Deterring Future Incidents of Intimate Partner Violence

Abstract

Few studies examine the comparative effectiveness of different formal interventions for domestic violence. Using arrest and civil protection order data, we compare three intervention scenarios (arrest, civil protection order, and both). Results suggest that intervention type has no substantive influence on the odds of reoffending. However, subsequent domestic violence is significantly associated with offender age, sex, and prior offense history as well as victim age and sex. We discuss our findings and their policy implications, noting that responding agencies should be sensitive to the characteristics that increase the odds of reoffending among those they come into contact with.

Keywords

intimate partner violence domestic violence protection orders

Introduction

There are an estimated 7 million people physically assaulted by an intimate partner annually in the United States (Tjaden & Thoennes, 2000). The majority of these assaults go unreported in official crime statistics, with victims who seek help recruiting it from family, friends, or local social service providers. Still, a non-trivial number of intimate partner violence victims seek some kind of formal intervention from law enforcement or the courts. The system of formal controls in place to intervene in cases of intimate partner violence can include victim and offender interaction with police, criminal courts, and civil courts. At times, offenders and victims involved in intimate partner violence are simultaneously engaged with all three of these systems. According to the National Violence Against Women Survey, approximately 30% of incidents involving rape, physical assault, or stalking by an intimate partner result in police notification, an estimated 1.9 million cases annually (Tjaden & Thoennes, 2000). Of incidents known to police, roughly 32% result in the arrest or detention of the offender. An estimated 7% of incidents result in criminal prosecution. Seventeen percent of female victims of intimate partner assault and 16% of female victims of rape in the United States procure a civil protection order (Tjaden & Thoennes, 2000). Victims are most likely to seek these orders (and possibly other forms of intervention) when exposure to violence is repetitive and/or particularly serious (Carlson, Harris, & Holden, 1999; Gondolf, McWilliams, Hart, & Stuehling, 1994; Jordan, 2004; Sabina & Tindale, 2008; Zoellner et al., 2000).

Intimate partner violence is a particularly acute problem in the state of New Mexico, with reported incidence rates close to 3 times the national average. In 2013, law enforcement agencies across the state responded to 18,954 incidents of intimate partner violence, an incidence rate of 9 per 1,000 persons (Caponera, 2014). Comparatively, the intimate partner violence incidence rate nationally was 3.6 per 1,000 persons in 2010 (Catalano, 2012). This national estimate is based on victimization data, which is likely a more complete estimate than the New Mexico rate based on law enforcement data. The comparatively high rate in New Mexico would likely be even more substantial if we had state-level victimization data. To put the personal toll of the New Mexico incidence rate in perspective, during 2013, law enforcement agencies in New Mexico identified more than 17,000 victims of intimate partner violence, with 43% of the related incidents involving injury to the victim (Caponera, 2014). Notably too, in 36% of these incidents, at least one child was present at the scene, further highlighting the broad toll of these incidents on families and communities in the state (Caponera, 2014). While there are a number of dynamics that contribute to the significant intimate partner violence problem in New Mexico, poverty and a related lack of intervention and prevention services are likely central, particularly in isolated, rural areas across the state. The importance of providing effective intervention for victims of intimate partner violence is difficult to overstate. The consequences of their victimization take both an immediate and long-term toll on these individuals, as well as their families and communities.

Prior research on the effectiveness of interventions for reducing intimate partner violence recidivism suggests that formal social control efforts can decrease the likelihood of subsequent instances of intimate partner violence. However, these studies have largely examined forms of intervention separately (for an exception, see Mears, Carlson, Holden, & Harris, 2001). These single intervention evaluations often compare various intervention modalities available from a single formal institution (e.g., mandatory vs. discretionary arrests) or assess variation in intervention success across populations. We add to this body of literature by comparing outcomes across three distinct formal intervention options. Using both law enforcement arrest data and civil court protection order data, we examine the relative effectiveness of three intervention scenarios for deterring subsequent offending among a sample of intimate partner violence offenders from Bernalillo County, New Mexico: arrest, civil protection order, and arrest in combination with a civil protection order. Our aim is twofold. First, we describe the characteristics of individuals who utilize (victims) or are subject to (offenders) each type of formal intervention. Second, we compare reoffending outcomes across these three intervention scenarios to assess the relative efficacy of each. In so doing, we control for the characteristics of those involved in each system to help account for selection issues that might shape utilization patterns.

Literature Review

Arrest and Domestic Violence (DV) Recidivism

The most common formal help-seeking response to intimate partner violence is to call the police (Tjaden & Thoennes, 2000). Evidence regarding the effectiveness of police intervention, however, is largely inconclusive. This may be because police response can take a number of forms. Sherman and Berk’s (1984) now classic Minneapolis experiment randomly assigned offenders in misdemeanor DV incidents reported to the police to one of three possible interventions: arrest, 8-hr separation of parties, or advice/mediation at the scene. They found arrest to be significantly more effective in deterring subsequent offending compared with more informal remedies such as “advice” from the officer or temporary separation. A number of researchers have replicated the Sherman and Berk protocol (Berk & Newton, 1985; Dunford, Huizinga, & Elliott, 1990; Hirschel, Hutchinson, & Dean, 1992; Pate & Hamilton, 1992), but with conflicting results. Berk and Newton (1985) found arrest to be a deterrent to future violence in Southern California, but the remaining studies found no significant difference in the effect of arrest on recidivism when compared with more informal police interventions. Consistent with this, Berk, Campbell, Klap, and Western (1992) used data from field experiments in four cities and find that the effects of arrest on subsequent DV offending do not vary from those of alternative police interventions. Maxwell, Garner, and Fagan (2001) pooled and re-analyzed the data from five studies replicating the Sherman and Berk methodology. They determined that arrest “may” reduce subsequent intimate partner violence, but so too might other police actions. Specifically, they found that although arrest does not increase subsequent violence, a “majority of suspects discontinued their aggressive behaviors even without an arrest” (Maxwell et al., 2001, p. 13).

More recently, Iyengar (2009) investigated the effects of mandatory arrest laws on rates of intimate partner homicide across states. She reports a 60% increase in such homicides among states that implemented mandatory arrest policies. Her findings do not necessarily implicate arrests in this increase, but rather suggest that the threat of mandatory arrest may reduce the overall likelihood that individuals call on the police to intervene in DV incidents. Moreover, it is not clear that the finding of increased homicide in response to mandatory arrest laws holds at lower levels of aggregation or for non-lethal intimate partner violence. Zeoli, Norris, and Brenner (2011) found mandatory arrest laws have no significant effect on intimate partner homicide in 46 U.S. cities. Similarly, Xie, Lauritsen, and Heimer (2012) used geocoded self-report data from the National Crime Victimization Survey for 40 metropolitan statistical areas and found no effect of such laws on non-lethal intimate partner violence.

In New Mexico, arrest is discretionary, so our focus on arrestees aims to evaluate the impact of arrest experiences on recidivism for individuals who the police, using their discretionary judgment, deem culpable. However, given the findings of Xie et al. (2012) and Zeoli et al. (2011), whether arrest is mandatory or discretionary may be substantively immaterial to subsequent outcomes. Our research cannot assess this question, but the point suggests that our findings may be generalizable to cities with or without mandatory arrest laws.

Other research focuses on the individual characteristics that might mediate the effects of arrest. In the 1984 reporting of the Minneapolis experiment results, Sherman and Berk suggested that deterrence was not just due to arrest, but may be the result of arrest within certain contexts or for certain types of offenders. They argue, for example, that the deterrent effect of arrest is stronger for individuals with social controls that generate a “stake in conformity” (Sherman, Smith, Schmidt, & Rogan, 1992). In a replication of the Sherman and Berk experiment in Dade County, Florida, Pate and Hamilton (1992) found that arrest decreased the likelihood of recidivism for employed offenders but increased the likelihood of recidivism for unemployed offenders. Other individual- or incident-level characteristics might reduce the deterrent effect of arrest. For example, among arrestees, Kingsnorth (2006) found that recidivism is more likely when the offender has a prior criminal history, is named as a respondent in a protection order at the time of the incident, and uses a weapon in commission of the abuse.

Furthermore, while it may deter recidivism in the short term, a growing body of research suggests that the deterrent effects of arrest decay over time. Studies tracking offenders for 6 months beyond the initial incident find that between 10-18% of offenders subjected to arrest re-offend (Dunford et al., 1990; Hirschel et al., 1992; Pate & Hamilton, 1992; Sherman & Berk, 1984). Studies examining recidivism for longer periods, 18-28 months, estimate that arrestees re-offend at a notably higher rate, between 15-30% (Berk & Newton, 1985; Kingsnorth, 2006; Tolman & Weisz, 1995). Klein and Tobin (2008) looked well beyond initial arrest and suggest that any short-term effects that arrest (and other law enforcement interventions) might have do not hold in the long term. Using longitudinal data, they found that 60% of DV offenders recidivate within the 10 years following law enforcement intervention. Still, the majority of incidents never come to the attention of the police. Rather, some of those who seek formal intervention may engage with the civil system and bypass law enforcement altogether. Studies comparing the range of police responses overlook the experiences and outcomes of those who seek formal intervention through the court protection order system.

Protection Orders and DV Recidivism

Victims who seek protection orders do so with the hope that the civil court and criminal justice systems can offer them a measure of protection. Prior research provides some evidence that protection orders can reduce the risk of subsequent victimization. Carlson et al. (1999) compared the number of physical abuse incidents reported to police 2 years before and after the issuance of the protection order. Among the 210 women in their sample (all of whom filed for a protection order against their intimate partner), they found fewer incidents reported to the police for the 2 years following the protection order when compared with the 2 years preceding the order. McFarlane, Willson, Lemmey, and Malecha (2000) reported similar outcomes. They collected interview data from 149 women who applied and qualified for a protection order (though not all were granted the order) at the special family violence unit of the Houston, Texas, district attorney’s office. Women who received a temporary or extended order all scored lower on a scale measuring the severity of subsequent and/or continued violence 18 months after applying for a protection order when compared with scores at intake. They conclude that simply invoking the system by applying for a protection order may serve to decrease subsequent abuse. Even so, Spitzberg’s (2002) review of 32 protection order studies focused on stalking behavior found that across studies, an average of 40% of those who secure a protection order experience a violation of that order. Logan and Walker (2009) found three in five women with a protection order experienced a violation. Furthermore, they found key risk factors for violations include respondent stalking behavior and petitioner intent to continue the relationship.

While studies suggest that protection orders might reduce odds of subsequent violence, the literature also cautions that a number of factors work to limit the potential effectiveness of these orders. Most notably, victims who secure a temporary restraining order (generally enforceable for 2 weeks) must return to court to request the order be extended. In one study, only 60% of women securing temporary orders returned for extended orders (Harrell & Smith, 1996). Another study found that 28% of women dropped protection order cases within 3 months of the initial petition, and another 18% did not receive a protection order due to problems such as the respondent never being served notice or because the court dismissed the case (Malecha et al., 2003). Indeed, in our data, we find that only 55% of women who received temporary restraining orders also secured extended orders (n = 363).

As is the case with police responses, civil court responses vary, in large part as a function of the problems noted above. Studies have examined the effect of different court responses on subsequent victimization. There is some evidence to suggest that, while extended orders of protection significantly decrease the risk of future victimization, temporary orders significantly increase this risk (Holt, Kernic, Lumley, Wolf, & Rivara, 2002). However, others have found that engaging the protection order process may be enough to deter offenders. Specifically, studies by McFarlane et al. (2000) and Carlson et al. (1999) find the likelihood of subsequent violence to be similar among victims with extended orders and those who begin the process but do not follow through. Zoellner et al.’s (2000) interviews with women who began the protection order process but did not complete it suggest that women were more likely to follow through with the process if their partner had threatened to kill them. Women were less likely to fully pursue a protection order if they had an emotional attachment to the abuser or if there were threats made to the victim’s child(ren). Clearly, and as we see with police interventions, it is not just the system itself but the characteristics of those invoking the system that shape responses and outcomes.

Assessing the Relative Effectiveness of Interventions

Studies of police intervention compare outcomes (subsequent violence) across various police responses. Studies of protection orders make a similar comparison across individuals who experience varying court responses (no order granted, temporary order, or extended order). These studies also assess how well these systems work for different populations. As a whole, this body of work helps us understand how these systems might best intervene with those at risk for ongoing intimate partner violence. Few studies have examined the effectiveness of arrest and protection orders relative to one another. This is important as social service providers and others in a position to recommend intervention options (and indeed even push for them) have a limited evidence base against which to make recommendations regarding one system or another.

Among the existing literature, some have tracked cases through the criminal justice system to examine the potentially compounded deterrent effects of increasingly serious system responses. Studies by Davis, Smith, and Nickles (1998) and by Tolman and Weisz (1995) treat police contact, arrest, and prosecution as three stages of intervention. Tolman and Weisz (1995) found arrest (and prior police contact with or without arrest) to be a significant deterrent to future involvement in intimate partner offending; however, neither study finds a significant relationship between successful prosecution for misdemeanor intimate partner assault and intimate partner violence recidivism. Wooldredge (2007) analyzed the effects of prosecution, jail time, and prison sentences on future violence among felony domestic assault offenders and found that while prosecution and jail time decrease the likelihood of subsequent intimate partner violence, the length of prison terms is unrelated to subsequent intimate partner violence.

To date, few studies have examined formal interventions distinguishing between cases in the criminal justice system and those in the civil court system. Holt, Kernic, Wolf, and Rivara (2003) compared the likelihood of subsequent abuse among women who reported partner violence to the police with those who filed for a protection order. Using interview data collected from women at 5 and 9 months post-incident, they found that victims with protection orders reported less contact, fewer threats, and less violence and injury from their abuser compared with those who engaged the police. Mears et al. (2001) used police and court data on intimate partner violence incidents to assess the effects of arrest only, protection order only, or both on prevalence and time to DV recidivism as measured by physical violence reported to police within 2 years. They found no difference in recidivism outcomes between the types of intervention. Rather, their findings suggest that individual-level characteristics are the strongest predictor of subsequent abuse. In particular, the likelihood of re-victimization is greatest for minority women from low-income areas, regardless of intervention type. This is not entirely surprising given evidence (summarized above) showing variation in intervention access and outcomes across demographic groups for both police and court intervention. Finally, Kothari et al. (2012) compared outcomes for women with a police reported intimate partner violence victimization coupled with a subsequent protection order with those with no subsequent protection order. They found that those who seek protection orders have more extensive victimization histories than those who did not seek protection orders, again suggesting variation across the populations with distinct intervention modalities.

However, they found that re-victimization levels among those women who sought a protection order (and had more extensive victimization histories) matched those of women with less extensive prior victimization who did not seek a protection order. Overall, our understanding of how various formal intervention options compare is limited.

Summary

The literature assessing the influence of arrest or court intervention on subsequent intimate partner violence is limited and results are mixed. The majority of victims never engage police or courts. Without case/control studies, it is difficult to discern whether interaction with formal institutions in response to intimate partner violence increases victim safety and reduces offender recidivism compared with outcomes for those who do not seek institutional intervention. However, given that a substantial number of victims do seek institutional intervention, it is important to assess how well those interventions serve this population and to understand the factors that increase or limit their effectiveness. The research on arrest suggests that under some circumstances, arrest may be the most effective police response to intimate partner violence but that under other circumstances, it may be just as effective for police to engage other response options. Similarly, protection orders seem to provide some measure of safety for those who seek them, but their effectiveness may depend on the length or type of order. Moreover, findings suggest that individual-level characteristics of offenders and victims are also relevant for assessing risk of subsequent violence, though it is not entirely clear whether these factors operate independent of intervention type. In the end, it appears that the degree to which formal system intervention reduces recidivism is dependent on the type of intervention and the characteristics of those seeking and those subject to intervention. Less clear from the literature is how various formal interventions compare in terms of both the populations they serve and their effectiveness at ensuring the long-term safety of those populations. It is this question that we address here by comparing the populations who seek law enforcement and/or court intervention for intimate partner violence and assessing the relative effectiveness of each intervention.

Current Study

The current study compares law enforcement and court interventions for intimate partner violence. Specifically, we compare the demographic and offending characteristics of offenders and victims served by each system or by both systems as well as the comparative likelihood of intimate partner violence recidivism among offenders following intervention. The sample of cases for the analysis includes those in which an offender was arrested for an intimate partner violence incident (n = 1,049), appeared as a respondent in a petition for an order of protection (n = 523), and those in which an offender was both arrested and named as a respondent in protection order proceedings for the same incident (n = 137), in Bernalillo County, New Mexico, in 2002.

Data Sources

We pulled our sample from two separate data sources. We used law enforcement data to identify all 2002 intimate partner violence incidents for which police made an arrest. The Albuquerque Police Department (APD) maintains automated data for both the City Police and the County Sheriff’s Office, covering all crime incidents reported to law enforcement in Bernalillo County, New Mexico. The data contain information about both the people involved in each reported incident and the incident itself. Focusing on intimate partner violence incidents, our data file includes the sex, race, age of both arrestees and victims, the incident specific crime code and statute, weapon code, incident location, and time of incident. Using the 2002 data, we then appended data on any additional incidents for which the arrestee was a suspect or was arrested in the 6 years prior to (1996-2002) and 4 years following (2002-2006) the sample selection incident. This allows us to control for prior offending and to assess subsequent offending outcomes.¹

We also obtained civil court data from the 2nd Judicial District in New Mexico, which serves Bernalillo County. From these records, we retrieved automated data on DV orders of protection (DVOP) issued in 2002. The DVOP court files include information on individual- and case-level characteristics for all petitions, temporary orders, and extended orders initiated in 2002, including prior case activity from 1997-2002, and subsequent case activity through 2006. The court records are divided into two files: one provides information on case events and the other provides information on each party involved in the case. Event files include a line for each case event with the court case number, date of event, an automated code for description of each event (i.e., openings, filings, hearings, closings), and a comment field describing the event. The party file includes a line for each party (i.e., petitioners and respondents) involved in a case. These data include party name, date of birth, sex, and the associated case number. We merged event and party data to produce a chronological event sequence dataset with both petitioner and respondent information.

Using both case and person identifiers (name, date of birth, social security number), we merged the law enforcement and court data. Unfortunately, neither data source contains relationship information, so we could not easily exclude non-intimate partner DV incidents (e.g., those involving siblings, roommates, parent/child). While not a perfect proxy for intimate partner status, we restricted our sample to only those cases with a single adult arrestee/respondent and a single adult victim/petitioner.² Each line of data represents one intimate partner violence offender and includes the associated incident information and demographic characteristics of both the offender and victim. In addition to demographic information and case information for the 2002 event, the dataset also includes prior criminal offending history, prior DVOP involvement, subsequent criminal offending, and any subsequent DVOP involvement for the offender/respondent.

Measures

Our recidivism measure aims to assess the relative effectiveness of each intervention type in promoting victim safety by reducing subsequent partner and family violence on the part of the offender. For this measure, we use subsequent police contact for any domestic incident. The measure of subsequent DV is coded as (1) if the offender has been cited as a suspect or arrestee in a domestic incident recorded by law enforcement sometime during the 4 years following the 2002 sample selection event. The victim of subsequent incidents may be different than the victim in the 2002 case. Indeed, we focus broadly on domestic incidents here and do not try to distinguish between subsequent intimate partner violence and other forms of DV. This is because we are interested in the influence of formal intervention, not just on victim safety but also on offender involvement in intimate partner and family violence more generally. In addition, our measure of subsequent DV includes incidents for which the offender is an arrestee or suspect in police records as we are more interested in prior and subsequent law enforcement contact for partner and family violence than in the specific nature of that law enforcement outcome. We constructed alternative measures for subsequent intimate partner violence, including whether the offender is a respondent in a DVOP petition. However, these measures yielded similar results in the final analysis, and including the DVOP group adds so few cases that it muddles interpretation rather than adding to the findings in any meaningful way. As such, we have chosen to present models using the known to law enforcement measure of recidivism. Alternative models are available on request.

Intervention type is a set of dichotomous variables that identify the sample group for each case. The study sample can be divided into three groups. The arrest sample consists of arrestees (referred to here as offenders) who were arrested by law enforcement, but for whom we find no DVOP filing within 60 days of the incident (n = 1,049).³ The protection order sample includes those cases for which there is a record of a DVOP filing and for which no incident involving the same respondent was recorded by law enforcement in the 60 days prior to the DVOP filing (n = 523). The dual intervention sample includes cases in which the offender was arrested and also named in a DVOP filing within 60 days of the incident (n = 137). We view each sample as representative of a distinct formal response to intimate partner violence.

Prior law enforcement contact is defined as any offense recorded by law enforcement for which the offender is listed as a suspect or arrestee before the 2002 sample selection event. The law enforcement data allow us to track prior offenses back to 1996. We examine prior law enforcement contact for a range of offense types: intimate partner violence, DVOP violations, drug, weapon, driving under the influence (DUI), family (non-violent), Part I violent index crimes (homicide, rape, robbery, and aggravated assault), Part II violent index crimes (other forcible sex and simple assault), and property crimes. For analytic purposes, we created a set of variables to capture the nature of prior law enforcement contact, which includes four mutually exclusive categories: no prior history, prior intimate partner violence only, prior non-intimate partner violence only,⁴ and both intimate partner and non-intimate partner priors. Each offender is assigned to one of these four categories.

Both sources of data provide some basic individual demographic information for offenders and victims. Only the police data include race/ethnicity indicators, so we focus here on age and sex, which are available for the full sample. Age at the date of case initiation is based on each person’s date of birth. Sex is based on the data from the automated law enforcement or DVOP court files. In the automated court files, the field for sex of the petitioner and respondent is not always populated. In these instances we relied on either the automated law enforcement data or, in the cases that involved DVOP only, individual names and coded individuals as male or female based on sex-normative given names when possible.⁵

Using these data we address two broad questions: (a) Is there variation in the characteristics of those who access different formal interventions and (b) are there differences between formal intervention scenarios, relative to one another, in terms of their deterrent effects on subsequent intimate partner violence? We first examine similarities and differences between intervention populations. Then we use logistic regression to estimate the odds ratios for these independent variables on the likelihood of intimate partner violence recidivism, where (1) represents a subsequent intimate partner violence charge and (0) no subsequent charges.

Results

We begin by examining sample descriptive statistics for offenders and victims in the full sample and then compare these same variables across intervention type subsamples (Table 1). As we would expect from official data, males dominate the offender category, making up 79% of offenders, while females are overrepresented in the victim category (77% of victims). The average age of both offenders and victims is 33 years. Seventy percent of offenders in the study sample have at least one prior offense; 3% have only prior intimate partner violence, 40% have only non-intimate partner priors, and 27% have both intimate partner and non-intimate partner prior offense histories. This suggests, as others have shown (Piquero, Brame, Fagan, & Moffitt, 2006), that among those arrested for intimate partner violence, family violence is often part of a broader offending trajectory.

Table 1.

Descriptive Statistics.

Variables	Total sample	Arrest sample	Protection order sample	Dual intervention sample
	n = 1,709	n = 1,049	n = 523	n = 137
	%	%	%	%
Dependent variable
Subsequent DV	23	25	21	22
Offender demographic characteristics
Sex^a,b
Male	79	77	80	88
Female	21	23	20	12
Age^a,c
18-24	27	31	21	20
25-44	61	58	63	68
45+	12	11	16	12
M (in years)^a,c	32.6	31.4	34.7	33.7
SD	10.0	9.6	10.5	9.7
Offender criminal history
No priors^a,c	30	37	19	19
DV only	3	3	3	3
Non-DV only^a,c	40	34	51	45
Both DV and non-DV	27	26	28	34
Victim demographic characteristics
Sex*
Male^a,c	23	26	18	12
Female	77	74	82	88
Age
18-24	29	31	25	26
25-44	57	55	60	58
45+	15	14	16	15
M (in years)^c	33.4	32.8	34.5	33.0
SD	11.4	11.4	11.6	10.4

Note. DV = domestic violence.

Law Enforcement and Dual Intervention samples significantly different, p < .05.

Court and Dual Intervention sample significantly different, p < .05.

Law Enforcement and Court samples significantly different, p < .05.

p < .05.

Comparing across intervention type, some differences emerge. Offenders subject to arrest only are younger on average when compared with respondents in the protection order and dual intervention samples. A larger percentage of offenders in this group fall into the young adult (18-24) age category when compared with the age distribution for the other two samples. Offenders in this sample are also significantly less likely to have a record of prior offenses when compared with other two samples.

Offenders in the protection order sample are older on average when compared with those in the arrest and dual intervention samples, and they have the highest representation of offenders in the older adult age category (45 years and older). Offenders in the protection order sample are also distinguishable from others by their prior offense histories. They are more likely than their arrest sample counterparts to have an arrest history, and it is the non-DV history that distinguishes them from the arrest sample. This is also the case for those in the dual intervention sample, whose general prior record, marked by both DV and non-DV incidents, also distinguishes them from those in the arrest sample. This dual intervention sample is distinct from the DVOP sample, however, in that it has a significantly higher percentage of male offenders when compared with the other two samples.

With respect to our outcome of interest, 23% of offenders in our sample have a subsequent intimate partner violence incident reported to law enforcement authorities (n = 393) within 4 years of the 2002 sample selection event. Among those with a subsequent intimate partner violence charge, offenders average 1.52 DV offenses during the period under analysis. As shown in Table 1, the percentage of offenders with a subsequent DV charge varies slightly, though not significantly, by intervention type. Twenty-five percent of offenders in the arrest sample have at least one subsequent DV offense within the 4-year follow-up period, compared with 22% of dual intervention sample offenders and 21% of offenders in the protection order sample. Not surprisingly, the majority of offenders also engage in non-DV offending during the 4-year follow-up. Fifty-nine percent of the sample has a subsequent non-DV arrest. There are no significant differences in subsequent arrests across subsamples.

While there are no differences across intervention type in the extent of subsequent DV activity, descriptive data do show variation across intervention type by individual characteristics. It is possible, then, that controlling for individual characteristics, intervention types may differentially affect subsequent DV. Table 2 provides the results of the logistic regression of intervention type and individual offender and victim characteristics on the likelihood of DV recidivism. Similar to the findings of Mears et al. (2001), these results suggest that relative to one another, intervention types are similarly correlated with the likelihood of subsequent intimate partner violence. When prior offending variables are included in Model 2, it appears that offenders in the arrest sample are significantly more likely than those in the protection order sample to have a subsequent intimate partner violence offense. However, once we include offender characteristics in the model, this effect becomes non-significant.⁶ This suggests that while intervention type is correlated with recidivism, this influence is largely driven by the characteristics of the individuals making up the intervention population. Age is particularly influential here with older offenders significantly less likely to have subsequent contact with the police for a domestic incident than younger offenders.

Table 2.

Logistic Regression of Intervention and Individual Characteristics on Subsequent DV.

Variables in equation	Model 1	Model 2	Model 3
Variables in equation	B (OR)	B (OR)	B (OR)
Sample
Arrest	0.175 (1.191)	0.292 (1.339)*	0.241 (1.272)
Dual intervention	0.028 (1.029)	−0.013 (0.987)	−0.068 (0.934)
Offender prior history
Prior DV Only	—	0.803 (2.232)*	0.862 (2.369)**
Prior DV and Other	—	1.302 (3.677)*	1.216 (3.374)*
Prior Non-DV Only	—	0.648 (1.913)*	0.574 (1.775)*
Offender demographics
Log of offender age	—	—	−2.096 (0.123)*
Offender is female	—	—	0.285 (1.330)
Victim demographics
Log of victim age	—	—	0.092 (1.097)
Victim is female	—	—	0.455 (1.576)*
Model
Constant	−1.300	−2.067	0.221
N	1,709	1,709	1,709
−2 Log likelihood	1,855.415	1,785.563	1,752.962
Df	2	5	9

Note. DV = domestic violence; OR = odds ratio.

p < .05.

We find statistically significant relationships in Model 3 between offender characteristics, victim demographics, and subsequent DV offenses. Offenders with prior offending histories are more likely than offenders without priors to engage in subsequent intimate partner violence activity. When offenders have both intimate partner violence and non-intimate partner violence priors, they are 3.4 times more likely than those with no priors to have a subsequent DV offense. Offenders having only DV priors are 2.4 times more likely than those with no priors to have a subsequent DV; and offenders with a non-intimate partner offense history are 1.8 times more likely to have a subsequent DV when compared with the no prior group. Younger offenders are more likely to engage in subsequent DV offending than older individuals. Interestingly, males are not significantly more likely than females to have a subsequent DV offense, but offenders who offend against a female victim are 1.5 times more likely than those with male victims to recidivate.

Combining protection order and arrest data restricts the range of incident and demographic variables we can consider as the court data have fewer incident and person-level details. A supplemental analysis using only the law enforcement and dual intervention samples (Table 3) suggests that incident characteristics may also have some influence on the likelihood of subsequent DV offending. Specifically, in incidents where a weapon was used, the likelihood of reoffending is lower. It may be that these are the offenders most likely to have spent time off the street given the seriousness of incidents involving a weapon. What we are seeing, then, may be the result of an incapacitation effect. However, our data do not allow us to test for this possibility. The addition of race/ethnicity variables suggests that Hispanic offenders are at higher odds of subsequent law enforcement contact for DV than Whites. More than race/ethnicity specifically, this finding likely implicates processes related to disadvantage and to policing of minorities and immigrant populations that we cannot test for with these data. Immigration aside, these processes are likely also relevant for New Mexico’s Black and Native American populations. That we see no significant differences for these populations compared with Whites may reflect more reluctance on the part of these groups to call the police in response to DV. It may also be an artifact of the data as these groups, while overrepresented in the sample, are still relatively small compared with Whites and Hispanics (who comprise 51% and 26%, respectively, of the total sample for which we have race/ethnicity data). Because Blacks comprise 6% of the sample and Native Americans 16%, we may simply lack the statistical power to detect any differences between these groups and the larger White sample.

Table 3.

Logistic Regression of Intervention, Incident, and Individual Characteristics on Subsequent DV.

Variables in equation	Model 1	Model 2	Model 3	Model 4
Variables in equation	B (OR)	B (OR)	B (OR)	B (OR)
Sample
Arrest	0.146 (0.864)	0.160 (1.174)	0.321 (1.379)	0.319 (1.376)
Incident characteristics
Use of a weapon	—	− 0.591 (0.554)	− 0.712 (0.491)*	− 0.739 (0.477)*
Juvenile victim	—	− 0.013 (0.987)	−0.173 (0.841)	− 0.171 (0.843)
Aggravated assault	—	0.448 (1.565)	0.474 (1.607)	0.542 (1.719)
Offender prior history
Prior DV Only	—	—	1.041 (2.833)*	1.052 (2.863)*
Prior DV and Other	—	—	1.270 (3.561)*	1.120 (3.066)*
Prior Non-DV Only	—	—	0.673 (1.959)*	0.631 (1.879)*
Offender demographics
Log of offender age	—	—	—	− 0.908 (0.403)*
Offender is female	—	—	—	−0.134 (0.874)
Offender is Hispanic	—	—	—	0.349 (1.418)*
Offender is Native American	—	—	—	0.279 (1.322)
Offender is Black	—	—	—	0.537 (1.711)
Victim demographics
Log of victim age	—	—	—	0.316 (1.371)
Victim is female	—	—	—	0.565 (1.760)*
Victim is Hispanic	—	—	—	0.059 (1.061)
Victim is Native American	—	—	—	0.151 (1.163)
Victim is Black	—	—	—	−0.432 (0.649)
Model
N	1,186	1,185	1,185	1,117
Constant	−1.142	−1.141	−1.141	−1.138
−2 Log likelihood	1,312.121	1,308.790	1,257.515	1,159.127
Df	1	4	7	17

Note. DV = domestic violence; OR = odds ratio.

p < .05.

Discussion

Our findings suggest that, overall, DV offenders respond to criminal and civil interventions similarly. The odds of reoffending do not vary by intervention type. However, we find considerable differences in the types of offenders subject to arrest, protection orders, or both intervention types. We conclude here by situating these findings in the existing literature on DV intervention and offering some relevant recommendations for research and practice.

Our study is similar to that of Mears et al. (2001), who examined re-victimization among victims following offender arrest, protection order, or both and found the odds of re-victimization to be similar across these groups. Examining the same three interventions, but focusing on offender outcomes, our analyses indicate that offenders subject to arrest, protection orders, or both are all similarly likely to have subsequent police contact. In the 4 years following formal intervention from the police or courts, 23% of offenders in our sample come into contact with police for another domestic incident. This is the same percentage reported by Mears et al. in their examination of re-victimization rates. Of course, given low reporting rates, this does not necessarily mean that the remaining offenders desisted, but it does suggest that the different types of system contact have similar effects on the likelihood of subsequent law enforcement involvement for DV. We find no evidence to suggest that one intervention is more effective than another for deterring future DV incidents. Mears et al. focused on the effect each intervention type has on the odds of re-victimization (as opposed to reoffending) and draw similar conclusions. However, they find that the odds of re-victimization do vary by individual and contextual factors. Although we cannot examine contextual factors as our court data are not geocoded, we do assess the influence of individual- and incident-level factors on reoffending and find that outcomes vary across some key indicators. Young male offenders, those with prior offending histories, and offenders with female victims are more likely to have subsequent police contacts for domestic incidents. This is notable because these are some of the same characteristics across which the populations served by the different interventions also vary.

These findings suggest that both court and law enforcement agencies should be mindful of the populations they serve and the particular risks associated with case, individual, and prior criminal history characteristics. Most notable here is the fact that the different modes of intervention we evaluate serve populations that vary along these dimensions of risk. Specifically, while the offenders in the protection order sample are older than those in the arrest sample, the civil court interacts with offenders who have more extensive criminal histories than those in the law enforcement sample, a characteristic that puts them at risk for reoffending. These individuals likely have more entrenched and varied criminal histories that are difficult to redirect. This population may be best served by interventions that target their offending patterns more broadly. Although we do not have information on alcohol or substance use, other work with these protection order data suggests that substance use is often implicated in the violence reported in protection order petitions (Denman, Albright, Broidy, & Kleymann, 2009). Indeed, Moracco et al. (2010) reported high rates of alcohol and drug use histories, along with mental health issues and prior intimate partner violence among a population-based sample of male DVOP respondents.

The population subject to arrest only is a young, male population with a less extensive criminal history than offenders subject to DVOP. For this group, police contact may represent a particularly important point of intervention that can redirect DV trajectories before they become entrenched. Coupled with informal social controls, the arrest may serve as a wake up call that motivates this group to change their behaviors. The dual intervention group represents the most high-risk population as the offenders in this sample are comparatively young, most likely to have female victims, and have histories of both domestic and non-DV arrests. For this group, courts and police need to be particularly mindful of the potential for reoffending. Although we see no significant difference in reoffending across samples, the personal and incident characteristics of this group suggest it to be at high-risk.

Although we note that system representatives and stakeholders should be mindful of the types of populations seeking particular interventions, how to use this information to improve outcomes is less clear. Good policy is contingent on the availability of reliable, evidence-based interventions targeting both offenders and victims. Unfortunately, the evidence base for intimate partner violence intervention (with both offenders and with victims) is thin. Day, Chung, O’Leary, and Carson (2009) suggested that batterer intervention programs have low efficacy because they often lack a well-articulated intervention model and have limited system-level support to ensure program compliance among program participants. Stover, Meadows, and Kaufman (2009) drew similar conclusions with respect to interventions aimed at perpetrators, victims, and couples, suggesting that there is limited evidence for the efficacy of current intervention modalities for these different groups. Although their review suggests substance abuse treatment is likely an important component of any intervention program with these populations, there is evidence of other program elements that are likely important. Stover, Berkman, Desai, and Marans (2010) found that home visit follow-ups by police advocates with victims who have made police contact serves to increase their system utilization more broadly and to increase their satisfaction with the system as well as their overall safety. Coulter and VandeWeerd (2009) noted high rates of attrition and recidivism among those enrolled in most batterer intervention programs but report notable success based on their evaluation of a multi-tiered, multi-systemic intervention. We echo the call of others to promote and fund the development and evaluation of interventions targeting the victims and perpetrators of intimate partner violence to build an evidence-based set of treatment and intervention protocols to improve victim safety and victim resources and to redirect perpetrator trajectories. With this information, the police and the courts could better target their responses to specific populations of offenders and victims they come into contact with and link them to the most appropriate supplemental services.

Limitations and Conclusions

Our use of administrative data to examine the comparative effectiveness of formal interventions for intimate partner violence has a number of strengths, particularly the substantial sample of offenders whose contact with the criminal justice system we can track over time. At the same time, it also introduces some problems and speaks to the importance of replication with different sampling strategies and methodologies. Although we can track offenders over time, this is not a prospective study and offenders enter the sample at different points in their offending trajectories. For some, the 2002 sample selection event is their first law enforcement contact while others have extensive criminal histories prior to this event. In some sense, this means we are comparing apples and oranges, and it would be beneficial to prospectively track a sample of offenders following their first formal contact to assess factors that shape how their subsequent offending trajectories unfold. That said, the current design allows us to compare outcomes across a range of intimate partner violence offenders at various stages in their trajectories. Our focus on offenders and offender outcomes also means we cannot track victims. Even among those offenders who evidence subsequent law enforcement contact for DV, the 2002 intervention may have protected the victim of that sample selection event. However, our research and that of Mears et al. (2001), which focuses on victim outcomes, lead to similar conclusions and, as such, provide an important addition to our understanding of the way in which these formal interventions influence subsequent offending and victimization. In addition, though our findings suggest individual and incident characteristics affect subsequent behavior, our data offer only a limited range of individual and incident-level variables. More detailed data on individuals, incidents, and, as Mears et al. suggested, the social structural context would help refine our understanding of who is best served by each system and who is most in need of focused treatment and intervention resources following formal system contact. Better race/ethnicity data across all systems would also help tease out the way in which various systems respond to minority populations. This is crucial because such populations are oftentimes isolated from services as a function of geography, poverty, or processes related to social isolation/exclusion.

For those who are subject to formal intervention for intimate partner violence, system contact can be an important and formative event in the context of an intimate partner violence trajectory. Each system needs to be mindful of how to best serve those they come into contact with. If we view formal system contact as the start of a chain of intervention events, we need to develop effective follow-up links on which the formal system can build. Future research should focus on the development and evaluation of effective treatment and intervention for offenders and victims involved in intimate partner violence. This is important because the formal system is not equipped to provide the kind of treatment and intervention that should follow formal system contact but it is well positioned to act as a key funneling agent linking offenders and victims to appropriate intervention. As Xie et al. (2012) suggested, collaboration between criminal, civil, and social services may be particularly effective. They find the lowest DV victimization rates for women in metropolitan statistical areas with the most sworn officers and social service workers per capita. Increased efforts to document effective, evidence-based practices for officers and social workers to draw from would likely further strengthen this relationship. It would also give those in places with fewer law enforcement and social service personnel more guidance as to how to intervene despite limited resources.

Footnotes

Acknowledgements

The authors thank the helpful staff at both the Second Judicial District Courthouse and the Albuquerque Police Department for facilitating data access. They also thank Tim Wadsworth for his contributions to the original project for which they collected the data they use in this article. Finally, thanks to Alana Cook for her editorial contributions.

Declaration of Conflicting Interests

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

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This project uses data collected for a project funded by the Bureau of Justice Statistics, Grant Number 2006-BJ-CX-K016.

Notes

Author Biographies

Lisa Broidy is an associate professor in the School of Criminology and Criminal Justice at Griffith University and a member of the Griffith Criminology Institute. Broadly, her research focuses on the etiology of crime with a particular focus on the influence of gender, life course transitions, institutional contact, and strains (including victimization and trauma). In addition, her work examines justice system policies related to domestic violence and offender reentry.

Danielle Albright is the coordinator of the New Mexico Intimate Partner Violence Death Review Team at the Center for Injury Prevention, Research, and Education, in the Department of Emergency Medicine, University of New Mexico. Her research interests include collective behavior, political activism, and public policy, with an emphasis on activities related to the prevention and intervention of interpersonal violence.

Kristine Denman is the director of the New Mexico Statistical Analysis Center located at the University of New Mexico where she engages in applied research and evaluation. Her areas of interest include prisoner reentry, case processing, and domestic violence.

References

Berk

R. A.

Campbell

Klap

Western

(1992). The deterrent effect of arrest in incidents of domestic violence: A Bayesian analysis of four field experiments. American Sociological Review, 57, 698-708.

Berk

R. A.

Newton

P. J.

(1985). Does arrest really deter wife battery? An effort to replicate the findings of the Minneapolis spouse abuse experiment. American Sociological Review, 50, 253-262.

Caponera

(2014). Incidence and nature of domestic violence in New Mexico XIII: An analysis of 2013 data from the New Mexico interpersonal violence data central repository. Retrieved from http://www.nmcadv.org/domestic-violence-in-nm-reports-by-betty-caponera/

Carlson

M. J.

Harris

S. D.

Holden

G. W.

(1999). Protective orders and domestic violence: Risk factors for re-abuse. Journal of Family Violence, 14, 205-226.

Catalano

(2012). Intimate partner violence, 1993-2010. Washington, DC: U.S. Department of Justice.

Coulter

VandeWeerd

(2009). Reducing domestic violence and other criminal recidivism: Effectiveness of a multilevel batterers intervention program. Violence and Victims, 24, 139-152.

Davis

R. C.

Smith

B. E.

Nickles

L. B.

(1998). The deterrent effect of prosecuting domestic violence misdemeanors. Crime & Delinquency, 44, 434-442.

Day

Chung

O’Leary

Carson

(2009). Programs for men who perpetrate domestic violence: An examination of the issues underlying the effectiveness of intervention programs. Journal of Family Violence, 24, 203-212.

Denman

Albright

Broidy

Kleymann

(2009). Petitioning for a domestic violence order of protection: An examination of abuse descriptions, outcomes and multiple petition filings (Final Report to Bureau of Justice Statistics). Albuquerque: New Mexico Statistical Analysis Center.

10.

Dunford

F. W.

Huizinga

Elliott

D. S.

(1990). The role of arrest in domestic assault: The Omaha police experiment. Criminology, 28, 183-206.

11.

Gondolf

E. W.

McWilliams

Hart

Stuehling

(1994). Court response to petitions for civil protection orders. Journal of Interpersonal Violence, 3, 503-517.

12.

Harrell

Smith

B. E.

(1996). Effects of restraining orders on domestic violence victims. In Buzawa

Buzawa

(Eds.), Do arrests and restraining orders work? (pp. 214-242). Thousand Oaks, CA: Sage.

13.

Hirschel

J. D.

Hutchinson

I. W.

Dean

C. W.

(1992). The failure of arrest to deter spouse abuse. Journal of Research in Crime & Delinquency, 29, 7-33.

14.

Holt

V. L.

Kernic

M. A.

Lumley

Wolf

M. E.

Rivara

F. P.

(2002). Civil protection orders and risk of subsequent police-reported violence. Journal of the American Medical Association, 288, 589-594.

15.

Holt

V. L.

Kernic

M. A.

Wolf

M. E.

Rivara

F. P.

(2003). Do protection orders affect the likelihood of future partner violence and injury? American Journal of Preventative Medicine, 24, 16-21.

16.

Iyengar

(2009). Does the certainty of arrest reduce domestic violence? Evidence from mandatory and recommended arrest laws. Journal of Public Economics, 93, 85-98.

17.

Jordan

C. E.

(2004). Intimate partner violence and the justice system: An examination of the interface. Journal of Interpersonal Violence, 19, 1412-1434.

18.

Kingsnorth

(2006). Intimate partner violence: Predictors of recidivism in a sample of arrestees. Violence Against Women, 12, 917-935.

19.

Klein

A. R.

Tobin

(2008). A longitudinal study of arrested batterers, 1995-2005: Career criminals. Violence Against Women, 14, 136-157.

20.

Kothari

C. L.

Rhodes

K. V.

Wiley

J. A.

Fink

Overholt

Dichter

M. E.

. (2012). Protection orders protect against assault and injury: A longitudinal study of police-involved women victims of intimate partner violence. Journal of Interpersonal Violence, 27, 2845-2868.

21.

Logan

T. K.

Walker

(2009). Civil protection order outcomes: Violations and perceptions of effectiveness. Journal of Interpersonal Violence, 24, 675-692.

22.

Malecha

McFarlane

Gist

Watson

Batten

Hall

Smith

(2003). Applying for and dropping a protection order: A study with 150 women. Criminal Justice Policy Review, 14, 486-504.

23.

Maxwell

C. D.

Garner

J. H.

Fagan

J. A.

(2001). The effects of arrest on intimate partner violence: New evidence from the spouse assault replication program. Washington, DC: U.S. Department of Justice.

24.

McFarlane

Willson

Lemmey

Malecha

(2000). Women filing assault charges on an intimate partner: Criminal justice outcome and future violence experienced. Violence Against Women, 6, 396-408.

25.

Mears

D. P.

Carlson

M. J.

Holden

G. W.

Harris

S. D.

(2001). Reducing domestic violence revictimization: The effects of individual and contextual factors and type of legal intervention. Journal of Interpersonal Violence, 16, 1260-1283.

26.

Moracco

K. E.

Anderson

Buchanan

R. M.

Espersen

Bowling

J. M.

Duffy

(2010). Who are the defendants in domestic violence protection order cases? Violence Against Women, 16, 1201-1223.

27.

Pate

A. M.

Hamilton

E. E.

(1992). Formal and informal deterrents to domestic violence: The Dade County spouse assault experiment. American Sociological Review, 57, 691-697.

28.

Piquero

A. R.

Brame

Fagan

J. A.

Moffitt

T. E.

(2006). Assessing the offending activity of criminal domestic violence suspects: Offense specialization, escalation, and de-escalation evidence from the Spouse Assault Replication Program. Public Health Reports, 121, 409-418.

29.

Sabina

Tindale

R. S.

(2008). Abuse characteristics and coping resources as predictors of problem-focused coping strategies among battered women. Violence Against Women, 14, 437-456.

30.

Sherman

L. W.

Berk

R. A.

(1984). The specific deterrent effects of arrest for domestic assault. American Sociological Review, 49, 261-272.

31.

Sherman

L. W.

Smith

D. A.

Schmidt

J. D.

Rogan

D. P.

(1992). Crime, punishment, and stake in conformity: Legal and informal control of domestic violence. American Sociological Review, 57, 680-690.

32.

Spitzberg

B. H.

(2002). The tactical topography of stalking victimization and management. Trauma, Violence, & Abuse, 3, 261-288.

33.

Stover

C. S.

Berkman

Desai

Marans

(2010). The efficacy of a police-advocacy intervention for victims of domestic violence: 12 month follow-up data. Violence Against Women, 16, 410-425.

34.

Stover

C. S.

Meadows

A. L.

Kaufman

(2009). Interventions for intimate partner violence: Review implications for evidence-based practice. Professional Psychology: Research and Practice, 40, 223-233.

35.

Tjaden

Thoennes

(2000). Extent, nature, and consequences of intimate partner violence: Findings from the National Violence against Women Survey. Washington, DC: U.S. Department of Justice.

36.

Tolman

R. M.

Weisz

(1995). Coordinated community intervention for domestic violence: The effects of arrest and prosecution on recidivism of woman abuse perpetrators. Crime & Delinquency, 41, 481-495.

37.

Wooldredge

(2007). Convicting and incarcerating felony offenders of intimate assault and the odds of new assault charges. Journal of Criminal Justice, 35, 379-389.

38.

Xie

Lauritsen

J. L.

Heimer

(2012). Intimate partner violence in U.S. metropolitan areas: The influence of police and social services. Criminology, 50, 961-992.

39.

Zeoli

A. M.

Norris

Brenner

(2011). Mandatory, preferred, or discretionary: How the classification of domestic violence warrantless arrest laws impacts their estimated effects on intimate partner homicide. Evaluation Review, 35, 129-152.

40.

Zoellner

L. A.

Feeney

N. C.

Alvarez

Waltington

O’Neill

M. L.

Zager

Foa

E. B.

(2000). Factors associated with completion of the restraining order process in female victims of partner violence. Journal of Interpersonal Violence, 15, 1081-1099.