Correctional Service Canada
Symbol of the Government of Canada
Skip to content | Skip to institutional links

Common menu bar links

Home > Research > Research Reports

Institutional links

Research Reports

Warning This Web page has been archived on the Web.

Assessing Security Reclassification with Male Aboriginal and Non-Aboriginal Offenders

2009 No R-203

Renée Gobeil

Correctional Service Canada

September 2008

Acknowledgements

Appreciation is extended to Karen Beattie for her gracious assistance with the early parts of the planning of this project. Thanks also to Colette Cousineau, Justin Gileno, and Mark Nafekh for their help in obtaining and preparing the data analysed in this report. Gratitude is also extended to Amelia Lamont for her analysis of the qualitative reasons for discrepancy data. Finally, Dena Derkzen’s editorial feedback is much appreciated, as is Kelley Blanchette’s methodological and editorial input.

Executive Summary

Security classification can influence involvement in programs, access to privileges, and discretionary release. For this reason, it is essential that security classification decisions be made using methods that are transparent, consistent, and valid. In order to achieve these goals, the Correctional Service of Canada’s approach to security classification includes the use of evidence-based assessment instruments together with clinical appraisals conducted by experienced caseworkers. Recently, however, criticisms have been levied against the security reclassification process used with male offenders. The Office of the Correctional Investigator has asserted that the process is inappropriate for Aboriginal offenders as it results in their over-classification to higher levels of security, relative to non-Aboriginal offenders. As a result of these comments, the Correctional Service of Canada committed to completing an analysis of the efficacy of its security reclassification approach with Aboriginal offenders.

It is important to acknowledge that greater proportions of Aboriginal offenders than of non-Aboriginal offenders are housed at higher levels of security. However, this higher representation does not directly suggest that Aboriginal offenders are being over-classified. A substantial body of research demonstrates that Aboriginal offenders, by and large, exhibit more characteristics relating to risk than do non-Aboriginal offenders. It is possible, therefore, that the representation of Aboriginal offenders at higher levels of security is attributable to genuine differences in risk, rather than to over-classification. Over-classification would be demonstrated by Aboriginal offenders being housed at levels of security higher than are appropriate given their risk profile.

It is within this context that the current study was undertaken. The study had three inter-related objectives: (1) to revalidate the assessment instrument used as part of the security review (the Security Reclassification Scale); (2) to contrast security reclassification recommendations produced by the assessment instrument with final reclassification decisions which reflect both the instrument and caseworkers’ clinical appraisals; and, (3) to examine whether the assessment instrument is appropriate and effective in conducting security reviews for both Aboriginal and non-Aboriginal men. Findings from this last objective will also allow an examination of whether Aboriginal offenders are being over-classified by the current security reclassification approach.

Data for this study were drawn from the Correctional Service of Canada’s automated offender database system. They corresponded to a total of 32,328 security reclassification reviews conducted in the four year period from November 1, 2002, to October 31, 2006.

Study findings demonstrated that the Security Reclassification Scale continues to be appropriate for use in the security reclassification process. As expected, offenders with higher Security Reclassification Scale ratings were higher risk (as demonstrated by lower rates of discretionary release and higher risk ratings) and less well adjusted (as demonstrated by more frequent involvement in institutional misconducts, higher needs ratings, and lower motivation and reintegration potential ratings) than those with lower Security Reclassification Scale ratings. Recommendations produced by the scale were also better able to predict involvement in institutional misconduct than would be expected under the previous reclassification system, which involved only clinical appraisal. However, the results suggest that the predictive validity of the scale could be improved.

Analyses revealed a very low rate of inconsistency between security reclassification recommendations produced by the scale and final security classifications. Given that this rate of inconsistency was so low, it is not surprising that scale recommendations and final security classifications were similarly able to predict involvement in institutional misconducts, escapes, discretionary releases, and post-release outcome.

Finally, examinations of the security reclassification approach for Aboriginal and non-Aboriginal offenders demonstrated that very slightly higher proportions of Aboriginal offenders were recommended to, and actually placed at, higher levels of security. That being said, both Security Reclassification Scale recommendations and final security classification decisions were usually equally related to relevant constructs and to outcomes of interest for Aboriginal and non-Aboriginal offenders. When small differences did exist, these differences suggested that the scale was slightly more predictive for Aboriginal offenders than for their non-Aboriginal counterparts, despite the fact that rates of involvement in major misconducts were virtually identical for the two groups of offenders at each level of security.

In other words, this study presents no evidence to suggest that the current security reclassification approach results in Aboriginal offenders being over-classified within Canada’s penitentiaries. Instead, this pattern of findings is consistent with the considerable literature that demonstrates that Aboriginal offenders tend to exhibit more characteristics associated with risk than do non-Aboriginal offenders.

Table of Contents

List of Tables

List of Appendices

Introduction

Given that security classification can influence involvement in programs, access to privileges, and discretionary release (Motiuk, 1997), it is essential that security classification be conducted in a fashion that is transparent, consistent, and valid. For this reason, the Correctional Service of Canada’s (CSC) approach to both initial security classification and security reclassification includes the use of evidence-based assessment instruments together with clinical appraisals conducted by caseworkers with training and experience working with federally sentenced offenders.

This security classification and reclassification approach has been criticized because the same assessment instruments are used for both Aboriginal and non-Aboriginal inmates. It has been argued that this procedure is inappropriate for Aboriginal men (Office of the Correctional Investigator, 2006). Indeed, relative to non-Aboriginal offenders, larger proportions of Aboriginal offenders are classified to medium and maximum security (Public Safety Canada, 2007). The Office of the Correctional Investigator (2007) has interpreted these discrepancies as an indication that the tools used in conducting security classification reviews for Aboriginal offenders are inappropriate and that these should be replaced.

Partly as a result of these criticisms, a study was undertaken to assess the security reclassification process used with federally incarcerated male offenders. Reclassification was focused on exclusively because several studies have already examined the instrument used in initial security classification (e.g., Grant & Luciani, 1998; Luciani, Motiuk, & Nafekh, 1996). The goals of the study were three-fold: (1) to revalidate the assessment instrument used as part of the reclassification process; (2) to contrast security reclassification recommendations produced by the assessment instrument with final reclassification decisions, which include use of both the instrument and caseworkers’ clinical appraisals; and, (3) to examine whether the assessment instrument is appropriate and effective for reclassification decisions for both Aboriginal and non-Aboriginal men. This latter goal of the study is especially important because the Correctional Service of Canada committed to completing an analysis of the efficacy of its security reclassification approach (Correctional Service of Canada [CSC], 2006) with Aboriginal offenders.

Security Classification

In Canada, federally sentenced offenders are assigned security classifications of minimum, medium, or maximum security. These classifications determine the institutions in which offenders are eligible to be housed, and, as a result, influence the physical conditions of their incarceration (e.g., community living style housing versus cell housing), their access to programs and temporary release opportunities, and the likelihood of their being granted discretionary release. Security classification decisions are legislatively required to consider three criteria: (1) institutional adjustment; (2) escape risk; and, (3) risk to the public in the event of an escape (Corrections and Conditional Release Act [CCRA], 1992). Given the important consequences associated with security classification decisions, Canadian legislation also dictates that these decisions be made in the least restrictive fashion while maintaining protection of the public.

Offenders’ security classifications are determined initially as part of the routine admission process, but are re-assessed periodically throughout the duration of their sentence in order to reflect changes in behaviour or circumstance which could influence the three criteria mentioned above. The initial security classification involves the administration of the Custody Rating Scale (CRS; Solicitor General of Canada, 1987), an empirically-derived actuarial tool which assesses institutional adjustment and security risk. The instrument, which has demonstrated its validity and reliability in research studies (Grant & Luciani, 1998; Luciani, Motiuk, & Nafekh, 1996), provides a security placement recommendation according to the offender’s highest score in these two areas. In conjunction with the administration of this tool, a caseworker completes a comprehensive clinical appraisal of the offender’s case in order to determine whether any unique case factors, not captured by the instrument, require consideration in determining security placement. When the clinical appraisal and the CRS recommendation are inconsistent, the recommendation from the clinical appraisal is applied and an explanation for the inconsistency is provided.

Given that one of CSC’s primary goals is to assist offenders to become law-abiding citizens through the provision of rehabilitative programs and interventions, it is not surprising that offenders’ security classifications can and do shift during their periods of incarceration. As a result, security classification reviews are legislatively required to be conducted at least annually for most offenders (CCRA, 1992).1 Correctional Service Canada policies require that reviews be conducted even more frequently in cases where there is reason to believe that the appropriate security level for an offender has changed or where decisions pertaining to work release, discretionary release, and institutional transfers are to be made (CSC, 2007b). Frequent reviews are necessary to reflect any changes in behaviour and to facilitate both the prompt cascading of offenders to lower levels of security, and, ultimately, their reintegration into the community.

For male offenders, the security reclassification review is anchored by the Security Reclassification Scale (SRS; Luciani, Taylor, & Motiuk, 1998),2 a second research derived actuarial instrument. This instrument includes primarily dynamic indicators - that is, indicators that can change in accordance with offenders’ behaviour, such as involvement in disciplinary incidents in the last year. As such, the instrument’s recommendations reflect current behaviour and circumstance. As is the case during initial classification, the SRS provides a recommendation that anchors the security reclassification decision, but it must be used together with a caseworker’s clinical appraisal to ensure that unique factors not captured by the scale are not missed. Therefore, security reclassification decisions ultimately reflect both the scale recommendation and the caseworker’s clinical appraisal.

Security Classification for Aboriginal Offenders

At present, the process of security classification is similar for both Aboriginal and non-Aboriginal offenders. This being said, when completing security classification decisions and reviews, staff attend to the Gladue decision (R. v. Gladue, 1999) and take into consideration unique systemic or background factors that may have played a part in Aboriginal offenders’ lives, including a history of overt and covert discrimination, dislocation (such as residential school experience or family history of residential school experience), and/or participation in Aboriginal traditional teachings, ceremonies, and activities (CSC, 2007a; 2007b). This sensitivity is integral to formulating a clinical appraisal as part of each security review. In practice, then, though the application of security classification assessment instruments is no different for Aboriginal and non-Aboriginal offenders, the classification approach as a whole is different. Regardless, the Office of the Correctional Investigator (2006) has argued that existing security classification approaches lack cultural sensitivity and are inappropriate for use with Aboriginal offenders.

Over-Classification or Differences in Risk?

Currently, institutional population profiles demonstrate that relative to non-Aboriginal offenders, higher proportions of Aboriginal offenders are classified to medium security (68.6% vs. 64.6%) and to maximum security (15.8% vs. 13.3%), while lower proportions are classified to minimum security (15.6 vs. 22.1%; Public Safety Canada, 2007). The Office of the Correctional Investigator (2007) has attributed these differences to over-classification, and has recommended that current security classification practices be modified in order to diminish the proportions of Aboriginal offenders classified to higher levels of security. Within this context, it is important to distinguish between over-representation, which signifies that one group is classified at certain levels of security in greater proportions than another group, and over-classification, which suggests that one group of offenders is housed at a level of security than is higher than necessary given the offenders’ institutional adjustment, risk of escape, and risk to the public in the event of an escape.

Therefore, while not contesting that greater proportions of Aboriginal offenders than of non-Aboriginal offenders are housed at higher levels of security, it is important to consider alternative explanations for this pattern. First, it is known that Aboriginal offenders tend to be younger than their Aboriginal counterparts (Brzozowksi, Taylor-Butts, & Johnson, 2006; Trevethan et al., 2002) and to comprise a relatively larger proportion of homicide offences (Motiuk & Vuong, 2005). Both of these factors directly contribute to security classification, through inclusion in the scales used for initial security classification (Solicitor General Canada, 1987) and security reclassification (Luciani, Taylor, & Motiuk, 1998), as well as through application of the so-called two-year rule whereby offenders convicted of first- or second-degree murder serve the first two years of their sentence in maximum security institutions. In other words, Aboriginal and non-Aboriginal offenders’ differences on these two variables may be sufficient in and of themselves to explain their differential security placements.

Second, it is possible that differences in the security classifications of Aboriginal and non-Aboriginal offenders reflect variation in risk (see Rugge, 2006, for a comprehensive review of this argument). A substantial body of literature has demonstrated that, in comparison to their non-Aboriginal counterparts, larger proportions of Aboriginal offenders exhibit characteristics shown to be linked to risk. More specifically, evidence demonstrates that by and large, Aboriginal offenders tend to have more extensive criminal histories (Holsinger, Lowenkamp, & Latessa, 2003; Moore, 2003), to have more serious needs pertaining to substance abuse (Finn, Trevethan, Carriere, & Kowalski, 1999; Trevethan, Moore, & Rastin, 2002), to have greater needs regarding family and marital attachments (Statistics Canada, 2001; Trevethan et al., 2002), and to have greater needs regarding criminal associates (Brzozowksi et al., 2006). Aboriginal offenders also tend to be younger (Brzozowksi et al., 2006; Trevethan et al., 2002), have lower levels of education (Brzozowksi et al., 2006; Moore, 2003; Moore, Trevethan, & Conley, 2004), and have less extensive employment histories and fewer employability skills (Finn et al., 1999; Moore, 2003).3 A large body of literature (e.g., Andrews & Bonta, 2006; Gendreau, Little, & Goggin, 1996) has demonstrated that each of these characteristics is associated with risk of re-offending.

Of course, this explanation is only appropriate if one assumes that the risk factors discussed above are equally relevant for Aboriginal and non-Aboriginal offenders. Since many of the conclusions regarding risk factors have been drawn from research conducted primarily with non-Aboriginal offenders, it cannot be assumed that these factors are linked to risk for Aboriginal offenders also. With this issue in mind, a number of researchers have examined differences in the relation between traditional risk factors and re-offending across ethnicity. Both primary studies and reviews show that the factors most strongly linked to risk of re-offending tend to be the same for Aboriginal and non-Aboriginal offenders (e.g., Bonta, LaPrairie, & Wallace-Capretta, 1997; Bonta, Lipinski, & Martin, 1992; Rugge, 2006). After reaching the same conclusion, one researcher posited that “before arguing for the necessity of modifying our underlying theoretical and methodological approach [to conceptualizing and measuring risk] to take into account Native American cultures and experiences, it is necessary to stress that what we have learned about risk … through studying the non-Indian population is likely very important for the Native population” (Pridemore, 2004, p. 53). It should be acknowledged, however, that these findings refer specifically to risk of re-offending, rather than risk in the three areas considered in making security classification decisions. While it is intuitively logical that this pattern would exist in the latter area also, empirical evidence specific to male offenders is lacking. Research conducted with women offenders, however, has found that the same factors were associated with outcomes relevant to security classification (e.g., involvement in minor and major institutional incidents) for both Aboriginal and non-Aboriginal women (Blanchette, 2005).

With this rationale in mind, the present study was conceived partly as a response to the Office of the Correctional Investigator’s claims that Aboriginal offenders’ greater representation at higher levels of security is “due to a large extent to the culturally insensitive security classification tools still used by CSC, despite being denounced as unfair and ineffective” (2006, p. 14). It is possible, instead, as argued by Rugge, (2006) that larger proportions of Aboriginal offenders than of non-Aboriginal offenders are classified to higher levels of security, not because they are Aboriginal, but because they exhibit more characteristics associated with risk. Within this context, whether security classification tools are effective or not with Aboriginal offenders can only be determined through empirical investigation.

Current Study

There are three inter-related objectives to this study. The first is to revalidate the SRS in order to examine the extent to which it continues to be effective. Periodic revalidation is an important component in the use of any actuarial instrument (Austin, 2003). The revalidation process involves examining the SRS’s reliability, the extent that SRS recommendations converge with evidence from other sources, and the extent that SRS recommendations actually predict relevant outcomes, such as involvement in institutional incidents and escapes.

The second goal involves expanding on the differences between SRS recommendations and final security classification decisions. Areas of interest relating to this goal are the reasons for discrepancy between SRS recommendations and final security classification decisions and whether SRS recommendations are more or less predictive of relevant institutional outcomes relative to final security classification decisions.

The study’s last goal is to further examine the applicability of the SRS to Aboriginal offenders. More specifically, the study will examine the extent to which SRS recommendations and actual security placements differ for each group, as well as the extent to which SRS recommendations and actual security placements converge with evidence from other sources and are predictive of relevant outcomes for both Aboriginal and non-Aboriginal offenders. This component of the study will contribute to answering the question of whether differences in security classification according to ethnicity represent over-classification or are reflective of true variation in risk.

Method

Sample

The data analysed in this report correspond to 32,328 security reclassification reviews conducted with male federally incarcerated offenders between November 1, 2002 and October 31, 2006. As many offenders were involved in more than one reclassification review during this period, the sample of reclassification events comprises 17,307 unique offenders. Given the research questions used in this project, data were analyzed separately for Aboriginal and non-Aboriginal offenders. Aboriginal ethnicity was determined according to self-reported data captured in the Offender Management System (OMS), the Correctional Service of Canada’s automated offender data system. Any offender who self-identified as First Nations (i.e., North American Indian), Métis, or Inuit was classified as Aboriginal.4

Data

For each security reclassification review, additional data were drawn from OMS regarding the offender’s demographic characteristics; his offence and sentence information, and his risk, need, motivation, and reintegration potential ratings at the most recent assessment period occurring prior to the security reclassification review. Data were also collected regarding the results of the application of the SRS and the final security classification decision, as well as the offender’s involvement in institutional misconducts, escapes and attempted escapes, releases, and post-release outcome.

The Security Reclassification Scale (SRS)

The SRS (Luciani et al., 1998) is an objective, transparent, actuarial reclassification instrument implemented in 1998. It includes 14 items, many of which are dynamic, which aim to assess institutional adjustment, escape risk, and risk to the public in the event of an escape. Items were selected for inclusion based on their relationships with these three constructs in a large sample of male offenders.5  The 14 variables identified for inclusion in the SRS were:

  1. Serious disciplinary offences (institutional disciplinary offences that resulted in conviction for a serious offence as defined by the court) within review period;
  2. Minor disciplinary offences (institutional disciplinary offences that resulted in conviction for a minor offence as defined by the court) within review period;
  3. Number of recorded incidents within review period;
  4. Pay grade as of date of review;
  5. Number of segregation periods during review period;
  6. Detention referral (actual or anticipated);
  7. Correctional plan progress during review period;
  8. Correctional plan motivation during review period;
  9. Drug and alcohol rating (current);
  10. Successful temporary absences / work releases during review period;
  11. Age at review;
  12. Psychological concerns (current);
  13. CRS escape history; and,
  14. CRS incident history.

Using actuarial methods, responses to each item are assigned a value. Values are then summed to produce a scale total, which is assessed against cutoff values to determine the security classification recommendation produced by the SRS. Higher scale totals correspond to higher security classification recommendations. Notably, a discretionary range exists at each cutoff, allowing flexibility in selecting security classification recommendations for those offenders whose scores fall near the threshold values.

In this study, various data drawn from the SRS were used, including item scores, scale totals, scale recommendations, and application (or not) of the discretionary range.

Actual Security Classification

Though the SRS recommendation is intended to anchor security reclassification decisions, caseworkers also conduct a clinical appraisal. As a result of this appraisal, they may ultimately make a recommendation that is either consistent with or different than the one produced by the SRS. The Institutional Head or Deputy Warden is then responsible for authorizing the final security reclassification decision, which may or may not be consistent with the caseworker’s recommendation (CSC, 2007a). As such, there are two stages within the reclassification process where inconsistency can occur. In this study, the actual security classification refers to the final security reclassification decision authorized by the Institutional Head or Deputy Warden.

Statistical Note

Throughout this study, statistics are primarily reported in descriptive terms rather than in terms of probability testing. This decision was made for two reasons. First, it was not appropriate to use inferential statistics (i.e., statistics that estimate the likelihood that a finding from the study sample represent the true value of the full population) because the full population of security reclassification reviews occurring during the study period was included. Second, the number of reviews and of individuals represented in the dataset being used in this study was so large that virtually any probability test would be statistically significant. In this context, it is important to differentiate between statistical significance and practical significance. The results in this study are interpreted predominantly in light of the latter.

Results

Sample

The total sample was composed of 32,328 reviews involving 17,307 federally incarcerated male offenders. Of these, 6,717 (representing 3,551 individuals) corresponded to Aboriginal offenders, while 25,611 reviews (13,756 individuals) corresponded to their non-Aboriginal counterparts. Aboriginal offenders were typically younger that their non-Aboriginal counterparts; the average age of offenders in this sample was 33.5 years (SD = 9.6), while that of non-Aboriginal men was 36.3 years (SD = 10.8). As shown in Table 1, most offenders in each group were widowed, divorced, separated, or single.

Table 1:

Demographic Characteristics of the Sample
  Number (%) of Offenders
Demographic Variable Aboriginal Non-Aboriginal
Ethnicity        
North American Indian 2407 (68)    
Métis 998 (28)    
Inuit 146 (4)    
Caucasian     11,785 (86)
Black     1,104 (8)
Latin American     98 (1)
Asian / East Indian     501 (4)
Other / Unknown     268 (2)
Marital Status        
Single / Widowed / Divorced / Single 2196 (62) 8,596 (63)
Married / Common-law 1325 (37) 5,113 (37)
Unknown 30 (1) 47 (0)

Note. NAboriginal = 3,551. NNon-Aboriginal = 13,756. Proportions may not sum to 100 due to rounding.

Fourteen percent (n = 502) of the Aboriginal offenders and 15% (n = 2,120) of non-Aboriginal offenders were serving indeterminate sentences (e.g., life sentence, dangerous offender designation). Of the remainder, the average sentence length was of 5.1 years (SD = 4.3) for Aboriginal offenders and somewhat longer, 5.8 years (SD = 5.6), for their non-Aboriginal counterparts.

Offenders’ convictions on the present sentence were also examined (see Table 2). Most offenders (86% and 82% of the Aboriginal and non-Aboriginal groups respectively) were convicted of at least one violent offence, with assault and robbery being the most common. Just over two-thirds of offenders (67% and 73% of the Aboriginal and non-Aboriginal groups respectively) were convicted of at least one non-violent offence, with break and enter and other property offences particularly common. A relatively large proportion of non-Aboriginal offenders were also convicted of drug offences.

Table 2:

Offence Characteristics of the Sample
  Number (%) of Offenders
Present Conviction(s) Aboriginal Non-Aboriginal
Violent Offences        
Homicide (murder, manslaughter) 732 (21) 2,404 (17)
Attempt murder 71 (2) 521 (4)
Assault 1316 (37) 3,828 (28)
Robbery 971 (27) 4,415 (32)
Kidnapping / Forcible confinement 247 (7) 1,254 (9)
Sexual assault 631 (18) 1,759 (13)
Arson 135 (4) 1,026 (7)
Utter threats 397 (11) 1,692 (12)
Weapon offences 493 (14) 2,692 (20)
Non-violent Offences        
Drug offences 422 (12) 2,966 (22)
Break and enter 850 (24) 3,433 (25)
Fraud 21 (1) 195 (1)
Other property offences 630 (18) 3,263 (24)
Obstruct justice 390 (11) 2,315 (17)
Other non-violent 1961 (55) 8,210 (60)

Note. NAboriginal = 3,551. NNon-Aboriginal = 13,756. Numbers sum to more than 3,551 and 13,756 respectively because many offenders were convicted of more than one offence.

SRS Descriptive Information

The mean SRS score was similar for Aboriginal offenders (M = 21.8, SD = 4.6) and non-Aboriginal offenders (M = 21.5, SD = 4.3). In both cases, this score corresponds to the middle range for medium security. Mean scores for individual items were comparable, though not identical, for Aboriginal and non-Aboriginal offenders (see Appendix A). Overall, in over two-thirds of cases, the SRS produced a recommendation of medium security (see Table 3). Notably, the scale produced recommendations to medium security for a slightly larger proportion of non-Aboriginal than Aboriginal offenders; conversely, it also produced recommendations to minimum and maximum security for slightly smaller proportions of non-Aboriginal offenders. These findings demonstrate that though mean scale scores were relatively alike for Aboriginal and non-Aboriginal offenders, the distribution of scores around the mean varied.

Table 3:

Percentage of SRS Recommendations to Each Security Level
  Number (%) of Offenders
Scale Recommendation Aboriginal Non-Aboriginal
Minimum 1,060 (16) 3,712 (14)
Medium 4,553 (68) 18,453 (72)
Maximum 1,104 (16) 3,446 (13)

Note. NAboriginal = 6,717. NNon-Aboriginal = 25,611. Proportions may not sum to 100 due to rounding.

Analyses were also conducted to examine SRS recommendations by pre-review security level. For most cases (65% for each of the Aboriginal and non-Aboriginal groups), the SRS resulted in a security recommendation that was the same as the offender’s pre-review security placement. A considerable proportion of SRS recommendations represented decreases in security (24% and 23% for Aboriginal and non-Aboriginal offenders respectively), while fewer represented increases in security (11% for each group).

The high concordance rate between pre-review security placement and SRS recommendations likely explains the differences found between the proportions of recommendations to each security level and the proportions of offenders known to be housed at each security level (that is, 20.8%, 65.4%, and 13.8% for minimum, medium, and maximum respectively; Public Safety Canada, 2007). Specifically, a somewhat lower proportion of offenders were recommended to minimum security in this study than are actually housed at minimum security. This discrepancy is likely due to the fact that offenders housed at minimum security only undergo security reclassifications when there is reason to believe that a change to their security level is needed. Thus, there are no periodic reviews conducted to confirm that minimum security is appropriate for these offenders, and therefore no corresponding reclassification recommendations.

SRS Recommendations and Offender Characteristics

A series of analyses was conducted to examine whether there were differences in SRS recommendations according to offenders’ age, region of incarceration, sentence length, and offence type. Given that age is known to be negatively associated with risk (Gendreau, Little, & Goggin, 1996), it was not surprising that offenders with higher SRS recommendations were younger (see Table 4) than those with lower SRS recommendations. This trend was apparent both overall and for Aboriginal and non-Aboriginal offenders separately.

Table 4:

Average Age by SRS Security Level Recommendation
  Mean (SD) Age
Scale Recommendation Aboriginal Non-Aboriginal Total
Minimum 37.0 (10.1) 39.7 (11.1) 39.1 (11.0)
Medium 35.2 (9.9) 38.2 (11.1) 37.6 (10.9)
Maximum 30.4 (8.1) 32.4 (9.0) 31.9 (8.8)

Note. NAboriginal = 6,717. NNon-Aboriginal = 25,611. NTotal = 32,328.

Table 5 presents the proportions of Aboriginal and non-Aboriginal offenders from each region for whom the SRS produced a recommendation of minimum, medium, and maximum security. Clear variation was apparent in the number of reviews completed per region, with the distribution of reviews being consistent with that of the offender population. As most Aboriginal offenders are housed in the Prairies region (Public Safety Canada, 2007), it was not surprising that most scales completed with Aboriginal offenders were completed in this region.

In contrasting recommendations across regions, it was apparent that reviews completed in Quebec were relatively likely to result in a recommendation to maximum security, most markedly for Aboriginal offenders. Reviews completed in the Prairies region were more likely than those completed in other regions to produce recommendations to minimum security for both Aboriginal and non-Aboriginal offenders. This pattern was consistent with the regional distribution of institutions of various security levels (CSC, 2008).

Table 5:

Percentage of SRS Recommendations to Each Security Level, by Region
  Region
Scale Recommendation Atlantic Quebec Ontario Prairies Pacific
Aboriginal Offenders          
n (248) (658) (909) (3,646) (1,256)
Minimum 11 6 8 21 12
Medium 72 69 76 64 72
Maximum 17 25 16 15 16
Non-Aboriginal Offenders          
n (2,796) (7,193) (7,538) (4,115) (3,969)
Minimum 14 10 15 23 15
Medium 71 73 74 66 72
Maximum 15 17 11 11 13

Note. NAboriginal = 6,717. NNon-Aboriginal = 25,611. Proportions may not sum to 100 due to rounding.

Scale recommendations were then examined for differences based on offenders’ sentence lengths. Differences across sentence lengths were small and unsystematic, but it generally appears that slightly higher proportions of offenders serving shorter sentences were recommended to minimum security. This was especially notable among Aboriginal offenders. Further, relatively few offenders serving indeterminate sentences, especially those of non-Aboriginal ethnicity, were recommended to maximum security.

Table 6:

Percentage of SRS Recommendations to Each Security Level, by Sentence Length
  Sentence Category
Scale Recommendation Up to 3 Yrs. 3 - 8 Yrs. 8+Yrs. Indeterminate
Aboriginal Offenders        
n (1,626) (2,484) (1,243) (1,364)
Minimum 20 17 12 12
Medium 65 66 69 73
Maximum 15 17 19 15
Non-Aboriginal Offenders        
n (5,330) (9,169) (5,820) (5,292)
Minimum 14 16 13 14
Medium 71 70 73 75
Maximum 15 14 14 11

Note. NAboriginal = 6,717. NNon-Aboriginal = 25,611. Proportions may not sum to 100 due to rounding.

Finally, differences in scale recommendations according to offence type were also examined; these are reported in Table 7. First, reviews corresponding to offenders who were and who were not convicted of violent offences were contrasted. Second, reviews corresponding to homicide offences were contrasted with those corresponding to non-homicide offences. For both Aboriginal offenders and their non-Aboriginal counterparts, reviews conducted with those convicted of a violent offence were less likely than those conducted with those not convicted of a violent offence to result in a recommendation to minimum security. Similarly, they were also more likely to result in a recommendation to maximum security. Differences were much less marked when examining reviews corresponding to homicide and non-homicide offences, though slightly higher proportions of reviews conducted with offenders convicted of non-homicide offences resulted in recommendations to maximum security.

Table 7:

Percentage of SRS Recommendations to Each Security Level, by Offence Type
  Offence Category
Scale Recommendation Violent Non-Violent Homicide Non-Homicide
Aboriginal Offenders        
n (6,027) (690) (1,776) (4,941)
Minimum 15 23 15 16
Medium 68 66 70 67
Maximum 17 11 15 17
Non-Aboriginal Offenders        
n (21,874) (3,737) (5,519) (20,092)
Minimum 14 19 16 14
Medium 72 71 73 72
Maximum 14 10 11 14

Note. NAboriginal = 6,717. NNon-Aboriginal = 25,611. Proportions may not sum to 100 due to rounding.

Reliability

A series of analyses were conducted to examine the reliability of the scale. These indices of reliability all measure the extent to which the items of the SRS are consistent or correlated with one another. However, there is not necessarily any reason for the items to be consistent (i.e., for the items to be measuring the same thing), as the SRS was not constructed to reflect a single theoretical construct. Instead, it was developed empirically based on the ability of items to predict three factors relevant to security classification - institutional adjustment, risk of escape, and risk to the public in the event of an escape.

The first of the reliability analyses was the calculation of standardized item-total correlations for each scale item (see Table 8). Overall, the mean correlation between standardized individual scale items scores and the sum of these scores (i.e., total score) was of r = .42. Considering cases corresponding to Aboriginal and non-Aboriginal offenders, the mean standardized item-total correlations were very similar (r = .44 and r = .42 respectively). Notably, however, there were four items for which the magnitude of the item-total correlation differed importantly for the two groups of offenders: number of recorded incidents, detention referral, drug and alcohol rating, and psychological concerns. For each of these items except drug and alcohol rating, the standardized item-total correlation was of greater magnitude for Aboriginal than for non-Aboriginal offenders. In other words, the relationship between the individual items and the total scale score was greater for Aboriginal offenders for these scale items.

Table 8:

SRS Component Items' Standardized Item-Total Correlations
  Item-Total Correlations
Scale Item Aboriginal Non-Aboriginal Total
1. Serious disciplinary offences .55 .55 .55
2. Minor disciplinary offences .47 .48 .47
3. Number of recorded incidents .62 .57 .58
4. Pay grade .48 .51 .51
5. Number of segregation periods .62 .60 .60
6. Detention referral  .23 .11 .14
7. Correctional plan progress .55 .53 .53
8. Correctional plan motivation  .63 .60 .60
9. Drug and alcohol rating .32 .38 .37
10. Successful temporary absences / work releases .24 .22 .22
11. Age at review .36 .34 .35
12. Psychological concerns .32 .25 .27
13. CRS escape history  .27 .25 .26
14. CRS incident history .43 .44 .44

Note. NAboriginal = 6,717. NNon-Aboriginal = 25,611. NTotal = 32,328.

Next, inter-item correlations were calculated (see Appendix B). For space reasons, these are not reported individually for Aboriginal and non-Aboriginal offenders. The mean magnitude of the inter-item correlations was r = .13. It was notable that item 10 (successful temporary absences / work releases) correlated only weakly with the other scale items, as also demonstrated by its weak item-total correlation.

Finally, Cronbach’s alpha, a measure of internal consistency, was calculated. The value obtained for the full sample was r = .64, while values derived from the cases corresponding to Aboriginal and non-Aboriginal cases were r = .67 and r = .64 respectively. These values are slightly below the commonly accepted standard in social science of .70. Follow-up analyses without item 10 (successful temporary absences / work releases) suggested that its removal would not substantially increase internal consistency, as a value of r = .66 was obtained for the full sample (and of r = .68 and r = .65 for the Aboriginal and non-Aboriginal samples respectively) without this item.

Altogether, these analyses demonstrate that the SRS is generally homogeneous, but that there is room for improvement. As mentioned, however, given that homogeneity was not a goal in the construction of the SRS, these results do not suggest that the scale is inappropriate for use.

Discretion and Inconsistencies

The SRS includes a discretionary range on either side of the cutoff values corresponding to different security levels, such that, where appropriate, caseworkers can decide to “bump” those cases on the threshold between levels. Approximately one-in-five reviews fell within this discretionary range, with no differences in this proportion between cases corresponding to Aboriginal offenders (n = 1,432; 21%) and non-Aboriginal offenders (n = 5,209; 20%). Of these, the discretionary range was applied in only a fifth of cases (Aboriginal offenders: 19%; non-Aboriginal offenders: 22%). All in all, then, the discretionary range was invoked in only 4% of the total number of reviews in this study. These were roughly evenly divided between applications that resulted in lower and higher levels of security being recommended.

Caseworkers also have the option of recommending a level of security different than that recommended by the scale, regardless of whether the scale score falls within the discretionary range.6  Typically, they do so when there exists a unique case factor that increases or decreases risk and is not be captured by the scale. Table 9 provides an overview of the cases where the SRS recommendation was and was not consistent with the final recommendation made by the caseworker. It should be noted that in this and all subsequent analysis, the discretionary range was applied to the SRS recommendations (i.e., if an offender’s score corresponded to a recommendation of maximum security but fell within the discretionary range for medium, and the discretionary range was applied, the scale recommendation was changed from ‘maximum’ to ‘medium’). This procedure was used in order to recognize the integrated flexibility of the SRS.

The proportion of cases in which the scale recommendation and the caseworker’s final recommendation differed was quite small: 7% for Aboriginal offenders and 9% for non-Aboriginal offenders. Of these cases, however, it was more common for the caseworker to recommend a higher level of security (in 63% and 69% of cases for Aboriginal and non-Aboriginal offenders respectively) than a lower level of security (37% and 31% of cases respectively).

Table 9:

SRS and Caseworker Recommendations
  Caseworker Recommendation
  Min. Med. Max Total
SRS Recommendation n % n % n % n %
Aboriginal Offenders                
Minimum 984 (15) 85 (1) - 1,069 (16)
Medium 116 (2) 4,207 (63) 219 (3) 4,542 (68)
Maximum - 65 (1) 1,041 (16) 1,106 (16)
Total 1,100 (16) 4,357 (65) 1,260 (19) 6,717
Non-Aboriginal Offenders                
Minimum 3,173 (12) 470 (2) 2 (0) 3,645 (14)
Medium 537 (2) 16,852 (66) 1,118 (4) 18,507 (72)
Maximum - 194 (1) 3,265 (13) 3,459 (14)
Total 3,710 (15) 17,516 (68) 4,385 (17) 25,611

Note. Concordant cases appear on the diagonal. Proportions may not sum to 100 due to rounding.

Next, differences between caseworker recommendations and final post-review security classifications (that is, final security classifications approved by the Institutional Head or Deputy Warden) were examined. As Table 10 shows, caseworker recommendations and final post-review classifications differed in only 7% and 6% of cases corresponding to Aboriginal and non-Aboriginal offenders respectively. Again, the largest proportion of inconsistent cases were ones wherein the post-review classification was higher than the caseworker recommendation (63% and 62% respectively).

Table 10:

Caseworker Recommendations and Post-Review Security Classifications
  Post-Review Classification
Caseworker Recommendation Min. Med. Max. Total
n % n % n % n %
Aboriginal Offenders                
Minimum 940 (14) 156 (2) 4 (0) 1,100 (16)
Medium 88 (1) 4,154 (62) 115 (2) 4,357 (65)
Maximum - 76 (1) 1,184 (18) 1,260 (19)
Total 1,028 (15) 4,386 (65) 1,303 (19) 6,717
Non-Aboriginal Offenders                
Minimum 3,113 (12) 592 (2) 5 (0) 3,710 (14)
Medium 281 (1) 16,896 (66) 339 (1) 17,516 (68)
Maximum - 290 (1) 4,095 (16) 4,385 (17)
Total 3,394 (13) 17,778 (69) 4,439 (17) 25,611

Note. Concordant cases appear on the diagonal. Proportions may not sum to 100 due to rounding.

Finally, differences between SRS recommendations and final post-review security classifications were examined (see Table 11). Of the analyses conducted, this contrast revealed the largest rate of inconsistency: 12% for both Aboriginal and non-Aboriginal offenders. Similar to the above, in most cases where differences existed, they corresponded to a post-review classification that was higher than the SRS recommendations (in 65% and 69% of cases for Aboriginal and non-Aboriginal offenders respectively). Together, the results of these three analyses demonstrate that relatively low rates of inconsistency exist at each level of the security review process, and that differences at the two stages (i.e., between the SRS and caseworker recommendations and between caseworker recommendations and final placements) have an additive effect. In the end, about one in eight reviews result in placements at a different level of security than recommended by the SRS.

Table 11:

SRS Recommendations and Post-Review Security Classifications
  Post-Review Classification
  Min. Med. Max. Total
SRS Recommendation n % n % n % n %
Aboriginal Offenders                
Minimum 858 (13) 209 (3) 2 (0) 1,069 (16)
Medium 170 (3) 4,068 (61) 304 (5) 4,542 (68)
Maximum - 109 (2) 997 (15) 1,106 (16)
Total 1,028 (15) 4,386 (65) 1,303 (19) 6,717
Non-Aboriginal Offenders                
Minimum 2,753 (11) 889 (3) 3 (0) 3,645 (14)
Medium 641 (3) 16,554 (65) 1,312 (5) 18,507 (72)
Maximum - 335 (1) 3,124 (12) 3,459 (14)
Total 3,394 (13) 17,778 (69) 4,439 (17) 25,611

Note. Concordant cases appear on the diagonal. Proportions may not sum to 100 due to rounding.

Reasons for Inconsistencies

Only the first of the three contrasts reported above - that is, between SRS and caseworker recommendations - lends itself easily to further exploration. As previously mentioned, caseworkers must document the reasons for their choice to recommend a security classification other than that resulting from the application of the SRS. In these cases, they can select from a list of possible reasons or choose the other option and input an explanation themselves. Logically, given the alternate categories, when the inconsistency resulted in the caseworker recommending a lower security classification, other was the only reason ever cited. When the inconsistency resulted in the caseworker recommending a higher security classification, being seriously involved in disruptive activity, committing an assault, and being seriously involved in the distribution of contraband were all relatively common (see Table 12).

Table 12:

Reasons Cited for Inconsistency when Caseworker Recommendations Were Higher than SRS Recommendations
  Percentage of Cases Citing
Reason for Inconsistency Aboriginal Non-Aboriginal
Escape / attempt / conspire from any level of custody or escort 2 2
Escape / attempt with violence from any level of custody or escort - 0
Serious escape history or escape on current sentence 0 1
Identified as a major source in the distribution of contraband 8 11
Identified as a major force in disruptive activity 13 15
Instigator in a disruption leading to confrontation or damage 6 3
Assault causing serious physical harm to staff, visitor, or offender 9 5
Other 61 62

Note. NAboriginal = 463. NNon-Aboriginal = 2,321. Cells with “0” are the result of rounding; those with “-” indicate that no cases had this reason cited. Proportions may not sum to 100 due to rounding.

For those cases where the other option was selected as the reason to either increase or decrease the recommendation produced by the SRS, the explanations provided by caseworkers were examined. As Table 13 shows, the most common theme cited in these explanations was current behaviour / attitude. Most frequently, this reflected poor institutional behaviour, but in about one case out of eight, the reason cited was positive or stable institutional behaviour. Other relatively recurrent explanations cited for suggesting a security level higher than that recommended by the SRS were risk to public safety (with no additional information provided), and escape risk. On the other hand, another common explanation for suggesting a lower security level was that the inmate was interested in and responsive to treatment.

Table 13:

Reasons for Inconsistency between SRS and Caseworker Security Recommendations
  Explanations Citing
Theme n %a
Current behaviour / attitude 111 56.6%
Positive or stable institutional behaviour 27 13.8%
Poor institutional behaviour 61 31.1%
More positive evaluation on three risk criteria (institutional adjustment, risk of escape, risk to the public in the event of an escape) than suggested by SRS 19 9.7%
Positive attitude/ high motivation 9 4.6%
Poor attitude/ low motivation 7 3.6%
Progress on programming / correctional plan / dynamic factors 33 16.8%
Interested in and responsive to programming 21 10.7%
Insufficient / no programming completed 8 4.1%
Failure to address dynamic needs 5 2.6%
Behavioural / attitudinal history 14 7.1%
Poor behaviour prior to incarceration (e.g., index offence) 8 4.1%
History of poor institutional behaviour 5 2.6%
History of positive institutional behaviour 3 1.5%
Inmate needs 37 18.9%
Inmate requires the greater structure of a higher security institution 17 8.7%
Inmate requires the structure or opportunities of alternate institution 10 5.1%
Inmate requires more support / assistance 7 3.6%
Inmate requires psychological assessment prior to transfer 5 2.6%
Other 55 28.1%
Risk to public safety (no other information) 27 13.8%
Escape risk (no other information) 17 8.7%
SRS ascribes too much weight to certain event(s) 12 6.1%
Other (e.g., maturity, emotional lability) 7 3.6%

Note.  Themes are non-exclusive.
a Percentages represent the proportion of cases including any given reason for inconsistency among cases for which the other option was endorsed as the reason for inconsistency.

Convergent Validity

Analyses were conducted to examine the association between SRS scores and scores on conceptually related measures. Such analyses represent a way of examining convergent validity, or the extent to which the scale’s recommendations are consistent with what other related measures suggest. In this case, SRS scores were examined in relation to offenders’ levels of risk, need, reintegration potential, and motivation, as assessed by caseworkers in their correctional plan progress reports. It was expected that SRS scores would be higher for offenders assessed as representing higher levels of risk and need and lower levels of reintegration potential and motivation. Indeed, this was the case both for the total sample and for cases corresponding to Aboriginal and non-Aboriginal offenders separately. Correlations between SRS total score and these related measures were also calculated. Correlations were in the expected directions and were of modest to moderate magnitude, as would be expected for related but not identical constructs (see Table 14).

Table 14:

Mean (SD) SRS Scores by Ratings on Related Measures
  Rating  
Measure Low Medium High r
Aboriginal Offenders        
Risk 19.8 (4.3) 20.4 (4.6) 22.2 (4.5) .17
Need 18.0 (3.9) 19.5 (4.4) 22.3 (4.5) .26
Reintegration Potential 23.0 (4.3) 19.8 (4.3) 18.2 (3.9) -.37
Motivation 24.3 (4.0) 21.4 (4.4) 18.3 (3.9) -.39
Non-Aboriginal Offenders        
Risk 19.6 (4.0) 20.8 (4.3) 22.0 (4.25) .16
Need 18.5 (3.6) 20.0 (4.0) 22.2 (4.2) .26
Reintegration Potential 23.2 (4.1) 20.5 (3.9) 19.0 (3.8) -.37
Motivation 23.8 (4.0) 21.2 (4.1) 18.6 (3.7) -.36

Note. NAboriginal = 5,360. NNon-Aboriginal = 20,448. Temporally relevant risk, need, reintegration potential, and motivation data (i.e., occurring in the time period between the current SRS review and any previous SRS review) were not available for all cases.

For non-Aboriginal offenders only,7 the association between SRS recommendations and scores on the Statistical Information on Recidivism - Revised 1 (SIR-R1) scale (Nuffield, 1982), was also investigated. This scale has a 57 point range (from -27 to 30), with higher scores suggesting a lower likelihood of recidivism. Again, results supported the convergent validity of the SRS, with cases corresponding to SRS recommendations to minimum security having the highest mean SIR-R1 scores (1.2; SD = 10.1), followed by those corresponding to SRS recommendations to medium security (-2.4; SD = 9.7) and finally to maximum security (-7.1; SD = 7.6). The correlation of r = -.28 also demonstrates that high SRS scores were moderately associated with low SIR-R1 scores (indicating a greater likelihood of recidivism).

Predictive Validity

The next series of analyses were conducted to evaluate the extent to which SRS recommendations were predictive of relevant outcomes. Of particular interest was the extent to which involvement in institutional misconduct was associated with preceding SRS recommendations. Similar investigations were also conducted to examine the association of SRS recommendations with subsequent escapes, discretionary release decisions, and post-release outcomes. For each of the outcomes, the association with actual post-review security classification was also examined. This allowed for a contrast of the predictive ability of SRS recommendations and actual post-review classifications.

Institutional Misconducts

Involvement in institutional misconducts was examined first. Each file was examined for any involvement in institutional misbehaviour in the six months immediately following a security reclassification review. Any cases for which the full six-month period was not available were omitted from these analyses, leaving 5,940 cases corresponding to Aboriginal offenders and 22,548 cases corresponding to non-Aboriginal offenders. Misconducts were coded according to their severity; analyses were conducted separately for major misconducts (i.e., homicide, assault, sexual assault, fighting, threatening behaviour, hostage-taking, inciting to riot or strike, possession of drugs, possession of weapons, escape, or attempting escape) and minor misconducts (i.e., all other misconducts). Rates of involvement in minor misconducts were slightly higher for non-Aboriginal offenders than for Aboriginal offenders (13% and 10% respectively), while 6% of each group was involved in a major misconducts.

As shown in Table 15, the rate of involvement in misconducts in the six month period after security review increased linearly with security classification both for Aboriginal and non-Aboriginal offenders. This relationship was apparent regardless of whether SRS recommendations or actual classifications were considered. Both SRS recommendations and actual classifications were able to discriminate among cases according to their future involvement in criminal activity

Table 15:

Involvement in Misconduct by Security Classification
  Cases Involved in Misconduct (%)
  Aboriginal offenders Non-Aboriginal Offenders
Security Classification Minor Major Minor Major
SRS Recommendation        
Minimum 7 2 9 2
Medium 10 6 12 6
Maximum 16 13 20 14
Actual Classification        
Minimum 7 2 9 2
Medium 10 6 12 5
Maximum 14 12 18 13

Note. NAboriginal = 5,940. NNon-Aboriginal = 22,548.

In order to further investigate the ability of security classifications to predict institutional misconduct, a series of receiver operating characteristic (ROC) curves were calculated. ROC curves are used to provide an overall estimate of the predictive ability of a measure, and the underlying framework includes consideration of both successful prediction and unsuccessful prediction. One of the most commonly used metrics derived from ROC curves is the area under the curve (AUC), which provides an estimate that the measure will assign a higher score to a case that will display the outcome of interest than a case that will not. In this context, then, the AUC is a measure of the likelihood that an offender who will be involved in misconduct will have a higher security classification than his counterpart who will not be involved in misconduct.

Table 16 shows that the AUC values obtained for each type of misconduct based on SRS recommendations and actual security classifications were virtually identical. Notably, however, AUC values were higher for major misconducts than for minor misconducts, both for Aboriginal, zSRS = 2.81, p < .01; zClassification = 3.33, p < .001, and non-Aboriginal offenders, zSRS = 6.00, p < .001; zClassification = 6.90, p < .001. This is logical given that major misconducts are more indicative of problematic institutional adjustment, one of the areas the SRS scale aims to predict. However, there was room for improvement in predictive ability for both groups of offenders and for both types of misconduct, as the AUC values, even for major misconducts, fell between the commonly established thresholds for “acceptable” and “good” prediction in social science (.60 and .70 respectively).

Table 16:

Results of ROC Analyses of the Predictive Ability of Security Classification on Institutional Misconduct
  ROC Analysis Results
  Aboriginal  Offenders Non-Aboriginal Offenders
Model AUC 95% C.I. AUC 95% C.I.
Minor Institutional Misconduct        
SRS Recommendation .57 .54 - .59 .56 .55 - .57
Actual Classification .56 .54 - .59 .56 .55 - .57
Major Institutional Misconduct        
SRS Recommendation .62 .59 - .65 .62 .61 - .64
Actual Classification .63 .60 - .65 .63 .61 - .64

Note. NAboriginal = 5,940. NNon-Aboriginal = 22,548.

Another way of examining predictive validity is to focus not only on whether an outcome did or did not occur, but also on how soon the event occurred. In this case, this involves considering the number of days elapsed prior to involvement in institutional misconduct. One technique used for such analyses, survival analysis, statistically accounts for different follow-up periods, and can therefore accommodate all offenders, regardless of the length of time between their security review and their subsequent review, release, or study end-date. For this reason, these analyses used the full sample of reviews, rather than only those with a six-month follow-up, as above. Again, results were very similar for Aboriginal and non-Aboriginal offenders, and demonstrated that those classified and placed to maximum security engaged in more misconducts than those classified and placed to minimum security, and took less time to do so. Given the similarity of results, these analyses are not reported in full.

Escapes

Though predicting escapes is an important goal of the SRS, these incidents occur quite infrequently. For this reason, it was not possible to conduct analyses for escapes using a fixed follow-up. Even when considering the full follow-up time available, only 179 cases (0.6%) corresponded to a subsequent escape-related behaviour (i.e., escape from custody, escape from escort, failure to return from a temporary absence, and attempted escape). Of these, approximately equal proportions were of Aboriginal (0.7%) and non-Aboriginal offenders (0.5%). Table 17 shows the proportions of cases at each level of security involved in escape-related behaviours. Notably, proportions were highest at both minimum and maximum security. There are likely two explanations for this. The first is that those housed at maximum security may be the ones most likely to attempt to escape (not surprising, given that security classification decisions are meant, in part, to reflect risk of escape). Second, those housed at minimum security may leave the facility (a behaviour labeled an escape) more frequently simply because it is much easier for them to do so. Typically, minimum security institutions have no perimeter fence and less staff supervision, relative to maximum security.

Table 17:

Escapes by Security Classification
  Cases Involved in Escape Incidents (%)
Security Classification Aboriginal offenders Non-Aboriginal Offenders
SRS Recommendation    
Minimum 1.7 0.9
Medium 0.3 0.2
Maximum 1.5 1.6
Actual Classification    
Minimum 1.5 1.4
Medium 0.3 0.2
Maximum 1.8 1.2

Note. NAboriginal = 6,717. NNon-Aboriginal = 25,611. Follow-up period in which an escape can occur varies.

ROC curves were the only method used to further analyse these cases given the very small proportion of cases involved in escapes. Unlike correlation-based statistical methods (including survival analysis), ROC analyses are relatively insensitive to base rates. They do not, however, account for differing follow-up periods, and the conclusions from these analyses must therefore be interpreted with caution. Nonetheless, initial review of the results of ROC analyses demonstrates that escape incidents can be predicted roughly as well as can incidents of other types and that prediction derived from SRS recommendations and from actual security classifications are approximately equally predictive (see Table 18). However, AUC values obtained in predicting escapes had very broad confidence intervals, including (in all cases but one) the value .50, and were generally statistically non-significant. This means that predictive confidence is poor, and is likely at least partially due to the fact that escapes were most common at both minimum and medium security.

Table 18:

Results of ROC Analyses of the Predictive Ability of Security Classification on Escape Incidents
  ROC Analysis Results
  Aboriginal  Offenders Non-Aboriginal Offenders
Model AUC 95% C.I. AUC 95% C.I.
SRS Recommendation .60 ns .49 - .70 .58 .52 - .65
Actual Classification .61 ns .50 - .72 .56 ns .49 - .62

Note. NAboriginal = 6,717. NNon-Aboriginal = 25,611. ns non-significant.

Discretionary Release

Analyses of release and on post-release outcome were conducted focusing on individuals rather than security reviews, and included only the first release occurring after a security classification review occurring within the study period. A total of 15,750 of the 17,307 individuals in this study (91%) had been released at least once during the study period; of these, 3,305 were Aboriginal and 12,445 were non-Aboriginal. Some were granted discretionary release (i.e., day or full parole) but most were released through non-discretionary channels (i.e., statutory release, warrant expiry). Since the granting of discretionary release suggests that the risk to the public is considered manageable, the proportions of offenders receiving this type of release were examined.

Table 19 shows these proportions by level of security. The rate of discretionary release decreased linearly with security level, both when considering SRS recommendations and when considering actual security classification.

Table 19:

Discretionary Release by Security Classification
  Cases Granted Discretionary Release (%)
Security Classification Aboriginal offenders Non-Aboriginal Offenders
SRS Recommendation    
Minimum 40 44
Medium 11 15
Maximum 2 2
Actual Classification    
Minimum 39 45
Medium 12 15
Maximum 2 3

Note. NAboriginal = 3,305. NNon-Aboriginal = 12,445. Follow-up period for discretionary releases varies.

Again, ROC analyses were used to examine the predictive association of security classification with type of release.8  Table 20 provides a summary of these analyses. Predictive accuracy was satisfactory when assessed against the usual cutoff of .70 for “good” prediction. This was true regardless of whether SRS recommendations or actual security classifications were considered. Moreover, predictive ability was again greater for Aboriginal than for non-Aboriginal offenders. This difference was statistically significant when considering SRS recommendations, z = 1.94, p < .05.

Table 20:

Results of ROC Analyses of the Predictive Ability of Security Classification on Type of Release (Discretionary, Non-Discretionary)
  ROC Analysis Results
  Aboriginal  Offenders Non-Aboriginal Offenders
Model AUC 95% C.I. AUC 95% C.I.
SRS Recommendation .72 .69 - .74 .69 .68 - .70
Actual Classification .71 .68 - .73 .69 .68 - .70

Note. NAboriginal = 3,305. NNon-Aboriginal = 12,445. 

Post-Release Outcome

Post-release outcome was the final area explored in relation to the predictive validity of the SRS. Again focusing solely on offenders’ first releases after a security review within the study period, a total of 3,837 (24%) of the 15,750 released offenders had their release revoked within one year of release. Of these, 920 were Aboriginal and 2,917 were non-Aboriginal. Table 21 presents the proportion of offenders who had their releases revoked by security level, both generally (i.e., due to breaches of conditions and to re-offending) and due solely to re-offending. For both Aboriginal and non-Aboriginal offenders, the proportions of offenders having their release revoked due to a new offence increased linearly with security classification, regardless of whether SRS recommendation or actual security classification was considered. However, this was not true of revocations generally. Security level seemed to be linearly associated with revocation for non-Aboriginal offenders, but no similar association was apparent for Aboriginal offenders.

Table 21:

Returns to Custody by Security Classification
  Cases Returned to Custody (%)
  Aboriginal Offenders Non-Aboriginal Offenders
Security Classification Any Revocation Revocation with Offence Any Revocation Revocation with Offence
SRS Recommendation        
Minimum 27 9 18 5
Medium 28 10 24 7
Maximum 30 13 30 9
Actual Classification        
Minimum 29 9 20 5
Medium 28 10 24 7
Maximum 29 12 28 9

Note. NAboriginal = 3,305. NNon-Aboriginal = 12,445.

Follow-up analyses were conducted using ROC analysis (see Table 22). These demonstrated that despite the apparent trends detected above, security classification was in fact only weakly predictive of post-release outcome. Specifically, ability to predict revocation (both generally and in respect to re-offence only) among Aboriginal offenders was virtually at chance levels. While AUC values were somewhat lower for Aboriginal offenders, the difference in predictive ability for Aboriginal and non-Aboriginal offenders was statistically significant only when considering SRS recommendations, zAny Revocation = 2.47, p < .01; zRe-Offence = 2.55, p < .01. Further, the rate of predictive success was so low as to render the SRS inappropriate for prediction of post-release outcome for either group.

Table 22:

Results of ROC Analyses of the Predictive Ability of Security Classification on Post-Release Outcome
  ROC Analysis Results
  Aboriginal  Offenders Non-Aboriginal Offenders
Model AUC 95% C.I. AUC 95% C.I.
Any Revocation        
SRS Recommendation .51ns .49 - .53 .54 .53 - .55
Actual Classification .50ns .48 - .52 .53 .52 - .54
Revocation with Offence        
SRS Recommendation .53ns .50 - .56 .54 .52 - .56
Actual Classification .53ns .49 - .56 .54 .52 - .56

Note. NAboriginal = 3,305. NNon-Aboriginal = 12,445. ns non-significant.

Discussion

The purpose of this study was to examine the security reclassification process used with federally incarcerated male offenders in Canada. More specifically, the study had three objectives: (1) to revalidate the SRS, the actuarial instrument used as part of the security review process; (2) to examine differences between recommendations derived from the SRS and final security classifications; and, (3) to investigate differences in the appropriateness and effectiveness of the security reclassification process for Aboriginal and non-Aboriginal offenders.

Revalidation of the SRS

Given that the SRS has been in use as part of the security reclassification procedure for federally incarcerated male offenders since 1998, a great deal of data was available for analysis.9  This allowed for a comprehensive revalidation, including calculation of indices of reliability, convergent validity, and predictive validity.

Reliability analyses involved calculating item-total correlations, inter-item correlations, and Cronbach’s alpha. Each of these indices is a measure of the extent to which the SRS items are consistent with one another. These were calculated for information purposes only, as the SRS was developed in such a way as to include items which are independently related to relevant outcomes, and there is therefore no reason to believe the items should be consistent with one another. Nonetheless, analyses revealed that the scale was moderately homogeneous. One item - successful temporary absences / work releases - seemed particularly weakly related to the others, but its exclusion from analyses did little to increase reliability. Altogether, while these indices of reliability did not quite meet conventional thresholds, they were satisfactory for a scale not expected to be homogeneous.

Arguably the most important indicators of the utility of a scale are its convergent and predictive validity. Results indicate that the SRS’s convergent validity - that is, the extent to which scores on the SRS are similar to those on related measures - was quite satisfactory. High SRS scores were associated with higher assessed levels of risk and need and lower assessed levels of motivation and reintegration potential. For non-Aboriginal offenders, higher SRS scores were also associated with scores demonstrating a higher likelihood of recidivism on the Statistical Information on Recidivism - Revised scale.10

Indices of predictive validity provided acceptable results. Analyses focused on the prediction of four types of events: involvement in institutional incidents, involvement in escape incidents, the granting of discretionary release, and post-release outcome. The most important of these is likely involvement in institutional incidents, as these are a relatively common institutional occurrence and hold the potential to be disruptive to both individuals (inmates and staff) and the institution as whole, in the event that a lock-down or similar outcome is necessary. While it could be argued that it would be as important for the SRS to be able to predict post-release outcome, the scale’s primary purpose is not related to recidivism, and other scales exist for this purpose.

In regard to prediction of involvement in institutional misconduct, the scale’s predictive power generally reached the level considered acceptable by social scientists. This was true both with respect to minor (e.g., possession of unauthorized items) and major incidents (e.g., homicide, assault), though predictive ability was better for major incidents. This is appropriate, as the SRS aims to measure institutional adjustment, and major incidents indicate more serious maladjustment than do minor incidents.

Predictive ability in regards to involvement in escape-related behaviours was less impressive. Specifically, the SRS was unable to predict this type of incident, likely due both to the very low frequency with which such events occurred in the study sample and to the fact that escape-related behaviours occurred most frequently for offenders at maximum security (escape attempts; likely due to risk) and at minimum security (so-called walk-aways; likely due to opportunity).

Prediction of discretionary release reached satisfactory levels, which is important as the granting of discretionary release represents that the National Parole Board feels that the risk presented by an offender is assumable in the community. This finding must be interpreted cautiously, however, because security placement also impacts the granting of discretionary release, with offenders housed at more restrictive levels less likely to be granted release (Motiuk, 1997). Given this, it is impossible to disentangle from the present results whether discretionary release decisions actually reflected risk, security level, or a combination of the two.

Finally prediction of post-release outcome was very poor. Prediction of conditional release revocation and of revocation with offence was very slightly above chance levels for non-Aboriginal offenders, but was still so low as to be of virtually no utility. This, however, is appropriate given that the scale was not developed to predict post-release outcome.

These results suggest that improvement to the SRS’s predictive ability is possible. Nonetheless, it is important to underscore the extent to which the present results represent an improvement over previous methods. Prior to the implementation of the SRS, security reclassification decisions were made solely using clinical appraisals of case information. However, a sizeable body of literature demonstrates that decisions made using solely clinical methods are virtually always inferior to those made using actuarial methods (e.g., Ægisdóttir et al., 2006; Grove & Meehl, 1996). Most importantly, clinical methods are more likely to be inaccurate - meta-analytic evidence demonstrates that the accuracy of clinical methods is equivalent to or lower than random chance (Ægisdóttir et al., 2006). Though the predictive accuracy of the SRS in regards to institutional incidents could still be improved, it was substantially better than chance odds.

Within this context, then, the predictive ability that was attained by the SRS was by no means negligible. When considering that over 3,500 men in this study were found to be involved in at least one minor misconduct within six months of their security review, the importance of even modest gains in predictive accuracy in this regard becomes more appreciable. Specifically, if even a fraction of such a number of incidents could be minimized or prevented, this would represent a significant operational gain. Thus results regarding prediction of institutional incidents, together with the very satisfactory convergent validity results, are supportive of the continued use of the SRS. Moreover, actuarial instruments such as the SRS have other advantages not studied here - they have been demonstrated to be more consistent, more transparent, and more defensible than approaches relying only on clinical appraisals (Austin, 1983; Grove, Zald, Lebow, Snitz, & Nelson, 2000; Zinger, 2004). These advantages only further reinforce the utility of the SRS.

Even so, future attention should be focused on the SRS’s inter-rater reliability, or the extent to which different individuals would assign the same scores to SRS items if faced with the same individual. Though only a portion of the SRS items are scored by the caseworker (with the rest of the items being automatically entered by the computerized system used, based on official file records), the regional variability in SRS recommendations in the present results underscored the importance of investigating this issue. While it was not surprising that the region with the most minimum security institutions (Prairies region; CSC, 2008) was the one with the most recommendations to minimum security, it is nonetheless impossible to be certain that the regional variability is fully attributable to this. It is possible that individuals are interpreting the SRS items differently, and that this contributes to variation in scoring.

SRS Recommendations and Final Security Classifications

The next goal of this study was to contrast security classification recommendations produced by the SRS and final security classifications actually approved and implemented by the Institutional Head or Deputy Warden. The rate of inconsistency between these was strikingly low. Indeed, only 12% of cases were inconsistent. This proportion falls comfortably below the limit of 20% considered acceptable by researchers in the area (Buchanan, Whitlow, & Austin, 1986). It is also markedly lower than the rates of 28% and 34% found in comparable analyses involving the Security Reclassification Scale for Women (Blanchette, 2005; Gobeil, 2007). As was the case with women, however, the majority of the discrepancies (two-thirds) resulted in final security placements at higher levels of security than recommended by the SRS. Common reasons for such discrepancies were involvement in disruptive activities or in the distribution of contraband or problematic program performance.

Given that the rate of inconsistency was so low, it is perhaps not surprising that SRS recommendations and final security classifications were similarly accurate in predicting relevant outcomes. Specifically, the ability of each of SRS recommendations and final security classifications to predict involvement in institutional misconducts, escapes, discretionary releases, and post-release outcome were statistically identical.

Certain researchers in the area of risk prediction have argued that inconsistencies where scale results and clinical appraisals are considered together and the clinical appraisal ultimately dictates final decision are associated with decreased predictive accuracy (Rice, 2007). The current results are therefore promising, as they did not show such decreases; instead, findings demonstrate that in this context, clinical appraisal is being appropriately combined with SRS recommendations.

Security Reclassification Reviews for Aboriginal Offenders

Though average SRS scores were virtually identical for Aboriginal and non-Aboriginal offenders, the distribution of SRS recommendations to levels of security differed, indicating that the distribution of scores around the mean was different for Aboriginal and non-Aboriginal offenders. Relative to their non-Aboriginal counterparts, a very slightly greater proportion of Aboriginal offenders were recommended to maximum security by the SRS (16% as compared to 14%). These proportions were paralleled by those for final security classifications (19% as compared to 17%).

This being said, findings demonstrated that both SRS recommendations and final security classification decisions were usually equally related to relevant constructs and to outcomes of interest for Aboriginal and non-Aboriginal offenders. In fact, where minor differences did exist, the pattern of differences suggested that the scale was more appropriate for Aboriginal offenders. In other words, this study presents no evidence to suggest that current security reclassification approaches result in Aboriginal offenders being over-classified within Canada’s federal corrections institutions. Instead, results suggest that the slightly higher rates of recommendations and placements to maximum security for Aboriginal offenders were supported by relationships with relevant constructs of risk and need and by associations with institutional behaviour. These results are consistent with the considerable evidence that demonstrates that Aboriginal offenders tend to exhibit more characteristics associated with risk than do non-Aboriginal offenders (e.g., Brzozowski et al., 2006; Finn et al., 1999; Moore, 2003; Statistics Canada, 2001; Trevethan et al., 2002), despite efforts to address the situation.

This conclusion is contrary to what has been suggested by the Office of the Correctional Investigator (2007). As the Office has itself stated, however, “to fully understand and appreciate why Aboriginal offenders suffer such a fate within the correctional system, one must look … beyond the confines of the institution” (2006, p. 14). In this quote, the Office of the Correctional Investigator was suggesting that components external to the institution comprised a piece of the puzzle to be addressed concurrently with changes to the security reclassification approach used by Correctional Service Canada; the present results, however, suggest that changes beyond the confines of the institution are a necessary first step. Indeed, the pattern of results in this study demonstrates that the current pattern of reclassification is reflective of appropriate management of risk presented by incarcerated offenders, Aboriginal or not.

As has been argued elsewhere, “recognizing the necessity of incorporating tribal culture and understanding of delinquency does not require throwing out what we have learned” (Pridemore, 2004, p. 54). The Correctional Service of Canada’s current security reclassification approach, which includes a scale based on variables found to be associated with institutional misconduct and risk, together with clinical appraisal of case-specific factors, including culturally relevant factors (CSC, 2007b), represents compliance with Pridemore’s suggestion.

Certain authors have argued that there may be factors associated with both risk (Holsinger, Lowenkamp, & Latessa, 2006; Sioui, Thibault, & Amiskou Groupe Conseil, 2001) and desistance (Pridemore, 2004) which are either unique to or different for Aboriginal offenders. Though the present results do not indicate a need to focus on these factors in the context of security reclassification, such a focus may nevertheless be of utility in the context of increasing understanding of the issue of Aboriginal offenders’ over-representation within the correctional system (Public Safety Canada, 2007). Since 2003, CSC has been focusing considerable resources on policy modifications and collaborations with other government agencies and with Aboriginal and non-Aboriginal community groups to close the gap between Aboriginal and non-Aboriginal offenders within the correctional system (see CSC Review Panel, 2007). Stakeholders involved in components of this challenge other than security classification may wish to conduct further research in the area of Aboriginal offenders’ unique risk and desistance factors.

Conclusion

Altogether, the results are supportive of continued use of the Security Reclassification Scale as an anchor in reaching security reclassification decisions for both Aboriginal and non-Aboriginal male offenders. Offenders with higher SRS recommendations were higher risk (as demonstrated by lower rates of granting of discretionary release and higher risk ratings) and less well adjusted (as demonstrated by more frequent involvement in institutional incidents, higher needs ratings, and lower motivation and reintegration potential). This result is particularly salient because there have been considerable changes to the offender population since the time of the SRS’s development (CSC, 2005).

Though very slight differences in the proportions of Aboriginal and non-Aboriginal offenders at each level of security existed, the current security reclassification approach produced recommendations as predictive of relevant outcomes for both groups. In fact, where small differences existed, the SRS and the security reclassification approach as a whole was more predictive for Aboriginal offenders than for non-Aboriginal offenders. In other words, the present study does not offer support for the contention that current reclassification approaches used with Aboriginal men result in their over-classification. This finding was especially important because the Correctional Service of Canada had committed to examining the efficacy of its reclassification approach with Aboriginal offenders, and revising this approach if results indicated that such a course were necessary (2006). Study results do not indicate that such a revision is currently needed.

The study presents results with all three major Aboriginal groups in Canada combined. Further research is needed to divide the results into the respective groups of First Nations people, Métis and Inuit. These further analyses would demonstrate if there are variable impacts across the groups.

References

Ægisdóttir, S., White, M.J., Spengler, P.M., Maugherman, A.S., Anderson, L.A., Cook, R.S., et al. (2006). The meta-analysis of clinical judgment project: Fifty-six years of accumulated research on clinical versus statistical prediction. The Counseling Psychologist, 34, 341-382.

Andrews, D.A. & Bonta, J. (2006). The psychology of criminal conduct (4th ed.). Cincinatti, OH: Anderson Publishing.

Austin, J. (1983). Assessing the new generation of prison classification methods. Crime and Delinquency, 28, 561-576.

Austin, J. (2003). Findings in prison classification and risk assessment. National Institute of Corrections Prisons Divisions - Issues in Brief. Washington, DC: U.S. Department of Justice, Federal Bureau of Prisons.

Blanchette, K. (2005). Field-test of a Gender-informed Reclassification Scale for Women Offenders. Unpublished doctoral dissertation, Carleton University, Ottawa, Ontario.

Bonta, J., LaPrairie, C., & Wallace-Capretta, S. (1997). Risk prediction and re-offending: Indigenous and non-Indigenous offenders. Canadian Journal of Criminology, 39, 127-144.

Bonta, J., Lipinski, S., & Martin, M. (1992). Native inmates: Institutional response, risk, and needs. Canadian Journal of Criminology, 31, 49-62.

Brzozowksi, J., Taylor-Butts, A., & Johnson, S. (2006). Victimization and offending among the Aboriginal population in Canada. Ottawa, ON: Canadian Centre for Justice Statistics, Statistics Canada.

Buchanan, R.A., Whitlow, K.L, & Austin, J. (1986). National evaluation of objective prison classification systems: The current state of the art. Crime and Delinquency, 32, 272-290.

Correctional Service of Canada. (2005). The changing federal offender population: Profiles and forecasts, 2005. Ottawa, ON: Author.

Correctional Service of Canada (2006). Response from the Correctional Service of Canada to the 33rd annual report of the Correctional Investigator 2005-2006. Ottawa, ON: Author.

Correctional Service of Canada. (2007a). Commissioner’s Directive 705-7: Security Classification and Penitentiary Placement. Ottawa, ON: Author.

Correctional Service of Canada. (2007b). Commissioner’s Directive 710-6: Review of Offender Security Classification. Ottawa, ON: Author.

Correctional Service of Canada (2008). National facility directory. Retrieved October 1, 2008 from http://www.csc-scc.gc.ca/text/region/nat-fac-dir-eng.shtml

Correctional Service of Canada, Performance Management. (2008). [Incarcerated and community Aboriginal offender population distribution over time.] Unpublished raw data.

Corrections and Conditional Release Act [CCRA]. (1992).

CSC Review Panel (2007). A Roadmap to Strengthening Public Safety: Report of the Correctional Service of Canada Review Panel. Ottawa, ON: Minister of Public Works and Government Services Canada.

Grove, W.M. & Meehl, P.E. (1996). Comparative efficiency of formal (mechanical, algorithmic) and informal (subjective, impressionistic) prediction procedures: The clinical / statistical controversy. Psychology, Public Policy and Law, 2, 293-323.

Grove, W.M., Zald, D.H., Lebow, B.S., Snitz, B.E., & Nelson, C. (2000). Clinical versus mechanical prediction: A meta-analysis. Psychological Assessment, 12, 19-30.

Finn, A., Trevethan, S., Carriere, G., & Kowalski, M. (1999). Female inmates, aboriginal inmates, and inmates serving life sentences: A one day snapshot. Ottawa, ON: Statistics Canada.

Gendreau, P., Little, T., & Goggin, C. (1996). A meta-analysis of the predictors of adult offender recidivism: What works! Criminology, 34, 575-607.

Gobeil, R. (2007). Revalidation of the Security Reclassification Scale for Women (SRSW) (R-191). Ottawa, ON: Correctional Service of Canada.

Grant, B.A. & Luciani, F.P. (1998). Security classification using the Custody Rating Scale (R-67). Ottawa, ON: Correctional Service of Canada.

Holsinger, A.M., Lowenkamp, C.T., & Latessa, E.J. (2003). Ethnicity, gender, and the Level of Service Inventory-Revised. Journal of Criminal Justice, 31, 309-320.

Holsinger, A.M., Lowenkamp, C.T., & Latessa, E.J. (2006). Exploring the validity of the Level of Service Inventory-Revised with Native American offenders. Journal of Criminal Justice, 34, 331-337.

Luciani, F., Motiuk, L., & Nafekh, M. (1996). An operational review of the Custody Rating Scale: Validity, reliability, and practical utility (R-47). Ottawa, ON: Correctional Service of Canada.

Luciani, F., Taylor, G., & Motiuk, L. (1998). National field test results of the security reclassification protocol: Final report. Unpublished manuscript. Correctional Service of Canada, Ottawa.

Moore, J.-P. (2003). First Nations, Métis, Inuit, and non-Aboriginal federal offenders: A comparative profile (R-134). Ottawa, ON: Correctional Service of Canada.

Moore, J.-P., Trevethan, S., & Conley, J. (2004). Program and service needs of federally incarcerated Metis offenders in Saskatchewan (R-155). Ottawa, ON: Correctional Service of Canada.

Motiuk, L.L. (1997). Classification for correctional programming: The Offender Intake Assessment (OIA) process. Forum on Corrections Research, 9(1), 18-22.

Motiuk, L.L. & Vuong, B. (2005). Homicide, sex, robbery and drug offenders in federal corrections: An end-of-2004 review (B-37). Ottawa, ON: Correctional Service of Canada.

Nuffield, J. (1982). Parole decision-making in Canada: Research towards decision guidelines. Ottawa, ON: Solicitor General of Canada.

Office of the Correctional Investigator. (2006). Departmental Performance Report for the Period Ending March 31, 2006. Ottawa, Canada: Author.

Office of the Correctional Investigator. (2007). Annual report of the Office of the Correctional Investigator 2006-2007. Ottawa, Canada: Author.

Pridemore, W.A. (2004). Review of the literature on risk and protective factors of offending among Native Americans. Journal of Ethnicity in Criminal Justice, 2, 45-62.

Public Safety Canada. (2007). Corrections and Conditional Release Statistical Overview 2007. Ottawa, Canada: Author.

R. v. Gladue. (April 23, 1999). Supreme Court of Canada.

Rice, M. (2007, June). Current status of violence risk assessment: Is there a role for clinical judgment? Invited presentation at the North American Correctional and Criminal Justice Psychology Conference, Ottawa, ON.

Rugge, T. (2006). Risk assessment of male Aboriginal offenders: A 2006 perspective (2006-01). Ottawa, ON: Public Safety and Emergency Preparedness Canada.

Sioui, R., Thibault, J., & Amiskou Groupe Conseil. (2001). The relevance of a cultural adaptation for Aboriginals of the Reintegration Potential Resassessment Scale (RPRS): Preliminary study (R-109). Ottawa, ON: Correctional Service of Canada.

Solicitor General of Canada (1987). Development of a security classification model for Canadian federal offenders. Ottawa, ON: Author.

Statistics Canada (2001). Aboriginal peoples in Canada. Ottawa, ON: Author.

Trevethan, S., Moore, J.-P., & Rastin, C.J. (2002). A profile of Aboriginal offenders in federal facilities and serving time in the community. Forum on Corrections Research, 14(3), 1-7.

Zinger, I. (2004). Actuarial risk assessment and human rights: A commentary. Canadian Journal of Criminology and Criminal Justice, 46, 607-621.

Appendix 1: Mean Item Scores on SRS Component Items

Mean Item Scores on SRS Component Items
  Mean (SD) Score
Scale Item Possible Range Aboriginal Non-Aboriginal Total
1. Serious disciplinary offences 0.5-2.0 0.7 (0.4) 0.7 (0.4) 0.7 (0.4)
2. Minor disciplinary offences 0.5-1.0 0.6 (0.2) 0.6 (0.2) 0.6 (0.2)
3. Number of recorded incidents 0.5-1.0 1.4 (1.0) 1.4 (1.0) 1.4 (1.0)
4. Pay grade 0.5-1.5 0.8 (0.3) 0.7 (0.3) 0.7 (0.3)
5. Number of segregation periods 0.5-3.0 1.6 (1.2) 1.5 (1.2) 1.5 (1.2)
6. Detention referral  0.5-2.0 1.1 (0.7) 1.0 (0.7) 1.0 (0.7)
7. Correctional plan progress 2.0-5.0 3.8 (0.9) 3.9 (0.9) 3.9 (0.9)
8. Correctional plan motivation  2.0-6.0 4.0 (1.4) 4.0 (1.4) 4.0 (1.4)
9. Drug and alcohol rating 0.5-1.5 1.1 (0.3) 1.0 (0.4) 1.0 (0.3)
10. Successful temporary absences / work releases 0.5-1.0 1.0 (0.1) 1.0 (0.1) 1.0 (0.1)
11. Age at review 0.5-1.0 0.7 (0.2) 0.6 (0.2) 0.6 (0.2)
12. Psychological concerns 0.5-1.5 0.9 (0.5) 0.8 (0.5) 0.9 (0.5)
13. CRS escape history  0.5-1.0 0.6 (0.2) 0.5 (0.1) 0.5 (0.1)
14. CRS incident history 0.5-3.0 1.2 (0.8) 1.2 (0.8) 1.2 (0.8)

Note. NAboriginal = 6,717. NNon-Aboriginal = 25,611. NTotal = 32,328.

Appendix 2: SRS Inter-Item Correlations

SRS Inter-Item Correlations

Items 		2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14.
1. .33 .32 .21 .35 -.08 .15 .20 .29 .04 .23 -.03 .06 .17
2. - .25 .21 .23 -.09 .13 .17 .14 .03 .25 -.02 .03 .12
3.   - .25 .41 -.01 .19 .27 .24 .01 .19 .03 .05 .24
4.     - .32 -.10 .25 .34 .14 .03 .20 .01 .02 .09
5.       - -.02 .25 .30 .20 .01 .18 .05 .08 .21
6.         - .04 .07 -.11 .01 -.20 .27 .01 .05
7.           - .61 .03 .08 .06 .17 .01 .15
8.             - .10 .06 .05 .17 .04 .16
9.               - -.03 .07 -.03 .07 .10
10.                 - .03 .03 -.01 .04
11.                   - -.11 .01 .09
12.                     - .01 .05
13.                       - .12

Note. NTotal = 32,328. Item 1: Serious disciplinary offences. Item 2: Minor disciplinary offences. Item 3: Number of recorded incidents. Item 4: Pay grade. Item 5: Number of segregation periods. Item 6: Detention referral. Item 7: Correctional plan progress. Item 8. Correctional plan motivation. Item 9: Drug and alcohol rating. Item 10: Successful temporary absences / work releases. Item 11: Age at review. Item 12: Psychological concerns. Item 13: CRS escape history. Item 14: CRS incident history.

1 Several exceptions exist: offenders serving life sentences for first or second degree murder undergo security reviews on a bi-annual basis, and offenders who are housed at minimum security undergo reviews only if involved in an incident which suggests that they may no longer be appropriately housed at this level of security.

2 A separate instrument, the Security Reclassification Scale for Women (Blanchette, 2005), is used with female offenders.

3 See Rugge (2006) for a comprehensive review of traditional risk factors as they apply to Aboriginal and non-Aboriginal offenders.

4 The data management system also allows for self-identification of Innu ethnicity, but this category was not endorsed by any of the offenders in this sample

5 The development sample was reflective of the population of federally incarcerated offenders undergoing security reviews. As of 1996-97 (presumably the time when scale development was occurring), Aboriginal offenders represented roughly 16% of the population of federally incarcerated offenders (CSC Performance Management, 2008); therefore, while the SRS was not explicitly constructed to be culturally-sensitive, the indices of association used in selecting scale items were calculated based on a representative Aboriginal sample. Nonetheless, results were aggregated across offenders and it is possible that associations derived from the larger proportion of non-Aboriginal offenders “washed out” those stemming from Aboriginal offenders.

6 Notably, this process has changed somewhat as of March 2007; the present report reflects the procedure used during the period under study. Since March 2007, recommendations to different security placements are made in separate documents; prior to March 2007, these recommendations were made within the SRS.

7 This analysis was not conducted for Aboriginal offenders because the SIR-R1 has only been psychometrically normed with non-Aboriginal offenders and the application of the scale to Aboriginal offenders is therefore not supported by CSC.

8 A fixed follow-up period was not used. Since minimum security offenders undergo reviews infrequently, lengthy periods can potentially elapse between security reviews and subsequent releases for minimum security offenders. Therefore, offenders housed at minimum security would be under-represented if a fixed follow-up were used.

9 This being said, analyses focused on reviews conducted between 2002 and 2006 because a minor change to the SRS was made immediately preceding this period and because ending in 2006 allowed time to follow-up offenders for any re-offences or readmissions.

10 This analysis was conducted with non-Aboriginal offenders only as the Statistical Information on Recidivism Scale - Revised is not used with Aboriginal offenders.

Date Modified: 2015-03-05

Top of Page
Important Notices