Abstract
Evidence that women have paid the price of what has been labeled the “shadow pandemic” is found in the increase in violence against them. The rates of femicide and lack of trust in the Turkish judicial system of the authoritarian government are worrisome. Protests have been made in the streets and through social media in response. Our goal was to determine the role of legacy news media and social media in bringing awareness of the femicide issue to the public and how affective publics function surrounding a relatively unaddressed societal problem using datasets created by women’s organizations. We conducted a content analysis on 150 sampled femicide cases in Turkey before and during the pandemic taken from online news sources. We investigated the quality of traditional news media coverage and the volume of social media users expressing emotional reactions to individual femicides over time. Results suggest that the journalistic performance of covering the issue of femicide fails to detail essential facts, and awareness and concern for the issue are evident in the Likes, retweets, shares, and expressions of emotional engagement provided to the victims, while online reactions to femicide have increased substantially since 2019, contributing to the formation of affective publics. The study makes a conceptual contribution to the understanding of legacy media and social media’s roles in spurring public engagement about serious social problems in autocratic political contexts while advancing the methodological tools of combining social scientific techniques with computational ones.
Keywords
Data collected by UN Women from four countries in the first months of 2021 revealed that half the women surveyed had personally experienced physical violence or verbal abuse or knew another woman who had since the start of the pandemic (“Emerging Data,” 2021). Dramatic increases in women’s calls to helplines or requests for emergency services have been reported in countries in Europe, North and South America, and Asia (“The Shadow Pandemic,” 2020). Our study focuses on the case of Turkey, an upper middle-income country facing recently serious economic challenges. Turkey is also an increasingly authoritarian country where the government is in control of more than 90% of the media (Murat, 2018) with a near-complete absence of press freedom (Reporters Without Borders, 2022) and an environment where women’s contributions to society are devalued unless focused on the role women play in birthing several children and supporting their families (Ogan & Baş, 2020). We seek to answer the question of how the rising cases of femicide (Cupolo, 2020; Durairaj, 2020) are portrayed in the mainstream media and what role social media, where a slightly more open climate exists, play in protesting those crimes.
The Turkish government does not regularly release data regarding violence against women. Consequently, two women’s organizations have published their own statistics on femicides and other violence, which in 2021 totaled 419 cases (Anıt Sayaç, 2022; Kadın Cinyetlerini Durduracağız Platformu, 2022). The We Will Stop Femicides Platform (Kadın Cinayetleri Durduracağız Platformu) issues monthly details of confirmed and suspicious cases of femicides taken from media reports of the events.
Not all women are equally likely to be subjected to violence. Following on Crenshaw’s introduction of the concept and its concern with the intersection between race and gender discrimination under the law to understand how the intersection shapes social and political life (Crenshaw, 1989; Weldon, 2008, p. 194), others have focused on gender identity and the ways it takes on meaning related to another social category, such as class or ethnicity. Brown et al. (2017) added that the perspective of intersectionality allows for the analysis of power as social identities are examined structurally, politically, and representationally (p. 1832). The authors researched the use of intersectionality in social media activism surrounding the Twitter campaign against violence directed at Black women by the police and other groups.
One recent study of violence against women in Turkey during the pandemic illustrates the ways disadvantaged women are frequently targeted for violence. Akalin and Ayhan (2022) surveyed 1,036 women between the ages of 18 and 59 from all seven geographical regions in Turkey. Being unemployed and experiencing reduced satisfaction in their relationships with their spouses or partners were factors in reporting increased intimate partner violence. Women living in rural areas experienced violence at higher levels than women in urban areas (65.2%–34.1%) (Akalin & Ayhan, 2022, p. 3), while the frequency of the violence was 1.4 times higher for those women reporting low-income status. For these women, gender was considered at the intersection of variables that made them more vulnerable to violence–poverty, income dependence on males, and rural residence. Though the study did not focus on immigrant women, our sample included femicides of refugees and immigrants. Being economically challenged because of their immigrant status possibly increases the likelihood of their vulnerability to abuse.
President Recep Tayyip Erdoğan’s authoritarian Justice and Development Party (AKP) views women’s primary role in society as wives and mothers, with a status unequal to men (Ogan & Baş, 2020), thus implicitly granting permission to would-be perpetrators of violence. The government has even chosen to leave a major European treaty (the Istanbul Convention) that set legally binding standards to prevent violence against women (“Istanbul Convention, 2019,” para. 6). Supporters of the government’s decision claimed that such policies are not popular in the country, arguing that any action to promote gender equality results from international commitments and the work of “submissive governments” and the “bureaucratic elite” (Altan-Olcay & Oder, 2021).
Long before the decision to withdraw from the Istanbul Convention, cases of violence against women were rising within a broader sociopolitical context during the COVID-19 pandemic. Based on the datasets from Anıt Sayaç 1 (2022) and Kadın Cinayetleri Durduracağız, 2 this study identifies the cases of femicides in Turkish news sources before and during the pandemic. We content analyzed 450 stories about the victims alongside responses to social media postings about the women’s deaths through Facebook and Instagram Likes, shares, comments, and emotional reactions available on Facebook posts as an alternative to Likes. Our investigation was conducted to determine the level of news media awareness of the growing femicide issue provided to the public and how social media users express concerns about individual femicides over a 3-year period. Specifically, we asked the following research questions: (a) What does the information in news stories about femicide cases reveal about the thoroughness of the femicide coverage? (b) What evidence do online emotional reactions provide concerning the existence of affective publics formed around the issue? and (c) How do collective public responses in social media change over time toward femicide cases?
The Conditions Surrounding Journalism in Turkey Today
Though complete media freedom never existed in Turkey (Coşkun, 2020), the degree of control by the government today severely limits the media’s ability to produce quality journalism. Coşkun (2020) identifies the three-pronged strategy used by the government to subdue that freedom. First, it has forced opposition owners out through financial sanctions and the effective use of government institutions working in concert to pressure those owners, replacing them with AKP businessmen; second, it has used the regulatory agency, Radio and Television Supreme Council (RTÜK), to fine the outlets they could not fully control; and third, it attacked critical journalists through legal actions and threats of imprisonment (Coşkun, 2020, p. 650). As a result, a large number of journalists have been jailed, left the profession, or work for small online news sources. Following the coup attempt in 2016, many media outlets were shut down leaving journalists without work. In 2020, an estimated 11,000 journalists were unemployed, while the press cards of thousands more were canceled and 90 were in jail according to a report from several professional journalists’ associations (“Record Increase,” 2020). These issues have only increased the difficulties for journalists during the pandemic, including new forms of censorship and large fines for reporting the actual severity of the Covid-19 crisis in the country (Yezdani, 2021). Some journalists fear returning to the profession, while local news outlets lack resources to train journalists (Coşkun, 2020; O’Donohue et al., 2020). These conditions can lead to the publication of unprofessional and incomplete stories, that is, of poor quality. A basic model for journalistic quality in a news story taught in reporting classes is one that includes information answering six questions—Who (is identified as the individuals the story is about); What (the details of the events described in the story); When (the time frame of the story); Where (the place where the story occurred); Why (possible explanations and interpretations of the circumstances of the story); and How the situation came about. As Dennis and Merrill (1991) categorize stories that are event or feature-oriented, three-story models (descriptive, analytical, and consequential) predominate. The descriptive story should answer the questions who, what, where, and when, and is used for event coverage, spot-news stories, and singular instance coverage (Dennis & Merrill, 1991, p. 121). Analytical stories should delve into the how and why of the story that focuses on the “forces at work, competing interests, and possible explanations” (Dennis & Merrill, 1991, p. 121). Finally, the consequential story answers the “so what” question, explaining the overall meaning of the event and its consequences for individuals, communities, and so on (Dennis & Merrill, 1991, p. 121). The femicide stories in our sample largely fall into the descriptive model, while a few adopted the analytical or consequential. Longer stories tended to be more analytical or consequential. Since we were specifically focused on the completeness of the news at the level of the story describing particular femicides, we included variables for the essential information needed to explain the victim’s story. We analyzed the amount of missing information across those variables in each story to answer the following question:
Research Question 1 (RQ1): What do the characteristics of the presentation of information about femicide cases in news stories reveal about the overall nature of reporting on femicide in Turkey?
Activism and Use of Social Media for Women’s Rights in Turkey
Yeşim Arat (1994) dates the origin of Turkish feminism to 1983 when Şule Torun added the topic to a weekly journal, Somut, sparking protests among women across the country—in person, in writing, and in institutions supporting their causes (p. 100). Men have more recently joined women in the quest for gender equality and women’s rights. As social media evolved, online protests were added to public demonstrations calling for changes in law, policy, and customs. The Gezi Park social movement in 2013 was one of the first to combine online with offline communication in the month-long protests (Ogan et al., 2016; Ogan & Varol, 2017; Varol et al., 2014). Though protest against the government in power was the main target in Gezi, many human rights issues were among the grievances. Women’s rights were at the forefront of those (Erhart, 2013). Gezi protests did not rid the country of institutionalized sexism or homophobia, but it was the largest civil disobedience in recent memory, as supporters protested government-assigned roles to women, rejecting government control of their bodies (Erhart, 2013, p. 302). Since that time, women have continued to protest the limiting of their choices (Ogan & Baş, 2020). The Erdoğan government and his ruling party, the AKP, have pressed women to fill traditional roles, preferring them to marry, bear at least three children; choose family over career; and have limited reproductive rights, including reduced access to abortion.
Turkish women’s struggle against the AKP government’s repressive policies and its failure to defend women’s rights is in line with the repertoire of political engagement enabled by digital communication technologies (the connective action as put forth by Bennett and Segerberg in 2012). The authors draw the distinction between connective and collective action as follows: the “logic of collective action (is) associated with high levels of organizational resources and the formation of collective identities,” while connective action relates to personal content-sharing in social media networks (Bennett & Segerberg, 2012, p. 739). Writing at the time of the Arab Spring the authors argued that digitally networked action adopted in that movement meant that it was remarkably similar to “action formation in decidedly undemocratic regimes” (Bennett & Segerberg, 2012, p. 743). Much has changed in the use of digital networks for social change since that time, however. In Turkey, the government has taken harsh measures to restrict the content and distribution available on the global Internet and also on communication through social media, thus controlling criticism and dissent for both in-person and online communication.
As documented in a recent report from İfade Özgürlüğü Derneği (Freedom of Expression Association), 58,809 websites were banned in 2020 alone (Fahrenheit 5651: Sansürün Yakıcı Etkisi, 2021). In addition, 7,500 Twitter accounts, 50,000 tweets, 12,000 YouTube postings, 8,000 Facebook postings, and 6,800 Instagram postings were blocked by the end of 2020 following Law 5651 and other provisions. In 2020, 15,832 news articles were removed from websites under government order (Fahrenheit 5651: Sansürün Yakıcı Etkisi, 2021). However, some efforts are also shown to be ineffective since users can use VPNs and other means of accessing information (Castro, 2022). A law passed on 1 October 2020, required social media companies serving more than one million users in Turkey to engage a domestic representative within Turkey. To date, Twitter, TikTok, and YouTube have complied with the law (İfade Özgürlüğü Derneği, 2021). Finally, Freedom House has ranked Turkey’s Internet in 2021 as “not free,” based on its creation of obstacles to access, limits to content, and violations of user rights (Freedom on the net 2021: Turkey, 2021). All these actions contribute to the reduction of the use of social media as a connective action tool for those who wish to end violence against women.
Given the changing political and structural context of the country since the Gezi Park protests, it is necessary to examine the current forms of online political activism as afforded by digital media within social networks. We focus on considering the use of affective publics or “networked public formations that are mobilized and connected or disconnected through expressions of sentiment,” as conceptualized by Papacharissi (2015, p. 125). Against this backdrop, this study analyzes online public responses and emotional reactions to social media posts given to a politically charged social issue, that is, violence against women, as a form of “affective publics” formation (Papacharissi, 2015).
Emotional Reactions to Femicides on Social Media as the Role of Affective Publics
The formations of affective publics are aligned with the logic of connective action and “driven by streams that are collections of opinions, facts, and emotion blended into one effusive stream” such that diverse distant publics are able to bring together individuals as they affectively focus attention on an issue of concern (Papacharissi, 2015, p. 129). Social media promote affective expressions through the Likes they allow on postings as well as comments and shares of those posts, thus allowing for otherwise unconnected people to learn about and then connect with others across geographic and social distances about an issue of interest to them all.
The research literature on the role of emotions in social media postings suggests that certain social media content gets more attention and reaction than others. Emotional Twitter messages are spread more often and quickly than neutral ones (Stieglitz & Dang-Xuan, 2013) and high arousal positive and negative messages that evoke awe, anger, or anxiety tend to be reposted more often (Berger & Milkman, 2012). Furthermore, social media users share photographs by considering the potential emotional impact of their posts on their audience and in turn are emotionally affected by the responses they receive (Stsiampkouskaya et al., 2021). However, the impact of the affect is limited unless, as Papacharissi (2015) writes, it “elicits feelings of community and identity” (p. 129). In the case of the struggle for women’s rights in Turkey, feelings of common identity and community have mostly been confined to the virtual world, given the government restrictions on public gatherings that promote women’s rights (Ogan & Baş, 2020).
Lünenborg (2019), in further explication, argues that affective publics are inherently performative in digital networks, and “temporally and situationally sustained in the mediated and localized co-presence of actors” (p. 323). The affective practices in these networks include producing and circulating memes, posting messages, and liking, sharing, and commenting on other users’ posts (Lünenborg, 2019, p. 325). As these affective practices of individuals increase, users may shift from personal to public communication and become part of affective publics previously considered part of the public sphere (Lünenborg, 2019, p. 326).
Some scholars dismiss affective practices, such as liking or sharing content, as slacktivism, defined as “low-risk, low-cost activity via social media, whose purpose is to raise awareness, produce change, or grant satisfaction to the person engaged in the activity” (Rotman et al., 2011, p. 821). Papacharissi (2015), however, presents the value of affective action in one of her five propositions for interpreting the civic gravitas afforded by technologies of premediation and remediation (p. 126). In her fourth proposition, she explains that affective publics produce “interruptions of dominant political narratives by presenting underrepresented viewpoints” (Papacharissi, p. 130). In Turkey, the support for the women who are victims of the violence perpetrated by men in their lives is an unpopular narrative of the ruling party. More often the court makes allowances for men who claim they were provoked by women, reducing their sentences for the provocation. When public demonstrations are banned by authorities, the affective publics produce and sustain emotional responses to the issue through their Likes and shares on social media. The reactions may be posted to a single case of femicide, but frequently users react to the many previous stories of women who have been targeted over time by repeating the victims’ names in reaction to the most recent femicide (Ogan & Baş, 2020). In this way, they keep the issue alive in people’s minds. Lünenborg (2019) calls for future research to study both the “specific implications and relevance of emotions in the constitution of these publics,” (p. 327) which we attempt to do in this research.
In a controlled media environment like Turkey’s, the media audience may depend more on social media for acquiring new information as well as sharing and commenting on that information, especially when there is minimal coverage of the topic in mass media sources, as there is on the femicide issue. They may also use social media more when there is high control over collective action in public spaces. In July 2021, Facebook users in Turkey numbered 62.29 million or about 73% of the population (“Facebook Users,” 2021), while Instagram users, who have been rapidly increasing, numbered 48.8 million or about 58% of the population (“Turkey: Instagram Users,” 2021).
In a study of independent journalism use and the corresponding social media use of citizens in Turkey, Kızılkaya and Ütücü (2021) found that despite the pervasive control of print and broadcast media by the ruling party, independent media’s use (media not directly controlled) is catching up with that of the pro-government media (about 34 million vs 48 million). Furthermore, the “independent media outlets receive 16.5% more interactions on social media and are closer to breaking through the echo chamber” (Kızılkaya & Ütücü, 2021, p. 4). Their findings indicate that social media posts, such as those about femicides, may be driven by interactions with both controlled and independent media and are having an increasing impact among users in the Turkish public sphere.
The popular use of Facebook and Instagram has generated numerous reactions through the Likes provided by both social media platforms. Social media also allows for the sharing of content with friends. In a study of incentives for sharing content with others, Fu et al. (2017) found that different incentives drive different types of content sharing practices (p. 132). Specifically, users who are driven by the psychological incentive of self-interest will share commercial messages and information about lifestyle affairs to bond social capital with other users, while those driven by a communal incentive (such as altruism and interest in connecting) will more often share lifestyle affairs and personal opinions to bridge social capital (p. 30). Another study examining behavioral incentives of shares, but also Likes and comments, classified Likes as affectively motivated, while comments were cognitively triggered, and shares could be either cognitively or affectively driven (Kim & Yang, 2017). They found that sensory and visual features of a post lead to Likes, while those that are rational and interactive result in comments. Sharing is the result of sensory, visual, and rational features in the posts (Kim & Yang, 2017, p. 441).
In the case of news about particular femicide cases, posts could be shared for a communal incentive, to spread the word (and bridge) to the larger community, and perhaps to people who might influence policy changes to prevent violence against women. We acknowledge that collective action surrounding the rise of femicide in Turkey has also taken place over the last several years, and while we are not highlighting the demonstrations in support of justice for the victims and their families, we analyze the online vocalizations of that support through user emotional engagements on Facebook and Instagram. Through their posts on Facebook and Instagram, the users adopt social and political action that is “strongly correlated with offline acts of political participation,” as argued by Piat (2019, p. 162). A study of the Turkish “Challenge Accepted” movement, a form of online activism where violence against women was protested on Instagram by sharing black and white profile pictures, finds that “social media offers its users an alternatively embodied political presence in public that minimizes harm and allows for plural political action and has tangible socio-political consequences and effects on protesters’ bodies and lives” (Ustel, 2021, p. 28). Given the current political context in Turkey where both offline and online forms of collective action are repressed, we investigate whether emotional reactions to social media posts can constitute affective publics that would serve as a legitimate form of political activism. To guide this investigation, the following research question is posed:
Research Question 2 (RQ2): How do individuals and groups (a) post on public Facebook pages and Instagram accounts; and (b) how do the acts of “liking,” commenting, and expressing other emotional reactions contribute to the formation of affective publics around the issue of violence against women changes over time in Turkey?
Methodology
Data Collection
Our study combines manual content analysis and online social network analysis to extract information and patterns from systematically gathered data from social network websites and online news articles.
Identification of Femicide Cases and Data Collection
We first built a collection of femicide victims for the past 10 years in Turkey. We relied on two major websites, Anıt Sayaç and Kadın Cinayetleri, for collecting a comprehensive list of victim names. Both organizations track the news and organize details about the cases, so they provide an invaluable record for our study. We implemented scrapers in Python to download and extract content from these websites. Our scrapers capture names of the victims and additional meta-data about the location of femicide, age of the victim, and information about the perpetrator. Our combined data consist of 3,486 cases that we incorporated to further data collection efforts. In our collection, three of the victims shared the same names with Turkish celebrities. The data collected for these individuals were affected by attention attributed to celebrities, so we excluded these cases from our analysis.
Social Media and Search Data Collection
We analyzed the online content from two distinct data sources: social media messages from Facebook and Instagram and search results from Google. We used social media posts and the number of estimated search results as proxies for public interest and content production, respectively.
In our analysis of social media posts, we preferred Facebook over other social media channels because Facebook posts provide eight distinct emotional responses spanning a wide range of reactions and we could access historical records through the Application Programming Interface (API) provided by the CrowdTangle platform. We queried each victim’s name and collected posts created on public Facebook pages in the past 10 years. These posts provide meta-data about the Facebook page, such as subscriber count, links to shared media, and aggregated statistics about posts. This dataset allows us to monitor temporal changes in emotional responses and the number of posts created from more than 27.5K Facebook pages.
Quantifying the content produced about a femicide case is a challenge, due to articles obviously copied from other news sources, copycat news articles, and the vast amount of news pages published daily. We relied on web search systems to estimate the amount of content on different websites and to retrieve the most relevant web pages. To gather a search dataset, we programmatically queried the names of each victim along with the keyword “cinayet” (Turkish for murder) on Google Search and collected the top 10 search results. We implemented scripts in the Python programming language, and we used default browser and user settings to prevent any personalization that might introduce biases to the data collection process.
Manual Content Analysis of News Articles
Automated analysis provides valuable insights into the networks in capturing complex patterns of diffusion and connectivity, sentiment about the content, and temporal changes in a debate or user behaviors. However, more nuanced analysis requires a combination of manual content (analysis) annotations and automated approaches. Existing research has analyzed social protest-related content using network and content analysis (Ogan & Varol, 2017). In this work, we combined these approaches and extracted information about victims, perpetrators, and the nature of and reactions to the femicides.
We conducted a manual content analysis on the top 3 of 10 news articles returned from a Google search based on a sample of 150 victim names selected at random between 2018 and 2020 (50 random cases for each year). Given the degree of repeated information and even the exact wording of that information, we limited our selection of news articles to the top 3. It should be noted that according to Kızılkaya and Ütücü (2021), Google’s top stories and Google News algorithms are biased toward pro-government Turkish media. Since the three top-ranking traditional media sources were selected for analysis for our study, it is likely that our study also carried that bias. That said, reports about femicides should not be partisan in nature and the highest-circulating press is nearly all owned by pro-government companies or individuals.
Our codebook consists of 42 variables for information about the event, perpetrator, victim, and details of journalistic practices. Two coders were trained on sample data and reliability tests were conducted multiple times. Krippendorf’s alpha ranged from .53 to 1.0 for the 18 variables tested. The criteria to include the variables in the intercoder reliability testing relied on the necessity to use judgment calls in coding (vs sheer reporting of the numbers such as the date of the event, and the age of the victim). It should be noted that the variables tested for reliability are in general those that are more complex to code, rather than coding manifest content. In other words, consensus on relatively more ambiguous variables was attempted to be achieved. Twelve of the 18 variables tested had a Krippendorf’s alpha of at least .70. Therefore, 61% of the variables tested were above the conventional levels of reliability (Riffe et al., 2014). Those variables with lower reliability levels than the conventional standards were mainly those that are more difficult for obtaining high reliability. We believe that the lack of news content about some important variables in our study and the extreme ambiguity for other variables resulted in some of the lower agreement levels.
Findings
Descriptive Findings
The descriptive findings of our news analysis of 450 stories written about 150 women who were killed over a 3-year period from 2018 to 2020 in Turkey covering the basic facts of the news stories alongside qualitative information related to the sensational nature of the crimes and, where possible, information about the court’s decisions regarding the perpetrator. The analysis revealed the following.
Information about the victims and the crimes: Their ages ranged from 1 year to 88 years old, with a mean age of 36.5, standard deviation of 17.09, and median age of 34. The perpetrators were primarily (63.5%) husbands, partners, boyfriends, or men with whom the victims had been romantically involved. The second most common perpetrator was a relative of the victim (19.8%), while only 1.7% were total strangers.
The perpetrators ranged in age from 16 to 85, with a mean age of 39.6, a standard deviation of 13.92, and a median age of 36. Jealousy or the breakup of the relationship was described as the motive in 43.7% of the stories, but other reasons—for example, a dispute over property or money—were cited in 44.1%. Family retribution was given as a motive in 9% of the stories. The perpetrators blamed the women for provoking them 14.4% of the time. One notable example was that of a man who said he killed his wife because she did not display any of his family’s photos in the living room.
The victims investigated in the news stories were generally young (42% were 30 or younger) and without the independence that employment could bring them. Many were likely dependent on their husbands (36.8%) or former husbands (13.6%) who killed them, given that no occupation was cited for 79.3% of the victims. Only 6.4% had requested a restraining order against their perpetrators, yet the orders did not prevent the women’s deaths. One case, likely a murder-suicide, was of two young women whose families forbid their relationship. LGBT (lesbian, gay, bisexual, and transgender) relationships are frowned upon by the AKP government and society for religious and cultural reasons. In the news stories, 80% of the women had no identifiable occupation, while an additional 0.4% were described as homemakers.
The published stories were about homicides occurring in the largest cities of the country (50.3%) or in lesser cities under 2 million (48.1%), leaving us to wonder how many other women were killed in rural areas, but not reported on in the press. In Emre Kızılkaya’s (2021) analysis of the news deserts in Turkey, he found that the places from which the least news is generated or published are the interior regions of the Aegean and the Anatolian provinces (excluding Ankara), alongside the Black Sea and Southeastern regions of the country. That would suggest that femicides taking place in these more rural or isolated areas may not be covered by the press.
Turkey is home to more than four million refugees and immigrants, most from Syria, or less than 5% of the total population of the country (Ten years on, Turkey continues its support for an ever-growing number of Syrian refugees, 2021). However, in this group of femicides, 13.6% of the stories were about immigrant women. Of those stories, 2.9% of the perpetrators were men of migrant origins. The immigrant women’s precarious positions in Turkish society led to their increased vulnerability. No occupation was mentioned for any of the immigrant women victims or the immigrant male perpetrators. Overall, our data support the concept of the intersectionality of femicides.
Journalistic Quality of the News Sources’ Coverage
To answer RQ1, we employed an analysis of the information presented compared to what is needed to answer basic questions about femicide. A content analysis of 450 news articles was conducted and the number of details missing in stories about 150 different femicides were quantified to assess the incompleteness of the news articles. Specifically, we examined the stories’ content based on the classic 5W’s and H necessary for a good fact-based story. We were particularly interested in the information that a reporter might have omitted from the story in answering these basic questions, as doing so leaves out essential facts. We also examined the length of the story as a possible measure of its completeness, while understanding that the reporter may not have been able to obtain complete information in time for a news deadline.
We were surprised to see how much information was missing across types of publication online—in the national versus local press, in broadcast media versus print media, and in Internet-based publications versus those that circulated a print version plus the online alternative. Typically, reporters rely on police reports for breaking news of violent crimes (Farhi & Izadi, 2020), so the information missing from the stories may have been a consequence of incomplete police reports. The quantity of missing information from the stories that answered the “who” (age, occupation, marital status, and pictures of the victim and the age of the perpetrator), “what” (the crime description), “when” (time and date), “where” (crime location), and “why” (motive and related details based on interviews) ranged from 19.5% for location of crime to 83.4% for information gleaned from interviewers or bystanders. Such basic data related to the crimes should have been available to the reporters and included in the news stories. Yet even in the national press where we would expect the highest levels of resources, large percentages of the stories were missing these and other facts (see Table 1 for the frequencies and percentages of different types of content missing in the coded news stories).
Missing Content in News Stories About Feminism.
Note. Total N of stories = 450.
When we examined the missing information broken down by the type of source, we found that the national news media provided no motive for the crime in 40.9% of the stories, while that percentage was 43.5% in the regional and local press, 54.4% in the online-only press, and 43.5% in online television news sites. The marital status of the femicide victims was not included in 30% of national news sources, 36.9% (local and regional), 38.8% (online press), and 41.9% (TV news sites). No specific location for the femicide was listed for 17.7% of national news sources, 23.9% (local and regional), 22.3% (online press), and 9.7% (television news sites).
Given the often complex nature of the crime, we coded for the length of each story, an indicator of its completeness (see Table 2 for the distribution of the textual word length of news stories). The story length varied from a mere 13 words to 1,675 words for the 413 print stories. The remaining 37 stories contained video clips only and were not coded for length, yet the content of the videos and the caption was holistically coded for the other variables in the codebook. The mean length of the text was 252 words, while the median number of words was 177. Only 3.6% of the stories were longer than 1,000 words. We have no information about what was accessible to reporters, but the shorter the story, the more likely that essential information was missing.
Femicide Online Text Word Count.
Note. Total N of stories = 413.
Mean number of words/story = 252.
Median number of words/story = 177.
The seriousness taken by the news organization about an issue can be measured through the source of the story (Lacy & Rosenstiel, 2015, pp. 7, 9). If femicide details were considered important to the news organization, an editor should assign the story to a reporter to investigate those details and the reporter’s byline would appear at the top of the published article. However, only 44 stories of the 450 carried a byline from the news source where it was published. The largest number of stories were attributed to a news agency without a byline and an additional 14 (3.1%) carried a byline from the agency, meaning that no local reporter was assigned to write the story. A total of 184 stories (or 40.9%) listed no source, so we concluded that a local reporter wrote the story without his or her byline. Overall, the level of missing information reflects the low journalistic standards of the news media in Turkey and the relative importance dedicated to such a pressing social issue.
Collective Emotional Reactions on Social Media as Indicators of Affective Publics
Social media provide a valuable lens to investigate public debate and quantify collective reactions to certain events. To answer RQ2a, we analyzed the social media responses collected from Facebook where each post contains the name of a femicide victim. Each post on Facebook contains reactions since early 2016. These reactions capture a wide range of emotions such as angry and sad to reflect negative emotions, haha and wow to show positive responses, and care and thankful to react empathetically. These reactions have been used to identify controversial topics (Basile et al., 2017), and researchers also show that reactions can effectively reflect the feelings of the users (Guintini et al., 2019).
Our analysis traced each post and accumulated emotions on a weekly basis in Figure 1. Not surprisingly, the major reactions toward femicide news are sad and angry. We also observe certain cases receiving massive attention reaching over 100K reactions. We identified the top 5 events based on the impressions in the overall activity timelines and searched for the events leading to these extreme activities. Each of these cases corresponds to distinct femicide cases, and in most cases, 70% of the reactions reflect sadness, while the murder of an LGBTQ+ (lesbian, gay, bisexual, transgender, and questioning) activist and a transwoman, Hande Kamer, led to the most angry responses (50%). Elevated activities in general observed after 2020 indicate a general awareness and reaction to femicides.
Collective emotional social media reactions to femicides.
Our Facebook data contain posts created on different public Facebook pages with each post concerning femicide cases of names in our collection. We built a network representation shown in Figure 2 that captures similarities and communities of Facebook groups identified through network analysis. For each Facebook page, we also store the number of posts published about each victim. Those post counts are represented as vectors for our further analysis. In this network representation, each node corresponds to a specific Facebook page, and the edge weights between different nodes calculated using cosine similarity of vectoral representations of pages that we calculated based on the occurrence of the victim’s name. Since we are interested in different Facebook pages and the communities in the network, we used the modularity maximization algorithm for community detection resulting in a modularity score of 0.512 leading to a meaningful separation of distinct communities.
Network of Facebook pages built by computing similarities of posts about femicides.
In our investigation of communities and the pages within these communities, we made the following observations: (a) women’s rights groups (yellow community) are active as indicated by the node sizes, and they cluster together meaning their posts share a similar set of victims; (b) popular news media (purple) are also present in the network, but their coverage of distinct femicide cases is lower as it is reflected by the node sizes; (c) Facebook groups for leftists and Kemalist pages 3 (light blue and red) are also active on this topic; and (d) surprisingly, we can also observe pages focusing on food, recipes, pets, and so on (dark blue) with a prominent location in the network.
To answer RQ2a, a network analysis on the public Facebook groups was conducted, which illustrates the increasing awareness of the social issue, arguably forming issue publics around femicides. The fact that victims’ names get posted by various Facebook groups with different characteristics and goals indicates that different segments of society share a common concern for the issue. In a media system where the traditional news outlets fail to sufficiently cover the femicide incidents (as our manual content analysis of news media coverage of femicides demonstrated) and the political regime punishes demonstrators for all forms of social protest, this activity of sharing victims’ names on public social media outlets and engaging with them through the expression of emotional reactions can be regarded as an alternative form of activism and the creation of affective publics in the sense that Papacharissi (2015) conceptualizes them.
Collective Public Responses to Femicides Over Time on Facebook and Instagram
RQ2b asks about the online emotional reactions to femicides and the temporalities of the online activities. To address it, we investigated the Facebook and Instagram posts and the collective reactions attached to the content for more than 2,000 femicide cases. The responses are operationalized in the form of Facebook Likes, shares, and comments, and the frequencies of each post were Liked and commented on Instagram.
Figure 3 illustrates the increasing volume of social media engagement with femicide victims’ names comparing Facebook and Instagram engagement. Although both of these platforms are owned by a company recently called Meta, they have different user demographics and engagement mechanisms. Information obtained from Facebook captures activities seen on public groups, while Instagram posts are created by individuals on their public profiles. We observe that on Facebook most responses are generated by Likes and a considerable amount of shares. Similarly, we observe that Likes dominate Instagram responses. We believe that passive reactions such as Shares and Likes help spread messages forward but comments on posts require more active participation and cognitive effort, so they are observed less frequently. It is evident that responses on Instagram are greater than those on Facebook with a significant increase after 2019. The overtime increase in the interest in Instagram can be attributed to the increase in the total number of Instagram users in Turkey as well as the shift in demographics of young social media users from Facebook to Instagram. As in most countries, social media have been popular for a long time. According to Statista, an organization monitoring social media users around the globe, Facebook use in Turkey has continued to increase in popularity over time. In May 2019, Statista reported 36.24 million users, and in 2021, that figure had increased to 62.29 million (“Facebook Users,” 2021). While Facebook is used by all age groups, more than one-third of the nearly 50 million Instagram users are between the ages of 25 and 34 (“Distribution of Instagram Users,” 2021). These statistics help us interpret the relative increase in the reactions to femicide postings on Instagram compared to Facebook.
Weekly responses on Facebook and Instagram platforms.
Conclusion
One of the main findings of the study reported here is the low quality of coverage of femicides in the Turkish press. This is by no means an isolated case, but rather a reflection of the diminishing quality of journalism and resources dedicated to reporting due to the structural political and economic conditions. Surprisingly, even in the national press where we would expect the highest levels of resources, large percentages of the stories were missing essential details. Therefore, we conclude that these women do not get the needed attention in major news outlets. Intersectionality is in evidence by the lack of detail about the victims, the relatively large number of migrant women who were targeted, and the details from the stories that indicate the marginality of the women’s roles in society.
Despite the arguable neglect of the issue in the traditional media outlets, the level of attention social media users pays to cases of femicides points to a spike in awareness about the issue, as well as the formation of affective publics. Specifically, the network analysis conducted on the public Facebook groups suggests an increasing awareness and engagement on the social issue, forming issue publics around femicides, and creating affective publics. The most common emotional reactions given to femicide cases are anger and sadness. Facebook was recently reported to assign five times more weight in its algorithms to “anger” reactions than any other reaction, (Morrow, 2021), thus potentially increasing the polarization of users over time.
Over-time analysis of social media responses in this study may likely indicate that Instagram responses to femicides exceed those of Facebook due to the increasing popularity of Instagram among users in Turkey. In the last few years, the volume of overall posts, as well as engagement with femicide posts, has increased, indicating a heightened awareness toward the social issue.
The study reported here combined manual content analysis with automated social media analysis to assess the level of awareness toward the increasing cases of femicides in Turkey. In a political context where offline and online social protests are repressed, examining the journalistic quality of such social issues as well as the volume of social media posts coupled with the emotional engagements, the posts shed light on the alternative expressions of citizen activism in the current media landscape. Furthermore, the study reported here strengthens the necessity to combine more traditional social scientific methods with computational techniques to advance the state of research.
The reported study focuses on the coverage of the traditional media sources as well as the user-generated content shared on social media about an increasingly dire social issue in Turkey, that is, femicides. While we propose a thorough examination of the media content circulating in the media sphere, we are in the dark regarding the intentions of the journalistic coverage as well as the motivations behind social media users’ responses to these cases. In an authoritarian society where citizens refrain from being actively involved in political protest (Dal & Nisbet, 2022), subtle expressions of emotions on social media can be regarded as alternative strategies adopted to fulfill their needs for political expression (Oz & Yanik, 2022). Yet, to definitively determine the motivations of emotional reactions to social issues on social media, future research should combine content analyses with pointed audience research to determine the intentions of user behavior.
An important limitation of the study is the incomplete dataset of the femicides. Emre Kizilkaya’s work on “deserts of news” suggests that reduced coverage of femicide cases particularly occurs in less populated parts of a country.
The 2022 year may mark a watershed for awareness and engagement with the topic of violence against women in Turkey as a consequence of the production of a documentary film submitted by the United Kingdom for Oscar consideration in the best international feature and best documentary categories. The film, Dying to Divorce (Fairweather, 2021), produced over a 5-year period, focuses on three women and their lawyer who fight for retribution for the brutal attacks they suffered at the hands of men in their lives. Films like this expose the issue to the public in many countries similarly experiencing a wave of violence and femicides in recent years. Perhaps renewed efforts will be made to rejoin the Istanbul Convention and Turkey will be recommitted to the protection of its women. However, recently, official efforts to close the Women’s Platform to Stop Femicides in the country on the grounds that it has broken the law by “disintegrating the family structure by ignoring the concept of the family under the guise of defending women’s rights” raise the stakes for the defense of women’s rights (Michaelson & Narli, 2022, para. 2).
Footnotes
Authors’ Note
We dedicate this work to the memory of Beril Varol, who died at the age of 27 following a femicide attack on 27 May 2022. All three authors contributed equally to the article.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
