Abstract
Estimates of sexual violence and partner violence rates among young women are generated primarily from college samples. Few studies have data to compare rates among similar-aged women attending college with those who never attended college. This study aims to estimate rates of partner violence by type (sexual, physical, and psychological) and severity (mild, moderate, severe), sexual harassment, and knowing or suspecting that someone put a drug in a drink (drugged drink) among a national sample of 959 young women aged 18 to 24 in an intimate relationship in the past 12 months who were either currently in college (college; n = 272) or never attended college (non-college; n = 687). After adjusting for demographic differences between these two groups, no significant differences were found in rates of sexual partner violence (28.4% non-college, 23.5% college), physical partner violence (27.9% non-college, 26.3% college), psychological partner violence (M score: 6.10 non-college, 5.59 college), sexual harassment (15.5% non-college, 14.1% college), or drugged drink (8.5% non-college, 7.8% college). Finding high rates of interpersonal violence among young women who are and are not currently attending college indicates the need to target all young adults with violence prevention interventions in educational, workplace, and other community-based settings.
Introduction
Rates of partner violence, sexual violence, and sexual harassment are highest among young women (Black et al., 2011; Tjaden & Thoennes, 2000), yet few population-based studies have evaluated whether rates of these forms of interpersonal violence against women differ among young women (ages 18-24) who are currently attending college (hereafter college) relative to those who have never attended college (hereafter non-college). Such an investigation is important because the majority of young women, ages 18 to 24, will not attend a 4-year college and the demographic characteristics of non-college young women are associated with higher risk of violence (e.g., lower income or unemployed families). Furthermore, young non-college women will not directly benefit from the sexual violence prevention programming now mandated for college campuses by the U.S. Department of Education (Ali, 2011). Thus, young non-college women may be at a disproportionately higher risk yet remain an under-served population with respect to prevention of interpersonal violence against women. This investigation of interpersonal violence rates among young women recruited from the same source and attending or not attending college begins to fill this knowledge gap.
The majority of studies providing estimated rates of interpersonal violence among young women have primarily been generated from samples of college students because this population has been perceived as being at increased risk of sexual violence (e.g., Fisher, Cullen, & Turner, 2000; Smith, White, & Holland, 2003). Another explanation for the focus on college students is that this population is easily accessed by researchers. However, these convenience samples, typically drawn from a single campus, cannot be generalized to young women not currently or never attending college. While large population-based surveys addressing partner and sexual violence among young adults have been conducted (Black et al., 2011; Tjaden & Thoennes, 2000), current college attendance information was not ascertained; thus, rate comparisons cannot be made to distinguish college from non-college peers (e.g., Fletcher, 2010; Whitaker, Haileyesus, Swahn, & Saltzman, 2007). Studies focusing on non-college young women have typically targeted economically disadvantaged women or those at increased risk of violence due to risk-taking behaviors (O’Donnell, Agronick, Duran, Myint-U, & Stueve, 2009; Tyler & Melander, 2009). Although illuminating as to the breadth of violence against women, these sub-samples are not representative of young women or those not in college.
Findings from the National Crime Victimization Survey (NCVS) have relevance for comparisons for violence rates by college attendance. Rennison and Addington (2014) reported significant differences in rates of violence between young women attending college and non-college women in data from the 1995-2011 NCVS, but could not distinguish violence occurring within or outside an intimate relationship. Higher rates of simple and aggravated assault were reported by non-college women (49.8/1,000 and 13.3/1,000, respectively) compared with college women (27.8/1,000 and 8.7/1,000). Again using NCVS data, Hart (2013) found significantly higher annual rates of rape/sexual assaults among those aged 18 to 24 currently attending college (2.6%) than non-college women (2.2%), yet simple and aggravated assault rates were significantly higher among non-college women. Again, these findings did not distinguish whether the violence was committed by a partner.
Women currently in college may be at lower risk of violence than non-college young women because of potentially risky lifestyle differences. Non-college young women may be more likely that those planning college attendance to have sexual relationships, marry or have children. Young women who begin relationships earlier may experience more relationship stress if they marry or cohabitate (Allard, Albelda, Colten, & Cosenza, 1997), become pregnant, or have a child (Moffitt & Caspi, 1999; Molidor, Tolman, & Kober, 2000). This non-college group of young women may face the economic challenges of low income (Carlson, Worden, van Ryn, & Bachman, 2000; Rennison & Welchans, 2002) or unemployment (Meisel, Chandler, & Rienzi, 2003). Finally, young non-college women may be at increased risk of violence because they may not receive prevention information or interventions directed at violence prevention or reduction in risk-taking behaviors such as alcohol or other substance abuse that are required of all Title IV institutions of higher education.
With the Crime Awareness and Campus Security Act, later renamed the Jeanne Clery Disclosure of Campus Security Policy and Campus Crime Statistics Act (20 U.S.C. § 1092(f)), Campus Sexual Assault Victims’ Bill of Rights (Pub. L. No. 102-325, § 486(c)), and the Violence Against Women Act (1994), Congress has begun to require universities and colleges to enumerate crime on campuses and provide education and awareness of crime and specifically gender-based violence on campuses. Because the Campus Sexual Violence Elimination Act (SaVE) (The Violence Against Women Reauthorization Act, 2013) mandates campus-based violence prevention efforts to include bystander programming, bystander intervention programs are becoming more prevalent on campuses. A reduction in sexual violence and potentially partner violence (including dating violence) among women attending college is the anticipated outcome of campus-based bystander interventions. In contrast, few programs targeting prevention of sexual violence and partner violence for young adults occur in the general community. If non-college women are experiencing a wide range of violence types at higher rates than college women, an important next step would be to determine whether existing programs implemented in institutions of higher education can be adapted as community-based or workplace-based prevention programs targeting young non-college populations.
In their recent review of methodologic challenges to measuring violence against college women, Rennison and Addington (2014) noted the lack of an appropriate comparison group of non-college women as one limitation to understanding the larger context of violence experienced by college women. As they correctly conclude, “Without a comparison group, it is difficult to appreciate whether college women are at greater, lesser, or similar risk for certain forms of violence” (Rennison & Addington, 2014, p. 6). Here we address this limitation by evaluating whether a wide range of violence types differ by current college attendance using data from a national sample of young women ages 18 to 24. College and non-college women’s recent (past 12 months) experiences with sexual harassment, drugged drinks (as a proxy for attempted sexual assault), and partner violence by three types (physical, sexual, and psychological) were compared.
Method
Participants
In spring 2010, young women between the ages of 18 and 24 living in the United States who are citizens and had been in a dating or intimate relationship in the past 12 months were eligible and invited to complete a survey online focusing on intimate relationships. The survey platform, www.Zoomerang.com, was utilized because Zoomerang has access to volunteers across the United States willing to complete online surveys to obtain compensation in the form of bonus points for credit cards and other shopping services. The 1,310 participants who completed the survey for this study received a modest compensation by Zoomerang for participation. The procedures for this study were approved by the University of Kentucky Institutional Review Board (#10-0412-P2H).
Measures
Outcomes
We used five measures to capture a wide range of violence experienced: sexual harassment, drugged drink, sexual partner violence, physical partner violence (in Table 2 by specific tactic), and psychological partner violence by scale (in Table 3). We opted to cover a wider range of violence experienced versus including more detailed questions on a more narrow definition of one specific form of violence (e.g., exclusively partner violence).
Sexual harassment
Two items (presented in Table 2) based on the Sexual Experiences Questionnaire (SEQ) were used to measure sexual harassment victimization occurring in the past 12 months (Fitzgerald, Magley, Drasgow, & Waldo, 1999). The items chosen represent the most common forms of sexual harassment. Response options ranged from 1 = never, 2 = once a year, 3 = few times a year, 4 = monthly, 5 = weekly, to 6 = almost daily. For the analyses, the two items were each scored dichotomously with 0 indicating it happened less than monthly and 1 indicating it happened monthly or more often.
Drugged drink
Because drug-facilitated sexual assault is now included in the definition of rape (National Intimate Partner and Sexual Violence Survey [NISVS]) and the experience of having alcoholic drinks either additionally dosed with more alcohol or drugs has become more widely reported (Taylor, Prichard, & Charlton, 2004), we created an item of reported attempts by others to drug their drink: “How many times do you know or suspect that someone put a drug into your drink?” (hereafter drugged drink). The drugged drink item is a proxy measure for potentially attempted drug-facilitated sexual assault. This item was treated dichotomously in analyses, such that any indication of having a drink drugged resulted in a score of 1 although this item’s response options were 0, 1 time, 2 times, 3 to 5 times, 6 to 10 times, and more than 10 times.
Partner violence
The remaining three sets of items refer to tactics used by the respondent’s current or former romantic partner during the prior 12 months.
Sexual partner violence was measured using four items modeled after NISVS items (Black et al., 2011). Cronbach’s alpha for the four items in this sample was .83. Item response options were never, once a year, few times a year, monthly, weekly, and almost daily, but items were used dichotomously in the analyses. Due to the severity of the items, any occurrence of these partner actions was scored as 1.
Physical partner violence was assessed by combining physically violent behaviors based on the Conflict Tactics Scale–2 (Straus, Hamby, Boney-McCoy, & Sugarman, 1996) into items representing severity levels of milder, moderate, and severe tactics. Cronbach’s alpha for the 12-item scale was .94. Item response options were never, once a year, few times a year, monthly, weekly, and almost daily, but items were used dichotomously in the analyses. Due to the severity of the items, any occurrence of these partner actions was scored as 1.
Three distinct measures of psychological partner violence were used. The Measure of Psychologically Abusive Behaviors (MPAB) was designed as a measure of serious psychological maltreatment with the additional stipulation that respondents believed that the partner intended psychological harm with his or her actions (Follingstad, 2011); Cronbach’s alpha was .96. Response options for this measure were never, once a year, few times a year, monthly, weekly, and almost daily. Scores for the total scale and the three sub-scales representing increasing severity levels of the categories of psychological abuse were determined by adding the total number of endorsed behaviors on the scale occurring at least monthly. The data from these scales were used as continuous variables and are presented in Table 3.
The Women’s Experience of Battering (WEB; Smith, Earp, & DeVellis, 1995), the second psychological abuse scale, was used as a measure of the impact that women were likely to experience as a result of ongoing physical violence (although the items require no verification that physical violence occurred in the relationship) as an alternative to measuring violence by counting physical violence incidents. Response options for the original 10-item scale ranged from “not at all” to “a lot” (five responses) with a total score ranging from 0 to 50; Cronbach’s alpha was .96. The score from this scale was used as a continuous variable.
Finally, for the third psychological abuse measure, we used a shortened version of Tolman’s (1989) Psychological Maltreatment of Women (PMWI) original scale (Tolman, 2001). For the PMWI to be comparable with the MPAB, we used the same response options as those for the MPAB. The final score reflects the total number of behaviors occurring on at least a monthly basis. Possible PMWI scores for this 21-item scale ranged from 0 to 63; Cronbach’s alpha was .96.
Covariates
Demographic characteristics
The following demographic characteristics were collected for all women: age, race, sexual attraction, highest educational attainment of either parent, current relationship status, number of children, and number of romantic partners in past 5 years (see Table 1).
Demographic Characteristics and Substance Use Covariates of Young Women Ages 18 to 24 by Current College Attendance.
Note. NS = Not statistically significant (p > .05); GED = General Educational Development.
Substance use
Because college students may have greater access to alcohol and drugs, we additionally assessed binge drinking in the past month (at least five drinks in 1 day) and drug use in the past 12 months. Drug use was defined as using (a) illicit opioids or methamphetamine, (b) cocaine/crack or other street drugs, and/or (c) anti-anxiety medications or pain pills. Prescription drug use was queried regardless of whether the use was with legitimate prescriptions or within medical guidelines. Brand names of prescription drugs were provided as well as official names to increase the likelihood that participants would recognize medications that they had used.
College versus non-college comparison
We compared rates of violence among women ages 18 to 24 who were either currently attending college or had never been to college. We did not ask attributes of the experienced college (e.g., 2- or 4-year institution, on- or off-campus residence, sorority membership). We excluded women who had graduated from college because we were interested in the effect of being in college on current violence rates. We could not determine from the data available how long the women have been out of college and therefore could not distinguish recent (e.g., within the past 12 months) from longer term graduates (beyond the 12-month recall window).
Of the 1,310 participants, 193 who had already graduated college were excluded because these women could not report on their experiences as college students in the past year, having already graduated, and they could not be in the comparison condition of participants never having attended college. Because women who do and do not attend college differed in our sample on sociodemographic attributes (e.g., parental education, race, having children, and being married), we further excluded 158 women who were missing data on one of these attributes. The final sample size for participants ages 18 to 24 was 959 women: 272 current college students (college) and 687 who had never attended college (non-college).
Statistical analysis
Data were analyzed using SAS Version 9.3; SAS Institute; Cary, North Carolina. Means and standard deviations were calculated to describe the distribution of each continuous variable whereas frequencies and percentages were used to describe the distribution of categorical variables; χ2 or t tests with associated p values were calculated (see Table 1). The proportion of young women experiencing sexual harassment, drugged drink, and sexual and physical violence is presented in Table 2 by current college attendance. Violence rates (%) are presented for all violence outcomes with the exception of psychological violence, which was reported as mean scale scores. Logistic regression was used to estimate the association between current college attendance and each dichotomized violence measure. To understand how both demographic characteristics and substance use covariates differ between women in college and non-college women and might affect associations with violence rates, we reported unadjusted rates (%) and χ2 and p values adjusted for demographic and substance use (Table 2). General linear regression models were used to estimate associations with college attendance and psychological violence frequency scores in separate models while adjusting for these same covariates (see Table 3).
Recent Interpersonal Violence (Sexual Harassment, Drugged Drink, Sexual and Physical Partner Violence) by Current College Attendance.
Note. NS = not statistically significant (p > .05).
Logistic regression models adjusted for age, race, highest educational attainment of either parent, number of romantic partners in the past 5 years, current relationship status, number of children, sexual attraction, drug use, and binge drinking.
Tactic had to occur at least monthly.
Recent Psychological Partner Violence by Current College Attendance.
Note. MPAB = measure of psychologically abusive behaviors; NS = not statistically significant (p > .05); WEB = Women’s Experience of Battering; PMWI = Psychological Maltreatment of Women.
General linear models with each psychological partner violence measure as the dependent variable in a separate model with the same covariates as above.
Results
As anticipated, the demographic characteristics of young college women differed from non-college women (Table 1). College women were more likely than non-college women to have parents with higher education, be single, and have no children. Controlling for these demographic characteristics, college women were also more likely than non-college women to binge drink (p = .04) but non-college women were more likely to use illegal drugs, prescription opiates, or benzodiazepines than college women (p = .006). Given differences in these demographic characteristics, subsequent analyses comparing violence rates by college attendance were adjusted for these characteristics. The addition of substance use covariates reduces the possibility that these forms of risk-taking behaviors may explain potentially higher violence rates in college versus non-college women.
When comparing the rates of sexual harassment, drugged drink, sexual partner violence, and physical partner violence, there were no significant differences in these proportions by current college attendance (Table 2). Similarly, no significant differences in psychological partner violence scores for any of the three measures were observed by current college attendance (Table 3).
Discussion
We found no statistically significant differences in the wide range of violence rates investigated when comparing college and non-college young women after adjusting for demographic characteristics and substance use. Because few studies have explored this range of violence types in the same sample of young women by current college attendance, we have little ability to compare these findings with others.
By comparing violence rates observed in this study for college and non-college women with other rates from other studies, we can approximate the reliability of our abbreviated measures for the wide range of interpersonal violence formed measured in this survey. Rates of sexual harassment (based on two items) in this sample were lower for both college (48%) and non-college women (44%) than the American Association of University Women (AAUW) reported rates (62%) for college women endorsing items indicative of sexual harassment (Hill & Silva, 2005), but that report included 15 different tactics of sexual harassment. The fact that the two items by themselves resulted in fairly high rates of harassment suggests that both groups of young women are frequent targets of sexual harassment. The proportion of women in this sample who knew or suspected that someone had purposefully put a drug in their drink (drugged drink in this sample: 7.8% in college and 8.5% in non-college women) was comparable with rates ranging from 8.0% to 9.7% among 4,203 18- to 24-year-old current college women surveyed electronically using the same question and a random sampling design (unpublished finding in three college samples, 2010). The rates of sexual (23.5% any and 5.5% forced sex) and physical (26.3% any and 8.5% severe) partner violence among young women currently in college were similar to other reports (e.g., 9.4% mean severe physical assault rate—Straus, 2004). The rate of sexual coercion (27.8%) observed by others (Sabina & Straus, 2008) was more similar to rates observed among non-college women (28.4%) than college women (23.8%). Rates of current (past 12 months) partner physical (4%) and sexual violence (3%) from the recent NISVS report (Black et al., 2011) were lower than we observed in our younger population (8% severe and 12.5% moderate physical partner violence, 9% coerced sex, and 6% forced sex). Our finding of no statistically significant difference in rates of forced sex by a partner among college and non-college women contrasts with others who found differences in rates between college and non-college populations (Kilpatrick, Resnick, Ruggiero, Conoscenti, & McCauley, 2007).
Limitations
Self-selection of respondents, the potential loss of interested women who did not have access to a computer, and self-report of responses may bias our estimates of violence rates. To address self-selection and the representativeness of our sample, we compared the demographic profile of this national sample with that of Census 2010 data (U.S. Census Bureau, 2011) and determined that our sample was reasonably similar. Among women ages 18 to 24 included in the 2010 Census, 77% identified as White race compared with 75% of our sample. In this same age and gender sub-group included in Census 2010, 42.9% had received some college or an associate’s degree compared with 28.6% of our sample; however, if we include the 193 women we excluded as 18- to 24-year-old women who had graduated college, our rate increases to 40.3%. We cannot directly compare marital status in our sample relative to Census data because we included living together with married and Census data does not. Because we were asking about physical, sexual and psychological partner violence, and sexual harassment experienced by young women, their own self-report was the only available source of these data.
Strengths
This work is among the first to compare rates for a range of physical, sexual, and psychological partner violence; sexual harassment; and drugged drink rates for young women currently in college relative to those never attending college. Inclusion of the wide range of interpersonal violence rates provides a more comprehensive appreciation of the scope of violence experienced by young women in all populations. As a large population-based study, the representativeness of the sample is relatively good. We were able to adjust for a range of demographic characteristics and substance use covariates that distinguish those currently in college relative to those who never attended college. Finally, we were able to provide estimates of a range of current interpersonal violence experiences that are disproportionately experienced by young women including sexual harassment, drugged drink (as a proxy for attempted sexual assault), and partner violence by type (physical, sexual, and psychological).
This study challenges the idea that college attendance is associated with higher risk of different types of partner violence, sexual harassment, and drugged drink. The range and specificity of the data analyses across different types of sexual harassment, drugged drink, and partner violence by type resulted in no statistical differences between young women currently in college and those never attending college. Although non-college women may more frequently experience external stressors associated with their marital, family, or work environments that could lead to their own experiences with physical, sexual, or psychological partner violence or sexual harassment, we find no statistical difference in the range of rates for these types of violence with those currently attending college. Few hypotheses suggesting that external elements appear “causal” of the use of physical violence in intimate relationships (with the possible exceptions of pregnancy and alcohol use) have been supported in existing literature. Indirectly, the lack of significant findings suggests that causal factors may be found more in relationship factors, beliefs, and individual differences than the structural differences of existing within a college setting versus “the real world.” Interestingly, Rennison and Addington’s (2014) analysis of the NCVS data from 1995 to 2011 concluded that college was a protective factor for young women, but these data covered a broader range of violent crimes rather than specific violence within relationships. Therefore, while living on a campus and being single appear to reduce exposure generally to crime for young women (especially non-Hispanic White women), the data from our study’s national sample suggest that violence from a partner or sexual harassment occurs at similar rates for college women and women never attending college.
While differences in partner violence, sexual harassment, and drugged drink rates did not differ among college and non-college young women, university administrators and others charged with young women’s and men’s safety while attending college need to commit consistent efforts to reduce the high rates of violence observed among current students (26.5% physical partner violence with 8.5% severe physical violence, 5.5% physically forced sex by a partner, and 48% sexual harassment with all estimates measured in the past 12 months). The Campus SaVE Act requires implementation of primary prevention interventions including bystander training programs on college campuses. While these programs are promising in their ability to increase bystander efficacy and behaviors (Banyard, Plante, & Moynihan, 2007; Coker, Cook-Craig, Williams, Fisher, Clear, Garcia, & Hegge, 2011), evaluation research has yet to establish consistent efficacy of these programs to reduce interpersonal violence.
These findings are perhaps most important for young women (and men) not attending college. Because the focus of measuring, preventing, and responding to different types of partner violence and sexual harassment has been primarily directed at young women attending college, very few prevention interventions or responses target those not currently attending college. Community and workplace interventions targeted to young adults are needed to reach this equally high risk population. Those charged with ensuring the safety of young women (and men) might find engaging young adults through their established peer social networks using social media to be an efficient and high-impact strategy. Furthermore, efforts to directly involve all young adults in educational and intervention programs to reduce risk of partner violence, sexual harassment, and drugged drink are encouraged.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
