NSW Bureau of Crime Statistics and Research

CRIME AND JUSTICE
Bulletin
NSW Bureau of Crime
Statistics and Research
Contemporary Issues in Crime and Justice Number 104
October 2006
The economic and social factors underpinning
Indigenous contact with the justice system: Results from the 2002 NATSISS survey
Don Weatherburn, Lucy Snowball and Boyd Hunter
This study uses the 2002 National Aboriginal and Torres Strait Islander Social Survey (NATSISS) to examine
the economic and social factors that underpin Indigenous contact with the criminal justice system. The analysis
shows that the Indigenous respondents to the NATSISS were far more likely to have been charged with, or
imprisoned for, an offence if they abused drugs or alcohol, failed to complete Year 12 or were unemployed.
Participating in the Commonwealth Development Employment Scheme (CDEP) appears to reduce the risk
of being charged (compared with being unemployed). Other factors that increase the risk of being charged
or imprisoned include: experiencing financial stress, living in a crowded household and being a member of
the ‘stolen generation’.
Introduction
Despite a concerted effort on the part
of all Australian Governments to reduce
Indigenous contact with the criminal
justice system, rates of Indigenous court
appearance and imprisonment are now
higher than they were at the time of the
Royal Commission into Aboriginal Deaths
in Custody. In New South Wales (NSW),
the rate of Indigenous appearance in
court on criminal charges is 13 times that
of non-Indigenous Australians (Snowball
& Weatherburn 2006). The rate of
Indigenous imprisonment in NSW is ten
times that of non-Indigenous Australians
(Australian Bureau of Statistics 2005a).
The high rate of Indigenous contact
with the criminal justice system is not
unique to NSW – it is found to a greater
or lesser extent in all Australian States
and Territories (Australian Bureau of
Statistics 2005a).
It is difficult to devise ways of reducing
Indigenous contact with the criminal
justice system without an understanding
of why Indigenous Australians are so
often prosecuted and imprisoned. One
way to approach this issue is to compare
Indigenous people who have had contact
with the justice system with those who
have not, in terms of factors already
known to increase the risk of prosecution
and imprisonment. The 2002 Australian
Bureau of Statistics (ABS) National
Aboriginal and Torres Strait Islander
Social Survey (NATSISS) provides an
opportunity to pursue this kind of research.
Apart from the fact that it is the only
existing nationally representative survey
that focuses explicitly on Aboriginal and
Torres Strait Islanders, it contains a
wealth of material highly pertinent to an
understanding of Indigenous involvement
in crime.1
earlier work by Hunter (2001), who used
The study reported here uses the 2002
NATSISS data (Australian Bureau of
Statistics 2005b) to identify factors which
are predictive of Indigenous contact with
the criminal justice system. It builds on
and the statistical methods used to
the 1994 National Aboriginal and Torres
Strait Islander Survey (NATSIS) data to
examine predictors of Indigenous arrest.
The present study differs from Hunter’s in
three main respects. Firstly, it examines
predictor variables that were not available
to Hunter in the 1994 NATSIS. Secondly,
rather than examine predictors of arrest,
the present study examines predictors of
being charged with an offence. Thirdly, in
addition to examining predictors of being
charged with an offence, we examine the
predictors of imprisonment.
The remainder of this bulletin is organised
as follows. The next section provides
some important theoretical and empirical
background. The third section provides
further detail about the NATSISS data
analyse it. We then present the results of
our analysis. The final section discusses
the results and outlines their implications
for policy.
B U R E A U
O F
C R I M E
Past research
The high rate of Indigenous contact
with the justice system is in large part a
reflection of the high rate of Indigenous
involvement in crime (Weatherburn,
Fitzgerald & Hua 2003; Snowball &
Weatherburn 2006). In searching for
possible predictors of Indigenous arrest
and imprisonment, it is useful to begin
by considering the factors that have
been shown to increase the risk of
involvement in crime. Although many of
the personal and family factors implicated
in offending (e.g. poor impulse control,
weak parental supervision, poor parental
disciplinary practices) are not measured
in the NATSISS, the survey does contain
a large number of questions that are
pertinent to offending. These include:
1.
age of person
2.
sex of person
3.
highest year of school completed
4.
labour force status
5.
principal source of personal income
6.
whether removed from natural family
7.
whether relatives removed from
natural family
8.
whether respondent is a member
of a sole-parent family
9.
presence of neighbourhood or
community problems
10.
whether respondent had days without
money for basic living expenses over
the previous 12 months
11.
large household
12.
crowded household
13.
whether respondent can call on
support in time of crisis
14.
social involvement
15.
social stress
16.
drug and alcohol use .2
A few brief comments on these factors
may facilitate an understanding of
their significance to crime, arrest and
imprisonment.
Age and sex
Age is important because the likelihood
of involvement in crime (and of being
S T A T I S T I C S
charged with an offence) increases rapidly
from the early teenage years, reaches a
peak between the ages of 20 and 24 years
and declines steadily after that (Baker
1998; NSW Bureau of Crime Statistics
and Research 2006). Gender is important
because studies of both self-reported and
officially recorded offending show that
males are more likely to offend, more likely
to be charged with a criminal offence and
more likely to receive a prison sentence
than females (Blumstein et al. 1986;
NSW Bureau of Crime Statistics and
Research 2006).
School performance/
retention
There is a large body of research showing
a close relationship between poor school
performance, early school leaving and
self-reported/officially recorded involvement
in crime (Blumstein et al. 1986; Baker
1998; Maguin & Loeber 1996; National
Crime Prevention 1999). Whether this is
because poor school performance/early
school leaving increases the risk of
offending, or because some other factor
(e.g. low academic ability) causes both,
is unclear (Maguin & Loeber 1996).
Measures that improve school performance
and/or retention, however, have been
shown to reduce the risk of juvenile
involvement in crime (MacKenzie 2002).
Unemployment
Studies tracking the behaviour of
individuals over time generally find a strong
relationship between unemployment and
crime, particularly where offenders from low
socio-economic status backgrounds are
concerned (Farrington et al. 1986; Good,
Pirog-Good & Sickles 1986; Thornberry
& Christensen 1984; Fagan & Freeman
1999). In their longitudinal study of 411
London boys, for example, Farrington et al.
(1986) found that low socio-economic status
offenders commit property crime at a higher
rate during periods of unemployment than
during periods when they are employed.
The Community Development
Employment Projects (CDEP) scheme
is one response to the chronically high
Indigenous unemployment rate. CDEP
participants get paid the equivalent of
their entitlement for unemployment benefit
in return for working, usually part-time,
A N D
R E S E A R C H
on a project that develops the local
Indigenous community (Altman, Gray &
Levitus 2005). CDEP scheme participants
have been found to be less likely to be
arrested than Indigenous persons who
are unemployed (Office of Evaluation and
Audit 1997). In this study, we compare
CDEP scheme participants both to those
who are employed in non-CDEP work and
to those who are unemployed.
Family disruption/
dissolution
Hunter (2001) found that Indigenous
Australians who were taken away from
their natural family were at significantly
higher risk of arrest. Although no other
study appears to have examined this
issue, Hunter’s finding is consistent
with other research showing that early
childhood trauma increases the risk of
juvenile involvement in crime (Loeber &
Stouthamer-Loeber 1986). A number of
studies have also shown that children in
sole-parent families are at heightened
risk of involvement in crime, particularly
where the sole caregiver is poor and/or
lacks a close friend, relative or neighbour
(Weatherburn & Lind 2001).
Neighbourhood problems
There is very little research into the
contribution of neighbourhoods to crime,
but Weatherburn and Lind (2001) found
that juveniles who are poorly supervised
by their parents are more likely to become
involved in crime if they live in a crimeprone neighbourhood than if they live
in a non crime-prone neighbourhood.
This finding was attributed to the greater
influence of delinquent peers in crimeprone neighbourhoods. A number of
studies have found that neighbourhoods
with a high percentage of unsupervised
peer groups generally have higher rates of
involvement in crime (Pratt & Cullen 2005).
Economic stress
Low socio-economic status and poverty
have long been known to be strong
correlates of both juvenile and adult
involvement in crime (Blumstein et al.
1986). For a while it was thought that this
correlation simply reflected bias in the
exercise of police discretion. It is now clear,
B U R E A U
O F
C R I M E
however, that the relationship between
economic well-being and offending,
although relatively weak for minor offences,
is quite strong for serious offences
(Blumstein et al. 1986). Recent research
suggests that financial stress increases
the risk of child neglect and abuse (and
other parenting problems) which, in turn,
increases the risk of juvenile involvement in
crime (Fergusson et al. 2004).
Large families/
household crowding
Children from large families have been
found to be more likely to get involved
in crime than children from families with
smaller numbers of children (Loeber &
Stouthamer-Loeber 1986), partly because
of the resource constraints that large
families face (Blumstein et al. 1986).
Although its causal status is unclear and
the avenue through which crowding might
effect crime is not obvious, the percentage
of ‘crowded households’ (i.e. households
with a large number of people relative
to the number of bedrooms) has also
been found to be strongly correlated with
percentage of residents in an area who
have a juvenile criminal record (National
Crime Prevention 1999; Weatherburn &
Lind 2001).
Lack of social support
and involvement
There is both direct and indirect evidence
suggesting that social support and social
involvement act to reduce the risk of
involvement in crime. The percentage
of residents who say they lack social
support is a strong independent predictor
of the level of crime in an area (Pratt &
Cullen 2005). Lack of social support and
lack of social involvement are also strong
independent predictors of child neglect and
abuse (Weatherburn & Lind 2001). Child
abuse and neglect, in turn, are known to
increase the risk of involvement in crime
(Loeber & Stouthamer-Loeber 1986).
Social stress
Since access to social support appears
to reduce the rate of involvement in
crime, one would expect social stress to
increase it. There is some evidence to
support this conjecture. Agnew and White
S T A T I S T I C S
A N D
R E S E A R C H
(1992) found that stressful life events
are strongly correlated with self-reported
involvement in crime even after controlling
for a variety of other factors known to
influence involvement in crime. Gendreau,
Little and Goggin (1996) have also found
interpersonal conflict and personal stress
to be strong independent predictors of
adult recidivism.
removed because they had a missing
value for the incarceration variable.
Drug and alcohol abuse
Our analysis proceeds in two stages. First
we examine the bivariate relationships
between the independent variables listed
above and the two dependent variables.
In the second stage, we conduct a
multivariate logistic regression analysis to
determine which independent variables
make an independent contribution to the
risk of being charged or imprisoned.8
The research literature on the relationship
between substance abuse and crime
is overwhelmingly supportive of the
hypothesis that drug and alcohol abuse
increase the risk of involvement in crime.
Illicit drug dependence increases the rate
of involvement in crime, at least in part
because of the high costs associated with
funding illicit drug dependence (Blumstein
et al. 1986). Alcohol abuse, on the other
hand, appears to exert a direct effect on
the proclivity of individuals to become
aggressive and violent in certain situations
(Exum 2006). Chikritzhs and Brady (2006)
have recently highlighted the problem
of Indigenous alcohol abuse. Delahunty
and Putt (2006) have recently documented
similar problems in relation to illicit drug use.
Data and Method
As already noted, the data for this study
are drawn from the 2002 NATSISS. This
survey, which was conducted from August
2002 to April 2003, involved interviews
with Indigenous people aged 15 years
or more living in private dwellings.3
The survey was administered in both
community and non-community areas.4
It had a response rate of 80 per cent
within non-community areas. In community
areas,5 78 per cent fully responded and
94 per cent partially responded. In total,
9,359 Indigenous persons living in 5,887
households were surveyed out of a total
Indigenous population of 282,205. In
other words, about one in 30 Indigenous
Australians took part in the survey.
Our analysis focuses only on adults
(respondents who were aged 18 years
or more at the time of the survey),
a group which accounted for 91.1 per
cent of the total Australian sample (8,523
respondents). Two respondents were
The dependent variables6 used in our
analysis were:
• whether the respondent had ever been
charged by police (‘Charged’); and
• whether the respondent had been
incarcerated in the five years previous
to the survey (‘Imprisoned’).7
Results
Bivariate Comparisons
The following bivariate comparisons
are statistically significant at the five per
cent level, unless otherwise stated. They
are also weighted using the appropriate
person weight included in the NATSISS
confidentialised unit record file (CURF).
For a point of comparison, Figure 1
shows the distributions of the ‘Charged’
and ‘Imprisoned’ variables across the
whole Indigenous adult population.
It shows that the likelihood of an
Indigenous person ever being charged 9 is
more than one in three, while the likelihood
of being imprisoned in the past five years
is one in 13.
Demographic variables
The first three variables we examine are:
age of respondent, sex of respondent,
and whether the respondent identified as
being of Torres Strait Islander origin.
Tables 1 and 2 consider the relationship
between the age and sex of an
Indigenous person and their probability
of being charged by the police or being
imprisoned. Table 1 shows that younger
respondents were more likely to have
been imprisoned but slightly less likely
to be charged than older respondents.
The relationship between a respondent’s
age and the ‘Imprisoned’ variable is easy
to interpret because the information on
B U R E A U
O F
C R I M E
imprisonment captures whether or not the
respondent had been imprisoned within
the previous five years. The ‘Charged’
variable, on the other hand, captures
whether or not the respondent has ever
been charged. Since older individuals
have been exposed to the possibility of
having been charged over a longer period,
the relationship between ‘Charged’ and
age is more difficult to interpret.
Table 2 shows that male respondents
were considerably more likely to be both
charged and imprisoned than female
respondents. The male-female ratio is
2.5:1 for charged and 4:1 for imprisoned.
The NATSISS includes a question that
asks respondents whether they identify
as being Aboriginal, Torres Strait Islander
or both. This question is only asked of
respondents who live in Queensland
(approximately 58 per cent of Torres
Strait Islander people), which makes
it difficult to draw conclusions about
the whole of the Torres Strait Islander
population. Table 3 shows the likelihood
of being charged or imprisoned according
to whether the respondent identifies as an
Aboriginal person, a Torres Strait Islander
or both. It can be seen that Torres Strait
Islander respondents were less likely
to be charged than Aboriginal people
(slightly under 1 in 4 compared with
more than 1 in 3) and to be imprisoned
(approximately 1 in 29 compared with
1 in 15). For the group who identified
as both Aboriginal and Torres Strait
Islander, the likelihood increased for both
characteristics, which suggests that they
need to be treated as a separate group in
the model.
Economic and labour
market indicators
The following tables explore the link
between contact with the criminal justice
system and a number of economic
indicators that were available through the
NATSISS, namely: labour force status
and involvement in a CDEP scheme,
principal source of income, days without
money for basic expenses, and highest
level of schooling completed.
Table 4 shows the percentage of
respondents charged and imprisoned
by their labour force status. For both
S T A T I S T I C S
A N D
R E S E A R C H
Figure 1: Percentage of respondents who were charged or imprisoned
Percentage (%)
40.0
35.0
30.0
25.0
20.0
15.0
10.0
5.0
0.0
Charged
Imprisoned
Table 1: Percentage of respondents who were charged or imprisoned by age
18 to 24
25 to 34
35 to 44
45 plus
%
%
%
%
Charged
36.9
38.3
42.1
31.7
Imprisoned
10.7
9.4
7.6
3.3
Table 2: Percentage of respondents who were charged or imprisoned by sex
Male
Female
%
%
Charged
54.1
21.6
Imprisoned
12.4
3.1
Table 3: Percentage of Queensland respondents who were charged
or imprisoned by Indigenous status
Charged
Imprisoned
Torres Strait Islander
Aboriginal
Both
%
%
%
23.9
36.5
44.3
3.5
6.8
8.3
B U R E A U
O F
C R I M E
the charged and imprisoned variables,
the ‘employed’ group have a similar
distribution to the ‘not in the labour force’
(NILF) group. Both have approximately
a one in three likelihood of having been
charged and a one in 17 chance of
previous imprisonment. The ‘unemployed’
group are more likely to be charged
(almost 3 in 5) and considerably more
likely to be imprisoned (1 in 5).
Of the employed group, 26 per cent are
in a CDEP scheme. Whether to treat
this group separately from the rest of the
employed group is a difficult question.
Table 5 considers the distribution of the
two variables of interest with respect to
whether a respondent, having stated they
were employed, was in a CDEP scheme.
The two groups have different likelihoods
for both variables. The CDEP group
have just over a two in five likelihood of
having received a formal charge and a
one in eight chance of imprisonment. The
non-CDEP group have just under a one
in three chance of charge and a one in 28
chance of imprisonment.
Table 6 considers the relationship
between a person’s principal source
of income and their contact with the
criminal justice system. The ‘Welfare’
category includes both CDEP payments
and government cash pensions and
allowances. Not included in this table
were 275 respondents (3.2%) for whom
the question was not applicable or the
response not stated.
The group receiving welfare as their
principal source of income was more
likely to be imprisoned (almost 1 in 10)
and charged (more than 2 in 5) than
either the group receiving income from
wages, business or property; or income
from another source.
The NATSISS asks households whether
they had days without money for basic
living expenses in the previous 12 months.
We used this as a measure of financial
stress. Table 7 looks at the relationship
between this variable and a person’s
likelihood of being charged or imprisoned.
It is clear that respondents who lived
within households that had experienced
financial stress were more likely to be
both charged and imprisoned.
S T A T I S T I C S
A N D
R E S E A R C H
Table 4: Percentage of respondents who were charged or imprisoned by their labour force status
Employed
Unemployed
Not in the
labour force
%
%
%
33.9
57.6
34.4
5.8
20.0
5.7
Charged
Imprisoned
Table 5: Percentage of respondents who were charged or imprisoned by whether they were involved in CDEP
Employed – CDEP
Employed – Non-CDEP
%
%
Charged
43.7
30.4
Imprisoned
12.0
3.6
Table 6: Percentage of respondents who were charged or imprisoned by their principal source of income
Welfare
Wages or
business/property
Other
%
%
%
41.2
29.7
30.4
9.8
3.3
4.1
Charged
Imprisoned
Table 7: Percentage of respondents who were charged or imprisoned by their principal source of income
Had days
without money
Did not have days
without money
%
%
Charged
45.4
30.6
Imprisoned
10.2
5.5
Table 8 explores the link between
For respondents who only completed
educational attainment and involvement in
either Year 10 or Year 11, the likelihood
the justice system. Clearly, respondents
rises to approximately one in 2.6 for being
who stayed longer at school were less
charged and one in 16 for imprisonment.
likely to be either charged or imprisoned.
It rises further for students who completed
For respondents who completed Year 12,
only Year 9 or below (or who did not
their likelihood of being charged is
attend school). This group stands a one
approximately one in five and their
in 2.4 chance of being charged and a one
likelihood of imprisonment is one in 30.
in ten chance of being imprisoned.
B U R E A U
O F
C R I M E
Social indicators
The following tables explore the link
between being charged and being
imprisoned, and a number of social
indicators. These are: whether the
household is ‘crowded’, number of
dependents in the household, whether the
household is a ‘one-parent’ household,
whether the respondent is a member of
the ‘stolen’ generation or has a relative
who was taken away from their natural
family, whether the respondent has social
support and whether the respondent is
socially isolated.
The link between household crowding,
the number dependents in a household
and a person’s involvement with the
criminal justice system is explored in
Table 9. We define a ‘large’ family as a
household with three or more dependents
and a ‘crowded’ household as one in
which the ratio of the number of people
per bedroom is more than two.10
The effect of living in a crowded
household is far more pronounced for
the ‘Imprisoned’ variable than for the
‘Charged’ variable. Living in a large
household has no effect on the risk of
being charged or imprisoned.
Table 10 looks at the effect of living in a
sole-parent household with dependent
children or students. Living in a soleparent household appears to reduce
the likelihood of being charged or
imprisoned. This is in contradiction to
most literature on the topic and will need
to be considered again when looking at
the results of the multivariate model.
Table 11 examines the effect on the risk
of being charged or imprisoned of either
having been removed from your natural
family or of having had a relative who
was removed from his/her natural family.
The risk of being charged or imprisoned
is higher in both cases but is more
pronounced, as would be expected, in
relation to the respondent’s own removal.
The likelihood of being charged rises
from slightly more than one in three for
respondents who were not removed to
over one in two for respondents who
were removed. For those with relatives
removed (irrespective of whether they
themselves were removed), the likelihood
S T A T I S T I C S
A N D
R E S E A R C H
Table 8: Percentage of respondents who were charged or imprisoned by the highest level of school completed
Charged
Imprisoned
Year 12
Year 10 or 11
Year 9 or below
%
%
%
21.0
38.8
42.5
3.3
6.6
10.4
Table 9: Percentage of respondents who were charged or imprisoned by crowded household and number of
dependents in the household
Number of dependents
in household
Crowded household
Yes
No
2 or less
3 or more
%
%
%
%
Charged
37.7
36.9
36.8
37.7
Imprisoned
12.7
6.6
7.4
7.9
Table 10:Percentage of respondents who were charged
or imprisoned by family type
Charged
Imprisoned
One parent with
dependents
Other family
type
%
%
35.9
37.3
7.2
7.6
Table 11: Percentage of respondents who were charged or imprisoned by removal from natural family
Removed from
natural family
Relative removed
from natural family
Yes
No
Yes
No
%
%
%
%
Charged
53.6
35.8
42.5
33.1
Imprisoned
16.5
6.7
10.3
5.8
rises from one in three to one in 2.4. The
effect on imprisonment is even stronger,
with the risk of being imprisoned rising from
one in 16 to one in six for those removed
from their natural family. For respondents
with a relative removed, the same risk rises
from one in 17 to one in ten.
In order to measure whether a
respondent felt they had social support,
we used the question asking whether
they felt they had support in the time
of a crisis, which required a simple
yes/no response. Table 12 looks at the
relationship between social support
B U R E A U
O F
C R I M E
S T A T I S T I C S
Table 12: Percentage of respondents who were charged or imprisoned by social support
Has support
Does not have support
%
%
36.0
46.8
6.9
13.5
Charged
Imprisoned
Table 13: Percentage of respondents who were charged or imprisoned by social isolation
Involved in
social activities
Not involved in
social activities
%
%
36.4
41.7
7.6
7.2
Charged
Imprisoned
Table 14: Percentage of respondents who were charged or imprisoned by stressors
Stressors in
last 12 months
No stressors in
last 12 months
%
%
37.7
36.9
7.6
7.5
Charged
Imprisoned
Table 15: Percentage of respondents who were charged or imprisoned by community problems
No problems
%
%
Imprisoned
39.0
31.5
8.0
6.1
Table 16: Percentage of respondents who were charged or imprisoned by remoteness
Charged
Imprisoned
R E S E A R C H
and whether the respondent has been
charged or imprisoned. For respondents
who felt they had social support, the
likelihood of having been charged was
just over one in three and their likelihood
of imprisonment was approximately
one in 14. For those who felt they did
not have support, the likelihood jumps
to almost one in two for being charged
and approximately one in seven for
imprisonment.
Social involvement is another factor of
interest. For this variable, we used the
NATSISS question asking whether the
respondent had been involved in social
activities in the last three months. Table
13 shows that respondents who were
involved in social activities were less
likely to be charged. The difference for
imprisonment is only just significant
(at the 5% level). Taken at face
value, Table 13 suggests social
involvement increases the likelihood
of being imprisoned, which is a rather
questionable finding in the light of past
research.
The NATSISS asks whether a respondent
has faced any type of stressor in the
previous 12 months and then asks them
to specify which type or types of stressors
they experienced. Because we measured
alcohol, drug use, employment and
crime separately, we only considered
respondents who indicated another
type of stressor. Neither difference is
significant at the five per cent level.
Geographic variables
Neighbourhood/
community problems
Charged
A N D
Major city
Regional
Remote or
very remote
%
%
%
33.2
41.1
34.8
7.5
6.6
9.0
Table 15 looks at the effect of living
in a crime-prone area on the risk of
having been charged or imprisoned.
We define a crime-prone area as one
where the respondent has stated that
neighbourhood or community problems
exist. This table does not include the
106 respondents (1.24% of the sample)
who did not know or did not state
whether they lived in an area with
problems.
Living in a crime-prone area increases
the likelihood of being charged and
imprisoned. The likelihood rises from
approximately one in three to two in five
for charged and one in 16 to one in 12 for
being imprisoned.
B U R E A U
O F
C R I M E
Areas are classified as remote/very
remote, regional or a major city using
the Australian Standard Geographic
Classification (ASGC) scale. Table 16
shows that location has a differing impact
on a person’s likelihood of coming into
contact with the criminal justice system.
Respondents living in remote areas are
about as likely as those in major cities to
be charged (1 in 3). Both groups are likely
to be charged than those living in regional
areas (2 in 5). However, for imprisonment,
those living in remote areas have the
highest likelihood (1 in 11 compared with
1 in 16 for regional respondents and
1 in 13 for respondents living in major
cities). Because of the high correlation
between location and the other variables
considered, the results may change when
controlling for other factors in the model.
Alcohol and substance use
The National Health and Medical
Research Council (NHMRC) guidelines
were applied to define the relative risk
levels for alcohol consumption used
in the NATSISS. We used the variable
tracking alcohol consumption over
the previous 12 months, as opposed
to the previous two weeks, in order
to examine a respondent’s long-term
S T A T I S T I C S
alcohol usage. Table 17 shows a clear
relationship between alcohol consumption
and involvement with the criminal justice
system. For high-risk users of alcohol,
the likelihood of being charged is
approximately three in five, compared with
one in four for non-consumers, two in five
for low-risk consumers and one in two for
medium-risk consumers. The same effect
is seen for imprisonment, with the
likelihood over one in five for high-risk
consumers as compared with slightly
under one in 20 for non-consumers,
approximately one in 13 for low-risk
consumers and one in 12 for medium-risk
consumers. Note that this table does not
include 58 respondents (0.68% of the
sample) who did not state their usage.
The substance abuse question was
administered separately for respondents
within the community and non-community
samples of NATSISS (with the latter
sample being concentrated in remote
Australia). Respondents living in noncommunity areas filled out a separate form
specifying their usage of illicit substances.
Respondents living in such communities
were required to respond verbally to the
interviewer. The ABS has cited a low
prevalence rate for people in community
areas and subsequently did not release
Table 17: Percentage of respondents who were charged or imprisoned by alcohol consumption
Does not
consume alcohol
Charged
Imprisoned
Low-risk Medium-risk High-risk
%
%
%
%
25.8
39.2
50.3
61.3
4.6
7.6
8.5
22.6
Table 18: Percentage of respondents who were charged or imprisoned by substance abuse
Charged
Imprisoned
Never abused substances
Substance abuser
%
%
26.9
52.4
3.5
11.8
A N D
R E S E A R C H
these results. Table 18 is based only on
respondents who lived in non-remote
areas.
We used the variable ‘ever used
substances for non-medical purposes’ as
a proxy for substance abuse. Table 18
looks at the relationship between being
a substance user and imprisonment or
charge. Just over one in four respondents
who had not abused substances had
been charged. This compares with more
than half of those respondents who had
abused substances at some stage.
For imprisonment, the relationship is
more pronounced. Approximately one in
29 respondents who said they were not
substance abusers had been imprisoned
compared with more than one in nine
respondents who had abused substances
at some stage. The following table does
not include 359 respondents (3.97% of
the sample) who did not state their drug
use or did not respond to this question.
Multivariate logistic
regression
The logistic regression models for the
‘Charged’ and ‘Imprisoned’ variables were
estimated using unweighted data from
the 8,521 respondents who met the study
criteria described previously.
All variables discussed in the above
bivariate analysis were tested for
explanatory significance in the model.
A number of variables had ‘Not stated’,
‘Don’t know’ and ‘Don’t want to answer’
responses. In order to retain as much
information as possible, we created
additional variables that reflected whether
the response for that question took one of
these three forms. In most cases, these
created variables were not significant.
Only those created variables that were
found to be significant at the five per cent
level were retained.
For the imprisonment model, two
variables were retained which were
found to be not significant at the five per
cent level. These were retained in order
to facilitate comparison across the two
models. For the same reason, the Torres
Strait Islander variable was omitted from
both models.11
B U R E A U
O F
C R I M E
• consumes alcohol in a manner which
is not considered high-risk
Charged
For the ‘Charged’ model, the base case is
a female who:
• is aged 25 years or over
• is employed or not in the labour force
• does not receive welfare as their
principal source of income
• is not experiencing financial stress
S T A T I S T I C S
• has never consumed illicit substances.
The parameter estimates and odds ratios
for the full model are presented in Table 19,
along with their associated confidence
intervals in brackets.
The model suggests the following:
• is not a member of the ‘stolen
generation’ and has no relatives who
were removed from their natural
family
• has a highest level of school completed
that is equivalent to Year 11 or less
• does not live in a sole-parent family
with dependents
• has social support
• lives in a remote area
• does not live in a crime-prone area
• Age, as in the bivariate analysis,
does not have a large marginal effect
on the probability of being charged.
The model suggests that being
under the age of 25 years slightly
reduces the probability. However, this
result could be due to the fact that
younger respondents have had less
opportunity to be charged, compared
with older respondents (i.e. ‘Charged’
is cumulative over time).
• Sex is a very powerful explanatory
variable, with males being much
Table 19: Results from the logistic regression model for the ‘Charged’ variable
A N D
R E S E A R C H
more likely to be charged with an
offence, holding other characteristics
constant. This is the largest effect in
the model.
• Being unemployed, as opposed to
being employed or NILF, has quite a
large effect on the probability of being
charged. Its effect is comparable to
that of the financial stress variable
and both have a greater impact
on being charged than whether a
person’s principal source of income is
a welfare payment.
• Inspection of the relevant parameter
estimates indicates that, by
comparison with being unemployed,
being a member of a CDEP scheme
reduces the probability of being
charged. However, the probability is
larger for this group than for those
who are employed in a non-CDEP
scheme or NILF.
• Education also has quite a large effect
on the probability of being charged,
with a reduction in probability for
those who finish Year 12.
• Belonging to a sole-parent family with
dependents and not being involved
in social activities both have a small
positive effect on the probability of
being charged. Living in a crime-prone
area has a similar positive effect on
the probability of being charged.
Comparison
Parameter
estimate
Odds Ratio
(with CI)
Intercept
-2.64 (0.13)
N/A
Under 25 years vs 25 years and over
-0.20 (0.07)
0.82 (0.72 - 0.94)
Male vs Female
1.54 (0.06)
4.69 (4.21 - 5.22)
Unemployed vs Employed or NILF
0.49 (0.09)
1.64 (1.38 - 1.94)
CDEP vs Employed or NILF
0.21 (0.08)
1.23 (1.06 - 1.42)
Welfare vs Other income source
0.44 (0.06)
1.55 (1.38 - 1.76)
Financial stress vs No financial stress
0.48 (0.05)
1.62 (1.46 - 1.79)
Completed Year 12 vs Did not complete Year 12
-0.66 (0.08)
0.52 (0.44 - 0.61)
Person or family member of ‘stolen generation’
vs Person or family not a member of the
‘stolen generation’
0.37 (0.05)
1.45 (1.30 - 1.60)
Sole-parent family vs Other family type
0.20 (0.07)
1.22 (1.07 - 1.40)
No social involvement vs Social involvement
0.30 (0.08)
1.35 (1.16 - 1.57)
Major city vs Remote
-0.47 (0.09)
0.77 (0.68 - 0.88)
Regional vs Remote
-0.26 (0.07)
0.63 (0.53 - 0.75)
Lives in a crime-prone area vs Does not live
in a crime-prone area
0.27 (0.06)
1.31 (1.16 - 1.48)
High-risk alcohol use vs Not high-risk alcohol use
0.96 (0.10)
2.60 (2.13 - 3.17)
Substance use vs No substance use
1.05 (0.00)
2.87 (2.49 - 3.31)
Imprisoned
Substance use missing vs No substance use
0.44 (0.13)
1.55 (1.20 - 2.01)
For the ‘Imprisoned’ model, the base case
is female who:
Hosmer-Lemeshow = 3.26 (p = 0.917)
-2 Log Likelihood = 9345.1
Pseudo R2 = 0.196
• The less remote a person’s location
is, the smaller the chance of being
charged. Living in a regional area, as
compared with a remote area, has a
small negative effect. Living in a major
city has quite a large negative effect.
• High-risk consumption of alcohol
and use of illicit substances exert
very large effects on the chance of
being charged. Respondents who
did not state whether they had used
substances were more likely to be
charged than those who stated they
had never used substances.
• Social support, large family, crowded
household and social stressors were
not significant predictors of being
charged.
• is aged 25 years or over
• is employed or not in the labour force
B U R E A U
O F
C R I M E
• does not receive welfare as their
principal source of income
• is not experiencing financial stress
• lives in a non-crowded household
(6 people or less)
• is not a member of the ‘stolen
generation’ and has no relatives who
were removed from their natural family
• has a highest level of school completed
that is equivalent to Year 11 or less
• lives in a remote area
• consumes alcohol in a manner which
is not considered high-risk
• has never consumed illicit
substances.
The parameter estimates and odds ratios
for the full model are presented in Table 20.
The model suggests the following:
• Sex is again a very powerful indicator
of whether or not someone has been
imprisoned in the last five years.
S T A T I S T I C S
As with the ‘Charged’ model, it has the
largest coefficient in the model.
• Age is not a significant predictor of
having been imprisoned within the
previous five years.
• Being unemployed, as opposed to
being employed or NILF, again has
a very large effect on the probability
of imprisonment. However, the effect
of being in a CDEP scheme on the
probability of being imprisoned is not
significant.
• Completing Year 12 reduces the
chances of imprisonment.
• Living in a ‘crowded’ household
increases the chances of having been
imprisoned.
• As with the ‘Charged’ model, being
a member or having a relative
who was a member of the ‘stolen
generation’ increases the probability
of imprisonment.
Table 20: Results from the logistic regression model for the ‘Imprisoned’ variable
Comparison
Parameter
estimate
Intercept
-4.78 (0.17)
N/A
Male vs Female
1.49 (0.10)
4.45 (3.65 - 5.44)
Under 25 years vs 25 years and over
0.17 (0.11)* 1.19 (0.96 - 1.47)
Unemployed vs Employed or NILF
0.63 (0.12)
CDEP vs Employed or NILF
0.15 (0.12)* 1.16 (0.92 - 1.47)
Welfare vs Other income source
1.07 (0.13)
2.92 (2.25 - 3.79)
Financial stress vs No financial stress
0.37 (0.09)
1.45 (1.21 - 1.74)
Completed Year 12 vs Did not complete Year 12
-0.59 (0.16)
0.56 (0.40 - 0.77)
Crowded household vs Non-crowded household
0.29 (0.12)
1.34 (1.06 - 1.69)
Person or family member of ‘stolen generation’
vs Not a member of the ‘stolen generation’
0.48 (0.09)
1.61 (1.34 - 1.93)
Major city vs Remote
-0.94 (0.19)
0.39 (0.27 - 0.56)
Regional vs Remote
-0.94 (0.15)
0.39 (0.29 - 0.53)
High-risk alcohol use vs Not high-risk alcohol use
1.00 (0.12)
2.71 (2.13 - 3.45)
Substance use vs No substance use
1.21 (0.15)
3.36 (2.49 - 4.53)
Substance use missing vs No substance use
0.57 (0.26)
1.77 (1.06 - 2.97)
Hosmer-Lemeshow = 7.40 (p = 0.495)
-2 Log Likelihood = 3664.6
Pseudo R2 = 0.0829
* Variable not significant at the five per cent level.
10
Odds Ratio
(with CI)
1.88 (1.48 - 2.39)
A N D
R E S E A R C H
• Living in either a regional area or
major city significantly reduces the
chances of having been imprisoned.
• High-risk consumption of alcohol
and illicit substance use has a
substantial effect on the probability
of imprisonment. Substance abuse is
the second largest effect in the model.
High alcohol use is the fourth largest
effect. As with the ‘Charged’ model,
people who did not state whether they
used substances were more likely to
have been sentenced to prison.
• Living in a one-parent family,
involvement in social activities, living
in a crime-prone area, having a large
family and social stress were not
significant predictors of imprisonment
at the five per cent level.
Largest marginal effects
for both models
In order to assess the effect of each
variable in relation to others, it is useful
to consider its marginal effect. To
determine marginal effects, we calculate
the risk of being charged (or imprisoned)
for an average respondent and then
examine the effect of changing one of the
respondent’s characteristics. Using the
median values for all of the characteristics
we examine, we define an ‘average
respondent’ as a respondent who is
female, aged 25 years or over, does not
participate in social activities and lives in
a regional area that is crime-prone. Each
other variable has been given the value of
zero, except for the variable of interest.
As mentioned above, the marginal effects
for the ‘Charged’ model are significantly
larger than those for the ‘Imprisoned’
model. Rather than comparing the
marginal effects for the same variable
between models, it is more appropriate
to compare the marginal effects for
each variable in the same model to the
probability of being charged or imprisoned
in the median case.
The most powerful predictors of being
charged or imprisoned are clearly
alcohol consumption and drug use. For
an average person (as defined above),
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
B U R E A U
O F
C R I M E
S T A T I S T I C S
Table 21: Marginal effects of selected variables for ‘Charged’ and ‘Imprisoned’ models
Marginal effect
Variable
Charged
Imprisoned
Substance abuse
12.89%
0.76%
High-risk alcohol consumption
11.40%
0.56%
Year 12 completion
-4.08%
-0.15%
Welfare as principal income source
4.26%
0.62%
Unemployed
4.84%
0.29%
CDEP
1.80%
not significant
Expected probability for the median case
8.90%
0.33%
being charged by over 11 percentage
points. The effects of alcohol and drug
use on the risk of imprisonment are much
smaller. However, the effect of substance
abuse is higher than any other effect in
the model. Alcohol is the third largest of
the five effects.
Most of the other factors in Table 21 exert
effects on the ‘Charged’ variable that
are roughly comparable in magnitude.
Completing Year 12 reduces a person’s
likelihood of being charged by 4.08
percentage points, while obtaining
welfare as the principal source of income
increases the risk of having been charged
by 4.26 percentage points. Being
unemployed increases the risk of having
been charged by 4.84 percentage points;
however, being in a CDEP scheme
only increases the likelihood by 1.8
percentage points. In other words, the
risk of being charged is less for those
participating in a CDEP scheme than for
those who are unemployed.
Year 12 completion and unemployment
exert similar effects on the risk of
imprisonment. Being on welfare, however,
has a bigger effect on the risk of being
imprisoned than high-risk alcohol
consumption. The CDEP variable is not a
significant predictor of imprisonment.
Discussion
It is always hazardous drawing causal
inferences from a study, like the NATSISS,
that seeks information from a group of
individuals at a single point in time. It is
even more hazardous when, as in the
present case, the dependent variables
measure aspects of the experience of
the respondent over the previous five
years or over their lifetime, whereas
the explanatory variables measure
characteristics of the respondent at the
time of the survey or in the preceding
12 months. In using the NATSISS to try
and identify the factors that influence
the risk of being charged or imprisoned,
we are assuming that the conditions we
examine were present and exerted an
influence at the time of, or before, the
respondent was arrested and charged or
imprisoned. This is a plausible assumption
but it should be noted that we have no
way of testing whether it is correct.
Setting this issue to one side, the present
study contains a number of findings
that may be of assistance in reducing
Indigenous over-representation in
the criminal justice system. The most
important finding concerns substance
use. The marginal effects of drug use are
stronger than those of any other factor,
with the exception of sex. Alcohol is the
third strongest factor for the ‘Charged’
model and fourth strongest in the
‘Imprisoned’ model. The suggestion that
drug and alcohol abuse is an important
cause of Indigenous contact with the
justice system is consistent with a large
body of other evidence linking drug
and alcohol abuse to increased risk of
involvement in property and violent crime
(Blumstein et al. 1986; Exum 2006).
Several studies have found evidence that
crime can be reduced through measures
that reduce the availability of alcohol
11
A N D
R E S E A R C H
(Gray et al. 2000; d’Abbs & Togni 2000)
and illicit drugs (Moffatt, Weatherburn
& Donnelly 2005). There is also strong
evidence that coerced treatment
programs (e.g. the NSW Drug Court)
reduce the rate at which drug dependent
offenders re-offend (Lind et al. 2002).
The present study strongly suggests,
therefore, that one of the key ways to
reduce Indigenous contact with the
criminal justice system is to reduce
Indigenous drug and alcohol abuse.
Although the other factors examined in
the study exerted smaller effects, those
effects are still quite significant. Failure
to complete Year 12 exerts only a small
direct effect on the risk of imprisonment
but it exerts a significant effect on the
likelihood of being charged. This is
consistent with a large body of evidence
linking school failure and poor school
performance to juvenile involvement
in crime (Baker 1998; Grunseit et al.
2005; Maguin & Loeber 1996.). As noted
earlier, there is some disagreement
amongst researchers about whether
the relationship between poor school
performance and offending is actually
causal or a reflection of some other
factor or factors (Maguin & Loeber 1996).
In the early 1990s, however, the Ford
Foundation established a program for
‘at risk’ youth which used a combination
of coaching and cash incentives to
promote school retention. Rates of arrest
for students participating in the program
were only three-tenths those of a control
group of students who did not participate
in the program (Greenwood et al. 1998).
This suggests that improving Indigenous
school performance and retention is
another potentially valuable point of
leverage on Indigenous contact with the
justice system.
The significant effect of unemployment on
the risk of being charged and imprisoned
mirrors that obtained by Hunter (2001) in
his analysis of the 1994 NATSIS survey
data. Longitudinal studies also generally
find a strong relationship between
unemployment and crime, particularly
where low socio-economic status
offenders are concerned (Farrington et al.
1986; Good, Pirog-Good & Sickles 1986;
Thornberry & Christensen 1984; Fagan
B U R E A U
O F
C R I M E
& Freeman 1999). Schochet, Burghardt
and Glazeman (2000) compared arrest
rates among Job Corps12 participants in
the United States with those of a control
group of young people deemed eligible for
Job Corps but who were (as a result of a
random ballot) not offered a place on the
program. Their study found substantial
and significant difference in arrest rates
favouring Job Corps participants. Labour
market programs or policies that reduce
the level of Indigenous unemployment
are therefore a third potentially fruitful line
of attack on Indigenous contact with the
justice system.
Given that unemployment increases the
risk of arrest and imprisonment, it is not
surprising to find that financial stress has
similar effects. Being employed but on
a very low income would be expected
to exert a strong effect on incomegenerating crime and there is strong
evidence from longitudinal studies that
it does (Grogger 1998). This may be
one reason why, relative to those who
are employed or not in the labour force,
respondents who reported being on
CDEP were more likely to have been
charged or imprisoned. The higher risk of
being charged and imprisoned for those
on CDEP might reflect other factors as
well. CDEP schemes are sometimes
used to assist illiterate and semi-literate
community members in dealing with the
justice system (see Kral forthcoming).
There may be a tendency, therefore, to
locate them in crime-prone communities.
Of course, for many Indigenous
Australians, the alternative to participation
in the CDEP scheme is unemployment.
In judging the contribution of the CDEP
scheme to Indigenous contact with
the justice system, it is therefore more
appropriate to compare the scheme
with unemployment. Viewed from this
perspective, the CDEP scheme appears
to provide a protective effect against the
risk of being charged.
In light of the ongoing political debate
about the causes and consequences
of welfare dependence, the finding
that being on welfare increases the
risk of being charged and imprisoned
is bound to be controversial. Some
S T A T I S T I C S
have argued that welfare dependence
encourages Indigenous involvement
in crime and have blamed Indigenous
welfare dependence on, for example,
the absence of private property rights
under native title legislation (Hughes
2005). It is possible, however, that
welfare dependence is simply acting as
a proxy for poverty and other forms of
social disadvantage (e.g. intellectual or
physical disability), which are already
known to be risk factors for involvement
in crime (Farrington 1997). All sides
agree that policies which reduce
Indigenous economic disadvantage
are likely to reduce Indigenous contact
with the criminal justice system. How
best to reduce Indigenous economic
disadvantage is not a question we
propose to discuss here.
A few comments are in order about the
variable ‘social support’, which was
found significant in the bivariate analysis
but not in the multivariate analysis. It
will be recalled that about one in three
respondents who felt they had social
support were charged, compared with
about one in two of those who felt that
they did not have social support. Similarly,
one in 14 of those who felt they had social
support had been imprisoned, compared
with one in seven of those who felt they
did not have social support. Although
this effect disappeared in the multivariate
analysis, past research has shown that
social support plays an important role in
buffering the effects of economic stress
on child maltreatment (Weatherburn
& Lind 2001), a common precursor
to juvenile involvement in crime. It is
possible that the protective effect of social
support on the risk of being charged or
imprisoned was simply obscured by its
close association with other factors such
as financial stress and involvement in
social activities. In light of this, we should
not dismiss the possibility that measures
which strengthen Indigenous social
support might reduce Indigenous contact
with the justice system.
Since an individual cannot be imprisoned
without first being charged with a criminal
offence, one would expect to see a fair
degree of overlap between the factors
12
A N D
R E S E A R C H
that predict being charged and those that
predict being imprisoned. A comparison
of Tables 19 and 20 shows some notable
differences. Living in a crime-prone area
increases the risk of being charged
(Table 19) but is not a significant
independent predictor of being
imprisoned (Table 20). Being socially
involved reduces the risk of being
charged but has no independent effect
on the risk of imprisonment. Household
crowding is a significant independent
predictor of imprisonment (Table 20) but
not of being charged (Table 19). Finally,
although drug and alcohol use, school
retention, welfare and unemployment
are significant predictors of both being
charged and being imprisoned, the
marginal effects of these factors are much
larger in the ‘Charged’ model than in the
‘Imprisoned’ model (Table 21).
These differences are probably a
reflection of both the sample size and
of differences in the factors that lead
to being charged and imprisoned. The
number of persons in the NATSISS who
reported having been imprisoned is far
smaller than the number who reported
having been charged. The power of our
analyses to detect significant effects
in relation to imprisonment would
therefore have been much smaller than
the corresponding power in relation
to being charged. The developmental
antecedents of violent offending, on the
other hand, differ to some extent from
those that lead to involvement in nonviolent crime (Crime Prevention Victoria
2003). Alcohol abuse, for example, is
more heavily implicated in violent crime
than in non-violent crime (Butler et al.
2003). Since violent offenders are
more likely to receive a prison sentence
than non-violent offenders (Snowball &
Weatherburn 2006), we would expect
to find some differences between the
NATSISS variables that predict being
charged and those that predict being
imprisoned.
In concluding the present study, we
note that it is one of only a handful so
far that have looked at the predictors
of Indigenous contact with the justice
system. Given that Indigenous
imprisonment rates are now higher
B U R E A U
O F
C R I M E
than they were at the time of the Royal
Commission into Aboriginal Deaths in
Custody, there is a pressing need for
further research in this area. It is no
easy task trying to identify the factors
underpinning Indigenous contact with the
justice system but, with all its limitations,
research of the kind reported here
provides a far better basis on which to
develop policy solutions than intuition,
guesswork and good intentions.
The ABS attempted to discount
imprisonment in protective custody,
for unpaid parking fines and other
infringements of good order, however
this could not be guaranteed.
8.
Because it is virtually impossible to be
imprisoned without first being charged,
for convenience in what follows we
refer simply to being ‘charged’ and
being ‘imprisoned’.
9.
The true fraction of Aboriginal and
Torres Strait Islanders who are ever
charged with an offence is almost
certainly higher than the figure we
present below because the data in
Table 1 are drawn from respondents
who are still alive and, in many cases,
quite young.
Notes
1.
2.
3.
The National Health Survey and
National Housing Survey have an
augmented Indigenous sample, but do
not focus specifically on Indigenous
issues and do not include information
on contact with the justice system.
The proposed Longitudinal Study
of Indigenous Children is still in the
pilot phase and will focus on issues
relating to child development.
Chikritzhs and Brady (2006) have
argued that the 2002 NATSISS
seriously underestimates the true
extent of Indigenous alcohol abuse.
That criticism does not affect the
present study because we only use
the NATSISS to rank respondents
in terms of alcohol consumption,
not to measure the absolute amount
of alcohol they consume.
10.
11.
The ABS defines private dwellings
as ‘houses, flats, home units and
any other structures used as private
places of residence at the time of
the [NATSISS]’ (Australian Bureau of
Statistics 2005b).
4.
See Australian Bureau of Statistics
(2004, p. 53) for details.
5.
The 80 per cent response rate for
non-community areas does not
include 12 per cent of households
who could not be contacted to
ascertain whether an Indigenous
person resided there.
6.
Appendix 2 provides a full list of
variables and their frequencies.
7.
The imprisonment variable included
all people who had spent any time
in prison in the five years previous.
S T A T I S T I C S
12.
13.
While this definition of family size would
not be considered large relative to
Indigenous norms, it is certainly large
compared to the Australian average
(Hunter, Kennedy & Smith 2003).
The Torres Strait Islander variable
was found to be significant for the
charged variable when treating the
‘both Aboriginal and Torres Strait
Islander’ group separately from the
‘only Torres Strait Islander’ group. For
the ‘Imprisoned’ model, it was found
to be significant when the two groups
were combined. In order to keep
consistency between the models, we
felt it more appropriate not to include
the Torres Strait Islander variable in
either model. In addition, the fact that
this variable was only collected for
Queensland made it difficult to draw
useful conclusions.
In 1999, Job Corps received $1.3
billion and enrolled 60,000 young
people in tailored one-year programs
that included classroom training in
basic education, vocational skills and
a wide range of supportive services
(including health care) at a cost of
roughly $15,000 per student.
Note that neither the frequency nor
percentage columns contain ‘not
stated’, ‘non-response’ and ‘not
applicable’ responses. For this
reason, the percentages may not add
to 100 per cent within each category.
13
A N D
R E S E A R C H
References
Agnew, R. & White, H.R. 1992, ‘An
empirical test of general strain theory’,
Criminology, vol. 30(4), pp. 475-499.
Altman, J.C., Gray, M.C. & Levitus, R.
2005, Policy Issues for the Community
Development Employment Projects
Scheme in Rural and Remote Australia,
CAEPR Discussion Paper No. 271,
CAEPR, ANU, Canberra.
Australian Bureau of Statistics 2004,
National Aboriginal and Torres Strait
Islander Social Survey 2002, Catalogue
No. 4714.0, Australian Bureau of
Statistics, Canberra.
Australian Bureau of Statistics 2005a,
Prisoners in Australia, Catalogue No.
4517.0, Australian Bureau of Statistics,
Canberra.
Australian Bureau of Statistics 2005b,
National Aboriginal and Torres Strait
Islander Social Survey: Expanded
Confidentialised Unit Record File,
Technical Paper, Catalogue No. 4720.0,
Australian Bureau of Statistics, Canberra.
Baker, J. 1998, Juveniles in Crime:
Participation Rates and Risk Factors,
NSW Bureau of Crime Statistics and
Research, Sydney.
Blumstein, A., Cohen, J., Roth, J.A. &
Visher, C.A. 1986, Criminal Careers
and Career Criminals, vol. 1, National
Academy Press, Washington DC.
Butler, T., Levy, M., Dolan, K. & Kaldor, J.
2003, ‘Drug use and its correlates in an
Australian prisoner population’, Addiction
Research and Theory, vol. 11(2),
pp. 89-101.
Chikritzhs, T. & Brady, M. 2006, ‘Fact
or Fiction? A critique of the National
Aboriginal and Torres Strait Islander
Social Survey 2002’, Drug and Alcohol
Review, vol. 25(3), pp. 277-287.
Crime Prevention Victoria 2003, Patterns
and Precursors of Adolescent Antisocial
Behaviour, Crime Prevention Victoria,
Melbourne.
d’Abbs, P. & Togni, S. 2000, ‘Liquor
licensing and community action in
regional and remote Australia: A review
of recent initiatives’, Australian and New
Zealand Journal of Public Health,
vol. 24(1), pp. 45-53.
B U R E A U
O F
C R I M E
Delahunty, B. & Putt, J. 2006, The Policing
Implications of Cannabis, Amphetamine
and Other Illicit Drug Use in Aboriginal
and Torres Strait Islander Communities,
National Drug Law Enforcement Research
Fund, Commonwealth of Australia,
Canberra.
Exum, M.L. 2006, ‘Alcohol and
aggression: An integration of findings
from experimental studies’, Journal of
Criminal Justice, vol. 34, pp. 131-145.
Fagan, J. & Freeman, R.B. 1999, ‘Crime
and Work,’ in M. Tonry (ed.), Crime and
Justice: A Review of Research, vol 25,
University of Chicago Press, Chicago,
pp. 225-290.
Farrington, D.P. 1997, ‘Human
Development and Criminal Careers’, in
M. Maguire, R. Morgan & R. Reiner (eds),
The Oxford Handbook of Criminology,
(second edition), Oxford University Press,
Oxford, pp. 361-408.
Farrington, D.P., Gallagher, B., Morley, L.,
St. Ledger, R.J. & West, D.J. 1986,
‘Unemployment, school leaving, and
crime’, British Journal of Criminology,
vol. 26(4), pp. 335-356.
Fergusson, D., Swain-Campbell, N. &
Horwood, J. 2004, ‘How does childhood
economic disadvantage lead to crime?’
Journal of Child Psychology and
Psychiatry, vol. 45, pp. 956-966.
Gendreau, P., Little, T. & Goggin, C.
1996, ‘A meta-analysis of the predictors
of adult offender recidivism: What works!’
Criminology, vol. 34(4), pp. 575-607.
Good, D.H., Pirog-Good, M.A. & Sickles,
R.C. 1986, ‘An analysis of youth crime
and employment patterns’, Journal of
Quantitative Criminology, vol. 2,
pp. 219-236.
S T A T I S T I C S
Grogger, G. 1998, ‘Market wages
and youth crime’, Journal of Labour
Economics, vol. 16, pp. 756-791.
Grunseit, A., Weatherburn, D. &
Donnelly, N. 2005, School Violence
and its Antecedents: Interviews with
High School Students, NSW Bureau of
Crime Statistics and Research, Sydney.
Hughes, H. 2005, The Economics of
Indigenous Deprivation and Proposals for
Reform, Issue Analysis No. 63, Centre for
Independent Studies, Sydney.
Hunter, B. 2001, Factors Underlying
Indigenous Arrest Rates, NSW Bureau of
Crime Statistics and Research, Sydney.
Hunter, B.H., Kennedy, S. & Smith, D.
2003, ‘Household composition,
equivalence scales and the reliability of
income distributions: Some evidence
for Indigenous and other Australians’,
Economic Record, vol. 79, pp. 70-83.
Kral, I. forthcoming, Change and
Transmission: Literacy and Learning in
a Remote Western Desert Region, Ph.D
thesis (Anthropology), The Australian
National University, Canberra.
Lind, B., Weatherburn, D., Chen, S.,
Shanahan, M., Lancsar, E., Haas, M. &
De Abreu Lorenco, R. 2002, NSW Drug
Court Evaluation: Cost-Effectiveness,
NSW Bureau of Crime Statistics and
Research, Sydney.
Loeber, R. & Stouthamer-Loeber, M.
1986, ‘Family factors as correlates and
predictors of juvenile conduct problems
and delinquency’, in M. Tonry & N. Morris,
(eds), Crime and Justice: An Annual
Review of Research, vol. 7, The University
of Chicago Press, Chicago, pp. 29-149.
Gray, D., Saggers, S., Sputore, B. &
Bourbon, D. 2000, ‘What Works? A review
of evaluated alcohol misuse interventions
among Aboriginal Australians’, Addiction,
vol. 95(1), pp. 11-22.
MacKenzie, D.L. 2002, ‘Reducing the
criminal activities of known offenders and
delinquents: Crime prevention in the courts
and corrections’, in L.W. Sherman, D.P.
Farrington, B.C. Welsh & D.L. MacKenzie
(eds), Evidence-Based Crime Prevention,
Routledge, London, pp. 330-404.
Greenwood, P.W., Model, K.E., Rydell, C.P.
& Chiesa, J. 1998, Diverting Children
from a Life of Crime: Measuring Costs
and Benefits, RAND, MR-699-1-UCB/RC/
IF, Santa Monica, California.
Maguin, M. & Loeber, R. 1996, ‘Academic
performance and delinquency,’ in M. Tonry,
(ed.), Crime and Justice: An Annual Review
of Research, vol. 20, The University of
Chicago Press, Chicago, pp. 145-264.
14
A N D
R E S E A R C H
Moffatt, S., Weatherburn, D. & Donnelly, N.
2005, What Caused the Recent Drop
in Property Crime?, Crime and Justice
Bulletin No. 85, NSW Bureau of Crime
Statistics and Research, Sydney.
National Crime Prevention 1999, Pathways
to Prevention: Developmental and Early
Intervention Approaches to Crime in
Australia, National Crime Prevention,
Attorney General’s Department, Canberra.
NSW Bureau of Crime Statistics and
Research 2006, NSW Criminal Courts
Statistics 2005, NSW Bureau of Crime
Statistics and Research, Sydney.
Office of Evaluation and Audit 1997,
Evaluation of the Community Development
Employment Projects Program: Final
Report September 1997, ATSIC, Canberra
Pratt, T.C. & Cullen, F.T. 2005, ‘Assessing
macro-level predictors and theories of
crime: A meta-analysis’, in M. Tonry (ed.),
Crime and Justice: A Review of Research,
vol. 32, Chicago, The University of
Chicago Press, pp. 373-441.
Sanders, W. 2005, Housing Tenure and
Indigenous Australians in Remote and
Settled Areas, CAEPR Discussion Paper
No. 275, CAEPR, ANU, Canberra.
Schochet, P.Z., Burghardt, J. &
Glazerman, S. 2000, National Job Corps
Study: The Short-term Impacts of Job
Corps on Participants’ Employment and
Related Outcomes, Mathematica Policy
Research, Princeton, NJ.
Snowball, L. & Weatherburn, D. 2006,
Indigenous Imprisonment: The Role of
Offender Characteristics, NSW Bureau of
Crime Statistics and Research, Sydney.
Thornberry, T. & Christenson, R. 1984,
‘Unemployment and criminal involvement:
An investigation of reciprocal causal
structures’, American Sociological
Review, vol. 56, pp. 609-627.
Weatherburn, D., Fitzgerald, J. & Hua J.
2003, ‘Reducing Aboriginal overrepresentation in prison’, Australian
Journal of Public Administration,
vol. 62(3), pp. 65-73.
Weatherburn, D. & Lind, B. 2001,
Delinquent Prone Communities,
Cambridge University Press, Cambridge.
B U R E A U
O F
C R I M E
Appendix 1
Because the substance abuse variable is
only determined for respondents living in
non-remote areas (for reasons outlined
above) it is necessary to carry out a form
of sensitivity analysis to determine the
effect this has on the other coefficients in
each model. We calculated coefficients
for the same variables in the models
outlined above for (a) all respondents;
and (b) only respondents in non-remote
areas. The results are outlined in
Table 22 for the ‘Charged’ model and
Table 23 for the ‘Imprisoned’ model.
Because of the way the geographical
variables were specified in the models,
we needed to respecify the model to have
‘living in a major city’ in the base case
(this accounts for the differences in the
intercept and geographic coefficients).
Apart from this, the base cases for the
two models are exactly as specified in the
results section above.
There is a large change in coefficients for
a number of variables, however most can
be explained by the differing experiences
of remote and non-remote Indigenous
people which are not completely
accounted for by the geographic
variables. For example, the effect of living
in a crowded house in remote and nonremote areas is likely to be fundamentally
different, given the different set of choices
and constraints facing Indigenous
householders in the respective areas
(Sanders 2005).
The drop in the effect of the alcohol
variable in the ‘Charged’ model is a cause
for concern, due to its high correlation
with the substance abuse variable. It
is impossible to determine the actual
coefficient for substance abuse across
remote and non-remote regions but these
results need to be taken into account
when examining the model.
S T A T I S T I C S
A N D
R E S E A R C H
Table 22: Parameter estimates for the ‘Charged’ models
Comparison
Total
model (a)
Non-remote
model (b)
Intercept
-3.10 (0.13)
-2.91 (0.18)
Under 25 years vs 25 years and over
-0.20 (0.07)
-0.19 (0.09)
Male vs Female
1.54 (0.06)
1.63 (0.08)
Unemployed vs Employed or NILF
0.49 (0.09)
0.41 (0.11)
CDEP vs Employed or NILF
0.21 (0.08)
-0.01 (0.15)*
Welfare vs Other income source
0.44 (0.06)
0.77 (0.09)
Financial stress vs No financial stress
0.48 (0.05)
0.44 (0.07)
Completed Year 12 vs Did not complete Year 12
-0.66 (0.08)
-0.74 (0.10)
Person or family member of ‘stolen generation’
vs Person or family not a member of the
‘stolen generation’
0.37 (0.05)
0.29 (0.07)
Sole-parent family vs Other family type
0.20 (0.07)
0.18 (0.08)
No social involvement vs Social involvement
0.30 (0.08)
0.07 (0.11)*
Regional vs Major city
0.21 (0.08)
0.18 (0.08)
Remote vs Major city
0.47 (0.09)
-
Lives in a crime-prone area vs Does not live in
a crime-prone area
0.27 (0.06)
0.18 (0.08)
High-risk alcohol use vs Not high-risk alcohol use
0.96 (0.10)
0.72 (0.14)
Substance use vs No substance use
1.05 (0.00)
1.09 (0.07)
Substance use missing vs No substance use
0.44 (0.13)
0.43 (0.13)
* Variable not significant at the five per cent level
Table 23: Parameter estimates for the ‘Imprisoned’ models
Comparison
Total
model (a)
Intercept
-5.73 (0.23)
-5.86 (0.22)
Male vs Female
1.49 (0.10)
1.49 (0.10)
Under 25 years vs 25 years and over
0.17 (0.11)*
0.19 (0.11)*
Unemployed vs Employed or NILF
0.63 (0.12)
0.60 (0.12)
CDEP vs Employed or NILF
0.15 (0.12)*
0.20 (0.12)*
Welfare vs Other income source
1.07 (0.13)
1.05 (0.13)
Financial stress vs No financial stress
0.37 (0.09)
0.38 (0.09)
Completed Year 12 vs Did not complete Year 12
-0.59 (0.16)
-0.57 (0.16)
Crowded household vs Non-crowded household
0.29 (0.12)
0.35 (0.12)
Person or family member of ‘stolen generation’
vs Not a member of the ‘stolen generation’
0.48 (0.09)
0.48 (0.09)
Regional vs Major city
0.01 (0.15)*
0.49 (0.10)
Remote vs Major city
0.94 (0.19)
-
High-risk alcohol risk vs Not high-risk alcohol use
1.00 (0.12)
1.02 (0.12)
Substance use vs No substance use
1.21 (0.15)
0.98 (0.13)
Substance use missing vs No substance use
0.57 (0.26)
0.32 (0.25)*
* Variable not significant at the five per cent level
15
Non-remote
model (b)
B U R E A U
O F
C R I M E
S T A T I S T I C S
A N D
R E S E A R C H
Appendix 2
Table 24: Frequency distribution of regressor variables13 Frequency
in sample
Weighted
per cent
Age
Under 25 years
25 years or over
6,927
1,594
20.7
79.3
Sex
Female
Male
4,919
3,602
52.6
47.4
Labour force status
Employed or NILF
Unemployed
Employed – CDEP
6,317
74.8
833
1,371
12.4
12.8
Principal income source
Welfare
Other than welfare
5,748
2,688
62.6
36.8
Financial stress
Experienced financial stress
Has not experienced financial stress
3,756
4,765
43.2
56.8
Education
Completed Year 12 or equivalent
Did not complete Year 12 or equivalent
1,285
7,236
18.5
81.5
Crowded household
Lives in crowded household
Does not live in crowded household
1,284
7,237
14.6
85.4
‘Stolen generation’
Relative or individual removed from family
No family removal
3,221
3,809
38.2
43.2
Family type
Sole-parent family
Other family type
1,744
6,777
20.2
79.8
Social isolation
Involvement in social activities
Not involved in social activities
7,482
1,039
89.5
10.5
Location
Major city
Regional area
Remote area
1,311
3,416
3,794
30.2
42.2
27.6
Crime-prone area
Lives in a crime-prone area
Does not live in a crime-prone area
6,485
2,036
74.2
25.8
Alcohol consumption
High-risk consumption
Non high-risk consumption
577
7,886
6.1
93.1
Substance abuse
Used substances
Never used substances
1,896
2,472
29.7
37.1
NSW Bureau of Crime Statistics and Research - Level 8, St James Centre, 111 Elizabeth Street, Sydney 2000
bcsr@agd.nsw.gov.au • www.lawlink.nsw.gov.au/bocsar • Ph: (02) 9231 9190 • Fax: (02) 9231 9187
ISSN 1030 - 1046 • ISBN 0 7313 2687 3
© State of New South Wales through the Attorney General’s Department of NSW 2006. You may copy, distribute, display, download and otherwise freely deal with this
work for any purpose, provided that you attribute the Attorney General’s Department of NSW as the owner. However, you must obtain permission if you wish to (a)
charge others for access to the work (other than at cost), (b) include the work in advertising or a product for sale, or (c) modify the work.