Tracing structural racism in Swedish policing: Laws,practices,and technologies of criminalized difference

Abstract

In 2008, the Swedish legislature removed any references to race and racial difference, formalizing a post-racial discourse. This has resulted in both the near impossibility of proving discrimination based on racialized difference, and a formal denial of racial discrimination as a social and structural problem. This article attempts to circumscribe these difficulties by suggesting an approach for tracing structural racism in Swedish policing. We argue that a claim of structural racism in policing cannot be studied by counting self-proclaimed racist officers or through individual experiences, for race is a structure in and of itself. Instead, we claim that structural racism can methodologically be studied through a combined reading of relevant (a) laws, (b) practices, and (c) technologies of policing, since these habitually produce racialized outcomes even in the absence of direct references to race or ethnicity. Reflecting this tripartite approach to the study of structural racism, we analyze data gathered from these three specific sites. We argue that developments in criminal policies and policing practices reproduce and reinforce structural racism as part of the wider penal structure, leading to a reproduction of the targets of punitive policies and police control in the intersection of class, place, race, and ethnicity.

Introduction

This article aims to trace racism in Swedish policing as a structural issue and suggests an approach to this form of inquiry in a context of widespread blindness to race as a structure in and of itself and of a dearth of research on this topic. Our objective is to take up the challenge of addressing this while stressing the need for a broader comprehension of race and racism that views racism as more than merely accounting for interpersonal experiences and attitudes. We view such an approach as crucial to understanding the contingencies of Swedish policing within the contemporary tough-on-crime penal landscape in Sweden. Let us start by providing a glimpse of this landscape.

In a press conference on 1 February 2024, the Swedish Minister of Justice introduced a new police mandate: visitation zones. Visitation zones are part of a broader legislative agenda—known as Tidöavtalet (Liberalerna, 2022)—driven by a tough-on-crime policy that is gradually being implemented by the current right-wing coalition government. Tidöavtalet (the Tidö Agreement) includes proposals such as lowering the age of criminal responsibility, removing young age as a mitigating factor, increasing sentences for violent and drug-related offences, increasing police numbers, building more prisons, and further linking migration management to criminal policy. Visitation zones are time- and place-specific measures that allow the police to stop and search individuals and cars without reasonable cause. Visitation zones have been pursued despite the Equality Ombudsman having raised serious concerns about the risk of discrimination and the institutionalization of ethnic profiling, arguing that such police powers give an official framing to already existing discriminatory police practices (Diskrimineringsombudsmannen, 2024).

The government's press conference (Justitiedepartementet, 2024) did little to alleviate concerns regarding the institutionalization of ethnic profiling. During his presentation, the Minister of Justice named style of clothing as a sufficient ground for police intervention. And during the subsequent Q&A, Martin Melin—member of parliament and former police officer—elaborated on clothing style as signifier of potential criminality: “We could say this… not everyone who wears a fake Gucci cap is a gang criminal, but a lot of gang criminals wear fake Gucci caps” (Justitiedepartementet, 2024: Minute 26:08).

The introduction of clothing style, coupled with Melin's elaboration of its practical meaning, fueled concerns regarding the criteria for being identified as a potential criminal in the eyes of the police. ¹ Is it the clothing or the bodies that wear it that are targeted?

Research (Browne, 2009, 2015; Nguyen, 2015) has already shown that the deployment of hyper-visual factors as truth-tellers about bodies often overlaps with processes that involve the dehumanization and degradation of those bodies—as criminal bodies, threatening bodies, violable bodies, or bodies that are asking for it. As far as institutionalizing ethnic discrimination and profiling through clothing is concerned, racialization has been argued to have moved away from mere references to skin and more towards projecting degrading meanings upon other surfaces (Nguyen, 2015). In the material world of analog visibilities, these are surfaces that cover the skin: hoodies, hijabs, turbans, fake Gucci or Burberry items, saggy pants, etc.

Although referencing the style of clothing seems to exclude skin and its socio-political attributes from the police's suspicion calculus, these proxy surfaces also index similar bodies and offer them as abstractions upon which racial optics can mount their epistemologies of risk and unworthiness. Furthermore, the transition from skin to other surfaces is not an anomaly but a feature of the liberal disavowal of race as a structural issue, which in turn goes hand in hand with the “proliferation of other abstractions as alibis” for race (Nguyen, 2015). In liberal democracies where race as a structure is silenced, references to culture, religion, food, smell, mannerisms, tone, and clothing “train individuals to look upon and react to race and racial cues in a manner that produces the visibility of ‘looking suspicious’ in ways that rationalize forms of violence that would otherwise be unjustifiable” (Obasogie, 2014: 167).

Later, when confronted with arguments pointing to clothing style as a placeholder for race and ethnicity, Melin categorically denied that ethnicity is a factor in police decision-making (ETC, 2024). In fact, the Swedish police have dismissed concerns regarding racism as a structural issue every time similar scandals have emerged. ² When disregarding racism as a structural issue, the Swedish police use a handful of arguments: citing police legislation that specifies equal treatment is a guiding principle, asserting that individual officers do not hold racist views, framing racism as an isolated problem linked to individual rotten apples, and a lack of statistical evidence indicating widespread racism in the police force (see Fundberg, 2022).

Setting aside self-proclamations of good intentions, the lack of statistical data on police racism is a red herring, because Swedish law prohibits the collection of data based on race or ethnicity—which has been criticized by the UN on several occasions (United Nations, 2018). Other arguments, about rotten apples and the non-recognition of racial bias in oneself or colleagues, overtly individualize race and racism as a matter of personal attitudes and behaviors, whereas race, as will be developed below, is a structure not a bad manner. Lastly, understanding structural racism as a pervasive problem among police officers is also misguided, since avidly racist people do not express such views regularly or publicly, and nor is a claim of structural racism a character judgment about all employees of the police institutions.

All this is to say that structural racism is an elusive concept to trace and a difficult claim to argue. In this article, we aim to take the difficulty of accounting for structural racism in police forces seriously while at the same time promoting an understanding of race and racism that goes beyond an individualized interpersonal matter exclusively based on skin color. Here we propose that structural racism in policing can be examined by tracing how race and racial thinking informs, organizes, and operationalizes the state's penal mechanisms and policing via three interrelated approaches: laws, practices, and technologies of policing. We understand structural racism as pointing to the macro level, to the social forces or arrangements, institutions, histories, ideologies, and processes, as well as their interactions, that generate and reinforce racial inequality (Powell, 2008). Methodologically, we contend that structural racism is an overall orientation of a particular institution as manifested in laws/policies that define the strategic objectives of the said institution, practices that operationalize forces to achieve those objectives, and tools/technologies that materialize the said practices.

In the following, we first engage with how racism is regulated and has been researched in Sweden in general and in the context of policing more specifically. We then present our tripartite model for examining structural racism as a tendency within the Swedish police, using Operation Rimfrost (Hoarfrost) as an example to show how elements of law/policies, practice, and technology all point to racial difference as a structural factor in policing. The empirical material used to this end consists of a tracing of policies, documents from the police and official state bodies, statistics, and news reports focused on Operation Rimfrost.

Post-racial Sweden

While contributing to the knowledge gap on structural racism and the police in Sweden, we situate this article within the field of Swedish research on racism and policing, as well as the broader international field of studies on post-racialism and structural racism. Our aim is to build on previous work and contribute to the understanding of race and racism beyond individual expressions and experiences, and in the field of policing, of their role as an indispensable part of the penal structure.

Racism is often understood and researched in terms of experiences and expressions of racism, particularly in the field of racism and policing in Sweden (e.g. Atak, 2022; Haller et al., 2020; Östlund, 2013; Pettersson, 2013; Schclarek Mulinari, 2017; Solhjell et al., 2019; Uhnoo, 2015; Wästerfors and Burcar Alm, 2020). In Swedish studies of racism in other areas, such as healthcare and social service institutions (e.g. Bradby et al., 2019, 2023; Lundström and Wendt Höjer, 2021; Storm, 2018, 2023), urban segregation and housing (e.g. Alinia, 2006; Andersson and Molina, 2018; Listerborn, 2021), and everyday racism (e.g. Kristofersson et al., 2021; Törngren, 2022), we find the same pattern; racism is approached via experiences or expressions. There are however key studies that have addressed racism primarily as structure. Due to the difficulties of obtaining empirical data about racism beyond experiences and expressions, as discussed above, various approaches have been used to showcase structural racism. In studies of structural racism related to the welfare state, labor unions and labor markets, for example, Mulinari and Neergaard (2023) explore the role of trade unions in racial capitalism and how they contribute to the “self-racialisation of white male workers in working life” (Mulinari and Neergaard, 2023: 48) by examining silences related to race in the media and trade union documents in the tradition of investigating colorblindness (cf. Bonilla-Silva, 2006). Grounded in a racial, gendered capitalism perspective, and analyzing political changes and previous literature, Mulinari and Neergaard (2022) argue that Sweden has reached a point where global racial capitalism has outgrown welfare state ideology and is currently dividing the working class along the color line. Using statistical material (limited to the place of birth as signifier for race) and intersectional class analysis, Neergaard (2021) similarly points to the dividing color line in the working class and the racialization of the Swedish class structure and working life. Structural racism has also been central in Swedish whiteness studies. Hübinette and Lundström (2014) analyzed historical material in the form of media, political, and cultural documents from three phases of Swedish nation-building with the aim of exploring how whiteness has been structurally reproduced in these phases. The authors disentangle the intricacies of Swedishness and whiteness and describe Sweden as a “white nation in crisis” suffering from “white melancholy.” Swedishness correlates strongly with whiteness, whereas non-white bodies are understood as non-Swedish. This attachment of whiteness to nationhood makes Swedish whiteness one of the most privileged whitenesses in the Western world, according to Hübinette et al. (2012: 60). Structural racism has furthermore been studied as a legacy of Swedish colonial complicity (Mulinari et al., 2009), while Groglopo et al. (2023) turn experience-based studies on their head by examining how structural racism impacts on attitudes towards immigrants. ³

Similarly, racism in policing, building on empirical material showing racist experiences and expressions, is also understood and theorized as structure. Based on interview methods, for example, Schclarek Mulinari and Keskinen (2022: 391) note, in line with Hübinette and Lundström (2014), that: “The equating of whiteness with nation-state belonging is a key characteristic of the Nordic region.” This is one of few studies that address policing, in the form of experiences of racial profiling, as a problem of structural racism, in which racial orders are reproduced by bordering practices that stem from a reproduction of normative Swedish whiteness (Schclarek Mulinari and Keskinen, 2022; see also Schclarek Mulinari, 2020). While Barker (2016) uses the Stockholm riots of 2013 to show how socially marginalized youth riot against the police because of the police's contribution to their social marginalization through over-control, violence, and racial profiling. Barker argues that: “the riots represent a conflict over the production of difference within Swedish society” (2016: 4), and that the police have a central role in reflecting and producing racial differences. In her 2017 article, Barker continues the argument of production of differences, this time regarding EU migrants (mostly Roma) in Sweden and based on a policy document and political debate study. Barker shows the violent and coercive effects of protecting the welfare state for Swedish citizens: “the welfare state deploys coercive means to protect and reproduce itself” (Barker, 2017a: 122) and argues that to understand race and racism in Sweden properly one must connect it to the welfare state (Barker, 2017a), as discussed in the above (see also Barker, 2017b).

These approaches have produced important results, which our study seeks to expand on by bringing in novel empirical material with an emphasis on laws, policies, practices, and technologies and their interrelation, for the study of structural racism in policing. We will do that by involving those inarticulable racist practices that constitute standard and routine outputs of a seemingly racially neutral system. ⁴ This problem is exacerbated if the social context of the research is a post-racial state such as Sweden.

In 2008, the Swedish legislature removed the word race from its legal corpus while enumerating the grounds for unlawful discrimination in its new anti-discrimination legislation. The government's expressed reason for this was among other things to reflect the “views and values of the time”—that is, the scientific baselessness of dividing humans into different racial categories (Regeringens Proposition, 2007: 119–120). In this way, race as a significant category shaping social relations was relegated to the past. The vocabulary of race has been proven to be scientifically flawed and race-conscious legislation or racial references must therefore be legally dismissed and considered morally reprehensible. As a result, race does not exist as a concept in public discourse and the use of the term is almost taboo (Hübinette et al., 2012). In turn, the said legislation defined ethnic affiliation in Article 1:5 as “national or ethnic origin, skin color or other similar circumstances.”

This is to say that Swedish racial governance relies on the moral gesturing of a post-racialist society—a “raceless” society (Goldberg, 2001)—in which race is deemed irrelevant both as a marker of group differentiation and as a force that shapes people's material realities. “Post-racialism marks the end of race history, and therefore the end for the need for race-conscious law and public policy” (Obasogie, 2014: 169). The self-understanding of a post-racial society, Obasogie (2014: 170) argues, is one in which “race is no longer a barrier to one's life opportunities. Race is simply no longer salient; society either no longer sees it or pays attention to it in any significant way.” What remains of racism in society is the interpersonal and vulgar referencing of a debunked signifier that would legally be processed under the category of ethnic affiliation.

The irrelevance of race as a category of reference in law would, of course, be logical if race were exclusively understood in terms of hyper-visualized biological attributes such as color and body type. However, this limited understanding ignores the social and material significance of racialization, its constant adaptability and attachment to changing signifiers across time and space (Hall, 2021 (1997)), and its presence as an underlying “system of classification,” albeit concealed, in the “modern western social and political self-vision” (Lentin, 2008: 490). What has made race historically and socially relevant and persistent is not a contestation over what truth—if any—hereditary attributes such as skin color, hair, or body type hold. Instead, race consciousness is significant because of the social, economic, and political consequences and meanings that accompany being associated with a racial category such as white, black, brown, etc. (Lopez, 2006). As Lentin (2008: 492) puts it, “race is significant, not in itself, but because of its inability to exist solely conceptually and in the absence of racism.” In other words, the existence of race lies in the work it does and not in identifying its biological trace. If race is to be merely a scientific notion, then we could not explain why “racial order” (Goldberg, 2001) has permeated so deeply into law and politics. Debunking biological race must go hand in hand with defining race as a social structure—as the “historically contingent social systems of meaning that attach to elements of morphology and ancestry” (Lopez, 2006: 10). More than being a referent to any meaningful biological attributes, race is a way of “making bodies meaningful” (Gilroy, 2000: 46) for social, political, and economic ends.

Post-racial Swedish governance does not account for such a structural understanding of race as a distributor of life chances and determiner of subject positions. This omission, however, is not an anomaly. Post-racial governance, like colorblind ideology, is predicated upon “a process of knowing designed to produce not knowing surrounding white privilege, culpability, and structural white supremacy” (Mueller, 2017: 220, italics in original). Here, Bonilla-Silva's (2006) insights into racism without racists are crucial to making sense of what this implies. Bonilla-Silva explains how, in a racially blind society, racial disparities persist without individuals acknowledging race or racism. Biological race, as the underlying ideological ground for discrimination, is removed from the public lexicon as the overt mechanisms of racism and racial ordering are replaced by “practices that are subtle, institutional, and apparently non-racial” (Bonilla-Silva, 2006: 3). Race-neutral governance, specifically in Europe, predicates its racial rule on the so-called “new racism,” which features non-racial dynamics and signifiers such as religion, culture, and migratory background (Lentin, 2008). Under this framework, segregation is rationalized by reference to individual preferences to keep to their own, economic disparity by reference to lack of ambition or laziness, and social exclusion by reference to cultural immaturity. In all these frames, racism emerges as actions of individuals and not the worldmaking function of racial rule.

The same frames that lead to colorblind racism—to racism without racists—are present to varying degrees in a post-racial society. Obasogie (2014) argues that colorblind ideology and post-racialism, despite significant differences, are synergetic modes of race governance predicated upon racial neutrality. Both modes of governance maintain material racial subordination by casting “all types of race consciousness—from Jim Crow forms of subordination to affirmative action types of remedies—to be equally reprehensible” (Obasogie, 2014: 170), thus denying the necessity of a structural perspective and framing inequalities as “the natural state of things that can only be rectified through individual efforts” (Obasogie, 2014: 173). In the end, post-racial and colorblind governance reduce race to “its lowest common denominator—visually obvious traits that define group membership” (Obasogie, 2014: 175).

Such a narrowing of race and exceptionalizing racism are blatantly evident in the Swedish context. In her analysis of all discrimination cases in Swedish courts relating to ethnic affiliation between 1994 and 2015, Brännström (2017) shows that in the absence of race as a ground for discrimination, Swedish courts have interpreted ethnic affiliation (like biological race) as an immutable, visually and morphologically discernible category. Consequently, Brännström notes, the Swedish courts almost exclusively understand unfavorable treatment as unlawful discrimination if such treatment is accompanied by the explicit verbal referencing of a person's appearance (e.g. a racial slur and/or a clear reference to skin color) (Brännström, 2017: 627). Strikingly, the race-neutral Swedish legislation ends up reemphasizing the idea of group membership based on visual and biological traits and the individualized referencing of these features. What remains underappreciated in the courts is the more subtle operation of race as a sorting and ordering mechanism.

Searching for traces of structural racism—that is “racial rule” and not racial attitudes—does not start with individuals but with the mechanisms, processes, and outcomes of taken-for-granted policies, norms, organizational practices, and historical continuities (see Bonilla-Silva, 1997). As such, laws, policies, and practices aimed at gang crime in Sweden constitute a relevant starting point for being able to uncover structural racism in Swedish contemporary penal policy and policing practices.

Laws: The construction of racialized crime

As outlined above, our first step in tracing structural racism in policing is to scrutinize laws and their preparatory works in the form of policy proposals. As such, we will outline the government's criminal policy-making and legislative strategy, as well as its targets, at the time of Operation Rimfrost.

In September 2019, the Social Democrat and Green Party coalition government introduced a 34-point program for combating organized crime as part of their flagship policy for a “safer Sweden.” At the time, this program was “the largest package of measures ever to combat gang crime in Sweden” (Socialdemokraterna, 2022). In a way, the 34-point program was the precursor to the above-mentioned Tidöavtalet (Liberalerna, 2022). The backdrop for both policies and legal agendas has been widespread concerns over the increasing frequency of violent crimes, including death resulting from shootings and explosions. Since 2016 Swedish police have kept statistics on shootings and bomb detonations across the country. Police data shows an alarming increase in the number of shooting cases from 25 to 363 between 2016 and 2023. As for detonations, the same data shows an increase from 90 cases in 2018 to 149 cases in 2023. ⁵ Moreover, since 2014 deadly violence in Sweden has increased from 87 cases to 121 in 2023 (Brå, 2024: 3–4). Another source of worry has been the number of victims of deadly crimes under the age of 18, which has steadily increased from 3 cases in 2018 to 17 cases in 2023 (Brå, 2024: 11). ⁶

Although the 34-point program seems narrowly focused on measures to combat gang crime, fighting gang crime was essentially the main focus of the Swedish government's general crime policy and legislative work at the time of Operation Rimfrost. The main measures and tools of the program were firstly the expansion of repressive measures in terms of their targets (by including children and youth), objects (e.g. property), consequences (imprisonment), and enforcers (cooperation between the police and other state bodies, such as the social services). And secondly, new laws enable expanded control technologies, such as the surveillance of encrypted communication, secret coercive measures, and camera surveillance. In relation to this, a range of new laws were proposed involving an expansion of criminalization and police and prosecutorial powers. In tandem with the program, there was also a promise of 10,000 new police employees and an SEK 2.4 billion investment in the justice system. ⁷

The 34-point program gestured towards the national goal of a safer Sweden. Operationally, however, the program directed its action and focus on interventions in a handful of geographically delimited areas framed as embodying crime and Sweden's national security and safety problem. These are referred to as socially vulnerable areas. The first hints of the inclination of the Swedish police towards structural racism are found in legal and policy efforts to frame crime and insecurity as a problem of a particular set of neighborhoods that are predominantly inhabited by the economic underclass and racialized people. The regulation of safety, through police interventions in designated spaces, works as a spatial sorting mechanism, which we will argue below constitutes a racialized sorting mechanism focused on spaces of the disorderly.

Abandoned neighborhoods

Operation Rimfrost took place in some of the most well-popularized so-called socially vulnerable areas in the city of Malmö. Socially vulnerable area (socialt utsatta områden) is police terminology that has found its way into the policy and legal discourses in Sweden. Based on the police definition, a socially vulnerable area is “a geographically delimited area characterized by low socio-economic status, where criminals have an impact on the local community” (Polisen, 2019: 4). The criteria for what constitutes a socially vulnerable area are rather vague. The bi-annual list of these areas, along with criteria for inclusion, is crafted by the police following subjective estimates of (1) inhabitants’ inclination to participate in legal proceedings; (2) the police's ability to perform its mission; (3) possible parallel structures; and (4) extremism, particularly violence-promoting Islamic extremism, none of which indicate any form of socio-economic measurement. Instead, these indicators exclusively focus on the police's own action possibilities in such areas. Even though the areas are termed socially vulnerable, the core focus of the police's definition and attention is directed at crime and repression rather than socio-economic measures.

The Swedish National Audit Office has directed criticism at the Police Authority's method for identifying these areas because “it is partly based on subjective estimates of the areas’ problems,” and there is no generally accepted definition (RIR, 2020: 8). The public presentation of these geographical areas as areas of problems and prioritization for policing goes hand in hand with a form of cognitive dissonance. For instance, in 2017 the Swedish government released a written communication, prepared by the Ministry of Justice, where it mentions that segregation, poverty, and social vulnerability characterize these neighborhoods, and are most noticeably manifested in overcrowding, high unemployment, and the deteriorating school performance and health of their inhabitants (Justitiedepartementet, 2017: 3). The cognitive dissonance surfaces when the same document interprets these socio-economic problems to be caused by criminality in these areas and not vice versa—that crime is an outcome of these, admittedly wide-reaching, forms of social exclusion and organized abandonment.

Definitional imprecision and cognitive dissonance are just one part of a number of problematic issues in the public and political framing of these neighborhoods. Although not including a language of race or ethnicity, the criteria are effectively racialized. Parallel structures, extremism, and Islamic extremism operate as signifiers for a racialized population, implying that the designated problematic areas in Sweden, that are in need of particular attention from the police, are racialized neighborhoods. Furthermore, the criteria relate to the police's ability to conduct their work in the areas, and the definition of these areas becomes a matter of how the inhabitants have failed in their duties toward the state, rather than a matter of how the state might have failed the inhabitants.

A cursory look at the available statistics from Malmö municipality (Malmö Stad, 2023) shows the depth of the socio-economic failings and offers a different framing of the areas—one that is centered on the material situation of the inhabitants of these areas and not the experience of the police in fulfilling its assignments. ⁸ As far as one of the central targeted neighborhoods of Operation Rimfrost, Rosengård, is concerned, this so-called vulnerable area ranks at the bottom of most socio-economically relevant data. As a city district, Rosengård has the highest population proportion that is either born outside Sweden or has two parents born outside Sweden (Malmö Stad, 2023). ⁹ Furthermore, Rosengård's inhabitants rank second lowest in the population in having a post-high school degree, have the lowest percentage of paid employment proportionate to the number of inhabitants (50%), and the lowest average disposable income in the city of Malmö (Malmö Stad, 2023). As to general health, the inhabitants are on sick leave more than any other residents in the city, and the pattern continues with regard to education quality (Malmö Stad, 2023). A report published by a local newspaper (Westerberg, 2022) shows that four of the lowest quality schools in Malmö are either in Rosengård or at its borders. The inhabitants of Rosengård are thus on average more unemployed, earn less when employed, are provided with the lowest quality education, and have poorer health than the rest of Malmö's population. Following Gilmore's (2022) conceptualization of “organized abandonment” by the state—a concept describing the results of austerity measures, where social and economic efforts are replaced by repressive measures—and framing vulnerability differently, we prefer to call these areas abandoned neighborhoods.

Areas such as Rosengård, and the racialized populations within them, are defined as a problem by the government and for the police—of unrest, disorder, crime, or propensity to crime—which in turn legitimizes increasing police efforts. This implies that the political and discursive construction of the neighborhoods necessary for an operation like Rimfrost to take place was already firmly established, and the operation was not a one-off incident but a continuation of the policy of concentrating police efforts in these neighborhoods.

What is clear from the ongoing trajectory of policing in Sweden, in terms of policy and legislative measures, is that it involves an increasing reliance on the concentration of police in abandoned neighborhoods, investment in the development of new policing technologies, and an increase in the number of police officers per capita. A 2020 report by the Swedish National Audit Office puts the extent of this trajectory into perspective and notes its unmistakable focus on abandoned neighborhoods as the target area for policing. From 2015 to 2020, the Swedish Police Authority received a budget increase of SEK 7.4 billion from the state, representing a 35% increase, while retaining the abandoned neighborhoods as their pronounced and specific focus site. The expanded budget has generated an increase in community police officers, with one-third of the increase being in abandoned neighborhoods to enable a greater police presence in these areas (RIR, 2020).

Practices: Operation Rimfrost

We now move on to the next step in our tripartite methodology: practices. The practices we explore are linked to the controversial police operation Rimfrost, which commenced in late 2019 in the city of Malmö. We contend that this operation showcases the features of the racialization of crime as reflected in the government's criminal policy program discussed above. More specifically, we show that this operation constitutes an example of another layer in the formation of structural racism in Swedish policing, and that it does so by deploying a narrow lens and easy fixes for complex social issues by focusing on neighborhoods inhabited by poor and racialized population as constituting the predominant site of crime.

Operation Rimfrost is only one example of a concentration and focus of police efforts on abandoned neighborhoods, but a significant one in its scope and use of resources. The operation lasted from November 2019 to June 2020, and took place in abandoned neighborhoods in the city of Malmö, among them the district of Rosengård mentioned above. Rimfrost had four major goals: reducing the number of shootings and explosions in criminal milieus; reducing the number of individuals in criminal networks through prosecution and gang exit initiatives; increasing the number of weapons and explosives seized; and enhancing public safety. As a special national operation, during Rimfrost the Swedish police redeployed massive resources from other regions to Malmö. To give one example, in a single raid conducted during the operation, in January 2020, 729 premises were searched, and more than 2000 vehicles controlled. In these and other such raids, more than 200 police officers and numerous police vehicles participated simultaneously. Additionally, other state bodies took part in the raid in cooperation with the police, such as the Swedish Enforcement Authority, the Tax Agency, and Customs (Galinon, 2020). Rimfrost concluded with the seizure of 96 weapons and 200 kilos of explosives, three million SEK in cash, 205 kilos of marijuana and hashish, 54 kilos of cocaine, six kilos of amphetamine, and half a kilo of heroin. Further, 209 persons were arrested and 53 wanted persons apprehended (Nygren, 2020). Given this, one could contend that the operation was a success in terms of one of its goals “increasing the number of weapons and explosives seized,” provided these seizures indeed constituted an increase in relation to normal practice. Regarding the amounts of various kinds of seized drugs, it is well-known that statistics on drug crime are almost solely produced by police priorities, so this was an expected result. The more an area is controlled, the more drug crime will be recorded in the crime statistics for that area, leading to a rationale regarding which areas police efforts should focus on and resulting in a reproduction of drug statistics. Deploying a massive police operation in any area is likely to produce results. As argued above, the areas targeted by Rimfrost had already been discursively constructed as disorderly and crime prone. The production of new statistics through Rimfrost bolsters this depiction of the areas in question, and legitimizes the continuing control of a racialized population with extraordinary measures.

Early on, the police publicly claimed that Rimfrost was a success (e.g. Dagens Nyheter, 2019; Expressen, 2019). Holgersson and Westman (2021: 27) note that claiming success was in fact part of the police's strategy from the outset of Rimfrost. This claimed success of the operation has been debated with regard to its stated goals, the goals as they relate to police strategies in general, and the operation's impact on the neighborhoods and people living in them. The Swedish National Audit Office found that the effects of Rimfrost were primarily temporary (RIR, 2020: 44). Holgersson and Westman (2021) did not even find a temporary effect, but rather that the operation hardly made any difference to statistics as compared to previous years. They also criticize the statistics presented. The number of weapons seizures, for example, includes both non-lethal weapons and weapons parts, and the goal of reducing the number of shootings was far from achieved (Holgersson and Westman, 2021). Holgersson and Westman conclude that the operation was put into effect to make the police appear empowered and in control of the situation, and, as such, was merely a symbolic operation (Holgersson and Westman, 2021). Holgersson (2020:15) has also noted that, “Many people believe the myth that more police resources are effective and therefore such an initiative can have an effect on public opinion,” and points out that strengthening resources in one place, as was done in Rimfrost, implies taking them away from somewhere else (Holgersson, 2020). The operation has also received criticism for its negative impact on inhabitants in the affected abandoned neighborhoods. One representative for an NGO working with young people in one of the neighborhoods stated that: “Many feel that it increases suspicion, both of themselves and the places in question” (Swedish Radio, 2020), something that is echoed in research on police controls of youth in Sweden (Schclarek Mulinari and Wolgast, 2020; Wästerfors and Burcar Alm, 2020). Rimfrost as a symbolic operation focused on a particular racialized area and contributing to its criminalization came at a high cost for the population in the area.

One of the goals of Rimfrost, “enhancing public safety,” is harder to measure than counting seized goods and persons. Over-control and under-protection in abandoned neighborhoods are well-known phenomena (Barker, 2016; Gilmore, 2022; Loader and Walker, 2007: Rios, 2011; Wacquant, 2010). Although these phenomena have not been extensively studied in Sweden's abandoned neighborhoods, there is a body of research on experiences and perceptions of over-control and under-protection, as well as racial profiling, particularly of youth (Schclarek Mulinari and Keskinen, 2022: Schclarek Mulinari and Wolgast, 2020; Wästerfors and Burcar Alm, 2020). Schclarek Mulinari and Keskinen (2022) argue that “experiences of racial profiling indicate the engagement of the police in racializing work and that the police in the Nordic countries are pivotal, and have long been so, in the upholding of hierarchies, reproducing boundaries of belonging, and functioning as pillars of the racial welfare state” (Mulinari and Keskinen, 2022: 392). By connecting this to the above discussion of the definition and understanding of so-called socially vulnerable areas as a police problem, having enhanced public safety as one of Rimfrost's main goals rests on a sorting mechanism of orderly (to be protected) and disorderly (to be controlled) spaces, and on symbolically asserting control over the “unruly” abandoned neighborhoods and their racialized inhabitants. This sorting mechanism is put in action via the idea of criminal networks and gang exit initiatives, as defined in the goals of Rimfrost, while the gang-label has consistently proven to be an effective legitimizer for over-policing and the othering of racialized communities and people, which again contributes to a popularized image of the racialized population as criminals (Williams, 2015).

Operations like Rimfrost produce visibility and easy victories and can be used to show political determination and police efficacy. Political victories are easy to collect when managing crime and safety means going to the same places to find the same things. The government's 34-point program clears the ground for policing that reproduces the racialization of crime in its predominant focus on crime that can be statistically localized and reproduced in abandoned neighborhoods. At the same time, these police-produced statistics and knowledge feedback to policymakers, which will again obtain new legitimacy for continuing their repressive priority focus on abandoned neighborhoods. Criminal policy in Sweden reproduces race and crime through a transformation of the utility of law and along the lines of class and race on behalf of the well-being and safety of already vulnerable populations. Race is not only constructed but distinctly produced, as disorderly through the criminal justice system and policing. Structural racism as ingrained in the policing of social order creates a self-perpetuating spiral in the reproduction of crime and its control, which reinforces and reproduces the image of “the criminal” and the “criminal neighborhood”—shaping and affecting the lives, well-being, and life chances of racialized persons.

Tools and technologies: Predictive policing

Another significant locus for tracing racial differentiation as a signifier of criminality is the development and use of the new digital technologies of predictive policing. Predictive policing refers to practices that rely on data analytics to predict future instances of crime, with the aim of facilitating efficient and preemptive intervention by the police (Browning and Arrigo, 2021). Typically, these technologies predict future crime by identifying patterns of activity and historical crimes while overlapping them with spatial and temporal distributions of crime (Kaufmann et al., 2019).

Predictive policing is predicated upon two major types of databases and algorithmic analytics: personal data and geo-spatial crime analytics (see Heaven, 2020). Producing personal information databases—of biographical and biometric data—as well as geo-spatial data connoting the spatial and temporal spread of crime, are therefore central to predictive policing. In practice, each form of data analytics produces a rating system that classifies people and places based on the risk of crime. Once the outcome of these data processes are superimposed on one another, a cohesive and updateable visualized map of people, places, and times of possible crime can be produced. These maps in turn help officers to determine where, when, and whom to police.

In Sweden, Rimfrost kick-started the national adoption of predictive policing technologies. Below we discuss the Swedish police development of personal data and geo-spatial analytics, and show how their use, function, and underlying logics contribute to the allocation of risk and suspicion of criminality specifically to the marginalized and racialized inhabitants of abandoned neighborhoods.

Personal data analytics

Soon after the conclusion of Rimfrost, a leaked customer list from a controversial tech startup—Clearview AI—disclosed the Swedish police as a customer of an AI-powered facial recognition technology (FRT) (Mac et al., 2020; Pettersson, 2020). It was also revealed that the Swedish police used and tested these automated facial recognition/biometric technologies during Rimfrost without acquiring the required legal permissions. ¹⁰ Hence, in its unannounced objective, Rimfrost emerges as a covert biometric data collection and technological testing operation, in which marginalized and racialized communities served as the testing ground.

Clearview AI is a private company based in the US that provides FRTs to law enforcement agencies. Like other FRTs, Clearview AI is a tool for identifying persons of interest by cross-scanning and matching an input image with images available in a database. Clearview AI is distinct for two main reasons. First, algorithms used by Clearview AI have collected the world's largest database of publicly available personal images in a process known as internet scraping. Second, Clearview AI sells its services not on a case-by-case basis but as packages and en masse to law enforcement services (Rezende, 2020: 376). This business model in and of itself poses serious threats to people's personal integrity and privacy rights (EDPB, 2020). Besides these concerns, FRTs like Clearview AI are notorious for reproducing racial and gendered biases (EDPB, 2022). As far as structural racial biases are concerned, FRTs disproportionately produce false-positive identifications when the scanned subject is a person of color and/or a woman (EDPB, 2022), in turn resulting in the “racialisation of risk” (DaViera et al., 2023). A major study at MIT has shown that even the FRTs developed by tech giants like IBM and Microsoft tend to misidentify Black women with an error rate of 34%. When compared with the 0.8% error rate for white male subjects, this study reveals a clear racial bias embedded in these technologies (Buolamwini and Gebru, 2018). One reason for this technological bias is that the databases upon which these AI tools are trained are predominantly composed of lighter skinned male subjects (79–85%) (Marks, 2021). Even so, on its homepage Clearview AI prides itself on possessing the world's most diverse database for training bias out of its algorithms.

What remains as an unbridgeable racial bias of FRTs, however, is the racial context in which visual technologies (including photography and video technology) have originally been developed. Put simply, visual technologies record images by capturing reflections of light from surfaces. The lighter the skin color of a subject, the better the reflection of light and consequently the more accurate the recorded image. The darker a skin color, the less distinguishable the subject, their specific features, and background. In fact, capturing the facial features of people with darker skin colors accurately has been a persistent problem in the cinema and television industries since their inception (Dyer, 1997). The same inaccuracy persists in the digital era (e.g. Sinclair, 2010). The techno-racial make-up of visual technologies also remains unresolved for both CCTV cameras and biometric scanners. Such remarkable and lasting discrepancies along racial lines in the context of police operations translate into frequent miss- or non-identification of people with darker skin colors, along with a subsequent allocation of higher risk values, higher rates of search and arrest and, above all, the justification of the over-policing of populations along racial lines.

In Sweden, the problem goes beyond the inherent biases of the technology and points to a racialized structure in which such technologies are deployed. Illegally testing a secretly acquired technology in an abandoned neighborhood, where migrant populations are in the majority, is to exclude them from the scope of legal protections granted to citizens with regard to personal integrity and privacy. The view that takes these bodies, their lives, and living spaces as being readily available for extra-legal interventions in the form of testing sites for new solutions and practices (be it medical, social, economic, or technological), goes hand in hand with the racial logic of extraction that sees non-white bodies and their spaces not as fully protectable bodies and places but as resources and sites for the extraction of data and knowledge with the purpose of creating prosperity and safety elsewhere and for other bodies.

This logic appears to be naturalized in the Swedish police, given that the decision to use and test Clearview AI was in fact based on the individual initiative of an unnamed employee at the National Operative Department (NOA) (IMY, 2021). During the approximately six months that the police were using Clearview AI on this one individual's initiative, not a single police officer or authority talked about the legal questionability of such practices. In fact, the Swedish Authority for Privacy Protection (Integritetsskyddsmyndigheten (IMY)) learned about Clearview AI through media reporting (IMY, 2021). In choosing Rimfrost as the occasion for testing such technologies, the police have on the one hand enacted their racialized assumptions regarding the location of criminality while on the other hand reinforcing it through a decision regarding from where the databases should begin collecting personal data.

Geo-spatial analytics of crime and the Swedish Crime Harm Index

As the scandal surrounding the unauthorized use of Clearview AI was brewing in Sweden, the Swedish police were simultaneously developing a new method of crime harm indexing—one that would both be machine readable and would fit the risk-oriented approach of predictive policing practices (see Lauffs, 2020). In August 2020, Sweden's national broadcasting service reported the introduction of a new crime statistics practice set to be operating by the end of that year (Lauffs, 2020).

This new method—the crime harm index—aims to devise indicators that could capture “differences in harm levels” between various types of crime (Kärrholm et al., 2020: 15). The material outcome of this method of indexing is visualized on maps of crime harm that are suited to pattern reading and the spatial-temporal risk classification of different parts of any given city.

The developers of this new method claim that conventional statistics lack sensitivity to various crime types. “[I]t is not sensible,” they write, “to equally value murder and shoplifting. Likewise, in the process of identifying geographical areas that are to be prioritized for patrolling, it is irrational to value serious assault and verbal harassment equally” (Kärrholm et al., 2020:16). The argument here is that the limited resources of the police should be efficiently concentrated on crimes that carry a higher “weight” or that produce “most harm.” In this view, then, how one defines harm determines where, how, and with what severity police forces will act. The harm of a crime in this new system is not defined via a social or intersubjective understanding of harm. It is instead determined by what the authors call the “penal value” of a crime (Kärrholm et al., 2020: 18).

In other words, the new Swedish indexing system understands harm in terms of the number of prison days that the legislature prescribes in the law (not the number of days decided by courts) for each crime. The higher the prescribed punishment for a crime, the higher the risk score of the location in which the crime is discovered or committed. As such, a socially grounded and empirical notion of harm is traded with the political and fluid level of punishment that the legislature allocates to each crime. One can already see how racial biases generated through legislative increases in punishment, as outlined in the 34-point program and later Tidöavtalet for example, and the over-policing of certain areas, in operations like Rimfrost, leads to statistical over-representation and then pours over into technological risk evaluation, which is presented as a technical, objective, machine decision. Deploying police resources based on the penal value of crimes reveals more of the same problem when we view this technological development together with other developments in Sweden, where harsher sentences and longer prison terms are being implemented for crimes that are engulfing abandoned communities and/or are investigated predominantly in those areas.

Together, these developments form a feedback loop between data and rationality, where the over-policing of racialized neighborhoods leads to ever more criminalization, leading to calls for harsher punishments and longer prison sentences, leading to the allocation of higher risk scores to these areas and, once again, leading to the justification of the over-policing of these neighborhoods and so on. From a holistic perspective, then, Operation Rimfrost and its technological trailblazing fall into a similar trajectory of developing laws, policies, and police practices that see criminality as a problem concentrated in particular areas, and among people who are manifestly defined and recognized in terms of their socio-economic abandonment and racial difference. These elements tend to reinforce, reiterate, and rationalize one another each step of the way.

Conclusion

In post-racial societies—where race as a force for organizing social relations and determining access to life chances is framed as passé—critiques of structural racism in policing are dismissed on the grounds of the lack or impossibility of empirical proof. The post-racial mode of governance reduces racism in policing to individually held beliefs, attitudes, and interpersonal relations. Hence, if no collectively held and publicly pronounced racialized view on crime amongst a vast number of officers is identifiable, any claim of structural racism will be framed as unfounded. In this article we took the challenge of tracing structural racism in such a context seriously and showed that, if taken as a structure, race operates through interrelated operations of laws/policies, practices, and technologies—none of which necessarily need to be explicitly or implicitly racial, but which nonetheless produce racialized outcomes. This tripartite model of tracing structural racism is more fitted to a legal and political context in which post-racialism is the dominant model of racial governance because tracing structural racism—that is “racial rule” and not racial attitudes—does not start with individuals but with the mechanisms, processes, and outcomes of taken-for-granted policies, norms, organizational practices, and historical continuities. In the context of Swedish policing, legal, and policy labeling of socio-economically abandoned neighborhoods as areas of risk inhabited by potential criminals lends justification to targeting racialized communities with oppressive police operations instead of providing needed social and economic support. At the same time, the specific—almost exclusive—targeting of racialized communities leads to further criminalization and sets in motion a cycle of oppressive measures. This cycle of overlapping racialization, abandonment, and crime gains a facade of objectivity and scientificity through the deployment of technological tools to predict and assess risks. Yet these technologies, in their historical development, conditions of operation, use and logic, consolidate the concurrence of racialization, abandonment, and criminalization.