Gender and Intimate Partner Violence

Abstract

Research on Intimate Partner Violence (IPV) has ignited considerable controversy about gender and IPV. Feminist scholars have viewed IPV as a manifestation of male dominance and control, with women primarily the victims and men primarily the perpetrators of this behavior (gender asymmetry). Conversely, family violence researchers have viewed IPV as emerging from conflicts in relationships, with both men and women being involved (gender symmetry). The present study was framed within the context of this controversy using new empirical evidence on one person arrested (single arrest) or both persons arrested (dual arrest) in IPV incidents. Data were acquired on the assessed risk of IPV recidivism immediately after arrest and actual IPV recidivism over an 18-month period postassessment across the state of Connecticut involving heterosexual partners (N = 2,155). Though not definitive, the findings showed persistent gender asymmetry on these behavioral characteristics across arrest categories.

Keywords

domestic violence legal intervention predicting domestic violence women offenders

Research on intimate partner violence (IPV) has sparked controversy, but no issue has been more controversial than the relation between gender and IPV. Central to the controversy has been whether women are as violent as men (e.g., Johnson, 2006, 2008, 2011; Meloy & Miller, 2011, pp.129-144; Straus, 2011; Winstok, 2011). Two groups have offered contending arguments and findings. Feminist scholars have viewed IPV as a manifestation of male dominance and control of women. As such, IPV is asymmetrical in that women are primarily victims and men are primarily perpetrators of this behavior (e.g., DeKeseredy, 2011; Dobash, Dobash, Wilson, & Daly, 1992; Kimmel, 2002; Tjaden & Thoennes, 2000). Conversely, family violence researchers have viewed IPV as emerging from conflicts in intimate relationships, with both men and women being involved. Hence, gender is not a significant factor in IPV, and the pattern of involvement by gender is symmetrical (e.g., Dutton, 2006, 2010; Felson, 2002; Straus, 2005, 2009, 2011).

The present study was framed within the context of this controversy using new empirical evidence drawn from one person arrested (i.e., single arrest) or both persons arrested in an IPV incident (i.e., dual arrest). A focus on dual arrest is important for two reasons. First, it has become increasingly common in the wake of legislation cascading across the United States that either mandates arrest with probable cause or prescribes arrest as the preferred police response to IPV (Hirschel, Buzawa, Pattavina, Faggiani, & Reuland, 2007). These mandatory and proarrest policies have swept both men and women into the net of the criminal justice system, but the increase has been greater for women (Hirschel et al., 2007). Second, a focus on dual arrest allows the analysis of novel data not previously brought to bear on the assumptions underlying these two contending perspectives on gender and IPV.

Background and Significance

States have utilized three main arrest policies over the past thirty years: officer discretion, mandatory arrest, or proarrest (Hirschel, 2008). Mandatory and proarrest policies have large implications for cases legally processed because a companion development during this period has been a broadening definition of IPV. Besides assaults, many states currently include behaviors such as kidnapping, breach of peace, or stalking (Hirschel et al., 2007). Additionally, the range of relationships covered in statutory definitions of family violence has expanded beyond conventional married partners to include cohabiting couples, dating couples, former partners, and individuals with children in common, regardless of whether they have ever resided together. The range of relationships has also included other household members, such as relatives and roommates. As the definitions of behavior and relationships have broadened, the incidents considered family violence in general and IPV in particular have expanded, leading to more arrests for this behavior.

Before the adoption of these policies, many feminist scholars and organizations critiqued police and lawmakers for failing to take IPV seriously. Even though recognized by the state as a crime, police often did not treat such violence accordingly, sometimes failing to arrest perpetrators even when requested by the victim (Ferraro, 1989; Hoctor, 1997; Langley & Levy, 1977). Partly in response to critiques of arrest policies, the Minneapolis Domestic Violence Experiment (MDVE) was conducted to determine empirically their deterrent efficacy (Sherman & Berk, 1984a, 1984b). This experiment marked a dramatic change in police responses to IPV, showing that arrested perpetrators were less likely to re-offend postarrest. Although Sherman and Berk (1984a) “suggest a cautious interpretation of the findings” (p. 8), the publicity sought by the researchers overshadowed the limited scope and location of the research (Binder & Meeker, 1988). Consequently, many states began passing mandatory and proarrest policies shortly after the findings were published (Berk, Campbell, Klap, & Western, 1992). These policies proliferated despite several forms of evidence citing problems with the MDVE (for extensive reviews, see Garner, Fagan, & Maxwell, 1995; Williams, 2005). For example, Binder and Meeker (1988) argued that only a few police officers issued treatment to a large proportion of the experiment’s sample. In addition, the sample consisted of largely minority and lower class males, which was not representative of all perpetrators. They also argued that treatment options were not standardized, as shown by the fact that the “advising” option included a wide range of actions, from counseling the parties to doing nothing (Binder & Meeker, 1988). These discrepancies may have introduced sample bias and unknown effects of the heterogeneous category of “advising.”

The Problem of Dual Arrests

The implementation of mandatory and proarrest policies brought about unintended consequences. Most relevant to the current study was an increase in dual arrests, meaning both intimate partners were arrested during an IPV incident. The legitimacy of this police practice has been widely debated. Some have argued that although women may engage in violent behavior during IPV incidents, that violence is self-defensive. Hence, “true” victims are being arrested needlessly, undermining the intent of mandatory and proarrest policies (e.g., Chesney-Lind, 2002). Others have argued that women perpetrate violence as much as men do, a fact which law enforcement is only beginning to realize (Straus 2005, 2009, 2011; Winstok, 2012). In any case, several studies have found that when these policies are in effect, dual arrest is much more likely (Finn, Blackwell, Stalans, Studdard, & Dugan, 2004; Frye, Haviland, & Rajah, 2007; Hirschel et al., 2007; Miller, 2001).

The Gender and IPV Debate

Two competing perspectives have dominated the debate about the role of gender in IPV (e.g., Winstok, 2011, 2012). The first perspective has been guided by feminist theory, asserting that IPV is largely a gendered crime. Proponents of this framework have asserted that males are usually the perpetrators of IPV, with the patriarchal notion that men dominate women residing at the core of IPV (e.g., DeKeseredy, 2011; Dobash et al., 1992; Kimmel, 2002; Tjaden & Thoennes, 2000).

The second approach, the family violence researchers’ perspective, has entailed arguments that little to no empirical evidence exists that patriarchy is the source of IPV (e.g., Dutton, 2006, 2010; Felson, 2002; Straus, 2011). Proponents of this perspective have viewed IPV as emanating from a variety of factors, and although acknowledging that IPV may result from a desire for power and control, they have contended that women have as much desire to dominate their partners as men (e.g., Dutton, 2006; Felson, 2002; Straus, 2005, 2009).

Dual Arrest and the Gender-IPV Controversy

The debate about gender symmetry or asymmetry in IPV is not likely to be resolved soon or perhaps ever on the basis of any empirical evidence. That conundrum is the result of the balkanization of both theoretical campus operating from divergent paradigmatic perspectives on the definition, measurement, and interpretation of IPV (e.g., Winstok, 2011, 2012). Nevertheless, comparing incidents resulting in a single arrest with those involving a dual arrest offers a novel empirical approach to this debate.

From a feminist perspective, women may become ensnared in a dual arrest situation because they engaged in self-defensive violence or what Johnson (2008) has identified as “violent resistance.” In support of the feminist perspective, researchers have found that women are more frequently arrested for IPV as part of a dual arrest situation; conversely, men are more likely to be arrested alone (Henning & Feder, 2004; Martin, 1997). Other researchers have found that among women involved in a dual arrest, most cannot be classified as primary aggressors in their relationships and are less likely to have previous criminal involvement than men (Henning & Feder, 2004; Henning, Renauer, & Holdford, 2006; Martin, 1997). Hence, gender asymmetry in the assessed risk of recidivism post arrest and actual recidivism post assessment should persist among men and women involved in dual arrest situations.

From a family violence researchers’ perspective, dual arrest may be a manifestation of “common couple violence,” resulting from conflicts in intimate relationships that escalated from arguments to violent outcomes. Such violence is typically episodic, infrequent, and minor in nature (Johnson, 2011; Straus, 2011). Hence, the assessed risk of recidivistic violence after arrest and the actual occurrence of recidivistic violence after assessment should be lower in dual arrest compared to single arrest situations, and assessed risk and actual recidivism should be similar between women and men (i.e., gender symmetry).

The present study sought to differentiate between these contending theoretical perspectives by empirically examining the gender differences between men and women arrested alone or with their partners. Two behavioral characteristics were used in making these comparisons: the assessed risk of persistent IPV (conducted shortly after arrest) and actual persistence in IPV postassessment (i.e., recidivistic violence during an 18-month follow-up period). Finding that assessed risk and actual IPV recidivism is higher for men in dual arrest situations would support the feminist perspective. Finding that such risk and recidivism is similar between men and women or that women have higher risk and recidivism than men in dual arrest situations would support the family violence researchers’ perspective. This study, therefore, brings new evidence to bear on the gender and IPV controversy by focusing on risk and recidivism in single versus dual arrest situations, an analysis not previously reported in this literature.

Method

Data on risk assessments conducted in February and March of 2007 for all persons arrested on family violence charges within the State of Connecticut were utilized. All assessed cases were tracked for 18 months postassessment (N = 3,569). The sample represented the full population of cases 16 years of age or older for the entire state during the assessment and follow-up period. Connecticut is an ideal location for this research because it has a high frequency of dual arrest and a mandatory arrest policy without a primary aggressor provision (Conn. Gen. Stat. § 46b-38a). That provision encourages police to determine a primary perpetrator and thus reduce the likelihood of dual arrest (Hirschel et al., 2007).

The statutory definition of family and household members in Connecticut included a wide variety of relationships (Conn. Gen. Stat. § 46b-38a). However, this study restricted the sample to heterosexual intimate partners, including current or former spouses, unmarried couples living together or previously living together, those currently or previously involved in a dating relationship, and those having a child in common regardless of whether they have ever lived together. This restriction eliminated 1,304 cases of family violence arrests involving other types of family and household members, such as aunts, uncles, cousins, siblings, parents and children, or roommates. Although incidents involving such persons are important forms of violence and worthy of empirical investigation, they fall outside the scope of the gender and IPV controversy. Same-sex relationships were also excluded for this reason, but they comprised only a small number of IPV cases (N = 110). These restrictions resulted in a total sample size of 2,155 perpetrators; 73.92% (1,593) were male and 26.08% (562) female. Restricting the analysis to heterosexual intimate partners resulted in only a three percent decrease in dual arrest cases (from 34% for the total sample to 31%). Additionally, 43.76% (943) of the heterosexual IPV perpetrators were non-Latino White and 56.24% (1,212) were ethnic minorities, predominantly Latinos and African Americans. Age of perpetrator ranged from 18 to 81 years, with an average age of 37.

Measurement of Behavioral Outcome Variables

Whenever an arrest was made by police on a family violence charge in Connecticut, a risk assessment was conducted by Family Relations Counselors (FRCs), most of which (slightly over 85%) have advanced degrees. Police did not conduct the risk assessments; Family Relations Counselors (FRCs) representing trained, clinical professionals independently performed them. This assessment was designed to estimate the likelihood that the perpetrator will persist in IPV against his or her partner. It was conducted with the Domestic Violence Screening Instrument, Revised (DVSI-R). This instrument has been used throughout Connecticut since 2004 and has been evaluated in two different predictive validation studies (Williams, 2011; Williams & Grant, 2006). Application of the DVSI-R yielded an objective total numeric risk score based on 11 items. Seven of these focused on the previous behavior of the perpetrator, such as prior nonfamily assaults, arrests, or criminal convictions or prior family violence assaults, threats, or arrests. The other four items addressed perpetrator and/or situational characteristics, specifically, employment status and substance abuse, children present, or objects used as a weapon at the time of the current or previous incident. The scoring for each of these items ranged from zero (no evidence) to two or zero to three, depending on the item, estimating the intensity of the risk factor, not simply whether it was present for the perpetrator being assessed.

FRCs coded the DVSI-R items during in-take interviews, and the total numeric risk score was automatically generated by a computerized system. The assessments drew from five sources of information: perpetrator interviews, a review of police reports, criminal history reviews, protective and restraining order registry reviews, and victim interviews usually conducted by victim advocates, with information passed on to FRCs if victims gave consent. The possible range of total numeric risk scores was from zero to 28. For this sample, the actual range was zero to 26, with a mean of 8.53 (SD = 5.71). In short, the DVSI-R total numeric risk score is a composite measure based on objective information obtained from multiple (five) sources of information. It is not a subjective measure based on a single evaluator’s perception of risk, particularly that of the police. The FRCs had been trained on risk assessment in general and the DVSI-R in particular, with booster trainings occurring periodically within the 23 FRC offices throughout Connecticut. A detailed discussion about the development of this instrument and its psychometric properties is provided elsewhere (Williams, 2008; Williams & Grant, 2006).

However, included in the DVSI-R were two more subjective assessment tools that allowed the FRCs to provide their own perception of the likelihood of recidivistic violence in the near future (i.e., next 6 months). The first item focused on the “imminent risk of violence” to the victim of the incident for which the perpetrator was arrested. The second addressed the “imminent risk of violence” to another person known by the offender or victim. These items were scored as follows: 0 = low risk, 1 = moderate risk, and 2 = high risk. They allowed the FRCs to provide their own subjective interpretations of the threat of recidivistic violence posed by the perpetrator, independent of the objective DVSI-R total numeric risk scores. Hence, the present study used both structured subjective assessments (i.e., interpretations informed by administering a statistical assessment of risk) and objective (i.e., statistical) assessments of risk.

The final outcome measure was the overall prevalence of recidivism for any IPV offense, excluding arrests involving other criminal offenses (N = 294). Whether or not a perpetrator recidivated was dummy coded, with 53.09% (988) not recidivating, meaning no arrests for any offenses during the 18-month follow up period, and 46.91% (873) committing some family violence offense within the 18-month follow up period; that is, either a new family violence crime, a violation of a protective or restraining order, or both. Arrests involving nonfamily violence criminal offenses (N = 294) essentially represent “ambiguous” cases concerning IPV recidivism, and recent research has shown such cases introduce estimation error in family violence risk and recidivism analyses (e.g., Hilton & Harris, 2009; Williams, 2011). Hence, they were excluded.

Analytical Procedures

The primary objective of the analysis was to determine the gender differences (if any) in the assessed risk of recidivism and actual IPV recidivism by arrest category (i.e., single or dual arrest). Scoring women perpetrators zero and male perpetrators one created a gender dummy variable. Since the sample consisted of heterosexual IPV incidents, by definition, male perpetrators had female victims and female perpetrators had male victims in both single and dual arrest situations. Hence, in all estimated models, this gender dummy variable was included to determine differences or similarities between male-to-female and female-to-male IPV. Additionally, all models included a dummy variable indicating single (scored zero) and dual arrest cases (scored one).

The analysis was conducted in three stages. First, regression models were estimated to determine the estimated effects of the gender dummy variable on the three measures of assessed risk of recidivism: (a) the DVSI-R total numeric risk score, (b) the perceived imminent risk to victim score, and (c) the perceived imminent risk to others score. Ordinary Least Squares (OLS) regression was used to estimate the effect of the gender dummy variable on the three measures of assessed risk, with perpetrator age and ethnicity (coded as non-Latino White = 1 and all other ethnic minorities = 0) included as control variables in all estimated models. However, supplementary analyses were conducted because of the distribution or measurement characteristics of the assessed risk measures.

Specifically, the distribution of the total DVSI-R numeric risk scores had a positive skew. Hence, negative binomial regression was conducted to determine whether the empirical results corroborated the OLS findings. Further, the perceived imminent risk items were scored using ordinal rankings; hence, ordered logistic regression was conducted to determine, again, whether the results corroborated the OLS findings. Substantive conclusions drawn from these supplementary analyses remained the same as those of the OLS analyses. Consequently and for simplicity, only the OLS findings were reported below (although the results of the supplementary analyses are available from the authors upon request). All analyses were conducted with StataSE 11.2.

Second, logistic regression was used to estimate gender effects on 18-month IPV recidivism because it is a dichotomous dependent variable. This logistic regression was hierarchal in structure, involving two steps. The first step included the gender dummy variable as well as perpetrator age and ethnicity. The second step included all the assessed risk measures to determine whether the estimated effects of gender on IPV recidivism persisted, controlling for the assessed risk of recidivism, in addition to perpetrator age and ethnicity. These measures constitute a composite of other characteristics associated with IPV but also related to perpetrator and victim gender. Hence, controlling for assessed risk reduces the likelihood of biased estimation.

Third, a dummy variable indicating whether the arrest involved both partners as perpetrators (i.e., dual arrest scored one) or only a single perpetrator (i.e., single arrest scored zero) was added to the OLS assessed risk models as well as the logistic regression IPV recidivism models to determine whether assessed risk and actual recidivism was higher or lower in dual arrest situations. An interaction term (i.e., cross-product) of dual arrest by gender was then added to determine whether the estimated effects of the gender dummy variable on risk and recidivism differed significantly between dual arrest and single arrest conditions.

Empirical Results

For this sample of heterosexual intimate partners, males committed the majority of IPV, with 73.92% of offenders being male and 26.08% being female. Dual arrest cases accounted for about 31% of all IPV arrests. Approximately 79% of the 1,591 arrests involving a male perpetrator and female victim were single arrests, while 21% were dual arrests. Of 561 arrests involving a female perpetrator and male victim, about 42% were single arrests and 58% were dual arrests. These results were consistent with previous research findings that single arrests were more likely for male perpetrators, while dual arrests were more likely for female perpetrators (Henning & Feder, 2004; Martin, 1997). The difference between these two types of IPV by dual arrest was significant (Chi² = 268.86, p = .00).

Estimating Effects of Perpetrator Gender on Risk and Recidivism

The empirical results of estimating the effects of the gender dummy variable on the three measures of assessed risk are reported in Table 1. Consider the results for the DVSI-R total numeric risk score. Male perpetrators with female victims resulted in a statistically significant increase of 2.48 units of assessed risk compared to female perpetrators with male victims, with an overall effect size of Beta = .19. The estimated effects of both perpetrator ethnicity and age, however, were not statistically significant. Consider now the results for the two perceived imminent risk scores. Once again, the gender dummy variable had statistically significant estimated effects. Holding perpetrator ethnicity and age constant, male perpetrators with female victims resulted in an increase of .38 units of perceived imminent risk to the victim compared to female victims with male perpetrators, with an overall effect size of Beta = 0.21. The estimated effects of perpetrator ethnicity and age were also not statistically significant in this model. Similarly, male perpetrators with female victims resulted in an increase of 0.28 units of perceived imminent risk to others compared to female perpetrators with male victims, with an overall effect size of Beta = 0.17. The estimated effects of perpetrator ethnicity and age were again insignificant in this estimated equation.

Table 1.

Ordinary Least Squares Estimated Effects of Gender on Assessed Risk.

	DVSI-R numeric			Imminent Risk to Victim			Imminent Risk to Others
	B	SE B	β	B	SE B	β	B	SE B	β
Gender^a	2.48***	0.28	0.19	0.38***	0.38	0.21	0.28***	0.04	0.17
Ethnicity^b	0.37	0.25	0.03	0.02	0.03	0.01	−0.02	0.03	−0.02
Age	0.00	0.01	0.01	−0.00	0.00	−0.01	−0.00	0.00	−0.03
Constant	6.39***	0.47		0.70***	0.06		0.47***	0.06
Adjusted R ²		.04			.04			.03

Note: DVSI-R = Domestic Violence Screening Instrument, Revised; SE = standard error.

1 = Male-to-Female IPV, 0 = Female-to-Male IPV. ^b1 = non-Latino White, 0 = Ethnic Minorities.

***

p < .001.

The empirical results of the logistic regression analysis involving gender and IPV recidivism are reported in Table 2. Holding perpetrator ethnicity and age constant, male perpetrators with female victims were almost twice (Odds Ratio = 1.97) as likely to recidivate as females with male victims. Perpetrator ethnicity did not have a statistically significant estimated effect on recidivism, but perpetrator age did have a statistically significant but small estimated effect, with a 1-year increase in age associated with about a 3% decline in the likelihood of recidivism (Odds Ratio = .97). Further, adding the assessed risk measures to the equations did not significantly change these results. Male perpetrators with female victims continued to be more likely to persist in IPV postarrest and postassessment (Odds Ratio = 1.62), with age continuing to have a small negative estimated effect on IPV recidivism (Odds Ratio = .97). As in previous analyses of the DVSI-R, the estimated effect of the total numeric risk scores, but not the perceived imminent risk items, had statistically significant and positive estimated effects on IPV recidivism (Odds Ratio = 1.09).

Table 2.

Logistic Regression Results of Gender on 18-Month IPV Recidivism.

	B	SE B	OR	B	SE	OR
Gender^a	.68***	.11	1.97	.48***	.12	1.62
Ethnicity^b	−.01	.10	.99	−.04	.10	.96
Age	−.03***	.01	.97	−.03***	.00	.97
DVSI-R total numeric risk score				.08***	.01	1.09
Perceived imminent risk to other				.17	.10	1.19
Perceived imminent risk to victim				.03	.11	1.03
Constant	.43*	.18			−.20	.20
Pseudo R²		.03				.08

Note: DVSI-R = Domestic Violence Screening Instrument, Revised; IPV = intimate partner violence; SE = standard error; OR = odds ratio.

1 = Male-to-Female IPV, 0 = Female-to-Male IPV. ^b1 = non-Latino White, 0 = Ethnic Minorities.

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

Estimating Effects of Dual Arrest on Risk and Recidivism

The final stages of the analysis were designed to estimate (a) the independent effects of dual arrest on risk and recidivism, and (b) the interaction effects of dual arrest by gender on risk and recidivism. The objective of estimating independent effects was to determine whether dual arrest reflects “common couple violence,” compared to more serious IPV or “intimate terrorism,” as reflected at least partially in single arrest situations. If so, dual arrests should be associated with lower risk and recidivism compared to single arrests. Estimating interaction effects was done to determine whether patterns of gender symmetry or gender asymmetry were revealed in dual arrest compared to single arrest situations.

The findings bearing on assessed risk are displayed in Table 3. The pattern of empirical results was similar between the DVSI-R total numeric risk scores and the perceived imminent risk items. Hence, for simplicity and brevity, the results reported in Table 3 were limited to the numeric risk scores (results involving the perceived imminent risk items are available upon request from the authors). Two findings are key in Table 3. First, dual arrests were significantly related with lower risk scores, with a shift from the single to the dual arrest category associated with almost a two-point decline in total numeric risk (B = −1.88). Second, no evidence was found of a statistically significant interaction effect involving dual arrests by gender. Recall that the findings in Table 1 showed statistically significant gender differences in assessed total numeric risk; therefore, finding no statistically significant interaction effects (see Table 3) suggests that such differences persisted across the two arrest categories (i.e., single and dual arrest situations).

Table 3.

Ordinary Least Squares Regression Results for the Independent and Interaction Effects of Dual Arrest on the DVSI-R Total Numeric Risk Score.

	Model 1			Model 2
	B	SE B	β	B	SE B	β
Gender^a	1.81***	.29	.14	2.07***	.40	.16
Ethnicity^b	.39	.25	.03	.40	.25	.03
Age	.00	.01	.01	.00	.01	.01
Dual arrest	−1.88***	.28	−.15	−1.50**	.48	−.12
Dual arrest × Gender				−.58	.59	−.04
Constant	7.50***	.50		7.26***	.55
Adjusted R ²		.06			.06

Note: DVSI-R = Domestic Violence Screening Instrument, Revised; IPV = intimate partner violence.

1 = Male-to-Female IPV, 0 = Female-to-Male IPV. ^b1 = Non-Latino White, 0 = Ethnic Minorities.

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

The empirical results bearing on IPV recidivism are presented in Table 4. Those findings revealed a similar pattern that can be summarized succinctly. Dual arrests, compared to single arrests were significantly associated with about a 29% reduction in recidivism during the 18-month follow-up period (Odds Ratio = .71), and again, no evidence was found of statistically significant interaction effects, meaning the previously reported gender differences in recidivism (see Table 2) held for both the single and dual arrest categories.

Table 4.

Logistic Regression Results for the Independent and Interaction Effects of Dual Arrest on 18-Month Intimate Partner Violence (IVP) Recidivism.

	Model 1			Model 2
	B	SE B	OR	B	SE B	OR
Gender^a	.56***	.12	1.75	.69***	.16	2.00
Ethnicity^b	.00	.10	1.00	.01	.10	1.01
Age	−.03***	.00	.97	−.03***	.00	.97
Dual arrest	−.34**	.11	.71	−.15	.19	.86
Dual arrest × gender				−.30	.23	.74
Constant	.62***	.20		.51*	.22
Pseudo R ²		.03			.03

Note: SE = standard error; OR = odds ratio.

1 = Male-to-Female IPV, 0 = Female-to-Male IPV. ^b1 = non-Latino White, 0 = Ethnic Minorities.

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

Discussion and Conclusion

Overall, the findings showed that a male perpetrator with a female victim was significantly associated with an increase in both assessed risk and IPV recidivism. Although the estimated effects of the gender dummy variable on these two outcome measures were slightly lower in the dual arrest compared to the single arrest category, that reduction was not statistically significant, as indicated by the insignificant interaction term of dual arrest by gender. In short, whether or not one person or both partners were arrested in an IPV incident, male perpetrators with female victims were both at higher risk of persistent IPV and actually were involved in persistent IPV during the 18-month follow-up period after initial arrest and assessment. This finding is striking since dual arrest appeared to reflect “common couple violence,” not more serious “intimate terrorism,” as indicated by the overall lower risk and recidivism found in the dual arrest, compared to the single arrest category.

Nonetheless, two caveats must be stated. First, advocates of both the feminist and the family violence researchers’ perspective have acknowledged that different sampling methods may yield biased estimates of the gender configuration of IPV, that is, whether it is symmetrical or asymmetrical by gender (for recent contending reviews, see Johnson, 2011; Straus, 2011). Stated briefly, general population surveys may overrepresent more minor, situational couple violence and underrepresent more serious IPV, such as intimate terrorism. Conversely, agency samples, such as emergency room data, batterer or victim intervention data, or police data may underrepresent the more minor IPV and overrepresent the more serious IPV because the more serious cases tend to come to the attention of official agencies charged with the responsibility of responding to such violence. However, IPV cases in Connecticut range in seriousness, as indicated by the official charges at arrest, from misdemeanors (about 78% of the cases), such as third degree assault (31%), breach of peace (13%), or disorderly conduct (14%), to felonies (about 22% of the cases), such as first or second degree assault (2%), criminal violation of a restraining order or violation of a protective order (25%), or risk of injury to children (25%). Straus (2011) has recently argued that the “bias” is more a function of whether research measures and analyzes the perpetration of IPV versus the effects of such perpetration rather than sampling methods used in research on this topic. He reviewed studies showing that regardless of the nature of the sample analyzed, gender symmetry tended to prevail when perpetration was the focus, but gender asymmetry was more commonly revealed when the focus was on the effects of such perpetration (typically injury or need for medical attention).

Regardless, the predominance of misdemeanor charges in the sample of IPV cases analyzed here suggests the sample is not “biased” in the direction of more serious cases. Moreover, the results that showed the dual arrest cases had significantly lower total numeric risk scores and lower persistent involvement in IPV over an 18-month period postarrest and postassessment; yet the gender asymmetry in the risk of recidivism assessed independently by trained FRCs and actual IPV recidivism during that period remained for these cases. An alternative research approach would be to address the gender symmetry or asymmetry issue through a general population survey in which data on risk assessments and persistent IPV measured independently (different sources of data) over time would be collected, but the logistics and costs of that approach would be daunting. Hence, although clearly not definitive, the findings of the present study, based on police arrest data, must suffice and were at least consistent with the feminist perspective but not the family violence researchers’ perspective.

The second caveat bears on the issue of geographic specificity. The data were gathered from only one state, and therefore, the findings should be replicated in other locations to establish empirically their generalizability. Given that Connecticut has a mandatory arrest policy but no primary aggressor provision, a particularly important extension of the present study would be to compare Connecticut to another state having neither such a policy nor such a provision as well as to another state having both a mandatory arrest policy and a primary aggressor provision. The purpose of those comparisons would be to determine the extent to which these legal prescriptions concerning arrest influence the frequency and gender configuration of IPV and especially dual arrests for this form of violence.

Conducting research in other states having different legal prescriptions for police work would be especially important because one could argue that such work is permeated by a gendered perception of IPV—that it is asymmetrical with men commonly perpetrators and women commonly victims (Winstok, 2012). However, placing more restrictive legal prescriptions on police work would, at least in principle, curb the influence of that perception, assuming it exists. Hence, that perceptual bias would be more likely to influence police work in states having either discretionary policies or even those having proarrest or mandatory arrest but also having a primary aggressor provision. That is the case because subjectivity is more likely to intrude with unrestricted police discretion and in the assessment of which partner to the relationship is the primary aggressor. Having only a mandatory arrest policy, as in Connecticut, reduces the likelihood of biased perception influencing police decisions about arrest. Granted, these are empirical issues that should be explored by further research in different states having different legal regulations concerning arrest for IPV.

It should be noted, however, that police in Connecticut exercised their judgment about IPV incidents only in making arrests. As described above, trained, professional FRCs using multiple sources of information conducted assessments of recidivistic risk independently. Hence, the gendered perception is less likely in these assessments. Evidence of this assertion was found in the estimated effects of the DVSI-R total numeric risk score, a composite score based on objective data drawn from five sources of information, compared to the perceived imminent risk of violence to victims or others on IPV recidivism in the 18-month follow up period. Only the total numeric risk scores had statistically significant and positive estimate effects on subsequent IPV.

Though not conclusive, one policy implication of the findings is that mandatory arrest policies may have negative, unanticipated consequences for the women victims they were intended to protect. If this is the case and if the findings of the present study are replicate elsewhere, such policies and their implementation should be carefully re-examined, monitored, and possibly modified.

Footnotes

Acknowledgements

Special appreciation is expressed to William Carbone, Executive Director of CSSD. Appreciation is also expressed (in alphabetical order) to Kathy Ceruti, Joe DiTunno, Steve Grant, Brian Hill, Deb Kulak, and Brian Sperry. CSSD owns the copyright of the Domestic Violence Screening Instrument (DVSI-R).

Authors’ Note

This research was conducted with assistance from the Judicial Branch, Court Support Services Division (CSSD) of the State of Connecticut.

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

Author Biographies

Caryn Bell Gerstenberger is a PhD student at the University of California, Riverside. Her research addresses juvenile delinquency and intimate partner violence.

Kirk R. Williams, PhD, recently moved to the University of Delaware to become the chair of the Department of Sociology and Criminal Justice. He has published widely on the causes and prevention of violence, particularly involving youth or adult intimate partners, with the most recent publications addressing bullying, juvenile offending, and domestic violence risk assessment.

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.

Binder

Meeker

J.W.

(1988). Experiments as reforms. Journal of Criminal Justice, 16, 347-358.

Chesney-Lind

(2002). Criminalizing victimization: The unintended consequences of pro-arrest policies for girls and women. Criminology and Public Policy, 2, 81-91.

DeKeseredy

W. S.

(2011). Feminist contributions to understanding women abuse: Myths, controversies, and realities. Aggression and Violent Behavior, 16, 297-302.

Dobash

R. P.

Dobash

R. E.

Wilson

Daly

(1992). The myth of sexual symmetry in marital violence. Social Problems, 39(1), 71-91.

Dutton

D. G.

(2006). Rethinking domestic violence. Vancouver, BC: UBC Press.

Dutton

D. G.

(2010). The gender paradigm and the architecture of antiscience. Partner Abuse, 1, 5-25.

Felson

R. B.

(2002). Violence and gender reexamined. Washington, DC: American Psychological Press.

Ferraro

K. J.

(1989). Policing woman battering. Social Problems, 36, 61-74.

10.

Finn

Blackwell

B. S.

Stalans

Studdard

Dugan

(2004). Dual arrest decision in domestic violence cases: The influence of departmental policies. Crime & Delinquency, 50, 565-589.

11.

Frye

Haviland

Rajah

(2007). Dual arrest and other unintended consequences of mandatory arrest in New York City: A brief report. Journal of Family Violence, 22, 397-405.

12.

Garner

Fagan

Maxwell

(1995). Published findings from the spouse assault replication program: A critical review. Journal of Quantitative Criminology, 11(1), 3-28.

13.

Henning

Feder

(2004). A comparison of men and women arrested for domestic violence: Who presents the greater threat? Journal of Family Violence, 19(2), 69-80.

14.

Henning

Renauer

Holdford

(2006). Victim or offender? Heterogeneity among women arrested for intimate partner violence. Journal of Family Violence, 21, 351-368.

15.

Hilton

N. Z.

Harris

G. T.

(2009). How nonrecidivism affects predictive accuracy: Evidence from a cross-validation of the Ontario Domestic Assault Risk Assessment (ODARA). Journal of Interpersonal Violence, 24, 326-337.

16.

Hirschel

(2008). Domestic violence cases: What research shows about arrest and dual arrest rates. Washington, DC: Office of Justice Programs, National Institute of Justice.

17.

Hirschel

Buzawa

Pattavina

Faggiani

Reuland

(2007). Explaining the prevalence, context, and consequences of dual arrest in intimate partner cases. Washington, DC: U.S. Department of Justice.

18.

Hoctor

M. M.

(1997). Domestic violence as a crime against the state: The need for mandatory arrest in California. California Law Review, 85, 643-700.

19.

Johnson

M. P.

(2006). Conflict and control: Gender symmetry and asymmetry in domestic violence. Violence Against Women, 12, 1003-1018.

20.

Johnson

M. P.

(2008). A typology of domestic violence. Boston, MA: Northeastern University Press.

21.

Johnson

M. P.

(2011). Gender and types of intimate partner violence: A response to an anti-feminist literature review. Aggression and Violent Behavior, 16, 289-296.

22.

Kimmel

M. S.

(2002). “Gender symmetry” in domestic violence. Violence Against Women, 8, 1332-1363.

23.

Langley

Levy

R. C.

(1977). Wife beating: The silent crisis. New York: E. P. Dutton.

24.

Martin

M. E.

(1997). Double your trouble: Dual arrest in family violence. Journal of Family Violence, 12(2), 139-157.

25.

Meloy

M. L.

Miller

S. L.

(2011). The victimization of women: Law, policies, and politics. Oxford: Oxford University Press.

26.

Miller

S. L.

(2001). The paradox of women arrested for domestic violence: Criminal justice professionals and service providers respond. Violence Against Women, 7, 1339-1376.

27.

Sherman

L. W.

Berk

R. A.

(1984a). The Minneapolis domestic violence experiment. Police Foundation Reports, 1, 1-8.

28.

Sherman

L. W.

Berk

R. A.

(1984b). The specific deterrent effects of arrest for domestic assault. American Sociological Review, 48, 261-272.

29.

Straus

M. A.

(2005). Women’s violence toward men is a serious social problem. In Gelles

R. J.

Loseke

D. R.

(Eds.), Current controversies on family violence (pp. 55-77). Newbury Park, CA: Sage.

30.

Straus

M. A.

(2009). Why the overwhelming evidence on partner physical violence by women has not been perceived and is often denied. Journal of Aggression, Maltreatment & Trauma, 18, 552-571.

31.

Straus

M. A.

(2011). Gender symmetry and mutuality in perpetration of clinical-level partner violence: Empirical evidence and implications for prevention and treatment. Aggression and Violent Behavior, 16, 279-288.

32.

Tjaden

Thoennes

(2000). Prevalence and consequences of male-to-female and female-to-male intimate partner violence as measured by the national violence against women survey. Violence Against Women, 6, 142-161.

33.

Williams

K. R.

(2005). Arrest and intimate partner violence: Toward a more complete application of deterrence theory. Aggression and Violent Behavior, 10, 660-679.

34.

Williams

K. R.

(2008). Domestic violence screening instrument (DVSI). In Cutler

B. L.

(Ed.), Encyclopedia of Psychology and Law (pp. 240-242). Newbury Park, CA: Sage.

35.

Williams

K. R.

(2011). Family violence risk assessment: A predictive cross-validation study of the domestic violence screening instrument-revised (DVSI-R). Law and Human Behavior. doi:10.1007/s10979-011-9272-6

36.

Williams

K. R.

Grant

S. R.

(2006). Empirically examining the risk of intimate partner violence: The revised domestic violence screening instrument (DVSI-R). Public Health Reports, 121, 400-408.

37.

Winstok

(2011). The paradigmatic cleavage on gender differences in partner violence perpetration and victimization. Aggression and Violent Behavior, 16, 303-311.

38.

Winstok

(2012). The effect of sex and severity of aggression on formal and informal social agents’ involvement in partner violence. American Journal of Men’s Health, 6, 136-145.