Femicide and Necrodata: Ways of Counting the Killing of Women

Abstract

Different disciplinary and socio-geographical approaches to counting, reporting, and categorizing death can yield divergent and often contradictory values of risk, harm, and mortality. This article examines historical and contemporary ways of recording and narrating the incidence of female homicide in South Africa (one of the countries with the highest femicide rates in the world), from state and police statistics to medical studies and media reports, and explores how each of these uses both common and discrete necrodata to construct specific narratives that reveal the [dead] body as a site of conflicting assertions of power.

Introduction

The gender/ed data gap is a major challenge facing contemporary efforts to identify and reduce violence against women (Buvinic & Levine, 2016; Maina & Rooney, 2024). As a data model and framework, this might include recognition that certain forms of violence are gendered, and that many of these types of violence in particular remain under-reported, even unreported, and are therefore under-counted or absent from official figures and statistics. The gender data gap also acknowledges that violence against women is universal but is unevenly distributed, and that regions with the highest rates of violence against women are also often those with little or no reporting on the phenomena. A recent UNODC and UN Women (2024) report on global estimates of femicide indicates that, at best, only around 75 out of 193 UN member states ever reported data on the killing of women and girls, with even this number declining by half following the COVID-19 pandemic. Currently, fewer than 10% of countries in Africa and Oceania and less than 20% of countries in Asia have available data on femicide (UNODC & UN Women, 2024, p. 28), despite Africa and Asia reporting the highest numbers of female murders, and Oceania recording the third-highest regional rate of female killings.

Acknowledging the gendered nature of data also considers that data collection can inherently exclude or fail to acknowledge women and/or gender as a category worth recording or incorporating (Ahn et al., 2024). Walby et al. (2014, online) highlight a number of obstacles in identifying violence against women “as a robust object for analysis,” which include not only differing theoretical assumptions about gender and equality, but also extend to how violence is defined (e.g., as specific acts, versus a continuum) and how these definitions vary across jurisdictions; what specific [gendered] information is considered relevant when recording an alleged crime; and even what “units” are used when recording crimes—for example, data on the number of incidents versus data on the number of victims—which makes it challenging to unify and analyze systems of measurement. The net effect of this gender/ed data gap is a framework of public knowledge where “violent crime against women is routinely made invisible in the public sphere” (Walby et al., 2014, online), and where missing (or inaccurate) data inhibits the ability to identify where such violence occurs, who is most at risk, who are common perpetrators, and how to develop effective responses that can not only identify but also prevent violence (Maina & Rooney, 2024).

More than just a feminist assessment of counting and classification, which observes that “what gets counted counts” (D’Ignazio & Klein, 2020), the push to address and fill these data gaps has also contributed—particularly in the global south—to a paradigm where “counting counts,” that is, where the measurement and quantification of violence against women has become essential not for understanding the phenomenon itself, but for driving and supporting institutional and state requirements for what is often described as research impact, or “data strong enough to make policy makers pay attention and act” (Bradley et al., 2020). In these contexts, violence is commonly requalified as an economic and public health issue (Brodie et al., 2023a) in which “the prevention of [violence against women and girls is seen to be] central to the economic development of developed and developing countries alike” (Vetten, 2016, p. 295) and where both the causes and consequences of violence against women are repositioned within economic frameworks that attempt to convert issues like women's rights and social justice “into neo-liberal objectives and strategies” (Vetten, 2016, p. 292) by rewriting “all aspects of existence in economic terms” (p. 295). In a study of post-rape care centers in South Africa, Vetten (2016) identifies how these conditions resulted in significant shifts in both private and state-sector funding for work addressing violence against women, and also contributed to changes in donor (funding) behavior as a result of revised priorities. Walby (2023, online), looking at femicide as a “global indicator of gender inequality and violence,” also acknowledges the tensions between strategies centered on “economic growth” and other, broader types of development as markers of progress, observing it within the conceptualization of gender and violence indicators adopted by the United Nations, particularly between the UN Millennium Goals and later Sustainable Development Goals (SDGs) —with Walby noting that when the United Nations “endorses a particular indicator, other entities adapt, even when they are not legally obliged to.” While Walby and others make the argument that pressure and interventions guided/initiated by feminist economists, feminist scholars, and feminist civil society organizations have succeeded in expanding and including gender dimensions, particularly in the SDG indicators, the imposed “universality” of these indicators continue to privilege an “instrumental rationality” (Esquivel, 2016; Vetten, 2016) which embeds these forms of neo-liberalism.

From this perspective, it is important to continue to identify the co-existence of multiple conceptualizations and actualizations—what Walby (2023) describes as a “fluid ecology of multiple contesting and cooperating epistemic systems”—of data, data collection and data analysis. These represent not only different measures, measurements, and ideologies, but also reveal parallel or even competing hierarchies and relations of power in relation to each other and to the human object—relationships Foucault (2003) conceived of as “biopolitics” and “biopower,” or a sovereign's “surveillance of ‘[p]ropagation, births and mortality, the level of health, life expectancy and longevity’” (Foucault in Führer & Eichner, 2015) in order to optimize life. Mbembe (2003) later described this sovereignty as that which is able to “exercise control over mortality and to define life as the deployment and manifestation of power” (2003, p. 11; emphasis mine), before proceeding to ask, instead, how, in these contexts, the “right to kill, to allow to live, or to expose to death” are exercised, what “place is given to life, death, and the human body (in particular the wounded or slain body)” and how are these “inscribed in the order of power” (2003, p. 12). Mbembe described this approach as “necropower” and “necropolitics,” terms that emphasize a sovereign power's core mandate (particularly in the colonial and post-colonial world) as the “right to kill” (droit de glaive; p. 17), where biopower is used to “regulate the distribution of death” (p. 17), particularly in relation to race, and necropower acknowledges the specific “subjugation of life to the power of death” (p. 39).

In the context of demographies, epidemiologies, and epistemologies of violence, we can extend these ideas into what might be called “necrodata,” a framework for inquiry into how phenomena and power-relationships of death are made visible (or obscured) through choices related to quantification, data, and statistics, and how these figures are explained or narrated. Necrodata asks what is counted as “death,” whose deaths are counted, and in which contexts and in which time/s; it also considers how the ways in which we count and recount death reflect specific socio-political contexts and even disciplinary approaches, and how these create specific meaning. This article attempts to demonstrate this approach, of necrodata, through an exploration of historical and contemporary ways of counting and recording the incidence of female homicide in South Africa (one of the countries with the highest femicide rates in the world), to show how different forms and formats of data are collected and applied to construct specific narratives around the [dead] female body. Similar frameworks of necrodata can potentially be applied to other data and types of deaths, and even perhaps to other (non-fatal) types of violence. Mbembe, for example, says death can be understood as the “violence of negativity” (2003, p. 14) and argues the human being only truly becomes a subject through struggle and confrontation with this violence. Other scholars, such as Eberhardt et al. (2016) and Ogude-Chambert (2025) have also described the idea of being “deathworthy,” particularly in relation to black male bodies, which includes both the consignment (future death) or abandonment (normative death) of these men to such a “fate,” demonstrating how indicators of “death” exist along a spectrum that often begins long before the coup de grace.

Separate Development, Separate Measurement

While death is inescapable, not all death is acknowledged. In many conditions, necrodata occupies a space somewhere in between biopolitics and necropolitics—which might be signified by the presence or the absence of data, particularly in the context of colonial and post-colonial geographies. In South Africa, the 20th century apartheid state's ¹ obsession with race and separate development (which was preceded by nearly three centuries of similarly prejudiced colonial rule) sought overtly to use techniques of scientific “measurement” to enable an “objective” basis for segregated social control. But the state's “mania for measurement” (Posel, 2000) expressed itself in what was primarily a preoccupation with issues of population growth, influx and labor control of black Africans, leading to a surfeit of data in some categories, and a dearth of information in others, including mortality. The “rituals of statistical measurement” were used to enable and enforce stringent controls over populations and individuals who had been “enumerated,” and created power structures directly related to routine “activities of counting” (Posel, 2000, p. 134). For example, the Population Registration and the Group Areas Acts of 1950 legislated, respectively, the racial classification of every South African subject, and the segregation of urban residential areas by race. Coupled with the introduction of Pass [identity] Books for “non-whites,” these were used to control movement within and between spaces. But comparatively little other everyday data was collected about black Africans, Coloreds, ² and Indians, and demographic information about people living in “homeland states” (separate ethnic states implemented by the apartheid government in an attempt to remove black populations from white urban areas) was excluded entirely until their re-integration in the 1990s. Data on crime and violence during apartheid was also collected and reported through uniquely racialized lenses—so, not just murder statistics but, rather, how many victims of each race had been assaulted or murdered by perpetrators from the same race versus perpetrators from another race group. A news report from October 1972 in the Rand Daily Mail, a large English-language newspaper published in Johannesburg, explained that, between 1967 and 1968: “16 Whites were murdered by Whites, seven Whites were murdered by Coloureds and 36 were murdered by Africans. One Coloured and six Africans were murdered by whites. But 1 286 Blacks were murdered by Blacks” (Menge, 1972).

The state, simultaneously attempting to hide the growing extent of violence across the country (e.g., by suppressing statistics in provinces with high murder rates) while actively seeking to position the bulk of the hostility as “black on black” aggression, put little effort into documenting anything that could not be categorized as either political or racially atavistic. Because of the social conditions, early independent attempts at studying violence, too, tended to focus almost exclusively on political violence in the context of apartheid, largely avoiding recognition of sustained and indeed far more numerous incidents of inter-personal violence (Butchart & Brown, 1991). With the singular exception of rape (see Bridger, 2023; Bridger & Hazan, 2022), reports on violence against women—particularly Black African women—were infrequent, and often described in veiled language, obscuring intimate partner violence in particular. Here again, burgeoning narratives on the prevalence of rape in South Africa in the 1970s and 1980s were promoted through different concerns, the first being a growing groundswell of global feminist activism which sought to redefine and gain greater recognition for the systematic and often invisibilized sexual violence experienced by women; the second being a “moral panic,” linked to features including rising Black student activism, school boycotts and the sustained use of sexual violence against women as a form of performative masculinity in South Africa's Black townships (Brodie, 2020). This latter version in particular suited the state's desired outcomes, seemingly underlining not only the extent of violence committed by Black men (this time against black women), but also appearing to support criminologists’ and social scientists’ assertions that “detribalization” and urbanization had created a “sub-culture of violence” within contemporary Black African culture—rather than identifying a culture of violence within the apartheid state itself.

The low levels of reporting on violence against women (Brodie, 2020, 2023) must also be seen in the context of the extent of violence being experienced by primarily black men in South Africa—a status that continues to this day. A news report, again in the Rand Daily Mail, from July 4, 1980 (Diktola, 1980) proclaimed that the murder rate in Soweto, the country's largest Black township some 20 km south of Johannesburg, was more than three times that of the “concrete jungle” of New York. Soweto, which had a then-estimated population of anywhere between 750,000 people and 1.4-million residents (again, the “counting” of the apartheid state never quite managed to achieve its ambitions) had reportedly recorded at least 500 murders in the first 6 months of that year, of which it stated only 23 victims were women (representing 4,6% of the total). In retrospect it seems likely that this proportion of female homicides was understated. Over the course of multiple subsequent national femicide studies (see next section), the total number of female murders has never made up less than 12% of all homicides in South Africa, with the majority of murders being perpetrated by current or former intimate partners (i.e., killings less likely to be directly impacted by political conflict). In separate studies, which estimated mortality rates in South Africa in the late 1980s, Dorrington et al. (1999, p. iii) showed that during this same period, just “44% [of] female deaths were recorded for the Black population group,” with only slightly more than half of Black male deaths similarly being recorded. This suggests that, when the very first South African femicide studies were initiated in the 1990s, the “discovery” of a much larger proportion of female deaths was not an artifact of the changed times but rather that, in order for the problem to be counted, first it had to be defined.

Femicide in South Africa

In 1991, a new Women's Health Project (WHP) was launched against the backdrop of political negotiations leading up to the country's first democratic elections [which took place in 1994]. The WHP was run by a “small group of academics and community health workers” (WHO, 1993, p.7), and its goals were to “develop and lobby for new policy proposals” (Klugman et al., 1995, p. 122) related to varied aspects that were determined to have an impact on women's health, including violence (although, at this stage, it should be emphasized the primary concerns with violence were with family or domestic violence, sexual assault, and the need for survivor support structures including women's shelters). In December 1994, the WHP organized a Women's Health Conference, held in Johannesburg, where 13 policy working groups were established, including one focused on violence, which produced a series of recommendations, one of which stated that research on violence against women was “necessary to understand the magnitude of the problem better and the numbers of women affected.” A similar approach was also supported in the articles of the new Women's Charter for Effective Equality, which was adopted by the country's first democratic parliament in 1994, and which explicitly listed violence as both a pervasive feature of South African life, and a threat to women's right to equality, explaining that “all women live under the threat of or experience violence.”

Both of these local initiatives should be viewed within a larger global context, where the need to name, reject and prevent violence against women was becoming increasingly explicit in doctrine, including the Declaration on the Elimination of Violence Against Women, adopted by the United Nations General Assembly in 1993, and the establishment of roles such as the UN Special Rapporteur on Violence against Women in 1994. These laid the foundation for the inclusion of violence against women in two of the platforms for action emerging from the 1995 UN World Conference on Women in Beijing—which also marked the first time the newly-democratic South Africa had participated in the world conferences, having been readmitted to the UN in 1994.

Both the WHP and Beijing Conference played an influential role in prompting early South African studies on what was increasingly being described as “femicide” or the murder of a woman; the use of the term “femicide” itself, as presented by sociologist Diana Russell in the late 1970s (see Brodie, 2024), had re-emerged particularly in feminist circles in the early 1990s following the publication of Russell and Radford's edited anthology Femicide: The Politics of Woman Killing (1992, Twayne Publishers). The early South African studies looked at intimate partner femicides in Johannesburg (Vetten, 1995, 1996), based on information obtained from newspaper record and court inquest reports, and at the epidemiology and injury patterns of rape and rape-homicide in Johannesburg and Cape Town, respectively (Martin, 1999).

Vetten's research in particular, which had identified 45 cases of intimate femicide in Gauteng in the pilot study, and 118 cases over a two-year period (between 1993 and 1994) in the second study, was “used extensively in advocacy campaigns” (Mathews et al., 2004)—not just because of its data but also because of the compelling way in which the data was presented. The studies contained important information: 55% of women had been murdered by men “known to them,” a quarter of murders had been motivated by a women's decision to end her relationship with the perpetrator, and a high proportion of perpetrators were policemen (Vetten, 1995). But the most often-mentioned finding was that, during the (second) study period, “a woman was killed by her partner every six days” (Vetten, 1996). This femicide “rate,” which Vetten noted at the time was likely an undercount, was specifically referenced in forensic pathologist Lorna Martin's (1999) subsequent research, and in the country's first national femicide study released a few years later (Mathews et al., 2004). This highlights an important feature, easily overlooked when considering how numbers and data are used in relation to violence/death, populations, or risk in the construction of narrative. In health sciences, morbidity and prevalence will often be discussed or described as a rate per 100,000 people in order to create information which can more easily be compared across populations, regions and times, and to avoid figures which are fractionalized and awkward to work with. So, a study might determine an intimate femicide “rate” of 8.8/100,000 women (as was the case for intimate femicide in South Africa in 1999). But this form of data representation is not intuitive for human “sensemaking” (Fjørtoft & Lai, 2021), nor is it particularly practical or compelling for storytelling (Ren, Wang & Zhao, 2023). By contrast, a woman being murdered “every six days” is something that is both accessible, and also a little shocking (what Boonzaier (2022) describes as the “spectacularization” of femicide). And features like this were extremely important in order for “data” to cross over from public health to public interest. When the results from the first national study of female homicide in South Africa (using 1999 data) were released in 2004, the initial policy brief from the South African Medical Research Council, which had found a dramatically higher femicide rate, was titled: “Every six hours a woman is killed by her intimate partner.”

Martin's (1999) rape/rape-homicide study, which included data from 152 clinical forensic cases (non-fatal incidents) and 59 forensic pathology cases (fatal incidents), had been motivated in part by what the author, a former District Surgeon, had observed as a systemic failure to recognize and manage the growing numbers of reported rape cases in South Africa, and the lack of official data being collated on rape-homicide. A further concern was the prevalence of intimate partner or domestic violence, particularly in fatal cases—an issue highlighted, Martin said, by Vetten's work. In the absence of police data on rape-homicide, Martin developed a research method that started from available mortuary data, including a medico-legal autopsy report and Death Register number, and worked backwards to identify a police station and case number, and then interviewed investigating officers to obtain more information about the salient features of the case such as the perpetrator status in relation to the victim. A similar research design would be used in the country's subsequent national femicide studies, the first of which was performed between 2002 and 2003, collecting mortuary and police docket data from female homicide cases recorded in 1999. The need for a retrospective study was driven by practical considerations: it would take a certain amount of time to allow for the actual recording of unnatural deaths, and also for there to have potentially been some progress or at least change in the status of the case within the police investigation and justice system; it could typically take months or even years for a matter to be heard before the court. In this way, necrodata can also be considered to be an artifact that is potentially affected by time, where data is understood as an indicator of a phenomenon (and not the phenomenon itself), and where the passing of time can erase some data, reveal others, or even change the nature of phenomena we wish to seek indicators for.

The release of the results of the 1999 South African national femicide study (Mathews et al., 2004) established a number of important categorizations and benchmarks, which would be used in much the same form across (to date) three subsequent femicide studies, spanning a period greater than 20 years, for the 2009 and 2017 calendar years (Abrahams et al., 2012, 2022), and for the 12 months between April 2020 and March 2021 (Abrahams et al., 2024a; the adjusted time frame was in order to study the impact of the COVID-19 pandemic and its associated lockdowns on femicide rates). The 1999 femicide study provided an initial working definition of femicide (rather than a theoretical one), which stated that femicide was the same as all female homicide—in contrast to the Russell school of thought, which insisted that femicide should only refer to a woman who murdered by a man because she is a woman (Russell, 2012)—and distinguished only between intimate partner killings (where the perpetrator was a current or former or rejected intimate partner/lover, including same sex partners) and non-intimate femicides. This approach demonstrated what Campbell and Runyan (1998, pp. 347–348) had described as a more “empirical perspective,” which relied on available data without stretching to derive an inference about motive. A second important feature about the femicide study's data collection was that, for its sample, it included the murders of “women aged 14 and over” (Mathews et al., 2004). Vetten's earlier work had identified a fairly large proportion of femicide victims who were younger than 20 years of age, and had commented that future research, in addition to expanding the definition of “intimate partners” beyond only marriage or formal relationships, should consider that young women began entering into relationships with (usually much older) men at quite an early age. Subsequent national homicide studies which included data collection on child homicides (Abrahams et al., 2012; Mathews et al., 2012) indicated that, indeed, both males and female children's risk profiles changed distinctly upon entering adolescence; where for boys the homicide risk increased as soon as the individuals started to participate in masculine types of social activities and socialization, for females the risk for being murdered increased once they started to date or enter into (or be seen as eligible to enter into) relationships with men. The 1999 femicide study produced a weighted sample [weighted to account for the mortuary sampling] of 3,798 female homicides, of which data could be collected on 3,296 victims [the balance meant that either dockets or information was missing], from which it was determined that intimate femicides and non-intimate femicides made up 41% and 40.5% of the sample respectively, with insufficient information (such as no perpetrator identified) in the remaining 612 cases. At the time, this reflected the highest rate of intimate femicide (victims per female population) ever recorded in the world.

The results of the 1999 national femicide study were discussed, expanded and analyzed in a number of subsequent publications, each utilizing or producing slightly different counts or necrodata (through focusing on narrower or specific criteria), including on intimate femicide-suicide (where the perpetrator of the femicide commits suicide after committing murder; see Mathews (2005) and Mathews et al. (2008)); on rape-homicides (Abrahams et al., 2008); and a study on the role of firearms in femicide (Abrahams et al., 2010). A number of unrelated studies on femicide were also produced during the period between the first and second national femicide studies, including studies on fatal strangulation (Suffla et al., 2008) and adolescent victims of homicide (Swart et al., 2018), which used mortality data from the country's National Injury Mortality Surveillance System (NIMSS), a parallel but separate surveillance system that stopped collecting national data after 2005. In 2009, the South African Police Services also published a report (through its Independent Complaints Directorate or ICD, now the Independent Police Investigative Directorate) investigating the high numbers of femicides involving members of the police services.

The over-representation of police members as perpetrators of femicide is significant, and requires particular attention. The South African Police Services (SAPS) are, in most senses, the official “frontline” for both the prevention and the reporting/investigation of unlawful killings, including female homicides, and for the arrests of suspected killers, these actions marking the compulsory entry point for any subsequent criminal justice proceedings to even be initiated. Multiple South African femicide studies over the past 25 years (e.g., Abrahams et al., 2024), read together with the annual and quarterly crime reports and press statements issued by SAPS, indicate that, despite formal national instructions to improve their handling of violence against women in general and domestic violence in particular, institutional police responses to female homicide can be considered as falling anywhere along a spectrum ranging from “well intentioned” to “ambivalent,” “indifferent,” and even “obstructive”—but rarely as proactive or effective. This relationship can also be explored through the police's use of data about femicides.

Policing Femicide

A useful starting point for understanding this relationship is to explore how the police collected, managed and described data about female homicides, which should in turn be considered within a broader professional context of deliberately obscure and inconsistent reporting on data from the justice system in general (Brodie, 2023). These issues can be understood in part as a result of the struggle to transition from a legacy racist justice system to a more democratic one, but it is also a dysfunction of the post-1994 democratic state—and an attempt to cover up broader state failures since the end of apartheid. In the framing of necrodata, this is a potentially valuable consideration: the extent to which even seemingly democratic states may wish to exert control over forms of human data, in order to not only manage populations (Breckenridge, 2014) but also to foster a strategic narrative which demonstrates that the state is at least in control of the data, if not necessarily in control of the phenomenon.

As mentioned above, data on all crimes including murder were unevenly counted in South Africa before the mid-1990s, and little to no public data was generally provided on the proportion of women affected by any crime other than rape. In mid-2000 (during a period when crime levels were still extremely high), the police minister abruptly suspended the publication of national crime statistics, only lifting the moratorium a year later, citing flaws in the police's data capture processes and systems. In February 2002, the police's “crime code list” was also amended in a way that made it possible to capture certain victim attributes (Minister for Safety & Security, 2003), and the ministry's annual report for that year indicated that, of the 21,533 reported murders, “14.1% of all murder victims were female,” together with an age breakdown of these victims.

The inclusion of this data was more than just a feature of new coding categories, it reflected a specific strategic trajectory that would align SAPS with repeated national government objectives to prioritize crimes against women and children. From 1996, the police services had managed crimes against women and crimes against children as part of the same cluster of units dedicated to investigating family violence, child protection and sexual offences (the Family Violence, Child Protection and Sexual Offences or “FCS” units, operating separately from, say, the “serious and violent crime” unit). But the police service saw its mandate/role for the FCS units as largely being unable to prevent the kinds of crimes they investigated, as they happened primarily within family circles. Instead, the units were supposed to “make the public aware of the existence of these crimes” (Minister for Safety & Security, 2003, p. 62) and encourage the public's role in “preventing and combating” them, echoing earlier similar statements from then-president Thabo Mbeki.

The same year that SAPS began to report data for female homicide was also, not coincidentally, the first year that the government began to implement its national “16 Days of Activism on No Violence Against Women and Children” program—a feature mentioned several times in the police service's annual report. But in real terms there seemed little urgency to engage differently with crimes against women. In a presentation to Parliament's Safety and Security Portfolio Committee a year later, representatives from SAPS told the committee that “violence against women and children usually fell under the categories of other crimes like rape, murder, assault, etc” and that “[s]eparating these out would be a tedious and time-consuming process.” (PMG, 2004) Nonetheless these numbers were eventually differentiated, and from around the mid-2000s at least (not all of the annual police reports or crime statistics reports are available anymore for the early 2000s) reports and figures on “crimes against women,” including the murder of women, became more readily available in the annual release of crime data. Perhaps related to this, the SAPS annual financial statements for 2007/2008 indicate the police service had received a donation from the European Union of around R250-million, over a nearly four-year period (from 2004 to 2007), to combat crimes against women and children.

SAPS data on the murder of women (femicide is not an “official” crime code or legal category, although the term is used in official reports) was—and still is—reported for women aged 18 years and older, unlike the Medical Research Council's national femicide studies’ criteria of 14 years and older—and is usually grouped with a separate table reporting on crimes against children (of any gender). The distinction created some initial confusion. In a 2008/9 annual police report, the number of reported crimes committed against “women” versus “children” had to be retrospectively adjusted because there was a mix-up on the crime management system between victims aged “18 years plus 364 days” (who had been counted as children in previous reports, but should have been documented as adult victims) and the view, taken by SAPS Crime Information Management, that a “child” should be a person who had not yet had their 18^th birthday (SAPS, 2009). In the same report, the results of a retrospective docket analysis for 1,348 murders committed between 2007 and 2008 (running from April 1 to March 31 corresponding to SAPS financial reporting rather than a calendar year) were also shared, in which a number of differences were observed related to the gender of the victim, particularly with regard to what SAPS had recorded as a motive, with female victims most-likely to be murdered during a “domestic-related” context, and males more likely to be killed as a result of a “misunderstanding/argument” (SAPS, 2009, p. 11). SAPS would only start reporting on “domestic violence” as a category several years later, around 2015/16 (Brodie, 2023). Without taking the discussion out of sequence, it is perhaps worth pointing out that perhaps because “femicide” originated as a largely socio-political term, it was erratically applied within both police and statistical services in South Africa, often with problematic results. For example, a request to the police statistical information services for data on femicide might be met with a response indicating that no such records existed as, citing Russell's definition, this required proof of motive (a woman who was killed “because she was a woman”), and that information was not available on the case management system (Brodie, 2019). The national statistical body, StatsSA, also produced a problematic report on “Crime against Women in South Africa” (Stats SA, 2018), repeating Russell's definition of femicide, and declaring that “‘femicide is a term that is often misused by activists and the media’ and that the ‘killing of females simply because they are females is a rare phenomenon in South Africa’” (Brodie, 2020, p. 20), before going on to determine that the brutal rape and murder of lesbian soccer player Eudy Simelane in 2008 was somehow not a femicide. Similar problems exist with phrases like “gender-based violence” which are narrative rather than empirical terms, and which may mean different things to different departments, depending on what is being considered or counted.

There were a number of other relevant contextual features that should be considered when looking at SAPS data on female homicide for the early 2000s. One was that, during this time the organizational structure of the FCS units that were supposed to deal with many of the crimes committed against women had been substantially changed (together with a number of other attempts at restructuring the struggling police force). Instead of operating as dedicated investigative divisions, the FCS had been moved first into another general unit, and then decentralized across the country. This, according to the police's own assessment as well as independent reviews undertaken by NGOs, had weakened the services provided by the units, had demoralized and demotivated existing FCS officers, and resulted in the deployment of many new officers who had insufficient training or experience to manage the requirements of the specialized unit (Rapcan, 2009). The police services also had a gender equity problem in terms of its workforce. At the time SAPS started to report national data about female homicides (c. 2002/3), just 27% of its personnel, from commissioners and superintendents to commissioned and non-commissioned officers, were women (Minister for Safety & Security, 2003).

It is almost certain that the gender makeup of the police service influenced the ways in which violent crimes against women were—or were not—recorded and investigated. A detailed analysis of the 1999 femicide study data, published in 2009 (Abrahams et al., 2009) commented on the significant amount of missing data on reported female homicides: dockets had disappeared in more than 6% of cases, with no police case found at all in 7% of female homicides identified in the mortuaries. In the first publication on the study (Mathews et al., 2004), the brief stated that dockets “frequently revealed that cases had not been properly investigated and leads had been inadequately followed up,” and that there were inadequate systems in place to file or track the movement of a docket, nor were backups made. In a later paper, the authors (Abrahams et al., 2009) commented that while it was not possible to know what data was absent as a result of the missing/incomplete dockets, it was possible “there were biases in the missing dockets” (p. 553), adding that not only was it common for money to change hands in exchange for dockets to disappear, but that this may be “more common with cases of murder where there was more police sympathy with the perpetrator.” The researchers also noted their “perception when collecting data that the deaths of [black] African women were often not thoroughly investigated, nor were known perpetrators pursued” (p. 554), reflecting historical inequalities in public service, and the historically low status of black women in hierarches of gender, race and power. This speaks to a persistent feature of the gendered and racialized nature of necrodata (and indeed most other data related to who is affected by or at risk from violence/crime/harm) which also highlights how citizens (and of course who is counted as a “citizen”) become visible or remain invisible to governments through data.

Related to all of these features was the SAPS's problem with femicide within its own ranks. A case study of femicide involving members of SAPS, published by the official oversight body, then the Independent Complaints Directorate (ICD) in 2009, reviewed 49 cases of female homicide committed by SAPS members between 2004 and 2007. The investigation was prompted for the large part by the realization that many of these murders had been committed using the officers’ own service weapons. This, in turn, had been specifically problematized (for the state) by a court case in which the Minister of Safety and Security had been sued by a widow whose (policeman) husband had attempted to shoot and kill her, in January 2003, and who had then proceeded to take his own life using the same service weapon—despite the fact that the woman had previously laid charges against her husband at the station where he served, which should have resulted in his access to his service weapon being removed. The data from the ICD study, which ultimately focused on 30 cases where information was available, showed that the majority of victims (where the perpetrator was a SAPS member) were married, were killed in their own homes, and were killed by the perpetrator's service firearm. Less than a third of cases were linked to formal reports of previous domestic violence (which correlated largely with the 1999 femicide study results, where just over 30% of intimate femicides had a reported history of IPV). The ICD recommendations included expanded support (for employees) within the organization, encouragement for female partners to report domestic violence at an earlier stage, and the suggestion that the presence of firearms in residential homes should possibly be reconsidered—although it was acknowledged this would create an additional administrative burden (to keep service weapons at the police station).

In the second national femicide study (for murders that occurred in 2009, with data first published in 2012 onwards; see Abrahams et al., 2012), while there were substantially fewer cases that were missing within the police system (i.e., no docket or not in the system at all), there had been no improvement in the recording of previous incidents of intimate partner violence, and also no change in the conviction rate for intimate partner killings. By the time the third national femicide study was conducted, for the 2017 calendar year (Abrahams et al., 2022), this proportion had remained largely unchanged, with only 29% of intimate partner femicide case dockets including any information on a history of partner violence. The 2017 study also noted that there had been a rise (from 1.6% to 15.7%; Abrahams et al., 2024) in the number of femicide cases where there was no information available, and an increase across all three study periods in the proportion of cases where the police had not identified a suspect—a feature in 30% of femicide cases in 2017, and 44% of cases in the fourth national femicide study conducted for cases in 2020/21 (although in the last instance it was considered that COVID-19 restrictions directly impacted on the ability of police to investigate matters). Following the release of the 2017 femicide data, the studies’ authors observed that the “decrease in the number of cases in which a perpetrator was identified and the increase in the number of missing dockets are evidence of a decline in the quality of case investigation and management” and that the systematic “problems with policing and undermining of the criminal investigation capabilities are well documented and reflect poor management and corruption” (Abrahams et al., 2024). This highlights a critical aspect of the collection / failure to collect data related to the killings of women, which should also be considered as necrodata: that, even if a matter is reported to the police (and murders tend to be the most-reported crimes), unless the matter is properly investigated and sufficient credible evidence is collected by the police for the prosecuting authority to charge and prosecute a suspect, there can be no court case, and no potential conviction. Even if the COVID-19 pandemic-period data is briefly excluded as anomalous, over the past two decades—and this only for cases where the perpetrator or suspect is known and the case is in the system (i.e., perhaps 60% to 70% of female homicides identified at mortuaries)—only a third of intimate partner femicide cases and an even lower proportion of non-intimate femicide cases end in a conviction [again, another form of necrodata]. This figure is still significantly higher than the average police success rate for all homicides (85% of which are male victims): recent data shows that only around 15% of murders in South Africa are ever solved (Faull & Bruce, 2023). But if one considers that the majority of female murders are committed by a male intimate partner, the police services’ sustained failure to identify a suspect in so many cases speaks to a more active obstruction of justice. Similarly, the continued failure of the police to collect data related to previous incidents of violence in intimate partner killings is one of the direct reasons for the relatively low conviction rate. In non-fatal cases, the police's own data has shown that 87% of offenders arrested for domestic violence are men, and more than half of such offenders have existing criminal convictions, both for other forms of contact crime as well for contraventions of the Domestic Violence Act (Brodie, 2023).

A Note on Necrodata, Femicide, and Gun Deaths

There are around three-million legally registered firearms in South Africa, licensed to some two-million gun owners (Lamb, 2024), with an unknown number of illegal weapons circulating (many of which start off as “legal” weapons; in 2022/23 over 8,200 guns were reported as stolen or lost). Guns account for over 40% of murders annually (Hussain & Morrison, 2025), and were the most common manner in which women were killed in 2020/21, according the fourth national femicide study (Abrahams et al., 2024b). The same study also showed that, after a period of decline in the proportion of firearm deaths between 1999 and 2009, the number of women killed by guns had been consistently increasing. The earlier decrease in gun-related femicide (and indeed a decrease in perpetrator suicide) was considered to have been a direct result of changed gun control laws in South Africa which were passed in 2000 and implemented from around 2004; since that time, however, there has been an increase in the proliferation of illegal firearms, and a deterioration of administration and oversight of gun controls, which has contributed to a resurgence in gun-related deaths. The positive effects (i.e., reduced gun deaths) of gun controls have been extensively documented in multiple studies (see for example Matzopoulos et al., 2016, Matzopoulos et al., 2018).

This is mentioned specifically in the context of femicide as the right to gun ownership and the role of firearms particularly in relation to gender-based violence remains highly politicized, and this has played out in ongoing public debates as to how gun data should be interpreted—with pro-gun lobbyists advocating for more women to be armed, in response to rising gun deaths… even though multi-year evidence clearly demonstrates that the majority of gun-related femicides are perpetrated by an intimate partner, usually with a legally owned firearm. That is, the proposal seems to be that if both spouses are armed this will somehow level the playing field. Again, local femicide studies show this is unlikely to be the case as guns are generally used to threaten or inflict harm in the context of ongoing domestic violence which typically includes varied levels of psychological as well as physical abuse, and extensive coercive control. Previous South African studies have shown that women who are killed by an intimate partner most often die from a single gunshot wound, mostly to the head or the face (Abrahams et al., 2010).

The role of firearms in both intimate and non-intimate femicide is consistently demonstrated through data, but is unfailingly contested by political and civil society initiatives advocating for gun rights. This is a reminder that even when seemingly indisputable figures are available about causes of death, this data can be rejected or modified on the basis of personal beliefs. Although not within the scope of this article, similar comments can be found with respect to the role of alcohol in the prevention of femicide and violence against women; during South Africa's extended COVID-19 lockdowns, the sale of alcohol was prohibited during specific periods (some of which coincided with stay-in-place lockdowns and curfews, which were retained to an extent even after alcohol sales were re-permitted). Data from the fourth national femicide study (Abrahams et al., 2024b) clearly show that the banning of alcohol sales had a clear and dramatic impact on reducing intimate partner femicide cases.

Necrodata and Datafication

There is an assumption, perhaps even an implicit expectation in the recognition of the “gender data gap” that the right solutions to the problem will emerge through making, building or populating better data. Nowhere is this mindset more evident in the current insistence that computational methods can “solve” almost any societal and anthropocenic problem, from poverty to the impacts of climate change. The datafication of violence against women and femicide has the potential to add rich and important detail to the mosaic of knowledge about the context and pathways of such violence, but there is also little evidence as yet that this is helping to reduce the prevalence of femicide—if anything, the act of creating data on femicide seems to have assumed a more prominent role as an indicator of “action,” allowing countries like South Africa to claim progress simply by counting the growing number of bodies.

Datafication, often closely linked to external, ambiguously instrumentalized indicators such as the SDGs, is also proving an effective vehicle for fundraising and for generating short-term popular and political support for the civil society movements and non-governmental organizations that typically rely on donor funding.

In 2022, a report published by non-profit health journalism outlet Bhekisisa presented a review of various programs, actions, and strategies aimed at combatting gender-based violence and femicide in South Africa (Ludolph, 2022). One of the “big wins,” the story said, was that a “more accurate picture” of the issue was emerging thanks to better, synchronized data—citing, in particular, a new “GBVF [gender-based violence and femicide] Statistics Dashboard,” funded by an also-new GBVF Response Fund, which was intended to utilize live SAPS crime information and compare this against the population of each province to identify crime “hot spots” by crime rate per population instead of by incident numbers alone. According to the article, results extracted from the dashboard had already assisted in identifying and providing funding to 110 community organizations, and identified four additional initiatives that would distribute money to support a further “52 smaller groups” (Ludolph, 2022, online), although neither the methods used to identify these communities nor the organizations selected for funding were described in the article. A review of the public GBVF Dashboard shows only limited insight into crimes, mapping publicly available police data at a provincial level (which is generally considered too broad to derive any indication of crime “hotspots”)—even though similar data had already long been processed and presented through existing tools like the Institute for Security Studies’ Crime Hub (crimehub.org), which has mapped crime at a police station level for well over a decade. At the time of writing, the GBVF Dashboard appears to not have been updated since 2022, suggesting either a loss of funding, or a loss of interest. It was deployed and promoted at a time when “data dashboards” had become extremely popular, primarily due to their use as visualization tools during the early period of the COVID-19 pandemic (Gleeson, Kitchin & McCarthy, 2022), where the numbers of daily new cases, deaths and recoveries could be tracked through health surveillance reporting. However, one of the central considerations of data about death, and which is fundamental in the mixed-methods framing-approach proposed by necrodata, is that even though we often refer to “epidemiologies of violence” or describe violence in disease-like terms (for example, the “shadow pandemic” of gender-based violence), neither violence nor death from non-natural causes behave in the same way as communicable or non-communicable illness, and cannot generally be mapped in the same way.

In the years preceding the COVID-19 pandemic, South Africa's femicide rate had settled at somewhere between 11 and 13 incidents per 100,000 women, which was incredibly high by global standards but marked a significant improvement over the rate of 24/100,000 recorded in 1999. The total number of femicides had also declined, from an average of eight women murdered each day in 1999 to between six and seven femicides a day by 2017 (a daily average does not take population growth into account, so the 2017 figure represents an even greater achievement as the country's population increased by around 11 million people during this same period).

During the first year of the pandemic (2020–21), data collected by the South African Medical Research Council (Abrahams et al., 2024) found that the overall femicide rate had again reduced slightly, likely because of the severe restrictions to people's movement for much of the year—but the intimate femicide rate had increased. This was considered a probable artifact of the stay-at-home restrictions, where women were locked up in the same home with their killers. After the easing of COVID-19 restrictions, however, all female homicide numbers in South Africa (together with all murders) began to increase, and soon substantially exceeded pre-pandemic levels. According to statistics from the SAPS, in 2021/22 there was a 15% increase in contact crimes against women (including murder); in 2022/23 this increased, again, by over 8% year-on-year (SAPS, 2023). During those crime reporting years, femicides in particular had grown from a figure of 2,695 murders in 2019 (pre-pandemic) to 3,198 femicides in 2021/22 and 3,422 femicides in 2022/23 (SAPS, 2020, 2023). By the 2023/24 crime year, SAPS quarterly statistics reports indicated that the rate of increase for femicides had slowed, although the exact number of female homicides seemed to vary (Vetten, 2025); when quarterly crime totals for the year were added, they indicated 3,839 femicides, whereas the official SAPS annual report gave a lower figure, of 3,485. This suggested between 9,5 and 10,5 femicides were taking place every day. A private lobby group called Women for Change (nd), however, surfaced even more horrifying claim: that more than 15 women were being killed every day. This calculation was based on a single SAPS presentation (since removed) on annual crime figures, which reported the total number of women killed in 2023/24 as 5,578, a figure substantially higher than any figures in the police service's other reports for the same year, or indeed for any year in the country's history.

If such a figure had been true, it would have marked an unprecedented 50% increase in femicide numbers, but this was something that was not borne out by any other crime data—even the police's own statistics. While SAPS later unofficially indicated the inflated number in the summary report was probably an error ³ , the police service did not take steps to publish any sort of formal or public correction and, instead, simply removed information on disaggregated data on crimes against women and children while it investigated the issue. While most violence experts took a cautious approach to the single report with improbably high figures (Gun Free South Africa, 2025, p. 2), Women for Change placed it prominently in the center of a large social media campaign, and used it, together with several other possibly inflated crime figures from the same report, to drive an online petition, an award-winning ad campaign, and, in late 2025, a series of nationwide “shutdown” protests that were covered widely in national and international media, which contributed to the South African government declaring femicide and violence against women as a “national disaster” in terms of the country's disaster management legislation. As gender violence experts subsequently explained (Vetten, 2025), the “national disaster” response was politically expedient, but would have little to no real short- or medium-term impact on access to funding or reducing violence. Vetten (2025) also pointed out that the presence of so much conflicting data on femicide—all provided by the same national police services—had the potential of making the data unusable, or at least less-useful to enable meaningful decisions about the allocation of resources, or appropriate response strategies. This highlights perhaps a final feature of applied necrodata, and which can be linked back to Corradi et al.'s (2016) important question as to whether femicide is a useful term. The datafication of femicide in South Africa is, clearly, having little impact on mitigation or prevention, but it is quite plainly being used to achieve other objectives.