research cj summary

article1research.pdf

Home >Education homework help >research cj summary

Police Practice and Research, Vol. 6, No. 3, July 2005, pp. 261–278

Beyond M andat ory Arrest : Developing a Comprehensive Response t o Domest ic Violence M ichael D. Whit e, John S. Goldkamp & Suzanne P. Campbell

Taylor and Francis LtdGPPR120611.sgm10.1080/15614260500206285Police Practice and Research1561-4263 (print)/1477-271X (online)Original Article2005Taylor & Francis Group Ltd63000000July 2005Dr MichaelWhiteDepartment of Law, Police Science and Criminal Justice AdministrationJohn Jay College of Criminal Justice899 Tenth AvenueNew YorkNY 10019USA+1 212 [email protected] wisdom about the appropriate police response to domestic violence has changed dramatically in recent years, characterized by an increasing move toward manda- tory arrest. This paper examines an effort by the Vacaville, California Police Department to provide a more comprehensive response to domestic violence, through an innovative law enforcement, clinical, and prosecutorial partnership (called FIRST). The paper employs interrupted time series analysis (ARIMA) with monthly domestic violence arrest data from 1990 to 2000 to investigate the impact of the program. ARIMA results are considered in the context of more general crime trends. Results suggest that the onset of the program coin- cided with an initial increase in arrests, followed by a longer-term decrease. Although other potential explanations could not be eliminated, findings indicate that the FIRST program played a contributing role in the reduction of domestic violence in Vacaville.

Keywords: Policing Domestic Violence; Domestic Violence; Mandatory Arrest

Introduction

This paper examines the impact of an effort by the Vacaville, California Police Depart- ment (VPD) to address domestic violence through a comprehensive and coordinated response. The multi-agency, multi-disciplinary team, called the Family Investigative Response Services Team (FIRST), seeks to address the intricate dynamics of domestic violence, increase successful prosecution of offenders, and reduce their recidivism. This paper examines whether this innovative domestic violence program affected the number of domestic violence arrests made by police. The paper first examines general

Correspondence to: Michael D. White, PhD, John Jay College of Criminal Justice, 899 Tenth Avenue, New York, NY 10019, USA. Email: [email protected]

262 M. D. White et al.

crime trends in Vacaville from 1990 to 2000 to provide an understanding of how crime has changed in the location over time. Interrupted time series analysis (ARIMA) is then employed with monthly domestic violence arrest data in Vacaville from 1990 to 2000 to examine the impact of the program on the target problem. Findings from ARIMA are considered in the context of general crime trends in order to determine if any changes in domestic violence arrests are part of larger shifts in crime prevalence.

Conventional wisdom about the appropriate police response to domestic violence has changed dramatically over the last few decades, from mediation and less formal responses (i.e., a private matter not requiring police intervention) in the 1960s and 1970s to an increasing move toward mandatory arrest (particularly for felony cases) in the 1980s and 1990s. The shift toward mandatory arrest of domestic violence offenders has raised additional problems for the police specifically and the justice system generally, and as a result, some jurisdictions, including Vacaville, have developed more compre- hensive approaches to domestic violence, coordinating law enforcement and prosecu- torial objectives with efforts to meet victims’ needs. This paper seeks to inform the debate about how the police should respond to domestic violence, and whether the adoption of a more comprehensive, coordinated approach can effectively reduce its prevalence.

Prior Research on the Police Response to Domestic Violence

The Traditional Police Response to Domestic Violence

Police have traditionally under-enforced laws prohibiting domestic violence (Black, 1980; Buzawa & Buzawa, 2001; Erez, 1986; Sherman, 1992). In 1967, the International Association of Chiefs of Police stated in its manual that arrest should only be employed as a last resort’ (International Association of Chiefs of Police, 1967; also Sherman, 1992). The American Bar Association developed a similar policy in their 1973 Stan- dards relating to the urban police function. The under-enforcement of domestic violence laws by police stems from a number of issues, including traditional limitations on police arrest powers, particularly for incidents involving misdemeanor offenses, the fact that most domestic dispute calls do not involve violence,1 victim preference against arrest, lack of training, and the belief held by most police that domestic disturbance calls are more dangerous than other types of calls (making arrest a less attractive option) (Buzawa & Buzawa, 2001; Sherman, 1992).

The Changing Police Response to Domestic Violence

The first effort at changing the traditional police response to domestic violence occurred in 1965, when the New York City Police Department created the Family Crisis Intervention Unit (FCIU) (Bard, 1970). Officers, specially trained in conflict manage- ment and mediation, were encouraged to take as much time as needed to negotiate the dispute, and to make referrals to appropriate social service agencies (Bard, 1970; Sherman, 1992). Bard (1970) later reported that the project led to lower rates of family assaults, homicides, and assaults on police officers, though the conclusions were later

Police Practice and Research: An International Journal 263

deemed unsubstantiated by the data (see Liebman & Schwartz, 1973). Nevertheless, a number of police departments across the country followed suit and created specialized domestic violence crisis or mediation units.

The Shift to Mandatory Arrest for Domestic Violence Offenders

During the late 1970s and early 1980s, a number of forces emerged that led to a shift toward mandatory arrest for domestic violence offenders. In the early 1970s, the femi- nist movement targeted spousal assault as a major women’s issue, and by the end of the decade, hotlines and shelters for battered women had been established across the coun- try (Sherman, 1992). The perception that police were discriminating against women victims led to two class-action law suits against the police, in New York City and Oakland (Sherman, 1992).2

In 1978, the US Supreme Court ruled in Monell v. Department of Social Services of the City of New York, 436 U.S. 658, when a representative of an agency violates an individ- ual’s constitutional rights because of the agency’s official custom and practice, the agency as well as the individual may be held liable (Skolnick & Fyfe, 1993).3 The ruling in Monell put local governments at risk of being forced to pay large civil judgments as a result of the behavior of individual officials (Skolnick & Fyfe, 1993).

In 1984, in Thurman v. Torrington, 595 F. Supp. 1521, the Monell ruling was applied to police behavior in domestic violence calls.4 The Thurman case extended police liabil- ity to acts of omission, indicating that failure to protect victims of domestic violence (arguably, the traditional police response) could result in civil litigation and large civil verdicts against the department and city (Skolnick & Fyfe, 1993).

Last, in 1984, results from the Minneapolis Domestic Violence study were published, which indicated that formal arrest led to significantly lower rates of recidivism among offenders (as compared to other less formal police responses) (Sherman & Berk, 1984). Within months of publication of the findings, ‘police departments across the nation began to adopt either mandatory or presumptive arrest policies for cases of marital assault’ (Gelles, 1996, p. 30). Drawing heavily on the results from the Minneapolis study, the Attorney General’s Task Force on Family Violence (US Department of Justice, 1984) recommended that family violence be recognized and treated as criminal activity (Gelles, 1996).

Moving Beyond Mandatory Arrest

The National Institute of Justice funded replications of the Minneapolis experiment in Atlanta, Omaha, Charlotte (North Carolina), Milwaukee, Colorado Springs, and Metro-Dade (Miami). Results from the replications clearly challenged the external validity of the Minneapolis findings;5 specifically, results in three studies indicated that arrest led to increased frequency of subsequent domestic violence offenses (Buzawa & Buzawa, 2001; Dunford, Huizinga, & Elliott, 1989; Garner, Fagan, & Maxwell, 1995; Hirschel, Hutchinson, Kelley, & Pesackis, 1990; Sherman, 1992; Sherman et al., 1991). Results from the original study and replications suggest that

264 M. D. White et al.

arrest has a differential impact on offenders, and Sherman (1992) argues that the impact of formal arrest is affected by the individual offender’s stake in conformity.6 However, in their review of domestic violence arrest evaluations, Garner et al. (1995) caution that the potential for reassault escalation, an increase in violence after program participation and probation has ended, should be considered to meaning- fully evaluate such programs.

Several researchers have identified inconsistent responses by prosecutors and courts as problematic for measuring the impact of arrest. Ferraro and Boychuk (1992) argue that requiring police to arrest offenders when prosecutors do not file charges and courts do not impose sentences establishes a contradictory and frustrating mandate. Hirschel, Hutchinson, Dean, and Mills (1992) note that in most of the cases in the Spouse Abuse Replication Project, prosecution did not follow arrest, and they argue that the failure to consistently prosecute batterers severely limits the potential deterrent effects of arrest. Berk, Campbell, Klap, and Western (1992) and Zorza (1994) concur, noting that arrest by itself does not work because more (rather than less) intervention is needed to produce a deterrent effect.

An unintended consequence of the shift to mandatory arrest is the police reliance on dual arrest. Research indicates that, in many jurisdictions, the percentage of women arrested for domestic violence offenses increased notably following adoption of a mandatory arrest policy (Martin, 1997; Miller, 2001). Miller (2001, p. 1340) argues, however, that reliance on dual arrest fails to take into account the different ‘context in which victims of violence resort to using violence themselves.’ Zorza (1992) notes that some jurisdictions have responded to this issue by adopting ‘primary aggressor’ laws and by modifying police training and policy.

Given the mixed findings on mandatory arrest and its unintended consequences, some attention has re-focused on the victim (and victim preference in resolving the encounter). Research shows that the majority of victims of domestic violence either refused to make a statement or recanted the statement later, often before the offender is charged (Buzawa & Buzawa, 1996; Cretney & Davis, 1995; Hoyle, 1998; Hoyle & Sanders, 2000).7 Hoyle and Sanders (2000, p. 19) also note the importance of victim empowerment, which seeks to

put in place those conditions which enable women to best understand what is in their interests, and encourage them to act accordingly, and then to support them in the choices which they have made—whether these choices include invoking criminal justice interven- tion or not.

Comprehensive Approaches to Domestic Violence

In several jurisdictions across the country, police have attempted to blend elements of each of these perspectives, employing mandatory arrest policies but also making an effort to establish a relationship with the victim and to connect her with a range of services that can help address her current situation. This approach often involves plac- ing police officers with social workers as a crisis intervention team, with police focusing on arrest and building a case for the prosecutor while the social workers provide

Police Practice and Research: An International Journal 265

services to the victim including counseling, advocacy, shelter, support, and legal aid (Kramer & Black, 1998).

Research on this type of domestic violence intervention has generally produced positive findings (Buzawa & Buzawa, 2001).8 Carr (1982) studied a crisis intervention team in Pawtucket, Rhode Island and reported that 79% of victims found the services helpful. Corcoran, Stephenson, and Perryman (2001, p. 397) surveyed police about their opinions of a Domestic Violence Response Team and found responses to be ‘overwhelmingly positive.’ Davis and Taylor (1997) evaluated a joint police–social services approach in New York City and found that the approach did not lead to a reduction in repeat violence, though victims receiving the experimental treatment were more likely to report new violence (which the authors attributed to increased victim confidence in the police response).

Gamache, Edleson, and Schock (1988) evaluated community intervention projects (CIPs) involving coordinated police, judicial, and social service intervention in three suburban communities. Intervention in all three communities resulted in statistically significant changes in police and court responses, including an increase in arrests and the use of mandated counseling. Similar positive results have been reported by Caputo (1988), Edleson and Grusznski (1988), Edleson and Syers (1991), Kramer and Black (1998), and Shepard (1992).

Although Buzawa and Buzawa (2001, p. 229) acknowledge that such programs can ‘dramatically affect the cycle of abuse,’ they note that the research suggests ‘there is a core group of abusers that are not, and apparently will not, be deterred’ (see also Gondolf, 2000). Buzawa and Buzawa (2001) also note that proactive comprehensive policies may have many negative, unintended consequences for the victim and offender (e.g., judicial intervention being used as a weapon against the victim, increased violence and harassment, victims are stigmatized), and for the agencies involved (e.g., displace- ment of resources, increased violence toward police and court personnel) that must be fully explored.

The Vacaville, California Response to Domestic Violence

Vacaville, California and the Target Problem

Vacaville, California is a relatively homogeneous community of 90,000 residents located approximately halfway between San Francisco and Sacramento. Often described as a ‘bedroom community,’ Vacaville is home to a large military base and state prison. Its population, according to the 2000 census, is 72.1% white, 17.9% Hispanic, and 10% African-American.

Throughout the early 1990s, it became clear to the VPD leadership that domestic violence comprised a substantial portion of the department’s calls for service.9 In 1995, for example, there were 988 domestic violence incidents comprising nearly 60% of the VPD’s calls for service involving violent crime. Also, Vacaville police officials noted that local prosecution of family violence cases was being hindered as victims would frequently recant their stories. Moreover, officials estimate that nearly 90% of the

266 M. D. White et al.

domestic violence cases filed in the first half of 1996 were affected by intimidation of either the victims or their children.

The Family Investigative Response Services Team (FIRST)

In June 1996, the VPD created the Family Investigative Response Services Team (FIRST), a multi-agency, multi-disciplinary team, designed to prevent batterers from intimidating victims, and to use master social workers to initiate a relationship and stabilize the victim and family while the District Attorney’s (DA) office pursued prosecution.10 Program goals included providing services for victims,11 even if they were hostile, establishing relationships with victims and children for at least one year including six months of intensive involvement, meeting basic needs of victims through ‘one-stop’ collaboration, preparing victims for court, and reducing repeat calls for service.

FIRST includes detectives from the VPD, master social workers, family support workers, and representatives from the DA’s office and Probation. Importantly, the master social workers and family support workers are employees of the VPD, located in the same office with detectives, working simultaneously with police on the cases that come to the unit with the same overall goals guiding their responses. Moreover, FIRST adopted victimless prosecution as a goal, which sought to allow the DA to prosecute domestic violence cases successfully without reliance on victim testimony or coopera- tion.12 Thorough investigation and evidence collection including pictures, medical reports, dispatch tapes, witness statements, and forensic interviews take the burden off of the victim and seek to reduce tension in the relationship, should the victim and the offender remain together.13

This research examines the impact of the FIRST program by examining the flow of domestic violence arrests over an extended period of time and considering the imple- mentation of the program in the context of long-term trends in arrests. Reductions in domestic violence arrests following the creation of FIRST would suggest that the program may have had an impact on the target problem, and more generally, would continue to shed light on the potential impact of comprehensive approaches on the prevalence of domestic violence.

Methodology

Interrupted Time Series Analysis

This paper studies all domestic violence arrests in Vacaville, California for a period of 11 years, from 1990 through 2000 (2,102 arrests).14 The paper employs interrupted time series analysis (ARIMA) with monthly domestic violence arrests to examine the impact of the FIRST program. Interrupted time series analysis, a strong quasi- experimental research design (see Taylor, 1994), is a well-established analytic tool that is often employed for such purposes. The goal of time series analysis is to account for or explain the values in the dependent variable, monthly levels of domestic violence

Police Practice and Research: An International Journal 267

arrests in this case (Babbie, 1992). Since monthly totals of domestic violence arrests serve as the unit of analysis, there are 132 data points in the time series.

ARIMA involves a two-stage process. The first stage seeks to identify a descriptive model that best captures or explains the implicit trend in the dependent variable (Box, Jenkins, & Reinsel, 1994).15 The second stage of ARIMA, called impact assessment, tests the impact of specific events, measured as interventions (an independent variable, the implementation of FIRST in this case), on the trend in the dependent variable.16 The analysis essentially seeks to determine whether the underlying trend model of the dependent measure, monthly domestic violence arrests, is consistent with changes that would be associated with the intervention being considered.

Methodological Weaknesses

The research has a number of weaknesses that must be acknowledged. First, as in other jurisdictions across the country, the VPD (and the state of California) moved from a narrow definition of domestic violence in the early 1990s to a more comprehensive definition in the later 1990s. As the state of California and the city of Vacaville adopted a more comprehensive understanding of domestic violence, expanding the definition to include relationships other than husband and wife, the number of domestic violence arrests naturally increased. The changing definition of domestic violence may also be coupled with general increases over time in violence among individuals involved in ‘domestic’ relationships. The specific contributions of each of these factors remain unknown at this point.

Second, the paper uses domestic violence arrests as its dependent variable, and there is a large body of literature describing the weaknesses of using formal arrest as a measure of the incidence of crime (Blumstein, Cohen, & Rosenfeld, 1991; Jensen & Karpos, 1993; McDowall & Loftin, 1992; Menard, 1991; Messner, 1984; O’Brien, 1985, 1990, 1996). As a result, domestic violence incidents not resulting in a police response (not reported) and incidents where police responded but did not make an arrest are not included in these data. Nevertheless, arrest remains an acceptable rough indicator of the prevalence of domestic violence in a given jurisdiction.

Third, although interrupted time series analysis is a strong quasi-experimental design, it still suffers from a number of threats to internal validity, most notably history. Ideally, this study would conduct a multivariate time series analysis, using other depen- dent variables such as city population levels, employment levels, and measures of socio- economic status. The research would chart changes in the other dependent variables over time, and their influence on levels of domestic violence arrests would become much clearer. However, in criminal justice research these data are difficult to find at the monthly level, and their use goes beyond the scope of this research, which seeks a more modest application of time series analysis to domestic violence arrest data. Nevertheless, it is possible that other, uncontrolled factors have influenced the levels of domestic violence arrests in Vacaville, and this weakness limits conclusions that can be drawn about the causal relationship between the onset of the FIRST program and changes in the level of the dependent variable (monthly domestic violence arrests).

268 M. D. White et al.

Analysis of General Crime Trends in Vacaville, 1990–2000

As an effort to address this weakness, general crime trends in Vacaville from 1990 to 2000 are examined. Data from the Federal Bureau of Investigation’s annual Uniform Crime Reports were recorded, including annual index offenses, population, and the index crime rate per 100,000 residents. This additional analysis, though superficial, helps to determine if change in domestic violence arrests is part of larger crime pattern changes in the jurisdiction, or if the domestic violence change (if there is one) is distinctive. If the trend in domestic violence arrests appears to be different from the general crime patterns in Vacaville, then the analysis provides further support for the argument that FIRST caused the change in domestic violence. If changes in domestic violence arrests appear to be related to larger crime patterns or if there is no change in the dependent variable, the evidence then contradicts the argument that FIRST affected the prevalence of domestic violence. This analysis is presented first to provide context and background for the more sophisticated ARIMA analysis.

Analysis

Figure 1 shows annual totals of domestic violence arrests in Vacaville, California, from 1990 through 2000. The annual number of arrests increases steadily from 89 in 1990 to 164 in 1992, is stable through 1994, then increases again from 1995 to 1997. After

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000

Year

100

150

200

250

300

350

N um

be r

of A

rr es

89 112

164 162 162

185

210

306

240 223

249

Crime and Justice Research Institute

Figure 1 Annual Arrests for Domestic Violence Offenses Made by the Vacaville, Califor- nia Police Department, 1990–2000. Source: Goldkamp et al. (2002).

Police Practice and Research: An International Journal 269

peaking at over 300 in 1997, the number of arrests drops notably in 1998 and 1999, before increasing to 249 in 2000.17 Table 1 shows the annual rate of domestic violence arrests per 100,000 residents during the study period, demonstrating the same general pattern. Figure 1 Annual Arrests for Domestic Violence Offenses Made by the Vacaville, California Police Department, 1990–2000. Source: Goldkamp et al. (2002).Interestingly, domestic violence arrests peak in 1997, one year after the start of the FIRST program. During the implementation stage of FIRST, the VPD leadership antic- ipated that their increased and focused attention on domestic violence may lead to temporary increases in domestic violence arrests, as victims became more willing to call for assistance and initiate a formal police response. However, as the program became more mature and began to effectively respond to domestic violence in the community, the police leadership expected a drop in overall domestic violence offenses. Superfi- cially, at least, the findings in Figure 1 and Table 1 seem to bear this out.

General Crime Patterns in Vacaville, 1990–2000

Table 2 shows index offenses known to the police in Vacaville from 1990 to 2000, as well as estimated annual population and the annual index offense rate per 100,000 resi- dents. Offenses known to the police represent a more general picture of crime trends in Vacaville while the later ARIMA analysis focuses on the more specific measure, domestic violence offenses resulting in arrest. Although there are differences in the measures being used, mandatory arrest laws in California may negate this difference (or at least minimize its impact). That is, the arrest measure for domestic violence offenses serves as a reasonable proxy for all domestic violence offenses (particularly felony offenses) because of state law and departmental rules that require officers to make arrests in such cases.18

A number of interesting findings emerge from this table. First, the population in Vacaville grew by over 17,000 during the 1990s, an increase of nearly 25%. Second, violent crime is relatively rare in Vacaville, as seen by the annual number of murders,

Table 1 Annual Rate of Domestic Violence Arrests per 100,000 Residents in Vacaville, California, 1990–2000

Year Rate per 100,000

1990 124.51 1991 153.49 1992 221.21 1993 205.85 1994 204.41 1995 221.75 1996 249.44 1997 359.08 1998 288.15 1999 263.64 2000 281.40

270 M. White et al.

Ta bl

e 2

In de

x O

ffe ns

es K

no w

n to

P ol

ic e

in V

ac av

ill e,

C al

ifo rn

ia , 1

99 0–

20 00

19 90

19 91

19 92

19 93

19 94

19 95

19 96

19 97

19 98

19 99

20 00

Po pu

la tio

n si

ze 71

,4 79

72 ,9

67 74

,1 36

78 ,6

98 79

,2 52

83 ,4

25 84

,1 88

85 ,2

17 83

,2 89

84 ,5

83 88

,4 86

C ri

m e

ra te

p er

1 00

,0 00

si de

nt s

3, 69

4. 79

4, 35

8. 13

4, 61

9. 89

4, 53

7. 60

4, 65

0. 99

40 09

.5 9

3, 68

9. 36

3, 69

1. 75

2, 98

5. 99

2, 78

3. 07

2, 57

4. 42

C ri

m e

in de

x to

ta l

2, 64

1 3,

18 0

3, 42

5 3,

57 1

3, 68

6 3,

34 5

3, 10

6 3,

14 6

2, 48

7 2,

35 4

2, 27

8 M

ur de

r/ no

n- ne

gl ig

en t

m an

sla ug

ht er

0 1

5 1

3 3

2 3

3 2

R ap

e 26

32 32

20 23

22 19

25 24

23 22

R ob

be ry

48 63

10 1

75 11

5 99

74 95

64 63

60 A

gg ra

va te

d as

sa ul

t 16

9 20

6 23

4 26

6 23

8 21

2 30

6 35

7 26

8 22

4 26

1 Bu

rg la

ry 40

3 48

8 55

4 56

3 72

4 60

2 54

0 52

7 39

2 33

3 30

1 La

rc en

y– th

ef t

1, 75

3 2,

20 6

2, 23

6 2,

40 1

2, 26

7 2,

08 9

1, 89

3 1,

84 7

1, 52

1 1,

49 4

1, 39

6 A

ut o

th ef

t 22

5 17

7 24

7 21

8 30

1 30

1 23

7 26

5 19

0 19

8 21

2 A

rs on

17 7

16 27

15 17

35 27

25 17

D at

a ar

e ta

ke n

fr om

t he

F ed

er al

B ur

ea u

of I

nv es

tig at

io n’

s U

ni fo

rm C

ri m

e R

ep or

ts , 1

99 0–

20 00

Police Practice and Research: An International Journal 271

rapes, robberies, and aggravated assaults.19 Third, the most common type of index crime in Vacaville is larceny/theft, typically representing 60–70% of the annual index offenses.

Fourth, the index crime rate per 100,000 residents shows sizeable changes in several different years.20 Specifically, the rate increases notably from 1990 to 1992, is stable through 1994, but decreases substantially in 1995 and continues to decrease through 2000. The important question for determining the impact of FIRST is whether these changes in more general crime patterns match the changes in domestic violence arrests, measured both superficially and through ARIMA.

Testing the Impact of FIRST with ARIMA

Interrupted time series analysis relies on monthly rather than annual totals of domestic violence arrests, shown in Figure 2. The first stage of the analysis involves identifying a model that ‘fits’ the data and explains trends over time.21 The identified model is (0,1,1)(1,1,0)12, regularly and seasonally differenced, with moving average and season- ally auto-regressive components.22 Figure 2 Monthly Arrests for Domestic Violence Offenses Made by the Vacaville, California Police Department, 1990–2000. Source: Goldkamp et al. (2002).In the second stage of the interrupted time series analysis, the potential impact of the onset of FIRST (the independent variable) is tested as an ‘intervention’ in time series modeling. In exploring possible impacts of FIRST, the analysis considered different types of impacts, including onset—whether gradual or abrupt—and duration— whether temporary or permanent (meaning comparatively long-lasting with no obvi- ous change in direction).23

Jan -90 Ap r-90Jul-

90 Oct -90 Jan -91 Ap r-91Jul-

91 Oct -91 Jan -92 Ap r-92Jul-

92 Oct -92 Jan -93 Ap r-93Jul-

93 Oct -93 Jan -94 Ap r-94Jul-

94 Oct -94 Jan -95 Ap r-95Jul-

95 Oct -95 Jan -96 Ap r-96Jul-

96 Oct -96 Jan -97 Ap r-97Jul-

97 Oct -97 Jan -98 Ap r-98Jul-

98 Oct -98 Jan -99 Ap r-99Jul-

99 Oct -99 Jan -00 Ap r-00Jul-

00 Oct -00

Month

N um

be r

of A

rr es

1 2

Crime and Justice Research Institute

Figure 2 Monthly Arrests for Domestic Violence Offenses Made by the Vacaville, California Police Department, 1990–2000. Source: Goldkamp et al. (2002).

272 M. D. White et al.

The analysis identified two distinct and related significant impacts. The first impact involves an increase in domestic violence arrests, beginning in June 1996 and lasting through September 1997. Given that FIRST officially started in June 1996, this impact can be considered abrupt, starting immediately after onset of the program. The analysis suggests that the abrupt increase in domestic violence arrests was temporary, continu- ing for 16 months and ending in September 1997.

This 16-month increase in domestic violence arrests is immediately followed by a permanent decrease, lasting from October 1997 through the end of December 2000 (the end of the study period). This impact is considered gradual in onset (starting 17 months after program implementation) and permanent in duration (lasting through the end of the study period). The results of the time series analysis are summa- rized in Table 3.

Interpreting ARIMA Findings in the Context of General Crime Trends

The first significant ARIMA finding indicates that there is a short-term (16 month) increase in domestic violence arrests starting in June 1996, the month FIRST was implemented. This spike in domestic violence arrests is not part of an increase in crime patterns in general, which were actually decreasing during this time (see Table 2). The second significant ARIMA finding indicates that the number of domestic violence arrests starts to decrease in late 1997 and continues through the end of the study period (December 2000). Table 2 shows that the index crime rate per 100,000 also drops nota- bly in 1998 (by more than 20%) and continues to drop in the last two years of the study period (an additional 14%). Thus, this shift in domestic violence arrests matches shifts in more general crime patterns in Vacaville.

Establishing a Causal Link between FIRST and Levels of Domestic Violence Arrests

The strong quasi-experimental design employed here clearly points to an association between the onset of FIRST and changes in levels of domestic violence arrests, but conclusions about a causal relationship between the independent and dependent variables are more difficult to draw. Although ARIMA is a sound methodological

Table 3 Summary Results from Interrupted Time Series Analysis with Domestic Violence Arrests in Vacaville, California, 1990–2000

Model Impact 1 Impact 2

Model characteristics (0,1,1) (1,1,0) 6/96–9/97 10/97–12/00 AIC 146.77 141.60 144.82 SBC 152.33 149.94 153.16 B – 0.520 −0.564

Impact probability – 0.007 0.049 Impact type – Abrupt temporary Gradual permanent

Police Practice and Research: An International Journal 273

tool, it suffers from a number of threats to internal validity that challenge causal inferences. The immediate increase in the dependent variable may be explained by a number of factors not associated with FIRST, such as changes in larger social forces in Vacaville including employment levels, racial makeup, age of residents, and the general health of the economy. In fact, review of Figure 1 suggests that annual domestic violence arrests had been increasing in the years prior to the start of the program. As a result, some portion of the impact may be explained by the increased incidence of domestic violence in Vacaville, independent of the start of FIRST. Also, the increase in arrests following program implementation may simply be a result of police officers being more aware of the problem and being more likely to make arrests.

The drop in domestic violence arrests identified through ARIMA may also be inde- pendent of the FIRST program. That is, in 1997 there were an unusually high number of domestic violence arrests, and perhaps the years 1998–2000 are marked by a return to normal levels of such violence in Vacaville. Or perhaps improvements in other domestic violence-related programs in the community led to the decrease in arrests (e.g., improved batterer intervention programs, changes in shelter programs).

One competing explanation to the causal impact of FIRST is that the changes in domestic violence arrests were part of larger shifts in crime patterns in Vacaville. Table 2 tested this alternative explanation, charting index offenses known to police and index crime rates per 100,000 residents in Vacaville from 1990 to 2000. Findings from this analysis were mixed. Although domestic violence arrests and the index crime rate per 100,000 residents appear similar through the early 1990s, domestic violence arrests increase notably from 1994 to 1997, by 90%. This shift in domestic violence arrests is clearly different from the index crime rate per 100,000 residents, which drops more than 20% over the same four-year period. As a result, the first ARIMA finding is not explained by shifts in more general crime patterns in Vacaville.

However, both domestic violence arrests and the index crime rate per 100,000 resi- dents drop notably after 1997. As a result, it is difficult to know whether the FIRST program caused the longer-term decrease in domestic violence arrests, or whether it is simply part of more general shifts in crime patterns.

Conclusion

The VPD’s FIRST program represents an effort to extend the response to domestic violence beyond mandatory arrest. Although the research did not provide definitive proof of a causal relationship between the program and the changing trends in domestic violence arrests, it appears that the increase in arrests is explained, at least in part, by an increased willingness of victims to call for police assistance and by an improved response by the VPD. Findings are less clear with regard to the subsequent decrease in arrests, but it seems reasonable to consider that decrease in the context of the primary goals of the FIRST program, which center on reducing the prevalence of domestic violence. Consequently, the Vacaville experience offers a number of lessons to police departments that continue to struggle to respond effectively and comprehensively to domestic

274 M. D. White et al.

violence. First, as state law (and police policy) requires in nearly all jurisdictions in the USA, formal arrest in felony domestic violence incidents, and in many cases misde- meanor incidents, is an essential first ingredient to an effective police response. This crit- ical first step initiates the criminal process and serves as a catalyst for implementing the rest of the elements of the response.

Second, FIRST recognizes the importance of engaging the victim in the process, addressing her needs, and honoring her decisions about the relationship.24 Impor- tantly, master social workers and family support workers are employees of the VPD, sharing office space and working cooperatively with detectives who are building cases for prosecution. Adding victim support workers with clinical training to the police team, rather than referring out for services, improves communication among counse- lors and detectives, improves recognition and understanding of the needs of each party by the other, and leads to an improved response to the needs of the victim.

Third, the cooperation between counselors and detectives is enhanced by the emphasis on victimless prosecution. The approach takes the burden for successful prosecution off of the victim. At the same time, the emphasis on collection of addi- tional evidence beyond the victim’s statements (such as photographs, medical reports, and witness statements) greatly improves the likelihood of successful prosecution as the issue of victim recanting is taken out of the equation.25

Fourth, a representative from the DA’s office should be a part of the police-led domestic violence response team. Although this element requires a substantial commitment by the DA’s office, the daily contact between police, counselors, and the Assistant District Attorney (ADA) results in open and frank discussion of cases, often identifying weaknesses in particular cases that require additional investigation and evidence collection. Full and open communication between the detectives and the ADA may increase the likelihood of a coordinated, consistent response to domestic violence and may serve to reduce tension among officials from two agencies that tradi- tionally have different goals and often do not see eye to eye on how domestic violence cases should be handled. Also, having an ADA on the team serves as an important avenue for feedback on the success of the police response, where detectives (and even patrol officers) can know about the final outcomes of cases and can analyze why successful prosecution was not achieved in some instances. Moreover, critiques of the Spouse Assault Replication Project cite low prosecution rates as a confounding factor in determining the impact of arrest. Collaboration between police and the prosecutor may enhance the potential deterrent effect of formal intervention (see Berk et al., 1992; Hirschel et al., 1992; Zorza, 1994).

Finally, the experience in Vacaville shows that police departments who adopt a comprehensive response to domestic violence should expect a short-term increase in the target problem (measured through arrests). This increase should be seen as a posi- tive sign, as victims become more willing to call for police assistance and police respond in a more effective, coordinated manner. If the comprehensive response is effectively achieving its objectives (as appears to be the case in Vacaville), this short-term increase should eventually subside and be followed by a decrease in the prevalence of domestic violence arrests.

Police Practice and Research: An International Journal 275

Notes 1 [1] Elliott (1989) found that one-third of domestic disturbance calls involved some form of

violence. 2 [2] In the New York case, police agreed that they would not take into account gender of the

victim or relationship to the offender when making arrest decisions (Bruno v. Codd, 74 N.Y. 2d 582, 393 N.E. 2d 976, 419 N.Y.S. 2d 901, 1979). Although the case resulted in little change to police practice, Sherman (1992) notes that it represented a major symbolic victory.

3 [3] Monell was a pregnant employee who was denied maternity leave and forced to resign because of department policy. However, the Court defined custom and practice broadly enough to include whatever officials routinely do, whether in line with official policy or not.

4 [4] Torrington police typically did not make arrests in domestic disputes. Police officers were at the scene and failed to stop the assailant from stabbing and kicking Thurman in the head. In fact, the husband was not taken into custody until after he approached his wife again as she lay on a stretcher, bleeding from knife wounds to the chest, neck, and throat (Sherman, 1992). Thurman sued the police department and was awarded $2.3 million.

5 [5] There were also a number of methodological weaknesses challenging the internal validity of the Minneapolis study. See Gelles (1996) and Sherman (1992) for a discussion of those weak- nesses.

6 [6] Specifically, Sherman (1992) argues that offenders with high stake in conformity (those with much to lose) are less likely to re-offend after arrest than those with low stake in conformity.

7 [7] There are a number of arguments against allowing victim preference to dictate the police response to domestic violence. See Hoyle and Sanders (2000) for a discussion of those arguments.

8 [8] Many of the evaluations of police interventions with domestic violence, including the current study, use arrest as the dependent variable. As a result, domestic violence incidents not coming to the attention of the police, or not resulting in arrest, are not captured in the data. See the discussion of this weakness in the Methodology section.

9 [9] As in other places around the country, the VPD adopted a more inclusive definition of domestic violence in the mid-1990s. This more expansive definition of domestic violence likely contributed to an increase in documented domestic violence arrests.

10[10] A portion of the funding for FIRST came from the Bureau of Justice Assistance’s Open Solici- tation Program. BJA created the program in 1997 to identify and support efforts to improve criminal justice on the local level. One of the objectives of the Open Solicitation Program was to support outstanding local initiatives and to help incorporate ways to measure the progress and impact of the local innovation. A demanding peer review process resulted in the selection of 37 jurisdictions for awards ranging from $50,000 to $150,000, including Vacaville (one site later declined the funding).

11[11] Clinical staff provide a range of services and interventions for victims such as Parent Project, an intensive 10-week course targeting impulsive, difficult-to-manage children; At-Risk ID Program, which seeks to identify children and adults at risk for mental or physical conditions, autism, and Alzheimer’s disease; Bridges to Change support group for female victims of domestic violence; a social skills group; a teen support group; women and children support groups; Spanish-speaking women’s group; 40-hour home visitor training; Homework Club; Positive Parenting Project; HUGS group, ‘Helping you grow safely,’ for children witnessing domestic violence; leadership skills group; and an anger management class. See Goldkamp, White, and Weiland (2002) for a more detailed discussion of the interventions. Although clin- ical staff are employees of the police department, the victim–counselor relationship is consid- ered privileged (i.e., information conveyed by the victim is confidential).

12[12] The emphasis on victimless prosecution arose from concerns regarding the prevalence of victims recanting and/or becoming hostile witnesses.

13[13] Although the VPD plays a leadership role in the comprehensive strategy, the police response to domestic violence still centers on ‘traditional’ activities (investigation and arrest). The real

276 M. D. White et al.

change for the police involves the degree of collaboration and cooperation with other agen- cies, as they carry out their traditional responsibilities (though not overly traditional for domestic violence). However, the Vacaville approach is unique in that the police department employs clinical staff to work directly with their detectives on domestic violence cases.

14[14] The VPD Crime Analyst provided the arrest data. These data represent all arrests in domestic violence cases during the study period.

15[15] The three-stage model building process includes identification, estimation, and diagnosis (Box et al., 1994). In the identification stage, the auto-correlation and partial auto-correlation functions are examined. In estimation, parameter estimates are identified. Finally, the last stage involves examining the error residuals.

16[16] Impact assessment is conducted by adding the intervention as a dummy variable to the exist- ing model identified in the first stage of ARIMA. Interventions vary by onset, abrupt or grad- ual, and duration, temporary or permanent.

17[17] The lower arrest totals in the early 1990s are at least partially explained by unclear and/or narrower definitions of domestic violence employed at that time.

18[18] Of course, the arrest measure fails to account for domestic violence offenses not coming to the attention of the police.

19[19] Some portion of the domestic violence offenses resulting in arrest, the dependent variable in the ARIMA analysis, is included in the aggravated assault category.

20[20] The index crime rate is calculated by dividing the number of index offenses known to police by the annual population, then multiplying that figure by 100,000.

21[21] The data are logged with no constant. 22[22] The seasonal component suggests that the number of domestic violence arrests seems to vary

during certain times of the year, and that variation is consistent over several years. Analysis of the monthly data indicates that arrests seem less common during the winter months but typi- cally peak during the summer months. Moving average and auto-regressive components describe the processes that affect each observation, suggesting that previous observations (i.e., arrest totals from previous months) help determine each month’s arrest total (see Box et al., 1994, for a more detailed discussion of ARIMA modeling).

23[23] Key measures of impact include the probability of the intervention component, and the AIC (Akaike Information Criterion) and SBC (Schwartz Bayesian Criterion) values. Significant impacts are characterized by an intervention with a probability below 0.05 and AIC and SBC values that are lower than the values in the original model. The size and direction of the B- weight describe the nature of the impact (strength and decrease or increase in the enrollment measure). See Box et al. (1984) for a more complete discussion of impact assessment.

24[24] In Vacaville during the study period, over 90% of domestic violence victims were female. 25[25] Victim satisfaction with the police response may still be problematic, since many victims

oppose arrest of the offender.

References

American Bar Association, Project on Standards for Criminal Justice. (1973). Standards relating to the urban police function. Chicago: American Bar Association.

Babbie, E. (1992). The practice of social research. Belmont, CA: Wadsworth. Bard, M. (1970). Training police as specialists in family crisis intervention. Washington, DC: US

Department of Justice. Berk, R. A., Campbell, A., Klap, R., & Western, B. (1992). The deterrent effect of arrest in incidents

of domestic violence: A Bayesian analysis of four field experiments. American Sociological Review, 57, 698–708.

Black, D. (1980). The manners and customs of the police. New York: Academic Press.

Police Practice and Research: An International Journal 277

Blumstein, A., Cohen, J., & Rosenfeld, R. (1991). Trend and deviation in crime rates: A comparison of UCR and NCS data for burglary and robbery. Criminology, 29(2), 237–263.

Box, G. E. P., Jenkins, G. M., & Reinsel, G. C. (1994). Time series analysis: Forecasting and control. Upper Saddle River, NJ: Prentice Hall.

Bruno v. Codd, 74 N.Y. 2d 582, 393 N.E. 2d 976, 419 N.Y.S. 2d 901 (1979). Buzawa, E., & Buzawa, C. (Eds.). (1996). Do arrests and restraining orders work? Beverly Hills, CA:

Sage. Buzawa, E., & Buzawa, C. (2001). Traditional and innovative police responses to domestic violence.

In R. G. Dunham & G. A. Alpert (Eds.), Critical issues in policing (4th ed.). Prospect Heights, IL: Waveland Press.

Caputo, R. (1988). Police response to domestic violence. Social Casework, 69, 81–87. Carr, J. (1982). The abusive partner: An analysis of domestic battering. New York: Van Nostrand

Reinhold. Corcoran, J., Stephenson, M., & Perryman, D. (2001). Perceptions and utilization of a police–social

work crisis intervention approach to domestic violence. Families in Society, 82(4), 393–398. Cretney, A., & Davis, G. (1995). Punishing violence. London: Routledge. Davis, R. C., & Taylor, B. G. (1997). A proactive response to family violence: The results of a

randomized experiment. Criminology, 35(2), 307–333. Dunford, F. W., Huizinga, D., & Elliott, D. (1989). The role of arrest in domestic assault: The Omaha

police experiment. Criminology, 28(2), 183. Edleson, J. L., & Grusznski, R. J. (1988). Treating men who batter: Four years of outcome data from

the Domestic Abuse Project. Journal of Social Service Research, 12(1/2), 3–22. Edleson, J. L., & Syers, M. (1991). The effects of group treatment for men who batter: An 18 month

follow-up study. Research on Social Work Practice, 1(3), 227–243. Elliott, D. S. (1989). Criminal justice procedures in family violence crimes. In N. Morris & M. Tonry

(Eds.), Crime and justice: An annual review of research (Vol. 11, pp. 427–480). Chicago and London: University of Chicago Press.

Erez, E. (1986). Intimacy, violence and the police. Human Relations, 39, 265–281. Ferraro, K. J., & Boychuk, T. (1992). The court’s response to interpersonal violence: A comparison

of intimate and nonintimate assault. In E. Buzawa & C. Buzawa (Eds.), Domestic violence. Westport, CT: Auburn House.

Gamache, D. J., Edleson, J. L., & Schock, M. D. (1988). Coordinated police, judicial, and social service response to woman battering: A multiple-baseline evaluation across three communities. In G. T. Hotaling, D. Finkelhor, J. T. Kirkpatrick, & M. A. Straus (Eds.), Coping with family violence. Newbury Park, CA: Sage.

Garner, J., Fagan, J., & Maxwell, C. (1995). Published findings from the spouse assault replication program: A critical review. Journal of Quantitative Criminology, 11, 3–28.

Gelles, R. J. (1996). Constraints against family violence: How well do they work? In E. Buzawa & C. Buzawa (Eds.), Do arrests and restraining orders work? (pp. 30–42). Thousand Oaks, CA: Sage.

Goldkamp, J. S., White, M. D., & Weiland, D. (2002). Implementing local criminal justice strategies: Final technical assistance report for the original 36 open solicitation sites. Philadelphia: Crime and Justice Research Institute.

Gondolf, E. W. (2000). A 30-month follow-up of court-referred batterers in four cities. International Journal of Offender Therapy and Comparative Criminology, 44(1), 111–128.

Hirschel, J. D., Hutchinson, C. D., III, Dean, C., & Mills, A. M. (1992). Review essay on the law enforcement response to spouse abuse: Past, present, and future. Justice Quarterly, 9, 247–283.

Hirschel, J. D., Hutchinson, C. D., III, Kelley, J. J., & Pesackis, C. E. (1990). Charlotte spouse assault replication project: Final report. Washington, DC: National Institute of Justice.

Hoyle, C. (1998). Negotiating domestic violence: Police, criminal justice and victims. Oxford: Oxford University Press.

278 M. D. White et al.

Hoyle, C., & Sanders, A. (2000). Police response to domestic violence: From victim choice to victim empowerment. British Journal of Criminology, 40(1), 14–36.

International Association of Chiefs of Police. (1967). Training key 16: Handling disturbance calls. Gaithersburg, MD: IACP.

Jensen, G. F., & Karpos, M. A. (1993). Managing rape: Exploratory research on the behavior of rape statistics. Criminology, 31(3), 363–385.

Kramer, L. C., & Black, H. (1998). DVERTing domestic violence: The domestic violence enhanced response team. FBI Law Enforcement Bulletin, 67(6), 22–26.

Liebman, D. A., & Schwartz, J. (1973). Police programs in domestic crisis intervention: A review. In J. R. Snibbe & H. M. Snibbe (Eds.), The urban policeman in transition (pp. 421–472). Springfield, IL: Charles C. Thomas.

Martin, M. E. (1997). Double your trouble: Dual arrest in family violence. Journal of Family Violence, 5, 51–64.

McDowall, D., & Loftin, C. (1992). Comparing the UCR and NCS over time. Criminology, 30(1), 125–132.

Menard, S. (1991). Encouraging news for criminologists (in the year 2050)? A comment on O’Brien (1990). Journal of Criminal Justice, 19(6), 563.

Messner, S. F. (1984). The ‘dark figure’ and composite indices of crime: Some empirical explorations of alternative data sources. Journal of Criminal Justice, 12, 435–444.

Miller, S. L. (2001). The paradox of women arrested for domestic violence. Violence Against Women, 7(12), 1339–1376.

Monell v. Department of Social Services of the City of New York, 436 U.S. 658 (1978). O’Brien, R. M. (1985). Crime and victimization data. Beverly Hills, CA: Sage. O’Brien, R. M. (1990). Comparing detrended UCR and NCS crime rates over time: 1973–1986.

Journal of Criminal Justice, 18, 229–238. O’Brien, R. M. (1996). Police productivity and crime rates: 1973–1992. Criminology, 34(2), 183–207. Shepard, M. (1992). Predicting batterer recidivism five years after community intervention. Journal

of Family Violence, 7(3), 167–178. Sherman, L. W. (1992). Policing domestic violence: Experiments and dilemmas. New York: Free Press. Sherman, L. W., & Berk, R. A. (1984). The specific deterrent effects of arrest for domestic assault.

American Sociological Review, 49(2), 261–272. Sherman, L. W., Schmidt, J. D., Rogan, D. P., Gartin, P. R., Cohn, E. G., Collins, D. J., et al. (1991).

From initial deterrence to long-term escalation: The effects of short-custody arrest for poverty ghetto domestic violence. Criminology, 29(4), 821–850.

Skolnick, J. H., & Fyfe, J. J. (1993). Above the law: Police and the excessive use of force. New York: Free Press.

Taylor, R. B. (1994). Research methods in criminal justice. New York: McGraw-Hill. Thurman v. Torrington, 595 F. Supp. 1521 (1984). US Department of Justice. (1984). Attorney general’s task force on family violence: Report. Washington,

DC: US Department of Justice. Zorza, J. (1992). The criminal law of misdemeanor domestic violence, 1970–1990. Journal of Criminal

Law and Criminology, 83, 46–72. Zorza, J. (1994). Must we stop arresting batterers? Analysis and policy implications of new police

domestic violence studies. New England Law Review, 28, 929–990.