The prediction of recidivism with Aboriginal offenders: a theoretically informed meta-analysis.

• Author: Gutierrez, Leticia
• Position: Canada

In Canada, Aboriginal peoples are significantly over-represented among offender populations. Nationally, Aboriginal peoples represent 3% of the adult Canadian population but account for 18-21% of admissions to custody and probation (Calverley 2010). The proportion of Aboriginal offenders in the federal/provincial/territorial correctional systems varies widely, with the highest rates found in the prairies and the northern territories. For example, in Saskatchewan, 11% of the general population is Aboriginal yet Aboriginal offenders represent 81% of sentenced admissions to provincial prisons (Perreault 2009). At the federal level, Aboriginal offenders constitute 17.9% of the total offender population, with Aboriginal women comprising 32.6% of incarcerated women (Public Safety Canada 2010). Furthermore, the rate of Aboriginal over-representation among the offender population has been growing (Perreault 2009).

The reasons for the over-representation of Aboriginal peoples among offender populations are many and have been discussed by scholars, political pundits, and governments for years (e.g., Canada 1996; Welsh and Ogloff 2008). The purpose of this article is not to add to this literature but to deal with one of the consequences of this over-representation, and that is the necessity to manage the growing numbers of Aboriginal offenders in a humane, rational, and effective manner. Correctional systems are entrusted by the public with the responsibility of housing offenders in facilities suitable to the offender's risk for disruptive and dangerous behaviour and to supervise them in the community at levels of supervision that balance community safety with the least restrictive intervention. Correctional systems are also expected to facilitate the rehabilitation of offenders. These mandates require the use of risk/need offender assessments, which should be comprised of factors that have demonstrated applicability to the populations with which they are used.

Towards a theoretically informed assessment of offender risk

The process of offender risk assessment involves judgements of the relevance of certain psychosocial and situational characteristics to future criminal offending. The major challenges in this process are, first, selecting the factors that may predict criminal behaviour, and then, demonstrating their predictive validity. The latter challenge is relatively straightforward and involves the measurement of the potential risk factors at Time 1 and linking these measures to criminal behaviour at Time 2. The first challenge, selecting the factors that may be predictive of criminal behaviour, is usually addressed in one of two ways. The first approach, "dustbowl empiricism," takes whatever information is readily available and then tests whether any of the factors are statistically associated with recidivism (Bonta 1997). The second approach, "theoretically informed," is to consider factors that are drawn from a conceptual understanding of criminal behaviour.

As any casual perusal of an introductory criminology textbook will attest, there are many theories of criminal conduct. However, most of the theories can be subsumed under one of three broad theoretical perspectives on criminal behaviour (Bonta 2002). First, there is sociological criminology, where the theories trace the causes of crime to a person's location within the social structure. For example, membership in a disadvantageous group (young, poor, racial/ethnic minority) is viewed as a risk factor. Despite the pervasiveness of sociological criminology's perspective on crime among academia and the public, the extant evidence shows that these factors are relatively minor predictors of criminal behaviour (Andrews and Bonta 2010; Gendreau, Little, and Goggin 1996). The second perspective, forensic mental health, posits criminal behaviour to be a consequence of psychological pathology (e.g., neurotic, low self-esteem, schizophrenic). Once again the evidence indicates that, except for antisocial personality and psychopathy, such factors are minor risk variables (Andrews and Bonta 2010; Bonta, Law, and Hanson 1998; Gendreau et al. 1996; Hanson and Morton-Bourgon 2005).

The third broad perspective is cognitive, social learning theory. One of the theories under this perspective is the General Personality and Cognitive Social Learning (GPCSL) model described by Andrews and Bonta (2010). The model incorporates distal and biosocial factors such as neighbourhood and race/ethnicity along with more proximal variables that influence the probability of criminal behaviour. The more proximal influences are the signalled rewards and costs in the immediate situation along with contextual and personal central eight risk/ need factors. Four of these risk/need factors, described as the big four, have the most direct and immediate influence on criminal behaviour. They are criminal history (reflecting behavioural habits), pro-criminal attitudes, pro-criminal associates, and antisocial personality pattern (e.g., impulsive, egocentric, feelings of hostility). Rounding out the central eight are the more moderate risk/need factors of employment/ education, family/marital, substance abuse, and leisure/recreation. These latter four risk/need factors exert their effect through the big four. For example, abusing drugs may lead to involvement with pro-criminal associates, and being raised by criminal parents may enhance the learning of pro-criminal attitudes. Meta-analytic summaries of these risk/need factors for general offenders find a mean effect size (r) of 0.26 for the big four and 0.17 for the remaining four risk/need factors of the central eight (Andrews and Bonta 2010: 65).

The GPCSL perspective is a general theory of criminal behaviour. In other words, the central eight risk/need factors are hypothesized to be relevant across offender types. Research has demonstrated that many of the risk/need factors drawn from the central eight appear relevant to women offenders (Smith, Cullen, and Latessa 2009), youth (Schwalbe 2009), mentally disordered offenders (Bonta et al. 1998), and sexual offenders (Hanson and Morton-Bourgon 2005, 2009). Further research is needed to strengthen this pattern of results among these offenders, and we have little knowledge of the applicability of the central eight to Aboriginal offenders.

Risk assessment with Aboriginal offenders

Actuarial risk assessment instruments have been available since Burgess's pioneering work in the 1920s (Burgess 1928) but widespread use of them was not seen until the 1980s. Today, many correctional systems around the world use some type of actuarial, evidence-based offender risk instrument. The development of these instruments has been based largely on Caucasian male offenders (e.g., the level of service instruments; Andrews 1982; Andrews and Bonta 1995; Debidin 2009; Hoffman 1996). Not surprisingly, when the instruments are applied to groups that differ significantly from the construction samples, they are criticized for making the assumption that the instruments, and the risk items comprising the instrument, would be equally valid to the new group. This criticism is prevalent among feminist scholars (Hannah-Moffat and Shaw 2001; Holtfreter, Reisig, and Morash 2004), and it has also been raised by those questioning their validity among African-American and Hispanic minority groups (Whiteacre 2006). This is clearly an issue in Canada, Australia, and New Zealand where Aboriginal peoples constitute a significant racial/ethnic minority.

In Canada, risk assessment instruments developed on non-Aboriginal male offenders are, for the most part, also administered to Aboriginal offenders. Critics have suggested that using these instruments introduces a cultural bias that may negatively and unfairly affect the risk assessment of Aboriginal offenders (LaPrairie 1995; Martel, Brassard, and Jaccoud 2011; Waldram 1992). Many risk instruments measure factors such as employment stability, educational achievement, and substance abuse; and while Aboriginal offenders may have more of these risk factors than non-Aboriginal offenders, the reason is that many Aboriginal communities have high rates of poverty and unemployment as a consequence of a long history of social marginalization. Nevertheless, risk instruments constructed on non-Aboriginal offenders are used in several correctional jurisdictions across Canada and Australia. In Canadian provincial jurisdictions, there are four major instruments used for adult offender classification. They are the community risk/ needs assessment (CRNA; British Columbia), the service planning instrument (SPIn; Alberta), the primary risk assessment (PRA; Saskatchewan) and, for the remaining provinces and territories, the level of service (LS) instrument (Wormith, Ferguson, and Bonta in press). For young offenders, a youth version of the LS instrument is the most common. All of these offender assessment instruments tap into indicators of most if not all of the central eight risk/need factors. Presently, research with these assessment tools is limited to the CRNA, PRA, and the LS (Alberta only implemented the SPIn in 2009).

All three instruments have demonstrated predictive validity with Aboriginal offenders. In a study of 374 males and 376 female Aboriginal probationers, the CRNA predicted recidivism over a 4-year follow-up (r = 0.30; British Columbia 2004). Furthermore, there were no statistically significant differences between male and female Aboriginal offenders in the predictive validity of the CRNA, leading to the conclusion that the risk factors for Aboriginal and non-Aboriginal offenders were similar. For the PRA, Bonta, LaPrairie, and Wallace-Capretta (1997) found that PRA intake scores predicted recidivism for Aboriginal offenders (r = 0.23, N = 390). They further broke down their sample into three groups: (1) M6tis (N = 153), (2) Aboriginal offenders with registered Indian status and...