Sexual Violence Among Postsecondary Students: No Evidence that a Low Response Rate Biases Victimization or Perpetration Rates in a Well-Designed Climate Survey

ABSTRACT

Purpose

The purpose of this study was to assess the accuracy and representativeness of common census-sampled campus climate surveys given the potential for misestimating sexual violence (SV) rates on campuses due to low response rates and self-selection bias in research (mixed findings in previous research).

Method

We compared SV rates obtained from a census-sampled campus climate survey with a lottery draw (a common method for collecting campus SV data) with those obtained from a gender-stratified random sample survey with individual incentives.

Results

We found no evidence that census-sampled campus climate surveys misestimate SV: our low response rate census-sampled survey produced very similar rates to our high response rate random sample survey.

Discussion and Conclusion

Our research suggests that less costly and labor-intensive census-sampled surveys, when well-designed, produce sufficiently accurate and representative SV estimates on campuses despite their lower response rates.

KEYWORDS:

• Sexual violence
• sexual assault
• university students
• random sample
• census sample
• climate survey

Acknowledgments

The authors would like to acknowledge Hio Tong Kuan and Carolyn Rauti for their support collecting and analyzing data for the census-sampled survey described in this paper.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. Students who do not identify as women or men; includes (based on study identification) nonbinary, two-spirit, genderqueer, gender fluid, androgynous, and trans students (however, counts of women and men in this study may include, based on self-reports, trans women and men). We compare only overall victimization rates between the two samples for gender expansive students because cell sizes within the victimization categories were too small and because no gender expansive students from the random sample reported any perpetration.

2. Fall semester = September to December. The field period was October 21 to November 18, 2019 (4 weeks).

3. Response options: “Yes, my data is as honest and accurate as possible” and “No, my data should not be used.”

4. Spring semester = January to April (referred to as the winter semester at Canadian postsecondary institutions). The field period was February 24–April 5, 2020 for the primary sample (6 weeks) and March 9–April 5, 2020 for the hold sample (4 weeks). Under 6% (roughly three weeks) of the 12-month SES-SFV and SES-SFP reference period was impacted by campus changes due COVID-19 for only the 174 participants (18%) who submitted the survey after March 16, 2020 (when classes were canceled for three days and then moved online). Thus, the timing of the survey was very unlikely to have impacted the SV rates in Sample 2.

5. All students of unknown gender who ultimately participated identified as women or men (not another gender) in our survey.

6. The undergraduate student population on campus is consistently smaller (~5%) in the spring semester than in the fall semester, likely due to first year dropouts between semesters.

7. Women and men were directed to the same items as in Sample 1. Gender expansive students were directed to modified versions of each item that used gender and anatomy inclusive terminology (see Jeffrey et al., 2022).

8. We did not use survey weights to account for differences in response rates across age and gender categories because weights can decrease precision (Pike, 2007) and would not change the conclusions of our sample comparisons (since results are reported by gender and the victimization rate did not differ significantly by age for either sample).

9. As with the total population comparison, we excluded students of unknown gender from the Registrar sample in this comparison (since we did not have these categories in our survey for comparison). Although the counts of women and men in the Registrar sample of 2,000 are likely underestimates since all students of unknown gender who ultimately participated identified as women or men (not another gender) in our survey, the results of the comparison here do not change if we redistribute the unknown gender students to the women and men categories in the Registrar sample of 2,000 based on their proportions in the respondent sample.

10. The victimization and perpetration rates for gender expansive students cannot be relied on given the small cell sizes. The perpetration data, in particular, highlight how small cell sizes can markedly impact estimates and comparisons (in this case, the small cell sizes meant a comparison of 25% and 0% perpetration rates).

11. Some studies cited here measured SV using slightly more restrictive definitions and slightly longer or shorter reference periods.

Additional information

Funding

This research was undertaken, in part, thanks to funding to the second author from the Canada Research Chairs Program and the University of Windsor.