Prev Sci DOI 10.1007/s11121-012-0332-z

Centrality of Control-Seeking in Men’s Intimate Partner Violence Perpetration M. Pippin Whitaker

# Society for Prevention Research 2013

Abstract Two million women and one million men experience rape, stalking, or physical assault by a current or former romantic partner each year in the U.S. Not only do women report twice the incidents, but intimate partner violence (IPV) that women experience is typically more severe. Explanations for IPV gender asymmetry include male dominance attitudes, hostile sexism, and men’s control-seeking. There are gaps in our knowledge of how attitudes and control-seeking co-relate to influence IPV. This study demonstrates a mediation analysis to investigate these relationships. Data were from a crosssectional online survey of male undergraduate students from a public Southeastern university. The survey measured attitudes of male dominance and hostile sexism, desire for control, and IPV perpetration. After including age and academic level in the model, male dominance remained a significant predictor of likelihood of physical IPV (OR01.16, p0.004) but not psychological IPV. The addition of control-seeking (physical OR01.65, p<.001) mediated the influence of male dominance on the likelihood of physical IPV perpetration (OR0 1.018, p0.753). Hostile sexism was a significant predictor of psychological and physical IPV (psychological IPV OR0 1.31, p<.001; physical IPV OR01.54, p<.001), over and above age and academic level. The addition of controlseeking (psychological IPV OR01.27, p<.001; physical OR01.53, p<.001) partially mediated the influence of hostile sexism on IPV (psychological IPV OR01.21, p0.001; physical OR01.34, p<.001). Results suggest control-seeking mediates the relationship between male dominance and physical IPV and partially mediates the relationship between hostile sexism and IPV. Practical implications for IPV prevention programs and theoretical implications are discussed. M. P. Whitaker (*) College of Social Work, DeSaussure College, University of South Carolina, Columbia, SC 29208, USA e-mail: [email protected]

Keywords Intimate partner violence . Prevention . Control . Sexism . Mediation Intimate partner violence (IPV) is a pervasive social problem with many plausible causes. Research studies seeking to understand why individuals perpetrate IPV often take a genderbased approach. This is valid in part because research on prevalence shows women are twice as likely as men to be victims of “rape, physical assault, and stalking perpetrated by current and former dates, spouses, and cohabiting partners” (Tjaden and Thoennes 2000, p. 5). In contrast, IPV research that includes psychological abuse and measures such as the Conflict Tactics Scale suggests women perpetrate IPV nearly as often as men (Straus 2008). Regardless of contentious differences in rates due to sample and measurement influences, there is greater agreement in the literature regarding the gendered consequences of IPV. While both women and men perpetrate IPV, evidence suggests the majority of IPV with severe health consequences and persisting over an extended period is perpetrated by men against women (Dobash and Dobash 2004; Johnson 2006; Johnson and Ferraro 2000). Among the most severe consequences, intimate partner homicide portrays a stark gender difference: According to the Bureau of Justice Statistics, men are 285 times as likely as women to murder their intimate partners (Hamby 2005). Sexism and IPV Given evidence of a gendered nature of IPV, the literature explores many plausible explanations for the nature and prevalence of men’s IPV perpetration. Prominent among these explanations are men’s desire for dominance over women and sexist attitudes. Male dominance definitions vary. At one end of the conceptual spectrum, male dominance is viewed as individual attitudes supporting men’s superiority over women and men’s preference that men be in positions of authority (Levant and Richmond 2007); on the other end of the spectrum, it is behaviors such

Prev Sci

as holding authority, being restrictive, and disparagement of a partner (Straus 2008). Straus found that dominance in male or female partners was equally associated with increased likelihood of abuse (Straus 2008); however, it is important to note that his concept of dominance pertains to dominating relationship behaviors versus beliefs in a maledominant social ideal. If the focus is on attitudes and beliefs about male dominance, research demonstrates consistent but moderate associations between male dominance and IPV perpetration across a range of studies (Harway and O'Neil 1999; Krug et al. 2002). Although traditional (first-generation) concepts of sexism, e.g. male dominance, remain empirically linked to IPV perpetration, more recent evidence suggests that alternate (secondgeneration) forms of sexism are increasingly relevant, at least among recent generations (Forbes et al. 2004). Forbes et al. (2004) contrasted several first and second-generation sexism measures in verbal IPV perpetration among 264 college freshmen (there was insufficient data on physical IPV). Forbes et al. found that after accounting for hostile sexism (second-generation), only one first-generation sexism was significantly associated with likelihood of perpetrating verbal IPV. Glick and Fiske’s (1996) hostile sexism scale is part of two dimensions of sexism: benevolent and hostile sexism. Benevolent sexism is associated with complementary gender differentiation and protective paternalism, whereas hostile sexism is associated with disdain for women’s progress toward equality with men (Glick and Fiske 1996). Recent work provides evidence that hostile sexism may originate via male gender role stress and ascription to toughness norms (Gallagher and Parrott 2011). However, hostile sexism differentially predicts IPV and represents a helpful discriminating concept (Forbes et al. 2004). Importance of Control-Seeking to IPV Beyond sexist attitudes, long-standing argumentation in the IPV literature contends that desire for control is central in motivating IPV perpetration. Dobash and Dobash (1992) argued that men’s perceived entitlement to control women and maintain authority, commonly referred to as “desire for power and control,” fuels IPV perpetration. Several studies corroborate that men’s control of the relationship predicts physical abuse of women (Kershaw et al. 2006; Raj et al. 2004; Tjaden and Thoennes 2000; Wingood and DiClemente 2000; Yang and Xia 2006). Results from the National Violence Against Women Survey show that women with emotionally or physically controlling partners are more likely to experience all forms of IPV (Tjaden and Thoennes 2000), and Raj et al. (2004) found that women whose (men) partners had high levels of relationship control were more likely to experience physical and sexual abuse. Others found greater health consequences for women in controlling relationships versus those in more egalitarian relationships (Lawrence et al. 1997; Pulerwitz et al. 2002; Wingood and DiClemente 2000).

Control-Seeking and Gender A number of studies examine whether control-seeking and its influence on IPV is gendered (Felson and Outlaw 2007; Stets and Hammons 2002; Straus 2008; Tanha et al. 2010). Stets and Hammons (2002) examined how a spouse’s controlling behaviors affect the respondent’s marital commitment. They found that wives reported using slightly more controlling behaviors and that a spouse’s control-seeking increased women’s marital commitment but decreased men’s. The results suggest that control-seeking is not unique to men but that the meaning of controlling behaviors may differ between men and women: a spouse’s control-seeking increased wives’ marital commitment, but the inverse occurred for men. In line with this, qualitative work on control-seeking in relationships suggests that men more often than women seek control to feel powerful in relationships (Hamberger et al. 1997; Harvey et al. 2002). This supports that control-seeking is not unique to men but that it has gendered meanings. Felson and Outlaw (2007) asked whether control-seeking behaviors were more closely associated with men’s IPV than women’s. Using data from the Survey of Violence and Threats of Violence Against Women and Men, Felson and Outlaw found that husbands and wives both had low and comparable rates of controlling behaviors, although there were gender differences in methods of control. Felson and Outlaw also found that controlling spouses were more likely to use violence, regardless of gender. Overall, they conclude that control-seeking does not mediate gender differences in IPV perpetration. Similarly, Tanha et al. (2010) examine control-seeking as the motivating factor in IPV. Tanha et al. hypothesized that physical IPV is used when other tactics to achieve control fail. Using a sample of 762 divorcing couples, the researchers found that victim reports of spouse’s coercive control-seeking were significantly linked to perpetration of IPV for both men and women. Again, the results show a different pattern of modes of violence related to control-seeking for men versus women. Specifically, Tanha et al. found that men were more often reported to perpetrate psychological IPV, sexual IPV, and severe physical IPV, while there were no gender differences in less severe physical IPV such as pushing and shoving. Notably, Straus’s dominance concept is more closely aligned with control than attitudes supporting a male dominant ideal. Straus’s dominance items (Straus et al. 1999) are very close to the control items used by Stets and Hammons (2002) and Felson and Outlaw (2007). For example, “I have a right to know everything my partner does, I insist on knowing where my partner is at all times” versus “I keep tabs on my spouse, I keep my spouse from doing things I do not approve” (Stets and Hammons 2002) or “Tries to limit your contact with family or friends, Insists on knowing who you are with at all times” (Felson and Outlaw 2007). Thus, it is applicable that Straus found that dominance, like control,

Prev Sci

was associated with both women’s and men’s IPV perpetration (Straus 2008). Dominance, however, had greater discriminant ability for male’s severe violence than for females (Straus 2008). The above research evidence supports a relationship between sexist attitudes and IPV and between control-seeking and IPV, as well as gendered meanings and practices of control-seeking. Given a gendered nature of IPV and debates about gender symmetry and asymmetry, the interplay of sexism and control-seeking on IPV appears relevant. Yet, the relative importance of control-seeking and sexism is untested. There is a gap in our knowledge of how attitudes and controlseeking co-relate to influence IPV. In particular, the potential mediating role of control-seeking in first-generation versus second-generation sexism is yet unexplored. If controlseeking is a central mediator in the relationship between sexism and men’s perpetration, this is vital knowledge for IPV treatment and prevention programs. IPV prevention programs typically focus on first-generation or second-generation concepts of sexism more than control-seeking (Offenhauer and Buchalter 2011). IPV treatment programs could also benefit from a better understanding of the role of controlseeking in the relationship between different forms of sexism and IPV. Further, the potential mediating role of controlseeking may contribute to our understanding of gender symmetry and asymmetry in the IPV literature.

Research Question and Hypotheses Based on the available evidence and literature, this study proposes a conceptual perspective that situates controlseeking as a link between sexist attitudes and IPV perpetration. This presupposes that attitudes about male dominance and hostile sexism precede control-seeking in relationships. The study asks whether control-seeking mediates the relationship between attitudes about male dominance and men’s IPV perpetration and the relationship between attitudes about hostile sexism and men’s IPV perpetration. This study focuses on physical and psychological IPV; other factors may explain sexual violence or stalking (Krug et al. 2002; Tjaden and Thoennes 2000). Drawing from this conceptual basis, this study tests two hypotheses. The first is that once respondent level of control-seeking is taken into account, there will no longer be an increased likelihood of psychological or physical IPV perpetration associated with higher levels of male dominance. The second is that once respondent level of controlseeking is taken into account, there will no longer be an increased likelihood of psychological or physical IPV perpetration associated with higher levels of hostile sexism. Figure 1 illustrates both hypotheses in one figure.

(a) Male Dominance

(c) ControlSeeking

Hostile Sexism

(e) IPV

(d) (b)

Fig. 1 Hypothesized mediation model

Method This study used data from a cross-sectional survey in the fall of 2009 of adult, in-state, degree-seeking undergraduate students at a large state-funded university in the Southeastern U.S. The present analysis includes undergraduate males aged 18 to 26 years to examine relationships among sexist attitudes, control-seeking, and IPV perpetration. In addition to attitude and IPV, the survey included relationship status, age, and academic level. College students are a suitable population among which to examine IPV because dating aggression is prevalent among students, with estimates of physical IPV ranging from 14 % to 42 % among U.S. universities (Straus 2004). After obtaining institutional review board approval, an invitation letter, informed consent, and link with an embedded unique identifier (ID) to take the survey were sent via campus email after the add/drop deadline, with three weekly (Sunday) reminders. The invitations and reminder emails included first names in the salutation. Invitation letters informed students that they were entered in a chance to win a $50 gift card with immediate notice of results upon completion.

Measures Male Dominance Attitudes about male dominance were measured with the dominance subscale of the Male Role Norms Inventory-Revised (MRNI-R) (Levant et al. 2007). Example items from the scale are “men should be the leader in any group” and “a man should provide the discipline in the family.” The MRNI-R measures the degree to which respondents agree with traditional male role norms. Although the MRNI has been used extensively over the past 15 years, it has moderate reliabilities (Levant and Richmond 2007). The revised scale has better discriminant construct validity and higher reliability (α0.96) (Levant et al. 2007). The MRNI-R dominance subscale is a seven-item scale with Chronbach’s α reported as .84. This study uses a six-point Likert scale (strongly agree, agree, slightly agree, slightly

Prev Sci

disagree, disagree, strongly disagree), which was coded 0 through 5 for all items. While the scale authors use a sevenpoint Likert scale, the present research omits the “no opinion” category to conform to the response format of the other two attitude scales in this survey in order to avoid response options as a confounding effect in data analysis. The male dominance subscale score was computed as the average of items, and was a continuous variable ranging from 0 to 5, with 0 indicating strong disagreement for male dominance and 5 indicating the most agreement with male dominance. Hostile Sexism Hostile sexism was measured with the 11item Hostile Sexism scale, which assesses respondent attitudes toward women’s authority and complaints (Glick and Fiske 1996). Example items are “women seek special favors under guise of equality” and “once a man commits, she puts him on a tight leash.” The reported scale reliabilities ranged from α0.80 to α0.92 across six studies (Glick and Fiske 1996) with strong evidence of concurrent validity with related measures such as the Hostility Toward Women Scale (HTW) (Forbes et al. 2004). HTW is similar, but focuses on interpersonal hostility as opposed to hostility toward women in general. The Hostile Sexism scale has six response options (strongly agree to strongly disagree) scored 0 to 5, with three reverse-scored (Glick and Fiske 1996). The hostile sexism score was an average of the items, resulting in a continuous variable ranging from 0 to 5, with 0 being strong disagreement with hostile sexism to 5 being the most agreement with hostile sexism. Control-Seeking Control-seeking was measured with five items from the ten-item Dominance Scale restrictive subscale (Hamby 1996). The subscale had α of .73. Five items tap insecurity (e.g., “I tend to be jealous”) and Hamby (1996) reported that they had factor loadings under 0.50 and removal had negligible effects on reliability. The remaining five items focused on desire for control (e.g., “I insist on knowing where my partner is at all times,” “It would make me mad if my partner did something I said not to do”) were retained. The items had six response options (strongly agree to strongly disagree). The control-seeking score was an average of five items, a continuous variable ranging 0 to 5, with 0 indicating low and 5 indicating high control-seeking. Relationship Status Respondents were asked whether they had at least one relationship partner in the past year (yes/no). Respondents were then asked to indicate whether “intimate/ dating relationship partner(s) in the past year were:” male (yes/no), female (yes/no). Respondents with only female partners in the past year were retained in this analysis. IPV Perpetration IPV perpetration was operationalized as reported psychological and physical abuse perpetrated

against an intimate partner in the past year. This was measured with the physical assault (e.g., I pushed or shoved my partner) and psychological aggression (e.g., I shouted or yelled at my partner) subscales of the Revised Conflict Tactics Scale (CTS2) for perpetration. Respectively, these subscales have twelve and eight items and Cronbach’s α of .86 and .79 (Straus et al. 1996). The CTS2 dichotomous response option (yes/no) was selected for simplicity. The CTS2 has comparable rates for the physical and psychological subscales compared to the seven chronicity response options (Hamby et al. 2006). Hamby et al. (2006) recommend mode scoring for the dichotomous subscales in which each subscale is “true” if any tactic on the scale was endorsed. Thus, a yes on any item results in a 1 (versus 0) for the subscale. IPV was a multinomial variable coded 0 for no physical or psychological IPV, 1 for psychological IPV only, and 2 for physical and psychological IPV. Only 28 respondents (1 %) indicated physical IPV with no psychological IPV. Those who indicated only physical IPV perpetration were coded 2 with physical and psychological IPN in order to conduct analyses of psychological versus physical IPV. Age, Academic Level, and Gender The survey included respondent age in years, academic level (freshmen00, sophomore01, junior02, senior03), and sex/gender (M/F). The sex labels were designed so that individuals could indicate both if desired. Twenty-one respondents indicated both, and two participants emailed the researcher independently to note appreciation for being able to select both. Respondents ages 18–26 who selected M were retained in this analysis.

Analyses Analyses were conducted using Stata/SE version 12 and included univariate descriptive and screening analyses, psychometric analyses, bivariate correlations, and multinomial regressions. Basic psychometric analyses including reliability and the standard error of measurement (SEM) were performed for the male dominance, hostile sexism, controlseeking, and IPV scales. The four-stage process of Baron and Kenny (1986) was used to test mediation. Baron and Kenny recommend a series of regression models and assessment of coefficient change to test mediation. The three models must support three relationships: first, the mediator is associated with the independent variable (IV); second, the IV is associated with the dependent variable (DV); and third, the mediator is associated with the DV, with the IV in the model. Once these three relationships are supported, the final test is to determine if the IV coefficient decreases significantly with the mediator in the model. Essentially, this process tests whether the addition of the mediator explains all or part of the relationship between the

Prev Sci

IV and DV. Because multivariate models are improved by including potential sources of spuriousness (Tabachnick and Fidell 2007), Baron and Kenny’s third step was adapted to include age and academic level, two widely cited risk factors for IPV perpetration (Stith et al. 2004). These variables could reveal spurious mediation if, for example, younger or freshmen respondents were more likely to perpetrate IPV and more likely to score high both IVs and mediators. Thus, the basic process outlined by Baron and Kenny was elaborated to include age and academic level in the third equation. In addition, correlation was substituted for the first regression for simplicity. Although bivariate correlations (Table 1) support all required model paths, correlation is not a sufficient substitute for the second or third of Baron and Kenny’s equations because the dependent variable is multinomial. Figure 1 illustrates the paths that must be significant to support mediation, with both models combined into one graphic. Steps used to test the first hypothesis are: 1) correlation tests whether male dominance is associated with control-seeking (significant path c); 2) multinomial regression tests whether male dominance predicts the likelihood of IPV perpetration with academic level and age in the model (significant path a); 3) multinomial regression tests whether control-seeking predicts the likelihood of IPV perpetration with male dominance, academic level, and age in the model (significant path e); and 4) observed change in the male dominance coefficient from the first to second models determines support for mediation or partial mediation. Mediation is supported if the male dominance coefficient goes from being significant in the first model to non-significant in the second. If the male dominance coefficient merely decreases, then a Wald Chisquare test determines whether it decreased significantly. If the coefficient decreased significantly, then partial mediation is supported. Both forms of mediation are illustrated in Fig. 1 using a dashed line for path a. Finally, the process repeats for hypothesis 2, substituting hostile sexism for male dominance.

Results The survey response rate for males was 25 %. Race was not obtained. However, the male undergraduate student population was 72 % white, which suggests that the sample was likely mostly white. Respondents were 28 % freshmen, 17 % sophomore, 29 % junior, and 26 % senior, with a mean age of 20 (SD02.45). Respondent age was close to the population (μ020, σ03.5), but academic level contained 9 % more freshmen, 3 % fewer juniors, and 5 % fewer seniors than population (X2 0146.34, df03; p<.001). The larger response from freshmen may be due to course incentives for research participation (freshmen research classes provide extra credit to students who take part in research). Eighty-six percent of male respondents had an intimate partner in the past year, and this proportion was evenly distributed among class levels (χ2 00.94, df03; p0.817). This study examined students aged 18 to 26 years with a relationship partner in the past year. Respondent ages dropped off dramatically after 26 years, from 27 respondents aged 26 years to 7 aged 27 years, with decreasing frequencies continuing for higher ages. Because age is an important risk factor in IPV, including older respondents would require adequate sampling. Because older ages were under-represented and the frequency clearly dropped off at 27, those respondents over 26 were excluded. After excluding 39 respondents 27 years and over, 48 respondents missing the second survey page (see below), and responses with improbable response patterns, there were 2421 respondents in the sample. Improbable responses were identified as the same response on all hostile sexism items, including the three reverse-scored items, along with uniform responses on control-seeking and male dominance items. Sexist Attitude Attitudes supporting male dominance were less prevalent than hostile sexism and control-seeking. The mean male dominance response was 1.76 (SD01.1), indicating an average between disagree and slightly disagree

Table 1 Bivariate correlations of model variables (N02421) Variable (Abbreviation)

psyIPV

phyIPV

MD

HS

CS

AL

Psychological IPV (psyIPV) Physical IPV (phyIPV) Male Dominance (MD) Hostile Sexism (HS) Control-seeking (CS) Academic Level (AL) Agea

– .343*** .044* .115*** .147*** .065** .073***

– .050* .090*** .117*** .001 −.005

– .607*** .355*** −.071** −.049*

– .334*** −.094** −.079***

– −.079** −.067**

– .799***

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

The residual of age regressed on academic level was used in multivariate analyses

Prev Sci

with male dominance statements. The hostile sexism mean was notably higher at 2.46 (SD00.8), indicating an average between slightly disagree and slightly agree. Controlseeking was nearly as high, with a mean of 2.39 (SD0 0.9). There were no significant departures from normality (skew00.174–0.336; kurtosis00.512–0.01). The attitude questions about male dominance, hostile sexism, and control-seeking were located on the first page of the survey and had at least a 99.8 % completion rate. Analysis of missing data patterns revealed a non-monotone (item-response) missing pattern among the attitude measures. The attitude scales were computed via a single imputation of missing items using ordinary least squares regression and a randomly selected residual added to the predicted value for missing item cases. Each full scale was used to predict missing items. When multiple items were missing, pared down models were used with at least two out of four predictors for control-seeking, seven out of ten for hostile sexism, and four out of six for male dominance. IPV Perpetration Fifty-five percent of male respondents with past-year relationships reported perpetrating some form of IPV within the past year, with 16 % reporting physical IPV and 53 % reporting psychological IPV. The physical IPV rate was within the range (14–42 %) identified by Straus (2004). Thirty-nine percent reported perpetrating only psychological IPV. The IPV perpetration items were on the second survey page and had a 98.5 % completion rate. Investigation of missing data patterns revealed a monotone missing pattern. The discrepant completion rates may have been due to difficulties in progressing between survey pages. Eleven respondents emailed the researcher during data collection regarding difficulty advancing pages. Because the missing rate was moderate and likely due primarily to Web survey platform failure, the missing responses were treated as missing at random and were excluded from analyses. Scale Reliabilities Reliabilities for the attitude and IPV measures support the present use. Scale reliabilities were within acceptable ranges for research purposes. Cronbach’s α was .92 for male dominance (SEM0.32), .86 for hostile sexism (SEM0.31), .78 for control-seeking (SEM0.41), .80 for physical IPV, and .73 for psychological IPV. The SEM is the scale standard deviation times the square root of 1 minus α. The SEM should be 5 % or less of the scale range, although this recommendation is based on scales with range 100. With relatively small ranges, the adequacy of the 5 % rule is ambiguous, and it does not apply to dichotomous scales. Results suggest that male dominance and hostile sexism scores varied moderately in proportion to the scale range. The SEM was elevated for control-seeking, indicating a larger than desired spread in scores.

Tests of Hypotheses The attitude measures were mean-centered so that predictor effects could be estimated compared to average levels. All correlations among independent variables were weak (Table 1) except for that between male dominance and hostile sexism (r0.61, p<.001), which were entered in separate models, and between age and academic level. Because age and academic level were highly correlated (r0.80, p<.001), age was regressed on academic level and the residual, which was uncorrelated with academic level (r 0.00; p 0.999), was retained in multivariate analyses. This allowed analyses to assess the influence of age after taking into account academic level. However, bivariate correlations provide limited tests of multicollinearity (Tabachnick and Fidell 2007). Squared multiple correlations (SMC) were obtained via R-squares with each independent variable regressed on the other predictors without the DV in the model. The largest SMCs were .36 and .38 for male dominance and hostile sexism, respectively. These SMCs values were well below the .80 multicollinearity cutoff (Menard 2001). The Box Tidwell transformation tested for linearity in the logit and logarithm product terms were nonsignificant (p0.083–.679). To identify outliers, deviance scores from full-model logistic regressions were computed for psychological IPV and physical IPV. Removing outliers with deviance scores above 4 had a negligible impact on model fit and did not change parameter interpretations. Thus, because poorly fit cases were not obviously or statistically outstanding, all 2421 observations were retained. Hypothesis 1 The results produced mixed support for the hypothesis that control-seeking mediates the relationship between male dominance and IPV perpetration. The results support that control-seeking mediates the relationship between male dominance and the likelihood that someone will perpetrate physical IPV. Mediation with psychological IPV was not tested because there was no significant relationship between male dominance and psychological IPV. A four-stage process tested the hypothesis. First, controlseeking was correlated with male dominance to establish a significant association for Fig. 1, path c (r0.36, p<.001). Second, IPV was regressed on male dominance, academic level, and the age residual via forced entry (Table 2). The model was significant (χ2 025.20, df06, N02421, p<.001), but with no significant relationship between male dominance and psychological IPV. Still, for every one-unit increase in male dominance above the mean, respondents were 16 % more likely to perpetrate physical IPV (OR0 1.16, p0.004) versus no IPV. Academic level and age increased risk of psychological IPV but were not significant predictors of physical IPV. The regression establishes a link between male dominance and physical but not psychological IPV (Fig. 1, path a).

Prev Sci Table 2 Multinomial regression test for mediation of male dominance and IPV (N02421)

IPV Type

Model Item

*Reference is no IPV Psychological IPV

Physical IPV

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

Intercept Male Dominance Control-Seeking Academic Level Age Residual Intercept Male Dominance Control-Seeking Academic Level Age Residual

In the third stage, IPV was regressed on control-seeking, male dominance, academic level and the age residual via forced entry (Table 2). The model fit improved after adding control-seeking (χ2 082.94, df08, N02421, p<.001) to the model. This and the significant coefficients for controlseeking support a relationship between control-seeking and IPV perpetration (Fig. 1, path e). For every one-unit increase in control-seeking above the mean, respondents were 36 % more likely to perpetrate psychological IPV (OR 01.36, p<.001) versus no IPV, taking into account male dominance, academic level, and age. For every one-unit increase in control-seeking, respondents were 65 % more likely to perpetrate physical IPV (OR01.65, p<.001) versus no IPV, taking into account the other variables. Given significant paths c, a, and e, the fourth stage compares the male dominance coefficient between models. After adding control-seeking to the model, male dominance no longer had a significant effect on likelihood of physical IPV perpetration (OR01.02, p0.753); this supports mediation of the influence on physical IPV.

Table 3 Multinomial regression test for mediation of hostile sexism and IPV (N02421)

IPV Type

Physical IPV

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

Mediation Model

Coefficient

OR

Coefficient

OR

−0.50*** 0.06 – 0.13** 0.10* −1.24*** .15** – 0.07 0.02

0.61 1.07 – 1.14 1.10 0.29 1.16 – 1.07 1.03

−0.53*** −0.02 0.30*** 0.15*** 0.10* −1.32*** 0.02 0.50*** 0.09 0.03

0.60 0.98 1.36 1.16 1.11 .27 1.02 1.65 1.10 1.03

Hypothesis 2 The results support that control-seeking mediates the relationship between hostile sexism and IPV perpetration. As with hypothesis 1, a four-stage process was used. First, control-seeking was correlated with hostile sexism to establish an association for Fig. 1, path d (r0.33, p<.001). Second, IPV was regressed on hostile sexism, academic level, and age residual (Table 3). The model was significant (χ2 060.55, df06, N02421, p<.001) and a one-unit increase in hostile sexism above the mean was associated with being 31 % more likely to perpetrate psychological IPV than no IPV (OR01.31, p<.001), with academic level and the age residual in the model. Further, for every one-unit hostile sexism increase, respondents were 54 % more likely to perpetrate physical IPV (OR0 1.54, p<.001) versus no IPV. Academic level and age increased psychological IPV risk but were not significant predictors of physical IPV. The results establish a link between hostile sexism and physical and psychological IPV (Fig. 1, path b).

Model Item

*Reference is no IPV Psychological IPV

Base Model

Base Model

Mediation Model

Coefficient

OR

Coefficient

OR

Intercept Hostile Sexism Control-Seeking Academic Level Age Residual Intercept

−0.54*** 0.29*** – 0.17*** 0.10* −1.30***

0.59 1.31 – 1.16 1.11 0.27

−0.55*** 0.188** 0.241*** 0.156*** 0.105* −1.36***

0.58 1.21 1.27 1.17 1.11 0.26

Hostile Sexism Control-Seeking Academic Level Age Residual

0.43*** – 0.09 0.03

1.54 – 1.10 1.03

0.29*** 0.42*** 0.11* 0.03

1.34 1.53 1.11 1.03

Prev Sci

In the third stage, IPV was regressed on control-seeking, hostile sexism, and other potential explanatory variables. Again, the model fit improved after adding control-seeking (χ2 0100.85, df08, N02421, p<.001). This and the significant coefficient for control-seeking for the psychological and physical IPV models support a relationship between control-seeking and IPV perpetration (Fig. 1, path e). For every one-unit increase in control-seeking, respondents were 27 % more likely to perpetrate psychological IPV (OR01.27, p<.001) versus no IPV, with the other variables in the model. Further, for every one-unit increase in controlseeking above the mean, respondents were 53 % more likely to perpetrate physical IPV (OR01.53, p<.001) versus no IPV, with the other variables in the model. Given support for paths c, a, and e, the fourth stage compares the hostile sexism coefficients from each model. After adding control-seeking to the model, there was a significant reduction in the coefficients for hostile sexism in predicting both psychological IPV (ΔWald χ2 013.19, df01, N02421, p<.001) and physical IPV (ΔWald χ2019.93, df01, N02421, p<.001), which supports partial mediation of effects on both types of IPV.

Discussion This study examined whether control-seeking mediates the relationships between two types of sexism and men’s IPV perpetration. Results support that control-seeking mediates the effect of male dominance on likelihood of physical IPV perpetration. There was no significant relationship between male dominance and likelihood of psychological IPV, thus no test for mediation. The non-significant effect of male dominance on psychological IPV perpetration is incongruent with findings from Forbes et al. (2004) and others. This may have resulted from use of a newer measure for male dominance. The present study uses the Male Role Norms Inventory-Revised (Levant et al. 2007) versus Forbes et al. who used the Old-fashioned Sexism Scale (Swim et al. 1995). Levant et al.’s inventory was developed more recently and was expected to have a higher rate in the sample versus Swim et al.’s scale, based on the argument that gender norms shift over time. However, the newer inventory has not been studied extensively in IPV. The low rate of attitudes supporting male dominance in this study may have produced insufficient power to find significant effects. Finally, this study examines psychological and physical IPV, whereas many previous studies have looked exclusively at one form of IPV or the other. This study’s focus on psychological and physical IPV may influence findings. It remains to be seen whether these results will be sustained in future studies with similar populations. The results also support that control-seeking partially mediates the influence of hostile sexism on the likelihood

of physical and psychological IPV perpetration. Notably, hostile sexism had a larger influence on physical IPV than did male dominance (Tables 2 and 3, base models), which is in line with previous studies that found hostile sexism to be a stronger predictor of IPV perpetration than first-generation sexisms, such as male dominance (Forbes et al. 2004). Limitations Before proceeding to describe the implications for theory and practice, this section reviews the limitations, their relevance to study findings, and remedies where applicable. Inferences based on self-report survey data are influenced by three threats to the validity of responses: false respondents, lying, and social desirability bias. The threat of false respondents occurs when the intended recipient is not the one completing the survey. The risk of false respondents was minimized by sending each respondent a unique link with an embedded unique identifier in an email. Identifiers were sixteen-digits, non-sequential, complex alphanumeric sequences with a low probability of accurately guessing a viable ID (there were 27,000 valid IDs in the full study sample but 7,607,872,110 potential sequences). The potential effect of lies, such as respondents’ giving the same answer for all questions, was mitigated by removing responses that were all the same including reverse-scored items. Social desirability bias results from the respondents’ completing their surveys with inaccurate responses that they believe are more socially desirable than the truth. If social desirability bias influenced the results, a likely result would have been a reduced number of respondents who truthfully reported IPV. This does not appear to be a substantial problem in this data because the prevalence rate in the sample was within ranges obtained from similar universities using similar measures (Straus 2004). Another limitation arises from the response rate (25 %). It is possible that the sample differed from other male students in important ways. This may have biased results. Indeed, male dominance had a low rate in this sample compared to a similar study (Levant et al. 2007). Thus, this sample may under-represent undergraduate males with higher levels of attitudes supporting male dominance. This study uses a cross-sectional college student sample, which carries two additional potential limitations in this study. First, the sample may not be representative of the general population. Although the study examines relationships among concepts versus prevalence, it is possible that relationships might be different for individuals who are underrepresented in this sample, for example those with lower high school academic achievement or racial minorities. In addition, Web-based surveys have been criticized for nonresponse error due to underrepresenting individuals without access to technology. This was unlikely to exclude students in the sample frame because the university required undergraduates to have computers and Internet access. The

Prev Sci

other limitation is that the cross-sectional sample provides a limited test of mediation. Mediation implies causality, but it is not possible to determine a time sequence with this data. The data are congruent with mediation, but it is impossible to determine whether sexist attitudes preceded controlseeking in time or, likewise, whether sexist attitudes and control-seeking preceded IPV. Conceptual Implications Hostile sexism has been suggested as a discriminating concept, compared to other attitudes, in understanding motivations for IPV as well as in better predicting perpetration. The findings above support controlseeking as a further discriminating concept. Overall, control-seeking had the strongest influence on the likelihood of both psychological and physical IPV perpetration, compared to hostile sexism or male dominance. Holding hostile sexism and other variables constant, increases in controlseeking continued to increase the likelihood for psychological and physical IPV perpetration compared to no IPV. That control-seeking mediated male dominance but only partially mediated hostile sexism suggests different underlying motives for IPV. This supports a central focus of the IPV literature, that IPV motivated by control-seeking is a significant and unique form of IPV. Notably, the present study does not explore the victim’s experience. There may be little difference in experiences due to motivation. This study’s results are most relevant to preventing and understanding perpetration. By understanding the underlying motivating factors, including control-seeking, we can better focus knowledge toward ameliorating motivational factors. Control-seeking as a central component in motivations for IPV provides an important link in the ongoing debate in the literature regarding gender symmetry and asymmetry. While control-seeking has been found to predict IPV perpetration for men and women, the mediating role of controlseeking in the gendered relationships between male dominance and hostile sexism and IPV perpetration helps to illuminate gendered factors that may undergird what appears to be gender symmetry in some studies. Also, the pathways to reduce control-seeking among, for example, men in heterosexual relationships may be different from the pathways to reduce control-seeking among others. Thus, an IPV prevention program seeking to reduce control-seeking among men may include sexism as part of the discussion, while addressing control-seeking among women would need to include factors that influence women’s control-seeking. Research Implications The findings show control-seeking is important in IPV perpetration, which underscores the need for further understanding of control-seeking. New theoretical approaches to control, such as Tanha et al. (2010), suggest that perpetrators seek control because they lack qualities such as empathy, interpersonal investment, and

desire for monogamy. The present results add to understanding how attitudes influence desire for control. Future research should explore the influence of Tanha et al.’s qualities together with attitudes on control-seeking and how control-seeking mediates relationships between these factors and IPV perpetration. The focus on control-seeking is also important to the study of victim outcomes. As Stark (2007) has shown through case studies of women survivors, the experience of the perpetrator’s control-seeking behavior is qualitatively different from IPV with other motives. The abusive partner’s control has been consistently linked to more severe and lasting negative victim outcomes, including physical injury, long-term health problems, and mental problems such as post-traumatic stress disorder (Kershaw et al. 2006; Lawrence et al. 1997; Pulerwitz, et al. 2002; Raj, et al. 2004; Tjaden and Thoennes 2000; Wingood and DiClemente 2000; Yang and Xia 2006). In addition, research should examine how the victim’s views of control-seeking, male dominance, and hostile sexism are linked to health outcomes. Finally, research on mutual violence should explore the balance of control-seeking among both partners. Future research should explore gender and sexuality differences in factors that influence control-seeking. Indeed, different influences on control-seeking may be linked to different types of IPV perpetration. Developing knowledge of controlseeking may reveal that different forms of control-seeking are linked to different IPV types and victim outcomes. Clearly, control-seeking is a key concept to be explored and developed to enhance understanding of IPV perpetration. Practice Implications In addition to conceptual development, the findings have practical implications. Overall, IPV prevention programs should consider increased attention to control-seeking. As noted above, most IPV prevention programs focus on sexism more than control-seeking (Offenhauer and Buchalter 2011). To the extent that adding content implies removing something else, adding controlseeking should likely be at the expense of male dominance rather than hostile sexism. This recommendation results from the finding, and consistent findings in the literature, that when traditional sexism is less prevalent, hostile sexism may be more relevant. The level of traditional sexism could be determined with a pre-assessment of the participants or participant population. Further, control-seeking mediated the influence of male dominance and had a larger influence on likelihood of IPV than male dominance. It should be reiterated that this finding does not imply male dominance is irrelevant, but instead suggests it is especially relevant to control-seeking in IPV. Male dominance content may be best situated among control-seeking content, and should only be reduced if necessary to accommodate controlseeking content.

Prev Sci

Thus, efforts to ameliorate control-seeking in IPV prevention or treatment programs should include content on the factors that influence control-seeking, such as empathy, interpersonal investment, desire for monogamy, hostile sexism, and male dominance. For example, male dominance could be moved from a central curriculum component to be enveloped in a component focused on control-seeking. Addressing first-generation sexism in the context of control-seeking helps to focus discussions around why certain attitudes, such as male dominance, are problematic. This may be especially helpful for some men who have difficulty identifying why a role model’s first-generation sexism is problematic. The link between attitudes supporting male dominance and an increase in the desire for control and, in turn, an increase in IPV perpetration provides a concrete, constructive criticism of often cherished maladaptive attitudes. In contrast, hostile sexist attitudes could be shown to be problematic because they both increase controlseeking and increase IPV directly. Although the evidence is premature on what motivates women’s control-seeking, the gendered natures of sexism and IPV suggest that content on control-seeking may need to differ for women and for those in relationships with other men. Overall, the findings promote conceptual development, future research, and new ideas for IPV prevention programs. The results show that control-seeking is a central concept in men’s IPV perpetration. Examining control-seeking separately from control tactics and taking into account the influence of sexism offers a new perspective on perpetrator typology. Practically, the findings provide evidence that controlseeking is at least as important a concept in discriminating IPV perpetration as hostile sexism. While further research is implicated, the findings are promising and prevention experts should consider the research and practice implications.

References Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 1173–1182. Dobash, R. E., & Dobash, R. P. (1992). Women, violence, and social change. New York: Routledge. Dobash, R. P., & Dobash, R. E. (2004). Women’s violence in intimate relationships: Working on a puzzle. British Journal of Criminology, 44, 324–349. Felson, R. B., & Outlaw, M. C. (2007). The control motive and marital violence. Violence and Victims, 22, 387–407. Forbes, G. B., Adams-Curtis, L. E., & White, K. B. (2004). First- and second-generation measures of sexism, rape myths and related beliefs, and hostility toward women. Violence Against Women, 10, 236–261. Gallagher, K. E., & Parrott, D. J. (2011). What accounts for men’s hostile attitudes toward women? The influence of hegemonic

male role norms and masculine gender role stress. Violence Against Women, 17, 568–583. Glick, P., & Fiske, S. T. (1996). The ambivalent sexism inventory: Differentiating hostile and benevolent sexism. Journal of Personality and Social Psychology, 70, 491–512. Hamberger, L. K., Lohr, J. M., Bonge, D., & Tolin, D. F. (1997). An empirical classification of motivations for domestic violence. Violence Against Women, 3, 401–423. Hamby, S. (1996). The dominance scale: Preliminary psychometric properties. Violence and Victims, 11, 199–212. Hamby, S. (2005). Measuring gender differences in partner violence: Implications from research on other forms of violent and socially undesirable behavior. Sex Roles, 52, 725–742. Hamby, S., Sugarman, D., & Boney-McCoy, S. (2006). Does questionnaire format impact reported partner violence rates? Violence and Victims, 21, 507–518. Harvey, S. M., Beckman, L. J., Browner, C. H., & Sherman, C. A. (2002). Relationship power, decision making, and sexual relations: An exploratory study with couples of Mexican origin. Journal of Sex Research, 39, 284–291. Harway, M., & O’Neil, J. (1999). What causes men to be violent against women? The unanswered and controversial question. In M. Harway & J. O’Neil (Eds.), What causes men’s violence against women? (pp. 5–11). Thousand Oaks: Sage. Johnson, M. P. (2006). Conflict and control: Gender symmetry and asymmetry in domestic violence. Violence Against Women, 12, 1003–1018. Johnson, M. P., & Ferraro, K. J. (2000). Research on domestic violence in the 1990s: Making distinctions. Journal of Marriage and Family, 62, 948–963. Kershaw, T. S., Small, M., Joseph, G., Theodore, M., Bateau, R., & Frederic, R. (2006). The influence of power on HIV risk among pregnant women in rural Haiti. AIDS and Behavior, 10, 309–318. Krug, E. G., Dahlberg, L. L., Mercy, J. A., Zwi, A. B., & Lozano, R. (2002). World report on violence and health. Geneva: World Health Organization. Lawrence, J. S. S., Eldridge, G. D., Shelby, M. C., Little, C. E., Brasfield, T. L., & O’Bannon, R. E., III. (1997). HIV risk reduction for incarcerated women: A comparison of brief interventions based on two theoretical models. Journal of Consulting and Clinical Psychology, 65, 504–509. Levant, R. F., & Richmond, K. (2007). A review of research on masculinity ideologies using the Male Role Norms Inventory. The Journal of Men’s Studies, 15, 130–146. Levant, R. F., Bryant, S. K., Aupont, M., House, A. T., Richmond, K., & Noronha, D. (2007). Initial validation of the Male Role Norms Inventory-Revised (MRNI-R). The Journal of Men’s Studies, 15, 83–100. Menard, S. (2001). Applied logistic regression (2nd ed.). Thousand Oaks: Sage Publications. Offenhauer, P., & Buchalter, A. (2011). Teen dating violence: A literature review and annotated bibliography. Washington: Library of Congress. Pulerwitz, J., Amaro, H., Jong, W. D., Gortmaker, S. L., & Rudd, R. (2002). Relationship power, condom use and HIV risk among women in the USA. AIDS Care, 14, 789–800. Raj, A., Silverman, J. G., & Amaro, H. (2004). Abused women report greater male partner risk and gender-based risk for HIV: Findings from a community-based study with Hispanic women. AIDS Care, 16, 519–529. Stark, E. (2007). Coercive control: The entrapment of women in personal life. Oxford: Oxford University Press. Stets, J. E., & Hammons, S. A. (2002). Gender, control, and marital commitment. Journal of Family Issues, 23, 3–25. Stith, S. M., Smith, D. B., Penn, C. E., Ward, D. B., & Tritt, D. (2004). Intimate partner physical abuse perpetration and victimization risk

Prev Sci factors: A meta-analytic review. Aggression and Violent Behavior, 10, 65–98. Straus, M. A. (2004). Prevalence of violence against dating partners by male and female university students worldwide. Violence Against Women, 10, 790–811. Straus, M. A. (2008). Dominance and symmetry in partner violence by male and female university students in 32 nations. Children and Youth Services Review, 30, 252–275. Straus, M. A., Hamby, S. L., Boney-McCoy, S., & Sugarman, D. (1996). The revised Conflict Tactics Scales (CTS2): Development and preliminary psychometric data. Journal of Family Issues, 17, 283–316. Straus, M. A., Hamby, S. L., Boney-McCoy, S., & Sugarman, D. (1999). The personal and relationships profile (PRP). Retrieved from http://pubpages.unh.edu/~mas2 Swim, J. K., Aikin, K. J., Hall, W. S., & Hunter, B. A. (1995). Sexism and racism: Old-fashioned and modern prejudices. Journal of Personality and Social Psychology, 68, 199–214.

Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate statistics (5th ed.). Boston: Pearson Education. Tanha, M., Beck, C. J. A., Figueredo, A. J., & Raghavan, C. (2010). Sex differences in intimate partner violence and the use of coercive control as a motivational factor for intimate partner violence. Journal of Interpersonal Violence, 25, 1836–1854. Tjaden, P. G., & Thoennes, N. (2000). Extent, nature, and consequences of intimate partner violence: Findings from the National Violence Against Women Survey. Washington: U.S. Department of Justice, National Institute of Justice. Wingood, G. M., & DiClemente, R. J. (2000). Application of the theory of gender and power to examine HIV-related exposures, risk factors, and effective interventions for women. Health Education & Behavior, 27, 539–565. Yang, X., & Xia, G. (2006). Gender, work, and HIV risk: Determinants of risky sexual behavior among female entertainment workers in China. AIDS Education and Prevention, 18, 333–347.