https://orcid.org/0000-0001-6009-2785
Abstract
Amongst growing public concern about climate change, right-wing populist actors often dismiss environmental concerns as part of an elitist, transnational agenda. At the same time, so-called alternative news media play an increasingly prominent role in spreading populist ideology. How do alternative news media cover environment-related topics, and what effects does such coverage have on news consumers? Guided by the notion of populism as a communication phenomenon, we argue that alternative media produce and disseminate content critical of environmental protections. Empirically, we combine a 3-wave online panel and web-tracking data collected during the 2021 German federal election campaign and proceed as follows. Firstly, we detect environment-related news in a corpus of articles from outlets previously classified as alternative news media. Secondly, we examine latent topics and key actors present in these articles in order to find out how alternative news media cover environmental news. Thirdly, we use a longitudinal model to examine whether consuming alternative news impacts individuals’ climate attitudes. We find that alternative outlets tend to politicize environmental news, though short-term attitudinal effects are mixed and limited.
Introduction
Alternative news media, once commended for diversifying established media landscapes and for creating platforms for democratic protest movements, have more recently raised concerns regarding associations with problematic protest forms and negative democratic changes (Fenton & Barassi, 2011; Guess, Nyhan, & Reifler, 2020; Müller & Schulz, 2021). The subgroup of right-wing populist alternative news media is believed to contribute to the electoral success of right-wing populist parties due to their distinct framing and coverage of political events. At the same time as these media grow in salience, voters in many Western countries are increasingly concerned about environmental issues. While climate change has been identified by the United Nation as the greatest threat the world has ever faced (United Nations General Assembly, 2022), right-wing populist political actors and voters in many countries are skeptical of climate protection measures (see, e.g., Huber, Fesenfeld, & Bernauer, 2020, 2021; Küppers, 2022; Schwörer & Fernández-García, 2023). Although some studies suggest that right-wing populist alternative news media also take a climate-skeptical stance (Peeters & Maeseele, 2023), empirical evidence on such coverage and potential attitudinal effects are still rare.
Against this background, our study is concerned with gaining a deeper understanding of how populist alternative media incorporate environmental topics into their antielite discourse and to what extent this impacts the environmental attitudes of their audience. We tackle this objective by using the combination of a three-wave online panel and web-tracking data collected during the 2021 German federal election campaign. This dataset provides us with measures of individuals’ climate and populist attitudes and news consumption behavior, allowing for robust and fine-grained analyses. We proceed as follows. Firstly, we detect environment-related news in a corpus of articles from outlets previously classified as alternative news media. Secondly, we examine latent topics and key actors present in these articles in order to find out how alternative news media cover environmental news. Thirdly, we use a longitudinal model to examine whether consumption of such news impacts individuals’ environment-related attitudes.
We focus our attention on Germany, which proves to be an interesting case for several reasons. First, right-wing populist alternative media have been shown to be on the rise in the so far little polarized German news media system (Müller & Schulz, 2021; Zimmermann & Kohring, 2020). These outlets form a close-knit network that amplifies and reinforces their messages (Hope Not Hate, 2021), ranging from antielite positions to conspiracy theories. Second, debates on energy transition and emissions are particularly salient given Germany’s strong automobile industry and its economic weight (Yan, Schroeder, & Stier, 2022). Third, the 2021 election was termed the “climate election” by some scholars (Korte, 2021), with environment-related topics topping the list of most pressing issues amongst voters in the run-up to polling day, surpassing Covid-19 (Demler, 2022). Both a climate-skeptical populist right-wing party (the AfD, Alternative for Germany) and a green party (Bündnis 90/Die Grünen) played a prominent role in the 2021 elections, with media reports about the election campaign featuring more climate-related content than ever before (Leidecker-Sandmann, Schafer-Hock, & Wilke, 2023). Finally, an extreme weather event occurred in Germany in July 2021, resulting in massive flooding with roughly 180 casualties. Whilst many political actors linked this event to human-made climate change, the AfD disputed this connection, further politicizing environmental issues (Thurau, 2021).
The remainder of this paper is organized as follows. We review the nexus between right-wing populism and climate change before we turn towards populist right-wing alternative news media, its contents, consumers, and effects. We then describe our data and methodology and present exploratory findings on mainstream media’s and right-wing populist alternative news media’s coverage of climate change and environmental topics. Based on differences detected between the two types of media, we then show how exposure to relevant content from alternative media affects individuals’ attitudes towards protecting the environment.
Background
Before we turn to the association between climate change skepticism and right-wing populism, we briefly introduce the conceptualization of populism that we will employ in this paper. We adopt de Vreese, Esser, Aalberg, Reinemann, and Stanyer’s (2018) understanding of populism as a communication phenomenon, which combines the ideas of populism as a “thin-centred ideology,” defined by people-centrism and antielitism (Mudde, 2004), with a particular discursive style used to communicate populist ideas (Hawkins, 2009). Crucially, the “pure people” and “corrupt elites” are so-called empty signifiers, which can be filled with different archetypes in different contexts (Waisbord, 2018). Concerning populism in news media, scholars differentiate between populism by and through the media. In the former, media actors proactively reproduce populist ideas by, for instance, criticizing political elites and promoting a Manichean worldview (populism by the media), whilst in the latter, they act as vehicles for populist actors to express their own ideas (populism through the media) (de Vreese et al., 2018).
We focus exclusively on right-wing populism and its nativist and socially authoritarian views. The increased salience of environmental topics has brought climate change to the populist agenda, with several studies—from a wide array of national contexts—indicating that right-wing populist politicians, outlets, and citizens tend to share a critical stance towards environmental concerns (Fiorino, 2022; Kulin, Johansson Seva, & Dunlap, 2021; Lockwood, 2018; Marquardt, & Lederer, 2022). In the following paragraphs, we explore the reasons and empirical evidence behind this association from the perspectives of, respectively, political elites, citizens, and so-called alternative media.
Explaining the Link between Climate Change Skepticism and Right-Wing Populism
Scholars typically view the rising electoral success of right-wing populists as a reaction to developments experienced by globalized, postindustrial societies. Amongst these, examples include the proliferation of transnational governance, economic modernization, demographic shifts related to migration, and mass attitudinal change on sociocultural topics (e.g., Lockwood, 2018; Norris & Inglehart, 2019). Against this background, right-wing populists tend to stake out nativist and conservative positions on transnational issues such as trade, immigration, and, crucial to our paper, environmental topics. From an economic perspective, right-wing populists often oppose constraints resulting from environmental protections championed by technocratic and international institutions. They also frame these economic constraints as resistance to a “cosmopolitan elite agenda” that supports climate protection and undermines national sovereignty through opaque, supranational decision-making bodies (Lockwood, 2018). By championing national solutions to transnational issues, right-wing populists equate “anti-elitism” (a central tenet of populist ideology) with “anti-globalism” and “anti-cosmopolitanism” (Huber et al., 2020; Lockwood, 2018; Waisbord, 2018), combining both economic and sociocultural arguments.
The actions of several populist leaders and parties serve to illustrate the points made above. Donald Trump’s government weakened environmental protections and cut funding for international organizations dedicated to climate change mitigation, often claiming that such policies were unfavorable to regular American “workers” and America itself in the face of conspiring global competitors (Fiorino, 2022). Brazil’s Jair Bolsonaro worked to expand agricultural interests in the Amazon rain forest, appeasing one of his core constituencies of support (agrobusiness) whilst also portraying the Amazon as integral to national identity and sovereignty, threatened by foreign interests (Marquardt, Oliveira, & Lederer, 2022). In Poland, the right-wing populist Law and Justice (PiS) government praised the country’s coal deposit for guaranteeing energy sovereignty for 200 years and labeled environmentalism as part of a Marxist culture (Lockwood, 2018). In the Philippines, Rodrigo Duterte portrayed the climate protection policy as an elitist, Global North project that was detrimental to the Philippines’ economic interests (Marquardt et al., 2022). Huber, Maltby, Szulecki, and Cetković (2021) analyze the rhetoric of populist parties in several European countries, concluding that right-wing populists in particular are more critical of the European Union’s Energy and Climate Policy, often employing arguments related to national sovereignty and the protection of key industries.
In the case of Germany, the populist right-wing party AfD illustrates a third link between right-wing populism and climate skepticism: the role of science and the contestation of established knowledge. As Marquardt et al. (2022) argue, right-wing populists challenge the scientific consensus put forward by global technocratic bodies, consolidating their self-image as being silenced by mainstream society. In that vein, Boecher, Zeigermann, Berker, and Jabra (2022) find that AfD politicians employ “common wisdom” rhetoric in disputing mainstream scientific findings, proposing their own alternative climate expertise. In line with previous examples, Küppers (2022) finds that the party often portrays the fight against climate change as a threat to Germany’s national sovereignty and economic interests.
There is less evidence of the nexus between populism and environment-related attitudes at the citizen level, though several plausible mechanisms may apply. First, citizens take cues from political elites and may therefore follow partisan allegiances on the matter, as found by (Huber, Greussing, & Eberl, 2020) in the United States. Second, populist attitudes correlate with lower levels of trust in science and institutions (Huber et al., 2022), which aligns with the “alternative climate expertise” promoted by right-wing populist actors. Third, Kulin et al. (2021) analyze data from 23 European countries and find that nationalist attitudes are a substantive predictor of populist right attitudes and voting for right-wing populist parties, which is in line with the “national sovereignty” frame employed by populist politicians. Fourth, and from an economic perspective, individuals who are affected by climate mitigation policies may constitute an important constituency for right-wing populist parties (Lockwood, 2018).
Finally, we note that the link between right-wing populism and the environment is not automatic. The examples above illustrate cases where political elites leveraged the politicization of environmental concerns as a way to further push their antielitist/globalist agenda. In contexts where environmental issues are not salient or cannot easily be instrumentalized for political gains, it is likely that right-wing populists may have a different focus. Indeed, what “fills” the empty signifier of global corrupt elites may change according to context (De Bruycker & Rooduijn, 2021), from immigration to general corruption (Cassell, 2023). “Given the salience of environmental concerns in the latest 2021 federal election (see Introduction section), we argue, Germany emerges as a ‘most-likely case’ for the connection between right-wing populism and environment-related topics.”
Having taken a brief excursus into the link between right-wing populism and climate change skepticism, we next turn to right-wing populist alternative news media, its content, consumers, and effects.
Right-Wing Populist Alternative News Media
Though there is no universally agreed-upon definition of right-wing populist alternative news media (Downing, 2003; Holtetal, 2019), defining elements encompass an antimainstream and antielitist stance as well as reactionary political positions (Müller & Schulz, 2021). Against missing a unified definition, we follow Holt, Ustad Figenschou, and Frischlich (2019) in using alternative media as a non-normative umbrella term that covers media outlets that position themselves as correctives and opponents to the dominant mainstream media from a populist right-wing stance. In keeping with our theoretical framework, we argue populist ideas are reproduced both by and through alternative news media. Alternative outlets provide a platform for coverage of right-wing populist actors and topics (populism through the media), whilst also embodying the populist tenets of antielitism.
Both these dynamics can be observed in the German context. Alternative outlets maintain a symbiotic relationship with the AfD, providing increased party coverage whilst AfD politicians help to disseminate alternative news via social media (Hope Not Hate, 2021). Whilst these outlets share core tenets of right-wing populism, they vary in their approaches and extremity (Müller & Freudenthaler, 2022). Popular outlets such as Junge Freiheit and Tichys Einblick stand closer to the mainstream, having been founded by actors with prior links to legacy conservative media. Towards the more extreme end of the spectrum, the blog-style Reitschuster gained popularity by spreading conspiracy theories related to the COVID-19 pandemic, whilst PI-News reproduces blatantly racist and xenophobic content (Hope Not Hate, 2021). Further, outlets such as the Epoch Times and RT deutsch have clear links with foreign countries and actors, being part of a “transnational informational ecology” of right-wing websites across Europe and the United States (Heft, Knüpfer, Reinhardt, & Mayerhöffer, 2021).
Several empirical works exemplify these patterns. Mirroring findings from Belgium and the United States on the coverage of extreme right-wing parties (Buyens & van Aelst, 2021; Kaiser, Rauchfleisch, & Bourassa, 2020), Müller and Schulz (2021) argue exposure to alternative outlets constitutes a type of partisan selective exposure, being strongly linked to the electoral rise of the AfD in Germany. Beyond covering populist actors, alternative outlets also pay significant attention to topics such as immigration (Müller & Freudenthaler, 2022) and, recently, the COVID-19 pandemic, common populist far-right talking points. At the same time, alternative media often play the self-perceived role of correctors of mainstream media content (Holt et al., 2019), often espousing ideologically extreme positions, conspiracy narratives, occasional disinformation, and an antagonistic approach to politics (Frischlich, Kuhfeldt, Schatto-Eckrodt, & Clever, 2022; Klumpp, 2023). Put together, populist alternative media are typically less diverse than mainstream media in terms of the topics and political parties they cover whilst also tending to politicize salient topics from a right-wing populist perspective (Freudenthaler & Wessler, 2022; Müller & Bach, 2021).
However, there is only limited empirical evidence on how alternative news media cover environmental topics, with existing studies suggesting a more critical stance. For example, Peeters and Maeseele (2023) show that climate change is often framed as a tool used by the leftist elite to exploit lower-class people. In contrast, Müller and Freudenthaler (2022) find that environmental topics make only a small portion of news content produced by alternative outlets in Germany. Given the limited scope of available evidence, we posit the following research question as a starting point: how do alternative news media differ from mainstream media in their coverage of environmental news?
The “alternative” content described above is likely to find a welcoming audience, as several studies on the profile of alternative news consumers suggest. Overall, users of such media tend to be more interested in politics, score higher on measures of authoritarianism and conspiracy mentality, and show low levels of trust in the political system in a country and in its mainstream media (Frischlich, Hellmann, Brinkschulte, Becker, & Back, 2021, 2022; Hameleers, Brosius, & de Vreese, 2022; Schulze, 2020; Wagnsson, 2022). Unsurprisingly, individuals who consume alternative news media tend to vote for right-wing populist parties and to hold populist attitudes (Guess et al., 2020; Müller & Bach, 2021; Müller & Schulz, 2021; Stier, Kirkizh, Froio, & Schroeder, 2020; Wagnsson, 2022; Yan et al., 2022). Again, however, we know little about the environmental attitudes of alternative media consumers and even less about how alternative media may shape attitudes. In an analysis of six Western countries, Yan et al. (2022) find that populist attitudes correlate both with climate-skeptical attitudes and consumption of climate change-related news, though the study does not focus on alternative outlets nor its effects.
In spite of limited evidence, we argue that it is plausible that differences in content between right-wing populist alternative media and mainstream media may translate into differences in media effects. That is, media effects documented for mainstream media (often limited in size) may not apply to alternative media (Dimitrova, Shehata, Strombäck, & Nord, 2014; Frischlich et al., 2021). Instead, given the populist rhetoric of right-wing populist alternative media, their antielitist, antimainstream, and often slanted reporting, we may expect that media effects be stronger than those documented for mainstream mass media. Transportation theory (Green & Brock, 2000) and the extended elaboration likelihood model (Slater & Rouner, 2002) back this assumption as certain types of narratives of media messages are assumed to lead to increased media effects (Valkenburg, Peter, & Walther, 2016). Not only are alternative news outlets ideologically cohesive, constituting a type of selective exposure (Müller & Schulz, 2021), but they are also characterized by their highly opinionated editorial agenda, which may be particularly effective amongst its core populist right-wing audience (Frischlich et al., 2022).
Against the background of a strong and biased framing of environment and climate change topics among right-wing populists and in like-minded media, paired with the popularity of environmental and climate change topics on both the right-wing populists’ agenda and in the German 2021 election campaign (see Introduction section), we hypothesize that exposure to such content will lead to a more skeptical stance towards environmental issue amongst news consumers. Given that alternative outlets also cover general news, we expect this attitudinal effect to be primarily driven by the opinionated editorial agendas of alternative media. To test this hypothesis empirically, we use combined web-tracking and panel survey data, which we describe in more detail in the next section.
Research Design
Data
We rely on the combination of a four-wave survey PINCET (Bach, Keusch, Areal, Pankowska, & Cernat, 2023) fielded before and after the 2021 German election (26 September) and its associated web-tracking data. In the period between July and December 2021, survey participants were required to run a plug-in recording of their browsing activity on both their personal computers and mobile devices. This plug-in recorded each time a respondent visited a particular URL and the duration of the visit. Note that this concerns only browser-based website visits. Activities on specific apps are not recorded. We restrict our analysis to the first three waves of PINCET (Bach et al., 2023), which contain attitudinal measures required for our study. Waves one and two were fielded in the weeks before the September 26 federal elections (wave one: August 30 to September 7; wave two: September 14–20), and wave three immediately after it (September 27 to October 4). Quotas (age, gender, and state of residence) were applied in order to match the population of adults living in Germany and eligible to vote in the 2021 election. Though respondents were always invited to subsequent waves, participation patterns vary significantly (Supplementary Table A1) across waves. After applying data filtering steps, i.e., no missing values on covariates and participation in at least two waves of the model for longitudinal estimates, we arrived at our final sample of 2,111 individuals and 5,704 observations (respondent-wave pairs). In the Appendix, we provide more details on the sampling procedure and sample characteristics, as well as balance checks comparing our analytic sample (individuals tracked for at least two waves) with the full PINCET (Bach et al., 2023) sample (Supplementary Figures A1, A2, and A3).
Operationalization of Key Variables
In order to analyze how mainstream and alternative news media cover environmental topics, and how such coverage may impact individuals’ attitudes, we need three types of information. First, we need data on the content of news articles published in mainstream and alternative news to then classify it as being environment-related. Second, we require information that captures which individuals were exposed to this content and when. Finally, we need several measures of these individuals’ political and environmental attitudes, as well as sociodemographic characteristics. All of this information is provided by the combination of the survey panel in combination with its associated web-tracking data. In the following, we provide detailed information for each data type and describe how we obtained them.
News Media Coverage of Environmental News
We obtain the content of news media coverage by scraping the content of URLs accessed by survey participants, focusing on URLs from German mainstream and alternative news outlets identified using lists from Scharkow, Mangold, Stier, and Breuer (2020) and Müller and Bach (2021). See Supplementary Tables A6 and A7 for full lists. For feasibility reasons, we scraped only mainstream news outlets that registered at least 1,000 visits by survey respondents. After excluding duplicates and articles whose full content could not be scraped (minimum of 50 words), the dataset for classification contained 98,695 unique articles, with 95,915 from mainstream news websites and 2,780 from alternative news outlets.
We then proceeded with two separate classification tasks: detecting articles about the environment and opinion pieces. This provides us with a direct measure of environmental news coverage, as well as allowing us to test the potentially stronger effects of opinionated news (Frischlich et al., 2022). For both tasks, we train our classifier, a pretrained BERT model (Devlin, Chang, Lee, & Toutanova, 2018) for the German language, bert-base-german-cased, on a fine-tuning dataset of labeled news articles. We obtained the fine-tuning data by scraping news articles from mainstream German newspapers that were clearly positioned in sections related to the environment and opinion columns, which act as labels for our classification. We took several validation steps to account for potential mislabeling of articles at the source (e.g., nonenvironment-related articles in “environment” sections in outlets used to build fine-tuning data) and discrepancies in classification between our model and the fine-tuning dataset (see Supplementary Appendix for more details). After two five-fold cross-validation procedures, we retrieved an F1 score of 0.97 and 0.98 for, respectively, environment and opinion pieces (Supplementary Tables A4 and A5). We restrict the data to articles consumed before the end of wave three and by respondents included in the survey models. Out of a total of 46,157 articles considered, 3,361 were classified as being environment-related (165 from alternative news outlets), of which 456 were also classified as being opinion articles (58 from alternative news outlets).
Individual-Level Measures
Dependent variables
We consider three environment-related attitudes to assess the potential effects of alternative news consumption. First, we measure attitudes towards the fight against climate change via a 5-point item that asks whether economic growth or climate action should take precedence, even if one harms the other. Second, respondents are asked how important it is to them that the environment be protected via a 4-item scale ranging from “not at all” to “very important.” Finally, we measure attitudes towards the Green Party via an 11-point like/dislike scale. All items are recorded such that higher values indicate a more pro-climate stance and z-standardized across the whole sample. The original wording and English translation of all survey items can be found in the Appendix (Supplementary Table A3).
Independent variables
We measure news consumption both in terms of exposure (dummy) and extent of consumption (relative count). For both types of outlets (mainstream and alternative), as well as news type (environment-related and opinion pieces) we create a dummy variable indicating if the individual registered at least one visit to a relevant news article in the period before each survey wave. We complement this measure with a relative count of visits by first calculating the number of visits to each outlet and article type and then dividing this raw count by the number of active days, that is, days on which the individual registered at least one log-on their browser (not necessarily news-related). This approach is inspired by Cronin, Clemm von Hohenberg, Goncalves, Menchen-Trevino, and Wojcieszak (2023), and it attempts to account for significant disparities in the number of tracked days per individual and per wave (see Supplementary Table A2 in the Appendix). As a robustness check, we also run alternative models with a logged version of these variables, which return similar results (Supplementary Figures A10 and A11). Further, by measuring news consumption both in terms of exposure and extent, we account for the fact that these are two distinct processes (i.e., consuming any versus consuming many), which speaks to the literature on the correlates of alternative news consumption (Müller & Bach, 2021).
Control variables
We control for populist attitudes as the average of three items developed by Schulz et al. (2018), targeting the dimensions of antielitism, popular sovereignty, and a view of “the people” as homogeneous. Given that populist attitudes are related to exposure to alternative news and a more critical stance on the environment, this acts as an important control in isolating media effects. We also control for left-right self-placement, political interest, party identification, and sociodemographics (age, sex, and education). With the exception of the last two, controls are time-varying to capture election dynamics.
Methods
We attempt to gain insights into how alternative and mainstream news outlets differ in how they report on environmental topics through two different methods. Firstly, we apply the BERTopic topic model (Grootendorst, 2022), an unsupervised technique that assigns individual documents (here, news articles) to broader categories based on text similarity, calculated via contextual word embeddings News articles that employ semantically similar content will thus be grouped together under the same “topic.” Secondly, we apply named entity recognition (Schweter & Akbik, 2020) to identify the relevant entities, such as persons and locations, commonly mentioned in environment-related news articles. Crucially, this method does not rely on a predefined list of entities and leverages contextual cues to aid in identification. In applying both techniques, we focus our analysis on the comparison between types of outlets. In other words, we study whether certain news topics or entities are more prevalent in alternative news outlets if compared to mainstream outlets.
To assess the potential effects of consuming environmental news on individuals’ attitudes, we estimate a series of random effects within-between models (Bell & Jones, 2015), which provide us with estimates for over-time effects (within individuals) and baseline differences between individuals. Within effects leverage only variation that took place between waves, thus reflecting the short-term effects of increased (or decreased) news consumption on environmental attitudes. Between effects, in turn, capture whether average differences in environmental attitudes are associated with average levels of news consumption, regardless of over-time variation. Substantively, within effects are better suited to studying the effects of news consumption, with individuals acting as their own control group, whereas between effects may reflect self-selection (i.e., different individuals consume different things in the first place).
Results
Basic descriptive results on news consumption across mainstream and alternative news outlets are shown in Table 1, which documents the number of unique URLs visited by respondents and the sum of visits, differentiating between classified URLs (i.e., fully scraped news articles used in the classification model), URLs classified as environment-related and, amongst those, opinion and descriptive pieces. This summarizes the data filtering steps, with counts referring only to URLs consumed in the period between waves one and three and by individuals who are part of the subsequent survey models. We also provide the number of individuals (users) who accessed each type of URL. Crucially, we note that data on news consumption is not necessarily representative of what news websites produced during this period but rather what was consumed by the respondents in our sample. In other words, the following findings say less about the outlets themselves and more about their audiences’ consumption preferences.
Descriptive Results on News Consumption by Outlet and Article Type
| Mainstream news | Alternative news | |||||
|---|---|---|---|---|---|---|
| Unique count | Visits | Users | Unique count | Visits | Users | |
| All URLs | 176,270 | 835,180 | 1,797 | 2,115 | 9,430 | 154 |
| Scraped URLs | 50,409 | 257,485 | 1,542 | 1,561 | 4,230 | 121 |
| Classified URLs | 44,618 | 167,628 | 1,424 | 1,539 | 4,165 | 121 |
| Environment-related | 3,195 | 12,241 | 726 | 166 | 455 | 36 |
| Opinion article | 398 | 1,993 | 298 | 58 | 246 | 26 |
| Descriptive article | 2,797 | 10,248 | 690 | 108 | 209 | 26 |
| Mainstream news | Alternative news | |||||
|---|---|---|---|---|---|---|
| Unique count | Visits | Users | Unique count | Visits | Users | |
| All URLs | 176,270 | 835,180 | 1,797 | 2,115 | 9,430 | 154 |
| Scraped URLs | 50,409 | 257,485 | 1,542 | 1,561 | 4,230 | 121 |
| Classified URLs | 44,618 | 167,628 | 1,424 | 1,539 | 4,165 | 121 |
| Environment-related | 3,195 | 12,241 | 726 | 166 | 455 | 36 |
| Opinion article | 398 | 1,993 | 298 | 58 | 246 | 26 |
| Descriptive article | 2,797 | 10,248 | 690 | 108 | 209 | 26 |
Note. Data are restricted to the period between waves one and three and to respondents who provided tracking data in that period. Classified URLs are news articles that went through the preprocessing steps detailed in the Appendix.
Descriptive Results on News Consumption by Outlet and Article Type
| Mainstream news | Alternative news | |||||
|---|---|---|---|---|---|---|
| Unique count | Visits | Users | Unique count | Visits | Users | |
| All URLs | 176,270 | 835,180 | 1,797 | 2,115 | 9,430 | 154 |
| Scraped URLs | 50,409 | 257,485 | 1,542 | 1,561 | 4,230 | 121 |
| Classified URLs | 44,618 | 167,628 | 1,424 | 1,539 | 4,165 | 121 |
| Environment-related | 3,195 | 12,241 | 726 | 166 | 455 | 36 |
| Opinion article | 398 | 1,993 | 298 | 58 | 246 | 26 |
| Descriptive article | 2,797 | 10,248 | 690 | 108 | 209 | 26 |
| Mainstream news | Alternative news | |||||
|---|---|---|---|---|---|---|
| Unique count | Visits | Users | Unique count | Visits | Users | |
| All URLs | 176,270 | 835,180 | 1,797 | 2,115 | 9,430 | 154 |
| Scraped URLs | 50,409 | 257,485 | 1,542 | 1,561 | 4,230 | 121 |
| Classified URLs | 44,618 | 167,628 | 1,424 | 1,539 | 4,165 | 121 |
| Environment-related | 3,195 | 12,241 | 726 | 166 | 455 | 36 |
| Opinion article | 398 | 1,993 | 298 | 58 | 246 | 26 |
| Descriptive article | 2,797 | 10,248 | 690 | 108 | 209 | 26 |
Note. Data are restricted to the period between waves one and three and to respondents who provided tracking data in that period. Classified URLs are news articles that went through the preprocessing steps detailed in the Appendix.
Seven percent of classified news articles were related to the environment in mainstream news outlets, whereas the figure is 11% for alternative news media. Though relatively high, consumption of environmental news on alternative outlets is driven by very few, very active individuals (n = 36), who are themselves only a fraction (23%) of all alternative news consumers. For contrast, 40% of mainstream news consumers read at least one article about the environment, even though only 7% of all visits were visits to environmental news. Thus, consumption of environmental news on alternative websites is rare but intense. We also note that opinion pieces attract more visits per article than descriptive news across both types of outlets. Finally, in the Appendix (Supplementary Figures A4 and A5), we display news consumption across time and types of outlets. Regarding environmental news, both trends and relative consumption levels are similar in mainstream and alternative news outlets. Lending face validity to our classification, environmental news consumption peaks in the weeks after the flood disaster in mid-July.
Text Analysis
Moving on to an analysis of the content of the environment-related news articles, we compare mainstream and alternative news coverage through a topic analysis using BERTopic (Grootendorst, 2022), which clusters text documents into broader categories in an unsupervised fashion, relying on text similarity. Topics are themselves characterized by keywords that best define the documents clustered together by the algorithm. We do not instruct the model to find any particular topics, nor do we prespecify the number of topics to be found in the corpus of environment-related news. In the Appendix, we detail the steps we took to preprocess the data and reduce the number of outliers and topics for a more cohesive overview, as well as validation steps (Supplementary Table A10; Supplementary Figure A6). The final model resulted in 40 topics (Supplementary Table A8 in the Appendix). For ease of interpretation, we further condensed 40 topics into manual categories created by aggregating topics with keywords and representative articles (as defined by BERTopic) that we judged to be semantically similar. The list of topics that form each manual category is also shown in the Appendix (Supplementary Table A9).
Table 2 displays the number of articles and the number of visits to articles belonging to the manually created categories, split by outlet type. Focusing on the relevant percentage for both visits and the number of articles in each category, we find that mainstream and alternative news websites are both primarily characterized by political news and articles related to natural disasters. Political articles typically concern party-related news, whereas the latter category focuses on the flood disasters of July 2021. Though the foci may be similar, however, this does not mean these outlets are also similar in stance, something we do not systematically analyze. For illustration purposes, we explore the most visited alternative news articles within the general “Politics” category. We find that they often take a critical stance, portraying environmental concerns as elitist projects with harmful economic consequences, with criticism leveled at political elites and the Green Party, whilst pitting concerns related to emissions and energy transition against environmental protection, understood as nature conservation (e.g., Haubold, 2021; Reitschuster, 2021). It should also be noted that the prevalence of political news is much greater in alternative outlets, with mainstream outlets being far more diverse in the topics covered.
Distribution of Unique Articles and Visits by Manual Categories Created from Topic Modeling
| Mainstream news | Alternative news | |||
|---|---|---|---|---|
| Manual category | N articles (%) | N visits (%) | N articles (%) | N visits (%) |
| Politics | 1,532 (50.07%) | 3,961 (54.65%) | 120 (72.29%) | 319 (75.41%) |
| Natural Disasters | 731 (23.89%) | 1750 (24.14%) | 22 (13.25%) | 61 (14.42%) |
| Mobility | 141 (4.61%) | 294 (4.06%) | 8 (4.82%) | 13 (3.07%) |
| Nature/Animals | 265 (8.66%) | 471 (6.50%) | 3 (1.81%) | 11 (2.60%) |
| Climate Activists | 83 (2.71%) | 168 (2.32%) | 5 (3.01%) | 6 (1.42%) |
| Other | 35 (1.14%) | 68 (0.94%) | 1 (0.60%) | 4 (0.95%) |
| Sustainability | 130 (4.25%) | 321 (4.43%) | 2 (1.20%) | 3 (0.71%) |
| Energy | 18 (0.59%) | 27 (0.37%) | 1 (0.60%) | 2 (0.47%) |
| Food | 23 (0.75%) | 34 (0.47%) | 2 (1.20%) | 2 (0.47%) |
| Env. Protection | 13 (0.42%) | 31 (0.43%) | 1 (0.60%) | 1 (0.24%) |
| Housing | 34 (1.11%) | 45 (0.62%) | 1 (0.60%) | 1 (0.24%) |
| Corona | 35 (1.14%) | 41 (0.57%) | 0 (0.00%) | 0 (0.00%) |
| Economy | 9 (0.29%) | 20 (0.28%) | 0 (0.00%) | 0 (0.00%) |
| General News | 11 (0.36%) | 17 (0.23%) | 0 (0.00%) | 0 (0.00%) |
| Mainstream news | Alternative news | |||
|---|---|---|---|---|
| Manual category | N articles (%) | N visits (%) | N articles (%) | N visits (%) |
| Politics | 1,532 (50.07%) | 3,961 (54.65%) | 120 (72.29%) | 319 (75.41%) |
| Natural Disasters | 731 (23.89%) | 1750 (24.14%) | 22 (13.25%) | 61 (14.42%) |
| Mobility | 141 (4.61%) | 294 (4.06%) | 8 (4.82%) | 13 (3.07%) |
| Nature/Animals | 265 (8.66%) | 471 (6.50%) | 3 (1.81%) | 11 (2.60%) |
| Climate Activists | 83 (2.71%) | 168 (2.32%) | 5 (3.01%) | 6 (1.42%) |
| Other | 35 (1.14%) | 68 (0.94%) | 1 (0.60%) | 4 (0.95%) |
| Sustainability | 130 (4.25%) | 321 (4.43%) | 2 (1.20%) | 3 (0.71%) |
| Energy | 18 (0.59%) | 27 (0.37%) | 1 (0.60%) | 2 (0.47%) |
| Food | 23 (0.75%) | 34 (0.47%) | 2 (1.20%) | 2 (0.47%) |
| Env. Protection | 13 (0.42%) | 31 (0.43%) | 1 (0.60%) | 1 (0.24%) |
| Housing | 34 (1.11%) | 45 (0.62%) | 1 (0.60%) | 1 (0.24%) |
| Corona | 35 (1.14%) | 41 (0.57%) | 0 (0.00%) | 0 (0.00%) |
| Economy | 9 (0.29%) | 20 (0.28%) | 0 (0.00%) | 0 (0.00%) |
| General News | 11 (0.36%) | 17 (0.23%) | 0 (0.00%) | 0 (0.00%) |
Note. Categories created by manually aggregating results of BERTopic model (see Supplementary Table A8 in Appendix). Percentages are relative to outlet type. Categories are sorted by the number of visits to alternative news outlets.
Distribution of Unique Articles and Visits by Manual Categories Created from Topic Modeling
| Mainstream news | Alternative news | |||
|---|---|---|---|---|
| Manual category | N articles (%) | N visits (%) | N articles (%) | N visits (%) |
| Politics | 1,532 (50.07%) | 3,961 (54.65%) | 120 (72.29%) | 319 (75.41%) |
| Natural Disasters | 731 (23.89%) | 1750 (24.14%) | 22 (13.25%) | 61 (14.42%) |
| Mobility | 141 (4.61%) | 294 (4.06%) | 8 (4.82%) | 13 (3.07%) |
| Nature/Animals | 265 (8.66%) | 471 (6.50%) | 3 (1.81%) | 11 (2.60%) |
| Climate Activists | 83 (2.71%) | 168 (2.32%) | 5 (3.01%) | 6 (1.42%) |
| Other | 35 (1.14%) | 68 (0.94%) | 1 (0.60%) | 4 (0.95%) |
| Sustainability | 130 (4.25%) | 321 (4.43%) | 2 (1.20%) | 3 (0.71%) |
| Energy | 18 (0.59%) | 27 (0.37%) | 1 (0.60%) | 2 (0.47%) |
| Food | 23 (0.75%) | 34 (0.47%) | 2 (1.20%) | 2 (0.47%) |
| Env. Protection | 13 (0.42%) | 31 (0.43%) | 1 (0.60%) | 1 (0.24%) |
| Housing | 34 (1.11%) | 45 (0.62%) | 1 (0.60%) | 1 (0.24%) |
| Corona | 35 (1.14%) | 41 (0.57%) | 0 (0.00%) | 0 (0.00%) |
| Economy | 9 (0.29%) | 20 (0.28%) | 0 (0.00%) | 0 (0.00%) |
| General News | 11 (0.36%) | 17 (0.23%) | 0 (0.00%) | 0 (0.00%) |
| Mainstream news | Alternative news | |||
|---|---|---|---|---|
| Manual category | N articles (%) | N visits (%) | N articles (%) | N visits (%) |
| Politics | 1,532 (50.07%) | 3,961 (54.65%) | 120 (72.29%) | 319 (75.41%) |
| Natural Disasters | 731 (23.89%) | 1750 (24.14%) | 22 (13.25%) | 61 (14.42%) |
| Mobility | 141 (4.61%) | 294 (4.06%) | 8 (4.82%) | 13 (3.07%) |
| Nature/Animals | 265 (8.66%) | 471 (6.50%) | 3 (1.81%) | 11 (2.60%) |
| Climate Activists | 83 (2.71%) | 168 (2.32%) | 5 (3.01%) | 6 (1.42%) |
| Other | 35 (1.14%) | 68 (0.94%) | 1 (0.60%) | 4 (0.95%) |
| Sustainability | 130 (4.25%) | 321 (4.43%) | 2 (1.20%) | 3 (0.71%) |
| Energy | 18 (0.59%) | 27 (0.37%) | 1 (0.60%) | 2 (0.47%) |
| Food | 23 (0.75%) | 34 (0.47%) | 2 (1.20%) | 2 (0.47%) |
| Env. Protection | 13 (0.42%) | 31 (0.43%) | 1 (0.60%) | 1 (0.24%) |
| Housing | 34 (1.11%) | 45 (0.62%) | 1 (0.60%) | 1 (0.24%) |
| Corona | 35 (1.14%) | 41 (0.57%) | 0 (0.00%) | 0 (0.00%) |
| Economy | 9 (0.29%) | 20 (0.28%) | 0 (0.00%) | 0 (0.00%) |
| General News | 11 (0.36%) | 17 (0.23%) | 0 (0.00%) | 0 (0.00%) |
Note. Categories created by manually aggregating results of BERTopic model (see Supplementary Table A8 in Appendix). Percentages are relative to outlet type. Categories are sorted by the number of visits to alternative news outlets.
Delving deeper into the content of environmental news, we analyze the entities (persons, organizations, places, and miscellaneous entities) commonly mentioned in these articles, applying a named entity recognition model (Schweter & Akbik, 2020). We compare the relative frequency of mentions of each entity between mainstream and alternative news outlets. This means we do not consider entities that are only mentioned in one of the two types of outlets, focusing, like the topic modeling, on finding differences in commonalities. We do so by drawing word clouds of the 30 entities with the highest relative proportion (Figure 1). Large words in red, for instance, are significantly more prevalent in articles from alternative outlets. Similar plots for the other categories (locations, miscellaneous, as well as a comparison of opinion versus descriptive articles) are shown in the Appendix (Supplementary Figures A7, A8, and A9).
Word cloud of named entities.
Mainstream outlets are much more likely to emphasize the leaders of the two largest parties in Germany in the run-up to the election (Armin Laschet and Olaf Scholz), as well as mainstream German parties (SPD, FDP, and CDU). The connection between environmental topics and the election period, which already emerged in our topic analysis and in previous research, is again seen here. In contrast, alternative news outlets seem to focus on individuals and agendas that are not necessarily in the mainstream of German politics. This includes, for instance, Alice Weidel, the controversial chairwoman of the AfD, and Karl Lauterbach, an advocate of strong COVID-19 containment measures and Health Minister since the 2021 election. Further, we note the international and antimainstream nature of alternative media content, with more frequent mentions of international organizations (such as NATO and the WEF) and individuals (e.g., Bolsonaro and Lagarde). This is also somewhat echoed in the locations mentioned, with mainstream media emphasizing German states affected by the floods (e.g., Nordrhein-Westfalen) and alternative outlets over-representing other countries.
Survey Models
We now turn to our second research question and investigate the effects of alternative news consumption on environmental attitudes. We consider the consumption of all alternative news articles and articles that were classified as environment-related by our model. Though we focus our attention on alternative news consumption, we also display results for the consumption of mainstream news merely as a point of comparison.
Figure 2 displays the results of random effects within-between models, one for each outcome of interest (see Supplementary Table A11 for regression tables). Given the complex structure of the model, with different types of news consumption variables and effects, we interpret each of these in turn. We start with within effects, which rely exclusively on over-time variation in news consumption and shifts in attitudes. Within effects allow us to assess whether individual shifts in environmental attitudes from wave to wave are associated with individual changes in news consumption observed in the tracking data, regardless of individual baseline levels.
Results of random effects within-between models.
The dummy within coefficient reflects the average change in attitudes that is associated with being exposed to alternative news (i.e., read at least one article) versus not being exposed at all. Individuals who were always or never exposed across waves are excluded in this estimate since they do not vary in the independent variable. We find mostly null results, indicating no discernible effects of news exposure on environment-related attitudes. However, we do find a statistically significant effect (p < .01) of exposure to general alternative news articles on attitudes towards environmental protection, though in an unexpected direction. Amongst individuals who varied their exposure to alternative news (n = 117), support for environmental protection was on average 0.16 standard deviations higher in waves with prior alternative news consumption. Note that this effect concerns all alternative news articles, regardless of environmental classification.
Moving on to the extent of the exposure, measured via the relative count of visits, we now compare the effects of increased (or decreased) news consumption amongst individuals who registered at least one visit to that particular outlet type. In simpler words, the count within effect compares higher versus lower extent of consumption rather than any versus no exposure. Again, we find mostly null results for higher consumption of alternative (and mainstream) news on environmental attitudes. However, we do find that increased consumption of environmental news on alternative news outlets is associated with lower support for environmental protection (p = .04). On average, an increase of one news article per day on the environment from alternative outlets is associated with a 0.62 standard deviation decrease in support for environmental protection. This result should be taken with caution, however, since this effect is solely driven by the 36 individuals who consumed environment-related news on alternative news platforms.
Regarding between effects, which capture average differences in attitudes between individuals that are associated with their overall levels of news consumption regardless of over-time variation, we find more consistent effects. Constant exposure to alternative news, denoted by the dummy reflecting the share of waves with exposure, is associated with more critical attitudes towards the fight against climate change (β = −0.48, p < .01), environmental protection (β = −0.57, p < .01), and the Green Party (β = −0.50, p < .01). We also highlight the differential effects of exposure to alternative news outlets on environmental protection: whilst individuals who were on average more constantly exposed to alternative news are critical of environmental protection (as per the within effect), a change (increase) in exposure is associated with greater concern for protecting the environment.
We now consider the possible effects of opinion articles on the environment from both mainstream and alternative news platforms by decomposing environment-related news into opinion and descriptive pieces. Running the same models as in Figure 2, we replace the variables related to environment-related news in general with measures of consumption of opinion and descriptive articles on the environment. This more fine-grained approach allows us to uncover potentially existing effects that are “cancelled out” due to differential effects of opinion and descriptive pieces. For the sake of simplicity, Figure 3 displays only the coefficients for the consumption of environmental news rather than “general” news, with full results in the Appendix (Supplementary Table A12). Note, however, that this entails a loss of statistical power, particularly relevant for alternative news consumption.
Results of random effects within-between models. Consumption of opinion and descriptive environmental news.
Whilst the pattern is one of mostly null results, we uncover two between effects. Individuals who, on average, were more constantly exposed to opinion articles on the environment from alternative news sources are more critical of the fight against climate change (β = −1.12, p < .05). Similarly, we now find that individuals who consume more opinion pieces about the environment from mainstream sources tend to be supportive of climate change action. This estimate is more reliable given the larger sample size, and it is substantively significant: one extra environment-related opinion article per day during the election cycle from mainstream sources is associated with a 0.77 standard deviation increase in support for actions against climate change (p < .01). Whilst these results may indicate that perhaps opinion pieces are stronger in shaping opinion, note that the negative association between alternative environmental news consumption and environmental protection found in Figure 2 is attributed to descriptive pieces (β = −1.53, p = .02), whereas opinion pieces return a null result.
Discussion
This paper examined how populist right-wing alternative media cover environment-related news and whether such coverage has any impact on individuals’ attitudes. We summarize and interpret the results as follows, starting with the alternative media coverage of environmental news. First, we find that environment-related news make up a sizeable share of news content consumed on alternative platforms, especially if compared to mainstream outlets. Second, findings from topic modeling on the content of such coverage indicate that alternative outlets are less diverse than mainstream media when reporting on the environment, focusing particularly on political topics. Third, alternative outlets tend to focus more on “alternative” political actors and international organizations. In other words, alternative outlets are more likely to link environmental concerns to the dynamics of policies and political competition, particularly from an antimainstream angle. This suggests that the greater topic diversity commonly found in mainstream media (Freudenthaler & Wessler, 2022) and the political focus of alternative news outlets (Müller & Freudenthaler, 2022) are both extended to environmental news. Further, the right-wing and international focus of actors mentioned in environment-related alternative news is in line with the “anti-globalist” populist framing of environmental concerns posited by previous studies (e.g., Lockwood, 2018; Marquardt et al., 2022; Peeters & Maeseele, 2023). In sum, we conclude that alternative media contribute to the politicization of environmental topics, which may render climate action a controversial partisan issue rather than a universally shared goal.
There are important caveats to the points made above. First, the generalizability of our findings may be conditional on certain contextual factors. The emphasis and politicization of environmental news may be specific to an election period where a natural disaster also took place. It is unclear if these characteristics will hold in a different period (see Müller & Freudenthaler, 2022). More broadly, we present Germany as the most likely case for the salience and politicization of environmental topics. As shown by cross-country research on populist communication (e.g., Cassell, 2023; De Bruycker & Rooduijn, 2021), populist actors from different countries emphasize different issues depending on opportunity structures. We also do not explore how left-wing populist parties and alternative media approach environmental news. Future research can therefore enhance our knowledge of the link between populist communication and the increasingly salient topic of climate change by expanding the scope of analyses to other time periods, parties, and countries. Second, we have only scratched the surface of alternative coverage of environmental topics. Future research can, for instance, provide more fine-grained analysis by systematically analyzing the stance present in alternative news articles. Further, our topic modeling is mostly informed by news from mainstream sources (due to sample size limitations), such that we may not be able to detect frames that are specific to alternative outlets. One such frame is the potential differentiation between “nature conservation,” framed in nationalistic terms, and climate change (Buzogany & Mohamad-Klotzbach, 2021; Hoerber, Kurze, & Kuenzer, 2021).
Turning now to our second research question, that is, the potential effects of alternative news coverage on environmental attitudes, we find limited and mixed effects. First, we find that alternative news consumers are on average consistently more critical of environmental policies and the Green Party, regardless of short-term changes. Regarding short-term attitudinal effects, we find that increased consumption (i.e., more articles read) of environment-related news on alternative websites is associated with lower concern for environmental protection, and contrary to our expectations, this effect is attributed to descriptive articles rather than opinionated pieces. In contrast, more constant exposure to alternative outlets in general is associated with greater concern for environmental protection. This contradictory result may be explained by different types of individuals accessing different types of articles since only a fraction of alternative news consumers also read articles about the environment. A positive effect of general topic articles could also be attributed to a framing of the environment as nature conservation, which our classifier, trained on mainstream news websites, may not have classified as environment-related.
However, in line with past research (Guess et al., 2020; Müller & Schulz, 2021), these findings are limited by the small sample size of alternative news consumers, which may return false positive results. Conversely, the short time-period under analysis does not allow for the detection of longer-term persuasion effects. Null results should not be taken as evidence of no effects, especially since populist communication may serve to maintain and reinforce prior attitudes (de Vreese et al., 2018). Further, there are potential issues pertaining to the data used to measure alternative news consumption. We do not consider offline sources of information, and recent methodological studies on web-tracking data point out potential biases arising out of, for instance, respondents not constantly tracking their devices (e.g., Bach, Kern, Bonnay, & Kalaora, 2022). Finally, though we use a carefully validated list of alternative outlets, the online media landscape is dynamic, and outlets themselves may change in the extent of their antimainstream stance (e.g., Heft, Mayerhoffer, Reinhardt, & Knüpfer, 2020). Future research efforts could thus employ a longer time-period, a larger sample size, and more precise measures of environmental attitudes in order to provide more robust results and a more nuanced understanding of the link between alternative media and climate change. Such expanded research design may also be able to test particular mechanisms that pit, for instance, economic versus noneconomic factors (Lockwood, 2018), as well as the potential causal effect of natural disasters.
Despite these shortcomings, this paper makes a significant contribution to the field of alternative news studies by providing evidence on its content and effects. We improve on past studies that classify environmental or opinion news based on keywords (e.g., Yan et al., 2022) and employ a sophisticated classification method based on a large and labeled fine-tuning dataset. The combination of longitudinal survey and web-tracking data also allowed for an objective (if imperfect) measure of individuals’ news consumption, which could be paired with three environment-related outcomes and a set of strong attitudinal controls. Further, Germany is one of many countries with a prominent populist right-wing party and an alternative media landscape. As climate change garners increasing public attention, environmental issues rise in salience both in the political and media agendas. Populist right-wing actors emerge as key opponents of climate action, with populist alternative media acting as an important channel for climate-skeptical views. Understanding the connection between populism, alternative media, and climate change is therefore crucial to our grasp of the current political and media landscape.
Biographical Notes
João Areal is a Doctoral Student at the Centre for Doctoral Studies in Social and Behavioural Sciences (CDSS) and a researcher at the Mannheim Centre for European Social Research (MZES) at the University of Mannheim, Germany. His research lies in the field of comparative political behavior, more specifically in the subfields of political identities and polarization.
Jan Behnert is a Master’s Student at the School of Social Sciences of the University of Mannheim and a research assistant at the Mannheim Centre for European Social Research (MZES) at the University of Mannheim. His main interest lies in data science approaches in the context of social sciences.
Dean Lajic is a Master’s Student at the School of Social Sciences of the University of Mannheim and a research assistant at the Mannheim Centre for European Social Research (MZES) at the University of Mannheim. His main interest lies in data science approaches in the context of social sciences.
Ruben Bach is a Research Fellow in the Data and Methods Unit of the Mannheim Centre for European Social Research (MZES) at the University of Mannheim. His research is rooted in computation social science and tackles questions of online news media consumption, AI-guided decision-making, and novel data for social science research.
Funding
R.B. gratefully acknowledges the support of this project through a postdoctoral fellowship with the Baden-Württemberg Foundation.
