The criminal is always the foreigner?!
An analysis of minority signification in German crime reporting
By Azade Esther Kakavand Student ID: 12846996
Master’s Thesis
Graduate School of Communication Journalism, Media and Globalisation
Supervisor: dr. Damian Trilling Date: May 29, 2020
2
Abstract
Specifying alleged criminals’ affiliations to minority groups – so-called minority
signification – can lead to stereotyping and prejudices. Nevertheless, in German crime reporting, ethnic affiliations are mentioned regularly. While minority signification has been studied after key events and between different outlets, no research has yet tackled the question of how minority signification changed over a longer time, including through different key events and across outlets. In a partly automated content analysis, I explore how the accentuation of minority signification changed over the years from 2014 to 2019. The study focuses on three aspects: Firstly, I set nationality mentions in reporting in context to actual crime numbers. I find that non-European nationalities are overrepresented while non-Europeans are underrepresented in crime news. Secondly, I examine whether key events can explain time periods with an especially high number of mentions. While the sexual assaults on New Year’s Eve 2015/16 had an influence on minority signification in the following time, the other sparks could not be linked to key events. Thirdly, I study how the usage of minority signification differs between three German newspapers. While the political-right newspaper mentions minority affiliations most, the tabloid is second and the political-left outlet mentions them least. Surprisingly, this pattern changed over the years.
3
Introduction
Sayings like “stolen today, in Poland tomorrow”1 reflect common stereotypes against foreigners in Germany. The picture of criminal foreigners is deep-seated in the head of many Germans. Migrants are evaluated as a risk and undesired neighbors who are associated with incivility2 and criminality (Schartau et al., 2018). Meanwhile, the contact with minorities is limited. Thus, the media plays an important role in creating an indirect experience with these out-groups (Bonfadelli, 2007). Generally, migrants are reported on less in frequency and more negative in tone than natives (Benson & Wood, 2015). Often, the coverage contains frames of criminality, crime, financial burden, and foreign infiltration (Tort, Guenther, & Ruhrmann, 2016). Such a biased representation can lead to prejudices in society (Bonfadelli & Moser, 2007;
Czymara & Schmidt-Catran, 2017; Dixon, 2008). To prevent this, the German press code
includes a non-discrimination rule accompanied by a guideline that limits the mention of alleged criminals’ affiliations to a minority group in crime reporting (German Press Council, 2017). Nevertheless, nationalities of criminals get mentioned often, inter alia because this is an easy way to provide information in fewer words (Hefner, Klimmt, & Daschmann, 2007).
So far, the existing research has mainly focused on the influence of key events on
accentuating minority affiliations in reporting (Arendt, Brosius, & Hauck, 2017; Tort et al., 2016; Yamamoto, 2012) as well as on the framing of specific minority groups (Alsultany, 2013; Dixon & Linz, 2000; Trilling, Kroon, & Raats, 2020). If and how the portrayal of criminals as part of minority groups in Germany changed over a longer timeframe is not known yet. Meanwhile, contemporary events, such as the refugee crisis, a rise in right-wing populism, and several terror
1 Original: „Kaum gestohlen, schon in Polen“.
4
attacks in Europe, may have changed the reporting on immigrants. My study fills this gap by analyzing minority signification – that is the mentions of alleged criminals’ ethnic affiliations – in German newspapers over the course of six years. The research question is: how did minority
signification change in German crime reporting from 2014 to 2019?
The research has high societal relevance. In various experiments and surveys, it was found that a negative image of minority groups such as Blacks (Dixon, 2008) Muslims (Czymara & Schmidt-Catran, 2017) or Kurds (Scheufele, 2002) influences the sentiment towards these groups in society. As every fourth person in Germany has foreign family ties, this affects large parts of society. Further, the research is academically relevant as it examines the change in reporting over several years and thus does not only fill a gap in academia but also provides a starting point for more detailed research, e.g. to follow up on Boomgaarden and Vliegenthart (2009) who showed that news coverage of ethnic minorities affects people’s attitude towards immigration in Germany from 1993 to 2005.
To answer the research question, I focus on three subtopics. Using exploratory research designs, I observe whether different nationalities get misrepresented in crime reporting in comparison to the actual criminality of the group. Further, I examine if and how key events can explain temporary changes in minority signification. Lastly, I assess how the use of minority signification in reporting differs between the researched newspapers.
The remainder of this paper is structured as follows: Firstly, I give an overview of related research and the theoretical background and introduce sub research questions and hypotheses. Secondly, I describe and discuss the methodological proceedings. Thirdly, I present the findings, and fourthly, I discuss these as well as general implications, limitations, and suggestions for future research.
5
Theoretical background and related research
Minority groups in society and media – the concept of minority signification
Pointing out minority affiliations can be seen as part of so-called othering, a concept that was first introduced in philosophy and became the subject matter of various disciplines such as the securitization theory (Buzan, Wæver, & Wilde, 1998), and post-colonialist studies (Said, 1987). A negative picture is painted of the others who differ from the norm, may that be upon race, religion, origin, and culture, or other factors. The resulting distinctions and stereotypes are introduced, spread, and reinforced through the media (Blumer, 1958). Therefore, in
communication science, several researchers engaged with the differentiation between groups in the media building on similar concepts but naming them differently.
In the USA, scholars especially studied the portrayal of Blacks and Latinos in the media landscape (Dixon & Linz, 2000; Entman & Rojecki, 2001). Other contexts are the portrayal of Muslims in the US (Alsultany, 2013) as well as in Germany (Hafez, 1999) and the association of foreign nationalities with negative attributes in Dutch reporting (Trilling et al., 2020) as well as the comparison between portrayal of foreigners and Flemish suspects in TV reporting (Jacobs, 2016).
Researchers such as Arendt et al. (2017) investigate the mention of foreigner-related attributes not giving the concept a name. While only concentrating on Chinese foreigners in her analysis, Yamamoto (2012) calls the concept minority signification that is defined as a process, “where the media amplifies the crime and criminality of a minority group” (p. 156). She broadly defines a minority group as a group that is associated with social stigma and limited political power. In her concept, she distinguishes between minority crime as the disproportionally frequent mention of affiliations, and minority criminality, as the use of affiliation as an explanation for the
6
alleged criminal’s behavior. I follow Yamamoto (2012) in calling the concept minority signification but focus here on minority crime.
Minority signification depending on ethnic affiliations
Crime reporting is popular because in general, events that are surprising, differing from the norm, or entail conflict, threat, and negativity are more newsworthy than others (Galtung & Ruge, 1965). Yamamoto (2012) argues that events that serve stereotypes about certain minority groups might be reported on because the news becomes more culturally meaningful. Especially if minority groups become associated with problematic pictures such as backwardness, criminality, and terror repeatedly, as it was shown for Kurds and the PKK (Esser, 2002; Scheufele & Brosius, 2001) and Islamic minorities (Hafez, 1999), the connection to a minority group then can be used as an explanation or short-cut to the behavior.
Czymara and Schmidt-Catran (2017) criticize former research for generalizing immigrants as one group and argue that they are not perceived equally but depending on the group they belong to. They find that immigrants from the Middle East and Africa are less accepted than immigrants from western Europe showing differences between culturally close and distant immigrants (see also Esser, 2002). Thiele (2015) points out that national stereotypes often overlap with ethnic, religious, or cultural stereotypes. Tort et al. (2016) find that in German reporting immigrants are often either portrayed as a homogeneous group or categorized by nationality. E.g. they find that East Europeans are framed as criminals more often than other groups in line with common stereotypes.
In an inter-reality comparison, Dixon and Linz (2000) find that Blacks are overrepresented as criminals in Californian crime reporting compared to actual crime numbers. Arendt et al. (2017) research how the incidents on New Year’s Eve 2015/16 (NYE) when a mob of young, North-African or Arab men allegedly sexually assaulted women in Cologne and other German
7
cities influenced reporting (Arendt, Brosius, & Hauck, 2017; Dürr, Märkl, Schiavone, & Verhovnik, 2016). While they do not have data to conduct inter-reality comparisons, they conclude from police status reports that the rise in minority signification is not explicable by crime statistics.
While these studies either focus on one event or limited minority groups, I use a longer timeframe and investigate the reference of several groups. By doing so, I examine whether or not the portrayal of foreigners as criminals in relation to the actual crime numbers differs between the nationalities. Sub-research question one reads: how does the representation of minority groups as
criminals differ from their actual committal of crimes? (Sub-RQ 1)
The review of existing literature shows that researchers find different groups to be affected by minority signification. This may be influenced by variances in the research design as well as the sampling period. Current events and long-term changes in societal composition may affect the general perception of a minority group. A change in reporting over a longer period has not been researched yet. To shed light on this, sub-research question two is: how does the portrayal of
alleged criminals with foreigner-related attributes change over time? (Sub-RQ 2)
Influence of key events on reporting
Blumer (1958) theorizes that big events can shape or reshape the perception of minority groups. But short-term changes in the attitude towards minorities might also be influenced by events. Several studies use the concept of key events to investigate changes in media
performance. The research supposes that extraordinary extensive reporting on a certain event alters the selection criteria for reporting temporarily. Selection criteria that are normally formed through news values (Harcup & O'Neill, 2001), as well as influences on the organizational or personal level (Reese & Shoemaker, 2016) then temporarily lose meaning (Arendt et al., 2017).
8
Key events can either be dramatic, unexpected or spectacular as such, or become key events by sensational reporting (Kepplinger & Habermeier, 1995).
In recent years, several incidents might have influenced how minority groups are portrayed in the media: In 2014, the Alternative for Germany (AfD)3 gained support as a new right-wing party moving in state parliaments for the first time and the Patriotic Europeans Against the Islamisation of the Occident (Pegida)4 formed and gained massive support by concerned citizens5 (Tort, Guenther, & Ruhrmann, 2016). To analyze the framing of immigrants in 2014, Tort et al. (2016) recognize the rising immigration numbers starting from 2009 onwards as the key event even though the stimulus does not represent a certain date. Arendt et al. (2017) find a significant rise in mentions of foreigner-related attributes after NYE. In the following years, other events could have had the same effect: In 2017, the press council eased its guideline for mentioning affiliations of alleged criminals in crime reporting (Presserat, 2017). Further, several terror attacks were carried out by Islamists all over Europe. The first big attack on German ground took place on December 19, 2016, when a fighter of the Islamic State (IS) drove a truck into a
Christmas market on the Breitscheidplatz in Berlin and killed 12 people. Also, crimes like the murder of young women or girls by refugees could affect since they generated national attention. It is not clear which of these events influenced the reporting. Consequently, the third sub-research question is: can months with high minority signification be linked to certain key events? (sub-RQ
3)
3Original: Alternative für Deutschland (AfD)
4Original: Patriotische Europäer gegen die Islamisierung des Abendlandes (Pegida)
9
The influence of media types
Different types of media can follow different news values and influences on the
organizational level (Reese & Shoemaker, 2016). For France, Norway and the US, Benson and Wood (2015) find that political-right news outlets portray immigrants more critically than political-left outlets. For Germany, Tort et al. (2016) examine newspapers on the political left-right spectrum and find that political-left outlets portray immigrants more positively than political-right outlets who also use the notion of threat scenarios more often. While former research focuses on various frames, different use of crime frames for different minorities was not found (Tort et al., 2016). I focus on this genre of reporting. Subsequently, my first hypothesis is:
political-right- newspapers include minority signification more often than political-left newspapers (H1).
Additionally, tabloids are generally known for reporting in a more lurid way and
simplifying connections, fueling stereotypes, and sensationalism (Schicha, 2008). While Arendt et al. (2017) examine a regional newspaper, a national quality newspaper, and a national tabloid, their sample is too small to find significant differences between the different outlets. Schlier and Lincoln (2014) research the portrayal of people with schizophrenia in German print news. They find that schizophrenic people are more stigmatized in tabloids than in broadsheet papers. I expect that their findings can be translated to this research that also deals with sometimes stigmatized minorities. The resulting hypothesis is: Tabloids are more likely to include minority
10
Methods
The sample
I collected the data using the Dow Jones Factiva database. To avoid biases due to the geographical focus or the publishing rhythm of the outlets, I chose daily national newspapers only. The selection of outlets is restricted to three: one conservative, one liberal, and one tabloid6. Among the outlets with high national importance, tageszeitung (taz) and Die Welt represent a political-left and a political-right outlet, respectively. Additionally, BILD represents not only the most sold newspaper but also the largest tabloid in Germany. A timeframe from January 1, 2014, to December 31, 2019, was chosen.
All selected articles are indexed in the subject “Crime/Legal Action” in the database and contain a synonym to criminal or other words related to committing a crime such as arsonist7. The region is set to Germany and the language to German. All articles were downloaded between March 17 and March 20, 2020, in RTF files containing maximally 100 articles. Further, the option to display “Full Article/Report plus Indexing” was chosen to obtain additional information and facilitate data analysis. In total, the search result included 15,452 articles. After the removal
6 The selection was done between “BILD” (approx. 1,357,000sold newspapers), “Süddeutsche Zeitung”
(approx. 327,000), “Frankfurter Allgemeine Zeitung” (approx. 220,000), “Die Welt“ (approx. 86,000), and “tageszeitung” (taz) (approx. 49,000) (Weidenbach, 2020). Despite its run of approx. 136,000 newspapers “Handelsblatt” was excluded due to its focus on economy and finance reporting.
7 Unfortunately, the Factiva indexing does not distinguish between criminality and lawsuits also including
articles about tenancy law or jurisdiction about tourism that are irrelevant for this research. To obtain an adequate sample, a test coding was conducted on 100 articles to distinguish between relevant and irrelevant articles. Consequently, a list of words related to the perpetrator was obtained using the words in the sample as well as synonyms. This list was then used to refine the search results.
11
of duplicates with the relevant Factiva setting on “similar” the sample consists of 15,044 news pieces.
Operationalization
Independent variables
The independent variables of my research are provided in the information I retrieved via the database. I use the publication date (sub-RQ1-3) and the outlet (H1-2) as independent variables. Due to big differences between publication dates, these are grouped to years (sub-RQ1-2, H1-2), calendar months (sub-RQ3), and 30 (or 31) days before and after an event (sub-RQ3).
Dependent variables
To measure minority signification in news reporting, I screen news texts for different affiliations. Yamamoto (2012) focuses on the mentions of Chinese citizens or international students. In German research, Tort et al. (2016) examine immigrants in general but exclude the following generations of immigrants. Meanwhile, Arendt et al. (2017) limit their research to the attributes of North African, migration background, foreigner, and asylum seeker. In my research, I use the operationalization by Arendt et al. (2017) as a guideline but modify the categories to assess crime reporting in on a broader spectrum that is independent of the NYE incidents.
Next to general terms like “foreigner”, “migrant”, “asylum seeker”, and “migration background”, I examine different nationalities. Analyzing German crime statistics from 2014 to 2019, I chose the 20 foreign nationalities with the most criminality for each year, resulting in a list of 26 nationalities (Bundeskriminalamt, 2014, 2015, 2016, 2017, 2018, 2019) (Figure 7). Further, I added regional attributes such as East European or North African into the codebook because it is possible that the media generalizes upon regional attributes especially if the act is portrayed in a line of crimes by criminals from the same region or serves a stereotype based on other characteristics such as culture (Arendt et al., 2017). Contrarily to Arendt et al. (2017), I
12
only counted the mention of Eastern Europeans, migration background, etc. if it was specifically mentioned. Accumulative regional variables were later created. Double nationalities were counted as the foreign one because these get associated with a migration background, foreign culture, and socialization. Refugees were counted if it was clear from the text e.g. when the alleged criminal lives in a refugee home.
Further, ethnic and religious attributes were coded. Because Tort et al. (2016) find that people with Islamic beliefs face higher resentments and are more often associated with
criminality, I include Islam in my variables. Also, Sinti and Romanies are often associated with criminality (Zentrum für Antisemitismusforschung & Institut für Vorurteils- und
Konfliktforschung e.V., 2014)8. Further, Kurds were included in the analysis because they represent a politized group of appr. 500,000 to 1 million people in Germany (Schürer, 2018). Esser (2002) finds Kurds often used as a synonym for the terror organization PKK. Thus, the dependent variables include nationalities, regional descriptions, Islam and Sinti and Romanies (a full list of examined ethnic minorities can be found in the appendix).
Research design
My approach is twofold. In the first step, I use automated content analysis (ACA) for a pre-selection of articles that might contain minority signification. In the second step, the identified articles were coded manually. This method allows me to benefit from the advantages of ACA but avoid pitfalls.
8 In a survey by the national anti-discrimination body, almost 50 percent of the respondents stated that Sinti and Romanies’ actions cause animosities. These strong and deep-seated prejudices might also be found in crime reporting.
13
ACA has advantages especially considering time and money restraints in large-scale analyses because no person has to manually code the content. On the downside, the complex construct of language where one word or term can have different meanings depending on the context is better understandable for human coders (Boumans & Trilling, 2015). Nevertheless, with an extensive codebook, the reliability of ACA can be high. In this research, codewords contain synonyms of criminal and ethnic attributes. These normally are no homonyms9, thus no confusion about meanings can occur here.
To establish validity, it has to be ensured that the keywords are used in connection to a description of the alleged criminal, not the victim, other protagonists, or circumstances. In the easiest case, the affiliation is directly connected to the word “criminal” or any synonym. But often, articles are written in a more complex way not giving this direct cue. Therefore, I code all articles by hand that are detected by the program.
The data was analyzed in Python using a combination of RegularExpression (RegEx) and SpaCy, a natural language processing tool (ExplosionAI GmbH). RegEx is a common text processing language that enables coders to detect patterns in text (López & Romero, 2014). I use RegEx for multiple steps. To clean and prepare the data, the raw text is split at the end of each article. I use the indexing to detect sequences such as the headline, main text, outlet, publication date among other information such as the subject code10.
Then keywords are searched for in the articles. These are based on stemming and full word recognition. The full word recognition includes words such as “migration background”. The
9 In the course of coding, I found that the stemming of nationalities led to other words being detected as well.
E.g. the stem of Serbia in German is “Serb”. This string occurs in “Serbe” (Serb) or “serbisch” (Serbian) but also in “Leserbrief” (letter to the editor). Such wrong detections were filtered out in manual coding.
14
stemming is used so the ACA recognizes e.g. “Turkish”, “Turk”, and “Turkey” all as related to Turks. With this scheme, all articles that contain one of the researched keywords for ethnic groups are selected.
The detected texts are then further used for operations with SpaCy. SpaCy provides functions that detect semantic connections between subordinate and superordinate words. The token.subtree and token.ancestors functions are combined to search for connections between a codebook containing words related to criminals or criminal actions and a second codebook containing foreigner-related attributes (Figure 6). 4,391 articles are found to include a semantic connection between foreigner attributes and criminal attributes. Comparing the outcome to a hand-coded test-sample of 50 articles, a recall of .84 and a precision rate of .57 is obtained. Thus, 84% of all articles that contain foreigner attributes as descriptions of the criminal are detected by the ACA. But only 57% of the detected articles are actually relevant. These are filtered out in the second step of the coding: the manual coding.
In manual coding, I analyze all paragraphs that mention one of the nationalities or ethnic attributes of interest. If the paragraph is not clear in its meaning, the whole article is read. Articles are considered as relevant if the crime was conducted in Germany or legal action took place in Germany. War crimes are coded as irrelevant11. Further, letters to the editor, articles about crime statistics, and interviews12 are not deemed relevant as well as articles about crimes that were
11 War crimes are often complicated cases that are on trial on basis of the Code of Crimes against
International Law which is derived from the Rome Statute of the International Criminal Court. Arendt et al. (2017) do not consider these cases either.
15
conducted before 200013. The coding unit is an article. If several crimes are mentioned in an article, all of those that were conducted by foreigners are coded. Thus, if an article refers back to other crimes by the same subject or subject group or to a similar crime by another subject group these are also coded. In a few cases14 a text or large parts of a text occurred twice in the sample. These are only coded once.
61 variables are coded. These include 26 nationalities and 13 regional attributes as well as 16 categories for the type of crime and variables that are not included in the analysis such as foreign reports, repeating, and ambiguous articles. All variables are dummies coded to 1 if the minority group or crime occurred and 0 if not. The type of crime is only coded if the criminal was a foreigner. Additionally, a variable for foreign reports, or other topics (e.g. broadcasting of a crime show), a variable for articles that occurred twice, and a variable for articles that were ambiguous were included. Further, two open variables are coded. Firstly, a variable for potential key events is filled if a crime happened to be described very often or sparked special attention (see appendix). Secondly, a variable for other nationalities is included. While these are not analyzed in the first instance, they happened to occur in combination e.g. with the reporting on refugees.
The coding was conducted by two coders. I analyzed 3,703 articles while the second coder analyzed 795 articles. If one of the coders was unsure about the coding of an article, the issue was discussed. To test intercoder-reliability, Krippendorff’s Alpha was calculated on 20% of the articles coded by the second coder (n=160). Alpha varied between 0.5 and 1.0 (Table 2). All crime type variables where Alpha was lower than 0.65 were excluded from the analysis. Due to
13 In this exclusion fell e.g. reports about the German terror organization Rote Armee Fraktion (RAF) that
cooperated with foreign terror organizations as well as crimes in the third Reich. 14 BILD: 23, taz: 1, Die Welt: 6.
16
the overall high number of variables, not every variable was present in the subsample and thus could be evaluated with Alpha. Nevertheless, the values for nationalities are important for the analysis because sub-RQ1 and sub-RQ2 directly build on the distinction between different
groups. Further, because the coding effort for all minorities is the same and Krippendorff’s Alpha values for the minority variables in the subsample lie between 0.74 and 1.0 (mean=0.94), I expect to have a similar quality of evaluation for the other minorities. Therefore, the variables are
accepted even though not all could be assessed directly.
Results
Over- and underrepresentation of criminal foreigners
In total, 1619 (out of 15044 = 10.76%) articles use at least one minority signification. To determine if some groups are overrepresented and others are underrepresented, I subtract the percentage of mentions for each of the 26 nationalities in my sample to the percentage of crimes in the crime statistics15. The values were normalized to represent the number of reported crimes of one group as the percentage of the reported crimes by all 26 observed groups (and the
percentage of actual crimes in relation to all these groups respectively). Figure 1 shows the difference between these values.
15 While I coded also for affiliations with cultural or religious groups as well as regional attributes and
societal standing, these are not represented in the crime statistics by the government. Thus, data to compare against could only be found for the nationalities. Also, articles that were coded as containing key events were excluded because they might confound the analysis.
17
Note. Difference between mentions in reporting in % of mentions of all 26 nationalities in reporting and crimes in crime statistics in % of crimes by all 26 nationalities. Key events were excluded. Nationalities with an average difference of <1% were excluded from the figure (Eritrea, France, Georgia, Iran, Kosovo, Macedonia, Nigeria, Pakistan, Somalia and Ukraine)
The figure shows that nationals from North Africa and the Middle East are generally rather overrepresented (Afghanistan: M=5.11%, Algeria: M=1.34%, Iraq: M=3.61%, Morocco:
M=3.34%, Syria: M=7.88%) but also Russia (M=4.83%) and Albania (M=1.34%). Meanwhile Europeans including Eastern Europe are rather underrepresented (Poland: M=-4.04%, Rumania: M=-7.38%, Serbia: M=-1.24%, Bulgaria: M=-2.45%, Bosnia and Herzegovina: M=-1.01%,
-15.00% -10.00% -5.00% 0.00% 5.00% 10.00% 15.00% 20.00% 2 0 1 4 2 0 1 5 2 0 1 6 2 0 1 7 2 0 1 8 2 0 1 9 F ig u r e 1 O ve r- a nd u nd e r r e p r e s e nt a t i on o f c r im i na ls i n g e r m a n c r im e ne w s b y na t iona l it y
Afghanistan Albania Algeria Bosnia Bulgaria Greece
Iraq Italy Croatia Morocco Poland Romania
18
Croatia: M=-1.00%, Italy: M=-3.02% and Greece: M=-1.80%)16. Concluding, sub-RQ1 can be answered: Criminal affiliations are not represented in relation to the actual committal of crimes by nationals but over- or underrepresented depending on the regional origin. Generally speaking, Europeans are underrepresented while non-Europeans are overrepresented.
These misrepresentations are not stable but change over the course of the years (sub-RQ2). While Syrians are always overrepresented, mentions rise more than the crime statistics would have suggested especially in the years from 2016 to 2018. Also, the overrepresentation of Moroccans and Algerians in crime reporting increases in 2015. Meanwhile, the data for Turkish criminals in crime news show a sharp rise in 2019, going from underrepresentation to
overrepresentation.
Influence of key events on reporting
To answer sub-RQ 3, I first displayed the ratio of articles that contain minority signification in contrast to total articles per month (see figure 5). The months with the highest ratios are
December 2016 (23.7%), August 2018 (21.8%), January 2016 (21.6%), and March 2018 (21.3%). I analyze these months exemplarily for the influence of key events on crime reporting. After finding a possible key event in the month or at the end of the previous month, I follow Arendt et al. (2017), in examining if the ratio of articles containing minority signification is higher in the first and second month after this key event in comparison to the month before the incident.
In January 2016, the reporting about sexual assaults against women on New Year’s Eve (NYE) in several German cities might have caused a rise of minority signification. The alleged criminals are generally described as North African or Arabic looking. In a timeframe from one
16 Eritrea, France, Georgia, Iran, Kosovo, Macedonia, Nigeria, Pakistan, Somalia and Ukraine are excluded
19
month before the incidents until two months after the incidents, in total 764 articles were found. Out of these, 126 were coded as containing minority signification and 638 were coded as not containing minority signification respectively. 27 articles were coded as containing the reporting about the assaults.
I examine if minority signification rises significantly after the key event NYE. All models base on the month before the key event and the first and second month after the key event as independent variables. The first model including all minority signification as the dependent variable is not significant (LLR p-value=0.079) and therefore will not be considered further. The second model includes minority signification of alleged criminals from the Middle East and North Africa (MENA) as the dependent variable. The minority signification here rises from December to January and from December to February. But the model does not fit the data well (McFadden Pseudo-R2 = 0.019). Thus, the key event can hardly explain any variance.
Separating the Middle East and North Africa further shows that the attributes for North Africa are the driving force here. North Africa (McFadden Pseudo-R2 =0.059) performs better than the Middle East (McFadden Pseudo-R2 =0.012) in explaining the variance in the data. Also, the odds ratio is much higher for North Africa than for the Middle East. The last model including the variable refugee/ asylum seeker as the dependent variable is not significant (LLR
p-value=0.586) and will thus not be considered further. Logistic regression models with several nationalities that can be associated with the suspect group as the dependent variable were either not significant as a model (LLR P-value > 0.05) or did not show significant changes (p>0.05). The results suggest that NYE has influenced the minority signification of alleged criminals from the MENA region and especially from North Africa.
20
Note. Minority signification in % of all articles per month, error indicator: standard error
0.00% 5.00% 10.00% 15.00% 20.00% 25.00% 30.00%
Figure 2
21
In December 2016, an Islamist terrorist deliberately drove a truck into a Christmas market on the Breitscheidplatz in Berlin. From one month before the attack until two months after the attack, 766 articles were published out of which 185 contained minority signification. 42 of these reported on the key event and are excluded. To test whether the attack could have been a key event that influenced reporting, I conduct logistic regression in the same manner as above (Table 34-48). Only the model with asylum seekers and refugees as the dependent variable is significant (LLR p-value=0.009). The minority signification for this group drops significantly in the second month after the event (Table 43). But the odds ratio is low. Further, logistic regression models for some nationality are significant. The mention of Afghans drops significantly after the attack compared to the month before the attack. However, the odds ratio is relatively low, indicating only a small effect. Meanwhile, the mentions of Syrians rise significantly in the first month after the attack, with a higher odds ratio.
Furthermore, two sparks in March and August 2018 are observed. By examining all relevant articles for these months, I found that two murder cases were heard in March 2018: A refugee raped and killed a student in Freiburg in October 2016 and another refugee stabbed several people in a supermarket in Hamburg in July 2017. Excluding all reporting about these events, the logistic regressions model for minority signification of asylum seekers and refugees is significant (LLR p-value = 0.043) but does not show significant changes in March (p = 0.094) or April (p = 0.559, Table 57). The regression models for foreigners, MENA, North Africa, and the Middle East were insignificant (Table 49-57).
In August 2018, an alleged rape of a girl in Hamburg by a refugee was reported on. Later in the same month, it turned out to be false. Excluding the reporting about this incident, the model including minority signification of the Middle East is significant (LLR p-value = 0.044). The mentions rise significantly in September compared to July (p = 0.022) but not in August (p = 0.509, Table 58-67).
22
As an answer to sub-RQ 3 it can be noted that while some sparks in the data might be explained by a single event as the sexual assaults on NYE, other sparks are ambiguous.
The influence of media types
Hypotheses 1 and 2 suggest that while the reporting of the examined newspapers is based on the same events and exclusively reporting about crimes in Germany was taken into account, the outlets differ in using minority signification. To answer H1 I notice that the political-right newspaper Die Welt mentions foreigner attributes of alleged criminals the most (12.4%), while the tabloid BILD mentions them less (10.3%) and the political-left newspaper taz mentions them the least (9.1%, see figure 3). H1 can only be confirmed for the years 2015, 2016, 2018, and 2019: the political-right newspaper Die Welt mentions foreigner attributes significantly more than the political-left newspaper taz in these years. In 2014 and 2017, differences are
insignificant.
The logistic regression model to test if tabloids are more likely to use minority signification (H2), is significant (LLR p-value < 0.001, Table 69). While Die Welt mentions foreigner
attributes significantly more than BILD (p < 0.001), the difference between taz and BILD is not significant (p = 0.143). As shown in figure 4, the data differs tremendously between the years. While Die Welt always uses minority signification most (though not always significantly), BILD even does so significantly less than taz in 2014 (p = 0.002, table 70). Only in 2018 does BILD use minority signification significantly more (p = 0.002, table 74). Therefore, H2 can only be partially supported.
23
Note. Percentages in comparison to all examined articles in that timeframe and outlet.
Note. Percentages in comparison to all examined articles in that timeframe and outlet.
0.00% 2.00% 4.00% 6.00% 8.00% 10.00% 12.00% 14.00%
SN|BILD SN|taz - die tageszeitung SN|Die Welt Figure 3
Ratio of relevant article per outlet
0.00% 2.00% 4.00% 6.00% 8.00% 10.00% 12.00% 14.00% 16.00% 18.00% 2014 2015 2016 2017 2018 2019
Figure 4
Distribution of minority signification per outlet and year
24
Conclusion and Discussion
Over- and underrepresentation of criminal foreigners
To answer sub-RQ one, I examined how the representation of minority groups as criminals differs from their actual committal of crimes. Further, conducting inter-reality comparison shows that the portrayal of foreigners as criminals in print media is far from representing actual crime numbers. Comparing the rate of minority signification in the sample to the actual crime statistics, I find that Europeans are underrepresented, and non-Europeans are overrepresented in the media. These nationalities mostly also represent Muslim countries. The overrepresentation of countries with Muslim majorities in crime reporting might also be a reason for this suspicion found by Czymara and Schmidt-Catran (2017) and supports Alikhani and Rommel (2017) who debate deep-seated racism against Muslims.
Sub-RQ2 asked how this representation changes from 2014 to 2019. Over the course of the years, the overrepresentation rises for some non-European nationalities. Reasons here can be that due to the refugee crisis, reporting about immigrants and thus also criminal foreigners rises. Bonfadelli (2007) points out that an increase in refugee numbers can lead to higher reporting on migrants from war-affected countries. My results also provide further explanation for the findings of Czymara and Schmidt-Catran (2017) who discover migrants from the Middle East and Africa are being accepted the least and detect a decrease of acceptance from April 2015 to January 2016. I cannot provide an explanation for the development of minority signification of Turks.
Influence of key events on reporting
To answer sub-RQ3, I examined if months with high minority signification can be
explained by certain key events. I find a minority signification ratio of over 20% in four months. But not all these sparks can be explained by the found key events.
25
The analysis of NYE partly confirms Arendt et al. (2017)’s findings that minorities signification increases after the event. But while they find foreigners, in general, to be more represented in January and February, I cannot find significance for these variables. I find a significant rise in mentions of North Africans in January and February but the model examining asylum seekers again is insignificant, while Arendt et al. (2017) report significant changes for North Africans only in January and for refugees in both months.
The findings are only partly comparable. Arendt et al. (2017) include different nationalities in their variables17. However, the operationalization of “asylum seeker” was the same. Different findings can further be explained by the different sample selection. Arendt et al. (2017) analyze Süddeutsche Zeitung, BILD, and Kölner Express, thus only BILD overlaps with my sample. Various reasons can play could play a role here: The ownership can have an influence as well as the regional location in the case of Kölner Express (Reese & Shoemaker, 2016). Also, their selection of articles is much stricter. They only consider articles as reporting on crime if they mention the criminal, victim, or element of the offense already in the title, subtitle, or lead.
For the attack on the Christmas market in Berlin, no key event study was done yet by other researchers, thus my findings cannot be compared to others. The minority signification of asylum seekers and refugees drops slightly while minority signification for Syrians shows a stronger rise. An explanation for these unsteady findings could be that in the selection of observed
nationalities, the nationality of Amri was not examined directly. Thus, if minority signification occurred for Tunisians, this was not tracked. This might have a slight effect on the findings, but I expect it to be only minor since it was an Islamist attack and Islamist was a keyword. The rising
17 Since Arendt et al. (2017) only study this event, their range of examined nationalities is concentrated on the
nationalities that play a role here. I selected far more nationalities but based on crime statistics of these years. Thus, my operationalization of the term “North Africa” differs from theirs.
26
minority signification of Syrians cannot be explained by the attack. While Amri was registered in German systems with 14 identities, he did not register as Syrian (Deutscher Bundestag, 2019). A man who was first suspected to have conducted the attack was Pakistani, not Syrian. But the regression for Pakistan was insignificant. Also, the drop in mentions of Afghans is suspicious here. Further research on this event specifically might find more evidence.
The reporting on the attack in Berlin shows similarities to the coverage of NYE. Both events were reported nationwide and sparked debates about national security and refugee policies. But while the sexual assaults on NYE were unexpected and conducted by many subjects, the threat of a terror act in Germany was long known. After attacks e.g. in France and Belgium, as well as smaller attacks in Germany such as a refugee attacking people in a regional train in Bavaria in summer 2016, the debate about IS fighters who register as refugees was already going on.
The spikes in minority signification in March and August 2018 are harder to explain. Here no key event on a single day was found. On March 23 an Islamist terrorist took hostages in a supermarket in France. At the end of August, forceful protests took place in Chemnitz, Germany, after a refugee allegedly stabbed a man. But while both events could be key events, they cannot explain the higher values for March and August respectively because the incidents only happened at the end of the month while higher values were already found before. After excluding articles about the lawsuits and the alleged rape18, no significant changes were found. Thus, while the high minority signification ratio in these months is driven by the events, these do not influence
minority signification in general. It is possible that other factors are omitted here. Also, I find no effect of the attack in France. No significant rise for any group was found in April either. But I
27
observed a significantly higher number of articles reporting about alleged criminals from the Middle East in September. In opposite to the events in 2016, the possible key events for March and August 2018 were considered as such because of extensive and sensational reporting. According to Kepplinger and Habermeier (1995), this can also mark a key event. Nevertheless, I do not find a significant change in reporting after these events.
This study can only be a first examination of the topic. Furthermore, the exploratory
research design is focused on giving a first overview of what could be a key event in recent years. In future research, it must be tested more extensively if e.g. terror attacks in other Western
countries can have an effect on German crime reporting. Focusing on single events here can have a couple of advantages as the codebook and the sample is fitted to that incident. Also, in this research, I compared relatively long timeframes of one month before the event to the first and the second month after the event. While Arendt et al. (2017) do the same, it is possible that an effect is only visible for a shorter timeframe. Future research might shed a more nuanced light on what events work as key events and how long the effect holds.
The influence of media types
Formulating H1, I expected the political-right outlet to use more minority signification than the political-left outlet. My findings support this hypothesis. But the comparison of different outlets still led to a couple of surprises. In H2, it was expected that the tabloid mentions foreigner attributes the most. This was not the case in comparison to the political-right outlet and not even compared to the political-left one at all times. In general, my findings support Benson and Wood (2015) while the finding by Schlier and Lincoln (2014) can only be partially supported. These findings could also be influenced by nuances in coding because I only counted mentions. Meanwhile, in the coding process, I noted that the political-left outlet also critically discusses issues of mentioning nationalities in reporting and points out the troubles in this. E.g. in an article
28
about a right-wing feminist movement the taz writes: “But there are reasons why such statistics [German crime statistics] are not very robust. For instance, sex offenses by foreigners are reported more often in percentage than those by Germans or by family members.”(Biller, 2018) 19.
Other researchers weigh the articles by the level of attention that they get in the newspaper. Arendt et al. (2017) use the space an article takes on a page and if a picture was included. Since this is not possible with the data provided by Factiva, such a measurement was not included here20. Further, Yamamoto weighs how present the signification is in the text by looking if this is used in the headline, lead, or main text21. Also, this was not done here but could make a
difference. Further, it was not measured how often the foreigner attribute was mentioned. This could indicate how present minority signification is within the article. Research on the effects of revealing minority belonging has shown that people who consume more biased news also have a more biased view of minorities (Dixon, 2008).
My research has some limitations. While the Factiva category “Crime/Legal Action” that was used for the selection of articles covers crime, I suspect it to be based on keywords in the text and thus includes also articles that touch the topic of a crime but primarily report on a political debate or entertainment. Thus, the data contains background noise that was not filtered out but
19Original: Es gibt jedoch Gründe dafür, weshalb solche Statistiken nicht sehr aussagekräftig sind. Etwa
werden Sexualdelikte von Ausländern prozentual häufiger angezeigt als die von Deutschen oder von Familienmitgliedern.
20 Factiva reports the page that the article was published on as well as the length but not how many other
articles were on the page. In theory this could be found out but for this large sample and considering that data from Factiva can only be downloaded for 100 articles at a time such an operation was not feasible.
21 While Factiva provides indexes for Headline, lead and main text, the separation of articles was not always
29
also not coded as relevant. I treat the data as if this noise is distributed equally across the outlets but due to the vast number of articles, it was not possible for me to test this assumption. Further, it diminishes the minority signification ratio. Another limitation arises because the selection of articles was based on keywords for criminals which were derived from a test sample of 100 articles as well as synonyms for “criminal” found in a synonym database on the internet. While all three outlets were used to create the list and it is supposed to be inclusive, it is theoretically possible that this selection is biased22. Because of the inconsistency in the Factiva data as well as the automatization that would require a very elaborate codebook it was not possible to establish these criteria in my study. Nevertheless, the trend I find goes in the same direction.
Assessing the outlet selection leads to further limitations. The availability of data led to the selection of Die Welt as the conservative outlet. According to the hierarchy of influences by Reese and Shoemaker (2016), the organization affects the selection and framing of news. Since BILD and Die Welt both belong to the publishing house Axel Springer SE, ownership could have an effect that is not controlled for here. Future research should take this into account and test if the same differences between outlets on the political spectrum or between outlet types also occur when other publishers are involved23.
Also, especially articles by BILD occurred to be incorrectly separated. A possible
explanation is the tabloid format of the newspaper, which might make the import of the articles more fault-prone. In 142 cases the page number was indexed as the headline of an article, most
22 Refining this list or coding all articles that Factiva holds in the category “Crime/Legal Action” was out of
scope for this work especially because only 100 articles can be downloaded at a time and it is not possible to automate this process.
23 A comparison to Frankfurter Allgemeine Zeitung for example is desirable but was not possible due to
30
probably collecting the whole page as one article. 26 of these pages contained foreigner attributes24. While in none of these pages two independently reported foreigner attributes
occurred (thus attributes are still coded per article, not per page) the number of articles that do not contain foreigner attributes might be skewed in the outlet.
I discovered several opportunities for further research: Following up on the issue that articles that discuss the naming of nationalities and here often refer to cases with foreigner as alleged criminals themselves, further analysis on the already collected data could be done to examine the framing in the articles that contain foreigner attributes. E.g. Jacobs (2016) compare the framing of foreigner groups to natives in Flemish crime reporting. Also, minority criminality as indicated by Yamamoto (2012) examines framing.
To fit the scope of this research, I had to restrict the number of nationalities that I evaluated. Future research could fill this gap and analyze also other groups than the ones here. Connecting content analysis and effect studies, further suggestions evolve. Fitzgerald, Curtis, and Corliss (2011) find that crime anxiety leads to a negative attitude towards immigration policies and that this connection is the strongest for politically interested people. It is possible that this effect does not only exist on the general level but also towards specific ethnic groups. Since politically interested people also tend to consume more news, the exaggeration of certain
minority groups as criminals could then have implications for the attitude towards immigration of these groups. Dixon (2008) examines the effects of an overrepresentation of Blacks in California crime reporting. He finds that people who consume news do not only evaluate Blacks more negatively but also as hypothetical criminals whose race is unclear. In Germany and especially in newspaper reporting, names could have a similar effect. Because names can give a clue about the
24 The error here was not distributed equally but concentrated on the years 2014 (56 of 106), 2015 (85 of 168)
31
ethnic origin of a person, it might not be necessary to mention the nationality of an alleged criminal explicitly to foster prejudices. An article by taz about the stabbing of a 14-year-old by a 15-year old German points this out: „what he [Gunnar Lindemann, AfD deputy from Berlin] and many other grumblers on the internet are missing is the name of the boy, at least a first name. So, one of the comments underneath the post [by the police] says ‚can one even pronounce the croak name of this passport German perpetrator?’”(Peter, 2018)25. This shows that also people who hold a German passport are not respected as truly German in the eyes of at least some people. Hefner et al. (2007) do an experiment to compare the evaluation of German and Turkish nationals as criminals and aides. While they mention the Turkish nationality explicitly, it might be
interesting to test whether the same effects occur if only the name gives a clue about the origin. Thus, to evaluate this connection, I suggest to first analyze how often and in which context names of alleged criminals are mentioned and second evaluate if this has the same effect as mentioning the nationality. Furthermore, Jacobs (2016) finds foreign criminals in TV news to have their name mentioned, and a picture is shown fewer times than Belgians which contributes to a homogenization of foreigners that leads to prejudice.
Lastly, it needs further research to determine in what contexts foreigner attributes of victims are mentioned. While foreigners as criminals were analyzed by several researchers (Arendt et al., 2017; Esser, 2002; Tort et al., 2016) no structured research was done yet on the representation of migrants as victims in recent years26. Especially since the refugee crisis began,
25 Original quote: „Was ihm und vielen anderen Pöblern im Netz fehlt, ist der Name des Jungen, mindestens ein Vorname. So heißt es in den Kommentaren unter dem Post etwa: „Kann man den Krechznamen dieses
passdeutschen Straftäters überhaupt aussprechen?““
26 Bade (2018) compares the 1990’s to 2010’s but rather on a theoretical level than using a consistent research
32
attacks on people with migration background in general and refugees are seen to have risen. The sad highlights of this development were attacks like the one in Halle in October 2019 or in Hanau in February 2020.
This study aimed to answer how the portrayal of alleged criminals changed in German crime reporting from 2014 to 2019. I found that generally minority signification depending on the minority group rises over the years. I found that while Europeans are underrepresented as
criminals in reporting, people from the MENA region are overrepresented. The
overrepresentation strengthens for some nationalities during the refugee crisis. The key event study only shows limited effects on minority signification of some groups. While the sexual assaults on NYE 2015/16 were found to affect the reporting on criminal foreigners, the terror attack on the Breitscheidplatz in December 2016 only showed marginal effects. Further, the hearings on two cases of murder committed by a refugee were found to affect reporting in March. But again, the murder and following protests in Chemnitz at the end of August 2018 might have influenced reporting. Thus, I suspect the found trend to be driven by an accumulation of events that were heavily debated in the media even though they were not analyzed as key events in this study.
Looking at different newspapers, the conservative outlet reported on criminals as members of a minority group the most. Overall, the liberal newspaper mentioned the origin of alleged criminals less than the conservative newspaper and the tabloid, this was not true for all years. Concluding it can be said that the reporting on criminals as part of a minority group is always present but changes from 2014 to 2019 and is also influenced temporarily by key events.
33
Reference List
Alikhani, B., & Rommel, I. (2017). Aufstieg des Kulturrassismus: Von Huntington zu Sarrazin.
Zeitschrift für Vergleichende Politikwissenschaft, 12(1), 9-24.
doi:10.1007/s12286-017-0363-8
Alsultany, E. (2013). Arabs and Muslims in the Media after 9/11: Representational Strategies for a “Postrace” Era. American Quarterly, 65(1), 161-169.
Arendt, F., Brosius, H.-B., & Hauck, P. (2017). Investigating the consequences of the key event “New Year’s Eve in Cologne” on crime reporting. Publizistik, 62(2), 135-152.
doi:10.1007/s11616-017-0322-z
Bade, K. J. (2018). From misswords to missdeeds: Cultural anxiety, populism, and enemy images in the German discussion on migration and asylum.; Von Unworten zu Untaten:
Kulturängste, Populismus und politische Feindbilder in der deutschen Migrations- und Asyldiskussion zwischen ‚Gastarbeiterfrage’ und ‚Flüchtlingskrise’. (s30).
doi:10.12759/hsr.suppl.30.2018.338-350
Benson, R., & Wood, T. (2015). Who Says What or Nothing at All? Speakers, Frames, and Frameless Quotes in Unauthorized Immigration News in the United States, Norway, and France. American Behavioral Scientist, 59(7), 802-821. doi:10.1177/0002764215573257 Biller, Z. (2018, 17.03.2018). petition der woche – Frauenrecht? Rechte Frauen! taz – die
tageszeitung, p. 18.
Blumer, H. (1958). Race Prejudice as a Sense of Group Position. The Pacific Sociological
34
Bonfadelli, H. (2007). Die Darstellung ethnischer Minderheiten in den Massenmedien. In Medien
und Migration : Europa als multikultureller Raum? (1. Aufl. ed.). Wiesbaden: VS Verlag
für Sozialwissenschaften.
Bonfadelli, H., & Moser, H. (2007). Medien und Migration : Europa als multikultureller Raum? (1. Aufl. ed.). Wiesbaden: VS Verlag für Sozialwissenschaften.
Boomgaarden, H. G., & Vliegenthart, R. (2009). How news content influences anti-immigration attitudes: Germany, 1993-2005. European Journal of Political Research, 48(4), 516-542. doi:10.1111/j.1475-6765.2009.01831.x
Boumans, J. W., & Trilling, D. (2015). Taking Stock of the Toolkit. Digital Journalism, 4(1), 8-23. doi:10.1080/21670811.2015.1096598
Bundeskriminalamt. (2014). Tabelle 61: Straftaten und Staatsangehörigkeit nichtdeutscher Tatverdächtiger. In P. K. 204 (Ed.).
Bundeskriminalamt. (2015). Tabelle 61: Straftaten und Staatsangehörigkeit nichtdeutscher Tatverdächtiger. In P. K. 2015 (Ed.).
Bundeskriminalamt. (2016). Tabelle 61: Straftaten und Staatsangehörigkeit nichtdeutscher Tatverdächtiger. In P. K. 2016 (Ed.).
Bundeskriminalamt. (2017). Tabelle 61: Straftaten und Staatsangehörigkeit nichtdeutscher Tatverdächtiger. In P. K. 2017 (Ed.).
Bundeskriminalamt. (2018). Tabelle 61: Straftaten und Staatsangehörigkeit nichtdeutscher Tatverdächtiger. In P. K. 2018 (Ed.).
Bundeskriminalamt. (2019). Tabelle 61: Straftaten und Staatsangehörigkeit nichtdeutscher Tatverdächtiger. In P. K. 2019 (Ed.).
Buzan, B., Wæver, O., & Wilde, J. d. (1998). Security : a new framework for analysis. Boulder, Colo: Lynne Rienner.
35
Czymara, C. S., & Schmidt-Catran, A. W. (2017). Refugees Unwelcome? Changes in the Public Acceptance of Immigrants and Refugees in Germany in the Course of Europe’s
‘Immigration Crisis’. European Sociological Review, 33(6), 735-751. doi:10.1093/esr/jcx071
Deutscher Bundestag. (2019). Zeugin schildert Klärung der Identität des Attentäters Anis Amri. Retrieved from https://www.bundestag.de/dokumente/textarchiv/2019/kw14-pa-ua-breitscheidplatz-631858
Dixon, T. L. (2008). Crime News and Racialized Beliefs: Understanding the Relationship
Between Local News Viewing and Perceptions of African Americans and Crime. Journal
of Communication, 58(1), 106-125. doi:10.1111/j.1460-2466.2007.00376.x
Dixon, T. L., & Linz, D. (2000). Overrepresentation and Underrepresentation of African
Americans and Latinos as Lawbreakers on Television News. Journal of Communication,
50(2), 131-154. doi:10.1111/j.1460-2466.2000.tb02845.x
Entman, R. M., & Rojecki, A. (2001). The black image in the white mind : media and race in America.
Esser, F. (2002). Fremdenfeindlichkeit als Medienthema und Medienwirkung : Deutschland im
internationalen Scheinwerferlicht. Wiesbaden: Westdeutscher Verlag.
ExplosionAI GmbH. spaCy – Industrial-strength Natural Language Processing in Python. Retrieved from https://spacy.io
Fitzgerald, J., Curtis, K. A., & Corliss, C. L. (2011). Anxious Publics. Comparative Political
Studies, 45(4), 477-506. doi:10.1177/0010414011421768
Galtung, J., & Ruge, M. H. (1965). The Structure of Foreign News: The Presentation of the Congo, Cuba and Cyprus Crises in Four Norwegian Newspapers. Journal of Peace
36
German Press Council. (2017). German Press Code. In Guidelines for journalistic work as
recommended by the German Press Council.
Hafez, K. (1999). Antisemitismus, Philosemitismus und Islamfeindlichkeit: ein Vergleich
ethnisch-religiöser Medienbilder. Christoph Butterwegge/Gudrun Hentges/Fatma Sarigöz
(Hg.), Medien und multikulturelle Gesellschaft, Opladen, 122-136.
Harcup, T., & O'Neill, D. (2001). What Is News? Galtung and Ruge revisited. Journalism
Studies, 2(2), 261-280. doi:10.1080/14616700118449
Hefner, D., Klimmt, C., & Daschmann, G. (2007). Typisch Türke? Die Folgen der
Nationalitätsnennung für die Bewertung von Akteuren in der Nachrichtenrezeption.
Medien & Kommunikationswissenschaft, 55(4), 575-594.
doi:10.5771/1615-634x-2007-4-575
Jacobs, L. (2016). Patterns of criminal threat in television news coverage of ethnic minorities in Flanders (2003–2013). Journal of Ethnic and Migration Studies, 43(5), 809-829.
doi:10.1080/1369183x.2016.1217152
Kepplinger, H. M., & Habermeier, J. (1995). The Impact of Key Events on the Presentation of Reality. European Journal of Communication, 10(3), 371-390.
doi:10.1177/0267323195010003004
López, F., & Romero, V. (2014). Mastering Python Regular Expressions. Birmingham – Mumbai.
Peter, E. (2018, 13.03.2018). Ein Mord reicht ihnen nicht mehr. taz – die tageszeitung, p. 22. Reese, S. D., & Shoemaker, P. J. (2016). A Media Sociology for the Networked Public Sphere:
The Hierarchy of Influences Model. Mass Communication and Society, 19(4), 389-410. doi:10.1080/15205436.2016.1174268
37
Schartau, L. K., Roy-Pogodzik, C., Gruß, J., Feltes, P. D. T., Goeckenjan, P. D. I., Hoven, P. D. E., . . . Singelnstein, P. D. T. (2018). Die Angst vor dem Fremden. Projekt "Flucht als Sicherheitsproblem". Bochum/Köln.
Scheufele, B. (2002). Mediale Kultivierung des Fremden. In F. Esser, B. Scheufele, & H.-B. Brosius (Eds.), Fremdenfeindlichkeit als Medienthema und Medienwirkung : Deutschland
im internationalen Scheinwerferlicht. Wiesbaden: Westdeutscher Verlag.
Scheufele, B., & Brosius, H.-B. (2001). Gewalt durch „Fremde“ – Gewalt gegen „Fremde“. Die Berichterstattung über PKK- und Kurden-Gewalt und fremdenfeindliche Straftaten.
Politische Vierteljahresschrift, 42(3), 447-473.
Schicha, C. (2008). Skandale in den Medien durch die Medien – Eine Typologie von Formen, Ausprägungen und Motiven. [Scandals in media by media – A typology of forms, characteristics and motives]. Zeitschrift für Kommunikationsökologie und Medienethik,
10, 73-83. Retrieved from
https://www.netzwerk-medienethik.de/wp-content/uploads/2012/01/ZfKM_2008.pdf
Schürer, S. (2018). Türkeistämmige Personen in Deutschland. Erkenntnisse aus der
Repräsentativuntersuchung "Ausgewählte Migrantengruppen in Deutschland 2015" (RAM). In: Forschungszentrum Migration, Integration und Asyl.
Thiele, M. (2015). Medien und Stereotype: Transcript Verlag.
Tort, M. N. G., Guenther, L., & Ruhrmann, G. (2016). Von kriminell bis willkommen. Wie die Herkunft über das mediale Framing von Einwanderern entscheidet. Medien &
Kommunikationswissenschaft, 64(4), 497-517. doi:10.5771/1615-634x-2016-4-497
Trilling, D., Kroon, A. C., & Raats, T. (2020). Guilty by Association: Using Word Embeddings to Measure Ethnic Stereotypes in News Coverage. Journalism and Mass Communication
38
Weidenbach, B. (2020). Überregionale Tageszeitungen in Deutschland nach verkaufter Auflage im 1. Quartal 2020. Retrieved from
https://de.statista.com/statistik/daten/studie/73448/umfrage/auflage-der-ueberregionalen-tageszeitungen/. from Statista
https://de.statista.com/statistik/daten/studie/73448/umfrage/auflage-der-ueberregionalen-tageszeitungen/
Yamamoto, R. (2012). Bridging crime and immigration: Minority signification in Japanese newspaper reports of the 2003 Fukuoka family murder case. Crime, Media, Culture: An
International Journal, 9(2), 153-170. doi:10.1177/1741659012461039
Zentrum für Antisemitismusforschung, & Institut für Vorurteils- und Konfliktforschung e.V. (2014). Zwischen Gleichgültigkeit und Ablehnung – Bevölkerungseinstllungen gegenüber
39
Appendix
Distribution of articles
Keywords to select articles
(Täter* or Mörder* or Verbrecher* or Gangster or Kriminelle* or Dealer* or Terrorist* or Attentäter* or Delinquent* or Dieb* or Räuber* or Einbrecher* or Ganov* or Langfinger* or Messerstecher* or Hochstapler* or Betrüger* or Verdächtig* or Vergewaltiger* or Randalierer* or Gewalttätige* or Unbekannte* or Brandstifter* or Streith?hn* or Beteiligte*)
0 50 100 150 200 250 300 350 400 2014 01 2014 03 2014 05 2014 07 2014 09 2014 11 2015 01 2015 03 2015 05 2015 07 2015 09 2015 11 2016 01 2016 03 2016 05 2016 07 2016 09 2016 11 2017 01 2017 03 2017 05 2017 07 2017 09 2017 11 2018 01 2018 03 2018 05 2018 07 2018 09 2018 11 2019 01 2019 03 2019 05 2019 07 2019 09 2019 11 num be r o f ar ti cle s Month
Figure 5
Distribution of articles
40 Figure 6
41 Coding scheme Table 1 Coding scheme Affiliation Example Nationality: Nationality, immigrational
background, regional generalization
Afghanistan Turkey Rumania Syria Poland Iraq Serbia Bulgaria ) Italy Iran Nigeria Russia Albania Kosovo Ukraine Croatia Morocco Georgia Greece Macedonia Eritrea
Bosnia & Herzegovina Pakistan
Algeria Regional attributes Maghreb,
North Africa, Africa, Southern,
Cumulated regional attribute MENA: North Africa, Maghreb, Algeria, Morocco, Arabic,
42
Middle East: Arabic, Southerner, Kurds, Afghanistan, Turkey, Iraq, Iran, Pakistan and Syria.
North African: North Africa, Maghreb, Southerner, Algeria, Arabic or Morocco
East European: Romanies and Sinti, Albania, Bulgaria, Bosnia, Georgia, Kosovo, Croatia, Macedonia, Rumania, Serbia, Ukraine Religion:
Affiliation to a religious group,
Muslim, Salafist, Islamist
Ethnicity:
Affiliation to a people
Arabs, Kurds,
Romanies & Sinti Social background:
Residence permit status, the reason for being in Germany
Asylum seeker, refugee,
migration background, Type of crime:
in categories used by the crime statistics
terror
organized crime/ clan/ mafia theft/ robbery/ burglary vandalism
sex crime drugs/ substances
illegal residence/ asylum fraud arson fraud harassment/ threat bodily harm homicide people smuggling forced prostitution kidnapping human trafficking Terror groups PKK Kurd
43
Selection of nationalities
Figure 1 Crime in Germany by Nationality excluding violations of residency and asylum laws and agreement of free movent [Straftaten insgesamt, jedoch ohne Verstöße gegen das Aufenthalts-, das Asylverfahrens- und das
Freizügigkeitsgesetz/EU (Schlüssel 725000)] 0 10,000 20,000 30,000 40,000 50,000 60,000 70,000 80,000 90,000 2 0 1 4 2 0 1 5 2 0 1 6 2 0 1 7 2 0 1 8 2 0 1 9 F ig u r e 7 C r im e b y na t iona l it y ( 2 0 1 4 - 2 0 1 9 )
Afghanistan Albanien Algerien
Bosnien und Herzegowina Bulgarien Eritrea
Frankreich Georgien Griechenland
Irak Iran Italien
Kosovo Kroatien Marokko
Mazedonien Nigeria Pakistan
Polen Rumänien Russische Föderation
Serbien Somalia Syrien
44 Krippendorff’s Alpha Table 2 Krippendorff’s Alpha Variable Alpha foreigner: 0.88 uncertain: 0.26 duplicate: nan othergenre_foreignreport: 0.55 asylumseeker_refugee: 0.83 migrationbackground: 1.0 africa: 1.0 north_africa: 1.0 arab: 1.0 southerner: 1.0 maghreb: nan east_europe_balkan: 1.0 romanies_sinti: nan kurds: nan islam_muslim: 0.66 islamist_salafist: 0.86 Afghanistan: 0.74 Albania: 1.0 Algeria: 1.0 Bulgaria: nan Bosnia: nan France: 1.0 Eritrea: 1.0 Georgia: nan Greece: nan Italy: 1.0 Iraq: 0.89 Iran: nan Kosovo: nan Croatia: 1.0 Morocco: 0.85 Macedonia: nan Nigeria: nan Pakistan: 1.0 Poland: 0.85 Rumania: 1.0 Russia: 0.85 Serbia: 1.0 Somalia: 1.0 Syria: 0.81 Turkey: 0.85 Ukraine: 1.0 terror: 0.83 organized_clan_mafia: 1.0 theft_robbery_burglary: 0.72 vandalism: 1.0 sexcrime: 0.95 drugs_substances: 0.5 illegalresidence_asylumfraud: 0.66 arson: nan fraud: 1.0 harassment_threat: 0.66 bodilyharm: 0.7 homicide: 0.8 peoplesmuggling: 1.0 forcedprostitution: nan kidnapping: nan humantrafficking: nan easteu: 1.0 mena: 0.85 nafri: 0.92 meast: 0.87 .
45
Analysis
New Year’s Eve
Table 3
Crosstabulation minority signification of foreigners one month before, one month after and two months after NYE
Not mentioned mentioned
n % n % Total
Month before key event 175 91.15 17 8.85 192 First month after key event 270 84.38 50 15.63 320 Second month after key event 197 87.17 29 12.83 226
Total 642 96 738
Table 4
Logistic regression predicting minority signification of foreigners after NYE
Foreigner B SE B exp(B) z p
constant -2.33 0.25 0.10 -9.18 <0.001
First month after key event 0.65 0.30 1.91 2.17 0.030 Second month after key event 0.42 0.32 1.52 1.29 0.198
