Assessment of the risk of reoffending by Indigenous male violent and sexual offenders

Foreword | Modern corrections practice requires an assessment of the risk of reoffending on at least two levels. First, risk assessments are necessary to decide which offenders should be targeted for rehabilitation. Second, risk assessment is necessary to deal with the increasing demand by the public and politicians that offenders who are at a high risk of reoffending, especially violent and sexual offending, should not be released prematurely. This paper reports on a risk assessment tool that has been developed specifically for Indigenous offenders. However the research has found that further work is required on developing tools for the various sub-groups of violent and sexual offenders.

Toni Makkai

Acting Director

The accurate assessment of offenders' risk of reoffending is a very important task in corrections. If risk is over predicted it will not only be a waste of expensive resources but the human rights of people are infringed if they are kept in preventative detention (Monahan 1981) while there are other appropriate methods of managing them in the community. This is of particular concern in respect to Indigenous people in Australia - it is well documented that they are overrepresented in the Australian criminal justice system - and that the situation is deteriorating (Carcach et al. 1999). On the other hand, the erroneous classification of an offender as at a low risk of reoffending can have grave physical consequences for the future victims of the person, as well as social and political consequences.

However, there have been reservations in the past about the assessment of the risk of harmful behaviour, in particular the risk of violent behaviour. Traditionally the assessment of the risk of violence was based on clinical judgement. In the late 1960s researchers demonstrated that the clinical judgement of the risk of violence (see for example, Monahan 1981) is inadequate and that actuarial assessments consistently match or surpass the accuracy of clinical assessments (Grove & Meehl 1996). This has more recently also been demonstrated in respect of sexual violence (see for example, Hanson & Bussiere 1998).

Consequently there has, since the 1970s, been a call for the development of actuarial tools based on statistics, and fixed and explicit algorithms that could be used to augment, or preferably replace, the clinical assessment of the risk of violence (Monahan 1991). In response a number of researchers and research teams developed assessment tools using primarily multivariate statistical methods, such as main effects linear regression and discriminant function analysis (Quinsey et al. 1998). Initially these instruments relied heavily on static predictors (Andrews & Bonta 1994), that is historical factors that cannot be changed through intervention, for example age, past criminal history and early family factors. However, while static factors are useful for assessing long-term risk potential, they are limited because they do not provide information that can be used to plan interventions to reduce a person's risk of reoffending. Nor can instruments based solely on them be used to determine whether change has taken place since a previous assessment.

Therefore researchers started taking clinical information into account in the form of dynamic risk factors, that is, factors that may be susceptible to change. Day and Howells (2002), for example, distinguish between two types of dynamic factors. They are criminogenic factors that are directly related to the offending (such as substance abuse and pro-offending attitudes and beliefs) and non-criminogenic factors (such as self-esteem, anxiety, unemployment) that in themselves may not be the cause of offending, but may explain offending to some extent. What is important about dynamic factors is that they can be targeted during interventions, for example self-esteem problems can be addressed, and it is possible to measure whether the intervention brought about the desired change. Dynamic factors are therefore included in most risk assessment instruments today.

Two important definitions

  • Reoffending was defined as being found guilty in a court of a similar subsequent offence
  • Time at risk was calculated from time of release into the community, or in the case of a community order, the commencement of the order

The problem

There is currently a range of risk assessment instruments available, but with some notable exceptions, they are based on North American research. The dilemma that faces assessors in Australia is that because of the over-representation of Indigenous people in the criminal justice system, they must assess a relatively large number of offenders from this group. Yet, to the best of our knowledge, no instrument specifically aimed at assessing the risk of reoffending by violent and sexual Indigenous offenders has been developed in Australia. Assessors must therefore use one of the international instruments. Alternatively, they could use one of the locally developed instruments produced for other purposes (e.g. general recidivism of offenders in the community) or non-Indigenous populations.

For a number of reasons we believe that it is inappropriate to use these risk assessment instruments with Australian Indigenous people, even if adapted, without a more thorough investigation.

Firstly, the literature (see for example Mals et al. 2000) suggests that the risk factors for Indigenous violence may differ from that in other culture groups. This was confirmed by a series of focus groups with people (including Indigenous people) working in the corrections field that we undertook prior to this study (Dawson & Allan 2000, 2001).

Secondly, available instruments with a clear theoretical basis invariably use constructs that are embedded in Western culture. For instance, the Psychopathy Checklist-Revised (PCL-R) (Hare 1991) that is frequently used to assess the risk of violent reoffending, and also forms part of a number of other risk assessment instruments, is based on the construct psychopathy as defined by Cleckley (1976) and Hare (1980). The construct of psychopathy may be valid in Australian Indigenous culture, but we are not aware of any research that has examined its existence, or usefulness, in Indigenous communities.

Thirdly, because the existing instruments are based on regression models they use predictors obtained from specific populations. The accuracy of such a risk instrument is dependent on risk markers that best characterise the population of interest. Care should be taken when these instruments are used to assess the risk of reoffending of people from other populations or cultures. We would even be reluctant to uncritically adopt an instrument that had been developed using a North American Indigenous group, as different Indigenous groups, even though they may share many commonalities, are still unique and distinct culture groups.

Additionally, even if these instruments prove to be very accurate in assessing the risk of reoffending of Indigenous males, it is always possible that there may be predictors, unique to Indigenous people, that are not included in them. Using these instruments may also limit our understanding of the dynamics of violent and sexual offending amongst Indigenous males, and restrict the development of strategies to prevent or reduce such offending.

Another reservation is the stage of development of risk assessment instruments. Even though there is some debate about this, it appears as if much work still has to be done before the existing instruments will be in a form that will be practical and satisfy ethical and legal expectations (Gendreau et al. 2002; Rogers 2000).

Finally, it is well known that Indigenous people question - with good reason - the appropriateness of using instruments and methods developed in other regions, (not to mention other countries and cultures), with Indigenous offenders (Worrall 2000). In fact, it can be questioned whether research findings about Indigenous people in a specific geographical location within Australia can be generalised to those in another area. A number of factors make it possible that there could be significant differences between the various Indigenous Australian groups. For example, Indigenous groups in Victoria and New South Wales had contact with Europeans for a much longer time than those in the Kimberley area. To ensure a risk assessment instrument used with Indigenous people is credible it is important that its development should take geographical and cultural factors into account.

Therefore, while there is a high regard for the available instruments and the research that underlies them, we believe that it would be inappropriate to use them, even if adapted, with Indigenous people without a more thorough investigation. Conceivably, it is possible that the available instruments may accurately assess the risk of violence by Indigenous males, but it may be that the appropriate approach is to assume that this is not the case until the contrary is demonstrated.

The study

The purpose of this study (Allan & Dawson 2002) was to identify factors that predict whether Indigenous adult male violent and sexual offenders in Western Australia (WA) would reoffend violently or sexually respectively, and to use these findings to construct a risk assessment instrument specific for them. A retrospective analysis was done of the file data of 1,838 adult male Western Australian Indigenous offenders who had since 1987 been identified as requiring either a violent or sexual offender program, including those who did not engage in or complete a program. The data of 525 offenders were used to isolate possible predictors and the data of 380 were used to build the various instruments. We examined the post release records of 258 offenders to determine the accuracy of the risk prediction instruments developed in the study.

The data of the other 675 offenders could not be used because of attrition due to death (the mortality rate of ex-prisoners is relatively high, see for example, Graham 2003), inability to confirm the ethnicity of offenders, and the unavailability of files.

It is important to acknowledge our awareness of the concerns about research in respect of Indigenous people (see for example, Smith 1999; Worby & Rigney 2002; Australian Psychology Society 2002). Therefore it was understood that the study and its results would only be accepted as reliable by the Indigenous community if it was designed in consultation with credible representatives of the Indigenous community who are well informed about the project. To complicate the situation, the assessment of risk of violent and sexual reoffending by Indigenous males is in itself controversial, and the community could easily perceive the research as prejudicial and discriminating.

It was therefore decided to involve an Indigenous advisory group to steer us through the whole research process. As the study involved the files of people across WA we could not only involve people from one specific community, but had to ensure that the group would represent Indigenous people from across the state. We were fortunate that a number of key Indigenous people with good knowledge of the corrections area initiated a process on our behalf that culminated in the formation of the group.

The group was involved in the research in a number of ways. At one level it guided the design of the project to ensure that the research findings would be relevant to the needs, problems and issues of the Indigenous communities, and contribute to the quality of life in these communities. The group also assisted in practical ways. For example, the use of archival file material meant that we had to ensure that cultural norms were not violated when selecting files (for example not using files of the deceased) and that we interpreted the file data within a cultural context (for example taking into account payback behaviour). The group suggested that Indigenous research assistants be employed and helped to identify appropriate people. Fong et al. (2002), who were members of the group, presented their experiences at an Australian Institute of Criminology conference.

The results

During the course of the study it became clear that male Indigenous violent and sexual offenders in WA formed two distinct groups and that it would serve no purpose to identify predictors for them as a joint group. The two groups were therefore separated for the purposes of this study with violent offenders, who had no record of sexual offences, forming a separate group from the sexual offenders. The sexual offender group included both violent and non-violent sexual offenders.

When the violent offenders were examined further we found that family violence offenders formed a discrete subgroup within the violent offender group and that there were also distinct differences between the violent sexual and non-violent sexual offenders. Ideally different instruments should be developed for each of these groups but the small number of family violence offenders and violent sexual offenders in the sample prevented us from doing this.

The instrument developed to assess the risk of violent reoffending (including family violence) was disappointing because while it correctly identified 95.4 per cent of the violent offenders who reoffended, the false positive rate was 55 per cent. The latter statistic is poor because it means that had the instrument been used, more than half of those who did not reoffend would either have been detained unnecessarily or may have needlessly undergone expensive rehabilitation programs.

For sexual offenders (violent and non-violent) the three factors that best predicted sexual reoffending are unrealistic long-term goals, unfeasible release plans, and poor coping skills prior to release. Both the true positive rate (92.3%) and the false positive rate (5.7%) of this 3-Predictor model are very good. This means that if the instrument is applied to a group of Indigenous male sexual offenders in WA it will identify 92 per cent of those who will reoffend correctly while only six per cent of non-reoffenders will be incorrectly identified as at risk of reoffending. The latter is very important in the context of Indigenous offenders because this will ensure that those who are not at risk of reoffending will not be kept in custody unnecessarily.

When we compared the 3-Predictor model with instruments currently used in WA, it outperformed all of the instruments in terms of predictive accuracy for both reoffenders and non-reoffenders. Though it should again be noted that the other instruments were either developed overseas or were local instruments that were neither developed for Indigenous offenders in particular, nor to assess the risk of sexual reoffending.


The study indicates that a risk assessment instrument developed specifically for Indigenous sexual offenders can be more accurate than instruments that were developed overseas, or for non-Indigenous populations in Australia, or to assess the risk of general recidivism. The overall accuracy of the 3-Predictor model is excellent, but it must be kept in mind that this was a retrospective study that used a relatively small sample and that the instrument cannot necessarily be used in other areas in Australia. Ideally this study should be followed by a prospective study in WA which uses the predictors identified during this study with a large sample. In practical terms, this will require the use of the 3-Predictor model to assess the risk of offenders currently in the system, and following them up for a substantial period of time after release to see whether the assessment made by the instrument was accurate. This will have to be a longitudinal study to ensure a large enough sample and a long enough period at risk, as offenders have less opportunity to reoffend while in custody and may be constrained while subject to community based orders. Once the predictive accuracy of the 3-Predictor model has been confirmed these studies should be replicated in other locations in Australia.

Unfortunately, this study suggests that it will not only be necessary to develop separate instruments for violent and sexual offenders, but also for subgroups of these two groups. Despite the predictive accuracy of the 3-Predictor model our findings suggest that separate instruments should be developed for sexual offenders who are violent and those who are non-violent as they appear to be two distinct groups. Our tentative findings in respect of the violent offender group likewise suggest that the profile of offenders who commit family violence differs from the profile of those who commit general violence. This is in accordance with the literature (see for example, Blagg 2000; Mals et al. 2000) and the findings of our focus groups with service providers in correctional settings (Dawson & Allan 2000, 2001). The implication of this is that it will be necessary to develop four specialised risk assessment instruments for Australian Indigenous male offenders, which will be costly and time consuming.

As the instrument developed for Indigenous violent offenders misclassified more than half of the people it was applied to, it is important to consider whether it is at all possible to develop such an instrument. We believe that a possible explanation for this finding is that much less data were available on the files of violent offenders in this study than on those of the sexual offenders. It may be possible to develop an instrument that can be used to assess the risk of reoffending by violent Indigenous male offenders if data are available about a broader range of potential predictors.

Despite our expectation that dynamic predictors would be prominent for Indigenous offenders, the fact that all three predictors in the 3-Predictor model were dynamic was a surprise. This finding confirms the contemporary view regarding the importance of dynamic predictors of risk. More importantly, from an intervention perspective, it implies that it is theoretically possible to reduce the risk of sexual reoffending of Indigenous males by attending to these three factors. Therefore, if the poor coping skills of Indigenous males could be dealt with it may reduce the probability of them committing a further sexual offence. Likewise, the risk of recidivism by Indigenous male sexual offenders with poor coping skills could, theoretically, be reduced by attending to this deficit while they are in prison or under supervised community orders, and supporting them after release from prison when they will have to cope with the stress of reintegrating into society.

However, the implications of these findings are less positive when examined closer. While coping and the formulation of pre-release plans and long-term goals involve the abilities of offenders, they are also a function of their external circumstances. For example, the historical and social context of Indigenous Australians makes it inevitable that many of them live in a social environment where people require relatively good coping skills. This will especially be true in the case of an offender who is released from prison and may be faced with cultural sanctions. The relationship between substance abuse and offending amongst Indigenous males is also well known (Mals et al. 2000). For many offenders abstaining from using substances is an essential goal if they want to desist reoffending. However, how realistic is it for most Indigenous males to aim not to use substances, if substance use is endemic in the community in which they live, and while financial and personal factors make it difficult for them to relocate? Likewise, formulating realistic long-term goals may be difficult for a person who lacks literacy skills and who comes from a community where employment opportunities are limited and where high unemployment is chronic.

The findings of this study therefore suggest that the predictors of risk of reoffending amongst male Indigenous sexual offenders in WA are closely linked to socio-economic factors that are beyond the control of the offenders and the corrections system as such, and requires societal changes. For rehabilitation in prisons to be truly effective it is necessary that Western Australian Indigenous offenders return to settings that are free of avoidable stressors and provide an environment that makes it possible for offenders to set realistic long-term goals and formulate feasible release plans.

In conclusion, it appears possible to develop risk assessment instruments for Indigenous male offenders in WA that are more accurate than those currently used by assessors. However, to ensure predictive accuracy it will be necessary to develop instruments for the various subgroups of violent (family violence versus other violence) and sexual (violent versus non-violent) offenders. Given the large number of different Indigenous groups in Australia (Raphael et al. 1998) and the large volume of data necessary to develop such instruments the feasibility of doing this should be carefully considered. The study further confirmed that instruments comprising of dynamic factors are useful predictors of risk, and suggests that dynamic factors may play an important role in the reoffending of Indigenous male sexual offenders. It identifies needs of Indigenous males in WA that could be targeted by intervention programs that aim to reduce sexual offending. However, such programs will also have to address the stressors and deprivations experienced by many Indigenous males in WA.


The authors acknowledge the assistance of the Indigenous Advisory committee and the Western Australian Department of Justice, and the comments of Dr Maria Allan, Mr Peter Marshall and an anonymous reviewer on an earlier draft of this paper. The opinions expressed in this paper do not necessarily reflect those of these entities or persons.


  • Allan A & Dawson D 2002. Developing a unique risk of violence tool for Australian Indigenous offenders CRC 6/00-01 Canberra: Criminology Research Council
  • Andrews D 1995. The psychology of criminal conduct and effective treatment. In J McGuire (ed.) What works: Reducing reoffending. Guidelines from research and practice. Chichester: John Wiley: 35-61
  • Andrews DA & Bonta J 1994. The psychology of criminal conduct. Cincinnati: Anderson
  • Australian Psychology Society 2002. Guidelines for the provision of psychological services for and the conduct of psychological research with Aboriginal and Torres Strait Islander Australians. Australian Psychology Society ethical guidelines: 5-11 Melbourne: APS
  • Blagg H 2000. Australian Aboriginal people and family violence. Paper presented at a Research and Evaluation Committee meeting of the West Australian Ministry of Justice Perth 9 February
  • Carcach C, Grant A & Conroy R 1999. Australian corrections: the imprisonment of Indigenous people. Trends & issues in crime and criminal justice no 137 Canberra: Australian Institute of Criminology
  • Cleckley H 1976. The mask of sanity. St Louis: Mosby
  • Dawson D & Allan A 2000. Practice informing research: focus group research and violent Indigenous offenders. Paper presented at Australian Institute of Criminology 'Reducing criminality: partnerships and best practice' conference Perth 31 July-1 August
  • --- 2001. Risk of violent re-offending: do Indigenous male offenders have different risk characteristics? Paper presented at Australian Institute of Criminology 'Best practice interventions in corrections for Indigenous people' conference Sydney 8-9 October
  • Day A & Howells K 2002. Psychological treatment for rehabilitating offenders: evidence-based practice comes of age. Australian psychologist vol 37: 39-47
  • Fong N, Merson N, Tittums J, O'Loughlin H & Yavu-Kama-Harathunian C 2002. Indigenous advisors: a 'blueprint' for the future. Paper presented at Australian Institute of Criminology 'Probation and community corrections: making the community safer' conference Perth 23-24 September
  • Gendreau P, Goggin C & Smith P 2002. Is the PCL-R really the 'unparalleled' measure of offender risk? A lesson in knowledge cumulation. Criminal justice and behaviour vol 29: 397-426
  • Graham A 2003. Post-prison mortality: unnatural death among people released from Victorian prisons between January 1990 and December 1999. Australian and New Zealand journal of criminology vol 36: 94-109
  • Grove WM & Meehl PE 1996. Comparative efficiency of informal (subjective, impressionistic) and formal (mechanical, algorithmic) prediction procedures: the clinical-statistical controversy. Psychology, public policy, and law vol 2: 293-323
  • Hanson RK & Bussiere MT 1998. Predicting relapse: a meta-analysis of sexual offender recidivism studies. Journal of consulting and clinical psychology vol 66: 348-362
  • Hare RD 1980. A research scale for the assessment of psychopathology in criminal populations. Personality and individual differences vol 1: 111-117
  • --- 1991. The Hare psychopathy checklist - revised manual. Toronto: Multi-Health Systems Inc.
  • Mals P, Howells K, Day A & Hall G 2000. Adapting violent rehabilitation programs for the Australian Aboriginal offender. Journal of offender rehabilitation vol 30: 121-135
  • Monahan J 1981. The clinical prediction of violence. Washington: National Institute of Mental Health
  • ---1991. The clinical prediction of dangerousness. Currents in affective illness vol 10: 5-12
  • Quinsey VL, Harris GT, Rice ME & Cormier CA 1998. Violent offenders: appraising and managing risk. Washington DC: American Psychological Association
  • Raphael B, Swan P & Martinek N 1998. Intergenerational aspects of trauma for Australian Aboriginal people. In Y Danieli (ed.) International handbook of multigenerational legacies of trauma. New York: Plenum: 327-339
  • Rogers R 2000. The uncritical acceptance of risk assessment in forensic practice, Law and human behaviour. vol 24: 595-605
  • Smith LT 1999. Decolonizing methodologies: research and Indigenous peoples. Dunedin: University of Otago Press
  • Worby G & Rigney D 2002. Approaching ethical issues: institutional management of Indigenous research. Australian universities review vol 45: 24-33
  • Worrall A 2000. What works at One Arm Point? A study in the transportation of a penal concept. Probation journal vol 47: 243-249

About the authors

Associate Professor Alfred Allan and Deborah Dawson are from the School of Psychology, Edith Cowan University.

Deborah is also the senior researcher, at the Department of Justice Offender Programs Edith Cowan Project


This paper was developed out of a larger project 'Developing a unique risk of violence tool for Australian Indigenous offenders', funded by the Criminology Research Council (CRC 6/00-01).

This paper is taken from the report of research undertaken with the assistance of a grant from the Criminology Research Council.