The influence of Indigenous status on the issue of police cautions

Abstract

Over the last 20 years, a body of research has emerged in the United States (US) revealing that the country’s sentencing courts to treat young male African American (and Hispanic) offenders more harshly than white offenders, even after adjusting for relevant legal and contextual factors. Similar research in Australia has generally found that the effect of Indigenous status on adult bail/sentencing outcomes is either non-significant, or significant, but very small. The aim of this study is to explore the impact of race, age, and gender on police decisions to prosecute rather than caution to a juvenile offender. We employ a multilevel model with random intercepts to explore the impact of race, age, and gender on police decisions to prosecute rather than caution to a juvenile offender. The first level controls for offender/offence factors a police officer may legally consider when deciding whether or not to caution a young offender. The second level controls for the police patrol to which the police officer is attached. After adjusting for the effects of legally relevant factors, we find Indigenous juvenile offenders (regardless of sex) are more likely to be prosecuted than cautioned, compared with their non-Indigenous counterparts. There is also wide variation across local area commands in willingness to caution juvenile offenders. We conclude that further research is needed to obtain a better understanding of the factors responsible for racial disparity in the use of police cautions.

Keywords

Multilevel model racial discrimination focal concerns Indigenous police caution juvenile offender

Introduction

Over the last 20 years, a body of research has emerged in the United States (US) revealing that the country’s sentencing courts treat young male African American (and Hispanic) offenders more harshly than white offenders, even after relevant legal and contextual factors have been accounted for (Bushway & Piehl, 2001; Mitchell, 2005). These findings are usually explained by reference to what has become known as focal concerns theory (Miller, 1958; Steffensmeier et al., 2006). According to this theory, sentencing judges are guided by three focal concerns: the blameworthiness and culpability of the offender; concerns about practical constraints and consequences; and a desire to protect the community. Because judges often lack the time and information required to adequately assess defendants in terms of these factors, focal concerns theory postulates that race, ethnicity, and sex come to function as a kind of perceptual shorthand for public safety. Young black (or Hispanic) males are treated more harshly, on this account, because they are perceived to be more dangerous.

A number of scholars in Australia have also expressed concern about racial bias in the operation of the criminal justice system (e.g., Allison et al., 2013; Anthony, 2015; Blagg, 2008; Blagg et al., 2005). After adjusting for factors that courts are required to consider, however, most studies find that the parameter measuring the effect of Indigenous status on bail/sentencing outcomes for adult offenders is either non-significant, or significant but very small (Bond & Jeffries, 2012a , 2012b; Broadhurst, 1997; Gallagher et al., 1997; Jeffries & Bond, 2009 , 2012; Luke & Cunneen, 1995; Snowball & Weatherburn, 2006 , 2007). The only exception to this pattern concerns Indigenous adult offenders charged with domestic violence offences. In that context, Indigenous defendants have been found to be significantly and substantially more likely to be imprisoned than non-Indigenous adult offenders, even after controlling for relevant legal factors (Bond & Jeffries, 2010; Donnelly & Poynton, 2015; Thorburn & Weatherburn, 2018).

The limited evidence of racial bias in the sentencing of convicted offenders to imprisonment is in some ways not surprising. There is little inter-racial violence in Australia compared with the United States. Judges and magistrates in Australia are not elected officials and are therefore rather more removed from public pressure than judges in some US States. They are also heavily constrained by statute law and the appellate review process. The limited evidence of racial bias in sentencing in Australia is also unsurprising for another reason. Focal concerns theory would not necessarily lead one to expect a general effect of Indigenous status on criminal justice outcomes. The theory posits an interaction between age, race, and gender; suggesting (in an Australian context) that differences between Indigenous and non-Indigenous defendants/offenders are likely to be limited to or concentrated among young Indigenous males. This fact may be obscured in past Australian research on racial bias which, up to this point, has mostly focused on adult offenders at the point of sentencing.

As elsewhere, police in Australia enjoy considerable discretion in how they choose to proceed against juvenile offenders. Not surprisingly, research has consistently shown that race effects are most likely to be found in the early stages of the formal juvenile process (Ericson & Eckberg, 2016). In New South Wales, for example, police can choose to warn or caution a young offender, refer the offender to a youth justice conference or charge the young offender and refer him or her to court. There is no appeal against a decision by a police officer to charge a child with a criminal offence and have the matter dealt with in court instead of issuing a formal caution. Decisions made in the early stages of the criminal justice process may have a compounding effect on outcomes at later stages. Gale et al. (1990), for example, have argued that because Aboriginal juveniles are more likely to be charged, and brought to court, they tend to acquire longer criminal records. In subsequent court appearances, this puts them at greater risk of bail refusal and imprisonment. Ericson and Eckberg (2016) found evidence of this effect in the US.

The aim of the current study is to examine the influence of Indigenous status, age, and gender on the police decision to issue a police caution or prosecute a young person suspected of having committed a criminal offence. We employ multilevel modelling to (a) to test whether there is any evidence that Indigenous juveniles are more likely than non-Indigenous juveniles to be prosecuted, than cautioned. Then, after controlling for factors that police are allowed to consider, we (b) assess whether the racial difference (if there is one) is moderated by age and gender and (c) quantify the scale of variation between police commands in willingness to issue a caution. The next section briefly reviews relevant past research in Australia before explaining the current study in more detail.

Relevant past research

A number of Australian studies have found that Indigenous young people are less likely to be cautioned by police and far more likely to be prosecuted than non-Indigenous young people (Allard et al., 2010; Gale et al., 1990; Loh & Ferrante, 2003; Ringland & Smith, 2013; Snowball, 2008; Wundersitz, 1997). In her three state comparison, for example, Snowball found that West Australian young Indigenous defendants were five times more likely to be prosecuted than diverted. Over the same period, in South Australia, the relative rates of prosecution rates for Indigenous versus non-Indigenous defendants ranged between 1.3 and 1.6. In New South Wales, they ranged between 1.9 and 2.5. Snowball noted that, while the likelihood of referral to court (as opposed to receiving a caution) rose with age for both Indigenous and non-Indigenous defendants, it rose much faster for Indigenous young people than for non-Indigenous young people.

Of course, the existence of a racial difference in cautioning rates is not by itself evidence of racial bias in the exercise of police discretion. The use of diversionary alternatives in all jurisdictions is subject to a number of legislative constraints. Those constraints in New South Wales include requirements on the part of police to consider the seriousness of the offence, the degree of violence involved in the offence, the harm caused to any victim, the number and nature of any offences committed by the child, and the number of times the child has been dealt with under the NSW Young Offender’s Act (YOA). If adherence to these constraints results in unwarranted differences between rates of Indigenous and non-Indigenous defendants in the use of police cautions, the fault lies with law rather than with the exercise of police discretion. The critical issue, then, is whether and to what extent the racial difference in cautioning rates persists after adjusting the factors police are obliged to consider.

Only three significant attempts have been made to address this issue. The first, by Luke and Cunneen (1995), involved a study of racial bias in the granting of police bail in NSW. They found no evidence of bias in the granting of bail once controls were introduced for prior imprisonment, prior criminal record, offence, age, and gender. However, they did find a significant difference (favouring non-Indigenous defendants) when examining the decision to charge rather than caution. They also found a significant difference when examining the decision to proceed by way of arrest, rather than summons (Luke & Cunneen, 1995, p. 77). The second, by Snowball (2008), examined the influence of Indigenous status on diversion in three Australian states (Western Australia, South Australia, and New South Wales). After adjusting for the effects of age, gender, offence type, number of prior court appearances, and whether or not the offender had previously received a custodial penalty, she found that Indigenous offenders were significantly less likely to be diverted from court than non-Indigenous offenders.

The third, by Ringland and Smith (2013), is the most sophisticated analysis so far conducted of Australian police use of their juvenile diversionary powers. Their study included controls based on the YOA legislative criteria governing the exercise of police discretion in relation to police cautions. Those controls included age, sex, number of current charges, number of prior cautions, whether or not the offender had previously been before a youth justice conference, whether the offender had previously committed a serious violent offence and whether the defendant had previously been placed on a control order (i.e., given a custodial penalty). They also excluded all offences not eligible to be dealt with under the YOA. Once again, even after controlling for these factors, Ringland and Smith (2013) found substantially lower odds of diversion amongst Indigenous offence.

Despite its sophistication, the Ringland and Smith (2013) study is now almost 10 years old. It is possible the difference in police willingness to caution Aboriginal as against non-Aboriginal defendants has diminished since then. For our purposes, the Ringland and Smith (2013) study has some limitations. Ringland and Smith (2013) formed a single measure of diversion that combined cautions and youth justice conferences. This made it impossible to determine which form of diversion police were reluctant to use. Moreover, instead of including a general measure of offence seriousness, they included a dummy variable indicating whether the principal offence involved serious violence or not. An offence may be serious without necessarily involving violence. Finally, because Ringland and Smith (2013) were primarily interested in the variation across police local area commands (LACs) in willingness to divert offenders (rather than in the willingness of police to divert Indigenous juvenile offenders), they did not explore the interaction between willingness to divert and other offender characteristics such as Indigenous status, age, and gender.

The current study

The aim of the present study is to replicate the Ringland and Smith (2013) study while overcoming the limitations noted above. The discretion to issue a YOA caution lies with the police officer who apprehends or detains a young offender. Section 8 of the YOA lays down a number of restrictions on the offences for which a juvenile may be cautioned. Those restrictions prohibit the issue of police cautions to juveniles who have committed a strictly indictable offence, an offence involving the death of someone, a domestic violence offence, a stalking offence, a graffiti offence (after 2014) and certain traffic, sex, and drug offences. It also prohibits the issue of a police caution if a child has been dealt with by caution on three or more occasions. The factors police must consider when deciding whether to caution a young person include the seriousness of the offence, the degree of violence involved in the offence, the harm caused to any victim, the number and nature of any offences committed by the child, the number of times the child has been dealt with under the YOA and, under s.20(2e), any other matter they think is appropriate in the circumstances.

The current study improves on Ringland and Smith (2013) in several ways. We employ a much larger sample of young offenders over a much longer time period, while controlling for possible changes over time in police willingness to issue a YOA caution. Rather than focussing on the association between Indigenous status and diversion in general, we focus on the association between Indigenous status and the issue of police cautions. This allows us to draw conclusions about the operation of one of the most important alternatives to prosecution under the YOA. In addition to including a control for whether or not the principal offence involved serious violence, we introduce a general control for offence seriousness. Finally, ours is the first Australian study to examine the interaction between Indigenous status, age, and gender in the disposition of juvenile offenders and only the second to conduct an analysis that properly controls for the factors police must consider when deciding whether to caution a juvenile offender.

Method

Approval for the current study was granted by the UNSW Human Research Ethics Committee (HC: 220226) on 20 May 2022.

Data sources

The data for the study were drawn from the NSW Re-Offending Database (ROD) maintained by the NSW Bureau of Crime Statistics and Research (for more details, see Hua & Fitzgerald, 2006). ROD contains linked data on all criminal court appearances in NSW since 1994, and all warnings, cautions and conferences under the YOA since 2010. For the purpose of this study, the data extracted from ROD included records of all 248,047 individuals who had been warned, cautioned, referred to a youth justice conference or referred to court between 1 January 2010 and 31 December 2021.

Sample

To align the dataset with the eligibility requirements for a police caution, the following restrictions were imposed. First, all those aged over 17 or who received an outcome other than a Children’s Court appearance or a YOA caution were removed from the dataset. Second, to avoid correlated observations, where a young person had multiple court appearances over this time period, one appearance was selected at random, and the records of all other contacts removed. Third, because young people must be prepared to admit the offence in order to receive a police caution, individuals referred to court were removed from the dataset if they pleaded not guilty to the principal offence. Fourth, records involving young people appearing in court for an offence that would have rendered them ineligible to receive a police caution were removed. This was accomplished by removing anyone charged with a strictly indictable offence or whose principal offence fell into one of the following Australian and New Zealand Standard Offence or NSW Judicial Commission law codes (Australian Bureau of Statistics, 2011; Judicial Commission of New South Wales, 2022):

An offence involving the death of someone (ANZSOC > = 111 & < = 132)

A sexual assault or related offence (ANZSOC > = 311 & < = 329)

A domestic assault offence (law codes: 64772, 64773, 64774, 64779, 64780, 64781, 48386, 64782).

A stalking offence (ANZSOC code 291)

Breach of a violence order (ANZSOC code 1531)

A traffic offence where the offender was old enough to obtain a learner’s permit (ANZSOC codes between 1411 and 1441 where the offender was aged 16)

A drug offence (any ANZSOC code > 1011 and < 1100)

A graffiti offence that occurred on or after 2014 (ANZSOC 1212).

Finally, due to the small number of cases where sex was unknown and the fact that all these cases were in the group that had been cautioned, the multivariate analysis proceeded without cases where the sex of the offender was unknown. After these adjustments, the study sample consisted of 38,439 records: 33,376 (86.83%) of which involved a police caution and 5063 (13.17%) involved a referral to court.

Measures

Because our focus is on whether Indigenous young people are more likely to be prosecuted, the dependent variable in the study is labelled prosecuted and is coded ‘1’ if the defendant was prosecuted and ‘0’ if cautioned. The controls (independent variables) comprised:

Age (coded “0” if 13 and under; “1” if 14, “2” if 15, “3” if 16 and “4” if 17).

Sex (coded “0” if female; “1” if male).

Area of residence (coded “1” if lives in a major city; “2” if lives in an inner regional area; “3” if lives in an outer regional area; and “4” if lives in a remote or very remote area).

Concurrent offences (coded “0” if none or one; “1” if more than one).

Offence seriousness (see below).

Serious violent offence (coded “1” if the principal offence involved an offence in ANZSOC categories 211, 311, 511 or 611 and “0” otherwise)

Prior warnings (coded “0” if none and “1” if one or more).

Prior cautions (coded “0” if none and “1” if more than one).

Prior conferences (coded “0” if none and “1” if more than one).

Prior court appearances (coded “0” if none and “1” if more than one).

Indigenous status (coded “0” if non-Indigenous and “1” if Indigenous).

Index year (the year in which the young offender was cautioned or prosecuted

We include age, sex, and residential location as controls because they are relevant to any assessment of the future risk of reoffending. We include offence seriousness, the number of concurrent offences and the number of previous warnings, cautions, conferences, and court appearances as controls, because they are relevant considerations in relation to the issue of cautions under section 20(3) of the YOA. Because sections 20(3b) and 20(3c) of that act require police to consider the degree of violence involved in the offence and the harm caused to any victim, we include a control for whether the juvenile has committed a serious violent offence. Finally, to control for possible variations in policing practice over the time period covered by the study, we include index year as a control.

Our measure of offence seriousness is based on the median seriousness ranking (MSR) developed by MacKinnell et al. (2010). Although the MSR ranking is continuous, the distribution of MSR scores is multimodal. Preliminary analysis revealed little variation in the likelihood of prosecution for the first six deciles of the MSR distribution, but a large increase for the next four. Two changes were therefore made to create the current measure of offence seriousness (labelled “offence severity”). First, the MSR was transformed by subtracting each seriousness score from the highest score possible (136). This created a scale in which more serious offences had higher scores. The resulting distribution was then collapsed into two categories. The first category (coded “0” if score is in the lowest six deciles of seriousness) comprised scores in the first six deciles of the distribution. The second category (coded “1”) comprised scores in the top four deciles of the distribution.

Analysis

Police officers in NSW for part of this study were organised into LACs, but in more recent years have been organised into a smaller number of police area commands (PACs). In what follows we refer to these police areas as PAC/LACs. Although the decision to caution a young person under the YOA rests with the arresting officer, police employed within the same PAC/LAC may share a similar disposition to prosecute rather than caution a young person. If this were the case, observations within each PAC/LAC would be correlated. We therefore employ a two-level logistic regression model with random intercepts as the vehicle for exploring the impact of Indigenous status on willingness to issue a YOA caution. The first level involves factors that the attending officer can legally consider when deciding whether to caution a young offender. The second level involves the officer‘s PAC/LAC. This second level allows observations within each PAC/LAC to be correlated. More importantly, it allows each PAC/LAC to have its own baseline (intercept) level of willingness to prosecute.

To test the appropriateness of a multilevel model we began by regressing the decision to prosecute against the variable measuring PAC/LAC. This allowed us to determine whether the variation between LACs/PACs in the willingness to prosecute is simply a result of variation in the types of offenders/cases coming to the attention of police (i.e., whether a multi-level model is warranted). If a multilevel model is warranted, there should be a significantly lower log likelihood for a model that includes a second (PAC/LAC) level than for a simple logistic regression model where PAC/LAC is the only covariate. The intraclass correlation coefficient (ICC), which measures the correlation among observations within the same PAC/LAC, should also be zero, or very small (e.g., less than 0.05).

If a multilevel model is warranted, we need to control for factors that police are required to consider when deciding whether to prosecute or caution a juvenile offender. The first 10 independent variables listed above measure factors police may consider. To identify relevant controls, we examined the bivariate relationship between the dependent variable (prosecution) and each of these variables. Significant factors were then included in the multilevel model at level 1. In our description of the sample below, we provide information on the offence profile of those cautioned or prosecuted by police. Rather than include offence type in the multilevel model, however, we rely on covariates 5 (offence seriousness) and 6 (serious violent offence) listed above to capture offence seriousness. There are two reasons for this. The first is that the multiplicity of offence categories would cause us to lose too many degrees of freedom. The second is that preliminary testing revealed no loss in explanatory power using offence seriousness and serious violent offence as controls instead of offence type.

Two other analyses were conducted: the first to explore the variation between PACs/LACs in willingness to prosecute a young offender; the second to explore the interaction between Indigenous status, age, and gender. To explore variation between PACs/LACs in willingness to prosecute a young offender we constructed what is known as a ‘caterpillar graph’. This is essentially a plot of the extent to which the likelihood of prosecution in each PAC/LAC deviates from the average likelihood of criminal prosecution for all PACs/LACs. Interaction effects were explored by deriving the predicted probability of prosecution for each observation and plotting the mean of these estimates as a function of race, age, and gender.

Results

Sample description

Table 1 shows the offence profile of those included in the study. The offences have been ranked from most to least prevalent.

Table 1.

Sample offence profile.

Principal offence	Freq.	Percent
Theft from retail premises	7360	19.2
Property damage	5410	14.1
Common assault	4501	11.7
Other theft	4056	10.6
Trespass	2191	5.7
Serious assault (causing injury)	1958	5.1
Prohibited weapons offence	1816	4.7
Break and enter	1805	4.7
Motor vehicle theft/illegal use	1280	3.3
Receive/handle crime proceeds	1155	3.0
Riot and affray	1093	2.8
Miscellaneous offences	1077	2.8
Offensive language/behaviour	690	1.8
Serious assault (no injury)	614	1.6
Traffic/vehicle regulatory offence	581	1.5
Obtain benefit by deception	578	1.5
Threatening behaviour	524	1.4
Resist/hinder police	523	1.4
Criminal intent	324	0.8
Theft from the person	208	0.5
Liquor and tobacco offence	195	0.5
Dishonest conversion	146	0.4
Public health and safety offence	143	0.4
Other justice	118	0.3
Transport regulatory offence	93	0.2

Table 2 provides additional descriptive statistics for other variables included in the study. All the variable values are well represented in terms of frequency. The proportion of the sample at each age group increases with age. The majority are male and live in major cities. Most have no concurrent offence, or only one. Most have no prior warning, no prior caution, no prior conference, and no prior court appearance. Fourteen percent of the sample were recorded as Indigenous at the time of their contact with police. In a large proportion (13.54%) of cases, however, the Indigenous status of the young offender is unknown. In what follows therefore, we include ‘unknown’ as a special category of Indigenous status. This obviates the bias that would result if observations involving an unknown Indigenous status were discarded. However, we report sensitivity tests in Appendices 1 and 2 to see whether the high percentage of missing values affects our results. The number of cases dealt with each year is not shown (for reasons of space) but ranged between 2398 (in 2016) and 5762 (in 2010); with an average of 3203. The majority (86.83%) of those eligible for a police caution received a caution. The remainder (13.17%) were proceeded against to court.

Table 2.

Sample description.

Variable	Freq.	Percent
Outcome
Caution	33,376	86.83
Court	5063	13.17
Age group
13 and under	5136	13.36
14	6272	16.32
15	8697	22.63
16	9043	23.53
17	9291	24.17
Sex
Female	13,377	35.01
Male	24,831	64.99
Area
Major cities	22,391	60.07
Inner regional	10,798	28.97
Outer regional	3517	9.44
Remote/very remote	568	1.52
Concurrent offences
None or one	30,069	78.23
More than one	8370	21.77
Offence severity
Not serious	26,929	70.06
Serious	11,510	29.94
Serious violence
No	36,469	94.87
Yes	1970	5.13
Prior warnings
None	34,077	88.65
One	3059	7.96
More than one	1303	3.39
Prior cautions
None	30,642	79.72
One	5305	13.8
More than one	2492	6.48
Prior conferences
None	37,216	96.82
One	958	2.49
More than one	265	0.69
Prior court apps
None	35,612	92.65
One or more	2827	7.35
Indigenous status
Non-Indigenous	27,965	72.75
Indigenous	5203	13.54
Unknown	5271	13.71

Bivariate analysis

Table 3 shows the percentage of non-Indigenous and Indigenous defendants who were prosecuted, broken down by offence type. Prior to adjusting for other differences between Indigenous and non-Indigenous offenders, the proportion of Indigenous offenders prosecuted is higher than the proportion of non-Indigenous juveniles prosecuted in every offence category.

Table 3.

Percent prosecuted by Indigenous status and offence type.

	Percent prosecuted
Principal offence	Non-Indigenous	Indigenous
Dangerous/negligent driving	58.33	88.24
Criminal intent	51.15	77.22
Other justice procedure offences	16.49	58.33
Serious assault (injury)	33.87	55.52
Riot and affray	32.29	52.12
Motor vehicle theft/Illegal use	25.13	43.43
Serious assault (no injury)	33.82	43.36
Resist/hinder police	19.78	39.42
Miscellaneous offences	14.66	36.63
Threatening behaviour	12.43	35.42
Break and enter	11.87	35.03
Offensive language/behaviour	14.82	32.7
Receive/handle crime proceeds	17.51	31.44
Theft from the person	18.56	23.33
Trespass	7.57	23.25
Property damage/graffiti	9.53	21.25
Fraud	10.34	20.97
Prohibited weapons offence	6.23	18.8
Common assault	9.18	17.76
Drive without a licence	7.28	17.14
Other theft	4.95	12.52
Public health and safety offences	2.33	11.76
Theft from retail pre	2.26	10.41

Table 4 shows the bivariate relationship between each of the independent variables in the study and the percentage who were prosecuted. All differences are statistically significant. Prior to controlling for any other factor, young offenders are more likely to be prosecuted if they are older, male, live in a major city or a remote/very remote area, commit a serious offence, engage in serious violence or have been previously warned, cautioned, referred to a youth justice conference or to court. Prior to controlling for any other factor, Indigenous juveniles are more than twice as likely to be prosecuted as non-Indigenous juveniles.

Table 4.

Bi-variate relationships with outcome.

Variable	Cautioned	Prosecuted	p-Value
Age group			<0.001
13 and under	96.22	3.78
14	92.03	7.97
15	88.44	11.56
16	84.13	15.87
17	79.24	20.76
Sex			<0.001
Female	90.04	9.96
Male	84.98	15.02
Area			<0.001
Major cities	88.91	11.09
Inner regional	86.61	13.39
Outer regional	82.88	17.12
Remote/very remote	77.82	22.18
Concurrent offences			<0.001
None or one	91.96	8.04
More than one	68.41	31.59
Prior conferences			<0.001
None	88.59	11.41
One	39.56	60.44
More than one	10.57	89.43
Offence severity			<0.001
Not serious	94.04	5.96
Serious	69.97	30.03	<0.001
Serious violence
No	88.03	11.97
Yes	64.57	35.43
Prior warnings			<0.001
None	88.4	11.6
One	78.69	21.31
More than one	64.93	35.07
Prior cautions			<0.001
None	91.84	8.16
One	78.72	21.28
More than one	42.5	57.5
Prior court apps
None	90.83	9.17	<0.001
One or more	36.43	63.57
Indigenous status
Non-Indigenous	88.94	11.06	<0.001
Indigenous	71.5	28.5
Unknown	90.74	9.26

Multivariate analysis

Table 5 shows the results of a multilevel model that includes no explanatory variables. There are 58 PAC/LACs, with an average of 621.4 observations per PAC/LAC (min = 38; max = 1357). The row labelled _cons is the grand mean of the PAC/LAC intercepts on prosecution. The row labelled PAC/LAC var (_cons) shows the estimated variance of the level 2 variable (PAC/LAC) intercepts.

Table 5.

Multilevel logistic regression model (level 2 only).

Court vs Caution	Coef.	Std. Err.	z	P > \|z\|	Lower 95 CI	Upper 95 CI
_cons	−1.832	0.049	−37.570	0.000	−1.927	−1.736
PAC_LAC
var(_cons)	0.118	0.026			0.077	0.181

LR test vs. logistic model: χ² = 342.88 Prob > = χ² < 0.001.

The chi-square test underneath the table compares a simple logistic regression model against a multilevel model. The difference is statistically significant; indicating that the multilevel model explains significantly more of the variation in prosecution rates. Although not shown here, the ICC (proportion of overall variance explained by PAC/LACs) is 0.035. This is below the conventional 5% threshold for a multi-level model recommended by some authors (Mehmetoglu & Jakobsen, 2017) but, given the significant improvement in model fit revealed in Table 5, we proceed with a multi-level approach.

Table 6 shows the results of a full multilevel model, including both level 1 and level 2 variables. The effect estimates are expressed in odds ratios rather than coefficients, for ease of interpretation.

Table 6.

Multilevel logistic regression model (levels 1 and 2).

Court vs Caution	Odds ratio	Std. Err.	z	P > \|z\|	Lower 95 CI	Upper 95 CI
Age group
14	2.111	0.220	7.180	<0.001	1.721	2.589
15	3.014	0.293	11.360	<0.001	2.492	3.646
16	3.665	0.350	13.590	<0.001	3.039	4.420
17	4.883	0.462	16.740	<0.001	4.056	5.879
Sex
Male	1.050	0.047	1.090	0.277	0.962	1.145
Area
Inner regional	1.089	0.079	1.170	0.241	0.944	1.256
Outer regional	1.065	0.103	0.650	0.517	0.881	1.287
Remote/very remote	0.781	0.155	−1.250	0.213	0.529	1.153
Concurrent offences
More than one	3.108	0.132	26.730	<0.001	2.860	3.378
Offence severity
Serious	4.507	0.201	33.760	<0.001	4.130	4.919
Serious violence
Yes	2.326	0.154	12.730	<0.001	2.042	2.648
Prior warnings
One	1.287	0.085	3.810	<0.001	1.130	1.466
More than one	1.862	0.170	6.810	<0.001	1.557	2.227
Prior cautions
One	1.769	0.091	11.070	<0.001	1.599	1.957
More than one	5.689	0.373	26.540	<0.001	5.004	6.469
Prior conferences
One	3.384	0.323	12.780	<0.001	2.807	4.080
More than one	9.647	2.318	9.430	<0.001	6.024	15.449
Prior court apps
One or more	5.535	0.331	28.620	<0.001	4.923	6.223
Indigenous status
Indigenous	1.830	0.102	10.850	<0.001	1.641	2.042
Unknown	0.830	0.058	−2.650	<0.001	0.723	0.952
Index year
2011	1.309	0.100	3.510	<0.001	1.126	1.522
2012	1.305	0.108	3.220	<0.001	1.110	1.534
2013	1.746	0.148	6.580	<0.001	1.479	2.061
2014	1.414	0.131	3.750	<0.001	1.180	1.696
2015	1.041	0.102	0.420	0.594	0.860	1.262
2016	1.386	0.134	3.380	<0.001	1.147	1.675
2017	1.198	0.118	1.840	<0.001	0.988	1.452
2018	1.287	0.124	2.610		1.065	1.555
2019	1.157	0.112	1.500	0.133	0.957	1.399
2020	0.991	0.099	−0.090	0.931	0.815	1.206
2021	1.106	0.103	1.090	0.275	0.923	1.327
Cons	0.005	0.001	−43.480	<0.001	0.004	0.006
PAC_LAC
var(_cons)	0.067	0.019			0.038	0.116

There are several points to note about Table 6. First, after controlling for other factors, the variables Area and Sex are no longer correlated with the likelihood of prosecution. Second, the direction of the association between likelihood of prosecution and the remaining variables is the same as that observed in the bivariate analysis. More specifically, the odds of prosecution increase with age and are higher for those who have more than one concurrent offence, committed more serious offences, committed a serious violence offence, or who have had a prior warning, caution, or court appearance. Thirdly, after adjusting for age, gender, offence seriousness, area of residence, concurrent offences, prior contact with the criminal justice system and index year, Indigenous status remains a strong predictor of whether a young offender will be prosecuted. The odds of being prosecuted for a caution-eligible offence are 1.83 times higher if the offender is Indigenous than if the offender is non-Indigenous.

PAC/LAC level variation

The ICC for the full model is 0.02 (down from 0.035), indicating that inclusion of the level 1 covariates reduced some of the variance attributable to PAC/LACs. In fact, after adjusting for level 1 covariates, the percentage of variation in willingness to prosecute a juvenile attributable to the officer’s PAC/LAC is comparatively small. Even so, the variation is noteworthy. This can be seen in Figure 1, which shows the deviation of each PAC/LAC from the mean likelihood of prosecution across all PAC/LACs after controlling for relevant legal factors. Approximately half (45%) of the PAC/LACs are significantly above or below the average for the group as a whole, with half of these significantly below the average for the group and half significantly above the average for the group.

Figure 1.

PAC/LAC effects on the likelihood of prosecution.

Interactions

We now turn to the question of whether there are any interaction effects. Since Sex is no longer significant in the multivariate model, we focus on the interaction between Indigenous status and age. The estimated probability of criminal prosecution as a function of age and Indigenous status (controlling for other factors) is shown in Figure 2. The upper curve shows the effect of age for young Indigenous offenders. The lower curve shows the effect of age for non-Indigenous offenders. The vertical lines drawn through each point estimate are the 95 confidence intervals surrounding the estimated effects at each age.

Figure 2.

Mean likelihood of prosecution by age and Indigenous status.

The likelihood of prosecution rises with age for both Indigenous and non-Indigenous offenders – however, the risk of prosecution is higher at every age for the former than for the latter, and the difference increases with age. For those aged 13 and under, the difference between the Indigenous and non-Indigenous probability of prosecution is 0.062 (viz., 6.2 percentage points). At age 17, the difference in the probability of prosecution rises to 0.266 (viz., 26.6 percentage points). Put another way, at age 13 and under, 8.5% of young Indigenous offenders are prosecuted (rather than cautioned) compared with 2.3% of non-Indigenous offenders. By age 17, 43.2% of young Indigenous offenders are prosecuted, compared with 16.6% of non-Indigenous offenders.

Robustness checks

To check the foregoing results, we conducted four robustness checks. The first two address the fact that the proportion of cases where the Indigenous status of the defendant is unknown is quite large. It might be argued that the results could have been different if the Indigenous status for all cases had been known. To address this concern, we run the model shown in Table 4 assuming that all cases where the Indigenous status was unknown actually involved Indigenous offenders (see Appendix 1). We then repeat the analysis on the assumption that all causes where Indigenous status was unknown were non-Indigenous. In both analyses, Indigenous defendants were found to be significantly more likely to be prosecuted than non-Indigenous offenders (see Appendices 1 and 2).

The third robustness check addresses the fact that, although we removed cases involving any principal offence that would have rendered an offender not eligible to receive a YOA caution, it is possible that some offence other than the principal offence would have rendered the offender ineligible for a YOA caution. If Indigenous offenders were more likely for some reason to have additional offences not eligible for a YOA caution than non-Indigenous offenders, we would have overestimated the percentage of cases where Indigenous offenders were prosecuted when they could have been cautioned. To test this possibility, we re-ran the model on offenders who had only one offence. This group makes up 78% of the offenders whose records are included in the study. Instead of reducing the odds ratio, the change increased it slightly, with the odds ratio for Indigenous prosecution rising from 1.83 to 1.9. The effect of Indigenous status remained significant (see Appendix 3).

The fourth check addresses the possibility that policing policy might have altered during the period of the COVID lockdown, somehow artificially inflating the prosecution rate of young Indigenous offenders relative to their non-Indigenous counterparts. To test this possibility, we run the model leaving out all court appearances that occurred after 2020. The results of this analysis return an odds ratio on Indigenous status of 1.91 (see Appendix 4).

Discussion and conclusion

The aim of the current study was to examine the influence of Indigenous status, age, and sex on the police decision to prosecute a young person who they believe to have committed a criminal offence, rather than caution them. After adjusting for the effects of factors police are required to consider, we found no effect of sex on the risk of prosecution. Indigenous juvenile defendants (regardless of sex), however, are more likely to be prosecuted than cautioned compared with their non-Indigenous counterparts. This effect holds up after controlling for age, sex, area, number of concurrent offences, offence seriousness and whether or not the offender has previously received a prior warning, a prior conference, a prior caution or had previously appeared in court. At age 13 and under, 9.3% of young Indigenous defendants are prosecuted (rather than cautioned) compared with 3.1% of non-Indigenous defendants. By age 17, 40.7% of young Indigenous offenders are prosecuted, compared with 16.9% of non-Indigenous offenders.

Although we focus on the contrast between prosecution and police cautions, our results are consistent with those obtained by Ringland and Smith (2013) in their analysis of the contrast between prosecution and diversion (in all forms). They found the odds of a young Indigenous defendant being diverted from court were roughly half (0.53) those of a young non-Indigenous defendant. We find the odds of a young Indigenous defendant being cautioned rather than prosecuted are roughly half (0.57) those of a young non-Indigenous defendant. These are quite substantial differences, and they raise the obvious question of why Indigenous young people are more likely to be prosecuted for a caution-eligible offence than non-Indigenous young people. One possibility is that young Indigenous defendants are treated more harshly simply because police harbour a prejudice against them. There is no direct survey evidence on police attitudes to Indigenous youth, but multiple reports have appeared in the media and in Government reports over many years documenting police and custodial officer mistreatment of young Indigenous defendants (Australian Human Rights Commission, 1996; Cunneen, 1991; Law Enforcement Conduct Commission, 2020; McPhee & Chain, 2021; Sentas & Pandolfini, 2017). A related possibility is that postulated by of focal concerns theorists; lacking reliable information on risk of reoffending, police may use Indigenous status as a signal of elevated future risk.

Direct racial discrimination, however, is not the only possible explanation for our findings. As noted in the introduction, under s.20(2e) of the YOA, a police officer contemplating whether to issue a caution can consider any other matter they think is appropriate in the circumstances. They could, for example, consider the response of the young person to being apprehended or detained by police, the attitude of the offender towards police, the people the defendant associates with, the family background of the young defendant, the size of the crime problem in the defendant’s neighbourhood or the attitude of senior officers towards the use of cautions with juvenile offenders. We do not know what factors police consider under the heading of s.20(2e). It is clear, however, that s.20(2e) of the YOA opens the door to a consideration of factors that are irrelevant and potentially prejudicial to the question of whether a young person should be cautioned. This may be one of the reasons behind the substantial variation across police commands in the rate at which YOA cautions are issued.

The findings reported here are a matter of concern. The Indigenous juvenile detention rate in Australia is currently 20 times that of non-Indigenous juveniles (Australian Institute of Health and Welfare, 2021). Australian, State and Territory Governments have set explicit targets for reducing the rate of Indigenous imprisonment. One of those targets involves reducing the detention rate of Indigenous and Torres Strait Islander young people (10–17) by at least 30% by 2031 (National Agreement on Closing the Gap, 2020). Reducing the number of Indigenous young people who end up with a criminal conviction and are thereby exposed to the risk of a custodial penalty, is likely to prove critical to achievement of this target. Past research has shown that a criminal conviction can function as a ‘snare’; narrowing an offender’s options and increasing the likelihood of further involvement in crime (Farrington, 1977 , 1978; Moffitt, 1993). Farrington (1977), for example, matched young offenders aged 14–18 on self-reported offending and compared those whose offending resulted in a criminal conviction with those who, up to that point, had received no criminal conviction. The results indicated higher rates of subsequent offending among those criminally convicted. This effect persisted up to age 21 (Farrington, 1978).

All research is subject to caveats, and the current study is no exception. Willingness to admit the offence is a pre-condition of being given a caution under the YOA. To ensure those who were referred to court would have admitted the offence, had they been offered a caution, we limited the prosecution group to those who pleaded guilty to the principal offence when they appeared court. It is possible that those who pleaded guilty at court, refused to admit the offence when first apprehended by police. To the extent that this occurred, our results would overstate the difference between Indigenous and non-Indigenous defendants in the likelihood of a police caution for a caution-eligible offence. The proportion of those who pleaded guilty at court but who refused to admit the offence when first apprehended by police is, unfortunately, unknown. It is worth noting, however, that police are required to contact the Aboriginal Legal Service (ALS) whenever they take a young Indigenous person into custody. This allows the ALS to raise the possibility of a ‘Protected Admission’ (NSW Police, 2019). If accepted by the young person, this admission cannot be used against the young person in court. The ALS maintains that they encourage a YOA disposal if police willing to do so, and the young person agrees with allegationⁱ (Clifford, 2022).

The current study provides strong evidence that young Indigenous defendants are more likely to be prosecuted for offending compared to non-Indigenous counterparts, even after controlling for a range of factors the law permits police to consider when deciding whether to caution or prosecute a juvenile offender. The present study, however, does not provide any insight into the reasons for this difference in treatment. Because the law itself is rather vague about the factors police can and cannot consider when deciding whether to caution a young offender, we cannot tell whether the difference in cautioning rates is justified or not. It is unlikely that further research of the kind reported here will help in resolving this issue. A more productive course of action might be to follow the approach adopted by Doob and Chan (1982) and interview police and young offenders to see what they have to say about the diversion process under the YOA. It is to be hoped that future research will accept this challenge.

Footnotes

Acknowledgements

Our thanks to Dr Suzanne Poynton, Deputy Director, NSW Bureau of Crime Statistics and Research and to James Clifford, A/Managing Solicitor, Children’s Criminal Practice, Aboriginal Legal Service (NSW/ACT) Limited, both of whom read and made a number of helpful suggestions on an earlier draft of this manuscript. Thanks also to Sara Rahman and Clare Ringland for providing the STATA code for the ‘caterpillar’ graph.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Don Weatherburn

Notes

Appendix 1: All Indigenous missing cases are Indigenous.

Variable	Odds ratio	Std. Err.	z	P > \|z\|	Lower 95 CI	Upper 95 CI
Age group
14	2.076	0.215	7.060	0.000	1.695	2.543
15	2.880	0.278	10.970	0.000	2.384	3.479
16	3.531	0.335	13.290	0.000	2.932	4.253
17	4.616	0.433	16.290	0.000	3.840	5.548
Sex	1.036	0.046	0.810	0.421	0.950	1.130
Area
Inner regional	1.113	0.082	1.460	0.144	0.964	1.285
Outer regional	1.116	0.109	1.120	0.264	0.921	1.352
Remote/very remote	0.945	0.189	−0.280	0.778	0.638	1.400
Concurrent offences
More than one	3.194	0.135	27.490	0.000	2.940	3.469
Offence severity
Serious	4.579	0.204	34.180	0.000	4.197	4.997
Serious violence
Yes	2.306	0.152	12.660	0.000	2.026	2.625
Prior warnings
One	1.313	0.087	4.120	0.000	1.154	1.495
More than one	1.888	0.172	7.000	0.000	1.580	2.256
Prior cautions
One	1.810	0.093	11.570	0.000	1.637	2.002
More than one	5.943	0.387	27.380	0.000	5.231	6.751
Prior conferences
One	3.377	0.322	12.780	0.000	2.802	4.070
More than one	9.618	2.309	9.430	0.000	6.008	15.396
Prior court (yes)	5.739	0.341	29.390	0.000	5.108	6.449
Indigenous	1.334	0.062	6.150	0.000	1.217	1.462
Index year
2011	1.307	0.100	3.490	0.000	1.124	1.518
2012	1.292	0.106	3.110	0.002	1.100	1.519
2013	1.716	0.145	6.390	0.000	1.454	2.025
2014	1.389	0.128	3.560	0.000	1.159	1.664
2015	1.016	0.099	0.160	0.869	0.839	1.230
2016	1.344	0.129	3.070	0.002	1.113	1.624
2017	1.136	0.111	1.310	0.191	0.938	1.376
2018	1.232	0.118	2.170	0.030	1.020	1.487
2019	1.088	0.105	0.880	0.381	0.901	1.314
2020	0.919	0.091	−0.860	0.392	0.756	1.116
2021	1.021	0.094	0.220	0.822	0.853	1.223
_cons	0.005	0.001	−43.350	0.000	0.004	0.007
PAC_LAC
var(_cons)	0.073	0.020			0.043	0.123

Appendix 2: All Indigenous missing cases are non-Indigenous.

Variable	Odds ratio	Std. Err.	z	P > \|z\|	Lower 95 CI	Upper 95 CI
Age group
14	2.117	0.220	7.210	0.000	1.727	2.596
15	3.015	0.293	11.370	0.000	2.493	3.647
16	3.676	0.351	13.630	0.000	3.048	4.433
17	4.887	0.463	16.760	0.000	4.059	5.883
Sex	1.053	0.047	1.160	0.247	0.965	1.149
Area
Inner regional	1.081	0.078	1.080	0.279	0.939	1.246
Outer regional	1.060	0.102	0.600	0.548	0.877	1.279
Remote/very remote	0.783	0.155	−1.240	0.216	0.531	1.154
Concurrent offences
More than one	3.124	0.132	26.880	0.000	2.875	3.395
Offence severity
Serious	4.512	0.201	33.780	0.000	4.134	4.924
Serious violence
Yes	2.320	0.154	12.700	0.000	2.038	2.642
Prior warnings
One	1.288	0.085	3.820	0.000	1.131	1.466
More than one	1.863	0.170	6.820	0.000	1.558	2.227
Prior cautions
One	1.769	0.091	11.080	0.000	1.599	1.957
More than one	5.694	0.373	26.570	0.000	5.009	6.474
Prior conferences
One	3.384	0.323	12.790	0.000	2.808	4.079
More than one	9.625	2.313	9.420	0.000	6.010	15.414
Prior court (yes)	5.533	0.331	28.640	0.000	4.922	6.220
Indigenous	1.887	0.103	11.630	0.000	1.696	2.100
Index year
2011	1.310	0.101	3.520	0.000	1.127	1.523
2012	1.304	0.108	3.210	0.001	1.109	1.532
2013	1.737	0.147	6.520	0.000	1.471	2.050
2014	1.406	0.130	3.680	0.000	1.173	1.685
2015	1.033	0.101	0.330	0.743	0.852	1.251
2016	1.372	0.132	3.280	0.001	1.135	1.658
2017	1.178	0.115	1.670	0.095	0.972	1.427
2018	1.266	0.122	2.440	0.014	1.048	1.528
2019	1.132	0.109	1.280	0.201	0.936	1.367
2020	0.965	0.096	−0.360	0.721	0.795	1.172
2021	1.073	0.099	0.770	0.442	0.896	1.285
_cons	0.005	0.001	−43.760	0.000	0.004	0.006
PAC_LAC
var(_cons)	0.067	0.018			0.039	0.113

Appendix 3: Model with no prior CJS contacts.

Variable	Odds ratio	Std. err.	z	P > \|z\|	Lower 95 CI	Upper 95 CI
Age group
14	2.482	0.376	5.990	0.000	1.843	3.341
15	3.831	0.542	9.490	0.000	2.903	5.056
16	4.850	0.680	11.270	0.000	3.686	6.383
17	6.537	0.904	13.570	0.000	4.985	8.573
Sex
Male	1.121	0.070	1.840	0.066	0.992	1.267
Area
Inner regional	1.050	0.100	0.510	0.608	0.871	1.266
Outer regional	1.027	0.133	0.210	0.835	0.797	1.325
Remote/very remote	0.769	0.217	−0.930	0.351	0.443	1.336
Concurrent
More than one	2.792	0.169	16.980	0.000	2.480	3.143
Offence severity
Serious	5.000	0.322	25.030	0.000	4.408	5.672
Serious violence
Yes	2.227	0.187	9.530	0.000	1.889	2.626
Indigenous status
Indigenous	1.901	0.162	7.530	0.000	1.609	2.248
Index year	0.929	0.086	−0.790	0.427	0.775	1.114
2011	1.728	0.201	4.690	0.000	1.375	2.171
2012	1.589	0.199	3.700	0.000	1.243	2.031
2013	2.527	0.312	7.520	0.000	1.984	3.218
2014	1.768	0.241	4.170	0.000	1.353	2.310
2015	1.474	0.209	2.740	0.006	1.117	1.945
2016	2.028	0.280	5.120	0.000	1.547	2.659
2017	1.661	0.236	3.570	0.000	1.257	2.193
2018	1.828	0.257	4.300	0.000	1.388	2.408
2019	1.722	0.240	3.900	0.000	1.310	2.264
2020	1.424	0.206	2.450	0.014	1.073	1.890
2021	1.514	0.204	3.070	0.002	1.162	1.973
_cons	0.003	0.000	−33.510	0.000	0.002	0.004
PAC_LAC
var(_cons)	0.094	0.028			0.053	0.169

Appendix 4: Model with no cases after 2020.

Variable	Odds ratio	Std. Err.	z	P > \|z\|	Lower 95 CI	Upper 95 CI
Age group
14	2.003	0.228	6.120	0.000	1.604	2.503
15	2.889	0.304	10.070	0.000	2.350	3.552
16	3.497	0.362	12.110	0.000	2.856	4.283
17	4.551	0.466	14.800	0.000	3.723	5.561
Sex	1.040	0.050	0.830	0.408	0.947	1.142
Area
Inner regional	1.168	0.090	2.010	0.045	1.004	1.359
Outer regional	1.112	0.113	1.040	0.298	0.910	1.358
Remote/very remote	0.897	0.192	−0.510	0.612	0.589	1.365
Concurrent offences
More than one	3.149	0.144	25.040	0.000	2.878	3.444
Offence severity
Serious	4.371	0.209	30.850	0.000	3.980	4.801
Serious violence
Yes	2.459	0.179	12.350	0.000	2.132	2.837
Prior warnings
One	1.243	0.092	2.950	0.003	1.076	1.437
More than one	1.916	0.200	6.240	0.000	1.562	2.350
Prior cautions
One	1.803	0.099	10.690	0.000	1.618	2.009
More than one	5.847	0.408	25.310	0.000	5.100	6.704
Prior conferences
One	3.445	0.353	12.080	0.000	2.819	4.210
More than one	10.220	2.755	8.620	0.000	6.026	17.334
Prior court (yes)	5.451	0.347	26.650	0.000	4.812	6.176
Indigenous	1.901	0.162	7.530	0.000	1.609	2.248
Index year	1.822	0.109	10.050	0.000	1.621	2.048
2011	0.928	0.073	−0.950	0.341	0.796	1.082
2012
2013	1.309	0.101	3.500	0.000	1.126	1.522
2014	1.306	0.108	3.230	0.001	1.111	1.536
2015	1.737	0.147	6.500	0.000	1.471	2.051
2016	1.409	0.131	3.690	0.000	1.175	1.690
2017	1.041	0.102	0.410	0.681	0.859	1.262
2018	1.382	0.134	3.340	0.001	1.143	1.671
2019	1.186	0.117	1.730	0.084	0.977	1.439
2020	1.284	0.125	2.570	0.010	1.061	1.552
2021	1.151	0.112	1.440	0.150	0.951	1.393
_cons	0.005	0.001	−40.740	0.000	0.004	0.007
PAC_LAC
var(_cons)	0.076	0.020			0.045	0.128

References

(2010). Diversion of Aboriginal and Torres Strait Islander young people in Queensland: The impact, use and alternatives to existing diversionary measures in addressing Indigenous over-representation. https://www.aic.gov.au/about-us/governance/criminology-research-advisory-council.

Allison

Schwartz

Cunneen

(2013). ‘That’s discrimination!’ Indigenous peoples’ experiences of discrimination in the Northern Territory. Indigenous Law Bulletin, 8(5), 8–12. UNSW Law Research Paper No. 2014-15.

Anthony

(2015). Indigenous people, crime, and punishment. Oxfordshire.

Australian Bureau of Statistics. (2011). Australian and New Zealand Standard Offence Classification (ANZSOC). Australian Bureau of Statistics. https://www.bocsar.nsw.gov.au/Pages/bocsar_publication/Pub_Summary/CJB/cjb95-Matching-Court-Records-to-Measure-Reoffending.aspx

Australian Human Rights Commission. (1996). Indigenous deaths in custody 1989–1996. Chapter 9. https://humanrights.gov.au/our-work/aboriginal-and-torres-strait-islander-social-justice/publications/indigenous-deaths.

Australian Institute of Health and Welfare. (2022). Youth Justice in Australia 2020-21. https://www.aihw.gov.au/reports/youth-justice/youth-justice-in-australia-2020-21/contents/summary.

Blagg

(2008). Crime, Aboriginality, and the decolonisation of justice. The Federation Press, Sydney.

Blagg

Morgan

Cunneen

Ferrante

(2005). Systemic Racism as a Factor in the Overrepresentation of Aboriginal People in the Victorian Criminal Justice System. Retrieved from: http://www.humanrightscommission.vic.gov.au/our-resources-and-publications/reports/item/download/8360_dd5d1fe14ac11c5eb730b6bba2b68a68. Accessed 12/04/2022.

Bond

Jeffries

(2010). Sentencing indigenous and non-indigenous women in Western Australia‘s higher courts. Psychiatry, Psychology and Law, 17(1), 70–78. https://doi.org/10.1080/13218710903089253

10.

Bond

Jeffries

(2012). Indigeneity and the likelihood of imprisonment in Queensland’s adult and children’s courts. Psychiatry, Psychology, and the Law, 19(2), 169–183. https://doi.org/10.1080/13218719.2010.543757

11.

Bond

Jeffries

(2012b). Harsher sentences? Indigeneity and prison sentence length in Western Australia’s highest courts. Journal of Sociology, 48(3), 266–286. https://doi.org/10.1177/1440783311413486

12.

Broadhurst

(1997). Aborigines and crime in Australia. Crime and Justice: A Review of Research, 21, 407–468. https://doi.org/10.1086/449255

13.

Bushway

Piehl

A. M.

(2001). Judging judicial discretion: Legal factors and racial discrimination in sentencing. Law and Society Review, 35(4), 733–766. https://doi.org/10.2307/3185415

14.

Clifford

(2022). A managing solicitor, children’s criminal practice aboriginal legal service (NSW/ACT) limited. Personal communication.

15.

Cunneen

(1991). Aboriginal young people and police violence. Aboriginal Law Bulletin, 17(6). http://www.bocsar.nsw.gov.au/Documents/BB/Report-2015-Prison-penalties-for-serious-domestic-and-non-domestic-assault-bb110.pdf

16.

Donnelly

Poynton

(2015). Prison penalties for serious domestic and non-domestic assault (Bureau Brief no. 110). http://www.bocsar.nsw.gov.au/Documents/BB/Report-2015-Prison-penalties-for-serious-domestic-and-non-domestic-assault-bb110.pdf.

17.

Doob

Chan

(1982). Factors affecting police decisions to take juveniles to court. Canadian Journal of Criminology, 24(1), 25–37. https://doi.org/10.3138/cjcrim.24.1.25

18.

Ericson

R. D.

Eckberg

D. A.

(2016). Racial disparity in juvenile diversion: The impact of focal concerns and organizational coupling. Race and Justice, 6(1), 35–56. https://doi.org/10.1177/2153368715594848

19.

Gale

Bailey-Harris

Wundersitz

1990). Aboriginal youth and the criminal justice system. Cambridge University Press.

20.

Gallagher

Poletti

MacKinnell

(1998). Sentencing Disparity and the Ethnicity of Juvenile Offenders. Judicial Commission of New South Wales. Research Monograph 17. https://www.judcom.nsw.gov.au/research-monograph-17-sentencing-disparity-and-the-ethnicity-of-juvenile-offenders/.

21.

Farrington

D. P.

(1977). The effects of public labelling. British Journal of Criminology, 17(2), 112–125. https://www-jstor-org-443.web.bisu.edu.cn/stable/23636356 https://doi.org/10.1093/oxfordjournals.bjc.a046802

22.

Farrington

D. P.

(1978). The persistence of labelling effects. British Journal of Criminology, 18(3), 277–284. https://doi.org/10.1093/oxfordjournals.bjc.a046913

23.

Hua

Fitzgerald

(2006). Matching court records to measure reoffending. In Crime and Justice Bulletin 95 (pp. 1–12). NSW Bureau of Crime Statistics and Research. https://www.bocsar.nsw.gov.au/Pages/bocsar_publication/Pub_Summary/CJB/cjb95-Matching-Court-Records-to-Measure-Reoffending.aspx

24.

Jeffries

Bond

(2009). Does indigeneity matter? Sentencing indigenous offenders in South Australia’s higher courts. Australian and New Zealand Journal of Criminology, 42(1), 47–71. https://doi.org/10.1375/acri.42.1.47

25.

Jeffries

Bond

(2012). The impact of indigenous status on adult sentencing: A review of the statistical research literature from the United States, Canada, and Australia. Journal of Ethnicity in Criminal Justice, 10(3), 223–243. https://doi.org/10.1080/15377938.2012.700830

26.

Judicial Commission of New South Wales. (2022). Lawcodes. https://lawcodes.judcom.nsw.gov.au/.

27.

Law Enforcement Conduct Commission. (2020). An investigation into the formulation and use of the NSW Police Force Suspect Targeting Management Plan on children and young people. Operation Tepito Interim Report pursuant to Part 6 LECC Act.

28.

Loh

Ferrante

(2003). Aboriginal involvement in the Western Australian Criminal Justice System: A statistical review, 2001. Crime Research Centre, University of Western Australia.

29.

Luke

Cunneen

(1995). Aboriginal overrepresentation and discretionary decisions in the NSW juvenile justice system. Juvenile Justice Council of NSW.

30.

MacKinnell

Poletti

Holmes

(2010). Measuring offence seriousness. In Crime and Justice Bulletin 142 (pp. 1–12). NSW Bureau of Crime Statistics and Research. https://www.bocsar.nsw.gov.au/Pages/bocsar_publication/Pub_Summary/CJB/cjb142-Measuring-offence-seriousness.aspx

31.

McPhee

Chain

(2021, May 4). Cop who ‘kicked an aboriginal teen to the ground face-first for threatening him’ is charged with assault and bodily harm after a year-long investigation. Daily Mail Australia. https://www.dailymail.co.uk/news/article-9539957/Cop-kicked-Aboriginal-teen-ground-face-threatening-charged-assault.html.

32.

Mehmetoglu

Jakobsen

(2017). Applied statistics using Stata: A guide for the social sciences. Sage Publications.

33.

Miller

W. B.

(1958). Lower class culture as a generating milieu of gang delinquency. Journal of Social Issues, 14(3), 5–19. https://doi.org/10.1111/j.1540-4560.1958.tb01413.x

34.

Mitchell

(2005). A meta-analysis of race and sentencing research: Explaining the inconsistencies. Journal of Quantitative Criminology, 21(4), 439–466. https://doi.org/10.1007/s10940-005-7362-7

35.

Moffitt

T. E.

(1993). Adolescent-limited and life-course-persistent antisocial behaviour: A developmental taxonomy. Psychological Review, 100(4), 674–701. https://doi.org/10.1037/0033-295X.100.4.674

36.

National Agreement on Closing the Gap. (2020). Closing the gap in partnership. Retrieved from: https://www.closingthegap.gov.au/sites/default/files/files/national-agreement-ctg.pdf. Accessed 12/4/2021.

37.

NSW Police. (2019). NSW police handbook. https://register.dunefun.com/?fl = m9&m = books&sub = %7Bsubid%7D&offer_id = 42729&campaign_id = 1010188&lid = 935adfc8-b4da-4d78-93df-b4cf6df39cd6&ap = 2&src = 143939&hash = 8cf0dc3b610ef273cda019605e3aca28.

38.

Ringland

Smith

(2013). Police use of court alternatives for young persons in NSW. In Crime and Justice Bulletin 167 (pp. 1–12). NSW Bureau of Crime Statistics and Research.

39.

Sentas

Pandolfini

(2017). Policing young people in NSW: A study of the suspect targeting management plan. In A report of the Youth Justice Coalition NSW (pp. 1–64). Youth Justice Coalition NSW.

40.

Snowball

(2008). Juvenile diversion of Indigenous offenders: A study examining juvenile offenders in Western Australia, South Australia, and New South Wales. Criminology Research Council.

41.

Snowball

Weatherburn

(2006). Indigenous over-representation in prison: The role of offender characteristics. In Crime and Justice Bulletin 99 (pp. 1–20). NSW Bureau of Crime Statistics and Research. https://www.bocsar.nsw.gov.au/Pages/bocsar_publication/Pub_Summary/CJB/cjb99-Indigenous-over-representation-in-prison-The-role-of-offender-characteristics.aspx

42.

Snowball

Weatherburn

(2007). Does racial bias in sentencing contribute to indigenous overrepresentation in prison? Journal of Criminology, 40(3), 272–290. https://doi.org/10.1375/acri.40.3

43.

Steffensmeier

Ulmer

Kramer

(2006). The interaction of race, gender, and age in criminal sentencing: The punishment cost of being young, black, and male. Criminology; An Interdisciplinary Journal, 36(4), 763–797. https://doi.org/10.1111/j.1745-9125.1998.tb01265.x

44.

Thorburn

Weatherburn

(2018). Effect of indigenous status on sentence outcomes for serious assault offences. Journal of Criminology, 51(3), 4343–4453. https://doi.org/10.1177/0004865817748179

45.

Wundersitz

(1997). Pre-court diversion: The Australian experience. In Borowski

O’Connor

(Eds.), Juvenile crime, justice, and corrections. Longman.