An Examination of a Risk/Needs Assessment Instrument and Its Relation to Religiosity and Recidivism Among Probationers in a Muslim Culture

Abstract

This study was conducted to assess the value of administering a risk/need assessment instrument to low-risk offenders in Pakistan. The Level of Service/Case Management Inventory (LS/CMI) and a measure of religiosity, the Muslim Religiosity-Personality Inventory (MRPI): Abridged Scale, were administered to probationers in this highly devout Muslim country that has little experience with risk/need assessment. In spite of the low recidivism rate, predictive validities based on correlation and receiver operating characteristic analyses were comparable with those of Western cultures overall, and for samples based on gender, geographic location, and type of crime. Although religiosity was negatively correlated with recidivism, it offered no incremental validity to the LS/CMI to predict recidivism because it was also correlated negatively with the LS/CMI. The findings have theoretical implications for the risk assessment of low-risk offenders and for the contribution of religiosity to offender risk and practical implications for the Pakistani justice system.

Keywords

offender risk assessment religiosity Pakistani probationers LS/CMI recidivism

The concept and practice of offender risk assessment has gained widespread use and popularity in correctional agencies across North America and Western Europe over the last two decades. Buttressed by a substantial research base, its application in correctional operations is commonly accepted as part of a large-scale movement to embrace evidence-based practice in the justice arena (Andrews, Bonta, & Wormith, 2006). Risk assessment is now used to aid a wide range of judicial (e.g., bail, sentencing, and detention) and administrative (e.g., custody level, parole, community placement, and supervision) decisions. Instruments that assess dynamic risks, or criminogenic needs, also have the capacity to inform correctional and programming decisions such as referrals to other agencies, selection of treatment, and aspects of case management supervision, including content and style. For the most part, empirical efforts have demonstrated the value of risk assessment-based decisions and activities in furthering public safety, crime prevention, and offender rehabilitation (Andrews, Bonta, & Hoge, 1990; Andrews, Zinger, Hoge, Bonta, Gendreau, & Cullen, 1990; Grove, Zald, Lebow, Snitz, & Nelson, 2000; Latessa & Lovins, 2010; Maloney & Miller, 2015).

Yet, there are critics of the offender risk assessment movement, particularly as it is applied to race, culture (Martel, Brassard, & Jaccoud, 2011; Whiteacre, 2006), and gender (e.g., Blanchette & Brown, 2006; Hannah-Moffat, 2013; Van Voorhis, Wright, Salisbury, & Bauman, 2010). Many of these concerns come from legitimate questions of fairness about the applicability of generic risk assessment tools to particular offender populations such as minorities and women. This argument comes in part from the fact that most instruments were developed and validated on male offender populations. Critics argue that they fail to attend to the unique characteristics that other demographically defined offender groups possess, the histories they have experienced, and the circumstances with which they must contend.

The current investigation examined the use of a common risk/need assessment tool with probationers in Pakistan. This may be described as a nontraditional culture, at least in terms of the scale’s current popular usage, which now includes large multicultural jurisdictions, including Hong Kong, Singapore, and Japan (Chng, Hong, & Misir, 2002; Chu et al., 2015; Takahashi, Mori, & Kroner, 2013). It is also a very religious culture, which makes it conceivable that such widespread devotion could disrupt or neutralize the well-established Western risk/need factors that are related to criminal behavior and offender recidivism. Hence, our interest was to explore the applicability of a Western risk/need instrument in a traditional Muslim culture.

Probation in Pakistan

The Probation of Offenders Ordinance (1960/Rules, 1961) was promulgated in Pakistan in 1960 to provide a statutory framework for courts to place or release eligible offenders on probation for no more than 3 years under certain probation conditions. Probation in Pakistan was the responsibility of the police, but now operates quite similarly to probation in Western countries, with one important exception. Eligibility in Pakistan is restricted to first time, nonviolent offenders. The ordinance aims to rehabilitate and reintegrate relatively minor offenders in the community by providing them with an opportunity to make themselves “honest, industrious and law-abiding” (p. 2) individuals of society under the supervision, assistance, and cooperation of probation officers (POs; Probation of Offenders Ordinance, 1960/Rules, 1961). POs function in all 36 districts in Punjab under the administrative control of the reclamation and probation (R&P) department.

Similar to many Western legal systems, according to the Probation Law (1960), assistance is sought from POs during the trial of petty offenders when making decisions to release them on probation (Probation of Offenders Ordinance, 1960). POs prepare and submit a “Presentence Report” (PSR) to the court about the offender that includes the nature of the offence, the antecedents, the character of the offender, home surroundings, and other important matters related to the commission of the offence (Bhutta, 2010). This report is expected to provide comprehensive information to the court when granting probation to the eligible offender. Notably, the R&P departments in Punjab and other provinces have not developed an objective risk/need assessment instrument or mechanism to determine the level of risk and needs of offenders. Yet, the identification of risk/need factors is often mandatory in Western countries for making decisions related to PSRs, offender classification and case management, and to design correctional strategies for offenders who are on probation (Andrews et al., 2006).

The legislation prescribes the duties and responsibilities of POs relating to the reclamation and probation system beginning before the probationer’s release and going through the planning, supervision, and guidance of the offender in his or her rehabilitation and reintegration (Probation of Offenders Ordinance, 1960). It also mandates a statutory obligation for POs to inform the court in the event of a probationer’s failure to observe the conditions of the probation order (e.g., probationer’s attendance in the PO office is obligatory, at least monthly in Punjab). The law also empowers the courts to revoke a probation order by passing a new order to rearrest the probationer who has failed to observe any of the listed conditions of the probation order (Probation of Offenders Ordinance, 1960).

Risk/Need Assessment with the Level of Service (LS) Instruments

The family of LS instruments consists of a number of offender risk assessment scales, all of which are designed to predict the likelihood of an offender recidivating and to identify particular target areas that might lower the probability of such recidivism. The LS scales have the capacity to achieve this second objective because they assess the presence of dynamic risk factors, or criminogenic needs, which may be affected by appropriate interventions, case management, and community supervision. Collectively, they are the most widely used offender risk/need assessment instrument internationally (Wormith, 2011).

The most substantial change in the tool was due to what the authors’ referred to as the emergence of “fourth generation” risk assessments (Andrews et al., 2006, p. 8), which extended the third-generation risk/need tools (e.g., Level of Service Inventory–Revised [LSI-R]; Andrews & Bonta, 1995) to instruments that went beyond risk and criminogenic need. Fourth-generation risk assessment tools assess risk, criminogenic needs, other client issues, and strengths and integrate the risk/need assessment process with offender case management to address the needs identified in the assessment (e.g., Level of Service/Case Management Inventory [LS/CMI]; Andrews, Bonta, & Wormith, 2004). Variations of the LS instruments have been developed for adults and youth, various cultures, and more than a dozen countries.

However, they all have a number of key features in common. First, they have been constructed from a general personality and cognitive social learning theory of criminal behavior and thus, have explanatory value. Second, their items cluster into various groupings or domains, now described as the “Central Eight” by Andrews and Bonta (2010, p. 58). Third, items are scored in a binary present–absent fashion and summed in simple arithmetic fashion. Finally, they have been extensively validated through numerous studies from a wide variety of jurisdictions and countries, a brief summary of which follows.

A series of meta-analyses nicely characterizes this research. First, Gendreau, Goggin, and Smith (2002) examined 33 studies of both adult and youth LS instruments and derived a mean effect size of .39 for general recidivism and .28 for violent recidivism. Subsequently, Yang, Wong, and Coid (2010) found numerous sources of variation for LS effect sizes, including country and type of sample.

A meta-analysis by Wilson and Gutierrez (2014) examined the predictive validity of the LS scales with respect to Aboriginal offenders. Although the LS predicted general recidivism for 13 samples of Aboriginal offenders, five of the eight subscales were less predictive for Aboriginal samples than they were for non-Aboriginal offenders. This finding is reminiscent of a more general meta-analysis of the Central Eight domains that comprise the more recent version of the LS scales (e.g., LS/CMI). Gutierrez, Wilson, Rugge, and Bonta (2013) also found that some domains (i.e., Criminal History, Alcohol/Drug Problems, and Antisocial Pattern) were less predictive for Aboriginal offenders, although they did conclude that the Central Eight risk factors are relevant predictors for Aboriginal offenders.

Yet, this study raises a legitimate concern that there may be other unidentified groups for whom the LS measures are less applicable. For example, there has been considerable research and debate about the applicability of the LS scales to female offenders because they fail to capture the full range of issues that are particular to women offenders (Hannah-Moffat, 2013). In this regard, some have suggested supplementing General Risk/Needs scales with a gender-informed supplement (e.g., Van Voorhis et al., 2010), while others have found that either it is not necessary as validities are comparable (Andrews et al., 2012), or it does not add any incremental predictive validity (Rettinger & Andrews, 2010).

The most definitive meta-analysis on LS instruments was completed by Olver, Stockdale, and Wormith (2014) on 151 independent samples consisting of various demographic groups. The mean predictive validity coefficient for general (any) recidivism was .29. Within-study comparisons revealed comparable predictive validity coefficients for ethnic minorities (.29) and nonminorities (.28) and for males (.29) and females (.29) on general recidivism. However, they found a substantial difference by country or region, with Canada generating the highest validities (.38), followed by outside North America (.30) and the United States (.20) for general recidivism. The order of these regions comes as some surprise. Reasonably, one would consider Canada and the United States to be more similar in language, custom, and culture than non-North American jurisdictions. The message that we derive from these findings is that nothing should be taken for granted when applying any offender risk assessment instrument to other cultures and justice systems.

Religiosity

Although the LS scales include domains that consider procriminal versus anticriminal attitudes, procriminal versus prosocial companions, and procriminal versus anticriminal use of leisure time, they do not consider any indicators of religiosity or spirituality. In fact, the only reference to religion is found in an item pertaining to participation in prosocial organized activity in which credit is given for attendance in church. Similarly, other popular offender risk/need assessment instruments do not include any direct measure of religiosity or religious affiliation. Some have suggested that this is an oversight (Baier & Wright, 2001) and that religiosity is the “forgotten factor” in criminological research (Johnson, De Li, Larson, & McCullough, 2000, p. 32). Intuitively, the idea that religiosity or the extent to which one professes and practices a particular religious faith has a certain appeal (Heaton, 2006). Religions from around the world hold to certain principles, such as the Golden Rule, that would seemingly promote prosocial appropriate interpersonal behavior and inhibit antisocial criminal behavior. Experts on criminal behavior have suggested that religion may strengthen one’s bonds with social institutions and society generally, which in turn may minimize one’s involvement in criminal behavior (Baier & Wright, 2001; Evans, Cullen, Dunaway, & Burton, 1995; Gottfredson & Hirschi, 1990), quite possibly through a (prosocial) socialization process and the development of self-control.

There is considerable empirical research that describes the relationship between religiosity and criminal behavior, although it is primarily on youth. Much of this literature is captured in a meta-analysis by Baier and Wright (2001). They found a small, negative correlation (−.12) between religiosity and criminal behavior across 60 studies. However, they did not elaborate on the various measures of religiosity that were used in these studies. This is an important issue as religiosity is a multidimensional construct. Moreover, a number of studies have determined that it is participation in religious activities per se that is related to delinquency and not religious attitudes or “hellfire” beliefs (Evans et al., 1995, p. 195). A more recent systematic review of 270 studies pertaining to the relationship between religiosity and crime found that religiosity had a protective effect in 90% of the studies reviewed (Johnson & Jang, 2010).

There has also been speculation about the mechanism that is responsible for the religiosity-criminal behavior relationship. One leading candidate is that self-control (Gottfredson & Hirschi, 1990) is responsible for a number of personal attributes, including religiosity, which when combined inhibit criminal behavior. This is called the confounding hypothesis. Others contend that religiosity promotes self-control, which in turn inhibits criminal behavior and consequently is described as the mediation hypothesis (Reisig, Wolfe, & Pratt, 2012). Yet, neither of these perspectives accounts for findings by Welch, Tittle, and Grasmick (2006) that Christian religiosity and self-control had independent, negative relationships with self-reported misbehavior among American adults. Regardless of the mechanism by which self-control relates to antisocial behavior, its correlation appears to be consistent across cultures. In a recent study of 16,000 adolescents in 11 different cultures, Vazsonyi and Huang (2015) demonstrated that behavioral and attitudinal measures of self-control were consistently related negatively to self-reported deviance.

Evidence for a religiosity–crime connection also comes from a body of research on offender intervention. Faith-based programs have been part of the rehabilitation landscape since the introduction of penitence to the prison environment by William Penn and the Quakers (Beaumont & de Tocqueville, 1833/1964). There is now an extensive list of faith-based programs in corrections with some encouraging results. For example, offenders in Minnesota who participated in the Inner Change Freedom program, which uses Christian philosophies to promote a crime-free lifestyle using volunteers from local churches, recidivated significantly less than their nonparticipating counterparts as measured by rearrest, reconviction, and reincarceration (Duwe & Johnson, 2013; Duwe & King, 2012). In this regard, any changes in religiosity or religious values that are subsequently related to lower rates of recidivism illustrate the “dynamic predictive validity” of religiosity in relation to future criminal behavior (Andrews et al., 2006). Although other program findings are also positive, including the training of prisoner ministers in Washington, D.C. (Young, Gartner, O’Connor, Larson, & Wright, 1995), systematic reviews and meta-analyses have ranged from cautious (Volokh, 2011), to mixed (Aos, Miller, & Drake, 2006), to supportive (Johnson, 2011). The finding that inmates who participate in religious programs are more “motivated” to succeed than those who do not take part complicates the interpretation of these studies (Camp, Klein-Saffran, Kwon, Daggett, & Joseph, 2006) and supports Volokh’s (2011) concern about a “self-selection problem” (p. 43).

Finally, it is important to acknowledge that most examinations of the relationship between religiosity and criminal behavior have been conducted in Western countries where Christianity is predominant. As noted previously, most religions have some key themes in common, but they also have many differences. In this regard, questions have been raised about the validity of applying measures of religiosity that have been developed in the context of a Judeo-Christian culture to Muslim countries, such as Pakistan where 97% of the population is Muslim (Amer & Bagasra, 2013; Krauss, Hamzah, & Idris, 2007; Krauss et al., 2006; Raiya, Pargament, Stein, & Mahoney, 2007). Moreover, the role and contribution of constructs such as self-control may also be related to religion differentially across cultures.

One large-scale, systematic investigation was conducted on a national sample of Christian and Muslim adolescents in Germany. Baier (2014) concluded that religiosity was not a central variable for explaining violent behavior, but it was differentially implicated in the two religions. He found that religiosity among Christian adolescents was negatively correlated with self-reported violent behavior, but this relationship was through control theory variables. Among Muslim adolescents, religiosity was related to lower alcohol consumption which in turn corresponded with less violence. However, this study was conducted in a predominantly Christian country (i.e., 91% of the current sample was Christian) and that Baier used a single generic, three-item measure (i.e., frequency of prayer, frequency of attending worship, and personal importance of religion) to assess religiosity.

The Current Study

The current study was designed to investigate the predictive validity of the LS/CMI and its various segments with recidivism among a sample of adult offenders released on probation in Pakistan. Another aim of the study was to determine the utility of a traditional risk/need assessment instrument with a more homogeneous group of low-risk offenders than is found in many correctional jurisdictions. Finally, given that religion plays a strong cultural role in Muslim countries, the study also considered religiosity as a possible risk/need factor for offender recidivism among the aforementioned sample. It was hypothesized that religiosity would be negatively correlated with risk for recidivism as measured by the LS/CMI and predictive of offender recidivism. Beyond that, however, the study was exploratory, asking the question: Is the LS/CMI missing an important risk factor by excluding religiosity or is religiosity sufficiently correlated with other measures of antisociality that its inclusion would be redundant in the prediction of offender recidivism?

To explore the criminal characteristics (i.e., risk/need factors) of the present sample of adult Pakistani probationers, the LS/CMI was selected based on two important considerations. First, the LS/CMI General Risk/Need section was chosen due to its capacity to determine the risk of reoffending among probationers from many jurisdictions. Second, it assists POs by providing a manual with comprehensive information about risk assessment and case management and how to apply them to individual offenders. Moreover, the instrument also includes guidelines that are useful for correctional programs and delivery of services, particularly for offenders in the community (Wormith, 1997).

Method

Participants

The study sample was comprised of 506 adult probationers (18 years and older) who were released on probation during 2010 in four districts of Punjab province. Punjab is the largest of four provinces and comprises 56% of the country’s population. As the assessment of probationers in the selected locations was part of the operational policy of the probation department, along with monthly reporting conditions, the sample represented virtually all probationers in the catchment area during the study. At the time of the assessment, offenders had been on probation for an average of 5.88 months (SD = 2.60 months). The sample consisted primarily of 475 Muslims (93.9%), while the remaining 31 were Christians (6.1%). The sample included 470 males (92.9%) and 36 females (7.1%). Most of the sample were from urban communities (n = 288, 56.9%), with the remainder living in rural settings (n = 218, 43.1%). With respect to the offense type of the offenders, the majority had been convicted of drug-related offenses (n = 260, 51.4%), followed by theft (n = 128, 26.9%), carrying weapons (n = 58, 11.5%), and miscellaneous offenses (n = 52, 10.3%). However, the latter three groups were merged into a nondrug group (n = 246, 48.6%) for statistical comparisons. The mean age of the adult probationers was 32.47 years (SD = 11.06 years). There was no significant difference in age between males and females, t(504) = 1.27, ns. However, urban probationers were significantly older than rural probationers, t(504) = 2.34, p < .02. Approximately 50% of the total sample (n = 226, 44.7%) had no education and were illiterate. The educational level of the male sample was significantly higher than the female sample, t(504) = 2.92, p < .005. The average monthly income of the sample was 6,665 PKR (SD = 5,077 PKR) or US$66.65. Most of the sample (93.9%) was on probation for 1 year, and the remaining probationers were either on probation for 2 years (5.7%) and 3 years less a day (0.4%).

Measures

The LS/CMI

The LS/CMI was developed by Andrews et al. (2004). It is a modification and extension of the previously developed LSI-R (Andrews & Bonta, 1995) with a number of innovations. It was pilot tested as the Level of Service Inventory–Ontario Revision (Andrews, Bonta, & Wormith, 1995) for a number of years before its official release. The LS/CMI consists of four data collection sections. The General Risk/Needs section (43 items) consists of eight subscales that are designed to identify both static risk and dynamic criminogenic need areas to determine an offender’s risk to recidivate. The eight subscales, or factors, also known as the Central Eight (Andrews & Bonta, 2010) are as follows: Criminal History (8 items), Education/Employment (9 items), Family/Marital (4 items), Leisure/Recreation (2 items), Companions (4 items), Alcohol/Drug Problems (8 items), Procriminal Attitude/Orientation (4 items), and Antisocial Pattern (4 items). A second section is devoted to a list of specific risk/needs that includes both personal characteristics and historical antisocial behavior. A third section, Other Client Issues, focuses on social, health, and mental health issues. Finally, a Responsivity section lists personal responsivity issues that may be an issue for the offender in treatment. All items are scored in a dichotomous (yes/no) format and are totaled for each section and subsections.

Psychometric attributes of the LS/CMI are provided in the instrument’s manual (Andrews et al., 2004) and other reviews (Andrews, Bonta, & Wormith, 2010). Internal consistency is consistently high (α = .86-.94) for the General Risk/Needs total score, but substantially lower for the eight subscales (e.g., .64-.86 for Criminal History and .42-.67 for Antisocial Pattern). A recent study of the interrater reliability of the youth version of the LS/CMI (Youth Level of Service Inventory/CMI [YLS/CMI]; Hoge & Andrews, 2002) produced interclass correlations of .78 among POs who use the YLS/CMI in the field with their clients (Rocque & Plummer-Beale, 2014). The predictive validity of the LS/CMI is summarized in a meta-analytic investigation of 12 studies that generated an unweighted average correlation of .42 (random effects; Olver et al., 2014). The predictive validities for the General Risk/Needs subscales in the LS/CMI and other versions of the LS instruments varied from .14 for Family/Marital to .31 for Antisocial Pattern. Various studies have demonstrated the predictive validity of the LS scales, and the LS/CMI in particular, across gender and ethnicity (Andrews et al., 2012; Olver et al., 2014; Wormith, Hogg, & Guzzo, 2012).

The LS/CMI was translated into Urdu, the national language of Pakistan, and back-translated to ensure accuracy of the translation. The Pakistani version of the LS/CMI was then pilot tested on a sample of 55 probationers. Some minor modifications were required to adapt the LS/CMI to the Pakistani culture and its justice system. First, the Criminal History subscale (8 items) was not removed, but in effect, amounted to a mean score of zero with the current sample of Pakistani probationers because only first time, nonviolent offenders are granted probation in Pakistan (Probation of Offenders Ordinance, 1960/Rules, 1961). Second, in the General Risk/Need section, two items (i.e., Items 12 and 13) in the Education/Employment subscale were modified in view of the very low literacy rate (58%) and the lack of grade equivalence in Pakistani schools compared with developed countries, such as Canada. Instead of achieving Grade 10 and 12, these items were scored for achieving Grades 5 and 10, respectively. Third, the pilot testing led to three changes in the Specific Risk/Needs section. Specifically, Items 2 (psychopathy), 11 (racist/sexist), and 13 (outstanding charges) were deleted and replaced with the following items: fine in lieu of probation, friendship outside age range, and achievement not on par with ability. Fourth, one item in the Other Client Issues section, Item 4 (immigration issues), was deleted and “family feuds” was added. These changes were made in consultation with the R&P department and reflected the social, economic, and cultural context of Pakistan. Finally, the Responsivity section was not completed in the current study due to cultural differences between the Western world and Pakistan.

The Muslim Religiosity-Personality Inventory (MRPI): Abridged Scale

To determine the degree of religiosity in the sample, a 14-item, short form of the Religious Personality subscale (Krauss et al., 2007) of the MRPI (Krauss, Hamzah, & Juhari, 2005) was created with the use of supplemental items from other scales. The original 102-item version of the MRPI (Krauss et al., 2005) and a slightly adapted version with wording changes to some items (Krauss et al., 2007) are comprised of two subscales, the Islamic Worldview subscale and the Religious Personality subscale. The former subscale represents the presence of one’s religious worldview in his or her righteous deeds (actions) and reflects the manifested aspects of an individual in a given society. The Religious Personality subscale operationalizes a religiously “refined character” (Krauss et al., 2005, p. 173) by capturing two components: special worship (ibadat), including rituals that reflect one’s direct relationship with God, and general worship that is illustrated in daily, religiously guided behavior (mu’amalat) toward other humans and the rest of creation (Krauss et al., 2005). However, Krauss et al. (2005) grouped the Religious Personality items of the MRPI into three subdimensions: Self (internal, self-directed virtues), Social (interpersonal religious-based strivings and behavior toward others), and Ritual (direct relationship with God through ritualistic acts). Considerable research on religiosity has been conducted with various modifications of the MRPI, including a shortened, 56-item version (Krauss & Hamzah, 2009) designed to facilitate its administration to samples varying in age, geographic region, faith, and religion (Annalakshmi & Abeer, 2011; Krauss et al., 2007; Krauss et al., 2012).

The MRPI was translated into Urdu by one of the authors, verified by a language expert, and pilot tested on a sample of 55 probationers. This led to the development of the MRPI: Abridged Scale. Specifically, 10 items of the MRPI: Abridged were derived from the Religious Personality subscale of the MRPI (Krauss et al., 2007). To examine attitudinal aspects (e.g., religious beliefs, importance of religion in life, etc.) of religiosity among the participants, the remaining 4 items of the New Religiosity scale were adapted from different empirical studies on the religion–crime context (e.g., Johnson et al., 2000; Sinha, Cnaan, & Gelles, 2007).¹ No effort was made to capture the Islamic Worldview section of the MRPI or to create specific subscales of the MRPI: Abridged Scale. Rather, the aim was to maintain the content validity of the Religious Personality subscale from the original instrument. Hence, 5 items were selected to reflect the Self component (e.g., “How important is your religion/faith to you?”), 5 items to reflect the Interpersonal component (e.g., “How caring are you regarding the rights of parents/neighbors/relatives?”), and 4 items to reflect the Ritual dimension (e.g., “To what extent do you offer prayers regularly?”). In so doing, the concepts of Special Worship and General Worship are also represented in a balanced fashion. Thirteen items were scored on a 5-point rating scale (1 = not at all, 2 = minimally, 3 = moderately, 4 = quite a bit, 5 = very much). Due to the nature of the question, 1 item was scored on a 3-point scale (1 = no, 2 = perhaps, 3 = yes). Therefore, the range of possible scores on the MRPI: Abridged is from 14 to 68. A high score on the MRPI: Abridged indicates a high degree of religiosity (see the appendix). The 14-item MRPI: Abridged Scale displayed moderate internal consistency with an α coefficient of .73.

Recidivism

Due to the nature of the sample and Pakistani justice regulations, only a single measure of recidivism was used, specifically, the cancelation of a probation order (Probation of Offenders Ordinance, 1960), which occurs when probationers breach the terms of their probation or are charged or convicted of a new offense. The recidivism variable was coded in binary format (0 = no recidivism, 1 = recidivism). Information on recidivism was provided by the Punjab R&P department at the end of follow-up period (i.e., December, 2011), which occurred 10 to 11 months after the LS/CMI assessment of the probationers (i.e., January and February, 2011).

Procedure

Four districts (i.e., Lahore, Kasur, Sheikhupura, and Nankana Sahib of Lahore division) from the Punjab province of Pakistan were identified as the geographical locations for the selection of the participants for this study. A quantitative survey method was used to collect data from the four districts of the Lahore division in the Punjab province of Pakistan. To gather information for the study, formal approval was sought from the R&P department of the Punjab. As discussed previously, POs in each district are responsible for the supervision and the correctional rehabilitation of all probationers in these communities. The research team obtained listings of all adult probationers in the four districts from the Directorate of Punjab at the R&P department, which is the agency’s headquarter office in Lahore.

A team of graduate students from Punjab University in Lahore was trained in the administration of the LS/CMI and then supervised and monitored by one of the authors. Training included a number of presentations about the theoretical and empirical bases of the LS/CMI, as well as a detailed review of the scoring manual and practice in its application to probationers in Pakistan. Face-to-face interviews of adult probationers, who consented to participate in this research, were conducted in the POs’ district offices. Interviews took 30 to 35 min. Offender files were also reviewed to collect socio-economic, demographic, and legal descriptions of the participants. Probationer files are maintained by POs in their respective districts. In accordance with cultural sensitivity, a female researcher assisted the team by conducting the interviews of the female probationers. A data file was created including legal and demographic information, LS/CMI items, total scores from all scales and subscales, and the recidivism outcome data.

Analytic Strategy

Frequency tables and cross-tabulations were used to create descriptive statistics about the sample and participants’ profiles on the LS/CMI. A series of t tests and ANOVA tests were conducted to examine possible variations in the LS/CMI scores for the various offender groups, including those defined by gender, marital status, geographic location, and offense type (i.e., drug vs. nondrug). As well, Cronbach’s alpha statistics were calculated to determine the reliability of the LS/CMI, its subscales, and the MRPI: Abridged Scale. Point-biserial correlations, receiver operating characteristic (ROC) analyses, and logistic regression analyses were conducted to determine the predictive validity of the LS/CMI and the MRPI: Abridged Scale for recidivism among the sample of adult probationers and the various subgroups of probationers. Also, logistic regression analyses were conducted to examine the LS/CMI subscales in relation to recidivism among various subsamples of probationers and to examine the combination of the LS/CMI General Risk/Needs section and the MRPI: Abridged Scale in relation to recidivism. The data were analyzed using SPSS (Version 21) and Lowry’s (2014) statistical calculator for comparing the differences between correlations.

Results

LS/CMI Scores by Gender, Location, and Offense Type

The total sample produced a mean score of 8.58 (SD = 4.46) on the LS/CMI General Risk/Needs. Male probationers (M = 8.24, SD = 4.29) scored significantly lower than female offenders (M = 13.00, SD = 4.43), t(506) = −6.22, p < .001. There was no significant variation between the urban and rural samples on the General Risk/Needs mean score, t(506) = 1.46, p = .145. Notably, drug offenders (M = 9.60, SD = 4.68) also generated significantly higher scores on the General Risk/Needs score than nondrug offenders (M = 7.50, SD = 3.95), t(506) = 5.42, p < .001. All of the preceding mean scores, except for female offenders, would be considered “low risk” in the traditional classification scheme for the LS/CMI (Andrews et al., 2004).

Revision of the LS/CMI Risk Levels

The LS/CMI assigns offenders to a level of risk based on the General Risk/Needs raw score. The five risk levels, which were derived from large samples of offenders in North America, are as follows: very low (0-4), low (5-10), medium (11-19), high (20-29), and very high (30-43; Andrews et al., 2004). Offenders in the current sample were allocated to these risk levels in the following manner: very low (18%), low (45%), medium (35%), high (2%), and very high (0%). It has been suggested by some researchers that the cutoff scores and the number of risk categories for a risk assessment tool may be determined by the correctional or criminal justice agency for its own clientele, keeping in mind the nature of their respective offender populations and the purpose for which the instrument is being applied (Hogg, 2011; Manchak, Skeem, & Douglas, 2008). Consequently, the original cutoff scores were deemed impractical for the present sample of first time offenders because they produced considerably lower scores than those cited in the research (e.g., M = 16.72, SD = 10.11; Andrews et al., 2004) and other samples of community offenders in North America (e.g., M = 13.04, SD = 7.65 in Girard & Wormith, 2004) and because of the highly skewed distribution of Pakistani probationers to risk level that is illustrated above. Therefore, an alternate set of LS/CMI risk levels were derived from the following cutoff scores: very low (0-4), low (5-7), medium (8-12), high (13-17), and very high (18-43) to distribute offenders more evenly across risk levels in anticipation that finer discriminations could be made in terms of recidivism rates among what traditionally had been referred to as low and moderate risk. These cutoffs were devised simply from a review of the distribution of scores and not with any reference to the recidivism outcome measure, hence avoiding the chance of capitalizing on chance variation by favoring one cutoff score over another. This modified risk-level scale generated the following distribution: very low (18%), low (28%), medium (36%), high (14%), and very high (4%). Although still skewed, this more balanced version of a risk-level scale was used in subsequent analyses.

LS/CMI Reliabilities

Reliability analyses were conducted by using Cronbach’s alpha statistic to examine the internal consistency of the items of the LS/CMI sections and the MRPI: Abridged Scale. By excluding the eight Criminal History items, the sample demonstrated a moderate level of reliability (α = .75) for the General Risk/Need items. The alpha coefficients for the LS/CMI subscales (Central Eight Risk/Needs) varied considerably, such as ranging from a high reliability coefficient for the eight-item Alcohol/Drug Problem subscale (α = .79) to a lower reliability coefficient for the four-item Antisocial Pattern subscale (α = .12). The reliability estimates for the Specific Risk/Needs section (14 items) and the Other Client Issues section (22 items) were moderate with alphas of .72 and .78, respectively.

Recidivism

True to the finding that the current sample consisted primarily of low-risk offenders, only 44 (8.7%) probationers recidivated by the end of the follow-up period, which is considerably lower than most samples of adult probationers who have been assessed with the LS/CMI (e.g., Girard & Wormith, 2004) or reported by other jurisdictions. There was no difference between the recidivism rates of the male and female probationers (8.7% vs. 8.3%, respectively), χ²(1) = 0.006, ns. Similarly, recidivism was not associated with geographic location, χ²(1) = 1.589, ns. However, the type of offense demonstrated a relationship with reoffending, such as drug offenders recidivating significantly more often (11.2%) than nondrug offenders (6.1%), χ²(1) = 4.07, p < .05. The length of time on probation at the time of the assessment was unrelated to the LS/CMI score (r = .02, ns) and to recidivism (r = .00, ns).

Predictive Validity of the LS/CMI

As Table 1 illustrates, the LS/CMI General Risk/Needs total score demonstrated a significant point-biserial correlation with recidivism. Among the Central Eight risk/need factors, the Alcohol/Drug Problem outperformed the others in predicting recidivism. With the exception of the Leisure/Recreation factor, the remaining Central Eight risk/need factors were also significantly related to recidivism (all p < .05). The LS/CMI Specific Risk/Needs section and the Other Client Issues section also produced a significant relationship with recidivism.

Table 1:

Point-Biserial Correlations and ROCs Between the LS/CMI Sections, Subscales, and Recidivism for the Complete Sample (N = 506)

LS/CMI, Sections and Subscales	r	AUC (95% CI)
General Risk/Need section	.383***	.820 [0.741, 0.899]
Original Risk categories (5-level)	.271***^a	.755 [0.679, 0.831]
Revised Risk categories (5-level)	.335***^a	.799 [0.719, 0.879]
Specific Risk/Need section	.244***	.693 [0.604, 0.782]
Other Client Issues	.217***	.680 [0.588, 0.772]
General Risk/Need subscales
Education/Employment	.254***	.701 [0.620, 0.798]
Family/Marital	.148***	.643 [0.556, 0.730]
Leisure/Recreation	.068	.558 [0.435, 0.652]
Companions	.228***	.700 [0.614, 0.785]
Alcohol/Drug Problem	.313***	.753 [0.670, 0.836]
Procriminal Attitude	.095*	.595 [0.507, 0.682]
Antisocial Pattern	.214***	.687 [0.609, 0.764]

Note. ROC = receiver operating characteristic; LS/CMI = Level of Service/Case Management Inventory; AUC = area under the curve; CI = confidence interval.

Steiger’s (1980) test for difference of correlations within a matrix, z = 3.15, p < .01.

p < .05. **p < .01. ***p < .001.

ROC analyses were also conducted to examine the validity of the LS/CMI sections on recidivism because ROC is relatively unaffected by base rates that deviate substantially from 50% (Rice & Harris, 2005) as they did in the current investigations. Table 1 presents the area under the curve (AUC) for each of the LS/CMI sections and subsections. The AUCs for the General Risk/Need section, the Specific Risk/Need section, and the Other Client Issues section were all significant as were six of the seven risk/need factors, the exception being the Leisure/Recreation factor. Therefore, the point-biserial correlation and ROC analyses supported the predictive validity of the LS/CMI sections in the prediction of recidivism among Pakistani probationers.

Point-biserial correlations and ROC analyses were also conducted on offender groups as defined by gender, location, and offense type (Table 2). The correlation for the General Risk/Need total score with recidivism was significant for both males and females, and there was no difference in the magnitude of these correlations, as determined by Fisher’s r-to-z transformation for independent samples (Lowry, 2014). For the male offenders, the Specific Risk/Needs section, the Other Client Issues section, and all of the subscales of the General Risk/Need section, except Leisure/Recreation, were significantly related to recidivism (p < .05). However, the pattern was quite different for the female probationers. Only the Alcohol/Drug Problem (r = .49, p < .001) and the Education/Employment (r = .35, p < .05) subscales were robust predictors of recidivism with the female sample. Predictive validity results for the urban sample on the General Risk/Needs were comparable with those of the rural sample, whereas the drug group of probationers revealed a significantly higher point-biserial correlation with recidivism compared with the nondrug offense group of probationers.

Table 2:

Point-Biserial Correlations and ROCs for the General Risk/Needs Section of the LS/CMI and Recidivism by Offender Groups

Offender Groups	n	r	Fisher r-to-z	AUC (95% CI)	Recidivism Rate (%)
Gender
Male	470	.402***	0.22	.825 [0.742, 0.908]	8.723
Female	36	.368*	(ns)	.914 [0.791, 1.000]	8.333 (ns)
Location
Urban	288	.398***	0.54	.825 [0.739, 0.911]	10.069
Rural	218	.356***	(ns)	.797 [0.632, 0.963]	6.880 (ns)
Offense type
Drug	260	.439***	2.18*	.859 [0.791, 0.927]	11.153
Nondrug	246	.269***		717 [0.536, 0.899]	6.098*

Note. ROC = receiver operating characteristic; LS/CMI = Level of Service/Case Management Inventory; AUC = area under the curve; CI = confidence interval; ns = nonsignificant.

p < .05. **p < .01. ***p < .001.

The ROC analysis generally paralleled these findings with a couple of exceptions. Although the AUC for female offenders was particularly high, it was not significantly higher than for the males, as noted by overlapping confidence intervals. The AUC for the urban probationers was similar to that of rural probationers. Finally, the AUC for the drug offenders was higher, but not significantly so, than for the nondrug offenders.

To determine the best combination of predictor variables among the LS/CMI Central Eight risk/need factors, stepwise logistic regressions were conducted for the total sample and various groups of offenders as identified by location and offense type (Table 3). The small sample of female offenders in the current sample precluded logistic regression analyses by gender. Again, one is reminded that these analyses exclude the static Criminal History section that is found in the LS/CMI and all other versions of the LSI. For the entire sample, the following four risk/needs predictors were significant: Companions, Alcohol/Drug Problems, Procriminal Attitudes, and Education/Employment, χ²(4) = 71.046, p < .001; Cox and Snell R² = .131; Nagelkerke R² = .294. The analyses of the urban, χ²(3) = 45.837, p < .001; Cox and Snell R² = .147; Nagelkerke R² = .307, and rural probationers, χ²(2) = 20.065, p < .001; Cox and Snell R² = .088; Nagelkerke R² = .223, and drug, χ²(3) = 55.319, p < .001; Cox and Snell R² = .192; Nagelkerke R² = .381, and nondrug offenders, χ²(2) = 19.516, p < .001; Cox and Snell R² = .076; Nagelkerke R² = .207, generated various combination of subscales in the prediction of recidivism. Unlike the full sample, only two or three subscales contributed to the prediction of recidivism within these subgroups, with Education/Employment and Antisocial Pattern being utilized in three of the four groups.

Table 3:

Stepwise Logistic Regressions With the LS/CMI Subscales to Predict Recidivism

Sample Predictor Variables (LS/CMI Subscales)	B	SE	Wald	df	p	OR (95% CI)
Total sample
Companions	0.387	.185	4.366	1	.037	1.472 [1.024, 2.115]
Alcohol/Drug Problem	0.370	.080	21.404	1	.001	1.448 [1.238, 1.694]
Procriminal Attitude	0.374	.142	6.897	1	.009	1.453 [1.100, 1.921]
Education/Employment	0.382	.101	14.318	1	.001	1.466 [1.202, 1.786]
Constant	−5.440	.553	96.772	1	.001	0.004
Urban group
Alcohol/Drug Problem	0.450	.097	21.557	1	.001	1.569 [1.297, 1.897]
Antisocial Pattern	0.753	.310	5.888	1	.015	2.122 [1.156, 3.898]
Education/Employment	0.274	.128	4.573	1	.032	1.315 [1.023, 1.689]
Constant	−5.062	.640	62.551	1	.001	0.006
Rural group
Companions	0.669	.323	4.666	1	.031	2.011 [1.067, 1.791]
Education/Employment	0.503	.160	9.902	1	.002	1.654 [1.209, 2.263]
Constant	−4.851	.720	45.377	1	.001	0.008
Drug group
Family/Marital	0.601	.208	8.358	1	.004	1.824 [1.213, 2.740]
Alcohol/Drug Problem	0.584	.108	29.125	1	.001	1.794 [1.451, 2.218]
Antisocial Pattern	0.623	.292	4.560	1	.033	1.865 [1.053, 3.305]
Constant	−5.385	.711	57.438	1	.001	0.005
Nondrug group
Antisocial Pattern	0.884	.378	5.042	1	.025	2.335 [1.114, 4.896]
Education/Employment	0.475	.162	8.620	1	.003	1.607 [1.170, 2.207]
Constant	−5.113	.744	47.272	1	.001	0.006

Note. LS/CMI = Level of Service/Case Management Inventory; OR = odds ratio; CI = confidence interval.

LS/CMI Revised Risk Levels and Recidivism

Both the original and revised five-level risk schemes were correlated with recidivism (Table 1). However, the latter version correlated significantly more highly with outcome than the original version, as determined by Steiger’s (1980) test for comparisons within a correlation matrix. Therefore, the practical utility of the revised risk levels on this low-risk sample of offenders was examined more closely. Recidivism rates with the revised version were as follows: very low risk (3%), low risk (2%), medium risk (4%), high risk (25%), and very high risk (55%). A one-way ANOVA was conducted to compare the recidivism rates of offenders in the different risk levels. Overall, there was a significant difference in the recidivism rates between the groups, F(4, 501) = 30.170, p < .001. Furthermore, Bonferroni post hoc multiple comparisons were conducted for all pairs of the risk levels. There were no differences in the recidivism rates between the very low, low, and medium risk groups at p < .05. However, these groups were all significantly different from the high and very high risk groups at p < .001, which were also different from each other at p < .001.

Religiosity and Its Contribution to the LS/CMI in Predicting Recidivism

The mean score on the MRPI: Abridged Scale for the sample was 50.56 (SD = 6.44). There was no difference between the Muslim (M = 50.61, SD = 6.49) and the Christian (M = 49.83, SD = 5.63) offenders with respect to their mean religiosity score, t(506) = .646, ns. The MRPI: Abridged score for the nonrecidivists (M = 50.86, SD = 6.32) was significantly higher than that of the recidivists (M = 47.43, SD = 6.88), t(506) = −3.410, p < .001. The point-biserial correlation reflected an inverse relationship (r = −.15, p < .001) between the religiosity score and recidivism. As expected, the religiosity score was also negatively associated with the General Risk/Need section of the LS/CMI, r = −.24, p < .001. All but one subscale (Procriminal Attitude) correlated with Religiosity (p < .10), with Employment/Education (r = −.22, p < .001), Alcohol/Drug Problems (r = −.20, p < .001), and Leisure/Recreation (r = −.16, p < .001) being most strongly correlated.

To examine whether the MRPI: Abridged Scale contributed additional information to the prediction of recidivism beyond the General Risk/Need section of the LS/CMI, the religiosity score and the LS/CMI General Risk/Need score were entered simultaneously into a logistic regression analysis (Table 4). In Model 1, the General Risk/Need section emerged as the only significant predictor in the logistic regression equation, indicating that religiosity as measured by the MRPI: Abridged Scale was redundant to the LS/CMI in predicting recidivism with this particular sample, χ²(2) = 70.606, p < .001; Cox and Snell R² = .130; Nagelkerke R² = .292. Neither the addition of control variables (gender, location, and offense type) in Model 2, χ²(5) = 77.277, p < .001; Cox and Snell R² = .142; Nagelkerke R² = .317, nor replacing the LS/CMI total risk/needs score with the Risk/Need subscales in Model 3, χ²(8) = 75.592, p < .001; Cox and Snell R² = .139, Nagelkerke R² = .311, had any impact on relevant coefficients.

Table 4:

Logistic Regression (Enter Method) With Religiosity Score and LS/CMI General Risk/Need Score to Predict Recidivism for the Complete Sample

Predictor Variable	B	SE	Wald	df	p	OR (95% CI)
Model 1
General Risk/Need	0.279	0.040	48.391	1	.001	1.322 [1.222, 1.430]
Religiosity Score	−0.043	0.028	2.462	1	.117	0.958 [0.907, 1.011]
Constant	−3.287	1.496	4.826	1	.082	0.037
Model 2
General Risk/Need	0.310	0.045	48.099	1	.001	1.364 [1.249, 1.489
Gender (0-1)	1.450	0.681	4.526	1	.033	4.261 [1.121, 16.200]
Location (0-1)	0.357	0.383	0.869	1	.351	1.429 [0.675, 3.029]
Drug (0-1)	0.171	0.391	0.191	1	.662	1.186 [0.551, 2.553]
Religiosity Score	−0.043	0.028	2.362	1	.124	0.958 [0.907,1.012]
Constant	−5.268	1.758	8.979	1	.003	0.005
Model 3
Education/Employment	0.379	0.104	13.241	1	.001	1.461 [1.191, 1.791]
Companions	0.336	0.191	3.109	1	.078	1.399 [0.963, 2.033]
Alcohol/Drug Problem	0.375	0.088	17.968	1	.001	1.455 [1.223, 1.730]
Procriminal Attitude	0.389	0.143	7.422	1	.006	1.476 [1.115, 1.952]
Gender (0-1)	0.738	0.666	1.227	1	.268	2.092 [0.567, 7.723]
Location (0-1)	0.275	0.396	0.483	1	.487	1.316 [0.606, 2.858]
Drug (0-1)	0.202	0.414	0.238	1	.626	1.224 [0.543, 2.756]
Religiosity score	−0.046	0.029	2.564	1	.109	0.955 [0.903, 1.010]
Constant	−4.118	1.716	5.756	1	.016	0.016

Note. LS/CMI = Level of Service/Case Management Inventory; OR = odds ratio; CI = confidence interval.

Discussion

The purpose of this research was to examine the LS/CMI General Risk/Needs section with an atypical offender sample, namely, low-risk adult probationers in a previously unexamined, non-Western culture and second, to assess its predictive validity. A third purpose was to assess the relationship between religiosity, the general risk/need factors, and recidivism in the aforementioned sample.

Scale and Sample Characteristics

The research supported the LS/CMI General Risk/Needs section as being moderately reliable. Its Cronbach’s alpha was comparable with the alpha estimates reported in prior research (Andrews et al., 2004). The reliability findings on the subscales were considerably varied. The Alcohol/Drug Problem subscale was the most reliable and the Antisocial Pattern subscale was the least reliable among the general risk/need factors. The Antisocial Pattern subscale is purposely designed to capture a range of heterogeneous items and therefore its reliability is often low (Andrews et al., 2004, 2010). However, the current finding was uncharacteristically so, possibly due to the very low frequency of its endorsed items for this low-risk sample. Regardless, the Antisocial Pattern subscale’s correlation with recidivism was at the median among the seven subscales, supporting its importance and continued use. The moderate reliabilities for the Specific Risk/Needs and Other Issues (social, health, and mental health problems) were comparable with the reliability estimates reported elsewhere for North American samples (Andrews et al., 2004; Girard & Wormith, 2004).

By virtue of the criteria required for probation in Pakistan, it was inferred that the sample came from a low-risk offender population. This was verified in two ways: the average General Risk/Needs score (8.58%), which is approximately one half of the published mean for North American probationers (16.72%; Andrews et al., 2004), and the overall recidivism rate (8.7%), which is also about one half the recidivism rate of federal probationers in the United States (18%) as measured by arrest or revocation within 1 year (Rhodes, Dyous, Kling, Hunt, & Luallen, 2012). This low-risk feature of probationers in Pakistan added a second reason, besides the country’s cultural uniqueness, for examining the applicability of the LS/CMI in this context. Specifically, is there any value in routinely administering a standard risk/need assessment instrument to a low-risk offender population, and if so, how might it contribute to the operation of the correctional agency, in this case probation services?

Predictive Validity of LS/CMI

Statistical tests established the predictive accuracy of the LS/CMI Risk/Needs score and the other scales and subscales among Pakistani adult probationers, irrespective of gender, geographical location, and the type of offense (i.e., drug or other). Moreover, the results were obtained without the benefit of any variability on the Criminal History section of the LS/CMI, which, along with Antisocial Pattern, routinely generates the highest predictive validity coefficients in North American samples (Olver et al., 2014). Hence, the instrument was reduced to its dynamic criminogenic need factors. The results were obtained also in spite of the fact that the sample was a relatively homogeneous group of low-risk probationers from an Asian country where the risk/need model was foreign. Nonetheless, the predictive validity estimates were comparable with the results of previous research, including meta-analytic investigations (Andrews et al., 2011; Gendreau et al., 2002; Girard & Wormith, 2004; Hollin & Palmer, 2003; Miles & Raynor, 2004; Olver et al., 2014; Smith, Cullen, & Latessa, 2009).

The predictive validity of the seven Risk/Need subscales factors on recidivism was examined through point-biserial correlations and stepwise regression analysis. Applied to the total sample, the regression analysis revealed four factors (i.e., Companions, Alcohol/Drug Problems, Procriminal Attitudes, and Education/Employment) as the best combination of predictors for recidivism. Two of these (i.e., Companions and Procriminal Attitudes) have been characterized as part of the “Big Four” in the risk, need, and responsivity (RNR) model, the other two being Criminal History and Antisocial Pattern (Andrews & Bonta, 2010, p. 58). The emergence of the Alcohol and Drug Problems subscale as being among the strongest Risk/Need subscales is contrary to the listing of the Big Four, but reminiscent of other Canadian (e.g., Andrews et al., 2012; Rettinger & Andrews, 2010) and worldwide studies (Olver et al., 2014). Coupled with the fact that different combinations of criminogenic need areas emerged in the analyses of offender subgroups, these results raise some question as to whether any of the Central Eight should be routinely declared the Big Four.

The finding that the Other Issues (social, health, and mental health) section was related to the outcome deserves attention. This so-called, noncriminogenic section was not designed to do so theoretically, but was included for case management planning more generally (Wormith, 1997). Nonetheless, these results are not unique as previous research has also found similar results (Girard & Wormith, 2004; Wormith et al., 2012). Although its repeated association with recidivism does not solely confirm its criminogenic function, further investigation on these Other Issues is suggested, to examine not only its contribution to treatment programming and case management, but also its function in the context of risk/need assessment.

The finding that female probationers yielded a considerably higher mean score on the general risk/needs score compared with their male counterparts is contrary to the traditional conclusions reported in previous research (Andrews et al., 2004; Austin, 2003; Coulson, Ilacqua, Nutbrown, Giulekas, & Cudjoe, 1996). Their relatively high General Risk/Needs scores, coupled with their very similar rate of recidivism as male probationers, as well as the differences in correlations for the Risk/Need subscales suggests an inconsistency by gender and raises some questions about the use of a standard risk/need assessment with women offenders (Belknap & Holsinger, 2006; Holtfreter, Reisig, & Morash, 2004; Reisig, Holtfreter, & Morash, 2006; Salisbury, Van Voorhis, & Spiropoulos, 2009). However, an alternative explanation of their higher LS/CMI scores in spite of comparable recidivism rates would be a cultural, gender bias in the assessment of female offenders in Pakistan.

The high correlation for alcohol and drug problems with recidivism for women is reminiscent of other LS findings (Olver et al., 2014), adding to the contention that, although risk factors may be the same, their relative strength may vary by gender, not only in North America, but across cultures. However, this in itself does not invalidate use of the LS/CMI with women offenders, and the small female sample necessitates caution when interpreting these findings.

Predictive validity analyses of the LS/CMI generated similar findings and patterns between the two statistical tests. On one hand, the relationship between the Leisure/Recreation subscale and recidivism was nonsignificant indicating that it was unrelated to recidivism regardless of the measure. On the other hand, in examining patterns of results, it is noted that male probationers generated a higher point-biserial correlation with recidivism, while females generated a higher ROC value. There was no difference between the recidivism rates for males and females, indicating that the difference was not a product of different base rates.

Religiosity, Recidivism, and LS Risk/Need Assessment

The modest negative relationship between religiosity and recidivism was consistent with previous research (e.g., Johnson, Larson, & Pitts, 1997), as well as the more general religion–crime relationship (Baier & Wright, 2001; Benda, Toombs, & Peacock, 2003; Johnson, 2004). Moreover, the current investigation extends the religiosity–crime relationship to a predominantly Muslim culture. In so doing, it supports the theoretical perspective and empirical evidence that religious beliefs, values, and practices are related to criminal behavior. This relationship applies not only to Christianity, but extends to Islam, and perhaps beyond, suggesting a nonspecific relationship between religious affiliation and prosocial behavior. This makes intuitive sense in that religious concepts that promote prosocial behavior, such as the Golden Rule, are found in the three major monotheistic religions. The possibility of such a relationship leads one to ask why measures of religious affiliation are not included in any of the common risk assessment instruments. The common rebuttal of course is, do they need to?

The answer to this question comes from a series of analyses conducted in the present study. As expected, there was a significant negative correlation between offender risk as measured by the LS/CMI General Risk/Need items and religiosity as measured by the MRPI: Abridged Scale. However, the MRPI: Abridged Scale did not add any incremental predictive validity to the LS/CMI Risk/Needs on recidivism. In other words, the variance in religiosity that was related to recidivism was also related to a portion of LS/CMI Risk/Needs that was also related to recidivism, and knowledge about probationers’ religiosity did not improve the prediction of recidivism beyond the capacity of the LS/CMI total risk/need score items in predicting recidivism among Pakistani probationers. The fact that this result was still found when gender, location, and offense type were added as control variables and when the Risk/Need subscales were used in place of the LS/CMI total risk/need score strengthens its credibility, at least with the two measures used in the current investigation.

Yet, this is not to suggest that religiosity is irrelevant to offender rehabilitation, as no causal pathway has been established. Rather, religious exposure and training could be very instrumental in lowering a number of criminogenic needs areas that are identified in tools such as the LS/CMI (e.g., lacking conventional aspirations for education and employment, alcohol and drug use, antisocial peers, criminal attitudes, and inappropriate use of leisure time). Resolving the nature of the relationship between risk/need, religiosity, and criminal behavior requires a longitudinal design or one that entails an experimental intervention.

Implications for the Probation System in Pakistan

The present study demonstrated the LS/CMI’s capacity to assess the risk of recidivism of adult probationers and to identify their specific risk/needs and noncriminogenic needs. Therefore, it is recommended that the R&P department in Punjab, Pakistan, extend the use of the LS/CMI and its risk/need factors to large samples of offenders and to translate the findings into practical applications such as PSRs, offender classification, case management, and, above all, correctional planning and the rehabilitation of offenders. This may be extended to its use by the courts to grant probation to eligible offenders in a more effective and judicious manner. Knowledge about the risk factors of first time offenders in Pakistan also has relevance for policy on crime prevention and offender management.

The recidivism rates for the revised risk levels were also of interest. In spite of our attempt to generate a revised risk-level classification in an attempt to assess variations in low risk among a group of already known low-risk offenders, the capacity of the instrument to differentiate the three lowest risk levels in terms of recidivism was futile. However, what was described as high and very high risk probationers proved to be very different from the lower risk groups and from each other. Moreover, use of the modified cutoffs that were developed in the current investigation is supported by their higher predictive validity coefficient with recidivism. Yet, it also appears reasonable to collapse the three lowest risk groups, resulting in three risk groups, perhaps labeled very low, low, and moderate, relative to offender recidivism rates generally. However, such a classification scheme would be limited to probationers in Pakistan. It remains to be seen how the instrument applies to (presumably) higher risk prisoners in the same country. Regardless, the empirical findings indicate a single distinguishable group of very low-risk offenders. In this regard, we present this exercise as a case study in developing one’s own local or agency-specific norms, a practice that has been endorsed (Wormith, 2014).

Limitations

The present study is not without its limitations. The first concern relates to the unique sample, at least as far as the application of the LS/CMI is concerned or, for that matter, any offender risk assessment instrument. The merits of studying this sample are discussed elsewhere. Our concern about the sample pertains to its geographically limited catchment area in a relatively large and diverse country. The participants of this research came from four districts in the Lahore Division of the Punjab province and do not necessarily represent adult probationers from the other 31 districts, or zillahs, of the Punjab province, let alone the three other provinces and more than 100 districts throughout the country. Consequently, the study should not be construed as an exercise that represents Pakistani probationers nationally. However, we do point out that the kind of local sampling used in the current study, actually across four probation sites, is not any different from the local samples used in many Western risk assessment studies. Second, representing only 7.1% of the sample, women made up a low percentage of the adult probationers, a rate that is even smaller than the proportion of women reported in most Western studies where their proportion has been rising for the last decade (e.g., 24% of probationers in the United States in 2012; Maruschak & Bonczar, 2013). Clearly, the analysis of female probationers’ data was hampered by the small sample, which limited the stability and power of the results, and therefore, the discussion of our current findings carries a special caveat for the women offenders.

Additional limitations are related to both of the prediction measures that were used. Considerable effort was made to ensure the integrity of the LS/CMI in its application to probationers in the Pakistani probation system. As noted previously, a few adaptations were made to specific items in the LS/CMI. This kind of modification has been made in a number of other countries, such as Singapore and Hong Kong, where translations of the LSI-R and LS/CMI have generated respectable validities, although we also note that international differences have been found in the predictive validity of the LSI-R and LS/CMI, the reason for which has been subject to speculation (Andrews et al., 2011; Olver et al., 2014; Yang et al., 2010). Consequently, the implications of the translation and adjustments made to the Urdu version of the LS/CMI are unknown. Also, in spite of several training sessions of the assessment team, the research may have been hampered by the quality of the LS/CMI assessments that were conducted as there was not an LS/CMI expert on-site to monitor the process, nor was there an opportunity, for administrative reasons, to conduct an interrater reliability check on the assessments.

Similarly, the current study represents the first use of the MRPI: Abridged Scale and further research is necessary to validate its use in the assessment of religiosity in Muslim cultures. The abridged version was developed, in part, because of the length of the original 102-item version with its two subscales (Krauss et al., 2005), which made it impractical for administration to the probationers. A more detailed psychometric investigation of the revision process, including factor analyses of the items, is available elsewhere (Bhutta, 2013) and further construct validation work is under way. Regardless, in considering the results of the current study pertaining to religiosity, one is mindful of the limited research on the Urdu version of the MRPI: Abridged Scale.

Moreover, the current findings may be specific to the manner by which religiosity and risk were captured in the current study. Not only did this investigation focus on religiosity in Muslim culture, but also it has operationalized religiosity primarily in a religious practice manner, rather than a sense of morality that some believe is a consequence of religion (e.g., Stark & Glock, 1968). Similarly, the current findings may not apply to other measures of offender risk. It is quite possible that instruments, particularly the so-called static tools that do not assess dynamic constructs such as antisocial attitudes and companions, are more independent of religiosity in which case religiosity could make an independent contribution to the prediction of offender recidivism.

As well, there are limitations pertaining to the outcome measure. A single measure (i.e., cancelation of a probation order for adult probationers) in a relatively short period of time (e.g., 10-11 months) might be considered less than inadequate to determine the recidivism of probationers in this study. However, because a new criminal charge or conviction automatically results in the cancelation of a probation order, charges and convictions were indirectly captured by this measure. The research team was only advised of the cancelation of the order and hence, was unable to differentiate between breaches of probation and new offenses or even the types of offenses. Consequently, we were unable to include violent recidivism in our analysis, although it is unlikely that there was a sufficient number of violent re-offenses to examine statistically, given the nature of this sample. Similarly, the researchers did not have access to the dates on which the assessments were conducted or recidivism occurred. Therefore, time to recidivate and corresponding statistics, such as survival analyses, and recidivism by length of follow-up, although minimal in terms of its variability, could not be calculated.

The short follow-up duration is likely to have been partly responsible for the low recidivism rate, although with a particularly low-risk sample of offenders, it is impossible to determine the extent to which the rate was a function of the low-risk sample or the short follow-up period. The brief follow-up may also have had an impact on the predictive validity coefficients as potential re-offenders require some time to commit their offenses and for their offenses to be detected and officially prosecuted by the justice system. In a meta-analysis of numerous risk instruments in the prediction of violence, Yang et al. (2010) found that predictive validity coefficients were positively correlated with the length of follow-up. This leads one to speculate that a longer follow-up period could generate higher validities.

Conclusion

This study has contributed three important findings that are of both theoretical and practical value. First, it has demonstrated the predictive validity of a traditional, Western, risk/need assessment instrument, the LS/CMI, in an Eastern, highly devout Muslim culture. Moreover, it did so across gender, urban and rural geographical locations, and by the type of offense (drug and nondrug). As such, the study generally supports the use of the Central Eight (Andrews & Bonta, 2010) as criminogenic needs in one particular Asian culture, although Criminal History could not be assessed and Leisure/Recreation did not prove to be predictive. Nonetheless, as offender risk assessments have not been implemented, or even introduced into the Pakistani justice system, the current study is encouraging and sets the stage for more widespread implementation and research in this area.

Second, irrespective of culture, the study demonstrates the utility of a risk/need assessment, such as the LS/CMI in a low-risk population of offenders. This point might be easily overlooked as so much of our attention in corrections is focused on the high risk, violent offenders, those who are most likely to commit the worst offenses. Yet in conjunction with Andrews et al.’s (1990) risk principle, it is also important to attend to the low-risk offenders so that they may be screened out of any excessive criminal justice processing and allocated to minimal conditions. In Pakistan, this amounts to probation, but in Western justice systems, where probation can include serious high-risk offenders, there is a plethora of diversion programs and alternative measures that entail minimal cost and are unlikely to have the iatrogenic effects of serious interference. The fact that the current study was capable of differentiating recidivists from nonrecidivists in this generally low-risk sample, albeit with adjusted cutoffs and no discrimination between very low, low, and medium risk, illustrates the other side of the double-edged, risk assessment enterprise, which is the capacity to identify recidivists in a low-risk, offender population. Again, these findings are both of theoretical and practical value, not only to Pakistan, but to Western correctional agencies as well, including those that are struggling to identify low-risk offenders to be released from overcrowded facilities.

Finally, the study brings the forgotten factor, religiosity, back into the dialogue. This statement may actually be an exaggeration as religiosity has never really been part of the offender risk assessment paradigm. Rather, the construct has been studied in quite different literature, largely by criminologists, to assist in explaining the underpinnings of their own perspectives of criminality, such as control theory. Not surprisingly, religiosity was correlated negatively, not only with recidivism, but also with risk/needs, such that it did not enhance the prediction of recidivism when the LS/CMI was in place. This is not to say that religiosity is irrelevant to our understanding of criminality or its prevention. The nonexperimental nature of the current study does not allow one to determine the extent to which religiosity may have contributed prospectively to a lack of problems in such risk/need areas as education, employment, alcohol, and drug use because the temporal order of their relationship is unknown. Moreover, religiosity’s lack of incremental validity in the current study cannot be generalized to secular societies where the church and state are independent. Yet, what we have learned is that the addition of religiosity to the LS/CMI would not likely improve the instrument’s capacity to predict recidivism, at least in this highly religious culture.

Footnotes

Appendix

J. S. Wormith receives royalty payments for sales of the Level of Service/Case Management Inventory (LS/CMI) from its publisher, Multi-Health Systems.

Notes

Mazhar H. Bhutta, PhD, is an assistant professor of sociology in the Department of Sociology and Criminology, Pir Mehr Ali Shah-Arid Agriculture University Rawalpindi–Pakistan. He has recently initiated an MSc Criminology Program at Pir Mehr Ali Shah-Arid Agriculture University Rawalpindi–Pakistan and has extensive research work on the probation system in Pakistan. His research interests include offenders risk assessment, crime prevention, correctional intervention/rehabilitation, criminal justice system, the sociology of religion, and religiosity and criminality.

J. Stephen Wormith, PhD, is a professor in the psychology department and the director of the Centre for Forensic Behavioural Science (), both at the University of Saskatchewan. He is a co-author of the Level of Service/Case Management Inventory (LS/CMI). His research interests include offender risk assessment, evidence-based intervention with offenders, mental health courts, victimization surveys, and crime prevention.

References

Amer

M. M.

Bagasra

(2013). Psychological research with Muslim Americans in the age of the Islamophobia: Trends, challenges, and recommendations. American Psychologist, 68, 134-144.

Andrews

D. A.

Bonta

(1995). The Level of Service Inventory–Revised: User’s manual. Toronto, Ontario, Canada: Multi-Health Systems.

Andrews

D. A.

Bonta

(2010). The psychology of criminal conduct (5th ed.). Newark, NJ: LexisNexis.

Andrews

D. A.

Bonta

Hoge

R. D.

(1990). Classification for rehabilitation: Rediscovering psychology. Criminal Justice and Behavior, 17, 19-52.

Andrews

D. A.

Bonta

Wormith

J. S.

(1995). Level of Service Inventory–Ontario Revision (LSI-OR): Interview and scoring guide. Toronto, Ontario, Canada: Multi-Health Systems.

Andrews

D. A.

Bonta

Wormith

J. S.

(2004). The Level of Service/Case Management Inventory: An offender assessment system—User’s manual. Toronto, Ontario, Canada: Multi-Health Systems.

Andrews

D. A.

Bonta

Wormith

J. S.

(2006). The recent past and near future of risk and/or need assessment. Crime & Delinquency, 52, 7-27.

Andrews

D. A.

Bonta

Wormith

J. S.

(2010). The Level of Service (LS) assessment of adults and older adolescents. In Otto

R. K.

Douglas

K. S.

(Eds.), Handbook of violence risk assessment (pp. 199-225). New York, NY: Routledge.

Andrews

D. A.

Bonta

Wormith

J. S.

Guzzo

Brews

Rettinger

Rowe

(2011). Sources of variability in estimates of predictive validity: A specification with level of service general risk and need. Criminal Justice and Behavior, 38, 413-432.

10.

Andrews

D. A.

Guzzo

Raynor

Rowe

R. C.

Rettinger

L. J.

Brews

Wormith

J. S.

(2012). Are the major risk/need factors predictive of both female and male reoffending? A test with the eight domains of the Level of Service/Case Management Inventory. International Journal of Offender Therapy and Comparative Criminology, 56, 113-133.

11.

Andrews

D. A.

Zinger

Hoge

R. D.

Bonta

Gendreau

Cullen

F. T.

(1990). Does correctional treatment work? A clinically relevant and psychologically informed meta-analysis. Criminology, 28, 369-404.

12.

Annalakshmi

Abeer

(2011). Islamic worldview, religious personality and resilience among Muslim adolescent students in India. Europe’s Journal of Psychology, 7, 716-738.

13.

Aos

Miller

Drake

(2006). Evidence-based adult corrections programs: What works and what does not. Olympia: Washington State Institute for Public Policy.

14.

Austin

(2003). Findings in prison classification and risk assessment. Washington, DC: U.S. Department of Justice, National Institute of Corrections.

15.

Baier

C. J.

Wright

B. R. E.

(2001). If you love me, keep my commandments: A meta-analysis of the effect of religion on crime. Journal of Research in Crime & Delinquency, 38, 3-21.

16.

Baier

(2014). The influence of religiosity on violent behavior of adolescents: A comparison of Christian and Muslim religiosity. Journal of Interpersonal Violence, 29, 102-127.

17.

Beaumont

G. D.

de Tocqueville

(1964). On the penitentiary system in the United States and its application in France. Carbondale: Southern Illinois University Press. (Original work published 1833)

18.

Belknap

Holsinger

(2006). The gendered nature of risk factors for delinquency. Feminist Criminology, 1, 48-71.

19.

Benda

B. B.

Toombs

N. J.

Peacock

(2003). Discriminators of types of recidivism among boot camp graduates in a five-year follow-up study. Journal of Criminal Justice, 31, 539-551.

20.

Bhutta

M. H.

(2010). Community based rehabilitation of offenders: An overview of probation and parole system in Pakistan. Pakistan Journal of Criminology, 2, 51-67.

21.

Bhutta

M. H.

(2013). Risk and need assessment of offenders on probation in Lahore (Unpublished doctoral dissertation). University of the Punjab, Lahore, Pakistan.

22.

Blanchette

Brown

S. L.

(2006). The assessment and treatment of women offenders: An integrative perspective. Chichester, UK: John Wiley.

23.

Camp

S. D.

Klein-Saffran

Kwon

Daggett

D. M.

Joseph

(2006). An exploration into participation in a faith-based prison program. Criminology & Public Policy, 5, 529-550.

24.

Chng

Hong

N. L.

Misir

(2002). Security classification revisited: Predicting institutional violence using LSI–R and SAQ. In Wong

Fang

T. T.

Sem

L. K.

Gob

(Eds.), Correctional research compendium (2006) (pp. 82-87). Singapore: Singapore Prison Service.

25.

Chu

C. M.

Lee

Zeng

Yim

Tan

C. Y.

Ang

. . . Ruby

(2015). Assessing youth offenders in a non-western context: The predictive validity of the YLS/CMI ratings. Psychological Assessment, 27, 1013-1021. doi:10.1037/a0038670

26.

Coulson

Ilacqua

Nutbrown

Giulekas

Cudjoe

(1996). Predictive utility of the LSI for incarcerated female offenders. Criminal Justice and Behavior, 23, 427-439.

27.

Duwe

Johnson

B. R.

(2013). Estimating the benefits of a faith-based correctional program. International Journal of Criminology and Sociology, 2, 227-239.

28.

Duwe

King

(2012). Can faith-based correctional programs work? An outcome evaluation of the Inner Change Freedom Initiative in Minnesota. International Journal of Offender Therapy and Comparative Criminology, 57, 813-814.

29.

Evans

T. D.

Cullen

F. T.

Dunaway

R. G.

Burton

V. S.

(1995). Religion and crime reexamined: The impact of religion, secular controls, and social ecology on adult criminality. Criminology, 33, 195-224.

30.

Gendreau

Goggin

Smith

(2002). Is the PCL-R really the “unparalleled” measure of offender risk? A lesson in knowledge cumulation. Criminal Justice and Behavior, 29, 397-426.

31.

Girard

Wormith

J. S.

(2004). The predictive validity of the Level of Service Inventory–Ontario Revision on general and violent recidivism among various offender groups. Criminal Justice and Behavior, 31, 150-181.

32.

Gottfredson

M. R.

Hirschi

(1990). A general theory of crime. Redwood City, CA: Stanford University Press.

33.

Grove

W. M.

Zald

D. H.

Lebow

B. S.

Snitz

B. E.

Nelson

(2000). Clinical versus mechanical prediction: A meta-analysis. Psychological Assessment, 12, 19-30.

34.

Gutierrez

Wilson

Rugge

Bonta

(2013). The prediction of recidivism with Aboriginal offenders: A theoretically-informed meta-analysis. Canadian Journal of Criminology and Criminal Justice, 55, 55-99.

35.

Hannah-Moffat

(2013). Actuarial sentencing: An “unsettled” proposition. Justice Quarterly, 30, 270-296.

36.

Heaton

(2006). Does religion really reduce crime? Journal of Law & Economics, 49, 147-172.

37.

Hoge

R. D.

Andrews

D. A.

(2002). The Youth Level of Service/Case Management Inventory: Manual and scoring key. Toronto, Ontario, Canada: Multi-Health Systems.

38.

Hogg

S. M.

(2011). The Level of Service Inventory–Ontario Revision Scale validation for gender and ethnicity: Addressing reliability and predictive validity (Unpublished master’s thesis). University of Saskatchewan, Saskatoon, Canada.

39.

Hollin

C. R.

Palmer

E. J.

(2003). Level of Service Inventory-Revised profiles of violent and non-violent prisoners. Journal of Interpersonal Violence, 18, 1075-1086.

40.

Holtfreter

Reisig

M. D.

Morash

(2004). Poverty, state capital, and recidivism among women offenders. Criminology & Public Policy, 3, 185-208.

41.

Johnson

B. R.

(2004). Religious programs and recidivism among former inmates in Prison Fellowship programs: A long-term follow-up study. Justice Quarterly, 21, 329-354.

42.

Johnson

B. R.

(2011). More God, less crime: Why faith matters and how it could matter more. West Conshohocken, PA: Templeton Press.

43.

Johnson

B. R.

De Li

Larson

D. B.

McCullough

(2000). A systematic review of the religiosity and delinquency literature: A research note. Journal of Contemporary Criminal Justice, 16, 32-52.

44.

Johnson

B. R.

Jang

S. J.

(2010). Religion and crime: Assessing the role of the faith factor. In Rosenfeld

Quinet

Garcia

(Eds.), Contemporary issues in criminological theory and research: The role of social institutions (pp. 117-150). Belmont, CA: Wadsworth.

45.

Johnson

B. R.

Larson

D. B.

Pitts

T. C.

(1997). Religious programs, institutional adjustment, and recidivism among former inmates in prison fellowship programs. Justice Quarterly, 14, 145-166.

46.

Krauss

S. E.

Hamzah

A. H.

(2009). The Muslim Religiosity-Personality Inventory (MRPI). Selangor: Putra University Malaysia.

47.

Krauss

S. E.

Hamzah

A. H.

Idris

(2007). Adaptation of a Muslim religiosity scale for use with four different faith communities in Malaysia. Review of Religious Research, 49, 147-164.

48.

Krauss

S. E.

Hamzah

A. H.

Ismail

I. A.

Suandi

Hamzah

S. R.

Dahalan

Idris

(2012). Parenting, community and religious predictors of positive and negative developmental outcomes among Muslim adolescents. Youth & Society, 46, 201-227.

49.

Krauss

S. E.

Hamzah

A. H.

Juhari

(2005). The Muslim Religiosity-Personality Inventory (MRPI): Towards understanding differences in the Islamic religiosity among the Malaysian youth. Pertanika Journal of Social Sciences & Humanities, 13, 173-186.

50.

Krauss

S. E.

Hamzah

A. H.

Suandi

Noah

S. M.

Juhari

Manap

J. H.

. . . Mahmood

(2006). Exploring regional differences in religiosity among Muslim youth in Malaysia. Review of Religious Research, 3, 238-252.

51.

Latessa

E. J.

Lovins

(2010). The role of offender risk assessment: A policy maker guide. Victims & Offenders, 5, 203-219.

52.

Lowry

(2014). VassarStats: Website for statistical computation. Available from vassarstats.net

53.

Maloney

Miller

(2015). The impact of a risk assessment instrument on juvenile detention decision-making: A check on “perceptual shorthand” and “going rates”? Justice Quarterly, 32, 900-927. doi:10.1080/07418825.2013.863961

54.

Manchak

S. M.

Skeem

J. L.

Douglas

K. S.

(2008). Utility of the Revised Level of Service Inventory (LSI-R) in predicting recidivism after long-term incarceration. Law and Human Behavior, 32, 477-488.

55.

Martel

Brassard

Jaccoud

(2011). When two worlds collide: Aboriginal risk management in Canadian corrections. British Journal of Criminology, 51, 235-255.

56.

Maruschak

L. M.

Bonczar

T. P.

(2013). Probation and parole in the Unites States, 2012 (NCJ 243826). Washington, DC: Bureau of Justice Statistics, U.S. Department of Justice. Retrieved from http://www.bjs.gov/content/pub/pdf/ppus12.pdf

57.

Miles

Raynor

(2004). Community sentences in Jersey: Risks, needs and rehabilitation. St. Helier: Jersey Probation and After Care Service.

58.

Olver

M. E.

Stockdale

K. C.

Wormith

J. S.

(2014). Thirty years of research on the Level of Service scales: A meta-analytic examination of predictive accuracy and sources of variability. Psychological Assessment, 26, 156-176.

59.

Probation of Offenders Ordinance. (1960). The probation of offenders ordinance, 1960 (Section 7). Retrieved from http://www.sparcpk.org/pdf/The%20Probation%20of%20Offenders%20Ordinance%201960.pdf

60.

Probation of Offenders Ordinance, 1960/Rules. (1961). West Pakistan probation of offenders rules. Retrieved from http://punjablaws.punjab.gov.pk/public/dr/PROBATION%20OF%20OFENDERS%20RULES,%201961.doc.pdf

61.

Raiya

H. A.

Pargament

K. I.

Stein

Mahoney

(2007). Lessons learned and challenges faced in developing the psychological measure of Islamic religiousness. Journal of Muslim Mental Health, 2, 133-154.

62.

Reisig

M. D.

Holtfreter

Morash

(2006). Assessing recidivism risk across female pathways to crime. Justice Quarterly, 23, 384-405.

63.

Reisig

M. D.

Wolfe

S. E.

Pratt

T. C.

(2012). Low self-control and the religiosity-crime relationship. Criminal Justice and Behavior, 39, 1172-1191.

64.

Rettinger

L. J.

Andrews

D. A.

(2010). General risk and need, gender specificity, and the recidivism of female offenders. Criminal Justice and Behavior, 37, 29-46.

65.

Rhodes

Dyous

Kling

Hunt

Luallen

(2012). Recidivism of offenders on federal community supervision (Award Number 2010-BJ-CX-K069). Cambridge, MA: Abt Associates. Retrieved from https://www.ncjrs.gov/pdffiles1/bjs/grants/241018.pdf

66.

Rice

M. E.

Harris

G. T.

(2005). Comparing effect sizes in follow-up studies: ROC area, Cohen’s d, and r. Law and Human Behavior, 29, 615-620.

67.

Rocque

Plummer-Beale

(2014). In the eye of the beholder? An examination of the inter-rater reliability of the LSI-R and the YLS/CMI in a correctional agency. Journal of Criminal Justice, 42, 568-578.

68.

Salisbury

E. J.

Van Voorhis

Spiropoulos

(2009). The predictive validity of a gender-responsive needs assessment: An exploratory study. Crime & Delinquency, 55, 550-585.

69.

Sinha

J. W.

Cnaan

R. A.

Gelles

R. J.

(2007). Adolescent risk behaviors and religion: Findings from a national study. Journal of Adolescence, 30, 231-249.

70.

Smith

Cullen

F. T.

Latessa

E. J.

(2009). Can 14,737 women be wrong? A meta-analysis of the LSI-R and recidivism for female offenders. Criminology & Public Policy, 8, 183-208.

71.

Stark

Glock

C. Y.

(1968). American piety: The nature of religious commitment. Berkeley: University of California Press.

72.

Steiger

J. H.

(1980). Tests for comparing elements of a correlation matrix. Psychological Bulletin, 87, 245-251. Retrieved from www.psychmike.com/dependent_correlations.php

73.

Takahashi

Mori

Kroner

D. G.

(2013). A cross-validation of the Youth Level of Service/Case Management Inventory (YLS/CMI) among Japanese juvenile offenders. Law and Human Behavior, 37, 389-400.

74.

Van Voorhis

Wright

E. M.

Salisbury

Bauman

(2010). Women’s risk factors and their contributions to existing risk/needs assessment: The current status of a gender-responsive supplement. Criminal Justice and Behavior, 37, 261-288.

75.

Vazsonyi

A. T.

Huang

(2015). Hirschi’s reconceptualization of self-control: Is truth truly the daughter of time? Evidence from eleven cultures. Journal of Criminal Justice, 43, 59-68.

76.

Volokh

(2011). Do faith-based prison programs work? Alabama Law Review, 63, 43-95.

77.

Welch

M. R.

Tittle

C. R.

Grasmick

H. G.

(2006). Christian religiosity, self-control and social conformity. Social Forces, 84, 1605-1623.

78.

Whiteacre

K. W.

(2006). Testing the Level of Service Inventory-Revised (LSI-R) for racial/ethnic bias. Criminal Justice Policy Review, 17, 330-342.

79.

Wilson

H. A.

Gutierrez

(2014). Does one size fit all? A meta-analysis examining the predictive ability of the Level of Service Inventory (LSI) with Aboriginal offenders. Criminal Justice and Behavior, 41, 196-219.

80.

Wormith

J. S.

(1997). Research to practice: Applying risk/needs assessment to offender classification. Forum on Corrections Research, 9, 26-31.

81.

Wormith

J. S.

(2011). The legacy of D. A. Andrews in the field of criminal justice: How theory and research can change policy and practice. International Journal of Forensic Mental Health, 10, 78-82.

82.

Wormith

J. S.

(2014). The risks in communicating sexual offender risk. Journal of Threat Assessment and Management, 1, 162-178.

83.

Wormith

J. S.

Hogg

Guzzo

(2012). The predictive validity of a general risk/needs assessment inventory on sexual offender recidivism and an exploration of the professional override. Criminal Justice and Behavior, 39, 1511-1538.

84.

Yang

Wong

S. C.

Coid

(2010). The efficacy of violence prediction: A meta-analytic comparison of nine risk assessment instruments. Psychological Bulletin, 136, 740-767.

85.

Young

M. C.

Gartner

O’Connor

Larson

Wright

(1995). Long-term recidivism among Federal inmates trained as volunteer prison ministers. Journal of Offender Rehabilitation, 22, 97-118.