Do Risk and Protective Factors for Chronic Offending Vary Across Indigenous and White Youth Followed Prospectively Through Full Adulthood?

Abstract

Although Indigenous youth are overrepresented in justice systems across North America, Australia, and New Zealand, explanations for this overrepresentation are principally theoretical as data at the individual level are lacking. Risk for offending among Indigenous youth may be overestimated because of their typically more negative socioeconomic outcomes tied to historical injustices perpetrated by governments across different nations. Data on 403 adolescent offenders followed from ages 12 to 29 were used to examine offending trajectories and associated risk and protective factors across Indigenous and White participants. A greater number of social adversities characterized Indigenous youth, yet they did not differ from White youth in their likelihood of assignment to the highest rate offending trajectory. Culturally sensitive assessment of risk for offending is recommended.

Keywords

criminal careers culture Indigenous Peoples offending trajectories

The overrepresentation of Indigenous youth in the criminal justice system is an ongoing concern across jurisdictions in the United States (Cross, 2008), Canada (LaPrairie, 2002), and Australia (Shepherd, Adams, McEntyre, & Walker, 2014). Despite this international concern, the risk factors associated with offending among Indigenous Peoples remain poorly understood, especially with respect to offending that persists through adulthood (Yessine & Bonta, 2009). This is principally because of a lack of individual-level offending data specific to Indigenous Peoples (Piquero, 2015). The key theory, research, and policy question consists of two parts: (a) do risk and protective factors vary across Indigenous and non-Indigenous youth, and (b) if differences do exist, are they important for better understanding offending outcomes? The latter question is critical as some researchers have challenged the validity of explanations of offending that lacked cross-cultural validation (see Hart, 2016). Explanations of offending using socioeconomic markers of disadvantage may be particularly problematic as these factors are disproportionately more common among Indigenous Peoples (Maurutto & Hannah-Moffat, 2007; Shepherd et al., 2014), and thus the risk for offending may be overestimated for this group. Specifically, Indigenous youth may have a greater likelihood of experiencing certain “risk” factors compared with non-Indigenous youth, yet the presence of such factors may not actually increase the odds of more negative offending outcomes. Nevertheless, this differential exposure may lead to prejudicial decisions resulting in a more serious type of sentence (e.g., prison vs. probation), a longer sentence, or placement in a more restrictive facility (e.g., maximum vs. medium security).

The current study addressed Piquero’s (2015) call for research on Indigenous Peoples and offending involvement across multiple developmental stages. Building from Yessine and Bonta’s (2009) study on Indigenous youth and offending trajectories, we considered potential ecological fallacy concerns in explanations of offending; that is, whether risk factors predicting offending within a general sample predicted offending across ethnic subgroups within this sample (e.g., Thorndike, 1939). To address these themes, offending trajectories between ages 12 to 29 were modeled for 403 participants from British Columbia, Canada. Although the sample is from Canada, explanations for individual-level patterns of offending may be similar across both American and Canadian Indigenous Peoples given the historical shared experiences of both groups, stemming from forced assimilation laws, which have been advanced as distal causes of tragic intergenerational cultural, familial, and criminal justice system consequences (Maurutto & Hannah-Moffat, 2007; Shepherd et al., 2014).

Overrepresentation and Explanations of Indigenous Involvement in Offending Behavior

Ethnic minority justice system overrepresentation is a prevailing issue, and in no justice system population is this type of policy concern greater than within samples of individuals adjudicated for their involvement in serious and violent offending (Piquero, 2014). Indigenous overrepresentation is especially salient in the prison context (Duran & Duran, 1995) and has persisted in Canada and the United States despite legislated attempts to reduce the use of custody among this population (Indian Affairs, 2016; Roberts & Reid, 2017). These attempts included focusing on the rights of Indigenous youth processed in the juvenile justice system (e.g., the right to legal counsel, restricted use of detention, and procedures to distinguish delinquent acts from need for services) and the rights of Indigenous groups to develop their own codes in response to youth arrested for alcohol-related and drug-related offenses, including diversion alternatives (Indian Affairs, 2016). The sociocultural impact of residential schools has resulted in marked structural disadvantage, including poverty, poorer educational and health outcomes, and difficulties with basic rights such as access to water (Ford, 2012). Specific to the criminal justice system, historically, prejudicial policies have helped explain Indigenous overrepresentation more broadly. However, less has been learned about what happens to Indigenous Peoples after they enter the justice system (cf. Reingle & Maldonado-Molina, 2012; Yessine & Bonta, 2009), including whether they are associated with more concerning offending outcomes relative to non-Indigenous Persons. Even less is known about risk and protective factors informative of offending trajectory involvement for Indigenous youth (cf. Sittner & Hautala, 2016), and we are not aware of any research that considered whether risk and protective factors for offending trajectories varied across Indigenous and non-Indigenous groups. Again, this latter question is especially important for assuring that sentencing and other criminal justice system practices are not culturally biased (Hart, 2016). First, we review the literature on risk and protective factors more generally and then consider the implications of such factors within an Indigenous context.

Sittner and Hautala (2016) provided guidelines for incorporating Catalano and Hawkins’s (1996) social developmental model into the study of risk and protective factors associated with offending trajectories among Indigenous youth. This model incorporates risk and protective factors from a range of social control and social learning theories and from across different levels (i.e., individual-level, familial-level, and peer-level risk factors). The family and school context are especially valued within Indigenous cultures (Sittner & Hautala, 2016) despite the historical injustices perpetrated against this group that directly affected family functioning and educational experiences. Familial factors may be particularly important to examine across ethnicity given that there are marked family structure differences between Indigenous and White youth, with the former group being more likely to have the extended family play a larger role in social upbringing (Whitbeck, Sittner, Hartshorn, & Walls, 2014). As such, widespread familial adversity may be a particularly salient factor for Indigenous youth in the unfolding of a more frequent offending trajectory.

Another important issue concerns the extent to which certain negative circumstances are (a) more common among Indigenous Peoples and (b) considered within the criminological literature to be risk factors for general offending, yet (c) do not actually increase the risk for offending within this group. For example, alcohol abuse is a concerning factor within Indigenous populations (e.g., Whitbeck, Adams, Hoyt, & Chen, 2004) and is also a risk factor for violent offending (e.g., Dawkins, 1997), yet is not necessarily a key risk factor for criminal justice system involvement among Indigenous Peoples (e.g., Rempel, Somers, Calvert, & McCandless, 2015). Untangling this type of relationship is critical because of the inclusion of multiple factors in various risk assessment tools and theoretical perspectives that may be more prevalent among Indigenous Persons despite not actually influencing offending within this group. Thus, it is important to consider factors that are potentially more common to Indigenous youth, considered risk factors for offending, yet not affect the likelihood of offending for this group. Factors that may be disproportionately represented among Indigenous youth include physical and sexual abuse, early sexual behavior, and negative self-identity (e.g., Pavkov, Travis, Fox, King, & Cross, 2010; Stewart, Livingston, & Dennison, 2008). The current study examined the impact of such factors on offending outcomes measured through adulthood within a sample of Indigenous and White youth.

Contribution of the Current Study to the Existing Literature

Only a few studies have utilized an individual-based trajectory approach to understand how traditionally or theoretically well-established risk factors are associated with Indigenous involvement in crime across several developmental stages. Focusing on, and comparing, the most serious Indigenous and White young offenders has several important policy implications regarding custodial sentencing and treatment intervention strategies. Thus far, three studies have examined the criminal career trajectories of Indigenous youth. Yessine and Bonta (2009) compared offending patterns of male Indigenous and non-Indigenous youth on probation. Their study contrasted from most trajectory research in that they reported just two offending trajectories for each subsample. However, these authors also only examined offending at five different measurement periods, which may have masked heterogeneity in offending patterns across multiple years of follow-up. Risk factors were examined at the bivariate level for both ethnic subgroups and showed that risk factors associated with chronic offending were not necessarily the same across the two ethnicities. Reingle and Maldonado-Molina (2012) identified three trajectories of violent offending over the period of adolescence for their sample of Indigenous male and female youth from the Add Health study. This study found an overlap between higher levels of violence and higher levels of victimization. Reingle and Maldonado-Molina (2012) indicated that substance use and lack of parental involvement were associated with a combination of offending and victimization experiences. Finally, Sittner and Hautala (2016) identified five offending trajectories during the period of adolescence for a large sample of Indigenous youth. Multinomial logistic regression (MLR) analyses revealed that peer delinquency, substance use, early dating, and male sex were particularly informative of chronic offending.

The current study adds to the three aforementioned studies in several ways, including the examination of self-reported risk and protective factors from a variety of different domains, the use of an offender sample, the inclusion of a comparison group of White youth from the same population, the extension of the follow-up period into full adulthood, and the use of multivariate analyses with interaction effects to explore whether the impact of risk and protective factors on offending trajectories varied across ethnic subgroup. Although some of these characteristics were captured by the three previous studies, none of the three studies included each of these characteristics. In addressing these themes, the aims of the current study were to (a) evaluate whether the prevalence of different risk and protective factors varied across Indigenous and White youth, (b) evaluate whether the heightened risk factor profiles of Indigenous youth corresponded with an increased likelihood of assignment to a more frequent offending trajectory, and (c) examine whether risk and protective factors associated with a more frequent offending pattern varied across ethnic subgroup. Taken together, the study aims are expected to help address ecological fallacy concerns recently raised in the risk assessment literature (Hart, 2016).

Method

Sample

Data were used from the Incarcerated Serious and Violent Young Offender Study (ISVYOS), which has been ongoing in British Columbia, Canada, since 1998. As part of this study, adolescent offenders (ages 12-19) were interviewed in open and secure custody facilities throughout the province. The ISVYOS consists of two cohorts: one was sampled at the time that Canada’s Young Offender Act represented the federal legislation in charge of youth justice and the second was sampled at the time that Canada’s Youth Criminal Justice Act represented the federal legislation in charge of youth justice. Interest within the current study was on offending outcomes in adulthood, and because many youth from the second cohort were not yet adults, only those that were recruited as part of the first cohort were included in the current study. The current study included all members of this cohort that were at least 30 years of age at the time of the most recent wave of data collection. This allowed for offending trajectories to be examined through age 29 for all members of the sample, meaning that all participants were of equal age at the end of data collection. Within this sample (n = 403), only participants that self-reported being White (n = 297) or Indigenous (n = 106) were included. Too few participants were from other ethnic groups (e.g., Chinese, Black, Middle Eastern, Indian, South East Asian) to allow for comparisons across ethnicities. Although only 6.2% of the population of British Columbia self-identifies as Indigenous (Statistics Canada, 2013), 22.2% of offenders in the full sample self-identified as such (i.e., including those self-identifying as non-White and non-Indigenous). Although this overrepresentation may affect generalizability of research findings for most research questions, it is conducive to addressing the primary research question in the current study.

Procedure

The Ministry of Child and Family Development (MCFD) acts as the legal guardian to all incarcerated youth in the province, and their consent allowed the research team to approach youth in custody centers throughout British Columbia. The study received research ethics approval from MCFD as well as the Simon Fraser University Research Ethics Board. Following ethics approval, participants were approached on their unit in the custody center by research assistants (RAs), if several criteria were met: (a) they were English-speaking, (b) they demonstrated an understanding of interview questions (e.g., had no noticeably severe deficits in IQ), and (c) they were willing to provide accurate information. If subjects persistently lied about known information (e.g., their age, offense resulting in incarceration), then they were permanently removed from the interview schedule at the discretion of the site manager. Halfway through the study, the prevalence of refusal was examined and revealed that approximately 5% of youth refused to participate when approached. Youth assented to participate after they were informed of the study’s purpose, how information would be collected (e.g., interview and file information), and that all information would be kept confidential apart from direct threats against themselves or someone else. Participants were told that their involvement or noninvolvement would not affect court outcomes, nor would it affect their stay in custody. RAs interviewed participants in an isolated interview room away from their living unit, other youth, and custody staff. Access to file information prior to interviews ensured that RAs were aware of discrepancies between interview responses and official records.

Measures

Demographic characteristics and risk and protective factors

All demographic characteristics, risk factors, and protective factors were based on self-reported information measured at the time of the participant’s interview as an adolescent. All youth indicated their gender as either male or female. Indigenous youth reported a First Nations or Metis background. Criminal trajectories were measured from ages 12 to 29 for all participants. Adolescents who were older at the time of their interview may have experienced greater opportunity for exposure to certain risk factors, and therefore, age at interview was included as a control variable.

Risk and protective factors in Table 1 were compared across Indigenous and White youth. Due to the severity of the sample, risk factors such as alcohol and drug use were overwhelmingly prevalent (e.g., 93.5% and 96.8%, respectively). As such, early onset of these factors was used instead. Early onset was defined as prior to age 12 to establish temporal order between risk/protective factors and offending outcomes (participants could not incur a criminal conviction until at least age 12). A total of four measures of substance use were included: substance use versatility (the sum of nine different substances: alcohol, marijuana, acid, mushrooms, ecstasy, cocaine, crack cocaine, heroin, and crystal methamphetamine), use of street drugs (either crack cocaine, heroin, or crystal methamphetamine), early onset of alcohol use, and early onset of drug use. For the substance use versatility scale, when a scale is comprised of dichotomous items, tetrachoric ordinal alpha provides a more accurate indication of scale reliability than Cronbach’s alpha (Gadermann, Guhn, & Zumbo, 2012). The nine items had a tetrachoric ordinal value of .88.

Table 1.

Risk Factor Comparisons Between Indigenous and White Adjudicated Youth.

	Indigenous (n = 106)	White (n = 297)	χ²/t, p, Φ/d
	M (SD)/% (n)	M (SD)/% (n)	χ²/t, p, Φ/d
Demographics
Male	73.6 (78)	79.5 (236)	χ²(1) = 1.57, ns, Φ = .06
Age at interview	16.29 (1.30)	16.38 (1.19)	t(401) = −0.61, ns, d = .07
Substance use
Substance use versatility	4.68 (2.30)	4.69 (2.23)	t(398) = −0.40, ns, d = .00
Street drug user	48.6 (51)	47.8 (141)	χ²(1) = 0.02, ns, Φ = .01
Drug use—before 12	39.2 (38)	32.8 (90)	χ²(1) = 1.27, ns, Φ = .06
Alcohol use—before 12	32.6 (31)	30.1 (81)	χ²(1) = 0.21, ns, Φ = .02
School behavior
Enrolled in school	49.1 (52)	50.3 (149)	χ²(1) = 0.05, ns, Φ = .01
Skip school—before 12	24.7 (21)	21.9 (49)	χ²(1) = 0.28, ns, Φ = .03
Number of different schools	6.03 (7.31)	6.29 (5.72)	t(321) = −0.33, ns, d = .04
Abuse
Physical abuse	55.7 (59)	41.7 (121)	χ²(1) = 6.08, p < .05, Φ = .12
Sexual abuse	29.5 (31)	19.9 (58)	χ²(1) = 4.08, p < .05, Φ = .10
Aggression and violence
Fighting—weekly basis	32.2 (28)	24.0 (54)	χ²(1) = 2.17, ns, Φ = .08
Angers easily	58.9 (53)	56.7 (131)	χ²(1) = 0.13, ns, Φ = .02
Bad temper	79.1 (72)	76.1 (178)	χ²(1) = 0.34, ns, Φ = .03
Identity
Self-perception—obedience	23.99 (4.97)	23.38 (5.18)	t(391) = 1.03, ns, d = .12
Self-perception—prosociality	14.72 (3.19)	14.65 (3.04)	t(389) = 0.19, ns, d = .02
Self-perception—ego	15.67 (2.65)	15.80 (2.69)	t(388) = −0.42, ns, d = .05
Others’ perception—obedience	27.02 (7.17)	25.78 (7.05)	t(355) = 1.46, ns, d = .17
Others’ perception—prosociality	16.28 (4.88)	15.99 (4.77)	t(355) = 0.51, ns, d = .06
Others’ perception—ego traits	13.48 (3.81)	12.88 (3.81)	t(351) = 1.30, ns, d = .16
Family problems
Ministry care	36.3 (37)	27.9 (78)	χ²(1) = 2.52, ns, Φ = .08
Living outside parents’ home	53.9 (55)	55.4 (155)	χ²(1) = 0.06, ns, Φ = .01
Family with history of alcohol abuse	2.08 (1.48)	1.45 (1.27)	t(389) = 4.16, p < .001, d = .46
Family with history of drug abuse	1.32 (1.07)	1.02 (1.14)	t(389) = 2.31, p < .05, d = .27
Family with history of physical abuse	0.99 (1.10)	0.76 (0.96)	t(376) = 1.92, ns, d = .22
Family with history of sexual abuse^a	0.51 (0.75)	0.30 (0.63)	t(135.2) = 2.40, p < .05, d = .30
Family with history of mental illness	0.29 (0.58)	0.33 (0.66)	t(380) = −0.63, ns, d = .06
Family with a criminal record^a	1.60 (1.74)	1.13 (1.06)	t(128.1) = 2.57, p < .05, d = .32
Criminal history
Convictions (ages 12-17)	12.78 (8.48)	13.14 (8.92)	t(399) = −0.35, ns, d = .04
Convictions (ages 18-29)	13.25 (13.22)	11.32 (12.26)	t(374) = 1.32, ns, d = .15
Days incarcerated (ages 12-17)	413.13 (321.05)	360.92 (301.50)	t(399) = 1.50, ns, d = .17
Days incarcerated (ages 18-29)	894.55 (1,040.04)	839.28 (1,044.78)	t(373) = 0.45, ns, d = .05

Levene’s test violated. Output for equal variances not assumed is shown.

Three items were examined to capture school-based risk factors, including whether the participant was enrolled in school prior to incarceration, whether they reported an early onset of skipping school, and the number of different schools they attended. Participants also self-reported whether they experienced any physical or sexual abuse during childhood or adolescence. Three different items captured aggressive tendencies of the participant, including self-reported weekly fighting, whether they thought they had a bad temper, and whether someone else had told them that they get angry easily. Family adversity was measured based on whether the participant had a history of placement in the care of MCFD (e.g., foster care, group home), whether the participant was living outside the home of both biological parents prior to incarceration, and six indicators of family adversity, defined by counts of the number of different types of family members (e.g., biological mother/father/sister/brother, etc.) with a history of alcohol abuse, drug abuse, physical abuse, sexual abuse, mental illness, and criminal record.

Schneider’s (1990) Good Citizen Scale was used to identify participants’ perceptions of themselves, as well as their perceptions of how others viewed them. Each scale includes 15 items. Principal components analyses (PCA) were performed for each scale. For self-perceptions, an initial PCA showed that two items, “wild/not wild” and “lazy/hardworking,” did not load onto any factor (factor loadings <.500). These two items were removed and a second PCA was performed. The second analysis resulted in one item (“poor/rich”) loading onto its own factor. This item was removed, and a third PCA with the 12 remaining items revealed three different components, which were named obedience, prosociality, and ego. Obedience was defined by the items “troublesome/cooperative,” “bad/good,” “breaks rules/obeys rules,” and “rude/polite.” Prosociality was defined by the items “unhelpful/helpful,” “dishonest/honest,” and “cruel/kind.” Ego was defined by the items “cowardly/brave,” “unattractive/attractive,” and “dumb/smart.”

Regarding the participants’ perceptions of how others viewed themselves, the initial PCA showed that two items, “wild/not wild” and “cowardly/brave” did not load onto any factor (factor loadings <.500). These two items were removed and a second PCA was performed. The result was three components, referred to as obedience, prosociality, and ego. Obedience was defined by the items “troublesome/cooperative,” “breaks rules/obeys rules,” “dumb/smart,” “mean/nice,” and “rude/polite.” Prosociality was defined by the items “bad/good,” “unhelpful/helpful,” “dishonest/honest,” and “cruel/kind.” The items that defined ego included “weak/tough,” “unattractive/attractive,” and “poor/rich.” Cronbach’s alpha values ranged across the six subscales (.39-.75), which was expected, given the small number of items (Cortina, 1993).

Measures of offending

Offending was measured using official data stored on a software system, CORNET, which contains information on an offender’s movement in and out of custody as well as the exact criminal offense, date of conviction, and sentence type received. For one participant, the criminal record in adolescence was sealed and thus offending between ages 12 and 17 was coded as missing. For another youth, the entire criminal record was sealed and thus this participant was excluded from the trajectory analysis. Adolescent and adult records were available for the remaining participants. Using CORNET, every criminal charge that resulted in a conviction was coded at each age from age 12, the age of criminal responsibility in Canada, to age 29. Although some participants were older than 30, criminal history information after age 29 was not coded to prevent differences in age affecting the interpretation of offending trajectories. Between ages 12 and 29, 15 offenders died (3.7% of the sample) and 11 (2.7%) moved outside the province (where data on offending were not available). For these offenders, convictions after the age of death or move were coded as missing. Frequency of convictions and number of days incarcerated during adolescence (12-17) and adulthood (18-29) are shown in Table 1.

Analytic Strategy

The analytic strategy proceeded in four steps. First, risk factors, protective factors, and criminal career parameters were compared across Indigenous and White youth to better understand the nature of differences, if any, across the two groups. Second, semiparametric group-based modeling (SPGM) was used to model offending trajectories between ages 12 and 29 for the full sample. Convictions at each person-period observation were used as the outcome of interest. Unlike cluster analysis and other techniques that identify groups ex ante, SPGM allows developmental trajectories to emerge from the data (Nagin, 2005). SPGM was performed using the STATA plugin. Exposure time was built into the model, which in the case of the current study was defined as the proportion of time spent in the community (as opposed to in a custody facility) at each person-period observation. Third, the resulting trajectories and their relationship with risk factors, protective factors, and demographic characteristics were examined at the bivariate level. Fourth, a series of MLR analyses were performed to (a) examine whether Indigenous status was predictive of association with a more serious offending trajectory, controlling for different risk and protective factors, and (b) examine whether risk and protective factors functioned differently across Indigenous and White participants to predict trajectory assignment.

Results

Risk and Protective Factor Comparisons Across Ethnicity

In Table 1, demographic characteristics and risk and protective factors from several domains of functioning (e.g., substance use, school behavior, aggression, self-identity, and family adversity) were compared across Indigenous and White participants. In no instances were these factors more common or more severe among White participants. However, there was a significant association between Indigenous status and physical abuse and sexual abuse (p < .05). As well, Indigenous youth averaged a significantly greater number of family members with a history of each of the following: alcohol abuse, drug abuse, sexual abuse, and a criminal record (p < .05). Despite being associated with a greater number of negative individual and familial circumstances, differences in the frequency of adolescent offending, adult offending, adolescent time incarcerated, and adult time incarcerated were not observed when compared between the two ethnic subgroups. Thus, traditional key risk factors were more common among Indigenous youth, yet did not result in the latter having more serious offending outcomes. That being said, these are rather simplistic measures of offending outcomes examined at only the bivariate level.

Trajectory Analyses

The first step in the SPGM analysis involved identifying the number and shape of the offending trajectories that best fit the sample.¹ A zero-inflated Poisson (ZIP) model with quadratic functional form was used to estimate the distribution of the offending trajectories. Based on Bayesian information criteria (BIC) values, a four-group quadratic model resulted in a BIC value closer to zero (–13,472.75) than both a three-group model (–13,714.42) and a five-group model (–15,011.28). The parameters of the four-group model are outlined in Table 2. Classification accuracy, based on the average posterior probability of accurately assigning individuals to a particular trajectory, was high for all four trajectories. Odds of correct classification (OCC) was used to help provide further confidence that individuals were assigned to the appropriate trajectory. As indicated in Table 2, OCC values were all greater than five, indicating high classification accuracy (Nagin, 2005). Very importantly, a potential ecological fallacy concern was that the nature of offending trajectory patterns differed between Indigenous and White youth. Given the disproportionate number of White participants, it was possible the trajectory groups identified were better representative of the offending patterns of White youth compared with Indigenous youth. However, using independent-samples t tests, inspection of posterior probabilities showed that for three of the trajectory groups, there were no differences in the posterior probabilities of correct assignment across ethnic subgroup (p > .05). For one trajectory group, Indigenous youth averaged a significantly (p < .05) higher posterior probability of correct assignment (.98) when compared with White youth (.94). As such, it did not appear that the disproportionate number of White participants in the sample affected the fit of the trajectory model to Indigenous participants.

Table 2.

Fit Statistics for a Zero-Inflated Poisson Model With Quadratic Functional Form (N = 403).

	Offending trajectories: Full adulthood (12-29)
	AL	Low rate stable	HRD	HRP
	(n = 108)	(n = 130)	(n = 57)	(n = 108)
Model parameters
Intercept	−23.46 (5.04)	1.97 (0.61)	−5.18 (0.77)	0.72 (0.34)
Linear	3.19 (0.63)	−0.00 (0.07)	0.90 (0.09)	0.12 (0.03)
Quadratic	−0.10 (0.02)	−0.00 (0.00)	−0.03 (0.00)	−0.00 (0.00)
Model fit
Median posterior probability	0.99	1.00	0.99	1.00
Average posterior probability	0.95 (0.10)	0.94 (0.13)	0.93 (0.13)	0.95 (0.12)
Range	0.51-1.00	0.51-1.00	0.50-1.00	0.44-1.00
OCC	18.82	15.55	13.05	18.82

Note. AL = adolescence limited; HRD = high rate desisters; HRP = high rate persisters; OCC = odds of correct classification.

The second step in the SPGM analysis involved interpreting the nature of each trajectory. Based on their shape, the trajectories shown in Figure 1 were labeled as follows: adolescent limited (26.8% of the sample), low rate stable (32.3%), high rate desisters (HRD; 14.1%), and high rate persisters (HRP; 26.8%). The adolescent limited group showed a pattern of desistance that began at age 17, with offending reaching a near-zero level by the early 20s. The low rate stable trajectory showed a similar pattern of offending in adolescence, but through the 20s showed a relatively stable, albeit low, pattern of offending. Part of the reason the groups appeared similar was because of the scale of Figure 1. Indeed, the adolescence limited (AL) trajectory averaged a significantly (p < .001) greater number of convictions compared with the low rate trajectory (17.72 [SD = 9.55] compared with 8.62 [SD = 5.64]). The difference was especially pronounced when examining the period of adulthood (ages 18-29; 6.97 [SD = 5.40] compared with 0.89 [SD = 1.42]; p < .001). The HRD group averaged the highest rate of offending in adolescence, but by the end of the study period was offending at a rate similar to the low rate stable group. The HRP group showed a pattern of offending in adolescence that was like the HRD group, but the HRP group continued to offend at a relatively high and stable rate through adulthood. Common among each of the four trajectories was that offending peaked in adolescence. The difference among each of the four trajectories was in terms of the pattern of offending in adulthood. These similarities and differences are unsurprising given that participants were recruited based on showing a pattern of serious and/or violent offending in adolescence, and thus, greater homogeneity in offending patterns was expected at this developmental stage.

Figure 1.

Trajectories of general offending measured from ages 12 to 29 for the full sample.

In Table 3, offender characteristics associated with each trajectory are shown. Self-reported Indigenous status was unrelated to trajectory-group membership. However, several risk and protective factors were informative of trajectory association. Chi-square analyses showed that being male, having a history of street drug use, and reporting fighting on a weekly basis were significantly associated with membership in a trajectory characterized by a higher rate of offending. On the contrary, being enrolled in school and reporting a history of sexual abuse were significantly associated with a lower frequency offending trajectory. A series of ANOVA analyses showed that, compared with the adolescent limited trajectory, individuals in the low rate stable, HRD, and HRP trajectories averaged significantly lower scores on measures of self-perception of obedience. At the group level, there were also significant differences in average scores on measures of participants’ perceptions of how others viewed their ego-based traits, although post hoc analyses showed that only one comparison trended toward being significant (p < .10), with the HRP group showing higher scores compared with the HRD group.

Table 3.

Association Between Offending Trajectories and Background Characteristics (N = 403).

	Offending trajectories				χ²/F, p, Φ/ω²
	ALM (SD)/% (n)	Low rate stableM (SD)/% (n)	HRDM (SD)/% (n)	HRPM (SD)/% (n)	χ²/F, p, Φ/ω²
Demographic characteristics
Indigenous	19.8% (21)	36.8% (39)	14.2% (15)	29.2% (31)	χ²(3) = 3.86, ns, Φ =.10
Male	20.4% (64)	30.6% (96)	17.5% (55)	31.5% (99)	χ²(3) = 44.60, p < .001, Φ = .34
Age at interview	16.39 (1.17)	16.34 (1.32)	16.30 (1.18)	16.37 (1.19)	F(3) = 0.82, ns, ω² = .01
Substance use
Substance use versatility	4.67 (2.15)	4.85 (2.26)	4.19 (2.25)	4.78 (2.31)	F(3) = 1.20, ns, ω² = .00
Street drug user	25.5% (49)	31.8% (61)	33.3% (64)	9.4% (18)	χ²(3) = 9.98, p < .05, Φ = .16
Drug use—before 12	23.4% (30)	35.9% (46)	13.3% (17)	27.3% (35)	χ²(3) = 2.08, ns, Φ = .08
Alcohol use—before 12	28.6% (32)	28.6% (32)	12.5% (14)	30.4% (34)	χ²(3) = 1.52, ns, Φ = .07
School behavior
Enrolled in school	33.8% (68)	27.9% (56)	13.9% (28)	24.4% (49)	χ²(3) = 10.44, p < .05, Φ = .16
Skip school—before 12	24.3% (17)	25.7% (18)	14.3% (10)	35.7% (25)	χ²(3) = 2.55, ns, Φ = .09
Number of different schools	5.45 (4.49)	6.43 (7.69)	6.66 (5.52)	6.58 (6.15)	F(3) = 0.67, ns, ω² = .03
Abuse
Physical abuse	24.4% (44)	35.0% (63)	16.7% (30)	23.9% (43)	χ²(3) = 4.07, ns, Φ = .10
Sexual abuse	32.6% (29)	39.3% (35)	11.2% (10)	16.9% (15)	χ²(3) = 8.12, p < .05, Φ = .14
Aggression and violence
Fighting—weekly basis	15.9% (13)	28.0% (33)	18.3% (15)	37.8% (31)	χ²(3) = 12.72, p < .01, Φ = .20
Angers easily	25.5% (47)	32.1% (59)	12.0% (22)	30.4% (56)	χ²(3) = 1.79, ns, Φ = .08
Bad temper	27.2% (68)	29.6% (74)	12.8% (32)	30.4% (76)	χ²(3) = 3.91, ns, Φ = .11
Identity
Self-perception—obedience	25.14 (4.75)^b,c,d	23.11 (5.09)^a,c,d	22.61 (4.84)^a,b,d	22.92 (5.41)^a,b,c	F(3) = 5.00, p < .01, ω² = .03
Self-perception—prosociality	15.53 (2.76)^c,d	14.55 (3.18)	13.93 (2.91)^a	14.35 (3.18)^a	F(3) = 4.42, p < .01, ω² = .03
Self-perception—ego	15.90 (2.70)	15.63 (2.73)	15.51 (2.47)	15.96 (2.72)	F(3) = 0.53, ns, ω² = .00
Others’ perception—obedience	27.37 (6.60)	25.73 (7.61)	26.32 (6.69)	25.25 (7.02)	F(3) = 1.55, ns, ω² = .00
Others’ perception—prosociality	16.22 (5.18)	16.42 (4.85)	15.36 (4.49)	15.89 (4.55)	F(3) = 0.67, ns, ω² = .00
Others’ perception—ego traits	12.72 (3.82)	12.80 (3.85)	12.39 (3.79)	14.02 (3.66)	F(3) = 2.99, p < .05, ω² = .02
Family problems
Ministry care	24.3% (28)	30.4% (35)	12.2% (14)	33.0% (38)	χ²(3) = 2.63, ns, Φ = .08
Living outside parents’ home	26.2% (55)	30.0% (63)	14.3% (30)	29.5% (62)	χ²(3) = 1.51, ns, Φ = .06
Family with history of alcohol abuse	1.36 (1.12)	1.75 (1.46)	1.68 (1.42)	1.65 (1.39)	F(3) = 1.72, ns, ω² = .01
Family with history of drug abuse	1.05 (1.14)	1.10 (1.16)	1.14 (1.21)	1.13 (1.03)	F(3) = 0.14, ns, ω² = .00
Family with history of physical abuse	0.82 (0.95)	0.90 (1.09)	0.70 (0.97)	0.78 (0.94)	F(3) = 0.60, ns, ω² = .01
Family with history of sexual abuse^e	0.34 (0.61)	0.42 (0.77)	0.35 (0.74)	0.30 (0.54)	F(3) = 0.55, ns, ω² = .00
Family with history of mental illness	0.40 (0.67)	0.31 (0.71)	0.33 (0.62)	0.25 (0.52)	F(3) = 1.03, ns, ω² = .00
Family with a criminal record^e	1.12 (1.02)	1.17 (1.16)	1.55 (1.78)	1.32 (1.35)	F(3) = 1.20, ns, ω² = .01

Note. AL = adolescence limited; HRD = high rate desisters; HRP = high rate persisters.

Significantly different from AL.

Significantly different from low rate stable.

Significantly different from HRD.

Significantly different from HRP.

Asymptotically F distributed, Welch statistic shown.

MLR Analyses

Demographic characteristics, all significant (p < .05) risk and protective factors from Table 3, and all factors that significantly differentiated Indigenous and White participants (see Table 1) were included in a MLR analysis with offending trajectory assignment as the outcome of interest. Most of the variables from Table 1 that differentiated Indigenous and White participants were related to problems within the family. As such, these variables were used to create an aggregate scale measuring total family adversity. Multicollinearity was not an issue as the strongest correlation between variables in the model was just .504 (the correlation between obedience and prosociality). Given the trend of differences in Table 3, the adolescent limited trajectory was used as the reference category to which the other three trajectories were compared. Gender was not included as a control variable due to the extremely low base rate of females in the HRD trajectory. As shown in Table 4, Indigenous status had no effect on the odds of trajectory-group membership. The odds of being in the HRP trajectory compared with the AL trajectory was approximately 2.5 times higher for youth with a history of street drug use. As well, youth who reported perceiving others as having a positive perspective on them (i.e., more attractive, rich, tough) were at higher odds of membership in the HRP trajectory. Youth enrolled in school at the time of their interview were significantly less likely to be in the HRP trajectory compared with the AL trajectory. Finally, the odds of membership in the HRD trajectory compared with the AL trajectory was approximately 3 times higher for participants reporting weekly fighting.

Table 4.

Multinomial Logistic Regression With Coefficients of Risk Factors by Trajectory Group (N = 403).

Covariates	Trajectories
	Low rate stable	HRD	HRP
	OR (95% CI)	OR (95% CI)	OR (95% CI)
Demographic characteristics
Indigenous	1.58 [0.69, 3.64]	0.88 [0.30, 2.64]	1.13 [0.46, 2.74]
Age at interview	0.88 [0.65, 1.18]	0.94 [0.64, 1.37]	0.84 [0.62, 1.15]
Significant predictors of offending
Street drug user	1.49 [0.71, 3.11]	0.54 [0.21, 1.41]	2.30 [1.06, 4.97]*
Enrolled in school	0.55 [0.26, 1.14]	0.78 [0.31, 1.99]	0.52 [0.24, 1.12]
Sexual abuse	1.52 [0.58, 4.00]	0.81 [0.23, 2.89]	0.51 [0.17, 1.57]
Fighting—weekly basis	0.87 [0.33, 2.27]	1.72 [0.58, 5.12]	1.11 [0.42, 2.94]
Self-perception—obedience	0.93 [0.85, 1.01]	0.91 [0.82, 1.02]	0.95 [1.00, 1.25]
Self-perception—prosociality	0.84 [0.72, 0.97]*	0.84 [0.69, 1.01]	0.85 [0.73, 0.99]*
Others’ perception—ego traits	1.04 [0.94, 1.15]	1.04 [0.91, 1.19]	1.19 [1.06, 1.33]*
Factors differentiating ethnic subgroups
Physical abuse	1.29 [0.59, 2.83]	1.98 [0.75, 5.25]	0.94 [0.41, 2.17]*
Family adversity	1.09 [0.98, 1.21]	1.12 [0.98, 1.27]	1.12 [1.00, 1.25]*
Model fit	LL = −298.77, χ² = 71.33, df = 33, p < .001

Note. Adolescence limited trajectory group is reference category. HRD = high rate desisters; HRP = high rate persisters; OR = odds ratio; CI = confidence interval; LL = log likelihood.

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

A second set of analyses were performed to examine whether ethnicity moderated the relationship between key risk factors and offending trajectories. All risk factors from Table 4 were tested for interaction effects. Thus, the principal interest within these analyses was to examine, not simply whether a given risk factor predicted an offending outcome, but whether the impact of a given risk factor on offending was contingent on a participant’s ethnic background. For the nominal variables included in the analysis, the reference category was always “White × Absence of Risk/Protective Factor.” Continuous variables were mean-centered before creating the interaction term, except for the measure of family adversity as this variable had a meaningful zero value (see Dalal & Zickar, 2012). Nine different models were run (see Table 5). In these models, the interaction term was tested while controlling for all other demographic characteristics and risk factors; however, only the interaction coefficient is shown.

Table 5.

Interaction Effects Between Ethnicity and Key Risk Factors (N = 403).

	Trajectory association as the outcome of interest
	M1	M2	M3	M4	M5	M6	M7	M8	M9
Interaction term
Street Drug User × Ethnicity	HRP > AL^a,b	—	—	—	—	—	—	—	—
Enrolled in School × Ethnicity	—	LRS < AL^c	—	—	—	—	—	—	—
Physical Abuse × Ethnicity	—	—	ns	—	—	—	—	—	—
Sexual Abuse × Ethnicity	—	—	—	ns	—	—	—	—	—
Fighting—Weekly Basis × Ethnicity	—	—	—	—	ns	—	—	—	—
Family Dysfunction × Ethnicity	—	—	—	—	—	HRP > AL^d	—	—	—
Self-Perception—Obedience × Ethnicity	—	—	—	—	—	—	ns	—	—
Self-Perception—Prosociality × Ethnicity	—	—	—	—	—	—	—	LRS < AL^e	—
Others’ Perception—Ego Traits × Ethnicity	—	—	—	—	—	—	—	—	ns

Note. AL trajectory group is reference category. Greater/lesser than signs indicate whether odds of trajectory-group membership were higher or lower relative to the reference category. HRP = high rate persisters; AL = adolescence limited; LRS = low rate stable; ns = nonsignificant interaction effects (p ≥ .05).

The odds of being in the HRP trajectory compared with the AL trajectory were 2.76 times higher for White youth with a history of street drug use versus White youth with no history of street drug use (p < .05).

The odds of being in the HRP trajectory compared with the AL trajectory were 3.56 times higher for Indigenous youth with a history of street drug use versus White youth with no history of street drug use (p < .01).

The odds of being in the LRS trajectory compared with the AL trajectory were 59% less likely for White youth attending school versus White youth not attending school (p < .05).

For Indigenous participants, a one-unit increase in family dysfunction resulted in a 15% increase in the odds of membership in the HRP trajectory compared with the AL trajectory (p < .05).

For White participants, a one-unit increase in mean-centered perceptions of prosociality resulted in a 20% decrease in the odds of membership in the LRS trajectory compared with the AL trajectory (p < .05).

Some significant interaction effects were observed for both Indigenous and White youth, whereas other interaction effects suggested that there were risk factors that functioned differently across Indigenous and White youth. Regarding the former, both Indigenous and White participants that reported street drug use were at significantly higher odds than White youth with no such history to be associated with the HRP trajectory compared with the AL trajectory. Some risk factors were specific to White youth. White youth attending school were significantly less likely than White youth not attending school to be in the low rate stable trajectory compared with the AL trajectory. As well, for White participants only, a one-unit increase in their mean-centered perceptions of their level of prosociality resulted in a 20% decrease in the odds of membership in the low rate stable trajectory compared with the AL trajectory (p < .05). In effect, school attendance and a more positive sense of self were protective factors for White youth but not Indigenous youth. Finally, family dynamics appeared particularly important for Indigenous youth. The interaction between family adversity and ethnicity revealed that, for Indigenous participants, a one-unit increase in level of family adversity resulted in a 15% increase in the odds of membership in the HRP trajectory compared with the AL trajectory. This relationship was not observed for White participants.

Discussion

The current study represented another step toward understanding the unfolding of criminal careers of Indigenous youth, the risk and protective factors associated with different patterns of offending among this group, and whether these factors differed when compared against White youth. Using a sample of adjudicated youth from British Columbia, all of whom were interviewed during a period of incarceration in adolescence and followed through age 29, the current study uncovered three important findings: (a) Indigenous youth were characterized by a greater number of social adversities compared with White youth, and in no instance were different risk factors more prevalent or prominent for White youth, (b) despite the conceptually greater likelihood of persistent offending for Indigenous youth given these risk factor differences, Indigenous and White youth did not differ in their association with a particular offending trajectory measured from ages 12 to 29, and (c) a series of interaction analyses revealed that the risk and protective factors informative of a more frequent offending pattern varied across Indigenous and White youth, with school attendance and higher levels of prosociality acting as protective factors for White youth only and familial adversity acting as a risk factor for Indigenous youth only.

Regarding family adversity, this measure examined the number of different types of family members that had experienced different adversities (substance abuse, abuse, mental health diagnoses, and criminal record). Indigenous youth scored significantly higher than White youth on this measure, and this measure also predicted a higher frequency of offending for Indigenous youth but not White youth. This finding is important to consider in the context of (a) the intergenerational transmission of adversity that is particularly salient for Indigenous Peoples and (b) the important role of the extended family within Indigenous cultures when it comes to helping raise children (Whitbeck et al., 2014). In effect, a key mechanism for helping raise children (i.e., through the extended family) has been jeopardized due to the plethora of adverse experiences that are more likely to characterize Indigenous Peoples. Consequently, it is possible that Indigenous children experience greater exposure to these family adversities, and that these familial adversities may influence offending in a variety of ways. For example, the high level of adversity may affect time and energy to help raise children and adolescents, may operate on the child or adolescent via social learning mechanisms, and/or may increase feelings of shame and emotional trauma within the child or adolescent.

In addition to factors that were disproportionately present within Indigenous participants that influenced offending, we also observed background factors that, despite being considered risk factors for offending (e.g., Catalano & Hawkins, 1996) and more common to Indigenous youth, did not increase the risk of persistent offending for this group. This observation has implications for risk assessment tools such as the SAVRY (e.g., Borum, Bartel, & Forth, 2002) because Indigenous youth may score higher on such tools compared with White youth despite not actually being at a higher risk of persistent offending. Failing to ensure that a risk assessment tool is cross-culturally valid may adversely and unethically (see Hart, 2016) affect Indigenous adolescents in conflict with the law.

Limitations and Future Research

The current study included a group of Indigenous youth, all of whom were incarcerated during a period of adolescence. Although offender samples help address key policy questions (Piquero, 2014), the incarcerated youth that the current study sampled from are likely different from the typical youth in the community as well as youth on probation. Future research can build upon some of the limitations of the measures used in the current study. For example, although neither physical abuse nor sexual abuse was related to persistent offending, Lowenkamp, Holsinger, and Latessa (2001) noted the importance of capturing different aspects of abuse (e.g., age of onset, severity, duration) in predicting offending. Furthermore, Stewart et al. (2008) found that early abuse experiences were related to more frequent offending during adolescence. The current study lacked measures of delinquent peer association, which Sittner and Hautala (2016) observed to be an important risk factor for chronic offending during adolescence among Indigenous youth. Moreover, given that socioeconomic markers of disadvantage are disproportionately more common among Indigenous Peoples, other measures of structural disadvantage such as poverty and neighborhood-level crime rates should be included in future research. Also missing from the current study were measures concerning Indigenous heritage (e.g., First Nations or Metis, membership to a specific band) that might reveal important within-group differences in Indigenous youth risk factor profiles and associated offending patterns. The inclusion of such factors may be helpful in developing more culturally informed risk assessment tools. Finally, family adversity was an important risk factor for Indigenous youth, but the specific dynamics of this risk factor were not examined (e.g., how this informed interactions between family members and child). Future research should examine the mechanisms in which family adversity influenced offending. Future research is also needed that examines the same research questions focused on in the current study, but with the use of existing risk assessment tools to evaluate the potential of the justice system to make biased decisions about Indigenous youth sentencing.

Footnotes

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: This work was supported by the Social Sciences and Humanities Research Council of Canada (435-2016-0962).

Notes

Author Biographies

Evan C. McCuish is an assistant professor in the School of Criminology at Simon Fraser University and is the project director of the Incarcerated Serious and Violent Young Offender Study, the largest and longest running study on young offenders in Canada. His research interests include criminal careers, desistance, developmental criminology, foster care, gang involvement, psychopathy, sexual offending, and violence. His work is published in the International Journal of Forensic Mental Health, Sexual Abuse: A Journal of Research and Treatment, Justice Quarterly, and Journal of Criminal Justice.

Raymond R. Corrado is a full professor in the School of Criminology at Simon Fraser University and was an associate faculty member in the psychology department and the faculty of health sciences. He is a visiting fellow at Clare Hall College and the Institute of Criminology, University of Cambridge. He is a founding member of the Mental Health, Law, and Policy Institute at Simon Fraser University. He has published more than 100 articles and book chapters on a wide variety of policy issues, including juvenile justice, violent young offenders, mental health, adolescent psychopathy, Aboriginal victimization, child/adolescent case management strategies, and terrorism.

References

Borum

Bartel

Forth

(2002). SAVRY: Manual for the structured assessment of violence risk in youth. Tampa: Florida Mental Health Institute, University of South Florida.

Catalano

R. F.

Hawkins

J. D.

(1996). The social development model: A theory of antisocial behavior. In Hawkins

J. D.

(Ed.), Delinquency and crime: Current theories (pp. 28-67). New York, NY: Cambridge University Press.

Cortina

J. M.

(1993). What is coefficient alpha? An examination of theory and applications. Journal of Applied Psychology, 78, 98-104.

Cross

(2008). Native Americans and juvenile justice: A hidden tragedy. Washington, DC: Poverty & Race Research Action Council. Available from http://www.prrac.org

Dalal

D. K.

Zickar

M. J.

(2012). Some common myths about centering predictor variables in moderated multiple regression and polynomial regression. Organizational Research Methods, 15, 339-362.

Dawkins

M. P.

(1997). Drug use and violent crime among adolescents. Adolescence, 32, 395-405.

Duran

(1995). Native American postcolonial psychology. Albany: State University of New York Press.

Ford

J. D.

(2012). Indigenous health and climate change. American Journal of Public Health, 102, 1260-1266.

Gadermann

A. M.

Guhn

Zumbo

B. D.

(2012). Estimating ordinal reliability for Likert-type and ordinal item response data: A conceptual, empirical, and practical guide. Practical Assessment, Research & Evaluation, 17, 1-13.

10.

Hart

S. D.

(2016). Culture and violence risk assessment: The case of Ewert v. Canada. Journal of Threat Assessment and Management, 3, 76-96.

11.

Indian Affairs. (2016). Roberts announces publication of final, updated BIA Model Indian Juvenile Code. U.S. Department of the Interior. Retrieved from https://www.bia.gov/cs/groups/public/documents/text/idc2-050414.pdf

12.

LaPrairie

(2002). Indigenous over-representation in the criminal justice system: A tale of nine cities. Canadian Journal of Criminology, 44, 181-208.

13.

Lowenkamp

C. T.

Holsinger

A. M.

Latessa

E. J.

(2001). Risk/need assessment, offender classification, and the role of childhood abuse. Criminal Justice and Behavior, 28, 543-563.

14.

Maurutto

Hannah-Moffat

(2007). Understanding risk in the context of the Youth Criminal Justice Act. Canadian Journal of Criminology and Criminal Justice, 49, 465-491.

15.

Nagin

D. S.

(2005). Group-based modeling of development. Cambridge, MA: Harvard University Press.

16.

Pavkov

T. W.

Travis

Fox

K. A.

King

C. B.

Cross

T. L.

(2010). Tribal youth victimization and delinquency: Analysis of youth risk behavior surveillance survey data. Cultural Diversity & Ethnic Minority Psychology, 16, 123-134.

17.

Piquero

A. R.

(2014). Moving from description to implementation of evidence-based research findings. Criminology & Public Policy, 13, 127-134.

18.

Piquero

A. R.

(2015). Understanding race/ethnicity differences in offending across the life course: Gaps and opportunities. Journal of Developmental and Life-Course Criminology, 1, 21-32.

19.

Reingle

J. M.

Maldonado-Molina

M. M.

(2012). Victimization and violent offending: An assessment of the victim-offender overlap among Native American adolescents and young adults. International Criminal Justice Review, 22, 123-138.

20.

Rempel

E. S.

Somers

J. M.

Calvert

J. R.

McCandless

L. C.

(2015). Diagnosed alcohol dependence and criminal sentencing among British Columbian Aboriginal offenders. Drug and Alcohol Dependence, 154, 192-198.

21.

Roberts

J. V.

Reid

(2017). Aboriginal incarceration in Canada since 1978: Every picture tells the same story. Canadian Journal of Criminology and Criminal Justice, 59, 313-345.

22.

Schneider

A. L.

(1990). Deterrence and juvenile crime. New York, NY: Springer.

23.

Shepherd

S. M.

Adams

McEntyre

Walker

(2014). Violence risk assessment in Australian Indigenous offender populations: A review of the literature. Psychology, Public Policy, and Law, 20, 281-293.

24.

Sittner

K. J.

Hautala

(2016). Aggressive delinquency among North American Indigenous adolescents: Trajectories and predictors. Aggressive Behavior, 42, 274-286.

25.

Statistics Canada. (2013). British Columbia (Code 59) (table). National Household Survey Aboriginal Population Profile. Ottawa, Ontario: Author. Available from www.statcan.gc.ca

26.

Stewart

Livingston

Dennison

(2008). Transitions and turning points: Examining the links between child maltreatment and juvenile offending. Child Abuse & Neglect, 32, 51-66.

27.

Thorndike

E. L.

(1939). On the fallacy of imputing the correlations found for groups to the individuals or smaller groups composing them. American Journal of Psychology, 52, 122-124.

28.

Whitbeck

L. B.

Adams

G. W.

Hoyt

D. R.

Chen

(2004). Conceptualizing and measuring historical trauma among American Indian people. American Journal of Community Psychology, 33, 119-130.

29.

Whitbeck

L. B.

Sittner Hartshorn

K. J.

Walls

(2014). Indigenous adolescent development: Psychological, social and historical contexts. New York, NY: Routledge.

30.

Yessine

A. K.

Bonta

(2009). The offending trajectories of youthful Aboriginal offenders. Canadian Journal of Criminology and Criminal Justice, 51, 435-472.