Predictive Accuracy of Dynamic Risk Factors for Aboriginal and Non-Aboriginal Sex Offenders

Abstract

Although Aboriginal offenders are overrepresented in Canadian prisons, there is limited research examining the extent to which commonly used risk factors and risk scales are applicable to Aboriginals. Aboriginal (n = 88) and non-Aboriginal (n = 509) sex offenders on community supervision were compared on the dynamic risk factors of STABLE-2007. Data on sexual, violent, any crime, and any recidivism (including breaches) were collected with an average follow-up of 3.4 years. Aboriginal offenders scored significantly higher than non-Aboriginal offenders on STABLE-2007 total scores and on several items measuring general criminality. STABLE-2007 did not significantly predict recidivism with Aboriginal offenders (although it did for non-Aboriginals). The general antisociality items were generally significantly less predictive for Aboriginals than non-Aboriginals, whereas items assessing sexual self-regulation and relationship stability predicted similarly for both groups. These exploratory results suggest that Aboriginal sex offenders are a higher-needs group but that some STABLE-2007 items are not predictive with this population.

Keywords

risk assessment dynamic risk factors Aboriginal offenders sex offenders predictive accuracy

Aboriginal offenders are overrepresented in the Canadian criminal justice system, despite the attempts of both the government and the Supreme Court to address this serious problem (Canadian Criminal Code § 718.2(e); R. v. Gladue, 1999). Currently, Aboriginals represent 3.7% of the adult male population (Statistics Canada, 2008) but 19.2% of the incarcerated male population (Public Safety Canada, 2009). Compared with non-Aboriginal offenders, Aboriginal offenders are also more likely to be incarcerated for violent offences (Trevethan, Moore, & Rastin, 2002) and to be sent to maximum security institutions (Public Safety Canada, 2009). Despite this overrepresentation, there has been insufficient research comparing Aboriginal and non-Aboriginal offenders on risk factors and examining the extent to which assessment, treatment, and management strategies are equally effective for both groups.

Risk Assessment

Most decisions throughout an offender’s progression in the criminal justice system involve risk assessment, including sentencing, security classification, parole decisions, treatment needs, and supervision intensity. Many of the most commonly used risk assessment instruments focus on static risk factors that are historical and unchanging. Dynamic risk factors are features related to recidivism, which can change, and when changed, should alter the likelihood of recidivism (Andrews & Bonta, 2006). For general offenders, empirically supported dynamic risk factors include antisocial personality pattern, procriminal attitudes, procriminal associates, work/school problems, family/marital problems, poor use of leisure/recreation time, and substance abuse (along with the static factor of criminal history, these are referred to as the “Central 8” risk factors; Andrews & Bonta, 2006). Important dynamic risk factors for sex offenders include the “Central 8” as well as deviant sexual interests, sexual preoccupations (Hanson & Morton-Bourgon, 2005), intimacy deficits, and emotional congruence with children (Mann, Hanson, & Thornton, 2010).

Dynamic risk factors have been found to add incremental predictive validity to static risk factors (e.g., Allan, Grace, Rutherford, & Hudson, 2007; Olver, Wong, Nicholaichuk, & Gordon, 2007). In addition, a recent meta-analysis found that dynamic risk scales predicted sexual recidivism with greater predictive accuracy than static scales (Hanson & Morton-Bourgon, 2009). Although there are several ways to identify and combine dynamic items into a global structured risk assessment (for a review, see Hanson & Morton-Bourgon, 2009), this article will focus on an actuarial risk assessment scale (i.e., STABLE-2007). Actuarial scales involve explicit rules to combine prespecified items into total scores, and they include empirically derived estimates of recidivism probability linked to each total score (Meehl, 1954).

Risk Assessment With Aboriginal Offenders

Although there is ample empirical support for static and dynamic risk factors and risk assessment scales, virtually all of this research is conducted with samples of primarily Caucasian offenders. Given the large number of Aboriginal offenders in the Canadian correctional system, knowing whether risk scales developed and validated on primarily non-Aboriginal offenders can generalize to Aboriginal offenders is imperative.

There is considerable debate regarding risk assessment with Aboriginal offenders, with some critics arguing that current assessment scales fail to consider important differences between Aboriginal and non-Aboriginal offenders and the possibility of unique risk factors for Aboriginal offenders (Maynard, Coebergh, Anstiss, Bakker, & Huriwai, 1999; Webb, 2003). These individuals advocate the development of risk scales specific to Aboriginal offenders.

The assertion of differences between Aboriginal and non-Aboriginal offenders is empirically well supported. On average, Aboriginal offenders are younger, less educated, and have higher rates of unemployment compared with non-Aboriginal offenders (Statistics Canada, 2006). In addition, they have lengthier criminal histories (Dell & Boe, 2000; Holsinger, Lowenkamp, & Latessa, 2003) and report more negative childhood histories (Johnston, 1997; Trevethan, Auger, Moore, MacDonald, & Sinclair, 2002). In adulthood, Aboriginal offenders are also slightly more likely to be single and are rated as higher need in the domain of family and/or marital problems (Trevethan, Moore, & Rastin, 2002). Not surprisingly based on the above-noted differences, Aboriginal offenders also have higher recidivism rates than non-Aboriginal offenders (Bonta, Rugge, & Dauvergne, 2003; Sioui & Thibault, 2002).

The fact that Aboriginal offenders exhibit more risk factors than non-Aboriginal offenders does not mean that risk factors predict differently between these groups. Although limited, research conducted thus far has found that history of antisocial behaviour, antisocial attitudes, and antisocial peers are equally predictive of future criminal behaviour for Aboriginal and non-Aboriginal offenders and that substance abuse may be a better predictor for Aboriginals compared with non-Aboriginal offenders (Bonta, 1989; Bonta, LaPrairie, & Wallace-Capretta, 1997; British Columbia Public Safety and Solicitor General, 2004). Psychopathy has also been found to predict recidivism similarly for Aboriginal and non-Aboriginal offenders (Boer, Couture, Geddes, & Ritchie, 2004; Schmidt, McKinnon, Chattha, & Brownlee, 2006). However, family/marital problems and school/work problems were not found to predict recidivism with Aboriginal offenders (Bonta, 1989; Bonta et al., 1997).

In terms of the predictive validity of risk assessment tools, several studies have demonstrated that scales such as the Level of Service Inventory (LSI) and its subsequent adaptations predict recidivism equally well for Aboriginal and non-Aboriginal general offenders (Bonta, 1989; Brews, Wormith, & Guzzo, 2009; Gossner & Wormith, 2007; Gross & Sroga, 2008; Tanasichuk, Wormith, & Guzzo, 2009). Similarly, positive results have been reported for the Statistical Information on Recidivism Scale (SIR; Bonta & Rugge, 2004; Hann & Harman, 1993; Nafekh & Motiuk, 2002). However, it is important to note that these findings have not always been replicated outside Canada. In a study in the United States, the LSI:R did not predict recidivism for Native American general offenders (Holsinger, Lowenkamp, & Latessa, 2006).

In general, given the limited research on Aboriginal offenders, it is not surprising to also discover few studies concerning Aboriginal sex offenders. In terms of social history, Ellerby and MacPherson (2002) found that when compared with non-Aboriginal sex offenders, Aboriginal sex offenders were more likely to have had an unstable upbringing, including separation from parents, exposure to general criminal activity, domestic abuse, and higher rates of neglect and sexual abuse. Also, consistent with the literature on general Aboriginal offenders, Aboriginal sex offenders had lower educational achievement, higher rates of unemployment, and greater rates of substance abuse than non-Aboriginal sex offenders. However, interestingly, Aboriginal sex offenders had lower levels of sexual deviance (e.g., coercive sexual fantasies, paraphilias) compared with non-Aboriginal sex offenders (Ellerby & MacPherson, 2002).

Regarding the types of offences committed, Aboriginals are more likely to have female victims (Ellerby & MacPherson, 2002; Nahanee, 1996; Rastin & Johnson, 2002; Rojas & Gretton, 2007) within their own ethnicity (Ellerby & MacPherson, 2002; Nahanee, 1996) and to abuse substances during the commission of the offence (Ellerby & MacPherson, 2002; Nahanee, 1996; Rastin & Johnson, 2002; Rojas & Gretton, 2007). Aboriginal offenders are also less likely to target younger victims (e.g., prepubescent) and victims with whom they hold a position of trust (Ellerby & MacPherson, 2002).

In addition, Aboriginal sex offenders show higher rates of sexual recidivism (Rastin & Johnson, 2002; Rojas & Gretton, 2007; Williams, Vallée, & Staubi, 1997), violent recidivism (Rojas & Gretton, 2007), and general recidivism (Rastin & Johnson, 2002; Rojas & Gretton, 2007) compared with non-Aboriginal sex offenders. The only study we identified that examined possible differences in the predictive accuracy of risk factors or a risk assessment scale found that Static-99 predicted recidivism comparably for Aboriginal and non-Aboriginal sex offenders (area under the receiver operating characteristic curve [AUC] = .67 for both groups; Nicholaichuk, 2001).

Purpose of Current Study

Given that Aboriginal offenders represent nearly 20% of the Canadian prison population, it is surprising that there is not more research examining predictors of recidivism for Aboriginal sex offenders. The purpose of this exploratory study was twofold: (a) examine whether Aboriginal sex offenders have different dynamic risk profiles than non-Aboriginal sex offenders and (b) examine whether the STABLE-2007 scale and its individual items predict sexual, violent, any crime, and any recidivism differently for Aboriginal and non-Aboriginal sex offenders. The hypotheses were that Aboriginal offenders would be significantly at higher risk than non-Aboriginal offenders on STABLE-2007 but that the items and risk scale would predict recidivism similarly for both groups.

Method

Participants

The sample for this study is from the Dynamic Supervision Project (Hanson, Harris, Scott, & Helmus, 2007). More detailed information on the study methodology is available in the original report, although it should be noted that the current analyses included only the Canadian offenders in this project and those who had at least one STABLE-2007 assessment. Also excluded were 4 female offenders and 12 offenders with less than 1 year of follow-up information. All offenders in this project were adult sex offenders starting a period of community supervision (e.g., probation or parole) between 2001 and 2005. In total, 597 offenders were included in this study; 88 (14.7%) were Aboriginal sex offenders and 509 (85.3%) were non-Aboriginal sex offenders. Offenders were identified as Aboriginal by the supervising officers. No explicit criteria were listed to define Aboriginal, and no information is available on the offender’s band or nation. The number of Aboriginal offenders living on reserves is unknown, although they would likely be a minority given that most of the supervising officers were located in major urban areas.

The offenders were supervised in all provinces and territories across Canada. Virtually all offenders were in the provincial/territorial correctional systems; only 13 offenders were supervised by the federal prison system (Correctional Service of Canada [CSC]). Similar to the general Canadian population, the proportion of Aboriginal offenders varied considerably across regions, with the highest proportions in the Prairie provinces and Northern Territories.

Measures

STABLE-2007

An earlier version of the dynamic risk scale, STABLE-2000, was developed for this project (see Hanson et al., 2007). However, based on the data received, three items were dropped from the scale (measuring attitudes tolerant of sexual offending), and the coding rules for three additional items were modified slightly (capacity for relationship stability, emotional identification with children, and deviant sexual interests). The revised version of the scale was called STABLE-2007. STABLE-2007 includes 13 items assessed using a three-point rating scale (from 0 to 2). The items are organized into five sections: capacity for relationship stability (one item), intimacy deficits (five items), general self-regulation (three items), sexual self-regulation (three items), and cooperation with supervision. Total STABLE-2007 scores are obtained by summing the items. The maximum possible score for child molesters is 26 and for rapists is 24 (because the emotional identification with children item is not scored for rapists). Offenders with scores of 0-3 are considered low risk/need, 4-11 are considered moderate risk/need, and 12 and higher are considered high risk/need.

Procedure

Data were collected as part of the routine supervision practices of the officers participating in the project. The STABLE-2007 assessment was to be completed within the first 3 months of supervision, and then every 6 months thereafter. Only the first STABLE assessment will be analyzed in this study because of the small number of offenders for which multiple assessments were received. Because STABLE-2007 was developed at the end of the project, the officers were actually trained on and submitted STABLE-2000 scores. STABLE-2007 scores were calculated from SPSS syntax using the original STABLE-2000 items, victim information submitted by the officers, and from an item on the Static-99 scale (ever lived with a lover for at least 2 years). Supervising officers attended a 2-day training session, although in rare cases, officers submitted data who had been trained by apprenticing with other local officers. Formal interrater reliability information is not available, but scores on training exercises and selective site visits suggested reasonable understanding and application of the scoring rules (for more information, see Hanson et al., 2007).

Recidivism

Information concerning new offences was gathered from reviews of provincial and national criminal history records, as well as from supervising officers and local police jurisdictions. In addition, one recidivist was identified from a newspaper article. National Canadian Police Information Centre (CPIC) records maintained by the Royal Canadian Mounted Police (RCMP) were received in August 2005 and June 2006. Provincial records were received from the following jurisdictions: British Columbia (January 2006), Manitoba (April 2005), and Ontario (December 2005). The Offender Management System of CSC was checked in May 2005 for recidivism information for the CSC offenders registered in the project.

Each recidivism source contained at least some unique information. Part of this can be attributed to time delays. For example, CPIC records only include information after charges have been formally processed (e.g., resulted in conviction or dismissal), whereas the other sources could contain information on charges in progress. In addition, some of the unique information is a reflection of the unreliability of the official criminal records. However, note that all the recidivism sources would be expected to underestimate the true rate of recidivism.

Information on the date and circumstances of new offences was obtained to classify them into one of the four recidivism categories (defined below). Recidivism was considered to have occurred if the agency reporting the information believed that the offence occurred. However, for breaches, an official record of parole revocation or a new conviction for violation of conditional release was required. Given that criminal history records were the major source of recidivism information, the majority of recidivism events were linked to an officially recorded charge or conviction.

Four types of recidivism were recorded. The first category was sexual recidivism, which included all crimes with a sexual motivation, whether or not the name of the offence was explicitly sexual. Violent recidivism was defined as all crimes that involved direct confrontation with the victim and included sexual recidivism. The third category was any criminal recidivism, which included all crimes but excluded breaches. The final category, any recidivism, included all crimes (sexual, violent, and nonviolent) as well as breaches. For breaches, if the offence description indicated that a crime had occurred, the incident was coded as a crime.

The follow-up period was calculated from the date that the first assessment information was collected to the date of the last recidivism information received (or until death or deportation). For the few cases that did not appear on any official record, the follow-up end date was set 1 month after the last assessment information was received. The offender start dates ranged from January 2001 to October 2005, with an average follow-up time of 3.4 years (SD = 1.0; range, 1.0–5.4 years).

Overview of Analyses

The AUC was used to examine whether Aboriginal offenders scored differently on risk factors than non-Aboriginal offenders and to examine the predictive accuracy of STABLE-2007. Cox regression was used to test whether predictive accuracy differed for Aboriginal and non-Aboriginal offenders. AUC is an effect size statistic for dichotomous outcome variables (Swets, Dawes, & Monahan, 2000) and can vary between 0 and 1, with .50 indicating the level of prediction that would be expected by chance. An AUC value of less than .50 indicates negative predictive accuracy (i.e., low scores on a risk factor are associated with higher recidivism rates). The AUC values between .50 and 1 indicate prediction exceeding chance levels, with higher numbers indicating greater accuracy. Also worth noting is that AUCs only examine relative predictive accuracy (e.g., are high-risk offenders more likely to reoffend than low-risk offenders?); they do not provide information about absolute recidivism rates.

As a rough heuristic, an AUC of .56 corresponds to a small effect size, whereas .64 reflects a moderate effect and .71 reflects a large effect size, as these values correspond to Cohen’s ds of 0.2, 0.5, and 0.8, when certain assumptions are satisfied (see Rice & Harris, 2005). An AUC value is statistically significant if the 95% confidence interval (CI) does not include .50. Given the small sample size for Aboriginal offenders, the AUC CIs tended to be extremely wide, indicating low power to detect statistically significant predictive accuracy, particularly compared with non-Aboriginal offenders. As such, our interpretation of the AUC results focused on magnitude and not on significance.

Although the AUCs are most commonly used to predict dichotomous outcome variables (e.g., recidivism), they can also be used to examine differences on a dichotomous grouping variable (e.g., do risk scores predict Aboriginal status?). Group differences could also be examined using t tests, but t tests assume interval data, and the risk scales are technically ordinal. In contrast, the AUCs are appropriate for ordinal data and provide a slightly more conservative test.

Cox regression analysis (Allison, 1984) was used to test whether the predictive accuracy of a STABLE-2007 item or total score was significantly different between Aboriginal and non-Aboriginal sex offenders. Cox regression estimates relative risk ratios (hazard rates) associated with one or more predictor variables from survival data with unequal follow-up times. To test for differences in predictive accuracy, Cox regression models were calculated using the predictor variable, Aboriginal status, and an interaction term for Aboriginal status and the predictor. A significant interaction means that the predictive accuracy of the item differs between Aboriginal and non-Aboriginal offenders. These analyses should be considered exploratory given that the number of tests conducted increases the Type I error rate.

In the Results section, only the test of the interaction was presented (Wald test). Although Wald values are always more than 0, we added positive and negative signs to indicate the direction of the findings, with positive values denoting higher predictive accuracy for Aboriginal offenders compared with non-Aboriginal offenders and negative signs denoting lower predictive accuracy for Aboriginals. The regression coefficients for the predictor variables were not reported because they are difficult to interpret in the presence of an interaction term, and they do not answer the primary research questions because they are combined for Aboriginal and non-Aboriginal offenders. Readers interested in the predictive accuracy of these items for the combined sample are encouraged to refer to the original report (Hanson et al., 2007).

Given variability in crime rates across Canada (Public Safety Canada, 2009), geographical region was entered as a strata variable, which allows Cox regression to estimate separate baseline hazard functions (i.e., recidivism rates) for each value of the stratified variable (in other words, it controls for regional differences in recidivism rates). This provides a more conservative test because the proportion of Aboriginal offenders differs across regions, so removing regional variability may also remove some variability due to Aboriginal status. To determine whether this test was too conservative, all analyses were run both with and without geographic region as a strata variable. Largely similar results were obtained, with a few additional significant findings without the strata variable. Given that the number of analyses increases the Type I error rate, only the more conservative analyses (with region as a strata variable) were reported.

Results

Descriptive Information

Of the 88 Aboriginal offenders, 10 reoffended sexually (11.4%), 28 reoffended violently (31.8%), 36 committed any new crime (40.9%), and 45 recidivated in any way, including breaches (51.1%). Of the 509 non-Aboriginal offenders, 37 reoffended sexually (7.3%), 64 reoffended violently (12.6%), 92 committed any new crime (18.1%), and 118 recidivated in any way, including breaches (23.2%).

Aboriginal offenders were significantly younger (M = 36.2, SD = 10.5) than non-Aboriginal offenders (M = 41.8, SD = 13.9; t (146) = 4.42, p < .001). Additional descriptive information for the two groups is presented in Table 1. There were significant differences between groups in offender type (x² = 11.06, df = 4, p = .026), with Aboriginal offenders more likely to be rapists or mixed offenders (committing offences against both adults and children) and less likely to have only noncontact offences. The proportion of child molesters (incest or extrafamilial) was similar in both groups. There were also significant differences between groups in the most serious victim injury (χ² = 9.13, df = 3, p = .028), with Aboriginal offenders more likely to cause victim injury than non-Aboriginal offenders (18.2% vs. 10.4%, respectively) and less likely to commit noncontact offences (3.4% vs. 11.4%, respectively). The proportion of offenders with intellectual impairments or major mental disorders was similar for Aboriginal and non-Aboriginal offenders; approximately 5% of offenders were developmentally delayed, and approximately 9% had previously been hospitalized for a major mental disorder.

Table 1.

Descriptive Information for Sample

	Aboriginal offenders (n = 88)		Non-Aboriginal offenders (n = 509)
	n (%)	n missing	n (%)	n missing	χ²	df	p
Offender type		1		20	11.06	4	.026
Noncontact	2 (2.3%)		54 (11.0%)
Extrafamilial child molester	22 (25.3%)		133 (27.2%)
Incest	17 (19.5%)		119 (24.3%)
Rapist	39 (44.8%)		155 (31.7%)
Mixed	7 (8.0%)		28 (5.7%)
Most serious victim injury		0		0	9.13	3	.028
Noncontact only	3 (3.4%)		58 (11.4%)
Physical contact	69 (78.4%)		395 (77.6%)
Victim injury	16 (18.2%)		53 (10.4%)
Life-threatening	0 (0.0%)		3 (0.6%)
Developmentally delayed	4 (4.9%)	6	24 (4.8%)	4	—	—	—
Major mental disorder	8 (9.4%)	3	48 (9.6%)	7	0.002	1	.965

Group Comparisons on STABLE-2007

Table 2 presents the means and standard deviations of Aboriginal and non-Aboriginal offenders on STABLE-2007 (items and total scores), and the results of the AUC analyses comparing the two groups. Means could range from 0 to 2 but most were less than one, indicating that the risk factors were generally low in both groups. The AUC values greater than .50 indicate that Aboriginals scored higher than non-Aboriginals. There were no significant differences between Aboriginal and non-Aboriginal offenders in their scores on significant social influences, emotional identification with children, nor for any of the items in the sexual self-regulation section. In the intimacy deficits section, Aboriginal offenders scored significantly higher on lack of concern for others. In the general self-regulation section, compared with non-Aboriginal offenders, Aboriginal offenders scored significantly higher on impulsive acts and poor cognitive problem solving. Aboriginal offenders also had significantly more problems cooperating with supervision than non-Aboriginal offenders. STABLE-2007 total scores were significantly higher among Aboriginal offenders compared with non-Aboriginal offenders, although the average score for both groups was within the moderate risk category.

Table 2.

Comparing Aboriginal (n = 88) and Non-Aboriginal (n = 509) Offenders on STABLE-2007 Item and Total Scores

	Aboriginal		Non-Aboriginal
	M	SD	M	SD	AUC	95% CI
Significant social influences	0.68	0.78	0.48	0.70	.569	[.502, .636]
Intimacy deficits
Capacity for relationship stability	1.09	0.70	1.02	0.76	.524	[.462, .587]
Emotional identification with children	0.15	0.36	0.25	0.52	.468	[.406, .531]
Hostility towards women	0.34	0.56	0.37	0.60	.494	[.429, .558]
General social rejection/loneliness	0.85	0.77	0.75	0.76	.538	[.474, .603]
Lack of concern for others	0.73	0.74	0.46	0.67	.602	[.537, .668]
General self-regulation
Impulsive acts	0.78	0.72	0.40	0.65	.650	[.587, .713]
Poor cognitive problem solving	1.06	0.73	0.69	0.73	.635	[.573, .696]
Negative emotionality/hostility	0.55	0.69	0.46	0.66	.533	[.467, .599]
Sexual self-regulation
Sexual preoccupations/sex drive	0.57	0.69	0.48	0.68	.538	[.473, .604]
Sex as coping	0.63	0.84	0.48	0.71	.536	[.468, .605]
Deviant sexual interests	0.66	0.71	0.78	0.74	.455	[.391, .519]
Cooperation with supervision	0.69	0.81	0.31	0.58	.627	[.558, .695]
STABLE-2007 total score	8.77	4.91	6.92	5.06	.618	[.558, .679]

Predictive Accuracy for STABLE-2007

Sexual recidivism

Table 3 presents the AUC values for the predictive accuracy of STABLE-2007 items and total scores as well as the Wald test from the Cox regression analysis used to examine group differences in predictive accuracy. The AUCs were higher for Aboriginals for only 2 of the 13 items (capacity for relationship stability and sex as coping). For Aboriginal offenders, the AUC values for the items ranged from .42 to .65, with the median effect size at chance levels (AUC = .48). For non-Aboriginals, the AUC values ranged from .51 to .70, with a median of .60. From the Cox regression analyses, four items had significantly lower predictive accuracy for Aboriginal offenders (hostility towards women, lack of concern for others, impulsive acts, and poor cognitive problem solving). For STABLE-2007 total scores, the predictive accuracy appeared lower for Aboriginal offenders (AUC = .529) compared with non-Aboriginal offenders (AUC = .701), but the difference was not significant (Wald = −2.78, df = 1, p = .096).

Table 3.

Predictive Accuracy of STABLE-2007 Items and Total Scores for Sexual Recidivism

	Aboriginal		Non-Aboriginal
	AUC	95% CI	AUC	95% CI	Wald^a
Significant social influences	.535	[.331, .738]	.593	[.497, .689]	−0.52
Intimacy deficits
Capacity for relationship stability	.628	[.471, .786]	.616	[.526, .707]	+0.12
Emotional identification with children	.473	[.288, .658]	.510	[.411, .609]	−0.47
Hostility towards women	.441	[.263, .619]	.640	[.540, .740]	−4.14*
General social rejection/loneliness	.506	[.282, .731]	.663	[.572, .754]	−2.00
Lack of concern for others	.420	[.257, .583]	.617	[.516, .717]	−4.96*
General self-regulation
Impulsive acts	.469	[.288, .650]	.699	[.609, .790]	−4.68*
Poor cognitive problem solving	.438	[.229, .647]	.688	[.605, .771]	−5.76*
Negative emotionality/hostility	.481	[.291, .670]	.564	[.462, .665]	−0.65
Sexual self-regulation
Sexual preoccupations/sex drive	.569	[.387, .752]	.566	[.467, .665]	−0.02
Sex as coping	.649	[.452, .847]	.599	[.502, .696]	+0.29
Deviant sexual interests	.556	[.362, .750]	.596	[.493, .699]	−0.26
Cooperation with supervision	.479	[.304, .653]	.593	[.490, .696]	−2.39
STABLE-2007 total score	.529	[.331, .726]	.701	[.611, .791]	−2.78

Interactions between Aboriginal status and STABLE-2007 items were tested using Cox regression, with 1 degree of freedom for each Wald test and using jurisdiction as a strata variable. Although Wald values are always more than 0, we added positive and negative signs to indicate the direction of the findings, with positive values denoting higher predictive accuracy for Aboriginal offenders compared with non-Aboriginal offenders and negative signs denoting lower predictive accuracy for Aboriginals.

p < .05.

Violent recidivism

Table 4 presents the AUC and Cox regression results. The AUC values were consistently lower for Aboriginal offenders compared with non-Aboriginal offenders, with only 2 of 13 items favouring Aboriginals (capacity for relationship stability and emotional identification with children). For Aboriginals, the AUC values for the items ranged from .48 to .59, with a median of .51. For non-Aboriginals, the AUC values ranged from .49 to .74, with a median of .61. Cox regression analyses identified five items with significantly lower predictive accuracy for Aboriginal offenders (hostility towards women, lack of concern for others, impulsive acts, poor cognitive problem solving, and cooperation with supervision). In addition, STABLE-2007 total scores had significantly lower predictive accuracy for Aboriginal offenders (Wald test = −5.24, df = 1, p = .022).

Table 4.

Predictive Accuracy of STABLE-2007 Items and Total Scores for Violent Recidivism

	Aboriginal		Non-Aboriginal
	AUC	95% CI	AUC	95% CI	Wald^a
Significant social influences	.582	[.451, .713]	.606	[.528, .683]	−1.09
Intimacy deficits
Capacity for relationship stability	.589	[.461, .717]	.580	[.506, .655]	+0.23
Emotional identification with children	.523	[.391, .654]	.493	[.417, .569]	+0.10
Hostility towards women	.508	[.380, .637]	.649	[.572, .726]	−7.12**
General social rejection/loneliness	.476	[.339, .614]	.611	[.540, .683]	−3.39
Lack of concern for others	.483	[.357, .609]	.644	[.567, .721]	−8.06**
General self-regulation
Impulsive acts	.500	[.369, .631]	.745	[.678, .812]	−11.26***
Poor cognitive problem solving	.487	[.360, .614]	.722	[.662, .781]	−10.37***
Negative emotionality/hostility	.496	[.366, .626]	.588	[.510, .667]	−1.55
Sexual self-regulation
Sexual preoccupations/sex drive	.516	[.384, .648]	.560	[.483, .637]	−0.17
Sex as coping	.535	[.402, .667]	.547	[.470, .623]	−0.002
Deviant sexual interests	.538	[.404, .672]	.551	[.471, .631]	−0.08
Cooperation with supervision	.479	[.351, .608]	.657	[.579, .734]	−8.88**
STABLE-2007 total score	.530	[.398, .663]	.722	[.660, .785]	−5.24*

Interactions between Aboriginal status and STABLE-2007 items were tested using Cox regression, with 1 degree of freedom for each Wald test and using jurisdiction as a strata variable. Although Wald values are always more than 0, we added positive and negative signs to indicate the direction of the findings, with positive values denoting higher predictive accuracy for Aboriginal offenders compared with non-Aboriginal offenders, and negative signs denoting lower predictive accuracy for Aboriginals.

p < .05. **p < .01. ***p ≤ .001.

Any crime

Table 5 presents the AUC and Cox regression results. Of the 13 item comparisons, the AUCs were higher for Aboriginal group for only 3 items (significant social influences, emotional identification with children, and deviant sexual interests). For Aboriginals, the AUC values for the items ranged from .47 to .61, with a median of .53. For non-Aboriginals, the AUC values ranged from .48 to .72, with a median of .61. Five items had significantly lower predictive accuracy for Aboriginal offenders (general social rejection/loneliness, lack of concern for others, impulsive acts, poor cognitive problem solving, and cooperation with supervision). For STABLE-2007 total scores, the predictive accuracy appeared lower for Aboriginal offenders (AUC = .581) compared with non-Aboriginals (AUC = .706), but the difference was not significant (Wald = −2.45, df = 1, p = .117).

Table 5.

Predictive Accuracy of STABLE-2007 Items and Total Scores for Any New Crime

	Aboriginal		Non-Aboriginal
	AUC	95% CI	AUC	95% CI	Wald^a
Significant social influences	.611	[.489, .733]	.604	[.537, .671]	+0.60
Intimacy deficits
Capacity for relationship stability	.604	[.482, .726]	.613	[.552, .675]	−0.26
Emotional identification with children	.516	[.392, .640]	.482	[.418, .546]	+0.20
Hostility towards women	.526	[.403, .649]	.608	[.541, .675]	−2.68
General social rejection/loneliness	.467	[.341, .592]	.611	[.550, .671]	−4.46*
Lack of concern for others	.530	[.408, .651]	.646	[.581, .711]	−7.30**
General self-regulation
Impulsive acts	.579	[.455, .702]	.725	[.664, .787]	−5.50*
Poor cognitive problem solving	.549	[.428, .670]	.708	[.651, .765]	−6.96**
Negative emotionality/hostility	.519	[.395, .643]	.594	[.527, .661]	−0.92
Sexual self-regulation
Sexual preoccupations/sex drive	.523	[.399, .647]	.545	[.478, .611]	−0.001
Sex as coping	.529	[.405, .654]	.538	[.472, .605]	−0.16
Deviant sexual interests	.537	[.414, .661]	.514	[.447, .582]	+0.01
Cooperation with supervision	.514	[.392, .636]	.617	[.550, .685]	−6.39*
STABLE-2007 total score	.581	[.459, .703]	.706	[.649, .763]	−2.45

p < .05. **p ≤ .01.

Any recidivism (including breaches)

Table 6 presents the AUC and Cox regression results. Only two AUC item comparisons favoured Aboriginal offenders (significant social influences and emotional identification with children). For Aboriginals, the AUC values for the items ranged from .47 to .63, with a median of .53, with two items demonstrating significant predictive accuracy (significant social influences and impulsive acts). For non-Aboriginals, the AUC values ranged from .51 to .73, with a median of .62. Six items had significantly lower predictive accuracy for non-Aboriginal offenders (hostility towards women, general social rejection/loneliness, lack of concern for others, impulsive acts, poor cognitive problem solving, and cooperation with supervision). In addition, STABLE-2007 total scores had significantly lower predictive accuracy for Aboriginals (Wald = −4.26, df = 1, p = .039).

Table 6.

Predictive Accuracy of STABLE-2007 Items and Total Scores for Any Recidivism (Including Breaches)

	Aboriginal		Non-Aboriginal
	AUC	95% CI	AUC	95% CI	Wald^a
Significant social influences	.621	[.504, .739]	.589	[.528, .650]	+0.26
Intimacy deficits
Capacity for relationship stability	.584	[.465, .703]	.643	[.588, .698]	−0.57
Emotional identification with children	.531	[.410, .652]	.508	[.448, .568]	+0.56
Hostility towards women	.508	[.387, .630]	.613	[.553, .674]	−5.09*
General social rejection/loneliness	.474	[.352, .595]	.625	[.568, .682]	−5.90*
Lack of concern for others	.516	[.393, .638]	.644	[.585, .704]	−9.05**
General self-regulation
Impulsive acts	.629	[.513, .746]	.731	[.676, .787]	−5.25*
Poor cognitive problem solving	.555	[.433, .677]	.713	[.661, .766]	−11.01***
Negative emotionality/hostility	.581	[.461, .700]	.620	[.560, .680]	−0.42
Sexual self-regulation
Sexual preoccupations/ sex drive	.520	[.399, .641]	.565	[.504, .626]	−0.12
Sex as coping	.526	[.405, .647]	.571	[.510, .632]	−0.18
Deviant sexual interests	.532	[.411, .653]	.536	[.475, .596]	−0.01
Cooperation with supervision	.537	[.416, .659]	.636	[.575, .697]	−7.68**
STABLE-2007 total score	.606	[.487, .725]	.731	[.680, .782]	−4.26*

p < .05. **p < .01. ***p ≤ .001.

Summary of predictive accuracy

Three items had significantly lower predictive accuracy for Aboriginal sex offenders for all four types of recidivism (lack of concern for others, impulsive acts, and poor cognitive problem solving). In addition, hostility towards women had significantly lower predictive accuracy for Aboriginals for the prediction of sexual, violent, and any recidivism. Cooperation with supervision had significantly lower predictive accuracy for Aboriginal offenders for all outcomes except sexual recidivism. General social rejection/loneliness was also less predictive for Aboriginals for any crime and any recidivism. Although several other items had lower AUC values for Aboriginal offenders compared with non-Aboriginal offenders, these differences were not statistically significant based on Cox regression.

For STABLE-2007 total scores, the predictive accuracy for Aboriginal offenders was significantly lower than non-Aboriginal offenders for the prediction of violent recidivism and any recidivism (including breaches) and approached significance for sexual recidivism. Differences in predictive accuracy between groups were not significant for any new crime.

Discussion

This study found several significant differences in the presence and predictive accuracy of dynamic risk factors for Aboriginal and non-Aboriginal sex offenders. In terms of demographic and offence characteristics, Aboriginal sex offenders were younger, more likely to be rapists or mixed offenders, less likely to be noncontact offenders, and more likely to cause victim injury. Consistent with the first hypothesis, Aboriginal sex offenders scored significantly higher than non-Aboriginal sex offenders on STABLE-2007, particularly on the items that assessed general antisociality (e.g., lack of concern for others, impulsivity, poor problem solving, and cooperation with supervision). Higher scores for Aboriginal offenders on these items make sense given that rapists generally exhibit more general criminality and antisociality than child molesters (Firestone, Bradford, Greenberg, & Serran, 2000; Harris, Smallbone, Dennison, & Knight, 2009).

Contrary to our second hypothesis, the same items on which Aboriginal offenders scored higher generally predicted recidivism with significantly lower accuracy than non-Aboriginal offenders. Conversely, Aboriginal and non-Aboriginal offenders had similar profiles on the sexual self-regulation items, and these items had comparable predictive accuracy for both groups. Significant social influences and capacity for relationship stability seemed to be equally predictive for Aboriginal and non-Aboriginal offenders.

The reason for the lower predictive accuracy for several of the general antisociality items is unclear. These results were surprising given that these items are strongly supported predictors of sexual recidivism (Hanson & Morton-Bourgon, 2005) and general recidivism (Andrews & Bonta, 2006), and other items and scales measuring similar constructs have been validated with general Aboriginal offenders (e.g., Gossner & Wormith, 2007; Schmidt et al., 2006). However, there are some key differences in this study. This is the first study we are aware of that examined the predictive accuracy of dynamic risk factors with Aboriginal sex offenders. In addition, most of the previous studies examined whether the items or risk scales predicted recidivism at all with Aboriginal offenders, whereas this study focused on whether the items predicted as well for Aboriginal offenders as they did with non-Aboriginal offenders. Impulsive acts, for example, significantly predicted any recidivism for Aboriginal offenders, but the effect size was still significantly lower than for non-Aboriginal offenders.

The poor performance of some of these items with Aboriginal offenders could be due to measurement issues. It is possible that the causal factors for recidivism are similar for Aboriginal and non-Aboriginal offenders but that the STABLE-2007 items are not adequately measuring these constructs, particularly for Aboriginal offenders. Although the static/dynamic conceptualization of risk factors has been dominant for over a decade, it has been challenged in recent years (e.g., Beech & Ward, 2004). Mann et al. (2010) have recently advocated for a new focus on psychologically meaningful risk factors. In this model, the risk factors generally measured in research studies are indicators of underlying propensities. For example, self-regulation problems may be an underlying propensity that causes offending and reoffending. Certain risk factors such as substance abuse, job instability, gambling, fighting, and poor problem solving may all be indicators of this propensity. Interestingly, this model downplays the importance of the static/dynamic distinction because underlying propensities can have both static and dynamic indicators.

It is possible that some of the risk factors measured in STABLE-2007 are not indicators of the same construct for Aboriginal and non-Aboriginal offenders. Perhaps for Aboriginal offenders, some of the behaviours measured in the impulsive acts item (e.g., substance abuse, fighting, and employment instability) are indicators of something other than self-regulation problems (e.g., cultural marginalization and poverty). A corollary of this is the possibility that similar underlying propensities may be expressed differently by Aboriginal and non-Aboriginal offenders due to social, economic, and cultural differences.

The strengths of this study are its prospective design and its field implementation of a risk assessment scale among trained supervision officers. This study also provides an important contribution by shedding light on an underresearched area. The main limitation of this study is the small sample size of Aboriginal offenders, which would likely lead to increased Type II errors (i.e., nonsignificant AUCs). However, alternately, the large number of analyses would increase the probability of Type I errors. Readers should consider the findings of this study as preliminary and exploratory; further replications with larger sample sizes are sorely needed.

Another limitation is that we used the same sample on which STABLE-2007 was developed. Although it is possible that this could inflate the AUC values, only three of the items (and the total score) were changed from STABLE-2000 (which was developed before these data were being collected) to STABLE-2007, so inflation is anticipated to be minimal. In addition, any inflation effect is unlikely to have a disparate effect on the two groups and is therefore unlikely to affect analyses of differences in predictive accuracy between Aboriginal and non-Aboriginal offenders. Nonetheless, independent validations of STABLE-2007 would be a valuable contribution to this field.

In addition, although the field use of the scale increases generalizability to real-world settings, it also introduces considerable variations in the quality of the STABLE ratings. The original report from this project identified substantial differences in the results for supervision officers considered conscientious (defined as those who submitted all the data that were requested of them) compared with less-conscientious officers (Hanson et al., 2007). Although the assessments in this study were completed by well-trained officers who demonstrated a good understanding of the coding rules, the lower levels of conscientiousness may reflect the tension between ideal practice and the realities of limited resources. Nonetheless, optimal tests of predictive accuracy for Aboriginal offenders (or for any group) should be conducted with risk assessments coded by raters who are properly trained, who have access to complete information, and who have sufficient time to adequately score the measures.

Implications for Researchers

This study provides preliminary data on the use of dynamic risk factors with Aboriginal sex offenders. This study did not find support for many of the general antisociality items of STABLE-2007, but additional studies are needed to replicate these findings, especially with larger samples of Aboriginal offenders. Research should also explore potentially unique risk factors for Aboriginal offenders (e.g., cultural identity) and examine possible differences within Aboriginal groups (e.g., different bands or whether the offenders live on a reserve).

Research should also examine the absolute predictive accuracy of actuarial risk scales (e.g., do the recidivism estimates overrepresent or underrepresent recidivism for Aboriginal offenders?). More research is also needed to understand what underlying propensities are related to sexual offending and to explore optimal ways of measuring these constructs. If the items on current risk assessment scales are not valid indicators of these constructs, it may be necessary to identify better indicators, such as self-report questionnaires or implicit tasks. Finally, another avenue of future research should explore the effects of rater prejudice on the scores and predictive accuracy of risk assessment scales with explicit coding rules.

Implications for Risk Evaluators

It is difficult to provide recommendations to practitioners on the basis of an exploratory study with a restricted sample size. Virtually all the STABLE-2007 items did not achieve statistical significance in predicting recidivism, which means that these items cannot be used with confidence for Aboriginal sex offenders. However, the statistical power of these analyses was quite small. Even for non-Aboriginal offenders (n = 509), some of the items did not reach statistical significance, even though they did in the original report with a larger sample size (Hanson et al., 2007). For these reasons, our interpretation of these exploratory findings focused on the magnitude of the effect sizes and the tests for differences in predictive accuracy, which provided some basis for optimism.

This study suggests that sexual self-regulation, relationship stability, and social influences predict recidivism similarly among Aboriginal and non-Aboriginal sex offenders. Although more research is needed to replicate these findings, we believe that practitioners can use these risk factors in their evaluations with some justification. Evaluators should be more cautious about using STABLE-2007 total scores and the items that had significantly lower predictive accuracy (e.g., hostility towards women, lack of concern for others, impulsivity, poor problem solving, and cooperation with supervision).

It is also worth noting that other risk assessment scales have already been validated for use with Aboriginal offenders, such as the LSI and its variants (e.g., Brews et al., 2009; Gossner & Wormith, 2007) and Static-99/R (Babchishin, Blais, & Helmus, 2012; Nicholaichuk, 2001). Therefore, the evaluations of Aboriginal sex offenders could include Static-99/R to assess static risk factors, the LSI-R or its variants to assess dynamic risk factors, and the sexual self-regulation items of STABLE-2007 (with caution) to supplement the LSI variables and to guide treatment intervention decisions.

Footnotes

Acknowledgements

We would like to thank R. Karl Hanson and Andrew Harris for allowing us to use the data for the current analyses and Karl for his helpful comments on an earlier draft of this article.

Authors’ Note

The views expressed are those of the authors and not necessarily those of Public Safety Canada.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Funding for this project was provided in part by the Social Science and Humanities Research Council of Canada.

References

Allan

Grace

R. C.

Rutherford

Hudson

S. M.

(2007). Psychometric assessment of dynamic risk factors for child molesters. Sexual Abuse: A Journal of Research and Treatment, 19, 347-367. doi:10.1177/10790632070190040210.1177/107906320701900402

Allison

P. D.

(1984). Event history analysis: Regression for longitudinal event data. Beverly Hills, CA: SAGE.

Andrews

D. A.

Bonta

(2006). The psychology of criminal conduct (4th ed.). Newark, NJ: LexisNexus/Anderson.

Babchishin

K. M.

Blais

Helmus

(2012). Do static risk factors predict differently for Aboriginal offenders? A comparison using Static-99R and Static-2002R. Canadian Journal of Criminology and Criminal Justice,54, 1-44.

Beech

A. R.

Ward

(2004). The integration of etiology and risk in sexual offenders: A theoretical framework. Aggression and Violent Behavior, 10, 31-63. doi:10.1016/j.avb.2003.08.00210.1016/j.avb.2003.08.002

Boer

Couture

Geddes

Ritchie

(2004). Yókw’tól: Risk management guide for Aboriginal offenders. British Columbia: Correctional Services of Canada.

Bonta

(1989). Native inmates: Institutional response, risk, and needs. Canadian Journal of Criminology, 31, 49-62.

Bonta

LaPrairie

Wallace-Capretta

(1997). Risk prediction and reoffending: Aboriginal and non-Aboriginal offenders. Canadian Journal of Criminology, 39, 127-144.

Bonta

Rugge

(2004). The prediction of recidivism with Aboriginal offenders. Unpublished manuscript.

10.

Bonta

Rugge

Dauvergne

(2003). The reconviction rate of federal offenders (Corrections Research User Report No. 2003-02). Ottawa, ON: Solicitor General Canada.

11.

Brews

Wormith

J. S.

Guzzo

(2009, June). An examination of the LSI-OR and its supplementary scales with female offenders. Paper presented at the Annual Convention of the Canadian Psychological Association, Montreal, QC.

12.

British Columbia Public Safety and Solicitor General. (2004). Female and male community risk needs assessment: Data from Aboriginal offenders assessed in 1998. British Columbia, Canada: Corrections Branch.

13.

Dell

C. A.

Boe

(2000). An examination of Aboriginal and Caucasian women offender risk and needs factors. Ottawa, ON: Correctional Service of Canada.

14.

Ellerby

L. A.

MacPherson

(2002). Exploring the profiles of Aboriginal sexual offenders: Contrasting Aboriginal and non-Aboriginal sexual offenders to determine unique client characteristics and potential implications for sex offender assessment and treatment strategies (Research Report No. R-122). Ottawa, ON: Correctional Service of Canada.

15.

Firestone

Bradford

J. M.

Greenberg

D. M.

Serran

G. A.

(2000). The relationship of deviant sexual arousal and psychopathy in incest offenders, extrafamilial child molesters, and rapists. Journal of the American Academy of Psychiatry and the Law, 28, 303-308.

16.

Gossner

Wormith

J. S.

(2007). The prediction of recidivism among young offenders in Saskatchewan. Canadian Journal of Police & Security Services, 5, 70-82.

17.

Gross

A. L.

Sroga

(2008, June). Establishing the needs of Aboriginal offenders using the Level of Service Inventory–Ontario Revision. Poster presented at the Annual Convention of the Canadian Psychological Association, Halifax, NS.

18.

Hann

R. G.

Harman

W. G.

(1993). Predicting release risk for Aboriginal penitentiary inmates (User Report No. 1993-12). Ottawa, ON: Solicitor General Canada.

19.

Hanson

R. K.

Harris

A. J. R.

Scott

Helmus

(2007). Assessing the risk of sexual offenders on community supervision: The Dynamic Supervision Project (Corrections Research User Report No. 2007-05). Ottawa, ON: Public Safety Canada. Retrieved from http://www.publicsafety.gc.ca/res/cor/rep/_fl/crp2007-05-en.pdf

20.

Hanson

R. K.

Morton-Bourgon

K. E.

(2005). The characteristics of persistent sexual offenders: A meta-analysis of recidivism studies. Journal of Consulting and Clinical Psychology, 73, 1154-1163. doi:10.1037/0022-006X.73.6.115410.1037/0022-006X.73.6.1154

21.

Hanson

R. K.

Morton-Bourgon

K. E.

(2009). The accuracy of recidivism risk assessments for sexual offenders: A meta-analysis of 118 prediction studies. Psychological Assessment, 21, 1-21. doi:10.1037/a001442110.1037/a0014421

22.

Harris

D. A.

Smallbone

Dennison

Knight

R. A.

(2009). Specialization and versatility in sexual offenders referred for civil commitment. Journal of Criminal Justice, 37, 37-44. doi:10.1016/j.jcrimjus.2008.12.00210.1016/j.jcrimjus.2008.12.002

23.

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. doi:10.1016/S0047-2352(03)00025-410.1016/S0047-2352(03)00025-4

24.

Holsinger

A. M.

Lowenkamp

C. T.

Latessa

E. J.

(2006). Exploring the predictive validity of the Level of Service Inventory-Revised with Native American offenders. Journal of Criminal Justice, 34, 331-337. doi:10.1016/j.jcrimjus.2006.03.00910.1016/j.jcrimjus.2006.03.009

25.

Johnston

J. C.

(1997). Aboriginal offender survey: Case files and interview sample (Research Report R-61). Ottawa, ON: Correctional Service Canada. Retrieved from http://www.csc-scc.gc.ca/text/rsrch/reports/r61/r61_e.pdf

26.

Mann

R. E.

Hanson

R. K.

Thornton

(2010). Assessing risk for sexual recidivism: Some proposals on the nature of psychologically meaningful risk factors. Sexual Abuse: A Journal of Research and Treatment, 22, 191-217. doi:10.1177/107906321036603910.1177/1079063210366039

27.

Maynard

Coebergh

Anstiss

Bakker

Huriwai

(1999). Ki te arotu—Toward a new assessment: The identification of cultural factors which may pre-dispose Māori to crime. Social Policy Journal of New Zealand, 13, 43-58.

28.

Meehl

P. E.

(1954). Clinical versus statistical prediction: A theoretical analysis and a review of the evidence. Minneapolis: University of Minnesota Press.

29.

Nafekh

Motiuk

(2002). The Statistical Information on Recidivism–Revised 1 (SIR-R1) scale: A psychometric examination. Ottawa, ON: Correctional Service of Canada, Research Branch.

30.

Nahanee

T. A.

(1996). A profile of Aboriginal sex offenders in Canadian federal custody. Forum on Corrections Research, 8, 8-9. Retrieved from http://www.csc-scc.gc.ca/text/pblct/forum/e082/082b_e.pdf

31.

Nicholaichuk

(2001, November). The comparison of two standardized risk assessment instruments in a sample of Canadian Aboriginal sexual offenders. Paper presented at the Annual Conference of the Association for the Treatment of Sexual Abusers, San Antonio, TX.

32.

Olver

M. E.

Wong

S. C. P.

Nicholaichuk

Gordon

(2007). The validity and reliability of the Violence Risk Scale–Sexual Offender version: Assessing sex offender risk and evaluating therapeutic change. Psychological Assessment, 19, 318-329. doi:10.1037/1040-3590.19.3.31810.1037/1040-3590.19.3.318

33.

Public Safety Canada. (2009). Corrections and conditional release statistical overview: Annual report 2009. Ottawa, ON: Author. Retrieved from http://www.publicsafety.gc.ca/res/cor/rep/2009-ccrso-eng.aspx

34.

Gladue

R. v.

, 1 S.C.R. 688 (1999).

35.

Rastin

C. J.

Johnson

S. L.

(2002). Inuit sexual offenders: Victim, offence and recidivism characteristics. Forum on Corrections Research, 14, 46-48. Retrieved from http://www.csc-scc.gc.ca/text/pblct/forum/e143/143n_e.pdf

36.

Rice

M. E.

Harris

G. T.

(2005). Comparing effect sizes in follow-up studies: ROC area, Cohen’s d, and r. Law and Human Behavior, 29, 615-620. doi:10.1007/s10979-005-6832-710.1007/s10979-005-6832-7

37.

Rojas

E. Y.

Gretton

(2007). Background, offence characteristics, and criminal outcomes of Aboriginal youth who sexually offend: A closer look at Aboriginal youth intervention needs. Sexual Abuse: A Journal of Research and Treatment, 19, 257-283. doi:10.1177/10790632070190030610.1177/107906320701900306

38.

Schmidt

McKinnon

Chattha

H. K.

Brownlee

(2006). Concurrent and predictive validity of the psychopathy checklist: Youth version across gender and ethnicity. Psychological Assessment, 18, 393-401. doi:10.1037/1040-3590.18.4.39310.1037/1040-3590.18.4.393

39.

Sioui

Thibault

(2002). Examining reintegration potential for Aboriginal offenders. Forum on Corrections Research, 14, 49-51.

40.

Statistics Canada. (2006). Victimization and offending among Aboriginal population in Canada. Juristat, 26(3) (Catalogue 85-002-XIE). Ottawa, ON: Author.

41.

Statistics Canada. (2008). Aboriginal peoples highlight tables: 2006 Census (97-558-XWE2006002). Ottawa, ON: Author. Retrieved from http://www12.statcan.ca/census-recensement/2006/dp-pd/hlt/97-558/index.cfm?Lang=E

42.

Swets

J. A.

Dawes

R. M.

Monahan

(2000). Psychological science can improve diagnostic decisions. Psychological Science in the Public Interest, 1, 1-26. doi:10.1111/1529-1006.00110.1111/1529-1006.001

43.

Tanasichuk

C. L.

Wormith

J. S.

Guzzo

(2009). The predictive validity of the Level of Service Inventory–Ontario Revision (LSI-OR) with Aboriginal offenders. Toronto, Canada: Ontario Ministry of Community Safety and Correctional Services.

44.

Trevethan

Auger

Moore

J. P.

MacDonald

Sinclair

(2002). Childhood experiences affect Aboriginal offenders. Forum on Corrections Research, 14, 7-9. Retrieved from http://www.csc-scc.gc.ca/text/pblct/forum/e143/143c_e.pdf

45.

Trevethan

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, 17-19. Retrieved from http://www.csc-scc.gc.ca/text/pblct/forum/e143/143f_e.pdf

46.

Webb

(2003). Risk factors, criminogenic needs and Maori. [Online]. In Crothers

(Ed.), Knowledge, capitalism and critique, Proceedings of the 2003 Sociological Association of Aotearoa New Zealand (SAANZ) conference (pp. 1-9). Retrieved from http://saanz.rsnz.org/webb.doc

47.

Williams

Vallée

Staubi

(1997). Aboriginal sex offenders: Melding spiritual healing with cognitive-behavioural treatment. Ottawa, ON: Correctional Service Canada. Retrieved from http://www.csc-scc.gc.ca/text/pblct/so/aboriginal/toce-eng.shtml