The risk–gang nexus in Sweden: Penal layering and the uneven topography of penal change

Abstract

This article maps the emergence of a risk–gang nexus in the Swedish correctional field. Using the concept of penal layering, the article analyses the configuration of risk-based practices to manage gang affiliates in the context of Swedish prison and probation services. By introducing the notion interludes of penal layering, this article stresses the synchronic/diachronic opposition in penal change processes and furthers ongoing discussions of the variegated, braided, and uneven patterns of penal change in and beyond Nordic penal institutions. Drawing on interviews with probation officers, prison staff, and other professionals working with gangs in Sweden, the study finds that new rules, practices, decisional hierarchies, and actors have been layered on top of or alongside existing institutional arrangements to manage gang affiliates. This penal layer enforces punitive conditions for incarcerated gang affiliates and creates tensions, agonism, and contradictions between the coexisting and multiple logics of rehabilitation, punishment, and risk in the penal field.

Introduction

No European country has experienced such an increase in gang-related lethal gun violence as Sweden has (Sturup et al., 2019). Political parties have tried to trump each other's plans to combat the situation, usually expressed through tough-on-crime policies. In September 2019, Sweden's Social Democratic government launched a 34-point gang intervention programme that included proposals for more surveillance, anti-gang legislation, and a greater focus on crime prevention (Regeringen, 2019). Katarina Påhlsson, Parliamentary Ombudsman, was doubtful about the sentencing enhancements and concluded: ‘If the proposals are carried through, they will entail far-reaching consequences on crimes without such a [gang] connection’ (Justitieombudsmannen, 2021). The proposed anti-gang policies were closely connected to the rise of new policing practices in marginalized communities. Since the mid-2010s, the Swedish police have targeted poor areas in Swedish cities, calling them ‘parallel societies’, ‘social risks’, and ‘particularly vulnerable areas’, illuminating not only the geography of social inequality, but also the transformation of welfare state strategies from universal ones addressing structural problems, to more selective measures based on risk governance. In the continuing debate, political parties have suggested a number of risk-informed strategies to identify youths at risk of joining gangs, including screening children's language skills at very early age. Such proliferation of penal power to fight gangs indicates a significant transformation of Sweden's patterns of punishment.

How and why penal institutions, practices, and ideas change is a key topic in criminology (Daems, 2008; Goodman et al., 2015; Rubin, 2016, 2018). Influential theories of penal change have linked macro-structural social changes – including industrial capitalism, the Keynesian welfare state, and neoliberalism – to the emergence of particular penal institutions, technologies, and practices (Foucault, 1995; Garland, 2001; Wacquant, 2009). In the so-called ‘punitive turn’ beginning in the 1970s, Garland (2001) linked major social changes in politics and the economy to the emergence of risk discourse, punitive rhetoric, and the social construction of dangerous offender groups that underpinned the cultural formation of a late-modern ‘crime complex’. Risk coupled with dangerousness formed a collective experience of crime, furthered the criminalization of social problems, and led to the invention of new governance arrangements for managing high-risk groupings (Andrews et al., 1990; Feeley and Simon, 1992; Garland, 2001; Maurutto and Hannah-Moffat, 2006). Harcourt (2007: 174), for instance, argues that ‘the social shift in our conception of just punishment in the 1980s and ‘90s can be traced specifically to the rise of actuarial methods and their implementation’. In a similar fashion, gang scholars Decker et al. (2022: 333) argue that rehabilitative gang interventions faded away in the 1980s, turning gang affiliation into a category that ‘could be exploited to result in an extra-legal penalty on charging decisions, pretrial detention, juvenile transfers to adult court, sentences to confinement, and longer sentences’.

Such rupture-based accounts of penal change have been criticized for downplaying the uneven, variegated, and braided character of penal change (Goodman et al., 2015; Hutchinson, 2006; Rubin, 2016) and for exaggerating the coherence of specific historical penal regimes (Hutchinson, 2006). Moreover, mechanical views of penal change tend to underestimate the ongoing struggle and tensions between different actors in the penal field striving to shape criminal policy objectives (Goodman et al., 2015).

Despite its considerable influence on crime policy discussions and law reforms in the present conjuncture, the role of gangs is a less explored theme in relation to penal change in the Swedish context. ¹ In a 2018 press release, the Swedish Prison and Probation Service (PPS) claimed that more than 80 criminal groups were represented in the Swedish prison population (Kriminalvården, 2018). Nordic prisons have been described as more inclusive, humane, and welfare oriented than those in other Western democracies (Pratt, 2008). Ugelvik (2016: 398), for example, argues that ‘prisoners in these institutions are given opportunities, resources and living conditions that are not available to prisoners in most other jurisdictions’. This Nordic exceptionalism thesis, however, has been contested by numerous scholars who have deconstructed the punishment–welfare nexus (Barker, 2013, 2018; Barker and Scharff Smith, 2021), traced the influences of neoliberal ideologies in crime policies (Shammas, 2016), and demonstrated the differential treatment of ‘members’ versus ‘non-members’ in penal institutions (Barker and Scharff Smith, 2021). Others argue that security concerns and risk assessments have been ‘elevated to a core part of everything that goes on in the Swedish prison estate’ (Scharff Smith and Ugelvik, 2017: 522). Risk assessment has become a key factor in decision-making throughout the correctional field, including in sentencing, prison placement, probation, early release, furlough, treatment programmes, and parole.

In 2016, PPS implemented the proprietary Risk, Need, and Responsivity Assessment Instrument (RNR-A), in which gang involvement is included as a specific risk/need item for recidivism (Bäckström Johansson et al., 2022). This micro-level change – the introduction of gang affiliation as a specific risk factor – provides an analytical entry point from which to map and analyse the emergence of a risk–gang nexus in Swedish corrections. In this article, we aim to analyse how gang affiliation is managed within PPS as a specific risk/need factor and discuss its significance for penal change. We apply Rubin's (2016) concept of penal layering to the different rules, ideas, actors, and practices that have been added on top of or alongside existing penal arrangements for managing gang affiliates as risk subjects. Penal layering describes a process in which several penal technologies coexist in a given system, producing an uneven topography and differential treatment within a penal population.

While penal layering refers to an uneven, diachronic process of gradual penal change, as opposed to wholesale and critical ruptures, we introduce the notion interludes of penal layering to provide a synchronic analysis that gives a detailed understanding of how additions, amendments, and revisions of penal technologies are configured, fought for, and resisted. We specifically ask what content (rules, actors, practices, or ideas) has been added and configured for managing gang-affiliated inmates and how these added layers might be understood in relation to changes in the conditions of incarceration and reintegrative crime policies.

In the following three sections, we introduce the study's method and the concept of penal layering, and review the risk-assessment literature from a penal-change perspective. Thereafter, we contextualize the ideational development of Swedish prison policies followed by an in-depth analysis of the configuration of the risk–gang nexus in relation to conditions of incarceration and community-based sanctions. In the final section, we discuss the relevance of the concept of penal layering as a mode of penal change and as a critical analytical tool more generally.

Method

The study is part of a multi-year research project (2016–2019) on processes of exiting gangs. Forty-two interviews were conducted with a total of 72 people who worked with gangs. Interviewees ranged from general policy and management administrators to those working directly with gang affiliates, such as municipal support service coordinators, probation officers, and those with various roles in the prison service or specific treatment programmes. All 42 interviews were analysed, but this article focuses on 26 interviews with 35 employees of PPS. We have also relied on other sources such as government documents, evaluations, and secondary sources to contextualize our findings.

The interviews were structured to focus on four areas: (1) activities, staff, organisation, and cooperation; (2) the gang affiliates with whom the respondents worked; (3) the significance of masculinity and gender in this work; and (4) factors and perspectives central to work with gang affiliates (Kvale and Brinkmann, 2009). The interviews were held at the respondents’ workplaces, namely prisons and probation offices.

We used a two-stage coding process in the analysis (Miles, Huberman and Saldaña, 2014). First, we sorted the entire corpus of interview material into either predetermined codes or codes created during the analytical process, using NVivo analysis software. The predetermined codes/categories were based on the overall research topics of 1) organizational forms and interventions to support gang affiliates, 2) representations of gang affiliates, and 3) the significance of masculinity in the support context. We noted early in the analysis that reasoning surrounding support and treatment initiatives strongly coincided with risk assessments and risk-based practices. In the second stage, we sorted the coded material again, based on three theory-informed questions: 1) What rules govern risk assessments for gang-affiliated inmates? 2) How are risk-assessment practices organized and performed? and 3) What ideas influence risk-based practices? We were attentive to how respondents conceptualized, narrated, and justified risk-based practices in relation to gang affiliates.

Penal layering

Rubin's (2016) concept of penal layering is based on the historical-institutionalist tradition. Institutional theorists have often attributed institutional change to ‘exogenous shocks’, ‘critical junctures’, or a ‘punctuated equilibrium’; layering, in contrast, is not characterized as a wholesale break with past approaches, the introduction of wholly new institutions, or a move from one dominant logic to another (Mahoney and Thelen, 2010). Mahoney and Thelen (2010: 17) argue that layering most likely takes place in political and institutional contexts when ‘challengers lack the capacity to actually change the rules’ or ‘to set up an alternative institution or system’. Layering is portrayed as a mode of gradual institutional change through small incremental steps, subtle revisions, and amendments to existing institutional frameworks (Mahoney and Thelen, 2010; van der Heijden, 2011). By attaching new rules, practices, ideas, and actors to those already in place, layering can change how existing actions are structured and given meaning within institutions.

Although the literature on penal layering aims to move beyond the ‘emphasis on the moment of change’, layering likewise consists of small changes that over time can accumulate and lead to more profound institutional change (Rubin, 2018: 193). We use the notion of interludes of penal layering to highlight these small changes and additions to existing frameworks, dominant logics, and prior trajectories. For this article, we provide a working definition of interludes of penal layering consisting of an amalgamation of two components: first, an analytical strategy for bracketing sequences in time to provide a synchronic analysis of how additions, amendments, and revisions of penal technologies are configured, fought for, and resisted in institutional spaces; second, a theoretical conceptualization that aims to overcome the synchronic/diachronic opposition in penal change processes by bridging local, situated micro-level changes with broader macro-level changes and historical patterns in order to understand more durable transformations in penality. By focusing on particular interludes, we acknowledge the micro-universe of relational dynamics between social forces, actors, technologies, and institutional innovations, revealing how penal layers are products of struggle and negotiations between the old and new, the near and far, and are adopted unevenly across time and space in the penal system.

Rubin (2016: 420) introduced the concept of penal layering to capture ‘the layering of new penal technologies over old technologies to describe the complicated coexistence of old and new penal technologies following significant moments of change’. Rubin conceptualizes penal change as a complex, variegated, and uneven process. The model of penal layers illustrates the multiple combinations of new and old penal technologies in a criminal justice system and emphasizes how new layers are shaped in relation to existing penal technologies (Rubin, 2016). Standardized risk assessments are a good example of the imperfect nature of penal technologies, since these instruments undergo constant revisions. The layering of new rules, risk categories, and organizational practices is most visible in the language of different ‘generations’ of risk assessment instruments (Andrews and Bonta, 2010).

Rubin (2016) uses the replacement of capital punishment with long-term incarceration in the USA in the late 1800s and early 1900s to exemplify penal layering through shifting technologies of punishment and more general penal reform. However, penal layering might also include various additions to existing frameworks. Based on the literature (Capano, 2019; Mahoney and Thelen, 2010; van der Heijden, 2011), we identify four types of content that are usually layered: rules/regulations, practices/instruments, ideas/values, and actors. Layering of rules or regulations refers simply to the articulation of legitimate institutional rules, formal or informal, that are expected to be followed (Streeck and Thelen, 2005), although they can be resisted, negotiated, or disobeyed. Layering of practices and instruments refers to the introduction of new institutional practices, in our case exemplified by the introduction of RNR-A. Risk-assessment practices are added onto multiple existing decision points, made part of policy objectives, and linked to specific rules that users must follow (see van Eijk, 2020). Layering of ideas or values captures the introduction and proliferation of new ideas that alter the ideational structure of institutional arrangements (Capano, 2019. In this context, ideas can be defined as ‘causal beliefs’ (Béland and Cox, 2010) that frame the definition of policy problems and their solutions. The layering of actors – or of governance arrangements more broadly – is closely connected to altered power hierarchies, such as new decisional levels, changed organizational roles, or removed discretionary powers.

While the content of penal layering is crucial for exploring additions to institutional arrangements, the specific characteristics of layers are important for understanding many different topics, including institutionalization, legitimacy, diffusion, and unevenness. Rubin (2016) uses four metaphorical descriptions to map the layers’ characteristics: depth, thickness, density, and horizontal space. These metaphors are powerful tools with which to understand why some layers persist over time, have strong support from different actors, and are unevenly spread across the penal landscape.

We adhere in general to Rubin's concept, but in the following outline of this framework suggest some slight modification. First, we emphasize the historical/temporal dimension of layers, namely their depth. Rubin (2016) uses depth to refer to older and more institutionalized practices. For instance, standardized risk assessments are based on historical and statistical records from a criminal justice system that has disproportionately targeted marginalized groups (Hannah-Moffat, 2013a). These deep layers can be reactivated and recycled to resurface in specific situations. The second dimension, thickness, relates to the frequency of actions within a specific layer (van der Heijden, 2011). Added actors, decisional levels, and organizations cause layers to become thicker and denser, resulting in institutional complexity. Finally, topography describes the uneven and variegated patterns of layers. As Rubin (2016) argues, penal layers are not necessarily identical in all places. Penal layering, therefore, can be restricted to certain segments of the population, the introduction of specialist roles, and other organizational functions operating on the fringes of institutions.

Risk instruments and penal change

Since the early 1990s, standardized risk instruments have proliferated across many jurisdictions and become ‘a key organizing principle of contemporary correctional practice and offender management’ (Maurutto and Hannah-Moffat, 2006: 438). Forerunners of modern actuarial methods can be traced to the 1920s through Chicago School sociologist Ernest Burgess and his work with statistical methods to predict risk of reoffending for various ‘social types’ (Harcourt, 2007). The construction of certain risk categories shows how risk assessments do not merely describe the world, but also have discursive effects as they constitute penal subjects (Hannah-Moffat et al., 2009).

In the 1980s, standardized risk-assessment instruments achieved a breakthrough in several Anglo-Saxon countries. Risk-assessment instruments such as the Salient Factor Score in the USA, the Offender Group Reconviction Scale in the UK, and the Statistical Information on Recidivism Scale in Canada were implemented. The common feature of these second-generation risk-assessment instruments was reliance on static variables and unchangeable items (Andrews and Bonta, 2010). The instruments were supposed to contribute to greater equality in the justice system, minimize subjective judgements, and identify ‘objective criteria’ for recidivism, including information about previous convictions, type of crime, length of sentence, age, and gender (Hannah-Moffat, 2013b). Psychologists Bonta and Andrews (2007: 4) discredited the single-minded focus on static risk factors and developed a third-generation instrument based on empirical research on recidivism and psychological theories of social learning (Andrews and Bonta, 2010; Hannah-Moffat, 2013a). The concept of dynamic risk factors, also known as ‘criminogenic needs’, was introduced to indicate how ‘assessed change is associated with subsequent criminal behaviour’ (Andrews and Bonta, 2010: 27).

The turn away from a static view of risk has resulted in a risk logic meant to prioritize high-risk groups and develop tailor-made treatment programmes (Hannah-Moffat, 2005; Hutchinson, 2006). For instance, in their fourth-generation risk-assessment instrument, Level of Service/Case Management Inventory, Andrews and Bonta (2010) include ‘gang participation’ as a specific risk/need item, aiming to recognize ‘the need to assess aspects of the offender and the offenders’ situation that may have criminogenic potential for that particular individual’ (p. 318). The evolution of standardized risk-assessment instruments represents a gradual transformation in which new layers of ideas, actors, and instruments are added on top of older generations of instruments. Despite the temporal development, Werth (2019: 4) notes that ‘all four forms are variously used in correctional settings today’.

While the constant revision and release of new generations of standardized risk-assessment instruments characterize an incremental process of penal change, the ‘new penology’ thesis of Feeley and Simon (1992) described the ascendency of risk-based practices as a ruptural transformation of penal governance. The new penology broke with former rehabilitative, reintegrative, and correctional strategies of penal modernism in favour of a more managerial orientation towards aggregated risk groups. Similarly, Garland (2001) viewed risk-based practices as central in the punitive turn to, and formation of, neoliberal crime policies. Many scholars have challenged the assumption that risk-based practices have displaced rehabilitative approaches and aligned risk solely with incapacitation. O’Malley (2004: 333), for example, sees the possibility of a socially just politics of risk based on ‘risk models that deliver preventive and restorative resources to disadvantaged sectors of the population’. Terms such as ‘risk configurations’, ‘risk assemblages’, and ‘transformative risk subject’ are used to conceptualize the deployment of risk for productive ends, with risk assessments mobilized in relation to multiple logics, strategies, and discourses (Hannah-Moffat, 2005; Hörnqvist, 2010; Maurutto and Hannah-Moffat, 2006; O’Malley, 2004). Furthermore, the ‘number-based logic’ of actuarial risk assessments and sentencing guidelines is often translated by legal actors into narrative form and biographical accounts, thereby making sense of penal subjects in sentencing procedures (Lynch, 2019).

The literature on risk assessment not only captures various examples of penal change over the past few decades, but also describes the distributional effects of risk assessments and the ways actors use, negotiate, and resist risk-based practices. First, standardized risk instruments influence decisions on correctional services, sentence plans, prison placements, and supervision levels (Monahan and Skeem, 2016; van Eijk, 2020). According to van Eijk (2020), risk-based practices have both inclusionary and exclusionary effects since they connect individuals to correctional services. Scholars have shown how standardized risk-assessment instruments may amplify social inequalities through their structure and internal logic (Hannah-Moffat, 2013a; Harcourt, 2007; Ugwudike, 2019). The statistical bases of risk assessments often represent a criminal justice policy in which marginalized groups have been and remain the subject of disproportionate profiling (Hannah-Moffat, 2013a; Harcourt, 2015).

Second, Hannah-Moffat et al. (2009: 407) describe how risk operates through a ‘negotiated process’ in which front-line practitioners with discretionary powers determine the kind of recommendations that can be made. Risk-assessment instruments incorporate numerous subjective decisions made by practitioners; they also introduce implementation problems and are made in alignment with organisational goals (Hannah-Moffat et al., 2009; Maurutto and Hannah-Moffat, 2006; Werth, 2019). Studies have shown practitioners’ limited understanding of the instruments’ theoretical premises, especially regarding dynamic risk factors (Hannah-Moffat et al., 2009). Risk-assessment instruments have also met resistance from some practitioners, who renegotiate the content of risk assessments based on their own professional logic and moral stance (Hannah-Moffat et al., 2009; Persson and Svensson, 2011). Nonetheless, practitioners’ discretionary power to frame and influence risk-based decisions also reveals how social inequalities, concerning, for example, class, race, and gender, are reproduced in the correctional field not only by the internal logic of the standardized risk instruments but also through a range of subjective judgements (Hannah-Moffat et al., 2009; Harcourt, 2007). The ambiguity between the clinical and actuarial is articulated by Harcourt (2007: 238), who argues for more randomization in policing and sentencing practices to counteract subjective judgments, profiling, and the unintended effects of predictions.

Ideational change: Evidence-based prison policies and the rise of risk governance

Since the mid-1990s, evidence-based prison policies, including cognitive–behavioural treatment programmes, individualized sentence plans, and the development of standardized risk-assessment instruments, have been a high priority in PPS (Hörnqvist, 2010). In Sweden, as in many other countries, offender treatment had previously been driven by a trial-and-error approach. In 1995, PPS launched a comprehensive project to initiate evidence-based treatment programmes aimed at reducing reoffending (Kriminalvården, 2014). This new policy agenda was not a wholesale replacement or a break with past approaches. Instead, it was seen as improving on an already progressive, humane, and rehabilitative prison agenda. Treatment programmes in the 1970s and 1980s had been wide-ranging and included a rich plethora of ideas, programmes, and ambitions to reintegrate inmates into society (Krantz, 2014; Nilsson, 2013).

In 1996, PPS implemented national guidelines with formal definitions and rules for treatment programmes based on individual needs, using cognitive–behavioural techniques, and tailor-made for certain offender categories (Kriminalvården, 2014). At the time, the scale and intensity of prison reform was massive. In October 1996, for example, PPS organized an exhibition at the Hall high-security prison, in which more than 1300 employees participated in the 50 showcased treatment programmes, demonstrating new ideas, developments, and practices in offender treatment (Kriminalvården, 2014).

The ideational developments at the end of the 1990s were facilitated by the introduction of new actors and governance arrangements. In 1999, a strategic network was created to oversee and facilitate the development of evidence-based treatment programmes, administrative procedures, and risk-assessment instruments (Kriminalvården, 2014). In 2002, the strategic network also introduced an accreditation panel of researchers to guarantee the effectiveness of treatment programmes. The accreditation process aimed at measuring whether new treatment programmes should be approved or discredited based on a set of predetermined indicators covering scientific validity, focus on criminogenic risk/need factors, programme consistency, and processing manuals (Kriminalvården, 2014).

Several penal practices have been closely aligned with the programmatic ideas of evidence-based prison policies. Standardized risk-assessment instruments and individual sentence plans are probably the most significant examples. The implementation of risk-assessment instruments, however, has not been straightforward; rather, it has been characterized by reluctance, failure, and revision (Bäckström Johansson et al., 2019). In the early 2000s, prison and probation officers’ risk assessments were unstructured, based on personal and professional experience, and lacked a scientific basis (Bäckström Johansson et al., 2022). In a study of 1320 pre-sentence reports (PSRs) between 2009 and 2010, Persson and Svensson (2011) showed that risk assessments were conducted in only 3 to 4 per cent of reports. When the National Audit Office undertook an audit of individual sentence plans and risk-assessment practices in 2009, it found major shortcomings (Riksrevisionen, 2009). PPS responded to the criticism by initiating an action plan in 2011, which resulted in the implementation of RNR-A.

Since 2016, RNR-A has been an integral part of the new sentence plan (Kriminalvården, 2016). RNR-A must be performed for all individuals who have a sentence plan – that is, 8500 in prison, 6500 on probation, and 1500 on electronic monitoring annually (Bäckström Johansson et al., 2019). RNR-A is informed by Andrews and Bonta's (2010) classification of risk factors for criminal conduct and focuses primarily on the four key risk factors of criminal behaviour: antisocial history, antisocial personality pattern, pro-criminal attitudes and values, and pro-criminal associates. The instrument has two parts. The first has 15 items covering static risk factors based on register data on prior criminal history and antisocial attitudes or behaviours. The second has 72 items and covers dynamic risk/need/responsivity domains in ten areas, which are assessed through a semi-structured interview. Part two also contains one item that deals with gang involvement as a specific risk/need factor, which notably is the only item scored dichotomously.

Moreover, RNR-A includes a computerized decision aid that sums the scores and suggests risk levels (low, medium, and high) for three separate domains: general criminality/violence, intimate partner violence, and sexual offending. The computer-generated scores also display the individual's need levels in all areas of the ‘Central Eight’ ² (see Andrews and Bonta, 2010).

The configuration of the risk–gang nexus

In the following two sections, we focus in greater depth on the configuration of a penal layer for managing gang-affiliated inmates as risk subjects. In PPS, the emergence of gang-related matters has encountered institutional complexity consisting of conflicting goals, specific rules and practices, and the presence of specialist roles. The configuration of a penal layer might be understood as the dominant way in which PPS has responded to this institutional complexity.

Penal layering and uneven conditions of incarceration

Although standardized risk assessments had been present for more than two decades in PPS, the introduction of RNR-A was a decisive turning point. RNR-A was integrated at multiple decision-making points: in PSRs, during prison placement, and in parole decisions at supervisory levels. Prison staff and probation officers described how the introduction of RNR-A had led to a more formalized and rationalized procedure for risk management and decision-making (see, e.g., Savelsberg, 1992). They described how the governance arrangements concerning RNR-A focused on prioritizing high-risk inmates and identifying criminogenic needs. One prison officer declared that the purpose of RNR-A had been to enhance equality in decision-making procedures and thereby minimize subjective judgments:

And it was developed only to make things as equal as possible for everyone, because you shouldn’t have any personal opinion. It's supposed to be, it should be, as equal as possible for everyone, and it should be more about what their needs are, not what they personally want to do. You really must always be careful not to form a personal opinion.

For those who were sentenced to prison, front-end risk assessments were made regarding their conditions of incarceration, that is, the prison administration's decisions about the inmate's conditions during the prison stay (Hörnqvist, 2010). A prison officer described these front-end assessments as a ‘baseline workup’:

We are supposed to do a baseline workup for everyone – RNR, risk, needs, and responsiveness. That is, an assessment has to be done and put into the computer system to see whether or not there's high risk of reoffending, medium, or low risk. And depending on that, a contact frequency is set. You are not supposed to invest in the low risks, you’re not supposed to do that. But if you say ‘gang member’ he is always going to be high.

Besides the proliferation of RNR-A, the interview excerpt shows how predetermined rules for classifying gang affiliates’ risk levels had been added to the front-end assessment. Such predetermined rules for particular groupings were contradictory to the idea of assessing individual criminogenic needs. The added rules illustrate how some institutional properties of RNR-A were displaced while others were left intact. Probation officers were critical of the contradictory logic found in RNR-A, according to which gang affiliates who fulfilled the RNR-A criteria of need/responsiveness were excluded from rehabilitative interventions. A probation officer reflected on the policy objectives – to work in accordance with RNR-A principles – and the specific rules created for gang affiliates:

And, in my experience, they have also ended up a bit excluded, because we see the need, of course, if we are supposed to work with these people who might want to change and maybe get out of this. So there is definitely a need for these services that, in some sense, they are deprived of the right to, that they aren't even considered the target group.

Prison placement had specific regulations for gang-affiliated inmates. According to one prison officer, the differences between the security classes of prisons could be understood in terms of ‘use of resources’, with high-security prisons allocating more time and resources to security issues and prisons with lower security classifications allocating more resources to reducing recidivism risks. Hence, risk-based decisions have distributional effects that shape uneven conditions of incarceration among the prison population (van Eijk, 2020). A prison officer told us that ‘they have made it a rule that if an inmate belongs to a gang or has been sentenced to deportation then they must be placed in Class One’. Shammas (2016) uses the term ‘dualization’ to describe a parallel system of punishment in penal institutions, one for citizens and one for non-citizens. Such specific regulations for gang affiliates’ prison placement were also institutionalized through the administrative category of ‘special conditions’.

For inmates with longer sentences, a special risk-assessment is conducted at the National Reception Unit at Kumla Prison detailing the specifics of the special conditions that will regulate the incarceration. The National Reception Unit, launched in 1997, had played a crucial role in the PPS's development of risk-assessment practices, as this institutional setting had made it possible to experiment with scientific ideas, develop risk-assessment practices, and construct governance arrangements based on risk logics (Hörnqvist, 2007). Whereas the standard rule for special conditions encompassed all those sentenced to a minimum of four years, gang-affiliated inmates could be assigned special conditions for any sentence over two years. The special conditions state the security class of the prison in which the inmate must be placed, whether furloughs are permitted, and whether or not the inmate has a right to early release. One prison officer told us that gang affiliates were often assigned ‘zero conditions’, meaning no furloughs, no placement in open prisons, no early release, and limited access to rehabilitative programmes:

The situation is, if you are sentenced to four years in prison, you are given special conditions. It is a matter of – actually, they control your entire sentence: what date you can be let out on furlough, whether you can be let out on furlough, what kind of furlough is allowed, which prison you can be sent to, when you must be incarcerated. So, they rule everything. And some, they don't have, often these young gang members, they don't get them right away, they have to show first, and then maybe they get them later.

The prison officer expressed the gravity and significance of the conditions: ‘So, they rule everything’. An alternative logic also emerges in the above excerpt, in which the special conditions can be changed if ‘young gang members … show first’, referring to access to correctional services and resources being contingent upon performance. Interviewees declared that gang members who did not officially state that they had left the gang were not considered responsive. This form of responsibilization was articulated in relation to the RNR-A logic and the idea of responsiveness (see e.g., Garland, 2001), revealing how the configuration of the penal layer incorporated braided logics activated under particular conditions.

Nevertheless, gang affiliates’ exclusion from access to reintegrative services was a matter of contention among the interviewees. Probation officers were sceptical about how early release was used in the prison system. A probation officer argued that early release and transition to treatment centres have become ‘a reward for good behaviour’. The probation officer used the logic of RNR-A to challenge this form of responsibilization, illuminating more broadly how probation officers struggled to shape the meaning of RNR-A in the management of gang affiliates.

Although we try not to look at early release … as a reward for good behaviour. Instead, you should base early release on the need – do you need this? The higher the risk of reoffending, the greater your need to connect with services. But then our hands are also tied, we are often frustrated because the prisons do not release the individual.

The introduction of specific rules for managing gang affiliates as risk subjects, the use of which became dominant, was the way PPS responded to the institutional complexity of gang-related matters. In most cases, these rules precluded the rehabilitative trajectories of the prison system and led to the dualisation of gang affiliates and other groups of inmates (Shammas, 2016). The topographical characteristics of penal layers are thus revealed through amendments to the existing penal arrangements that shaped uneven patterns of punishment.

Penal layering and exclusion from community-based sanctions

The management of gang affiliation as a specific risk/need factor can be considered in relation to the more welfare-oriented features of the correctional field. Probation is one of the institutional prototypes of penal welfarism and community-based sanctions (Simon, 2013). In the 1970s, the idea of probation and the ‘principle of normalization’ flourished in the Swedish criminal justice system. Since the 1998 organizational reform that integrated the Probation Services and the Prison Services, the autonomous role of probation services has gradually diminished (Svensson, 2001). Probation officers expressed frustration over their limited discretionary powers and their gradual diminishment in recent years, which became increasingly evident with the formalization of RNR-A in PSRs and the rules for gang affiliates’ access to community-based sanctions.

In Sweden, probation decisions are made by the courts, based on information in PSRs covering personal background, socio–psychological profile, present life conditions, and the assessment of treatment needs (Persson and Svensson, 2011). The preparation of PSRs is intended to be guided by RNR-As (Kriminalvården, 2016).

Although probation officers are meant to identify rehabilitative needs in individual gang affiliates and recommend appropriate treatment programmes, those interventions are often prohibited to individuals with continuing gang affiliations. Probation officers explained that the current rule was that gang members were not eligible for community-based sanctions. A probation officer declared that ‘a gang member cannot be sentenced to contract treatment, [and] he cannot be sentenced to care and treatment instead of prison’. PSRs revealed a process in which specific rules, decisional hierarchies, and organizational structures were added on top of the probation officers’ ordinary methods of casework, interviewing, and risk assessments. Most apparent was the launch of a security organization function, which had a security coordinator retrieving and distributing information about gang affiliation. This intelligence work was framed within a risk discourse and articulated as ‘horizon scanning’: it incorporated updates on the national gang situation, the gang environments of certain cities, and present conflicts, networks, and relations within and between gangs. It also introduced a decisional hierarchy into the pre-sentence context, which removed the discretionary powers of probation officers. A security coordinator described the process of confirming gang members:

So, we are talking about someone being a confirmed member of a network or gang, and there is information from – this is a bit sensitive – but there is information from two sources that such is the case. Just that someone is part of something. Then we stick out our necks and say that in the Prison and Probation Service's judgement, Joe is part of a criminal network based in x city or x neighbourhood or whatever. And we make that statement in contexts including our pre-sentence reports.

Probation officers described how the specific rules for gang affiliates had made it almost impossible to carry out individualized treatment programmes, since predetermined rules stood in conflict with the aim of individual sentence planning. In 2016, PPS launched the gang-treatment programme Entré. Created for the probation and parole context, it could be seen as a continuation of PPS's rehabilitative ideas of providing evidence-based treatment for certain high-risk offender groups. A probation officer reported that the implementation of Entré was novel. ‘This group [gangs] has never had any treatment before … with Entré you have succeeded in making a programme that targets a high-risk group that is very hard to reach’. Because gang affiliates were not eligible for community-based sanctions, probation officers faced conflicting goals in relation to Entré. A probation officer labelled this situation ‘a huge dilemma’:

Entré has a target group, gang members, and the programme consists of non-custodial sentencing, but if clients cannot be sentenced to probation with Entré, simply because they have the gang label, what do you do then? It is impossible.

While gang affiliates were excluded from rehabilitative services in the probation context, the frequency with which they faced supervision under parole was significant. Back-end risk assessments determined the level of supervision, which for gang affiliates was often weekly for a year. A probation officer described the use of RNR-A in the parole setting: ‘This new RNR-A makes us come into frequent contact with those at high risk, which includes those with gang affiliations’. This finding illustrates how gang affiliation as a specific risk/need factor was configured to fit different decision-making points in the criminal justice system, where the operationalization of RNR-A followed different logics.

Probation officers repeatedly emphasized the importance of needs and responsiveness in managing gang affiliates. However, the probation officer's power to influence the purpose and character of gang interventions was severely limited by the specific rules created to regulate gang-related matters. The probation officers’ struggles reveal how particular interludes of penal layering can include agonistic dimensions and a struggle over the configuration and meaning of penal layers (Goodman et al., 2015).

Discussion

Using the concept of penal layering (Rubin, 2016), we have shown that introducing gang affiliation as a specific risk/need factor has given rise to a layer of specific rules, practices, actors, and decisional hierarchies for managing gang affiliates. While this variegated layer produced institutional complexity concerning gang-related matters, we could identify how this layer was configured in PPS through a number of additions, including specific rules for risk classification, intelligence work, decisional hierarchies, constraints on discretionary powers, and displacement of some institutional properties. Conceptually, we understand this configurative act as an interlude of penal layering. In most cases, this layer enforced punitive penal change through assigning gang affiliates to high-security prisons, restrictive conditions of incarceration, and exclusion from rehabilitative services.

We conclude with three points about penal layering as a mode of penal change, a critical analytical tool, and a useful concept for explaining duality in penal institutions. Penal layers are configured rather than preordained. In highly regulated institutions, such as prisons and probation services, less room is left to set up alternative institutions, displace dominant logics, and propose wholesale changes (Mahoney and Thelen, 2010). Therefore, it is more likely that actors will work within the existing system by adding amendments on top of original rules and finding ways to cope with tensions. Although the introduction of gang affiliation as a specific risk/need factor triggered institutional changes, this interlude was never preordained. Rather, we identified an interlude of layering that configured different contents, additions, and institutional trajectories into a distinctive penal layer. Rubin's (2016: 432) account, that new penal layers are formed in an ‘ongoing relationship with existing layers’, became evident in the configuration of the multiple, coexisting, and braided logics of risk, incapacitation, and rehabilitation for managing gang affiliates. Moreover, we noted how standardized risk assessments were a key organizing principle for managing this institutional complexity, formalizing organizational functions, and institutionalizing specific rules for decision-making (Maurutto and Hannah-Moffat, 2006). Nonetheless, the use of standardized risk assessments is flexible and can advance different logics, interests, and policy objectives (Maurutto and Hannah-Moffat, 2006), a point that became especially apparent in the probation officers’ struggle over the application of RNR-A in relation to gang affiliates. Interludes of penal layering, therefore, recognize the decentred struggles among different actors in the penal field who shape the content and characteristics of particular incremental steps of penal change (Goodman et al., 2015).

Penal layering is a critical analytical tool for challenging monolithic explanations of penal change. In the Introduction, we noted a critical juncture in the Swedish criminal justice system. Criminologist Tham (2022) claims that the scale and intensity of punitive law reforms in Sweden – more than 50 policy proposals and law reforms concerning criminalization and sentencing enhancement in the past eight years – must be ‘some kind of historical record’. Although this interlude signals a ruptural transformation, the role of critical research must be to destabilize monolithic and mechanical explanations and excavate the conditions of possibility for such a development. Like Sewell's (2005: 227) conceptualization of events, the bracketing of particular interludes of penal layering for synchronic analysis could be useful to describe these interludes’ inner workings and their part in the durable transformation of structures. That being said, penal layering might serve as a critical analytical tool with which to map the variegated mixture of particular instances and historical trajectories that might accumulate into broader patterns of macro-scale penal change. We show, for instance, how ideational changes in prison policies over the past three decades were important for understanding why particular offender groups have become the target of selective measures and the development of risk governance. Furthermore, our focused analysis revealed the topographical characteristics of penal layers, and showed how uneven conditions of incarceration and exclusion from reintegrative services were shaped, anticipating the broader penal reforms and policy proposals of anti-gang legislation. Besides tracing historical–institutional developments, our notion interludes of penal layering could further discussions of methodological approaches to operationalizing Rubin's concept by, for example, analysing interconnections between concurrent processes of penal layering and working laterally between different cases, contexts, and institutional spaces (see Rubin, 2016: 436).

Penal layering is useful to explain contradictory patterns of penal change and the duality in penal institutions. Because this study was conducted in Sweden, penal change in societies characterized by egalitarian welfare states and more humane penal regimes is of particular interest. We aim to advance scholarly discussion by putting forward penal layering as particularly valuable for explaining the coexistence of different conceptions of punishment and penal technology, and the uneven treatment of different groups in penal institutions (Barker, 2018; Barker and Scharff Smith, 2021; Shammas, 2016). The targeting of high-risk subjects characterizes the topographical character of penal layers, where practices, rules, and actors have been introduced to control, discipline, or punish certain segments of the offender population. This theme recurs in scholarly discussions of the contradictory logics found in Nordic penal regimes, namely ‘equality and freedom for some but not others’ (Barker and Scharff Smith, 2021: 1553). The concept of penal layering could be useful to explain the endogenous development of an uneven, parallel system of punishment for different offender groups in Nordic penal institutions, responding to Barker and Scharff Smith's (2021) call to find analytical categories that can capture the apparent duality.

Policy implications

In addition to the article's conceptual elaborations, its findings have important policy implications for risk governance in the correctional field. This analysis of penal layering identified the mechanisms that displaced institutional properties of the RNR-A logic and precluded rehabilitative services for gang affiliates, a displacement effect that is not arbitrary. The depth of penal layers revealed here indicates the path-dependency of institutional properties that can be reactivated and used across various contexts and times, including, in risk assessments, for the overestimation of static risk factors and alignment with selective incapacitation. Dichotomously scoring gang affiliation in RNR-A construes gang affiliation as a static risk factor, one that undermines the idea of dynamic risk/need factors. More flexible RNR-items should be developed that could capture the ambiguities of gang life and the individual's responsiveness regarding leaving the gang, thereby also ensuring socially just risk-assessment procedures. As we have argued elsewhere, acknowledging that gang life is filled with tensions, contradictions, and unfulfilled promises is crucial for supporting individuals to leave gangs.