Cale, Jesse; Burton, Melanie --- "Factors Associated with Breaches of Home Detention and Returns to Custody Post-Home Detention in South Australia" [2018] CICrimJust 10; (2018) 30(1) Current Issues in Criminal Justice 35

	Home \| Databases \| WorldLII \| Search \| Feedback Current Issues in Criminal Justice

Factors Associated with Breaches of Home Detention and Returns to Custody Post-Home Detention in South Australia

In 2016, the South Australian Government proposed expanding the use of Home Detention (‘HD’) sentences. A key concern of government and community alike regarding HD sentences has to do with the eligibility of prisoners for the sanction and, more specifically, the likelihood that individuals serving HD sentences will pose a threat to public safety when they are serving their sentences in the community. The aim of the current study was to establish baseline empirical evidence about: (a) the profile of prisoners serving HD sentences in the state; (b) factors associated with breaches of HD sentences; and (c) the nature and extent of reoffending by prisoners serving HD sentences. The sample consists of a cohort of prisoners released to HD in South Australia between June 2014 and June 2015 who were followed until June 2017. The results show that the most robust predictors of breaches of HD orders and returns to custody were risk assessments while prisoners were in custody. Furthermore, prisoners convicted of violent offences who received a HD sentence were less likely to return to custody compared to those who committed non-violent offences. Finally, the vast majority of offences for which prisoners returned to custody following a HD order were administrative offences.

Keywords: home detention – sentencing – eligibility – prisoner profiles –electronic monitoring – recidivism – South Australia – Australia

Until 2016 in South Australia (‘SA’), home detention (‘HD’) was applied only as a ‘back-end’ sentencing option (referred to as ‘Release Ordered Home Detention’ (‘ROHD’)) available only to certain prisoners who had served at least half of the non-parole period of their sentences in prison. In mid-2016, there were a number of legislative, policy and program changes to expand current HD provisions, which included the expansion of the existing ROHD scheme under the Correctional Services Act 1988 (SA) and the introduction of Court Ordered Home Detention (‘COHD’) under the Criminal Law (Sentencing) Act 1986 (SA).

Similar to other Australian states (for example, New South Wales (‘NSW’), Northern Territory (‘NT’)), HD orders are not available for the most serious/violent offences such as homicide, sexual, or terrorist offences. Home detention is also not typically available in SA for ‘low-risk’ individuals who would not be eligible for a custodial sanction for their offence to avoid potential net-widening effects. Typical conditions for HD include: suitability of residence; regular contact and compliance with parole officers; and movement restriction conditions of HD orders. Additional conditions can also be imposed, such as drug testing and certain program attendance, on a case-by-case basis.

A common concern for governments and communities alike surrounding HD has to do with public safety and the likelihood that individuals serving HD sentences will commit offences while they are in the community. To investigate this issue, the current study provides empirical evidence on: (a) the profile of prisoners on HD orders; (b) the nature and extent of reoffending patterns of prisoners on HD orders; and (c) factors associated with breaching HD orders and reoffending post HD, in the state of SA.

Although non-custodial alternative sanctions to prison sentences are gaining increasing popularity in Western nations, due in large part to inflating prison populations and associated increasing costs in corrective services, they are far from novel in theory or practice. In the early-to-mid 20^th century, a key focus of early non-custodial alternative sanctions was based on a shift away from punishment as a key goal of sentencing towards rehabilitation-oriented sanctions that promote community reintegration (Ball & Lilly 1986). This represented the first generation of contemporary alternatives to custody in the modern criminal justice system and involved the increased application of probation sentences where an offender avoids prison and is closely monitored in the community. Another example is parole, where an offender spends a portion of his or her sentence in prison and, if eligible, the rest in the community to promote reintegration. Suspended sentences involve a prison sentence, but one that is suspended so that, like probation, an offender can remain in the community to promote reintegration but if the offender commits any other offence in the interim, the original jail term is immediately applied. It is well established that alternatives to custody such as these are substantially less costly than prison (for example, see Byrne, Lurigio & Petersilia 1992).

Despite promoting community reintegration of ex-prisoners and offenders, alongside the economic benefits, alternatives to custody are a particularly divisive issue that draw largely on the perception among the public of the appropriate balance between retribution, rehabilitation and public safety. Like most criminal justice issues, the issue of alternatives to custody is highly politicised and, in conservative political climates, is often framed as being too lenient or ‘soft’ on crime (for example, see Bartels & Martonivic 2017; Robinson, McNeill & Maruna 2013). This was particularly evident in the late 1970s in the United States (‘US’) and other Western and industrialised nations when Martinson (1974) published his scathing critique of rehabilitative practices in criminal justice entitled Nothing Works. In the context of alternatives to custody, one of the effects of the 1970s/1980s shift toward more retributive criminal justice policies involved supplementing these sanctions with mechanisms aimed to promote accountability on the part of the offender. Therefore, a second generation of alternatives to custody emerged that sat somewhere between alternative sanctions, such as probation, and incarceration (see Beck & Klein-Saffran 1989) typically referred to as intermediate sanctions (Roberts 2004). One of the most significant practices along these lines was the introduction of HD.

Home detention was introduced in the US in the 1980s as an alternative to incarceration amid prison overcrowding and increasing criminal justice costs that would also promote offender accountability (Clear & Braga 1995; Doherty 1995; Martinovic 2007; Morris & Tonry 1990). This coincided with the introduction of electronic monitoring (‘EM’) technology that created the ability to monitor the whereabouts and presence of convicted offenders in certain places at specific times (Vanhaelemeesch, Beken, & Vandevelde 2014). Following from this, United Nations Congresses adopted resolutions for implementing

non-custodial alternatives to incarceration, resulting in the increasing application of HD since the 1980s around the world (Joutsen & Zvekic 1994). As other countries trialled and implemented HD programs, they typically have been designed and uniquely tailored to specific jurisdictions. While HD was, in part, intended to promote offender accountability, it is also commonly accepted as a more humane alternative to imprisonment with less harmful side effects (Black & Smith 2003; Worrall 2006).

Since the widespread implementation of HD since the 1980s, several jurisdictions have tailored HD programs to suit their respective criminal justice systems. Home detention is also referred to as: home detention curfew; home confinement; house arrest; EM; electronically monitored home confinement; electronic detention house arrest; and home confinement and EM, to name a few (Martinovic 2013). These different terms all describe very similar sanctions that essentially restrict the offender to his or her home as a form of incapacitation and punishment (Nellis 2013). The US Sentencing Commission defines HD as ‘a program of confinement and supervision that restricts the defendant to his place of residence continuously, except for authorised absences, enforced by appropriate means of surveillance by the probation office’ (US Sentencing Commission 2013, p. 427).

The contemporary rationale for HD is that it provides a cost-effective alternative to imprisonment that also avoids many of the negative consequences associated with incarceration in correctional facilities. Here, a key policy view is that allowing the offender to remain in the community promotes successful reintegration because it allows for individuals to retain connections with employment, family relationships and housing (see, for example, Ball, Huff, & Lilly 1988). At the same time, HD avoids the known criminogenic aspects of incarceration, such as the loss of human capital/personal agency, stigmatisation, and accumulation of criminal capital (Grogger 1995; LeBel 2012; Lopoo & Western 2005; Pager 2003; Petersilia 1987). Therefore, HD has the potential to address different sentencing objectives in the criminal justice system.

Similar to Western nations, HD first emerged in Australia in the 1980s as an alternative to incarceration amid prison overcrowding and increasing criminal justice costs that also promotes offender accountability through routine monitoring (Palmer, De Lint & Dalton 2015). Home detention takes different forms and can be utilised at various stages of the criminal justice process. In the pre-trial process, HD can be used as a component of bail designed to ensure the defendant’s appearance at trial and non-interference with witnesses. In Australia this is typically referred to as home detention bail. The use of HD in sentencing is intended as a sanction following an individual’s conviction and is applied either as a

‘front-end’ or ‘back-end’ alternative to incarceration. ‘Front-end’ or ‘primary sentencing’ HD occurs when offenders have their sentences of imprisonment fully suspended and are instead sentenced to serve their time at home. In effect, an offender is sentenced to imprisonment, and then, if deemed suitable (for example, low risk and suitable conditions), the term may be served through HD. More commonly, HD is available for eligible offenders on the ‘back end’ of sentences after a specific period of incarceration, as part of parole or a distinct stage in the sentence. Here, post-trial HD is most frequently targeted to low-to-medium risk, non-violent offenders. In some jurisdictions, offenders who have completed a specified non-parole period in custody and meet eligibility criteria can apply for HD. This is referred to as ‘pre-release’ or ‘court-ordered’ HD in some Australian states.

The application of HD in Australia varies from state to state. Four states (NSW, NT, Western Australia, SA) have legislative provisions for the use of HD in different contexts (for example, pre-trial, front-end, back-end). The remainder of the states do not have specific legislative provisions for the use of HD, although they use EM in different capacities with offenders on community orders. The state of Victoria abolished the use of HD as a sentencing option in 2012, and currently Tasmania is considering phasing out suspended sentences in lieu of other intermediate sanctions, including HD.

In NSW, HD is primarily used as a front-end sentencing option where it can be used as a substitute for a sentence of imprisonment. It is available for offenders sentenced to a maximum of 18 months imprisonment who meet specific suitability criteria (for details, see NSW Justice State Parole Authority 2015). Generally, offenders who have committed the most serious crimes (for example, homicide, sexual assault, armed robbery) are not eligible. Having certain prior convictions along these lines also precludes eligibility for HD. Offenders serving HD orders in NSW are monitored using EM devices and there are general conditions applicable to all orders such as abstinence from drugs and alcohol. Home detention orders in NSW are tailored according to an offender’s level of risk. This means that ‘higher-risk’ offenders on a HD order typically should receive more conditions and priority for certain treatment programs compared to ‘lower-risk’ offenders who are only monitored. Breaches of conditions can result in an offender being placed back into custody, but also may involve warning depending on the nature and extent of the breach. In 2013, the NSW Law Reform Commission noted that use of HD in NSW has substantially declined since 2005 in the state (NSW Law Reform Commission 2013).

Similar to NSW, HD in the NT is used as a front-end sentencing option (for details, see Northern Territory Government 2017). In the NT, HD orders are not available for sentences with mandatory minimum terms of imprisonment or for certain serious and violent offences. It is available as a back-end sentencing option in some cases, where the court can partly suspend a sentence and impose a HD order. Home detention orders can only be made for a maximum of 12 months so, in the front-end context, they can only be imposed in cases where the total term of imprisonment on is 12 months or less, and the court specifies where the offender can live while on HD. General conditions that apply to all HD orders in the NT include: regular contact and compliance with probation or parole officers and movement restriction conditions of HD orders; wearing an EM device; and regular drug and alcohol testing.

In SA, until 2016, HD was applied primarily as a ‘back-end’ sentencing option for individuals who have served at least half of the non-parole period of their sentence, referred to as COHD. These HD orders were not available for the most serious/violent offences such as homicide, sexual, or terrorist offences, nor was HD a sentencing option for individuals convicted of minor offences and considered to be ‘low-risk’. Typical conditions for HD included: suitability of residence; regular contact and compliance with parole officers; and movement restriction conditions of HD orders. Additional conditions could also be imposed — such as drug testing and certain program attendance — on a case-by-case basis. Legislative changes in 2016 provided an expansion of eligibility criteria removing the requirement that offenders complete half of their non-parole period and have less than 12 months to serve to their parole eligibility date. In addition, SA has recently funded the implementation of a wraparound support program that seeks to address the causes of offending and support an offender’s successful rehabilitation and/or reintegration into the community (see Hilferty et al. 2018). While certain Australian states have adopted the use of HD in sentencing, the implementation of HD also varies from state to state. Critically, no recent studies have examined the effectiveness of HD in Australia.

In international studies that have examined the effectiveness of HD, key outcome measures include program cost, rates of program completion, and reoffending. Not surprisingly, although the costs of HD vary by program and jurisdiction, there is widespread consensus that that HD is substantially less costly than incarceration (see, for example, Dodgson et al 2001; Dodgson, Mortimer & Sugg 2000; Fox 1987). There is less consensus when it comes to the impact of HD on reoffending (see Renzema 2003; Renzema & Mayo-Wilson 2005 for reviews). One of the key reasons for this is that outcome measures vary across studies (for example, program completion, breaches on HD, reoffending during and post HD, length of time to reoffending and how long individuals are followed post-HD). There is also substantial variability across programs in terms of conditions (that is, who is eligible for HD), implementation (for example, EM or EM with supervision) and, when HD is implemented through the sentencing process (for example, pre-trial, post-trial), and how (front-end versus back-end) (Gibbs & King 2003). In line with the aims of the current study, below we review the empirical evidence on the effectiveness of ‘back-end’ HD on reoffending.

Evaluations of the effectiveness of HD in the back-end context on reoffending outcomes have produced mixed results. Dodgson et al. (2001) conducted a large-scale evaluation of a back-end HD program utilising EM in England and Wales that was introduced in 1999. In this program, offenders who received short term custodial sentences (that is, anywhere between three months and four years) and considered to be minimal risk of reoffending upon release were eligible for early release from prison on electronic monitored HD. In the first 16 months of the program, more than 21,000 offenders were released to serve the remainder of their sentences on HD in the community. Of those released, only five per cent were returned to prison and the main reasons were due to breaches in curfew conditions (that is, just over

Marie, Moreton and Goncalves (2011) examined data on 63,384 offenders discharged from prison receiving HD between 2000 and 2006 in England and Wales. Using a

quasi-experimental evaluation design, they reported that ten per cent of the sample were recalled to prison while they were on HD; eight per cent were recalled for breaching the terms of their HD, and two per cent were recalled for reoffending while on HD. The specific index offence an individual was convicted of, number of previous offences, and the number of previous breaches, were key factors associated with recalls to prison.

In the US, Padgett, Bales and Blomberg (2006) examined the impact of EM on 75,661 offenders placed on HD in Florida between 1998 and 2002. In addition to individuals serving ‘back-end’ HD sentences, the study also included front-end HD placements. Their results also showed that EM was effective at reducing the likelihood of reoffending and absconding while on HD for violent and property, and drug offenders, among those under EM. Similarly, Marklund and Holmberg (2009) examined the effects of an early release from prison program that used EM and HD with offenders in Sweden. They compared participants in the program with a control group that was not eligible for early release. Electronic monitoring participants were required to have a daily occupation and were subjected to regular drug testing. Examining three-year post-sentence recidivism rates between the two groups, they found that those on EM were significantly less likely to be reconvicted and sentenced to a term of incarceration and/or probation. However, it was not possible in their study to assess the extent to which differences in recidivism rates were associated with EM specifically or other elements of the program.

In contrast, looking at medium-to-high-risk offenders, Finn and Muirhead-Steves (2002) examined the effectiveness of EM by comparing violent offenders on parole with EM to a random sample of offenders on parole not on EM. In this sample, placement on EM required parolees to remain at home except during scheduled activities. They examined returns to prison over a four-year period post-release from custody. The results showed that there was no direct effect of EM that differentiated the two groups. In addition, drug problems and low parole success likelihood scores were associated with an increased likelihood of returning to custody. However, the results did suggest that offenders on EM remained in the community longer before returning to custody than those who were not on EM. In a meta-analysis of recidivism studies of the effects of EM on reducing crime among moderate-to-serious offenders, Renzema and Mayo-Wilson (2005) also concluded that there was no evidence that EM had positive impact on recidivism up to three years across the studies they included. However, their stringent inclusion criteria for the meta-analyses meant that this conclusion was based on only three studies.

More recently, Giess and colleagues (2012, 2013) reported positive impacts of GPS monitoring in two separate studies of parolees who committed serious crimes: one examining high-risk sex offenders and another examining high-risk gang offenders in the US. In the first, Giess et al. (2012) examined parole compliance and reoffending outcomes (that is, re-arrests, reconvictions, returns to custody) of high-risk sex offenders placed on EM compared to matched offenders in the community who were not between 2006 and 2009 in the state of California. Their findings indicated that high-risk sex offenders on EM supervision displayed significantly better outcomes in terms of both compliance with parole conditions and recidivism measures compared to those not monitored on EM. Using the same outcome measures, in the second study, Giess et al. (2013) found that high-risk gang offenders placed on EM were less likely to be rearrested for any offences and violent offences, but were more likely to breach parole conditions compared to matched controls not on EM.

Taken together, studies on the effectiveness of HD from international jurisdictions have produced mixed findings in terms of its impact on recidivism. Research is needed in the Australian context as expanding HD is being explored as an intermediate sanction in some states. With this in mind, the aim of the current study is to provide a baseline picture of a cohort of prisoners released on HD in SA between June 2014 and June 2015, prior to legislative changes described earlier. First, this involves a statistical profile of the cohort of prisoners detailing their demographic characteristics, offence history and index offence characteristics, risk assessment data, and program participation in custody. Next, this cohort was tracked for breaches of HD orders, as well as returns to custody post-HD orders up to June 2017. Finally, a statistical profile of factors associated with breaches of HD orders and returns to custody is outlined.

The current study was based on all prisoners released to HD in SA from June 2014 to June 2015. Up until 2016, prisoners in SA could only receive ROHD, meaning they have all served a period of time in custody prior to being released to serve part of their sentence on HD. Between June 2014 and June 2015 this included a total of 317 prisoners who were approved for HD following assessment by South Australia Department for Correctional Services (‘DCS’) HD Committee, were released onto a HD order and either completed their HD order or breached conditions of the order.

The research protocols for the current study were conducted according to the ethical guidelines stipulated by the Research Ethics Board of the University of New South Wales and the SA DCS. Data for the current study were obtained from the DCS of SA. These data were de-identified by DCS and included: demographic characteristics, current and historical offence information, risk assessment scores, information about involvement in different programs while in custody, sentence details, and information about breaches while on HD orders and returns to custody up until June 2017.

In the current study, five demographic covariates included: (1) age; (2) sex (0=female, 1=male); (3) Aboriginal or Torres Strait Islander (0=no, 1=yes); (4) level of education (0=less than high school and 1=high school or greater); and (5) employment status prior to incarceration (0=unemployed, 1=employed). Offence history variables included the type of offence (based on most serious offence in cases of multiple charges) for which they were serving the current sentence were coded as a: violent (0=no, 1=yes); theft (0=no, 1=yes); drug related (0=no, 1=yes); administrative (0=no, 1=yes); or fraud (0=no, 1=yes). The number of prior sentences was also included as well as the number of non-parole period days that were part of the initial sentence, and the number of days sentenced to HD. Initial and final security rating assessments (that is, level of security environment while in custody) were coded as: 0=low, 1=medium, 2=high, and risk assessment information (Risk of Reoffending (‘RoR’) and Offender Risk Needs Inventory-Revised (‘ORNI-R’)^[1] Scores) were also included. Finally, participation in programs in custody was coded as: prison employment (0=no, 1=yes); prison education (0=no, 1=yes); behavioural change programs (0=no, 1=yes).

Two different outcome variables were measured in the current study. The first was whether prisoners’ records indicated they had breached their HD conditions (0=discharged from HD, 1=breached HD). The second was whether prisoners returned to custody for a new offence (that is, with a new sentence) at some point following the actual discharge date associated with their sentence that included a HD order (0=did not return to custody by June 2017, 1=returned to custody by June 2017).

First, bivariate analyses were used to provide a description of the sample according to demographic characteristics, covariates and outcome variables. These analyses were conducted comparing males and females. Next, two sets of logistic regression models were estimated, the first predicting breaches of HD orders, the second predicting returns to custody by June 2017. Variables for prediction models were selected based on their bivariate association with the respective outcomes (that is, if they were significantly associated with breaches of HD or returns to custody) to determine the value of adjusted odds ratios identifying which variables predicted breaches and returns to custody.

Table 1 (below) provides a bivariate description of the sample characteristics stratified by gender. The vast majority of the sample were male (84.2% male, 15.8% female). The average age of the entire sample at the time they were released to HD was 37.5 (sd=11.5) years, and there were no differences in age between males and females. Approximately nine per cent of the sample were Aboriginal or Torres Strait Islander, and again there were no differences between the proportions of males or females who were Aboriginal or Torres Strait Islander. Just over one-quarter of the sample (28.8%) had high school education or above, with a significantly higher proportion of females (40.5%) compared to males (27.0%) having high school education or above. In contrast, more than one-third of the sample (38.2%) reported being employed prior to their most recent custody episode, and here the proportions were reversed: nearly twice as many males (40.4%) reported employment prior to custody compared to females (20.7%). Approximately one-third of the sample (29.0%) had more than one prior sentence. The average number of prior sentences in the entire sample was 1.8 (sd=1.8, range=1–13) and there were no statistical differences between males and females.

Table 1 also displays risk assessment information and involvement in prison programs. Approximately one-quarter of the entire sample (26.5%) received an initial security rating of ‘high’ upon arrival into custody. However, females were four times more likely to receive a ‘high’ initial security rating compared to males (72.0% compared to 18%).^[2] In contrast, a minority of the sample had a ‘high’ risk rating upon release to HD (1.9%). The average RoR score for a sample of 254 prisoners for which data were available was 12.2 (sd=5.8,

range=1–20).^[3] Males scored significantly higher on the RoR compared to females (x=12.7, sd=5.6 compared to x=9.4, sd=6.5), and the effect size of this difference was moderate (d=0.53). In contrast, there were no differences in ORNI-R scores between males and females among the 113 prisoners for which ORNI-R assessment data were available. Finally, most of the sample (91.2%) had at some point been involved in employment programming in their custodial histories: 100 per cent of females compared to 89.5 per cent of males had prior involvement in these programs. Far fewer had prior involvement in education programming in custody (45.7%) and behavioural change programming (30.3%), and there were no statistical differences in the proportions of male or female involvement in these two types of programs.

Table 1: Demographic and criminal justice related characteristics of prisoners

Demographics	Total sample (n=317) %/x(sd)	Males (n=267) %/x(sd)	Females (n=50) %/x(sd)	X²(df), φ / t(df), Cohen’s d
Age at release to HD^log	37.5 (11.5)	37.3 (11.6)	38.4 (10.7)	n.s
Aboriginal or Torres Strait Islander^a	9.0%	8.2%	13.6%	n.s
≥ High school education^b	28.8%	27.0%	40.5%	X²(1)=2.9⁺, 0.10
Employed prior to most recent custody^c	38.2%	40.5%	20.7%	X²(1)=4.3^*, 0.13
Offence history
Multiple prior sentences (yes)	29.0%	30.3%	22.0%	n.s
Average number of prior sentences	1.8 (1.8)	1.9 (1.8)	1.5 (1.4)	n.s
Risk ratings
Initial security rating (high)	26.5%	18.0%	72.0%	X²(2)=63.1^***, 0.45
Final security rating (high)	1.9%	2.2%	0.0%	X²(2)=4.9⁺, 0.09^f
RoR Score^d	12.2 (5.8)	12.7 (5.6)	9.4 (6.5)	t(50.2)=-2.9^**, 0.53
ORNI-R Score^e	24.4 (5.1)	24.1 (4.7)	25.9 (7.3)	n.s
Prison programs (ever)
Employment	91.2%	89.5%	100.0%	X²(1)=5.8^*, 0.14
Education	45.7%	46.0%	45.7%	n.s
Behavioural change	30.3%	26.0%	31.1%	n.s
p<.10+, p<.05, p<.01, p<.001** Non-parametric bivariate comparison for average number of prior sentences: Mann-Whitney U. a. (n=301) d. (n=254) b. (n=267) e. (n=113) c. (n=251) f. low expected cell counts

Table 2 (below) displays sentencing information, HD characteristics, breaches of HD and returns to custody following HD. The types of offences for which the sample were incarcerated for at the time of data collection were: drug related offences (30.9%) (for example, drug trafficking, manufacturing, and one individual for possession of drugs); violent offences (21.8%) (for example, assaults, robbery); administrative offences (19.2%) (for example, offences against justice procedures, licence and registration offences); theft (11.7%)

(for example, theft, break and enter, receiving stolen goods); and fraud (11.4%) (for example, fraud and extortion). Here, males were more than twice as likely as females to have been incarcerated for a violent offence (24.3% compared to 8.0%), whereas females were more than three times more likely than males to have been incarcerated for a fraud related offence (28.0% compared to 8.2%).

Table 2: Sentence and HD characteristics, breaches of HD, and returns to custody by

Index offence (HD sentence)	Total sample (n=317) %/x(sd)	Males (n=267) %/x(sd)	Females (n=50) %/x(sd)	X²(df), φ / t(df), Cohen’s d
Drug	30.9%	31.1%	30.0%	n.s
Violent	21.8%	24.3%	8.0%	X²(1)=6.6^*, 0.14
Administrative/driving	19.2%	19.9%	16.0%	n.s
Theft	11.7%	10.9%	16.0%	n.s
Fraud	11.4%	8.2%	28.0%	X²(1)=16.3^***, 0.23
Public order/property	5.0%	5.6%	2.0%	n.s
Sentence characteristics
Non-parole period (days)^log	542.8 (483.5)	562.1 (492.9)	439.9 (418.8)	t(315)=-2.3^*, 0.34
Average length of HD Sentence (sentenced days)	148.4 (105.8)	149.8 (103.7)	141.3 (117.5)	n.s
Average length of HD Sentence (actual days)	143.2 (106.7)	144.4 (105.4)	134.7 (114.4)	n.s

HD breaches
Breached HD conditions	15.8%	17.2%	8.0%	n.s
Average time to breach of HD (days) (n=50)	103.7 (94.3)	103.4 (91.2)	107.5 (142.6)	n.s

Returns to custody (‘RTC’) post-HD
RTC (re-offence) by June 2017	19.6%	19.5%	20.0%	n.s
Average time to RTC (days) (n=62)	295.7 (172.9)	308.7 (168.0)	228.3 (191.5)	n.s
Administrative offence^a	82.3%	80.8%	90.0%	n.s
Non-violent (not including administrative) offences^b	22.2%	21.2%	30.0%	n.s
Violent offence^c	8.1%	7.7%	10.0%	n.s
p<.05, p<.001** a. Administrative offences include: offences against justice procedures. b. Offences include: theft, break and enter, fraud, offences against good order, and ‘other’. c. Offences include: sexual assault, assault, unlawful possession of a weapon.

The average length of the non-parole period attached to HD sentences was 542.8 (sd=483.5) days, and males had significantly longer non-parole periods compared to females (562.1 days compared to 439.9 days). The average number of sentenced HD days was 148.4 (sd=105.8) and there were no differences in the lengths of HD sentences received between males and females. Accounting for breaches/early terminations of HD sentences, the number of actual HD days served for the entire sample was 143.3 (sd=106.7) and again there were no differences between males and females.

In terms of breaches of HD, only 15.8 per cent of the sample breached a HD order. While this reflected 17.2 per cent of males and 8 per cent of females, this difference was not statistically significant. A slightly higher proportion of the sample (19.6%) returned to custody by June 2017 following their discharge from a HD sentence. Similarly, there were no differences in the proportion of males and females who returned to custody by June 2017. In terms of the types of offences for which individuals returned to custody, the vast majority were for new administrative offences (82.3%) followed by non-violent offences, and only 8.1 per cent of those individuals who returned to custody by June 2017 committed a violent offence. Again, there were no statistical differences between males and females in terms of the type of offences they committed that resulted in their return to custody.

Table 3 (below) displays bivariate comparisons between individuals who breached HD orders compared to those who did not. In total, out of 317 individuals who received a HD order between June 2014 and June 2015, 52 (16%) breached the conditions of the order. In terms of demographic characteristics, individuals who breached HD were significantly younger on average than those who did not breach their HD order (34.4 years old compared to 38.0 years old) and the effect size of this difference was low to approaching moderate (d=0.29). Individuals who did not breach their HD order were nearly twice as likely to have more than high school education compared to those who did (31.3% compared to 16.3%). There were no statistical differences in the demographic profile of individuals who breached their HD order compared to those who did not in terms of gender, whether they were Aboriginal or Torres Strait Islander, or whether they had been employed prior to their most recent custody episode. Similarly, there were no statistical differences between individuals who breached HD in terms of the number of prior sentences in their history.

There were no statistical differences between individuals who breached HD orders and those who did not in terms of initial security ratings they received upon entry into custody. However, a larger proportion of individuals who breached HD had received a final security rating (that is, upon release to HD) of high (8.0%) compared to those who did not (0.7%). However, these differences should be interpreted with caution due to the low base rate of individuals who received a ‘high’ final security rating (1.9%). In contrast, in the sample of 254 individuals for whom RoR assessment data were available, those who breached HD had a significantly higher RoR score (15.7, sd=3.3) compared to those who did not (11.4, sd=6.0) and the effect size was strong (d=0.90). For the subsample of individuals for whom ORNI-R assessment data were available, there were no statistical differences between ORNI-R scores of individuals who breached HD and those who did not.

There were no differences in participation in employment or education programs while in custody between individuals who breached HD orders and those who did not. However, half of individuals who breached HD orders had participated in behavioural change programs in custody at some point (50.0%) compared to only one-quarter of individuals who did not breach HD orders (26.6%).

	Total sample (n=317) %/x(sd)	No HD breach (n=267) %/x(sd)	Breached HD (n=50) %/x(sd)	X²(df), φ / t(df), Cohen’s d
Demographics
Male gender	84.2%	82.8%	92.0%	n.s
Age at release to HD^log	37.5 (11.5)	38.0 (12.0)	34.4 (7.8)	t(86.3)=2.1⁺, 0.29
Aboriginal or Torres Strait Islander^a	9.0%	8.8%	10.0%	n.s
≥ high school education^b	28.8%	31.3%	16.3%	X²(1)=3.9^*, 0.12
Employed prior to most recent custody^c	38.2%	40.5%	26.8%	n.s
Offence history
Multiple prior sentences (yes)	29.0%	27.7%	36.0%	n.s
Average number of prior sentences	1.8 (1.8)	1.8 (1.7)	2.0 (2.0)	n.s
Risk ratings
Initial security rating (high)	26.5%	25.5%	32.0%	n.s
Final security rating (high)	1.9%	0.7%	8.0%	X²(2)=4.9⁺, 0.09_f
RoR Score^d	12.2 (5.8)	11.4 (6.0)	15.7 (3.3)	t(120.8)=-6.8^***, 0.90
ORNI-R Score^e	24.4 (5.1)	24.4 (5.3)	24.2 (4.5)	n.s
Prison programs (ever)
Employment	91.2%	90.6%	94.0%	n.s
Education	45.7%	43.8%	56.0%	n.s
Behavioural change	30.3%	26.6%	50.0%	X²(1)=10.9^**, 0.19
Index offence (HD sentence)
Violent	21.8%	19.1%	36.0%	X²(1)=7.1^**, 0.19
Theft	11.7%	10.5%	18.0%	n.s
Drug	30.9%	32.2%	24.0%	n.s
Fraud	11.4%	12.7%	4.0%	X²(1)=3.2⁺, 0.10
Administrative/driving	19.2%	20.2%	14.0%	n.s
Public order/property	5.0%	5.2%	4.0%	n.s
NPP (days)^log	542.8 (483.5)	535.1 (506.3)	584.1 (336.9)	t(87.5)=-2.7^**, 0.37
Days Sentenced to HD (days)^sqrt	148.4 (105.8)	142.2 (105.9)	181.6 (100.1)	t(78.4)=-3.1^**, 0.45
p<.10+, p<.05, p<.01, p<.001** note. Non-parametric bivariate comparison for average number of prior sentences: Mann-Whitney U. a. (n=301) d. (n=254) b. (n=267) e. (n=113) c. (n=251) f. low expected cell counts

Finally, in terms of the offences for which individuals were incarcerated, those incarcerated for violent offences were almost twice as likely as those who were not to breach their HD order (36.0% compared to 19.1%). However, the reverse pattern was true in cases where individuals were incarcerated for fraud related offences: only 4 per cent of individuals who breached HD were in custody for fraud-related offences compared to 12.7 per cent of individuals who did not breach HD. While these two patterns were statistically significant, the effect sizes were relatively low (φ=0.19 and φ=0.10 respectively). There were no other differences between individuals who breached their HD order and those who did not in terms of whether they were incarcerated for theft, drug, or administrative/driving related offences. However, individuals who breached HD had significantly longer non-parole periods compared to those who did not breach HD orders (584.1 days compared to 535.1 days), as well as significantly longer HD order sentences (181.6 days compared to 142.2 days).

	B	SE(B)	Wald	Exp(b)	95%CI
Demographics
Age at HD	0.01	0.02	0.16	1.01	(0.97–1.05)
≥ high school education	-.68	0.50	1.87	0.51	(0.19–1.34)
Risk rating
Final security rating (med-high)	1.25	0.45	7.68	3.48**	(1.44–8.38)
RoR Score	0.19	0.06	9.59	1.21**	(1.07–1.36)
Index offence
Violent	0.44	0.48	0.87	1.56	(0.61–3.96)
Fraud	-0.39	1.11	0.12	0.68	(0.08–6.02)
NPP (days)	0.00	0.00	3.19	1.00	(1.00–1.00)
Prison programs
Behavioural change	0.03	0.47	0.00	1.03	(0.41–2.60)
Days sentenced to HD	0.01	0.00	14.46	1.01***	(1.01–1.02)
²(9) = 48.5, p < .001
Nagelkerke R² = .32
Cox & Snell R²= .20 % Correct class. = 83.0%

Given these bivariate differences, next logistic regression models were estimated to determine the relative contribution of covariates to the likelihood of an individual breaching their HD order. Table 4 (above) displays a logistic regression model predicting breaches of HD based on differences in the profiles identified in Table 3 (above). The results show that net of other factors, a final security rating (coded as 0=low, 1=medium/high) and RoR score, and the length of the HD sentence significantly predicted the likelihood of breaching HD. More specifically, individuals who received a final security rating of medium or high were 2.6 times more likely to breach HD than those who received a final security rating of low. In addition to this, for every unit increase in RoR score an individual was approximately 15 percent more likely to breach a HD order, and for every unit/day increase in the length of the HD sentence handed down there was a one percent increase in the likelihood of individuals breaching HD. These multivariate findings regarding risk level also explain why at the bivariate level, individuals who engaged in behavioural change programs were more likely to have breached HD. In effect, once risk level is controlled for, the association between participating in behavioural change programs and breaches of HD disappears.

Description: Individuals who returned to custody post-HD orders by June 2017

Table 5 (below) displays bivariate comparisons between individuals who returned to custody following the discharge date of their HD order compared to those who did not return to custody by June 2017. Out of 317 individuals who received a HD order between June 2014 and June 2015, 62 returned to custody by June 2017. Similar to individuals who breached HD, those who returned to custody were younger on average (34.0 years old, sd=12.0) compared to those who did not (38.9 years old, sd=12.0) and the effect size of this difference was moderate (d=0.44). Aboriginal or Torres Strait Islanders were over-represented in terms of returns to custody: 14.8 per cent of individuals who returned to custody were Aboriginal or Torres Strait Islander compared to 7.5 per cent of those who did not return to custody. The effect size of this difference was in the low range (φ=0.10). There were no other differences in terms of demographic characteristics (gender, level of education, employment prior to previous custodial episode) between individuals who returned to custody and those who did not. However, more than twice as many individuals who returned to custody had multiple prior sentences (53.2%) compared to those who did not (23.1%) and the effect size of this difference was low approaching moderate.

There were no statistical differences between individuals who returned to custody and those who did not in terms of initial security ratings they received upon entry into custody. However, similar to those who breached HD orders, a larger proportion of individuals who returned to custody had received a final security rating (that is, upon release to HD) of high (3.2%) compared to those who did not (1.6%). Again, these differences should be interpreted with caution due to the low base rate of individuals who received a ‘high’ final security rating (1.9%). In contrast, in the sample of 254 individuals for whom RoR assessment data were available, those who returned to custody had a significantly higher RoR score (15.8, sd=3.6) compared to those who did not (11.2, sd=5.9) and again the effect size was strong (d=0.93). In addition, for the subsample of individuals for whom ORNI-R assessment data were available, the average score of those who returned to custody was 26.7 (sd=3.3) compared to an average of 23.3 (sd=5.4) for those who did not and this association was also strong (d=0.75).

	Total sample (n=317) %/x(sd)	No RTC by 2017 (n=255) %/x(sd)	RTC by 2017 (n=62) %/x(sd)	X²(df), φ / t(df), Cohen’s d
Demographics
Male gender	84.2%	84.3%	83.9%	n.s
Age discharge from HD^log	37.9 (11.6)	38.9 (12.0)	34.0 (12.0)	t(315)=2.9^**, 0.44
Aboriginal or Torres Strait Islander^a	9.0%	7.5%	14.8%	X²(1)=3.1⁺, 0.10
≥ high school education^b	28.8%	30.0%	24.1%	n.s
Employed prior to most recent custody^c	38.2%	39.7%	32.7%	n.s
Offence history
Multiple prior sentences (yes)	29.0%	23.1%	53.2%	X²(1)=21.9^***, 0.26
Average number of prior sentences	1.8 (1.8)	1.5 (1.3)	2.9 (2.8)	U=5281.0***
Risk ratings
Initial security rating (high)	26.5%	24.7%	33.9%	n.s
Final security rating (high)	1.9%	1.6%	3.2%	X²(1)=6.3^*, 0.14^f
RoR Score^d	12.2 (5.8)	11.2 (5.9)	15.8 (3.6)	t(132.9)=-7.1^***, 0.93
ORNI-R Score^e	24.4 (5.1)	23.3 (5.4)	26.7 (3.3)	t(99.8)=-4.1^***, 0.75
Prison programs (ever)
Employment	91.2%	92.2%	87.1%	n.s
Education	45.7%	45.5%	46.8%	n.s
Behavioural change	30.3%	26.3%	46.8%	X²(1)=9.9^**, 0.18
Index offence (prior HD sentence)
Violent	21.8%	23.9%	12.9%	X²(1)=3.6⁺, 0.11
Theft	11.7%	9.4%	21.0%	X²(1)=6.5^*, 0.14
Drug	30.9%	32.9%	22.6%	n.s
Fraud	11.4%	12.2%	8.1%	n.s
Administrative/driving	19.2%	16.5%	30.6%	X²(1)=6.4^*, 0.14
Public order/property	5.0%	5.1%	4.8%	n.s
NPP (days/prior HD-related sentence)^log	542.8 (483.5)	567.1 (517.9)	442.9 (285.0)	n.s
Breached prior HD order	15.8%	11.8%	32.3%	X²(1)=15.8^***, 0.22
p<.10+, p<.05, p<.01, p<.001** note. Non-parametric bivariate comparison for average number of prior sentences: Mann-Whittney U. a. (n=301) d. (n=254) b. (n=267) e. (n=113) c. (n=251) f. low expected cell counts

A similar pattern to HD breaches also emerged in terms of custodial program involvement: nearly half of individuals who returned to custody had prior involvement in behavioural change programs (46.8%) compared to just over one-quarter of those who did not return to custody (26.3%). No differences were evident between individuals who returned to custody and those who did not in terms of prior participation in employment or education programs.

Compared to individuals who breached their HD orders, a slightly different profile emerged for individuals who returned to custody in terms of their prior offences. First, individuals who returned to custody were less likely to have previously committed a violent offence (12.9%) compared to those who did not return to custody (23.9%), although this relationship was statistically marginal and the effect size was low (φ=0.11). In contrast, more than twice as many individuals who returned to custody had previous theft offences (21.0%) compared to those who did not return to custody (9.4%). Similarly, twice as many individuals who returned to custody had previous administrative/driving offences compared to those who did not return to custody (16.5%). Finally, individuals who breached their prior HD order (32.3%) were nearly three times more likely to return to custody than those who did not breach their prior HD order (11.8%).

	B	SE(B)	Wald	Exp(b)	95%CI
Demographics
Age at discharge from HD	-0.04	0.02	2.17	0.97	0.92–1.01
Aboriginal or Torres Strait Islander	0.31	0.62	0.25	1.37	0.40–4.65
Number of prior sentences	0.40	0.11	14.38	1.50***	1.22–1.85
Risk rating
Final security rating (med-high)	0.51	0.41	1.52	1.66	0.74–3.69
RoR Score	0.11	0.06	3.38	1.11⁺	0.99–1.24
Prison programs
Behavioural change	-0.05	0.43	0.01	0.95	0.41–2.20
Index offence (prior HD sentence)
Violent	-1.58	0.61	6.85	0.21**	0.06–0.67
Theft	-0.62	0.60	1.06	0.54	0.17–1.75
Administrative/driving	0.62	0.49	1.64	1.87	0.72–4.85
Breached prior HD order	1.01	0.43	5.55	2.75*	1.19–6.40
2(10) = 62.08, p < .001
Nagelkerke R2 = .35
Cox & Snell R2 = .23 % Correct class. = 78.5%

Given the bivariate differences between individuals who returned to custody and those who did not, Table 6 (above) displays a logistic regression model of the variables associated with returns to custody. A different profile for returns to custody emerged than variables predicting breaches of HD orders. Given that ORNI-R data were available for less than half of the sample, it was not possible to include this variable in the models. The number of prior sentences, RoR score, a prior non-violent offence, and breaching the previous HD^[4] order were associated with a higher likelihood of returning to custody post-HD. More specifically, each prior sentence in an individual’s history was associated with a 50 per cent increase in the likelihood of a return to custody. RoR score was marginally associated with returns to custody net of other factors. In terms of the prior offence, those individuals who committed a violent offence were 79 per cent less likely to return to custody compared to those individuals who did not commit a violent offence. Finally, individuals who breached their previous HD order were 2.75 time more likely to return to custody than those who did not breach their prior HD order.

A reduced form logistic regression model is presented in Table 7 (below). The results are consistent with those in Table 5 (above), when removing the non-significant variables from the equation RoR also predicts return to custody; each unit increase in RoR score results in a 19% increase of the likelihood of a return to custody.

	B	SE(B)	Wald	Exp(b)	95%CI
Number of prior sentences	0.31	0.09	11.83	1.36**	1.14–1.62
RoR Score	0.17	0.05	14.11	1.11***	1.09–1.30
Index offence = violence (prior HD sentence)	-1.50	0.51	8.56	0.22**	0.08–0.61
Breached prior HD order	0.95	0.41	5.46	2.57*	1.17–5.69
²(4) = 57.8, p < .001
Nagelkerke R² = .32
Cox & Snell R²= .20 % Correct class. = 79.1%

The findings from the current study provide a statistical profile of a cohort of 317 prisoners sentenced to HD between June 2014 and June 2015 in SA. The proportion of males compared to females serving HD orders in 2014–15 reflects the over-representation of males in crime and the proportion of male and female involvement in the justice system more broadly (approximately 75–80% males compared to 20–25% females). Aboriginal and Torres Strait Islander individuals were slightly over-represented in HD relative to the proportion of the general population they make up, but far lower than the proportion of Aboriginal and Torres Strait Islander individuals involved in the justice system more broadly.

In total, less than one-fifth of all prisoners sentenced to HD in this time frame breached the conditions of their HD order. Only a slightly higher proportion of individuals (approximately one-fifth) returned to custody at some point by June 2017 following the conclusion of their HD order. The latter figure falls far below overall recidivism rates of released prisoners returning to custody in South Australia (36.2%) and across all Australian jurisdictions (36.2% to 57.1%) over the same time period (Australian Productivity Commission, 2018). Less than ten percent of returns to custody (that is, five individuals) were for a violent offence. This contrasts with the fact that approximately one-fifth of the cohorts’ HD orders in 2014–15 (21.8%) were associated with a violent index offence. In fact, the multivariate analyses in the current study indicated that those individuals whose HD sentence was associated with a violent offence were less likely to return to custody by 2017 compared to those individuals whose HD order was associated with a non-violent or administrative offence.

Prior studies have shown that EM and HD can be implemented effectively even with

high-risk offenders (see, for example, Giess et al. 2012; Giess et al. 2013). In the current study, comparable ‘high-risk’ offenders were not eligible for HD (for example, sex offenders). In effect, for those who received a HD sentence for a violent offence, such as an assault, the offence likely was not part of a broader pattern of violence in their history. To add to this, the length of non-parole period was not associated with breaches of HD or returns to custody by 2017 net of other factors considered. This points to another important pattern — the vast majority of offences for which individuals returned to custody post-HD order were administrative offences (that is, offences against justice procedures). In other words, while the prevention of serious and violent offences committed by individuals on HD orders should be a paramount concern for policymakers and those working in the criminal justice system, in the current cohort they would constitute rare exceptional cases.

This then raises the question of what are the key factors associated with non-compliance of HD orders, be it breaches of HD conditions or reoffending? The length of the HD sentence was a significant predictor (in addition to a high security rating and RoR score) of breaches of HD and for those who breached, the average amount of time from HD sentence commencement to breach for those who did was just over three months. This suggests that, not surprisingly, the longer someone is on HD the more likely they are to breach. When breaches did occur, they typically happened in the latter portions of HD sentences.

Given the low base-rate of violent recidivism, it was not possible to look at this outcome specifically. Nonetheless, different profiles emerged pertaining to individuals who breached conditions of their HD order and those that returned to custody following a HD order. In short, the key predictors of HD breaches were the length of the HD sentence, custodial security ratings, and risk assessment scores. This is in line with prior studies that have found risk scores are important determinants of eligibility for HD (for example, Dodgson et al. 2001), in this case even in ‘lower risk’ samples (that is, those eligible for HD in SA). Net of the other factors considered (that is, demographic characteristics, index offence associated with the HD sentence, programs in custody), the results show that the practice of risk assessment in the current context provides knowledge about the likelihood of who succeeds (or not) while on HD orders. Final security custody ratings were associated with breaches of HD in addition to risk assessment scores. This suggests that, at least to some extent, security ratings in custody that are administered prior to release into the community provide some information about whether or not individuals are likely to breach conditions of their HD order.

In terms of returns to custody for new offences following the completion of a HD order, a slightly different profile emerged. The number of prior sentences, risk assessment score, a prior non-violent index offence, and prior HD breach were all associated with returns to custody. Again, the majority of returns to custody were for administrative offences, and it is possible that some individuals may have been returned to custody for breaching their HD order. These findings parallel those of Marie et al. (2011) and others, who found that most of their sample on HD who were returned to custody had committed administrative offences

(for example, breaches). However, the profile stands that returns to custody were not predicted by violent index offences.

Taken together, these findings suggest that, rather than the nature of the index offence for which an individual received a HD sentence, chronicity or, in other words, the frequency of prior offending (here measured as number of prior sentences) is a more robust predictor of reoffending, in addition to risk assessment scores. Given that most returns to custody following HD orders were for administrative offences, this may also be indicative of difficulties faced by ex-prisoners meeting on-going requirements associated with community orders. The findings from Marklund and Holmberg’s (2009) Swedish study suggest that monitoring and support (for example, stable employment) while on HD likely have a positive impact on success. Padgett et al. (2006) showed that using EM with HD also had a positive impact on reoffending rates. In effect, the use of EM technology in addition to other support mechanisms can help ex-prisoners meeting the requirements of HD and reduce the likelihood of reoffending. While this is a likely possibility, this relationship could not be assessed directly in the current study.

In Australia, there has been some reluctance to adopt HD programs compared to other Commonwealth and European jurisdictions (Bartels & Martinovic 2017). The findings from the current study provide the first baseline evidence of the nature and extent of reoffending of individuals sentenced to HD orders in SA, as well as factors associated with breaches of HD orders and returns to custody. The findings suggest that violent reoffending by individuals on HD was rare, and that security and risk assessment tools utilised by corrective services were among the best predictors of who breached HD orders as well as who returned to custody, based on the variables available for analysis. Furthermore, the results also showed that the vast majority of returns to custody were for administrative offences — specifically, breaches of justice orders — suggesting the possibility that ex-prisoners face challenges meeting requirements placed on them by the justice system while in the community. Future research needs to explore this critical question more closely.

Importantly, this study suffered from methodological limitations. First, it is based on a relatively small sample of prisoners which limited the scope of analysis. Nonetheless, it also reflected the population of prisoners sentenced to HD in 2014–15 in the state of SA. Second, the scope of independent variables was limited, and missing data characterised some of the demographic and risk assessment variables, which in addition to the small sample size limited the power of the statistical analysis conducted. Despite these limitations, the findings provide some insights into the profile of individuals sentenced to HD in SA, and factors associated with breaches and reoffending of prisoners sentenced to HD orders. Future studies examining the impact of HD on reoffending in Australia should include matched control/comparison groups where possible, as well as details pertaining to supervision on HD orders to allow for tests of potential net-widening effects of the sanction.

Australian Productivity Commission 2018, Report on Government Services, Part C, Table CA.4, viewed 21 July 2018, http://www.pc.gov.au/research/ongoing/report-on-government-services/2018/justice#

Ball, RA, Huff, CR & Lilly, JR 1988, House arrest and correctional policy: doing time at home, Sage Publications, Newbury Park.

Ball, RA & Lilly, JR 1986, ‘A theoretical examination of home incarceration’, Federal Probation, vol. 50, pp. 17–24.

Bartels, L & Martinovic, M 2017, ‘Electronic monitoring: the experience in Australia’, European Journal of Probation, vol. 9, no. 1, pp. 80–102.

Beck, JL & Klein-Saffran, J 1989, Community control project: interim report, United States Parole Commission, Washington.

Black, M & Smith, RG 2003, Electronic monitoring in the criminal justice system, Australian Institute of Criminology, Canberra.

Byrne, JM, Lurigio, AJ & Petersilia, J 1992, Smart sentencing: the emergence of intermediate sanctions, Sage, Thousand Oaks.

Clear, T & Braga AA 1995, ‘Community corrections’ in JQ Wilson & J Petersilia (eds.), Crime:

twenty-eight leading experts look at the most pressing problem of our time, Institute for Contemporary Studies, San Francisco.

Dodgson, K, Goodwin, P, Howard, P, Llewellyn-Thomas, S, Mortimer, E, Russell, N & Weiner, M 2001, Electronic monitoring of released prisoners: an evaluation of the Home Detention Curfew scheme, Home Office Research, Development and Statistics Directorate, London.

Dodgson, K, Mortimer, E & Sugg, D 2000, Assessing prisoners for Home Detention Curfew: a guide for practitioners, Home Office Research, Development and Statistics Directorate, London.

Doherty, D 1995, ‘Impressions of the impact of the electronic monitoring program on the family’ in K Schulz (ed.), Electronic monitoring and corrections: the policy, the operation, the research, Simon Fraser University, Canada, pp. 129–39.

Finn, MA & Muirhead-Steves, S 2002, ‘The effectiveness of electronic monitoring with violent male parolees’, Justice Quarterly, vol. 19, no. 2, pp. 293–312.

Fox, RG 1987, ‘Dr Schwitzgebel’s machine revisited: electronic monitoring of offenders’, Australian & New Zealand Journal of Criminology, vol. 20, no. 3, pp. 131–47.

Geis, SV, Gainey, R, Cohen, MI, Healy, E, Duplantier, D, Yeide, M, Bekelman, A, Bobnis, A & Hopps, M 2012, Monitoring high-risk sex offenders with GPS technology: an evaluation of the California Supervision Program final report, Development Services Group Inc, Bethesda.

Geis, SV, Gainey, R, Cohen, MI, Healy, E, Yeide, M, Bekelman, A & Bobnis, A 2013, Monitoring

high-risk sex offenders with GPS technology: an evaluation of the California Supervision Program final report, Development Services Group Inc, Bethesda.

Gibbs, A & King, D 2003, ‘The electronic ball and chain? The operation and impact of home detention with electronic monitoring in New Zealand’, Australian & New Zealand Journal of Criminology, vol. 36, no. 1, pp. 1–17.

Grogger, J 1995, ‘The effect of arrests on the employment and earnings of young men’, The Quarterly Journal of Economics, vol. 110, no. 1, pp. 51–71.

Hilferty, F, Lafferty, L, Valentine, K, Cale, J & Zmudski, F 2018, ‘Changes to home detention in South Australia: evaluation design and early implementation findings’, Advancing Corrections, vol. 5, pp. 31–44.

Joutsen, M & Zvekic, U 1994, ‘Noncustodial sanctions: comparative overview’ in U Zvekic (ed.), Alternatives to imprisonment in comparative perspective, Nelson-Hall Publishers, United States.

LeBel, TP 2012, ‘Invisible stripes? Formerly incarcerated persons’ perceptions of stigma’, Deviant Behavior, vol. 33, no. 2, pp. 89–107.

Lopoo, LM & Western, B 2005, ‘Incarceration and the formation and stability of marital unions’, Journal of Marriage and Family, vol. 67 no. 3, pp. 721–34.

Marie, O, Moreton, K & Gonclaves, M 2011, The effect of early release of prisoners on home detention curfew (HDC) on recidivism, Ministry of Justice, London.

Marklund, F & Holmberg, S 2009, ‘Effects of early release from prison using electronic tagging in Sweden’, Journal of Experimental Criminology, vol. 5, no. 1, pp. 41–61.

Martinovic, M 2007, ‘Home detention: issues, dilemmas and impacts for detainees’ co-residing family members’, Current Issues Criminal Justice, vol. 19, no. 1, pp. 90–105.

Martinovic, M 2013, The evolution of home detention based sanctions frameworks in the USA and Australia up to 2013: a comparative case study, RMIT University, Melbourne.

Martinson, R 1974, ‘What works? Questions and answers about prison reform’, The Public Interest, vol. 35, pp. 22–55.

Morris, N & Tonry, M 1990, Between prison and probation: intermediate punishments in a rational sentencing system, Oxford University Press, New York.

Nellis, M 2013, ‘Surveillance, stigma and special constraint: the ethical challenges of electronic monitoring’ in M Nellis, K Beyens & D Kaminski (eds.), Electronically monitored punishment: international and critical perspectives, Routledge, London.

New South Wales Law Reform Commission 2013, Sentencing, viewed 15 July 2018, http://www.lawreform.justice.nsw.gov.au/Documents/Publications/Reports/Report-139.pdf.

New South Wales Government State Parole Authority 2015, Home detention orders, viewed 15 July 2018, http://www.paroleauthority.nsw.gov.au/Pages/Function/home-detention.aspx.

Northern Territory Government 2017, Home detention, viewed 15 July 2017, https://nt.gov.au/law/

Padgett, KG, Bales, WD & Blomberg, TG 2006, ‘Under surveillance: an empirical test of the effectiveness and consequences of electronic monitoring’, Criminology and Public Policy, vol. 5, no. 1, pp. 61–91.

Pager, D 2003, ‘The mark of a criminal record’, American Journal of Sociology, vol. 108, no. 5, pp. 937–75.

Palmer, D, De Lint, W & Dalton, D 2015, Crime and justice: a guide to criminology, 5th edn., Thomson Reuters, Pyrmont.

Renzema, M 2003, Electronic monitorings impact on reoffending, Campbell Collaboration, Kutztown.

Renzema, M & Mayo-Wilson, E 2005, ‘Can electronic monitoring reduce crime for moderate high risk offenders?’, Journal of Experimental Criminology, vol. 1, pp. 215–37.

Roberts, JV 2004, The virtual prison: community custody and the evolution of imprisonment, Cambridge University Press, Cambridge.

Robinson, G, McNeill, F & Maruna, S 2013, ‘Punishment in society: the improbable persistence of probation and other community santions and measures’ in J Simon and R Sparks (eds.), Handbook of Punishment and Society, Sage, London.

United States Sentencing Commission 2013, United States sentencing commission guidelines manual, Government Printing Office, Washington.

Vanhaelemeesch, D, Beken, TV & Vandevelde, S 2014, ‘Punishment at home: offenders’ experiences with electronic monitoring’, European Journal of Criminology, vol. 11, no. 3, pp. 273–87.

Worrall, A 2006, ‘Community sentences’ in E McLaughlin & J Muncie (eds.), The sage dictionary of criminology, 2nd edn., Sage, London.

^* Senior Lecturer in Criminology, School of Social Sciences, UNSW, The University of New South Wales NSW 2052, Australia. Email: j.cale@unsw.edu.au.

^† PhD Candidate, Faculty of Law, UNSW, The University of New South Wales NSW 2052, Australia. Email: melanie.burton@unsw.edu.au.

^[1] ORNI-R scores are typically assessed only for those individuals who receive high RoR scores.

^[2] While these differences are statistically significant, they are due, at least in part, to DCS procedural differences in how females and males are processed upon entry in custodial facilities.

^[4] It was not possible to determine with absolute certainty if an individual returned to custody as a result of breaching their HD order. Of all individuals who breached a HD order, 32.3% also returned to custody. It is possible that some of these individuals returned to custody because of breaching their HD order, which is also why this is included as a covariate in the current model.