A Washington Post/Kaiser Family Foundation survey found that 20 percent of current and recent female college students living on or near campus report being sexually assaulted while attending school. How did the survey arrive at this statistic?

Measuring the prevalence of sexual assault is a tricky task in surveys for two reasons. The first is definitional: Asking plainly whether a person has been sexually assaulted can produce unreliable results, as people disagree about what exactly that entails. The second is that respondents might not be willing to report a highly sensitive event such as sexual assault in a survey. In both cases, the most common potential error is that prevalence of sexual assault will be underestimated.

“Terms like sexual assault and rape — there are legal definitions for what those things are, and people’s understandings of what those things are often are not consistent with the legal definitions,” said Christopher Krebs, a researcher at RTI International who led federally sponsored studies of sexual assault on college campuses. “If you use those terms, you’ll dramatically undercount those incidents.”

To avoid ambiguity about terminology, the Post-Kaiser survey asked respondents about specific types of unwanted sexual contact or sexual assault and the scenarios in which they occurred. Respondents were each provided the same definition of relevant types of sexual contact, which included five different types of assault: forced sexual touching, oral sex, sexual intercourse, anal sex and sexual penetration with a finger or object. (Exact question wording and descriptions can be found after Question 49 in the survey instrument.) These descriptions were modeled off of the 2007 Campus Sexual Assault Study sponsored by the National Institute of Justice and led by Krebs.

After receiving this definition, respondents were asked about circumstances in which they might have experienced these types of unwanted sexual contact. These included physical force or threat of force (Q52) and sexual contact when incapacitated and unable to provide consent or stop it from happening due to being passed out, asleep, drugged or drunk (Q54), including only incidents they are certain occurred. Both of these scenarios fit common legal definitions of sexual assault, and they are the source of the Post-Kaiser 20 percent estimate of sexual-assault prevalence among college women.

Beyond these situations, the survey also asked about attempted sexual contact by physical harm, suspected (but uncertain) contact while incapacitated, and sexual contact stemming from non-physical threats or promises of rewards. In total, 25 percent of college women reported either sexual assault or another type of unwanted sexual contact.

Reluctance of respondents to report sexual attacks is another challenge for measuring prevalence accurately.

Sensitivity is “based on the shame a woman or a man might feel, the fear of being isolated from your social groups,” according to Ken Rasinski, an expert on measuring sensitive issues in surveys. “There is a lot of publicity around the fear of not being taken seriously or being humiliated by officials who you might report this to or other authorities who might be trying to wash it under the table.”

In surveys about sexual assault, who is asking can matter in whether a victim is willing to report an incident. Rasinski, who did not review the survey prior to publication, described a worst-case scenario in asking about rape as “you’re a man and you’re talking with a woman, and you come off as an authoritarian man and you say ‘have you ever been raped recently,’ and it may be even worse in-person asking the same thing.”

“Even under ideal circumstances, we still expect under-reporting,” Rasinski says of surveys measuring sexual-assault experiences. But he notes that good measures of sexual-assault experiences will respect a person’s privacy and self-esteem, carefully establishing a time frame in which incidents occurred and avoiding terms that might bring up fears of stigmatization. An ideal way to accomplish this is through an anonymous, self-administered survey, where respondents fill out an online survey or paper-and-pencil questionnaire in private.

The Post-Kaiser survey was conducted by telephone (almost entirely reaching respondents on their cellphones) with professionally trained interviewers, as it was determined to be the best method to reach a random national sample of current and recent college students. Nearly all female respondents were matched to female interviewers to reduce sensitivity of reporting an unwanted sexual experience, and most male respondents were interviewed by men. Questions about unwanted sexual experiences were asked later in the questionnaire to ensure interviewers had an opportunity to establish rapport with respondents.

Interviewers also received training on how to administer the questions accurately and handle respondents who were apprehensive about answering questions about sexual assault. For instance, trigger warnings were given noting that wording is explicit and could be upsetting and offering contact information for support services if needed. pre-testing of the questionnaire found interviewers could administer questions clearly and respondents were able to answer them. Breakoffs were relatively rare for respondents who began the survey: 89 percent of respondents completed the full interview. Most breakoffs occurred during the middle of the 25-minute interview, while about one-tenth of breakoffs occurred following questions about individual sexual-assault experiences.

A potential limitation to the Post-Kaiser survey includes “non-response” bias, or the idea that survey respondents are different from non-respondents. As Krebs of RTI noted after reviewing the survey, low response rates can threaten a survey’s ability to represent a population. The survey used random sampling procedures to identify young adults from national weekly polls of adults, for which the typical response rate is 8 percent to 10 percent. This is a common response rate for Random Digit Dial telephone surveys. To combat potential non-response bias, the final sample was weighted to population estimates for current and recent students at colleges and universities.

A second potential limitation noted by Krebs is question order. Questions about sexual-assault experiences were asked after many questions about campus life, effective prevention methods for sexual assault and what qualifies as consent for sexual activity as well as demographics. While this ordering was aimed at building respondent rapport prior to asking sensitive questions on sexual experiences, Krebs notes it also could have primed respondents to think only about their preconceived definition of sexual assault rather than the definition provided in the survey. “Since it used lots of those terms, it sent the message to respondents that this is what you’re interested in,” Krebs said. Asked how this might impact results, Krebs suggested “it could have resulted in a suppression of the prevalence estimate.”

Despite challenges in measuring sexual-assault prevalence, one striking finding is the similarity in estimates from the Post-Kaiser survey reaching a nationally representative sample of respondents from more than 500 schools and the 2007 Campus Sexual Assault Study conducted at two unidentified universities in the Midwest and South. The Post-Kaiser survey found 20 percent of current and recent female students report being assaulted by force or while incapacitated, compared with 13.7 percent in the 2007 survey among current college students only. In a more apples-to-apples comparison, the Post-Kaiser survey found 21 percent of seniors and those who have left school said they were assaulted, compared with 19.8 percent of seniors in the two-university study. Given the margin of sampling error of both surveys, the difference between these two estimates is not statistically significant.

Peyton M. Craighill contributed to this report.

Get updates on your area delivered via e-mail