Australian Institute of Criminology

Skip to content

Risk factors for Indigenous violent offending - Part 2

Physical health and disability

Extent of community disadvantage

Over many decades, both media and public attention has focused on the poor health outcomes for Indigenous adults and children. A handful of statistics from the wealth of data now available are summarised below.

Life expectancy at birth

  • In 2001, Indigenous life expectancy was around 17 years lower than for the general population of Australia. More recent estimates indicate a life expectancy of 59 years for Indigenous males and 65 years for Indigenous females, compared with 77 years and 82 years for non-Indigenous males and females respectively (SCRGSP 2007).

Infant birth weight and mortality rates

  • Indigenous babies are two to three times more likely to be born with a low birth weight, which is predictive of future problems (Coorey 2001: 92)
  • Between 2003 and 2005, the combined infant mortality rates for New South Wales, the Northern Territory, Queensland, South Australia and Western Australia were two to three times as high as those of Australian infants in general (SCRGSP 2007).

Disability, chronic disease and injury

  • The proportion of the Indigenous population aged 15 years and over reporting a disability or long-term health condition in 2002 was almost twice as high as that reported by non-Indigenous people. In 2001, Indigenous people reported higher rates of asthma, diabetes and kidney disease than did non-Indigenous people (SCRGSP 2007).
  • In 2004–05, Indigenous children under four years of age were twice as likely as non-Indigenous children to be hospitalised for potentially preventable diseases and injuries. In that same year, the hospitalisation rate for Indigenous adults with a potentially preventable chronic illness was eight times the rate of non-Indigenous adults (SCRGSP 2007).
  • In 2001–02, deaths due to diabetes were two to four times higher among Indigenous than non-Indigenous people (SCRGSP 2007).
  • In 2002, in Western Australia, the prevalence of foetal alcohol syndrome was 2.76 per 1,000 Aboriginal children compared with only 0.02 per 1,000 non-Aboriginal children (O'Leary 2002).
  • In that same state, 29 percent of Aboriginal children aged 0–17 years had a perforated eardrum and 65 percent had experienced hearing loss as a result of otitis media (Zubrick & Silburn 2006).

These national findings are replicated in various Indigenous communities across Australia. For example, in the Cape York region (see Fitzgerald 2001):

  • the median age of death among Indigenous people was 20 years below that of the general Queensland population, while mortality rates were two to three times higher;
  • there were high rates of sexually transmitted diseases, particularly among pre-adolescent and adolescent girls;
  • the rates of premature births and low birth weights (both of which constitute risk factors for infant mortality and health problems in later life) were 1.5 times as high as that of the general Queensland population; and
  • there were high rates of diseases, including glaucoma, ear infections, heart disease and kidney failure.

Link to offending

There is very little empirical data on the extent to which poor health outcomes act as a risk factor for violent offending. That such a link may exist comes from the 2002 NATSISS which found that:

  • those Indigenous persons who had been charged by police at some stage in their lives were more likely to report fair to poor health status than those who had never been charged (30% compared with 20% respectively; ABS 2002); and
  • of those who had been arrested in the previous five years, 20 percent indicated they suffered from a disability, which was significantly higher than the 14 percent recorded by those who had not been arrested during that period (ABS 2002).

Overall, though, it is likely that the link between physical health and violence is an indirect one, mediated by other risk factors, such as the impact of poor health on educational and employment opportunities.

Psychological distress and mental health issues

Extent of community disadvantage

A mental health problem has been defined as 'diminished cognitive or social abilities but not to the extent that the criteria for a mental illness are met' while a mental illness is a 'clinically diagnosable disorder' (Department of Health & Ageing 2009: 29). However, data on the prevalence of either mental health problems or mental illness within the Indigenous context are 'glaringly deficient' (Hunter 2003: 150). One data source is hospitalisation records but, as with criminal justice data, such information reflects only the 'tip of the iceberg' because many Indigenous people either do not have access to, or prefer not to use, these services. Nevertheless, these records indicate that Indigenous persons experience considerably higher levels of psychological distress than non-Indigenous persons. For example:

  • The chances of an Indigenous person being admitted for involuntary psychiatric care is three to five times higher than for Australians in general and is even higher for disorders relating to substance use, psychotic disorders and dementia (Response Ability 2009)
  • In 2003–04, the rate of hospitalisation for those diagnosed with mental disorders due to psychoactive substance use was over four times as high as for Indigenous than non-Indigenous males and over three times as high for Indigenous females compared with non-Indigenous females. Indigenous male and female rates for schizophrenia, schizotypal and delusional disorders were more than double those of their non-Indigenous counterparts (ABS & AIHW 2005).

In terms of deaths resulting from mental illness:

  • From 1999 to 2003, just over two percent of Indigenous deaths in Queensland, Western Australia, South Australia and the Northern Territory were due to mental disorders, which was over five times that of non-Indigenous Australians (ABS & AIHW 2005)
  • Over this same period, male Indigenous death rates due to mental illness were 5.5 times as great as non-Indigenous male rates, while among females the Indigenous rate was 2.2 times the non-Indigenous rate (AIHW 2006).

Findings from the 2002 NATSISS showed that Indigenous persons aged 18 years and over were significantly more likely to have experienced at least one stressor in the preceding 12 months (83%) than non-Indigenous persons (57%). The stressors most frequently cited were death of a family member or close friend (46%), serious illness or disability (31%) and inability to get a job (27%; ABS 2002).

Somewhat similar findings emerged from the 2004–05 NATSIHS (SCRGSP 2007). It noted that:

  • Indigenous people were twice as likely to be hospitalised for mental and behavioural disorders than non-Indigenous people.
  • After adjusting for age differences, 27 percent of Indigenous people reported high to very high levels of distress compared with only 13 percent of non-Indigenous people, with these inter-group differences applying across all age categories. These levels did not vary significantly between those Indigenous persons living in major cities, regional areas or remote Australia.
  • Abuse or violent crime was the most prominent stress factor cited (causing high to very high distress in 42% of respondents), a finding which illustrates the existence of a 'vicious circle', with stress potentially increasing the risk of violence which, in turn, increases the risk of higher stress levels.
  • Other stress factors identified included drug-related problems (cited by 41% of Indigenous respondents), alcohol problems (39%), divorce or separation (38%) and gambling (39%).

High levels of mental health problems and psychological distress have also been identified in Indigenous children. The Western Australian Aboriginal Child Health Survey (WAACHS) conducted between 2000 and 2002 among young people aged under 18 years found that:

  • Twenty-four percent of Aboriginal children were at high risk of clinically significant emotional or behavioural difficulties compared with 15 percent of non-Indigenous children (Zubrick et al. 2005).
  • The factor most strongly associated with these difficulties was 'life stress events' (Zubrick et al. 2005), with over one in five Aboriginal children (22%) aged 0 to 17 years living in families who indicated they had experienced seven or more major life stress events in the 12 months prior to the survey, compared with less than one percent of non-Aboriginal young people. Among the children themselves, 70 percent of those identified as Aboriginal had experienced three or more life stress events in 12 months, compared with 14 percent of non-Aboriginal children (Blair, Zubrick & Cox 2005).
  • Those children cared for by adults who had been forcibly separated from their natural families in childhood were more than twice as likely as other children to be at high risk of clinically significant emotional or behavioural difficulties (Zubrick & Silburn 2006).

Given these results, it is not surprising that according to the same survey:

  • One in five Aboriginal young people aged 12 to 17 years were 'at high risk of clinically significant emotional behaviours...compared with 7%…of a contemporaneous sample of non-Aboriginals 12 to 17 years old' (Blair, Zubrick & Cox 2005: 435).
  • One in three Aboriginal young people were at 'high risk of clinically significant conduct problems compared with 13.1% of the non-Aboriginal sample' (Blair, Zubrick & Cox 2005: 435).

Interestingly, the proportion of children at high risk of clinically significant emotional or behavioural problems was lowest in areas of extreme isolation. This may suggest that stronger adherence to traditional culture and lifestyle that occurs in these isolated areas may act as a protective factor (Zubrick & Silburn 2006).

Another potential indicator of depression and poor mental health among Indigenous people is their very high levels of suicide and non-fatal self-harm (which, according to Memmott & National Crime Prevention 2001 and others, constitute forms of violence in their own right). In 2005, in those four states for which data were available (Queensland, South Australia, Western Australia and the Northern Territory), Indigenous suicide rates were markedly higher than non-Indigenous rates (SCRGSP 2007: 73–75). More specifically, when compared with non-Indigenous figures, Indigenous suicide rates were:

  • 2.6 times as high in Western Australia (18.8 suicides per 100,000 Indigenous population compared with 10.8 per 100,000 non-Indigenous population);
  • 3.1 times as high in the Northern Territory (48.2 compared with 15.5);
  • 3.9 times as high in South Australia (45 compared with 11.4 respectively); and
  • 2.6 times as high in Queensland (32.2 compared with 12.2 respectively).

These inter-group differences are even larger in some communities. For example, in the remote Cape York region, Indigenous male suicide rates in 2001 were over six times as high as that of all Queensland males, while Indigenous female rates were two times higher than statewide rates for the total female population (Fitzgerald 2001).

There is also some evidence that Indigenous suicide rates may be increasing. Research has indicated that, not only has the Northern Territory recorded the highest Indigenous suicide rate since the mid 1990s, but that this rate has escalated over the past two decades. In the early 1980s, the age-adjusted suicide rate for NT Indigenous males was about one-third of that of NT non-Indigenous males, while no suicides were recorded for the female Indigenous population. But between 1981 and 2002, the suicide rate of both Indigenous males and females increased substantially (by an average of 17% and 26% respectively each year). In contrast, non-Indigenous male and female suicide rates increased by only one percent and six percent respectively over this same period. As a result of these different trends, by 2001–02, Indigenous male and female suicide rates were almost two times that of non-Indigenous rates (Measey et al. 2006).

Statistics from the AIHW National Hospital Morbidity database on incidents involving self-harm that did not result in death also point to higher levels among Indigenous than non-Indigenous people, with an age-standardised hospitalisation rate for intentional self-harm in 2004–05 of 2.9 per 1,000 Indigenous population compared with 1.6 per 1,000 non-Indigenous population (SCRGSP 2007).

The risk of suicide among Indigenous persons is higher among males than females. This is particularly true in the Northern Territory where, between 2001 and 2005, the Indigenous male suicide rate was more than seven times that of Indigenous females (84.9 per 100,000 Indigenous male population compared with 11.7 for females). In contrast, the gender gap was lowest in South Australia and Queensland (with males 3.3 and 3.6 times respectively more likely to suicide than females; SCRGSP 2007: attachment Table 3A.8.1). Interestingly, however, the opposite applied to incidents involving non-fatal self-harming behaviour. Here, the statistics indicate that female rates are higher than those of males. In 2004–05, self-harm hospitalisation rates among Indigenous females was three per 1,000, compared with 2.7 per 100,000 male Indigenous population (SCRGSP 2007).

The risk of suicide among Indigenous persons also varies according to age. It is highest among persons aged 25–34 years, which is also the peak age for involvement in violent offending. Within this age group, suicide rates ranged from 35.6 per 1,000 in Western Australia, 54.6 in Queensland, 85.2 in South Australia, to 92.6 in the Northern Territory. By comparison, the risk is lower among those aged under 25 years (11.5 per 1,000 Indigenous population in Western Australia, 15.5 in Queensland, 23.3 in South Australia and 30.5 in the Northern Territory (SCRGSP 2007).

Link between mental health issues and violent offending

The limited amount of empirical evidence relating to the link between mental health and violent offending suggests the presence of high stress levels/mental health problems among Indigenous offenders. However, most of these data apply to all offenders rather than to violent offenders per se and pertain only to those individuals who have been 'caught' by police.

  • According to the 2002 NATSISS (ABS 2002), a higher proportion of Indigenous persons who had been charged by police at some stage in their lives (86%) had experienced at least one stressor in the previous 12 months than those who had never been charged (81%). Interestingly though, when stress associated with alcohol, drugs and unemployment were excluded, those persons charged by police were no more likely to experience a stressor than those who had never been charged (38% compared with 37%).
  • A study conducted among prisoners in Western Australia found that, when compared with that state's Indigenous population as a whole, rates of hospital admissions for mental disorders were approximately twice as high among Indigenous male prisoner and three times as high for Indigenous female prisoners (Hobbs et al. 2006).
  • Results from a survey of female prisoners in six Australian jurisdictions conducted as part of DUCO found that Indigenous females imprisoned for a violent offence were more likely (78%) to have mental health problems than non-violent Indigenous female offenders (63%; Johnson 2004). Similar differences were observed when property offenders were compared with non-property offenders.
  • A survey of offenders serving community-based orders in Queensland (Mazerolle & Legosz 2007) found that over half of the Indigenous respondents were suffering either moderate to severe depression. However, although the differences were not statistically significant, depression seemed to be more pronounced among non-Indigenous respondents, with seven percent classified as severely depressed compared with only three percent of Indigenous respondents. Non-Indigenous offenders were also significantly more likely than their Indigenous counterparts to have been previously diagnosed by a doctor with having a mental health disorder. This applied to all of the disorders measured, with the exception of alcohol and drug dependence (see Table 14). However, while these figures may point to lower levels of mental disorders among Indigenous offenders, they may also reflect their lower access to medical support and less willingness to access such services.
Table 14: Proportion of respondents previously diagnosed as having a mental health disorder, by Indigenous status, Queensland (%)
Type of disorder Indigenous Non-Indigenous
Depression 31.9 46.3*
Bipolar disorder 1.1 9.4*
Personality disorder 1.1 8.3*
Schizophrenia 1.1 9.4*
Anxiety 10.6 27.3*
ADHD 7.4 9.1*
Drug dependence 28.2 20.2*
Alcohol dependence 10.6 14.7*

* Differences were statistically significant

Source: Mazerolle & Legosz 2007: 58


That Indigenous persons involved with the criminal justice system experience mental health problems and psychological stress has been identified by various government inquiries. The RCIADIC (1991), for example, drew a strong link between undiagnosed mental and social distress and high rates of incarceration among Indigenous people. It also noted that incarceration may, in itself, either trigger mental illness or, if already present in the individual, exacerbate it—an outcome which is particularly likely in those situations where appropriate treatment programs are not made available to prisoners.

However, the link between mental health issues and offending (including violent offending) is likely to be complex and multidimensional. Psychological stress, for example, may not be directly causative of violence, but may lead to alcohol abuse which, in turn, may result in violence. There is also a growing body of evidence (outlined earlier in this report) of a link between illicit substance abuse (particularly marijuana and methamphetamines), mental illness and violence.

In recognition of the failure of the criminal justice system to deal effectively with those persons whose offending is linked to their mental health status, a number of Australian jurisdictions have now established specialised diversionary courts (colloquially referred to as Mental Health Courts) which aim to use the offender's contact with the criminal justice system as a lever for engaging that person in effective treatment programs. As yet, however, these courts operate in only a handful of locations (predominantly in capital cities) and are limited to a relatively small number of offenders. They are therefore not readily accessible to Indigenous offenders (Hunter & McRostie 2001). Other specialist or problem-oriented courts, such as drug courts or Indigenous courts, may on occasion identify and seek to address mental health issues but this is not their primary purpose.

Geographic location and remoteness

Spatial variations in Indigenous population distribution and characteristics

The geographic distribution of the Indigenous population is substantially different from that of non-Indigenous Australians. While the latter are predominantly urban dwellers, Indigenous Australians mainly reside in rural and remote areas. Yet even within the Indigenous population, the characteristics of Indigenous persons living in remote regions vary considerably from those resident in rural areas or in major cities. The 2002 NATSISS, for example, found that persons aged 15 years and over living in remote areas were more disadvantaged across a range of indicators than their non-remote counterparts. In particular:

  • They had lower education levels. Fewer had a post-school qualification (17% compared with 3%) while a higher proportion had left school in Years 6, 7 and 8 or had never attended school in the first place (6% compared with less than 1% of non-remote Indigenous residents; ABS 2002).
  • While a significantly higher proportion were employed (52% compared with 44%), their main source of employment was CDEP. In remote areas, 33 percent of Indigenous people were employed under this scheme while only 19 percent were engaged in other work. In comparison, only five percent of non-remote Indigenous residents were employed under CDEP, while 40 percent were engaged in non-CDEP work. The prominent role played by CDEP in remote communities meant that the proportion of the population officially designated as unemployed was actually lower than in non-remote areas (6% compared with 17%; ABS 2002). But given the range of criticisms now being levelled at CDEP (as outlined earlier), this may not be a positive indicator of economic wellbeing.
  • A significantly higher proportion of remote residents had experienced at least one stressor in the past 12 months (86% compared with 81% of non-remote dwellers; ABS 2002) while almost three-quarters (73%) reported experiencing financial stress compared with under one-half (47%) of non-remote dwellers (ABS 2002).
  • A significantly lower proportion believed they could get support in a time of crisis (87% compared with 92%; ABS 2002).
  • A significantly lower proportion were home owners, while a significantly higher proportion lived in dwellings that had major structural problems (50% compared with 33% of non-remote residents) or were overcrowded (ie needed additional bedrooms; 52% compared with 16%; ABS 2002).

In relation to some characteristics, however, Indigenous persons living in remote regions seemed to fare better than those in non-remote locations. For example, remote residents were less likely to have been removed from their natural family (6% compared with 9%) or have a relative who had been removed (28% compared with 39%) and were more likely to have retained a commitment to traditional values and lifestyle, including identifying with a clan or tribal/language group, currently living in their homeland or traditional country, speaking an Aboriginal language at home and being more heavily involved in Indigenous cultural events (ABS 2002).

Overall though, Indigenous persons in remote areas seem to rank higher on a range of indicators widely considered to be risk factors for violence.

Link between geographic location and violent offending

Given what seems to be a concentration of risk factors for violence in remote communities, combined with the greater difficulties involved in accessing services and programs, it may be expected that these communities would exhibit higher levels of violence. This perception has frequently been reinforced by the media, particularly in their reporting of the NTER.

However, empirical evidence to this effect is sparse, with most information derived either from the NATSISS, or official criminal justice data sets, both of which measure contact with the system rather than offending per se. The findings from these different sources are also somewhat contradictory.

On the one hand, there is some indication that those in remote areas have higher offending rates than those in non-remote areas. For example, between 1999–2000 and 2004–05, the Indigenous homicide rate was lowest in major cities (4.6 per 100,000 Indigenous population), increasing to 4.9 in inner regional areas, to 13.4 in outer regions and 16.7 in remote areas, before dropping slightly to 13.1 in very remote areas. In contrast, non-Indigenous homicide rates were not only much lower across all regions, but showed little variation from one geographic setting to another, with figures ranging from 1.4 to approximately 1.6 per 100,000 non-Indigenous population. Only in the very remote areas did the non-Indigenous rate exceed two per 100,000 population (SCRGSP 2007: attachment Table 3A.10.5). As a result of these different trends, the discrepancy between Indigenous and non-Indigenous homicide rates increased as the degree of remoteness increased. In major cities, the Indigenous rate was approximately 3.3 times that of the non-Indigenous rate, but in remote areas it was 10 times as high.

Other data, however, do not support the view that Indigenous persons in remote communities have a higher risk of offending or involvement with the criminal justice system. According to the 2002 NATSISS, Indigenous people aged 15 years and over in remote or very remote areas of Australia were no more likely that those in a major city to have been charged by police at some stage in their lives (35% and 33% respectively) or to have been arrested in the preceding five years (17% and 16% respectively; ABS 2002). This lack of difference may, however, be due to factors other than an absence of variation in offending behaviour. First, because the data measure contact with the system, rather than actual offending, they are likely to be influenced by regional variations in police strength and availability. In addition, the use of a highly generalised dichotomised variable that groups all respondents into either 'remote' or 'non-remote' (the latter of which combines those living in major cities, inner regional and outer regional areas) may be too crude a measure to detect differences at a local level.

One analysis which used a more fine-grained geographic differentiation to investigate regional variations in Indigenous violent apprehension rates was undertaken in Western Australia in 2001 (Fernandez 2003). While the results (a selection of which are outlined below) are now somewhat dated, they nevertheless point to significant differences in apprehension rates at both a regional and a sub-regional level. They therefore indicate that the link between geographic location and the risk of violence is far more complex than a simple remote/non-remote dichotomisation.

Regional variations in Indigenous apprehension rates for violent offences, Western Australia

Indigenous apprehension data for 2001 were extracted for a number of ABS postal areas within seven key regions in Western Australia: Gascoyne-Murchison, Goldfields, Great Southern, Kimberley, Pilbara, North Metropolitan and South Metropolitan (Fernandez 2003). The data showed marked variations between and within these seven regions, in both the rate of violent offending, as well as the relative positioning of violence compared with property and good order apprehensions.

In the Kimberley region, for example, rates of apprehension for violent offences varied from 68.4 per 1,000 Indigenous population in Kununurra to 18.5 in Wyndham. In all six Kimberley sub-regions, apprehension rates for violent offences were consistently lower than apprehension rates for good order offences and (with the exception of Fitzroy Crossing) for property offences (see Figure 10). Of particular note, however, is that these rates do not appear to be linked to the relative size of the Indigenous population in each region. Although over 60 percent of the population in Fitzroy Crossing identified as Indigenous, compared with only 24 percent in Broome, apprehension rates for violent offences were relatively similar in both locations. In Kununurra, where apprehension rates for violence were higher, only 14 percent of the population was Indigenous.

Figure 10: Indigenous apprehension, Kimberley Region, Western Australia 2001 (rates per 1,000 Indigenous population)

Indigenous apprehension, Kimberley Region, Western Australia 2001

a: These figures represent the percentage of the total population in each region that identifies as Indigenous

Source: Fernandez 2003

Figure 11: Indigenous apprehension offence comparisons, North Metropolitan Region, Western Australia 2001 (rates per 1,000 Indigenous population)

Indigenous apprehension offence comparisons, North Metropolitan Region, Western Australia 2001

a: These figures represent the percentage of the total population in each region that identifies as Indigenous

Source: Fernandez 2003

Interestingly, apprehension rates for Indigenous violent offending in the remote Kimberley region were no higher than those recorded in more southerly locations, such as the North Metropolitan region, which includes sections of Perth and parts of that state's wheat belt (Fernandez 2003). As shown in Figure 11, in the North Metropolitan region, rates of violent apprehensions exceeded 60 per 1,000 Indigenous population in five of the eight sub-regions listed and reached a high of 159.6 per 1,000 in Kellerberrin. As was the case in the Kimberley, both property and good order offences were more prominent than apprehensions for violent offences.

This considerable intra-regional variation, and the fact that the highest apprehension rates for Indigenous violent offences do not necessarily occur in semi-remote or remote areas, are further illustrated in Table 15. This lists all sub-regions where the violence apprehension rate exceeded 70 per 1,000 Indigenous population. The highest rate of violent apprehensions (250 per 1,000 population) was recorded in the Great Southern district of Wagin, followed by Kellerberrin (159.6) in the North Metropolitan region. Interestingly, no sub-region in the Kimberley or the South Metropolitan regions recorded such high apprehension levels. Yet, arguably, these two regions sit at quite different ends of the geographic and Indigenous cultural spectrum in Western Australia.

Table 15: Regions/sub-regions with an apprehension rate for violent offences of 70 per 1,000 Indigenous population or above, Western Australia, 2001
Region Indigenous Indigenous persons
Rate per 1,000 persons As a % of total population
Carnarvon 106.9 12.5
Mount Magnet 100.0 16.7
Wiluna 151.5 22.0
Esperance 72.8 3.9
Kalgoorlie-Bolder 78.4 6.3
Laverton 195.1 32.2
Great Southern
Albany 85.5 2.7
Katanning 120.4 7.2
Mount Barker 93.8 3.7
Narrogin 81.6 5.3
Wagin 250.0 5.1
Kimberley 0.0 0.0
Port Hedland 70.3 15.2
North Metropolitan
Kellerberrin 159.6 8.1
Merredin 113.8 4.6
Northam 102.5 6.6
York 72.9 3.2
South Metropolitan 0.0 0.0

Source: Fernandez 2003

These figures suggest that the size of the Indigenous population, when viewed as a percentage of the total population in each region, does not affect Indigenous apprehension rates. Regions with large Indigenous populations relative to the size of the non-Indigenous population do not have higher rates of offending than areas where the proportion of Indigenous persons in the population is relatively small (although absolute numbers are likely to be higher). Nor does the rate of violent offending seem to vary in any consistent way according to distance from Perth.

There was, however, one element of consistency across all of these regions, irrespective of the size of the Indigenous population or its geographic location—namely that Indigenous apprehension rates for violent offending were higher than those recorded by non-Indigenous persons. This issue is explored in more detail below.

Regional variations in Indigenous/non-Indigenous rates of apprehension for violent offences, Western Australia

That Indigenous apprehension rates for violent offences in Western Australia were consistently and substantially higher than non-Indigenous rates is illustrated in Figure 12, which focuses on a selection of locations in the Gascoyne-Murchison and South Metropolitan regions. However, the patterns illustrated here are consistent with those observed across all other regions. Figure 12 also indicates that the extent of difference between Indigenous and non-Indigenous individuals varies from one sub-region to another. Within the Gascoyne-Murchison region, for example, Indigenous apprehension rates for violent offences in Wiluna were 194 times that of non-Indigenous apprehension rates for violence, while in Denham they were only 13 times as high. To fully understand the factors in Wiluna and Denham that lead to such large discrepancies between the two, in-depth community-based studies are required.

Figure 12: Gascoyne-Murchison region—apprehensions for offences of violence, 2001 (rates per 1,000 population by Indigenous status)

Gascoyne-Murchison region - apprehensions for offences of violence, 2001

Source: Fernandez 2003


While both the incidence and nature of Indigenous violent offending varies considerably from one region to another, the data do not point to an association between remoteness and levels of violence. An analysis of victim-based rather than offender-based data is also inconclusive. While the 2002 NATSISS did not find any variation between remote and non-remote areas in levels of self-reported violent victimisations, the survey did find that those living in remote areas were more likely than non-remote dwellers to perceive violence as a bigger problem in their community, with a significantly higher percentage citing abuse and violent crime (17% compared with 9% of non-remote residents) and the witnessing of violence (30% and 10% respectively) as stressors in their lives. In addition, when asked to identify problems that they considered to be a serious issue in their neighbourhood, a significantly higher proportion of remote than non-remote respondents aged 15 years and over listed family violence (41% compared with 14%), assaults (41% compared with 12%), sexual assaults (17% and 5% respectively) and levels of neighbourhood conflict (31% and 9% respectively; ABS 2002). While perceptions may not necessarily reflect reality, these findings, combined with what seem to be the more pronounced presence in remote communities of risk factors for violent offending, contributes to an expectation of higher levels of violence in these settings. That this is not reflected in the offender-based data may be due to several factors:

  • Most of the available information on geographic variations in offending relates to the individual's contact with the criminal justice system. It may be that, although levels of violence are higher in remote than non-remote areas, such violence is either never reported to police or does not result in the official apprehension or charging of the perpetrator.
  • Alternatively, as suggested earlier, the differentiation between remote and non-remote may simply be too crude. Indeed, it is more probable that certain risk factors for violence will vary from one community to another even within the same geographic setting, while other risk factors may remain constant, irrespective of geographic location. A 1998 study by the Queensland Criminal Justice Commission (cited in Memmott & National Crime Prevention 2001), which analysed the rate of violent offences reported to police broken down by police division, found that, rather than geographic location, it was their history as mission centres that characterised those four Indigenous communities which recorded the highest incidence of violent crime. This finding accords with the expectation of Memmott and National Crime Prevention (2001: 13) that 'Indigenous communities which are most affected by violence [are likely to be] those with a long history of functioning as removal centres or missions and where maximum dysfunctional cultural change has occurred'. An earlier study by Trigger et al. (1983) of 14 Aboriginal reserves in Queensland also found that levels of violence varied depending on the characteristics of the community involved, and argued that the history of government policy and the style of administration experienced by a community was one of the key factors in determining its characteristics. Fitzgerald (2007) citing Pearson (2001b), suggests a different argument for the absence of regional variations in violence from one setting to another within Cape York. He attributes it to the fact that many of these communities, irrespective of their location, are dependent on welfare, which, in turn, links to high levels of dysfunctional behaviours, including alcohol abuse and violence.

A better understanding of the relationship between geographic location and risk factors for violence therefore requires a more community-specific approach, rather than trying to use broad locational groupings as a proxy for violence.

Access to services

Extent of disadvantage within the community

The extent to which individuals and their communities have access to appropriate services and support programs may influence both the initiation as well as the continuation of violence. These services include:

  • broader community-wide programs designed to improve education, employment, health and housing opportunities for Indigenous people;
  • strategies or interventions which aim to redress particular types of dysfunctional behaviour within the individual, such as alcohol abuse, illicit drug use, anger and aggression; and
  • criminal justice and related services that are specifically geared to respond to both victims and perpetrators of violence. This includes those agencies with responsibility for providing counselling and support services to victims, as well as criminal justice agencies that are required to investigate, apprehend, prosecute and sentence the alleged perpetrators of such violence.

It is beyond the scope of this report to canvas the range and effectiveness of either the generic services or those designed to respond to dysfunctional behaviours situated within the individual that represent risk factors for violence. Instead, consideration will be limited to those agencies whose work brings them into direct contact with offenders. While access to these services will inevitably vary from one community to another both within and across different geographic settings, numerous reports indicate that remote and semi-remote communities are the most severely disadvantaged (eg see Aboriginal and Torres Strait Islander Women's Task Force on Violence & Department of Aboriginal and Torres Strait Islander Policy and Development 2000; Fitzgerald 2001; Gordon, Hallahan & Henry 2002; Mullighan 2008; Wild & Anderson 2007). To illustrate this, information from only one of these inquiries will be summarised below, although the findings apply to many other regions in Australia.

The Mullighan Inquiry (2008) in the APY Lands of South Australia detailed case after case where government services failed to respond adequately, particularly to issues of child abuse. Criticisms of Families South Australia, the agency with legislative responsibility for protecting children at risk, included:

  • failure to locate welfare and child protection workers within the communities themselves, therefore preventing them from responding quickly to critical situations;
  • failure to investigate notifications of abuse and neglect, delays in investigation or failure to undertake follow-up work after initial contact;
  • lack of effective strategies to resolve child abuse and neglect situations. In many cases, the only solution authorities could offer was to remove the victim from his/her family and the community, while leaving the alleged perpetrator in situ. Not only did this appear to punish the victims, but it also meant that when they later returned to the community (as most did), they often found themselves back in the same risky situation as before; and
  • a complex and cumbersome mandatory notification system which reportedly discouraged some notifiers from lodging reports.

Criticisms were also levelled against workers in other agencies—notably health—for their failure to report instances of child abuse and neglect. The Inquiry identified many cases where teenage girls had sought medical treatment for a sexually transmitted infection (which, at the very least, indicates under-age, and probably unlawful, sexual behaviour or at the worst, a potential child abuse situation) but these were either not reported to the relevant government agency or were reported only after lengthy delays. Various reasons given for non-notification included that the young girl was involved in 'consensual sex', or that notification would create further problems in the community and potentially heighten tension between individuals and families.

The provision of mental health services for both young victims and young perpetrators of sexual assault was also considered inadequate. The Inquiry noted that the agency with statewide responsibility for providing mental health services to abused children and youth (the Children's and Adolescents Mental Health Service) had no workers based in the APY Lands and, at one point, refused to follow up on referrals from that region because of insufficient funding (Mullighan 2008). This was despite acknowledgement that child victims of sexual abuse require access to therapeutic services, not only to deal with the immediate trauma, but to reduce the risk that they will become perpetrators of such abuse themselves. Nor were any services provided by the main public health agency responsible for supporting adult rape and sexual assault victims in South Australia. Such services could only be accessed by transferring the victim to Adelaide, or through the two Adelaide-based psychiatrists who visited the APY Lands approximately four times per year (Mullighan 2008). The Inquiry also drew attention to the absence of any medical practitioners or psychologists who had the training required to undertake assessments of child sexual abuse allegations involving children under the age of seven years.

Criminal justice agencies came in for their own share of criticism. In relation to the police, Mullighan (2008) noted that:

  • at the time of the Inquiry, there was no permanent police presence in any of the Indigenous communities on the APY Lands, with the nearest 24 hour police station located hundreds of kilometres away, therefore making it difficult for residents to contact police after hours;
  • police responses were slow or, at times, non-existent; and
  • often when a police investigation did occur, no action resulted. Overall, very few alleged perpetrators were proceeded against, with SA police data indicating that between 2000 and 2007, there were only six apprehensions for child sexual assault on the APY Lands, despite the fact that the Inquiry found evidence that the incidence of such abuse was high.

Community dissatisfaction with police services is not new and is not limited to the APY Lands. Over a decade ago, the 1994 NATSIS found that almost one-quarter (22%) of Indigenous people surveyed believed that police did not do a good job when dealing with violence. Levels of dissatisfaction were slightly higher in other urban centres and capital cities than in rural areas (24%, 23% and 20% respectively). Reasons given for these negative views mirrored those found by the Mullighan Inquiry—namely that the police response was too slow, that they did not understand Indigenous people or their culture, that they failed to fully investigate the incident or that they failed to respond at all (Mukherjee et al. 1998).

The APY Lands Inquiry also noted that responses from other sections of the criminal justice system, notably prosecutions and court, were inadequate. Quite often, following the apprehension of a suspect, the individual was either never prosecuted or was never convicted. This observation has been substantiated by empirical evidence from other areas of Australia. An NT inquiry found that only half (52%) of those Indigenous persons apprehended for child abuse (as well as only 48% of non-Indigenous apprehensions) actually progressed through to the final stages of processing, with the largest attrition levels occurring at the point of sentencing after a matter had been finalised (Wild & Anderson 2007). Although not specific to Indigenous offenders, a study of sexual assaults against children in South Australia found that, of all such incidents reported to police in 2000–01, only 16 percent resulted in the alleged perpetrator experiencing some type of consequence, which could include either diversion to a family conference (in the case of juvenile perpetrators) or a court sentence (Wundersitz 2004). Even when an Indigenous perpetrator is found guilty and imprisoned, upon release, they often return to the same community and take up residence in close proximity to the victim. This has the potential to further traumatise that victim.

Other complaints about the criminal justice system include a lack of interpreters, an absence of victim support services to help people negotiate their way through the criminal process and a lack of specialists to deal with and provide support to child victims (Mullighan 2008; Wild & Anderson 2007). The dearth of appropriate prison-based treatment programs for incarcerated violent offenders, including anger management and sex offender programs, was also noted, as was the limited support and treatment provided during the post-release period.

Link between inadequate service responses and violent offending

Again, there is very little empirical data linking access to appropriate interventions, support services and programs with either the initiation, or the continuation, of violent behaviour. One of the few pieces of information comes from a study of repeat violent offending among Indigenous offenders in Western Australia (Allan & Dawson 2002). That study found that, after controlling for a range of factors, one of the key predictors of the likelihood of subsequent violent offending by Indigenous persons was the failure by the system to implement feasible release plans for these individuals upon exiting from prison. This variable was significant in predicting both future violent offending and future sexual offending among Indigenous releasees.


The lack of an adequate response by service providers to acts of violence within Indigenous communities may contribute to the continuation of that violence in several ways. For example:

  • Failure to intervene effectively at the time of the incident itself means that a particular act of violence will not be curtailed and may potentially spread to involve other members of the community.
  • Low apprehension levels and the absence of effective deterrence measures, particularly for child sexual assault, convey the impression to perpetrators that they will not be held accountable for their actions and so they have no incentive to change their behaviour (Wild & Anderson 2007).
  • The absence of effective treatment programs for violent offenders, particularly during the early stages of their offending careers, means that any underlying issues (such as alcohol or drug abuse) are not being addressed. Similarly, victims (particularly child victims) who do not have access to appropriate counselling and support services may, themselves, go on to become offenders or be re-victimised.

Equally important, however, is the fact that agency inaction may engender attitudes within the community which, in themselves, contribute to the continuation and even escalation of violence. Various inquiries have identified what seems to be a degree of Indigenous acceptance of violence as either normative or inevitable, together with a refusal to report or condemn such violence. However, this apparent acceptance may, at least in part, stem from a sense of futility and powerlessness—a sense that even if they do report an incident to police or 'welfare', nothing will be done about it. Allied with this may be fear of retaliation if they do take action.

As one witness to the Mullighan Inquiry (2008: 46) expressed it:

Imagine...being [assaulted] for refusing sex and the person gets effectively a slap on the knuckles. Why would you go through years of vilification, abuse and ostracism in small communities? This is educative for communities…you can be violent and destructive to get your own way and anybody who stands up to you gets publicly vilified and hounded.

Non-Aboriginal workers also suffer consequences for taking action. The APY Lands Inquiry, for example, cited an incident where students refused to attend school because they believed that school personnel had been responsible for the arrest of a popular male member of the community on child assault charges (Mullighan 2008). In other situations, the decision by agency workers not to report a potential incident, or the decision by police not to proceed against the suspected perpetrator was based on their perception that bringing the matter into the open would generate further violence in the community (Mullighan 2008). More effective and more timely agency responses to individual acts of violence could reassure residents that they would be protected from retaliation if they did report an incident to police.

Again, however, there is very little empirical evidence linking access to appropriate interventions, support services and programs with the continuation of violent behaviour. One of the few pieces of information comes from a study of repeat violent offending among Indigenous offenders in Western Australia (Allan & Dawson 2002). That study found that, after controlling for a range of factors, one of the key predictors of the likelihood of subsequent violent offending by Indigenous persons was the failure by the system to implement feasible release plans for these individuals upon release from prison. This variable was significant in predicting both future violent offending and future sexual offending among Indigenous releasees.

Summary of univariate analyses

A broad range of variables have been identified as potential risk factors for Indigenous violence. A number of these characterise the community as a whole as well as impacting on discrete individuals (ie they operate as both distal and proximal risk factors for violence). An examination of the empirical evidence linking some of these risk factors to violence indicated the following:

  • gender—Indigenous males are more likely to engage in violence than Indigenous females. While this trend is also evident within the non-Indigenous population, it is noteworthy that Indigenous females generally record higher levels of violence than non-Indigenous males. Many of the reasons put forward to explain this gender imbalance in the general population are also likely to apply in the Indigenous setting. However, additional factors may also be operating, such as the displacement of Indigenous males attendant upon the loss of their traditional roles in the post-colonisation era. This, it is argued, has led to problems such as a lack of self-esteem, unresolved anger and high alcohol consumption within this group, which in turn, may find expression in violent behaviour;
  • age—the risk of perpetrating violence varies according to age, with those in the mid range of 18–34 years being the most likely to engage in such behaviour. While this pattern is characteristic of both the Indigenous and the non-Indigenous population, Indigenous violent offenders tend, on the whole, to be slightly younger than their non-Indigenous counterparts, a finding which cannot be entirely explained by differences in the age profiles of the two groups;
  • Aboriginal/Torres Strait Islander—persons who identify as Aboriginal have a higher risk of contact with the criminal justice system for a violent offence than do Torres Strait Islanders. To some extent, this may be due to different levels of exposure to certain risk factors for violence, with Aboriginal persons recording lower education and employment levels, for example;
  • alcohol misuse—writers such as Pearson (2001a: np) argue that 'grog and drug epidemics' are now the most important underlying issue confronting Indigenous communities. Interestingly, though, at a community level, contrary to popular perception, the percentage of the Indigenous population who consume alcohol is no greater than in the non-Indigenous population. The main difference lies in the fact that, among those who do drink, proportionately more Indigenous than non-Indigenous people consume alcohol at risky to high risk levels or engage in episodes of binge drinking than their non-Indigenous counterparts. That there is a clear link between Indigenous alcohol consumption and violent offending is indicated by the fact that, at an individual level, a much higher proportion of Indigenous offenders consume alcohol at risky to high risk levels than either Indigenous non-offenders or non-Indigenous offenders. Moreover, as consumption levels increase, so too does involvement in acts of verbal or physical abuse. Prisoner surveys also indicate that Indigenous offenders are more likely than their non-Indigenous counterparts to be under the influence of alcohol at the time of their offending and to attribute their offending to alcohol consumption. This is particularly true of Indigenous males. Various reasons have been put forward to explain the apparent link between Indigenous alcohol misuse and violence, including the pharmacological impact that alcohol has on an individual's sense of what is and is not appropriate behaviour.
  • illicit drug use—in contrast to alcohol, illicit drug use is less prevalent within the Indigenous than the non-Indigenous population. Similarly, among offender groups, Indigenous persons who have been arrested or imprisoned are far less likely than non-Indigenous detainees to report use of most types of illicit drugs, with the exception of marijuana, where the pattern is reversed. Indigenous adult offenders are also less likely to be dependent on illicit drugs, to be under the influence of an illicit drug at the time of their most recent offending or to attribute their most recent offence to illicit drug use than their non-Indigenous counterparts. In contrast to adults, both Indigenous and non-Indigenous young offenders reportedly used similar types of drugs at similar frequencies. Interestingly, though, Indigenous youth were more likely to nominate illicit drug use as a factor in their most recent offending episode. Apart from these age differences, there are also some gender differences. Although alcohol use is high among both Indigenous women and men, the former are equally likely to attribute their offending to either alcohol or illicit substances, whereas Indigenous males are far more likely to blame alcohol only. However, while illicit drug use at a community level and within specific offender groups is still lower among Indigenous than non-Indigenous people, of particular concern is the rapid escalation in marijuana use within Indigenous communities over the past 10 years or so. This trend is particularly evident among young people in remote communities, where the introduction of non-sniffable fuel (OPAL) has resulted in a decrease in petrol sniffing but a concomitant increase in marijuana use. This may have serious implications, given the growing body of evidence that heavy, long-term use of marijuana may lead to or exacerbate pre-existing mental illnesses, which in turn, may trigger violent episodes;
  • child victimisation—child protection data indicate that Indigenous children experience relatively high levels of child abuse and neglect, although this data has to be interpreted with caution as it is a record of reported incidents. In turn, such maltreatment appears to constitute a risk factor for subsequent involvement in violent offending. This is indicated by the high proportion of Indigenous adult offenders who report that, as a child, they were either subjected to sexual or other forms of maltreatment or witnessed violence within their family and/or community. Various explanations for the link between childhood violence and subsequent offending observed within the general community—such as the effect that these early experiences have on the child's cognitive and emotional development, and the increased risk that the child will grow up believing that violence is normal—are also likely to apply to Indigenous children. However, the relationship is not a simple one, as evidenced by the fact that the majority of victims of childhood abuse do not grow up to be perpetrators themselves;
  • exposure to pornography—anecdotal evidence suggests that access to pornography is widespread in some Indigenous communities, particularly in remote areas, and that such exposure increases the risk of sexualised violence, particularly among children and adolescents. However, empirical evidence to this effect is lacking;
  • education, employment, income and housing—it is well documented that Indigenous communities exhibit lower education levels, higher unemployment rates, lower incomes, higher rates of welfare dependency and poorer housing conditions than their non-Indigenous counterparts. Similar differences emerge when Indigenous offenders are compared with either Indigenous non-offenders or non-Indigenous offenders. However, it is likely that the relationship between these factors and violence is an indirect one, mediated by their association with other factors such as self-esteem, levels of resentment and alcohol abuse. Pearson (2001a; 2001b), for example, links violence to alcohol misuse, which he, in turn, ascribes to passive welfare dependency. The role of CDEP is also criticised because it is part of the artificial economy now underpinning many Indigenous communities. And while evidence suggests that those Indigenous persons on CDEP have a higher risk of offending than those in fully paid employment, they have a lower risk of offending than those who are unemployed. In some respects, it may therefore act as a protective, rather than as a risk factor for violence;
  • physical health—that Indigenous people have significantly poorer health outcomes than non-Indigenous people across a broad range of indicators has also been well documented. However, empirical evidence linking physical health and disability to an increased risk of becoming a violent offender is sparse. One indication comes from the 2002 NATSISS, which found that those persons who had experienced formal contact with police were more likely than those who had had no such contact to be suffering from poor health or a disability. Again, however, the link, if it does exist, is likely to be an indirect one;
  • mental health—within Indigenous communities, a relatively high proportion of both adults and juveniles experience distress and mental illness. At least some of this has been attributed to the trauma and unresolved grief stemming from the loss of traditional country and culture. One expression of this psychological distress is the high rates of Indigenous suicide and non-fatal self-harm incidents, particularly among males, which are consistently higher than corresponding rates within the Australian community. The nature of the relationship between mental illness and violence is, however, unclear. Some data indicate that Indigenous persons charged by police are more likely than those who have not been charged to experience high stress levels. But when stress associated with alcohol, drugs and unemployment are excluded, these differences disappear. Other research suggests that, although many Indigenous and non-Indigenous offenders suffer from depression, levels are actually higher among non-Indigenous offenders (Mazerolle & Legosz 2007);
  • geographic location—there seems to be a general assumption that levels of violence are higher in remote than in non-remote communities, based in part, on a perception that the latter suffer from a greater range of social, economic and other disadvantages than their urbanised counterparts. However, the relationship between geographic location and the risk of violence is not straightforward. While data from the NHMP indicate higher levels of violence in remote communities, findings from the 2002 NATSISS suggest that those in remote areas were no more likely to experience contact with police than those in major urban centres. However, using a highly generalised dichotomous variable (ie remote/non-remote) as a measure of geographic location may be too crude. As evidence of this, WA police data indicate that Indigenous apprehension rates for offences of violence vary just as much from one location to another within a particular region as they do between regions. It is, therefore, more likely that community characteristics including policing, coupled with region-wide patterns of movement, rather than geographic location per se, are the critical factors in understanding violence and that these will differ even between so-called 'remote' communities; and
  • access to services—linked, in part, to geographic location is the level of community access to services. Numerous inquiries have criticised the lack and/or the ineffectiveness of the services provided to both perpetrators and victims of Indigenous violence, particularly in semi-remote and remote areas of Australia. Criticisms include a failure by government agencies to locate appropriate police, welfare and support workers within the communities themselves; slow and, at times, ineffective responses to violent incidents; and a failure by some non-Indigenous professionals to notify relevant authorities about suspected instances of abuse. The criminal justice system is also criticised because of its low success rate in apprehending and convicting offenders, particularly in cases involving child sexual abuse, although this criticism applies to all such cases and not just those involving Indigenous children. In the absence of high quality, responsive services, offenders are not held accountable for their actions, community engagement in developing appropriate responses to violent offenders is undermined and there are few strategies in place to either prevent the onset of violence, or to reduce the incidence of repeat offending.

The variables outlined above do not include all of the potential risk factors for violence. Yet they do serve to illustrate the complexity of the issue and the problems generated by a lack of appropriate data.

Identifying the predictors of Indigenous violence using multivariate analysis

Many of the factors that seem to be associated with an increased risk of Indigenous violent offending are themselves interrelated. By focusing on one variable at a time, it is not possible to identify those that are directly related to violence and those that are indirectly related through their association with one or more intervening variables. An alternative approach is to use more sophisticated statistical techniques to determine those factors that remain strongly associated with the likelihood that an individual will offend while simultaneously controlling for the effects of a range of other variables. This approach, which is based on comparing offenders and non-offenders within the Indigenous community, goes a long way to overcoming the limitations of univariate analyses described in the preceding section.

However, only a handful of studies have so far applied this methodology to Indigenous offending, as described below. In so doing, however, it should be noted that, as with the univariate studies, they too have limitations. In particular:

  • because of the type of data available for analysis, they are only able to identify factors associated with the likelihood of contact with the criminal justice system, rather than the likelihood of actually offending;
  • most focus on offending in general rather than violent offending in particular; and
  • they do not have access to statistics on, and so are unable to test the impact of, the full range of factors potentially influencing an individual's behaviour.

Predicting the likelihood of arrest using NATSIS data

A comparatively early study (Mukherjee et al. 1998) used data from the 1994 NATSIS to predict the probability of arrest among Indigenous males and females. The national survey found that 20 percent of Aboriginal and Torres Strait Islander people reported they had been arrested at least once during the five years immediately prior to interview in 1994. More than three times as many males were arrested than females. Almost half of Indigenous young men aged 18 to 24 years reported being arrested at least once, and for both males and females strong associations with the likelihood of arrest were found with age, state of residence, whether taken away from family as a child, labour force status and perceptions of relationships with police. These variables, along with urban or rural place of residence, recognition of own homeland and whether they had a place to meet for cultural activities made up the set of reference characteristics that were used in the logistical regression to predict the likelihood of an arrest. Results indicated that, for both men and women, five factors remained strongly predictive of the likelihood of being arrested by police at least once during the previous five years, once the effect of other relevant variables had been controlled for:

  • labour force status—the probability of arrest was significantly higher among unemployed males and females than among those in employment other than CDEP. This was the strongest predictor of male arrests (adjusted odds-ratio (OR) of 2.39) and the second strongest predictor of female arrests (OR=2.11);
  • relationship with police—those males and females who believed that relationships with police had deteriorated over the previous five years had a significantly higher probability of arrest than those individuals who did not hold this view (however, it is unclear whether their assessment of police relations impacted on their offending behaviour or whether the fact of being arrested coloured their views of police). This variable was the second strongest predictor of male arrests(OR=2.04) and the third strongest predictor of female arrests (OR=1.92);
  • whether the individual had been taken away from the family as a child—those Indigenous males and females who had been removed from their families were significantly more likely to be arrested compared with those who had not been taken away. This factor was ranked third in order of predictive capacity for males (OR=1.82) and fourth for females (OR=1.75);
  • age—for both sexes, the probability of arrest was significantly higher among young adults aged 20–24 years (males OR=1.39 and females OR=1.63) than was the case for those aged 25–44 years. In turn, those individuals in the younger (13–14 years) and older age groups (45 years and over) had lower predicted probabilities of arrest when compared with a standardised reference group. Among males, age was the fourth strongest predictor, while for females it was the fifth strongest; and
  • state of residence—among Indigenous males, state of residence was the fifth strongest predictor of arrest, with those living in Queensland (OR=0.53) and Tasmania (OR=0.94) recording a significantly lower risk than those from the reference state of New South Wales. Among females, state of residence was the most important predictor of arrest, with those living in South Australia (OR=2.19) having a significantly higher probability than those in the reference state of New South Wales, while those living in Queensland (OR=0.62) had a significantly lower probability. Again, it is unclear whether these results point to state-based variations in offending rates or whether they are the product of different policing procedures, legislation etc.

Two factors were predictive of female, but not male, arrests:

  • living in an urban area—those Indigenous women living in capital cities and other urban areas were marginally (OR=1.36) (but still significantly) more likely to have been arrested over the preceding five years than women living in rural areas, once the effect of other variables had been controlled for; and
  • having a place to meet for cultural activities—those women who had access to a meeting place had a marginally (OR=1.31) (but still significantly) lower probability of arrest than those who lacked such access.

A subsequent analysis of NATSIS data, this time disaggregated according to the type of offence involved in the most recent self-reported arrest in the previous five years, of those aged 13 years and over, was undertaken by Hunter (2001). The offence categories used were assault, theft, drinking-related offences (ie drinking in public or drink driving) and total offences. Of these, the most common offence for which Indigenous respondents had been arrested was drinking-related offences (16% of male and 5% of female arrests) while a smaller proportion (5% of males and 2% of females) had been arrested for assault.

Hunter (2001) goes into considerable detail to explain the variables selected for the model as indicators for a range of individual and household factors likely to increase the probability of contact with the criminal justice system. Several individual variables such as sex, age, labour force status and being taken away as child, were also examined by Mukherjee et al. (1998) but he tests for a wider range of individual-level (eg alcohol consumption) and household-level variables (eg number of residents) as well as a number of what he termed institutional variables (eg whether there were Indigenous police aides or liaison officers in the community and whether a respondent lived within 50 kilometres of a police station). In contrast to Mukherjee et al. (1998), Hunter excludes perceptions of police because of concerns of 'endogeneity bias' and the respondent's state of residence, as he assumes the influence of jurisdictional differences is adequately picked up by other variables. He also raises concerns about the 'current location' questions given the high rates of geographic mobility (Hunter 2001: 11).

Using a standard probit regression analysis, marginal effects were calculated as the difference in probability of arrest for a person with or without the specified characteristic, with all other characteristics fixed at average values. Males were 13.1 percentage points more likely to be arrested than females, the oldest age groups were 9.5 percentage points less likely to be arrested than other respondents (the peak age group for the probability of arrest was 18 to 24 years),Torres Strait Islanders were 7.7 percentage points less likely to be arrested than Aboriginal persons and those who were unemployed were 13.1 percentage points more likely to be arrested, Other important factors were educational outcomes, ever drinking alcohol (12.8 percentage points) and having been physically attacked or verbally threatened. Smaller marginal but significant (to the 5% level) effects were found for urban residence, the policing variables, long-term health condition, being taken from one's natural family and living in a crowded house.

Overall, the study found a relatively high degree of consistency in predictive variables across the different offence types. The factors that remained significantly and independently associated with the probability of being arrested for an assault, once the influence of other variables had been taken into account, were similar to those risk factors for arrests in general. However, the size of the marginal effect of each variable on the likelihood of arrest varied from one offence category to another, with alcohol consumption and being the victim of a physical attack or verbal threat being particularly important predictors of drinking-related arrests and assault arrests.

Six factors proved to be strongly predictive of arrests for assault. In order of magnitude, these were:

  • alcohol consumption—of the range of factors tested, this factor exerted the greatest effect, with those who reported 'ever' consuming alcohol being significantly more likely to be arrested for assault than the 'average' Indigenous person who had never consumed alcohol;
  • education levels—as the number of years of secondary schooling increased, the likelihood of being arrested for assault decreased, with the greatest likelihood of an arrest being concentrated among those individuals who had completed six to nine years of schooling only. Contrary to this trend, however, those who had no formal schooling or primary school education only had the least chance of an assault arrest. According to Hunter (2001: 21) this variable may be 'picking up the detrimental effect of imposing a largely alien education system onto Indigenous peoples with the consequent impact on their cultures and social cohesion';
  • gender—being male was associated with a higher probability of being arrested for assault, but the effect was lower than for the other offence types analysed;
  • victimisation experience—individuals who had been physically attacked or verbally threatened had a significantly greater likelihood of being arrested for assault than other Indigenous offenders. Hunter (2001: 22) concluded that 'this would seem to confirm the suspicion that there is a cycle of violence and abuse in Indigenous communities which is probably related to drinking related behaviour';
  • age—of the four age groups considered, those aged 45 years and over had the lowest probability of an assault arrest, while those aged between 25–34 years had the highest likelihood, although for the latter group the marginal effect was not statistically significant; and
  • labour force status—those who were unemployed had a greater likelihood of being arrested for assault than other Indigenous persons. Interestingly, those on CDEP had a lower risk of an assault arrest than those who were unemployed. Hunter (2001: 25) concluded that 'the continued expansion of the CDEP scheme is likely to play a role in mitigating the Indigenous over-representation in arrest statistics'.

The probability of being arrested for an assault was also slightly (but still significantly) higher for those who:

  • lived in a crowded house (defined as having two or more residents per bedroom); and
  • had a long term health condition.

In contrast, probabilities of an assault arrest were slightly, but significantly, lower for those who:

  • identified as Torres Strait Islander rather than Aboriginal—a finding which, as noted earlier, may be due to the smaller range of disadvantages experienced by Torres Strait Islander persons and that many of these individuals live in very remote parts of Australia;
  • lived in a rural area rather than elsewhere. In contrast, living in a capital city or in a remote area had no significant effect on the probability of arrest; and
  • lived in a community that had Indigenous police aides, which may indicate the availability of a culturally appropriate police service.

Variables that were not significantly predictive of the likelihood of an assault arrest included:

  • living with non-Indigenous people (used as an indicator of better economic prospects);
  • the quality of the housing stock (as measured by whether household utilities were available and working);
  • living with householders who themselves had been arrested (considered to be a proxy for peer group pressure);
  • living within 50 kilometres of a police station (indicative of access to policing services);
  • living in a family with at least one dependant (which may imply greater carer responsibilities); and
  • living in a household where other members had voted (a variable which may act as a proxy measure of the extent of an individual's social networks; Hunter 2001: 8).

Of particular note was the finding that being removed from their natural family had no marginal effect on an individual's probability of being arrested for assault. This is in marked contrast to the results obtained when all arrests were considered, irrespective of offence type. When all Indigenous persons who reported they had been arrested in the previous five years were considered, the likelihood of arrest for those persons who had suffered removal from family was higher than for those who had not been removed. Hunter (2001) noted that this variable's lack of statistical association with the likelihood of an assault arrest may be due to the relatively small number of respondents arrested for this type of offence and the attendant reduction in the power of the statistical analysis. Larger sample sizes may have produced different results.

Predicting the likelihood of being charged by police

Several later studies, this time using self-report data collected by the 2002 NATSISS, also sought to identify those factors that seemed to predict Indigenous contact with the criminal justice system (Weatherburn, Snowball & Hunter 2008, 2006). In contrast to Hunter (2001) these two studies did not examine different types of offences (so it is not possible to single out factors that increase the probability of arrest or charges for a violent offence) and a wider range of t measures of contact with the criminal justice system are employed—whether or not the individual had 'ever' been charged by police, whether they had been imprisoned in the previous five years and whether they had been arrested in the last five years and the number of arrests in that period. Additional predictor variables, which were not available in the earlier survey, were also examined.

Of the 8,523 adults aged 18 years and over surveyed by the NATSISS, approximately 36 percent indicated they had been charged at least once by police at some stage in their lives and the likelihood of being imprisoned in the past five years was one in 13. In the first study, an initial univariate analysis identified a wide range of variables (such as gender, educational attainment, alcohol use, employment status, principle income source, financial stress etc) that, when analysed separately, were significantly associated with the likelihood of ever being charged or being imprisoned in the previous five years. In the next stage of analysis, multivariate logistic regression was used to determine which of this long list of factors remained predictive of police charging (model's pseudo R2=0.196) and imprisonment (model's pseudo R2=0.0829) when the influence of the other factors had been partialled out (Weatherburn, Snowball & Hunter 2006).

Of the variables tested, the ones that exerted a large negative effect on (ie substantially increased the likelihood of) being charged by police included:

  • being male rather than female—of all the factors tested, this one exerted the strongest marginal effect on the likelihood of being charged (OR=4.69);
  • being a substance user and a high risk user of alcohol—these were the second (OR=2.86) and third (OR=2.6) strongest predictors. The authors noted that 'for an average [Indigenous] person, being a substance user increases the probability of being charged by almost 13 percentage points. Being a high risk user of alcohol increases the risk of being charged by over 11 percentage points' (Weatherburn, Snowball & Hunter 2006: 10–11); and
  • being unemployed rather than being employed or not in the labour force—those who were unemployed had a greater likelihood of being charged (OR=1.64) than those on CDEP, who, in turn, had a greater likelihood of being charged than those in other types of employment or not in the labour force (OR=1.23). The report therefore concluded, as did the earlier study by Hunter (2001), that being in a CDEP scheme 'appears to provide a protective effect against the risk of being charged when compared with those who are unemployed' (Weatherburn, Snowball & Hunter 2006: 12).

Smaller positive effects were found for age (18 to under 25 years versus those 25 years or over; OR 0.82), not completing Year 12 (OR= 0.52), sole parent with dependent children (OR=1.22), living in a crime-prone area (OR= 1.31), welfare income source (OR=1.55), financial stress (OR= 1.62), person or family member of 'stolen generation' (OR=1.45), no social involvement (OR= 1.35) and living in a major city versus remote (OR=0.77). Social support, large family, crowded household and social stressors were not significant predictors of being charged. The marginal effects for the 'charged' model were significantly larger than those for the 'imprisoned' model. However, the authors concluded that the 'most powerful predictors' of being charged or imprisoned (other than the sex of the respondent) were alcohol consumption and drug use (Weatherburn, Snowball & Hunter 2006: 10). Although the effect for substance abuse on the risk imprisonment is lower, it is higher than any other effect in the imprisonment. Year 12 completion and unemployment exerted similar effects on the risk of imprisonment, while being on welfare exerted a bigger effect than high-risk alcohol consumption; the CDEP variable was not significant. The differences in the strength and significance of predictor variables between the charged and imprisoned models were attributed by the authors to both the sample size and to differences in the factors that lead to being charged and imprisoned. Importantly, they note that violent offenders are more likely to be imprisoned than non-violent offenders, which could partly explain the differences.

Predicting frequency of arrest

A subsequent study sought to identify some of the 'main predictors and correlates of Indigenous arrest frequency' (Weatherburn, Snowball & Hunter 2008: 310). However, rather than investigating the likelihood of being charged by police (as was the case in the first report), it focused on whether or not 8,523 respondents to the NATSISS had been arrested, as well as the number of arrests in the five years preceding the survey of 1,390 respondents. The variables included as potential predictive variables were similar to those used in their first study and a probit modelling approach ensured the methodology was consistent with the Hunter (2001) study. Of the significant independent variables in the two models, the main caveat related to drug abuse because, although 90 percent of respondents answered the question, ABS had concerns about data quality and did not release results for remote areas. However, based on further analysis, the authors decided to include this variable but caution is urged when considering the results of this variable (Weatherburn, Snowball & Hunter 2008). Both models provided good fit to the data, having non-significant Hosmer and Lemeshow tests.

The frequencies of the variables used in the models were as follows: aged 18 to under 25 years 18.7%, male 42.3%, one parent family 20.5%, urban residence 15.4%, regional residence 40.1%, remote 44.5%, community or neighbourhood problems 76.1%, social engagement 87.8%, welfare dependence 67.5%, unemployed 16.1%, CDEP 9.8%, stolen generation 37.8%, alcohol abuse 6.8%, drug abuse 22.2%, financial stress 44.1% and Year 12 completion 15.1%.

In terms of the risk of arrest, no significant interaction effects between alcohol or illicit drug abuse and welfare, unemployment or financial stress were identified. Fewer than one in 10 respondents indicated they engaged in risky alcohol consumption but the most powerful predictive factor apart from gender was alcohol abuse (parameter estimate b=0.64). However, drug abuse (b=0.59), welfare dependence (b=0.42), unemployment (b=0.36), financial stress (b=0.36), being a member of a one parent family (b=0.22) or part of the stolen generations (b=0.19), being less than 25 years of age (b=0.24), living in a crime prone area (b=0.12) and being on CDEP rather than in other forms of employment (b=0.19) all increased the risk of arrest (again, though, being on CDEP reduced the risk when compared with those who were unemployed). Completing Year 12 and social involvement significantly reduced the risk of arrest. Finally, respondents living in urban and regional areas of Australia were less likely to be arrested than those in remote areas.

In terms of the number of arrests (excluding those who had never been charged by police), no significant interaction effects were identified between alcohol or illicit drug use and the socioeconomic variables tested. Alcohol abuse remained the most powerful predictor other than sex, followed by welfare dependency. Drug abuse, being unemployed and having limited social involvement also had an effect on the number of arrests. In contrast, living in a crime-prone area, being a member of the stolen generation, financial stress in the past 12 months, completing Year 12 and region of residence had no significant effect. The factors that were predictive of the likelihood of arrest were, therefore, somewhat different from those that predicted the actual number of arrests among those who had been apprehended at least once. Two possible explanations are offered for this difference—differences in sample sizes or that some variable play a role in shaping 'the risk of involvement in crime, but play little or no role in shaping the frequency of contact with the criminal justice system among active offenders' (Weatherburn, Snowball & Hunter 2008: 318)

Overall, this study confirmed the previous findings that alcohol use was a stronger predictor of both the likelihood and frequency of being arrested by police than were factors such as drug abuse, unemployment, welfare dependency, financial stress, failure to complete Year 12 and lack of social support. Importantly, the absence of any apparent interaction effect between drug/alcohol abuse and factors indicative of socioeconomic disadvantage suggests that substance abuse is not simply a product or symptom of Indigenous disadvantage. This means that reducing Indigenous economic and social disadvantage and reducing Indigenous alcohol and drug use are potentially quite separate issues, with a reduction in one not necessarily leading to a reduction in the other. These findings, according to the authors, reinforce Pearson's (2001a) argument that drug and alcohol abuse, rather than being a symptom of cultural, social and economic disadvantage, now constitute problems in their own right and play an independent role in explaining Indigenous violence as well as in perpetuating Indigenous socioeconomic disadvantage.

Predicting the likelihood of violent recidivism among violent Indigenous offenders: A Western Australian study

Another empirical study relevant to this issue is that by Allan and Dawson (2002). The aim of their research was to identify the risk factors associated with violent re-offending among Indigenous persons in Western Australia and, in turn, develop a predictive risk assessment instrument for this offender group. Unlike the studies of Hunter (2001) and Weatherburn, Snowball and Hunter (2008, 2006), this study concentrated specifically on Indigenous violence. It also differed from the others in that it sought to identify risk factors associated with violent re-offending, rather than initial offending. Hence, its target group was those Indigenous offenders who had already experienced some contact with the criminal justice system.

The sample used for the analysis comprised 525 adult male Indigenous offenders in Western Australia who had been found guilty by the court of a violent or sexual offence and who had been identified by correctional services as requiring either a violence or sexual offender intervention program (Allan & Dawson 2002). Re-offending was defined as any subsequent finding of guilt for a violent or sex offence.

Of these 525 offenders, 48 percent had a violent offence recorded as their most serious index offence, while 21 percent had a family violence offence, 22 percent had a non-violent sexual offence and eight percent had a violent sexual offence. Information on 67 potentially predictive variables was extracted for each person from relevant Department of Justice files. These variables were then categorised into specific 'predictor domains' using three different methods of categorisation, each of which embodied a different conceptualisation of violence. This multifaceted approach allowed the study to test the predictive strength of a particular variable when it was combined with different arrays of factors based on different conceptual models. To take the variable of offence severity as an example, under what was defined as a 'static criminogenic and non-criminogenic' approach, this variable was combined with previous violent offences, previous non-violent offences, age of first offence, age of index offence, juvenile violence, previous prison term, previous sexual offences and history of perpetrating violent offences against family members. According to the Violence Risk Appraisal Guide (VRAG) instrument, offence severity was grouped with age at the time of the:

  • index offence;
  • offence related to payback behaviour;
  • offence related to active involvement in inter-family, inter-community or inter-regional feuding;
  • offence related to debts or money issues; and
  • unfeasible release plans.
Table 16: Predictors of violent re-offending among Indigenous persons in Western Australia
Predictors of violent re-offending Static, criminogenic and non-criminogenic grouping VRAG grouping Factors specific to Indigenous violence (after Memmott & National Crime Prevention 2001)
Concordance across all three conceptual groupings
Previous violent offences Yes Yes Yes
Victim died Yes Yes Yes
Alcohol misuse Yes Yes Yes
Previous male victim Yes Yes Yes
Poor anger/behavioural control Yes Yes Yes
Unfeasible release plans Yes Yes Yes
Exposure to violence/family violence from an early age Yes Yes Yes
Relationship instability Yes Yes Yes
Unresponsive to or non compliance with treatment (exclude mental illness) Yes Yes Yes
Victim received medical attention Yes Yes Yes
Concordance across two conceptual groupings
Irresponsibility and not caring about the needs of significant others Yes Yes
Lack of realistic long-term goals Yes Yes
Juvenile history of violent behaviour Yes Yes
Denial (ie won't accept responsibility for actions/minimisation/victim takes responsibility) Yes Yes
Drug misuse Yes Yes
Affect (ie restricted emotional responses/unable to deal with strong emotions?) Yes Yes
Age at time index offence was committed Yes Yes
History of perpetrating family violence or related to victims Yes Yes
Childhood problem behaviour, aggression and offending Yes Yes
Concordance across one conceptual grouping
Previous non-violent offences Yes
Active involvement in inter-family/community/region feuding Yes
Grandiose sense of self-worth Yes
Age of first offence Yes
Employment problems/status Yes
Criminal associates Yes
Solvent use Yes
Previous prison term Yes
Breach of any order Yes
Female victims Yes
Impulsivity (unplanned behaviour without thought for consequences) Yes
Offence severity Yes

Source: Allan & Dawson 2002

Under a third grouping, based on the causes of violence outlined by Memmott and National Crime Prevention (2001)—namely underlying causes, situational factors and precipitating causes—offence severity was combined with age of first offence, age at the time of the index offence, previous violent offences, previous non-violent offences, juvenile violence, previous prison term, breach of orders, history of sexual offences, history of family violence, child victims, male victims, female victims, animal victims, victim died and victim required medical attention.

Preliminary analysis indicated that the predictive factors varied depending on whether the offence involved was a violent or a sexual offence. Moreover, within the violent offender category, family violence perpetrators formed a distinctive sub-group, while among sexual offenders, those who did not use violence seemed to differ from those who were violent. The authors therefore noted that, ideally, separate analyses should be undertaken for each of these four groups. However, because of the small number of individuals available for analysis, the study was only able to differentiate between violent offenders and sex offenders.

The factors that proved to be predictive of violent re-offending within each of the three categorisation schemas are summarised in Table 16. As shown, 10 items proved to be predictive across all three conceptual groupings, while nine were predictive in two of the three groupings and 12 were found to be predictive in only one of the conceptual groupings.

The factors that proved to be predictive of sexual offending are summarised in Table 17. As shown, 11 variables were predictive across the three conceptual groupings, while four were predictive in two of the approaches and four factors were identified as significant in one approach only.

Table 17: Predictors of sexual re-offending among Indigenous persons in Western Australia
Predictors of sexual re-offending Static, criminogenic and non-criminogenic grouping VRAG grouping Factors specific to Indigenous violence (after Memmott & National Crime Prevention 2001)
Concordance across all three conceptual groupings
Age at time index offence was committed Yes Yes Yes
Juvenile history of violent behaviour Yes Yes Yes
Previous male victims Yes Yes Yes
Poor anger/behavioural control Yes Yes Yes
Unfeasible release plans Yes Yes Yes
Exposure to violence/family violence from an early age Yes Yes Yes
Lack of realistic long-term goals Yes Yes Yes
Denial (ie won't accept responsibility for actions/minimisation/victim takes responsibility) Yes Yes Yes
Unresponsive to or non-compliant with treatment (exclude mental illness) Yes Yes Yes
Had treatment prior to re-offending Yes Yes Yes
Sexual abuse during childhood Yes Yes Yes
Concordance across two conceptual groupings
Relationship instability Yes Yes
Impulsivity (unplanned behaviour without thought for consequences) Yes Yes
History of sexual offences Yes Yes
Childhood problem behaviour Yes Yes
Concordance across one conceptual grouping
Poor coping skills Yes
Age at first offence Yes
Previous violent offences Yes
Restricted emotional response/unable to deal with strong emotions Yes

Source: Allan & Dawson 2002

A long list of factors did not have any predictive value for either Indigenous violent or sexual re-offending, including whether the victim was a child or animal, whether the offender was a member of the stolen generation, lived with his/her primary caregiver until aged 16 years, had an absent father during childhood, had experienced problems at school, had been sexually abused while in an institution, had been physically or emotionally abused during childhood or while in an institution, was assessed as having superficial charm, became bored/needed stimulation, was manipulative or a pathological liar, perceived violent/sex offending to be acceptable behaviour, lacked remorse, exhibited intimacy problems, led a parasitic lifestyle, engaged in promiscuous sexual behaviour, had low education status, exhibited self-harm or suicidal ideation, exhibited identity issues or over-identification with masculine roles/stereotypes, engaged in paranoid behaviour, exhibited low self-esteem or stress associated with deaths in custody, had high levels of stress, had received treatment for mental illness or had an untreated mental illness, was involved in a relationship characterised by jealousy/jealous behaviour, where the violence was related to payback behaviour or to debts and money issues, or was associated with the viewing of pornographic material.

In the second stage of analysis, the study used the predictor variables identified in Stage 1 of the project to construct a risk-assessment instrument that would accurately differentiate between Indigenous re-offenders and non-re-offenders. Additional variables routinely collected by WA's Department of Justice were also included.

In the final model designed to predict violent re-offending, only four variables were retained:

  • age at first offence;
  • unfeasible release plans;
  • impulsivity (ie unplanned behaviour without thought for consequences); and
  • personal/emotional orientation.

However, while the model was able to accurately classify re-offenders using these four variables (with a classification accuracy of 95%), its ability to classify non-re-offenders (at 55%) was only marginally better than chance. The authors therefore concluded that their attempt to construct a risk assessment tool for Indigenous violent re-offending had failed.

In constructing a risk assessment model for re-offending among sexual offenders, the three factors that were the best predictors of such behaviour were:

  • unrealistic long-term goals;
  • unfeasible release plans; and
  • poor coping skills.

In contrast to the predictive model developed for violent offenders, the sex re-offending model was able to accurately classify both re-offenders and non-re-offenders (with a classification accuracy of 92% and 94% respectively).

The authors considered that the retention of these three factors within the sex re-offending model was particularly pertinent for policy development, in that at least two of them were what they termed 'dynamic factors', that is, factors that were susceptible to change. They argued that poor coping skills could be addressed while offenders were in prison, with follow-up assistance provided after release. Similarly, they argued that it should be possible to devise more feasible release plans for each individual. They did acknowledge, however, that the task of addressing unrealistic long-term goals may require more extensive intervention, not only with the individual, but also with his broader community to bring about long-term improvements in opportunity.

In terms of the limitations of this study, the authors drew attention to the relatively small sample sizes and the fact that their analyses were limited to those variables collected by criminal justice agencies. Potentially, there may be other factors impacting on violent and sexual behaviour for which they could not test because of a lack of data. They also pointed to their inability to test for what they considered to be 'subtle but fundamental' differences between different Indigenous communities (Allan & Dawson 2002: 22). They argued that, ideally, risk factors should be identified for each community separately to take account of variations in aspects such as geographic location, levels of de-culturation and acculturation, and language differences. Finally, they also acknowledged that, by focusing on risk factors, their analysis was one-sided because it failed to investigate the protective factors that may help to prevent recidivism among violent and sex offenders (Allan & Dawson 2002).

Despite these limitations, one of the most significant outcomes of this study was the finding that risk factors differed depending on whether generalised violence, family violence, non-violent sexual offending or violent sexual offending was being considered. The research, therefore, highlights the need for further investigations that differentiate between the types of violence involved rather than, as has been the tendency so far, to talk about violence as a single form of behaviour.

Summary of findings from the multivariate analyses

Although small in number, the multivariate analyses described above confirm that Indigenous offending (including violent offending) is multicausal. A large number of variables remain independently predictive of Indigenous offending (or more accurately, contact with the criminal justice system) after the effects of other factors have been partialled out. Nevertheless, as summarised in Table 18, there is some consistency in the predictive factors identified. The studies by Hunter (2001) and Weatherburn, Snowball and Hunter (2006) found that alcohol use/abuse, gender, education levels, age, labour force status and residential location were all significantly predictive of contact with the criminal justice system, with four of these variables also predictive of the frequency of contact (Weatherburn, Snowball & Hunter 2008). As additional variables were available in the NATSISS, Weatherburn, Snowball and Hunter (2008, 2006) also found that drug abuse, financial stress, welfare dependency, involvement in social activities and living in a crime prone area were also important.

There were, however, some differences between the studies. Hunter (2001), for example, found that family removal was not associated with the likelihood of an assault arrest although this variable did appear to be related to police contact when 2002 data were used.

The fact that results varied from one study to another is to be expected, given differences in the data sources (NATSIS versus NATSISS), the range of predictor variables tested, the age range of the respondent group (Hunter (2001), for example, included all persons aged 13 years and over, whereas Weatherburn, Snowball and Hunter (2008, 2006) focused on adults aged 18 years and over), the types of offending involved (assaults versus all offences) and the offending indicator used (ie 'ever charged', arrested in last five years, or number of arrests). Definitions also varied for what seemed to be the same predictive variable. For example, in terms of alcohol use, Hunter (2001) focused on whether or not a person had ever consumed alcohol, whereas the studies by Weatherburn, Snowball & Hunter (2008, 2006) used the extent of consumption in the previous 12 months.

One multivariate study not included in Table 18 is that by Allan and Dawson (2002) because, unlike the others, it focused on violent re-offending by those individuals already involved with the criminal justice system and used a much larger and broader array of potential predictors that the other analyses. Its findings again point to the multitude of factors that remain predictive of violence, once the effects of other variables have been controlled for. The study also indicated that, not only do the predictive variables change depending on the conceptual frameworks used to group them during the testing phase, more importantly, they also vary depending on the type of violence considered.

Table 18: Summary of findings from multivariate analyses of NATSIS and NATSISS data
Risk factor Murkherjee et al. (1998) Hunter (2001) Weatherburn, Snowball & Hunter (2006) Weatherburn, Snowball & Hunter (2006) Weatherburn, Snowball & Hunter (2008)
NATSIS: arrest in last 5 years, males/females, all offences NATSIS: arrest in last 5 years by most serious offence at last arrest=assault NATSISS: Charged ‘ever’, persons aged 18 years and over NATSISS: arrested in last 5 years, persons aged 18 years and over NATSISS: number of arrests in last 5 years for those aged 18 years and over who had been arrested
Alcohol misuse Significant Significant Significant Significant
Gender Significant Significant Significant Significant
Education level Significant Significant Significant Not significant
Age Significant Significant Significant Significant Significant
Labour force status (employed/unemployed Significant Significant Significant Significant Significant
Place of residence Significant – female only Significant Significant Significant Not significant
CDEP/non-CDEP Significant Significant Not significant
Illegal drug use Significant Significant Significant
Financial stress Significant Significant Not significant
Welfare dependence Significant Significant Significant
Involvement in social activity Significant Significant Significant
Living in crime prone area Significant Significant Not significant
Removal from family Significant Not significant Significant Significant Not significant
Living in family with one dependent/sole parent family Not significant Significant Significant Not significant
Living in crowded (large) household Significant Not significant Not significant
Victim experiences Significant
Long term health condition Significant
TSI/Aboriginal Significant
Social stressors Not significant Not significant
Lack of social support Not significant Not significant
Living in non-Indigenous household Not significant
Quality of housing Not significant
Other members of household arrested Not significant
Living within 50 kms of police station Not significant
Other household members had voted Not significant
Indigenous police aides Not significant
Relationship with police Significant
State of residence Significant
Access to cultural meeting place Significant – female only