Does electronic monitoring as a means of release preparation reduce subsequent recidivism? A randomized controlled trial in Germany

Abstract

This study analyses the recidivism-reducing effect of electronic monitoring (EM) in the context of early work release and home detention as a means of release preparation. We tested the hypothesis that EM reduces recidivism after the termination of EM. The results are based on a randomized controlled trial in Baden-Württemberg, Germany, where a pilot project between 2010 and 2012 enabled different forms of EM. The participating prisoners were randomly assigned either to the experimental group under EM or to the control group, whose participants had to continue their regular sentence behind prison walls. Qualitative data and data of a matched-pair sample complement the analyses. There was no statistical significant difference between the recidivism rates of the EM and control group subjects.

Keywords

Electronic monitoring GPS randomized controlled trial recidivism rehabilitation reoffending

Introduction

In contrast to the United States and parts of Europe such as England and Wales, Germany has taken a cautious approach to the use of electronic monitoring (EM). Of the 16 German states, only Hesse regularly uses EM, in order to avoid (re)imprisonment in the contexts of pre-trial detention and probation (Mayer, 2004). At the national level, EM has been available since 2011 for use on high-risk sex and violent offenders after their release from prison; recently, the group of persons subject to EM was expanded to include convicted (or suspected) terrorists.¹ Although acknowledging the misgivings and uncertainties associated with EM, the state of Baden-Württemberg passed the ‘Act on Electronic Monitoring during the Enforcement of Imprisonment’ in 2009 in order to scientifically evaluate EM as an innovative method. The goals of EM as foreseen in the law were twofold: (1) avoiding the incarceration of low-risk offenders who were unable to pay a fine and (2) facilitating preparation for release from prison (that is, home detention and early work release). The law, which was designed as a pilot project, called for the undertaking of a study – including a randomized controlled trial (RCT) – to evaluate the programme. Conducted by the Criminology Department of the Max Planck Institute for Foreign and International Criminal Law, the study consisted of two phases: the first phase examined whether EM was a viable tool for avoiding imprisonment and facilitating preparations for release and thus rehabilitation (see Schwedler and Woessner, 2017); the second phase took the form of a study of recidivism. The results of this second phase are presented here. The Baden-Württemberg EM law, originally enacted for a period of four years, expired in August 2013. Owing to a change in government and unfulfilled expectations concerning the efficacy of EM, it was not extended (Schwedler and Wößner, 2015).

The monitoring of offenders with electronic devices, which can be used to pursue various goals, was originally introduced in the United States to counteract prison overcrowding (Hucklesby, 2008; Renzema and Mayo-Wilson, 2005). Operating under the premise that EM is less punitive than imprisonment, policymakers soon introduced EM as an intermediate measure. Thus, EM became an inherent part of the sanctioning systems of many countries. Until now, however, there has been a dearth of empirical evidence on the rehabilitative potential of EM (Renzema and Mayo-Wilson, 2005; Vanhaelemeesch et al., 2014; Yeh, 2010). Critics emphasize that EM is generally used as a method of control rather than as a rehabilitative measure (Kornhauser and Laster, 2014; Renzema, 2003). In addition, owing to the plurality of available EM technologies (GPS, radio frequency and alcohol monitoring using tracking devices with transdermal alcohol sensors) and the existence of various application schemes that focus on persons of differing risk levels, it is difficult to draw general conclusions about the effect of EM. Moreover, most of the research on EM has focused on its implementation (Killias et al., 2010; Nellis, 1991) whereas studies on the crime preventive and rehabilitative potential of EM have been quasi-experimental. Likewise, EM is usually compared with conventional sanctions – if it is compared at all. Lastly, a reduction in recidivism may be caused by more than one factor and not necessarily just EM. Given these uncertainties, it is difficult to arrive at conclusive findings about EM. A major shortcoming of previous research on EM is the lack of a theoretical foundation. The study here is one of the few to incorporate empirical analysis of recidivism after EM into a theoretical framework. It tests the hypothesis that EM has a recidivism-reducing impact after its use is terminated. As mentioned above, the study is based on data from the Baden-Württemberg pilot project. It must be noted, however, that most of the participants in the pilot project were low- to medium-risk offenders; thus, the hypothesis must be narrowed to the research question of whether early release EM reduces recidivism in low- to medium-risk offenders. These findings on GPS-assisted EM can help us better understand the potential of EM as a preventive measure as well as its limitations.

Theoretical background

The rationales for why EM might prevent reoffending are manifold, as are the findings of past and current studies on EM and the likelihood of recidivism.

A widespread assumption is that continuous monitoring stimulates the feeling of a higher detection risk, thus referring to Rational Choice Theory. This could lead to a deterrent effect (Cornish and Clarke, 1986) that inhibits the offender from committing new offences on the basis of cost–benefit considerations. According to the logic of Routine Activity Theory (Felson and Clarke, 1998), EM helps the offender maintain a daily structure, thus minimizing opportunities to get involved in crime. Building on the idea of minimizing crime opportunities and according to Incapacitation Theory (Zimring and Hawkins, 1995: 44), EM keeps the offender away from criminogenic influences and places by forcing him/her to stay at home. Although these mechanisms mainly refer to the period of EM, a long-term effect of leading a more structured lifestyle might be measurable after EM has ceased. Moreover, some proponents assume a fairly long-term rehabilitative effect associated with EM resulting from the gradual internalization of the external control function of EM (Bottos, 2007). However, to date there is no empirical evidence that EM alone influences a habit or personality. This hoped-for effect can only be expected – if at all – if there are additional measures of reintegration and rehabilitation. In fact, the Council of Europe’s recommendation on EM² stresses the need to include EM in a comprehensive concept of rehabilitation. Within this context, some authors claim that EM fosters the supervised person’s compliance to participate in rehabilitative measures (Hucklesby, 2009). Moreover, EM – it is claimed – enables a more favourable release situation. According to social control theories, supportive social relations and a workplace are important factors regarding rehabilitation and reintegration (Hirschi, 1986). Thus, the use of EM to replace or shorten a prison sentence would help a convicted person maintain or restore these social ties, thus attenuating negative effects of imprisonment (Fallesen and Andersen, 2017). Especially for low-risk offenders, prison is perceived as a highly criminogenic environment where criminal behaviour can literally be learned due to the influences of the criminogenic subculture (Clemmer, 1958; Goffman, 1981). The avoidance or shortening of the prison sentence may therefore reduce a prisonization effect (Yeh, 2010). However, it should be noted that an electronic ankle bracelet might also have a stigmatizing effect. For instance, its visibility on the ankle could compromise access to the job market (Nellis, 2015; Renzema, 2003).

Research on whether EM reduces recidivism is barely guided by theoretical considerations. This might well be because it is difficult to identify an EM-inherent recidivism-reducing effect other than a control or managerial approach to offenders. In 1990, Petersilia and Turner conducted one of the few RCTs on EM in three Californian counties. They compared the recidivism rates of offenders on probation with and without EM during a one-year follow-up period. Their study yielded no significant EM-based effect on recidivism rates. However, several other risk factors had an impact on the risk of reoffending (for example, drug abuse). Meta-analyses by Renzema and Mayo-Wilson (2005) and MacKenzie (2006) substantiated the finding that EM does not reduce recidivism. However, a more recent quasi-experimental Norwegian study (Andersen and Telle, 2016) yielded findings of a recidivism-reducing effect of EM. The authors attribute this positive effect to the fact that offenders were able to maintain their links to the labour market ‘while serving their sentence’ (Andersen and Telle, 2016: 38). They did not, however, find any support for other theoretical mechanisms such as deterrence or situational/strain theory. Fallesen and Andersen (2017) also concluded that EM enables the offender to live in his/her social environment, which could prevent the breakdown of social bonds and reduce the stigmatization associated with incarceration. Other studies yielded unambiguous results on the recidivism-reducing effect of EM (Henneguelle et al., 2016; Marklund and Holmberg, 2009). Similar to Andersen and colleagues, Henneguelle et al. (2016) see the preventive impact of EM especially in the possibility of avoiding or shortening incarceration and of preventing deprivation and prisonization. However, these studies are based on quasi-experimental study designs. Thus, the possibility that the claimed recidivism-reducing effects are associated with pre-existing differences between the compared groups cannot be excluded. A study based on an RCT conducted in Switzerland (Killias et al., 2010) with low-risk offenders found a significant impact of EM on reoffending.

Most of the recidivism research associated with EM focusing on high-risk offenders is concentrated on the measurement of reoffending while still under EM, mostly after prison release. Padgett et al. (2006) addressed the effectiveness of EM during its use and claimed that EM reduces recidivism for serious offenders, regardless of the type of offence. Offenders with the highest risk of reoffending were, at the same time, under the highest level of surveillance within the group of serious offenders. Gies et al. (2012) also stated that GPS-based EM combined with traditional parole lowers the risk of recidivism for high-risk offenders. They trace the outcome to the deterrent effect of EM, particularly rational calculation and the increased risk of apprehension because of real-time surveillance. Therefore, the study concentrated, as so often, on recidivism and violations during EM. Likewise, Bales et al. (2010: 148) found evidence for the compliance-fostering effects of EM with regard to terms and conditions of supervision ‘resulting in lower levels of absconding, violations of court imposed conditions of supervision, and re-offending’. Henneguelle et al. (2016) also assume a deterrent effect for low-risk offenders. However, the meta-analysis conducted by Renzema and Mayo-Wilson (2005) – which included 154 EM studies (of which only three fulfilled their quality criteria) – showed that EM has no impact on recidivism for high-risk offenders. In addition, several authors (for example, Armstrong and Freeman, 2011; Finn and Muirhead-Steves, 2002; Nellis, 2015) have questioned a deterrent effect even for high-risk offenders under EM. Thus, each study must be interpreted in light of the risk level of the target group and the application scheme. The aim of the study presented in this paper is to contribute to the understanding of the effectiveness of EM with regard to its recidivism-reducing impact following the termination of EM.

Method

Background of the trial (pilot project on EM in Baden-Württemberg)

The pilot project on EM in Baden-Württemberg ran from October 2010 until March 2012. The law that called for the study provided for the use of EM in three areas: (1) home detention for up to six months prior to regular prison release (including conditional early release after serving one-half or two-thirds of a sentence); (2) home detention instead of imprisonment if a fine could not be paid; and (3) early work release in the context of day parole. In all three cases, the offender spent a certain amount of his prison sentence outside the correctional facility.³ The prisoners under supervised early work release were referred to halfway houses (open prison). Offenders placed on home detention had to stay at home except for approved activities, for example work, treatment or meetings with their probation officers. Additional prerequisites for home detention included a permanent residence and the consent of other persons living in the residence. Furthermore, in order to participate in any of the three areas of application, an offender had to have either a job or an appropriate alternative to work for at least 20 hours a week. The GPS-operated monitoring device allowed for the definition of inclusion and exclusion zones. In theory, it facilitated the exact monitoring of an offender’s whereabouts around the clock except for movements inside the offender’s home or halfway house.⁴ All offenders who agreed to participate in the study had to be able to cope with the burdens of EM. In addition, only those offenders who could be expected to refrain from exploiting the advantages of the programme (that is, to escape or to commit additional offences) were eligible to participate.

As EM is still an under-researched topic and its benefits are anything but undisputed, the state law also stipulated the evaluation of the pilot project within the framework of an RCT. Eligible prison inmates were identified by prison staff and reported to the research institute. The institute then randomly assigned them either to the experimental group (electronically monitored subjects) or to the control group (subjects served the remainder of their sentence in prison). The evaluation was conducted in two phases: the first phase consisted of an analysis of whether EM exhibited the hoped-for outcomes with regard to avoiding or shortening imprisonment and to improving the release situation (for the results, see Schwedler and Wößner, 2015; Schwedler and Woessner, 2017); the second phase, from 2015 to 2017, consisted of the completion of a recidivism study, the results of which are presented in this article.

Procedure

Five prisons in the federal state of Baden-Württemberg participated in the pilot project. The prison staff approached eligible inmates and informed them about the project and the procedure. Inmates who volunteered to participate were reported to the indepen-dent research team and randomly assigned to one of the two trial groups: the experimental group under EM or the control group (those who stayed in the regular prison). Unfortunately, a high number of subjects originally assigned to one of the three EM schemes dropped out of the trial. The dropout rate was 37.8 percent in total for the experimental group, mainly owing to the strict legal requirements for participation. Additionally, subjects dropped out because of interim changes (for example, released before project start, violations of prison rules that made them ineligible) or withdrawal of consent.⁵ Because of the low subject-intake for home detention due to inability to pay a fine (three subjects), this target group proved to be not suitable for this EM scheme and hence was excluded from the analysis. Consequently, the findings concentrate on the two remaining areas of application (home detention and early work release). To compensate for the drop-outs in the experimental group, the block randomization method with an adaptive allocation rate was applied. One-fifth of the subjects who were randomly assigned to the control group ultimately refused to participate in the scientific investigations.

In light of the aforementioned drop-outs following assignment to the experimental or control group, a decision was made to create a second control group in the two remaining areas of application using one-to-one matching; that is, for each subject in the experimental group we recruited a matched subject from the Baden-Württemberg prison population (matching variables: German versus non-German citizenship, migration background, age at the time of the offence, previous convictions, type of offence and sentence). With the inclusion of this second control group as an addition to the RCT, it was possible to take into account potential selection biases (for example, effects of motivational issues) or potential violations of the ‘stable unit treatment value assumption’ (SUTVA, Sampson, 2010). Figure 1 gives an overview of the groups per application field.

Figure 1.

Overview of sample groups.

Sample

A total of N = 92 offenders participated in the pilot project. The recidivism analysis included these participants in the RCT and the matched-pair sample. Thus, a total of N = 137 subjects were analysed. All subjects were male. Eighty percent of the subjects had German citizenship. The majority of the overall sample had committed property crimes (57 percent, n = 79) followed by traffic offences (12 percent, n = 17). Nearly all of the offenders had a job and a permanent residence at the time of release. The ‘Level of Service Inventory – Revised: Screening Version’ (LSI-R:SV; Andrews and Bonta, 1998) was used to assess the subjects’ risk level. Almost two-thirds of the subjects were classified as low-risk offenders (medium-risk: one-third; no high-risk offenders). No differences were observed with regard to the two areas of application (home detention/early work release). Table 1 shows the descriptive statistics of the sample in detail.

Table 1.

Description of the sample.

	Home detention			Early work release
	Experimental group	Control group	Matched group	Experimental group	Control group	Matched group
Number of subjects	19	19	19	24	30	26
Age (time of offence)	M = 34.1 (12.5)	M = 37.8 (10.3)	M = 35.2 (12.4)	M = 32.5* (8.4)	M = 38.6* (9.8)	M = 33.2 (8.1)
Age (first offence)	M = 29.4 (15.6)	M = 28.6 (11.4)	–	M = 26.8 (10.6)	M = 27.4 (11.9)	–
Substance abuse (time of offence)	32%	26%	37%	54%	37%	46%
Psychological treatment (before incarceration)	11%	5%	11%	13%	13%	0%
Civil status (married)	32%*	58%*	32%	25%	33%	27%
Previous convictions	M = 3.8 (4.2)	M = 5.4 (5.3)	M = 5.9 (7.2)	M = 5.9 (6.0)	M = 9.4 (8.7)	M = 5.5 (5.7)
Sentence (month)	M = 23.5 (19.5)	M = 23.8 (16.0)	M = 27.2 (21.9)	M = 24.4 (17.7)	M = 19.9 (18.9)	M = 23.7 (16.8)
Privileges granted prior to release	84%*	89%*	74%	9%*	73%	50%*
Early release	95%*	84%*	79%	92%	67%	65%
Length of EM (days)	M = 92.1 (50.7)	–	–	M =84.6 (49.0)	–	–

Note: Standard errors are indicated in parentheses, significant differences are presented in italics and marked with *. The level of significance is p ⩽ 5 percent.

Data collection and analysis

During the overall evaluation project, data collection occurred in multiple waves. Both experimental and control group subjects received financial compensation of €10 after having successfully participated in the first point of data collection; another €10 was paid to those who successfully participated in the last point of data collection. For this paper, we used data from the analysis of the prison files and official crime records of all 137 subjects after a follow-up period that ranged from 12 to 36 months. Missing data were handled with the single imputation technique. Missing observations were replaced with the overall unconditional mean of the (sub)sample (see Pérez et al., 2002: 3888). We also conducted a semi-structured interview with the subjects under EM (18 of the home detention group and the total of 24 in the early work release group) and five control group subjects (n = 47) at the end of the measure in order to gather information on how they experienced EM and why they decided to participate. An independent researcher visited the participants in their home or halfway house to conduct this interview. These interviews were analysed with a qualitative methodological approach using a thematic analysis. The aim of this analysis was to shed light on potential recidivism-reducing effects that the subjects associated with the measure. To achieve this, we employed deductive and inductive coding. The guiding principle for the deductive coding procedure was an open-ended analysis of whether the subjects associated EM with effects that are plausible from a theoretical perspective. The inductive coding was an exploratory analysis looking for unexpected categories and dimensions. This qualitative analysis was conducted by the authors and a research assistant. A constant comparison of the coding yielded a consensual understanding of the categories and enabled us to adjust the coding procedure when necessary.

We will begin by presenting the descriptive recidivism statistics and the results of logistic regression. We will then provide an overview of survival analysis, Kaplan-Meier estimates and Cox regression. In accordance with the initial development of the study, we will first present the analysis of the RCT, followed by the matched-pair analysis. Finally, we will summarize the relevant qualitative results.

Findings

Randomized controlled trial

Given the controversy over operationalization issues of relapse in recidivism studies, we defined recidivism in this study as every new criminal conviction, regardless of the type of crime, committed within a period of three years after release from prison or after termination of EM. Given the structure of the participants’ index offences (mostly less severe crimes) this procedure seemed reasonable. In the following, we compare the experimental group under EM with the control group within the same application field. The experimental group is the reference group.

Recidivism rates are presented in Table 2. No significant group differences were found in either the home detention (χ² (1, N = 38) = 0.18; p = .68) or the early work release scheme (χ² (1, N = 54) = 2.4; p = .12). The results of logistic regression supported this finding (electronically monitored home detention: B = −0.35, SE = 0.84; p = .68; early work release: B = 0.89, SE = 0.58; p = .13).

Table 2.

Recidivism statistics.

	Home detention		Early work release
	Experimental group (EM)	Control group	Experimental group (EM)	Control group
Follow-up period (months)	36	36	24	24
Sample size (N)	19	19	24	30
Recidivists (n)	4	3	7	15
Recidivists (percent)	21.1	15.8	29.2	50

To verify whether other factors influenced these results, we included multiple vari-ables gathered from the file analysis in the regression model. Stepwise regression (combination of backward elimination and forward selection) revealed that certain factors other than EM affected the likelihood of recidivism. Table 3 exhibits the multivariate final model of the logistic regression with the highest explanatory power in terms of recidivism for home detention.

Table 3.

Multivariate final model (home detention).

Variable	B	SE	OR	p
Control group	0.53	1.57	1.71	0.733
Substance abuse	2.99	1.5	19.97	0.046*
Age (first offence)	−0.17	0.11	0.84	0.137
Privileges	−4.64	1.99	0.01	0.02*
Constant	4.28	3.02	72.41	0.157

Notes: Obs: 38; Pseudo-R²: 0.56⁶; *p ⩽ .05; AIC: 26.06; BIC: 34.25.

The final model substantiated the results of the univariate logistic regression that EM had no significant impact. Furthermore, it showed that subjects with (former) substance abuse problems had a significantly higher risk of reoffending. In addition, there was a negative relationship between age of first (recorded) offence and the risk of recidivism. Lastly, subjects who were granted privileges had a significantly lower risk of relapsing than subjects who were not granted privileges.

Similar results can be reported for the application field of early work release (see Table 4). The recidivism risk depended not on whether the subjects were electronically monitored or not, but on the age of the offender at the time the crime was committed, the number of previous convictions and whether or not they were granted privileges.

Table 4.

Multivariate final model (early work release).

Variable	B	SE	OR	p
Control group	0.71	0.92	2.04	0.44
Age (time of offence)	−0.17	0.06	0.84	0.004**
Privileges	−3.76	1.65	0.02	0.023*
Previous convictions	0.25	0.82	1.28	0.003**
Constant	0.19	1.88	1.22	0.09

Notes: Obs: 54; Pseudo-R²: 0.45; *p ⩽ .05; **p ⩽ .01; AIC: 50.52; BIC: 60.46.

In both application fields, no significant effect was observed for additionally incorporated covariates such as type of offence, sentence length, civil status, level of education, clinical psychological symptoms (for example, suicide attempts, depressive symptoms, personality disorder) and psychological treatment. In sum, subjects who did not benefit from privileges in prison, who had substance abuse problems and who committed their (first) offence at a young age exhibited the highest risk of reoffending – irrespective of their participation in the electronically monitored early release scheme.

Next, we analysed how recidivism evolved over time without controlling for covariates. To this end, we used survival analysis illustrated by the nonparametric Kaplan-Meier estimate model to assess the efficiency of EM. Figure 2 shows the comparison of the estimated survival curves of the experimental group and the control group of home detention with the confidence interval (CI).

Figure 2.

Kaplan-Meier survival estimates: Home detention.

In Figure 2, the CIs overlap and both curves proceed within both CIs. Thus, the groups were near equivalent with regard to the course of reoffending. EM did not affect the likelihood of recidivism. The Log-Rank test proves this result (p = .63). Figure 3 illustrates the survival analysis of the early work release groups.⁷

Figure 3.

Kaplan-Meier survival estimates: Early work release.

Most of the new offences occurred within the first 12 months after prison release. In this area of application, the Kaplan-Meier curves show no significant difference. After 12 months, a slightly higher recidivism risk for the control group can be observed. However, the Log-Rank test (p = .27) and the Wilcoxon–Breslow Test (p = .37) showed that these differences did not reach statistical significance. EM did not have a significant impact on the recidivism risk.

The Cox proportional hazards model was also used to assess the impact of EM and other variables on time to recidivism. The Cox model allows for the consideration of single and multiple reoffending within a given follow-up time. The Cox regression substantiated the preceding results. Considering the time of the re-arrests within the follow-up period, no difference regarding EM could be observed (p = 1.0 in each application field). In summary, the RCT showed no effect of EM on the likelihood of recidivism.

Matched-pair analysis

The follow-up period in the additional analysis was one year after release from prison or after EM. The groups had a high level of comparability. In accordance with the matching procedure, there were no significant differences between the groups regarding the matching variables. Table 5 shows the recidivism statistic.

Table 5.

Recidivism statistic.

	Home detention		Early work release
	Experimental group (EM)	Matched group	Experimental group (EM)	Matched group
Follow-up period (months)	12	12	12	12
Sample size (N)	19	19	24	26
Recidivists (n)	3	2	6	3
Recidivists (percent)	15.8	10.5	25	11.5

The results of the chi-square analysis and the univariate logistic regression in these additional analyses correspond to the findings from the RCT. The chi-square analysis showed no significant difference in the recidivism risk for home detention (χ² (1, N = 38) = 0.23; p = .63) or early work release (χ² (1, N = 50) = 1.53; p = .22). The results of the logistic regression confirmed these findings (home detention: B = −0.47, SE = 0.98, p = .63; early work release: B = −0.94, SE = 0.77, p = .23). Here again, the multivariate final model showed different variables that affected the likelihood of recidivism in comparison with the experimental analysis (see Tables 6 and 7). Additionally, in neither the application field of home detention nor the application field of early work release did EM influence the recidivism risk significantly.

Table 6.

Multivariate final model: Home detention.

Variable	B	SE	OR	p
Matched group	−2.31	2.26	0.09	0.306
Age (time of offence)	−0.52	0.29	0.60	0.077
Psychological treatment	5.05	2.55	156.36	0.047*
Civil status	4.11	2.56	61.17	0.108
Constant	6.07	4.44	432.94	0.171

Notes: Obs: 38; *p ⩽ .05; Pseudo-R²: 0.49; AIC: 25.01; BIC: 33.2.

Table 7.

Multivariate final model: Early work release.

Variable	B	SE	OR	p
Matched group	−1.10	0.96	0.33	0.253
Age (time of offence)	−0.18	0.1	0.84	0.071
Previous convictions	0.34	0.13	1.4	0.008*
Constant	1.67	2.31	5.33	0.468

Notes: Obs: 38; *p ⩽ .05; Pseudo-R²: 0.35; AIC: 38.55; BIC: 46.2.

The Kaplan-Meier estimates support the outcome that EM did not affect the likelihood of recidivism significantly (see Figures 4 and 5; Log-Rank test, p = .51, for the home detention comparison and Log-Rank test, p = .71, for the early work release comparison).

Figure 4.

Kaplan-Meier survival estimates: Home detention – matched-pair.

Figure 5.

Kaplan-Meier survival estimates: Early work release – matched-pair.

Thus, the additional analyses with the matched-pair subgroups confirmed the finding that EM does not have an influence on the risk of reoffending after the termination of the intervention. The hypothesis that EM exerts a recidivism-reducing impact has to be rejected.

Qualitative results

Although we cannot claim that the results of the qualitative analysis are representative, they did yield some interesting findings with regard to the discussed theoretical underpinnings. Some interviewees stressed a deterrent effect, but solely for the duration of EM: ‘It is a deterrent at first. You think twice before doing something.’ But it also became evident that this deterrent effect vanished over time: ‘… in the meantime, I’ve forgotten about it, almost as if it were nothing.’

The most prominent finding pertained to the possibility of maintaining family ties and going to work because of EM. ‘For me, the most important thing was to not lose my job,’ said one of the interviewees. One subject claimed that, due to imprisonment, ‘most of the people lose their jobs. They lose social bonding and social contacts,’ which leads to ‘getting into trouble and then getting sent right back to prison’. He based this on the assumption that ‘no job means alcohol and drugs, no money, going out to steal, getting in trouble’. Another subject was convinced ‘that if I hadn’t been released now, then I would have lost my family. I’m sure. That this didn’t happen is entirely due to the electronic bracelet.’

As indicated above, the sample consisted mainly of low- to medium-risk subjects. For them, prison may be a highly criminogenic place where low-risk offenders in particular may ‘end up being more criminal after they are released’, as one respondent claims. Even though incarceration is likely to be more stigmatizing than wearing an ankle bracelet, EM entails the risk of stigmatization as well. Almost every offender expressed something similar to the following statement made by another subject: ‘I worried a lot that when people saw me they would think that I am one of those’ (with ‘one of those’ being a reference to sex offenders). Still, the majority of offenders who expressed concern about stigmatization did not actually experience it.

Some participants made remarks with regard to a lasting internalizing effect. ‘If a person doesn’t change, then this electronic ankle bracelet doesn’t guarantee that nothing will happen … More is required than just an electronic ankle bracelet,’ one respondent said, while another claimed that his life ‘changed completely because of it [EM]’. He stated that the lack of structure in his daily routine had led to all the offences and that EM enabled him to ‘have a structured daily routine and an apartment’.

Discussion

The aim of this study was to test whether EM of low-risk prisoners in two areas of application contributed to reduced recidivism as compared with prisoners who did not have the opportunity to shorten their prison sentence by early work release or home detention. To answer this question, we employed an RCT supplemented by a matched-pair sample. Our results show that EM did not have a recidivism-reducing impact on the target groups under scrutiny. Neither in the early work release nor in the home detention scheme did EM lead to a decreased rate of recidivism in the experimental group. With regard to the goal of avoiding the incarceration of low-risk offenders who were unable to pay a fine, EM could not meet its expectations at all.

Like other studies on EM, the present study focused on a very specific target group. In our case, the statutory requirements led to the recruitment of low-risk offenders. With an average age of 27 at the time of their first offence, the participants did not belong to the high-risk group of ‘early starters’, a risk factor that has been proven to be a strong predictor of repeat offending (Farrington, 2003). Thus, the present sample did not represent a particularly relapse-prone population, a factor that has to be kept in mind when interpreting the findings and comparing this study’s findings with other research results. Compared with the general prison population in Germany, the subjects had a relatively favourable release situation, especially considering factors such as housing, work and family situation.

A deterrent effect after the termination of EM that is associated with the Rational Choice approach should not be assumed. In a study with medium- to high-risk offenders in the US state of Florida, 85 percent of the offenders said the likelihood of absconding was not influenced by EM (US Department of Justice, 2011). Gies et al. (2012) in contrast argued in favour of the deterrent effect of EM and found that it had a recidivism-reducing effect during its use on high-risk offenders. The deterrent effect of EM – if it exists at all – might already begin to decrease during the surveillance period: on the one hand, offenders are likely to become habituated to the electronic device; on the other hand, ‘GPS technology is currently too underdeveloped’ (Armstrong and Freeman, 2011: 181) and the numerous false alarms will counteract a deterrent effect (while under EM). With regard to long-term recidivism-reducing effects in terms of Routine Activity Theory, Killias et al. (2010) claim a recidivism-reducing potential of EM emerging from the elimination of opportunities to commit an offence. Furthermore, the offender might internalize self-control as mentioned in the commentary to the recommendation of the Council of Europe.⁸ Yeh (2010) stated that EM’s long-lasting effect originates from the long-term change in the offender’s personality that is learned during the time under surveillance. However, one ‘would not expect any crime control impact of this “self-control assist” to last beyond its application’ (Renzema, 2003: 8). Yet, this thought should not inevitably lead to net-widening in the sense of prolonging the amount of time under EM. To date, there is no evidence supporting the assumed effect of the internalization of external control. Likewise, our results do not support an internalizing impact of the measure. Even the interviewee who stressed the benefits of the structuring effect of EM relapsed.

Regarding the time course of recidivism, most of the offences were committed within the first 6 to 12 months after release from prison. The high risk of recidivism especially in the first year after prison release is confirmed by numerous criminological studies (for example, Bush and Moore, 2001; Jehle et al., 2016). If EM has any effect at all, it can be assumed that it is most effective for release preparation or directly after prison release in this risky phase of re-entry. It should be noted that the speed of reoffending does not differ significantly between the experimental group and the control group. This should also be kept in mind with regard to proponents of EM who argue that EM will eventually lead to an internalization of control (Henneguelle et al., 2016).

Another theoretical link between EM and recidivism discussed in the literature is its potential prevention of prisonization effects. According to Clemmer (1958), prisonization increases linearly during the course of incarceration. Shortening a prison term by means of EM could thus reduce prisonization. Wheeler (1961) claimed that prisonization is at its highest midway through the term of imprisonment. EM, as a means of mitigating or avoiding negative effects of imprisonment, would have an effect only if used to replace a prison sentence completely or if used in conjunction with conditional early release during the first half of the sentence. According to Yeh (2010), the avoidance as well as the shortening of the prison sentence might have the anticipated impact. Likewise, Schwedler and Woessner’s (2017) results portended a prisonization-reducing effect among the experimental group of the present study (which also included the transfer to electronically monitored home detention before conditional release during the first half of the fixed-term prison sentence). In addition, some respondents stressed the prisonization-preventing effect that was associated with early release under EM.

One of the strongest recidivism-reducing mechanisms of EM is seen in its potential to facilitate early release or to avoid incarceration entirely and thus to preserve protective prosocial factors, such as the workplace or social ties (see Andersen and Telle, 2016; Larsen, 2017). According to Padgett et al. (2006: 73), for instance, the workplace has a structuring effect and improves ‘lifestyle stability’, a factor that may support desistance from crime. Additionally, the offender does not lose important work skills during incarceration (Western et al., 2001). Participants in the present study also stressed this positive effect. It must be taken into account, however, that participation in the study was purely voluntary, which might have led to a positive bias. A positive impact with regard to family and job might depend on the length of time under EM and the quality of the relationship with both family and employer. Furthermore, some eligible prisoners rejected participation because they feared a negative impact of EM on family and workplace. An electronically supervised offender has to cope with the frustration and restrictions on autonomy associated with the measure. This might in turn put strain on a partner (Payne, 2014). In addition, certain jobs such as truck-driving might be especially difficult under EM supervision. Not only the supervised person him/herself but employers and partners as well might fear the stigmatizing effects of a perceivable ankle bracelet (Gainey and Payne, 2000; Nellis, 2015; Vanhaelemeesch et al., 2014). This fear of humiliation applies regardless of whether participants are low- or high-risk offenders (see Gainey and Payne, 2000; Vanhaelemeesch et al., 2014) and regardless of whether the offender actually experiences exclusion or stigmatization owing to the electronic device. Almost all respondents in the present study voiced such feelings, even though only two subjects actually experienced stigmatization. Fallesen and Andersen (2017) stated that EM lowers the risk of stigmatization and therefore supports social inclusion, which is an important factor for noncriminal trajectories. However, this result has to be put into perspective: obviously, a general de-stigmatizing effect of EM must be rejected.

Factors that are generally predictive of criminal recidivism such as substance abuse, age and prior criminal history (Gendreau et al., 1996; Hanson, 2009) predict the risk of reoffending more accurately than EM (for example, Marklund and Holmberg, 2009; Petersilia and Turner, 1990). Since subjects who are granted privileges are very likely to exhibit positive prognostic features, it comes as no surprise that privileges are associated with a reduced likelihood of reoffending. Likewise, age at first offence and age at the time of the index offence as well as a high number of previous convictions are well-known predictive criteria (Farrington 1988; Hirschi and Gottfredson 1983). With regard to the influence of substance abuse, the results of several studies suggest a strong relationship between substance abuse and offending (Hakansson and Berglund, 2012). Individuals who are exposed to multiple risk factors are particularly vulnerable to substance abuse as a coping strategy, while at the same time these risk factors contribute to the pathway to delinquency. According to the General Theory of Crime of Gottfredson and Hirschi (1990), both delinquent behaviour and substance use represent difficulties in self-control and thus have common roots. Renzema and Mayo-Wilson (2005), in their meta-analysis, claimed that treatment of substance abuse contributes significantly to reduced recidivism. According to Looman and Abracen (2011), low-risk offenders are likely to recidivate at relatively high rates if they exhibit a history of alcohol abuse. EM could support monitored persons by helping them keep appointments with drug counsellors or by using exclusion and inclusion zones to keep them away from places known for drug trading and drug-related crimes. However, drug abuse problems also increase the likelihood of programme failures, that is, they increase the likelihood that the participant will fail with the EM order (Bonta et al., 1999). In a study by Deuchar (2012: 119), EM home curfew worsened drug problems since for ‘some young men, substance abuse was used as a coping mechanism to deal with the perceived strains associated with restrictions of liberty’. The majority of subjects under EM in a study by Hucklesby (2008) reported unmodified substance use while under EM. The findings of Barton and Roy (2008) support the assumption that the success of EM in combination with drug programmes depends on sociodemographic features such as age and employment status. Technological advances enable the use of remote alcohol monitoring, which allows for testing and reporting of alcohol and drug consumption based on the transdermal measure-ment of the substances; this tool is mainly used for DUI (driving under the influence) offenders (McKnight et al., 2012).

Overall, it seems that the positive impact associated with EM takes effect only under certain conditions. Convicted persons who are imprisoned face the risk of losing their job and social ties, losses that entail adverse financial and social consequences. Although EM does not eliminate these risk factors, it might enable the offender to sustain these protective factors or facilitate a stable life during release preparation. Andersen and Andersen (2012) stated that the opportunity to participate in the labour market is negatively affected by the stigma of incarceration in contrast to the use of EM, which facilitates returning to one’s job (see also Henneguelle et al., 2016). Again, however, this effect might apply only to certain offender groups, since some studies found that EM had detrimental effects on reintegration into the job market (for example, Bales et al., 2010).

The study is subject to several limitations. The small sample size and the considerable dropout rate hamper internal validity. Thus, even the gold standard of an experimental study design encounters methodological shortcomings. A third control group with the same privileges and treatment as the EM subjects but without electronic surveillance would have been extremely valuable.

Even though there are plenty of implementation studies on EM, there is still a lack of evidence-based research on the impact of EM on reoffending. Of course, with the analysis presented here, we were not able directly to test the assumed associated theoretically based effects of EM on recidivism. However, the study provides new evidence on the theoretical foundation of EM and on its preventive effect on reoffending. A major objection might be that the study did not really measure the impact of EM since EM only facilitated the realization of early release measures and – thus – it was these measures that were tested. This is, of course, an objection that can be raised about all EM research.

Although this concern may be partially true, the study showed that EM indeed does not reduce recidivism on its own, but that it might be a necessary condition for further rehabilitative measures (such as early release), taking the safety requirements of society into consideration. It must be emphasized, especially with regard to the present findings, that this ‘soothing’ side effect should not be the main driving force of its application.

Conclusion and outlook

A major purpose of criminal sanctions is to reduce the risk of reoffending. At the same time, sanctions must be proportionate and appropriate with regard to their punitiveness. Therefore, the intensity of EM must reflect the necessity in the individual case, the severity of the offence committed and the risk of reoffending.⁹ This means that the risk levels of persons subject to EM must be taken into account. However, in order for this to be done, it is first necessary to thoroughly understand how EM works in various settings, namely, in various application schemes and with various target groups.

The present findings have some significant implications for the EM discussion. First, they underline the exchangeability of sanctions. According to Boehm (1996), the risk of committing a new offence does not depend on the sanction (see also Tetal, 2018). Whether or not a person is likely to reoffend depends solely on the load of risk factors combined in one person. Thus, the decision to impose EM requires an in-depth evaluation of the situation, a thorough understanding of the goals of the sentence and the aims of EM and a judgement as to whether the disruptive effects on social ties and job opportunities caused by the sanctions are proportionate to the crime committed. Against this background, the findings of our study could also lead to the conclusion that a comparable risk of reoffending for both schemes (early release with EM as compared with regular imprisonment) might in some cases speak for the EM scheme (since early release with EM for this target group does no harm in terms of recidivism).

Second, these assessments will depend on the target group. For subjects who, like those in the present pilot study, are able to decide voluntarily for or against the use of EM, EM exhibits different dynamics and effects than it does for high-risk offenders who are subject to mandatory electronic supervision. In fact, even in the present study, not all approached potential subjects were as eager to wear an electronic device as were those for instance in Killias et al.’s study (2010).

This leads to a third aspect, namely ethical issues. Even if participation in the pilot project was voluntary, approached inmates might feel pressure to participate out of the fear of negative consequences should they refuse. In addition, critics might argue that offenders who undergo EM experience a greater degree of freedom notwithstanding any validated benefit. However, as long as the positive effects of EM are not sufficiently and clearly proven, this objection is not entirely convincing.

Fourth, there is still no theoretical support for a recidivism-reducing effect of EM other than deterrence during the period of EM. If EM produces any recidivism-reducing effects beyond that, they can, at best, be attributed to accompanying measures such as probation sessions, drug counselling and early release itself – rather than EM alone. It is questionable whether EM needs to be employed to foster compliance in a low-risk offender group for which the rehabilitation needs are minor. EM may have a prosocial impact with regard to family and the workplace, but a recidivism-reducing effect has not yet been proven. The theoretical assumptions as to why EM should have a recidivism-reducing effect have to be interpreted against the background of the sample’s overall characteristics. Thus, we still need more research and debate about the different effects of EM (and its various technological applications – GPS versus radio frequency) on different target groups. Even though EM might increase public acceptance for early release, it also entails a very real danger of net-widening.

Footnotes

Acknowledgements

We thank Dr Volker Grundies, Dr Chris Murphy and Emily Silverman, JD (Berkeley Law, LL.M.) for their thoughtful comments and suggestions. We also wish to thank Dr Joachim Obergfell-Fuchs and Bernadette Schaffer, from the Academy for Prison Officers and the Criminology Service Baden-Württemberg, for their efforts in providing the data for the matched group.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship and/or publication of this article: With the exception of this recidivism study, the overall project evaluation was partially funded by the Ministry of Justice Baden-Württemberg.

Notes

References

Andersen

(2012) Losing the Stigma of Incarceration. Does Serving a Sentence with Electronic Monitoring Causally Improve Post-release Labor Market Outcomes? Odense: University Press of Southern Denmark.

Andersen

Telle

(2016) Electronic monitoring and recidivism. Quasi-experimental evidence from Norway. Statistics Norway, Research Department, August.

Andrews

Bonta

(1998) Level of Service Inventory – Revised. North Tonawanda, NY: Multi-Health Systems.

Armstrong

Freeman

(2011) Examining GPS monitoring alerts triggered by sex offenders: The divergence of legislative goals and practical application in community corrections. Journal of Criminal Justice 39(2): 175–182.

Bales

Mann

Blomberg

Gaes

Barrick

Dhungana

McManus

(2010) A quantitative and qualitative assessment of electronic monitoring. Florida State University, Tallahassee, January.

Barton

Roy

(2008) Convicted drunk drivers in an electronic monitoring program: A preliminary study. International Journal of Criminal Justice Sciences 3(1): 28–43.

Boehm

(1996) Die spezialpräventiven Wirkungen der strafrechtlichen Sanktionen. In: Jehle

(ed.) Kriminalprävention und Strafjustiz. Wiesbaden: Kriminologische Zentralstelle, 263–290.

Bonta

Wallace-Capretta

Rooney

(1999) Electronic Monitoring in Canada. Solicitor General Canada.

Bottos

(2007) An Overview of Electronic Monitoring in Corrections: The Issues and Implications. Correctional Service Canada.

10.

Bush

Moore

(2001) Recidivism report. Florida Department of Corrections, May.

11.

Clemmer

(1958) The Prison Community. Boston: Harcourt Brace College Publishers.

12.

Cornish

Clarke

RVG

(1986) The Reasoning Criminal. Rational Choice Perspectives on Offending. New York, Berlin, Heidelberg, Tokyo: Springer.

13.

Deuchar

(2012) The impact of curfews and electronic monitoring on the social strains, support and capital experienced by youth gang members and offenders in the west of Scotland. Criminology and Criminal Justice 12(2): 113–128.

14.

Fallesen

Andersen

(2017) Explaining the consequences of imprisonment for union formation and dissolution in Denmark. Journal of Policy Analysis and Management 36(1): 154–177.

15.

Farrington

(1988) Advancing knowledge about the onset of delinquency and crime. In: Lahey

Kazdin

(eds) Advances in Clinical Child Psychology. Boston: Springer, 283–342.

16.

Farrington

(2003) British randomized experiments on crime and justice. The Annals of the American Academy of Political and Social Science 589(1): 150–167.

17.

Felson

Clarke

(1998) Opportunity Makes the Thief. London: Home Office, Policing and Reducing Crime Unit, Research, Development and Statistics Directorate.

18.

Finn

Muirhead-Steves

(2002) The effectiveness of electronic monitoring with violent male parolees. Justice Quarterly 19(2): 293–312.

19.

Gainey

Payne

(2000) Understanding the experience of house arrest with electronic monitoring. An analysis of quantitative and qualitative data. International Journal of Offender Therapy and Comparative Criminology 44(1): 84–96.

20.

Gendreau

Little

Goggin

(1996) A meta-analysis of the predictors of adult offender recidivism: What works! Criminology 34(4): 575–607.

21.

Gies

Gainey

Cohen

Healy

Duplantier

Yeide

Bekelman

Bobnis

Hopps

(2012) Monitoring High-Risk Sex Offenders with GPS Technology: An Evaluation of the California Supervision Program. Final Report. Bethesda, MD: Development Services Group.

22.

Goffman

(1981) Asyle. Frankfurt am Main: Suhrkamp.

23.

Gottfredson

Hirschi

(1990) A General Theory of Crime. Stanford, CA: Stanford University Press.

24.

Hakansson

Berglund

(2012) Risk factors for criminal recidivism – a prospective follow-up study in prisoners with substance use. BMC Psychiatry 2012 12: 111–118.

25.

Hanson

(2009) The psychological assessment of risk for crime and violence. Canadian Psychology/Psychologie canadienne 50: 172–182.

26.

Henneguelle

Monnery

Kensey

(2016) Better at home than in prison? The effects of electronic monitoring on recidivism in France. Journal of Law and Economics 59(3): 629–667.

27.

Hirschi

(1986) On the compatibility of rational choice and social control theories of crime. In: Cornish

Clarke

(eds) The Reasoning Criminal: Rational Choice Perspectives on Offending. New York: Springer, 105–118.

28.

Hirschi

Gottfredson

(1983) Age and the explanation of crime. American Journal of Sociology (89)3: 552–584.

29.

Hucklesby

(2008) Vehicles of desistance? The impact of electronically monitored curfew orders. Criminology and Criminal Justice 8(1): 51–71.

30.

Hucklesby

(2009) Understanding offenders’ compliance: A case study of electronically monitored curfew orders. Journal of Law and Society 36(2): 248–271.

31.

Jehle

Albrecht

Hohmann-Fricke

Tetal

(2016) Legalbewährung nach strafrechtlichen Sanktionen. Mönchengladbach: Forum Verlag Godesberg GmbH.

32.

Killias

Gillieron

Kissling

Villettaz

(2010) Community service versus electronic monitoring – What works better? British Journal of Criminology 50(6): 1155–1170.

33.

Kornhauser

Laster

(2014) Punitiveness in Australia. Electronic monitoring vs the prison. Crime, Law and Social Change 62(4): 445–474.

34.

Larsen

BØ

(2017) Educational outcomes after serving with electronic monitoring. Results from a natural experiment. Journal of Quantitative Criminology 33(1): 157–178.

35.

Looman

Abracen

(2011) Substance abuse among high-risk sexual offenders: Do measures of lifetime history of substance abuse add to the prediction of recidivism over actuarial risk assessment instruments? Journal of Interpersonal Violence 26(4): 683–700.

36.

MacKenzie

(2006) What Works in Corrections? Reducing the Criminal Activities of Offenders and Delinquents. Cambridge: Cambridge University Press.

37.

McKnight

Fell

Auld-Owens

(2012) Transdermal Alcohol Monitoring: Case Studies. Washington DC: National Highway Traffic Safety Administration.

38.

Marklund

Holmberg

(2009) Effects of early release from prison using electronic tagging in Sweden. Journal of Experimental Criminology 5(1): 41–61.

39.

Mayer

(2004) Modellprojekt elektronische Fußfessel. Freiburg: edition iuscrim.

40.

National Institute of Justice (2011) Electronic monitoring reduces recidivism. U.S. Department of Justice, NCJ 234460.

41.

Nellis

(1991) The electronic monitoring of offenders in England and Wales. British Journal of Criminology 31(2): 165–185.

42.

Nellis

(2015) Standards and Ethics in Electronic Monitoring. Handbook for Professionals Responsible for the Establishment and the Use of Electronic Monitoring. Council of Europe, June.

43.

Padgett

Bales

Blomberg

(2006) Under surveillance. An empirical test of the effectiveness and consequences of electronic monitoring. Criminology & Public Policy 5(1): 61–91.

44.

Payne

(2014) It’s a small world, but I wouldn’t want to paint it: Learning from Denmark’s experience with electronic monitoring. Criminology & Public Policy 13(3): 381–391.

45.

Pérez

Dennis

Jacky

Rondón

López

(2002) Use of the mean, hot deck and multiple imputation techniques to predict outcome in intensive care unit patients in Colombia. Statistics in Medicine 21(24): 3885–3896.

46.

Petersilia

Turner

(1990) Comparing intensive and regular supervision for high-risk probationers. Early results from an experiment in California. Crime & Delinquency 36(1): 87–111.

47.

Renzema

(2003) Electronic monitoring’s impact on reoffending. Review. Department of Criminal Justice and Social Work, Kutztown University, March.

48.

Renzema

Mayo-Wilson

(2005) Can electronic monitoring reduce crime for moderate to high-risk offenders? Journal of Experimental Criminology 1(2): 215–237.

49.

Sampson

(2010) Gold Standard Myths: Observations on the Experimental Turn in Quantitative Criminology. Journal of Quantitative Criminology 26(4): 489-500.

50.

Schwedler

Wößner

(2015) Elektronische Aufsicht bei vollzugsöffnenden Maßnahmen. Berlin: Duncker & Humblot.

51.

Schwedler

Woessner

(2017) Identifying the rehabilitative potential of electronically monitored release preparation. A randomized controlled study in Germany. International Journal of Offender Therapy and Comparative Criminology 61(8): 839–856.

52.

Tetal

(2018) Die Wirkung strafrechtlicher Sanktionen auf die Legalbewährung. In: Walsh

Pniewski

Kober

Amborst

(eds) Evidenzorientierte Kriminalprävention in Deutschland. Ein Leitfaden für Politik und Praxis. Berlin: Springer, 533–556.

53.

Vanhaelemeesch

Vander Beken

Vandevelde

(2014) Punishment at home. European Journal of Criminology 11(3): 273–287.

54.

Western

King

Weiman

(2001) The labor market consequences for incarceration. Crime & Delinquency 47(3): 410–427.

55.

Wheeler

(1961) Socialization in correctional communities. American Sociological Review 26(5): 697–712.

56.

Yeh

(2010) Cost–benefit analysis of reducing crime through electronic monitoring of parolees and probationers. Journal of Criminal Justice 38(5): 1090–1096.

57.

Zimring

Hawkins

(1995) Incapacitation. New York: Oxford University Press.