A Longitudinal Examination of Offending and Specialization Among a Sample of Massachusetts Domestic Violence Offenders

Abstract

A paucity of existing research focuses on longitudinal examinations of criminal trajectories among reoffenses committed by domestic violence offenders. Specifically, few studies have longitudinally assessed whether domestic violence offenders specialize, recidivating in domestic violence assault, or generalize, committing a range of personal and property crimes. Acknowledging these research deficiencies, the current study uses longitudinal data from a cohort of 317 batterers who were processed in a domestic violence court to investigate the trajectories of domestic violence arrests and nondomestic violence arrests over a 10-year period. The degree of overlap between domestic and nondomestic violence arrest trajectory groups is examined through a cross-tabulation and chi-square analysis. Logistic and multinomial regression models are applied to identify risk factors that distinguish trajectory groups. A PROC TRAJ procedure identifies two trajectory groups for domestic violence arrests (low and high rate) and three trajectory groups for nondomestic violence arrests (very low, low, and high rate). Results indicate that specialization among domestic violence offenders is rare—prior alcohol and drug crimes predict membership in the high-rate domestic violence arrest trajectory group and prior domestic violence arrests predict membership in both the low-rate and high-rate nondomestic violence arrest trajectories. Implications for future research and policy are discussed in this article.

Keywords

domestic violence criminal careers specialization trajectories

Partner violence is a common type of assault in the United States with 11% of the population reporting striking a partner in their lifetime (Klevens, Simon, & Chen, in press). Partner violence involves the use of coercive behaviors with the intent of controlling one’s intimate partner, and may comprise of psychological, physical, and sexual abuse, as well as stalking (Gover, 2011). Research estimates that approximately 20% of women and 11% of men are physically assaulted by a current or former partner during their lifetime (Breiding, Black, & Ryan, 2008). These estimates translate to 1,200 deaths in the United States each year as well as 2 million injuries among women and almost 600,000 injuries among men (Centers for Disease Control and Prevention, 2003).

There are significant negative health outcomes for victims of partner violence. Psychological and behavioral consequences include depression, anxiety, PTSD, eating disorders, disruptive sleep patterns, gastrointestinal disorders, self-harming behaviors, unsafe sexual behaviors, and substance abuse (Barnett, 2000; Coker et al., 2002; Ellsberg, Jansen, Heise, Watts, & García-Moreno, 2008; Leserman & Drossman, 2007). Research suggests that many domestic violence victims report living in chronic fear (Belknap & Sullivan, 2002; Fischer & Rose, 1995), which can lead to severe anxiety and avoidance behaviors (Herman, 1992). Additional physical health problems that may also emerge include migraines, stroke, ill health, and chronic disease (Brewer, Roy, & Smith, 2010; Coker et al., 2002). These problematic health outcomes also have associated economic costs. For example, the Centers for Disease Control and Prevention (2003) estimates that intimate partner violence costs the US $5.8 billion each year because of the costs of medical care and lost labor.

Despite the prevalence and severity of domestic violence, the criminal justice system’s response to this social problem was dormant until about 25 years ago (Olson & Stalans, 2001). The 1980s brought about punitive sanctions for perpetrators of domestic violence, such as mandatory and/or proarrest policies. The implementation of criminal and civil protection orders provided additional safeguards for victims of chronic violence. The Violence Against Women Act (VAWA) granted further funds for communities and police departments to coordinate a response to intimate partner violence. Because of a combination of rising arrest rates and reluctance by the criminal justice system to sentence domestic violence offenders to jail time, courts increasingly relied on court-ordered offender treatment. Offender treatment programs commonly employ cognitive behavioral techniques and a feminist perspective, confronting offender denial and accountability, victim blaming, and sexist attitudes toward women and relationships (Gondolf, 2001). The use of treatment as a sentencing option necessitates specialized knowledge about the unique aspects of domestic violence perpetrators in contrast to other violent offenders to implement effective treatment and prevent subsequent violence (Olson & Stalans, 2001).

Research on batterer intervention programs documents a minimal deterrent effect on recidivism (Babcock, Green, & Robie, 2004; Feder & Wilson, 2005; Peterson, 2008). This lack of effectiveness may be because of the variability among such programs (Gover, 2011). Although many states have developed standards and guidelines for such programs, many view state standards as insufficiently research-based (Austin & Dankwort, 1998). Further, many states rely on a “one-size-fits-all” strategy, although recent research indicates that domestic violence offenders are fairly heterogeneous (Gover, 2011). In addition, batterer intervention programs have remained homogenous because of the limited research on the criminal careers of domestic violence perpetrators (Piquero, Brame, Fagan, & Moffitt, 2006).

To examine whether long-standing assumptions about the specialization and escalation of violence among domestic violence offenders are accurate, it is necessary to look at longitudinal arrest trajectories among domestic violence perpetrators. To date, little such research exists. Furthermore, common outcome evaluations of treatment programs that focus on counts or binary indicators of domestic violence offending overlook the importance of deviations in the scope and intensity of violence over time (Jones, Heckert, Gondolf, Zhang, & Ip, 2010). Furthermore, the existing, minimal research suggests that domestic violence offenders do not specialize (Piquero et al., 2006), but in fact commit a range of violent and nonviolent crimes against multiple victims.

The criminal justice system’s reliance on offender treatment programs may reflect bias such that perpetrators of domestic violence are sanctioned differently than perpetrators of stranger violence (Dobash & Dobash, 1979), or a response to a societal preference to protect the family as a societal institution (Harvard Law Review, 2003). Regardless, offender treatment programs have become an institutionalized response to domestic violence, and the criminal justice system is obligated to deliver this treatment effectively. Consequently, recognizing the diversity among offenders and performing a time-sensitive assessment of offending will inform intervention programs. Specifically, the present research aims to investigate whether partner violence perpetrators are generalists rather than specialists in that perpetrator commit a range of crimes rather than specializing in partner violence only.

Literature Review

The Specialization Debate

For several decades, criminologists have debated whether criminal offenders typically commit a range of crimes, rather than perpetrating one type of crime exclusively. Gottfredson and Hirschi (1990) and Sampson and Laub (1993) both predict in their general theories of crime that offenders infrequently specialize (Piquero, Brame, Mazerolle, & Haapanen, 2002). Alternatively, criminologists who adhere to a developmental perspective anticipate a combination of specialized and generalized criminal behaviors (DeLisi et al., 2011; Piquero, 2000; Piquero et al., 2002). Specifically, Moffitt (1993) describes two unique categories of offenders, one group who offends throughout life (life-course-persistent offenders) and another group who offends only during adolescence (adolescence-limited offenders). Although life-course-persistent offenders are believed to commit a range of crimes, adolescence-limited offenders are described as primarily committing nonviolent crimes (Piquero et al., 2002). Similarly, Loeber and Stouthamer-Loeber’s (1998) developmental perspective distinguishes offenders into multiple groups and proposes unique offending trajectories among each group over time (Piquero et al., 2002).

Many perceive the generalists’ perspective as the predominant theoretical approach (Hindelang, 1981). For instance, Klein (1984) conducted a meta-analysis composed of 33 studies on criminal behavior and found only four studies that supported the notion of specialization. In contrast, other research has indicated specialization may occur with age (Piquero, Farrington, & Blumstein, 2003). In regards to violent crime in particular, Blumstein, Cohen, Das, and Moitra (1988) observed that property crimes were conducted with more consistent specialization than violent crimes. Likewise, Piquero et al. (2006) assessed a sample of domestic violence offenders and found that one-third had no criminal history, one-third had a nonviolent criminal history, and one-third had histories of violent crimes—thus, only a portion of offenders sampled exclusively specialized in violent crime. In addition, Piquero et al. (2006) observed variation in offending, finding some offenders displayed a stable, high-level of aggression, some offenders escalated in their violence, some offenders de-escalated in their violence, and some offenders sustained low levels of aggression.

Hilton et al. (2004) demonstrated that among their sample of domestic violence offenders, there was an average of 0.39 prior domestic violence offenses. Five percent had previously violated a no-contact order and inflicted an average of 1.19 injuries to prior domestic violence victims. Twenty-four percent of the sample was previously given correctional sentences, 4% reported a history of violence against nonpartners, and 15% had violated a previous conditional release; overall, the sample averaged 3.3 previous criminal charges. Regarding prior violence against nonpartners, there was an average of 0.09 assaults and 1.09 injuries to victims. Furthermore, four percent of domestic violence victims indicated that the offender was also violent to others (Hilton et al., 2004). This sample was emblematic of offenders with generalized behavior—many had abused substances, perpetrated previous partner violence, and previously committed other types of crime—all offenders had injured someone in a violent offense (Hilton et al., 2004).

Results from another study of domestic violence offenders using experimental methods found that between 30% and 52% had previously been arrested for a felony, 94% to 100% had been arrested for a misdemeanor, 22% to 27% had been convicted of a felony, 45% to 57% had been convicted of a misdemeanor, 1% to 2% had been arrested as juveniles, and 10% to 14% had been previously incarcerated (Feder & Dugan, 2002). In addition, the majority of offenders had not been previously arrested for domestic violence (84-86%) (Feder & Dugan, 2002). In addition among a sample of individuals on probation for domestic violence offenses and offenders on probation for other violent crimes, Olson and Stalans (2001) observed no significant differences in previous violent crime convictions and prior illegal drug use. Overall, 52% had previous adult convictions, 64% had previous convictions for violent crime, 66% reported a substance abuse history, and 31% indicated prior drug abuse (Olson & Stalans, 2001). This finding supports the notion of generalization among domestic violence offenders and emphasizes the importance of considering varying criminal histories within domestic violence offender treatment standards. In addition, it is important to recognize that domestic violence is only one manifestation of a variety of criminal conduct (Piquero et al., 2006). Furthermore, the diversity among offending propensities supports the validity of differentiated treatment for domestic violence offenders (Gover, 2011).

The Importance of Examining Offending Among Domestic Violence Offenders Longitudinally

One of the major contributions to the domestic violence literature is the development of the cycle of violence, which emphasizes the variability in batterer behavior over time (Walker, 1984). The recognition of the importance of monitoring domestic violence offenders over time is significant, with consideration for both research and policy. The limited longitudinal data on batterers (Jones et al., 2010) contributes to a gap in the literature and, in turn, weakens domestic violence offender treatment boards’ abilities to create research-based standards that consider time as a variable. In some research, domestic violence offenders have been found to be different when compared to other offenders. For example, Olson and Stalans (2001) found that 18% of domestic violence offenders revictimized the same individual during probation, whereas only 5% of offenders of other violent crimes recidivated. History can also influence time to reoffending, as research demonstrates that offenders with prior domestic violence offenses recidivate earlier than first-time domestic violence offenders (Frantzen, San Miguel, & Kwak, 2011).

Research examining recidivism among perpetrators who have completed offender treatment programs reveals that between 23% and 40% of men reoffend (Dutton, Bodnarchuk, Kropp, Hart, & Ogloff, 1997; Shepard, 1992) with rates varying dependent on sample size, length of study, and operationalization of the outcome variable. For example, one recent meta-analysis on the effectiveness of court-mandated batterer intervention treatment programs found that among experimental studies examining official reports, the mean effect size was 0.26 and the mean effect size among experimental studies relying on victim reports of reoffending was 0.01 (Feder & Wilson, 2005). Researchers have also noted that high rates of attrition in batterer treatment programs (as high as 84%; Murphy, Musser, & Maton, 1998) may contribute to difficulties in designing effective offender treatment programs (Babcock et al., 2004).

Extant research observes that among perpetrators of domestic violence that recidivate, generally reoffenses occur fairly quickly after the first arrest. One study found that 22% of offenders were rearrested for committing domestic violence within 2 years after their first offense, with an average of 635 days (or 21 months) to recidivism (Frantzen et al., 2011). Conviction status may also influence time to recidivism. Frantzen et al. (2011) found that during the first year of their study, 18% of convicted domestic violence offenders were rearrested, versus 10% of offenders who had been arrested but not convicted. At the 2-year observation point, 24% of convicted offenders had recidivated versus 16% of offenders who had not previously faced conviction (Frantzen et al., 2011). Hilton et al. (2004) documented that 30% of their sample had reoffended over 51 months, with an average of 15.1 months to recidivism. The overwhelming majority of reoffenses (95%) involved the same victim as the prior offense (Hilton et al., 2004). In a two-year study of domestic violence victims from four cities, the bulk of reoffenses occurred during the year following the first assault (32%), with 42% of victims reporting assaults within 2 years (Gondolf, 2001). Piquero et al. (2006) also reported domestic violence recidivism rates between 20% and 50% at 6 months.

Shepard, Falk, and Elliott (2002) grouped the men in their domestic violence offender sample within the following categories: (1) low-risk offenders with no domestic violence history, (2) moderate-risk offenders who perpetrate intermittent violence, (3) serious-risk batterers who have a history of domestic violence, and (4) high-risk offenders which are dangerous to victims and the community. Their results indicated that the bulk of reoffenses were committed within the first six months, with 50% of category four offenders, 45% of category three offenders, 31% of category two offenders, and 18% of category one offenders’ recidivating. At the final 18-month observation point, 50% of category four, 64% of category three, 45% of category two, and 36% of category one offenders were investigated, charged, or convicted of at least one additional domestic violence crime (Shepard et al., 2002). Other research interviewing both offenders and victims found that batterers’ violent behavior did not change over time (Feder & Dugan, 2002). Although the bulk of research on time to recidivism among domestic violence offenders indicates that most offenses occur fairly quickly after the first offense, research following offenders beyond 2 years is limited, restricting our understanding of time to recidivism to a 24 month-span.

Current Study

Few studies have examined the criminal careers of domestic violence offenders longitudinally with an emphasis on specialization versus generalization (Piquero et al., 2006). Furthermore, domestic violence offending is often measured in the form of a binary outcome, rather than considering the frequency of domestic violence offending over time (Jones et al., 2010). Recognizing these limitations, the present study addresses these gaps in the literature by longitudinally examining a cohort of batterers processed in a domestic violence court using a novel methodological technique, the trajectory methodology. The cohort’s domestic violence arrests and nondomestic violence arrests are tracked over a 10-year period (1995-2004), with the trajectory analysis distinguishing groups within the cohort based on arrest rates. In addition, the degree of overlap between domestic violence arrest and nondomestic violence arrest trajectory groups is examined, as well as the risk factors that distinguish trajectories groups.

Method

Sample

These data are from the “Longitudinal Study of a Cohort of Batterers Arraigned in a Massachusetts District Court, 1995-2004” (Wilson & Klein, 2006). The original study included 342 men arrested for domestic violence within the jurisdiction of an eastern Massachusetts’ district court between April 1994 and March 1996. Using criminal history records, this study followed the criminal behavior of participants from their arrest until December 2004. Because of the longitudinal nature of the current study, the sample used for the current analyses includes only the 1995 arrestees (n = 317) as these offenders represent the entire cohort of arrestees who were arraigned for domestic violence in the same calendar year (1995) and have an equal length of follow-up time at risk (10 years).

Measures

Dependent Variables

This study included two dependent variables: the number of arrests for nondomestic violence offenses (including both nonviolent crimes and violent crimes that were not perpetrated against a partner such as assault with a deadly weapon, operating a vehicle under the influence of alcohol and/or drugs, motor vehicle theft, larceny, crimes against person, and breaking and entering) and the number of arrests for domestic violence offenses. The number of arrests per year for nondomestic violence crime was summed for each participant from 1995 to 2004. Also, the number of arrests per year for domestic violence were identified and summed for each participant from 1995 to 2004.

Independent Variables

This study included the following four independent variables: victim-offender relationship, age of the offender at the time of arrest, being a previous domestic violence offender, and prior alcohol and drug offenses. The victim–offender relationship is a dichotomous variable distinguishing between married intimate partners (1) and nonmarried intimate partners (0). Age at the time of arrest was measured as a continuous variable and prior domestic violence and prior drug and alcohol offenses were each dichotomous measures, where 1 = yes and 0 = no.

Analytic Strategy

Analyses proceeded in three stages. First, the trajectory solutions are identified using the PROC TRAJ procedure, which is available as a macro in SAS (Jones, Nagin, & Roeder, 2001). Second, a cross-tabulation and chi-square analysis are presented to demonstrate the degree of overlap between the nondomestic violence arrest trajectories and the domestic violence arrest trajectories. Finally, a logistic regression model is estimated to determine which risk factors distinguished membership in the domestic violence arrest trajectory groups, and a multinomial regression model was estimated to identify which risk factors distinguished the nondomestic violence arrest trajectory groups.

Results

The present research employs a novel method for investigating longitudinal data, the group-based trajectory model introduced by Nagin and Land (1993). Trajectory analysis shares some similarities with other more traditional group-based modeling approaches such as hierarchical linear modeling (HLM) and latent curve analysis (LCA), but also possesses some marked differences. Departing from HLM and LCA’s underlying assumption of a continuous distribution, trajectory models assume there is heterogeneity within the population that manifests as different clusters of individuals who exhibit distinct longitudinal patterns of behaviors or trajectories (Nagin, 2005).

Trajectory models are based on a maximum likelihood function and demonstrate the qualities of maximum likelihood parameter estimates (e.g., consistent and asymptotically normally distributed). The best fitting model is identified by maximizing the Bayesian Information Criteria (BIC), with lower values indicating a better fit, and analyzing the posterior probabilities for group assignment. The BIC favors more parsimonious models and exacts a penalty for introducing additional parameters (additional groups) to the model. Posterior probabilities for group assignment represent the probability that an individual can be assigned to a particular group-based trajectory with a certain degree of precision. The posterior probabilities range from 0 to 1, with mean group-based probabilities greater than .70, indicating good model precision (Nagin, 2005).

In regards to the domestic violence trajectories, a two-group solution was the best fitting model (Figure 1). The mean posterior probabilities were both well above the .70 cutoff provided by Nagin (2005); specifically, the mean posterior probability for the first trajectory group was .92 (median = .99) and .93 (median = .98) for the second trajectory group. The first trajectory group (G1) is composed of 78.5% of the sample and represents a low rate domestic violence arrest trajectory group with almost no domestic violence arrests from years 3 to 10. Group 2 (G2) can be considered a high-rate domestic violence arrest trajectory and is composed of 21.5% of the sample. Group 2 begins with 1.2 domestic violence arrests in year one and then declines to an average of 0.5 domestic violence arrests by year 2 with little change through year 10.

Figure 1.

Domestic violence arrest trajectories across 10 years (1995-2004)

Turning toward the nondomestic violence arrest trajectories, a three-group solution provided the best fitting model (Figure 2). Similar to the trajectory solution described above, the mean posterior probabilities were all well above the .70 cutoff provided by Nagin (2005); specifically, the mean posterior probability for the first trajectory group was .84 (median = .84), .82 (median = .83) for the second trajectory group, and .90 (median = .97) for the third trajectory group. The first trajectory group (G1) is the largest group encompassing 42.6% of the sample and represents a very low-rate group for nondomestic violence arrests across the 10 years. Group 2 (G2), encompassing 40.4% of the sample, begins at approximately 0.2 nondomestic violence arrests at year 1, peaks at 0.45 nondomestic violence arrests at year 2, and then decreases and holds steady at 0.15 nondomestic violence arrests per year from year 5 to year 10. Comparatively, Group 3 (G3) is the most dynamic and smallest of the nondomestic violence arrest groups and represents approximately 17% of the sample. This group averages approximately 0.5 nondomestic violence arrests at year 1, increases to 0.9 nondomestic violence arrests at year 2, and then decreases from years 3 to 4. At year 5, Group 3 again increases to 0.9 nondomestic violence arrests before decreasing steadily to 0.6 nondomestic violence arrests per year at year 8. Finally, Group 3 increases in year 9 to .75 nondomestic violence arrests before decreasing again to approximately 0.5 nondomestic violence arrests at year 10.

Figure 2.

Nondomestic violence arrest trajectories across 10 years (1995-2004)

The next stage of analysis involves a cross tabulation that assesses the degree of overlap among the domestic violence arrest trajectories and the nondomestic violence arrest trajectories. Table 1 indicates that there is significant association between domestic violence arrest trajectory group membership and nondomestic violence arrest trajectory group membership (χ² = 20.63, p < .001). The majority of individuals (87%) assigned to the low-rate domestic violence arrest trajectory group (G1) were also assigned to either a very low-rate (G1) or low-rate (G2) nondomestic violence arrest trajectory group. Comparatively, 41.2% of those assigned to the high-rate nondomestic violence arrest trajectory group (G3) were assigned to the low-rate domestic violence arrest trajectory group (G1) and greater than one-third of the individuals in the high-rate domestic violence arrest trajectory group were also assigned to the high-rate nondomestic violence arrest trajectory group (G2).

Table 1.

Cross-Tabulation of Nondomestic Violence Arrest and Domestic Violence Arrest Trajectories Across 10 years (1995-2004)

	Very Low Rate Non-DV Offenders (G1)	Low Rate Non-DV Offenders (G2)	High Rate Non-DV Offenders (G3)	Total
Low rate DV offenders (G1)	118 (47.4%)	100 (40.2%)	31 (12.4%)	249 (78.5%)
High rate DV offenders (G2)	17 (25.0%)	28 (41.2%)	23 (33.8%)	68 (21.5%)
Total	135 (42.6%)	128 (40.4%)	54 (17.0%)	317 (100.0%)

Note: Number of participants and column percentages in parentheses for each cell indicated χ² = 20.63, p < .001; Φ = .255 (p < .001).

The results from the logistic regression model distinguishing the high-rate domestic violence arrest trajectory from the low-rate domestic violence arrest trajectory is presented in Table 2. The odds of a domestic violence offender being assigned to a high-rate domestic violence arrest trajectory nearly doubled if they had previously committed a domestic violence offense prior to their current domestic violence offense (OR = 1.825; CI = 1.005–3.312; p < .05). Similarly, the odds of a domestic violence offender being assigned to high-rate domestic violence arrest trajectory were two times greater if they had previously been arrested for alcohol/drug crimes (OR = 1.941; CI = 1.053–3.576; p < .05). In contrast, age was inversely related to being assigned to a high-rate domestic violence arrest trajectory (OR = 0.946; CI = 0.915–0.979; p < .001). Although the victim–offender relationship was in the expected direction with domestic violence offenders who were married to their victims having decreased odds of being assigned to a high-rate arrest trajectory, the relationship was not statistically significant.

Table 2.

Distinguishing Domestic Violence Arrest Trajectories

Variables	High Rate DV Offenders OR (CI)
Victim–offender relationship (married/not married)	0.738 (0.365–1.489)
Prior domestic violence offender (yes/no)	1.825* (1.005–3.312)
Prior alcohol/drug crimes (yes/no)	1.941* (1.053–3.576)
Age (at time of arrest; continuous)	0.946*** (0.915–0.979)
Model fit statistics
–Chi 2 Log likelihood = 306.510
χ² = 23.089***
Nagelkerke R² = 0.109

Note: OR = odds ratio; CI = 95% confidence interval.

p < .05. **p < .01. ***p < .001.

Table 3 illustrates the results of the multinomial regression model distinguishing the low- and the high-rate nondomestic violence arrest trajectories from the very low-rate nondomestic violence arrest trajectory. Virtually identical results were found in this particular regression model as those that were produced in the previous regression model distinguishing the domestic violence arrest trajectories. Specifically, having committed a prior domestic violence offense before the current domestic violence offense and having been previously arrested for alcohol/drug crimes significantly increased the odds of being in the low- and high-rate trajectory group relative to the very low rate nondomestic violence arrest trajectory group. In fact, the odds of being assigned to a high-rate nondomestic violence arrest trajectory was nearly twice as large for those domestic violence offenders who had committed a domestic violence offense prior to their current domestic violence offense (OR = 3.447; CI = 1.593–7.458; p < .01) compared to the odds of being assigned to a low-rate trajectory (OR = 1.867; CI = 1.004–3.469; p < .001) relative to a very low-rate trajectory. Similarly, age decreased the odds of being assigned to either the low-rate (OR = 0.945; CI = 0.918–0.973; p < .001) or the high-rate trajectory group (OR = 0.896; CI = 0.856–0.937; p < .001). Once again, domestic violence offenders who were married to their victims had decreased odds of being assigned to either a low- or high-rate nondomestic violence arrest trajectory rather than a very low-rate arrest trajectory, albeit nonsignificant.

Table 3.

Distinguishing Nondomestic Violence Arrest Trajectories

Variables	Low Rate Non-DV Offenders OR (CI)	High Rate Non-DV Offenders OR (CI)
Victim–offender relationship (married/not married)	0.940 (0.521–1.695)	0.680 (0.282–1.644)
Prior domestic violence offender (yes/no)	1.867* (1.004–3.469)	3.447** (1.593–7.458)
Prior alcohol/drug crimes (yes/no)	3.565*** (2.074–6.129)	4.797*** (2.201–10.453)
Age (at time of arrest; continuous)	0.945*** (0.918–0.973)	0.896*** (0.856–0.937)
Model fit statistics
−2 Log likelihood = 576.621
χ² = 75.783***
Nagelkerke R² = 0.244

Note: OR = odds ratio; CI = 95% confidence interval.

p < .05. **p < .01. ***p < .001.

Discussion

Existing research on domestic violence has largely operated under the assumption that domestic violence offending takes place in a vacuum such that batterers specialize in domestic violence. As a result, unique theories of domestic violence, batterer intervention treatment programs, and dedicated domestic violence courts have been created to respond to this unique offender—the domestic violence offender. At the same time, limited existing research suggests that domestic offenders rarely specialize in domestic violence or even violence (Piquero et al., 2006). As such, the present research investigated whether trajectories of domestic violence arrests and nondomestic violence arrests can be identified among a cohort of batterers who were processed in a domestic violence court. In addition, this study examined if (and how) domestic violence arrest and nondomestic violence arrest trajectories overlap and whether evidence suggests that risk factors predict membership in domestic violence arrest and nondomestic violence arrest trajectory groups more similarly of differently.

The trajectory analyses revealed two distinct trajectory groups for domestic violence arrests and three distinct trajectory groups for nondomestic violence arrests. The two-group trajectory model suggested one trajectory of low-rate offenders (G1) and a second trajectory that exhibited a high-rate of domestic violence arrests over time. Comparatively, the nondomestic violence arrests demonstrated three distinct groups, with a very low-rate trajectory (G1), a low-rate trajectory (G2), and a high-rate trajectory (G3). Furthermore, the results examining the overlap between domestic violence and nondomestic violence arrest trajectories indicated that group assignment for one outcome was significantly associated with group assignment for the other. For example, the majority of individuals assigned to the low-rate domestic violence arrest trajectory group (G1) were also assigned to either a very low-rate (G1) or low-rate (G2) nondomestic violence arrest trajectory group. Comparatively, assignment to the high-rate domestic violence arrest trajectory group was associated with assignment the high-rate nondomestic violence arrest trajectory.

The multivariate results suggested that prior domestic violence offenses, prior alcohol and drug crimes, and age significantly distinguished trajectory group membership for both domestic violence and nondomestic violence arrests. Having a prior domestic violence and/or alcohol/drug offense significantly increased an individual’s likelihood of being assigned to a high-rate domestic violence arrest trajectory and significantly increased their likelihood of being assigned to a low- or high-rate nondomestic violence arrest trajectory. In addition, older individuals were significantly less likely to be assigned to a high-rate domestic violence arrest or a low- or high-rate nondomestic violence arrest trajectory.

One important covariate, victim–offender relationship, was not found to significantly distinguish trajectory membership for domestic violence or nondomestic violence arrests. Specifically, although being married to the victim did decrease the odds of membership in both the high-rate domestic violence arrest trajectory and the low- and high-rate nondomestic violence arrest trajectories the association was not statistically significant. This finding is inconsistent with prior literature suggesting that marriage is a salient predictor of treatment success (Gover, Jennings, Davis, Tomsich, & Tewksbury, 2011) and that batterers who are married have fewer domestic violence arrests compared with unmarried batterers (Akers & Kaukinen, 2009). Similarly, for nondomestic violence arrests, these findings diverge from extant literature demonstrating a deterrent effect of marriage on offending behavior over time (Sampson, Laub, & Wimer, 2006).

The present study has several important implications for theory and policy but several limitations should also be noted. First, although the sample is unique in that it encompasses a cohort of batterers and includes information on their domestic violence and nondomestic violence criminal careers over 10 years, the extent of the generalizability of these results outside of the state of Massachusetts and/or the northeastern United States is unknown. In addition, although this research provides an important first step in exploring specialization versus versatility among domestic violence offenders, future research should attempt to link the psychosocial variables that may affect offending trajectory membership—an investigation beyond the scope of the present data. Finally, the current study was unable to control for periods of incarceration, which may affect the shape of the offending trajectories especially among the high-rate arrest groups.

The influence of prior drug and alcohol crimes on membership in both the high-rate domestic violence and nondomestic violence trajectory groups may reveal differential treatment needs among offenders with substance abuse disorders that call for intensive substance abuse counseling outside the realm of traditional batterer intervention treatment. Similarly, individuals with criminal histories including violent offenses against both strangers and intimates might be better served by combining treatment services that target violence generally, in addition to the incorporation of traditional programs for batterers that focus on divisions of power and control in intimate relationships. In addition, first-time domestic violence offenders or batterers who do not offend outside of intimate settings may be amenable to batterer intervention programs because of the potential influence of “turning points” such as the loss of their spouse and/or children of perpetrators of domestic violence specifically (Sheehan, Thakor, & Stewart, 2011).

Overall, the present results are consistent with theoretical orientations indicating that specialization among offenders is rare (DeLisi et al., 2011; Gottfredson & Hirschi, 1990; Piquero, 2000; Piquero et al., 2002; Sampson & Laub, 1993) and support the limited research demonstrating that domestic violence offenders commit both violent and nonviolent offenses (Piquero et al., 2006). Such findings foster the need for a reexamination of the current conceptualization of homogeneity among domestic violence offenders and the corresponding “one-size fits all” response to domestic violence. A “one-size fits all” approach usually refers to all offenders receiving the same type of treatment for the same period of time. Findings from the current study identified different groups of domestic violence offenders. Differential treatment would call for high-rate offenders to receive the most intensive treatment. In this same vein, Andrews and Bonta (1994) have reported that providing high levels of treatment to low-risk offenders can have an adverse effect on low risk offenders. In addition, Carey (1997) reported that domestic violence arrests are significantly reduced when offenders receive appropriate levels of treatment. Thus, perhaps providing differential treatment regimens for domestic violence offenders that are versatile (e.g., generalists) in their offending and exhibit a high risk of domestic violence arrest specifically and nondomestic violence arrest more generally may prove to be more effective treatment alternatives.

Footnotes

Declaration of Conflicting Interests

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

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Bios

Tara N. Richards is an Assistant Professor in the Department of Government and Justice Studies at Appalachian State University and a Research Coordinator for ongoing funded research in the Department of Mental Health Law and Policy at the Louis de la Parte Mental Health Institute at the University of South Florida. She earned her PhD in criminology from the University of South Florida in 2011. Her major research interests include violence against women; mental health, substance abuse, and trauma/violence; and evaluation research. Some of her published work has appeared in Violence Against Women, Feminist Criminology, and Crime and Delinquency. She is the recipient of the 201 American Society of Criminology Division of Women and Crime’s graduate scholar award.

Wesley G. Jennings, PhD, is an Assistant Professor in the College of Behavioral and Community Sciences in the Department of Criminology and has a Courtesy Assistant Professor Appointment in the Department of Mental Health Law and Policy at the University of South Florida. He received his doctorate degree in criminology from the University of Florida in 2007. He has published over 75 peer-reviewed articles, and his major research interests include longitudinal data analysis, semi-parametric group-based modeling, sex offending, gender, race/ethnicity, and the victim-offender overlap. He recently completed research on a National Institute of Justice funded project examining sex offender recidivism and collateral consequences, for which he was a Co-Lead Investigator. In addition, he is the current Editor of the American Journal of Criminal Justice, a recent recipient of the 2011 William S. Simon/Anderson Publishing Outstanding Paper Award from the Academy of Criminal Justice Sciences, and a Lifetime Member of the Academy of Criminal Justice Sciences. Finally, in a recent forthcoming publication to be published in the Journal of Criminal Justice Education (Copes et al.) he was recently recognized as the number 1 ranked Assistant Professor in the country who is currently employed in a department with a doctoral program in criminology based on his publication record and citations.

Angela R. Gover is a Professor in the School of Public Affairs at the University of Colorado Denver. She earned her PhD in criminology and criminal justice from the University of Maryland. Her research interests include policy relevant family violence issues, violence against women, gender and crime, and evaluation research. Some of her published work has appeared in Violence and Victims, Victims and Offenders, Deviant Behavior, Journal of Criminal Justice, and Violence Against Women. She is currently engaged in numerous research projects focused on safety and relationship issues, law enforcement and judicial responses to domestic violence, and law enforcement officer attitudes about domestic violence.

Elizabeth A. Tomsich is a PhD student in the School of Public Affairs at the University of Colorado at Denver. She earned her master’s degree in social psychology from Purdue University. Currently, her research interests involve violence against women, partner violence, and reproductive health policy.

References

Akers

Kaukinen

(2009). The police reporting behavior of intimate partner violence victims. Journal of Family Violence, 24, 159-171.

Andrews

D. A.

Bonta

(1994). The psychology of criminal conduct. Cincinnati, OH: Anderson.

Austin

Dankwort

(1998, October 5). A review of standards for batterer intervention programs. National Online Resource Center on Violence Against Women: Applied Research Forum. Retrieved from http://new.vawnet.org/applied-research-papers/print-document.php?doc_id=393&find_type=Intervention%20Services.

Babcock

J. C.

Green

C. E.

Robie

(2004). Does batterers’ treatment work? A meta-analytic review of domestic violence treatment outcome research. Clinical Psychology Review, 23, 1023-1053.

Barnett

O. W.

(2000). Why battered women do not leave, part 1: External inhibiting factors within society. Trauma, Violence, and Abuse, 1, 343-372.

Belknap

Sullivan

C. M.

(2002). Longitudinal study of battered women in the system: The victims’ and decision-makers perceptions (Final report).Rockville, MD: National Institute of Justice/NCJRS Paper Reproduction Sales.

Blumstein

Cohen

Das

Moitra

S. D.

(1988). Specialization and seriousness during adult criminal careers. Journal of Quantitative Criminology, 4, 303-345.

Breiding

M. J.

Black

M. C.

Ryan

G. W.

(2008). Prevalence and risk factors of intimate partner violence in eighteen U.S. states/territories, 2005. American Journal of Preventive Medicine, 34, 112-118.

Brewer

Roy

Smith

(2010). Domestic violence: The psychosocial impact and perceived health problems. Journal of Aggression, Conflict, and Peace Research, 2, 4-15.

10.

Carey

(1997). Cog probation. American Probation and Parole Association Perspectives, 21, 27-42.

11.

Centers for Disease Control and Prevention. (2003). Costs of intimate partner violence against women in the United States. Atlanta, GA: Author.

12.

Coker

A. L.

Davis

K. E.

Arias

Desai

Sanderson

Brandt

H. M.

Smith

P. H.

(2002). Physical and mental health effects of intimate partner violence for men and women. American Journal of Preventive Medicine, 23, 260-268.

13.

DeLisi

Beaver

K. M.

Wright

K. A.

Wright

J. P.

Vaughn

M. G.

Trulson

C. R.

(2011). Criminal specialization revisited: A simultaneous quantile regression approach. American Journal of Criminal Justice, 36, 73-92.

14.

Dobash

R. E.

Dobash

(1979). Violence against wives: A case against the patriarchy. New York, NY: Free Press.

15.

Dutton

D. G.

Bodnarchuk

Kropp

Hart

S. D.

Ogloff

J. R. P.

(1997). Wife assault treatment and criminal recidivism: An 11-year follow-up. International Journal of Offender Therapy and Comparative Criminology, 41, 9-23.

16.

Ellsberg

Jansen

H. A. F. M.

Heise

Watts

C. H.

García-Moreno

(2008). Intimate partner violence and women’s physical and mental health in the WHO multi-country study on women’s health and domestic violence: An observational study. Lancet, 371, 165-172.

17.

Feder

Dugan

(2002). A test of the efficacy of court mandated counseling for domestic violence offenders: The Broward experiment. Justice Quarterly, 19, 343-375.

18.

Feder

Wilson

(2005). A meta-analytic review of court-mandated batterer intervention programs: Can courts affect abusers’ behavior? Journal of Experimental Criminology, 1, 239-262.

19.

Fischer

Rose

(1995). When enough is enough: Battered women’s decision making around court orders of protection. Crime and Delinquency, 41, 414-429.

20.

Frantzen

San Miguel

Kwak

D. H.

(2011). Predicting case conviction and domestic violence recidivism: Measuring the deterrent effects of conviction and protection order violations. Violence and Victims, 26, 395-409.

21.

Gondolf

E. W.

(2001). An extended follow-up of batterers and their partners (Final report). Indiana, PA: Mid-Atlantic Addiction Training Institute.

22.

Gottfredson

Hirschi

(1990). A general theory of crime. Stanford, CA: Stanford University Press.

23.

Gover

A. R.

(2011). New directions for domestic violence offender treatment standards: Colorado’s innovative approach to differentiated treatment. Partner Abuse, 2, 95-120.

24.

Gover

A. R.

Jennings

W. G.

Davis

Tomsich

Tewksbury

(2011). Factors related to the completion of domestic violence offender treatment: The Colorado Experience. Victims & Offenders, 6, 137-156.

25.

Harvard Law Review. (2003). Developments—The law of marriage and family. Harvard Law Review, 116, 1996-2122.

26.

Herman

J. L.

(1992). Complex PTSD: A syndrome in survivors of prolonged and repeated trauma. Journal of Traumatic Stress, 5, 377-391.

27.

Hilton

N. Z.

Harris

G. T.

Rice

M. E.

Lang

Cormier

C. A.

Lines

K. J.

(2004). A brief actuarial assessment for the prediction of wife assault recidivism: The Ontario domestic assault risk assessment. Psychological Assessment, 16, 267-275.

28.

Hindelang

(1981). Variations in sex-race-age-specific incidence rates of offending. American Sociological Review, 46, 461-474.

29.

Jones

A. S.

Heckert

D. A.

Gondolf

E. D.

Zhang

(2010). Complex behavioral patterns and trajectories of domestic violence offenders. Violence and Victims, 25, 3-17.

30.

Jones

B. L.

Nagin

D. S.

Roeder

(2001). A SAS procedure based on mixture models for estimating developmental trajectories. Sociological Methods & Research, 29, 347-393.

31.

Klein

M. W.

(1984). Offense specialization and versatility among juveniles. British Journal of Criminology, 24, 185-194.

32.

Leserman

Drossman

D. A.

(2007). Relationship of abuse history to functional gastrointestinal disorders and symptoms. Trauma, Violence, and Abuse: A Review Journal, 8, 331-343.

33.

Loeber

Stouthamer-Loeber

(1998). Development of juvenile aggression and violence: Some common misconceptions and controversies. American Psychologist, 53, 242-259.

34.

Moffitt

T. E.

(1993). Adolescence-limited and life-course persistent antisocial behavior: A developmental taxonomy. Psychological Review, 1, 674-701.

35.

Murphy

C. M.

Musser

P. H.

Maton

K. I.

(1998). Coordinated community intervention for domestic abusers - intervention system involvement and criminal recidivism. Journal of Family Violence, 13, 263-284.

36.

Nagin

D. S.

(2005). Group-based modeling of development. Cambridge, MA: University of Harvard Press.

37.

Nagin

D. S.

Land

K. C.

(1993). Age, criminal careers, and population heterogeneity: Specification and estimation of a nonparametric, mixed Poisson model. Criminology, 31, 327-362.

38.

Olson

D. E.

Stalans

L. J.

(2001). Violent offenders on probation: Profile, sentence, and differences among domestic violence and other violent probationers. Violence Against Women, 7, 1164-1185.

39.

Peterson

R. R.

(2008). Reducing intimate partner violence: Moving beyond criminal justice interventions. Criminology and Public Policy, 7, 537-545.

40.

Piquero

A. R.

(2000). Frequency, specialization, and violence in offending careers. Journal of Research in Crime and Delinquency, 37, 392-418.

41.

Piquero

A. R.

Brame

Fagan

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, 4, 409-418.

42.

Piquero

A. R.

Brame

Mazerolle

Haapanen

(2002). Crime in emerging adulthood. Criminology, 40, 137-171.

43.

Piquero

A. R.

Farrington

D. P.

Blumstein

(2003). The criminal career paradigm. In Tonry

(Ed.), Crime and justice: A review of research (Vol. 30, pp. 359-506). Chicago, IL: University of Chicago Press.

44.

Sampson

R. J.

Laub

(1993). Crime in the making: Pathways and turning points through life. Cambridge, MA: Harvard University Press.

45.

Sampson

R. J.

Laub

J. H.

Wimer

(2006). Does marriage reduce crime? A counterfactual approach to within-individual causal effects. Criminology, 44, 465-508.

46.

Sheehan

K. A.

Thakor

Stewart

D. E.

(2011). Turning points for perpetrators of intimate partner violence. Trauma, Violence, & Abuse, 13(1), 30-40.

47.

Shepard

(1992). Predicting batterer recidivism five years after community intervention. Journal of Family Violence, 7, 167-176.

48.

Shepard

M. F.

Falk

D. R.

Elliott

B. A.

(2002). Enhancing coordinated community responses to reduce recidivism in cases of domestic violence. Journal of Interpersonal Violence, 17, 551-569.

49.

Tjaden

Thoennes

(2000). Extent, nature, and consequences of intimate partner violence: Findings from the National Violence Against Women Survey. (DOJ Publication No. NCJ 181867). Washington, DC: US Department of Justice.

50.

Walker

L. E.

(1984). Battered woman syndrome. New York, NY: Springer.

51.

Wilson

Klein

(2006). Longitudinal study of a cohort of batterers arraigned in a Massachusetts district court, 1995-2004. ICPSR04543-v1. Cambridge, MA: BOTEC Analysis Corporation, Ann Arbor, MI: Inter-University Consortium for Political and Social Research. 2007-02-23. doi: 10.3886/ICPSR04543