Sexual Violence on Campus: No Evidence that Studies Are Biased Due to Self-Selection

Abstract

Numerous research studies suggest that at least one in five female college students is sexually assaulted while enrolled. However, many studies exploring sexual violence prevalence on campus use methodology permitting students to self-select into the study based on interest in the topic (i.e., students receive an email offering them the opportunity to participate in a study on sexual violence). Self-selection may bias these prevalence estimates of campus sexual violence. To explore this issue, we surveyed two samples of college women on their experiences of sexual assault. We recruited Sample 1 in a typical way: by emailing a randomly selected subset of students provided by the university registrar and inviting participation with information about the survey topic. We recruited Sample 2 using a human subjects pool where students in introductory psychology and linguistics courses sign up for studies without prior knowledge about the topic of the research they will participate in (hence greatly minimizing the risk of self-selection). The two samples yielded nearly identical victimization rates. Over a quarter of participants in both our samples had experienced sexual contact without consent, consistent with recent research from the Association of American Universities. College victimization estimates do not appear to be biased by self-selection based on knowledge of the survey topic.

Isn’t high school bad enough already? Rates of gender harassment and institutional betrayal in high school and their association with trauma-related symptoms

“…Second, we did not account for possible confounders like childhood trauma and other types of high school sexual misconduct, which may reduce the share of trauma-related symptoms that are explained by gender harassment and institutional betrayal [ 32 ]. Third, though participants were blinded to the study’s contents and therefore unlikely to be biased by self-selection, we did draw from a sample of college students, which may introduce bias and severely limit generalizability [ 17 , 33 ]. Fourth, our survey lacked attention checks, which may have inflated the reliability and degraded the validity of the data [ 34 ].…”

“…Students receive course credit for completing studies, and they are permitted to end their study participation at any time. Because of the design of the Human Subjects Pool, participants are not aware of the topic of the research prior to signing up to participate, which protects against biased self-selection [ 17 ]. This process was approved by the institution’s Office of Research Compliance (Institutional Review Board), as evidence suggests that trauma research is considered minimal risk [ 18 , 19 ].…”

Sexual Harassment at Institutions of Higher Education: Prevalence, Risk, and Extent

“…In interpreting and contextualizing these findings, limitations due to measurement issues, at the intersection of sexual orientation and race, must be noted. The low survey response rate, which is typical of similar studies, can be improved but still represent student groups with minimal response bias concerns (Cook, Heath, & Thompson, 2000; Rosenthal & Freyd, 2018). Increased promotion and recruitment, especially in minority populations, would help with gaining significant statistical power to represent all groups in analysis.…”

A Trauma-Informed Approach to Sexual Violence Research Ethics and Open Science

“…Finally, research should continue to explore the degree to which advertising that a study is about sexual assault actually does bias study samples and results. Rosenthal and Freyd (2018) recruited two samples of college women from the same university, one recruited with and one without knowledge that the study was about sexual assault, and they found nearly identical rates of victimization between the samples. Their finding suggests that sexual assault survivors are not dissuaded from participating in a study that they know is about sexual assault and, if replicated, indicates that vague recruitment language may not be as necessary as is currently believed.…”

“…To that end, we encourage more research on how sexual assault survivors perceive different recruitment methods and specific recruitment language. Rosenthal and Freyd’s (2018) finding that transparent recruitment language may not, in fact, bias a study’s sample merits replication with both college student and community populations. In addition, research is needed on how survivors react to recruitment language that does not tell them the study will be about sexual assault: Do they feel upset, do they feel vulnerable to retraumatization when they sign up for a sexual assault study unknowingly, or do they express no such concerns?…”