Europe against the people: does eurosceptic news exposure relate to populist attitudes? Evidence from a linkage study across nine European countries

ABSTRACT

The media’s attention to populism is oftentimes associated with its electoral success. In this paper, we propose that the relationship between media coverage and populist beliefs among citizens can be understood as an over-time cultivation of support for populist ideology due to the media’s attention to populist ideas and policies. The media may devote attention to ideas high on the populist agenda, such as Euroscepticism, which may prime populist worldviews among individuals. To test this expectation, we first use a comparative study in nine EU countries for which we link content analytic data on negativity toward the EU in the media to panel survey data measuring populist attitudes. Overall, we find that exposure to negative news on the EU does not trigger populist attitudes. In a more detailed case study in the Netherlands across multiple panel waves, we do find that social media exposure is more likely to be associated with populist attitudes than traditional news exposure. Together, our study offers limited support for the notion that attention to Euroscepticism in established media can fuel populist attitudes among voters.

KEYWORDS:

• Euroscepticism
• populist attitudes
• news media
• social media

Over the past decades, populist parties have gained electoral success in many European countries. Media coverage on populist parties and their ideas has been regarded as an important factor explaining their (electoral) success and visibility (Mazzoleni 2008). Arguably, the media devote disproportionate attention to the views and political agendas of populist parties – without critically scrutinizing their viewpoints (Boukes and Hameleers 2020). The media’s attention to populism can be regarded as disproportionate as, relative to other parties with a similar vote share, populist ideas may be covered more saliently in the press because they resonate with prominent news values focusing on conflict and negativity (e.g. Bos and Brants 2014; Mazzoleni 2008). In support of this media logic argument, experimental research (Bos et al. 2019; Matthes and Schmuck 2017) and panel survey studies (Müller et al. 2018) have offered evidence for a causal link between populist news exposure and populist attitudes.

Despite the important role assigned to the media, content analyses found that explicit populist ideas that emphasize an antagonism between the virtuous people and the unvirtuous elite only occur seldomly in news content (Bos and Brants 2014; Wettstein et al. 2018). Against this backdrop, we propose that the link between media coverage and populist support can best be understood as an association between the media’s attention to populist ideas and policies and over-time support for populist ideas among recipients. Specifically, the media may devote (disproportionate) attention to issue positions owned by populists, such as Euroscepticism (e.g. Aalberg et al. 2017; Arzheimer 2015; Pirro, Taggart, and van Kessel 2018), which may activate populist interpretation frames among people consuming such content. In support of this mechanism, a strong correlation between populist attitudes and Eurosceptic views has been established in prior survey research (Santana-Pereira and Cancela 2020; Tsatsanis, Andreadis, and Teperoglou 2018). In this paper, we rely on a panel survey study linked with content analytical data to explore the causality of this association, and the media’s role in priming populist associations.

Euroscepticism is salient in the discourses of populist movements across the European continent (Pirro, Taggart, and van Kessel 2018). Both left-wing and right-wing populist movements can relate their anti-establishment rhetoric and blame attributions to widely accepted negative sentiments toward the EU, and connect EU negativity to salient crises, such as the refugee crisis or economic recession. Media coverage promoting this Eurosceptic agenda may consequentially activate populist attitudes among news consumers by promoting the idea of a central antagonism between the ordinary people and the unresponsive or corrupt elites in (supra)national politics. As populist attitudes are strongly related to populist party preferences (e.g. Akkerman, Mudde, and Zaslove 2014) and considering the correlational evidence on the link between populist attitudes and Euroscepticism (e.g. Tsatsanis, Andreadis, and Teperoglou 2018), we believe that exposure to negativity towards the European Union may prime individual-level support for populist viewpoints.

To study these dynamics, we pose the following research question: Does exposure to Eurosceptic news coverage fuel populist attitudes among citizens in Europe? Offering empirical support for a relationship between populist news exposure and populist attitudes, Müller et al. (2018) found that exposure to fragmented elements of the populist ideology in news coverage can make prior populist worldviews more pronounced. In this paper, considering the fact that explicit elements of populist ideology are only seldomly expressed in print news (Bos and Brants 2014; Wettstein et al. 2018), we look at more general content elements of negativity toward the EU. Hence, different from Müller et al. (2018) who looked at the effects of exposure to populism by the media (i.e. explicit elements of the populist ideology in media content), we explore the relationship between populism for the media (i.e. attention to issues owned by populist parties) and populist attitudes. In this paper, we rely on an extensive operationalization of media exposure, incorporating exposure to TV news, newspapers, online news and social media use. To match media content exposure to populist attitudes measured in original survey data, we rely on representative data collected in ‘most different’ European countries.

In the sections that follow, we first of all review extant literature on the relationship between populist media coverage and populist attitudes. After this, we explain the general set-up of our studies that test the association between exposure to Euroscepticism and populist attitudes. Then, we report the methods and findings of a nine-country linkage study in which we explore whether exposure to negative media content on the EU corresponds with higher levels of populist attitudes. In a second case study on the Netherlands, we offer more robust causal evidence for this relationship by using more detailed media exposure measures connected to over-time changes in populist attitudes. By exploring the relationship between negativity towards a salient scapegoat in European populists’ discourse – the EU – and populist attitudes on the individual level, our paper aims to offer new insights into the relationship between attention to populist issues in the media and support for populist ideas among voters.

Populism and populist attitudes

Populism can be understood as a style, strategy, or (thin-centered) ideology in which the antagonistic divide between the ‘ordinary’ people and the culprit ‘corrupt’ elites is central (e.g. Aalberg et al. 2017; Canovan 1999; Mudde 2004; Taggart 2000). Populist ideas hold that politics should revolve around the ordinary people’s general will. Populism can be regarded as a social identity frame in which the ‘good’ ordinary people are pitted against the generalized ‘evil’ elites that deprive the people’s in-group of what it deserves (Bos et al. 2019). In this paper, we specifically look at the association between populism for the media (e.g. Bos and Brants 2014) and populist attitudes among voters (see e.g. Akkerman, Mudde, and Zaslove 2014; Schulz et al. 2018). Populism for the media can be understood as the media’s receptivity to populist ideas, policies and actors (Bos and Brants 2014). This relationship has been regarded as a consequence of the market orientation of the media in Europe and beyond (e.g. Aalberg et al. 2017; Mazzoleni 2008). The media’s focus on negativity, conflict and other styles that attract audiences in a competitive market aligns with populist ideas and policies that focus on an antagonism between the ‘good’ people and ‘evil’ elite actors.

We specifically assess how the attention for populist policies in the media – in this case Euroscepticism – relates to populist attitudes among voters. Populist attitudes are defined as the perceived socio-political opposition between ordinary or pure people (an in-group) and corrupt and self-interested elites (the out-group blamed for the people’s problems) (e.g. Akkerman, Mudde, and Zaslove 2014; Schulz et al. 2018). We herewith follow existing conceptualizations that have measured populist ideas among citizens as support for the core ideas of populism: The antagonism between ordinary people and corrupt elites and the homogeneity of the ordinary people’s will (Akkerman, Mudde, and Zaslove 2014; Schulz et al. 2018). It has been found that these populist attitudes are a strong predictor of populist party preferences (e.g. Akkerman, Mudde, and Zaslove 2014). In other words, the more pronounced people’s populist attitudes are, the more likely they are to vote for populist parties. Populist party preferences may, however, be rather stable and not affected by exposure to negative news on the EU on the short term. We therefore focus on populist attitudes that offer an indicator for the extent to which attention to populist ideas in the media may relate to support for populist worldviews.

Different from our approach, extant research has predominately approached the media-populism relationship as the effect of populism by the media on populist attitudes (e.g. Hameleers, Bos, and de Vreese 2017; Müller et al. 2018). Populism by the media can be understood as the use of explicit populist frames and populist ideas by the media, which implies that journalists actively interpret issues within a populist frame of reference (e.g. Bos and Brants 2014). Deviating from this approach, we look at a more indirect and over-time association between attention to populist policies (populism for the media) and individual-level support for populist ideas. We specifically postulate that media diets containing a higher dose of negativity toward elite actors scapegoated in populist discourse – such as the EU – may prime negative perceptions related to elite actors in general, which are captured with populist attitudes.

Populist attitudes and euroscepticism

Extant survey research has found a strong relationship between populist attitudes and Euroscepticism (e.g. Santana-Pereira and Cancela 2020; Tsatsanis, Andreadis, and Teperoglou 2018). Here, we define Euroscepticism as all forms of skepticism, negativity and critique toward the European Union and its institutions (also see e.g. Pirro, Taggart, and van Kessel 2018). Importantly, Euroscepticism is a catch-all concept that differs greatly across national settings (e.g. Leconte 2010). It may generally include hostility against the EU, but its critique can be related to different aspects, such as identity-driven considerations related to integration and European unification or economic concerns related to EU-membership (Leconte 2010).

In line with extant literature, we regard Euroscepticism as central to the policies and styles of different populist parties in Europe (e.g. Pirro, Taggart, and van Kessel 2018; Vasilopoulou 2018). Considering that Euroscepticism is just as adaptable as populist ideas, and because skepticism toward the EU can relate to a wide range of left-wing and right-wing issue positions as well as different crisis sentiments, it can be regarded as a unifying feature of the many different populist parties in Europe (e.g. Pirro, Taggart, and van Kessel 2018). Thus, although populism and populist parties in Europe are different when it comes to their issue positions and ‘thicker’ ideologies (e.g. Mudde 2004), they share their opposition toward the EU project and the elite institutions governing it.

Although Euroscepticism can connect to more hostile and uncivil discourse resonating with an explicit anti-elitist framing of Europe, we do not differentiate between different degrees of negativity and critique in this paper. Rather, we explore whether overall negative evaluations toward the EU, its institutions and the European project can ‘awaken’ populist associations in the minds of recipients. Herewith, we aim to capture a wide range of negative expressions related to the EU, which may include both ‘hard’ and ‘soft’ forms of skepticism (Taggart and Szczerbiak 2002). We take into account that evaluations of the EU mostly do not include a clear assessment of whether the EU is good or bad, especially in quality media that mostly discuss policies (Bijsmans 2021). Against this backdrop, we regard overall negatively biased evaluations of the EU, its institutions and project (as compared to neutral, balanced, positive or no valanced evaluations) as the minimal and universal defining feature of Euroscepticism.

The core idea of populism is ‘thin’ regarding the ideology it conveys. Yet, most research in Europe has focused on right-wing populism, which could be explained by the relative popularity of these movements across the continent during the past decades (Aalberg et al. 2017). Right-wing populist ideas salient in many European nations oftentimes combine populism’s antagonistic worldview with an exclusionist understanding of the people (Jagers and Walgrave 2007). However, as also noted by Pirro, Taggart, and van Kessel (2018), Eurosceptic views can relate to both left- and right-wing populist ideas, albeit in different ways. Whereas right-wing Euroscepticism focuses on European integration and immigration, left-wing interpretations focus on the EU’s failures regarding socio-economic issues. In the populist democracy of Greece, it has been argued that populist challenger parties may emphasize extreme positions toward the EU when confronted with a crisis (Vasilopoulou 2018). Others studies also found that critique toward the EU may take on a more substantial form in election times, when (populist) challengers may clearly express their opposition toward European integration (e.g. De Vreese et al. 2006). In light of this variety across the European continent, where both left- and right-wing populists have successfully integrated an Eurosceptic agenda in their rhetoric (e.g. Aalberg et al. 2017; Pirro, Taggart, and van Kessel 2018; Vasilopoulou 2018), our study looks at a variety of European regions and approaches populist attitudes as individual-level support for a thin ideology that can attach itself to both left- and right-wing issue positions.

Extant literature on the link between media exposure and the increasing success of populist movements in Europe has argued that the media may have contributed to the success of populism (Aalberg et al. 2017; Krämer 2014; Mazzoleni 2008). Arguably, the media’s role in fueling the success of populist parties and ideas is based on an alignment between the issue positions and styles of populists and the media: Populist actors raise conflict, promote sensation, dramatize political issues and focus on emotionalized styles – which may resonate with the current media logic and dominant news values (Hameleers, Bos, and de Vreese 2017; Mazzoleni 2008). News media may thus devote disproportionate attention to populist parties and their agendas as these are newsworthy and conducive to a logic that attracts a large audience in a competitive market. Especially in liberal democracies and multi-party systems, the media are ought to devote attention to all different viewpoints and ideologies – which also makes populist ideas visible in mainstream news coverage. Hence, as populist parties have become part of the political landscape, the media cannot ignore them and have to pay attention to their ideology – which may be amplified due to the resonance between news values and populist styles of communication.

Despite this theoretical assumption, empirical evidence on the salience of populism in the news media is scarce at best (see e.g. Bos and Brants 2014; Wettstein et al. 2018). In this paper, we therefore argue for a more indirect relationship based on the associations between Euroscepticism and populist attitudes: By devoting more attention to the issues and policies addressed by populists, support for populist ideas among voters may be cultivated. In line with this, the media only seldomly promotes populist ideologies, but increasingly devotes attention to (right-wing) populist issue positions (Hameleers and Vliegenthart 2020), potentially normalizing and legitimizing populist worldviews among the electorate.

The activation of populist worldviews among European news consumers

Research on the effects of (right-wing) populist communication has been growing in recent years (see e.g. Bos, van der Brug, and de Vreese 2013; Busby, Gubler, and Hawkins 2019; Hameleers, Bos, and de Vreese 2017; Matthes and Schmuck 2017; Müller et al. 2018). Most studies have relied on a (forced exposure) online experiment in which people are exposed to the ideational core of populism after which their populist worldviews are measured (e.g. Bos et al. 2019; Busby, Gubler, and Hawkins 2019). These studies have offered important insights into the persuasiveness of populist messages: Populist messages can activate populist attitudes (Bos et al. 2019), activate right-wing populist party support (Hameleers, Bos, and de Vreese 2017) or augment polarized divides in society (Müller et al. 2018). As argued by Müller et al. (2018), experimental approaches also bear important limitations. Importantly, such designs ignore the fragmented nature of populist messages’ production and consumption (e.g. Engesser et al. 2017).

In a more ecologically valid assessment – pairing self-reported news exposure and populist attitudes with an extensive content analysis of populist communication – Müller et al. (2018) found that populist news exposure may have a polarizing impact on society. In this setting, it was found that populist news exposure reinforces increasing cleavages between people supporting and opposing populism’s antagonistic worldview. This effect can be explained using the ideational approach to populism (Busby, Gubler, and Hawkins 2019; Hawkins et al. 2018), According to the ideational approach, people with (moderate levels of) prior populist attitudes are affected most by populist messages, as these messages cultivate or activate populist interpretations of reality that are existing as ‘dormant’ traits that need to be ‘awakened.’ In other words, when people have existing but dormant evaluations related to populism available, exposure to media content in which elements of populist ideas, policies or styles are emphasized may activate or trigger a more encompassing set of evaluations in line with populism’s ideology.

Theoretically, this corresponds to schema theory, which postulates that people’s cognitions are structured by different clusters of topics and associations (Brewer and Nakamura 1984). In line with this mechanism, Krämer (2014) has argued that exposure to populism in the media should activate related populist schemata among recipients. In the context of our paper, we regard Eurosceptic elements in news coverage as cues related to populist elements. Exposure to these elements should, in line with the mechanism of schema theory, activate more comprehensive networks of associations related to populism in the minds of receivers. In this paper, we regard these related cognitive networks as populist attitudes (also see Müller et al. 2018). Hence, exposure to media content containing elements of populist ideas or policies may activate congruent populist schemata among receivers, which are typically measured as populist attitudes among voters (see Krämer 2014 for a theoretical assessment).

The underlying mechanisms operating in the ‘black box’ of the activation of schemata can be understood as the priming of existing and available topic clusters related to populism. Hence, exposure to populist messages may prime populist attitudes mostly in conditions where populist interpretations are – to some extent – relevant and mentally available (Busby, Gubler, and Hawkins 2019). Priming can generally be understood as the activation of mental constructs that affect individuals’ evaluations of other ideas and concepts (Domke, Shah, and Wackman 1998). More specifically, to evaluate political issues and the performance of politicians, people use the issues that are most salient and retrievable from memory (Iyengar and Kinder 1987). The issue ownership literature further postulates that citizens will prefer political parties that they consider to be best able to deal with the most salient issues (e.g. Budge and Farlie 1983). Thus, when Euroscepticism is foregrounded as a salient issue through news coverage, people may prefer populist parties that own this issue, and promise to deal with the issue that has been deemed as crucial. Although we do not directly look at populist vote choice or party preferences, extant research has found that populist attitudes are very strongly related to populist party preferences (e.g. Akkerman, Mudde, and Zaslove 2014). Therefore, the priming of populist attitudes may eventually increase the likelihood that people vote for populist parties whose ideology aligns with these beliefs.

Applying this understanding of issue ownership and priming to electoral volatility, Geers and Bos (2017) found that favorable news coverage on a party makes voters more likely to switch to that party. They explain this as a confirmation of a priming hypothesis: Getting a party’s owned issue in the media may foster electoral success as the most salient attributes of political parties that are emphasized in media coverage are considered when people evaluate political parties (e.g. Kim and McCombs 2007). Extrapolating these mechanisms to the relationship between exposure to Euroscepticism in the media and populist attitudes, we propose that by making EU negativity – an issue owned most by populist parties – more salient in the minds of voters through media coverage, these primed evaluations are taken into account when people arrive at political judgements. By contributing to the heightened accessibility of populist issues and ideas among news users, exposure to (news) media that devotes attention to populist issues may thus activate populist attitudes, even beyond the conscious awareness of citizens exposed to populist media content (Hoewe 2020).

Priming has been connected to framing effects (e.g. Domke, Shah, and Wackman 1998; Scheufele 1999). Frames in communication can be understood as patterns of interpretation in a text that increase the salience of certain elements of a phenomenon or event whilst leaving out others (e.g. Entman 1993). Through these patterns, emphasis frames suggest an interpretation to audiences, for example, regarding the causal interpretation or moral evaluation of an issue. A framing effect would occur when the frames in communication (i.e. emphasizing certain problems and causes) prime congruent or related interpretation frames (e.g. Chong and Druckman 2007). Here, we follow the argument of Domke, Shah, and Wackman (1998) that the media framing of issue information, in this paper presented as negativity toward the EU, can activate particular cognitions that recipients use to evaluate other elements in their political environment. More specifically, exposure to negativity and critique on the EU in news coverage may activate related anti-elite associations which may consequentially be used to evaluate political issues with a populist frame of reference – which we measure as individual-level populist attitudes. Against this backdrop, we formulate the following central hypothesis of this study: Higher exposure to Eurosceptic views in the media results in higher levels of populist attitudes among citizens (H1).

Acknowledging the fragmented spread of populist ideas in the media (Engesser et al. 2017) as well as limited support for the existence of media populism or populism by the media (Bos and Brants 2014; Wettstein et al. 2018), we thus postulate that exposure to viewpoints and issue positions associated with populist movements in the media can trigger or activate related cognitive schemata and more comprehensive populist schemata – also referred to as populist attitudes (e.g. Akkerman, Mudde, and Zaslove 2014; Schulz et al. 2018). Here, we should acknowledge that populist messages or Eurosceptic content may not have the same effect for all recipients. Specifically, as indicated in extant literature on the effects of populist communication, populism may have positive effects on populist attitudes for some citizens, whereas others reject or even counter-argue populist arguments (see e.g. Aalberg et al. 2017; Reinemann et al. 2019). Factors such as political cynicism, relative deprivation or ideological alignment may result in stronger effects, whereas a pre-existing ideological disagreement with left- or right-wing populist ideas may diminish effects (e.g. Reinemann et al. 2019). Against this backdrop, although we control for individual-level differences that may influence the hypothesized relationship, we offer an overall assessment of the link between exposure to Eurosceptic views and populist attitudes.

We should also point out that the causal relationship forwarded here may be part of a reinforcing spiral, instead of a one-directional impact of exposure to Eurosceptic views on populist attitudes. As also indicated by Galpin and Trenz (2017), a negativity bias towards the EU in the media can amplify negativity towards the EU in public opinion, which may subsequentially have an impact on political parties’ coverage of the EU. As the media may be inclined to devote more attention to anti-establishment populist viewpoints out of strategical considerations and market orientations (e.g. Mazzoleni 2008), they may pay more attention to populist policies when populist attitudes are perceived as more salient. At the same time, it can be argued that the higher levels of populist attitudes among voters, the more likely they are to expose themselves to content that resonates with their populist beliefs (e.g. Hameleers, Bos, and de Vreese 2017; Müller and Schulz 2021). This confirmation-biased selection may consequentially result in higher levels of populist attitudes, creating a reinforcing spiral of congruent populist news exposure and populist attitudes.

Euroscepticism’s impact across different media platforms

Empirical research has demonstrated that populist communication is most likely to be expressed on online (social) media platforms (e.g. Jacobs and Spierings 2019). The affordances of interactivity and direct (unmediated) communication should offer an ideal contextual opportunity for populist discourse that circumvents elite actors and journalist gatekeepers (Engesser et al. 2017). In addition, the resonance between right-wing populism and media critique (i.e. accusing the media of spreading disinformation with the use of ‘fake news’ labels) indicates that populism frequently delegitimizes mainstream information and news channels (e.g. Egelhofer and Lecheler 2019) – making online and social media in particular a more likely forum for the expression of populist elements and ideas than mainstream (news) media.

The resonance between populism and online news media formats has also been established in public opinion research (Müller and Schulz 2021). More specifically, people with stronger populist attitudes are found to consume more alternative media – and rely heavily on Twitter and Facebook to receive information about politics. As people with populist attitudes also tend to perceive the mainstream media as an enemy of the people (e.g. Fawzi 2018), it can be argued that there is a specific affinity between exposure to online (alternative) and social media that circumvent gatekeepers of the mainstream press and populist attitudes.

On social media, people have a higher likelihood to encounter populist views and negativity targeted at elite actors (Waisbord and Amado 2017). Corresponding with their role performances striving toward balance, objectivity and rationality, established media may be more supportive and less hostile toward the EU than online media. On the receiver-side, people with higher populist attitudes are more likely to get their news from alternative information sources and non-established news formats (Müller and Schulz 2021). As these information sources are more likely to contain anti-establishment views and negativity toward elite actors (e.g. Holt 2018), people consuming online and social media as opposed to established news sources may be exposed to the highest dose of Euroscepticism. As illustrated by Michailidou’s (2015) qualitative analyses of Euroscepticism across more professional and user-driven digital media, the tone of EU-driven negativity and critique is more hostile and explicitly negative on social media compared to traditional journalistic content (Michailidou 2015). In the online public sphere, disenchantment and anti-elite sentiments targeted at EU policy prevailed, which is different from the more civil critique in professional news coverage (Michailidou 2015). The centrality of identity and security threats associated with the EU on social media more strongly resonates with populist interpretations compared to the more general and substantive critique on the EU in traditional media. Based on this, the association between Eurosceptic news coverage and populist attitudes should be most pronounced for social media exposure.

Considering that people with populist attitudes may select online and social media as this reinforces their anti-establishment views (e.g. Müller and Schulz 2021), we expect that the relationship between exposure to Euroscepticism and populist attitudes is especially pronounced for online and social media use: The relationship between exposure to Euroscepticism and populist attitudes is more pronounced for online or social media compared to mainstream or established media exposure (H2). This second hypothesis is tested in one single country (the Netherlands, Study 2) where more detailed media exposure data is available across different waves.

Data and methods

For this paper, we rely on two studies that are part of a bigger project. As the two studies are similar in their approach and the data used, we will explain the wider methodological framework that applies to both studies before delving deeper into the separate studies. Specifically, the data were collected as part of a larger multi-country and longitudinal project in the context of the 2019 European Parliament elections (Goldberg et al. 2021). This project included both a content analytical component as well as a survey among the public. Both components were linked to each other and followed an over-time logic by covering media content of up to almost two years prior to the elections and in parallel up to seven survey panel waves (albeit not for all countries as we explain below). Especially relevant for this study, the covered media outlets were harmonized across both components as to allow for linking media content to news exposure as reported by the survey respondents.

For our analysis, we rely on data from nine member states of the European Union (EU): Denmark, France, Germany, Greece, Hungary, the Netherlands, Poland, Spain, and Sweden. Although not representing the whole EU-27, the nine countries are a representative sample of the EU by consisting of smaller and larger EU member states, geographically spread across Europe and comprising both founding EU members as well as countries from later EU enlargement rounds. Furthermore, the countries differ in their experiences in and positions towards the EU, which may be relevant for the related EU media coverage, and also with regards to the strength of populist parties that may be reflected in the different prominence of populist attitudes among the respective populations (e.g. Aalberg et al. 2017).

The content analysis

In order to cover the variety of news sources that people may be exposed to, capturing the ideological diversity and most prominent news sources in the different countries, the content data in most countries include the two most watched TV news shows from public and private broadcasters, at least two newspapers (a broadsheet and a tabloid), and at least two widely used online news websites. Yet, due to differences in media systems and restricted availability of some news sources, there are exceptions to this rule by including less or additional media sources (see Brosius et al. 2021 for a more detailed documentation including an overview of all media sources sampled in each country; and consult the Reuters Institute Digital News Report for the national reach of the respective media under https://www.digitalnewsreport.org/). The content analysis was limited to news coverage that was related to the European Union. In practical terms, every fourth TV news show was sampled and every third article for newspapers and online news (in chronological order). This made the coding by native-speaking student coders feasible but still yielded a sufficiently large sample size. As a consequence of this coding procedure (and the respective time periods covered), our content measure across all nine included countries is mostly based on outlets with at least 30 news items coded, on average 65 news items coded per outlet. However, see some few exceptions in the final list of included news outlets per country in Table A1 in the appendix. When zooming in on the Netherlands (see explanation below), the data covers between 53 and 241 news items per outlet. Table A2 in the Appendix provides further details about the overall N of coded news items (and share of coded EU negativity).

The content analytical data include a large battery of items on the European Union and its evaluation. We specifically look at one key variable in this study: Negativity expressed towards the EU. Here, it should be noted that we apply a broad operationalization of Euroscepticism, referring to any criticism, negativity or attack on the EU and European integration. Considering the fact that explicit populist frames are an extremely rare event in media coverage (e.g. Hameleers, Bos, and de Vreese 2017) and our aim to test a more indirect linkage between exposure to skepticism targeted at the EU and populist worldviews, we did not distinguish between explicit populist and non-populist frames of Euroscepticism. We thus use a general measure that is ought to incorporate the various frames that may be emphasized through Euroscepticism.

Our independent variable more specifically measures the overall evaluation of the EU and/or its institutions with three categories representing (rather) negative, balanced/mixed or (rather) positive evaluations in addition to the absence of any explicit evaluation. This variable was only scored when the EU or related institutions were mentioned in the article (including synonyms), and coders had to score negativity on the article level, and had to arrive at an overall verdict in cases where the EU was mentioned more than once (explicit coding instructions are included in Appendix A; inter-coder reliability based on Fretwurst’s Lotus varied mainly between 0.73 and 0.81, for more details see also Appendix A). We operationalize negativity towards the EU as the share of negative evaluations per outlet. We thus rely on a relative measure of Eurosceptic news coverage in the media content of the nine countries we include in our sample.

The survey data

All country surveys were conducted by the company Kantar using Computer Assisted Web Interviewing (CAWI). Light sampling quotas were enforced to ensure representative samples according to age, gender, region and education (checked against information from the National Statistics Bureaus or Governmental sources). The sample composition followed each country’s distribution of key demographics as close as possible (age, region, gender and education), and deviations between sample and census data were not substantial. The data collection followed a panel logic with usually one to two panel waves collected before and two waves collected after the 2019 EP elections in each country (see Goldberg et al. 2021 for more details). The only exception to this rule is the Netherlands, for which seven waves cover the longitudinal development in more detail (we explain this in more detail in the following). The completed responses in the final wave (total N = 5,910) per country are: N_DE = 518, N_DK = 563, N_ES = 552, N_FR = 776, N_GR = 494, N_HU = 586, N_NL = 1067, N_PL = 857, N_SE = 497.

Dependent variable: measuring populist attitudes

Our dependent variable of interest measures populist attitudes. The variable is the mean score of three agreement items measured on a 7-point answer scale ranging from fully disagree (1) over neither agree nor disagree (4) to fully agree (7). The statements focus on the anti-establishment component of the perceived divide between the ordinary people and the corrupt elites (also see Akkerman, Mudde, and Zaslove 2014; Schulz et al. 2018). The three items are worded as follows: The ordinary people instead of politicians should make our most important policy decisions, Politicians in the government are corrupt and Politicians make decisions that harm the interests of the ordinary people (similar items are used in e.g. Akkerman, Mudde, and Zaslove 2014). Cronbach’s alpha for the three combined items in the pooled sample is 0.82, which points to sufficient reliability (see Table A3 for more detailed information). As we aimed for a general measure of the populist divide between ordinary people and corrupt elites, we refrained from using more extensive multidimensional measures of populist attitudes in this study (Schulz et al. 2018). As also noted by Van Hauwaert, Schimpf, and Azevedo (2020), a reduced scale of three items may accurately capture populist attitudes in a comparative set-up, although nuances and extremities of populist beliefs may not be captured.

Independent variables: measuring media exposure

Our independent variable of media exposure is measured in two ways. First, we measure the exposure to the respective media outlets per country (see Table A1 for all outlets per country). For this, we sum the individual frequency of news exposure across all three media types, that is the self-reported exposure to each of the media outlets on a 0–7 days a week scale. Given the partly different number of outlets per country, and to make the resulting media exposure comparable across countries, we standardize the measure per country (mean of zero and standard deviation of one). As a second step, we combine the survey-based media exposure with media content data. For this, we first multiply the individual frequency of news exposure per outlet with the EU negativity share of the respective outlet. Subsequently, we sum up the resulting values per respondent. In a final step, we divide this sum score by the respondent’s overall EU news exposure to arrive at the ratio of negative EU news in the respondent’s respective news diet (all respondents with no media exposure to any of the included outlets, who are assigned a zero in the numerator of the resulting division, were assigned a zero in the final ratio score). As for the simple media exposure measure, we standardize this linked measure of negative EU news exposure to make it comparable across countries.¹ Finally, we also include news exposure on social media, although not linked to respective content. Respondents were asked how often they encounter political news on the social media platforms Facebook, Twitter, Instagram, Snapchat, YouTube and Reddit on 0–7 days/week basis. We created a simple sum score for these six platforms.

Control variables

We include a number of control variables in our analytical models. Next to basic socio-demographics (age in years, gender as a female dummy and education measured in 7 ordinal ES-ISCED categories), we control for left-right self-placement (in linear and squared terms), national party voting in the EP elections recoded into EP party groups (seven official party groups plus categories for other/smaller parties and non-voters) and satisfaction with democracy at the EU level. Furthermore, and as our media exposure measure does not capture all available media outlets per country, we include controls for general news exposure via newspapers, TV news shows and online news separately via more complete country-specific lists of outlets (as simple sum scores measured on a 0–7 days/week logic). To control for potentially remaining country-specific differences, we include country-fixed effects in our comparative models.

Study 1: populist news consumption across different European countries

We test our hypotheses in two subsequent steps. First of all, we aim to offer the ‘bigger picture’ by focusing on the impact of news coverage on populist attitudes in all of the nine countries simultaneously. For this first step, we use a cross-sectional setup and focus on populist attitudes as measured in the final survey panel wave conducted between 1–12 July 2019. To reduce the potential of reverse causality as much as possible, the related media exposure measures stem from the previous wave which was conducted between 27 May – 10 June 2019. The linked media content data covers at least the two months before this panel wave (maximum time period of available content data in countries FR, GR, PL and SE), i.e. was collected between 1 April – 2 June 2019, or covers an even longer time period by going back until 20 November 2018, to increase the validity of the respective content data, as available in the remaining five countries (DE, DK, ES, HU and NL). All of the other control variables are collected before the final panel wave, that is, to again reduce the potential of reverse causality. The statistical models are simple OLS regression models with country-fixed effects.

Results of study 1

The general hypothesis of this study postulated that higher levels of exposure to Eurosceptic views in the media corresponds to higher populist attitudes among participants. Table 1 displays the outcome of the regressions on populist attitudes across the nine pooled countries. Starting with the simple media exposure variable measured in the survey and not linked to media content, we observe a negative relationship between media exposure and populist attitudes. Overall, this means that the more news people consume, the less populist attitudes they hold. This relationship holds in the first model setup with only basic controls (1), but also in the full model (2) containing all other variables including the potentially most relevant control of populist attitudes at t-1.

Table 1. Regression models displaying estimates across nine countries.

	DV: populist attitudes
	(1)	(2)	(3)	(4)
Media exposure (not linked)	−0.066***	−0.058**
	(0.017)	(0.020)
Media exposure (linked w content data)			−0.171***	−0.068***
			(0.017)	(0.013)
Controls
Female	0.035	0.025	0.025	0.022
	(0.034)	(0.024)	(0.033)	(0.024)
Age	−0.002	0.001	−0.001	0.001
	(0.001)	(0.001)	(0.001)	(0.001)
Education	−0.114***	−0.027***	−0.106***	−0.025***
	(0.010)	(0.007)	(0.010)	(0.007)
Country-fixed effects	✓	✓	✓	✓
General media exposure
Newspaper		0.026		0.024
		(0.013)		(0.013)
TV news		−0.025*		−0.024*
		(0.012)		(0.011)
Online news		0.016		0.004
		(0.008)		(0.008)
Social media		0.018		0.019
		(0.015)		(0.015)
Populist attitudes t-1		0.629***		0.624***
		(0.009)		(0.009)
Left-right		−0.027		−0.029
		(0.018)		(0.018)
Left-right^2		0.002		0.002
		(0.002)		(0.002)
EP party voting (ref. non-voter)
EPP		−0.035		−0.035
		(0.041)		(0.041)
S&D		−0.073		−0.070
		(0.042)		(0.042)
Renew		−0.065		−0.062
		(0.046)		(0.046)
Greens-EFA		−0.172**		−0.149**
		(0.057)		(0.057)
ID		0.244***		0.244***
		(0.061)		(0.061)
ECR		0.010		0.024
		(0.053)		(0.053)
GUE-NGL		0.021		0.029
		(0.051)		(0.051)
Other		0.066		0.075
		(0.052)		(0.052)
Satisfaction democracy		−0.074***		−0.077***
		(0.009)		(0.009)
Constant	4.207***	1.827***	4.166***	1.868***
	(0.097)	(0.104)	(0.097)	(0.102)
Observations	5910	5910	5910	5910
R^2	0.206	0.602	0.218	0.604

Standard errors in parentheses,; p < 0.05, ** p < 0.01, *** p < 0.001

To formally test H1, we repeated these two model setups for the linked media exposure variable in models 3 and 4. Contrary to expectations, we observe a negative coefficient in these models, which means that exposure to a larger share of negative EU news is related to weaker populist attitudes. Hence, being exposed to a relatively higher dose of negative views related to the EU in the media corresponds with lower rather than higher levels of populist attitudes among participants. Running the models with the absolute exposure to negative EU news confirms the negative coefficients (see Table A4 in the appendix). Looking at some of the controls as further robustness checks, we observe negative coefficients for education, satisfaction with democracy and green voting (the Greens-EFA group), and positive coefficients for right-populist party voting (the ID group). All these relationships are in line with what one could expect based on extant literature. Furthermore, we did not detect any problems of multicollinearity in the models, for instance, due to the inclusion of various media exposure variables as controls (all VIF factors are clearly below the potential critical threshold of 5).

Given the unexpected negative relationships for our linked media exposure variable, we ran two more robustness checks with more general attitudes towards the EU as the dependent variables, for which the link between EU negativity in news coverage and related EU attitudes might be more direct. Tables A5 and A6 in the appendix include the results of the same model setups with once EU trust and once EU performance as dependent variables (both measured as mean scales of three items each), now expecting negative relationships given the positive measure of EU attitudes. While we again do not find the expected relationships between EU negativity exposure and EU attitudes in terms of significant effects, we do see a significant difference between the simple and the linked media exposure variables, much more so for the simpler model setups 1 and 3. Importantly then, these additional models show that the negative relationship we found between Eurosceptic news and populist attitudes in our main models might be more specific and does not equally hold across related attitudes towards the EU. All in all, we do not find support for H1: Being exposed to Eurosceptic news in the media does not correspond to more pronounced populist attitudes on the individual level. If anything, lower rather than higher populist attitudes are found for participants exposed to critical news on the EU.

Study 2: the impact of populism for the media over time

The second study aims to arrive at a more detailed assessment of the causal relationship between news exposure and populist attitudes. Hence, whereas the advantage of Study 1 is the large scope of the data collection in nine countries, it comes with potential issues for endogeneity. Therefore, Study 2 tests the same hypothesis within a single country setting where we use more robust measures of media exposure and over-time changes in populist attitudes. We specifically focus on one single country known to have been a prototypical case of successful right-wing populism: The Netherlands (e.g. Aalberg et al. 2017). In this second study, we rely on a more refined and comprehensive measure of media exposure, in which we differentiate exposure to newspapers, TV news shows, online news and social media use. This refined measure is possible to use due to the higher overall number of coded news items and a more comprehensive coverage of news outlets in the media content data in the Netherlands (see Table A1 in the appendix). For the other eight countries out of study 1, the shortened time period of both the survey and media data collections, the smaller overall number of news items and less extensive number of outlets to represent the respective national media system prevent the inclusion of these countries for the second study. The more detailed measures for the Netherlands are calculated in the same way as previously discussed, but now distinguishing between various types of media exposure on the source level.

We can profit from the longitudinal logic of the data and rely on repeated measures of both populist attitudes and news exposure. In detail, we have three time points available for populist attitudes (wave 2 (N = 2648): 16–28 January 2018; wave 5 (N = 1637): 5–18 April 2019; wave 7 (N = 1067): 1–10 July 2019) and the respective news exposure measured always one wave before (wave 1: 13 September – 1 October 2017; wave 4: 7–27 December 2018; wave 6: 27 May – 9 June 2019) to correctly model the causal order (not all measures were repeated in all waves to, e.g. run alternative model setups with both variables measured in the same waves).² The media content is linked to these three time periods, by covering the time in between the waves when we measure populist attitudes, our DV (period 1: 13 September 2017–28 January 2018; period 2: 17 January 2018–18 April 2019; period3: 6 April – 2 June 2019).³

Given the repeated measurements of our key variables, we present cross-sectional models at the respective three time periods (for the latter ones additionally controlling for the populist attitudes at t-1), with again the media exposure variables measured in the wave preceding the measure of populist attitudes. We furthermore run models to explain change in populist attitudes by creating simple change variables representing the difference between populist attitudes at time t and t-1. For both the explanation of stable populist attitudes and the change in populist attitudes, we run simple OLS models. Importantly, again, all control variables are measured at least in the wave preceding the measure of our DV. For the control of vote choice, this means we use vote intentions as measured before the elections took place.

Results of study 2

For our more detailed analyses in the Dutch setting, we expected that the relationship between Eurosceptic news exposure and populist attitudes is stronger for participants with higher levels of online and social media consumption compared to participants with higher mainstream media exposure (H2). Hence, we expected that the relationship between Eurosceptic news exposure and populist attitudes would be contingent upon the type of news media people consume. Table 2 shows the respective regression outcomes. The table includes the same (full) model setups across three panel waves, once for the simple exposure variables (models 1-3) and once for the linked variables (models 4-6). Overall, we observe less clear-cut and less stable coefficients than in the comparative setup with the coefficients of TV news and online news exposure going in opposite directions. While TV exposure is related to stronger populist attitudes in most models, the opposite is true for online news exposure, albeit related coefficients are not significant among all three panel waves. Newspaper exposure is negatively related to populist attitudes in the first panel wave under study and positively related with populist attitudes in the final two waves, although the coefficients are not always statistically significant. The most striking and stable relationship is actually the one for social media exposure (albeit not linked to any content data). Respondents with stronger exposure to social media platforms hold stronger populist attitudes. Control variables again show the expected relationships, that is, confirming the correct model specification.

Table 2. Regression models showing detailed estimates for the Dutch setting.

	DV: populist attitudes
	(1)	(2)	(3)	(4)	(5)	(6)
	Wave 2	Wave 5	Wave 7	Wave 2	Wave 5	Wave 7
Newspaper exp	−0.012	0.070	0.099*
	(0.037)	(0.040)	(0.041)
TV news exp	0.059	0.050	−0.068
	(0.033)	(0.035)	(0.040)
Online news exp	0.004	−0.059*	−0.064*
	(0.025)	(0.027)	(0.029)
Newspaper exp				−0.628**	0.413*	0.063
(linked w content)				(0.225)	(0.195)	(0.107)
TV news exp				0.525**	0.167	0.060
(linked w content)				(0.183)	(0.358)	(0.152)
Online news exp				−0.260	−0.892**	−0.405*
(linked w content)				(0.355)	(0.294)	(0.175)
Social media exp	0.096***	0.037	0.120**	0.113***	0.041	0.111**
	(0.029)	(0.035)	(0.039)	(0.028)	(0.033)	(0.036)
Controls
Female	−0.069	0.022	0.078	−0.081	0.018	0.072
	(0.044)	(0.045)	(0.049)	(0.044)	(0.045)	(0.050)
Age	−0.002	−0.004**	0.002	−0.000	−0.004*	0.001
	(0.002)	(0.002)	(0.002)	(0.001)	(0.002)	(0.002)
Education	−0.101***	−0.051***	−0.058***	−0.093***	−0.048***	−0.053***
	(0.012)	(0.013)	(0.014)	(0.012)	(0.013)	(0.014)
Populist attitudes t-1		0.489***	0.607***		0.489***	0.600***
		(0.021)	(0.024)		(0.021)	(0.024)
Left-right	−0.088*	0.053	−0.039	−0.096*	0.046	−0.045
	(0.044)	(0.047)	(0.047)	(0.044)	(0.047)	(0.048)
Left-right^2	0.010*	−0.002	0.004	0.010*	−0.002	0.005
	(0.004)	(0.004)	(0.004)	(0.004)	(0.004)	(0.004)
EP party voting (ref. non-voter)
EPP	−0.053	−0.086	−0.076	−0.024	−0.072	−0.066
	(0.089)	(0.093)	(0.095)	(0.089)	(0.093)	(0.096)
S&D	−0.434***	−0.189	−0.054	−0.391***	−0.175	−0.048
	(0.104)	(0.104)	(0.085)	(0.103)	(0.104)	(0.085)
Renew	−0.508***	−0.233**	−0.183*	−0.483***	−0.227**	−0.197*
	(0.068)	(0.075)	(0.088)	(0.068)	(0.075)	(0.088)
Greens-EFA	−0.485***	−0.222*	−0.100	−0.460***	−0.217*	−0.089
	(0.099)	(0.100)	(0.117)	(0.099)	(0.100)	(0.118)
ID	0.417***	0.161	0.070	0.439***	0.168*	0.105
	(0.082)	(0.085)	(0.152)	(0.081)	(0.084)	(0.153)
ECR	−0.024	0.004	−0.010	−0.002	0.019	0.004
	(0.085)	(0.081)	(0.085)	(0.085)	(0.081)	(0.085)
GUE-NGL	0.085	0.078	0.106	0.101	0.092	0.114
	(0.085)	(0.086)	(0.106)	(0.085)	(0.086)	(0.106)
Other	0.279	−0.991	−0.226	0.350	−1.036*	−0.233
	(0.504)	(0.518)	(0.212)	(0.503)	(0.518)	(0.212)
Satisfaction democracy	−0.383***	−0.189***	−0.138***	−0.382***	−0.188***	−0.138***
	(0.017)	(0.019)	(0.021)	(0.017)	(0.019)	(0.021)
Constant	6.125***	2.624***	2.268***	6.032***	2.651***	2.319***
	(0.163)	(0.214)	(0.225)	(0.166)	(0.216)	(0.228)
Observations	2648	1637	1067	2648	1637	1067
R^2	0.363	0.548	0.616	0.366	0.549	0.614

Standard errors in parentheses; * p < 0.05, ** p < 0.01, *** p < 0.001

For the wave 2 models, satisfaction with democracy is measured at the national level as the EU level measure is not available at this time point.

When looking at change in populist attitudes as the dependent variables in Table 3, we observe almost no significant relationships, especially not in models 3 and 4 including the linked media exposure variables. Independent of their significance, we again observe coefficients with opposite signs depending on the type of media exposure, for newspaper and TV news exposure even significant in model 2. Unlike for stable populist attitudes in Table 2, the social media exposure coefficient does not turn significant in any of the change models either. In general, considering the amount of explained variance of the models (R²), our models explain very little of the changes in populist attitudes over the different panel waves.

Table 3. Regression models with change in populist attitudes as DV (Netherlands).

	DV: change in populist attitudes
	(1)	(2)	(3)	(4)
	Wave 2 → 5	Wave 5 → 7	Wave 2 → 5	Wave 5 → 7
Newspaper exp	0.052	0.104*
	(0.047)	(0.046)
TV news exp	0.006	−0.124**
	(0.041)	(0.045)
Online news exp	−0.015	−0.053
	(0.032)	(0.033)
Newspaper exp			0.362	0.006
(linked w content)			(0.228)	(0.120)
TV news exp			0.518	0.255
(linked w content)			(0.418)	(0.170)
Online news exp			−0.298	−0.199
(linked w content)			(0.343)	(0.196)
Social media exp	−0.068	0.074	−0.067	0.051
	(0.040)	(0.043)	(0.038)	(0.040)
Controls
Female	0.027	0.046	0.027	0.045
	(0.053)	(0.055)	(0.053)	(0.056)
Age	−0.002	0.006**	−0.003	0.003
	(0.002)	(0.002)	(0.002)	(0.002)
Education	0.012	−0.009	0.012	−0.005
	(0.014)	(0.016)	(0.015)	(0.016)
Left-right	0.078	−0.039	0.075	−0.050
	(0.054)	(0.053)	(0.054)	(0.053)
Left-right^2	−0.004	0.002	−0.004	0.003
	(0.005)	(0.005)	(0.005)	(0.005)
EP party voting (ref. non-voter)
EPP	0.063	−0.038	0.058	−0.051
	(0.109)	(0.107)	(0.109)	(0.108)
S&D	0.097	0.044	0.085	0.018
	(0.121)	(0.094)	(0.121)	(0.096)
Renew	0.114	0.043	0.110	0.014
	(0.086)	(0.097)	(0.086)	(0.098)
Greens-EFA	0.034	0.143	0.033	0.151
	(0.117)	(0.130)	(0.117)	(0.131)
ID	−0.032	−0.044	−0.033	−0.016
	(0.098)	(0.170)	(0.098)	(0.171)
ECR	0.048	−0.022	0.049	−0.013
	(0.095)	(0.095)	(0.095)	(0.095)
GUE-NGL	0.136	0.060	0.134	0.062
	(0.101)	(0.118)	(0.101)	(0.119)
Other	−0.628	0.081	−0.676	0.053
	(0.606)	(0.236)	(0.605)	(0.237)
Satisfaction democracy	0.014	0.009	0.014	0.010
	(0.020)	(0.022)	(0.020)	(0.022)
Constant	−0.495*	0.012	−0.513*	0.000
	(0.201)	(0.199)	(0.202)	(0.201)
Observations	1637	1067	1637	1067
R^2	0.013	0.023	0.015	0.014

Standard errors in parentheses; * p < 0.05, ** p < 0.01, *** p < 0.001

We conducted a final robustness check in which we replace the general populist attitudes by specific EU populist attitudes (only available for waves 2 and 5). This means that respondents were asked the same populist items, yet always specifying that it concerns EU politicians (i.e. politicians in the EU make decisions that harm the interests of the ordinary people). Results of this final robustness models are displayed in Table A7 of the appendix. While various of the coefficients for the types of media exposure again display opposite signs, albeit most not being significant, we again observe a positive TV news coefficient, which is significant for the wave 2 models. Statistically significant in models 1–4 are furthermore the coefficients of the social media exposure variable, which again confirms the relevance of social media for populist attitudes. As in our main models for changes between waves 2 and 5, the change models in Table A7 display no significant coefficients for any of the media exposure variables.

Taken together, we only find limited support for H2: social media versus mainstream news exposure does not universally result in a stronger relationship between Eurosceptic news exposure and populist attitudes. However, we do find support for the notion that participants with a stronger social media diet are more likely to hold populist worldviews – which offers some support for the specification of outlet type and populist attitudes forwarded in H2. At the very least, we have to conclude that mere exposure to negative views on the EU across all outlets is not conducive to support for the populist ideology among individual news users.

Discussion

The media arguably contribute to the success of populist parties by paying attention to newsworthy populist movements (Aalberg et al. 2017; Hameleers, Bos, and de Vreese 2017; Mazzoleni 2008). In line with this, media attention has been regarded as an important supply-side explanation of populist success in Europe and beyond (e.g. Bos, van der Brug, and de Vreese 2011). Yet, to date, empirical research has not offered strong support for a direct relationship between populist ideas expressed by the media and support for the populist ideology among voters (e.g. Bos and Brants 2014). Although elements of populism may have become more salient in the news over time (Hameleers and Vliegenthart 2020; Rooduijn 2014), there is only limited support for a high salience of explicit populist discourse expressed by the media (e.g. Bos and Brants 2014; Wettstein et al. 2018). As a more nuanced test of the media logic explanation of populist support, we postulate that increasing attention to ideas and viewpoints owned by populist parties – in this case Eurosceptic views – may contribute to support for the populist ideology by the media’s legitimization of negativity expressed toward elites on the supranational level.

To assess the relationship between news exposure emphasizing populist issue positions and populist attitudes among voters, we rely on an externally valid methodological approach that links content analytic data to media exposure data and populist attitudes measured in a multi-wave survey (also see Müller et al. 2018). Across nine European countries that offer most different settings regarding the success of populist movements on the left and right-wing of the ideological spectrum (Aalberg et al. 2017), we found no support for our central hypotheses. Hence, exposure to Eurosceptic news coverage did not correspond to more pronounced levels of populist attitudes among voters. If anything, exposure to negativity towards the EU in established news outlets resulted in lower rather than higher levels of populist attitudes.

The only key finding in line with our theoretical expectations is that social media use corresponds to more pronounced populist attitudes – which is in line with extant research that points to an affinity between alternative and social media use and populist attitudes (Müller and Schulz 2021). We can explain this as a resonance between populist worldviews and a negative evaluation of the legacy media in populist discourse (e.g. Fawzi 2018). Hence, people with populist attitudes tend to perceive the established media as part of the ‘corrupt’ elite (Fawzi 2018), and may therefore approach alternative outlets and social media channels that more directly represent the voice of the ordinary people whilst circumventing elite sources of information (Hameleers, Bos, and de Vreese 2017).

We can forward different potential explanations for the unexpected findings related to the relationship between exposure to Euroscepticism in the media and populist attitudes. First of all, we looked at a rather indirect relationship between populism expressed in the media and populism among voters. Although we forwarded the argument that exposure to ideas related to a populist worldview may prime or activate congruent mental schemata among voters (also see e.g. Krämer 2014; Müller et al. 2018), the elements of media coverage we looked at were not inherently populist. Therefore, populist associations may not have been triggered sufficiently by media content that does not explicitly point to a causal and moral opposition between the ordinary people and corrupt elites in the EU. This could explain why studies looking at more direct indicators of populist content have established a causal relationship between populist framing and populist attitudes (e.g. Bos et al. 2019; Hameleers, Bos, and de Vreese 2017).

Second, although Euroscepticism is a central feature of dissatisfied populist citizens’ profile across Europe (Kaltwasser and Van Hauwaert 2019; Pirro, Taggart, and van Kessel 2018), negativity toward the EU is a quite common media interpretation that may not always contain populist interpretations. Negativity toward the EU can mean different things, ranging from a well-funded critique on its performance to fact-free hostile sentiments that frame the EU as an enemy of the people. Although the latter may fuel populist sentiments due to its stronger resonance with populist schemata, other forms of critique and negativity may also correspond to non-populist worldviews. Hence, it may be the case that the indicators of negativity captured by our content analyses are reflexive of a more widely shared critique on the European Union rather than a direct representation of populist references to the EU.

Third, populist attitudes may be considered as relatively stable traits that are difficult to influence with media exposure that does not directly include populist references. Populist attitudes may be driven by dissatisfaction, relative deprivation or other socio-political and psychological factors, and mere exposure to low doses of negativity toward elite actors may not have a short-term effect on making these stable attitudes more pronounced. This is not to say that negativity biases toward elite actors in the media do not have an influence: Their impact may be more gradual and based on an over-time activation of cynicism and distrust that is not directly measurable with the set-up of this study.

Our findings suggest that there may be boundary conditions for the relationship between Eurosceptic news exposure and populist attitudes: Negativity in general does not directly fuel support for the populist ideology, which contradicts the theoretical notion that mere attention for populist ideas, actors and policies may contribute to the popularity and success of populists (e.g. Mazzoleni 2008). Yet, we stress that these boundary conditions should be investigated in more detail in future empirical studies that make a distinction between different degrees of negativity and anti-elite sentiments in the media: How ‘populist’ should media content be in order to activate populist attitudes among voters? Although Müller et al. (2018) suggest that references to ‘fragmented’ features of populist ideology may be enough to activate populist attitudes, we leave it up to future research to further explore the nuances of this relationship.

An important implication of our findings is that the relationship between the current media logic – focusing on negativity, anti-elitism and conflict-oriented discourse – and the success of populist parties allegedly profiting from this logic is more complex than expected. On a positive note, this means that the increasing attention to populist issue positions in the established press (e.g. Hameleers and Vliegenthart 2020; Rooduijn 2014) may not universally fuel support for populist worldviews by legitimizing the positions of populist parties. However, although it reaches beyond the scope of our empirical data, it could be argued that the relationship is part of a spiral toward cynicism and dissatisfaction with democratic institutions that cannot be directly assessed by looking at media exposure and populist attitudes. For more distrusting and cynical segments of the audience, the media’s negativity toward elites may reinforce their dissatisfaction with (supra)national institutions, which may in turn also result in the avoidance of established media (e.g. Hameleers, Bos, and de Vreese 2017).

Our findings indicate that social media use is – irrespective of the dose of Eurosceptic views people are exposed to – associated with increases in populist attitudes. In an information era when relativism towards facts is mounting (Van Aelst et al. 2017), and populist parties increasingly blame the established media in their anti-elitist discourse (Egelhofer and Lecheler 2019), the question is whether the media logic argument as supply-side explanation of populist success still holds: People with a tendency to support populist ideas may not be informed by the established media’s interpretation of issues in the first place, as these outlets are distrusted altogether. Likewise, in this age of factual relativism and digitalization, populists may be less dependent on the established media to offer them with the oxygen of publicity when they can use more direct social media channels to speak to the ordinary people (e.g. Engesser et al. 2017). This may also explain why the mainstream media did not have the agenda-setting and priming effects that we hypothesized.

Our findings bear limitations that can be addressed in future research. First, our content analysis only focused on one element that may be associated with populist issue positions – negativity towards the EU – whereas there are more elements that resonate with populist positions on the left- and right-wing of the political spectrum. Future research may, among other things, look at anti-immigration sentiments, welfare chauvinism and economic inclusionism. Related, our content analysis did not explicitly look at more direct indicators of populism’s fragmented nature (Müller et al. 2018). We also included a rather broad measure of general negativity toward the EU, whereas Euroscepticism can include more explicit anti-elitist sentiments and more severe forms of critique that oppose EU integration altogether (Leconte 2010). Arguably, more severe forms of critique that are closer to populist framing have a stronger impact on populist attitudes than ‘softer’ forms of Euroscepticism. Future research may need to rely on a more extensive set of indicators differentiating between populist ideas, styles and issue positions, and further look at a more fine-grained differentiation between different forms of Euroscepticism in the media. Similar limitations can be identified for our survey data: We looked at populism as a one-dimensional concept, which may overlook some of the more nuanced relationships between elements of populist ideology and news coverage. In addition, we rely on a rather general measure of self-reported news exposure, which next to the reliability issues haunting self-reported media use does not exclude that people consume the mentioned outlet, but always skip the political section which corresponds to our content data. The availability of more reliable and detailed tracking data of news consumption, ideally on the article level, could result in a more comprehensive picture of which specific news media diets may correspond to the activation of populist worldviews. It should also be noted that we did not directly look at priming in terms of affecting support for populist parties through making salient mental associations related to this party. Rather, we assessed how the media’s emphasis on an issue owned by populist parties primes related associations captured in populist attitudes. Future research may delve deeper into this relationship by considering how the media’s attention to and framing of populist issues may prime populist vote intentions.

Despite these limitations, we believe that our study offers an externally valid assessment of the relationship between the media’s attention to issue positions conducive to populist parties and populist attitudes among voters. We show that mere references to issue positions resonating with populist ideas and policies does not directly correspond to strengthened populist attitudes, which nuances some assumptions about the unprecedented impact of (mainstream) media coverage on populist success.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 As robustness check, we also repeat our analyses with the linked measure in an absolute sense, that is, without conducting the third step in the operationalization to calculate a ratio measure by dividing the sum score by the individual news diet.

2 The retention rates across all seven waves in the Dutch panel data varied between 72 and 88%. The sample composition concerning sex, age and education shows highly stable values.

3 The actual field periods of the survey panel wave to measure our DV is included as overlapping days in the three time periods for the media content data. Depending on when the respondent was interviewed, this person could have seen the respective content before the interview took place for this wave or only before the interview of the following wave.