Closer to the people: A comparative content analysis of populist communication on social networking

A comparative content analysis explores stylistic and contextual factors that resonate with populist communication on social networking sites (SNSs). Advancing the field, we investigate differences in politicians’ use of populist communication in preand post-election periods by analyzing a comprehensive sample of Facebook posts and tweets (N = 1,010) of the leading candidates in the 2017 national parliamentary elections in Austria and the Netherlands. We identify important stylistic elements that resonate with populist political communication such as negative emotions or an us-versus-them rhetoric. Our findings further suggest that although populist communication is more prevalent in the SNS communication of right-wing and left-wing populist parties, political actors across all parties are more likely to use populist communication before than after elections. In line with recent conceptualizations, we argue that populism can be understood as a framework of communication with measurable sub frames, which can be expressed by different political actors with different goals. Understanding populism as a discursive framework of communication can ultimately help to reconcile existing divergent conceptualizations of populism. ARTICLE HISTORY Received 26 August 2018 Accepted 26 February 2019 Politicians around the world rely heavily on so-called social networking sites (SNSs) such as Twitter and Facebook to reach out to their voters. In particular, populist politicians were early adopters of SNSs, many of them using such platforms as their primary tool of communication, as the example of Donald Trump on Twitter demonstrates. The success enjoyed by populists in the digital space raises questions about the specific communication strategies and stylistic elements employed in populist political communication as well as the specific contextual structures, which may provide a breeding ground for the success of populist communication on SNSs. Essentially, populist ideas prioritize a pervasive dichotomy in politics and society: good, ordinary people versus elites who, in failing to represent the people, threaten their interests © 2019 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/ licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. CONTACT Desirée Schmuck desiree.schmuck@univie.ac.at Advertising and Media Effects Research Group, University of Vienna, Währinger Strasse 29, 1090, Vienna INFORMATION, COMMUNICATION & SOCIETY 2020, VOL. 23, NO. 10, 1531–1548 https://doi.org/10.1080/1369118X.2019.1588909 (Mudde, 2004; Taggart, 2000). SNSs may play a key role in facilitating the direct communication of populist actors and the people, as populist actors can circumvent journalists to establish and strengthen unmediated relationships with the people (e.g., Engesser, Ernst, Esser, & Büchel, 2016). Such self-communication aligns with the populist argument that traditional media represent the corrupt elites, as demonstrated by the populist scapegoating of traditional media for disseminating fake news. Yet, although these actors may have a stronger tendency not to talk with journalists of certain traditional media outlets, they may still profit from the oxygen of publicity provided by journalists to their populist self-communication on SNSs. The SNS accounts of US President Donald Trump and European politicians Geert Wilders and H.C. Strache, among others, provide clear examples of populist self-communication distributed via Twitter and Facebook. As Trump himself has argued, by Twitter no less, ‘If the press would cover me accurately, I would have far less reason to “tweet.” Sadly, I don’t know if that will ever happen!’ (Trump, 5 December 2016 at 08:00 h). With that tweet, Trump brought home the idea that traditional media should be regarded as inaccurate, unreliable sources that, as such, force him to use alternative platforms to communicate to the people. The successful adoption of SNS use by populist politicians is also mirrored in Austria, where the right-wing populist candidate, Heinz-Christian Strache, is at the forefront of using SNSs for his political purposes with more than 790,000 followers on Facebook. In the Netherlands, the right-wing populist political leader Geert Wilders is the most successful and active politician on Twitter with currently more than 808,000 followers on Twitter and up to 4–20 times more activities on Twitter than other major party leaders in the Netherlands (Jacobs & Spierings, 2018). Given the alleged prominence of SNS as forums for populist communication, empirical research has enhanced current understandings of contextual conditions (e.g., Ernst, Engesser, Büchel, Blassnig, & Esser, 2017; Jacobs & Spierings, 2018) as well as elements of presentation style (Bobba, 2018; Bracciale & Martella, 2017; Engesser et al., 2016; Waisbord & Amado, 2017) that resonate with populist political communication on SNSs. Yet to our knowledge, there is a paucity of research on the specific stylistic and contextual determinants that predict populist political communication in a comparative setting. We need to identify important stylistic elements of presentation that may be associated with populist political communication in order to truly understand its appeal. In so doing, we distinguish between the substantial content of populist communication (i.e., sub frames such as people-centrism, antielitism, and popular sovereignty) in light of discursive framing and stylistic elements (i.e., emotionality, negativity, us-them-rhetoric, and references to common sense) used to express those messages on SNSs. While the discursive framework of populist political communication and its sub frames refer to what is expressed in populist political communication, stylistic elements describe how those messages are presented. Additionally, we need to pin down the structural conditions such as the type of SNS, the type of party, the time period, and country-specific factors. Previous research has in particular insufficiently accounted for the contextual factor of election periods in predicting politicians’ use of populism. However, only by looking at both preand post-election periods, we can determine (a) whether mainstream politicians are also tempted to flirt with populism in election times and (b) whether politicians, who are commonly classified as populist, also engage in populist communication in non-election times. To that end, we assessed the interaction of time period (i.e., preand post-election periods) and type of political actor (i.e., 1532 D. SCHMUCK AND M. HAMELEERS

. SNSs may play a key role in facilitating the direct communication of populist actors and the people, as populist actors can circumvent journalists to establish and strengthen unmediated relationships with the people (e.g., Engesser, Ernst, Esser, & Büchel, 2016). Such self-communication aligns with the populist argument that traditional media represent the corrupt elites, as demonstrated by the populist scapegoating of traditional media for disseminating fake news. Yet, although these actors may have a stronger tendency not to talk with journalists of certain traditional media outlets, they may still profit from the oxygen of publicity provided by journalists to their populist self-communication on SNSs. The SNS accounts of US President Donald Trump and European politicians Geert Wilders and H.C. Strache, among others, provide clear examples of populist self-communication distributed via Twitter and Facebook. As Trump himself has argued, by Twitter no less, 'If the press would cover me accurately, I would have far less reason to "tweet." Sadly, I don't know if that will ever happen!' (Trump, 5 December 2016 at 08:00 h). With that tweet, Trump brought home the idea that traditional media should be regarded as inaccurate, unreliable sources that, as such, force him to use alternative platforms to communicate to the people.
The successful adoption of SNS use by populist politicians is also mirrored in Austria, where the right-wing populist candidate, Heinz-Christian Strache, is at the forefront of using SNSs for his political purposes with more than 790,000 followers on Facebook. In the Netherlands, the right-wing populist political leader Geert Wilders is the most successful and active politician on Twitter with currently more than 808,000 followers on Twitter and up to 4-20 times more activities on Twitter than other major party leaders in the Netherlands (Jacobs & Spierings, 2018). Given the alleged prominence of SNS as forums for populist communication, empirical research has enhanced current understandings of contextual conditions (e.g., Ernst, Engesser, Büchel, Blassnig, & Esser, 2017;Jacobs & Spierings, 2018) as well as elements of presentation style (Bobba, 2018;Bracciale & Martella, 2017;Engesser et al., 2016;Waisbord & Amado, 2017) that resonate with populist political communication on SNSs.
Yet to our knowledge, there is a paucity of research on the specific stylistic and contextual determinants that predict populist political communication in a comparative setting. We need to identify important stylistic elements of presentation that may be associated with populist political communication in order to truly understand its appeal. In so doing, we distinguish between the substantial content of populist communication (i.e., sub frames such as people-centrism, antielitism, and popular sovereignty) in light of discursive framing and stylistic elements (i.e., emotionality, negativity, us-them-rhetoric, and references to common sense) used to express those messages on SNSs. While the discursive framework of populist political communication and its sub frames refer to what is expressed in populist political communication, stylistic elements describe how those messages are presented. Additionally, we need to pin down the structural conditions such as the type of SNS, the type of party, the time period, and country-specific factors. Previous research has in particular insufficiently accounted for the contextual factor of election periods in predicting politicians' use of populism. However, only by looking at both pre-and post-election periods, we can determine (a) whether mainstream politicians are also tempted to flirt with populism in election times and (b) whether politicians, who are commonly classified as populist, also engage in populist communication in non-election times. To that end, we assessed the interaction of time period (i.e., pre-and post-election periods) and type of political actor (i.e., populist and mainstream actors on the left and right of the ideological spectrum) for the first time in extant literature. In so doing, we aim to contribute to ongoing debates which argue that different kinds of political actors strategically use populist political communication to maximize votes or increase power (Jagers & Walgrave, 2007;Weyland, 2001). Altogether, our research marks the first study in a comparative context that investigates the populist communication of mainstream and populist actors online, both before and after elections, by taking features of presentation style as well as context into account.
In what follows, we underline the relevance of populist political communication on SNSs before we offer a definition of populist political communication as a discursive framework and describe the three fundamental sub frames of populist political communication: people-centrism, antielitism, and popular sovereignty. In a next step, we hypothesize important stylistic and contextual predictors of populist political communication on SNSs based on previous research. Finally, we present and discuss the results of a comparative content analysis of Twitter and Facebook posts (N = 1,010) issued by the accounts of 13 leading political candidates before and after the national parliamentary elections in Austria and the Netherlands in 2017.

Populist self-communication on social networking sites
SNSs have been regarded as a principal supply-side factor in populism's recent worldwide ascent (e.g., Engesser et al., 2016;Ernst et al., 2017;Waisbord & Amado, 2017). In the current era of reduced trust in mass media and, consequently, political actors' use of alternative, self-controlled media, a shift toward direct populist self-communication has become clear. Today's politicians, with an alternative discursive platform at their disposal allowing them to communicate their positions directly to the electorate, have become empowered to not only circumvent traditional media but also strengthen bonds with their followers (e.g., Engesser et al., 2016;Hameleers & Schmuck, 2017).
Of all politicians, populists seem particularly suited to take advantage of SNSs (e.g., Ernst et al., 2017). In the populist dichotomy, elites include not only political actors but also traditional mass media, which, in catering exclusively to the will of elites, neglect the people's will. In response, by mobilizing their SNS accounts, populist politicians can bypass traditional media outlets (Engesser et al., 2016). The technological affordances of SNSs allow politicians to establish a direct, close, and personal relationship with the people whom they claim to represent without the interference of journalists (Bartlett, Birdwell, & Littler, 2011;Hameleers & Schmuck, 2017). Additionally, through self-communication on SNSs, populist communication may receive attention from the mass media due to controversial posts, which are reproduced by journalists because they fit today's media logic (Mazzoleni & Bracciale, 2018).
However, given recent empirical evidence suggesting that populist communication on SNSs spreads in fragmented ways (Engesser et al., 2016), the various elements of such communication may not always surface altogether in a single post to Facebook or Twitter. Instead, populist political communication can be understood as a discursive framework that has several sub frames, which may appear together or separately in politicians' selfcommunication. In the next section, we describe three major components of populist political communicationpeople-centrism, antielitism, and popular sovereigntythat can be understood as the fundamental sub frames of the discursive framework of populism.

Populist communication as a discursive framework
At base, populism imagines a dichotomy in society between ostensibly good, ordinary citizens whose interests are threatened by ostensibly evil, corrupt elites (Canovan, 1999;Mudde, 2004;Taggart, 2000). More particularly, however, populism has been conceived in different ways that have been used separately or combined in theory and empirical research (see e.g., Engesser, Fawzi, & Larsson, 2017): as a thin-centered ideology (Mudde, 2004), a political communication style (Cranmer, 2011;Jagers & Walgrave, 2007;Moffitt, 2016), as a strategy seeking to gain or maintain power (e.g., Weyland, 2001), and as a discursive framework (e.g., Aslanidis, 2016). Although those different conceptualizations might seem irreconcilable at first glance, they are not mutually exclusive, and it is possible to accommodate them in the concept of populist political communication (Aalberg, Esser, Reinemann, Strömbäck, & de Vreese, 2017;Engesser et al., 2017). The thin-centered ideology of populism understood as the ultimate separation of society into two antagonistic groups, 'the pure people' and 'the corrupt elite', reflects the content of populist communication. Furthermore, populist communication can be presented using specific stylistic elements, which refer to how the content is presented (de Vreese, Esser, Aalberg, Reinemann, & Stanyer, 2018). The populist ideological content and stylistic elements can further be employed strategically by different actors with different goals (e.g., to maximize votes, gain or maintain power). Lastly, conceptualizing populism as a discursive framework defines manifest artefacts (Aslanidis, 2016), which allow measuring populism as a matter of degree within a framework that various actorsvoters (e.g., Akkerman, Mudde, & Zaslove, 2014), mass media (e.g., Mazzoleni, 2008), and politicians (e.g., Jagers & Walgrave, 2007)can apply.
Drawing from the thin-centered ideology of populism, three major ideas can be understood as the key sub frames, which define the content of populist communication's framework: the centrality of ordinary people, antielitism, and popular sovereignty that leaves or returns power to the people (Schulz et al., 2017; see also Ernst et al., 2017). First, populist communication constructs a sense of in-group favoritism by framing ordinary people as central to political decision making (Canovan, 1999). Populist communication postulates that though the people's will should be central to politics, it is largely neglected. The ingroup is therefore framed as the silent majority of everyday citizens (Caiani & della Porta, 2011). As the second core component of populist communication, antielitism refers to the dichotomy between corrupt elites and ordinary people; elites stand accused of taking care of themselves only and at the expense of the people, whose will they have been entrusted to protect (Mudde, 2004). Last, popular sovereignty advocates the empowerment of the people while denying the sovereignty of elites, that is, populist political communication recognizes the people as the legitimate sovereign entity and democracy as the mechanism for expressing the vox populi (Canovan, 1999). Taken together, the three core ideas of populist communication can be understood as frame-elements that emphasize the problem situation defined as a threat or crisis facing the ordinary people, a causal interpretation in which blame towards corrupt elites is highlighted, a moral evaluation that separates the 'good' people from the 'evil' others, and a potential treatment evaluation proposing that the elites should be removed from the people's in-group (see Entman, 1993).

Stylistic elements of populist communication
Researchers have amply addressed the specific style by which populist ideas are presented (e.g., Fieschi & Heywood, 2006). Populism has been called highly emotional, negative, oversimplifying, dramatizing, common-sense, and centered on conflict (e.g., Bos, Van der Brug, & de Vreese, 2010;Mazzoleni, Stewart, & Horsfield, 2003;Mudde, 2007;Rooduijn, 2014). Here, we distinguish the substantial subframes of populist communication (i.e., people-centrism, antielitism, and popular sovereignty) in light of discursive framing and stylistic elements (i.e., emotionality, negativity, us-themrhetoric, and references to common sense) used to express those messages in content. Thus, we are interested in how the discursive framework of populist political communication is presented on SNSs.
To begin, populist communication has been described as highly emotional (Aalberg et al., 2017;Mazzoleni et al., 2003) and, at that, with emotions typically negative in valence (Fieschi & Heywood, 2006). Populist actors have been characterized, for example, as drawing upon anger and fear to intensify the divide between ordinary people and evil elites (Hameleers, Bos, & de Vreese, 2017). In terms of blame attribution, populist communication shifts fault from the people to elites by stressing ordinary people's anger at the culprits and fear of their own impending deprivation (Mols & Jetten, 2014). Despite the salience of negative emotions, populist communication can also rely on positive emotions, including in-group attachment, nostalgia, and pride (Aalberg et al., 2017), by emphasizing in-group favoritism, loyalty to the heartland, and the achievements of hardworking, ordinary citizens. Accordingly, we hypothesized: H1a: The discursive framework of populist communication is presented with negative and positive emotions.
Related to populism's often negative emotionality, negative tonality has also been identified as a specific aspect of populist political communication (Mazzoleni et al., 2003). Although indeed closely connected, the two concepts refer to contrasting message characteristics. Unlike emotionality, which describes a communicant's use of positive (e.g., happiness) and negative (e.g., sadness, anger) emotions, tonality refers to the messages' overall tone. For instance, describing positive (e.g., achievement and improvement) or negative (e.g., failure and fiasco) events or developments (De Vreese, Esser, & Hopmann, 2016;Heiss, Schmuck, & Matthes, 2018). Populist messages frequently reconstruct a sense of crisis or urgency and assign culpability to elites or societal out-groups for negative outcomes (Mudde, 2004;Taggart, 2000). Therefore, we assumed: H1b: The discursive framework of populist communication is presented in a negative rather than a positive tonality.
Researchers have additionally observed that populist actors emphasize their closeness to ordinary people and their distance from elites with the means of certain elements of presentation style (e.g., Canovan, 1999;Mudde, 2007;Taggart, 2000). Closeness to the people can be expressed by referring to oneself as one of the people (e.g., Engesser et al., 2016). In populist communication, politicians thus speak on behalf of the ingroup of ordinary people, often by using the first person (e.g., 'we', 'our') to identify themselves with the people whom they claim to represent. At the same time, populist politicians have been described as positioning themselves as reluctant politicians: ordinary people who have entered politics only due to the urgent need to alter or dismantle the established political system (Taggart, 2000). That tendency surfaces in the stylistic element of emphasizing distance from elites by speaking of them in second or third person (e.g., 'they', 'them'). We thus hypothesized: H1c: The discursive framework of populist communication is presented using the first person when speaking about the people and the second or third person when speaking about politicians.
Last, concerning populism's emphasis on common sense (Bos et al., 2010;Mazzoleni et al., 2003;Rooduijn, 2014), scholars have argued that populist political communication appeals to the so-called 'man on the street' by using plain language and a slogan-based rhetoric (e.g., Mazzoleni et al., 2003). Using such a presentation style, populist political communication is meant to appeal to an audience of working-class people who regard themselves as alienated from politics and its technocratic language (Kriesi et al., 2006). Thus, our next hypothesis states: H1d: The discursive framework of populist communication is presented using plain, common-sense references.

Context-level predictors of populist political communication
Along with the mentioned stylistic predictors, researchers have identified important contextual factors that can shape the expression of populist communication on Twitter and Facebook.

Type of SNS
Although few studies have involved differentiating SNSs as predictors of populist political communication, their results suggest that the SNS type is an important variable in the likelihood of populist political communication (Cranmer, 2011;Ernst et al., 2017). The most common and intensively used SNSs for political purposes in Europe are Facebook and Twitter (e.g., Dolezal, 2015;Ernst et al., 2017). This also holds true for the two countries of investigation, where all political candidates used Facebook and Twitter during the national elections campaign in 2017 to varying degrees, whereas only a few candidates used other SNSs such as Instagram and YouTube. Apart from the high penetration of Facebook and Twitter among the political candidates, these two SNSs also rank among the most frequently used SNSs among voters in both countries. In Austria, around 63% use Facebook, while around 12% of the population use Twitter on a regular basis (Reuters Institute for the Study of Journalism, 2018). In the Netherlands, Facebook and Twitter are the most popular SNSs. More than half (56%) of all internet users are active on one of those SNSs (Centraal Bureau voor de Statistiek, 2018), which underlines the importance of comparing these two SNSs in Austria and the Netherlands.
Overall, Facebook provides a more attractive environment for populist political communication than Twitter does, for it affords greater proximity to voters and more reciprocal communication. Moreover, Facebook is more popular and socially diverse than Twitter, which is more often used by journalists and other professionals (Jacobs & Spierings, 2018;Vis, 2013). Although not including Austria and the Netherlands, Ernst et al. (2017) found within a comparative study that political candidates were more likely to use populist political communication on Facebook than on Twitter, which they attributed to the greater reciprocity of messages on Facebook as well as to the primary use of Twitter for professional reasons, and the character limit it imposes on tweets. Accordingly, we proposed: H2: The discursive framework of populist political communication is more prevalent on Facebook rather than Twitter.

Party type
Furthermore, characteristics of the politicians' party may affect the degree they use the framework of populist communication. Researchers have indicated that extreme right-wing parties are particularly liable to use populist political communication online (Ernst et al., 2017;Krämer, 2014). However, populist communication may also be employed successfully by parties on the left, as the Greek party Syriza and the Spanish movement Podemos have exemplified (Jacobs & Spierings, 2018). Researchers have shown that populist political communication is more pronounced in press releases (Schmidt, 2017) and the SNS communication (Ernst et al., 2017) of members of political parties on the fringes of the political spectrum. Yet, existing research lacks a systematic comparison whether politicians of previously defined populist and non-populist parties differ significantly in the degree to which they use populist communication on SNSs. Thus, we aimed to examine whether populist communication is indeed more pronounced in the SNS communication of politicians affiliated to parties commonly classified as populist. Specifically, we hypothesized: H3: Political actors in left-and right-wing populist parties are more likely to use the discursive framework of populist political communication than their counterparts in mainstream parties.

Pre-versus post-election
Because research on populist communication has primarily focused on communication during election periods (Rooduijn, 2014;Schmidt, 2017), differences in populist communication before and after elections have largely been neglected (but see Jagers & Walgrave, 2007). Yet, the various motives of politicians before, during, and after elections also encourage the comparison of data from pre-and post-election periods. Elections stress the need to challenge opponents, blame other politicians for failures, and establish close relationships with followers. Such motives clearly resonate with the tendencies of populist discourse to stress the people's opposition to failing elites, the centrality of ordinary people, and the (unfulfilled) promise of representation. Given the alleged resonance of populist communication with politicians' self-presentation during elections, we assumed that: H4: The discursive framework of populist political communication is more salient before than after elections.
Acknowledging the strategic motives of mainstream politicians to appeal to ordinary people during election seasons (Bale, Green-Pedersen, Krouwel, Luther, & Sitter, 2010;Bos, van der Brug, & de Vreese, 2013), it is also conceivable that the increased use of populist communication's discursive framework during election seasons is more salient among mainstream than populist politicians. Indeed, several scholars have argued that mainstream political parties use populist communication before elections in order to compete with populist actors who promote antielitism, people-centrism, or popular sovereignty (Bos et al., 2013;Mudde, 2004). At those times, the strategic use of populist communication can conceivably help mainstream politicians to recover votes from populist challengers. Put differently, if the discursive framework of populist political communication is strategically used by politicians to generate votes, the discrepancy between the use of populist political communication before and after the elections should be more pronounced among mainstream political candidates compared to politicians in populist parties. It follows: H5: Political candidates in mainstream parties are more likely to use the discursive framework of populist political communication before than after the elections than their counterparts in populist parties.

Country
We conducted our comparative content analysis with data from Austria and the Netherlands. We chose those two countries, because national parliamentary elections were held in 2017 in both countries, which allowed a systematic comparison of the use of populist political communication in pre-and post-election campaigns across two different countries. Yet, Austria and the Netherlands make a good case for a comparative study for several other reasons: (1) Both are western European nations that have witnessed the fierce rise of right-wing populism in the last few decades. In both countries, rightwing populist parties -Freiheitliche Partei Österreichs (FPÖ) in Austria and Partij voor de Vrijheid (PVV) in the Netherlandsare highly successful. (2) Both countries fit the democratic corporatist model (Hallin & Mancini, 2004), which means they have comparable political systems and media systems. For instance, both countries are parliamentary representative democracies with a comparable involvement of the state in media regulation. Therefore, the motivation and opportunity structures for politicians to use SNSs might be similar in both countries. Despite the striking similarities of the countries, with the comparative scope of our paper we aim to furnish insights into how politicians' populist self-communication resonates with variations in national opportunity structures. We therefore posed a research question: To what extent does the salience of the discursive framework of populist communication differ in Austria and the Netherlands?

Method
We conducted a quantitative content analysis of the SNS communicationnamely, Facebook and Twitter postsof the leading candidates of parliamentary parties in Austria and the Netherlands.

Sample
Our sample included Facebook and Twitter posts (N = 1,010) of all 13 leading candidates of each parliamentary party before the 2017 national parliamentary elections in both countries (Appendix A). We collected posts made from 6 weeks before to 4 weeks after the elections. In the Netherlands, because the national parliamentary elections occurred on 15 March, the pre-election period spanned from 1 February to 15 March and the post-election period from 16 March to 13 April. In Austria, by contrast, the national parliamentary elections occurred on 15 October 2017, which meant a pre-election period from 4 September to 15 October and a post-election period from 16 October to 12 November. We constructed artificial weeks to ensure that each week and each weekday was represented in the sample. To achieve an equal number of posts during each period, we gathered posts on four days in each pre-election week and six days in each post-election week for each candidate. If more than one post was available on those days, we collected one post on Facebook and Twitter for each candidate on those days. If no post or tweet was available for the candidate on that day, then we collected one on the next day instead. This procedure resulted in a sample of maximum 24 tweets and 24 Facebook posts for each candidate before and after the elections (i.e., a maximum number of 96 units of analysis for each candidate, see Appendix A). In total, we collected a sample of 1,010 posts: 249 tweets and 264 Facebook posts from the Netherlands and 255 tweets and 242 Facebook posts from Austria. By period, pre-election posts totaled 322 from the Netherlands and 269 from Austria, whereas post-election posts totaled 191 items from the Netherlands and 228 from Austria.

Coding procedure
The unit of analysis was the textual content of a post or tweet, and we did not code images or videos. In each country, two independent human coders fluent in English and native speakers in their respective language (i.e., Dutch or German) performed data collection. After intensively training the coders, we conducted several rounds of intercoder reliability tests for each variable to assess both within-and between-country intercoder reliability. For within-country intercoder reliability tests, the coders analyzed the posts and tweets of each political candidate in each country across both pre-and post-election periods. For the between-country reliability tests, the coders analyzed only English-language Facebook posts and tweets. After each round of testing, we discussed inconsistencies with the coders and reran tests with different subsamples. The total sample for the intercoder reliability tests consisted of 220 Facebook posts and tweets (21.8% of the total sample), and for each variable, the four coders ultimately achieved acceptable levels of reliability. We computed the average Brennan and Prediger's kappa for each variable, which ranged between .71 and .98 and had a mean value of .91 (Appendix B).

Dependent variable
We based our measurement of populist communication as a dependent variable on Ernst et al.'s (2017) conceptualization, which comprises three sub frames of the framework of populist communication and nine frame elements (Table 2). We merged the nine frame elements into a formative index of populist political communication. For each category, we first coded whether a given populist statement was present. Next, we considered a dimension to be present when one of the frame elements was present. Last, we considered the index of populist political communication as a central dependent variable to be present when one of the three dimensions was present. We also coded the target of the populist message as elites, the people, or the candidate respective of his or her party. Following Ernst et al. (2017), we considered a sub frame of the framework of populist communication to be present when a frame element and the respective target were present as well (Table 2).

Stylistic predictors
To code the predictors of presentation style for each post, we first coded the use of negative and positive emotions. We designated certain keywords to identify positive emotions (e.g., hope, joy, and love) versus negative emotions (e.g., anger, hatred, and disgust; see Heiss et al., 2018;Schmuck, Heiss, Matthes, Engesser, & Esser, 2017). Second, we coded the post's overall positive or negative tonality (De Vreese et al., 2016;Heiss et al., 2018); in reference to the tone of the post's content, designations for positive tonality were e.g., political success, achievement, and improvement, whereas designations for negative tonality were e.g., political failure, disaster, and crisis. Third, to assess us-versus-them rhetoric, we coded whether the politician used the first person when talking about the people and whether he or she used the second or third person when talking about other politicians. Last, we operationalized common sense as a specific type of reasoning postulating that everyone should know a certain idea regardless of expertise. We formulated specific keywords and phrases for both countries to identify the use of common sense (e.g., 'Every child knows that' and 'It's very easy to understand').

Context-Level predictors
Regarding context, we coded the SNS type, the affiliated political party, the date, and the country of origin of each post. To assess party type, we created three party categories for each country: right-wing populist, left-wing populist, and mainstream. We identified Austrian candidate H.C. Strache (FPÖ) and Dutch candidate Geert Wilders (PVV) as rightwing populist candidates and Austrian candidate Peter Pilz (Liste Pilz) and Dutch candidate Emile Roemer (Socialist Party, SP) as left-wing populist candidates. We based our categorization on findings in recent literature showing that whereas the FPÖ and PVV are clearly described as right-wing populist parties (e.g., Aalberg et al., 2017;Rooduijn, 2014), the Dutch socialist party can be considered a left-wing populist party (e.g., Jacobs & Spierings, 2018;Rooduijn, 2014). In Austria, the representative left-wing populist party was the Liste Peter Pilz, which, founded in 2017 by a former member of the Green Party as a left-wing populist counterpart to the FPÖ, entered parliament after the 2017 elections. The party is regarded as a 'left-wing populist party' (e.g., Eberl, Zeglovits, & Sickinger, 2017), which combines 'less liberal positions on migration and integration policies with a strong leftist social agenda reminiscent of Podemos or Syriza' (Buzogány & Scherhaufer, 2018, p. 568). Additionally, its leader, Peter Pilz, has a well-known reputation as 'corruption fighter' (Plescia, Kritzinger, & Oberluggauer, 2017, p. 188). We identified all other parties as mainstream or moderate and grouped them into one category.

Data analysis
To analyze the predictors of populist political communication, we ran logistic regression analyses. We included fixed effects of the 13 politicians (dummy variables with Sebastian Kurz as reference category) as covariates in the analysis. We dummy coded all other predictors prior to entering them into the analyses. To account for interaction effects between party type and election period, we ran stepwise logistic regression analyses with two-way interactions in Model 2. Last, we included mean-centered differences in vote shares of each party before and after the election as a control variable to ensure that our effects are independent of the party's success in the election. We computed the difference by subtracting the pre-election vote share from the post-election vote share, meaning that higher levels indicated an increase and lower levels a decrease in vote share after the elections.

Results
In total, 138 of 1,010 Facebook posts and tweets (13.7%) contained at least one sub frame of the framework of populist communication. Of those posts, 59.4% (n = 82) were issued on Facebook. Antielitism was most prominent (7.6%), closely followed by the sub frame of people-centrism (7.4%), whereas the sub frame of popular sovereignty was only half as prevalent (3.8%). On Facebook, 43 posts contained the antielitism sub frame, 53 posts the people-centrism sub frame, and 17 the popular sovereignty subframe. On Twitter, 34 posts contained the antielitism subframe, 22 posts the people-centrism subframe, and 21 the popular sovereignty subframe. Detailed results appear in Tables 1 and 2. To analyze our hypotheses and research question, we ran logistic regression analyses (Table 3). Concerning stylistic elements (Table 3, Model 1), the framework of populist communication was more likely to be presented with negative emotions (b = 0.85, SE = 0.43, p = .049), but not with positive emotions (b = -.04, SE = 0.35, p = .903). Therefore, H1a was only supported for negative but not positive emotions. Furthermore, the populist framework was more likely to be related to negative tonality compared to neutral tonality (b = 2.53, SE = 0.35, p < .001). Additionally, although not hypothesized, the use of subframes of the populist framework was also significantly more likely to be presented in a positive compared to a neutral tonality (b = 1.07, SE = 0.32, p = .001). Therefore, H1b found partial support as well. As expected, we also found support for the assumption that first-person references to the people (b = 1.57, SE = 0.31, p < .001) and third-person references to politicians (b = 1.31, SE = 0.25, p < .001) strongly resonated with subframes of the populist framework. Hence, H1c found satisfactory support. However, because 1.2% (n = 1) references to common sense did not significantly relate to the framework of populist political communication (b = −0.55, SE = 0.71, p = .437), we rejected H1d.
Regarding variables of context, subframes of the populist framework were more prevalent on Facebook than on Twitter (b = 0.82, SE = 0.25, p = .001), which fully supported H2. Furthermore, right-wing (b = 1.46, SE = 0.64, p = .022) and left-wing populist candidates (b = 1.92, SE = 0.62, p = .004) were more likely to use subframes of the populist framework than mainstream politicians were. Thus, H3 also received support. Additionally, pre-election posts were significantly more likely to contain subframes of the populist framework than post-election posts (b = 0.68, SE = 0.25, p = .006), which provided support for H4. 564.80 568.74 + p < 0.1; * p < 0.05; ** p < 0.01; *** p < 0.001. a The political candidate was controlled by using dummy variables for each political candidate with the first candidate as reference category (not shown here for clarity reasons). Finally, we tested whether mainstream candidates as opposed to right-wing or left-wing populist candidates were more likely to use the populist framework in pre-compared to post-election periods (Table 3, Model 2). However, because no significant interaction effect emerged between election period and mainstream candidates as opposed to rightwing (b = −0.04, SE = 0.56, p = .945) or left-wing candidates (b = 0.16, SE = 0.75, p = .827), we had to reject H5. Overall, subframes of the populist framework were slightly more prevalent in Austria than in the Netherlands (b = 1.29, SE = 0.70, p = .064). However, this difference failed to reach statistical significance. Therefore, answering our research question, we did not find significant country differences with regard to the use of populist political communication.

Discussion
Despite politicians' increased use of SNS for their political communication, few researchers have examined predictors of the use of populist political communication in such networks. Accordingly, we aimed to identify the stylistic and contextual factors that can predict the use of politicians' populist communication on Facebook and Twitter.
Our results revealed that negative emotions and negative tonality have informed politicians' use of the populist framework, which resonates with previous findings that have characterized populist communication as highly emotionalized and oriented toward conflict (e.g., Fieschi & Heywood, 2006;Taggart, 2000). At the right end of the ideological spectrum, the populist discourse often blames immigrants for taking native citizens' jobs or for threatening the nation's safety, all of which engenders a sense of conflict, crisis, and negativity (Jagers & Walgrave, 2007;Schulz et al., 2017;Taggart, 2000). Additionally, populist communication associated with traditional left-wing political topics such as critique of the political or economic elites are also highly likely to be presented in a negative way. However, populist subframes also appeared in posts with an overall positive tonality. That finding is understandable, since some components of populism, including an emphasis on the virtues and achievements of the people and politicians' closeness to them, are quite likely to be presented in positive ways (Aalberg et al., 2017). Thus, our findings suggest that politicians engaging in populist political communication do not use explicit positive emotions such as hope, joy, or love, but they do describe positive developments or political successes. As such, populist political communication is not merely restricted to negativity.
The use of the populist framework also resonated with the use of us-versus-them rhetoric of the political candidates, who, to convincingly position themselves as members of the populace, used stylistic means to assert their closeness to the people and emphasize their distance to the elites. However, references to common sense did not relate to the use of the framework, and in fact, the politicians whom we studied collectively made such references in less than 2% of their posts. It is possible that using common-sense references is less important for politicians when communicating via SNS than via other channels, since the former allow them to communicate directly to the people. In sum, our results identifiedfor the first time in a comparative contextspecific elements of presentation style that are likely to co-occur with the framework of populist communication on SNSs, which provide politicians with tremendous potential for personalization and freedom to shape their messages compared to traditional means of communication (Engesser et al., 2016).
Regarding contextual predictors, the framework of populist communication was more prevalent on Facebook than on Twitter during both the pre-and post-election periods. The specific characteristics of Twitterthe character limit imposed upon tweets, the medium's largely professional audience of journalists and other specialists, and the messagerelated characteristics of tweetsseem to provide a less suitable environment for populist political communication (Ernst et al., 2017;see Jacobs & Spierings, 2018). In fact, our descriptive findings suggest that the difference between Facebook and Twitter was most pronounced for the people centrality framepresumably because politicians assume that the 'ordinary people' are more likely to be found on Facebook than Twitter, the audience of which is often described as more professional or higher educated (Ernst et al., 2017). As expected, right-wing and left-wing populist politicians were generally more likely than mainstream or moderate ones to use elements of populist communication, which is in line with previous results suggesting that populist candidates are more likely than others to attack elites (e.g., Engesser et al., 2016;Ernst et al., 2017;Krämer, 2014).
By period, all candidates were more likely to make populist statements before than after elections, which suggests that populist political communication is indeed used as a strategic method or instrument of maximizing votes during election season (Weyland, 2001). However, the increased use of populist communication's discursive elements pre-election compared to post-election was not more salient among mainstream than populist candidates. Rather, all politicians were more likely to use elements of populist communication before compared to after the elections.
Last, we found no notable differences with regard to the prevalence of populist political communication on SNS between the two countries, Austria and the Netherlands. The comparable media and political systems (Hallin & Mancini, 2004), which may lead to similar motivations and opportunity structures for political candidates to express themselves via SNS may explain the similarity of the findings.
Altogether, our findings indicate that populism cannot be exclusively understood as political ideology, strategy, or style. Instead, our findings suggest that the ideological core of populist political communication operationalized as the subframes antielitism, people-centrism, and popular sovereignty (see Ernst et al., 2017;Schulz et al., 2017) cooccurs with stylistic elements such as negative emotions, negative and positive tonality, or us-versus-them rhetoric in politicians' self-communication on SNSs. Additionally, the more frequent use of the populist framework during election periods suggests that politicians from all political camps employ populist communication strategically presumably with the goal to maximize votes (Weyland, 2001). In sum, we argue that understanding populism as a discursive framework of communication helps to reconcile existing divergent conceptualizations of populism by referring to the communicative act through which the populist dichotomy between 'the pure people' versus 'the corrupt elite' can be expressed by different actors with different political goals (Aslanidis, 2016).

Limitations and future research
Our study involved some notable limitations. First, the percentage of posts, which actually contained populist communication was rather low, although we coded a high number of posts and tweets for each political candidate across two countries. The small sample size did not allow a detailed analysis of the stratification of populist political communication among populist and mainstream actors, in two countries and on two SNSs. Future research should therefore extend the research presented here to a larger sample of politicians and countries.
Furthermore, our analysis focused on written text only, not pictures or videos, because those cues are multifaceted and contain various meanings that are more difficult to code or interpret in a valid, reliable way. In the same vein, we focused only on two specific SNSs: Facebook and Twitter. Scholars should analyze additional SNSs used for populist messages, including YouTube and Snapchat. Moreover, because we focused on populism from the supply side (i.e., from political actors), researchers should build upon our findings by investigating how the specific stylistic elements that we identified influence attitudes on the demand side (i.e., populist attitudes) both before and after elections.
Those limitations notwithstanding, our findings contribute to contemporary research on populist political communication online by shedding light on both the stylistic elements of populist communication as well as the contextual structures that shape political candidates' populist communication. Last, our findings underscore the importance of distinguishing election and non-election periods in research on populist political communication online.