The German Media Discourse on
Immigration
A Relational Mixed Methods Approach to
Corpus-Assisted Discourse Studies
Master Thesis
Hannah-Marie B¨uttner Amsterdam, 13 August 2020
Research Master Social Sciences
hannah-marie.buttner@student.uva.nl UvA ID: 12292311
Supervisor: Prof. Dr. Jan Willem Duyvendak Second reader: Dr. Alex van Venrooij
Abstract
Immigration is a central topic in the public discourse in Germany, especially since the so-called ‘European refugee crisis’. Mass media (re)produce this discourse and hence play an essential role in the construction of social realities. This study provides reconstruction and critical reflection of the German media discourse on immigration for the period from late 2017 to early 2020 from a Critical Discourse Analysis point of view. Using an innovative approach of Corpus-Assisted Discourse Studies (CADS) more than 16,000 newspaper articles were analyzed. With a method mix of machine learning techniques, network analysis, and qualitative text analysis, the study was able to reveal the dominant topics and actors in the discourse and the contents resulted from their relations to each other. The findings highlight both a pronounced dominance of political actors as well as a distinct marginalization of migrants. Furthermore, the paper demonstrates the strength of CADS and expands them by a novel mixed methods approach capable of uncovering patterns in discourses that only materialize in an abundance of text, while at the same time revealing more complex embedded meanings in these discourses.
Contents
Abstract i
List of Figures iii
List of Tables iv
1 Introduction 6
2 Theoretical Framework 8
2.1 The German Media Discourse on Immigration: A Literature Review . . 8 2.2 Access, Power, Manipulation: Theoretical Foundations of Critical
Dis-course Analysis . . . 11
3 Research Design 13
3.1 Methodology: Corpus-Assisted Discourse Studies . . . 13 3.2 Data . . . 15 3.3 Methodical procedure . . . 16 4 Results 20 4.1 Descriptive Findings . . . 20 4.2 Cluster analysis . . . 21 5 Discussion 36 6 Conclusion 40
List of Figures
1 Relative frequencies of the identified discourse topics’ occurrences over time. . . 20 2 Network of topics and actors. . . 22 A1 Coherence scores calculated for different topic models and varying numbers
of topics k. . . 51 A2 Relative frequencies of occurrence of the ten most prevalent persons over
List of Tables
A1 Overviews of newspaper corpus composition. . . 49 A2 Definition of discourse topics based on keyword search with regular
1
Introduction
Since the beginning of the so-called ‘European refugee crisis’1 in 2015, discussions about
immigration have dominated the public discourse in Germany and other European countries consistently. Far-right populist and racist acts and expressions have increased in the meantime. In Germany, the right-wing populist party AfD has managed to enter the parliament and publicly manifested far-right extremism has become more prevalent, as exemplified by the many supporters of the xenophobic and Islamophobic organization Pegida. The consequences of immigration, asylum policy, and fear of ¨Uberfremdung (‘over-foreignization’) are issues that are not exclusive to the right edge but permeate wide areas of the entire public discourse in Germany. Mass media (re)produce this discourse decisively and thus bear a high responsibility for the construction of social reality (Yildiz 2006, p. 37). What the majority of the German society knows about immigration and migrants, they know almost exclusively through the mass media (Van Dijk 2016). The realities constructed in newspapers, television programs, and other media channels influence public opinion about immigration and affect the life situations of immigrants who are typically marginalized in the debate. To counter the rise of xenophobia in Germany, and the damaging and destabilizing effects that racism has on all social coexistence (Bonnett 2005; Pedersen, Walker, and Wise 2005), it is therefore of utmost importance to identify and reflect on the constructions of immigration in the media discourse.
This study critically examines the media discourse on immigration in Germany by analyzing articles of 17 German newspapers from December 2017 until February 2020. Since Germany is not merely the most populous country in the European Union but is also perceived by many as a leading power, it is of great value even for a larger European context to examine the German discourse on immigration issues. Focusing on the discursive elements of topics and actors, the study aims to answers the following research questions:
1 The term is written with quotation marks because as it is typically used in public, political, and
academic discourse, the notion refers to a eurocentric perspective which frames “the one million arrivals in Europe in 2015 as a sudden and unexpected event with exclusively negative effects for the continent” (Chouliaraki and Zaborowski 2017, p. 632). In line with Chouliaraki et al. (2017), I argue that the term disregards the realities of non-European hosting countries – Turkey and Lebanon, for example, are the countries where most refugees from Syria have been accommodated by the end of 2019 (together almost 4.5 million), and Pakistan has hosted over 1.4 million people from Afghanistan by then (UNHCR 2019).
(1) Which are the dominating themes in the discourse and how is this changing over time?
(2) Who are the dominant actors in the debate?
(3) How are these actors and topics connected to each other and which discourse contents result from these relations?
To answer these questions, a mixed methods approach that combines theories and methodologies of Critical Discourse Analysis and Corpus Linguistics is employed. Using machine learning algorithms and other dimension-reducing corpus-based methods, the new potentials resulting from today’s abundance of accessible text data are exploited for investigating the construction of meaning through natural language. Through the additional use of a network approach, the individual discursive elements are examined from a relational perspective, which allows the discourse to be analyzed as a connected social system consisting of these interrelated components. Qualitative text analysis of small samples is used for an in-depth analysis, which enables the discursive context of the identified topics and actors to be grasped.
The use of corpus-based methods for discourse analysis overcomes a major weakness of traditional discourse and content analysis, namely the enormous amount of time required to analyze relatively small text corpora. At the same time, large-scale text analyses enable the identification of prevalent language patterns which help to uncover underlying structures in the discourse, and which small-scale analysis may have missed (Baker and McEnery 2005). Hence, the project also provides a decisive methodological contribution to the relatively young field of Corpus-Assisted Discourse Studies (CADS).
The paper starts by discussing previous findings about the media discourse on immigration in Germany and providing an overview of existing corpus-based studies in this field. Subsequently, the theoretical foundations of Critical Discourse Analysis (CDA) underlying the investigation are explained. In the following section, the methodological approach of CADS, the data collection, and the employed methods are explained in more detail. After presenting the results, a critical discussion of the findings takes place.
2
Theoretical Framework
2.1
The German Media Discourse on Immigration: A
Litera-ture Review
Numerous studies have repeatedly demonstrated that the media maintain and dis-seminate prejudices and stereotypes against migrants and ethnic minorities. Previous research on media coverage of immigration in Germany consists mainly of qualitative and quantitative content and discourse analyses. The studies have focused on the analysis of television programs (Kr¨uger and Simon 2005; K¨uhne-Scholand 1987; Ruhrmann and Sommer 2005; Ruhrmann, Sommer, and Uhlemann 2006; Schorb, Echtermeyer, Lauber, and Eggert 2003) and newspaper articles (Abadi, d’Haenens, Roe, and Koeman 2016; Addicks et al. 2012; Berry, Garcia-Blanco, and Moore 2016; Delgado 1972; Gr¨af 2008; H¨omberg and Schlemmer 1994; J¨ager, Cleve, Ruth, and J¨ager 1998; Meißner and Ruhrmann 2001; Merten and Ruhrmann 1986; Predelli 1995; Ruhrmann and Kollmer 1987; Vollmer and Karakayali 2018).
Almost all of these studies have shown that immigration is reported on in an event-related manner and especially in negative contexts. Immigration is primarily framed in the media as a threat. That this is especially the case when it comes to the topic of asylum was already recognized in the 1980s by K¨uhne-Scholand (1987) and later confirmed by Berry et al. (2016). The latter also demonstrated that the framing of immigration as a threat was in the period under study (June 2014–April 2015) mainly based on the argument that IS terrorists and other extremists could be hiding among refugees coming to Europe.
In general, it can be distinguished between more or less desired migrants in the media discourse, as shown by Merten et al. (1986). This differentiation depends primarily on the intention of the stay and the cultural affiliation: invited guests with a temporary stay (e.g. athletes) as well as people from countries that are culturally similar to Germany are much more desired than uninvited asylum seekers whose duration of stay is uncertain and who come from culturally foreign regions. However, the authors also observed a predominantly negative reporting about migration.
A large number of previous studies have also identified certain over- and under-representations that lead to a distorted picture of immigration in the German media. On the one hand, specific nationalities such as Turks are generally mentioned significantly more often and not in accordance with the actual proportion of these nationalities among immigrants in Germany (Delgado 1972; Predelli 1995; Ruhrmann 2002; Trebbe and K¨ohler 2002), especially when they are perceived as particularly conspicuous and
alien (Ruhrmann and Sommer 2005). On the other hand, migrants as a whole are disproportionately strongly represented in the media when it comes to the topic of crime, i.e. the ratio of migrants and non-migrants in reporting on criminal offenses does not correspond to the crime statistics (J¨ager et al. 1998; Ruhrmann and Sommer 2005; Schorb et al. 2003). Furthermore, migrants are portrayed in a more dangerous and brutal way compared to criminals who are considered non-migrants, and they are often accused of being involved in organized criminality (J¨ager et al. 1998).
Another finding of virtually all previous studies is that migrants appear merely as passive and collectivized objects that are hardly ever given a voice in media discourse. They appear extremely rarely as interview partners and, as Gr¨af (2008) observed, even perpetrators of xenophobic attacks have a greater say in the media than their victims of foreign origin. Especially asylum seekers are portrayed as a collective and amorphous mass (H¨omberg et al. 1994).
A counter-example of the predominantly passive and negative connotations of mi-grants in German media is provided by Kr¨uger et al. (2005) with their investigation of the entire television program of Westdeutscher Rundfunk for four weeks in 2003. Positive and negative reports on migration were balanced here and people with a ‘migration background’ were given a voice in almost¾ of the program. However, this applies above all to fictional and other entertainment genres, not so much to news coverage.
Moreover, previous research, especially on German newspapers, have found that the discourse on immigration is highly political and that representatives of the government and other political institutions are the key actors in the debate (Berry et al. 2016; Gr¨af 2008; H¨omberg et al. 1994; Ruhrmann and Sommer 2005). Ruhrmann, Sommer, and Uhlemann (2006), on the other hand, demonstrated a thematic shift away from the classical, political migration issues towards increasingly sensationalist reporting on terrorism and crime. For the year 2003, they observed that more than a third of the reporting on migrants and migration issues is part of a debate on terrorism in Germany, which is according to the authors much more frequently than in earlier periods. The long-term study by Gr¨af (2008) also confirmed for the period 1995-2005 that terror and crime are the most dominant topics in the media discourse on immigration.
Discursive shifts in the German immigration discourse were also shown by Vollmer et al. (2018) in their historical-critical discourse analysis of several German newspapers for the period between March 2015 and March 2016. The Cologne New Year’s Eve 2015/16, during which many presumed migrants sexually harassed women, has led to a shift in the discourse away from vulnerable children and families who deserve help to “the young male adult [who] ranges at the opposite end of the vulnerability-deservingness scale and
more easily fulfills the requirements of a discursive pattern in which migrants are framed as evil-doers”. (Vollmer et al. 2018). According to the authors, the discursive changes have led to alterations in the meaning of categories such as ‘migrants’ or ‘refugees’. Refugees have been re-demonized as undeserving migrants (e.g. ‘economic migrants’), which has opened a “window of ideological opportunity” for right-wing parties.
A limitation of the majority of the mentioned studies is the small number of cases, especially in qualitative studies, which makes it difficult to permeate the discourse comprehensively. Besides, traditional quantitative content analyses use a deductive approach which hardly allows to examine the discourse outside of predefined categories. Corpus-assisted research on the immigration discourse in Germany, which is able to analyze a much larger amount of text through the use of Corpus Linguistics methods, is still scarce. As Mattissek and Schopper (2019) specify “Corpus linguistics aims to identify structural characteristics of discourses that appear in regularities of language use that are typically too broad and numerous to be discovered through close reading (or hearing) of singular texts” (Mattissek et al. 2019, p. 253). The following summarized studies of this field all use a combination of the methods frequency2, concordance3,
collocation4, and keyword analysis5.
The study by Becker (2016) is one of the very few that analyzed data from various media channels (newspaper articles, discussions in online forums, and political talk shows). Using a corpus-based approach, she examined the relevance of fear in the media discourse on immigration in the years 2013 and 2014. Through a comparative analysis of texts containing the root Asyl (asylum) with a general corpus of news, she showed that forms of the word Angst (fear) are increasingly used in the context of asylum issues and that fear is always a matter of ‘the Germans’, and never of people seeking asylum in Germany.
Rada (2016) combined cluster and collocation analysis as well as qualitative text analysis to investigate the concept of Willkommenskultur (welcome culture) in the discourse on immigration based on newspapers. She argued that the term was a keyword
2 Frequency analysis provides an initial overview of a corpus by indicating how frequently a term or
combination of terms appears in a given corpus, but it can also be used to identify characteristic word scatter, e.g. between periods or stances of speakers (Mattissek and Schopper 2019).
3 Concordance analysis is used to examine the immediate co-text of words by identifying the terms
that occur right next to a search term.
4 Collocation analysis is used to investigate the joint occurrence of words. Noticeable features are
determined by comparing the frequencies of co-occurrences with a random distribution.
5 Keyword analysis identifies terms that are typical for certain corpora, i.e. words that appear more
in the migration discourse from early 2015 to early 2016 and revealed an overall positive connotation of the concept as openness towards migrants, which influences Germany’s self-image.
Mattissek et al. (2019) also applied a corpus linguistic approach to investigate how identities are constructed in the refugee discourse with regard to culture(s). By analyzing German newspaper articles published from May 2015 to April 2016 they examined the mutual collocations of the term Kultur (culture) with other corpus-specific words and its concordance lines. They concluded: “culture is frequently used to constitute an antagonistic division between the ‘self’ and the ‘other’, thereby creating the illusion of a homogenous German political identity which is threatened by refugee immigration.” (p. 261) As a result of their “cultural incompatibility” with the German society, refugees are thus presented as essentially foreign.
Finally, Griebel and Vollmann (2019) studied the dominance of various “cultural-ization regimes” that tend either to societal openness or closure for the period from January 2015 to September 2017. Through a comparative analysis of the keywords in two German newspapers and the collocates of lemma such as Fl¨uchtling/Gefl¨uchtete (refugees) and Asylbewerber (asylum seekers), they demonstrated that the left-leaning die tageszeitung generally advocates an open society and solidarity with those referred to as Gefl¨uchtete, whereas Fl¨uchtlinge are also addressed in the context of exclusion from society. The conservative Die Welt mainly problematizes migration in general and tends much more towards a closed society.
The present study expands the corpus-assisted research on the immigration discourse in Germany through a novel methodological approach. Instead of relying only on the classical methods of corpus linguistics, machine learning algorithms, and network analysis combined with qualitative text analysis are used, which will be explained in more detail below. From a theoretical point of view, the study is grounded in CDA.
2.2
Access, Power, Manipulation: Theoretical Foundations of
Critical Discourse Analysis
The German media discourse on immigration is studied using an approach inspired by CDA. As Hajer (2006) points out, the assumption underlying all discourse studies “is that language profoundly shapes our view of the world and reality, instead of being merely a neutral medium mirroring it.” (p. 66). Language is thereby in the position to change power relations by altering socially shared symbols and signs, which can affect policy-making. Discourse is thereby understood “as an ensemble of ideas, concepts,
and categories through which meaning is given to social and physical phenomena, and which is produced and reproduced through an identifiable set of practices” (p. 67). Hajer establishes the concept of “story lines” referring to complex narratives compressed in concise claims and which are supposed to be activated by cues in the mind of the recipient. They have profound organizational value for social interaction. However, it is often not the case that a cue elicits in the recipient what the speaker had in mind.
One specific type of such cues, which are intended to convey complex meanings, are metaphors. The metaphor of the ‘wave’, for example, which is frequently used in the discourse on immigration, especially in relation to refugees, not only articulates a large number of migrants, but also stimulates mental models associating waves with overwhelming masses of water in which the society risks drowning (Van Dijk 2016).
How we think and act is structured to a large extent by a system of concepts of which we are not aware and which is predominantly metaphorical (Lakoff and Johnson 1980). The very absence of awareness of this system makes it an important factor in the spread of racism because metaphors have the power to influence world views and actions and can thus become a dangerous weapon of political propaganda (Wasniewska 2017, p. 46f.).
Turning to the theoretical framework of critical studies, according to Van Dijk (1995), CDA is characterized by a strong focus on the (re)production of power relations. It considers consent, inequality, and injustice through text and/or talk, as well as the processes of legitimation and manipulation thereby. CDA is not a specific method but can involve any that contributes to the critical study of social inequality. Often conducted as inter- or multidisciplinary research, it is employed to study implicit or hidden structures of “the discursive conditions, components and consequences of power abuse by dominant (elite) groups and institutions” (Van Dijk 1995, p. 24). Discursive power and dominance are determined by the access to and control over discourse, which is provided by institutional resources (e.g. job positions) and group membership (e.g. racialized groups). Through the dominance of the voices of powerful groups and by marginalizing other viewpoints “discourse itself becomes a ‘segregated’ structure” (Van Dijk 1993, p. 260).
As Van Dijk (1993, 1995) further explains, control over discourse is accompanied by the manipulation of the minds of recipients of this discourse. Dominant groups are thus able to influence social cognition, “involving the influence of knowledge, beliefs, understanding, plans, attitudes, ideologies, norms and values” (Van Dijk 1993, p. 257). The processes of influencing the recipients’ mindsets are highly complex and depend on the characteristics of the text, the context, and in particular on the recipients’ prior
ideologies and knowledge.
The present study on immigration in Germany identifies dominant actors and topics in the media discourse. By this means, various discursive fields are revealed, i.e. the “dynamic terrain in which meaning contests occur” (Steinberg 1999, p. 748). These dynamic and not definitely delimitable fields develop in discussions between protagonists and antagonists as well as bystanders (Snow 2004). They constitute “public battlegrounds” on which various actors vie for the interpretative dominance over a certain topic (T¨ornberg and T¨ornberg 2016a).
Which issues are addressed in these discursive fields is largely determined by the elites controlling the discourse and “by emphasizing specific topics at the expense of others . . . or by preventing others to address other topics, they may influence the overall (mental) model structures that are involved in discourse comprehension” (Van Dijk 1995, p. 23). It is therefore assumed that the discursive fields consisting of dominant actors and themes identified in the discourse fundamentally influence the German society’s knowledge about and attitudes towards immigration. Actors who are absent from the discourse, or only passively occurring without having a voice themselves, are considered to be marginalized by the powerful. This study aims to uncover these power relations in the German immigration discourse to counteract them. Therefore, the project is explicitly political which is in line with critical studies because they are “unabashedly normative: any critique by definition presupposes an applied ethics” (Van Dijk 1993, p. 253).
3
Research Design
3.1
Methodology: Corpus-Assisted Discourse Studies
Classical approaches to CDA often involve solely qualitative text analysis techniques (sometimes supported by the analysis of images, film, or sound material). They are criticized because the analyzed material consists only of exemplary fragments of the discourse (Fowler 1996). Others claim that CDA lacks academic rigor (Widdowson 1996; Widdowson 2000), especially because data selection and analysis are inclined to confirm the subjective prejudices of the analyst (Orpin 2005). Moreover, a weakness of purely qualitative discourse analyses that only cover small N is that they easily overlook patterns of language use, since these are realized through a huge amount of text and individual documents contain only small snippets of the underlying discourses (Stubbs 1997).
Today’s abundance of produced and digitized text data creates new possibilities to study the construction of meaning through natural language. Yet it requires innovative methods to be able to grasp these opportunities. Using computer-based methods, the ‘concrete text corpus’, i.e. the sample to be analyzed from the virtual corpus which contains all texts produced on a topic (Busse, Teubert, Angermuller, Maingueneau, and Wodak 2014), can be much larger than with classical approaches of discourse analysis. The combination of discourse analysis with automated text analysis methods that are typically used in Corpus Linguistics has been established in recent years under the term Corpus-Assisted Discourse Studies (CADS) (Cheng 2012; Hardt-Mautner 1995; Partington 2006; Stubbs 1996; Wodak and Meyer 2009). There seems to be a growing recognition of the potential of these methods, as there are several recent studies using techniques of Corpus Linguistics for CDA (Baker and McEnery 2005; Baker, Gabrielatos, et al. 2008; Hellsten and Dawson 2010; Nelson 2006; Orpin 2005; T¨ornberg et al. 2016a; T¨ornberg and T¨ornberg 2016b)
Corpus Linguistics comprise different methods of the linguistic analysis of text corpora and aims to find “probabilities, trends, patterns, co-occurrence of elements, features or groups of features” (Teubert and Krishnamurthy 2007, p. 6). Scholz (2019) summarizes the benefits of using such methods in discourse analyses:
Corpus methods applied in discourse studies can help to gain an overview over dominant discourse participants, discursive positions, topics, and arguments, and their interrelations relatively quickly. They can trace discursive dynamics over time within these discursive structures. And they permit the analyst to ‘zoom in’ into parts of the data which have proven to be of particular interest within the research process. (p. 12)
CADS can be performed as a methodical triangulation consisting of separate quali-tative and automated text analysis elements, as is the case in Baker and Levon (2015). Triangulation refers to the conventional idea that qualitative and quantitative approaches can be combined to validate results by mutually confirming them (Bryman 2012). How-ever, CADS can also be carried out in a way in which qualitative and quantitative approaches interact as a “methodological synergy” (Baker, Gabrielatos, et al. 2008). The present study uses the various methods in such a synergy, with analytic and interpretive steps taking place alternately, building on and intertwining each other. In this way, the project seeks to dissolve the boundary between the seemingly opposing paradigms of qualitative and quantitative research, a line that is still strongly advocated in the social sciences. The goal of CDA, to uncover hidden power structures, is pursued with this large-scale text analysis, in the sense that it is based on the assumption that every single text can influence the reader, but that the dominance of certain discourse elements only
becomes apparent in the mass of texts.
3.2
Data
To study the immigration discourse in Germany, a text corpus of German newspaper articles was generated based on a keyword search. Newspaper articles are highly suitable as a unit of analysis for the investigation for several reasons. Firstly, newspapers are easily accessible due to extensive digitization. Every prominent German newspaper nowadays features an online version. Secondly, newspapers reach a large audience. Print and online papers together currently reach 89 percent of the German-speaking population (DieZeitungen 2019)6. Finally, newspapers function as important political
agenda-setters (Meraz 2007). They continue to be an important source of information for the “symbolic elites”, i.e. those who control public communication, influence public opinion, and legitimize politics (Van Dijk 1993; Van Dijk 2016).
The articles analyzed in this study were scraped using INCA, a Python module developed for the analysis of media content (Trilling et al. 2018). Articles related to the German immigration discourse were gathered by searching for the German equivalents of the keywords ‘migrant’, ‘migration’, ‘immigration’, ‘immigrant’, ‘refugee’ and ‘asylum’ in all German-speaking newspapers and magazines covered by the database that INCA accesses. Since a single occurrence of one of the keywords was sufficient for the article to be considered relevant, the corpus contains both texts in which immigration plays a central role and others in which the subject is only peripherally addressed. The latter can be of great value in discourse analysis because it is precisely those incidental references that echo the dominant assumptions often better than the texts that deal explicitly with the subject and are therefore omitted by readers who are not clearly interested in it (DiMaggio, Nag, and Blei 2013). Concerning the publication dates, no restrictions were determined. However, as the applied database has only covered German newspapers for a couple of years, all articles were published in the period from December 19, 2017, until February 2, 2020, the day of data collection.
Since INCA uses a database that only covers a selection of German newspapers and periodicals, additional articles were scraped from the Nexis Uni database for Bild, which is by far the newspaper with the highest circulation in Germany. However, as ultimately only articles containing more than 500 words were used for the analysis,
6 Referring to the German population from the age of 14, each reader was counted only once, no
Bild was not represented in the corpus after all, since only keyword-relevant articles of shorter length were found. T¨ornberg et al. (2016a, 2016b) argue that the Latent Dirichlet Allocation (LDA) (Blei, Ng, and Jordan 2003), the topic modeling algorithm applied in this study, performs best for documents of at least 1000 words. Since no evidence for this threshold could be found in further literature and only 15 percent of the corpus consisted of texts of the recommended length, a compromise of a minimum length of 500 words was reached, which applied to half of all gathered articles.
Besides the criteria of text length, three factors affected the completeness of immigration-related articles. Firstly, as mentioned above, the database used did not include all German newspapers, so that only a part of the relevant media was represented in the corpus. Secondly, many articles could only be scraped partially with INCA, so that data as title and publication date were available but not the content of the article. These articles were excluded from the corpus. Thirdly, there was a significant data gap in October 2018, which is very likely due to a failed server and could therefore not be filled within the scope of this project. The final corpus used for the analysis consisted of 16,321 articles from 17 different newspapers and magazines (mainly their online versions), of which the vast majority are national daily and weekly papers, except of Der Tagesspiegel which is regional for Berlin (see Table A1 in the appendix for an overview of the corpus composition). Since the corpus contains a balanced number of newspapers politically oriented to the left, right and center, it largely reflects the political spectrum of the German media landscape.
3.3
Methodical procedure
The combination of methods employed to study the German media discourse on im-migration comprises different techniques of automated text analysis, social network analysis, and qualitative text analysis. Analytic and interpretive steps alternated and built on each other in a dynamic and interdependent manner. In the following, the individual steps of the methodical procedure are explained in more detail.
Named entity recognition
For automatic identification of the actors in the discourse a named entity recognition was carried out, a method that is based on supervised machine learning to detect persons, organizations, locations, etc. within a text corpus. To this end, the newspaper articles were first pre-processed by removing any punctuation. The analysis was finally carried out with the Python package spaCy, leading to a recognition of 322,123 occurrences of
persons, out of which 87,781 were unique (after normalization of names7), and 243,026 occurrences of organizations, out of which 56,278 were unique. To further reduce the dimensionality, the 200 most frequently occurring persons and organizations were then counted based on the detected actors. These lists were then manually cleared of false identifications, such as Mann (man) and Asyl (asylum).
Topic modeling
To investigate the discourse themes, topic modeling, a form of unsupervised machine learning, was applied. Topic modeling algorithms find inductively, thus without pre-defined keywords, themes that occur in the corpus by returning clusters of co-occurring terms. The method provides “an overview or ‘content map’ of immense sets of documents, revealing small but systematic patterns and tendencies in the data” (T¨ornberg et al. 2016a, p. 404).
The topic modeling was preceded by several pre-processing steps to improve the performance of the algorithm. With the Python library NLTK, the articles were cleaned of punctuation and all words were transformed to lower case. The documents were tokenized, i.e. split into separate words, and stopwords, which have no relevance for capturing the content, were removed. Furthermore, a part-of-speech tagging was performed to identify adjectives and nouns and to perform the topic model only with these most meaningful words. Inspired by Konrad (2016), for this purpose, a tagger was created using the Python class ClassifierBasedGermanTagger, developed by Philipp Nolte in 20118, and trained with the TIGER Corpus which consists of annotated German
newspaper texts provided by the University of Stuttgart. After filtering out all words apart from nouns and adjectives, these were lemmatized to reduce derived and inflected forms of a term to a common basis.
For modeling the topics, the popular LDA algorithm (Blei et al. 2003) was used, which is based on Bayesian statistical theory (Gelman et al. 2013). In a hierarchical probabilistic model, the topics and ratio of topics per document are considered as latent variables (T¨ornberg et al. 2016b). The model produces a collection of topics and
7 Since the algorithm identified different persons when they were sometimes called by their full name
and sometimes only by their first or last name, the names were normalized in such a way that if a person’s full name occurred at least once in a document, all other occurrences of the first or last name were replaced by the full name for that specific document.
8 Nolte’s code to create the tagger can be retrieved on
estimates for each document the ratio of these topics and which words are associated with them.
During the technical implementation of topic modeling, for which the Python library gensim was used, several hyperparameters are adjustable, which can influence the performance of the model significantly. These include amongst others the number of topics k, the concentration parameter alpha determining the document-topic intensity, and the beta parameter determining the topic-word density. Furthermore, the corpus, which serves as input for the model, can be modified in various ways.
To define the best possible combination of these factors, four differently adapted corpora were initially created. For the base model, the word id was simply mapped with the word frequency per document. For the second corpus, extremes were filtered out, i.e. all words that occurred in less than five documents or more than 50 percent of the texts. The third corpus was modified with a tf-idf (term frequency-inverse document frequency) vectorizer, which weights the words according to their relevance in the corpus. The fourth input corpus was finally created as a combination of filtering out the extremes with the tf-idf vectorizer.
These corpora were then used for models estimated with different k in the interval [10,100] in steps of ten. The comparison of the coherence scores, a measure of quality for these models, has shown that it is the tf-idf model with 20 topics that performs best (see Figure A1 in the appendix). For k = 20, models with every possible combination of alpha and beta parameter values in the interval [0.01,1) were calculated in steps of 0.1. The optimal model determined from these calculations achieved a coherence score of 0.62.
Topic definition
The decisive step in defining a topic is the interpretation of the model’s output, i.e. the word sets that each represents a topic. Since the model did not provide any directly interpretable themes – possibly caused by a too heterogeneous corpus –, the top words from each topic were selected and their concordance lines and a selection of the articles in which they occurred were thoroughly read to grasp the meaning, relevance, and context of these words. For this purpose, the freeware corpus analysis toolkit AntConc (Anthony 2019) was used. For each term and combination of words, a varying number of articles were read until theoretical saturation was reached, i.e. until reading additional articles no longer provided substantial new insights for the definition of a theme. In the end, fourteen topics were defined based on keywords that were considered to be particularly meaningful. The entire corpus was then automatically scanned for these
words to identify which themes appeared in which documents. For example, the topic Activism was considered as present in an article if at least one of the root words Aktivist, Demonstr, or Protest occurred. For other topics, however, a condition was defined that a combination of certain words must occur (see Table A2 in the appendix for the exact keywords per topic). Out of these fourteen topics, three were finally excluded because they occurred only in a very small number (at most one percent) of the articles.
Social network analysis
To analyze the connections between the extracted elements of the immigration discourse, namely actors – in the sense of persons and organizations – and topics, a social network was constructed with them. Through this relational perspective, the discourse could be approached in a way that, despite the strong reduction of dimensions, allowed it to be analyzed as a social system consisting of the mutual influence of its individual elements. Network analysis can reveal not only the direct connections between different actors and topics in the discourse but also those where several entities are indirectly related to each other through the common connection to another entity. Through these chains of links and nodes, distant parts of a system can influence each other (Borgatti, Everett, and Johnson 2018). Using network analysis can also disclose the discursive power of actors which results partly from their network position (Uitermark, Traag, and Bruggeman 2016).
The network of topics and actors were constructed using igraph for Python and visualized with the network analysis software Gephi. Based on their correlation in the documents, weighted links were assigned between nodes whenever they had a positive correlation value of at least 0.29. This provided a network with 229 nodes, 540 edges, and
107 weakly connected components, thus a rather sparse network graph10. Subsequently,
however, all nodes with less than two links to other nodes were excluded from the network, which resulted in a much denser network11 of only four weakly connected components with 115 nodes and 506 edges. Clusters of topics and actors were initially identified using the community detection algorithm (based on modularity) provided by Gephi. Afterward, some classes were merged to reduce dimensionality.
9 Different thresholds were tested and 0.2 seemed to work best to capture significant connections
between actors/topics but also in an appropriate dimensionality that was still interpretable.
10 Graph density = 0.021. 11 Graph density = 0.07.
Analysis of network clusters
The clusters of actors and topics revealed by the network analysis were analyzed with a second round of qualitative text analysis. The appearance of the actors and topics represented in a cluster was inspected considering their co-texts as well as contexts. For this purpose, the software AntConc was used again to read the concordance lines for a given cluster, i.e. the direct co-text that related to an actor or topic occurrence, as well as entire articles to grasp the context. Again, the principle of theoretical saturation was applied.
4
Results
4.1
Descriptive Findings
Figure 1: Relative frequencies of the identified discourse topics’ occurrences over time.
As illustrated in Figure 1, by far the most common topic in the German media discourse on immigration during the entire period under review was Right-wing. At the end of 2017 and the beginning of 2018, the prominence of the topic increased continuously and from then on appeared almost uninterruptedly in about 60 percent of the immigration-related newspaper articles. The topics Left-wing and Activism occurred on average in about one third of the articles. Their fluctuations over time were relatively consistent, which speaks for a frequent common occurrence of the two topics in the media discourse. For most of the period under study, they remained the most dominant
topics in the immigration discourse after Right-wing. The themes were merely overtaken in their prominence by the peaks of the strongly fluctuating topic European. This happened most noticeably in June 2018 and April 2019 – of which the latter peak is probably due to the European Parliament election in May 2019.
The topics Crimes and Events – the latter represents highly publicized crimes of (alleged) immigrants addressed via the city names of the respective crime scenes – fluctuated strongly in their dominance in the immigration discourse. This may suggest that the topics were discussed especially in the course of ‘real events’ (e.g. the murderous knifing in Chemnitz in late August 2018 and the demonstrations that followed).
Populism, Youth, Turkey and USA were the least common among the defined topics examined in detail. Their prominence in the immigration discourse remained very constant over the period under study, except for an extraordinary occurrence frequency of Turkey in October 2019 – due to the much-discussed Turkish military offensive in northern Syria during that time.
4.2
Cluster analysis
Figure 2 illustrates the correlation network of the actors and topics identified in the discourse. The position of the nodes resulted from the ForceAtlas algorithm provided by the network analysis software Gephi. However, positions were slightly adjusted to prevent overlap of labels. The node size is proportional to the node degree, i.e. the number of links an actor or topic has to other nodes. While nodes representing topics are denoted white, the color of the actors’ nodes marks their cluster membership. Seven clusters that can be interpreted thematically could be detected: German, European, and Austrian politics, Crimes, Turkey, USA, and Deportation. In the following, the results of the analysis of these network clusters are presented.
German politics
The largest cluster of the network, colored in Figure 2 in purple, comprises prominent German politicians and the parties represented in the Bundestag during the period under study – except FDP which does not appear in the correlation network. At the center of the cluster are political organizations, especially parties, with which the persons in the cluster are connected according to their membership. The parties with the most links are AfD and SPD, followed by CDU, CSU, Gr¨unen, and Linke.
Three of the previously defined topics are part of the cluster: Populism, Right-wing, and Left-wing. Interestingly, the AfD shows a strong correlation to all these topics
Figure 2: Network of topics and actors.
(Populism: 0.21, Right-wing: 0.22, Left-wing: 0.27) and is even the only actor through which the themes Populism and Right-wing are connected to the network. It is hardly surprising that the two topics increasingly appeared in the discourse in conjunction with the AfD, as it is the most discussed and successful right-wing and populist party in Germany.
Who is also closely related to the AfD in the network, and even exclusively to it, is the Islamophobic, racist and right-wing populist organization Pegida, which has increasingly attracted attention in Germany, and even internationally, with its regular demonstrations since the end of 2014. Analyzing the co-occurrences of AfD and Pegida in the newspaper articles has revealed that they appear extensively in the direct co-text of each other as, for instance, “AfD, Pegida and Co.”1213 or “Pegida activists and AfD
members”14. In the majority of the discourse, Pegida and AfD are to some extent
12 All excerpts of newspaper articles have been translated from German into English by the researcher. 13 E.g. die tageszeitung, 2019-04-16, “Ein taz-Mitgr¨under, der die AfD verteidigt: Als ‘der Ulli’ rechts
abbog”.
equated as the right-wing movement in Germany. Parts of the AfD, however, try to distance themselves from the organization. Nikolaus Kramer, member of the AfD, for example, claimed in an interview in Die Welt that there is no cooperation with Pegida “because the work of the AfD takes place in parliaments, Pegida on the street. ‘We are a party, and Pegida is a movement,’ said Kramer – but did not distance himself from Pegida in terms of content.”15 The same article, which bears a title translatable as
“AfD and Pegida – Side by side far to the right”, portrays other AfD members as being obviously sympathetic to Pegida.
The proximity of the AfD to the topic Left-wing, on the other hand, seemed un-expected at first. The close reading of the articles in which this part of the cluster is predominant revealed that left-wing and right-wing positions are constantly juxtaposed and (re)negotiated in the discourse on immigration. Although the political orientations were typically presented as contrasts, the parties Die Linke and AfD were also discussed in a strong analogy. They were both considered the extreme parties of the political spectrum in Germany, often labeled populist – although this was much more frequently the case with the AfD. The media discourse presented the interests and, above all, the concerns of the voters of both parties far more similarly than would be expected from parties at politically opposite extremes. As a Focus article stated: “For 59 percent of supporters of Die Linke and 90 percent of AfD voters injustice prevails in our social order.”16 The figures are framed as a commonality between the two parties as they show that in both electorates social injustice seems to be an important issue for more than half of the voters. However, as the numbers are quite divergent, they could well be presented as a difference instead. Thus, co-occurrence does not necessarily mean substantive proximity, and qualitative textual analysis is necessary to grasp the specific meaning of this mutual appearance.
The question of social justice and the failure of Die Linke to address it adequately was mentioned in the media as the reason for the success of the AfD. Especially in the eastern states, where many people still feel socially disadvantaged in a nationwide comparison, the left party is replaced by the AfD which is promising voters more social justice with its nationalist strategy. Concerns about the refugee crisis, or rather the media representations of it, further support the success of the right-wing party. Die Zeit wrote in the spring of 2018: “Since many in the East believe that the refugees could
15 Die Welt, 2018-02-20, “AfD und Pegida — Seit an Seit ganz weit nach rechts”.
16 Focus, 2018-08-18,“#aufstehen: Warum vor allem die politischen R¨ander auf Wagenknechts
take away from them what they have laboriously built up, they are turning more to the AfD.”17 It was further stated in the article that Die Linke no longer functions as a protest party that rebels against the political elite because it has become part of that elite itself through its governmental participation. Thus, Die Linke is replaced because the AfD is the better protest party. It becomes apparent that left and right positions are blurring in the competition between the two parties, and that it is rather a question of who is seemingly more vigorously addressing the concerns of citizens which are similar in both electorates: “For some East Germans the AfD is the better Linke because it is not left. Because it focuses hard and distinctly on social issues, but in a way that is more appropriate for a society that tends to focus on isolation: The social for us first”18.
According to an article in Der Tagesspiegel, the sociologist Wilhelm Heitmeyer said that in Germany xenophobic and authoritarian attitudes as well as the willingness to demonstrate have increased considerably and become more radical already since 2009 after the financial crisis. He also stated that “decisive parts of the established politics did not care, because it had not affected the mandates until then”19. The dominance of
the AfD as a political actor in the discourse on immigration and the related appearance of the organization Pegida demonstrate that the right-wing movement with its roots in East Germany has entered the national political agenda:
Because of Pegida and the AfD, more than just a few East German publicists, politicians and intellectuals have now realized that the Germans have to talk about the East again – and must look at and understand the people there in a new and different way, if they are not to become AfD voters in even greater numbers.20
There is another issue at stake in the struggle between Die Linke and AfD, and between left and right, according to an article in die tageszeitung. Although the growing pluralization of German society and other nations that the German media are looking at has made it possible for many people to live their identity freely nowadays, for others it has led to a lack of identity: “Where I can enjoy the advantages of a free and plural society today, there is another part of society that has been stripped of its identity. My liberation is a threat to others. Therefore, a significant part of society wants to return to the narrative of the nation”21 Those who feel threatened are those who were previously
17 Die Zeit, 2018-05-06, “Ostdeutschland: Ist die Linke schuld am Aufstieg der AfD?”. 18 Ibid.
19 Der Tagesspiegel, 2018-10-19, “Soziologe Wilhelm Heitmeyer ‘Die AfD ist ein Meister der emotionalen
Ausbeutung’”.
20 Die Zeit, 2018-03-22, “Ostdeutschland: Und mittendrin die Ostdeutschen”.
privileged (e.g. “white, heterosexual men”). The AfD responds to this demand for nationalism by relying on “a new narrative of exclusion that denies a large part of the new society its belonging to the nation anew.”22 Die Linke, on the other hand, has not yet been able to offer an alternative to exclusionary nationalism that nevertheless meets the patriotic need. According to the author, this would be possible with an “inclusive patriotism”, in which people do not identify themselves with the nation, but with the basic democratic rights of their society. This could create a new common identity.
Another weakness of the left party, cited as one of the reasons why voters migrated from left to right to the AfD, is the moral traps into which leftist standpoints have repeatedly fallen in the course of the immigration debate. In an article in Junge Welt, the author Bernd Stegemann is quoted as speaking of a “left-wing universalist morality ... that demands open borders for everyone, [and thus] ignores the social consequences”23, such as short-term housing scarcity. Left-wing positions were generally strongly criticized in the discourse, even independently of the party Die Linke and the right-wing opponent AfD. They were repeatedly accused of inconsistency in their substantive principles and political demands concerning dealing with immigration. For example, according to an article in Die Welt, the left sees Islam as “a community of socially marginalized people worthy of protection because it is threatened by racists”, yet on the other hand “religion in general and Islam in particular stands for regression and barbarism”24. In the magazine Cicero, the left is also criticized for having chosen Islam “as the most worthy of protection of all identities of the others”25, although at the same time, leftists have made it their task to oppose religions.
Leftist positions were also attacked discursively with regards to the themes of sexism and sexual violence, which were repeatedly portrayed in the media as inherent to the culture of several immigrant groups, especially Muslims. In Cicero, for example, the left is accused of defending the disregard for gender equality among immigrants as an expression of their own culture. They would also argue that “the Universal Declaration of Human Rights need not apply to Muslims because they have a different cultural identity”26.
What is particularly interesting here is that conservative voices on immigration in
22 die tageszeitung, 2018-05-27, “Essay zu Nation, Werten und Grundgesetz: Generation Weltb¨urger”. 23 Junge Welt, 2019-01-24, “Politische Tabus”.
24 Die Welt, 2018-07-15, “Boykottaufruf nach Auftritt: Warum ein Islamhasser zu Gast bei Linken
ist”.
25 Cicero, 2019-08-15, “Identit¨atspolitik: Deutsch sein ist eh ganz bl¨od”. 26 Ibid.
Germany employ values typically associated with the left for their argumentation. They claim that those positioned further to the political right are better advocates of gender equality because they demand it for immigrants as well, who would have to leave the country if they failed to conform to Germany’s egalitarian values. The representatives of the right in Germany act as opponents of the “predominance of left-wing opinion”27
complained of by many from the conservative spectrum. The newspaper Junge Welt sums up the contradictions in the right-wing argumentative strategy: “AfD and Pegida claim in a particularist way to protect the existence of their own ‘people’ (Volk ). But when they see a headscarf or hear ‘Cologne’, they immediately think of universalist women’s rights, which they forget as soon as the word ‘gender’ appears.”28 As a fundamental contradiction the right is accused here of serving particularistic interests while at the same time using universalistic values for their argumentation.
European politics
The cluster on German politics is connected in the network with one surrounding European politics, marked in yellow in Figure 2. The link is established by the correlation between Angela Merkel and Emmanuel Macron (0.21), who both occupy a leading role in the EU. The main issues that are discussed regarding European politics are a common European asylum policy, promoted inter alia by Macron, the 2019 European Parliament election, and the influence of the Italian government on migration policy.
In the debate about common European migration policy, many political actors, e.g. French President Macron, and also the media themselves, were calling for a European solution. There were discussions about what should be regulated nationally and what at the European level, but overall the desire for more European solutions has become apparent in the media discourse on immigration. The term Europ¨aisch (European) increasingly appeared in newspaper articles as a demand to act, solve, think, live, and harmonize in a European way. Some media, such as die tageszeitung, for example, position themselves clearly for the implementation of a European asylum policy, even against those who are not (sufficiently) committed to it: “There is a chance for a European solution in refugee policy. If not everyone wants to participate, then some must lead the way. The slowest must not set the pace. With this principle – the
27 Der Tagesspiegel, 2018-03-27, “Kommt raus aus der Schweigespiralen-Schmollecke!”. 28 Junge Welt, 2019-01-24, “Politische Tabus”.
willing lead the way – we can solve problems more quickly, in a pragmatic, flexible and efficiency-enhancing way. That is also Emmanuel Macron’s approach.”29
Another subject that frequently appeared in the media discourse in the context of European politics is the 2019 European Parliament election, in which the leading issue was again a common European migration policy. As an article in Der Tagesspiegel has pointed out, this is a matter that will occupy the EU’s policy for a long time to come, and which it must tackle internally as a union and externally in cooperation with other continents, above all with Africa:
The question of migration is also a topic for the European election. But we should not just talk about the shameful odyssey of refugees between Malta, Italy and Spain that is currently affecting us. It is not only about the refugees of 2018, but also about the refugees of 2025 – that is, the migration movements that the next EU Parliament will have to deal with. We need a wise, collective EU Africa strategy.30
During the European election campaign, the dispute between the European People’s Party (EPP, in German EVP) and Hungary’s Prime Minister Viktor Orb´an was frequently discussed – surprisingly, however, the correlation of Orb´an with the other actors in European politics was still not strong enough for him to appear in the cluster. In a poster campaign, Orb´an defamed the then EU Commission President Jean-Claude Juncker together with billionaire George Soros as supporters of illegal migration, after which he and his party were suspended by the EPP. With the controversy, Orb´an also drew attention to George Soros: “Hungary’s Prime Minister Viktor Orb´an accused the Hungarian Jewish billionaire George Soros of a ‘plan’ to open the borders of Europe and ‘mix the populations’ in the interest of capital”.31 Soros, who advocates for an
open society and has donated large sums of his private fortune to individuals and organizations campaigning for freedom and democracy, has been declared an enemy of the conservatives. As an article in Der Spiegel quoted a US magazine, the radical right is using “his name as a single-word answer to the question of left-liberal supremacy”32.
In connection with the European asylum policy, the role of Italy was widely discussed as well. The Italian government, represented in the network by Matteo Salvini, Giuseppe Conte, and Luigi Di Maio, is portrayed as an influential actor in EU policy when it comes to migration. The right-wing populist, separatist and EU-sceptical coalition, in power
29 die tageszeitung, 2018-06-08, “Bundestagspr¨asident Wolfgang Sch¨auble: ‘Ich bin nicht die
Sprach-polizei’“.
30 Der Tagesspiegel, 2018-08-12, “Die EU vor der Europawahl: ‘Die Gefahr ist allen bewusst’”. 31 Die Welt, 2018-02-18, “Die unselige Traditionslinie des polnischen Antisemitismus”.
until August 2019, began closing Italian ports for boats carrying refugees and banning private sea rescue services in the summer of 2018. These measures were interpreted in the German media as a call for more European solidarity, for the equitable distribution of refugees among the member states, and for Italy not to be left on its own any longer. At the same time, Italy’s isolation strategy and generally eurosceptical attitude were emphasized as obstacles to common European immigration policy.
Austrian politics
A third cluster, which ties in thematically with European migration policy, is formed by the Austrian Federal Chancellor Sebastian Kurz, the Minister of the Interior Herbert Kickl and the parties ¨OVP, FP ¨O and SP ¨O, colored in Figure 2 in red. The media stressed Austria’s significant influence on European immigration policy. With the coalition of the conservative ¨OVP and the right-wing populist FP ¨O, which had been in government since 2017 until the “Ibiza affair” was revealed in May 2019, Austria has experienced a shift to the right. As with Italy, this has increased the pressure on European migration policy. The central issue of the government, publicly represented primarily by the then Minister of the Interior Kickl, was security. As Die Zeit, for example, wrote: “For the FP ¨O, security is primarily about one issue: migration.” The article went on to claim that security against (unwanted) immigration is “the core issue of the Austrian government, and it is the central campaign promise that brought the conservative ¨OVP together with the right-wing national FP ¨O to power.”33
Austria’s influence on Europe was strengthened by its EU Presidency in the second half of 2018, but also by the common line in migration policy with German Interior Minister Horst Seehofer and the then Italian Interior Minister Matteo Salvini. Austria’s Chancellor Sebastian Kurz calls the trio “axis of the willing”, which was to “ensure improved border protection – and, in case of doubt, also reject migrants at the respective internal borders of the EU countries if the external borders are not sufficiently secured”34 The shift to the right in Austrian politics was furthermore presented in the German newspapers as a concrete example of where other EU countries could also be heading, possibly oriented specifically towards the rise of Sebastian Kurz, as suggested, for instance, in an article of Die Zeit : “Because Kurz’ meteoric rise served as a blueprint for
33 Die Zeit, 2019-05-22, “Sebastian Kurz: Die Leere der Macht”.
34 Focus, 2018-06-15, “W¨ahrend die Union an der Asylpolitik zu zerbrechen droht, schafft Kurz
a whole generation of young European conservatives”35. His alliance with the FP ¨O at the far right also raised questions about how right-wing a government can be for democracy to still function, a consideration that the same article of Die Zeit labeled as “one of the great questions of European politics”.
As the media coverage of Austrian politics has shown, the actors would be expected to occupy a network position within or in proximity to the cluster of European migration policy. Instead, the cluster around Austrian politicians is linked to other groups in the network via the actor AFP. However, this must be ascribed to a weakness in the actor analysis: Since all persons detected by the Named Entity Recognition were identified as such even if only their surname appeared in the articles, Sebastian Kurz was identified also if only the word Kurz (German for ‘short’) appeared and this was disproportionately often the case in texts quoting the AFP news agency.
Crimes
Another cluster, which is marked in orange in Figure 2, is formed around the topic Crimes with links to the topic Events, to prominent actors of right-wing terror in Germany, and to the police. While the connection of crimes with the police is obvious, only the close reading of the articles revealed in which contexts the topic Events and the actors of right-wing terror appeared in the discourse.
The topic Events covers highly publicized crimes committed by (putative) immigrants – addressed via the city names of the crime scenes. Occurrences of the subject in the media discourse can be distinguished into three categories: Reports on specific crimes and the reactions of the public to them, the instrumentalization of such cases to create right-wing sentiment, and criticism directed at such instrumentalization.
In the media coverage of specific felonies and the reactions of the public to them, some cases were repeated time and again, such as the case of a group rape in Freiburg, in which the majority of the suspects were Syrians, and the murder of a young woman in Wiesbaden, which was confessed by a man who had fled from Iraq. Another such case is a murderous stabbing in Chemnitz. This city, together with Cologne, where on New Year’s Eve 2015/16 sexual assaults on women were committed by presumed migrants, was most frequently mentioned in the newspaper articles. Due to its high prominence,
the case of Chemnitz forms a small network cluster consisting of its three main actors. However, it is thematically classified here as belonging to the topic of Events.36
Regarding public responses to offenses perpetrated by (supposed) immigrants, Der Tagesspiegel, for instance, reported in the context of the case of Susanna F. in Wiesbaden about demonstrators who carried a banner with the inscription “‘Kandel is everywhere’. An allusion to another murder case, also committed by a refugee. And a hint that these demonstrators are not about Susanna F., not even really about her killer.”37 The
newspaper’s commentary alludes to the generalization and instrumentalization of such crimes for right-wing propaganda.
The instrumentalization of offenses allegedly perpetrated by immigrants was also carried out by the German newspapers themselves, most obviously by Cicero and Junge Freiheit. An article in Cicero claimed:
The excesses of New Year’s Eve in Cologne 2015/16, as well as rapes and murders of young women by migrants have changed public perception. Even the masters of ‘good attitude’ are now reluctantly forced to admit that their noble unprejudicedness was often only due to their ignorance, and that knowledge of the mentality and values of members of minorities can sometimes help to avoid danger.38
By linking offenses to “the mentality and values of members of minorities”, a propensity for crime, especially for sexual assaults against women, is portrayed as inherent to migrant men. At the same time, the magazine refers to the caving in by those who have so far resisted such generalization because even for them it is now apparent that the general suspicion towards certain immigrant groups is not based on prejudice but on “knowledge”.
The stigmatization of migrants as criminals permeated a series of articles in Cicero. The magazine claimed that migrants are disproportionately represented in serious violent crimes. As evidence, one article refers (without stating the source) to “a response by the state government to a small inquiry by the AfD parliamentary group”39. The authors cleverly created the impression that they have proof that migrants are more likely to commit crimes than non-migrants in Germany. The perception that this is simply the truth and bears no relation to right-wing propaganda was reinforced by writing
36 The actors in the Chemnitz case are all positively correlated on the topic Events, although the
correlation values are just below the threshold of 0.2, which is why they are not connected in the present network illustration in Figure 2.
37 Der Tagesspiegel, 2018-06-11, “Wie der Mord an Susanna F. eine Stadt spaltet”.
38 Cicero, 2018-06-22, “Auswirkungen der Willkommenskultur: ‘Unsere Frauen sind uns heilig’”. 39 Cicero, 2019-12-17, “Kriminalit¨atsstatistik: Migranten sind zehnmal ¨ofter mordverd¨achtig”.
about “the danger of generalizations and a general suspicion” and “an interest to protect minorities”:
It has always been a good custom that precise information on perpetrators of crimes be withheld. The interest of media users in knowledge is counterbalanced by the interest to protect minorities, in this case foreign minorities. Because there is indeed the danger of generalizations and a general suspicion that can strike the innocent, especially as resentments against foreigners are not an imagination, but unfortunately exist. This is not to be encouraged by media coverage without necessity.40
However, that there is a “necessity” to report on the proportion of migrants in criminal offenses was to be illustrated in the article with seemingly unequivocal figures revealing the share of migrants among suspects in murder and manslaughter, which was indicated at around 20 per cent. According to the magazine, the need for naming the nationality of perpetrators in media coverage arises from these statistics:
There is a pattern emerging. And to show a pattern, to make it recognizable, is not only a journalistic motivation. It is a journalistic duty. To make it tangible with an example. Manufacturers of crashed buses are actually not worth a mention per se. But if several bus accidents prove that it is always the same model from one manufacturer, then one MUST report about it. And so it MUST be in the case of serious violent crimes, if it turns out that these local events and individual cases show a pattern, in this case, the markedly disproportionate share of a certain group of perpetrators41.
Here, Cicero skilfully merged statistics on suspects that were official (because they were allegedly published by the German government) with the resentment towards certain groups of migrants that they tend to perpetrate violent crimes. However, the fact that a disproportionately high number of people who are identified as migrants are suspected of having committed these crimes is not necessarily proof of their greater potential for violence, but may also indicate racial profiling by the police: suspicions based on racialized or ethnicized characteristics such as skin color – and other phenotypical features – or presumed religious affiliation, often intersecting with gender, age, socio-economic
status and other dimensions of inequality.
Racial profiling was also addressed in a few of the newspaper articles. In an article published by Focus, J¨urgen Schlicher, anti-racism expert and trainer in diversity management, was quoted on this issue:
‘When police officers increasingly control migrants because they have the prejudice that this group is more criminal than another, there are consequences.’ On the one hand, the
officers would detect more violations of the law, more checks would also result in more hits, Schlicher explains. On the other hand, the migrants might get angry because they notice that they are controlled more often than other people. They then in turn develop resentment against the police. This in turn encourages them to control more migrants. ‘This is a cycle, a self-fulfilling prophecy’42.
The instrumentalization of crimes in which (alleged) migrants are suspects or actual perpetrators to create right-wing sentiment was also clearly named and thereby criticized in some articles. Quoting a sociologist, Der Tagesspiegel wrote: “Crimes committed by Muslims – such as the murder of 15-year-old Mia in Kandel – would be used to create a sentiment against minorities.” Another article discussed the increasingly diverging perspectives on the issue of immigration and crime, which lead to the fact that the “subject is in danger of being buried under more and more unspeakable things”43.
Besides the discussion about highly publicized crimes perpetrated by (supposed) immigrants, grouped under the topic Events, the network cluster around the topic Crimes consists mainly of actors of right-wing terror in Germany. In the newspaper articles, the murders committed by the NSU terror organization were repeatedly mentioned as the prime example of right-wing terror in Germany. The extremist underground network, which has been committing racist murders for many years, was uncovered in 2011. Its connections to the police and the German secret service uncovered the institutional racism deeply embedded in the German society.
More recent articles focused on the murder of Walter L¨ubcke in the context of right-wing extremist violence. The politician, who had been campaigning against right-right-wing extremism for years, was murdered by the radical right-winger Stephan E. on June 1, 2019. Other racially motivated attacks were then almost exclusively addressed in the immigration discourse in association with the L¨ubcke case. The reporting media thereby conveyed concern about the increasing tendency of right-wing extremism in Germany to commit serious violent crimes. Reporting on the attempted murder of an Eritrean in Frankfurt, an article in Frankfurter Allgemeine Zeitung stated: “The hatred of migrants by right-wing populists is now apparently turning into assassination attempts in broad daylight.”44
The two most prominent issues surrounding the topic Crimes in the immigration discourse were thus offenses committed by (alleged) migrants on the one hand, and
42 Focus, 2019-03-21, “Rassismus: Experte erkl¨art, wie sie Vorurteile erkennen und bek¨ampfen”. 43 Der Tagesspiegel, 2018-06-07, “Wiesbaden: Der Mord an Susanna ist Gift f¨ur die Gesellschaft”. 44 Frankfurter Allgemeine Zeitung, 2019-07-23, “Nach Sch¨ussen: W¨achtersbach ruft zu Mahnwache
