Abstract
Research shows that U.S. college students are at risk for a variety of victimization types. College students in Canada also experience similar types of victimization compared to U.S students, yet recent research shows that Canadian students are at a greater risk than U.S. college students of being victimized. Little is known, however, as to whether college-level factors influence victimization risk and how these risks may be different for U.S. and Canadian college students. Using the American College Health Association’s National College Health Assessment data (ACHA-NCHA II), the current study builds upon past research by exploring the role of both individual and school characteristics and whether they are similarly related to risk of victimization in the two populations. The findings show that Canadian students are at higher risk for victimization and that some school-level factors play a role in predicting this risk for all students, with some variation in ones that matter for each country.
Introduction
Decades of research has documented the victimization risk that college students in the United States face (Baum & Klaus, 2005; Carey et al., 2015; Daigle et al., 2008, 2018; Fisher & Daigle, 2009; Fisher et al., 1998, 2002; Krebs et al., 2009; Martin & Hummer, 1989; Mustaine & Tewksbury, 1998). Within this research, scholars identify U.S. college students as being at risk for a variety of forms of victimization such as theft, rape/sexual assault, physical assault, and stalking (Daigle et al., 2008, 2018; Fisher & Daigle, 2009; Fisher et al., 1998, 2002; Krebs et al., 2009; Martin & Hummer, 1989; Mustaine & Tewksbury, 1998). The reasons that U.S. college students are at risk for victimization vary but stem from the college culture and individual factors. These factors are in line with expectations from two theoretical perspectives—ecological systems theory and routine activities/lifestyle-exposure theory (RAT/L)—that have been used to study college student victimization. According to ecological systems theory, human behavior is influenced by the ecological system in which a person is embedded (Bronfenbrenner, 1977). In this way, individual, interpersonal, and institutional-level features shape behavior. In addition, routine activities theory suggests that a person’s exposure to crime and motivated offenders will increase victimization risk when that person is deemed a suitable target and lacks capable guardianship (Cohen & Felson, 1979; Hindelang et al., 1978). U.S. college students may not be unique in that their counterparts in other nations also experience victimization (Barnes et al., 1991; Chan et al., 2008; DeKeseredy & Kelly, 1993; Mirsky, 2003; Phipps & Smith, 2012).
Despite knowledge about victimization risk in nations other than the U.S., little work has been done comparing the risk of U.S. college students to others. One exception is a study by Daigle et al. (2019) that examined individual-level predictors of college student victimization in the U.S. and Canada. They found that risk factors frequently identified in past research—such as gender, mental health problems, disability, binge drinking, drug use, and sexual history—predicted violent and sexual victimization for students in both the U.S. and Canada. When controlling for these factors, however, they found that Canadian students had higher risks of violent and sexual victimization. This finding suggests that there were factors beyond those in the model that could explain varying risks of victimization for U.S. and Canadian college students. In particular, macro-level factors, such as institutional characteristics, may help explain the difference in risk between the two populations. Beyond this empirical finding, there are several reasons to believe that the college experience and context may be different for U.S. and Canadian college students. First, drinking culture differs in the U.S. and Canada. Second, U.S. and Canadian colleges are structured differently; thus, students attending these institutions may have qualitatively different experiences. Third, policies and practices in place to prevent and respond to victimization are different for U.S. and Canadian schools. To further explore differences in victimization between U.S. and Canadian college students, the current study examines institutional-level risk factors while controlling for individual-level risk factors.
School-Level Predictors of Victimization Risk
The Campus Context
Focusing only on individual-level risk factors is insufficient for understanding the extent of problem behavior on university campuses. Research suggests that a broader institutional perspective may help explain differences in behaviors across universities and why some deal with more issues than others (Bronfenbrenner, 1977; Byrd & McKinney, 2012; Lizzio et al., 2002; Moylan & Javorka, 2020; Moylan et al., 2019). To understand the potential link between school-level characteristics and individual experiences, one may consider the institution as part of a greater ecological system. Therefore, various parts of any given institution are interconnected and influence one another.
According to ecological systems theory, human behavior is formed and shaped through a number of interconnected environments that exist at multiple levels in society (Bronfenbrenner, 1977; Larochette et al., 2010; Way, 1990). One must consider the whole ecological system to fully understand human behavior and why certain negative behaviors are more prevalent in some places than others (Bronfenbrenner, 1977; Byrd & McKinney, 2012). Negative behavioral outcomes are a product of several risk factors operating at the individual, interpersonal, and institutional levels. The ecological environment is made up of nested structures that range from the microsystem, referring to the direct environment where the person interacts with immediate family or friends, to the macrosystem, which encompasses frameworks found in the culture or subculture that affect social structures and activities at the other levels. More concretely, the macrosystem is made up of the various economic, social, educational, legal, and political institutions within a given society that influence all other systems (Bronfenbrenner, 1977; Byrd & McKinney, 2012; Larochette et al., 2010; Way, 1990).
Differences in how these systems interact with one another at various levels within an institution play a role in socializing its members (Byrd & McKinney, 2012; Moylan & Javorka, 2020; Way, 1990). In this way, the institutional-level features of a university are expected to be salient in establishing general behavioral norms. Institutional-level factors encompass the educational setting, teaching practices, and the social and institutional climate of a college or university. Specific institutional-level characteristics should influence values and behavioral norms that exist on campus, giving rise to different sub-cultures across schools (Bronfenbrenner, 1977; Larochette et al., 2010).
In conjunction with ecological systems theory, it is instructive that college in general brings large numbers of young people together into proximity which, acknowledging that society contains individuals who have high potential to offend, includes bringing potential offenders and victims together. However, the type and degree to which colleges and universities foster, and interfere with, victimization opportunities most likely vary. For example, effective policies, programs, and practices should decrease victimization at schools that have them in place. Routine activities theory and lifestyle-exposure theory (RAT/L) are perspectives often used to articulate individual and institutional-level risk such as behaviors that bring individuals in contact with potential offenders in environments lacking capable guardians (e.g., heavy drinking at male-controlled parties). However, since institutional characteristics plausibly shape interactions between potential offenders and victims and presence of capable guardians, RAT/L can also offer reasons to expect that school-level factors influence students’ risk of victimization (Cass, 2007; Stotzer & MacCartney, 2016).
Using these two perspectives together, it would seem that geographic location and the size of school are some factors that influence components of the institution (Bronfenbrenner, 1977; Gottfredson & DiPietro, 2011; Moylan & Javorka, 2020). A large, urban state university will have different values and traditions compared to a small, rural liberal arts college. A private university likely also differs from a public university in the norms and values its members endorse. Variations in institutional-level characteristics are not only shown to affect the level of school disorder (Bayer & Astin, 1969; Hurtado, 1992) and violence (Armstrong et al., 2006; Bayer & Astin, 1969; Fisher et al., 1998; Hurtado, 1992; Moylan et al., 2019; Stage & Downey, 1999), but also influence student perceptions and academic outcomes (Lizzio et al., 2002).
Empirical Tests of Institutional-Level Factors Related to Victimization
Despite the possibility that school-level factors influence behaviors on college campuses, including victimization, relatively few studies have specifically explored these factors. Without doing so, the ways in which college context may influence risk cannot be understood. Research that has incorporated school-level factors has found that alcohol policy and alcohol use of college students may be important. Wechsler et al. (2001) found that students at colleges/universities that ban alcohol for all students on campus were less likely to be heavy episodic drinkers and more likely to abstain from alcohol, although they engaged in as much extreme drinking as students at non-ban schools. Critical in this study is that among students living on campus, those at ban schools were less likely to experience unwanted sexual advances. Despite this difference, there were no statistically significant differences in physical assault or sexual assault/rape victimization for on-campus students. Subsequent research found that higher college-level heavy episodic drinker rates (percent of students self-reporting heavy episodic drinking at a school) increased a woman’s risk of being raped while intoxicated (Mohler-Kuo et al., 2004). Similarly, the risk for sexual assault increases on campuses that have high rates of binge drinking among students, a greater proportion of sexual minority students, and an overall younger student body (Moylan et al., 2019). This finding suggests that characteristics of the student body impact victimization risk.
Other research examining school-level factors either did not identify specific factors that predict sexual assault and other forms of sexual victimization or found few campus-level factors to matter. Fisher et al. (1998) found that the percent of students living on campus was associated with a reduction in theft victimization and that campus theft rate increased risk, yet no other campus-level factors predicted either theft or violent victimization. In addition, Cass (2007) found that individual but not school-level factors predicted sexual assault victimization. This lack of effect is surprising since many of the school-level factors were relevant to RAT/L theory, a theory frequently used to explain victimization in past research. In addition to number of undergraduates, size of campus, and local crime rate, the study included measures of interventions used to interfere with motivated offenders (e.g., limited access to campus), reduce target vulnerability (e.g., escort services), and increase guardian capabilities (e.g., number of police officers). More recently, Cantor et al. (2017) found wide variation in sexual victimization rates and campus climate perceptions among American colleges/universities, which suggests that school differences could help explain victimization. However, analyses revealed that university characteristics predicted little victimization. For example, enrollment size, whether school is public or private, and percent undergraduate were not related to sexual victimization or stalking.
Considering this body of work, several key points emerge. It appears that alcohol use across college campuses may be particularly relevant in explaining sexual victimization risk. Additionally, other school-level factors may not be relevant for this type of victimization. Still, the role that school-level factors play in college students’ risk for other types of victimization has not been fully explored.
The Current Study
The U.S. and Canada differ in fundamental ways that can uniquely influence the institutional-level features of colleges and universities in each country (Lipset, 1986; Ross et al., 2000). For example, drinking culture differs between them. In Canada, the minimum drinking age is either 18 or 19 years old compared to 21 for the U.S. (Arbour-Nicitopoulos et al., 2010). Research shows that Canadians drink 50% more than the global average (Shield et al., 2013), but heavy alcohol use is more prevalent in the U.S. (Kuo et al., 2002). Evidence suggests that the different legal drinking age influences drinking behavior in the two countries (Kuo et al., 2002), and differences in drinking culture potentially influence drinking norms on university campuses. Since heavy alcohol use is more prevalent in the U.S., there may be more campuses with climates supporting it compared to Canada. Moreover, there are many differences between the countries in how colleges/universities are structured. For instance, most colleges and universities in Canada are located in urban areas (Phipps & Smith, 2012; Quinlan et al., 2016) and tend to be commuter based (Pascarella & Terenzini, 2005). Those in the U.S. are often located in rural or suburban areas (Carnegie Foundation for the Advancement of Teaching, 2011), and are mainly campus-based with a focus on student services and social events (Pascarella & Terenzini, 2005).
In addition, U.S. and Canadian colleges may respond to victimization in different ways and often have different policies in place to prevent, detect, and handle victimization. For example, there is little standardization of polices that address sexual violence across Canadian college campuses. Sexual assault is usually bundled with sexual harassment policies that are more focused on staff rather than student victimization (Quinlan et al., 2016; Sheehy & Gilbert, 2015). Investigations into the public universities in Canada conducted in 2014 (a year after the data were collected for the current study) found that only nine of the 78 Canadian universities had adopted a special policy to deal with sexual violence (Mathieu & Poisson, 2014). This same investigation noted that none of the universities located in Ontario had such a policy, although Bill 132 was passed in 2016 requiring a stand-alone sexual assault policy for every Canadian college or university that receives ongoing federal funding (Lopes-Baker et al., 2017). Comparatively, all U.S. institutions of higher learning that receive federal funding are required to provide prevention programming to their students in accordance with the Campus SaVE Act of 2014 (Griffin et al., 2017). In 2017, no federal legislation existed in Canada to reduce sexual victimization on college campuses (Lopes-Baker et al., 2017). Also, unlike the U.S., campus crime data are not systematically collected, analyzed, and distributed in Canada (Quinlan et al., 2016). Although these policies focus on sexual, dating, and stalking violence, they suggest that U.S. colleges may be more engaged in actively providing prevention programming as well as resources for their students who have been victimized. It is possible that having such policies, as well as resources and services which mitigate trauma associated with victimization (Eisenberg et al., 2016), signals to students that violence will not be tolerated. These differences may influence the schools’ institutional-level features and types of behavior and activities supported by the campus climate (Byrd & McKinney, 2012; Lizzio et al., 2002).
Also of consideration is the reporting behaviors of victims in the U.S. and Canada. In 2014, only 5% of Canadians indicated that they reported their sexual assault to the police (Conroy & Cotter, 2017), whereas 33.6% of rapes and sexual assaults were reported to the police by residents of the U.S over the age of 12 (Truman & Langton, 2015). Other research shows that Canadian college women may be even less prone to report their sexual victimization to the police than U.S. college women (DeKeseredy et al., 1993). This lack of reporting may also contribute to an environment on Canadian college campuses where offenders feel they can sexually assault with impunity.
Further, it is possible that these features vary across campuses and create differential risk for victimization reflective of, in part, institutional differences. Evidence from U.S. campuses shows that many institutions of higher-education continue to fall behind in full compliance of Title IX legislation. Estimates show that about 30% of campuses do not have a Title IX officer, 40% of campuses do not have preventative programming in place, and 10% of campuses do not have a Title IX coordinator (as cited in Moylan & Javorka, 2020). Additionally, regarding SaVE Act requirements, Griffin et al. (2017) found that of the colleges and universities in their sample, 30.7% met less than six criteria, 23.8% met 6–11 criteria, 35.7% met 12–17 criteria, and only 11.2% met all 18 criteria (full compliance). The variation in legislative compliance in the U.S. and Canada, as mentioned above, suggests that certain institutional factors may shape the way campuses respond to the policy context surrounding crime reporting. The ways in which campus policy implementation occur is likely moulded by university administrators’ calculation of the risk and benefits of certain institutional responses that could affect liability and public image (Moylan & Javorka, 2020). Such calculations regarding policy implications could dictate the likelihood that someone will report campus victimization.
Although it is likely that there are differences in the experiences of students enrolled in U.S. and Canadian colleges, school-level characteristics have not been included in studies examining risk for victimization across these countries. The current study examines school-level characteristics of U.S. and Canadian colleges along with individual-level risk factors to determine if school-level characteristics are related to victimization. In doing so, multiple types of victimization are explored to see if certain school-level factors attribute risk for different types of violent victimization—violent, sexual, and stalking. In this way, the common features that may explain victimization risk for a variety of types of victimization across campuses can be explored.
Methods
Data and Sample
Data were drawn from the Spring 2013 American College Health Association’s National College Health Assessment II (ACHA-NCHA), which was administered in the U.S. and Canada. Designed to assess health behavior and health-related experiences of college students, the ACHA-NCHA is a national survey of college students. The ACHA-NCHA was first administered in 2000 and continued until Spring 2008, at which time the revised version of the ACHA-NCHA II was implemented. The survey is administered one time in the Fall and again in the Spring (ACHA, 2013). The Spring 2013 data are used for the current project since both U.S. and Canadian data were collected during this administration.
Data from 123,078 students enrolled in 153 institutions are included in the Spring 2013 U.S. data. Only those institutions that used a random sampling design or surveyed all enrolled students were included in the data. Most institutions—122—used a web-based survey, while the others used a paper survey. The Spring 2013 Canadian data include information on 34,039 students enrolled in 32 institutions. To be in the sample, institutions had to survey all of their students, use a random sampling technique, and agree to have their data included (ACHA, 2013). 1 After listwise deletion, our final analytical sample is 120,332. The analytical sample includes 26 Canadian schools. No single variable had missing cases for more than 5% of the sample.
Measures
Dependent variables
Three victimization variables were examined: violent (nonsexual) victimization, sexual victimization, and stalking. Two measures comprise the violent victimization variable. Students were asked if they experienced physical assault and verbal threats during the previous 12 months. Responses indicating that either of these occurred were coded as 1 while those indicating that neither occurred were coded as 0. For sexual victimization, students were asked about four types that could have occurred during the previous 12 months: sexual penetration without consent, attempted sexual penetration without consent, sexual touching without consent, and someone having sex with them without consent while drinking alcohol. Responses indicating that that any of these occurred were coded as 1 while those indicating that none occurred were coded as 0. For stalking, one item is used. Students were asked if they had been the victim of stalking during the previous 12 months. Responses indicating that stalking victimization occurred were coded as 1 and then 0 if not.
Independent variables
Individual-level variables
Several measures of risk factors derived from the RAT/L perspective—that reflect target suitability, exposure to crime and criminals, and capable guardianship—were included. For binge drinking, responses that indicated that students drank five or more alcoholic drinks in a sitting during the past 2 weeks were coded as 1 while those who did not were coded as 0. A measure to capture underage drinking was also included. This measure reflects whether a student is under the age of 21 years and also indicated that they had drank alcohol during the previous 30 days (coded as 1) or not (coded as 0). Drug use indicates whether (coded as 1) or not (coded as 0) students used marijuana, methamphetamine or other amphetamines, sedatives, hallucinogens, anabolic steroids, inhalants, ecstasy and other club drugs, or other illegal drugs within the last 30 days. Students were also asked if they had been treated for or diagnosed with 14 different mental health issues such as depression and anxiety. Those who indicated they had been treated for or diagnosed with at least one mental health issue were coded as 1 and coded as 0 otherwise. Disability is a measure of whether or not a student has one of five disabilities (deafness/hearing loss, mobility/dexterity disability, partial sightedness/blindness, speech or language disorder, or other). Students who indicated having any of these were coded as 1 and then 0 if not. Greek is a measure of whether (coded as 1) or not (coded as 0) a student is affiliated with a social Greek fraternity or sorority. Involvement in sports measures whether a student is a part of a varsity, club, or intramural sports team (coded 1 for yes, 0 for no). Risky behavior measures whether (coded as 1) or not (coded as 0) a student wore a seatbelt or helmet in the past 12 months while driving, biking, motorcycling, or skating. Single is a measure of whether (coded as 1) or not (coded as 2) a student is in a romantic relationship. Living location measures where a student lives. Dummy variables were coded to indicate if a student lives on campus, with a parent, in a fraternity house, or off-campus (on campus serves as the reference group).
Institutional-level variables
These measures were included to account for risky environments. Student population measures were created by aggregating individual-level variables by school so that they reflect the proportion of students with each characteristic in each school. Included are the proportion of students who reported engaging in binge drinking on campus, illicit drug use on campus as well as the proportion of students who live off campus. General risk taking behaviors indicates the proportion of students who in the past 12 months did not take safety precautions while driving, biking, motorcycling, and skating. 2 Finally, four measures of school characteristics were also included. School size measures the population of students enrolled at a particular school as whether the school has 20,000 students or more (coded as 1) or fewer (coded as 0). Also measured are whether a school is private (coded as 1) or public (coded as 0) and is a 2-year (coded as 1) or 4-year (coded as 0) institution. 3 Finally, a measure of the size of the locale that the school is located was included and is coded as 1 if the town has a population of 500,000 or greater and coded as 0 if it is smaller.
Control variables
Finally, individual-level demographic controls are included. Gender is included such that females are coded as 0 and males are coded as 1 (transgender individuals were excluded from analyses because of the small number in the total sample). Year indicates if a student is in the first year of college (coded as 1) or not (coded as 0). Sexual orientation indicates whether a student self-identifies as bisexual, gay, lesbian, or unsure (coded as 1) or heterosexual (coded as 0). Race/ethnicity indicates if a student self-identifies as White (the referent category), Black, Hispanic, Asian, Multiracial, or another race.
Analytical Procedure
Analysis for this project included a series of multilevel binary logistic regression models conducted in Stata 15 using the mixed effects command. This strategy allows for the modelling of the clustering of data to produce standard errors that account for the potential correlation between cases from the same school, and the examination of both individual and contextual effects on victimization. Models were run to examine three types of victimization: violent, sexual, and stalking victimization. Models were split to examine Canadian and U.S. college schools separately, allowing for the examination of whether the same factors are related to the three types of victimization for students enrolled in both school locations. Full models with all students including interaction terms between U.S./Canada and all other independent and control variables were also run to examine if factors were differentially related to U.S. and Canadian student victimization.
Results
Table 1 presents the descriptive statistics for the U.S. and Canadian college student samples. In the U.S. sample, 18.8% reported violent victimization, 6.9% reported sexual victimization, and 5.5% reported stalking victimization. Prevalence of all three victimization types was higher in the sample of Canadian college students: 20.8% reported violent victimization, 7.8% reported sexual victimization, and 5.8% reported stalking victimization. Descriptive statistics regarding individual-level predictors and socio-demographic control variables are similar to those reported in Daigle et al. (2019).
Descriptive Statistics of Sample by College Location, U.S. or Canada.
Regarding school-level factors, to begin, across U.S. schools the mean proportion of binge drinkers at a school is 33% compared to 36% across Canadian schools. The mean percentage of students who use drugs at a school is close to 40% across schools in each sample. The mean percentage of risk-takers at a school is also roughly equal—close to 25%. The average U.S. school has 44% of its students living off-campus compared to 55% of the average Canadian school. Among Canadian students, about 53.6% belonged to student bodies of 20,000 or more compared to about 43.9% of U.S. students. Among U.S. students, close to 32.4% attended a private school compared to only 8.1% of Canadian students. None of the Canadian schools were 2-year institutions, but 12.9% of U.S. students attended one. Finally, 34.7% of Canadian students attended schools located in areas with populations of 500,000 or more, compared to about 29% of U.S. students.
Before conducting logistic regression analyses, we first examined the extent to which level 2 factors account for variation in the three types of victimization. For each type, there was preliminary support for level 2 effects. The amount of variation in the outcome that can be explained at the school-level is demonstrated by the intra-class correlation coefficient. Level 2 factors account for 7% of the variation of violent victimization for U.S. students and 3% of this variation for Canadian students. Level 2 factors account for 9% of the variation of sexual victimization for U.S. students and 3% of this variation for Canadian students. Level 2 factors account for 7% of the variation of stalking victimization for U.S. students and 6% of this variation for Canadian students.
Table 2 presents the results of the multilevel logistic regression analysis predicting violent victimization. Models were run for U.S. and Canadian college students separately. To examine if any variables significantly differed in their influence on violent victimization a full model with all students including interaction terms between country of school and all variables was also run. Before each multivariate analysis was conducted, each of the level 1 variables—individual-level predictors and demographic controls—was group mean centred. Regarding school-level factors—level 2, in the U.S. model, proportion of binge drinkers and attending a 2-year school was related to an increase in the odds of violent victimization while attending private schools decreased them. The proportion of students living off campus was also associated with a reduction in the odds for violent victimization. Only two level 2 factors were significant in the Canadian model: proportion of students living off campus and attending a large school both are associated with a decrease in the odds of violent victimization.
Multilevel Logistic Regression Predicting Violent Victimization.
p < .05. **p < .01. ***p < .001.
The first set of rows indicate which level 1 predictors are related to violent victimization. In the U.S. model, binge drinking, being an underage drinker, drug use, having mental health issues, having a disability, membership in a Greek social organization, involvement in sports, and risk taking behavior were all associated with an increase in the odds of violent victimization. Also, living in a fraternity house and living off campus compared to living on campus was linked to an increase in the odds of violent victimization. Results of the Canadian model were similar to those of the U.S. model, except that Greek social group membership was not significant and living with a parent (relative to living on campus) was associated with an increase in the odds of violent victimization while living in a fraternity house or off campus were insignificant.
The relationship between control variables and violent victimization are also shown. In the U.S. model, males, freshman, non-heterosexuals, Black students, multiracial students, and students of other races faced higher odds of violent victimization than females, non-freshman, heterosexuals, and White students, respectively. However, Hispanic and Asian students faced lower odds than White students. Results of the Canadian model were similar to those of the U.S. model except that being Black and Hispanic were not significant. None of the interactions between country of school and Level 2 variables were significant, which indicates that Level 2 factors operate similarly for U.S. and Canadian college students. One interaction with a control variable was significant—country of school and gender, which shows that effect of gender on violent victimization differs for U.S. and Canadian college students.
Table 3 presents the results of the multilevel logistic regression analysis predicting sexual victimization. Two models are included for the students because both proportion binge drinking and proportion drug use could not be included in the same model due to multicollinearity issues. Table 3 reflects models in which binge drinking is included at level 2.
Multilevel Logistic Regression Predicting Sexual Victimization.
p < .05. **p < .01. ***p < .001.
Only two level 2 factors were significantly related to sexual victimization. In the U.S. model, proportion of binge drinkers at a college was associated with an increase in the odds of sexual victimization. The proportion of students living off campus was linked to a reduction in the odds of sexual victimization. In the Canadian model, the same two level 2 factors were significantly related and in the same direction to the odds of sexual victimization.
In the U.S. model, most level 1 variables were significant. Binge drinking, underage drinking, drug use, having mental health issues, having a disability, Greek social group membership, sports involvement, risk taking behavior, and being single all were associated with an increase in the odds of experiencing sexual victimization while living with a parent or off-campus was linked to a decrease in these odds as compared to living on campus. Results of both Canadian models were similar to those of the U.S. model except that drug use and living off-campus were not significant and living in a fraternity house (compared to on campus) was tied to an increase in the odds of sexual victimization.
Regarding control variables, in the U.S. model, males face lower odds of sexual victimization while non-heterosexual college students and those who are Black, multi-racial or other race (compared to White) face greater odds of sexual victimization. These results are found in both Canadian models except that race is statistically insignificant. When examining whether any of the variables significantly interact with country, no level 2 variable interactions are significant, which suggests that effects are similar for U.S. and Canadian college students. Three interactions with level 1 variables were significant. The effect of binge drinking, living in a fraternity house, and gender are all differentially related to sexual victimization
Table 4 presents the results of the multilevel logistic regression analysis predicting sexual victimization with drug use at Level 2. In the U.S. model, four level 2 variables are significantly related to sexual victimization—proportion drug users, proportion of students who live off campus, proportion of students engaging in risky behavior, and 2-year schools. Only proportion drug users was related to sexual victimization for Canadian students.
Multilevel Logistic Regression Predicting Sexual Victimization.
p < .05. **p < .01. ***p < .001.
In the U.S. model, most level 1 variables are significant. Binge drinking, underage drinking, drug use, having mental health issues, having a disability, Greek social group membership, sports involvement, risk taking behavior, and being single all were associated with an increase in the odds of experiencing sexual victimization while living with a parent or off-campus was associated with a decrease in these odds as compared to living on campus. Results of both Canadian models were similar to those of the U.S. model except that living in a fraternity house (compared to on campus) was connected to an increase in the odds of sexual victimization and living off campus was not significant.
Many of the control variables are related to sexual victimization in the U.S. model. Males face lower odds of sexual victimization while non-heterosexual college students and those who are Black, multi-racial or other race (compared to White) face greater odds of sexual victimization. These results are found in both Canadian models except that race (except for being multiracial) is statistically insignificant. None of the level 2 variable interactions were significant, which suggests that the measured school-level effects are similar for U.S. and Canadian college students. Three interactions with level 1 variables were significant. The effect of binge drinking, living in a fraternity house, and gender are all differentially related to sexual victimization.
Table 5 presents the results from the multi-level logistic regression analysis predicting stalking victimization. In the U.S. model, several level 2 variables are related to stalking. Attending 2-year schools was linked to an increase in the odds of stalking victimization while proportion of students living off-campus and attending a private school is linked to a decrease in the odds. However, in the Canadian model, proportion of students living off-campus, the proportion of students engaging in risky behavior, school size, and location size were significant. Proportion of students living off-campus and school size were negatively related to stalking victimization, while proportion of students engaging in risky behavior and size of location were positively related to stalking victimization.
Multilevel Logistic Regression Predicting Stalking Victimization.
p < .05. **p < .01. ***p < .001.
In the U. S. model, many of the level 1 variables are significant. Binge drinking, underage drinking, drug use, having mental health issues, having a disability, sports involvement, and risk taking behavior increased the odds of stalking victimization. Living off campus reduced the odds of stalking victimization, and living with a parent increased the odds. Results for the Canadian model differ in that Greek affiliation was significant and involvement in sports was not. None of the location living variables were significant in the Canadian model.
Several control variables are also significant in the U.S. model, being female, non-heterosexual, Black, Hispanic, multi-racial or other race (compared to White) increased the odds of stalking victimization. Results for the Canadian model differed in that none of the race variables were significantly related to the odds of stalking victimization. One of the interaction terms with level 2 variables is significant—the interaction between country and school size indicates that the effect of school size on the odds of stalking victimization is different for U.S. and Canadian students. In addition, the interaction term of country and Greek membership, country and involvement in sports, and country and being multiracial are all significant.
Discussion
Research examining school-level factors and their relationship to victimization among college students has produced equivocal results. The difference in findings regarding whether school-level factors are related to victimization and which matter may be attributed to the difference in samples and the types of victimization explored. In addition, no comparison across U.S. and Canadian colleges and how school-level factors may influence victimization has been performed. Our research adds to the existing literature by doing that—we examine both individual- and school-level factors and their relationship to victimization. Further, we do so by examining three types of victimization: violent, sexual, and stalking victimization.
Our analyses reveal three key findings. First, a greater percentage of Canadian college students are victimized than U.S. college students for three types of victimization. In this way, Canadian students are more likely to be violently victimized, to be sexually victimized, and to experience stalking. These findings are in line with previous work showing Canadian college students faced greater risks of both violent and sexual victimization (Daigle et al., 2019). This finding suggests that an exploration such as ours as to why Canadian students face greater risks is needed.
Second, unlike some past studies (Cantor et al., 2017; Cass, 2007; Fisher et al., 1998), the current study found that some school-level factors play a significant role in predicting college student victimization. School-level predictors were found in both the U.S. and Canadian college student samples, although only three factors were found in both samples—none of them consistently. Among variables tested in both samples, proportion of binge drinkers at a school, proportion of students living off campus, and attending a private school played the largest roles in predicting U.S. student victimization while proportion of students living off campus and attending a larger school played the most consistent roles in predicting Canadian student victimization. In the U.S. models, 2-year school was related to all three types of victimization as was proportion of students living off campus. Proportion binge drinkers and attending a private school were significantly related to two types of victimization while the proportion of students engaged in risky behavior and proportion drug users each were related to only one (stalking and sexual, respectively). Proportion of students living off campus was related to each type of victimization for Canadian students, while school size was related to both violent and stalking victimization. Thus, it appears that school-level factors are more consistently related to U.S college student victimization. We should note that only one interaction term between country of school and a level 2 variable was significant to indicate that school-level variables operate differently for U.S. and Canadian students. School size reduced the effects of stalking victimization in the Canadian sample but was not significant in the U.S. sample. These findings suggest that level 2 factors, when relevant, are largely influencing victimization the same for U.S. and Canadian students. It also suggests the need for future analysis with a larger number of Canadian schools to explore additional level 2 factors.
We found that student body characteristics stood out as victimization correlates. Proportion of binge drinkers was the most consistent significant variable overall. Having a high proportion of binge drinkers at a school was associated with an increase in the risk of violent and sexual victimization in the U.S. and sexual victimization in Canada. Also, having a high proportion of students who use illicit drugs at a school was linked to an increase in the odds of sexual victimization in both the U.S. and Canada. These results are consistent with the RAT/L perspective, as having a high proportion of students rendered more vulnerable to victimization through substance use increases opportunities for motivated offenders (Cass, 2007; Mohler-Kuo et al., 2004; Stotzer & MacCartney, 2016). Another finding is that having a high proportion of students at a school who live off campus decreased the odds of all victimization types for U.S. and Canadian students. This finding is consistent with arguments that residence hall and/or Greek living environments can foster risk factors for victimization (Armstrong et al., 2006). It may be that having more students living off campus reduces the likelihood that motivated offenders come into contact with suitable targets (Franklin et al., 2011). In other words, students may spend less time on campus around other students they can harm or be harmed by.
Third, characteristics of the school itself also were related to some types of victimization. Attending a private school was associated with a decrease in the odds of violent and stalking victimization in the U.S. but not in Canada, though the interaction terms suggests these effects are not significantly different. Although speculative, it could be that private schools are more likely to have policies and practices in place that reduce victimization risk (Craig et al., 2017; Riley, 2005; Vanderwoerd & Cheng, 2017). For example, in our study, private colleges were more likely to be religiously-affiliated than public schools, which may shape the activities of students and the culture of the institution. Private and public schools differed across U.S. and Canada as well—private schools in Canada were more likely to be larger in size. Also, attending a larger school decreased the odds of violent and stalking victimization in Canada but not in the U.S. Larger schools may create diffuse social networks, where students do not closely interact or know other students well. Research demonstrates that first-year students who commute (which is related to school size) have fewer friends than other students (Buote et al., 2007). If this is the case, then students may not come into close contact with other students beyond the classroom; thus, reducing the likelihood that they would be violently victimized or stalked by their peers. Indeed, in our data, larger schools have a greater proportion of students living off-campus, which shapes victimization risk. It is also possible that the nature of large schools in the U.S. is different than what is found in Canada. Many large schools in the U.S. have a well-developed Greek system and have sports teams that shape the culture of the institution. Because these factors are tied to victimization, it makes sense that attending large U.S. colleges does not reduce risk of victimization (Gallagher et al., 1994).
Also, attending a 2-year school increased the odds of each type of victimization among U.S. college students, but this variable was unavailable in the Canadian sample. This finding may be because 2-year schools have more first-year students attending them. In our data, about a third of students attending a 2-year school were freshmen. Research shows that college students are more likely to be victimized during the early years of college (Cranney, 2015; Kimble et al., 2008; Lasky et al., 2017). Also, in our data, those students who attend 2-year colleges are younger on average than those who attend 4-year schools. With a greater proportion of young students at these colleges, it could be that there are a greater number of motivated offenders, as perpetration of crime decreases with age (Barry, 2010; Richardson & Budd, 2003). More research is needed to test the consistency of private vs. public school, school size, and 2-year schools as correlates of victimization and to understand why they are associated with an increase in risk for victimization. Further, research that can incorporate measures of culture to see if U.S. colleges and Canadian colleges have different contexts that may be linked to victimization risk needs to be undertaken.
The findings concerning rates of substance abuse and living on-campus have practical implications. Intensive education programs may inform students of the dangers of alcohol and drugs in an attempt to prevent abuse. They may also inform students of harm reduction strategies in drinking alcohol (Daigle et al., 2016; Johnson et al., 2017). Both may reduce vulnerability and exposure to potential offenders. However, these alone will not sufficiently reduce victimization. As Wechsler et al.’s (2001) finding that schools banning alcohol tend to have lower sexual assault rates suggests, control policies may also reduce victimization. Strategies having to do with policing/security and environmental crime prevention are also important. However, institutions must consider possible inadvertent effects of policies and practices, even those that are well-intentioned. As Armstrong et al. (2006) point out, students may respond to strict rules by partying in dangerous off-campus environments. Thus, it is important to provide students with stimulating alternative activities. Regarding risk associated with residence and Greek life, institutions must also look for ways to counteract circumstances like peer influences favorable to aggression and the mingling of potential victims and offenders, such as exposing students to others with more diverse backgrounds and increasing guardianship capabilities at parties. Because the effects of most of the variables did not differ significantly across county, efforts to reduce victimization through targeting risk factors such as problematic drinking should be taken in both U.S. and Canadian schools.
Our study is not without limitations. As noted above, although we include a wide variety of level 2 variables to capture school characteristics, unavailable or limited data prohibited the inclusion of some school-level factors which theoretically predict victimization. Thus, our study includes a very limited application of ecological systems and RAT/L to understand institutional culture and structure differences between Canadian and U.S. schools and how these differences impact victimization risk.
We were unable to assess institutional culture characteristics such as school climate and norms, values, beliefs, and attitudes concerning drinking, sex, and various forms of violence. We suspect that these variables “set the stage” for individual risky behavior and offending. Some of the variables we included such as proportion of students who binge drink we argue are proxies for cultural norms on drinking, but it is possible that the overall drinking behavior of students on a particular campus is driven by other factors. Thus, although we found that proportion binge drinking and proportion drug use increased the odds of some types of victimization, we cannot definitively explain why this is the case.
We were also unable to assess differences in policies and practices concerning alcohol and drug use, victimization prevention, victims’ assistance, and victimization disclosure that may impact students’ chances of being victimized and willingness to report it. Stricter interventions may reduce both opportunities and motivation to victimize (they signal that violence will not be tolerated), victim services may increase awareness of and sensitivity toward the problem, and policies and practices may facilitate or inhibit students’ willingness to report victimization. Also, because there were fewer Canadian schools included in the data compared to the U.S. data, we were unable to include the full variety of structural school variables. For example, we could not include the measure of whether a school was a 2- or 4-year school or if religious-affiliated in our analysis because all of the Canadian students in the sample were attending 4-year schools and few were attending a religious-affiliated school.
Additionally, although we examined three types of victimization outcomes, the data do not include measures of other types that students may experience such as theft. Finally, we also are unable to establish causality. Our data are cross-sectional; thus, our findings are correlational in nature and do not allow us to examine the causes of victimization across school type. Thus, future research is needed to examine direct measures of culture, more school-level variables, policy and reporting differences, and other types of victimization and employ longitudinal designs to better explore causality.
In conclusion, our study provides a glimpse into the factors that increase risk for three types of victimization on U.S. and Canadian college campuses. We find several college-level factors to be related to an increase in the odds of victimization, with some factors emerging as relevant for U.S. but not Canadian college students and vice versa. However, our findings suggest that, although there were variables relevant in one model and not the other, that the effects operated similarly for most Level 2 factors. In this way, the current study lays the foundation on which future research can build. To develop effective policies and interventions surrounding college student victimization, research is needed on 2-year intuitions in the Canadian context and how it compares to the U.S as well as the role of institutional policies surrounding drinking, sex, and violence and their link to victimization. Addressing these gaps will enhance our understanding of the similarities and differences of college student victimization across these two nations and how best to address the problem.
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.
