Assembling Risk

doi:10.1093/bjc/azi073

BRIT. J. CRIMINOL. (2006) 46, 438–454 Advance Access publication 27 July 2005

ASSEMBLING RISK AND THE RESTRUCTURING OF PENAL CONTROL P A U L A M A U R U T T O and K E L L Y H A N N A H -M O FF A T * In this paper, we draw attention to new assemblages with risk in order to highlight the multiple forms of knowledge and logics at work in new risk assessment practices. We seek to complicate the theoretical explanations of risk by highlighting how risk logics merge and shift in tandem with various rationalities. For example, when risk is merged with need, needs are reconfigured as criminogenic needs but, in this process, risk becomes a fluid concept that can be treated, altered and transformed. When risk is merged with more welfare and disciplinary-based logics, such as rehabilitation and clinical assessments, new forms of risk management are produced, such as targeted treatment. Through these processes, risk’s association with actuarial calculations is weakened by other judgments and appraisals. As well, risk takes on more productive ameliorative possibilities, associated with risk minimization. These new assemblages enable new forms of risk-based governance as evident in contemporary correctional case management planning and the accreditation of programmes. This analysis is developed through an examination of the Level of Service Inventory (LSI)—an internationally used risk assessment instrument. The past few decades have witnessed the ascendancy of risk as a key organizing principle of contemporary correctional practice and offender management. International scholars have documented the significance and effects of risk on institutional practices (Feeley and Simon 1992; 1994; O’Malley 2000; 2001; 2004; Stenson and Sullivan 2001; Robinson 2002; Baker and Simon 2002: Ericson and Doyle 2003; Hudson 2003; Kemshall 2003). Analyses of risk that evolve from Feeley and Simon’s (1992) archetypal ‘new penology’ and ‘actuarial justice’ tend to articulate risk as a static, unified, homogeneous process that comparably crosses continents and criminal justice sites. Actuarial risk has taken on a hegemonic dominance that supercedes other models of governance, such as welfare and disciplinary forms of regulation. Yet, critiques of the ‘new penology’ suggest the coherence and global character of risk governance and the associated erosion of welfare practices are overestimated (Simon 1995; 1996; Kemshall et al. 1997; Kemshall 1998; 2002; 2003; Lynch 1998; Robinson 1991; 1999; 2001; 2002; Sparks 2000; Kemshall and Maguire 2001; Miller 2001; Leacock and Sparks 2002; Hannah-Maffat 2005). Present analyses of risk have not fully explored the complexity, ambiguity and impact of recent developments in risk identification and management strategies on institutional regimes. Increasingly, the actuarial form of risk that typifies the ‘new penology’, and is closely aligned with punitive penal politics, is advanced as only one form of risk governance on the penal landscape. Recent examinations (O’Malley 2000; 2002; 2004; Hannah-Moffat 2005) of local institutional practices demonstrate that not only is the grip of actuarial justice more tenuous and less coherent than previously envisioned but * Department of Sociology–UTM, University of Toronto, Canada; [email protected]. Kelly Hannah-Moffat, Department of Sociology–UTM, University of Toronto, Toronto, Canada; [email protected].

438 © The Author 2005. Published by Oxford University Press on behalf of the Centre for Crime and Justice Studies (ISTD). All rights reserved. For permissions, please e-mail: [email protected]

ASSEMBLING RISK AND THE RESTRUCTURING OF PENAL CONTROL

that multiple applications and interpretations of risk exist within institutional frameworks. Critical scholars are now concerned with the fluidity of risk technologies and how risk-management practices incorporate an assortment of logics and frameworks. In this paper, we draw on O’Malley’s (2004: 8) latest theorization of risk as ‘a fluid, inclusive, heterogeneous array of practices with diverse effects and implications’ to analyse how ‘risk is assembled into complex configurations with other technologies’ (O’Malley 2004: 27). The concept of assemblages, adapted from Deleuze and Guattari’s (1998) work, is used to explore the interface between risk and other logics, such as clinical assessment, rehabilitation and welfare practices that are increasingly shaping criminal justice sites. The focus on assemblages with risk enables us to counteract the presumption that risk technologies necessarily operate in opposition to rehabilitation or more welfare-based approaches. Rather, we underscore how new penal technologies combine, merge and continually reassemble risk with other logics in response to various institutional agendas. As risk technologies are being reinvented and retrofitted, often in unpredictable ways, the concept of risk itself shifts. Its association with actuarial calculations is weakened as it takes on more ameliorative, productive and rehabilitative connotations. Risk increasingly is conceived of as a fluid concept that can be minimized, treated and continually reassembled. More specifically, we examine how various assemblages with risk have structured a popular ‘avant-garde’ internationally used series of risk/need assessment tools—the Level of Service Inventory1 (originally named the Level of Supervision Inventory, and more commonly known as the LSI). Attention to the technical transformations made to the LSI provides insight into the convergence of risk with other logics, in this case, clinical assessment, rehabilitation and need. It also underscores how these varied LSI assemblages are enabling new forms of governance. The widespread use and modification of the LSI tools make them a logical focus of analysis. The LSI was originally developed in Ontario, Canada in the late 1970s and quickly developed international notoriety. Currently, the tool is used in jurisdictions throughout Canada, the Untied States,2 the United Kingdom and Australia, among others. The tool, originally written in English, is available in Spanish, Croatian and French (French European and French Canadian), and it is in the process of being translated into Dutch and Icelandic.3 It is one of the most extensively researched offender classification instruments. Versions of the LSI are used at several stages of the criminal justice system to assess men, women and young offenders for pre-sentencing and sentencing decisions, as well as for institutional placement. They are used in both community (probation/parole) and institutional correctional settings to determine institutional placement, programme needs and the threat of an offender to him/herself or others. They are also routinely reapplied during the custodial sentence to inform significant administrative decisions regarding case management, transfers, programme completion and release. This tool has undergone several modifications and enhancements; over ten versions of the tool 1 Although we refer here to the LSI, readers need to be aware of the fact that several versions of this tool exist. The paper illustrates the evolution of the tool to make a broader conceptual argument but it does not detail the specific changes in each version of the tool. For a more detailed description of the contents of these tools and the science informing them, see Andrews and Bonta (1998). 2 Multi Health Systems—the company that markets the LSI–R—indicates that more than 600 agencies in the United States currently use this risk/need tool (Lowenkamp et al. 2004). 3 Information obtained from correspondence with Multi-Health Systems Inc., 22 September 2003.

439

MAURUTTO AND HANNAH-MOFFAT

are in circulation. For the purposes of this paper, we group the various versions into two main categories: (1) the Level of Supervision Inventory (LSI) group, which was initially developed for supervision purposes; and (2) the newer Level of Service/ Case Management Inventory (LSI/CMI) series of tools, designed to target treatment interventions.4 New risk configurations inherent in tools like the LSI are important not only for theorizing the concept of risk, but also for understanding new penal strategies. The production of varied assemblages with risk creates the conditions for new forms of governance which are often more pervasive and intrusive. Each formulation of the tool enables it to take on new meanings, to be deployed for ever more assorted purposes, to be transported to new institutional sites and be used to assess more diverse populations. For example, the repackaging of the LSI has enabled the tool to be deployed for purposes other than supervision or security classification. Retrofitted risk-assessment tools are currently used to target treatment, to minimize risk and to direct case management plans; they are used to structure accredited treatment programmes and to make decisions about institutional resource allocation. In essence, the reassembling of risk technologies allows for new possibilities of governance. A Genealogy of Risk Assessments Our observations of the genealogy of risk assessment tools and their fluid role in offender management illustrate the coexistence of a range of risk assemblages. It is these diverse risk configurations and their corresponding effects that we explore by paying attention to how risk assessments have been subjected to an array of modifications. This genealogical approach differs from most historical accounts of risk assessments that chart their progressive evolution from clinical to actuarial risk prediction. For example, Bonta (1996) describes a three-stage generational development of risk assessments, beginning with clinical and advancing to more sophisticated actuarial models. Such depictions connote a linear development in which advancements in actuarial approaches eventually displace clinical models. By contrast, our project is to unravel how different technological forms embody multiple logics and practices that converge to produce a particular form of risk-mediated governance. More specifically, we examine how calculative actuarial technologies have been merged with clinical, social, rehabilitative and therapeutic strategies. The article pays specific attention to what may appear as minute tinkering or insignificant changes in the assessment of risk but which, in fact, reveal the conflation of new modes of governing and the production of new assemblages. In so doing, we are able to highlight how risk technologies, far from being monolithic, are constantly being reinvented, retrofitted and reassembled in response to institutional agendas. Actuarial risk assessments are often presumed to have developed much later than, and to be more objective tools that supersede, clinical risk strategies. Both approaches, however, have a long history, dating back to the 1920s (cf. Hart 1923; Burgess 1928). 4 Various versions of the tool have been created, including the Level of Supervision Inventory (LSI) versions I–VI; Level of Service– Revised (LSI–R); Level of Service–Ontario Revised (LSI–OR)—an international version of this will soon be released, named the Level of Service–Revised/Case Management Inventory (LSI–R/CMI); Level of Service–Revised–Screening Version (LSI–R–SV); Youth Level of Service–Case Management Inventory (YLS/CMI); Young Offender–Level of Supervision Inventory (YO–LSI).

440

ASSEMBLING RISK AND THE RESTRUCTURING OF PENAL CONTROL

They represent different conventions in the identification and classification of risk, but they are by no means the only means of producing risk. Clinical approaches rely on psychological tools and personality inventories that survey individual characteristics and traits. By contrast, psychologists trained in cognitive behaviouralism and social learning developed actuarial approaches premised on aggregate statistical calculations (using multiple regressions, correlations and meta-analyses) that predict an offender’s likelihood of future recidivism. Neither clinical nor actuarial approaches are technically neutral assessments; they are both imbued with cultural assumptions and, thereby, moral and political strategies of governing (Latour 1987; O’Malley 2004), which can be mobilized in support of a particular political agenda. Actuarial assessments, while in existence since the 1920s, rise to prominence in the 1970s because they are easily aligned with the dominant political and administrative priorities of the time (Glasner 1985; O’Malley 2004). By the 1980s, actuarial tools were widely used in correctional settings to predict recidivism and level of risk. Some of the more prominent tools included the Salient Factor Score (SFS) in the United States (Hoffman 1983), the Risk of Reconviction (ROR) scale in the United Kingdom (Copas et al. 1996) and the Statistical Information Recidivism scale (Nuffield 1982) in Canada. Actuarial assessments appealed to correctional administrators because they focused less on individual characteristics and more on ‘objective’ criteria (i.e. type of offence, prior criminal history, age, gender and sentence length). The criteria contained in these tools are easily identifiable and verifiable, and hence can be quickly scored by correctional staff, thereby reducing the need for extensive, laborious assessments by professionally trained clinicians. They also maintain a veneer of objectivity and ‘truthfulness’ because of their reliance on and production of numerical calculations. Criminologists such as Feeley and Simon (1992; 1994), who observed the proliferation of actuarial tools and the seeming decline of more clinical approaches as well as the rise of incapacitative policies, argued that these trends signalled a new form of penal managerialism or ‘actuarial justice’. The form of actuarial justice identified by Simon and Feeley is predictive, statistical and concerned with efficiently managing the ‘risk’ offenders represent. This particular actuarial assemblage of risk is exclusionary and typically juxtaposed to more modernist socially inclusive and discretionary forms of penal welfarism. The scholarly emphasis on actuarial risk assemblages tends to overlook some of the significant distinctions and shifts occurring in risk governance. While continuing to inform decision making, the use of ‘pure’ actuarial risk tools were discredited by researchers and correctional workers in the mid-1980s for relying too heavily on static (unchangeable) variables that measured criminal history, and for ignoring ‘criminogenic need factors’ (also known as dynamic variables) that change over time. These static assessments produced a fixed level of risk that facilitated supervisory decisions, but offered little by way of long-term management strategies. For instance, Wormith (1997: 1) notes that the risk logic embedded in actuarial tools provides ‘no instruction or direction for the type of management and treatment of an offender most likely to bring about positive change, therefore limiting the capacity to help staff lower an offender’s degree of risk’. Static risk models offered practitioners little guidance in the determination of treatment needs or programme allocation which many practitioners regard as central to risk minimization and management. 441

MAURUTTO AND HANNAH-MOFFAT

Many practitioners were deeply invested in the belief that offenders can change and, consequently, they continue to defend the merits of rehabilitation. This led correctional researchers to experiment with assessment models that merged actuarial and clinical approaches.

Recalibrating Risk: The Production of Risk/Need Assessments In the mid-1980s, newly revised risk/need scales were advanced that combined static (unchangeable) variables with dynamic indicators (changeable usually with treatment interventions) that are said to directly bear on recidivism (Andrews and Bonta 1998: 224). Examples of dynamic items include substance abuse, use of leisure time, relations with family and antisocial attitudes or behaviours, among others. These items are typically identified as ‘criminogenic needs’ which, if ‘treated’, lower recidivism and, consequently, risk. For example, substance abuse is known to increase recidivism, but treatment programmes can cure or diminish such problems, thereby lowering recidivism. Unlike previous understandings of risk, risk/ need assemblages reassert the archetypal assumption of correctional treatment— that offenders can change and that positive changes in behaviour will reduce the risk of recidivism.5 This new configuration of risk/need makes it possible for practitioners to adjust risk levels over time as offenders’ circumstances change in response to treatment participation. The LSI group of risk/need tools represents perhaps the most wisely used and researched assessment tools. Research on the LSI began in the 1970s, when probation and parole officers were required to make more transparent decisions (regarding levels of supervision and the type of services clients needed) and to distribute scarce resources in an efficient and fiscally responsible manner (Andrews and Bonta 1998: 229). The LSI incorporates the findings of the recidivism literature, professional opinions of probation officers and a social learning perspective on criminal behaviour (Andrews and Bonta 1998). The merger of static with dynamic risk factors is promoted as the strength of the LSI tools and is what presumably makes them superior to psychologically based classification systems like the Megargee MMPI (Megargee and Bohn 1979), Quay’s AIMS (Quay 1984) and the I-level (Sullivan et al. 1957) that only classified offenders according to their treatment needs and assume that ‘needs assessment were some how fundamentally different from risk assessments’ (Bonta 1996). The tool is designed to decrease correctional controls for those who do not require it and to identify the dynamic risk factors that can be targeted in treatment and monitored for change (Andrews and Bonta 1998: 232). The LSI group of tools has undergone several modifications. When first developed, the tool was referred to as ‘The Level of Supervision Inventory’ and was advanced as a ‘guide’ for determining appropriate ‘levels of supervision’ and ‘actual amount of supervision’ (Andrews 1982: iii). Over time, both the name and purpose of the tool changed. By the time the Level of Service-Revised (LSI-R) was released in the 1990s, 5 Examples of such tools include the Wisconsin Risk and Needs assessment instrument (Baird et al. 1979), the Community Risk– Needs Management Scale, developed for the Correctional Service of Canada to assist in the supervision of parolees (Motiuk and Porporino 1989) and the Level of Service Inventory (LSI) (Andrews and Bonta 1995).

442

ASSEMBLING RISK AND THE RESTRUCTURING OF PENAL CONTROL

the tool had been renamed ‘The Level of Service Inventory’ in an effort to underscore the new priority placed on treatment above and beyond that of supervision. A noteworthy observation is that despite the shift in purpose, the different versions of the tool look relatively similar. The number of items changed slightly; the original LSI contained 62 items, whereas the LSI-R includes 54. These items are grouped into the following ten factors: Criminal history (ten items), Education/Employment (ten items), Financial (two items), Marital/Family (four items), Accommodation (three items), Leisure/Recreation (two items), Companions (five items), Alcohol Problems (nine items), Emotional/Personal (five items), Attitudes/Orientation (four items). In versions one to six of the LSI, each item was scored according to a ‘yes/no’ format and the final risk/need score was used to classify offenders on a five-point scale, ranging from ‘very low’ to ‘very high’ risk. The LSI-R also uses a ‘yes/no’ scoring system to determine overall risk/need levels, but it allows certain dynamic risk factors (those which are amenable to interventions) to be further ranked on a three-point scale. The final risk score indicates presence or absence of an item, whereas the ranking aids in the determination of case-management strategies. These rankings are used to profile the criminogenic needs of a client, in order to direct intervention. Scores are conveniently plotted on a ColorPlot Profile form that allows for quick comparison with a large normative base and provides criteria for making decisions regarding institutional classification, probation, parole and halfway-house placement (Andrews and Bonta 2002: 5). The assessment tool includes open, unstructured text space for the assessor to document ‘special circumstances’ and reasons for ‘professional discretion overrides’. Here, the importance of professional judgment is acknowledged and, supposedly, ‘permission’ is given to override the final assessment. The combination of static risk with dynamic need factors allows for a fluid, malleable understanding of risk. Pure actuarial tools produced a ‘fixed risk subject’ that denied the possibility of change in the offender and thus threatened and delegitimized correctional interventions. The application of risk was confined to supervision and security classifications. By contrast, the merger of risk with dynamic need factors provides for a ‘transformative risk subject’ who is amenable to change via targeted treatment strategies (Hannah-Moffat 2005). It becomes possible to identify effective interventions which target aspects of an offender’s life that are identified as risky yet amenable to change. Here, risk is understood as a fluid concept; it can be adjusted, treated and manipulated. This new configuration allows for an intensification of risk-based technologies within criminal justice institutions. Moreover, the transient nature of risk produces the need for constant re-evaluation, as offenders must be monitored for potential changes to overall risk/need scores—a fortuitous outcome for the developers of the tool who are remunerated for each use of the tool. In an interview with one of the developers of the LSI, it was suggested that, ideally, offenders should be reassessed every four months (Hannah-Moffat and Maurutto 2004). The understanding of what constitutes an offender’s need likewise shifts when need is conflated with risk. Need is reconfigured as criminogenic need—a need that is a priori determined on the basis of its correlation, and hence potential impact on, recidivism (Hannah-Moffat and Maurutto 2004; Hannah-Moffat 2005). Criminogenic needs are narrowly defined as problems which influence the chances of recidivism, rather than a statement of resource entitlements or self-reported offender needs (Hannah-Moffat 2005). Variables that are significant but not related to recidivism, yet require intervention, 443

MAURUTTO AND HANNAH-MOFFAT

are deemed non-criminogenic needs (i.e. poverty, health) and are considered a low priority in terms of intervention, except for humane consideration. An ‘intervenable need’ is not an individual’s self-perceived need,6 but rather is a characteristic an individual shares with a population that is statistically correlated with recidivism.7 An intervenable need is defined not only through the availability of resources and structural arrangements that allow for intervention and possible amelioration, but also through statistical knowledge of it as a variable that is predictive of an undesirable and preventable outcome: recidivism.8 Hence, individual circumstances that may very well lead to recidivism but cannot be shown to be statistically significant for the larger sample population (i.e. mental health) are excluded. Criminogenic needs are characterized as universal predictors of risk that can easily be transported internationally, at least within Western countries. Little attention is paid to gender, ethnic or racial differences or the differing social, economic or political contexts in which these tools are deployed. This raises a concern when these tools, originally tested and developed for use with male offenders in Ontario, are deployed with women and youth internationally. The assumption is that criminal behaviour and criminogenic need that precipitate recidivism are universal phenomena that are detached and unaffected by legal or criminal justice practices or political, economic or social relations. Differences in criminal law, incarceration rates and criminal justice policies would presumable have marginal impacts on LSI scores. Recently, new studies examining the validity of these tools have begun to question their geographical portability. Research produced in Canada by the authors of the tools and their students have reported that LSI tools can predict recidivism reasonably well (Andrews 1982; Andrews and Bonta 1994; 1998; Bonta and Motiuk 1995; Motiuk et al. 1986; Bonta and Motiuk 1987; Bonta 1989; Bonta and Motiuk 1990; Coulson et al. 1996; Loza and Simourd 1994). These studies have yielded statistically significant Pearson correlation coefficients in the area of 0.30 and 0.40. Most of these studies were conducted on male offenders within the province of Ontario, Canada. Similar validation studies in the United States, however, have produced less convincing results (Dowdy et al. 2002; O’Keefe et al. 1998; Philbrick et al. 1993). These studies conducted on community correctional clients in Colorado not only question the predictive validity of LSI assessments, but also raise concerns over whether such tools can be easily transported to other countries, or, for that matter, other provinces in Canada itself (O’Keefe et al. 1998; Philbrick et al. 1993; see Dowdy et al. (2002) for an extensive summary of validation studies on the LSI conducted in Canada and the United States). More importantly, the statistical calculations that make up the LSI risk scores obscure a range of subjective and moral judgments. A number of academics have identified how probability scores and, in particular, risk assessments are not inherently objective; they obscure a range of subjective and arbitrary decisions—a process that Rose refers to as the ‘black box’ (Rose 1998: 187). Risk assessments include a number of arbitrary decisions, ranging from the selection of items, the number of risk levels and the threshold range, all of which can be adjusted, altered and modified to fit the 6 However, efforts are under way to develop reliable self-report instruments for the assessment of criminogenic needs (Serin and Mailloux 2001). 7 For detailed information on the statistical determination of variables as needs, see the Forum on Corrections Research, September 1998, Vol. 10, No. 3—special issue on dynamic factors. 8 See Hannah-Moffat (2004) for a more detailed discussion of an ‘intervenable need’.

444

ASSEMBLING RISK AND THE RESTRUCTURING OF PENAL CONTROL

needs of criminal justice settings (Hannah-Moffat and Maurutto 2004; Kemshall 2003).9 For instance, risk levels and thresholds can vary from institution to institution.10 According to Wormith (1997), ‘the number of risk levels in any scale or instrument is decided arbitrarily by the developer or the agency using it’. As resources shrink or swell, the tools can be adjusted accordingly. The practitioner literatures (Kreamer 2004; Dal Pra 2004) clearly state that the development and adoption of particular risk tools (actuarial risk versus risk/need) are decisions that ought to be made in concert with and evaluation of correctional mission statements and resources revealing that risk technologies are used instrumentally by correctional organizations. Clearly, corrections organizations select the risk technologies that ‘fit’ their agency’s vision and mission. Moreover, the logic of risk/need calculation and the meaning of the LSI-R scores require careful consideration. The items in these and other similar tools are said to represent and ensure that ‘a reasonable comprehensive survey of attributes of offenders and their situations has been conducted prior to decisions being made’ (Andrews 1982: 25). They represent risk and criminogenic indicators that have been shown to be correlated with recidivism—and not, necessarily, danger. However, the correlation between any one indicator and recidivism is generally quite small, ‘which means that knowledge about a single factor will assist one only minimally in the prediction of criminal behaviour’ (MSGCS 1995: 1). It is the combination of loosely correlated items into a total risk score that enables the prediction of risk. Given this specific logic of tabulation, the number of items can vary considerably from scale to scale, as eliminating one item should not significantly impact the overall risk score. This flexibility is one of the fundamental strengths of the LSI approach; the tool is easily amenable to further development and refinement. The ease with which this tool can be manipulated would be substantially reduced if the items were weighted. As one of the developers of the tools notes, the ‘scoring format makes it very easy to add, delete or modify items when experience suggests that the modifications would increase the validity or utility of the instrument’ (Andrews 1982: 2). Hence, LSI instruments have and can be easily adapted, reworked and revised in response to new psychological innovations or institutional requirements. The technical computation makes it difficult to separate risk analysis from managerial decisions. The degree to which subjective decisions and institutional agendas determine probability calculations are, as Hacking argues, achieved ‘by covering opinion with a veneer of objectivity’ (Hacking 1990: 4). Discretion is not done away with, but it becomes boxed into categories and systematized and, thereby, presumed to be more neutral. The technical computation lends an authority to the risk score that is difficult to contest and appeals to practitioners. Interviews with practitioners (Hannah-Moffat and Maurutto 2004) consistently revealed that they believed that the advent of risk-assessment tools has resulted in more defensible and accountable practices of assessment. Practitioners maintain that decisions made using structured risk-assessment tools were more defensible than gut feelings, which were seen as the basis of discretionary judgments. 9 See Kemshall (2003: 67) for a more detailed analysis of the statistical limitations of the tool, including a discussion on the limits of meta-analysis and the problem of low base rates. 10 LSI tools typically have anywhere from three to five risk levels, ranging from very low, low, medium, high to very high. Moreover, the threshold level can also be arbitrarily set. Threshold levels determine the numeric range that differentiates, for example, a low- from a medium-risk score (i.e. scores of greater or less than 11 are considered medium- and low-risk, respectively). As resources shrink or swell, the cut-off mark for low-, medium- or high-risk offenders can be adjusted accordingly.

445

MAURUTTO AND HANNAH-MOFFAT

This is consistent with the claims of Rose, who suggests that risk assessment may be used, not so much in order to make accurate predictions, but to ensure that the decision made was defensible if something should go wrong (Rose 1998). In the end, the final score appears to be logical, mechanical and free of judgment. Yet, a close examination of the criteria and data used to produce a risk score reveals the moral and political nature of risk. Assembling Actuarial and Clinical Assessment: the LSI/CMI By the mid-1990s, a new group of tools emerged under the label Level of Service/Case Management Inventory (LSI/CMI). While initial versions of the LSI were limited to sorting for the purpose of supervision, containment of security risks and identification of criminogenic need, the repackaging and introduction of the LSI/CMI tools is designed to ensure productive interventions that have an overall effect of minimizing risk. The goal is not simply to normalize subjects, nor is it merely about containing dangerousness; it is to transform risky subjects to make them less risky. This is achieved through a complex assessment process that reassembles aspects of actuarial justice and criminogenic need and combines them with a range of clinical assessments. Clinical appraisals of offenders’ personalities, characters, strengths and cultures are merged with risk/criminogenic need assessments to formulate a supervision and case-management plan. For organizations where incapacitative models are not tenable, this LSI/CMI assemblage enables a form of governing that revitalizes rehabilitation through the integration of clinical and risk-based technologies. Within this configuration, supervision remains important, but priority is placed on efficient forms of service delivery, targeted correctional treatment and risk/need minimization. We are not suggesting that incapacitation or actuarial justice is rejected or replaced, but rather that multiple models of penal governance coexist and govern simultaneously. In order to understand the implications of this new assessment model, it is useful to provide a brief description of the tools and their underlying logic. The LSI/CMI group encompasses at least three new tools, including the Level of Service–Ontario Revised (LSI–OR), the Youth Level of Service/Case Management Inventory (YLS/CMI) and the soon to be released Level of Service–Revised/Case Management Inventory (LSI–R/ CMI). Although there are significant differences among the tools, they all share a basic format. The LSI/CMI group typically includes ten sections, which combine actuarial risk indicators with more traditionally clinical diagnoses and treatment-based indices. This cumulative information is used to direct and inform a case-management plan. The first of the ten sections, entitled ‘General Risk/Need Factors’, comprises the original LSI risk/need assessment. This now forms just one component of the overall assessment and has been modified to include a measure of ‘strength’. Rather than identifying items as negative correlations, the identification of ‘strength’ flags a ‘positive’ risk and/ or criminogenic need that should be considered when devising a case-management plan. For example, a positive family situation could be used to support and enhance treatment strategies. These are often referred to in the criminal justice literature as ‘protective factors’ or ‘resiliency’; they are social relations that can mitigate or potentially counteract risk factors and lower chances of recidivism. Here, the actuarial calculation of riskiness becomes conflated with social protective variables that cannot be easily inserted into statistical equations but should be considered in the evaluation of 446

ASSEMBLING RISK AND THE RESTRUCTURING OF PENAL CONTROL

risk. Strengths and criminogenic needs are then plotted on a ‘Risk/Need Profile’—a grid that is used to identify areas to target for treatment. With this new profile mapping, individual criminogenic needs, rather than the over-all risk score, become more salient and central to case management. The subsequent sections are new additions that address more traditionally clinical variables. They typically include information of the history of the offence and document clinical information that is not significantly correlated with recidivism but might be relevant in adjusting supervision levels or in planning a course of treatment. For example, one section, curiously entitled ‘Specific Risk /Need Factors’ (as opposed to general risk/need factors), is divided into two subsets, labelled ‘Personal problems with criminogenic potential’ (a 14-item checklist, including diagnosis of psychopath or personality disorder, deficits in problem solving/self-management skill, poor social skills, peers outside age range, racist/sexist behaviour, etc.) and ‘History of Perpetration’ (a nine-item list, classifying sexual assaults, escapes, fire setting, weapons, etc.). Interestingly, the specific crimes committed by offenders are here listed as non-criminogenic needs; they are not significant predictors of recidivism and hence were never included in the original LSI. The section on ‘Prison Experience’ captures previous classification levels, including assignments to protective custody, past treatment recommendations, misconducts, security or management concerns, among others. Social and cultural variables are captured by the sections entitled ‘Social, Health and Mental Health’ (a 20-item list, documenting physical ability, depression, homelessness, financial problems, low self-esteem, shy/withdrawn, etc.) and ‘Special Responsivity Considerations’ (eight variables, including motivation, denial/minimization, culturally issues, ethnicity issues, low intelligence, communication barriers, etc.). These again are variables that can mitigate criminogenic need/risk scores and should be considered in the casemanagement plan. Responsivity factors have emerged as a key consideration in the development of case management. According to the responsivity principle, intervention must be matched to the learning style of each offender, and potentially also to their ethno-cultural backgrounds. The remaining sections include a number of supplementary pages for text related to offence information, case notes and discharge summaries, as well as administrative decisions. These in-house, ministry-specific administrative sections were introduced to maximize the connection between the offender’s risk/needs assessment, the practitioner’s case management and the administrator’s decision making (Wormith 1997). The LSI/CMI hence marks a turning point for offender classification where treatment is prioritized and emphasis shifts from security to case management. This has significant implications for the restructuring of institutional practices. The LSI/CMI assemblage merges multiple logics that go well beyond the actuarial. Actuarial calculations remain a central feature; however, the tool increasingly encapsulates an individualized and clinical evaluation that cannot simply be subsumed under the traditional banner of risk. Ultimately, actuarial calculations are weakened by other judgments and appraisals. The inclusion of measures like strengths, health and responsivity that document clinical, social and cultural uniqueness weaken the primacy of actuarial calculations. Nonetheless, these variables are mediated by risk and inserted into the LSI/CMI in very specific ways. Underlying the LSI/CMI is a logic wherein offenders’ backgrounds, behaviour and needs are repackaged into criminogenic (those variables statistically correlated with recidivism) and non-criminogenic needs (those 447

MAURUTTO AND HANNAH-MOFFAT

variables not easily amenable to computation because of their variance within the population). Variables not amenable to calculation are difficult to insert into actuarial probability calculations. These non-criminogenic variables, however, do not drive the formation of treatment programmes. They are used to both mitigate security classification and to tailor interventions to an offender’s learning and behavioural style. For example, when developing an overall plan, the general risk/need profile may identify a need for drug treatment programmes, but the specific type of programme, who administers it, its frequency and whether it occurs in a group or one-on-one setting are determined on the basis of responsivity, along with social and health considerations. Andrews and Bonta (1998: 245) explain how ‘insight-oriented therapy delivered in a group format may not connect very well for a neurotic, anxious offender with minimal intelligence’ (emphasis original); an individual programme would be better suited. Van Voorhis suggests that high-anxiety inmates may not respond well to treatment strategies that involve confrontation (Patricia Van Voorhis 1997). Treatment success is affected by factors like intelligence, cognitive maturity, attention deficit disorder, psychological disorders, self-esteem and learning style that while not correlated with criminogenic need, mitigate treatment amenability (Patricia Van Voorhis 1997). While this is nothing new, the LSI/CMI tools attempt to package these individual assessments as non-criminogenic and hence irrelevant in developing the types of treatment but central to the mode of treatment. Essentially, non-criminogenic needs ensure that intervention strategies are determined according to an offender’s specific personality and circumstances, but they, as will be discussed, do not inform the possible areas requiring intervention. Destabilizing the Welfare/Risk Binary The empirical analysis of the LSI/CMI affords an opportunity to question the conventional criminological and sociological explanations of risk. Typically, ‘the risk society’ is juxtaposed to and deemed a response to the disappointment of ‘the welfare state’ and its rehabilitation project. Risk is characterized as a departure from universal, therapeutic, transformative, welfare-based strategies. Rather, risk is associated with ‘actuarial justice’ and attempts to classify, manage, contain and incapacitate dangerous offenders and threats to society (Hudson 2001; Garland 2001; Feeley and Simon 1992; Kemshall 1998). Within this discursive construction, risk is conceived of as a negative strategy that incapacitates and manages but never produces productive transformations (Feeley and Simon 1992: 452). Accordingly, Ulrick Beck notes how ‘risks only suggest what should not be done, not what should be done’ (Beck 1997: 141; 1992). Likewise, Mary Douglas comments on how ‘now risk refers only to negative outcomes …. The language of risk is reserved as a specialized lexical register for political talk about the undesirable outcomes’ (emphasis original) (Douglas 1992: 24). Within this framework, risk becomes a calculable negative approach to governing offenders. More recently, a new body of criminological literature is emerging that seeks to destabilize the association between risk and its negative implications (O’Malley 2000; Garland 2003). This literature documents the genealogy of risk and its association with uncertainty, insurance and the welfare state. If one were to trace the origins of risk meanings, it would become apparent that risk was first associated with the throw of a dice (Douglas 1992: 24)—in other words, with chance and uncertainty. In the field of insurance, risk becomes a probability calculation. These associations have a long trajectory 448

ASSEMBLING RISK AND THE RESTRUCTURING OF PENAL CONTROL

that pre-date the advent of the welfare state (Rosanvallon 2000; Hacking 1990; Ewald 1991). According to O’Malley, ‘risk is a core characteristic of all modern liberal and capitalist societies, dating back to the about the end of the eighteenth century’ (O’Malley 2000: 17). Indeed, the welfare state can be conceived of as the ‘risk management state’ that sought to minimize risk through techniques of social distribution (Garland 2003: 61–2). This association of risk with welfare makes it possible to envision the deployment of risk for productive ends. O’Malley has advanced a vision for a ‘progressive politics of risk’ in which he argues that ‘there is no obvious reason why risk cannot be inclusive rather than exclusionary, why risk should pit “offenders” against “victims”, or why it cannot be unifying rather than polarizing. Likewise we may ask whether risk might be associated with therapeutic and restorative regimes rather than with those focused on punishment or incapacitation, or with programs of social justice rather than individual or market justice’ (O’Malley 2004: 8). He contrasts actuarial justice with other forms of risk-centred governance like drug harm minimization programmes that seek to ‘empower’ users. Hannah-Moffat (2005) also discusses how risk/need-based technologies produce transformative subjects. And many crime-prevention approaches are premised on the understanding that risk can be altered and minimized. These ‘ameliorative’ risk associations are not meant to suggest that risk-based practices or technologies are any less regulatory; indeed, they open up new spaces of governance. The LSI/CMI, is one part of a larger correctional strategy designed to identify, target and lessen risk in a manner that goes beyond merely managing risky populations. It is a product of the ‘What Works’ framework that has gained international prominence, in part because, on the surface, it appears as a promising and progressive ‘common sense’ approach to offender treatment and case management. The success of this approach resides in its ability to revitalize and restore correctional legitimacy to the rehabilitative agenda. Ferguson (2002: 4) points to how ‘the shift in research away from “nothing works” to a focus on “what works” means that correctional practice also must shift from a get tough law-order focus relying strictly on sanctions, to a more balanced approach that includes both sanctions and treatment’. What works rejects incapacitative rationales of punishment that assume nothing works. It condemns harsh punishment approaches (i.e. electronic monitoring, militaristic boot camps, intensive supervised probation and warehousing) as ineffective and unsupported by the research literature. The larger goal is to minimize risk and to teach offenders how to be risk-averse. The tool is designed to identify areas for intervention in order to reduce and not simply contain risk. This approach does not abandon those deemed high-risk or high-need; rather, the objective is to ensure that they receive more intensive support. As one of the developers of the tools, Don Andrews notes ‘the important task of corrections is to manage the sentence in such a way that low risk cases remain low risk, and high risk cases move in a low risk direction’ (Andrews 1991: 11). Risk minimization, in this context, is accomplished through merging actuarial risk/need categories with a clinical diagnoses of individual circumstances. For organizations where incapacitative models are not tenable, the LSI/CMI assemblage enables a form of governing that revitalizes rehabilitation through a complex integration of clinical and risk-based technologies. Rehabilitation, however, when understood within the context of risk, takes on the form of targeted governance (Valverde and Mopas 2004). Targeted treatment is achieved by matching the level and type of intervention to the risk and criminogenic need posed by an offender. Here, risk and criminogenic need become the core 449

MAURUTTO AND HANNAH-MOFFAT

principles underlying rehabilitative programmes. According to the National Institute of Corrections, the assessment of offender risk and need is the ‘basic building block on which organizations can provide effective interventions to change offender behaviour and more effectively protect communities’ (Dal Pra 2004: 12). Increasingly, criminogenic needs, as developed in risk assessments, are being used to streamline and structure interventions. The result is that institutional programmes become dictated not by what an offender ‘may need’ or want, but rather through the identification of core correctional programmes that can be demonstrated to lower risk levels and that are amenable to repeat evaluations. Given the cost of treatment services, universal access for all offenders to all available programmes is no longer deemed tenable, nor an efficient use of resources. Instead, only ‘essential services’ are to be delivered, and only to those deemed most in need. The LSI/CMI assessment is used not only to calculate and categorize who would most benefit from a particular service, but also to determine which services to offer. Programmes that can be easily subject to evaluation, like substance abuse programmes, become core. Services not readily amenable to evaluation or for which improvements may take a considerable length of time, like those that target self-esteem or psychiatric symptoms, are devalued and cut. Likewise, general or client-centred approaches that attempt to diagnose what might be ‘good for the offender’ are depicted as ‘unstructured’ and ‘vague’. In the end, the cost of treatment programmes is reduced, but so too is the important range of programmes available to offenders. The influence of this risk/need assemblage extends beyond the micro level of assessment to the accreditation and delivery of correctional programmes. Arguably, the emphasis on ‘evidence-based principles’ of risk/need assessment and programme delivery is part of an extensive organizational restructuring that attempts to efficiently and effectively allocate scarce treatment resources to suitable and responsive offenders. The what works framework embodied in contemporary risk assessments, like the LSI/ CMI, inform not only the criteria contained in assessment tool, but also the types and delivery of correction treatment programmes as well as the mission and purpose of correctional agencies. Using the analogy of an emergency room, White (2004: 42) indicates that correctional managers are ‘in a triage business … within the constraints imposed on us by both our internal and external stakeholders, we need to base our decisions on evidence based practice. Assessing risk and needs and allocating resources accordingly are thus perhaps the most critical functions of any correctional agency’. There are too many elements to the development of an ‘efficient and effective’ correctional regime to address in this paper. Nevertheless, the development of targeted interventions and the accreditation of correctional programmes are two recent initiatives, which are intended to streamline correctional services. Conclusions The modern crisis in penality, in which the framework of risk is mobilized, is one that advocates an overt punitiveness coupled with continued demands on the state to ‘do something about crime’ beyond the warehousing of offenders. As Sparks notes (2000: 136), ‘The state can not any longer simply perform punishment as a sovereign right. It must also thereby promise something’. Obviously, such promises are highly contingent on how more universal rationalities of governing are deployed in local socio-political 450

ASSEMBLING RISK AND THE RESTRUCTURING OF PENAL CONTROL

contexts. In Canada, while is clear that the actuarial logic of risk is pervasive in penal discourse, it is embedded in a national context which is not simply a mirror reflection of punitive, over-the-top American penal politics. Our research shows how new assemblages with risk produce new sites and possibilities for governing. The actual uses of these tools vary and often depart from the articulated purpose of the tools and organizational policies (Hannah-Moffat and Maurutto 2004). Nevertheless, the purpose of this paper was to illustrate changes in how risk and need are assembled and how such changes enable new strategies of penal governance. Notwithstanding the claims of many, the proliferation of risk assessments is not the result of a ‘loss in faith in’ or ‘abandonment of’ the rehabilitative ideal. On the contrary, the Canadian project, which has resulted in the development of internationally used risk-assessment tools, is integrally linked to the reaffirmation of rehabilitation. A close reading of the Canadian correctional risk-assessment literature, which has had a significant international impact, clearly connects the effective assessment of an offender’s level of risk (and need) to treatment interventions. The logic behind the LSI/CMI is that recidivism is predictable and that risk/needs assessment and correctional treatment can reduce the likelihood of recidivism. Risk is no longer simply associated with safety or security; new assemblages enable risk-based practices to adopt more productive connotations that operate to target treatment, inform case-management plans and, ultimately, minimize risk. Each new reassemblage of risk produces new possibilities of governing that are often more pervasive and intrusive and that structure and limit discussion of penal reform. The LSI/CMI is a poignant example of how new assemblages can operate to limit and streamline the range of institutional programmes and choices for offenders. These tools are increasingly being used to structure and limit which treatment programmes to offer and to make decisions about institutional resource allocations. These new forms of penal governance are consistent with neo-liberal models of governing. The ways in which new risk technologies are revitalizing neo-liberal discourses and governing strategies require our careful attention. REFERENCES ANDREWS, D. A. (1982), The Level of Supervision Inventory (LSI): 1. The First Follow-up. Ottawa: Ministry of Correctional Services. ———(1991), ‘Recidivism Is Predictable and Can Be Influenced: Using Risk Assessments to Reduce Recidivism’, Practitioner’s Forum, 2/3: 8–14. ANDREWS, D. and BONTA, J. (1989/1994) The Psychology of Criminal Conduct. Cincinnati, OH: Andersen Publishing. ANDREWS, D. A. and BONTA, J. (1998), The Psychology of Criminal Conduct. Ohio: Anderson Publishing. BAIRD, S. C., HEINZ, R. C. and BEHMUS, B. J. (1979), Project report 14: A two-year follow up. Wisconsin: Department of Health and Social Services, Case Classification/Staff Deployment Project, Community Corrections. BAKER, T. and SIMON, J. (2002), Embracing Risk: The Changing Culture of Insurance and Responsibility. Chicago: University of Chicago Press. BECK, U. (1997), World Risk Society. London: Polity Press.

451

MAURUTTO AND HANNAH-MOFFAT

BONTA, J. (1996), ‘Risk/Needs Assessment and Treatment’, in T. Harland, ed., Choosing Correctional Options that Work: Defining the Demand and Evaluating the Supply. Thousand Oakes: Sage Publications. BURGESS, E. W. (1928), ‘Factors Determining Success or Failure on Parole’, in A. Bruce, A. Harno, E. Burgess and J. Landesco, eds, The Workings of Intermediate-Sentence Law and the Parole System in Illinois. Springfield, IL: State Board Of Parole. COPAS, J., MARSHALL, P. and TARLING, R. (1996), Predicting Reoffending for Discretionary Conditional Release, Home Office Research Study 150. London: HMSO. DAL PRA, Z. (2004), ‘In Search of a Risk Instrument’, in M. Thigpen and G. Keiser, eds, Topics in Community Corrections: Issues for Managers. Washington: National Institute of Corrections. DOUGLAS, M. (1992), Risk and Blame: Essays in Cultural Theory. London: Routledge. ERICSON, R. V. and DOYLE, A. (2003), Risk and Morality. Toronto: University of Toronto Press. EWALD, F. (1991), ‘Insurance and Risk’, in G. Burchell, C. Gordon and P. Miller, eds, The Foucault Effect. Chicago: University of Chicago Press. FEELEY, M. and SIMON, J. (1992), ‘New Penology: Notes on the Emerging Strategy of Corrections and its Implications’, Criminology 30/4: 449–74. ———(1994), ‘Actuarial Justice: The Emerging New Criminal Law’, in D. Nelken, ed., The Futures Of Criminology. London: Sage. pp.173–201. FERGUSON, J. (2002), ‘Putting the What Works Research into Practice: An Organizational Perspective’, Criminal Justice and Behaviour, 29/4: 472–92. GARLAND, D. (2001), The Culture of Control: Crime and Social Order in Contemporary Society. Chicago: University Of Chicago Press. ———(2003), ‘The Rise of Risk’, in R. Ericson and A. Doyle, eds, Risk and Morality. Toronto: University of Toronto Press. GLASNER, D. (1985), ‘Who Gets Probation and Parole: Case Study versus Actuarial Decision Making’, Crime and Delinquency, 31: 367–78. HACKING, I. (1990), The Taming of Chance. Cambridge: Cambridge University Press. ———(2005), ‘Criminogenic Needs and the Transformative Risk Subject: Hybridizations of Risk/Need in Penality’, Punishment and Society, 7/1: 29–51. HANNAH-MOFFAT, K. and M AURUTTO, P. (2004), Youth Risk/Need Assessment: An Overview of Issues and Practices.Ottawa: Research and Statistics Division, Department of Justice Canada. HANNAH-MOFFAT, K. and SHAW, M. (2001), Taking Risks Incorporating Gender and Culture into Classification and Assessment of Federally Sentenced Women. Ottawa: Status Of Women Canada. HART, H. (1923), ‘Predicting Parole Success’, Journal of Criminal Law and Criminology, 14: 405–13. HOFFMAN, P. B. (1983), ‘Screening for Risk: A Revised Salient Factor Score’, Journal of Criminal Justice, 11: 539–47. HUDSON, B. (2001), ‘Punishment, Rights and Difference: Defending Justice in the Risk Society’, in K. Stenson and R. Sullivan, eds, Crime, Risk and Justice: The Politics of Crime Control in Liberal Democracies. Cullompton, Devon: Willan Publishing. ———(2003), Justice in the Risk Society. London: Sage Publications. Kemshall, H. (1998), Risk in Probation Practice. Aldershot: Ashgate. ———(2002), Risk, Social Policy and Welfare. Buckingham: Open University Press. ———(2003), Understanding Risk in Criminal Justice. Buckingham: Open University Press.

452

ASSEMBLING RISK AND THE RESTRUCTURING OF PENAL CONTROL

KEMSHALL, H. and MAGUIRE, M. (2001), ‘Public Protection, Partnership and Risk Penality: The Multi-Agency Risk Management of Sexual and Violent Offenders’, Punishment and Society, 3/2: 237–64. KEMSHALL, H., PARTON. N., WALSH, M. and WATERSON, J. (1997), ‘Concepts of Risk in Relation to Organizational Structure and Functioning within the Personal Social Services and Probation’, Social Policy and Administration, 31/3: 213–32. LATOUR, B. (1987), Science in Action. Cambridge, MA: Harvard University Press. LEACOCK, V. and SPARKS, R. (2002), ‘Riskiness and At-Risk-Ness: Some Ambiguous Features of the Current Penal Landscape’, in N. Gray, J. Laing and L. Noaks, eds, Criminal Justice, Mental Health and the Politics of Risk. London: Cavendish Publishing. LOWENKAMP, C., LATESSA, E. and HOLSINGER, A. (2004), ‘Empirical Evidence on the Importance of Training and Experience in Using the Level of Service Inventory—Revised’, Topics in Community Corrections. Colorado: National Institute of Corrections–US Department of Justice. LOZA, W. and SIMOURD, D. (1994), ‘Psychometric evaluation of the level of supervision inventory (LSI) among male Canadian federal offenders’, Criminal Justice and Behaviour, 21/4: 468–80. LYNCH, M. (1998), ‘Waste Managers? The New Penology, Crime Fighting and the Parole Agent Identity’, Law and Society Review, 32/4: 839–69. MEGARGEE, E. I. and BOHN, M. J. (1979), Classifying Criminal Offenders: A New System Based on the MMPI. Thousand Oaks, CA: Sage Publications. MILLER, L. (2001), ‘Looking for Postmodernism in all the Wrong Places: Implementing a New Penology’, British Journal of Criminology, 41: 168–84. MINISTRY OF THE SOLICITOR GENERAL AND CORRECTIONAL SERVICES (MSGCS) (1995), Level Of Service Inventory—or—Fact’s Behind Risk Assessment. Ontario: Ministry of the Solicitor General and Correctional Services. NUFFIELD, J. (1982), Parole Decision-Making in Canada. Ottawa: Solicitor General of Canada. O’MALLEY, P. (2000), ‘Risk Societies and the Government of Crime’, in M. Brown and J. Pratt, eds, Dangerousness, Risk and Modern Society. London: Routledge. ———(2001), ‘Risk, Crime and Prudentialism Revisited’, in K. Stenson and R. Sullivan, eds, Crime, Risk and Justice: The Politics of Crime Control in Liberal Democracies. London: Willan. ———(2002), ‘Globalizing Risk? Distinguishing Styles of ‘Neo-Liberal’ Criminal Justice in Australia and the USA’, Criminal Justice, 2/2: 205–22. ———(2004), Risk, Uncertainty, Government. London: The Glasshouse Press. ROBINSON, G. (1999), ‘Risk management and rehabilitation in the probation service: Collision and collusion’, The Howard Journal, 38/4: 421–33. ROBINSON, G. (2001), ‘Power, knowledge and “what works” in probation’, The Howard Journal, 40/3: 235–54. ———(2002), ‘A Rationality of Risk in the Probation Service: Its Evolution and Contemporary Profile’, Punishment and Society, I 4/1: 5–25. ROSANVALLON, P. (2000), The New Social Question: Rethinking the Welfare State. Princeton, NJ: Princeton University Press. ROSE, N. (1998), ‘Governing Risky Individuals: The Role of Psychiatry in New Regimes of Control’, Psychiatry, Psychology and Law, 5/2: 177–95. SERIN, R. and MAILLOUX, D. (2001), Development Of A Reliable Self-Report Instrument For The Assessment Of Criminogenic Need. Ottawa: Research Branch of the Correctional Service of Canada. SIMON, J. (1995), ‘They Died with their Boots On: The Boot Camp and the Limits of Modern Penality’, Social Justice, 22/2: 25–48.

453

MAURUTTO AND HANNAH-MOFFAT

———(1996), ‘Criminology and the Recidivist’, in D. Schichor and D. Schrest, eds, Three Strikes and You’re Out: Vengeance as Public Policy. London: Sage. SPARKS, R. (2000), ‘Perspectives on Risk and Penal Politics’, in T. Hope and R. Sparks, eds, Crime, Risk and Insecurity. London: Routledge. STENSON, K. and SULLIVAN, R. (2001), Crime, Risk and Justice: The Politics of Crime Control in Liberal Democracies. London: Willan. VALVERDE, M. and MOPAS, M. (2004), ‘Insecurity and the Dream of Targeted Governance’, in W. Larner and W. Walters, eds, Global Governmentality. London: Routledge. VAN VOORHIS, P. (1997), ‘Correctional Classification and the ‘Responsivity Principle’, Forum, 9/1. WORMITH, S. (1997), ‘Research into Practice: Applying Risk/Needs Assessment to Offender Classification’, Forum on Corrections Research, 9/1.

454