All party leaders in parliament have active Facebook pages; hence, populist party leaders do not differ in that respect. They are, however, more active than party leaders from other parties.

Party leaders from post-material parties are much less active compared to the rest. The mean number of posts from party leaders from populist parties (N=5) was 114 and the mean for post-material party leaders was only 22 (N=3).iii For the rest of the party leaders (N=12) the mean was 84, which is lower than for populist party leaders. There seems to be a difference here, but the samples are too small for statistical testing.

To see if the proliferate posting by populist leaders is robust, we look at the variation among populist party leaders, for instance to see if the higher mean is not driven by only one leader. We do see considerable variation among party leaders from populist parties, but a clear pattern remains. In the Netherlands, Wilders (PVV) posted 178 times in the studied period (#4 most posts in our entire sample of MPs) and Marijnissen (SP) 93 times, whereas

Baudet (FvD) only had five posts. The FPÖ deputy party leader Johann Gudenus (and leader of FPÖ’s parliamentary group) in Austria had the highest number of posts in the period studied, with 217 posts. Finally, the party leader from the Sweden Democrats (SD) posted 77 times. Overall, particularly leaders of populist parties seem to make use of Facebook pages 17

(except for Baudet in our sample) supporting Hypothesis 2, but more studies are needed to delve deeper into these mechanisms, preferably with larger samples.

Emotions

Having more insight into MPs Facebook adoption, we now shift our attention to the second part of this study: the emotions evoked. Table 3 shows few significant relationships with emotional responses as the control variables, but post-materialist MPs had significantly more love responses, whereas female MPs gain more ‘love’ on Facebook but significantly less

‘haha’.

Crucially, Table 3 also shows that populist MPs’ mass communication garners more and stronger emotions: anger, love, haha, wow and sad were all positively correlated with populism. So not only did we find a much larger number of likes on the posts of populist MPs

(on average 253 compared to 121 for post-materialist politicians and 128 for other parties’

MPs), but more of the people responding (in both absolute and relative terms) took the effort of not giving a default like, but actually selected a more specific emoji. In other words, despite people’s general tendency to use the simple like, the stronger emotional reactions overpowered the more subtle like option for populists’ posts quite clearly. With the exception of sadness all the coefficients are also statistically significant. Hypothesis 3 is thus supported by these results.

Furthermore Table 3 shows that there are considerable differences between which emotions are elicited. While populist actors evoke more emotional responses in all forms

‘Anger’ stands out, with a coefficient that is roughly four times as large as those of the other emojis. Following at a considerable distance are the ‘Haha’ and ‘love’ emoji’s, with ‘love’ being less strongly significant, indicating greater variance. It thus seems that the posts from populist actors primarily evoke feelings of anger and humor or sarcasm (‘haha’), as well as love, which we theorized on the grounds of pride or loyalty.

Table 3. Emotional charge: Number of Anger, love, Haha, Wow, Sad or likes emoji’s

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Anger Love Haha Wow Sad Likes B B B B B B coefficient coefficient coefficient coefficient coefficient coefficient Populists 3.963*** 0.990* 0.928*** 0.255* 0.088 -8.731*** Post-materialist 0.490 1.626*** 0.068 0.281 0.168 -0.775 Another party Ref Ref Ref Ref Ref Ref Female MPs -0.690* 1.306*** -0.505*** -0.168* 0.067 -2.181 Age -0.002 -0.026 -0.006 0.005 0.007 -0.096 Seniority 0.052 0.079* 0.008 0.009 -0.007 -0.198 Reactions (count) 0.000* 0.001*** 0.000* 0.000 0.000 0.002***

Intercept 2.242** 2.779 1.854*** 1.326*** 1.600*** 86.833*** Variance Country level 0.201 3.890 0.051 0.034 0.109 29.453 MP level 2.721*** 4.506*** 0.820*** 0.163*** 0.847*** 153.788***

BIC 73920.99 60469.81 57201.91 50788.58 62860.08 92613.679 *p < 0.05 **p < 0.01 ***p < 0.001

The results for the elicited emotional reactions are in line with our expectations (Hypotheses

4-6), but the prominence of ‘love’ was not as strongly linked to populist messages as anger or laughter. As the latter can be related to both humor and sarcasm this result is explored deeper below by zooming in on the posts receiving most ‘haha’ responses.

Exploring populist postings receiving more laughter

Wilders’ two posts with the most ‘haha’ responses are not simply humorous. While he makes a humorous comment about an MP from his own party in one, he writes sarcastically about other politicians and immigrants in the other (341 and 302 haha’s respectively). Laughter is thus also evoked against the other politicians and outsiders by making clear how ridiculous some situations are and by stressing that other parties do not take this seriously. In a way it could even be said that this humor paves the way for anger.

Figure 1. Two posts from Geert Wilders Facebook page

Note: Text says ‘Fleur sticks out her tongue!’ and ‘Eight Eritreans in a drunken state. Probably on the way to the New Year's drink of D'66 Amsterdam at the invitation of Alexander Penthouse.’, with the headline of the news article reading ‘Police takes drunken migrant from truck at highway truck stop: “They consumed some of the beer being transported”.’ (Wilders, 23/01/2018 and 21/01/2018).

In Sweden, the post with the most ‘haha’ reactions was also the one with the angriest ones, again suggesting the interrelatedness of the two emotions. The party leader of the Sweden

Democrats invited the other party leaders to talk about healthcare, being described as an

‘acute crisis’, but the other parties declined this invitation and two did not reply. The picture in the post contain these answers (Figure 2.). The anger can be interpreted as a critique of the other parties not being serious about fixing problems relating to healthcare due their refusal to deal with the Sweden Democrats, a cordon sanitaire practice which has been upheld since the party entered parliament (see e.g. Strömbäck & Jungar, 2016). ‘Haha’ seems a bit more out of place here, but could be interpreted as a mockery of the other parties’ responses and portraying the others almost behaving as children. Åkesson explicitly cites their e-mails with quotes like ‘We thank you for your invitation’ and ‘We do not intend to work with Jimmie

Åkesson’.

Figure 2. Post from the Sweden Democrats party leader

‘Last week, Jimmy Åkesson invited the other party leaders together with the spokespersons for healthcare, to discuss the urgent situation of Swedish healthcare and what we together can do as soon as possible to change this negative development. Here are their answers:’

When Jimmie Åkesson invites the other party leaders to talk about the acute crisis in healthcare, they answer:

“The prime minister does not intend to participate.”, “We decline your invitation.”, “We collaborate and develop our policies with the Alliance so we decline your invitation.”, “Thank you for the invitation. However, we decline to participate.”, “We do not intend to carry out dialogue with Jimmy Åkesson.”, “(MP and V did not even reply.)” ’

Note: Posted by Jimmie Åkesson 12/12/2017

In Austria, the post with the most ‘haha’ reactions contains a link to a news article with the headline: ‘For children abroad: Kern wants to cut family allowances.’ In the post Gudenus writes ‘Like this, mister Kern?’ (Figure 3. The post to the left). We interpret this as Gudenus basically supporting the measure, while at the same time pointing to a perceived inconsistency in the position of SPÖ (Kern). As it is not formulated as a joke or pun but targeting the political opponent (Kern), it can be interpreted as a kind of gloating or malicious delight.

Similarly, the post with the angriest reactions targets a Social Democratic politician seen at what is claimed to be a radical left demo. Gudenus writes ‘Where will you run into social democratic [SPÖ] politicians? At radical left demonstrations.’ The post can be argued to eliciting anger as a form of blaming the Social democrats of betraying the people. At the same time it received almost a third ‘haha’ emoji’s compared to the angry ones (911 versus 289). It would thus seem as these two emotions partly go hand in hand and that the elicited emotions

from these posts are rooted in related feelings (disappointment or perceived betrayal of conventional parties and the policies they represent).

Figure 3. Two posts from the FPÖ deputy party leader

Note: Text says "Like this, mister Kern?" and "Where will you run into social democratic [SPÖ] politicians? At radical left demonstrations." Posted 14/01/2018.

Conclusion and discussion

This study set out to examine to what extent populist MPs use Facebook and whether they elicit more emotional responses, two core elements undergirding the claim that social media is especially beneficial to populists, partly because of the latter’s emotional loaded language and the direct connection with the people.

Given their leader-focused and tightly controlled party structure, we expected populist

MPs to be less likely to have a Facebook profile and page. This turned out to be the case: populist MPs were less likely to have a Facebook profile or page. However, if they did have an account, they were more likely to post more messages. This difference was partially due to the populist leaders, as their pages were among the most active. Afterwards we zoomed in on the degree and type of emotions these messages triggered. Overall, posts on populist pages

elicited more emotional reactions. Among these emotional reactions, anger stood out. It was much more likely to be elicited by populist MP’s Facebook posts compared to their non- populist counterparts. Next to anger, ‘haha’ and ‘love’-reactions were also clearly more likely to occur. Together with results from prior work that populists post more emotionally loaded messages (Engesser, Fawzi, Larsson, 2017; Heiss & Matthes, 2019; Wirz 2018), these results lead to the overall observation that populist MPs, particularly their rather prolific leaders, use

Facebook to activate people’s emotions. In this respect our framework connecting social media architecture to populist ideology and party structures was useful to create a better understanding of populists’ Facebook uptake and reception. Our results are not conclusive, however, and also point to some specific follow up questions to further complement the chain from uptake to impact. Most prominently, future research could dig deeper into which elements of posts trigger specific emotions and whether these responses mainly come from supporters of opponents. This is currently close to impossible to establish with the accessible information on Facebook.

More generally, this study contributes to our understanding of the role of emotional engagement on social media in populist mass self-communication (Hameleers, 2018a).

Facebook communities take part in actively shaping the populist discourse on social media.

Emotions seem to have a significant influence in building ties between the populist politician and his or her followers. Moreover, it can be expected that politicians are receptive to the types of messages receiving the most engagement on their social media pages and thus purposely post messages that will attract similar attention. If this is the case, then a self- reinforcing cycle of a certain type of messages will play out, where the ‘people’s platform’ gives populists access to generate outrage directed towards perceived threats of immigrants, refugees or the ‘corrupt’ elite.

In terms of scope, our analyses covered three countries with similar political and media systems: Austria, the Netherlands and Sweden. Our findings were robust across these countries and thus likely to be generalizable to other countries with such systems. Future studies can examine different media and political systems to check whether our findings can also be generalized to other settings.

Regarding media systems, one can expect that Facebook is used as an alternative to traditional media to an even larger extent in more polarized pluralist or liberal media systems. When populists are outsiders and have limited access to traditional media, Facebook can be an especially promising and powerful alternative communication channel.

As for political systems, one can expect that populists also use Facebook more in presidential systems. Indeed, we found that the populist leader was more likely to be on

Facebook than the backbenchers were. This pattern is likely to be even more pronounced in settings that have a stronger focus on leaders, such as presidential systems. Countries with presidential systems and polarized pluralized and liberal media systems are thus important cases for future studies of populists' Facebook use.

In addition to expanding the geographical scope, future studies can also apply a perspective combining platform architecture with populist ideology and party structure to other types of communication. While our study suggests that populist MPs use some elements of Facebook differently than other politicians, it remains to be seen whether this holds for other features. For instance, are they more likely to use micro-targeting and who do they target? Which types of posts do they send and does that content differ for instance by containing more disinformation? If one wants to examine the communication of parties who claim to be speaking directly to the people, one needs to examine their communication on what we have described as the ‘people’s platform’: Facebook.

References

Bracciale, R., & Martella A. (2017). Define the populist political communication style: the

case of Italian political leaders on Twitter. Information, Communication & Society

20(09), 1310–1329.

Bodlos, A., & Plescia, C. (2018). The 2017 Austrian snap election: a shift rightward. West

European Politics, 41(6), 1354-1363.

Bolin, N. (2016). Green parties in Finland and Sweden: Successful cases of the North?. In: Van

Haute, E. (ed.). Green parties in Europe. Milton: Routledge.

Bossetta, M. (2018). The Digital Architectures of Social Media: Comparing Political

Campaigning on Facebook, Twitter, Instagram, and Snapchat in the 2016 U.S.

Election. Journalism & Mass Communication Quarterly, 95(2), 471-496.

Buzogány, A., & Scherhaufer, P. (2018). Austrian Greens: from pyrrhic presidential victory to

parliamentary exit. Environmental Politics, 27(3), 566-571. de Vreese, C. H., Esser, F., Aalberg, T., Reinemann, C., & Stanyer, J. (2018). Populism as an

Expression of Political Communication Content and Style: A New Perspective. The

International Journal of Press/Politics, 23(4), 423–438.

Dolezal, M. (2015). Online campaigning by Austrian political candidates: Determinants of

using personal websites, Facebook, and Twitter. Policy & Internet, 7(1), 103–119.

Engesser, S., Ernst, N., Esser, F., & Büchel, F. (2017a). Populism and social media: how

politicians spread a fragmented ideology, Information, Communication & Society, 20(8),

1109-1126.

Engesser, S., Fawzi, N., & Larsson, A. (2017b). Populist online communication: introduction

to the special issue. Information, Communication & Society, 20(9), 1279-1292.

Ernst, N., Engesser, S., Büchel, F., Blassnig, S., & Esser, F. (2017). Extreme parties and

populism: an analysis of Facebook and Twitter across six countries. Information,

Communication & Society, 20(9), 1347-1364.

Ernst, N., Blassnig, S., Engesser, S., Büchel, F., & Esser, F. (2019). Populists Prefer Social

Media Over Talk Shows: An Analysis of Populist Messages and Stylistic Elements Across

Six Countries. Social Media + Society.

Fieschi, C., & Heywood, P. (2004). Trust, cynicism and populist anti‐politics. Journal of

Political Ideologies, 9(3), 289-309.

Gerbaudo, P. (2018). Social media and populism: an elective affinity?. Media, Culture &

Society, 40(5), 745-753.

Gibson, R., & McAllister, I. (2015). Normalising or equalising party competition? Assessing

the impact of the web on election campaigning. Political Studies, 63(3), 529–547.

Grill, C. (2016). How anti-European, populist parties campaigned in the 2014 EP election. In

C. Holtz-Bacha (Ed.), Europawahlkampf 2014. Internationale studien zur Rolle der

Medien (pp. 75–96). Wiesbaden: Springer VS.

Hameleers, M., Bos, L., & de Vreese, C. H. (2017). “They did it”: The effects of emotionalized

blame attribution in populist communication. Communication Research, 44(6), 870–900.

Hameleers, M. (2018a). A Typology of Populism: Toward a Revised Theoretical Framework

on the Sender Side and Receiver Side of Communication. International Journal of

Communication, 12(20).

Hameleers, M. (2018b). Augmenting polarization via social media? A comparative analysis of

Trump’s and Wilders’ online populist communication and the electorate’s interpretations

surrounding the elections. Acta Politica. https://doi.org/10.1057/s41269-018-0119-8

Heinisch, R., & Mazzoleni, O. (2016). Understanding Populist Party Organisation. The radical

right in Western Europe. London: Palgrave MacMillan.

Heiss, R., & Matthes, J. (2019). Stuck in a Nativist Spiral: Content, Selection, and Effects of

Right-Wing Populists’ Communication on Facebook. Political Communication, 1-26.

Jacobs, K., & Spierings, N. (2016). Social Media, Parties, and Political Inequalities. New

York: Palgrave MacMillan.

Jacobs, K., & Spierings, N. (2018). A populist paradise? Examining populists’ Twitter adoption

and use. Information, Communication & Society, 1-16.

Jagers, J., & Walgrave, S. (2007). Populism as political communication style: An empirical

study of political parties’ discourse in Belgium. European Journal of Political Research,

46(3), 319–345.

Jungherr, A., Schroeder, R., & Stier, S. (2019). Digital Media and the Surge of Political

Outsiders: Explaining the Success of Political Challengers in the United States, Germany,

and China. Social Media+ Society, 5(3), 2056305119875439.

Kreiss, D., Lawrence, R. G., & McGregor, S. C. (2018). In Their Own Words: Political

Practitioner Accounts of Candidates, Audiences, Affordances, Genres, and Timing in

Strategic Social Media Use. Political Communication, 35(1), 8–31.

Kriesi, H. (2014). The populist challenge. West European Politics, 37(2), 361–378.

Larsson, A. O., & Kalsnes, B. (2014). ‘Of course we are on Facebook’: Use and non-use of

social media among Swedish and Norwegian politicians. European Journal of

Communication, 29(6), 653-667.

MacMillan, C. (2017). Welcome to the carnival? Podemos, populism and Bakhtin’s

carnivalesque. Journal of Contemporary European Studies, 25(2), 258-273.

Milburn, K. (2019). The comedian as populist leader: Postironic narratives in an age of cynical

irony. Leadership, 15(2), 226-244.

Mudde, C. (2004). The populist zeitgeist. Government and Opposition, 39(4), 542–563.

Mudde, C. (2007). Populist radical right parties in Europe. Cambridge: Cambridge University

Press.

Rieder, B. (2013). Studying Facebook via data extraction: the Netvizz application. In

Proceedings of the 5th annual ACM web science conference (pp. 346-355). ACM.

Rooduijn, M. (2015). The rise of the populist radical right in Western Europe. European

View, 14(1), 3-11.

Rooduijn, M. (2018). What unites the voter bases of populist parties? Comparing the electorates

of 15 populist parties. European Political Science Review, 10(3), 351-368.

Sandberg, L. A. C., & Öhberg, P. (2017). The role of gender in online campaigning: Swedish

candidates’ motives and use of social media during the European election 2014. Journal of

Information Technology & Politics, 14(4), 314-333.

Spierings, N., & Jacobs, K. (2018). Political parties and social media campaigning. Acta

Politica, 1-29.

Stier, S., Bleier, A., Lietz, H., & Strohmaier, M. (2018). Election Campaigning on Social

Media: Politicians, Audiences, and the Mediation of Political Communication on Facebook

and Twitter. Political Communication, 35(1), 50–74.

Strömbäck, J., & Jungar, A. C. (2016). Sweden: No Longer a European Exception. In Populist

Political Communication in Europe (pp. 78-92). Routledge.

Taggart, P. (2000). Populism. Buckingham: Open University Press.

Thompson, D., Mackenzie, I. G., Leuthold, H., & Filik, R. (2016). Emotional responses to irony

and emoticons in written language: evidence from EDA and facial EMG.

Psychophysiology, 53(7), 1054-1062.

Tromble, R. (2018). Thanks for (actually) responding! How citizen demand shapes politicians’

interactive practices on Twitter. New media & society, 20(2), 676-697.

Vergeer, M., & Hermans, L. (2013). Campaigning on Twitter: Microblogging and online social

networking as campaign tools in the 2010 general elections in the Netherlands. Journal of

Computer-Mediated Communication, 18(4), 399-419.

Wirz, D. S. (2018). Persuasion Through Emotion? An Experimental Test of the Emotion-

Eliciting Nature of Populist Communication. International Journal of Communication, 12,

25.

Zulianello, M., Albertini, A., & Ceccobelli, D. (2018). A Populist Zeitgeist? The

Communication Strategies of Western and Latin American Political Leaders on Facebook.

The International Journal of Press/Politics, 23(4), 439-457.

Xenos, M. A., Macafee, T., & Pole, A. (2017). Understanding variations in user response to

social media campaigns: A study of Facebook posts in the 2010 US elections. New media

& society, 19(6), 826-842.

Appendix Table 1. Uptake of Facebook pages and profiles in the Netherlands, Sweden and Austria (logistic regression) NL SE AU (B) (B) (B) Active page Populist -2,547*** -1,135** -0,727* Post materialist -0,786 0,045 0,122 Female 0,111 -0,164 -0,167 Seniority (years) 0,092 0,004 -0,072* Age -0,057** -0,034** -0,020 Intercept 3,311*** 0,928 1,826* Profile Populist -2,059*** -1,018** -0,003 Post materialist -0,962 0,32 -0,953 Female 0,297 -0,146 -0,444 Seniority (years) -0,107* 0,024 -0,058* Age -0,042 -0,034** -0,009 Intercept 4,548*** 5,841*** 1,384 *p < 0.05 **p < 0.01 ***p < 0.001

i FvD was not represented in parliament at the time data were collected for these seminal studies classifying the Dutch parties; it is considered populist in scientific blogs though. E.g.: http://stukroodvlees.nl/zijn-vnl-fvd-en-geenpeil-populistisch/ ii Some have suggested that PILZ is left-wing populists (Buzogány & Scherhaufer, 2018), while others emphasize PILZ as a spinoff from the Green party, also because Peter Pilz, the party’s founder was also founding member of and MP for the Green party (Bodlos & Plescia, 2018). The support of Green and SPÖ MPs for PILZ and its spinoff status, made us consider the party as post-materialist rather than populist. iii The FPÖ was in government and so was the party leader, we therefore included the deputy party leader, Gudenus, as our focus is on MPs. If we instead calculate the mean replacing Gudenus with the party leader, Strache, the mean for populist party leaders is still highest at 95.

Socially Mediated Issue Ownership

Abstract: Given the growing importance of issue competition, and the growing use of social media during elections, this study seeks to create a better understanding of how issue dynamics relating to political parties play out on social media. It tests whether issue ownership theory can explain how parties and issues are being discussed on Twitter, and to what extent a mediated form of issue ownership aligns with citizens’ perceptions of issue ownership. The results indicate that perceptions of issue ownership as measured in representative surveys correlate with variations of what issues parties are linked with on

Twitter. Some deviations also emerged, which possibly reflect short-term changes in parties’ issue competition. Understanding how issue ownership mediates through social media platforms is important in order to evaluate the role of social media in contemporary opinion forming processes, and sheds light on the issue competition among political parties in online fora.

Keywords: Issue ownership, public opinion, social media, Twitter

Introduction

Party politics in Western Europe has become issue-based which implies an increased focus on how parties present their policy views in the media (Green-Pedersen, 2019). Voters’ perceptions of important issues and which parties they trust to handle a particular issue the best (issue ownership), are crucial for parties to ensure electoral support (Van der Brug, 2004;

Green-Pedersen, 2007; Bélanger and Meguid, 2008). Factors determining the outcome of this issue competition are to some extent relatively stable, such as the structure of party competition (Ansolabehere and Iyengar, 1994), but also dependent on other unpredictable and short-term factors, such as media attention, key events, and skillful political communication

(Walgrave and De Swert, 2007; Walgrave, Lefevere, and Nuytemans, 2009). Perception of issue ownership is therefore not static, and the increased influence of social media in political information and news flows make these processes even more dynamic. As social media and other web-based platforms contribute to opinion formation and agenda setting, these processes can no longer be understood as communication that merely passes through press organizations to affect the opinions of citizens. Twitter is a particularly attractive venue for parties to influence media coverage and thus an arena where we can expect dynamics of issue competition to be observable, in terms of both stability and change.

Twitter has become important in the political agenda setting as it influences journalists’ news judgment (McGregor and Molyneux, 2018). The intermedia agenda-setting dynamic of political news sharing that is integrated on Twitter and in news networks, makes an interesting case for examining the dynamics of issue competition. This article presents and argues for a socially mediated dimension of issue ownership, which should be understood in relation to the nature of political discussion on Twitter.

While recent work makes important contributions to the issue competition literature, analyzing parties’ issue strategies on Twitter (Van Ditmars, Maggini, and van Spanje, 2020;

Vaccari, Smets, and Heath, 2020), the discursive linkage between parties and political issues, has not yet been examined. In bringing together the concept of issue ownership with the idea of a social media logic that intermediates parts of public opinion and traditional media content, this paper measures a socially mediated issue ownership dimension. This dimension captures a discursive form of ownership by focusing on which issues parties are discussed together with on Twitter. Studying this “receiver” side of political messages further provides information about whether parties are successful in pushing their political agendas on social media.

Traditional mass media coverage has a significant influence on the perceived link between parties and certain issues, influencing both issue ownership stability and change

(Tresch and Feddersen, 2019). News media coverage is also a strong predictor of what issues are likely to evoke discussion on Twitter. The intermedia dynamics between news media and

Twitter has been suggested to follow a logic in which Twitter is more likely to influence news medias’ agenda in terms of breaking news, whereas news media are more likely to lead

Twitter’s agenda in terms of ongoing discussions (Su and Borah, 2019). This mutual responsiveness to the news media agenda may therefore represent a condition where survey and Twitter data are particularly likely to correspond (Pasek et al., 2018). This implies that citizen’s perceptions of issue ownership are reflected in discussions on Twitter as a result of news media’s influence. However, parties themselves use Twitter strategically to communicate their issue positions and priorities, and to influence their media coverage (Van

Ditmars, Maggini, and van Spanje, 2020). Additionally, the direct feedback provided through social media facilitate party responsiveness to discussions concerning how the party is perceived as well as which issues evoke the most engagement (Barberá et al., 2018). Thus, it is of interest to examine not only if but also how the aggregated agenda on social media deviates from perceptions of party issue ownership as understood through representative

3 surveys of citizens. This article therefore seeks to identify and assess the influences of mediating factors on issue ownership. In so doing, the following empirical research questions are addressed: To what extent are voters issue ownership perceptions correlated with discursive patterns in how parties are linked with the same issues on Twitter? What similarities and differences emerge in the mediation of issue ownership on Twitter?

Examining how an established theoretical concept such as issue ownership plays out on Twitter gives insight to not only the extent social media represents a similar or different issue agenda, but also the potential it holds for observing dynamics of issue competition.

Despite the potential of linking surveys and digital trace data, these two data types have rarely been connected in social science (Stier, Breuer, Siegers, and Thorson, 2019). Compared with surveys, data from Twitter are “naturally occurring” and disclose how expressive behaviors

(such as tweets) provide attention to the issues parties wish to be associated with or not.

Since Twitter is the medium of opinion leaders (Jacobs and Spierings, 2016), this might further yield insights to “elite evaluations” of party performance.

Data used in this study is comprised of Twitter content for eight months leading up to the 2014 Swedish general election. A measure of convergence or divergence in relation to issue ownership perceptions is then established by conducting a secondary analysis of existing survey data for the Swedish general electorate in the Swedish National Election Studies

Program (SNES). Sweden is an appropriate and illuminating case for the exploration of issue ownership on social media because of its highly ideologically motivated voters and historically stable issue ownership perceptions (Martinsson, Dahlberg, and Christensen,

2013). While the mainstream parties have maintained their dominance on socio-economic issues, newer parties have established themselves representing environmental and anti- immigration policies. In contrast to the socio-economic cleavages that up until recently

4 dominated party competition, this post-materialist dimension represented by the “new” parties provides additional insights into issue competition, both online and offline.

The paper will proceed as follows: It will first discuss the concept of issue ownership and related changes in contemporary party competition. From there, it presents a rationale of how a mediated issue ownership dimension can be understood. Sweden as a case study will then be discussed, followed by a presentation of the data and methods used in the study. The results of the study comprise the next section, and the paper closes with a discussion and conclusion.

Party competition and issue ownership

Socioeconomic cleavages no longer condition partisanship in ways that were once the norm.

Some prominent scholars of party systems argue that parties have become disconnected from wider society and that the competition they pursue therefore lacks meaning. In Ruling the

Void (2013), Mair declares that the age of party democracy has passed and that parties are incapable of sustaining democracy in its present form. Several trends are identified that together strengthen the notion of a growing disenchantment with political parties: declining voter turnout, decreased party attachment, and increased electoral volatility. The parties themselves have reacted to these changes by dedicating more of their programs to a wider variety of issues unrelated to the traditional left-right dimension of party competition, such as the environment and immigration (Green-Pedersen, 2007). Parallel to the decline of established political parties and an overall decrease of the traditionally dominant economic cleavage in favor of newer salient issues, challenger parties have emerged. These parties reshape the political landscape by placing new issues on the agenda, such as immigration and

European integration (Kriesi et al., 2008). Contemporary issue competition within a multiparty system is therefore multifaceted; it also plays out on a variety of media platforms.

Since socioeconomic cleavages have become less important in forming partisanship, the explanatory power of social–structural models of voting behavior have declined over the past few decades. Thus, scholars must now closely consider the role of issue preferences in party choice (van der Brug, 2004). Issue ownership is a theoretical framework used for explaining both party and voter behavior. It implies that “parties and their candidates attempt to mobilize voters by emphasizing issues on which they hold a reputation of competence.

Political parties in turn receive support on the basis of those issues that they are perceived to own at election time” (Bélanger and Meguid, 2008, p. 477). The effect of issue ownership is in turn conditioned by the perceived salience of certain issues. A party’s competence in relation to a particular issue only influences voter behavior when the issue is also considered important (Bélanger and Meguid, 2008). The concept of issue ownership has been argued to involve both a competence and associative dimension (Walgrave, Lefevere and Nuytemans,

2009). While the competence dimension reflects that voters consider some party or parties to be better able to deal with specific issues, associative issue ownership is defined as “the spontaneous identification between specific issues and specific parties in the minds of voters”

(Walgrave et al., 2012, p. 771). Both dimensions are considered determinants of voting behavior when combined with the relative importance of issues. Voters may think that one party has the best policy, or they may associate a certain party with a specific issue. However, whether they cast their vote for this party also depends on the relative importance voters attribute to said issue.

Voters’ partisanship tends to dominate perceptions of issue ownership. However, their attitudes and performance evaluations also matter (Stubager and Slothuus, 2013). Perceptions of party reputation on a given issue are relatively stable, but the importance voters ascribe to certain issues is subject to more rapid changes. If parties strategically emphasize certain issues or successfully frame issues that have become salient during the campaign, they can influence

6 perceptions of ownership in the short-term (Blomqvist and Green- Pedersen, 2004; Holian,

2004; Bélanger and Meguid, 2008; Walgrave et al., 2009). Mass media coverage determines perceptions of issue salience to a large extent (agenda setting) but also influence the perceived link between parties and certain issues, particularly in relation to the dynamics of short-term issue ownership (Page and Shapiro, 1992; Thesen, Green-Pedersen, and Mortensen, 2017;

Walgrave and De Swert, 2007). Even if voters’ positions are relatively stable over time, the dynamic nature of issue salience and increased volatile voting behavior render it necessary for parties at each election to ensure they are associated with issues they are trusted to be capable of handling in order to make relevant policy changes.

Recent work has incorporated issue ownership as one component into larger frameworks such as the issue incentive model (Green-Pedersen, 2019) or the issue yield model (D’Alimonte, De Sio, and Franklin, 2020), in order to capture the more dynamic processes behind issue competition. Some of the major criticism of the issue competition literature revolves around a partial neglect of the diverse nature of policy problems that drive attention to issues (Green-Pedersen, 2019), and to what extent parties’ selective issue emphasis implies that other issues are de-emphasized (Dolezal et al., 2014). While acknowledging that issue ownership cannot fully explain parties issue strategies or the issue content of party politics, it can provide an explanation of variation (or a lack thereof) in linkages between issues and parties on various media platforms. Instead of assessing voter evaluations or party strategy, a socially mediated issue ownership dimension emphasizes the interpersonal communication and inter media agenda setting.

Issue ownership theory offers little insight into the role of media or interpersonal political communication in shaping what issues are perceived to be more important (Kiousis,

Strömbäck, and McDevitt, 2015). Contemporary perspectives derived from social media analysis can therefore complement this traditional theoretical framework to create a more

7 comprehensive understanding of issue dynamics around elections. In the next section, this paper will bring together the concept of issue ownership with the idea of a social media logic that intermediates parts of public opinion and traditional media content, what this article refers to as a socially mediated issue ownership dimension.

A socially mediated issue ownership dimension on Twitter

Voters are more likely to operate on the basis of short-term considerations and influences when disengaging from the arena of conventional politics. When the electorate becomes progressively destructured, it affords more scope to the media to play the role of agenda-setter and requires a much greater campaign effort from the parties (Mair, 2013; Green-Pedersen,

2019). This points to an increased importance of media in shaping issue ownership perceptions among the population. The extent to which social media reinforces these processes has yet to be explored as well as what role social media fill in contemporary policy agenda setting. In policy agenda setting, the allocation of attention to issues is understood as an agenda-setting process, since “attention is important for politics because it is consequential and, at the same time, scarce” (Green-Pedersen, 2019, p. 25). Parties therefore compete by presenting their views on the issues dominating the political agenda, which can be understood as a hierarchy of attention to issues existing at a given time (Green-Pedersen, 2019, p. 28).

While parties try to influence the issue agenda, they are at the same time swayed by it. When an issue climbs up the media agenda or becomes prominent on social media, parties need to be responsive. Policy agenda setting dynamics on Twitter are however formed by different mechanisms than the traditional news media agenda.

Issue dynamics on Twitter are shaped by this platform’s specific digital architecture

(Bossetta, 2018), be that technical, spatial (for example limitations in number of characters) or participatory (who involves themselves in political discussions). This might then result in an

8 aggregated issue agenda in which some issues become systematically marginalized, whereas others appear more salient. Previous research has shown that public, political, and media agendas mutually influence each other (Walgrave et al., 2008; Soroka, 2002; Vliegenthart and

Walgrave, 2011). These mutual influences are particularly visible on Twitter as prominent political topics in the public’s mind, in news and on Twitter often correspond (Jungherr,

Schoen, and Jurgens, 2015). Thus, Twitter offers a mediated image of political reality that is highly interconnected with traditional media and public opinion on salient issues.

Due to Twitter’s speed and flexibility, discussions can be expected to be responsive both to changes in the media agenda and public opinion. Issue ownership measurements on

Twitter may therefore reflect short-term changes in issue salience in the media or for the public in general, more noticeably, since those discussing politics on Twitter react to real-time events (Pasek et al., 2018; Jungherr, Shoen, and Jurgens, 2015). Political events and topics covered by mass media are more likely to catch the attention of many users simultaneously, and thus be reflected in pattern shifts in the aggregates of Twitter messages (Jang and Pasek,

2015; Jungherr et al., 2015). Many of the issues covered on Twitter derive from the mainstream media agenda, and therefore Twitter would amplify certain issues depending on the media’s reporting rather than put forth an alternative issue agenda. That is, when some issues are salient in the news media it will evoke discussions on Twitter. Similarly, priming effects of events reported in the news media influence survey responses (Iyengar and Simon,

1993). Traditionally, agenda-setting influence descend from the media or politicians to citizens. On Twitter, however, there is a reciprocal relationship between news outlets and

Twitter content as well as between the whole spectra of parties, politicians, journalists, and regular social media users that is likely to be reflected.

The mediation of issue ownership through social media therefore captures both aspects of interpersonal political communication that occur on Twitter as well as the reciprocal

9 agenda setting relationship with the mass media. This measurement is more closely related to an associative dimension of issue ownership than a competence dimension. What is being measured, however, reflects what issues are discussed and addressed in relation to which parties on social media; hence, it involves the logic pertaining to political discussions on

Twitter. It is therefore more accurate to define measurements of issue ownership on Twitter as socially mediated rather than merely relating to an association. Issue ownership has been identified as a determinant of political parties’ media coverage (Van Camp, 2017; van der

Brug and Berkhout, 2015). On social media, there is no indication as to why this media logic would not resonate similarly. If anything, the social media logic would contribute to a reinforced mainstream agenda and create an accentuated association between issues and parties. In considering how sensitive Twitter discussions are to the news media agenda and newsworthy events, the following expectation has been formulated:

Hypotheses 1: Parties are addressed more frequently in relation to issues they are perceived to own, compared to other issues on Twitter.

Due to the logic of political coverage on Twitter and the above-mentioned characteristics of

Twitter with a discursive component, a fluxional news component, and a political content component, it is expected that Twitter mediates a dimension of issue ownership in a way that is sensitive to short-term changes during election campaigns. Thus, it might reveal issue competition dynamics or variations in the overall policy agenda during election campaigns. It can therefore be expected that parties’ perceived ownership does not reflect similarly for all the issues analyzed. In line with the hypotheses above, the linkages between parties and issues are likely to be accentuated on Twitter for those areas where some parties dominate issue ownership perceptions, compared with issue areas characterized by weaker ownership

10 perceptions. In situations with weak associations between parties and issues, Twitter discussions might therefore reproduce a more ambiguous association in how parties are addressed together with these issues on Twitter. Some deviations might therefore be expected:

Hypotheses 2: Political topics characterized by strong issue ownership perceptions are more likely to be discussed on Twitter, together with the mention of parties "owning" these issues, than issues that are less linked to specific parties.

The issue competition literature contributes with important insights on the importance of the specific type of policy problem (Green-Pedersen , 2019) as well as the distinction between positional and valence issues (D’ Alimonte, De Sio, and Franklin, 2020). Instead of offering a potential explanation of parties’ issue strategies, this second hypothesis is formulated based on expectations as to what extent the Twitter discussions are more likely to be in flux when it comes to issue areas with weaker ownership perceptions. Similar to the mainstream news media agenda, it is likely that the attention is directed towards bigger mainstream parties in economic issues, for example when these parties make statements or present new initiatives, whereas issues characterized by weaker ownership perceptions are addressed in relation to several parties on Twitter.

Case selection: Sweden as a two-dimensional multi-party system

In multiparty systems, voters can have confidence in more than one party to handle a specific issue. Issue ownership might therefore accumulate over several issues, meaning the more issues a party is trusted to handle within a comprehensive policy set the more likely that party is to receive votes (Karlsen and Aardal, 2016). In multiparty systems, it is also possible for several parties to be internally united on a similar issue goal, which could mean multiple parties pursuing ownership of a particular issue (D’ Alimonte, De Sio, and Franklin, 2020). 11

Sweden has historically been seen as one of the most unidimensional political systems since the traditional left–right dimension strongly structured party competition and voting behavior.

In this party-centered system with its highly ideologically motivated voters, parties’ policy positions and voters’ issue standpoints have had large explanatory power in models of voting behavior (Oscarsson and Holmberg, 2015). Left–right ideological predispositions are still strong determinants of voters’ party choice, even though new conflicting issue dimensions, such as immigration, the environment, and gender equality, have gained importance. Judging by its recent elections, Sweden can therefore be characterized by a multidimensional cleavage structure that forms a two-dimensional party system (Berg and Oscarsson, 2015). An overview of how the parties are placed within these two dimensions is presented in Figure 1.

There is a clear divide between parties’ positions on the left–right political scale. On a libertarian–authoritarian scale, the Sweden Democrats (SD) are distinguished by their high placement in terms of authoritarian values, followed at some distance by the Christian

Democrats (KD). Issue ownership in Sweden thus relates to issues on the left–right cleavage structure alongside a second cleavage revolving around issues such as the environment and immigration.

Figure 1. Party positions in Sweden on a two-dimensional, left–right, libertarian– authoritarian scale.

Note: The data presented in this figure is taken from the Chapel Hill Expert Survey (2014) using the variables galtan and lrgen.

In Sweden, few regular citizens use Twitter and even fewer tend to activley partcipate in discussions.1 Political discussions in particular tend to be dominated by a smaller group of active users that make up a substantial part of the activity; these are politicians themselves and established journalists and opinion-makers (Larsson and Moe, 2015). The possibility to reach out to influential actors such as journalists provides Swedish parties strong incentives to use

Twitter stratgically during election campaigns.

Data and methods

The Twitter data used in this paper were gathered from February 2014 to September 2014 using the Twitter Streaming API (Dokoohaki et al., 2015). Three filters were applied when

1 In Sweden, approximately 20 % of Internet users are on Twitter according to The Swedish Internet Foundation: https://svenskarnaochinternet.se/rapporter/svenskarna-och-internet-2018/sociala-medier/

13 gathering the data: one for location, which was set to Sweden, another for political and election-related hashtags (e.g. #svpol), and a third for political party accounts. Approximately seven million tweets were gathered in total. A subsample of data that mentioned any of the

Swedish political parties was used to analyze the dimension of issue ownership on Twitter. Of the tweets that mentioned Swedish parties, the frequency of different topic discussions per party was calculated as a measure of ownership. In order to compare the deviations and similarities in how issue ownership forms on Twitter with how voters perceive ownership, survey data from the Swedish National Election Survey (2014) was used. The question posed in the survey was phrased as follows: “Are there, in your opinion, any party or parties with a good policy on [issue area]?” The analyzed issue areas consisted of nine different policy domains—namely, education, social welfare, the labor market, immigration/integration, taxes, health care, pensions, the economy, and the environment. These issues/policy domains were the most salient issues in the minds of the voters during this election (Statistics Sweden,

2015).

Natural language processing

A dictionary-based automated lexical approach was used to structure data (i.e., the textual content of individual posts) from Twitter. This is a common method within the semantic and computational linguistic to analyze the content of large corpora of text, using either data- driven or bottom-up lexical approaches (Schober, Pasek, Guggenheim, Lampe, and Conrad,

2016). This approach allows for an analysis of the frequency of keywords from a particular semantic category for each post in the corpus (Schober et al., 2016). As part of the data preprocessing stage, the stem and lemma of each word were detected using stemming and lemmatization libraries specific to the Swedish language, separating text into separate words and removing inflectional endings only. A lemma, the base dictionary form of a word, can be

14 used in different ways in a sentence (e.g., as a noun or verb), thus providing different meanings to a word. As a part of speech tagging, each word in a sentence is determined, and it is then possible to analyze tweets mentioning a word as a political topic and not a verb. For example, in the Swedish language, “job” can have several inflections and be used as the verb

“to work.” In addition, stop words were removed from these tweets (dropping common terms).

A manual classification of the preprocessed words in the whole subsample was then conducted. Based on the phrases and words used in the coding scheme of the SNES2 for the nine policy areas, the base dictionary form of the words in the total corpus of tweets was manually matched to these predefined categories of political issues to allow for comparison.

In the coding scheme of the national election survey, the economy for instance has nine sub- issues for coding the answer of survey respondents such as economic growth or banks. For example, the occurrence of “banks” in the Twitter data set corresponds to one of the subcategories of economy. Any preprocessed word that did not correspond to any of the subcategories used to code survey answers was left out3.

Comparing issues between surveys and tweets, capture different aspects of issue ownership. Words on Twitter do not equate to survey respondents’ answers concerning the question of which party has the best policies in a particular area. The competence dimension, as measured in the survey, demonstrates that a positive relationship between party and issue is present. On Twitter, the dimension of issue ownership is more closely related to an associative connection, since the mention of an issue in relation to a party on Twitter can be positive, negative, or neutral. Nonetheless, the assumption is that voters’ perceptions of issue

2 The coding scheme can be found at https://som.gu.se/digitalAssets/1481/1481065_kodbok-riks-som-2013--v1- .pdf (p. 67-69). Data from the Swedish election studies (1956–2018) are deposited at the Swedish National Data Service, www.snd.gu.se 3 Python was used for preprocessing the textual Twitter data whereas the output in the form of base dictionary words was analyzed using Excel and structured per party. 15 ownership will be reflected in the differences between what party is discussed in connection to which issue. In order to analyze how similarly voter’s issue ownership perceptions are mediated on Twitter, the Pearson correlation coefficient was used to compare the two different measurements of issue ownership. Since this approach cannot capture the differences in the two data sets, the results should be interpreted with this in mind. However, the correlation coefficient does signal whether a similar ratio in the relationship between issues and parties on Twitter and in surveys can be found, which is helpful for interpreting these results.

Results

Issue ownership in the 2014 Swedish general election

Table 1 shows the electorate’s perceptions of issue ownership in the 2014 Swedish election.

The Social Democratic Party (S) had the best reputation on most issues and is currently and historically the biggest party. The Conservative Party (M) became the second biggest party after the 2014 election and had the strongest issue ownership on the economy and taxes. The anti-immigration party, Sweden Democrats (SD), became the third biggest party after the election and had the strongest ownership on immigration/integration. The Sweden Democrats’ electoral gains in the 2010 and 2014 elections is likely explained by the increased salience of the immigration issue. The Green Party (MP) was regarded as owning the environmental issue and almost half of the voters perceived the party to have the best policy in this area. Voters thought that the Centre Party (C) had the second best policies for environmental issues. The

Liberal Party (L) has since the 1998 election, maintained issue ownership over education and school (Martinsson, Dahlberg, and Christensen, 2013). In the 2014 election, however, they lost part of their hold on this main profiling issue, with more than a quarter of voters instead

16 thinking that the Social Democrats had the best policies in that domain. The Left Party (V) had the strongest ownership on social welfare (by 20% of the voters). During the 2014 election, the major issue that the Left Party campaigned on was limiting profits for private companies in the welfare sector. The majority of voters were in favor of the Left Party’s position (Nilsson, 2017). The main profiling issue of the Christian Democrats (KD) was health care. However, three other parties experienced higher levels of trust from voters in dealing with this issue.

Table 1. Issue ownership perceptions in the 2014 Swedish election (%).

M C L KD S V MP SD (N) Taxes 40 3 4 2 33 11 4 3 (9644) Environment 7 23 3 1 11 4 48 2 (9883) Labor market 36 5 5 2 37 8 4 3 (9480) Economy 47 2 3 1 32 7 3 4 (9364) Immigration 15 5 10 6 25 11 7 21 (9051) Health care 14 5 9 12 38 14 4 5 (9273) Education 13 4 25 4 31 11 8 3 (9645) Social welfare 20 5 8 6 32 20 4 4 (9302) Pensions 20 4 7 7 40 12 3 7 (8251)

Note: The data originates from the Swedish National Election Survey (2014). The question posed in the survey was phrased as follows: “Are there, in your opinion, any party or parties with a good policy on [issue area]?”

The highest and second highest shares are highlighted. Other parties and missing values are excluded in the analysis. Each row adds up to 100%.

In relative terms, a stronger issue ownership dimension characterizes some issue areas over others. Measured as the standard deviation of each issue as it relates to the parties, issues with weaker ownership perceptions are immigration, education, social welfare, pensions, and health care. For example, ownership over immigration related issues are distributed across

17 more parties and therefore have lower standard deviation.4 Policy areas with relatively strong perceptions of ownership are taxes, environment, labor market and economy.

Issue ownership dimension on Twitter

The analysis revealed a strong issue ownership dimension on Twitter, thus supporting

Hypothesis 1. When the political issues were addressed and coupled with a party, this resulted in 0–47% issue association for a particular party (see Table 2). The Social Democrats and the

Conservative Party dominated on Twitter in discussions of taxes, the labor market, the economy, and pensions. These are issues that they are considered to also “own” in the representative survey. When correlating the two measures, all of these issues had a strong, positive correlation.5 The traditional economic dimension has strongly structured party competition and voting behavior in Sweden and elsewhere, issues usually owned by large mainstream parties (Green-Pedersen, 2019), which is consistent with the pattern on Twitter.

Social welfare, also belonging to a traditional left–right dimension, had the strongest association with the Left Party and Social Democrats (Table 2) on Twitter, which was positively correlated with the survey results.6 On Twitter, the Left Party was mentioned most often when issues concerning social welfare were addressed, particular in relation to discussions of limiting profits in the welfare sector. Searching tweets for the whole sentence

“profits in welfare” gave 139 results in tweets also mentioning the Left Party but only between 0-17 for all other parties. The word “welfare” alone accounted for most mentions of the Left Party with this issue.

4 Standard deviation: Taxes (15.18), environment (16.07), labor market (14.92), economy (17.3), immigration (7.29), health care (11.03), education (10.39), social welfare (10.39) and pensions (12.34). 5 In relation to taxes, the value of R is 0.94, and the value of R2 is 0.88; labor market: R is 0.90, and R2 is 0.80; economy: R is 0.93, and R2 is 0.86; pensions: R is 0.92, and R2 is 0.84. 6 In relation to social welfare, the value of R is 0.91, and the value of R2 is 0.83. 18

Table 2. Issue ownership on Twitter before the 2014 Swedish election (%).

M C L KD S V MP SD (N) Taxes 28 9 3 3 34 9 11 4 (4409) Environment 9 30 4 0 8 5 44 0 (5345) Labor market 47 7 6 2 23 6 7 1 (15306) Economy 30 12 5 1 26 13 9 3 (2065) Immigration 16 4 13 2 9 7 20 28 (6548) Health care 23 6 15 21 21 10 2 3 (1373) Education 16 5 17 3 25 16 17 1 (18326) Social welfare 17 5 9 4 27 30 6 2 (2890) Pensions 27 3 9 11 33 5 4 9 (519)

Note: The table displays the share of party mentions when a particular issue is addressed on Twitter. The highest and second highest shares are highlighted. Each row adds up to a 100%.

Issue ownership over education also displayed a similar pattern on Twitter as in the survey.7

On Twitter, this issue was most strongly associated with the Social Democrats, followed by the Liberal Party and the Greens. As discussed previously, the Social Democrats took over the

“ownership” of education from the Liberal Party, which had ownership over the issue following the 1998 election. This shift in ownership was reflected on Twitter, where the

Social Democrats were clearly associated with the issue. In terms of party mentions and topic discussion, the Social Democrats stood for 25% compared with 17% for the Liberal Party and the Greens.

Health care did not generate as clear differences between the parties as those mentioned above did.8 The Social Democratic Party is considered to have the best policies regarding health care by survey respondents (38%). On Twitter, this association was weaker.

7 In terms of education, the value of R is 0.84, and the value of R2 is 0.71. 8 In terms of health care, the value of R is 0.64, and the value of R2 is 0.42. 19

Health care was discussed equally in relation to the Conservative Party, the Social Democrats, and the Christian Democrats.

Both environment and immigration have emerged as important issues for the electorate, introducing another policy dimension in Swedish politics. The parties who are mobilizing voters on primarily one of these two issues are the Green Party and the Sweden

Democrats, which is reflected in their ownership of these issues. In environmental politics, the

Green Party is considered to have the best policies by 48% of survey respondents. On Twitter, this association was present, and the results show a strong correlation between issue ownership in the survey and on Twitter regarding environmental issues. Both in the survey and on Twitter, the Centre Party came second to owning this issue, with a big gap between them and all remaining parties.9 Immigration/integration is a polarizing topic in Sweden. Not surprisingly, the issue ownership over this issue is also divided: 25% of survey respondents consider the Social Democratic Party to have the best policy, whereas 21% consider the

Sweden Democrats to have the best policy. On Twitter, the Green Party shared ownership with the Sweden Democrats in terms of receiving the most mentions with regard to discussions of this issue, which caused a weaker correlation.10

Overall, there is a strong resemblance between a socially mediatied issue ownership dimension on twitter and ownership perceptions measured in the survey. In seven of nine political areas, the correlation between the two came out as strong and significant. Two topics, however, showed a clear deviation. These were immigration/integration and health care, where one or two parties were mentioned more on Twitter than what would be expected from the survey results. In the national election study, areas with relatively weaker ownership perceptions were immigration/integration, education, social welfare, pensions, and health care

9 In terms of the environment, the value of R is 0.98, and the value of R2 is 0.95. 10 In terms of immigration/integration, the value of R is 0.44, and the value of R2 is 0.19.

(compared to taxes, the environment, the labor market and the economy). Areas with the weakest correlations between the survey and Twitter were immigration/integration and health care, followed by education. This result renders some support for Hypothesis 2 but cannot fully explain the differences in issue ownership between the survey and Twitter.

In understanding why these differences emerge on Twitter, it is relevant to examine these deviations a little closer. In the next paragraphs, the two issue areas displaying the largest differences in ownership will be discussed in light of the Swedish 2014 election.

Issue ownership deviations between the survey and Twitter

Debates revolving around immigration/integration played a major role in the 2014 election. In the news media, topics concerning immigration and refugees featured most prominently during the election campaign for the first time ever (Johansson, 2017). A cordon sanitaire practice of isolating the anti-immigration party, the Sweden Democrats, was also upheld

(Rydgren and van der Meiden, 2018). In terms of the liberal–authoritarian dimension, the

Greens and the Left Party were most strongly opposed to the Sweden Democrats. Before the

2014 election, the Green Party campaigned on an anti-racist agenda. At the time, the Green

Party’s secretary said the party was active in analyzing which issues that resonated best on social media; the party noticed that they generated the most engagement on social media when they opposed the Sweden Democrats. This response then inspired the design of their broader campaign (Stockholm University, 2014). The engagement on social media did not translate into votes, however, and when seen from the perspective of issue ownership, a stronger focus on environmental issues in the election campaign might have garnered better results. Therefore, how the Green party themselves acted to put emphasis on integration/immigration and oppose the Sweden democrats, in particular on social media,

21 might explain why the party was mentioned frequently with this issue, which resulted in a weaker correlation between the survey and Twitter measurement.

In total, 20% of the electorate viewed health care as one of the most important issues:

It placed sixth on the scale of the most salient issues (Statistics Sweden, 2015). Party system attention to health care has increased in several countries and parties tend to focus on this issue regardless of ownership perceptions which make ownership less clear (Green-Pedersen,

2019). The high proportion of tweets mentioning the Christian Democrats in relation to this issue could therefore be explained by the party’s effort to compete over this issue during the campaign. Examples of this can be found in the data, when the Christian Democrats were mentioned together with health care, a large share referred to nationalizing health care, which was the policy position the Christian Democrats took to differentiate themselves from other parties’ position on the issue. The party’s own campaign efforts might therefore explain why they were discussed more in relation to health care on Twitter than what would be expected given the ownership perceptions of this issue.

Discussion and conclusion

This paper has brought together the concept of issue ownership with the idea of a social media logic that intermediates parts of public opinion and traditional media content, forming a socially mediated issue ownership dimension. This dimension of issue ownership is discursive rather than associative or evaluative and reflects mutual influences of public, political, and media agendas on Twitter. The results suggest that using Twitter as a data source holds potential to observe dynamics of issue competition, in terms of both stability and change. In line with the expectations, parties’ perceived competence over issues were largely reproduced on Twitter as some issues were discussed more in relation to a party “owning” that issue. That citizens’ perceptions of issue ownership strongly correlate with the variation

22 of socially mediated issue ownership on Twitter is likely related to the logic of political coverage on Twitter, which in part replicates the mainstream news media agenda in terms of what issues are addressed together with specific parties. Studying communication patterns on

Twitter as the “receiver” side of political messages therefore provides an indication of whether or not parties are successful in pushing their policy agendas. In this respect, Twitter portrays an issue-friendly media agenda (Thesen, Green-Pedersen, and Mortensen, 2017), as the issues parties wish to be associated with corresponded to discursive linkages.

News media coverage strongly influences perceptions of issue ownership but also functions as a strong predictor of what issues are likely to evoke discussion on Twitter.

Therefore, the deviant cases for which perceptions of issue ownership did not align with the discursive patterns make them valuable for assessing the influences of other factors shaping the mediation of issue ownership on this specific platform.

Issues of immigration and health care represented two policy areas where measures of ownership on Twitter did not align as clearly with public perceptions. Tentative explanations as to why some issues were addressed on Twitter more often with some parties that did not

“own” these issues suggests that the parties’ efforts to compete on these issues during the election campaign was a contributing factor. If this explanation holds true, Twitter messages would have a unique potential to evaluate parties’ efforts to compete over issues prior to election. The two deviant cases presented here also indicate that the issue strategies employed by the Green Party and the Christian Democrats differed in terms of cause and effect. While the Christian Democrats seemed successful in their attempts to politicize the health care issue and make it a profiling issue of the party on Twitter, the Green Party instead shifted the party’s issue agenda as a response to what engaged their supporters on social media, only this attempt backfired in terms of issue yield at election time.

Modern election campaigns increasingly revolve around issue competition and gaining trust rather than presenting clear-cut, competing ideologies (Green-Pedersen, 2019). Parties must therefore try to get as much attention as possible for issues in which they are perceived to be the most trustworthy and competent. Mass media coverage has a significant influence on the perceived link between parties and certain issues, particularly for short-term issue ownership dynamics. Since traditional broadcasters have lost part of their hold on forming perceptions of issue ownership, social media opens new ways to influence the issue agenda.

This paper has taken a first step to deepen our understanding of how issue ownership forms on social media and the potential changes that may arise due to parties’ increased efforts in issue competition on these platforms. The results suggest that a socially mediated issue ownership dimension on Twitter captures dynamics of issue competition in terms of both stability and change. Future work on the role Twitter plays in contemporary policy agenda- setting dynamics and issue competition should seek to further establish how much parties’ own efforts explain shifts in the aggregate issue agenda on Twitter.

Reference list

Ansolabehere, S., & Iyengar, S. (1994). Riding the wave and claiming ownership over issues:

The joint effects of advertising and news coverage in campaigns. Public Opinion

Quarterly, 58, 335‐357.

Barberá, P., Casas, A., Nagler, J., Egan, P. J., Bonneau, R., Jost, J. T., & Tucker, J. A. (2018).

Who leads? Who follows? Measuring issue attention and agenda setting by legislators and

the mass public using social media data. American Political Science Review,

https://doi.org/10.1017/S0003055419000352.

Bélanger, É., & Meguid, B. M. (2008). Issue salience, issue ownership, and issue‐based vote

choice. Electoral Studies 27(3), 477‐491.

Berg, L., & Oscarsson, H. (2015). The Swedish general election 2014. Electoral Studies, 38,

91–93.

Blomqvist, P., & Green‐Pedersen, C. (2004). Defeat at home? Issue‐ownership and social

democratic support in Scandinavia. Government and Opposition, 39(4), 587-613.

Bossetta, M. (2018). The Digital Architectures of Social Media: Comparing Political

Campaigning on Facebook, Twitter, Instagram, and Snapchat in the 2016 U.S. Election.

Journalism & Mass Communication Quarterly, 95(2), 471–496.

D’Alimonte, R., De Sio, L., & Franklin, M. N. (2020). From issues to goals: a novel

conceptualisation, measurement and research design for comprehensive analysis of

electoral competition. West European Politics, 43(3), 518-542.

Dokoohaki, N., Zikou, F., Gillblad, D., & Matskin, M. (2015). Predicting Swedish elections

with twitter: A case for stochastic link structure analysis. In Advances in Social Networks

Analysis and Mining (ASONAM), 2015 IEEE/ACM International Conference on (1269-

1276). IEEE.

Dolezal, M., Ennser‐Jedenastik, L., Müller, W. C., & Winkler, A. K. (2014). How parties

compete for votes: A test of saliency theory. European Journal of Political

Research, 53(1), 57-76.

Green-Pedersen, C. (2007). The growing importance of issue competition: The changing nature

of party competition in Western Europe. Political Studies, 55(3), 607-628.

Green-Pedersen, C. (2019). The Reshaping of West European Party Politics, Oxford, Oxford

University Press.

Green, J. & Hobolt, S. (2008). Owning the issue agenda: Party strategies and vote choices in

British elections. Electoral Studies, 27(3), 460‐476.

Iyengar, S., & Adam S. (1993). News coverage of the Gulf crisis and public opinion: A study

of agenda-setting, priming, and framing. Communication research, 20 (3), 365-383.

Jacobs, K., & Spierings, N. (2016). Social Media, Parties, and Political Inequalities. New

York: Palgrave MacMillan.

Jang, S. M., & Pasek, J. (2015). Assessing the carrying capacity of Twitter and online news.

Mass Communication and Society, 18(5), 577-598.

Johansson, B. (2017). ”Medievalrörelsen 2014 Invandringsfrågan i fokus för balanserad

bevakning”. In B. Johansson (red) & L. Truedson (Eds.) När makten står på spel:

journalistik i valrörelser. Stockholm: Institutet för Mediestudier.

Jungherr, A., Schoen, H., & Jürgens, P. (2015). The mediation of politics through Twitter: An

analysis of messages posted during the campaign for the German federal election

2013. Journal of Computer-Mediated Communication, 21(1), 50-68.

Karlsen, R., & Aardal, B. (2016). Political values count but issue ownership decides? How

stable and dynamic factors influence party set and vote choice in multiparty systems.

International Political Science Review, 37(2), 261-276.

Kiousis, S., Strömbäck, J., & McDevitt, M. (2015). Influence of issue decision salience on vote

choice: Linking agenda setting, priming, and issue ownership. International Journal of

Communication, 9 (22), 3347-3368.

Kriesi, H., Grande, E., Lachat, R., Dolezal, M., Bornschier, S., & Frey, T. (2008). West

European politics in the age of globalization. Cambridge: Cambridge University Press.

Larsson, A. O. & Moe, H. (2015). From emerging to established? A comparison of Twitter use

during Swedish election campaigns in 2010 and 2014. In A. Bruns, G. Enli, E. Skogerbø,

A.O. Larsson & C. Christensen (eds.), The Routledge companion to social media and

politics (pp. 311–324). London: Routledge.

Mair, P. (2013). Ruling the void: The hollowing of Western democracy. London; New York:

Verso.

Martinsson, J., Dahlberg, S., & Christensen, L. (2013). Change and stability in issue ownership:

The case of Sweden 1979-2010. In Dahlberg, S., Oscarsson, H., & Wängnerud, L (editors),

Stepping Stones: Research on Political Representation, Voting Behavior, and Quality of

Government. Gothenburg Studies in Politics: 133. ISBN 978-98-89246-59-31.

McGregor, S. C., & Molyneux, L. (2018). Twitter’s influence on news judgment: An

experiment among journalists. Journalism. https://doi.org/10.1177/1464884918802975

Nilsson, L. (2017). Väljare och valda om vinster i välfärden. In Andersson, U., Ohlsson, J.,

Oscarsson, H. & Oskarson, M. (Eds.) Larmar och gör sig till. University of Gothenburg:

SOM-institute.

Holian, D. B. (2004). He’s stealing my issues! Clinton’s crime rhetoric and the dynamics of

issue ownership. Political Behavior, 26(2), 95‐124.

Oscarsson, H.. & Holmberg, S. (2015). Issue voting structured by left–right ideology. In Pierre,

J. (Ed.), The Oxford Handbook of Swedish Politics: Oxford: Oxford University Press.

Page, B., & Shapiro, R. Y. (1992). The rational public: Fifty years of trends in Americans’

policy preferences. University of Chicago Press.

Pasek, J., Yan H. Y., Conrad, F. G., Newport, F., & Marken, S. (2018). The Stability of

Economic Correlations over Time: Identifying Conditions under Which Survey Tracking

Polls and Twitter Sentiment Yield Similar Conclusions. Public Opinion Quarterly, 82(3),

470-492.

Rydgren, J., & van der Meiden, S. (2018). The radical right and the end of Swedish

exceptionalism. European Political Science, 1-17.

Sartori, G. (2002). Homo Videns. Rome: Economia Laterza.

Schober, M. F., Pasek, J., Guggenheim, L., Lampe, C., & Conrad, F. G. (2016). Social media

analyses for social measurement. Public Opinion Quarterly, 80(1), 180-211.

Soroka, S. N. (2002). Issue attributes and agenda‐setting by media, the public, and policymakers

in Canada. International Journal of Public Opinion Research, 14(3), 264-285.

Statistics Sweden. (2015). The Super Election Year 2014. Democracy Statistics Report no 20.

Stier, S., Breuer, J., Siegers, P., & Thorson, K. (2019). Integrating Survey Data and Digital

Trace Data: Key Issues in Developing an Emerging Field. Social Science Computer

Review, https://doi.org/10.1177/0894439319843669

Stockholm University. (2014, October 9). #VemsVal14? Sociala Medier, Journalistik &

Svenska Valet 2014. Seminar at the Department of Media Studies, Stockholm, Sweden.

Stubager, R., & Slothuus, R. (2013). What are the sources of political parties’ issue ownership?

Testing four explanations at the individual level. Political Behavior, 35(3), 567-588.

Su, Y., & Borah, P. (2019). Who is the agenda setter? Examining the intermedia agenda-setting

effect between Twitter and newspapers. Journal of Information Technology & Politics, 1-

14.

Thesen, G., Green-Pedersen, C., & Mortensen, P. B. (2017). Priming, issue ownership, and

party support: The electoral gains of an issue-friendly media agenda. Political

Communication, 34(2), 282-301.

Tresch, A., & Feddersen, A. (2019). The (in) stability of voters’ perceptions of competence and

associative issue ownership: The role of media campaign coverage. Political

Communication, 36(3), 394-411.

Vaccari, C., Smets, K., & Heath, O. (2020). The United Kingdom 2017 election: polarisation

in a split issue space. West European Politics, 43(3), 587-609.

Van Camp, K. (2017). Issue ownership as a determinant of political parties’ media coverage.

Communications, 43(1), 25-45.

Van der Brug, W. (2004). Issue ownership and party choice. Electoral Studies, 23(2), 209‐233.

Van der Brug, W., & Berkhout, J. (2015). The effect of associative issue ownership on parties’

presence in the news media. West European Politics, 38(4), 869–887.

Van Ditmars, M. M., Maggini, N., & van Spanje, J. (2020). Small winners and big losers:

strategic party behaviour in the 2017 Dutch general election. West European

Politics, 43(3), 543-564.

Vliegenthart, R., & Walgrave, S. (2011). Content matters: The dynamics of parliamentary

questioning in Belgium and Denmark. Comparative Political Studies, 44(8), 1031-1059.

Walgrave, S., & De Swert, K. (2007). Where does issue ownership come from? From the party

or from the media? Issue‐party identifications in Belgium, 1991–2005. The Harvard

International Journal of Press/Politics, 12(1), 37‐67.

Walgrave, S., Soroka, S., & Nuytemans, M. (2008). The mass media's political agenda-setting

power: A longitudinal analysis of media, parliament and government in Belgium (1993 to

2000). Comparative Political Studies, 41(6), 814-836.

Walgrave, S., Lefevere, J., & Nuytemans, M. (2009). Issue ownership stability and change: how

political parties claim and maintain issues through media appearances. Political

Communication, 26(2), 153‐172.

Walgrave, S., Lefevere, J., & Tresch, A. (2012). The associative dimension of issue ownership.

Public Opinion Quarterly, 76(4), 771-782.

Wlezien, C. (2005). On the salience of political issues: The problem with ‘most important

problem. Electoral Studies, 24(4), 555‐579.

421 Start Sharing the News Exploring the Link Between Right-wing Alternative Media and Social Media During the Swedish 2018 Election

Linn A.C. Sandberg & Karoline Andrea Ihlebæk

Abstract This study examines the influence of right-wing alternative media on social media during the Swedish 2018 general election, by analyzing links shared on Facebook and the engagement around these links. The results indicate that content originating from such sites had notable visibility on social media in comparison to traditional news-media content. Of the news stories shared in our sample, 28 % originated from right-wing alternative media outlets and the amount of engagement around links shared was comparable to that of mainstream news media. Immigration and integration-related issues dominated the covered content shared from right-wing alternative media. Content in links that evoked highest engagement on Facebook similarly revolved around immigration, crime, and the Sweden Democrats. In this regard, Facebook might contribute to giving voice to contested views of these controversial news sites and the platform, to broaden their reach.

An important question in all democracies concerns who manages to set the agenda during election time. While traditional editorial-driven media still plays a crucial role when it comes to informing people about political and societal issues, the role of social media has enabled new actors to compete for visibility and impact. In this article, we will focus particularly on the extent to which right-wing alternative news media managed to gain visibility and create engagement on Facebook during the election campaign in Sweden, in comparison to established news media. The type of media actors in question includes a variety of amateur or semiprofessional news outlets that aim to represent an alternative to the mainstream-media coverage (Figenschou & Ihlebæk 2018; Holt 2016a). Even though studies have shown that the Swedish mainstream press predominantly frames immigration negatively (Strömbäck Andersson & Nedlund 2017), right-wing alternative media actors often claim that what is perceived as left-wing journalism does not cover immigration fairly

Linn A.C. Sandberg and Karoline Andrea Ihlebæk work at the Department of Media and Communication, University of Oslo. E-mail: [email protected] (corresponding author); [email protected]

Statsvetenskaplig tidskrift · Årgång 121 · 2019 /3 422 Linn A.C. Sandberg & Karoline Andrea Ihlebæk

(Figenschou & Ihlebæk 2018). Furthermore, anti-system and anti-elitist attitudes often characterize such sites, mirroring the rhetoric of far-right populism (Holt 2018). Arguably, the Internet and social-media platforms have lowered the threshold for populist or non-elite actors to enter the arenas of communication (Atton 2004; Engesser, Fawzi & Larsson 2017). The rise of right-wing alternative media can therefore be seen in relation to the growth of far-right populist political actors in Sweden, as well as in Europe and the U.S. (Engesser et al. 2017; Ernst et al. 2017; Haller & Holt 2018; Krämer 2017; Schierup & Ålund 2011). The concern is that by promoting values like anti-elitism and anti-immigration on social-media platforms, the influence from such actors will lead to polarization, mistrust, and racism (Krämer 2017). The partisan and often hostile reporting found on the right-wing alternative media sites in Sweden have been highly controversial (Holt 2016a; 2016b), and they have been framed as “hate sites,” “racist sites,” or “Sweden Democrat sites” in the Swedish press (Andersson and Hammarlund 2016). As the latter name implies, some right-wing alternative news sites have had close links to representatives of the Sweden Democrats; Avpixlat (the predecessor of Samhällsnytt), for instance, was economically founded by Kent Ekeroth from the Sweden Democrats, while Nya Tider was the party paper of the now dissolved Nationaldemokraterna. Recent numbers show that right-wing alternative news outlets in Sweden appear relatively modest in comparison to national established news media when it comes to readership. However, many sites have nevertheless managed to attract a solid audience base. Reuters Digital News Report (Newman et al. 2018) shows that between 6% and 11% say they have used alternative media in the last week (Nya Tider, Fria Tider, Samhällsnytt, Nyheter Idag, Ledarsidorna). In comparison, 46% say they have used Aftonbladet online, 32% have used Expressen online, and 17% have used regional or local newspapers. Furthermore, studies from Norway and the U.S. have indicated that right-wing alternative media often successfully manage to elicit user engagement on social media, which in turn might influence the wider public agenda (Benkler, Faris & Roberts 2018; Larsson 2019). Therefore, it is important to examine the sharing potential of a right-wing agenda in news articles on Facebook and engagement around this content, especially in Sweden, since many of these sites have attracted a relatively large audience. By analyzing a unique dataset of links posted on Facebook, we examine to what extent they originated from right-wing alternative or established news sites, as well as what type of political topics evoked most engagement. Furthermore, we apply an explorative approach to examining the news reports originating from right-wing alternative media and receiving the highest level of engagement, focusing on topics covered and their framing. Start Sharing the News 423

The study contributes to knowledge about right-wing alternative media and their visibility on social media in a Swedish context. The aim is to inform the discussion concerning the role of new agenda-setters in a high-choice media environment, and how platforms like Facebook reinforce the position of marginal actors. In the following discussion, we will first outline theoretical perspectives on “alternative media” before we present the hypotheses tested in the article.

Perspectives on Alternative Media A central precondition of our study concerns the distinction between established and alternative forms of journalism. Scholars within the field of media and communication have long grappled with questions concerning what con- stitute the key differences between the two. Several authors have argued that the boundaries between professional and amateur journalism must be understood as blurry and fluid, rather than something static and fixed (Kenix 2013; Carlson 2015; Hujanen 2016; Singer 2015). That being said, some key characteristics that have been identified as significant markers in the literature include: (1) alternative media are often run by a small number of activist or semiprofessional actors motivated by political, ideological, or economic interest; (2) alternative media are funded on scarce economic means since they operate outside the structures of the established commercial media industry; (3) alternative media are structured as nonhierarchical and participatory organizations; (4) people working for alternative media reject or have an unclear relationship to press ethics or editorial responsibilities; (5) the journalistic content is characterized by a more personal and biased style (Atton 2002; Atton and Hamilton 2008; Atton 2015; Figenschou & Ihlebæk 2018; Nygaard 2019; Wickenden & Atton 2005). Another somewhat more disputed component is the normative dimension of the term “alternative media.” Historically, alternative media have been studied in relation to left-wing media connected to social-protest movements and grassroots activism. From this perspective, alternative media is defined as democratic projects, mobilizing citizens to actively participate in news production or political activism and consequently empowering them (Fuchs 2010; Haas 2004). Haas’s definition of alternative media builds on this approach, identifying them as “media devoted to providing representations of issues and events which oppose those offered in the mainstream media and to advocating social and political reform” (Haas 2004: 115). While studies of left-wing alternative media dominated the field for some time, the role of right-wing and far-right alternative media gained less attention. Atkinson and Berg (2012a) argue that this “flaw” has “left scholars half-blind to the political realities of the twenty- first century” (p. 117), pointing to how far-right populist parties and movements 424 Linn A.C. Sandberg & Karoline Andrea Ihlebæk

have managed to engage an active follower-base. However, a number of studies have recently explored how right-leaning online news sites have managed to set the agenda, often promoting critical stories against immigration and Islam, as well as fronting explicit and extensive media criticism and skepticism toward the cultural elites (Benkler, Faris & Roberts 2018; Figenschou & Ihlebæk 2018; Atkinson & Berg 2012a; Atkinson & Berg 2012b; Nygaard 2019; Rauch, forthcoming; Haller & Holt 2018; Holt & Haller 2017). Based on the traditional normative benchmarks of “alternative media”, questions have arisen as to whether this kind of alternative media should be termed “alternative” at all (Atton 2006). Some scholars have instead labeled far-right groups as “communities with closure” (Atton 2006), “anti-publics” (Cammaerts 2009),”anti-democratic publics” (Alvares & Dahlgren 2016), “repressive radical media” (Downing 2001), “populist counter-media” (Noppari, Hiltunen & Ahva 2019), “hyperpartisan news” (Bastos & Mercea 2019; Larsson 2019) or simply “partisan media” (Newman et al. 2018). Whether “alternative media” can include right-wing news media is consequently contested. Studies have acknowledged that right-wing alternative media cater to participatory involvement and mobilize for political activism. At the same time, they also fuel cultural divisions and promote exclusionary viewpoints (Padovani 2016; Atkinson & Berg 2012a). Furthermore, it is imperative to take into account that what has been referred to as right-wing or far-right (alternative) news media might differ greatly in terms of their ideological position (ranging from controversial to extreme positions), production practices, diversity of content, and participatory possibilities. While some sites produce and distribute misinformation or disinformation that could be harmful and polarizing, others are more moderate in their stance and style, aiming to present alternative viewpoints to their audiences. Conceptualizing this phenomenon by considering the similarities, but also acknowledging the differences among the variety of sites in question, is a complex and difficult exercise. In this context, Holt points to the importance of also looking at the role of alternative media as “self-perceived correctives” for traditional media. Holt proposes the following definition: “A self-assumed term that signals an opposition to traditional media (‘old media’), which many of the writers in this field regard as failing to report properly on important societal issues, for example, by avoiding reporting on social problems related to immigration” (2018: 52). In other words, the self-defined outsider role as alternative or independent is emphasized. While Holt’s definition is useful to escape the strong normativity that has dominated the scholarship on “alternative media”, it surely also has some weaknesses. The self- ascribed role cannot be the only parameter; it should be seen in combination with other dimensions identified in the literature, as outlined above. Acknowledging that the term “alternative media” is disputed and that “right-wing alternative media” might be controversial, we nevertheless believe Start Sharing the News 425

the term is useful for analytical purposes. By “right-wing alternative media”, we mean online news sites that more or less explicitly position themselves as an alternative or corrective to the established media; are run by amateur or semiprofessional actors motivated (more or less explicitly) by conservative-to- far-right values; and resist or have an unclear relationship with press ethics.

Setting the Agenda on Social Media The competition for attention and visibility is fierce in high-choice media environments. Old and new content producers experiment with how they can utilize social media to spread their content. National news outlets still have the highest readership and strongest agenda-setting power; consequently, they arguably also have the strongest presence on Facebook in terms of user engagement. However, recent research indicates that alternative news sites have an active audience base that eagerly shares content, and many right-wing groups seem to be highly knowledgeable about the dynamics of social media and attaining visibility online (Benkler, Faris & Roberts 2018; Marwick & Lewis 2016; Krämer 2017; Haanshuus & Jupskås 2017). A study by Larsson (2019) compared news use and audience engagement on Facebook with a cross-section of national, regional, and what he terms “hyperpartisan news” in Norway. He found that national news outlets have a much higher number of followers on Facebook; however, hyperpartisan actors “emerge as more successful in terms of the degree to which their followers engage as news users” (p. 19). This follower-base both shares and comments to a higher degree, “resulting in the amplification of the perspectives purveyed by these pages” (p. 19). Similarly, it has been argued that these sites have “a political or ideological agenda and their user base tends to passionately share these views” (Newman et al. 2018, p. 15). Therefore, it seems plausible that the recent emergence of right-wing alternative-media sites and their rapid growth in some countries has been, if not made possible, at least facilitated by social-media distribution. Fewer total users consume content directly from right-wing media outlets; however, those outlets’ reach on Facebook might be greater, since their audience is more active in sharing and engaging around this content. In addition, the content itself might evoke more engagement, as it deals with controversial topics and uses contested frames or a “mobilizing tone” (Larsson 2019). Many of these sites position themselves in opposition to the journalism found in the established press, which they perceive as failing in the responsibility to report on issues of immigration and integration. The self-ascribed role of producing content omitted from mainstream media might, by definition, lead to news coverage more likely to be engaging on Facebook. In established theories from communication research, the elements that enhance the popularity of a political message include persuasive elements or news factors (Porten-Cheé 426 Linn A.C. Sandberg & Karoline Andrea Ihlebæk

et al. 2018). Empirical research on social media seems to support this claim; journalistic content that could be characterized as newsworthy (i.e., connected to geographical and cultural closeness, conflict, and human-interest stories) correlates positively with content that becomes “shareworthy” on Facebook. Moreover, traditional news-agency articles receive less engagement, indicating the importance of exclusivity (Trilling, Tolonchko & Burscher 2017). We therefore expect the following relationship between right-wing alternative media and national news outlets to characterize engagement around news content:

H1: The mean engagement score per link originating from traditional news outlets versus right-wing alternative media will be equivalent.

We have previously outlined that questions related to immigration and Islam have been some of the topics that have motivated the rise of right-wing alternative media (Benkler, Faris & Roberts 2018; Holt 2016a; Holt 2016b; Figenschou & Ihlebæk 2018), and that news items on issues pertaining to immigration and Islam resonate well with the follower base of partisan outlets on Facebook (Larsson 2019). At the same time, there is widespread mistrust of traditional- media reporting on immigration-related issues (Sandberg & Demker 2014). Therefore, we expect immigration to score high, particularly for the alternative-media sites.

H2: Compared with other topics, immigration will dominate the coverage in shared links originating from right-wing alternative media.

H3: News stories on immigration-related issues from right-wing alternative- media outlets will similarly evoke the most engagement, compared with other topics.

In the following section, we will discuss the data and methods used in this study before we move on to present results on how content from alternative media was shared, and how Facebook groups engaged with it during the Swedish election, compared to the traditional-news-media content, and the level of engagement related to political issues covered.

Data and Methods

Data selection The Swedish public broadcaster (SVT) collected the data used for analyzing engagement around news links (URLs) on Facebook during the three weeks prior to the Swedish national election and one week after the election on September 9, 2018. SVT selected groups on Facebook that in some way discussed election-related issues based on searches in the social-media tracking tool, Crowdtangle. After a manual classification by relevance and size, a list of Start Sharing the News 427

850 open Facebook groups was compiled. This selection is not comprehensive; nonetheless, it comprises a substantial number of the links shared on Facebook. Crowdtangle does not access links shared from private accounts or closed Facebook groups, and the information gathered from Facebook is com- pletely anonymous.1 The dataset used in this study consists of shared links that originated from either national news media or what we have classified as right- wing alternative-media outlets. Therefore, using secondary-data analysis, we can only assess the relationship between national news media and right-wing alternative media, not the total number of links shared in this time span and the sources behind them. News articles shared on Facebook that were liked, shared, and commented on were aggregated into a “total engagement” score. SVT measured engagement by using a point system in which “shares” and comments were each given two points, and “likes” and other reactions (i.e., love, anger, sad, wow, and haha emojis) were each given one point. Both sharing and commenting signals a greater interest in a post than merely liking or another emoji that expresses feelings toward a post. Users make a greater effort in sharing or commenting on a post, reflected in a higher engagement point value, as not all interactions with a post indicate equal commitment. “Likes” and “shares” function as popularity cues, i.e., indicators of the perceived relevance and positive evaluation of news items and political messages (Porten-Cheé et al. 2018). Based on the aggregation of engagement, we cannot conclude that all engagement is positive. Comments in particular might express both negative and positive attitudes toward the content. Engagement on Facebook does not necessarily reflect public opinion or citizens’ issue engagement. Rather, it reflects exposure and visibility of issues and news content on this specific platform. On Facebook, the activities leading up to increased visibility of a post are largely driven by Facebook’s algorithms, and the engagement around posts can be driven by automatic or semi-automatic accounts (so-called “bots”) or by single individuals. In this way, actual engagement and visibility can be manipulated, and it is difficult to examine the extent to which this affects the result. Greater numbers of interactions with a post are nonetheless rewarded by algorithms and increase visibility and exposure (DeVito 2017). Thus, aggregated engagement with links shared on Facebook reflect what users see and what content those users perceive as prominent.

Issue engagement Journalists at SVT manually coded the content of links shared on Facebook. The choices of issues coded were based on the most important topics for the

1 SVT describes their data collection and methods here: https://www.svt.se/opinion/ metod-facebook-kartan. 428 Linn A.C. Sandberg & Karoline Andrea Ihlebæk

electorate, as reported in a survey conducted by Novus from July to August 2018.2 In addition to the survey, SVT added a category of election/government by which to classify articles about government formation, election campaigning, and party strategy. A category of “other” was also used for content not suitable to any of the predefined categories. It was not possible to see from what group or page the link was shared in the coding procedure. An inter-coder reliability test of a 10% randomly selected sample revealed 94% and 89% inter-rater agreement in coding of link topics, between the coding done by SVT and the sample coded by the authors, which is rather high.3

Classification of right-wing alternative and mainstream media The focus in this paper is to compare shared links and engagement around content originating from right-wing alternative media and traditional media. Traditional mainstream media in our sample consist of well-established national news-media sites in Sweden; local newspapers are excluded, to limit the sample size of traditional media. More specifically, the national news sites are Dagens Nyheter, Sveriges Television, Svenska Dagbladet, Aftonbladet, Expressen, TV4, Metro and OMNI. Table 1 outlines some key characteristics of the sites classified as right- wing alternative. We have based our classification on previous research (Holt 2016a; Nygaard 2019; Newman et al. 2018) as well as a qualitative interpretation emphasizing their self-ascribed role and stated motivation. We have also considered if they report following the ethical codes for the Swedish press; however, we do not know if this is implemented in practice and in the content presented on their website.

2 The topics were Environment/climate, Economy, Labor market, Pensions, Healthcare, Elderly care, Education, Immigration/Integration, and Crime. Additional categories added were Election/Government formation and a category for other content. 3 Out of all 1,214 shared links from traditional media, 121 links were randomly selected in order to test the inter-coder reliability score. Correspondingly, 47 links shared from alternative-media sites were coded by a second coder based on content. Cohen’s kappa was used for measuring interrater reliability and resulted in K= 0.92 for national news media and K= 0.86 for content from partisan media outlets. Start Sharing the News 429

Table 1. Overview of Right-Wing Alternative Media

Stated Ascribed commitment to editor press ethics Self-description on their website Nyheter Idag Yes Yes Independent news site in tabloid format focusing on politics, gossip, social media and foreign policy by creating own news through interviews, be on site to film/photograph and follow developments on social media. Rests on a libertarian outlook. Rapport 24 No No Swedish political discussions and satire daily. Do not support racism or hate speech. Critical towards the Swedish government and its handling of the country. Ledarsidorna Yes Yes Independent web based newsmagazine. Critics of the system, “consequence- neutral” and constantly in opposition. Nyhetsbyrån Yes No Nyhetsbyrån is needed in a media buzz and news flow that increasingly comes to consist of values and emotions. We created NB to select news in areas important to the development of society and its survival. Samhällsnytt Yes No Samhällsnytt takes on a broad publicist approach. We specifically cover the areas in which “establishment media” prove to neglect or to lapse, but also focus broadly on common news journalism. We aim to be the publicist product that tomorrows’ media consumers interested in society, wants and have the right to demand. Fria Tider No No Media in Sweden needs to have the screws put to it. [Mediesverige behöver en rak höger] Samtiden Yes Yes Independent news site with social conservative editorial tendencies. We seek to offer our readers news, relevant analyzes and a debate around conservative values. Nyatider No No The only printed-paper to refute the “press- system’s” lies in their own format: on paper. Since its founding in 2012, Nyatider scrutinize the political correctness and report how reality looks like beyond “system- media’s” distorted view. Svegot No No Daily digital newspaper that aims to broaden the Swedish media landscape and at the same time raise issues important to our non-profit organization and try to push the opinion in a direction more in line with our organization’s ideas. Our editorials political profile is libertarian-nationalist. Nyadagbladet Yes No Independent, humanist and ethno-pluralistic daily newspaper with an anti-globalist stance and an international perspective. Founded in 2012 as a new media platform publishing news, analyses, stories etc. in all media formats. It specifically covers issues on ethnical and cultural rights, integrity and surveil- lance, environment and health.

Note: Self-presentation in the table is a summarized version of how these sites present themselves on their webpage in December 2018 and translated to English. Accountable editor or publisher, and commitment to press ethical norms is noted if it has been stated on the site. 430 Linn A.C. Sandberg & Karoline Andrea Ihlebæk

We argue that even though the included sites differ in many ways, the uniting features are that they are run by amateur or semiprofessional actors, that they employ a more or less explicit anti-establishment rhetoric and critical approach toward traditional media outlets. The ideological position also varies within the selection of sites, and some are more conservative than far right. Our classification further demonstrates the limitations of characterizing these sites only by taking into account their self-ascribed role. For instance, Nyheter Idag does not explicitly state an oppositional role, rather claiming to be “independent,” and furthermore claiming to follow the ethical standards of the press. Consequently, its self-ascribed role in opposition to the mainstream media or as anti-establishment is less explicit. However, previous research has classified this site as alternative, based on its partisan and anti-immigration reporting (Holt 2016a; Newman et al. 2018). At first look, the content on some sites might also appear noncontroversial; a closer reading can reveal, however, a potentially extreme underlying political agenda. The site Nyadagbladet, for example, provides a broad range of news articles and presents itself as Sweden’s broadest and most independent daily newspaper, with the aim of becoming “Tomorrow’s Newspaper!” However, it also defines itself as ethno-pluralistic, and journalists have described it as extreme right.4 Its publisher and news director were previously active in the extreme-right party, Nationaldemokraterna, until the party dissolved. Similarly, the organization behind Svegot has a stated purpose to act as an association for ethnic Swedes, “by standing up for the Swedes and working against the growing hostility by ethnic minorities towards everything Swedish.” “The organization will also work long-term by building networks, purchasing real estate, creating Swedish zones and laying the foundation for an alternative to the failing multicultural state.”5 Like the two examples above, the majority of these sites can be described as far right, based on their coverage and stated purposes of covering news.6

Results From all content shared on Facebook in our material, 1,214 links originated from national news media and 466 from right-wing alternative media. Engagement with links shared from national news media scored mean per link of 1,742, compared to 1,352 per link originating from right-wing media (the total engagement score for national news media was 2,114,849, and 628,666 for right-wing media). Engagement per link compared between the two media

4 https://www.dn.se/webb-tv/klipp/nyheter/det-har-extremsajterna-gemensamt/. 5 https://www.detfriasverige.se/english/ Accessed 23.04.2019. 6 Far right is used here as a unifying concept ranging from right-wing populists to neo-Nazi terrorists, following Mudde’s (2017) classification. Start Sharing the News 431

types was thus similar, even though more links from national news media were shared in our sample of Facebook groups, supporting Hypotheses One (H1). Distribution of engagement with links shared however, varied considerably between the two media types. Links shared from national news media had higher variation in total engagement points. The top link from traditional news media with highest engagement had 31,024 points, and the equivalent link shared from a right-wing outlet had 9,391 points (see Figure 1), suggesting that the total reach is greater for national news media. Nonetheless, the link from right-wing media evoking the highest engagement came in at thirteenth place for total engagement, compared with links shared from national news media.

Figure 1. Engagement on Facebook Around Links Originating from National News Media and Right-Wing Alternative Media Outlets

40,000

30,000

20,000 Engagement

10,000

National news media Right-wing media

Media outlet

Comment: Boxplot showing the spread of engagement around news links, originating from either national news media or right-wing media outlets. Engagement is measured as the total sum of shares (2 p), comments (2 p), likes (1 p) and other reactions (1 p).

Notably, as much as 28% of the content originated from right-wing alternative media sites in our sample of links shared. Given the profile of these sites summarized previously (Table 1), the results not only point to considerable visibility of partisan alternative media on Facebook, but influence from sites specifically hostile to traditional media and the government. Many of these are far-right sites with connections to extreme-right movements. The amount of engagement evoked and the number of links shared is therefore noteworthy in the Swedish context. The results tell us something about the media landscape in Sweden today and the importance of Facebook for spreading and potentially broadening the audience reach of right-wing alternative media. 432 Linn A.C. Sandberg & Karoline Andrea Ihlebæk

We will return to a more in-depth discussion of the content shared in prominent posts to examine further the appeal of right-wing news sites. First, we discuss what political topics evoked most engagement, in the form of shares and total engagement (including number of shares, likes, other reactions and comments) on Facebook, comparing traditional news media and right-wing alternative media.

Table 2. Political Topics Covered in Relation to Shares and Engagement (percentage and mean)

Topics covered in shared links, % (n) Total engagement per topic, Mean (Std.D) Traditional Right-wing Traditional Right-wing news media alternative media news media alternative media Immigration 20 (247) 41 (190) 1721 (1958) 1479 (1344) Election and government 27 (327) 28 (131) 1754 (2273) 1340 (1191) Law and order 13 (154) 11 (52) 1652 (1638) 1467 (1359) Environment 6 (75) 2 (8) 1894 (3713) 1039 (425) Elderly care 0 (3) 2 (8) 1421 (343) 1681 (1217) Healthcare 6 (69) 2 (9) 1263 (1112) 1420 (855) Education 3 (34) 2 (8) 1718 (1674) 898 (645) Pension 2 (20) 1 (3) 2031 (1402) 1108 (518) Economy 7 (85) 0 (11) 2182 (2810) 799 Labor market 1 (8) 0 (1) 995 (522) 524 (-) Defense 1 (9) 0 (0) 988 (614) - Other 15 (183) 10 (45) 1788 (2225) 943 (581) Total 100 (1214) 100 (466) Comment: There was a positive correlation between topics in links shared from traditional news media and right-wing media (Pearson). The value of R2, the coefficient of determination, was 0.7472; yet, no correlation between mean engagement scores for which the value of R2 was 0.0017.

Of the unique links shared on Facebook, most covered topics of immigration/ integration, election and government formation, and law and order, originating from either traditional or alternative right-wing media. The other political topics were covered in less than 10% of shared links. In links originating from right-wing alternative media, 41% covered topics concerning immigration/integration, whereas the corresponding percentage for traditional media was 20%; thus, these topics dominated the coverage in links shared from alternative media, in particular. Topics of election, government formation, and party strategy were covered in most of the links shared from traditional media (27%) and were also the second-most-covered topic in links originating from alternative media (28%). “Law and order” was the third-most-prominent topic in links shared both from traditional news media and from right-wing Start Sharing the News 433

alternative-media sites. Other topics were covered considerably less, and regarding content from right-wing alternative media, the other topics were only covered in 0–11 links, compared with 3–85 links from established media. Content classified as covering topics other than those mentioned in Table 2 appeared to a similar degree in content shared from alternative or traditional media (10% and 15%, respectively). The results indicate that similar political topics in the unique links activate sharing practices on Facebook. These topics thus appear to be salient and “shareworthy,” regardless of source. Topics that the different outlets themselves cover is, of course, also likely to influence type of content shared, especially since partisan alternative news sites tend to have a narrower focus (Newman et al. 2018). Although there was a correlation between the topics featured in most links originating from national news media and right-wing alternative media, the result gives some support to Hypothesis Two (H2), since issues relating to immigration dominated the coverage in links shared from right-wing alternative media. Apart from content concerning “election and government,” topics other than immigration only appeared in 0–11% of the news stories shared. On the other hand, comparing the topics’ different degrees of engagement reveals less similarity, and there was no correlation in topic engagement between traditional and right-wing alternative media. Among the news from right-wing alternative media, topics concerning elderly care and immigration, followed by law and order, received highest engagement value (measured as mean engagement score). However, differences in mean were small for most topics, with respect to content shared from right-wing alternative media. On the other hand, the mean engagement score varied considerably between the topics evoking highest engagement value (elderly care) versus lowest (labor market), ranging between 524 and 1,681. Topics evoking the highest engagement in shares from traditional news media were the economy, followed by pension and the environment. Election and government topics had the fourth highest mean score, followed by immigration, education, law and order, elderly care, health care, labor market, and, lastly, issues regarding Swedish defense. Defense-related issues were not covered in any of the links shared from right- wing alternative media. Compared with other political topics, immigration coverage evoked the second-highest mean engagement score in links from right-wing alternative media. Although immigration was not among the topics that evoked most engagement in links from national news media, it did evoke a higher mean engagement score (1,721) than it did in links from right-wing alternative media (1,479). Therefore, the coverage of immigration did not evoke the highest engagement value on average, compared to other topics, nor did it evoke higher mean engagement value compared with immigration coverage in national news 434 Linn A.C. Sandberg & Karoline Andrea Ihlebæk

media. Our results, therefore, do not give full support to our third hypothesis that immigration-related issues would generate highest engagement around content originating from right-wing alternative media. Next, we will take a closer look at the content in links shared that received highest engagement value, to further explore the link between right-wing alternative media and its engagement appeal on Facebook.

Immigration, crime, and the Sweden Democrats Among the links originating from right-wing alternative media sites with highest engagement value, scores for the top eight news articles shared varied between 9,391 and 5,481 (see Appendix). Content in these most engaging links was published on three different alternative news sites (Samhällsnytt, Fria Tider, and Nyheter Idag). Samhällsnytt published the news article evoking the highest engagement. The article describes a candidate for the Social Democratic party and his criminal record, focusing on his background as an immigrant and statements he had made about reducing crime rates. Another article from Fria Tider with the third-highest engagement value also focuses on a Social Democratic politician with an immigrant background who fled the scene of a car accident with “a minivan full of immigrant-politicians.” The article explains this accident in the author’s own words: “Was in a rush to get to the airport as they were flying home to the countries they fled from on vacation.” The article with second-highest engagement points also concerns immigration and integration themes. In the article, a Social Democratic Danish politician is quoted as encouraging the Swedish population to vote for the Sweden Democrats: “He believes that Sweden’s prime minister lacks the correct ‘reality view’ and that immigration is so severe that it can bring down the Nordic welfare state.” The three links with the highest engagement of all concerns immigration and integration, as well as Social Democrats, either committing criminal acts as in the two first examples, or as in the last example, using a quote from a Social Democrat encouraging voting for the Sweden Democrats. Moreover, the links with the fourth- and fifth-highest engagement rates deal with immigration- related issues and Islam—more specifically: “Residential area in Västerås is taken over by Salafists—the police deeply concerned” and “Poland rather leave the EU than being forced to admit Muslims.” Another theme in the links with highest engagement (6th through 8th place) concerns the Sweden Democrats and fraud or crime allegedly commit- ted against them. In one, a journalist is perceived as boasting about sabotaging ballot papers belonging to the Sweden Democrats while she expresses support for the Center party on Instagram. In a comment, the party leader thanks her for her support, which is interpreted as the party leader supporting the act of sabotage. Another link reports on a false call against the Sweden Democrats in Start Sharing the News 435

which several celebrities supposedly signed this list, but in fact had not. Lastly, a link with the title “Sweden Democratic youth heavily assaulted by Somali gang” evoked high engagement. Immigration, the Sweden Democrats, and crime dominated the content in the links evoking the highest engagement value on Facebook. In some articles, the themes overlap. For example, regarding the Social Democratic politician’s alleged misconduct, both angles of immigration and crime are used while the main message is aimed at the faulty behavior of Social-Democratic politicians. When articles concern the Sweden Democrats, they express either support for or wrongdoings against the party. Also, worth noting is that while some of the articles evoking high engagement are neutral in tone and reporting, others are depreciatory and ridiculing toward political opponents. While there is no indication of fabricated news content, some of the articles give the impression of exaggerating and, to some extent, tailoring the facts to fit a cause. News reporting in these articles features both, in the form of self-produced content and news from established media, recontextualized to make a political statement.

Discussion and Conclusion In this study, we explored the role of right-wing alternative media on Facebook during the Swedish general election, in particular, the degree to which they managed to create visibility and engagement in relation to the established media. A common line of argument in the self- presentation of these sites is anti-establishment rhetoric, such as “system critical,” “constantly in opposition,” “establishment media,” refuting “system-media’s” distorted view and the “press-system’s” lies. Fria Tider, one of the most well-known and read sites included in our analysis, has the slogan, “Media in Sweden needs to have the screws put to it [Mediesverige behöver en rak höger].” Another example takes this rhetoric a step further by arguing for the need of alternative news in order to cover areas “important to the development of society and its survival.” Uniting features of these sites are criticism or hostility toward legacy media that are believed to censor important information, anti-establishment, and a strong focus on immigration-related issues and what are perceived as negative immigration policies or effects of immigration. Our study finds that right-wing media sites seem to have had a substantial influence on social media during the Swedish 2018 election, compared with national news media. Of the news stories shared in our sample, 28% originated from these sites, compared to national news media on Facebook. As many of these sites are far-right with stated motives going beyond merely covering what traditional news media are perceived to neglect, the reach and potential influence are made more striking. 436 Linn A.C. Sandberg & Karoline Andrea Ihlebæk

In line with our expectations, the engagement with links shared was similarly high, despite the fact that traditional news media generally reaches a larger segment of the population. This tells us something about the audiences of alternative right-wing media and supports previous findings of this reader base being particularly active in sharing and engaging with this content (Larsson 2019). Concerning our second and third assumption, we believed that immigration would dominate the coverage in links shared on Facebook and that immigration-related issues would generate the highest engagement scores, which our study partly supported. Immigration-related issues were covered in the majority of links shared from right-wing alternative media and evoked high engagement; however, this topic did not evoke the highest engagement score on average. Compared to traditional news media, the mean engagement score was also lower. This suggests that immigration-related issues evoke relative high engagement, regardless of the news media actor. Immigration-related issues were covered in a substantial portion of links shared from national news media (20%). Negative reporting of immigration-related issues in traditional media might therefore generate engagement similar to the coverage of right-wing alternative media, although this content is more controversial and uses contested frames. Taking a closer look at the content in the most engaging links revealed quite a homogeny in the right-wing agenda, both in type of content and partisan or one-sided angles. Content concerned immigration in particular, but also crime, and indicated support for the Sweden Democrats. In several, a sarcastic under- tone can also be detected, possibly aimed at undermining political opponents or expressed as contempt for politicians with an immigrant background. The links receiving highest engagement value originated from three different sites (Samhällsnytt, Fria Tider, and Nyheter Idag). This also indicates a widespread supply of alternative news, as not only one or two outlets seem to have dominated the “alternative agenda”. In this study, we have only examined the relationship between national news media and right-wing alternative-media outlets. We can therefore not assess how right-wing content was shared and engaged with, compared to all content shared on Facebook. Other outlets, particularly political blogs, would be of value to include in future studies. Moreover, due to data restriction, we only examined open Facebook groups and can therefore not know how this distribution differs from that in private profiles or closed Facebook groups. Sweden received significant international attention around its 2018 election, in particular, misinformation and the influence of bots on social media. These bots expressed support for the Sweden Democrats, criticism of immigration Start Sharing the News 437

and refugees, and critique of the elites and of the media.7 Compared with the result presented in this study, a similar agenda can be found in content originating from right-wing alternative media. Since many of the sites in question have a fundamentally radical or extreme agenda, the influence of content from these sites on the political news flow on Facebook might contribute to increasingly polarized debates and reinforcing political beliefs about a lying press and harmful immigration. In future research, it will be of importance to examine these outlets more closely and the extent to which social-media dynamics reinforce these types of messages.

References Alvares, Claudia & Dahlgren, Peter, 2016. “Populism, Extremism and Media: Mapping an Uncertain Terrain”, European Journal of Communication 31(1), pp 46–57. Andersson Schwarz, Jonas & Hammarlund, Johan, 2016. “Kända men inte erkända: Skildringen av Avpixlat och andra invandringsfientliga nätmedier i fem svenska dagstidningar 2013–2015”, pp 174–191 in Truedson, Lars (red.), Migrationen i medierna: men det får en väl inte prata om?. Stockholm: Institutet för mediestudier. Atkinson, Joshua & Berg, Suzanne V.L., 2012a. “Right Wing Activism: The Next Challenge for Alternative Media Scholarship”, pp 117–133 in Sützl, Wolfgang & Hug, Theo (eds),Activist Media and Biopolitics: Critical Media Interventions in the Age of Biopower. Innsbruck: Innsbruck University Press. Atkinson, Joshua & Berg, Suzanne V.L., 2012b. “Narrowmobilization and Tea Party Activism: A Study of Right-Leaning Alternative Media”, Communication Studies 63(5), pp 519–535. Atton, Chris, 2002. Alternative media. London: Sage. Atton, Chris, 2004. An alternative internet. Edinburgh: Edinburgh University Press. Atton, Chris, 2006. “Far-right media on the internet: culture, discourse and power”, New Media & Society 8(4), pp 573–587. Atton, Chris & Hamilton, James, 2008. Alternative journalism. London: Sage. Atton, Chris, 2015. The Routledge companion to alternative and community media. New York: Routledge. Bastos, Marco T. & Mercea, Dan, 2019. “The Brexit Botnet and User-Generated Hyperpartisan News”, Social Science Computer Review 37(1), pp 38–54. Benkler, Yochai, Faris, Robert & Roberts, Hal, 2018. Network Propaganda Manipulation, Disinformation, and Radicalization in American Politics. Oxford: Oxford University Press. Cammaerts, Bart, 2009. “Radical pluralism and free speech in online public spaces: the case of North Belgian extreme right discourses”, International Journal of Cultural Studies 12(6), pp 555- 575. Carlson, Matt, 2015. “Introduction: The many boundaries of journalism”, pp 1–19 in Carlson, Matt & Lewis, Seth C. (eds), Boundaries of Journalism. London: Routledge.

7 https://edition.cnn.com/2018/08/30/europe/sweden-twitter-bots-elections-intl/index.html 438 Linn A.C. Sandberg & Karoline Andrea Ihlebæk

DeVito, Michael, A., 2017. “From Editors to Algorithms”, Digital Journalism 5(6), pp 753–773. Downing, Robert, 2001. ”Repressive Radical Media”, pp 88–96 in Downing, Robert (ed.), Radical Media. Rebellious Communication and Social Movements. Thousand Oaks: Sage. Engesser, Sven, Ernst, Nicole, Esser, Frank & Büchel, Florin, 2017. “Populism and social media: how politicians spread a fragmented ideology”, Information, Communication & Society 20(8), pp 1109–1126. Engesser, Sven, Fawzi, Nayla, & Larsson, Anders Olof, 2017. ”Populist online communication: introduction to the special issue”, Information, Communication & Society 20(9), pp 1279–1292. Ernst, Nicole, Engesser, Sven, Büchel, Florin, Blassnig, Sina & Esser, Frank, 2017. “Extreme parties and populism: an analysis of Facebook and Twitter across six countries”, Information, Communication & Society 20(9), pp 1347–1364. Figenschou, Ustad, Tine and Ihlebæk, Karoline Andrea, 2018. “Challenging Journalistic Authority”, Journalism Studies, pp 1–17. doi: 10.1080/1461670X.2018.1500868. Fuchs, Christian, 2010. “Alternative Media as Critical Media”, European Journal of Social Theory, 13(2), pp 173–192. Haller, André & Holt, Kristoffer, 2018. “Paradoxical populism: how PEGIDA relates to mainstream and alternative media”, Information, Communication & Society, pp 1–16. Holt, Kristoffer, 2016a. “«Alternativmedier»? En intervjustudie om mediekritik och mediemisstro”, pp 113–150 in Truedson, Lars (ed.), Migrationen i medierna – men det får en väl inte prata om?. Stockholm: Institutet för mediestudier. Holt, Kristoffer, 2018. “Alternative Media and the Notion of Anti-Systemness: Towards an Analytical Framework”, Media and communication 6(4), pp 49–57. Holt, Kristoffer, 2016b. “Skilda verkligheter?: ‘Internets undervegetation’ vs ‘PK-maffian’”, pp 150–172 in Truedson, Lars (ed),Migrationen i medierna: Men det får en väl inte prata om? Stockholm: Institutet för mediestudier. Holt, Kristoffer & Haller, André, 2017. “What Does ‘Lügenpresse’ Mean? Expressions of Media Distrust on PEGIDA’s Facebook Pages”, Politik, 20(4), pp 42–58. Hujanen, Jaana, 2016. “Participation and the Blurring Values of Journalism”, Journalism Studies 17(7), pp 871–880. Haanshuus, Birgitte Prangerød & Ravik Jupskås, Anders, 2017. “Høyreklikk! En analyse av ytre høyre på sosiale medier i Norge”, Tidsskrift for samfunnsforskning, 2, pp 145–165. Haas, Tanni, 2004. “Research note. Alternative media, public journalism and the pursuit of democratization”, Journalism Studies, 5(1), pp 115–121. Kenix, Linda Jean, 2013. “A Converging Image? Commercialism and the visual identity of alternative and mainstream news websites”, Journalism Studies, 14(6), pp 835–856. Krämer, Benjamin, 2017. “Populist online practices: the function of the Internet in right-wing populism”, Information, Communication & Society 20(9), pp 1293–1309. Larsson, Anders Olof, 2019. “News Use as Amplification – Norwegian National, Regional and Hyperpartisan Media on Facebook”, Journalism & Mass Communication Quarterly, online early. Start Sharing the News 439

Mezzofiore, Gianluca & Smith, Rory, 2018. “Sweden: Twitter bots have doubled ahead of elections”, CNN, Retrieved 01.01.2019 from: https://edition.cnn.com/2018/08/30/ europe/sweden-twitter-bots-elections-intl/index.html. Marwick, Alice & Lewis, Rebecca, 2016. Media Manipulation and Disinformation Online. New York: Data & Society Research Institute. Mudde, Cas, 2017. The far right in America. Routledge. Newman, Nic, Fletcher, Richard, Kalogeropoulos, Antonis, Levy, David A. L. & Nielsen, Rasmus Kleis, 2018. Digital News Report 2018. Oxford: Reuters Institute for the Study of Journalism. Noppari, Elina, Hiltunen, Ilmari & Ahva, Laura, 2019. ”User profiles for populist counter-media websites in Finland”, Journal of Alternative and Community Media, 4(1), pp 23–37. Nygaard, Silje, 2019. “The Appearance of Objectivity: How Immigration-Critical Alternative Media Report the News”, Journalism Practice, pp 1–17. doi: 10.1080/17512786.2019.1577697. Padovani, Cinzia, 2016. “The Media of the Ultra-Right: Discourse and Audience Activism Online”, Journal of Language and Politics 15(4), pp 399–421. Porten-Cheé, Pablo, Haßler, Jörg, Jost, Pablo, Eilders, Christiane & Maurer, Marcus, 2018. “Popularity cues in online media: Theoretical and methodological perspectives in political communication research”, Studies in Communication and Media, 7(2), pp 208–230. Rauch, Jennifer. Forthcoming. “Right-Wing Audiences for Alternative Media: Comparing Progressive and Conservative Attitudes towards Journalism”, in Atkinson, Joshua & Kenix, Linda Jean (eds), Activist Nation Rising. New York: Rowman & Littlefield. Sandberg, Linn & Demker, Marie, 2014. ”Starkare oro för främlingsfientlighet än för invandring”, pp 71–82 in Bergström, Annika & Oscarsson, Henrik (eds), Mittfåra & marginal. Göteborg: SOM-institute. Schierup, Carl-Ulrik & Ålund, Aleksandra, 2011. “The end of Swedish exceptionalism? Citizenship, neoliberalism and the politics of exclusion”, Race and class 53(1), pp 45–64. Singer, Jane B., 2015. “Out of bounds. Professional norms as boundary markers”, pp 21–37 in Carson, Matt & Lewis, Seth C. (eds), Boundaries of Journalism. Professionalism, Practices and Participation. New York: Routledge. Sköld, Josefin, & Carlsson, Mattias, 2017. “Det har extremsajterna gemensamt”, Dagens Nyheter. Retrieved 01.01.2019 from: https://www.dn.se/webb-tv/klipp/nyheter/ det-har-extremsajterna-gemensamt/. Sköld, Josefin, Urisman, Otto & Ewald, Hugo, 2018. “Så tog högerpopulistiska sidor över valrörelsen på Facebook – med hjälp av SD”, Dagens Nyheter. Retrieved 01.01.2019 from: https://www.dn.se/nyheter/politik/ sa-tog-hogerpopulistiska-sidor-over-valrorelsen-pa-facebook-med-hjalp-av-sd/. Strömbäck, Jesper, Andersson, Felicia & Nedlund, Evelina, 2017. Invandring i medierna: Hur rapporterade svenska tidningar 2010–2015? Stockholm: Delegationen för migrationsstudier. Trilling, Damian, Tolochko, Petro & Burscher, Björn, 2017. ”From Newsworthiness to Shareworthiness: How to Predict News Sharing Based on Article Characteristics”, Journalism & Mass Communication Quarterly, 94(1), pp 38–60. Wickenden, Emma & Atton, Chris, 2005. “Sourcing Routines and Representation in Alternative Journalism: a case study approach”, Journalism Studies 6(3), pp 347–359. 440 Linn A.C. Sandberg & Karoline Andrea Ihlebæk

Appendix

Table 1. Links Originating from Alternative Media Sites With Highest Engagement Scores

Engagement- score Outlet and date URL link 9391 Samhällsnytt https://samnytt.se/forsokte-strypa-kassor-kandiderar- (27/08/2018) for-socialdemokraterna-i-valet/ 8864 Fria Tider http://www.friatider.se/ (25/08/2018) dansk-socialdemokrat-till-svenska-folket-r-sta-p-sd 8278 Fria Tider http://www.friatider.se/smet-fr-n-trafikolycka-och-lj-g- (04/09/2018) sig-till-l-gre-b-ter-sen-blev-haxhimustafa-32-l-fvens-nya 7603 Samhällsnytt https://samnytt.se/bostadsomrade-i-vasteras-tas- (29/08/2018) over-av-salafister-polisen-djupt-oroad/ 7071 Nyheter Idag https://nyheteridag.se/ (08/08/2018) polen-lamnar-hellre-eu-an-tvingas-ta-emot-muslimer/ 7054 Nyheter Idag https://nyheteridag.se/det-stammer-inte-kandisar- (06/09/2018) tar-avstand-fran-kandislista-mot-sd/ 6486 Nyheter Idag https://nyheteridag.se/journalist-skryter-om- (25/08/2018) valsabotage-mot-sd-annie-loof-tackar-for-stodet/ 5481 Nyheter Idag https://nyheteridag.se/sverigedemokratiska- (03/09/2018) ungdomar-grovt-misshandlade-av-somaliskt-gang-de- fortsatte-sparka-pa-honom-nar-han-lag-ner/