Sage Journals: Discover world-class research

Abstract

Previous research on victim-offender mediation (VOM) and reoffending rates often lack valid control groups and might suffer from a self-selection bias. Moreover, VOM’s role in the criminal justice system, particularly the effect on sanctioning, has not been considered in explaining the lower reoffending rates. This research addressed these gaps, by comparing four offender groups and using propensity score matching (N = 2,907): those who participated in VOM, those unwilling, those unable due to victim refusal, and those not referred. Results replicated that VOM participation predicts lower reoffending rates, primarily due to the mediation process itself rather than self-selection. While VOM reduced the likelihood of sanctions, this did not explain the observed decrease in reoffending rates.

Keywords

restorative justice victim-offender mediation offender behavior propensity score matching recidivism

Introduction

One aim of the criminal justice system is reducing future criminal behavior (Carlsmith et al., 2002; Claessen, 2019). This means that it is important for governments that the programs used in the criminal justice process have the possibility to impact reoffending rates. One such program is victim-offender mediation (VOM). In VOM victims and offenders of an offense get the opportunity to discuss the crime, ask questions, and come to an agreement on how to best solve the crime (Hansen & Umbreit, 2018). VOM is an example of a restorative justice program. Whereas in the conventional justice system crime is seen as a violation of rules and norms, restorative justice views crime as a violation of the relationship between people, which can only be solved by active involvement of the main involved parties of a conflict—victim, offender, and the community (Bohmert et al., 2018; Wong et al., 2016). It aims to focus on the needs of the victim while actively holding the offender accountable. According to restorative justice principles, crimes create obligations of which one is to make things right (Zehr, 2015). Through restorative practices, offenders are facilitated and encouraged to take responsibility, and repair the harm that has been done to the victim, the community and their own dignity (Braithwaite, 1998). The general mean used within restorative justice is a dialogue between the parties, such as VOM. In VOM, a voluntary conversation is organized between the victim and the offender in presence of a trained mediator (Hansen & Umbreit, 2018). The mediator structures the process so parties can work towards an agreement. When parties are prepared well and respectfully listen to each other in this process, there is an increased change of apologizing and forgiveness (Braithwaite, 2003).

The application of restorative justice as a response to crime continues to grow (D’Souza & L’Hoiry, 2019). Restorative justice programs can be applied at every moment of the criminal justice process. It can be used as either supplemental to or as alternative for the contemporary criminal justice system (Claessen & Roelofs, 2020; Zebel et al., 2017). Although restorative justice is not primarily applied to reduce reoffending rates, but to restore the damage done, it would be beneficial when restorative justice can have an inhibitory impact on reoffending rates. Since crime reduction is an important aim of the criminal justice system, it is important for law enforcement agencies that, when restorative justice is part of the criminal justice process, this program is able to reduce reoffending rates.

Research does indeed suggest that restorative justice programs, such as VOM, impact reoffending (Bergseth & Bouffard, 2013; Claessen et al., 2015; Hansen & Umbreit, 2018; Kimbrell et al., 2023; Wong et al., 2016). On the other hand, scholars, have argued that the effects of VOM on reoffending might be explained by a self-selection bias, since very few studies have adopted a study design that can infer a causal relationship (Kimbrell et al., 2023; Latimer et al., 2005). To be able to infer causality, random assignment of offenders to either VOM or the conventional criminal justice system would be necessary. However, such a design is hard to realize. Due to the voluntary nature, it is practically hard and ethically undesirable to randomly assign people to VOM. As a result, in most studies examining VOM, researchers compared a group of offenders who participated in VOM, with a group of offenders who did not participate in VOM (Bradshaw & Roseborough, 2005; Latimer et al., 2005). Considering the voluntary nature of VOM, this might mean that researchers are comparing offenders who are willing to participate in VOM to offenders who are not willing to participate or for whom it is unknown if they are willing to participate, because they were not referred. Offenders who participate in VOM might have different motivations and psychological needs compared to offenders who do not participate (Jonas-van Dijk, 2024). Hence, a self-selection bias might explain the relation found between participation in VOM and reduced reoffending rates (Latimer et al., 2005). For this reason, it is important to include valid control groups to examine the effects of VOM on reoffending more rigorously.

Recently studies have been published with a better research design that allow for more robust conclusions. Kimbrell et al. (2023) performed a meta-analysis including both studies with a randomized controlled trail (RCT) and studies with a quasi-experimental design. They examined the overall impact of all type of programs that included a restorative component for juveniles (e.g., conferencing, mediation, restitution, and teen court), but also looked at the specific impact of programs like conferencing and mediation, in which victim and offender had some form of mediated, (in)direct contact. For this specific group of restorative justice programs, the outcomes showed that conferencing/mediation was more effective in reducing reoffending compared to the traditional criminal justice process without victim-offender contact, with a moderate effect size. However, Kimbrell et al. (2023) do indicate that even though the outcomes are promising, more randomized controlled trials are necessary for strong conclusions. Especially since it seems that for VOM only one RCT has been performed so far and in that study there was a problem of (selective) dropout (Urban & Burge, 2006).

A different study, from Stewart et al. (2018) focused on mediation with adult offenders only, during or after release from prison. They concluded that offenders who participated in mediation after release had lower reoffending rates compared to offenders who did not participate. They did not find this effect when offenders participated in mediation while incarcerated. In their study they made use of one-to-one matching, to include a better control group, that matched on co-variates. However, as indicated by Stewart et al. (2018) there still might be a change that offenders who participated in mediation had different motivations compared to those who did not participate—thus there might still have been a selection bias. It would be interesting and important to include a control group of offenders who were willing to participate in VOM but did not experience VOM, and compare this with a group who participated and a group who was not willing to participate. Such a research design would offer more insight into the role of the self-selection bias in explaining lower reoffending.

Jonas-van Dijk et al. (2020) adopted such a research design and focused on whether the impact of VOM on reoffending was due to a self-selection or due to the VOM process by including a control group of offenders willing to participate in VOM, but unable to do so because the other party declined the option. In their research a comparison was made between (a) offenders who participated in VOM, (b) offenders who were willing but unable to participate, and (c) offenders who were unwilling to participate—while controlling for several case and offender characteristics. They observed significantly lower reoffending rates of offenders who participated in VOM compared to those who were unwilling to participate. However, the group of offenders who were willing but unable to participate did not significantly differ from the other two groups. The rates of this group were in between the groups of offenders who participated in VOM and the group of offenders who were unwilling to do so. They therefore cautiously concluded that the impact of VOM on reoffending in their study was explained by a combination of a self-selection bias and the VOM process itself.

Lauwaert and Aertsen (2016) also suggest a self-selection bias explaining the impact of VOM on desistance, since they indicate that VOM supports a desistance process that already started, suggesting that offenders might already have a specific motivation that non-participators might not have. In addition, Jonas-van Dijk et al.’s (2022) findings suggest that the VOM process itself, over and beyond self-selection effects, brings about a psychological impact, such as increased responsibility taking and higher feeling of guilt and shame. These psychological variables are shown to be related to prosocial behavior and reduced reoffending (Gausel et al., 2016; Hosser et al., 2008). The findings from previous research to date therefore suggested to us that a combination of the VOM-process and a self-selection bias would offer the best explanation for lower reoffending rates observed after VOM.¹

Although the approach of Jonas-van Dijk et al. (2020) helped to gauge the impact of self-selection and the VOM process itself on reoffending, their analytical approach using controls still cast doubt on the internal validity of this impact of the VOM process. A stronger analytical approach when a true experiment with random allocation is not possible or feasible, is to use a propensity score matched control group. With propensity score matching the aim is to statistically mimic an experimental design with random allocation (Beal & Kupzyk, 2014). Propensity score matching means that for every case the chance of getting assigned to VOM is calculated based on background or confounding variables. This is the propensity score. Subsequently every referred case is matched, based on the propensity score to a non-referred case. Due to the balance in confounder variables, effects that are found in the analyses can then with more certainty be ascribed to the VOM process the case went through (compared to the conventional criminal justice process without VOM). However, the conclusions that can be drawn about the effects, is as strong as the matching that was used (Wermink et al., 2010). To the degree that more (potentially) confounding variables are used for matching, the stronger the matching becomes and the conclusions that can be drawn. The first aim of this quasi-experimental study was therefore to provide an advanced, more rigorous test of the impact of VOM participation on reoffending, by comparing offenders who participated in VOM to (a) two groups of offenders referred to but without VOM, who were willing or unwilling to participate and (b) to a propensity-score matched control group of offenders who were not referred to VOM.

Another factor that has not been examined in previous research, that has been taken into account as well here, is the role of the manner in which the criminal case is finalized (i.e., the type of sanctioning). When VOM is part of the criminal justice system, the judge or public prosecutor has a final say in how the case is solved (Hansen & Umbreit, 2018; Jonas-van Dijk et al., 2022). This means that when parties come to an agreement during mediation, and the judge or public prosecutor takes this into account, a sentence can still be imposed, or they can decide that a sanction is no longer required given the outcomes of the mediation (Claessen et al., 2015). In general, receiving an official sanction can be experienced as stigmatizing (Bernburg & Krohn, 2003), which can predict increased reoffending (Braithwaite, 2001). A sanction and a criminal record, also negatively impact future job positions (Cerda-Jara & Harding, 2024; Nur & Monaghan, 2024) and therefore increases recidivism (Uggen, 2000). Therefore, the absence of sanctioning, or receiving a different type of sanction as a result of participating in mediation might be a factor as well contributing to the lower reoffending rates after mediation (Claessen et al., 2015; Jonas-van Dijk et al., 2020). Due to the absence of sanctioning or receiving a lower sentence, the mediation process could be perceived as less stigmatizing and more reintegrative to offenders, supporting the reintegrative shaming theory (Braithwaite, 2001). This means that (absence of) sanctioning might be part of the explanation of the lower reoffending observed after VOM programs offered within criminal procedures. Hence, the second aim of this study was to examine if the nature and severity of sanctions after VOM differ from the sanctions that are imposed in a criminal justice process without VOM and to examine to what extent this explained lower reoffending rates.

Hypothesis

In this research the reoffending rates among four different offender groups is compared. Similar to the work of Jonas-van Dijk et al. (2020), three groups of offenders that were referred to VOM were examined in this study: (a) offenders who participated in VOM (mediation group), (b) offenders who were unwilling to participate in VOM (unwilling group), and (c) offenders who were willing to, but unable to participate, because the victim declined the option (unable group). Additionally, (d) a group of offenders who were not referred to VOM was also added to this study, which was composed using propensity score matching.

A comparison of these four groups might result in four patterns of findings. One potential pattern would be “only the mediation process explains lower reoffending rates.” For this pattern the reoffending rates should be lower among the mediation group compared to all the other groups and the other groups do not significantly differ. A second pattern would be “only a self-selection bias explains reduced reoffending.” In this case the willing offenders (the mediation and unable group) would have lower reoffending rates compared to offenders who were unwilling to participate. A third pattern, “selection at referral explains lower reoffending rates,” would mean that non-referred offenders have significantly higher reoffending rates compared to those offenders who were referred (mediation, unwilling, and unable group).

The fourth pattern, “self-selection bias and the VOM process itself explain lower rates of reoffending,” would show a pattern in which the mediation group has significantly lower reoffending rates compared to the unwilling and non-referred group of offenders. The unable group is expected to have reoffending rates in between the other groups, but do not significantly differ from these groups. This fourth pattern constitutes our hypothesis.

Method

VOM in Criminal Cases in the Netherlands

After a 4-year pilot, VOM has been implemented across all court jurisdictions in the Netherlands in 2017. This program is called mediation in criminal cases (Mediation in Strafzaken; MiS) and will be the program under investigation in this research. In 2021, 1.346 cases were referred to MiS nationwide. Of the cases in which mediation actually started (N = 787), 83% was successful (Raad voor de Rechtspraak, 2022). MiS is the largest mediation practice in the justice context in the Netherlands. 42% of all referred mediation cases within the justice context are cases referred to MiS (Hoekstra et al., 2024).

For MiS to be applied in a case, the public prosecutor or judge must refer a case to mediation. While offenders or victims can indicate their preference for mediation, the formal party that decides on referral is always the public prosecutor or judge. Once both parties agree to participate in MiS, the case is handed over to the mediation bureau of the court. A mediation officer from the bureau needs confirmation from both parties that they are (still) willing to participate. It sometimes happens that parties withdraw after receiving more information or after having had time to think about their participation. When both parties agree to participate, the mediation officer assigns two independent mediators to the case. There are no exclusion criteria in terms of type of case; in principle every type of case can be referred to mediation, but the referral party is always the public prosecutor or judge (Hoekstra et al., 2024). However, when an offender is denying any involvement or part in the criminal act, the case cannot be referred to mediation (Hoekstra et al., 2024; Uitslag & Kalter, 2024).

Before the actual encounter between the victim and the offender takes place, the mediators, often on the same day as the actual encounter, first have separate meetings with the offender and the victim, to explore if VOM would be beneficial for both parties. If both the parties and the mediators are convinced that a meeting would be helpful, the actual encounter takes place. Sometimes parties bring supporters to mediation, however the main focus of the VOM-encounter is on the victim and the offender. When the parties come to an agreement during mediation, this agreement is, with consent from the victim and the offender, communicated back to the referring agency. When a party does not want the mediator to communicate the content of the agreement, the only information communicated is then that a successful mediation process took place. The public prosecutor or judge consequently decides which punishment to impose, if one is (still) deemed necessary. This means that the referral agency has a final say in how the case will be solved. This implies that this type of VOM program does not constitute a diversion program, but is an integral part of the criminal justice process (Claessen & Roelofs, 2020). Participation in MiS and the outcome agreement are not intended to be an alternative sanction. However, taking into account the outcome agreement, the public prosecutor or judge can decide to dismiss the case. In a recent evaluation of the Dutch MiS program public prosecutors and judges indicated that participation in mediation often results in a lenient outcome for offenders: a lower sanction, a conditional sanction or a guilty verdict without sanction (Hoekstra et al., 2024).

Dependent Variable

Reoffending Rates

Different methods exist to examine reoffending. In this study the reoffending rates of different offender groups were examined, which should not be confused with risk of reoffending. Risk of reoffending can be defined as the possibility of an offender to reoffend, based on different risk factors, such as family circumstances, peer influence and personality (Villanueva et al., 2014). Reoffending rate is the actual observable reoffending behavior of an offender. In this study reoffending is measured 5.5 to 11 years after the case was referred to mediation as a dichotomous variable indicating whether the offender had been a suspect in a new criminal case or not (i.e., yes (one or more new criminal cases) versus no (no new criminal cases).

Type of Sanction

Type of sanction was derived from the recidivism data extracted from the WODC-recidivism monitor (Wetenschappelijk Onderzoek- en Datacentrum; Scientific Research- and Data Centre). It was categorical variable with 11 categories. However, due to low counts one some of these categories, we have merged categories to: unconditional custodial sentence, community service, conditional custodial sentence, money penalty, no sentence/measure, acquittal, technical dismissal and other/unknown, which included policy dismissals and guilty without sanction or measure.

Data Collection Procedure and Type of Offender Groups

The criminal cases that were selected for this study were referred to the mediation bureaus between October 2013 and October 2016 and only concerned cases with adult offenders. During this period, a pilot ran at six court jurisdictions in The Netherlands: Amsterdam, Breda, Den Bosch, The Hague, Noord-Holland, and Rotterdam. For all cases that contained a valid case-number and a date of birth of the offender, the researchers requested recidivism data (N = 582).

The encrypted dataset with the data retrieved from the mediation bureaus was sent to the WODC-recidivism monitor. The WODC-recidivism monitor is a longitudinal research project with the aim of unravelling the course of offenders and the effects of interventions to prevent crime and/or lower reoffending. Based on the case-number and date of birth of the offender, the WODC is able to retrieve the recidivism data from the “research and policy database judicial documentation” (Onderzoek- en Beleidsdatabase Justitiële Documentatie, OBJD). The OBJD is an encrypted copy of the official Dutch judicial documentation system (JDS). The JDS stores data of over three million persons and more than ten million criminal cases. New data is automatically added to the system. Every 3 months the OBJD is refreshed with the most recent data (Wartna et al., 2011). In addition to the recidivism data, the WODC also added the following information from the OBJD about the offender to our datafile: gender, age during MiS case, country of birth, age during first judicial contact, criminal record, type of offense, and type of punishment. The datafile was encrypted and sent back to the researchers. This resulted in a total N of 357 of which 221 offenders participated in VOM, 51 offenders were unwilling to participate in VOM, and 85 offenders were unable to participate in VOM, because the victim declined the option.

From the national pool of over three million offenders in the OBJD, the non-referred sample of offenders (group D) was also drawn. Preferably, we would have used a sample from the same court districts and same time period as the offenders who were referred to VOM during the pilot period. However, the WODC-recidivism monitor does not register if a criminal case has been referred to mediation or not. This means that we could not use a control group with cases from the same court districts in which the pilot ran, since cases in which mediation was offered could not be filtered out of the selection. In addition to that, the pilot ran in the districts with the highest crime prevalence and where the highest proportion of people with a migration background lived (AlleCijfers, 2024; CBS, 2023). As a consequence, this made it more difficult to match cases from districts in which VOM did not ran during that time period to cases from the pilot districts. Given these constraints, we decided to use two control groups. One control group in which cases were documented in the judicial database between October 2013 and October 2016, but only cases that were initially registered in the districts that were not part of the mediation pilots: Gelderland, Overijssel, Midden-Nederland, and Noord-Nederland. In this way, the changes are brought to a minimum that this control group included cases that were referred to mediation. The second control group consisted of cases that were documented in the judicial database between January 2011 and December 2012, right before the start of the mediation pilots. In this second group we only included the districts where the pilots did run later in time between October 2013 and October 2016. We reasoned that these two control groups enabled us best, given the above constraints, to compare the impact of a case being referred to mediation or not on the rates of reoffending.

Propensity Score Matching

Based on the above criteria, The WODC-recidivism monitor selected 277.030 cases that were not referred to mediation for the two control groups. Together with the cases that were referred to mediation (N = 357), this resulted in a total data set of 277.387 cases (i.e., 99.9% were cases not referred to mediation). Recidivism data were retrieved in March 2022, meaning that the period-at-risk analyzed for all cases varied between 5.5 and 11 years.

With t-tests and chi-square tests it was examined whether differences existed between the referred group of offenders and the non-referred groups of offenders in terms of offender and case-related variables. Table 1 illustrates that significant differences existed between the referred group and non-referred group on country of birth, number of previous contacts with the criminal justice system, age of the first contact with the criminal justice system, and the type of offense. Considering the imbalance in background variables and sample size between the groups, it was decided to use propensity score matching. With this method, the chance of being part of the referred (vs. non-referred)² group is estimated with a logistic regression analysis, including the background variables: gender, country of birth, criminal record, age of the first contact with the criminal justice system, age at moment of referral of the (non-) referred case, and type of offense. These background variables were based on research from Nagin (2009) in which it was argued that to have an acceptable comparison, variables that explain who the person is (in terms of age, race, and gender) and what the person has done (criminal history, and offence type) should be included in the matching process. In addition, according to Shadish (2013) and Shadish and Steiner (2010) the variables used for propensity score matching should actually predict the chance of getting assigned to the experimental group. Although this has not been examined before in the context of mediation, it is plausible—based on, amongst others, the study of Nagin (2009)—that the background variables used in our matching are predictive for being referred to mediation. However, we are aware that other predictive variables might exist, which we did not have in our current study. This will be elaborated on in the discussion.

Table 1.

Pre- and Post-matching Referred and Non-Referred Cases to Mediation t-Test and Chi-Square Test, Using Many-to-One (3–1) Matching.

	Unmatched			Matched
	Referred n = 357			Referred n = 357
Variable	Non-referred n = 276.793			Non-referred n = 2,509
T-test	F	t	p	F	t	p
Criminal record	36.84	4.38	<.001	1.09	.94	.17
Age first contact criminal justice system	5.38	−2.216	.013	.47	.02	.492
Age at moment of (non-)referral	2.01	−.09	.465	.04	−.27	.394
Average convictions per year	9.04	3.98	<.001	.98	1.21	.114
Chi-square test	X²	df	p	X²	df	p
Gender	.052	1	.820	.25	1	.616
Country of birth	34.05	7	<.001	23.19	7	.002
Type of offense	550.24	9	<.001	335.16	9	<.001

Note. p-values in bold are significant.

The score between 0 and 1 is called the propensity score: the higher the score, the higher the change of being referred to VOM. Based on the propensity scores three non-referred cases were matched to each referred case, to remain enough power to do the statistical analyses (Beal & Kupzyk, 2014). The caliper width was set to 0.0001, to get the best balance in background variables. A non-replacement method was used, which means that a non-referred case could only be matched once to a referred case. As can be seen in Table 1, matching led to more balance between the referred and non-referred group. However, on two variables a significant difference remained, namely on country of birth and type of offense. In the analyses that will follow in this study the influence of these variables was controlled for statistically. Cases that were not matched were excluded from data-analysis. Type of sanction was not used as confounding variable in the matching procedure, since type of sanctioning is used as an outcome variable as well in this study.

This study was preregistered at OSF. The expectations that were tested in this study are in accordance with the preregistration. However, at the preregistration we did not state that propensity score matching would be used. This was decided on later for multiple reasons. Only after receiving the non-referred control group, we observed the disbalance in background variables and in sample sizes. In addition, propensity score matching would result in even more robust outcomes and conclusions in comparison to the analyses that were planned without propensity score matching, because we would have a better (matched) control group.

Participants

After the matching procedure, the sample consisted of 2,907 offenders. Most of the offenders were male (N = 2,330, 80%) and were born in the Netherlands (N = 2,267, 78%). This ratio is also true for the four subsamples, as shown in Table 2. The mean age of offenders at the moment of referral was 33.4 (Range 12.52–85.42). On average, offenders had a criminal record with five offenses before the (non-)referred case. The mean age offenders had their first contact with the criminal justice system was 26. The mean age at the moment of non-referral was higher (M = 33). For the types of offense, a number of differences existed between the groups. For the total sample, most cases were cases of violence (N = 897, 31%) or property crimes without violence (N = 826, 28%). These cases were also largely present in the non-referred groups. In the mediation and unable group by far the largest group of cases were cases of violence (N = 164, 74% and N = 56, 66%). What stood out is that in the unwilling group 20% of the cases (N = 10) concerned cantonal facts (which are minor violations, such as speeding or vandalism).

Table 2.

Overview of Sample Characteristics of the Matched Sample.

	Total	Mediation	Unwilling	Unable	Non-referred
Variable	n = 2,907	n = 221	n = 51	n = 85	n = 2,508
Gender (male)	2,330 (80%)	184 (83%)^a*	38 (75%)^a	68 (80%)^a	2,009 (80%)^a
Country of birth
The Netherlands	2,267 (78%)	178 (81%)^a	41 (80%)^a	70 (82%)^a	1,946 (78%)^a
Morocco	76 (3%)	4 (2%)^a	—^a	4 (5%)^a	67 (3%)^a
Former Netherlands Antilles	59 (2%)	1 (0.5%)^a	—^a	—^a	55 (2%)^a
Suriname	74 (3%)	11 (5%)^a	2 (4%)^a	4 (5%)^a	56 (2%)^a
Turkey	53 (2%)	7 (3%)^a	1 (2%)^a	—^a	45 (2%)^a
Other Western Countries	197 (7%)	7 (3%)^a	7 (14%)^b	3 (4%)^a,b	176 (7%)^a,b
Other non-western countries	138 (5%)	13 (6%)^a	—^a	4 (5%)^a	120 (5%)^a
Unknown	43 (1%)	—	—	—	43 (2%)
Criminal record	4.7 (SD = 8.9)	3.8 (SD = 5.6)	4.6 (SD = 12.7)	5.3 (SD = 9.6)	4.7 (SD = 9.0)
Age first contact criminal justice system	26.0 (SD = 13.5)	27.0 (SD = 13.1)	27.7 (SD = 13.8)	22.3 (SD = 10.1)	26.0 (SD = 13.6)
Age at moment of (non-)referral	33.4 (SD = 14.7)	34.8 (SD = 14.4)	34.1 (SD = 13.6)	30.1 (SD = 12.6)	33.4 (SD = 14.8)
Type of offense
Violence (exc. sex offences and property crimes with violence)	897 (31%)	164 (74%)^a	25 (49%)^a,b	56 (66%)^b	626 (25%)^c
Sex offences	42 (1%)	1 (0.5%)^a	—^a	2 (2%)^a	39 (2%)^a
Property crimes with violence	43 (1%)	5 (2%)^{a, b}	—^a,b	7 (8%)^b	30 (1%)^a
Property crimes without violence	826 (28%)	10 (5%)^a	2 (4%)^a	1 (1%)^a	813 (32%)^b
Vandalism, aggression and public order	386 (13%)	26 (12%)^a	4 (8%)^a	9 (11%)^a	337 (13%)^a
Drugs	154 (5%)	1 (0.5%)^a	—^a,b	-^{a, b}	153 (6%)^b
Traffic	206 (7%)	6 (3%)^a	10 (20%)^b	4 (5%)^a	183 (7%)^a
Other	193 (7%)	3 (1%)^a	—^a,b	3 (4%)^{a, b}	187 (8)^b
Canton fact	147 (5%)	3 (1%)^a	10 (20%)^b	3 (4%)^a	129 (5%)^a
Unknown/not classifiable	13 (0.5%)	2 (1%)^a	—^a	-^a	11 (0.4%)^a
Average convictions per year		0.46 (SD = .44)	0.47 (SD = .40)	0.61 (SD = .50)	0.52 (SD = .46)

Each superscript letter denotes a subset of the four offender groups whose column proportions do not differ significantly from each other at the .05 level, based on the Bonferroni method.

Results

Controlling for Time at Risk, Demographic and Offense Related Variables

In total, 51% of the offenders reoffended. The recidivism rate of the mediation group (35%) was comparable to the rates of the unwilling group (37%). This rate seems considerably lower than for the unable group (53% reoffend) and the non-referred group (51% reoffended).

Since the time at risk varied between 5.5 and 11 years for the offenders, a cox survival analysis was used to statistically control for any variation in time at risk between the groups. The time at risk was set as time indicator. The status variable used was whether or not someone reoffended (i.e., was implicated in one or more new criminal cases or not). The four different groups of offenders were used as predictor. In this analysis we controlled for the demographic (gender, country of birth, age at moment of (non-)referral) and offense-related (criminal record, age first criminal contact, average convictions per year, and type of offense) background variables.

In the first step the background variables were entered as predictors in the model, which resulted in a significantly improved model compared to the model without predictors (X²(7, N = 2,542) = 767.52, p < .001). In the second step the offender groups were added as predictor to the model. Another significant improvement of the model was found (X²(3, N = 2,542) = 16.53, p = .009). A significant difference was found between the non-referred offender group and the mediation group (exp(B) = 1.55, p < .001). The non-referred group had a significant higher predicted rate of reoffending. The difference between the mediation and unable group was not significant (exp(B) = 1.37, p = .095). Also, no significant differences were found between the mediation group and the unwilling group (exp(B) = 1.27, p = .355), between the unable and unwilling group (exp(B) = .77, p = .923), or between the non-referred group and the unable (exp(B) =.89, p = .44) and unwilling group (exp(B) = .82, p = .386). The regression model can be found in Table 3.

Table 3.

Overview of the exp(B) and p Values of the Cox Survival Analysis Overview of the exp(B) and p Values of the Cox Survival Analysis Controlling for Time at Risk, Demographic and Offense-Related Variables.

Model	Exp(B)	Sig.	95% CI lower, 95% CI upper
Step 1
Gender (0 = male)	0.81	.009	[0.69, 0.95]
Country of birth^a		.232
The Netherlands	0.77	.242	[0.49, 1.17]
Morocco	0.74	.296	[0.43, 1.30]
Former Netherlands Antilles	1.11	.713	[0.63, 1.96]
Suriname	0.71	.234	[0.40, 1.19]
Turkey	0.60	.122	[0.32, 1.10]
Other Western Countries	0.90	.671	[0.55, 1.44]
Other non-western countries	0.91	.697	[0.55, 1.42]
Age at case of referral	0.99	<.001	[0.98, 0.99]
Age at first judicial contact	0.98	<.001	[0.97, 0.99]
Criminal history	1.00	.789	[0.99, 1.01]
Type of offense^b		<.001
Violence (exc. sex offences and property crimes with violence)	1.26	.651	[0.47, 3.38]
Sex offences	1.57	.422	[0.53, 4.67]
Property crimes with violence	1.80	.278	[0.62, 5.18]
Property crimes without violence	1.55	.383	[0.58, 4.18]
Vandalism, aggression and public order	1.32	.586	[0.49, 3.56]
Drugs	1.03	.956	[0.37, 2.83]
Traffic	1.04	.936	[0.38, 2.86]
Other	0.93	.903	[0.34, 2.59]
Canton fact	1.21	.718	[0.44, 3.33]
Average convictions per year	3.73	<.001	[3.21, 4.34]
Step 2
Offender group^c
Unable	1.40	.079	[0.93, 1.96]
Unwilling	1.07	.782	[0.78, 2.15]
Non-referred	1.42	.003	[1.16, 1.87]

Reference group = unknown.

Reference group = unknown/not classifiable.

Reference group = mediation group.

The Role of Sanctioning

Before we examine the impact of sanctioning on the predicted rates of reoffending, it was first examined with a chi-square test if differences exist in the type of sanctions the offender groups received. Testing this difference was not preregistered, but was an explorative analysis. This analysis was included, because we observed in the data differences in types of sanctioning, that might indicate that after a VOM process different types of sanctions are imposed.

This test showed to be significant (X² = 341.41, p < .001). Looking at the differences between the four groups, a Bonferroni test showed that the mediation group significantly more often did not receive a sanction, compared to the other three groups (Table 4). The mediation group also received community service less often, compared to the unable and non-referred group. The unwilling and non-referred group received monetary sanction more often and their cases resulted in technical dismissals more often compared to the mediation group. No other significant differences existed between the mediation and the other groups.

Table 4.

Overview of the Type of Sanction of the Matched Sample Based on the (Non-)referred Case.

Type of sanction	Total	Mediation	Unwilling	Unable	Non-referred
N	2,907	221	51	85	2,907
Unconditional custodial sentence	202 (7%)	8 (4%)^a*	—^a	6 (7%)^a	187 (8%)^a
Community service	690 (24%)	30 (14%)^a	11 (22%)^a,c	38 (45%)^b	600 (24%)^c
Conditional custodial sentence	63 (2%)	3 (1%)^a	—^a	—^a	59 (2%)^a
Money penalty	784 (27%)	15 (7%)^a	12 (24%)^b,c	13 (15%)^a,b	736 (29%)^c
No sentence/measure**	572 (20%)	140 (63%)^a	16 (31%)^c	18 (21%)^b,c	388 (16%)^b
Acquittal	163 (6%)	8 (4%)^a	4 (8%)^a	2 (2%)^a	145 (6%)^a
Technical dismissal	285 (10%)	6 (3%)^a	7 (14%)^b	5 (6%)^a,b	263 (11%)^b
Other/unknown^***	148 (5%)	11 (5%)^a	1 (2%)^a	3 (4%)^a	130 (5%)^a

Each superscript letter denotes a subset of the four offender groups whose column proportions do not differ significantly from each other at the .05 level, based on the Bonferroni method.

This includes policy dismissals and guilty without sanction or measure.

***

This includes the following sanctions: conditional measure, learning punishment, other transaction, and technical decision.

Another cox survival analysis was conducted in which next to the previously mentioned demographic and offense related variables, the type of sanctioning was controlled for. In the first step we again included the background and offense related variables. In the second step the offender groups were added as variable, just as with the previous cox survival analysis, resulting in a significant improvement of the model (X²(3, N = 2,542) = 10.97, p = .012). Lastly, type of sanction was then added to the model. This did also significantly improve the model (X²(7, N = 2,542) = 15.22, p = .033). When sanctioning was added to the model the patterns of differences between the offender groups remained unchanged. That is, offenders in the mediation group still had significant lower predicted reoffending rates compared to the non-referred group (exp(B) = 1.40, p = .008). But no other significant differences were found between the mediation group and the unwilling group (exp(B) = 1.28, p = .345), and between the mediation and the unable group (exp(B) = 1.29, p = .194). In Table 5 an overview is given of the outcomes of the last cox survival analysis. These findings indicate that sanctioning does not explain the difference we observed in reoffending between the mediation group and the non-referred group. The pattern of these findings, as shown in Figure 1, is mostly in accordance with our expectations: offenders who participated in VOM less often reoffended compared to offenders who were not referred to VOM and thus did not participate, and the unable group scored in between these two groups, but did not significantly differ from any of the groups. Only the reoffending rates of the unwilling group of offenders were not as expected.^3,4

Table 5.

Overview of the exp(B) and p Values of the Cox Survival Analysis Controlling for Time at Risk, Demographic, Offense-Related Variables, and Sanctioning.

Variable	Exp(B)	Sig.	95% CI lower, upper
Gender (0 = male)	.82	.014	[0.70, 0.96]
Country of birth^a		.225
The Netherlands	.77	.250	[0.49, 1.20]
Morocco	.73	.272	[0.42, 1.28]
Former Netherlands Antilles	1.13	.668	[0.64, 1.99]
Suriname	.70	.227	[0.40, 1.24]
Turkey	.60	.128	[0.32, 1.16]
Other western countries	.90	.679	[0.55, 1.47]
Other non-western countries	.90	.687	[0.54, 1.50]
Age at case of referral	.99	<.001	[0.98, 0.99]
Age at first judicial contact	.98	<.001	[0.97, 0.99]
Criminal history	1.00	<.62	[0.99, 1.01]
Type of offense^b		.002
Violence (exc. sex offences and property crimes with violence)	1.57	.379	[0.58, 5.25]
Sex offences	1.92	.245	[0.64, 5.78]
Property crimes with violence	2.11	.173	[0.72, 6.20]
Property crimes without violence	1.90	.207	[0.70, 5.16]
Vandalism, aggression and public order	1.65	.330	[0.60, 4.50]
Drugs	1.24	.675	[0.45, 3.45]
Traffic	1.29	.626	[0.47, 3.57]
Other	1.21	.711	[0.44, 3.39]
Canton fact	1.51	.434	[0.54, 4.21]
Average convictions per year	3.64	<.001	[3.13, 4.24]
Offender group^c
Unable	1.29	.194	[0.88, 1.88]
Unwilling	1.28	.345	[0.77, 2.13]
Non-referred	1.40	.008	[1.09, 1.79]
Type of sanction^d
Unconditional custodial sentence	.97	.820	[0.76, 1.25]
Community service	1.14	.554	[0.74, 1.74]
Conditional custodial sentence	.87	.283	[0.68, 1.12]
Money penalty	.80	.110	[0.61, 1.05]
No sentence/measure	.92	.596	[0.66, 1.27]
Acquittal	.82	.165	[0.61, 1.09]
Technical dismissal	1.23	.182	[0.91, 1.67]

Reference group = unknown.

Reference group = unknown/not classifiable.

Reference group = mediation group.

Reference group = other/unknown.

Figure 1.

Predicted risk of recidivism per offender group, controlled for the time at risk, demographic, and offense-related variables and type of sentence.

Discussion

This research had two aims. The first aim was to revisit the impact of participation in VOM on reoffending, using an improved and more rigorous research design compared to previous research. The second aim of this research was to examine if the nature and severity of sanctions after a VOM-process differ from the sanctions that are imposed in cases that go through a criminal justice process without VOM; and to what extent this explains lower reoffending rates.

The finding that offenders who participated in VOM have significant lower predicted rates of reoffending compared to offenders who were not referred to mediation is in accordance with previous research, which also shows that participation in VOM is related to reduced reoffending rates (e.g., Bergseth & Bouffard, 2007; Bouffard et al., 2017). Since this non-referred group of offenders was set up using propensity score matching and was matched on multiple variables that are considered to be essential for a good matching procedure, we tried to lower the risk of a confounding (self-)selection bias.

We hypothesized that the relation between participation in VOM and lower reoffending rates was due to self-selection processes as well as the mediation process itself. Our findings offer partial support for this hypothesis and suggest that the role of a self-selection bias might be smaller as initially thought based on research from Jonas-van Dijk et al. (2020). Jonas-van Dijk et al. (2020) concluded that lower reoffending rates are explained by both the self-selection bias and the VOM process based on the outcome that the offenders who participated in VOM had significantly lower rates of reoffending compared to the unwilling offenders and that the unable offenders scored in the middle and did not significantly differ from both groups. In this research we found that offenders who were unable to participate in VOM did not have a higher or lower reoffending rates compared to offenders who received VOM or offenders who did not receive VOM because they were not referred. However, in our study reoffending for the unable group does not fall in between the rates of the groups of offenders who were not referred and who participated in VOM. The rates for the unable group of offenders in this current study was more comparable to the offenders who were not referred to VOM. So, based on these outcomes it seems plausible that the relation between participation in VOM and lower reoffending in the current sample is mostly due to the VOM process itself and that the role of self-selection bias seems limited.

Unexpectedly, we did not find a difference between reoffending rates of offenders who participated in VOM and those who were unwilling to participate. This is not in accordance with findings from for example, Bergseth and Bouffard (2013), Kimbrell et al. (2023), and Stewart et al. (2018). What might be an explanation for this is the observation that a relatively high number of cantonal facts (violations that are considered minor criminal offenses) that were present in this “unwilling” group. Potentially these offenders were unwilling to participate because they thought their cases were too trivial and/or did not consider themselves to be “guilty”—as a requirement for participation in VOM is that an offender takes responsibility for the offense. Based on the type of offense in the unwilling group of offenders this seems to be a different type of offender group containing cases that might be less suited for mediation, which could explain why no difference was found between this group and the mediation group.

Role of Sanctioning

The second aim of this study was to examine if the nature and severity of sanctions after VOM differ from the sanctions that are imposed in a criminal justice process without VOM and to examine to what extent this can account for the lower reoffending rates observed after VOM. In previous research it has been suggested that the relationship between VOM and reoffending might potentially be explained by the decision to not impose a sanction after a successful mediation process (and agreements made) between parties (Jonas-van Dijk et al., 2020). When VOM is part of the justice system the agreement made between the offender and the victim is communicated to the public prosecutor or judge of case. This professional takes the agreement into account when deciding which sanction to impose, if any is still necessary.

First of all, this research indeed showed that sanctioning after VOM is different compared to cases in which VOM did not take place. After VOM more often it is decided not to impose a sanction or a policy dismissal follows. This suggests that VOM is considered as an alternative manner of dealing with a criminal case, instead of sanctioning. Second, and most importantly, when the type of imposed sanction was taken into account in the analyses, the patterns remained the same: offenders who participated in VOM still showed lower reoffending compared to offenders who were not referred to VOM, with again the unable group showing a higher but (non-significant) recidivism rate. This means that the relation between participation in VOM and lower reoffending is not explained by the (absence of or kind of) a sanction. Even when no additional sanction is given, offenders who participate in VOM have robust lower reoffending rates.

Limitations

Although propensity score matching used in this research resembles random allocation and this is to our knowledge the first study in the domain of VOM to have used this method to form a good comparison group, the quality of the matching depends on the background variables used in the matching procedure. In this research variables were included that explained who the person is (age, sex, race) and what the person has done (case type and criminal history). Research for example shows that minor cases are more often referred to mediation, as well as cases in which parties knew each other (Hoekstra et al. 2024). However, according to Nagin (2009) the matching procedure would have been even stronger when variables were included related to individual characteristics (mental, personality, or substance use disorders), since these variables might be related to the chance of being referred to mediation.

In this research we did not have access to or collected that kind of data. In this research only static variables were included, whereas dynamic risk factors should also be considered. These dynamic factors may offer an alternative explanation for reoffending rates, since they might impact the risk of reoffending. Information about someone’s life circumstances would have benefited the robustness of the matching procedure (Wermink et al., 2010). It could be that offenders in non-referred group had different mental, personality, or substance abuse disorders or different life circumstances compared to offenders who participated in mediation. For example, life course theory explains that certain characteristics in life contribute to the desistance process. These are strong social connections, structured routine activities and purposeful human agency (Sampson & Laub, 2017). Also, social control theory indicates the importance of social bonds in refraining people from (re-)offending (Hirschi, 2017).

In future research these different static and dynamic risk factors should be taken into account to draw stronger conclusions about the relation between participation in mediation and lower reoffending. When more variables are included that are known to be predictive for being assigned to the mediation group, propensity score matching would be more reliable (Shadish, 2013; Shadish & Steiner, 2010).

Related to the previous point, it is suggested to use a longitudinal cohort research design in future studies to examine if these different static and dynamic factors offer an alternative explanation for the findings. When these different factors are measured at multiple time points over a longer period of time, it will be become clearer which role these factors have, as well if they offer an alternative explanation for a lower reoffending after mediation. Again, it is important, just as in this current research, to have multiple groups of offenders that will be compared, to control for a (self-)selection bias and the role of sanction. This type of research design would offer stronger conclusion about the relationship between participation in mediation and a lower reoffending.

Implications

This study started with arguing that it is important for governments to respond to crime with programs that could possibly reduce reoffending rates. The findings of this study, as well as previous studies, show that VOM might be such a program. This study adds to the literature that VOM might result in less sanctioning. This also implies that VOM might result in lower costs, since sanctioning costs money. A crime reducing impact, in combination with less sanctioning, resulting in lower costs might be good reasons for governments to implement VOM as a restorative justice program in the criminal justice system and for judges and public prosecutors to refer to mediation. One critical note that should be made in this light, is that it is unknown if the fewer and less severe sanctioning we observed followed because of VOM or if, without the use of VOM, also fewer and less severe sanctioning would have occurred. It might be that minor cases that would already end with a dismissal are referred to VOM by the criminal prosecutor. Although VOM might then still impact the offenders, it is important for practitioners and policy makers to realize that VOM is a program that is not only effective in minor cases but in more severe cases as well (Umbreit et al., 1999; Zebel et al., 2017). Since VOM was found to be effective across different types of crime, including more severe cases, this suggest that VOM should not be solely applied in minor cases.

Even though this research is still limited in inferring causal relations, we believe that it constitutes a significant improvement compared to previous research. That is, the research design adopted here offers new insights and stronger conclusions regarding the favorable relationship between mediation and a lower reoffending than before. Importantly, the current study also suggest that this relationship cannot be explained by (the absence of) sanctioning. Less severe sanctions that are imposed in criminal cases where mediation was adopted compared to cases without mediation appears to be a parallel outcome to reduced reoffending.

Footnotes

Data Availability Statement

The anonymous dataset is available on request.

Declaration of Conflicting Interests

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

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was funded by a Research Talent Grant (406.17.555) from the Dutch Organization for Scientific Research (NWO).

Ethical Approval

Ethical approval for this study was obtained from the Ethics Committee of the Faculty of Behavioural, Management and Social Sciences at the University of Twente (File number: 191033) and from the Dutch public prosecution office (File number: 552007).

Informed Consent

Participants had to offer online written informed consent before participating in the study.

Pre-Registration

Pre-registration DOI:

ORCID iD

Jiska Jonas-van Dijk

Notes

Author Biographies

Jiska Jonas-van Dijk works as an assistant professor at the Department of Psychology of Conflict, Risk and Safety. In her research, she examines the dynamics of restorative encounters, the (psychological) impact of these encounters on parties, as well as the public perception of restorative justice. With her research she aims to contribute to a broader application of restorative justice.

Sven Zebel currently works as an associate professor at the Psychology of Conflict, Risk and Safety section at the University of Twente, and is endowed professor of mediation at the Vrije Universiteit Amsterdam. As a social and forensic psychologist, he aims to advance the scientific underpinnings and application of strategies that foster sustainable conflict resolution and emotional restoration after serious conflicts and crime.

Jacques Claessen is an endowed professor of restorative justice and an associate professor of criminal law at the Faculty of Law of Maastricht University, The Netherlands. He holds a degree in law (with honors) and a PhD at the same institution. He has a special interest in restorative justice, criminal sanctions, and the position of the victim in criminal law. Furthermore, he is editor-in-chief of the Dutch-Flemish legal periodical Tijdschrift voor Herstelrecht (Journal on Restorative Justice), an honorary judge at the District Court of Limburg and a guest lecturer at the Anton de Kom University of Suriname.

Hans Nelen is a professor of criminology at the Department of Criminal Law and Criminology in the Faculty of Law of Maastricht University, The Netherlands. He has an academic background in both criminology and law. He has conducted research and published extensively on a variety of criminological subjects, including drugs, corruption, fraud, organized crime, corporate crime, and the societal reaction to crime.

References

AlleCijfers. (2024, September 16). Ranglijst: Autochtoon en migratieachtergrond van de inwoners per provincie in Nederland. AlleCijfers.Nl [Ranking: Native Dutch and migration background of residents per province in the Netherlands]. https://allecijfers.nl/ranglijst/autochtoon-en-migratieachtergrond-per-provincie-in-nederland/#google_vignette

Beal

S. J.

Kupzyk

K. A.

(2014). An introduction to propensity scores: What, when, and how. The Journal of Early Adolescence, 34(1), 66–92. https://doi.org/10.1177/0272431613503215

Bergseth

K. J.

Bouffard

J. A.

(2007). The long-term impact of restorative justice programming for juvenile offenders. Journal of Criminal Justice, 35(4), 433–451. https://doi.org/10.1016/j.jcrimjus.2007.05.006

Bergseth

K. J.

Bouffard

J. A.

(2013). Examining the effectiveness of a restorative justice program for various types of juvenile offenders. International Journal of Offender Therapy and Comparative Criminology, 57(9), 1054–1075. https://doi.org/10.1177/0306624X12453551

Bernburg

J. G.

Krohn

M. D.

(2003). Labeling, life chances, and adult crime: The direct and indirect effects of official intervention in adolescence on crime in early adulthood. Criminology, 41(4), 1287–1318. https://doi.org/10.1111/j.1745-9125.2003.tb01020.x

Bohmert

M. N.

Duwe

Hipple

N. K.

(2018). Evaluating restorative justice circles of support and accountability: Can social support overcome structural barriers? International Journal of Offender Therapy and Comparative Criminology, 62(3), 739–758. https://doi.org/10.1177/0306624X16652627

Bouffard

Cooper

Bergseth

(2017). The effectiveness of various restorative justice interventions on recidivism outcomes among juvenile offenders. Youth Violence and Juvenile Justice, 15(4), 465–480. https://doi.org/10.1177/1541204016647428

Bradshaw

Roseborough

(2005). Restorative justice dialogue: The impact of mediation and conferencing on juvenile recidivism. Federal Probation, 69(2), 1–9.

Braithwaite

(1998). Restorative justice. In Tonry

(ed.), The handbook of crime and punishment (pp 323–344). Oxford University Press.

10.

Braithwaite

(2001). Reintegrative shaming. In Paternoster

Bachmann

(Eds.), Explaining criminals and crime: Essays in contemporary criminological theory (pp. 242–251). Roxbury.

11.

Braithwaite

(2003). The fundamentals of restorative justice. In Dinnen

(Ed.), A kind of mending (pp. 35–44). ANU Press.

12.

Carlsmith

K. M.

Darley

J. M.

Robinson

P. H.

(2002). Why do we punish? Deterrence and just deserts as motives for punishment. Journal of Personality and Social Psychology, 83(2), 284–299. https://doi.org/10.1037/0022-3514.83.2.284

13.

CBS. (2023). Geregistreerde criminaliteit; soort misdrijf, regio. https://opendata.cbs.nl/statline/#/CBS/nl/dataset/83648NED/table?ts=1687354531063

14.

Cerda-Jara

Harding

D. J.

(2024). Criminal record stigma in the labor market for college graduates: A mixed methods study. Sociological Science, 11, 42–66.

15.

Claessen

(2019). Theories of punishment. In Keiler

Roef

(Eds.), Comparative concepts of criminal law (pp. 11–34). Intersentia.

16.

Claessen

Roelofs

K. J. M.

(2020). Herstelrecht(voorzieningen) en mediation in strafzaken [Restorative justice (services) and mediation within criminal cases]. In Boksem

Mevis

P. A. M.

Paridaens

D. J. M.

Waling

Woslwijk

H. D.

(Eds.), Handboek Strafzaken—online. Wolters Kluwer.

17.

Claessen

Zeles

Zebel

Nelen

(2015). Bemiddeling in strafzaken in Maastricht II. Onderzoek naar de samenhang tussen bemiddeling en recidive [Mediation within criminal cases in Maastricht II. Examining the relation between mediation and recidivism]. Nederlands Juristenblad, 29.

18.

D’Souza

L’Hoiry

(2019). An area of untapped potential? The use of restorative justice in the fight against serious and organized crime: A perception study. Criminology and Criminal Justice, 21(2), 224–241. https://doi.org/10.1177/1748895819858379

19.

Gausel

Vignoles

V. L.

Leach

C. W.

(2016). Resolving the paradox of shame: Differentiating among specific appraisal-feeling combinations explains pro-social and self-defensive motivation. Motivation and Emotion, 40(1), 118–139. https://doi.org/10.1007/s11031-015-9513-y

20.

Hansen

Umbreit

(2018). State of knowledge: Four decades of victim-offender mediation research and practice: The evidence. Conflict Resolution Quarterly, 36(2), 99–113. https://doi.org/10.1002/crq.21234

21.

Hirschi

(2017). On the compatibility of rational choice and social control theories of crime. In Cornish

Clarke

(Eds.), The reasoning criminal (pp. 105–118). Routledge.

22.

Hoekstra

M. S.

Bak

R.-D.

Teeuwen

(2024). Plan-en procesevaluatie Mediation in Strafzaken. WODC

23.

Hosser

Windzio

Greve

(2008). Guilt and shame as predictors of recidivism: A longitudinal study with young prisoners. Criminal Justice and Behavior, 35(1), 138–152. https://doi.org/10.1177/0093854807309224

24.

Jonas-van Dijk

. (2024). Opening the black box of victim-offender meiation: Does participation in VOM reduce offenders’ reoffending and, if so, how? Maastricht University.

25.

Jonas-van Dijk

Zebel

Claessen

Nelen

. (2020). Victim–offender mediation and reduced reoffending: Gauging the self-selection bias. Crime and Delinquency, 66(6–7), 949–972. https://doi.org/10.1177/0011128719854348

26.

Jonas-van Dijk

Zebel

Claessen

Nelen

. (2022). The psychological impact of participation in victim-offender mediation on offenders: Evidence for increased compunction and victim empathy. Frontiers in Psychology, 12(2022), 6531. https://doi.org/10.3389/fpsyg.2021.812629

27.

Kimbrell

C. S.

Wilson

D. B.

Olaghere

(2023). Restorative justice programs and practices in juvenile justice: An updated systematic review and meta-analysis for effectiveness. Criminology and Public Policy, 22(1), 161–195. https://doi.org/10.1111/1745-9133.12613

28.

Latimer

Dowden

Muise

(2005). The effectiveness of restorative justice practices: A meta-analysis. The Prison Journal, 85(2), 127–144. https://doi.org/10.1177/0032885505276969

29.

Lauwaert

Aertsen

(2016). With a little help from a friend: Desistance through victim-offender mediation in Belgium. Restorative Justice- an International Journal, 4(3, SI), 345–368. https://doi.org/10.1080/20504721.2016.1245913

30.

Nagin

D. S.

Cullen

F. T.

Jonson

C. L.

(2009). Imprisonment and reoffending. Crime and Justice, 38(1), 115–200.

31.

Nur

A. V

Monaghan

(2024). Occupational attainment and criminal justice contact: Does type of contact matter? Crime & Delinquency, 70(2), 573–600.

32.

Raad voor de Rechtspraak. (2022). Jaarverslag 2021. https://www.rechtspraak.nl/SiteCollectionDocuments/Jaarverslag-Rechtspraak-2021.pdf

33.

Sampson

R. J.

Laub

J. H.

(2017). A general age-graded theory of crime: Lessons learned and the future of life-course criminology. In Farrington

(Ed.), Advances in criminological theory: Testing integrated developmental/life course theories of offending (Vol. 13, pp. 165–181). Transaction Publishers.

34.

Shadish

W. R.

(2013). Propensity score analysis: Promise, reality and irrational exuberance. Journal of Experimental Criminology, 9, 129–144.

35.

Shadish

W. R.

Steiner

P. M.

(2010). A primer on propensity score analysis. Newborn and Infant Nursing Reviews, 10(1), 19–26.

36.

Stewart

Thompson

Beaudette

J. N.

Buck

Laframboise

Petrellis

(2018). The impact of participation in victim-offender mediation sessions on recidivism of serious offenders. International journal of offender therapy and comparative criminology, 62(12), 3910–3927. DOI: https://doi.org/10.1177/0306624X1775227

37.

Uggen

(2000). Work as a turning point in the life course of criminals: A duration model of age, employment, and recidivism. American Sociological Review, 67, 529–546.

38.

Uitslag

Kalter

(2024). Het nemen van verantwoordelijkheid door de verdachten [offender responsibility taking]. In van Mazijk

(Ed.), Handboek mediation in strafzaken: Handleiding voor de strafmediationpraktijk (2nd ed., p. 184). Hogeschool Utrecht.

39.

Umbreit

M. S.

Bradshaw

Coates

R. B.

(1999). Victims of severe violence meet the offender: Restorative justice through dialogue. International Review of Victimology, 6(4), 321–343. https://doi.org/10.1177/026975809900600405

40.

Urban

L. S.

Burge

S. E.

(2006). Victim/offender mediation in St. Louis: An assessment. Justice Research and Policy, 8(2), 89–114.

41.

Villanueva

Jara

García-Gomis

(2014). Effect of victim-offender mediation versus dispositions on youth recidivism: The role of risk level. Journal of Forensic Psychology Practice, 14(4), 302–316. https://doi.org/10.1080/15228932.2014.950508

42.

Wartna

B. S. J.

Blom

Tollenaar

(2011). De WODC-Recidivemonitor-4e herziene versie. WODC.

43.

Wermink

Blokland

Nieuwbeerta

Nagin

Tollenaar

(2010). Comparing the effects of community service and short-term imprisonment on recidivism: A matched samples approach. Journal of Experimental Criminology, 6(3), 325–349.

44.

Wong

Bouchard

Gravel

Bouchard

Morselli

(2016). Can at-risk youth be diverted from crime?: A meta-analysis of restorative diversion programs. Criminal Justice and Behavior, 43(10), 1310–1329. https://doi.org/10.1177/0093854816640835

45.

Zebel

Schreurs

Ufkes

E. G.

(2017). Crime seriousness and participation in restorative justice: The role of time elapsed since the offense. Law and Human Behavior, 41(4), 385–397. https://doi.org/10.1037/lhb0000242

46.

Zehr

(2015). The little book of restorative justice: revised and updated. Good Books.

Participation in Victim-Offender Mediation Predicts Less Reoffending: A Propensity Score Matched Case-Control Study

Abstract

Keywords

Introduction

Hypothesis

Method

VOM in Criminal Cases in the Netherlands

Dependent Variable

Reoffending Rates

Type of Sanction

Data Collection Procedure and Type of Offender Groups

Propensity Score Matching

Participants

Results

Controlling for Time at Risk, Demographic and Offense Related Variables

The Role of Sanctioning

Discussion

Role of Sanctioning

Limitations

Implications

Footnotes

Data Availability Statement

Declaration of Conflicting Interests

Funding

Ethical Approval

Informed Consent

Pre-Registration

ORCID iD

Notes

Author Biographies

References