Do Victim Impact Panels Have Sustained Effects on DUI Recidivism?

This study examines whether Victim Impact Panels reduce DUI recidivism 5 to 8 years postsentence. Original 2-year data on 410 DUI offenders who attended a Victim Impact Panel and 373 DUI offenders from the same court system who did not attend a Victim Impact Panel indicated slight recidivism reduction effects. Logistic regression results at the 5-year period reveal that not attending a Victim Impact panel increases the odds of another DUI by a factor of 1.5 with an upper 95% confidence odds level of 2.2. At the 8-year mark, non-VIP participants were 1.8 times more likely to record another DUI with an upper 95% confidence odds level of 2.6. Offenders with a prior DUI who attended a Victim Impact Panel had significantly lower recidivism rates after 8 years relative to non-attendees. Prior research suggested that males in the 26–35-year age group benefitted more from Victim Impact Panel participation. These data indicate that the effects of age are relatively uniform across male age groups. Overall, these recidivism results indicate that the emotional messages communicated by victims to DUI offenders might carry sustained effects.


Introduction
In 2019, the National Highway Traffic Safety Administration (NHTSA) reported 9943 alcohol-impaired driving fatalities and close to 300,000 road-related alcohol-impaired injuries (National Highway Traffic Safety Administration 2020). In its "Report to the Nation", MADD, (Mothers Against Drunk Driving) reported that drunk driving statistics remain stubbornly unmovable with surveys continuing to show that over 300,000 Americans drive drunk every day (Mothers Against Drunk Driving 2019). As a result, policy experts and special interest groups worry that efforts to prevent drunk driving have become stagnant (Mothers Against Drunk Driving 2019).
In the past few decades, the justice system has increasingly responded to lawbreaking behavior such as drunk driving by giving voice to the concept of reintegrative shame, the intent of which is to introduce guilt and remorse in offending populations (Harris 2006). One process, referred to as a Victim Impact Panel (VIP), is facilitated by allowing victims to communicate to offenders how their actions produced harm in their lives. While Victim Impact Panels are somewhat compatible with the tradition embodied in restorative justice and reintegrative shaming, they differ in singularly giving voice to the victim. Thus, the most the most common employment of Victim Impact Panels has been geared toward combatting drunk driving (Shinar and Compton 1995;Fors and Rojek 1999;Polacsek et al. 2001).
In Victim Impact Panels, convicted drunk drivers are sentenced to attend a session in which 4-5 prior victims communicate how a drunk driving incident affected their lives. Speakers can include individuals who have lost loved ones in a DUI accident. This exposure is intended to personalize the loss experienced by victims. The nonconfrontational setting is designed to allow victims to communicate the depth of trauma and grief they experienced at the hands of a drunk driver. In so doing, VIPs expose the emotional, physical and financial consequences of drinking and driving to offenders (Fors and Rojek 1999).
In traditional court sentencing, the sanctions of fines and fees imposed by the courts are theorized to be impersonal, with the state functioning as the offended party (Rojek et al. 2003). Offenders are theorized to feel empowered by VIP sessions through a process that makes them feel reintegrated, as opposed to stigmatizing them as drunken wrongdoers. Through this process, offenders are urged to consider modeling more responsible driving behavior in the future. Consequently, VIPs target future drinking and driving behavior among offenders. Embodied in the messages communicated by victims are emotions which allegedly should function to influence the cognition of offenders and deter them from drinking and driving.
The chief cheerleader for VIPs has been Mothers Against Drunk Driving (MADD). The MADD website includes emotional testimony from offenders regarding the impact victim stories have had on their lives. One of these statements is, "it only took a few hours but I will remember their messages forever." Another example is, "I didn't see how many people my actions impacted until I heard from the victims" (https://maddvip.org/, accessed on 14 August 2021). In short, MADD has fostered the idea that VIPs are an effective vehicle for reducing the probability that offenders will drive drunk in the future.
In their systematic review of DUI interventions, Miller et al. (2015) reported mixed findings regarding the effect of VIPs on recidivism. Their review identified seven quasiexperimental studies. These studies included a quasi-experiment by Fors and Rojek (1999) who found that VIP attendees recorded a 10% lower rearrest rate than non-VIP attendees. This effect was particularly pronounced among white males in the 26-35-year age group and appeared to be more effective among drivers with one previous DUI arrest. Sprang (1997) measured 1-year recidivism rates comparing a VIP group to a comparison group and revealed a 9% lower recidivism rate among VIP participants. Our own previous study reported a significant reduction in the probability of a DUI 2 years later among VIP attendees compared to non-VIP attendees (Joyce and Thompson 2017).
The Miller et al. (2015) review included several studies in which the investigators reported no significant differences between a VIP and comparison group. One study, encompassing a 2-year follow-up of VIP attendees, a matched comparison group, and VIP no-shows (Shinar and Compton 1995). The investigators scrutinized the driving records of 2092 offenders in Oregon and California and found no differences among the three groups (Shinar and Compton 1995). A smaller study of 100 DUI offenders in New Mexico also failed to find significant differences between a standard DUI sanctioned group and a VIP group (Wheeler et al. 2004).
For the studies for which we were able to identify VIP outcomes (including those cited by Miller et al. 2015), the average follow-up period was 3 years. Thus, we know very little about the long-term recidivism impact of Victim Impact Panel exposure beyond 2 to 3 years. Some studies in the recidivism literature reveal that longer-term follow-up periods increase the effect size in favor of the programmatic element (Nagayama Hall 1995). Other studies indicate a pattern of diminishing returns of programmatic effects on compliant behavior (Ray 2014;Johnson 2004). Ray's (2014) study of a mental health court revealed that a majority of both completers and non-completers of the program had recidivated by year 3. Krebs et al.'s (2007) study of drug courts also reported short-lived program effects at around 18 months. In a study of prison fellowship programs, Johnson (2004) similarly reported no group differences between participants and non-participants at the 8-year mark. Other studies indicate that, controlling for age, there is a point in time when the risk level of a person with a criminal record converges with the risk level of a person with a prior criminal record (Blumstein and Nakamura 2009;Kurlycheck et al. 2007).
If victims' stories create empathy and vicarious grief, they should have a lasting impact on drunk drivers. It is not unusual during a VIP session for a victim to break down in describing the loss they experienced at the hands of a drunk driver. Indeed, surveys of VIP participants demonstrate that attendees reported feeling increased empathy for victims of drunk drivers following their participation (Schaaf 2008). Other studies have shown that attendees report a lower likelihood of drinking and driving in the future (Sprang 1997;Badovinac 1994). Nevertheless, the courts are more intent on changing drinking and driving behavior than in transforming attitudes. Thus, recidivism continues to be the VIP measure of intended outcomes.

The Current Study
This study examines the sustained effectiveness of a Victim Impact Panel in North Dakota and its role in deterring DUI recidivism. Rural states such as North Dakota tend to struggle with controls on drunk driving, as in rural states, DUI enforcement may be less visible and the volume of driving, often on deficient roads, can be problematic (Miller and Zaloshnja 2009). One report on per capita alcohol indicators placed North Dakota near the top in terms of alcohol consumption and sales of alcohol (Haughwout et al. 2015). According to the Behavioral Risk Factor Surveillance System survey (Center for Disease Control and Prevention 2014), North Dakotans are roughly twice as likely to report driving after drinking too much than other residents nationally. A website that employed both FBI and U.S. Department of Transportation statistics placed North Dakota's DUI severity score at 10.6 compared to the national average of 6.6 (Best and Worst States for Drunk Driving 2018) https://backgroundchecks.org/which-states-have-the-worst-dui-problems.html, accessed on 14 July 2021. Indeed, in 2019, North Dakota led the nation in alcohol-impaired driving fatalities, at 41% of all fatalities. In contrast, New Jersey, which is a relatively urban state, reported an alcohol-impaired driving fatality rate of 23% (U.S. Department of Transportation 2020).
Consequently, one of the ways in which the present study may differ from the earlier Miller et al. (2015) review is our focus of the effect of a Victim Impact Panel in a more rural jurisdiction, situated in a state with higher prevalence and incidence estimates of drunk driving. The authors were initially interested in assessing the association between VIPs in this jurisdiction and DUI arrests in 2015 (Joyce and Thompson 2017). That evaluation revealed a reduction in DUIs of VIP participants relative to non-participants of 162% after 1 year and 77% after 2 years. The fact that the differences between the VIP group and the comparison group declined the second year suggests that the effect of VIPs on drunk driving convictions may not have a lasting impact.
Based on previous Victim Impact Panel studies, we assess the following research questions: (1) Does the Victim Impact Panel have any effect toward reducing the probability of a DUI conviction beyond the period of 2 years? (2) Does the Victim Impact Panel have any DUI reduction effect for offenders with a prior DUI? (3) Is there a gender and age subgroup that benefits the most from VIP exposure?

Method
Data for this study were gathered with assistance from the Public Health Department and the District Court system in the county. The Judicial District encompasses three counties and employs nine judges. The population for the jurisdiction is a little more than 200,000 and most of its residents are white. The jurisdiction includes a modestly sized university, enrolling approximately 12,000 students.
The Public Health Department scheduled Victim Impact Panels on the second Wednesday of every month. DUI offenders were required to register for a VIP session to fulfill their sentence. Participants were charged USD 50 which was due at sign-in. Prior to COVID-19, each session lasted roughly 1 h and 45 min. A variety of different victims of drunk driving incidents were listed as the main speakers at these sessions. Participants were required to attend the entirety of the presentation, lest they risk losing court credit. The VIP Coordinator then notified the courts as to the status of each participants' attendance.
In the first iteration of this study (Joyce and Thompson 2017), we discovered that a comparison group existed that could be employed to assess the impact of the VIP. Among the nine criminal sentencing judges in the jurisdiction, one judge elected not to sentence DUI offenders to a Victim Impact Panel because he did not feel that this program would be an effective vehicle for preventing further DUIs. Thus, during this period of time in which he adjudicated on the criminal bench (2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014), he simply sentenced DUI offenders to probation and fined them. Because criminal cases are assigned on a rotating basis by the court clerks, we can assume that the any differences between the two groups would be random.
When the study commenced in 2015, we were provided with a list of VIP attendees by the Public Health Department. Arrest dates for this group ranged from April 2006 to September 2014. VIP attendance dates ranged from January 2009 to April 2015. A total of 1739 offenders appeared in the Public Health register as having attended a VIP group during this period of time. Because of the large number of attendees, a decision was made to employ systematic sampling to whittle down this list and more closely approximate the number of offenders in the comparison group. Consequently, a decision was made to select every fourth VIP case from the chronological list. This systematic sampling method resulted in a total of 410 VIP cases selected for analysis.
The comparison group consisted of DUI offenders from the same court jurisdiction who were not sentenced to attend a Victim Impact Panel during the same time period. This list consisted of 373 cases. Comparison group offender's complaint dates ranged from April 2004 through January 2014. Thus, a total of 783 DUI offenders were available for analytical purposes.

Control Variables
As a result of the Public Health and District Court data set, we were able to glean the following data regarding offender characteristics: (1) Demographics (gender, age at complaint date); (2) Criminal background-prior DUI (0, 1), number of prior DUIs, criminal conviction at the Class B misdemeanor level or higher (0, 1), and number of previous criminal convictions.

Outcome Variable-Recidivist DUI Conviction
Recidivism was defined as a new DUI conviction following either VIP attendance or initial sentencing for the non-VIP group. All criminal records for this study were publicly available online. Institutional Review Board approval was not needed due to the data being obtained through public records. Online public records searches were conducted using both North Dakota (http://publicsearch.ndcourts.gov/default.aspx, accessed on 14 November 2021) and Minnesota (http://pa.courts.state.mn.us/Search.aspx?ID=100, accessed on 14 November 2021) data bases due to the North Dakota county being located adjacent to Minnesota. The public records search provided conviction dates and sentencing dispositions. For this second iteration, data-gathering efforts ceased in February of 2020.
Our 2017 study examined DUI recidivism between 6 months and 2 years, following VIP attendance or complaint date for non-VIP attendees. For the present study, we were able to elongate the follow-up period to 5 and 8 years following VIP attendance or complaint date (comparison group). To achieve this, we allowed all 783 DUI offenders in the study to have the same timeline opportunity to acquire another DUI conviction: 2, 5, and 8 years following either their attendance at a VIP program or the sentencing date for the comparison group. Thus, time was relatively controlled in the analysis. Table 1 displays the data for the entire sample. A little more than three-fourths (75.6%) of the sample consisted of males. A little less than half (42%) of the sample recorded a previous DUI (e.g., previous to their initial court date in this study). In total, 69.3% of the sample recorded a previous criminal conviction for any type of crime. The mean age at the time of arrest/complaint was 33.8 years of age (range = 18-73). At the 2-year mark, a little more than 11% of the sample recorded another DUI. At the 5-year mark, the DUI recidivism rate was almost 21% and at the 8 year mark, almost one-fourth (24.8%) of this sample recorded another DUI. Thus, it is evident that another DUI conviction was fairly common following a DUI court appearance. The question we are seeking is whether the VIP recidivism rates differ significantly from the comparison group. In our initial study, we noted several significant differences between the two groups, despite the lack of intentional selection bias. Table 2 displays the percentages and means for each of these groups. There were roughly equal percentages of males and females in each group (chi-square = 2.84, 1 df, probability = 0.09). At the time of the initial DUI complaint, there were no significant age differences between the two groups. The comparison group recorded a slightly higher prevalence of having been convicted of a prior DUI by 12% (chi-square = 13.4, 1 df, p < 0.001). The comparison group also recorded a higher percentage of previous criminal convictions by 11% (chi-square = 10.02, 1 df, p < 0.001). Consequently, the comparison group could exhibit biased recidivism estimates based on their prior DUI and criminal history. As a result, these variables will be controlled in the logistic regression analysis.

Results
We begin with the bivariate recidivism results. Table 3 displays the associations between group assignment and DUI recidivism at 2, 5, and 8 years. At 2 years, the VIP group recorded a 6% lower DUI recidivism rate than the comparison group. Translated to an odds ratio, this means that the VIP group recorded a 1.88 lower odds of recidivism. This difference is significant at the alpha 0.01 level. At 5 years, the group difference expands to 9% with an odds ratio of 1.73 and a significance level of 0.01. In examining the 8-year recidivism rate, we note that the VIP group recorded a 13-point lower DUI recidivism rate with an odds ratio of 2.04. This difference is significant at the alpha 0.001 level. Thus, it appears that the VIP program is having its greatest bivariate impact beyond 5 years. Nevertheless, evidence that the comparison group had higher odds of a previous DUI as well as criminal convictions could be facilitating these differences. To test this possibility, logistic regression equations were run to estimate the log odds of another DUI conviction controlling for gender, age, prior DUI and prior criminal conviction. These equations were run using stepwise procedures. In the first step, gender and the current age of the offender was entered. In step two, we entered the variables for prior DUI and prior criminal conviction. The group assignment variable (VIP or comparison group) was entered at step three. Prior to running these equations, we assessed whether our models were unbiased. We began by testing whether our three continuous variables (age, prior DUIs, and prior criminal convictions) were linearly associated with the log of the three recidivism outcomes. To assess linearity, we entered the three continuous variables and then created interaction terms with the log of itself. These data (not shown) reveal that none of the nine interaction terms for the three recidivism outcome years were statistically significant. This indicates that these variables are linearly associated with recidivism. 1 Table 4 displays the log odds of DUI recidivism at 2 years, 5 years, and 8 years. While variables were entered stepwise, we only present the final model in the table to conserve space. At 2 years, gender was significant at the alpha 0.05 level (b = 0.60, SE = 0.30) but not age at arrest. At step two, the variables for prior DUI and prior criminal conviction revealed that having a prior criminal conviction was significantly associated (b = 1.51, SE = 0.39) with another DUI but having a prior DUI did not significantly elevate recidivism odds. When the group assignment variable was entered at step three, we observed that not attending a VIP program raised the odds of another DUI by a factor of 1.72 over those who attended a VIP (b = 0.540, SE = 0.24). The 95% confidence interval indicates that these odds varied between 1.08 and 2.73. Adding the group assignment variable also significantly improved upon the fit of the model over step 2 (chi-square = 27.1 − 22.7 = 4.4, 1 df, p < 0.05). At 5 years, we observe similar patterns. The only significant variables in the final model was whether or not the offender had a prior criminal conviction (b = 1.19, SE = 0.28) and whether or not they attended a VIP program (b = 0.44, SE = 0.18). At 5 years, those not attending a VIP had recidivism odds that were 1.5 times greater than those who attended a VIP. Adding the group assignment variable at step three improved upon the model fit over step two (chi-square = 37.1 − 32.4 = 4.7, 1 df, p < 0.05). At 8 years following VIP attendance or arrest date (comparison group), the only significant variables were again prior criminal conviction (b = 0.95, SE = 0.26) and group assignment (b = 0.61, SE = 0.17). Offenders who were not court-ordered to attend a VIP were 1.8 times more likely to record another DUI after 8 years. Entering the group assignment variable at step three again improved upon the model fit (chi-square = 56.9 − 45.9 = 11.0, 1 df, p < 0.001). Thus, these data suggest that there is a recidivism benefit when judges sentence DUI offenders to attend a Victim Impact Program.
We also wished to run logistic regressions while controlling for the number of prior DUIs and criminal convictions in each offender's record. We felt that because offenders in the comparison group had more lengthy DUI and criminal histories, their prior offenses were likely inflating the estimates for DUI recidivism. Consequently, we ran similar equations at 2, 5 and 8 years using the same stepwise ordering. These patterns were identical to the patterns in which DUI and criminal conviction were entered as binary variables. Not attending a Victim Impact Panel raised the odds of 2-year, 5-year, and 8-year DUI recidivism by 1.64, 1.50, and 1.79, respectively. The Nagelkerke R value (0.091) was also highest at year 8, indicating that the contribution of these variables to recidivism was greater after 8 years. Table 5 presents data on offenders who had recorded a previous DUI prior to the study. The Fors and Rojek (1999) study suggested that the VIP program could be effective for offenders with a prior DUI. There were 329 offenders who qualified for this analysis. We were interested in whether the VIP program had a short-, moderate-, or long-term impact on this group, relative to non-VIP attendees. For these equations, we controlled for the frequency of prior criminal convictions as well as gender and age at arrest. The logit regressions clearly show a monotonic pattern with the prior DUI group. At 2 years, the VIP program failed to differentiate DUI odds between the two groups (b = 0.37, SE = 0.34, p = 0.27). At 5 years, the differences between the groups become more marked but the coefficient barely failed to reach significance (b = 0.49, SE = 0.26, p = 0.059). However, at 8 years, offenders attending the VIP program recorded significantly lower DUI odds than non-attendees (b = 0.76, SE = 0.25, p = 0.003). The exponent for this coefficient indicates that the odds of acquiring another DUI conviction were 2.1 times greater for non-VIP offenders than for VIP attendees. Thus, it appears that as time elapses, the magnitude of the beneficial effects of the Victim Impact Panel increases among a group of offenders who recorded a previous DUI. Finally, we wished to ascertain whether the Victim Impact Panel program exhibited larger effects for a specific age and gender group. Fors and Rojek (1999) reported that males in the 26-35-year age group benefitted the most from VIP exposure. To test this assumption, we dummy-coded age among males by uncoupling 26-35-year-olds from all other age groups. We then entered the main effects for age at complaint date, group assignment, and then interaction terms for group assignment by age. These data (not shown) indicate that none of the group assignment by age interaction terms was significant for any of three recidivism outcome measures. Consequently, we surmise that the effects of the VIP program are roughly uniform across age groupings among males.

Discussion
This study of a Victim Impact Panel was carried out in a more rural jurisdiction in a state that has traditionally wrestled with a large volume of drinking and driving. We wished to determine whether DUI offenders exposed to a Victim Impact Panel were able to avoid an additional DUI arrest after participating in a VIP 5 to 8 years earlier. To date, little is known as to whether the stories conveyed by victims to DUI offenders carry weight over a sustained period of time. While the short-term evaluations of Victim Impact Panels reveal mixed results, we lack data on the effect of VIP programs beyond 2-3 years.
The data employed in this study show that not attending a Victim Impact Panel program raised the odds of a subsequent DUI by 1.5 times at the 5-year mark and by 1.8 times at the 8-year mark. Stated another way, attending a Victim Impact Panel reduced the odds of a subsequent DUI by 49% and 77% at 5 and 8 years, respectively. Similar patterns emerged at the 8-year mark for offenders who had recorded a prior DUI. While these overall effects are fairly modest, they mirror the effect size of a meta-analytic review of 32 restorative justice programs (Latimer et al. 2005). While we acknowledge that Victim Impact Panels may not be completely compatible with restorative justice procedures, it is of interest that the phi coefficient effect size of 0.07 in the studies reviewed by Latimer et al. (2005) is similar to our phi coefficients of 0.09, 0.11, and 0.15 at the 2-year, 5-year, and 8-year mark, respectively. Why the effect sizes might increase over time is subject to speculation, but it is possible that VIP attendees could be more prone to seek chemical dependency treatment over time. If so, this could partially account for the expanded difference.
Regarding gender and age, males were more likely than females to acquire another DUI at the 8-year mark but gender was not significant at the 2-and 5-year timelines. We also examined whether a subgroup of males in the 26-35-year age group would benefit more from attending a VIP than males who were simply fined. We detected no age by group assignment interaction effects, suggesting that at least for these data, the effects of the VIP are uniform across age groups for males.
The secondary nature of these official data obviously limits the volume of variables that we could entertain in our analysis. We could be missing important covariates such as socioeconomic status, marital status, and employment, but there is little likelihood that a sitting judge would base drunk driving sentencing on whether an offender is poor, divorced or employed. As for selection bias, Latimer et al. (2005) reported that most restorative justice studies they reviewed were fraught with biased estimates due to the voluntary nature of these types of programs. We were able to avoid this problem as a result of the random assignment of rotating court cases. We were informed that the sole judge who fined DUI offenders rather than requiring VIP attendance simply did so because he was not a believer in their recidivist impact. We would also be reluctant to generalize these findings to other court jurisdictions as the bulk of these DUI offenders were white, so we do not know if minority groups would be equally affected by Victim Impact Panels.
We now have evidence from one study that the messages imparted by drunk driving victims carries some staying power, at least with respect to reducing the odds of another DUI. This of course is the hope on the part of the court; that compelling drunk drivers to listen to the tragic and stirring stories of drunk driving victims will trigger some kind of response that tempers willingness to drive drunk in the future. We do not know if this decision implicates the increased use of a designated driver, an Uber call, or simply controlling their alcohol quantity-frequency. If so, we can possibly ascertain that somehow the grief and loss stories shared by victims could have an emotional impact that facilitates more rational thinking surrounding drinking behavior. However, we have no way of determining which mechanism is functioning to reduce further drunk driving.
We can report that the Public Health Administrator charged with administering the VIP program in this jurisdiction revealed that 90% of post-VIP attendees agreed or strongly agreed that drinking and driving is a mistake and posed an enormous risk. Further, 97% of attendees during the same period of time agreed or strongly agreed that they would consider stopping or decreasing their alcohol consumption if they have to drive. So, recent attendees have echoed sentiments attaching risk to drinking and driving.
Victim Impact Panels were not meant to replace other DUI related sanctions. The courts should continue to employ sentencing programs such as 24/7 sobriety, ignition interlocks, and DUI Courts. VIP programs are generally reserved for first-time DUI offenders while the aforementioned sanctions target multi-convicted offenders. Nevertheless, a second exposure to a Victim Impact Panel should not be excluded as a sentence. For some offenders, it may take more than one emotional victim's story to have them rethink their drinking and driving behavior. At least in this jurisdiction, this less-stigmatizing approach works better than traditional fines and probation.
Author Contributions: K.T. and S.J. have agreed to the conceptualization, methodology, software, validation, formal analysis, investigation, resources, data curation, writing-original draft preparation, writing-review and editing, visualization, supervision, project administration and funding acquisition of this manuscript. All authors have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.

Institutional Review Board Statement:
The original study fell under Exempt review status (December 2016) by the North Dakota State University Institutional Review Board as no respondents were contacted and official documents were employed as a data base.
Informed Consent Statement: Respondent consent was not required for this study as no respondents were contacted for the study purpose.
Data Availability Statement: Data for this study are available upon request from the principle author.

Conflicts of Interest:
The authors declare no conflict of interest.