Can a “Psychosocial Model” Help Explain Violence Perpetrated by Female Batterers?

Abstract

Objective:

The purpose of the study was to evaluate the psychosocial predictors of propensity for abusiveness among a large sample of women ordered into a 26-week batterer intervention program (BIP).

Method:

The study employed a nonequivalent, control group design (comparing program completers to dropouts) in a secondary analysis of 485 women.

Results:

Analysis indicated that (1) there were no significant differences between program completers and dropouts on their propensity for abusiveness. (2) There are specific family of origin indicators associated with higher levels of propensity for abusiveness among women batterers. (3) Family of origin indicators leading to increased propensity for abusiveness were significantly different among racial groups.

Conclusion:

These findings reveal characteristics of women batterers, as they relate to the propensity for abusiveness and provide preliminary evidence suggesting exposure to various forms of violence in family of origin helps explain the use of violence by adult women against their intimate partners.

Keywords

domestic violence batterers female offenders propensity for abusiveness

Introduction

Since the start of the women’ shelter movement in the 1970s, the dominant force in conceptualizing domestic violence later evolving into the term intimate partner violence (IPV) has been the feminist movement. By way of a summary, in overly simplistic terms, the feminist paradigm sees acts of domestic violence as a physical manifestation of patriarchy. Men hit women because they can and their violence in these relationships is endorsed by larger cultural norms (Dobash & Dobash, 1979). There is no way to overstate the importance of this paradigm on shaping public policy and cultural mores. The policy culmination of this position at the federal level was the violence against Women Act, originally signed by President Clinton in 1994 and renewed in 2013, which has moved billions of dollars to domestic violence shelters and agencies working to assist women and families abused by male perpetrators.

Despite the obvious value the feminist paradigm has had on shaping both our ideas about how violence gets played out in intimate relationships and the roles of the players, there is the increasing evidence that the feminist position is not monolithic. As the previously mentioned discussion of federal policy illustrates, one of the easiest ways to identify gaps in the model is to interrogate how the feminist model could explain violence occurring in nonheterosexual relationships. Specifically, if IPV is a consequence of patriarchy, it stands to reason that lesbian relationships should be violence free and gay relationships should be replete with violence. Although the literature on IPV in lesbian, gay, bisexual, and transgender (LGBT) relationships is limited (for an excellent review of the available literature, see Canon & Buttell, 2015), there is increasing evidence that violence is occurring in same-sex relationships at rates similar to heterosexual couples (Blosnich & Bossarte, 2009; Carvalho, Lewis, Derlega, Winstead, & Viggiano, 2011; Messinger, 2011).

In addition to the limitations of the feminist model in capturing and adequately explaining the violence occurring in LGBT relationships, there has long been a disagreement between the feminists and what have been termed the “family violence” scholars in explaining the violence occurring in heterosexual relationships. The gender paradigm of IPV argues that domestic violence is a result of patriarchal social systems where men are exclusively the batterers and females are exclusively the victims of male dominance and privilege. Consequently, female violence in intimate relationships is initiated reactively, purely as a form of self-defense or in retaliation for violence that was initiated by men (Dobash, Dobash, Wilson, & Daly, 1992). Where adherents of the gender paradigm may acknowledge female participation in IPV, they are largely concerned with the specificities of social context and differences in physicality. They argue that gender symmetry in domestic violence is a myth created and perpetuated by family violence researchers’ methodological failure to address the structural components of patriarchy. In addition, gender paradigm scholars argue that family violence research lumps disparate forms of abuse and injury into the same broad category of “domestic violence” (Dobash et al., 1992; Winstok, 2011).

In contrast to the feminist paradigm, family violence scholars endorse a definition of violence that lacks gendered connotations—one that is predicated on the idea that violence, aggression, and exploitation in its variant forms is dangerous, destructive, and consequential regardless of the method of violence or the gender of the perpetrator (Felson, 2006). Family violence scholars do not dismiss gender as a factor in IPV. Rather, they dismiss the utilization of gender as the sole focal point of policy formation and scholarly inquiry. Family violence scholars privilege a paradigm that acknowledges IPV to be the result of more than one demographic characteristic and endorse the construction of studies and policies that seek to appreciate, acknowledge, and assist both male and female victims (e.g., Carney, Buttell, & Dutton, 2006; Dutton & Nicholls, 2005; ). Certainly, one of the most salient differences between the feminist paradigm and the family violence paradigm is that, for the former, male offensive actions are unilaterally considered violent while every aggressive action perpetrated by women is committed in self-defense. Family violence scholars reject the idea that patriarchy alone explains violence occurring in intimate relationships and argue instead that gender is just one potential determinant within a broad of array of psychosocial factors contributing to IPV (Langhinrichsen-Rohling, 2010; Winstok, 2011). In other words, for family violence scholars, the existence of patriarchal mechanisms of stratification and subordination are not the issue in dispute. Rather, the debate centers on the nature of the relationship between patriarchy and IPV. Unfortunately, this difference is the source of heated contestation since billions of dollars in funding are at stake.

Recently, in an effort to bridge some of the divide, Buttell and Starr (2013) have advocated for a psychosocial model for explaining IPV for women in Batterer Intervention Programs (BIPs). Such a model allows for a broader and nuanced understanding of IPV that would both acknowledge multiple pathways to violent perpetration in adult relationships and capture the experiences of currently marginalized groups (e.g., LGBT relationships, ethnic minorities). The gender paradigm, as it has been understood as a male violent expression of patriarchal systems (Dobash et al., 1992), forecloses this opportunity because it cannot help us understand why, for example, heterosexual women assault their male intimate partners or a femme lesbian abuses her femme lesbian partner.

The purpose of this study is to investigate the relationships between psychosocial variables and propensity for abusiveness among a large sample of women arrested, prosecuted, convicted of IPV criminal charges, and sent to a BIP as part of a criminal sentence. Specifically, it addresses the following research questions: (a) are the women who fail to complete the treatment program more likely to be abusive than those that complete treatment? and (b) what is the ability of various demographic and family of origin psychosocial variables to predict propensity for abusiveness in this sample?

Methods

Data Collection

This study employed a posttest-only design with nonequivalent groups (Cook & Campbell, 1979), which is diagrammed in Figure 1.

Figure 1.

Nonequivalent, control group design.

In Figure 1, X represents completion of the BIP and the bottom group represents program dropouts. They serve as a non-equivalent control group for treatment completers, which are represented on the top of the diagram. Furthermore, the study utilized a secondary analysis of data collected by the Domestic Abuse Center (DAC), a nonprofit agency in Columbia, SC. Since 1982, the DAC has been providing batterer intervention services based on agency and court-based referrals. DAC staff interviewed participating clients using a questionnaire covering a variety of demographic topics. This study utilized verbal one-on-one interviews rather than self-report on written questionnaires, which yielded more complete responses. Participating clients also completed two psychological instruments: the Balanced Inventory of Desirable Responding (BIDR), (Paulhus, 1984) and the Propensity for Abusiveness Scale (PAS), (Dutton, 1995). The BIDR measures socially desirable responding. The PAS measures the likelihood of a given respondent to utilize physical force in an intimate relationship. For the purpose of this study, the term dropout refers to those clients who completed the intake interview but failed to complete the entire treatment program.

Clients completed this assessment process in the first two intake sessions. Denying researchers direct access to client files preserved client privacy. Instead, an employee of the agency deleted personal information prior to providing researchers with the data obtained from the intake process. This process ensured anonymity of clients, meaning the researchers have no ability to either identify clients or link clients to their responses.

The DAC has been using the previously mentioned assessment procedure since January 2006. The purpose of the assessment package is to assist agency staff in creating a pretreatment profile of clients. This study population encompasses women referred to the BIP between the period of January 1, 2006, and June 30, 2009. Women comprised 15.6% of the total number of people referred during this period. In that time, 1,039 women were court mandated to the 26-week BIP at the DAC. Of the 1,039 women referred, 495 dropped out of the program (48%) before completing the initial intake sessions. Of the remaining women, 136 (13%) dropped out before completing the program, 335 (32%) completed the program, and 73 were still active in the program (7%) when the study was concluded. The sample for this study included all women referred during this period who attended at least two sessions and completed the assessment (N = 544).

In total, 59 women originally selected for the sample were omitted from the analysis due to high scores (>14) on either the BIDR Self Deception or Impression Management subscales. Due to social desirability biases, it was concluded that high scores for either of these measures threaten the accuracy of the responses. Thus, the final sample consisted of 485 women, of whom 284 (58%) reported being Caucasian, 180 (37%) reported being African American, and 20 (4%) reported “Other” for race.

Instruments

Propensity for abusiveness

The PAS measures the likelihood of a given respondent to utilize physical force in an intimate relationship. This scale, designed by Dutton (1995), has a score range from 32 to 160, with higher scores correlating with an increased likelihood of IPV. It measures three factors: affective labiality, recalled negative parental treatment, and trauma symptoms. The PAS was previously determined internally reliable, yielding a Cronbach’s α of .88 (Dutton, 1995). A study investigating its utility in distinguishing violent from nonviolent men found a mean score for nonviolent men of 44.7 (standard deviation [SD] = 11.7), while violent men had a mean score of 59.2 (SD = 17.1; Dutton, 1995). Although there is no research indicating the appropriateness of this measure for women offenders, the available research supports the use of the PAS in evaluating BIP comprised of diverse populations (e.g., Buttell & Carney, 2006; Carney & Buttell, 2006; Dutton, Landholt, Starzonmski, & Bodnarchuk, 2001).

Social desirability

Socially desirable responding was measured using the BIDR scale (Paulhus, 1984). The BIDR is a two-factor measure used to determine the level of socially desirable responding in the sample population. The 40-item Likert-type scale is divided into two subscales: the BIDR-SD (self-deception) and the BIDR-IM (impression management). Scores range from 0 to 20 on each of these subscales. High scores on the BIDR-SD (>14) have been shown to correspond with a lack of insight concerning problematic behaviors, while high scores on the BIDR-IM (>14) correspond with a desire to manipulate the perception of others (Kroner & Weeks, 1996). According to Kroner and Weeks (1996), these measures are effective in measuring social desirability among violent offenders. Internal consistency reliability testing of each of these dimensions of the BIDR resulted in a Cronbach’s α of .73 for the self-deception scale and .84 for the impression management scale (Kroner & Weeks, 1996). A comparison of scale differences was used to confirm external validity (Kroner & Weeks, 1996). However, it should be noted that there is no research indicating the appropriateness of these measures for women offenders and/or ethnic minorities, although they have been used in many studies of domestic violence offenders (e.g., Buttell & Carney, 2005, 2006).

Referral Sources

The majority of BIP participants (37%) were referred by a summary court that handles misdemeanors, criminal domestic violence (CDV) court (27%), or were participants in a pretrial intervention (PTI) program (25%). A small percentage (10%) reported being referred from a governmental agency, such as probation or Department of Social Services, or listed “other” as the referral source. It should be noted that each of these referral sources has a different level of client supervision and different consequences for program dropout.

Participants in the PTI receive the most active supervision. They are first time offenders, with no significant criminal history, who have been approved by both the prosecutor and the victim. Successful program completion in the PTI results in the dismissal of the pending criminal charges. Participants are required to check in with their case managers every month, as well as perform community service, test negative on drug screens, and complete a written report on domestic violence laws, and pay any required fees. Failure to meet any of these requirements is grounds for dismissal from the PTI program. If dropped from the program, clients of the PTI are returned to the regular court docket for adjudication and sentencing.

Moderate levels of supervision and support are provided for clients referred by the CDV court, with even less support provided for clients referred by the summary court. CDV courts typically have designated prosecutors, investigators, and advocates trained in the handling of domestic violence cases. The court focuses on offender accountability throughout the proceeding (Gover, Brank, & MacDonald, 2007). As a result, there is more oversight and judicial involvement in the sanctioning of offenders than one finds in summary court. Theoretically, noncompliance when referred by either of these courts should result in jail time; however, higher case loads and lack of staff charged with monitoring treatment compliance makes sanctioning less likely for those clients referred by summary court than those referred by CDV court.

The BIP

The site for this study was the DAC, a nonprofit organization that has been providing counseling services to the Columbia, South Carolina community since 1982. The intervention program is cognitive-behavioral in orientation and is consistent in organization and focus to those programs described in the literature presented (for an excellent review, see Eckhardt et al., 2013). Perhaps most important, the intervention program is similar to the national model in that it was designed for male batterers. Women who volunteered for the program and women who were arrested, prosecuted, and mandated into the program as part of a criminal sentence participate in the same curriculum as the male batterers. The intervention program is a structured, intensive, 26-week, group treatment program that focuses primarily on anger management and skills development. The intervention program incorporates three phases: (a) orientation and intake interview (2 sessions), (b) psychoeducational classes (20 sessions), and (c) group therapy regarding termination (4 sessions). Groups consist of approximately 15 batterers, female only, and meet one night each week for approximately 2 hours. This batterer treatment program incorporates confrontation, therapy, and educational components. In this setting, the common proximal events of domestic violence are directly addressed with clients, and they are given an opportunity to make changes that will positively affect their personal relationships with others. The 20-week psychoeducational program curriculum can be broken up into three successive series of group experiences. Because most offenders share a common set of defenses (minimization, denial, and blame) that foster aggressive behavior, the first series of group sessions helps participants to recognize and overcome these defense mechanisms. In this series, participants are assisted in overcoming their natural resistance to change by helping them achieve insight into their use of defense mechanisms. Thus, the first step toward modifying behavior occurs when clients recognize and accept their behavior as the problem. The second series of sessions are designed to help clients restructure their thinking by modifying the beliefs that promote violent behavior. The final series of sessions is designed to help clients increase interpersonal skills by providing them with a repertoire of alternate and appropriate behaviors. In this series, skills such as problem solving, assertiveness, and negotiation are taught and practiced in the group setting.

Results

The sample of 485 women consisted of 58.8% (N = 285) Caucasian women, 37.1% (N = 180) African American women, and was made up by a small percentage of Other women (i.e., Latino/Hispanic, Asian, American Indian) 4.1% (N = 20; Table 1). Caucasians had the highest percentage of program completion with 75.8% (N = 216), compared to other 75% (N = 15) and African American 70.0% (N = 126) completers. The mean age for Caucasian women in the study was 33.6 years (SD = 10.65), compared to 31.5 years (SD = 10.14) for African American women and 28.0 years (SD = 7.23) for other women. High school completion or higher was the highest among African American women 71.1% (N = 128) followed by Caucasian women 65.9% (N = 188) and the other women group with 55% (N = 11). Employment was the highest among other women with 70% (N = 14) being employed, followed by 53% (N = 151) for Caucasian and 51.7% (N = 111) for African American women.

Table 1.

Demographic Characteristics.

	Program Participants (N = 485)
Characteristic	Caucasian % (n)	African American % (n)	Other % (n)	Total % (N)
Race	58.8 (285)	37.1 (180)	4.1 (20)	100 (485)
Client status
Completer	44.5 (216)	25.9 (126)	3.1 (15)	73.5 (357)
Dropout	14.3 (69)	11.1 (54)	1.1 (5)	26.5 (128)
Relationship to victim
Married	67.2 (112)	28.7 (48)	4.1 (7)	100 (167)
Living together	57.9 (99)	38 (65)	4.1 (7)	100 (171)
Child in common only	32.8 (22)	64.2 (43)	3 (2)	100 (67)
Separated/divorced	70.3 (45)	23.4 (15)	6.3 (4)	100 (64)
Other	43.8 (7)	56.2 (9)	0 (0)	100 (16)
Children
Yes	58.1 (231)	37.7 (150)	4.3 (17)	100 (398)
No	62.1 (54)	34.5 (30)	3.4 (3)	100 (87)
Still with victim
Yes	61.7 (164)	34.6 (92)	3.7 (10)	100 (266)
No	55.1 (119)	40.3 (87)	4.6 (10)	100 (216)
Relationship length (months)
M	83.65	84.19	70.31	83.30
N	274	170	19	463
SD	83.92	84.27	54.27	82.98
Social services involvement
Yes	12.9 (52)	21.1 (16)	15 (5)	100 (73)
No	56.6 (233)	39.8 (164)	3.6 (15)	100 (412)
Education level
Less than high school graduate	60.9 (95)	33.3 (52)	5.8 (9)	100 (156)
High school graduate	59.6 (112)	36.7 (69)	3.7 (7)	100 (188)
Greater that high school graduate	54.7 (76)	42.42 (59)	2.9 (4)	100 (139)
Employed
Yes	54.7 (151)	40.2 (111)	5.1 (14)	100 (276)
No	27.6 (131)	14.6 (68)	42.3 (6)	100 (205)
Uses alcohol
Yes	65.7 (190)	31.5 (91)	2.8 (8)	100 (289)
No	48.5 (95)	45.4 (89)	6.1 (12)	100 (196)
Uses illegal drugs
Yes	77.2 (44)	17.5 (10)	5.3 (3)	100 (57)
No	56.1 (239)	39.9 (170)	4.0 (17)	100 (426)
Past criminal history
Yes	47.8 (111)	50.0 (116)	2.2 (5)	100 (232)
No	68.8 (174)	25.3 (64)	5.9 (15)	100 (253)
Past violent behavior
Yes	59.6 (90)	38.4 (58)	2.0 (3)	100 (151)
No	58.4 (195)	36.5 (122)	5.1 (17)	100 (334)
Abused as a child
Yes	67.2 (135)	28.4 (57)	4.4 (9)	100 (201)
No	52.8 (150)	43.3 (123)	3.9 (11)	100 (284)
Victim of sexual abuse
Yes	72.2 (96)	21.8 (29)	6.0 (8)	100 (133)
No	53.7 (189)	42.9 (151)	3.4 (12)	100 (352)
Witnessed domestic violence as a child
Yes	35.1 (135)	14.9 (29)	50 (8)	100 (172)
No	47.9 (150)	48.2 (151)	3.9 (12)	100 (313)
Age
M	33.6	31.52	28.1	32.59
N	283	179	20	482
SD	10.66	10.14	7.24	10.42
Score on PAS
M	71.22	65.66	69.33	69.08
N	285	180	20	485
SD	22.98	19.56	18.10	21.71

Note. PAS = Propensity for Abusiveness Scale; SD = standard deviation.

Conceptually, family of origin exposure to violence was captured by including the variables (1) abused as a child, (2) child victim of sexual abuse, and (3) witnessing domestic violence as a child. Among Caucasian women, 47.4% (N = 135), other women 45% (N = 9) and 31.7% (N = 57) African-American women reported being abused as a child. Victims of childhood sexual abuse were the highest among other women 40.0% (N = 8), followed by Caucasian women 33.7% (N = 97) and 16.1% (N = 29) for African American women. Witnessing acts of domestic violence as a child was the highest among other women 60% (N = 12), followed by Caucasian women 52.6% (N = 150) and 44.4 (N = 80) for African American women.

Social system involvement was conceptualized by including variables on (1) past mental health treatment, (2) past criminal history, (3) past violent behavior, and (4) social services involvement. The highest percentage of previous mental health treatment was found among Caucasian women with 54.7% (N = 156) receiving services followed by other women 40% (N = 8) and 25% (N = 45) of African American, respectively. Past criminal history was very similar for Caucasian women and African American women; 38.9% (N = 111) of Caucasian women reported past criminal history followed by 35.6% (N = 64) African American women and 25% (N = 5) among other women. Past violent behavior was nearly identical for Caucasian women and African American women. Among Caucasian women 31.6% (N = 90), 32.2% (N = 58) of African American women and 15% (N = 3) of other women reported past violent behavior. Social services involvement was the highest among the Other women group with 25% (N = 5) reporting social services involvement, followed by Caucasian women with 18.2% (N = 52) and African American women with 8.9% (N = 16) for involvement with social services. Table 1 contains the remainder of the demographic information on the sample.

An independent t-test indicated that there was no significant difference between program completers and program dropouts on mean propensity for abusiveness scores (Table 1), although numerically higher PAS scores were observed among the client dropouts 70.19 (SD = 23.66) compared to completers 68.68 (SD = 21.00). Consequently, treatment completers and dropouts were aggregated for the remainder of the analyses.

Significant differences were observed between PAS scores and a number of demographic variables. The effect size was calculated for all appropriate analyses. The first significant difference observed was for race. A one-way between-subjects analysis of variance (ANOVA) indicated significant differences between racial groups based on PAS scores (F[2] = 3.65; p < .05; d = 0.26; CI_.95: [0.06, 0.44]). The mean PAS scores for Caucasian women were 71.22 (SD = 22.99), other women 76.20 (SD = 18.10) and 65.66 (SD = 19.56) for African American women. Since the two largest groups were Caucasian women and African American women, effect size was calculated for these two groups. The second significant difference observed was for level of education where an inverse relationship existed between level of education and propensity for abusiveness. Less than high school graduate scored 70.97 (SD = 21.45), high school graduate scored 69.57 (SD = 21.45), and greater than high school scored the lowest on PAS scores 66.79 (SD = 20.69). Since the two largest groups were less than high school and high school graduate, the effect size was calculated for these two groups. A one-way between-subjects ANOVA indicated significant differences between PAS scores based on education level (F[3] = 2.66; p < .05; d = 0.06; CI_.95: [−0.14, 0.28]). The final demographic variable that yielded a significant difference was whether an individual was employed or not. Being employed resulted in a lower level of propensity for abusiveness, whereas being unemployed was associated with higher propensity for abusiveness scores. An independent samples t-test indicated significant differences between propensity for abusiveness scores based on being employed or not. There was a significant difference in PAS scores and not being employed (M = 71.89, SD = 22.86) and being employed (M = 66.83, SD = 20.63) and PAS scores; t = 2.54; df = 479; p < .05; d = 0.23; CI_.95: [0.05, .0.41]). No significant differences were found for age, having children, relationship status, and length of relationship as indicators of propensity for abusiveness.

Based on the observation of significant differences among racial groups on PAS, separate regression models were modeled for Caucasians and African Americans. This step was done to ensure that variability was not distributed across the groups since the other group consisted of only 4% of the overall sample. A total of four multiple regression models (family of origin, social system involvement, substance abuse, and demographics) were run for Caucasian and African American women participating in a BIP, to ensure that variance was correctly attributed the set of predictor variables within each group.

Model 1 (Family of Origin Factors)

Two separate standard multiple regression models were performed to investigate if family of origin factors (abused as a child, child victim of sexual abuse, child witness to domestic violence) resulted in an increased level of propensity for abusiveness among Caucasian women (Model 1a) and African American women (Model 1b).

Model 1a: Family of origin predictors of propensity for abusiveness among Caucasian women

A standard multiple regression analysis was performed to investigate whether family of origin factors (abused as a child, victim of sexual abuse, witnessing domestic violence as a child) resulted in an increased level of propensity for abusiveness among Caucasian women. The R² statistic was statistically significant, F(3, 281) = 25.905, p = .0001, R² adjusted = .208, indicating that potentially 20.8% of the variance in propensity for abusiveness among Caucasian women can be explained by family of origin factors (abused as a child, victim of sexual abuse, and witnessing domestic violence as a child). A summary of the regression coefficients is presented in Table 2 and indicates that being abused as a child and witnessing domestic violence as a child contributed significantly to the prediction of the level of propensity for abusiveness in Model 1a.

Table 2.

Standard Multiple Regression Analysis for Family of Origin Predictors of Propensity for Abusiveness Among Caucasian Women.

	B	β	t	p	Bivariate r	Partial r
Abused as child**	12.730	.277	3.998	.0001	.429	.232
Victim of sexual abuse	5.658	.117	1.928	.055	.305	.114
Witnessing DV as child*	7.855	.171	2.671	.008	.364	.141

Note. DV = domestic violence. n = 281; df = 3.

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

Model 1b: Family of origin predictors of propensity for abusiveness among African American women

A standard multiple regression analysis was performed to investigate if family of origin factors (abused as a child, victim of sexual abuse, and witnessing domestic violence as a child) resulted in an increased level of propensity for abusiveness among African American women. The R² statistic was statistically significant, F(3, 176) = 8.766, p = .0001, R² adjusted = .115, indicating that 11.5% of the variance in propensity for abusiveness among African American women can be potentially explained by family of origin factors (being abused as a child, victim of sexual abuse, and witnessing domestic violence as a child). A summary of the regression coefficients is presented in Table 3 and indicates that being abused as a child contributed significantly to the prediction of the level of propensity for abusiveness in Model 1b.

Table 3.

Standard Multiple Regression Analysis for Family of Origin Predictors of Propensity for Abusiveness Among African American Women.

	B	β	t	p	Bivariate r	Partial r
Abused as child**	12.364	.295	3.551	.0001	.346	.259
Victim of sexual abuse	−.946	−.018	−.238	.812	.106	−.018
Witnessing DV as child	4.450	.116	1.425	.156	.257	.107

Note. n = 176; df = 3.

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

Model 2 (Social System Involvement)

The second model consisted of two separate standard multiple regression models to investigate if social system involvement factors (mental health treatment, past criminal history, past violent behavior, and social services involvement) resulted in an increased level of propensity for abusiveness among Caucasian women (Model 2a) and African American women (Model 2b).

Model 2a: Social system involvement predictors of propensity for abusiveness among Caucasian women

A standard multiple regression analysis was performed to investigate if social system involvement predictors (mental health treatment, past criminal history, past violent behavior, and social services involvement) resulted in an increased level of propensity for abusiveness among Caucasian women. The R² statistic was not statistically significant, F(4, 280) = 1.261, p = .285, R² adjusted = .004, indicating that only 0.04% of the variance in propensity for abusiveness can be explained by social system involvement predictors.

Model 2b: Social system involvement predictors of propensity for abusiveness among African-American women

A standard multiple regression analysis was performed to investigate if social system involvement predictors (mental health treatment, past criminal history, past violent behavior, and social services involvement) resulted in an increased level of propensity for abusiveness among African American women. The R² statistic was not statistically significant, F(4, 175) = 1.956, p = .103, R² adjusted = .021, indicating that only 0.21% of the variance in propensity for abusiveness among African American women can be potentially explained by social system involvement predictors.

Model 3 (Substance Abuse)

The third model consisted of two separate standard multiple regression models to investigate if substance abuse factors (uses alcohol, uses illegal drugs, and alcohol/drug use at incident) resulted in an increased level of propensity for abusiveness among Caucasian women (Model 3a) and African American women (Model 3b).

Model 3a: Substance abuse predictors of propensity for abusiveness among Caucasian women

A standard multiple regression analysis was performed to investigate if substance abuse (uses alcohol, uses illegal drugs, and alcohol/drug use at incident) resulted in an increased level of propensity for abusiveness among Caucasian women. The R² statistic was not statistically significant, F(3, 281) = 1.286, p = .279, R² adjusted = .003, indicating that only 0.003% of the variance in propensity for abusiveness among Caucasian women can be explained by substance abuse.

Model 3b: Substance abuse predictors of propensity for abusiveness among African American women

A standard multiple regression analysis was performed to investigate if substance abuse (uses alcohol, uses illegal drugs, and alcohol/drug use at incident) resulted in an increased level of propensity for abusiveness among African American women. The R² statistic was not statistically significant, F(3, 176) = 1.208, p = .308, R² adjusted = .003, indicating that only 0.003% of the variance in propensity for abusiveness among African American women can be explained by substance abuse.

Model 4 (Demographics)

The fourth model consisted of two separate standard multiple regression models to investigate if demographic variables (age, relationship to victim, length of relationship in months, children, still in relationship with victim, education level, and employed or not) resulted in an increased level of propensity for abusiveness among Caucasian women (Model 4a) and African American women (Model 4b).

Model 4a: Demographic predictors of propensity for abusiveness among Caucasian women

A standard multiple regression analysis was performed to investigate if demographic predictors (age, relationship to victim, length of relationship in months, children, still with victim, education level, and employed or not) resulted in an increased level of propensity for abusiveness among Caucasian women. The R² statistic was statistically significant, F(7, 262) = 3.310 p = .002, R² adjusted = .057, indicating that 5.7% of the variance in propensity for abusiveness among Caucasian women can be explained by demographic factors. A summary of regression coefficients is presented in Table 4. Age, relationship to the victim, and employed or not contributed significantly to the prediction of the level of propensity for abusiveness among Caucasian women in Model 4a.

Table 4.

Standard Multiple Regression Analysis for Demographical Predictors of Propensity for Abusiveness Among Caucasian Women.

	B	β	t	p	Bivariate r	Partial r
Age*	−.413	−.185	−2.872	.004	−.182	−.175
Relationship to victim*	3.556	.175	2.249	.025	.078	.138
Length of relationship in months	.006	.022	.321	.748	−.066	.020
Children	−5.204	−.090	−1.482	.140	−.072	−.091
Still with victim	6.393	.139	1.800	.073	−.036	.111
Education level	.657	.023	.369	.712	−.021	.023
Employed or not*	−7.354	−.168	−2.742	.007	−.140	−.167

Note. n = 262; df = 7.

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

Model 4b: Demographic predictors of propensity for abusiveness among African American women

A standard multiple regression analysis was performed to investigate if demographic predictors (age, relationship to victim, length of relationship in months, children, still with victim, education level, and employed or not) resulted in an increased level of propensity for abusiveness among African American women. The R² statistic was not statistically significant, F(7, 161) = .588 p = .765, R² adjusted = −.017, indicating that only .017% of the variance in propensity for abusiveness among African American women can be explained by demographic factors.

Discussion and Application to Social Work

The results indicate that there were no significant differences between program completers and dropouts on the PAS. This is interesting for two reasons. First, it seems that dropout in this sample was lower than the national average (50%) due to the way in which the data were collected. Because the data were collected at the first two intake sessions in an interview format, many of those likely to dropout may have already done so prior to either the first session or before completing the entire assessment package. This suggests that efforts to retain women in BIPs should focus almost exclusively on factors that prevent initial engagement with the program. Second, given the finding that there were no differences between dropouts and completers in terms of propensity for abusiveness, it may be that structural variables rather than lack of interest in the program explain attrition. For example, it may be that transportation issues (a variable not assessed in this study) prevented women from initially attending and which may have had a long-term impact on their ability to successfully complete the program as well. The same may be true for other variables not investigated here like childcare access while they attend the program. These results suggest that more attention should be paid to getting women successfully transitioned from the court (or referral source) to the program, so that these kinds of issues can be assessed and addressed.

The second research question addressed the investigation of whether psychosocial variables might serve as an accurate predictor of propensity for abusiveness among women arrested for IPV offenses and mandated into treatment. Prior to discussing those results, it is important to contextualize the PAS scores for this sample. Dutton (1995), in his description of the development of the instrument, suggests that nonviolent men score 44.7 (SD = 11.7) and male batterers score 59.2 (SD = 17.1). The scores here were higher than those of Dutton’s battering sample. Also, the scores for this sample were higher than those of a similarly large sample of male batterers drawn from the same location (M = 59.6, SD = 20.32; Buttell & Carney, 2008). It is important to note that the PAS was administered with the BIDR to serve as a check on socially desirable responding. As such, it seems unlikely that these scores are inaccurate. It may be that the women that comprise this sample are even more forthcoming about their violence and endorsement of abusiveness as a conflict resolution strategy, which would be a strength-based interpretation of this situation.

There has been much written about the failure of BIPs to address issue specific to women offenders (for a recent review, see Buttell & Starr, 2013). In brief, as part of the criminalization of domestic violence, many states passed mandatory and pro-arrest policies in calls to the police that involved domestic violence. Virtually, all of the laws were gender neutral with the assumption that those arrested would be predominantly if not exclusively male (Kernsmith & Kernsmith, 2009). However, the arrest data suggest that women constitute 50% of arrests in some jurisdictions (Buttell, Powers, & Wang, 2012). The problem this has created for BIPs is that these women are being sent to their programs as part of a criminal sentence and they are being treated with curriculums designed for male batterers (e.g., the Duluth Program). This problem is compounded by the fact that 45 states have now legislated standards that institutionalize this “one-size-fits-all” treatment approach (Batterer Intervention Services Coalition Michigan, 2014). Unfortunately, a significant part of this standardized program focuses on patriarchy as the causal explanation for men initiating violence in intimate relationships. Obviously, this explanation has questionable relevance to women offenders. The results of this study suggest that these women may benefit from an intervention that focuses on family of origin issues.

The results here suggest that exposure to violence in the family of origin is a strong predictor for developing a propensity for initiating abusiveness in their adult intimate relationships for Caucasian women, explaining 20.8% of the variance among these women. The relationship was less strong for African American women, accounting for only 11.5% of the variance. This difference is hard to reconcile. It may be that African American women are treated differently by the police and judicial systems, which results in their arrest, prosecution, and referral to the BIP for incidents and offenses that would not result in the same process were they Caucasian. Given the location of the BIP in the rural south, this explanation certainly seem possible, which was a point made in a similar discussion of demographic and psychological differences between African American and Caucasian male batterers (Buttell & Carney, 2005).

Conversely, these results suggest that other demographic and background variables are unrelated to explaining the PAS scores. The models for social system involvement and substance abuse were insignificant for both groups and the demographic model was only significant for Caucasian women and explained just 5.7% of the variance. In this way, these women appear to be different from male batterers, where these types of demographic variables predict treatment completion, severity of relationship violence, and recidivism (for an excellent meta-analysis of the BIP attrition studies, see Jewell & Wormith, 2010).

Regardless of the different amount of variance explained in the family of origin models for African American and Caucasian women batterers, it is clear that the constellation of family of origin variables explain a significant amount of the variance in this sample of women’s PAS scores. This is an interesting development and may serve as an alternative way to explain women’s developmental trajectories toward using abuse as a conflict resolution strategy in their adult intimate relationships, as distinguished from men. In the earlier section, the limitations of the dominant feminist model for explaining violence perpetrated by women against their intimate partners were discussed. These results suggest that women in BIPs may benefit from the development of a model that takes a process-oriented approach to their treatment, one in which the focus is on making sense of their childhood exposure to violence and victimization and that also explores the connection between these experiences and their adult behaviors. As social workers and treatment providers, this approach would certainly make more intuitive sense to us than this model, which simply switches the gender pronoun in the Duluth model for perpetrators and asks them to consider their actions from a patriarchal perspective.

Conclusion

There has been a recent accumulation of empirical evidence that suggests that (a) women’s motivations for violence are similar to men (e.g., Straus, 2011); (b) women have a constellation of family of origin issues that help explain their use of violence in their intimate relationships (Ehrensaft, Cohen, & Johnson, 2006; Graham-Kevan, 2009), and (c) that this legislated intervention program, designed for male batterers, does not target these issues successfully (Buttell & Starr, 2013). This information, when viewed in light of the findings from this study, suggests that it may be profitable to create a separate curriculum for women arrested for IPV offenses and mandated into a BIP and that this intervention should be distinguished from the intervention program offered to male batterers. These results suggest that elements of an effective intervention for these women should, at a minimum, target and treat the IPV exposure and victimization issues that occurred when they were children. These results also suggest that it may be important to consider structural barriers to women’s participation in a BIP (e.g., transportation and childcare) and that addressing these issues might increase program retention.

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.

References

Batterer Intervention Services Coalition Michigan. (2014). Other state’s standards. Retrieved from http://www.biscmi.org/other_resources/state_standards.html

Blosnich

J. R.

Bossarte

R. M.

(2009). Comparisons of intimate partner violence among partners in same-sex and opposite-sex relationships in the United States. American Journal of Public Health, 99, 2182–2184.

Buttell

Carney

(2005). Do batterer intervention programs serve African American and Caucasian batterers equally well? An investigation of a 26-week program. Research on Social Work Practice, 15, 19–28.

Buttell

Carney

(2006). A large sample evaluation of a court-mandated batterer intervention program: Investigating differential program effect for African American and Caucasian men. Research on Social Work Practice, 16, 121–131.

Buttell

Carney

(2008). A Large sample investigation of batterer intervention program attrition: Evaluating the impact of state program standards. Research on Social Work Practice, 18, 177–188.

Buttell

Powers

Wang

(2012). Evaluating predictors of program attrition among women mandated into batterer intervention treatment. Research on Social Work Practice, 22, 20–28.

Buttell

Starr

(2013). Lifting the veil: Foundations for a gender-inclusive paradigm of intimate partner violence. In Russell

B. L.

(Ed.), Perceptions of female offenders: How stereotypes and social norms affect criminal justice responses (pp. 117–132). New York, NY: Springer Science + Business Media.

Canon

Buttell

(2015). Illusion of inclusion: The failure of the gender paradigm to account for IPV in LGBT relationships. Partner Abuse, 6, 65–75.

Carney

Buttell

(2006). An evaluation of a court-mandated batterer intervention program: Investigating differential program effect for African American and white women. Research on Social Work Practice, 16, 571–581.

10.

Carney

Buttell

Dutton

(2006). Women who perpetrate intimate partner violence: A review of the literature with recommendations for treatment. Journal of Aggression and Violent Behavior, 12, 108–115.

11.

Carvalho

Lewis

Derlega

Winstead

Viggiano

(2011). Internalized sexual minority stressors and same-sex intimate partner violence. Journal of Family Violence, 26, 501–509.

12.

Cook

Campbell

(1979). Quasi-experimentation: Design & analysis issues for field settings. Chicago, IL: Rand McNally College Publishing.

13.

Dobash

R. E.

Dobash

R. P.

(1979). Violence against wives. New York, NY: The Free Press.

14.

Dobash

R. P.

Dobash

R. E.

Wilson

Daly

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

15.

Dutton

(1995). A scale for measuring propensity for abusiveness. Journal of Family Violence, 10, 203–221.

16.

Dutton

Landholt

Starzonmski

Bodnarchuk

(2001). Validation of the propensity for abusiveness scale in diverse male populations. Journal of Family Violence, 16, 59–75.

17.

Dutton

D. G.

Nicholls

T. L.

(2005). The gender paradigm in domestic violence research and theory: Part 1—The conflict of theory and data. Aggression and Violent Behavior, 10, 680–714.

18.

Eckhardt

C. I.

Murphy

C. M.

Whitaker

D. J.

Sprunger

Dykstra

Woodard

(2013). The effectiveness of intervention programs for perpetrators and victims of intimate partner violence. Partner Abuse, 4, 196–231.

19.

Ehrensaft

M. K.

Cohen

Johnson

J. G.

(2006). Development of personality disorder symptoms and the risk for partner violence. Journal of Abnormal Psychology, 115, 474–483.

20.

Felson

R. B.

(2006). Is violence against women about women or about violence? Contexts 5, 21–25.

21.

Gover

A. R.

Brank

E. M.

MacDonald

J. M.

(2007). A specialized domestic violence court in South Carolina: An example of procedural justice for victims and defendants. Violence Against Women, 13, 603–626.

22.

Graham-Kevan

(2009). The psychology of women's partner violence: Characteristics and cautions. Journal of Aggression, Maltreatment & Trauma, 18, 587–603.

23.

Jewell

L. M.

Wormith

J. S.

(2010). Variables associated with attrition from domestic violence treatment programs targeting male batterers: A meta-analysis. Criminal Justice and Behavior, 37, 1086–1113.

24.

Kernsmith

(2009). Treating female perpetrators: State standards for batterer intervention services. Social Work, 54, 341–349.

25.

Kroner

D. G.

Weeks

J. R.

(1996). Balanced inventory of desirable responding: Factor structure, reliability, and validity with an offender sample. Personality and Individual Differences, 21, 323–333.

26.

Langhinrichsen-Rohling

(2010). Controversies involving gender and intimate partner violence in the United States. Sex Roles, 62, 79–193.

27.

Messinger

A. M.

(2011). Invisible victims: Same-sex IPV in the national violence against women survey. Journal of Interpersonal Violence, 26, 2228–2243.

28.

Paulhus

D. L.

(1984). Two component models of social desirability responding. Journal of Personality and Social Psychology, 46, 598–609.

29.

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.

30.

Winstok

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