The relationship between risk factors, protective factors and women’s use of sexual coercion

ABSTRACT

Researchers have found that both women and men engage in verbal and physical sexually coercive behaviours (Black et al. [2011]. The National Intimate Partner and Sexual Violence Survey (NISVS): 2010 summary report. National Center for Injury Prevention and Control, Centers for Disease Control and Prevention; Bouffard et al. [2016]. Examining the correlates of women’s use of sexual coercion: Proposing an explanatory model. Journal of Interpersonal Violence, 31(13), 2360–2382. https://doi.org/10.1177/0886260515575609). Risk factors such as rape myth acceptance and prior victimisation related to sexual coercion have been identified however, no research has examined how potential protective factors may impact engaging in this kind of behaviour. The current study examined how risk and protective factors related to the incidence of sexual coercion among a sample of 713 college women from a southern university. The results of the study supported previous findings on risk factors such as higher rape myth acceptance, having experienced prior sexual victimisation, and having a lower belief in the law, were significantly related to engaging in sexual coercion for the women in this sample. Findings indicated several protective factors were significantly related to lower chances of having engaged in sexual coercion, such as increased ability to control thoughts and behaviour and having positive social support. The significance of these findings and implications are discussed.

Practice impact statement:

This article can assist educators and professionals with developing sexual assault prevention programmes by explaining what factors are significant to college-aged women’s willingness to engage in coercive behaviour.

KEYWORDS:

• Protective factors
• Women’s sexual coercion

Sexual coercion includes a range of behaviours that are used by an individual to get another person to engage in sexual activities with them. These behaviours include verbal tactics, the use of drugs and alcohol, and physical tactics (Struckman-Johnson et al., 2003). Recent research has found evidence that women’s use of sexual coercion is more common than societal perceptions would suggest (Stemple et al., 2017). Research has demonstrated that some women do engage in verbal and physical sexual coercion against men, despite most societal and empirical attention being paid primarily to men as those who engage in such coercive behaviours (e.g. Black et al., 2011; Bouffard et al., 2016; Fiebert & Tucci, 1998; O'Sullivan & Byers, 1993; Struckman-Johnson et al., 2003). For example, the National Intimate Partner and Sexual Violence Survey report that about 6% of men experienced sexual coercion during their lifetimes (Black et al., 2011). Among the men who reported experiencing sexual coercion, 81.6% of these incidents were reportedly committed by women (Black et al., 2011). Furthermore, research that used data from multiple large-scale agency surveys found that of the men who experienced sexual coercion, more reported that the individual engaging in the coercive behaviour towards them was a woman (Stemple et al., 2017).

The existing research has primarily focused on the risk factors that predict an increased likelihood of engaging in sexual coercion, especially among men. From the limited research that exists on female sexual coercion, the known risk factors include various antisocial attitudes, having experienced prior sexual victimisation, and having engaged in prior risky sexual behaviours (Bouffard et al., 2016; Russell & Oswald, 2001). At the same time, there are several protective factors that could be potentially useful in understanding sexual coercion, which may mitigate the criminogenic effects of these risk factors (e.g. antisocial attitudes), and have been used to predict desistance from crime, more generally. In particular, some existing research has examined the role of various protective factors (e.g. emotional control skills, attachment to family/peers) among samples of known sex offenders. This emerging research has established linkages between these protective factors, and the likelihood of sexual offending, specifically among male samples (Miller, 2015; Ullrich & Coid, 2011). While it is known that protective factors may relate to a reduced likelihood of male sexual offending, how these factors may relate to the probability of women engaging in sexually coercive behaviours have yet to be empirically explored.

The current study then adds to the literature by examining how potential protective factors and other known predictors (e.g. risk factors like prior victimisation and rape myth acceptance) may be related to having engaged in each of several sexually coercive behaviours, among a sample of college-aged women. As previously mentioned, there is considerable evidence that women do in fact engage in sexual coercion tactics, however, less is known about the risk factors for such behaviour by women. In addition, we believe this is the first study to explore whether certain protective factors may be related to women’s use of sexual coercion. It is important to understand how various individual-level characteristics, including a range of potential risk and protective factors, may relate to female sexual offending (Struckman-Johnson et al., 2003). Identifying similarities between the predictors of sexual coercion among men and women may help to better understand the underlying nature of these acts among women, as distinct from those factors that may be unique predictors of men’s sexually coercive behaviour. The results of the study will be able to provide insight into possible strategies for reducing sexual coercion among college age populations.

Sexual coercion tactics and prevalence

Sexual coercion is a term that refers to a wide range of behaviours used in order to have sexual contact with someone against their will. These often include behaviours such as using verbal tactics (like saying things you don’t mean or pressuring someone with arguments), lying to someone, having them consume drugs or alcohol and even the use or threat of physical force (Struckman-Johnson et al., 2003). In the existing scientific research on the topic, it is not uncommon to see the terms sexual coercion, sexual aggression and sexual perpetration used interchangeably to refer to this wide range of behaviours (Lyndon, White, & Kadlec, 2007). The current study examines a series of coercive behaviours, both verbal and physical, that represent the same kinds of domains that have been examined in previous studies (e.g. Abbey et al., 2001; Abbey et al., 2011; Bouffard et al., 2016; Mouilso & Calhoun, 2013).

Research has examined the prevalence of verbal and physical sexual coercion engaged in by both college-aged men and women (Anderson, 1996; Hines, 2007; Struckman-Johnson et al., 2003). The existing research on sexual coercion tactics used by either men or women has generally examined two broad categories of sexually coercive behaviours; the use of various verbal coercion tactics and the use of physically coercive strategies. For instance, measures of verbal coercion tactics typically include behaviours such as lying, insisting, arguing, and threatening the individual, while physical coercion tactics can include acts of physical force, and attempting to get an individual drunk or high in order to obtain sex (Hines, 2007). Struckman-Johnson et al. (2003) found that 70% of their male and female college-aged sample (n = 656) had experienced sexual coercion from the opposite sex, and one-third reported they had engaged in these types of sexually coercive behaviours towards a person of the opposite sex. The most commonly reported tactics were attempting to increase the other person’s level of sexual arousal, along with emotional manipulation and lying, with 40.4% of men saying they had used sexual arousal tactics in comparison to 25.5% of women. Likewise, more men (32.4%) in the sample had engaged in emotional manipulation and lying than women (15.2%; Struckman-Johnson et al., 2003). A much smaller number of men (2.9%) and women (1.1%) reported that they have engaged in coercion that involved either the use of physical force or threats.

Focusing on college-aged women, Anderson (1996) examined sexually aggressive behaviours that women in his sample reported engaging in, by using a 26-item scale called the Sexually Aggressive Behaviours Scale. Among the sample of 212 women, 28.5% reported engaging in verbal coercion, and 7.1% reported having used physical coercion on men. Among a sample of female college students, Bouffard et al. (2016) found that 30% of their sample reported engaging in some type of sexually coercive behaviour, including behaviours such as saying things they did not mean to obtain sex, getting someone drunk or high to have sex with them, and using force to obtain sex (Bouffard et al., 2016). Stemple et al. (2017) found evidence that there is a significantly larger number of women who engage in sexual coercion than is commonly thought. Specifically, their results indicated that men who experienced sexual coercion were more likely to report that it was a woman who used coercive tactics toward them.

Research suggests that some of the most common sexual coercion tactics used by women are attempts to increase sexual arousal in their partner (e.g. persistent kissing touching, taking off own clothes) and use of verbal coercion tactics, such as emotional manipulation and lying (Struckman-Johnson et al., 2003). From a sample of college students (who reported currently being in romantic relationships) collected from universities around the world, Hines (2007) found that 3% of the male sample reported experiencing physical coercion, while 22% reported experiencing verbal coercion. The existing research generally suggests that between 2% and 6% of college-aged men reported experiencing sexual coercion by use of physical force (Anderson & Aymami, 1993; Krahé et al., 2000; O'Sullivan & Byers, 1993). In a qualitative study of men who reported being sexually coerced by women, the researchers found themes of verbal and emotional coercion were common, which included the use of emotional pressure and blackmail (Platt & Busby, 2009).

Another study found that 87% of their male sample reported a situation where a female partner wanted additional sexual activity, and in some cases the women used verbal coercion tactics, such as humour, pleading and attempting to induce feelings of guilt (O'Sullivan & Byers, 1993). French et al. (2015) used a sample of high school and college-aged males, and found that 43% of the sample had experienced coercion, and the most frequent coercive tactic was verbal pressure. Overall, within the current research on female sexual coercion, there is a pattern of results to suggest that verbal coercion tactics are more commonly used than are physical coercion tactics.

Predictors of sexual coercion

While most research on sexual coercion has focused on predictors of men engaging in such behaviours, a smaller volume of research has found a number of predictors that appear to be relevant for both men and women (Schatzel-Murphy et al., 2009). One study with a large sample (n = 13,877) of college-aged men and women, examined risk factors for engaging in sexual coercion and found that having experienced past sexual abuse and possessing more antisocial attitudes were each significantly related to the use of verbal sexual coercion tactics by both men and women (Gámez-Guadix et al., 2011).

Looking at social desirability and engaging in coercive or sexually aggressive behaviours, Sierra et al. (2009) used a 33-item social desirability scale and found that those college-aged men who scored higher on the scale were less likely to engage in sexually aggressive behaviours. A study that examined a sample of both men and women found that for men there was no significant correlation between social desirability and having engaged in coercion, while for women those who had higher social desirability were less likely to report engaging in sexual coercion (Bell & Naugle, 2007). Thus, some emerging research suggests a series of risk factors that may be equally useful in explaining patterns of sexual coercion use among both men and women.

Research has also examined the relationship between various attitudinal factors and sexual coercion. For example, Hines (2007) examined potential predictors including past victimisation, the perceptions of the status of women and hostile attitudes toward men, using a sample of men and women from different regions of the world. Specifically, they examined the “Status of Women Index” (e.g. women’s level of representation in government, education and the workforce) and “Gender Hostility towards Men and to Women” scales (i.e. adversarial sexual beliefs). Among the men in the sample, 3% reported experiencing forced coercion, while 22% reported experiencing verbal coercion. The results indicated that in regions where women had higher status and exhibited more hostile attitudes towards men, men were more likely to have reported experiencing physical sexual coercion (Hines, 2007). Another factor that has been related to sexual coercion in college students is sexual self-esteem. In a study of college age students from Turkey and Chile, Schuster and Krahé (2019) found that having high sexual self-esteem was a risk factor for coercion among the sample of students from Turkey. These authors’ sexual self-esteem measures captured participants’ perception of their own sexual behaviours, as well as their feelings about their own level of control, adaptiveness, skills and experience related to sexual encounters.

Researchers focusing specifically on female sexual coercion have identified several significant predictors, such as having previously experienced sexual victimisation. Krahé et al. (2003) examined characteristics of women in Germany who reported being sexually aggressive toward men. Using a definition of the term “aggression” that included verbal and physical behaviours used to pressure someone into sexual activity, their findings indicated that the women who were sexually aggressive were twice as likely to have experienced sexual abuse as a child (Krahé et al., 2003). Russell and Oswald (2001) examined predictors of sexually coercive behaviour among a sample of 285 college-aged women and found that 18% had engaged in some form of sexual coercion (e.g. verbal or physical coercion). Moreover, the women who reported previously engaging in sexually coercive behaviours were more likely to have a history of victimisation and a higher tolerance (approval) of sexual harassment (Russell & Oswald, 2001). This finding of past sexual abuse being related to coercive behaviour among women has also been found in research by Anderson (1996), and also by Bouffard et al. (2016).

In a more recent study, Bouffard et al. (2016) examined correlates of women’s use of sexually coercive behaviour in a sample of 582 female college students. The results indicated that women with a history of sexual victimisation, early sexual activity, and possessing more antisocial attitudes and rape myth acceptance (RMA) were more likely to self-report engaging in sexually coercive behaviour (Bouffard et al., 2016). For instance, having higher levels of rape myth acceptance was related to women in their sample having previously used more coercive tactics, whereas stronger beliefs in conventional values (i.e. belief in the legitimacy of law) were related to having used fewer sexually coercive tactics (Bouffard et al., 2016). While RMA is traditionally used as a predictor of men’s use of sexual coercion/aggression tactics against women, there is thus at least some research suggesting that it can be a useful predictor of women’s use of coercion as well. Specifically, one could conceptualise the measure as representing the extent to which an individual holds callous views toward interpersonal and/or sexual relations. For instance, RMA may also reflect individual’s views that sex is something impersonal and that some resistance to engaging in sexual behaviour is common (even among men), and as such it is acceptable to engage in coercion to overcome such “token resistance” (see Bouffard et al., 2016 for a discussion). In general, then, the few existing studies that have examined predictors of sexual coercion by women have suggested that factors such as past sexual abuse, a number of antisocial attitudes including RMA, hostility towards men and previously engaging in risky sexual behaviours were potentially useful in predicting sexual coercion.

Protective factors

Protective factors have more recently become a focus of research examining potential predictors of violent re-offending and desistance among individuals who had previously been convicted of a sexual offense (de Vries Robbe, Mann, et al., 2015; Miller, 2015). Within the existing research literature, however, there is no consistent conceptualisation of what constitutes a protective factor. For instance, some authors have posited that protective factors are social and psychological factors that reduce recidivism likelihood and can work to counteract the detrimental effects of certain risk factors (Clayton et al., 1995; Miller, 2006). It has been suggested that protective factors can be used to explain why someone who might otherwise be subject to the detrimental effects of various risk factors, does not in fact engage in criminal activity (Polaschek, 2017). Protective factors have also sometimes been conceptualised to operate in a similar fashion as do risk factors, such that protective factors are seen as simply any “feature of a person that lowers the risk of offending” (de Vries Robbe, Mann, et al., 2015, p. 18). In the current study, protective factors are understood as measurable constructs whose presence reduces risk. Protective factors can reflect the opposite role of risk factors (e.g. self-control vs. impulsiveness, negative social influence vs. pro-social influence), remedies to risk factors (e.g. medication to manage high hostility levels), or exist somewhat independent of extant risk factors (e.g. goal-directed living). Importantly, protective factors are not simply the absence of risk factors, they must reflect the positive presence of internal, social, or environmental resources that reduce risk and promote desistance.

Regardless of the debate about the nature of protective factors, some authors have conceptualised them to include elements such as the availability of social and emotional supports, as well as engagement in prosocial activities (e.g. religious activities, legitimate employment; see Serin et al., 2016). Criminological theorists (see for example, Hirschi, 1969; Sampson & Laub, 1993) have also hypothesised that these kinds of social bonds can serve as protective factors that are important to the desistance process for at least some offenders. For instance, Ullrich and Coid (2011) found support for the important role of protective factors (i.e. social and emotional supports) in desistence from crime. Their research suggested for instance that spending spare time engaged in prosocial activities was related to decreased general recidivism amongst a sample of known offenders.

When it comes to the applied utility of empirical linkages between protective factors and offending, with few exceptions the potential role of protective factors (e.g. personal strengths, environmental factors) has received significantly less attention than risk factors. For instance, the role of protective factors in risk assessment, especially for sexual offenders, has been largely ignored in favour of a focus on uncovering influential factors that increase risk (de Vries Robbe, de Vogel, et al., 2015). Nonetheless, some researchers have examined protective factors in relation to sexual offending and desistance, specifically among men who sexually offend. For instance, Miller (2006) examined a sample of 110 men who had sexually offended using the Inventory of Offender Risk, Needs, and Strengths (IORNS; see Miller, 2006 for a thorough description), an assessment that includes both risk and protective strength factors. Accounting for static and dynamic risk, the results indicated that several specific protective factors were significantly related to recidivism rates. In particular, the Protective Strength Index (including items related to factors such as cognitive behaviour regulation skills, support from family and friends) was shown to predict likelihood of sexual, violent, and general reoffending, as those who had higher protective strengths scores were significantly less likely to recidivate (Miller, 2015). There are no current studies assessing the utility of the protective strengths scales on the IORNS with females, whether they are females that are justice-involved or in the community.

Likewise, de Vries Robbe, Mann, et al. (2015) proposed the use of Structured Assessment of Protective Factors (SAPROF) to predict desistance from sex offending. The SAPROF includes a number of protective domains which include healthy sexual interests, goal directed living (e.g. self-control skills), and engagement in employment or constructive leisure activities. The SAPROF includes a number of specific sub-scales including those measuring protective internal factors, motivational factors, and external factors (Yoon et al., 2018). In 2015, de Vries, de Vogel, et al. (2015) found that the SAPROF had short-term predictive validity for 83 men who had sexually offended; scores on the SAPROF were significantly related to rates of both general and sexual recidivism. On the other hand, Yoon et al. (2018) conducted an evaluation of the SAPROF’s utility among a sample of 467 adult men who had sexually offended. Their results indicated that the SAPROF scale was a modest predictor of non-sexual recidivism, but that the tool was not a significant predictor of sexual recidivism.

A study of first-year college males examined risk and protective factors related to the men’s previous perpetration of sexual violence (Salazar et al., 2018). Two specific protective factors (having knowledge of effective consent strategies and higher levels of family functioning) were related to lower likelihood of having engaged in sexual violence. At the same time, pornography use, hypermasculinity, alcohol and drug use and perceived peer support for sexual violence were each associated with an increased likelihood of having engaged in sexual violence.

From the research discussed above there is emerging evidence that a number of protective factors may be useful predictors of offending behaviours, however further research is warranted to understand their utility in predicting sexual coercion and sexual offending, especially among women. There are few studies that have examined the assessment of protective factors among women overall, whether they have been convicted of a sexual offence or in the general population and admit to engaging in sexually coercive behaviour. For example, although the authors state that the SAPROF can be used with females who have sexually offended, they state this use is not well validated or studied. Increased attention to the role of protective factors could also be valuable in that including measures of these protective factors in existing assessment tools could improve predictive accuracy of reoffending risk among individuals who have sexually offended. Likewise, such information could also be potentially useful in improving treatment efficacy by providing information to clinicians about additional targets for treatment services (de Vries Robbe, Mann, et al., 2015). Within the current empirical literature that examines protective factors and sexual offending there is evidence to support significant relationships between protective factors and sexually aggressive behaviour among men. However, existing research has not yet investigated these relationships among women.

Current study

Although decades of research have produced extensive knowledge on the predictors of male sexual coercion, much less is known about the role of various risk, and especially protective factors in predicting female sexual coercion. While sexual coercion research has concluded that women do engage in coercive behaviours, albeit at generally lower rates than men, few studies have specifically incorporated potential protective factors into their analyses. At the same time, the few existing studies of female sexual coercion have suggested that predictors of sexual coercion include having anti-social attitudes, and having experienced previous victimisation (Bouffard et al., 2016; Russell & Oswald, 2001). Research has also indicated that protective factors are significant predictors of desistance from sexual offending (Miller, 2015). While there is limited research about the predictors of female sexual coercion, there is even less known about how protective factors may influence female sexual coercion.

To address the gap in the research, the current study examined whether a series of known risk factors, as well as several potentially useful protective factors were predictive of several sexual coercion tactics among women. Specifically, we examine a number of potentially useful protective factors as measured on the Inventory of Offender Risks and Needs (IORNS, Miller, 2006) including measures of Cognitive Behaviour Regulation, Anger Regulation, and Training and Education, as well as the level of support from one’s family and friends. Previous research has found these protective factors are important in predicting sexual recidivism for men (Miller, 2006), however these factors have not yet been examined as potential predictors of sexual coercion use among a sample of college age women. The current study thus attempts to determine which, if any of these protective factors are potentially relevant within a population of college women. As such, the current study attempted to fill these gaps in the literature by examining what risk and protective factors were related to female sexual coercive behaviour, and whether these protective factors continue to predict sexual coercion, after a number of known risk factors were controlled for.

Methods

Sample

The data for the current study were collected as part of a larger study that examined the influence of sexual arousal on male and female decision making in relation to the use of sexually coercive behaviours (see Bouffard, 2013; Bouffard & Miller, 2014 for details on the original study). The original sample included 1166 men and women enrolled in several introductory social sciences courses (i.e. criminal justice/criminology and psychology). As part of the original study, participants were randomly assigned into conditions where they viewed either a sexually arousing video segment (i.e. heterosexual couple engaging in sexual activities) or a control video (depicting part of an academic lecture). A total of 713 female undergraduate students completed the survey including the variables of interest in this study. Among the female students in this sample, the average age was 21.8 years (range = 18–62; SD = 5.03). The sample was predominantly White (76.3%), 17.7% of the sample were African American, and 20.3% were self-described as Latina. The sample was predominantly heterosexual (91%), with smaller proportions identifying themselves as bisexual or lesbian (about 4%, each) or questioning their orientation (1%). Table 1 provides further description of the analytic sample of the female students who participated in this study.

Table 1. Participant characteristics (N = 713).

Variables	Mean (SD) Range	% (n)
Age	21.81 (5.03)	–
White	–	56.00% (314)
Rape Myth Acceptance	1.41 (.37) 1–2.95	–
Sexual victimisation history		25.70% (183)
Engaged in sexual coercion	–	10.20% (64)
Belief in law	3.06 (.65) 0–3	–
Protective Strength Index (PSI)	21.34 (2.92) 0–25	–
Personal Resource Scale	15.73 (2.39) 0–19	–
Education & Training	3.79 (.73) 0–5	–
Cognitive Behaviour Regulation	8.29 (1.25) 0–9	–
Anger Regulation	3.62 (1.09) 0–5	–
Environmental Resource Scale	5.63 (.97) 0–7	–

While the purpose of the current study is exploratory in nature, the sample does generally represent the wider university’s population on several factors. For instance, the university’s population contains 74% White students, 19% African American students and 18% of students who self-identify as Latina, with an average age of 24 years. The current sample’s somewhat lower average age (21.8 vs. 24) is likely related to the sampling frame of introductory level social science classes, which would tend to contain somewhat younger students.

Procedure

Potential student participants were solicited to volunteer for an Institutional Review Board approved study on “College Student Decision Making”. Students from criminal justice/criminology classes make up approximately 37% of the total female sample, while the remaining 63% are students recruited from the psychology undergraduate student subject pool maintained by that department. After volunteering, participants were randomly presented with either the arousal (i.e. erotic video) or non-arousal (control) video segments, each about 6 min in length. This was done in order to reduce any potential selection bias produced by participants withdrawing from the study after learning of which group they were being assigned. Immediately after viewing the experimental or control videos, participants were asked to read a hypothetical scenario depicting a date between an adult male and adult female who were prior acquaintances, and then to respond to several questions related to the scenario. Then, at least 24 h after completing this first portion of the survey materials, participants were sent a link to a second part of the survey which contained all the items used in the current study. This 24-hour delay between the experimental manipulation and the completion of the items used in the current study was designed and implemented so that situational arousal levels induced by the experimental manipulation would not influence participants’ responses to these items (e.g. rape myths, IORNS, prior victimisation items, and so on).

This data was originally collected as part of a project to examine the impact of sexual arousal on the use of sexual coercion, however, there was only one significant (but small) difference between those in the arousal and control groups in terms of the variables examined in this study. Specifically, 29% of those in the Control group reported prior sexual victimisation compared to 23% of those in the Arousal condition (Χ² = 4.682, p < .05). None of the other variables examined in this study differed between the two groups, as would be expected in light of the intentional delay between the experimental manipulation and the completion of these items. For instance, there was no significant difference in the percent of the two samples who reported having used physical force to obtain sex (1.7% and 1.3% in the Control and Arousal conditions respectively, Χ² = .228, n.s.). Likewise, the average score on the Rape Myth Acceptance scale was 1.42 for each group and the average score on the Personal Resources Scale among those in the Control group (mean = 15.71) was statistically similar to the average score for this particular protective factor among those in the Arousal group (mean = 15.76, t = –.294, n.s.). The results for each of the other protective factors, demographic and background variables and each of the four other past sexual coercion measures were also not significantly different across the Arousal and Control groups (results available upon request). As such, the pooled dataset (including participants from both conditions together) is used for the analyses that follow.

Measures

Outcome measure: prior use of sexually coercive behaviours.

The dependent variable in this study is a measure of whether the participant reported having previously engaged in any of five specific sexually coercive tactics. This type of retrospective measure has been used in several published studies on predictors of sexual coercion (Abbey et al., 2001; Abbey et al., 2011; Bouffard et al., 2016; Mouilso & Calhoun, 2013). Specifically, these five items include the following, with each item being answered with a Yes (1) or No (0) response.

1. I have threatened to end a relationship in order to obtain sex.

2. I have pressured someone with continual arguments to obtain sex.

3. I have said things I did not mean to obtain sex.

4. I have gotten someone drunk or high to obtain sex.

5. I have used physical force to obtain sex.

Responses to these five questions were used to compute a dichotomous measure. Specifically, if a participant reported having ever engaged in any of these five behaviours, they were scored a “1” on our dichotomous outcome variable, and a “0” if they reported that they had never engaged in any of these behaviours. Overall, 10.2% of the sample reported that they had engaged in at least one of these five types of behaviour. The most common type of coercion was “saying things you did not mean to obtain sex” (reported by 7.6% of the sample; see Table 1). The least common type of sexual coercion was “using physical force” (reported by 1.4% of the sample).

Independent variables

Risk Factors. Previous research has supported the link between factors such as rape myth acceptance, having a prior sexual victimisation experience, and having antisocial attitudes and the likelihood of engaging in sexual coercion (Abbey et al., 2001; Bouffard et al., 2016; Forbes et al., 2004; Gámez-Guadix et al., 2011). Several potentially relevant predictor variables were examined in this study, including one’s level of endorsement of various rape myths (from the Illinois Rape Myth Acceptance scale; Lonsway & Fitzgerald, 1995). In addition, sexual victimisation experiences prior to the time of the data collection were measured with a modified set of questions from the Sexual Experiences Scale (Koss et al., 2007). Finally, antisocial attitudes were measured with a set of questions related to one’s endorsement of several items related to “belief in the law”, (i.e. conventional values; Hirschi, 1969).

Illinois Rape Myth Acceptance scale. The Illinois Rape Myth Acceptance Scale contains 19-items, including statements such as “When women talk and act sexy, they are inviting rape” (Lonsway & Fitzgerald, 1995, p. 707). Participants responded to each of these 19 items using a 4-point Likert scale (4 =  Strongly Agree, 1 =  Strongly Disagree). Beyond capturing information about these women’s endorsement of rape myths as they relate to men’s sexual aggression again women, we also conceptualise this measure as reflecting some more general views about interpersonal and sexual relations. For instance, viewing sex as impersonal and competitive, as well as holding beliefs that it is common for people of offer “token resistance” to sexual advances. The average score for all 19 items is used in later analyses and this set of items exhibited good internal consistency (Cronbach’s alpha = .874). Across the sample of female students examined here, the average score on this scale was 1.42 (S.D. = .37), suggesting that in general this sample of female college students tended not to agree with these statements.

Sexual Experiences Scale. The measure of sexual victimisation is derived from the Sexual Experiences Scale (SES, Koss et al., 2007) and includes three questions. Specifically, we asked participants about whether someone had ever (1) had oral sex with you or made you have oral sex without your consent, (2) vaginally penetrated you (by a penis, fingers or other objects) without your consent, or (3) anally penetrated you (by a penis, fingers or other objects) without your consent. In the SES developed by Koss et al. (2007) the prompts (e.g. someone had oral sex with you or made you have oral sex without your consent) are followed by a set of several specific tactics the individual may have used to accomplish this (e.g. by telling lies … , by taking advantage of me when I was too drunk … , and so on). In our study, the modified measures reflect whether the person reported ever having been a victim of each of these three types of unwanted contact at some point in their past, regardless of the specific tactic used (i.e. we did not query about the specific tactics used). Responses to each of these three questions were coded Yes (1) or No (0), and the variable used in later analyses is a dichotomous variable indicating whether the person has been subjected to any of these forms of sexual victimisation at some point prior to the data collection (see Table 1). Because this measure is retrospective (and we do not know exactly when in the woman’s past this victimisation may have occurred) we cannot definitely say that any given participant’s sexual victimisation occurred prior to their own use of sexual coercion tactics. Prior research (e.g. Bouffard et al., 2016; Krahé et al., 2003) has demonstrated a link between sexual victimisation and use of sexual coercion, and thus we include this measure in our statistical models to at least examine the extent of their correlation, acknowledging that our data do not allow firm causal conclusions.

Protective Factors. Several items were also included that we hypothesised would be negatively related to the use of sexual coercion, including belief in the law (Hirschi, 1969), and several measures of personal and environmental resources measured using the IORNS tool (Miller, 2006). Specifically, Hirschi’s control theory includes the element of belief in conventional norms (1969), such that individuals who have a belief in conventional norms, such as believing in the “correctness” of laws set by society, are less likely to break the law and we include this item to represent one’s general tendency toward conformity and adherence to the law.

Belief in the Law. Indicators of Hirschi’s (1969) measure of “belief in the law” included four questions each answered on a 5-point Likert scale (0 = strongly disagree, 4 = strongly agree). These items included the following items: (1) “Many things called crime do not really hurt anyone,” (2) “When parents set down a rule, children should obey,” (3) “It is ok to sneak into a ballgame or movie without paying,” and (4) “Even though it is against the law it is ok to sell alcohol to minors.” Item 2 (children should obey rules) was reverse coded for consistency. The measure for this concept is the average score on these four items (average score 3.07, SD = .65) this likely reflects the disparate nature of the specific behaviours that each item refers to. These scale items have been successfully used to predict criminal behaviour in previous studies, with higher belief in the law scores being related to less involvement/likelihood of engaging in various forms of criminal behaviour (see Bouffard & Petkovsek, 2014; Longshore et al., 2004, for examples).

IORNS. The small amount of existing research on protective factors has indicated that those with higher levels of these protective factors are less likely to offend and/or recidivate (Miller, 2015; Ullrich & Coid, 2011). The current study utilised the Offender Risk, Needs, and Strengths (IORNS; Miller, 2006) which includes a number of measures indicative of the presence of individual and environmental protective factors. In particular, the IORNS is a risk and treatment need assessment tool that assesses several static, dynamic, and protective factors related to offending behaviour. For this study, the Protective Strength Index and its component subscales were used. Specifically, the Protective Strength Index (PSI) is derived from items on both the Personal Resources Scale and the Environmental Resources Scale. Questions on the Personal Resources Scale include items from three specific domains; Cognitive Behaviour Regulation (e.g. “I often think before I act”), Anger Regulation (e.g. “I can control my anger more than I used to”), and Training and Education (e.g. “I have special skills for a job”). The Environmental Resources scale includes items that assess the level of positive environmental support, such as support from family and friends, and having family and friends who do not engage in antisocial behaviours (e.g. “I have friends that support me no matter what is going on.”). The IORNS assessment tool has previously been validated on samples of male and female offenders (Miller, 2006; 2015).

Planned analysis

In order to assess whether any of these various risk and protective factors related to whether the participants had previously engaged in sexually coercive behaviours, a series of t-tests will be used to compare differences between those who had and had not engaged in previous sexual coercion.

Next, we estimate two logistic regression models to examine the relationship between our set of protective factors, several well-known risk factors and whether the respondents reported previously engaging in any of the sexually coercive tactics (see Table 4). In the first model, we included the Personal Resource Scale, and then in the second model we included the three separate subscales that make up the Personal Resource Scale to examine whether any of these specific types of protective factor were more salient in predicting sexually coercive behaviour.

Results

Bivariate analysis

As can be seen in Table 2, there were significant differences between those who have engaged in prior sexual coercion and those who have not. Specifically, prior sexual coercion was related to higher levels of rape myth acceptance, less belief in the law, and lower scores on the Protective Strength Index. Looking at the specific subscales of the Protective Strength Index, there were also significant differences between the two groups for the Personal Resources scale, as well as the Cognitive Behavioural Regulation, Education and Training and Anger Regulation scales, where each of these protective factors was higher among those who had not been coercive in the past. There was no statistically significant difference between the groups on the Environmental Resource scale. The results of the chi-square analyses comparing these two groups in terms of the categorical variables examined here are presented at the bottom of Table 3. The results indicated that there are statistically significant differences between the groups on sexual victimisation history and race. In particular, there was a higher proportion of white students among the prior sexual coercion users, and also a higher proportion of students who had reported experiencing sexual victimisation in the past among those who also reported prior use of sexual coercion themselves.

Table 2. Participant sexual coercion history.

Engaged in past sexual coercion
Yes	10.20%
No	89.80%
Count of past sexual coercion tactics	.15
Threatened to end relationship	1.80%
Pressured with arguments	1.80%
Said things they did not mean	7.60%
Got someone drunk	1.80%
Used physical force	1.40%

Table 3. Group comparisons of factors related to coercion.

	Prior sexual coercion		No prior sexual coercion		Test statistics
	Mean	SD	Mean	SD	t-test	p
Age	21.77	5.19	21.94	5.15	.26	.793
Rape Myth Acceptance	1.62	.49	1.38	.34	−3.83	.000**
Belief in the law	2.77	.64	3.10	.64	−3.94	.000**
Protective Strength Index (PSI)	19.9	2.85	21.45	2.90	3.73	.000**
Personal Resource Scale	14.55	2.35	15.82	2.36	3.88	.000**
Education & Training	3.53	.81	3.79	.72	2.40	.019*
Cognitive Behaviour Regulation	7.91	1.34	8.34	1.21	2.54	.011*
Anger Regulation	3.04	1.09	3.67	1.06	4.42	.000**
Environmental Resource Scale	5.38	1.51	5.66	.97	1.81	.074^†
	%		%		X^2	p
White	59.6		43.8		5.93	.015*
Sexual victimisation history	46.9		24.2		15.16	.000**

Note: The “White” race variable is coded 1 = White, 0 = Non-White; For “Sexual Victimization History” Any = 1, None = 0.

^† p < .10.

*p < .05.

**p < .01.

Multivariate analysis

Model 1 examined whether these independent variables are related to having reported engaging in any of our sexual coercion behaviours in the past. Given that the bivariate analysis indicated that the Personal Resource Scale was significantly related to prior use of coercion, while the Environmental Resources was not, only the Personal Resource Scale was included in this regression model. The results indicated that for every one unit increase in the Personal Resources Scale there is a 13% decrease (1–.87) in the odds of having engaged in sexual coercion previously. Rape myth acceptance and having experienced a sexual victimisation were also significant predictors. Specifically, a one unit increase in the score on the rape myth acceptance scale corresponds to a 239% increase in the odds of having engaged in sexual coercion. Having experienced a sexual victimisation increased the odds of having engaged in sexual coercion by 196%. Finally, a one unit increase in the score on the belief in the law scale corresponds to a 41% lower likelihood of having engaged in sexual coercion. Of these three significant coefficients, rape myth acceptance explained the largest amount of variance in the model. Neither age nor race are significant predictors in this regression model. In this first model, 17% of the total variance in the dependent variable was explained (Table 4).

Table 4. Logistic regression models predicting sexual coercion.

	Model 1			Model 2
Variables	b	SE	Exp(b)	b	SE	Exp(b)
Rape Myth Acceptance	1.22**	.34	3.39	1.33**	.36	3.79
Prior Sexual Victimisation	1.08**	.29	2.96	1.08**	.29	2.95
Belief in the law	–.52*	.22	.59	–.55*	.22	.57
Age	.03	.02	1.03	.03	.02	1.04
White	–.60	.29	.54	–.56^†	.29	1.75
Personal Resources Scale	–.13**	.04	.87	–	–	–
Education & Training	–	–	–	–.21	.20	.80
Anger Regulation	–	–	–	–.37**	.13	.68
Cognitive Beh Reg	–	–	–	.06	.12	1.07
Nagelkerke r^2	.17			.19
Model X^2	50.65**			54.92**

Note: The “White” race variable is coded 1 = White, 0 = Non-White participant.

^† p < .10.

*p < .05.

**p < .01.

In the second model, we attempted to examine the relative contribution of the three components that make up the Personal Resource Scale (i.e. Cognitive Behaviour Regulation, Anger Regulation, Education and Training) as they relate to past use of sexual coercion, while still controlling for the set of risk factors included in Model 1. Recall that our bivariate results suggest that each of these three subscales is significantly, negatively related of prior use of sexual coercion. As might be expected, results from Model 2 again reveal that rape myth acceptance and sexual victimisation at some point in the past are each significant predictors. For instance, a one unit increase in the score on the rape myth acceptance scale means there is a 279% increase the odds of having engaged in past sexual coercion, while having experienced a prior sexual victimisation was related to a 195% higher likelihood of having engaged in sexual coercion previously. As in model one belief in law was significant, for every one unit increase in the score on the belief in the law scale, there was a 43% decrease the odds of having engaged in sexual coercion. Examining the three protective strength subscales that make up the Personal Resources Scale, Anger Regulation scores are significant, while the Cognitive Regulation subscale and the Education and Training subscale are not significantly related to past coercion use. The results indicated that a one unit increase in scores on the anger regulation scale relates to a 32% lower likelihood of having engaged in sexual coercion.

This second model only explains a slightly higher amount of the variance in prior use of sexual coercion (19%) relative to Model 1, such that adding the three specific subscales that comprise the Personal Resource Scale does not seem to have a substantial effect on the overall model fit. At the same time, this more detailed examination of the Personal Resources scale seems to suggest that the important component here, in terms of explaining prior use of sexual coercion, may well be the inability of the individual to engage in appropriate Anger Regulation, while other Personal Resource components are not related.

Discussion

Researchers have reported that women engage in sexually coercive behaviours (Bouffard et al., 2016; Struckman-Johnson et al., 2003). Additionally, prior studies have found that a number of specific risk factors, such as sexual victimisation, antisocial attitudes, and rape myth acceptance are related to women’s use of sexual coercion (Abbey et al., 2001; Bouffard et al., 2016; Hines, 2007; Krahé et al., 2003). While some specific risk factors related to coercive behaviour in women have been identified, considerably less is known about the relationship between potential protective factors and sexual coercion. The examination of how certain protective factors may relate to the use of sexually coercive tactics is important for a better understanding of the relevant predictors of these behaviours. This study tried to fill some of that knowledge gap through an examination of how relevant risk and protective factors were related to female use of sexual coercion.

In the current study, we asked a sample of college-aged women to self-report whether they had ever previously engaged in each of five specific types of sexually coercive tactics to obtain sex. Among this sample of college women, 10.2% reported engaging in some type of sexual coercion. This self-reported rate of coercion use is lower than in several previous studies (see Struckman-Johnson et al., 2003). On the other hand, we found that the most common type of coercion among this sample of women was “saying things they did not mean to obtain sex”, while “using physical force” was the least commonly reported. These results are consistent with existing studies (Anderson & Aymami, 1993; Hines, 2007; Krahé et al., 2003; Struckman-Johnson et al., 2003) that show women are more likely to use verbal than physical forms of sexual coercion.

Our findings regarding potential risk factors for engaging in sexual coercion were also consistent with previous research (e.g. Bouffard et al., 2016; Hines, 2007; Krahé et al., 2003). For instance, having been sexually victimised at some point in the past, endorsement of rape myths and less belief in the law were all significantly associated with having engaged in sexually coercive behaviours in the past, as would be expected from previous research findings. The finding that rape myth acceptance accounted for the most variance in the models, even within a sample of women, suggests that they may view sexual aggression as something normalised within society. Previous research has indicated that rape myths are higher among men and are related to men normalising sexual aggression and rape proclivity (Boakye, 2009; Bohner et al., 2006; Reling et al., 2018). Given the relatively consistent findings, from both the current study and previous research about the role of rape myth acceptance, sexual assault awareness and prevention programming at universities should continue to target the endorsement of rape myths as a predictor of sexual coercion/assault across gender.

As mentioned above, experiencing sexual victimisation at some point in the past was also a significant correlate of past coercion use, which is consistent with several existing studies (Bouffard et al., 2016; Gámez-Guadix et al., 2011; Krahé et al., 2003; Russell & Oswald, 2001). Not only is this finding consistent with previous research on female sexual aggression, but sexual victimisation has also been found to be a significant correlate of coercion use by men (Carr & VanDeusen, 2004; Malamuth, 1998). Because both of our measures of sexual victimisation and use of sexual coercion tactics are retrospective (and we did not have information about when either behaviour/event occurred) it was not possible to determine for each woman in our sample which of those two things may have occurred first in their past. As such our results cannot confirm that sexual victimisation has a causal relationship with past coercion use, however our results are consistent with past research showing that these two variables are related to one another. More research is needed looking at the impact of prior sexual victimisation on engaging in sexual coercion and other forms of aggression to better understand the nature of this association.

Turning to the role of protective factors, there were significant overall differences in the scores on Protective Strength Index from the IORNS between participants who had engaged in past sexually coercive behaviours compared with women who had not engaged in sexual coercion. In all the bivariate results there were significant differences in scores on each protective factor between the past sexual coercion use groups. Additionally, the Personal Resource scale significantly predicted the use of sexual coercion in the multivariate analysis. These results indicate that among women, protective factors are significantly related to sexual coercion, and to an extent that is at least similar to the strength of the relationships seen for several risk factors. These findings are important in that they expand our understanding of the scope of relevant influences on sexually aggressive behaviour in women, beyond those predisposing risk factors like rape myth acceptance and prior victimisation. By adding consideration of the role of potential protective factors, researchers can gain a more comprehensive understanding of the pathways to women’s use of sexual coercion, as well as those factors that may act to limit the use of such tactics. At the same time, beyond improving the amount of explained variance in statistical models of women’s use of sexual coercion, these results also point to new areas for prevention and treatment efforts, particularly around improving emotional regulation, as suggested by the significant relationship seen between coercion use and Anger Regulation. Future research should attempt to unpack the exact nature of this relationship and whether, for instance it is limited to control of anger or also applies to one’s (in)ability to cope with other negative emotions that might otherwise serve to precipitate the use of sexual coercion.

Indeed, the specific examination of the role of protective factors in the second of our multivariate models revealed that women with higher scores on the Personal Resource Scale (PER) of the IORNS, and in particular those with better anger regulation skills, were less likely to report previously engaging in sexual coercion. The PER items reflect personal attributes, traits, and accomplishments that research has shown to be protective against engaging in antisocial behaviour. Individuals who score higher on the items and subscales of PER endorse the ability to control their emotions and behaviour more effectively and have training or education that aids in employment (Miller, 2006). The current results indicate that protective factors may indeed play a significant role in reducing women’s use of sexual coercion, especially internal characteristics such as ability to control one’s emotional responses like anger. Prior research has examined the linkage between anger and aggression more generally (see Roberton et al., 2014 for instance), so our results showing a link to forms of sexual aggression are consistent with this existing work. Indeed, other research has found a relationship between sexual offending and anger associated with emotional dysregulation (Velotti et al., 2017). Craig et al. (2020) found that among a sample of men, those who had lower emotional regulation were less likely to be able to control impulses when their sexual advances were turned down. The study further found that emotion regulation moderated the effect of sexual arousal when it came to men reporting likelihood of engaging in sexually coercive behaviour (Craig et al., 2020). As this research suggests that emotional regulation is salient to the likelihood of engaging in sexual coercion for men, our results indicate that this may also to true for women.

Implications and future research

In general, our results for these specific protective factors (e.g. personal resources, and anger regulation in particular) are similar to other research that has shown protective factors as related to desistence from sexual offending in samples of individuals who had already engaged in a sexual offense (De Vries Robbe, De Vogel, et al, 2015; Miller, 2015). On the other hand, the lack of significant relationship between the Education and Training subscale scores and use of sexual coercion may have resulted from the fact that there is relatively little variation in the education characteristic within our sample of women currently enrolled in a university. Likewise, our sample of college women may not exhibit sufficient variation in levels of thought and behaviour regulation skills (as measured on the Cognitive Regulation subscale) for this to be a useful predictor of sexual coercion use among this sample. Further research with samples exhibiting varying ranges of these traits and accomplishments is warranted to examine potential relationships between these protective strengths and sexually coercive behaviours in community populations.

An important contribution of the current study is the finding that protective factors can be useful in explaining sexually coercive behaviour in populations other than groups of known sexual offenders. In addition, our results suggest that these protective factors exhibit significant influence on sexual coercion use, even when controls for several known risk factors are also included. As such, it will be important for additional research to examine whether and how these risk and protective factors may work to counteract one another, as well as whether these types of protective factors may operate differently among different populations. At the same, our current study examined a small number of potential protective factors, such that there is room for additional empirical investigations of the role of other factors including other social bonds and other measures of internal factors such as coping skills, attitudes toward relationships and many others. In addition, future research should also examine these potentially interactive effects of risk and protective factors in relation to a wider variety of specific sexually coercive/aggressive behaviours not examined here. Lastly, as the current study focused on individual factors in the future examining contextual factors, such as their relationship with the person they coerced should be examined.

While our results are preliminary and largely exploratory in nature, one potential implication would involve the development and provision of programming for college students aimed at reducing the prevalence of sexual coercion/aggression through training that focuses on decreasing rape myth acceptance among both men and women, as well as the need to improve regulation of anger and other negative emotions, especially in the context of interpersonal relationships. As this study indicates that protective factors may work to reduce the use of sexually coercive tactics among college women, consistent with protective factor investigations among general justice-involved women and men, developing university programmes that increase effective coping skills, especially around managing negative emotion and developing social support, could work to further reduce sexual aggression on college campuses. Similar to the treatment focus with individuals who have been convicted of a sexual offense, where treatment standards have now shifted to include both the mitigation of risk factors and the development of protective strengths (see ATSA practice standards for adults and juveniles, 2014; 2017), college campus programmes could follow suit to include efforts to reduce risk factors such as rape myth beliefs, but also focus to increase protective factors such as negative emotion management and building healthy coping skills.

With research consistently showing that prior sexual victimisation is a significant factor in understanding sexual coercion, prevention efforts would be wise to consider how to incorporate components that address the role of prior victimisation to raise awareness of these issues among victims in ways that are will not revictimise and help these individuals. At the same time, such awareness training programmes might also need to increase their efforts to provide referrals for more individualised counselling to those who may self-identify as victims after participating in such campus-based prevention programming. Having such services available can help victims process the trauma and provide support, as well as help avoid this past trauma manifesting in negative behaviours being exhibited by these victims themselves (i.e. engaging in sexual coercion). Additionally, current results indicate that helping women effectively cope with negative emotions, especially anger, may serve as a protective factor against engaging in sexually coercive tactics.

In light of research (including ours) that supports the link between rape myth acceptance and sexual coercion, as well as the high rate of sexual victimisation among college students, university prevention programming that raises awareness about campus sexual assault that attempts to decrease college students’ endorsement of rape myths is clearly merited. At the same time, it is important for those developing and delivering these programmes to recognise that women tend to engage in different forms of sexual coercion than do men and that they may have come to those behaviours differently. Thus, the effectiveness of these interventions may be expected to vary by gender and may need to be tailored to account for these somewhat different sets of risk and protective factors. It would be useful for future studies on the effectiveness of these kinds of interventions to examine if there are in fact differences in effectiveness for men and women.

Limitations and conclusions

While our results are useful first steps to begin filling important gaps in the existing research literature related to correlates of women’s use of sexual coercion, the study is not without limitations. For instance, one limitation involves the temporal ordering of relevant variables, in particular the use of prior sexual coercion as a dependent variable and the lack of specific information on the timing of prior sexual victimisations. Future research should attempt to overcome this temporal ordering limitation through the use of alternative designs including longitudinal data collection, as well as those making use of hypothetical vignettes to assess the role of various predictors (risk and protective) on one’s intentions to engage in sexual coercion. In addition, our dependent variable does not measure the frequency or recency of the participants’ past use of these five sexual coercion tactics. Future research on female sexual coercion could be advanced by examining how frequently the women engaged in such behaviours. The items used to measure sexual coercion ask about tactics used “to obtain sex”, and we do not have the ability to know if they completed the sexual act, they intended to engage in. However, the use of this type of measure does allow us to know that some respondents engaged in the various coercive tactics with the intent of obtaining some form of sexual activity from another person. In general, our results do support the utility of future research on women’s sexual coercion and the potentially important role of personal and environmental protective factors, as this study suggests that protective factors are related to women’s engagement in sexual coercion. Additional research should continue this line of inquiry by examining other types of protective factors, and by sampling more diverse populations, to examine whether other specific protective factors may influence sexual coercion tactics used by broader samples of women. As different populations may have different risk factors, it is important to explore the range of protective factors that may be related to the use of various sexually coercive behaviours by diverse samples of women.

In conclusion, this study added to the literature on predictors of female sexual coercion and examined the potential utility of a number of relevant protective factors. To the authors’ knowledge this is the only study to date that has examined protective factors in relation to women’s use of sexually coercive behaviours. As the results of the study indicate, protective factors are salient for predicting the likelihood of engaging in sexual coercion among college-age women. Thus, protective factors should be included in future research on sexual coercion, as well as included in assessment tools to help get a more accurate picture of the factors that are related to sexual coercion.

Acknowledgements

This study received approval from SHSU IRB on 23 April 2011. Protocol # 2011-03-060.

Disclosure statement

One of the authors, Holly A. Miller, is the developer of the IORNS and has financial interest in that measure.