29
\\server05\productn\C\CPP\7-4\CPP402.txt unknown Seq: 1 29-DEC-08 10:17 REDUCING INTIMATE PARTNER VIOLENCE: AN EVALUATION OF A COMPREHENSIVE JUSTICE SYSTEM- COMMUNITY COLLABORATION* CHRISTY A. VISHER University of Delaware ADELE HARRELL The Urban Institute (retired) LISA NEWMARK George Mason University JENNIFER YAHNER The Urban Institute Research Summary In 1999, three communities were selected to participate in a research demonstration designed to test the feasibility and impact of a coordi- nated response to intimate partner violence that involved the courts and justice agencies in a central role. The primary goals of the Judicial Oversight Demonstration (JOD) were to increase victim safety, hold offenders accountable, and reduce repeat offending using coordinated community services and integrated justice system policies in intimate partner violence court cases. The partnerships differed from earlier coordinated community responses to domestic violence by placing spe- cial focus on the role of the court, specifically the judge, to facilitate offender accountability in collaboration with both nonprofit service providers and other criminal justice agencies. This article presents the results of an impact evaluation of this demonstration in all sites. The demonstration received mostly positive responses from justice system agencies, service providers, offenders, and victims. Improvements were * This research was supported by Grant 99-WT-VX-K005 from the National Institute of Justice. Direct correspondence to Christy A. Visher, Department of Sociology & Criminal Justice, University of Delaware, Newark, DE 19716 (e-mail: [email protected]). CRIMINOLOGY & Public Policy Volume 7 Number 4 Copyright 2008 American Society of Criminology 495

REDUCING INTIMATE PARTNER VIOLENCE: AN EVALUATION OF A COMPREHENSIVE JUSTICE SYSTEM-COMMUNITY COLLABORATION

Embed Size (px)

Citation preview

\\server05\productn\C\CPP\7-4\CPP402.txt unknown Seq: 1 29-DEC-08 10:17

REDUCING INTIMATE PARTNERVIOLENCE: AN EVALUATION OF ACOMPREHENSIVE JUSTICE SYSTEM-COMMUNITY COLLABORATION*

CHRISTY A. VISHERUniversity of Delaware

ADELE HARRELLThe Urban Institute (retired)

LISA NEWMARKGeorge Mason University

JENNIFER YAHNERThe Urban Institute

Research SummaryIn 1999, three communities were selected to participate in a researchdemonstration designed to test the feasibility and impact of a coordi-nated response to intimate partner violence that involved the courts andjustice agencies in a central role. The primary goals of the JudicialOversight Demonstration (JOD) were to increase victim safety, holdoffenders accountable, and reduce repeat offending using coordinatedcommunity services and integrated justice system policies in intimatepartner violence court cases. The partnerships differed from earliercoordinated community responses to domestic violence by placing spe-cial focus on the role of the court, specifically the judge, to facilitateoffender accountability in collaboration with both nonprofit serviceproviders and other criminal justice agencies. This article presents theresults of an impact evaluation of this demonstration in all sites. Thedemonstration received mostly positive responses from justice systemagencies, service providers, offenders, and victims. Improvements were

* This research was supported by Grant 99-WT-VX-K005 from the NationalInstitute of Justice. Direct correspondence to Christy A. Visher, Department ofSociology & Criminal Justice, University of Delaware, Newark, DE 19716 (e-mail:[email protected]).

CRIMINOLOGY & Public PolicyVolume 7 Number 4 Copyright 2008 American Society of Criminology

495

\\server05\productn\C\CPP\7-4\CPP402.txt unknown Seq: 2 29-DEC-08 10:17

496 VISHER ET AL.

made in offender monitoring, consistent sanctioning, and increasedsupervision. However, these changes did not translate into gains in vic-tim perceptions of their safety or into reductions in repeat violence inall sites.

Policy ImplicationsThe demonstration had minimal impact on changing offender attitudesand behavior. The mixed results of the evaluation indicate that the mosteffective justice system responses to intimate partner violence mustinclude a focus on protecting victims, close monitoring of offenders,and rapid responses with penalties when violations of court-orderedconditions are detected. Indications were found that JOD strategieswere effective for some subgroups, including younger offenders withfewer ties to the victim and offenders with extensive arrest histories. Theobserved reductions in intimate partner violence in selected subgroupsin the JOD sites may suggest a fruitful way to begin designing newintervention strategies, including prevention programs for men andwomen.

Keywords: intimate partner violence, JOD, evaluation

Intimate partner violence results in nearly half a million crimes commit-ted against women each year, which accounts for about one fifth of allviolent crimes against women (Catalano, 2007). For many women, this vio-lence is not a single, isolated event. Many evaluation studies have beendesigned to assess whether legal or social service interventions reducerevictimization (Roberts, 2002). Yet results from evaluations of theseinterventions are limited by inadequate theoretical guidance, poor evalua-tion designs, and lack of solid implementation of the program or policy(Lyon, 2005; Mears, 2003).

Moreover, prior research overwhelmingly represents a single interven-tion approach (e.g., prosecution or batterer treatment or victim services)and has not addressed the effectiveness of comprehensive, multicom-ponent strategies for reducing intimate partner revictimization. A broadreview of intervention research suggested that community efforts toreduce intimate partner violence among individuals charged with domesticviolence offenses were likely to benefit from a coordinated criminal justiceand community response that provides protections for victims and strongsanctions for offenders (National Research Council, 1998). Thus, the U.S

\\server05\productn\C\CPP\7-4\CPP402.txt unknown Seq: 3 29-DEC-08 10:17

REDUCING INTIMATE PARTNER VIOLENCE 497

Department of Justice designed the Judicial Oversight Demonstration(JOD) to provide a much-needed opportunity to evaluate a complex,coordinated approach to intimate partner violence that incorporated fullparticipation by the courts.

The central hypothesis of JOD was that strong judicial oversight ofdomestic violence offenders, together with extensive graduated sanctionsfor offenders and comprehensive services for victims, would reduce reof-fending, increase accountability of the defendant and the system, andenhance victim safety. This article presents the findings from a state-of-the-art research demonstration that tested whether such a collaborativeresponse would reduce intimate partner victimization. It begins with anoverview of the JOD initiative and its essential components in three dem-onstration sites, describes the quasi-experimental design of the impactevaluation, and presents the results and their implications for policy andpractice.

JOD Initiative

In the late 1990s, staff from the Office of Violence Against Women andthe National Institute of Justice sat down together to design a research-based demonstration project that would build on prior research on crimi-nal justice and community models for addressing intimate partnerviolence. An explicit focus of the demonstration would be to increase therole of the judge and the court in the response to such violence and, partic-ularly, to strengthen processes to hold offenders accountable for theiractions through greater surveillance, sanctions, and treatment. It was alsorecognized that community partners, including victim service agencies andbatterer intervention programs, needed to be more closely coordinatedwith the actions of the criminal justice system.

The JOD was intended to consolidate the gains in legal protections fordomestic violence victims made in the 1980s and 1990s within specificareas and justice agencies. For example, new laws had been establishedthat prohibited stalking, repealed spousal exceptions in rape cases,expanded the use of protection orders in the criminal courts, and widenedthe range of offenses and severity of penalties for domestic violence(Miller, 1997). In law enforcement, model codes as well as written policiesand procedures replaced wide local variation in legislation and enforce-ment. In addition, prosecution of these cases became increasinglyspecialized, with formal protocols for domestic violence cases, no-droppolicies (often flexible), and use of evidence to support or replace victimtestimony. Courts also began to develop specialized processes for domestic

\\server05\productn\C\CPP\7-4\CPP402.txt unknown Seq: 4 29-DEC-08 10:17

498 VISHER ET AL.

violence cases, including screening to coordinate case processing, special-ized calendars, and court-ordered batterer treatment (Keilitz, Guerrero,Jones, and Rubio, 2000). However, few comprehensive domestic violencecourts that handled intake through sentencing and postconviction follow-up were in operation in the late 1990s.

Moreover, an important missing link was coordination among courtsand other justice and community agencies. Many experts felt the complexand recurring nature of domestic violence required a coordinated, sys-temic response. Early efforts toward this end occurred in Minnesota(Domestic Abuse Intervention Project), Virginia (Alexandria DomesticViolence Intervention Project), Nebraska (Domestic Violence Coalition),and Illinois (DuPage County), among other places, and showed promisingreductions in domestic abuse of victims served by the programs (Orchow-sky, 1999; Shepard and Pence, 1999; Steinman, 1990; Syers and Edleson,1992; Tolman and Weisz, 1995; also see Gamache, Edleson, and Schock,1988; Shepard, Falk, and Elliott, 2002). Early results also indicated that themechanism for coordination played a role in determining the effectivenessof domestic violence response efforts, with criminal justice leadershipyielding benefits as or more promising than those produced by advocacyorganization or coordinating council leadership (Shepard, 1999). Accord-ingly, Hofford and Harrell (1993) identified these six essential features ofsuccessful implementation of a coordinated approach to domestic vio-lence: (1) clear policies that define roles and responsibilities of partners;(2) designated personnel to lead coordination in each agency; (3) strongleadership, especially by judges; (4) cross training of staff from multipleagencies; (5) vigorous prosecution; and (6) formal monitoring of partner-ship performance. Above all, the importance of collaboration inresponding coalitions was affirmed in the Services-Training-Officers-Prosecutors (STOP) Grant, a component of the Violence Against WomenAct, which required that states engage in a collaborative planning process;divide the funds among law enforcement, prosecution, and victim serviceagencies; and encourage coordinated community responses. Evaluation ofthe program indicated that STOP subgrantees attributed the most signifi-cant changes in their communities to increased collaboration (Burt et al.,1999).

Thus, based on prior research and evaluation of intimate partner vio-lence prevention, several theories of change were incorporated into theJOD model at the level of the community, the offender, and the victim.First, and most broadly, JOD tested whether coordinated communityresponses that linked criminal justice agencies with community serviceproviders would result in decreases in repeat partner violence. Second,JOD was aimed at increasing the accountability of individuals charged

\\server05\productn\C\CPP\7-4\CPP402.txt unknown Seq: 5 29-DEC-08 10:17

REDUCING INTIMATE PARTNER VIOLENCE 499

with intimate partner violence principally by enhancing and strengtheningthe roles of the courts (and judges), pretrial and postrelease supervisionagencies, and batterer intervention programs. These shifts in practice wereintended to deter repeat partner violence by increasing perceptions of therisk of arrest and penalties for subsequent offenses. Third, the modeltested whether enhanced offender accountability under JOD negativelyaffected offender attitudes about the fairness of the justice systemresponse and decreased their willingness to comply with court orders (i.e.,aspects of procedural justice). Finally, JOD proposed that the coordina-tion of victim services by government and nonprofit agencies wouldimprove victim access to services and assist them in staying safe.

The research demonstration explicitly included an evaluation compo-nent that would examine the feasibility and effectiveness of the model sothat other jurisdictions could benefit from the experiences of the demon-stration programs and would provide an evaluation of the impact of JODon victim safety and offender recidivism. This article presents the resultsof the impact evaluation.

Components of the JOD Intervention

In 1999, after an extensive search for jurisdictions with the resources,infrastructure, and commitment necessary to implement the envisioneddemonstration, three sites were selected for the implementation of JOD:Dorchester, Massachusetts; Milwaukee County, Wisconsin; and Washte-naw County (Ann Arbor), Michigan.1 In each of the demonstration sites,criminal justice and community-based agencies serving victims and offend-ers formed partnerships to work collaboratively to support an effectiveresponse to intimate partner violence incidents. The partnerships differedfrom earlier coordinated community responses to domestic violence byplacing special focus on the role of the court, specifically the judge, tofacilitate offender accountability in collaboration with both nonprofit ser-vice providers and other criminal justice agencies. The core JOD elementswere as follows:

• Uniform and consistent initial responses to domestic violenceoffenses, including (1) proarrest policies, (2) arrest of primary aggres-sor, and (3) coordinated responses by law enforcement and victimadvocates

1. The sites were selected by the Office of Violence Against Women in the Officeof Justice Programs. Technical assistance was provided by the Vera Institute of Justice,and the National Institute of Justice awarded the 5-year process and impact evaluationto the Urban Institute.

\\server05\productn\C\CPP\7-4\CPP402.txt unknown Seq: 6 29-DEC-08 10:17

500 VISHER ET AL.

• Coordinated victim advocacy and services, including (1) contact byvictims’ advocates as soon as possible after an intimate partner vio-lence incident, (2) individualized “safety plans” for the victims, and(3) provision of needed services

• Strong offender accountability and oversight, including (1) intensivecourt-based supervision, (2) referral to appropriate batterer interven-tion programs, and (3) administrative and judicial sanctions andincentives to influence offender behavior

Figure 1 illustrates the network of agencies that collaborated to improveresponses to intimate partner violence. Each site implemented the coreJOD elements within the context of their local resources, needs, and pri-orities, which expanded their existing coordinated community response toinclude criminal justice agencies, and established regular meetings todevelop and implement strategies for interagency coordination. Guided bytechnical assistance teams and the needs of their jurisdictions, each sitereviewed and developed model policies and programs based on exper-iences in other jurisdictions, recent research, and other best practices forintimate partner violence cases. The core elements of the demonstrationtranslated into specific policies and procedures in justice system and com-munity-based agencies. Sites differed in the nature of the procedures thatwere developed, although more similarity than difference was found in theimplementation of JOD across sites. In practice, JOD sites focused onthese five elements: coordination of court and community agencies, proac-tive law enforcement, specialized prosecution and court procedures,specialized probation and batterer intervention services, and enhancementof victim services.2

Coordination of court and community agencies. The design of JODand its specific elements in each site reflected principles identified in ear-lier demonstrations focused on building coordinated community responsesto domestic violence. Mechanisms to ensure coordination in JOD sitesincluded hiring a project director to coordinate and oversee implementa-tion of JOD; establishing an executive committee, with working

2. Detailed information about JOD implementation and the strategies and inno-vations adopted by JOD sites can be found in Harrell, Newmark, Visher, and Castro(2007).

\\server05\productn\C\CPP\7-4\CPP402.txt unknown Seq: 7 29-DEC-08 10:17

REDUCING INTIMATE PARTNER VIOLENCE 501

Figure 1. JOD Network of Partners

FamilyCourt

Community Residents/ Victim Families

CivilProtectionOrder Court

OffenderSupervisionAgencies

LawEnforcementAgencies

Prosecutor’sOffice

Other Agencies/ Institutions

Court-BasedVictim Services

Defense

TheJudge

OffenderInterventionAgencies

Community-BasedVictim Advocacy/ Assistance Services

subcommittees, that met regularly to identify and troubleshoot emergingissues; developing standard policies and protocols to improve the consis-tency of case handling practices; and improving communication amongagencies through shared databases.

Specialized law enforcement, prosecution, and court procedures. In allsites, the law enforcement components of the JOD initiatives includedexpanding training, introducing new policies related to domestic violencearrests and protection order enforcement, and improving interagencycommunications. The JOD initiatives included strategies such as special-ized prosecution units, no-drop policies, vertical prosecution, andevidence-based prosecution. JOD prosecutors expanded their use of inde-pendent evidence, such as photographs of victims’ injuries, hospitalrecords, excited utterances, expert testimony, 911 audiotapes, and otherevidence, to support or replace victim testimony. Milwaukee placed spe-cial focus on prosecuting bail violations such as bail jumping and witnesstampering.

Milwaukee and Washtenaw County included several improvements inpretrial processes for domestic violence cases. Washtenaw County stan-dardized bond conditions across judges and ordered defendants to attendgroup meetings to review bond conditions within 1 week of arraignment.

\\server05\productn\C\CPP\7-4\CPP402.txt unknown Seq: 8 29-DEC-08 10:17

502 VISHER ET AL.

Milwaukee created a commissioner-led domestic violence court to processall pretrial matters and implemented an intensive pretrial monitoring com-ponent for offenders accused of a repeat domestic violence offense.Dorchester could not introduce specialized pretrial procedures.

The JOD demonstration sites also included specialized court calendarsthat reserved domestic violence cases for specific days of the week, judicialreview hearings for probationers, and, in Milwaukee, a domestic violenceintake court. Specialized calendars made it possible for all personnelinvolved in a particular case to be present in the courtroom together. Thisarrangement resulted in improved coordination and information-sharingamong criminal justice personnel (i.e., between judges and probationofficers) and among community service providers and the court.

Specialized probation and batterer intervention services. The JODdemonstrations in Washtenaw County and Dorchester included special-ized probation officers, enhanced staffing to reduce caseloads, referrals tocertified batterer intervention programs, enhancement of communicationamong probation officers, and batterer intervention program staff to pro-vide information for judicial review hearings. In Milwaukee, a specializedprobation response was not possible, although probation officers weregiven training in supervising domestic violence offenders.

Enhancement of victim services. JOD sites expanded in-court victimservices from both justice- and community-based providers in variousways, including civil legal services, programs to increase victim autonomy,and victim waiting rooms in the courthouse. Victim services in Milwaukeeand Dorchester were more difficult to coordinate than in WashtenawCounty because of the multiple agencies involved: Washtenaw County hasonly one nongovernmental victim service agency.

Evaluation Design and Analytic Methodology

Two JOD sites—Dorchester and Washtenaw County—participated in acontemporaneous, quasi-experimental evaluation of the impact of theintervention. Intimate partner violence cases that reached court disposi-tion during the demonstration period were compared with similar casesthat reached disposition in Lowell, Massachusetts, and Ingham County(Lansing), Michigan. The sample consisted of victims in intimate partnerviolence cases that reached disposition and offenders in those cases whowere placed on probation. The offender JOD and comparison samplesexcluded defendants found not guilty, those in dismissed cases, and thoseconvicted and sentenced to 6 or more months of incarceration because ofthe intervention’s focus on offenders placed on probation. All cases in thesample involved (1) criminal intimate partner violence charges, (2) victims

\\server05\productn\C\CPP\7-4\CPP402.txt unknown Seq: 9 29-DEC-08 10:17

REDUCING INTIMATE PARTNER VIOLENCE 503

and offenders aged 18 years or older, and (3) victims and offenders wholived in the target jurisdiction at the time of case disposition.3 The incidentthat led to the court case sampled for the evaluation was typically a physi-cal assault, often a severe physical assault, but only rarely a sexual assault.Weapons were used in about 2 out of 10 incidents, and assault and batterywas by far the most frequent top arrest charge (77–80%), although aggra-vated assault and battery charges were not uncommon (12–14%).

Data for the impact evaluation in these sites included (1) in-personinterviews with victims and offenders conducted 2 months after case dispo-sition or sentencing and again 9 months later, (2) criminal history recordsfrom state and local law enforcement records on arrests before and afterthe sampled case, and (3) data on JOD victim services and probationsupervision.4,5 Interviews were completed with 1,034 victims (526 fromJOD sites and 508 from comparison sites) 2 months after case disposition(50% of eligible victims) and with 914 victims (90% of the initial interviewsample) 11 months after case disposition. Furthermore, interviews werecompleted with 454 offenders (229 from JOD sites and 225 from compari-son sites) 2 months after case disposition (39% of eligible offenders) andwith 366 offenders (84% of the initial interview sample) 11 months aftercase disposition.6

In the third JOD site—Milwaukee—the impact evaluation was based ona quasi-experimental comparison of offenders convicted of intimate part-ner violence and ordered to probation during the JOD period withintimate partner violence offenders ordered to probation in the yearbefore JOD. This design was selected when early plans for an experimen-tal design had to be abandoned and no comparable contemporaneouscomparison group could be identified. In Milwaukee, data for the evalua-tion were collected from court and prosecutors’ records of case and

3. Cases that reached disposition more than 1 year after the incident wereexcluded to limit loss of data because of poor recall of the facts of the incident andpolice response.

4. Atlantic Research and Consulting (now Guidelines) conducted the in-personinterviews in Massachusetts. The Center for Urban Studies at Wayne State Universityconducted the in-person interviews in Michigan.

5. In Michigan, the Michigan State Police Department of Information Technologyprovided the criminal history records. In Massachusetts, criminal offender informationrecords from the Massachusetts Criminal History Systems Board were supplemented bychecks of warrants that resulted in arraignments after case disposition to verify that thenew incidents occurred during the year after case disposition.

6. As expected, it was difficult to locate the selected sample, particularly theoffenders, using data available from court and police records. Often, the victim’saddress information was incomplete or the victim had moved; also the offenders fre-quently moved as they were no longer living with the victim.

\\server05\productn\C\CPP\7-4\CPP402.txt unknown Seq: 10 29-DEC-08 10:17

504 VISHER ET AL.

defendant characteristics, probation files on offender supervision prac-tices, and official records of rearrest 1 year after case disposition.Interviews with victims and offenders were not attempted because of thedifficulty in locating the individuals in the comparison group whose casescame to court in the year before JOD; moreover, their recollections ofspecific court events were likely to be poor. Thus, findings from Milwau-kee are limited to analyses of arrest records for offenders during the JODperiod and for comparable offenders in the year before JOD.

Victim, Offender, and Case Characteristics

Tables 1 and 2 present the characteristics of victim and offender samplesin Dorchester and Washtenaw County as well as at the comparison sites,and Table 3 presents offender characteristics in Milwaukee. Regardless ofstate or type of site (JOD vs. comparison), victims in the eligible criminaldomestic violence cases were typically young women with several children.Various race/ethnicity groups were included in the samples, with represen-tation varying considerably by site. Although specific socioeconomicindicators tended to differ by site, income levels were generally low, anduse of public assistance was widespread (see Table 1).

The victims in these cases had significant relationships with their abus-ers at the time of the incident that led to the court case. Patterns weregenerally consistent across sites, with only a few differences noted.Approximately half of the victims had children in common with theabuser, and half or more were in a current, cohabiting relationship at thetime of the incident. Relationship length averaged more than 6 years.These relationships generally involved substantial prior abuse. Overall,approximately two thirds of the victims had experienced threats or intimi-dation, and half to three quarters had experienced a physical assault in theyear before the incident. On average, the victims had been abused fornearly 4 years, and approximately one quarter of them had obtained a no-contact order at some point before the incident. By the time of the follow-up interview (9 months later), only approximately one third of the victimswere still in a relationship with the abuser, although more than half hadlived with the abuser at some point between the two interviews.

As Tables 2 and 3 indicate, most offenders in both samples were male,and most were in their early 30s. Although approximately three quartersof the offenders had graduated from high school, nearly half of the pooledsample reported in the initial interview that they were not employed(either full time or part time), and more than half reported incomes below$20,000 per year. The tables also attest to some differences between the

\\server05\productn\C\CPP\7-4\CPP402.txt unknown Seq: 11 29-DEC-08 10:17

REDUCING INTIMATE PARTNER VIOLENCE 505

Table 1. Victim Sample Characteristicsa

Dorchester Lowell Washtenaw Ingham(n = 307) (n = 286) (n = 219) (n = 222)

Female 89% 88% 92% 91%Average age 33.6 34.2 32.1 31.8Race/ethnicityWhite 10% 67% 50% 49%Black 64% 4% 39% 32%Asian 1% 9% 1% 1%Hispanic 7% 13% 1% 6%Other/multiracial 18% 6% 10% 13%Has children 86% 83% 76% 80%High school graduate 78% 75% 88% 77%U.S.-born 79% 78% 93% 96%Employed 47% 58% 74% 61%Housing assistance 67% 84% 87% 87%Income < $10,000/year 49% 44% 42% 45%Victim-defendantrelationshipChildren together 55% 53% 47% 52%Lived together 52% 64% 69% 70%Relationship length(years) 6.4 7.2 6.2 5.6Any physical assault inpast year 57% 65% 70% 74%Length of abuse(years) 3.8 4.3 3.8 3.3Ever had no-contactorder 26% 30% 17% 19%aSamples of victims who completed both interviews.

JOD and comparison offenders by race, income, number of prior arrests,and living arrangements with their victims. Weights using these variableswere constructed to increase sample comparability between JOD andcomparison samples and to adjust for the sample selection process.

The analyses presented here used inverse probability weighting (IPW)to make the results generalizable to the population targeted by JOD andto control for differences in selection of cases in the JOD and comparisonsites. IPW methods provide an intuitive and general approach to controlfor such effects. The IPW method is a logical extension of inverse weight-ing methods used in survey sampling, in which the sample at hand may bea nonrepresentative subset of the population of interest (i.e., members ofthe population of interest are either overrepresented or underrepresentedin the observed sample). To correct for any nonrepresentativeness, it is

\\server05\productn\C\CPP\7-4\CPP402.txt unknown Seq: 12 29-DEC-08 10:17

506 VISHER ET AL.

Table 2. Offender Sample Characteristics: Michigan and Massachusettsa

Dorchester Lowell Washtenaw Ingham(n = 97) (n = 82) (n = 83) (n = 103)

Average age 33.9 35.6 32.7 35.1Male 79% 84% 84% 90%RaceWhite 8% 57% 49% 52%Black 65% 1% 45% 36%Other/multiracial 27% 41% 6% 12%High-school graduate 74% 66% 84% 76%Not employed atinitial interview 60% 46% 30% 38%Income less than$20,000/year 69% 48% 58% 57%Number of priorarrests 8.3 3.7 1.9 2.9Lived with victim attime of incident 61% 77% 66% 79%a Samples of offenders who completed both interviews.

customary to weight the sampled observations by their inverse probabilityof being selected. Therefore, the IPW methods can be used to correct gen-eral forms of sample selection problems, including selection, attrition, andstratification (Hirano, Imbens, and Ridder, 2003; Wooldridge, 2007).7 Theremainder of this article examines the impact of JOD on (1) the exper-iences of intimate partner violence victims with service and justice systemagencies, including case outcomes and court requirements for offenders;and (2) victim well-being and offender recidivism.

7. The IPW approach is particularly appropriate for the JOD analysis becausetwo types of sample selection adjustment (sample representativeness and sample com-parability) can be modeled distinctly and then can be combined using Bayes’ theoremto obtain a final selection probability. This composite probability is the joint probabilityof agreeing to participate and being assigned to a particular condition (i.e., treatment orcontrol). Inverting these composite probabilities yields a weight. When these weightsare used to perform weighted estimation, they effectively “control” for the selectionprocess by assigning relatively higher weights to those sampled members who have alower probability of being sampled—similar to assigning higher weights to under-represented sample members in analyzing survey data.

\\server05\productn\C\CPP\7-4\CPP402.txt unknown Seq: 13 29-DEC-08 10:17

REDUCING INTIMATE PARTNER VIOLENCE 507

Table 3. Offender Sample Characteristics: Milwaukee

Pre-JOD JOD(n = 289) (n = 333)

Age in years 35 34Male 96% 93%Race

White 32% 32%Black 56% 49%Other/multiracial 13% 19%

Number of prior arrests 5.1 5.2

Impact of JOD on Victim and Offender Experienceswith Services and Justice Agencies

Enhanced victim services and strong offender accountability were keyelements of JOD. To expand services and improve outcomes for victims,JOD provided increased advocacy in the courthouse and in the communityto help victims with civil and criminal cases and to help address the needsof victims for safety and autonomy through culturally and linguisticallycompetent services. Collaboration among victim service providers wasimproved, and victim sensitivity became more of a focal point in the ser-vices of other agencies that had contact with victims. These enhancedvictim services and improved interactions with other agencies are hypothe-sized to improve victims’ perceived safety and their sense of well-being.

To enhance offender accountability, JOD introduced changes in the pol-icies and practices of criminal justice and community agencies and focusedon improving victim safety through close monitoring of offenders,required attendance at batterer intervention programs, and consistent useof administrative and judicial sanctions as well as incentives to influenceoffender behavior. JOD offenders were expected to be influenced by thegreater attention to accountability and to anticipate more consistent andsevere penalties for future intimate partner violence. To test these hypoth-eses regarding the impact of JOD on victim and offender experiences withservice agencies and the justice system, we analyzed JOD and comparisonvictim and offender reports about their contacts with these agencies, whichwere gathered in one-on-one interviews.

As shown in Table 4, JOD victims reported more contact with nongov-ernmental victim service agencies in the JOD sites (46%) than in thecomparison sites (27%). Furthermore, significantly more JOD victims pur-sued follow-up contact with victim services (28% to 13%). Nearly allvictims (more than 80%) reported an immediate police response to the

\\server05\productn\C\CPP\7-4\CPP402.txt unknown Seq: 14 29-DEC-08 10:17

508 VISHER ET AL.

Table 4. Victims’ Contact with Service Agencies and Justice SystemAgencies for JOD and Comparison Respondents

Type of contact JOD Comparison

Service agency contactContact before sampled incident*** 25% 14%Contact since sampled incident*** 46% 27%Any follow-up contacta*** 28% 13%Average number of services provided* 2.7 2.2Police and prosecutor contactPolice responded during or immediately afterincident* 83% 88%Had contact with prosecutor’s office*** 65% 52%Victim went to court for a case hearing 66% 67%Victim testified 30% 28%Victim felt pressured to change testimony or nottestify 65% 71%Probation contactb

Contact with probation if offender onprobation*** 71% 25%Average number of contacts** 3.0 2.3Contact with BIP if offender ordered to BIPc** 37% 23%

Notes. Sample sizes are from initial interview sample of victims, except wherenoted (JOD sample = 526; comparison sample = 508). BIP = Batterer InterventionProgram.a Sample sizes for follow-up sample: JOD = 453; comparison = 461.b Samples are based on cases ordered to probation: JOD = 257; comparison = 177.c Samples are based on cases ordered to BIP: JOD = 209; comparison = 125.* p ≤ .05. ** p ≤ .01. *** p ≤ .001.

incident that led to the court case, with victims in comparison sites slightlymore likely to report a response. Not surprisingly, victim contact with theprosecutor’s office and the court was less widespread than contact with thepolice: Approximately two thirds of JOD victims had contact with theprosecutor’s office (65%), whereas just more than half of the comparisonvictims had contact. Approximately two thirds of both groups reportedthat they went to court for hearings in their cases, and less than a thirdtestified in their cases; JOD did not increase court appearances by thevictim or her likelihood of testifying. Disturbingly, approximately twothirds of victims in both groups reported that they felt pressured to changetheir testimony or not testify, most often by the defendant or defendant’sfamily. The largest differences in contact with justice agencies betweenJOD and comparison victims occurred in their interactions with probationofficers. More than two thirds of JOD victims (71%) reported contact with

\\server05\productn\C\CPP\7-4\CPP402.txt unknown Seq: 15 29-DEC-08 10:17

REDUCING INTIMATE PARTNER VIOLENCE 509

probation officers, which was almost three times the number of compari-son victims who reported such contact (25%). The number of contacts wasalso higher among JOD victims. In summary, the results support the JODstrategy of increasing victim contacts with services and justice systemagencies within the JOD sites: JOD victims did report greater contactswith victim service agencies and dramatically higher likelihood of interac-tions with probation officers than comparison victims, although JODvictims did not report greater involvement in the court cases.

We tested the impact of JOD on efforts to increase offender accounta-bility by examining case outcomes and court requirements for JOD andcomparison offenders (see Table 5). It is common in many jurisdictions fordefendants in intimate partner violence cases to have their cases deferredfrom prosecution for some specified period of time, after which point thecharges are dropped. This practice makes it difficult to identify repeatoffenders. During the demonstration, a review of court records showedJOD offenders were significantly more likely to be convicted and sen-tenced than comparison offenders (82% compared with 69%), more likelyto have the case continued without a finding (CWOF) (18% comparedwith 11%), and less likely to be granted deferred prosecution (0% com-pared with 20%).8

JOD offenders received more severe sentences, with 96% receiving pro-bation alone or with jail, compared with 63% of comparison offenders.The probation requirements ordered by JOD courts clearly reflect theJOD strategy of heightened supervision of intimate partner violenceoffenders and attention to victim safety. Among probationers, JODoffenders were much more likely to be required to attend a batterer inter-vention program (and attend more sessions), to be ordered to drug testing,and to have weapons restrictions. JOD offenders were not more likelythan comparison offenders to have a strict “no victim contact” order (32%compared with 35%), but they were more likely to have an “only consen-sual contact” order (44% compared with 0%) and to be ordered to haveno abusive contact with the victim (31% compared with 12%). Thus, JODsuccessfully increased offender accountability overall by adding to court-ordered supervision and monitoring requirements for offenders convictedof intimate partner violence.

8. In Dorchester, CWOF dispositions for first-time offenders in selected casesinvolved probation supervision before conviction. Violations of CWOF conditionsresulted in probation violation hearings, and the judge responded with sanctions. InLowell (the comparison site), similar cases often were dismissed.

\\server05\productn\C\CPP\7-4\CPP402.txt unknown Seq: 16 29-DEC-08 10:17

510 VISHER ET AL.

Table 5. Case Outcomes and Court Requirements of Offenders

JOD ComparisonCase outcomes (n = 180) (n = 185)

Case disposition***Convicteda 82% 69%CWOF 18% 11%Deferred prosecution 0% 20%Top conviction charge lower than top arrestcharge (if convicted) 10% 16%Sentence, imposed or deferred***Probation, alone or with jail 96% 63%Jail (served or deferred), no probation 3% 10%No jail or probation 1% 27%Court requirementsBatterer intervention program*** 80% 42%Number of required sessionsb*** 39.2 25.8Drug testing*** 57% 34%No contact with victim 32% 35%Only consensual contact with victim*** 44% 0%Weapons restriction*** 42% 1%No abuse/assaultive behavior*** 31% 12%a This measure includes offenders placed on deferred sentencing.b Based on offender self-report.*** p ≤ .001.

Impact of JOD on Victim Well-Beingand Offender Recidivism

The final set of analyses examined the key questions of whether JODimproved victim safety and reduced the reoccurrence of domestic violence(i.e., offender recidivism). Data on the reoccurrence of violence comefrom these sources: (1) interviews with JOD and comparison victims andoffenders in Michigan and Massachusetts, and (2) records on arrests ofoffenders in all three JOD sites (including Milwaukee).

Victims were asked at initial and follow-up interviews about how safethey perceived themselves to be from the defendant and about their over-all sense of well-being using multiple items to form two scales. Theperceived safety scale consisted of four Likert-type response items thatindicated how safe respondents perceived themselves to be from thedefendant. The global well-being scale consisted of three Likert-typeresponse items that indicated how satisfied and confident respondents felt

\\server05\productn\C\CPP\7-4\CPP402.txt unknown Seq: 17 29-DEC-08 10:17

REDUCING INTIMATE PARTNER VIOLENCE 511

about their lives and their futures.9 Scale scores ranged from 1 to 4, withhigher scores indicating greater perceived safety or well-being. As shownin Table 6, both JOD and comparison victims’ reported sense of safety andwell-being was moderately high (3.2 or greater), but JOD victims did notreport greater overall safety or well-being than comparison victims. Infact, victims in comparison sites reported significantly greater perceivedsafety at the follow-up interview (3.6) than JOD victims reported (3.4).

Table 6. Victims’ Ratings of Perceived Safety and Well-Being for JODand Comparison Respondents

JOD Comparison

Initial interview sample n = 526 n = 508Perceived safety scale (1–4) 3.4 3.4Global well-being scale (1–4) 3.2 3.3Follow-up interview sample n = 453 n = 461Perceived safety scale** 3.4 3.6Global well-being scale 3.3 3.4

** p ≤ .01.

The impact of JOD on offender recidivism is examined first throughvictim self-reports of new threats, intimidation, or violence at both the ini-tial and follow-up interviews. Threats or intimidation by the offender afterthe sampled incident were measured using seven questions asked only ofvictims.10 Using these questions, the prevalence and frequency of threatsor intimidation were computed. Prevalence indicates whether any threats

9. The items for the perceived safety scale included how likely victims felt thatthe offender would physically hurt them in the next 6 months; how afraid victims werethat the offender would try to kill them; how safe from the offender victims felt now;and whether victims felt more safe, less safe, or about the same as at the time of theincident. The JOD global well-being scale included these three items: how satisfied vic-tims felt with their lives now; whether they felt better, worse, or about the same aboutlife in general since the incident; and how confident victims were that they would beokay in the future. Both scales achieved reliabilities of .70 or greater in the follow-upinterview.

10. The threat questions measured how many times the offender had (1) madeserious threats to hurt or kill the victim or the victim’s children, family members, orother loved ones; (2) hurt, killed, or damaged the victim’s pets or things important tothe victim; (3) made serious threats to take the victim’s children away or turn themagainst the victim; (4) gotten the victim in trouble with the police, courts, immigration,or other government agencies; (5) done things to frighten the victim such as following,spying, or sending intimidating messages; or (6) done things to control the victim suchas not letting the victim have money, have a job, or talk with family and friends. Thequestions had response options of never, once, twice, 3 to 5 times, 6 to 10 times, 11 to 20

\\server05\productn\C\CPP\7-4\CPP402.txt unknown Seq: 18 29-DEC-08 10:17

512 VISHER ET AL.

or intimidation were reported, whereas frequency shows the average num-ber of threats or intimidation reported.11 Physical assaults by the offenderafter the sampled incident were measured using 12 questions asked ofboth victims and offenders. Using these questions, prevalence and fre-quency measures of physical assaults were computed. The questions wereadopted from the physical violence component of the revised Conflict Tac-tics Scale (Straus, 2004).12 The most serious types of assaults, includingbeing hit, kicked, choked, or struck by a weapon, were combined into asecond measure of severe physical assault.

Table 7 presents the simple bivariate analyses of revictimization basedon victim reports at the initial interview 2 months after the case had beendisposed and on a composite measure that used responses from both inter-views and covered 11 months after case disposition. More than a third ofthe victims in the JOD (35%) and comparison (38%) groups reportedexperiencing threats or intimidation in the 2 months following case dispo-sition and at a relatively high average frequency of eight times since theincident. No significant difference was found between JOD and compari-son victims’ reports of threats and intimidation, however. Significantlyfewer JOD victims reported a new physical assault by the offender (13%)than did the comparison victims (21%). However, when examining onlysevere physical assaults (the first seven items of the assault measure),these types of assaults did not differ between groups: Approximately 1 in10 of the victims in both groups reported a new severe physical assault bythe offender in the initial interview.

times, and more than 20 times. Following Straus (2004), the frequency score assigned toeach respondent is the midpoint of the range they selected for the specific item (25 ifrespondent selected more than 20 times), summed across all items. The frequency mea-sure in the tables is the mean of these midpoints.

11. The internal reliabilities for the prevalence and frequency versions of this scalewere high, reflected in Cronbach’s alphas of .79 and .79 at victims’ initial interview and.81 and .77 at follow-up.

12. The assault questions measured how many times the offender did the followingthings to the victim: (1) punched or hit with something that could hurt, (2) beat up, (3)kicked, (4) choked, (5) slammed against a wall or other hard surface, (6) used a knife orgun, (7) burned or scalded, (8) slapped, (9) grabbed, (10) threw something that couldhurt, (11) pushed or shoved, and (12) twisted arm or pulled hair. The questions hadresponse options of never, once, twice, 3 to 5 times, 6 to 10 times, 11 to 20 times, andmore than 20 times. For the victim sample, the Cronbach’s alpha reliabilities for theprevalence and frequency versions of this scale were .90 and .91 at initial interview and.92 and .89 at follow-up. For the offender sample, reliabilities were .77 and .81 at initialinterview and .82 and .79 at follow-up.

\\server05\productn\C\CPP\7-4\CPP402.txt unknown Seq: 19 29-DEC-08 10:17

REDUCING INTIMATE PARTNER VIOLENCE 513

Table 7. Victim Reports of Offender Revictimization Since Incident

JOD Comparison(n = 453) (n = 461)

Victim reports at initial interviewAny threats or intimidation since incident 35% 38%Frequency of threats or intimidation 8.3 8.3Any physical assault since incident** 13% 21%Frequency of physical assault 4.9 3.1Any severe physical assault since incident 9% 12%Frequency of severe physical assault 1.9 0.9Victim reports at initial and follow-up interviewsAny threats or intimidation since incident 53% 56%Frequency of threats or intimidation 17.7 18.9Any physical assault since incident* 28% 35%Frequency of physical assault 11.4 7.9Any severe physical assault since incident 21% 23%Frequency of severe physical assault 4.0 2.8

* p ≤ .05. ** p ≤ .01.

The second half of Table 7 shows the same bivariate comparisons ofvictim-reported repeat intimate partner violence, but measured over alonger period of time—from the time of the incident until 11 months aftercase disposition. Within this time frame, more than half of all victimsreported recurrent threats or intimidation by the offender at a frequencyof 18 to 19 times, with no significant differences between groups. As withearly reports of violence, a significant difference was found between JODand comparison victims in the prevalence of physical assault in the overallsample: Approximately one fourth of JOD victims (28%) reported a newphysical assault compared with just more than one third of comparisonvictims (35%). Approximately one in five victims in each sample exper-ienced a new severe assault, but no statistically significant differences werefound between JOD and comparison victims’ reports.

To examine in more detail the impact of JOD on victims’ reports of newassaults, multivariate models were estimated that controlled for victim,offender, and case characteristics, including features of the relationshipthat theoretically could influence offender recidivism.13 Separate models

13. The following variables were included in multivariate models for each measureof revictimization: offender age, offender race, offender socioeconomic status, offenderprior arrests, offender alcohol/drug use, victim financial independence, victim socialsupport, victim has child with other partner, severe incident, aggravated assault charge,on-scene arrest, prior violence in relationship, prior police response, lived togethersince incident, and frequent interaction since incident.

\\server05\productn\C\CPP\7-4\CPP402.txt unknown Seq: 20 29-DEC-08 10:17

514 VISHER ET AL.

were estimated using victim reports of any physical assault at the initialinterview and the combined report of any assault from both interviews.After including the control variables, the effect of JOD in reducing victimreports of physical assaults both in the initial 2-month period and in the 11months since the incident remained statistically significant (results notshown).14

Offender self-reports of repeat violence against the victim also wereexamined during the two interviews but, as expected, few offenders admit-ted any repeat violence (data not presented in tabular form). In the initialperiod, 8% of JOD offenders and 6% of comparison offenders reportedcommitting a new physical assault against the victim. Using a combinedmeasure from both interviews, 17% of JOD offenders and 13% of com-parison offenders reported a new physical assault since case disposition.Compared with victim reports for the same time period (refer to Table 7),offender reports represent one third to one half of the prevalence ofvictim-reported physical assault.

Table 8. Official Records: Offender Recidivism and ProbationRevocation 1 Year Since Case Disposition

JOD Comparison

Victim sample (MA and MI) n = 453 n = 461Any arrest 1 year since case disposition (notspecific to DV) 25% 30%Number of arrests 0.4 0.5Probation revoked (MA, MI) 12%, 1% Unknowna

Offender sample (MA and MI) n = 180 n = 185Any arrest 1 year since case disposition (notspecific to DV) 22% 25%Number of arrests 0.3 0.3

Offender sample (Milwaukee) n = 289 n = 333Any arrest 1 year since case disposition (notspecific to DV) 24% 25%Number of arrests 0.4 0.3Domestic violence arrest* 4.2% 8%Number of domestic violence arrests* 0.09 0.04Probation revoked 27% 2%a Probation records were not available in Massachusetts and Michigan comparisonsites.* p ≤ .05.

14. The multivariate models are available from the authors by request and are alsoavailable in Harrell et al. (2007).

\\server05\productn\C\CPP\7-4\CPP402.txt unknown Seq: 21 29-DEC-08 10:17

REDUCING INTIMATE PARTNER VIOLENCE 515

The final indicator of offender recidivism measured whether theoffender was arrested during the year after disposition; such data wereavailable in all three JOD sites. In Massachusetts and Michigan, offenderarrests in the following year were assessed using criminal history recordsobtained from the Massachusetts Criminal History Systems Board and theMichigan State Police Department of Information Technology. Recidivismwas defined as any arrest that occurred within 1 year of the sampled casedisposition. In these two states, inconsistencies and vagueness in crimeclassifications made it impossible to distinguish arrests for new domesticviolence incidents in the available official records. Thus, the arrest mea-sure refers to any new arrest after disposition that appeared in the officialrecords. In Milwaukee, data were collected from the National CrimeInformation Center (NCIC), a nationwide information system that pro-vides criminal history records to law enforcement agencies.15 Theseofficial records permitted classification of arrests by offense type, includ-ing domestic violence charges (although not limited to intimate partnerviolence). Records of probation revocation for offenders in the study werealso available in all three sites, except for comparison offenders in Massa-chusetts and Michigan.

With remarkable consistency, approximately one quarter of sampledoffenders in all three JOD sites were rearrested during the year after casedisposition. As shown in Table 8, JOD offenders in Massachusetts andMichigan were slightly, but not significantly, less likely to be arrested thancomparison offenders (25% compared with 30%). Estimated rearrest ratesfrom a multivariate model that controlled for victim, offender, and casecharacteristics were 22% for JOD offenders and 28% for comparisonoffenders. Although this result is suggestive of JOD reducing arrests 1year after case disposition, it is not statistically significant. Notably, proba-tion revocation rates for JOD offenders were much higher inMassachusetts (12%) than in Michigan (1%). The implications of thesesite differences are discussed in the next section.

In Milwaukee, where it was possible to identify specific domestic vio-lence arrests, about 4% of JOD offenders were rearrested for a domesticviolence charge compared with 8% of offenders in a comparison samplefrom a period before JOD, which is a statistically significant difference. Asin Massachusetts and Michigan, no significant difference was foundbetween JOD and comparison offenders on overall (any charge) arrests

15. Data contained in NCIC are provided by the Federal Bureau of Investigation;federal, state, local, and foreign criminal justice agencies; and authorized courts. NCICrecords were checked against Consolidated Court Automated Program records main-tained by the Wisconsin Circuit Courts and the Milwaukee Police Department’s internaldatabase to ensure all arrests were recorded in NCIC.

\\server05\productn\C\CPP\7-4\CPP402.txt unknown Seq: 22 29-DEC-08 10:17

516 VISHER ET AL.

(24% and 25%, respectively). Probation revocation seemed to haveexpanded dramatically during JOD in Milwaukee: 27% of JOD offendershad their probation revoked compared with only 2% of pre-JOD offend-ers. It is possible that the lower domestic violence arrest rates may havebeen attained primarily through early detection and incarceration of pro-bationers who had continued their patterns of domestic violence.

Interactions and Site Differences

To explore in more detail the impact of JOD on offender recidivism, weexamined possible site differences and interaction effects between JODand theoretically important victim, offender, and case characteristics. Ofparticular interest was whether the impact of JOD on offender recidivismwas similar in Massachusetts and Michigan. In addition, we examinedwhether JOD had larger impacts for some types of victims and offenders.

As discussed, multivariate analyses were conducted to predict victims’reports of new victimization since case disposition using six measures ofthreats, intimidation, and violence. Examining the data by site revealedthat, in Massachusetts, JOD victims reported significantly less revictimiza-tion for four of the six measures than comparison victims. However, inMichigan, JOD victims reported similar levels of revictimization as theircomparisons on all measures. Thus, the observed effect of JOD on physi-cal assault in the pooled sample (refer back to Table 7) was actuallybecause of a significant effect of JOD in Massachusetts but not inMichigan.

This site difference in the impact of JOD was surprising because theprocess evaluation indicated that JOD was implemented well in Michigan.Taken together with the findings from Milwaukee, the most plausibleexplanation is that reductions in repeat violence occurred in the jurisdic-tions that revoked probationers for noncompliance. In WashtenawCounty, the county jail was at capacity, which may have prevented judgesfrom revoking probation for offenders who were not complying with pro-bation conditions. Offender attitudes also differed in the two JOD sites:Dorchester offenders scored much higher on a measure of perceived cer-tainty of legal sanctions for repeat intimate partner violence thanWashtenaw offenders, and JOD significantly increased the perception byDorchester offenders that future violence would have negative social con-sequences in the form of loss of employment or negative responses fromfamily, friends, children, or the victim.

\\server05\productn\C\CPP\7-4\CPP402.txt unknown Seq: 23 29-DEC-08 10:17

REDUCING INTIMATE PARTNER VIOLENCE 517

To explore the possibility that JOD’s effects on offender recidivismvaried across different subgroups of victims and offenders, tests for inter-actions between JOD and other independent variables were conductedusing victims’ reports at both time periods (2 months and 11 months aftercase disposition) for all six self-report measures of revictimization. Severalsignificant interactions consistently emerged by type of measure and bysite. In summary, JOD had its strongest effect on reducing repeat intimatepartner violence among victims characterized as follows:

• Victims whose offenders were young (age 18 to 29 years)• Victims with moderate-to-high social support• Victims whose offenders had a high number of prior arrests (seven or

more)• Victims who did not have children in common with the offender• Victims in a relatively short relationship (less than 3 years)

Thus, a close examination of the impact of JOD on offender recidivismrevealed important differences by site and by characteristics of victims andoffenders. These findings suggest various policy implications for communi-ties that engage in criminal justice and community partnerships to reduceintimate partner violence and to improve the safety of victims. Beforeturning to these policy implications, we briefly comment on the primarylimitations of this evaluation.

Study Limitations

The study was designed to measure the overall impact of the JOD inter-vention and was not designed to assess the impact of individual strategiesor component services. The primary reason for this design is that individu-als received various JOD interventions based on their needs and theirparticular circumstances, which makes comparisons with individuals whodid not receive the same particular intervention inappropriate. In addition,considerable variation was found within intervention components pro-vided to sample members. For example, in each site, offenders could bereferred to one of several batterer intervention programs, which varied incontent and duration. Moreover, victims received services based on theirneeds and interest in participation.

Another caution is that the samples were carefully selected to createsimilar JOD and comparison groups, but group members were not ran-domly assigned to JOD as in a true experiment. With random assignment,sample groups can be assumed to vary only by chance. With the quasi-experimental design in this study, the validity of the results depends on the

\\server05\productn\C\CPP\7-4\CPP402.txt unknown Seq: 24 29-DEC-08 10:17

518 VISHER ET AL.

extent to which differences in sample characteristics can be adequatelycontrolled in the statistical analysis. In the outcome analyses, weights andmultivariate modeling techniques were used to control for observed groupdifferences and to minimize any bias because of selection effects, but theycannot control for unobserved differences. In addition, the impact evalua-tion was based on experiences at three carefully selected sites. The extentto which results from these locations can be generalized to other commu-nities cannot be determined.

Another potential threat to the internal validity of the quasi-experimental comparisons in Massachusetts and Michigan is that preexist-ing differences between JOD and comparison sites—not the JODintervention—might account for differences in outcome. However, in thisstudy, a careful process evaluation documented few differences in theresponse to intimate partner violence in each site, which provides littlesupporting evidence for differences in policies and practices hypothesizedto impact the outcomes.16 The threat that external features of the settingaffect the outcomes is minimized in the Milwaukee evaluation by compar-ing outcomes within a single site before and during JOD. However, thisdesign opens the possibility that changes other than JOD during the dem-onstration period could account for differences in outcomes. Monitoringof the court and other agency responses to intimate partner violence dur-ing the Milwaukee demonstration period did not identify events otherthan JOD that were likely to affect the measured outcomes.

Finally, our official measure of the reoccurrence of violence based onarrest reports in two sites (Michigan and Massachusetts) is not limited torearrests for domestic violence but includes arrests for any offense. Thisgeneral measure of arrest may have obscured the effects of JOD onoffender recidivism in these sites. When the arrest measure was limited todomestic violence events in the Milwaukee site, JOD offenders were rear-rested less often than a comparison group. However, our use of victims’reports of repeat violence in two sites—which did show reductions in vio-lence among JOD victims—provides a key measure of revictimization.The combined use of both measures in these analyses balances the reliabil-ity and validity difficulties of the separate measures.

16. See Visher, Newmark, and Harrell (2007).

\\server05\productn\C\CPP\7-4\CPP402.txt unknown Seq: 25 29-DEC-08 10:17

REDUCING INTIMATE PARTNER VIOLENCE 519

Discussion and Policy Implications

The JOD initiative produced substantial changes in the response to inti-mate partner violence in all three sites, with enhanced collaborationamong justice agencies that respond to intimate partner violence andamong law enforcement officers, prosecutors, judges, and probationagents. Data from multiple sources show that the JOD initiatives led tocourt and probation improvements and to greater leadership, includingincreased court specialization, initiation of pretrial monitoring and post-trial compliance reviews, coordination with victim service agencies,dedicated domestic violence probation agents, increased supervision, com-pliance review preparations for the court, and outreach to victims.

However, these system changes produced mixed results in terms of pro-ject goals and hypotheses about theories of change embedded in the JODmodel. The results are complicated by a different research design in one ofthe three sites and by inconsistent findings across outcome measures. It isimportant to note that JOD and comparison victims continued to experi-ence considerable repeat violence in the year after disposition. During aperiod when most offenders were still on probation, approximately half ofthe victims in JOD sites had experienced threats or intimidation andalmost one quarter had experienced a severe assault. In Massachusetts,JOD victims reported significant reductions in threats and assault duringthe year after case disposition. In Michigan, no significant reductions werefound in victim reports of repeat violence. Reductions in arrests fordomestic violence charges were found in Milwaukee, but no reductions inoverall rearrests were found in the two states in which type of offense wasunavailable. In addition, JOD did not achieve gains in victim participationin the court process or in victim perceptions of her safety or well-being,using survey measures for assessment. Gains in offender accountabilitywere significant, but they did not translate into significantly lower rates ofrepeat violence reported by victims in both sites or by the offenders them-selves. Thus, the JOD model had much smaller effects on offenders (andvictims) than the developers envisioned. None of the theories of changethat underlie the JOD model were supported.

The reductions in repeat violence occurred in jurisdictions that revokedprobationers for noncompliance, which supports an incapacitation theoryof change. The implication is that the reduction resulted from incapacitat-ing abusers who failed to comply with probation conditions rather than bydeterring offenders. Despite the implementation of strategies for holdingoffenders accountable—through judicial review hearings, specialized pros-ecution and probation, police training, and increased requirements forbatterer intervention—JOD offenders did not differ from comparison

\\server05\productn\C\CPP\7-4\CPP402.txt unknown Seq: 26 29-DEC-08 10:17

520 VISHER ET AL.

offenders in their perceptions of risk of legal sanctions for future violence.However, in one site (Dorchester, Massachusetts), offenders scored muchhigher on a measure of perceived certainty of legal sanctions for repeatviolence than offenders in the Michigan site. To some extent, these percep-tions may be related to the higher rate of actual revocation in Dorchester(12%) compared with Washtenaw (1%). In Milwaukee during JOD, muchhigher revocation rates (27% in the first year of probation) were accompa-nied by a dramatic drop in rearrest rates for domestic violence, probablybecause of the incarceration of offenders most likely to be arrested. Thesefindings suggest that research is needed on the effectiveness of selectiveincarceration through probation revocation or other strategies for increas-ing the perceived threat of legal sanctions in this population.

Like many other studies, JOD found efforts to change offender percep-tions and reduce the reoccurrence of intimate partner violencechallenging. The results also suggest that referral to batterer interventionprograms does not have a powerful effect in reducing intimate partner vio-lence. Until progress is made in changing offender beliefs and behavior,the implication is that the justice system must continue to focus on protect-ing victims and using the authority of its agencies to monitor offendersclosely and to respond rapidly with penalties when violations of court-ordered conditions are detected.

The success of JOD in reducing intimate partner violence in selectedsubgroups may be a fruitful way to begin designing new intervention strat-egies, including prevention programs for men and women. Results did notsupport some previous research that offenders with a greater stake in con-formity are more likely to be deterred by criminal justice interventions,especially arrest. Rather, in this study, results suggested that the JODmodel was more effective for younger offenders with fewer ties to the vic-tim and for offenders with extensive arrest histories. Thus, the stake-in-conformity hypothesis may not extend to offenders who are more exper-ienced with the criminal justice system (an average of five arrests) andwho have been adjudicated by the court. In JOD sites, offenders with sub-stantial arrest histories may have been particularly affected by theunusually strong emphasis on offender accountability, which may have ledto reductions in their patterns of intimate partner violence. More researchis needed to confirm these findings.

The benefits of a coordinated system response for victims of intimatepartner violence are less obvious. Even in Michigan, where most JOD vic-tims received a wide range of quality services, victims did not reporthigher levels of well-being or safety than comparison victims. Surveyresults indicate that interventions intended to improve victims’ safety andoverall well-being need to go beyond services centered on cases in the

\\server05\productn\C\CPP\7-4\CPP402.txt unknown Seq: 27 29-DEC-08 10:17

REDUCING INTIMATE PARTNER VIOLENCE 521

court system to include services that address issues in the victims’ livesoutside the realm of the court case. The efforts of victim service providersmay be most fruitful when they focus on helping victims strengthen theirsocial support networks and augment the positive consequences, whileattenuating the negative impacts of abuse and its aftermath; such focus canbe on financial impacts (finding a job), practical issues such as moving, andhelping the victims and their children cope with emotional trauma. How-ever, despite the efforts of victim service agencies to provide support andencouragement, victims may be unwilling to take actions that wouldincrease their safety. In all sites (JOD and comparison sites), victims whoreported that they lived with their offender or had frequent contact withtheir offender after the case was disposed were more likely to reportrepeat victimization, including intimidation, threats, and assaults.

Overall, the evaluation points to the need for research in several criticalareas: how to build stronger links between courts and nongovernmentalvictim service providers given the high levels of satisfaction with serviceswhen they are received, how to motivate offender compliance and desis-tance from violence using sanctions and treatment in combination, how tochange offender perceptions of the risks of future violence, and how toengage victims in services that will assist them in staying safe.

References

Burt, Martha R., Adele Harrell, Lisa J. Raymond, Britta Iwen, Kathryn Schlicter,Lauren E. Bennett, and Kim M. Thompson. 1999. Evaluation of the STOP formulagrants to combat violence against women: 1999 report. Washington, DC: The UrbanInstitute.

Catalano, Shannon. 2007. Intimate partner violence in the United States. Bureau ofJustice Statistics webpage. Retrieved October 23, 2008 from ojp.usdoj.gov/bjs/intimate/ipv.htm.

Gamache, Denise J., Jeffrey L. Edleson, and Michael Schock. 1988. Coordinatedpolice, judicial, and social service response to woman battering: A multiplebaseline evaluation across three communities. In (Gerald T. Hotaling, DavidFinkelhor, John T. Kirkpatrick, and Murray A. Straus, eds.), Coping with familyviolence: Research and policy perspectives. Thousand Oaks, CA: Sage.

Harrell, Adele, Lisa Newmark, Christy A. Visher, and Jennifer Castro. 2007. Theimpact of JOD in Dorchester and Washtenaw County. Final report on theevaluation of the Judicial Oversight Demonstration (Vols. 1–4). Washington, DC:National Institute of Justice.

Hirano, Keisuke, Guido W. Imbens, and Geert Ridder. 2003. Efficient estimation ofaverage treatment effects using the estimated propensity score. Econometrica,71:1161–1189.

\\server05\productn\C\CPP\7-4\CPP402.txt unknown Seq: 28 29-DEC-08 10:17

522 VISHER ET AL.

Hofford, Meredith and Adele Harrell. 1993. Family violence: Interventions for thejustice system. Bureau of Justice Assistance Program Brief (NCJ-144532). Washing-ton, DC: U.S. Department of Justice.

Keilitz, Susan, Rosalle Guerrero, Anne M. Jones, and Dawn Marie Rubio. 2000.Specialization of domestic violence case management in the courts: A nationalsurvey. Williamsburg, VA: National Center for State Courts.

Lyon, Eleanor. 2005. Impact evaluation of special session domestic violence:Enhanced advocacy and interventions. Final report prepared for the NationalInstitute of Justice (NCJ 210362). Washington, DC: U.S. Department of Justice.

Mears, Daniel P. 2003. Research and interventions to reduce domestic violencerevictimization. Trauma, Violence, Abuse, 4:127–147.

Miller, Neal. 1997. Domestic violence legislation affecting police and prosecutorresponsibilities in the United States: Inferences from a 50-state review of statestatutory codes. Presentation to the 5th International Family Violence Conference,University of New Hampshire.

National Research Council. 1998. Violence in families: Assessing prevention andtreatment programs. Washington, DC: National Academy Press.

Orchowsky, Stan J. 1999. Evaluation of a coordinated community response todomestic violence: The Alexandria Domestic Violence Intervention Project, finalreport (NCJ 179974). Washington, DC: U.S. Department of Justice, NationalInstitute of Justice.

Roberts, Albert R. 2002. Handbook of domestic violence intervention strategies:Policies, programs, and legal remedies. New York: Oxford University Press.

Shepard, Melanie F. 1999. Evaluating coordinated community responses to domesticviolence. National Online Resource Center on Violence Against Women, VAWnetApplied Research Forum. Retrieved May 17, 2008 from new.vawnet.org/Assoc_Files_VAWnet/AR_ccr.pdf.

Shepard, Melanie F., Dennis R. Falk, and Barbara A. Elliott. 2002. Enhancingcoordinated community responses to reduce recidivism in cases of domesticviolence. Journal of Interpersonal Violence, 17:551–569.

Shepard, Melanie F. and Ellen Pence. 1999. Coordinating community responses todomestic violence: Lessons from Duluth and beyond. SAGE Series on Violenceagainst Women. Thousand Oaks, CA: Sage.

Steinman, Michael. 1990. Lowering recidivism among men who batter women.Journal of Police Science and Administration, 17:124–132.

Straus, Murray. 2004. Scoring the CTS2 and CTSPC. Durham, NH: Family ResearchLaboratory, University of New Hampshire. Retrieved July 7, 2006 from pubpages.unh.edu/~mas2/CTS28a3.pdf.

Syers, MaryAnn and Jeffrey L. Edleson. 1992. The combined effects of coordinatedintervention in woman abuse. Journal of Interpersonal Violence, 7:490–502.

Tolman, Richard M. and Arlene Weisz. 1995. Coordinated community interventionfor domestic violence: The effects of arrest and prosecution on recidivism ofwoman abuse perpetrators. Crime & Delinquency, 41:481–495.

\\server05\productn\C\CPP\7-4\CPP402.txt unknown Seq: 29 29-DEC-08 10:17

REDUCING INTIMATE PARTNER VIOLENCE 523

Visher, Christy A., Lisa Newmark, and Adele Harrell. 2007. Findings and lessons onimplementation (Vol. 2). Final report on the evaluation of the Judicial OversightDemonstration. Washington, DC: National Institute of Justice.

Wooldridge, Jeffrey M. 2007. Inverse probability weighted estimation for general miss-ing data problems. Journal of Econometrics, 141:1281–1301.

Christy A. Visher, Ph.D., is a professor in the Department of Sociology and CriminalJustice at the University of Delaware and a principal research associate at the UrbanInstitute in Washington, DC. She is also codirector of the Center for Drug and AlcoholStudies at the University of Delaware. She was coprincipal investigator of the evalua-tion of the Judicial Oversight Demonstration Initiative at the Urban Institute. Herresearch includes evaluations of strategies for crime control and prevention, prisonerreentry, substance use, and communities and crime. She is currently coprincipal investi-gator of the Multi-Site Evaluation of the Serious and Violent Offender ReentryInitiative, an examination of state and local reentry programs funded by the federalgovernment. She received her Ph.D. in sociology from Indiana University.

Adele Harrell, Ph.D., formerly the director of the Justice Policy Center at the UrbanInstitute, is currently a member of the Sentencing and Criminal Code Reform Commis-sion of the District of Columbia and is serving as a member of a National Academy ofSciences Panel. She was the lead investigator for the evaluation of the Judicial Over-sight Demonstration Initiative. She is a member of the American Society ofCriminology and a fellow of the Academy of Experimental Criminology. She receivedher Ph.D. in sociology from the George Washington University.

Lisa Newmark, Ph.D., is a term assistant professor in the Administration of JusticeDepartment at George Mason University. She was coprincipal investigator of the evalu-ation of the Judicial Oversight Demonstration Initiative at the Urban Institute. While asenior research associate at the Urban Institute, Newmark conducted evaluations ofvictim assistance and compensation programs and evaluations of specialized programsfor victims of domestic violence, sexual assault, trafficking, and prison rape. Herresearch interests focus on criminal justice and social services to assist victims of a widerange of crimes.

Jennifer Yahner, M.A., is a research associate in the Justice Policy Center at theUrban Institute. She has been conducting criminal justice research for nearly a decadeand has experience with issues surrounding prisoner reentry, intimate partner violence,and high-risk youth program evaluation. She is skilled in large-scale data managementand statistical analysis using a variety of quantitative methodologies. Yahner receivedher Master’s degree in criminology and criminal justice at the University of Maryland.