The Under-Reporting of College Sexual Assault and The Role of Social Norms

I designed a 2x2 online survey experiment in which subjects complete two primary tasks on social norms and experience with sexual assault. In the social norm task, subjects complete an incentivized social norm elicitation exercise. I present subjects with two statements: one suggesting that women reporting sexual assault are partly responsible for what happened, and the other questioning whether the woman is lying about the assault. Participants are asked to predict the percentage of previous survey respondents (men and women) who agreed with each statement. In the experience task, subjects answer binary questions on their experience with either sexual assault perpetration or victimization. All men are asked about sexual assault perpetration, and all women are asked about sexual assault victimization. I vary: 1) whether survey responses on victimization and perpetration are recorded using direct elicitation or “hard-garbling,” and 2) the order in which subjects complete the experience and the social norm tasks. In the hard-garbling survey technique, a subset of responses to a binary question (on either victimization or perpetration of sexual assault) are randomly recorded as “Yes” responses, regardless of the true response. This randomization provides respondents with plausible deniability, as researchers – or anyone else viewing the data – cannot know for sure whether a “Yes” response was truly a “Yes” or an intended “No.” I also vary the salience of individual perceptions of social norms related to reported assaults by experimentally manipulating whether subjects complete the social norm task before or after answering the questions on personal experiences of sexual assault.

Edited on 2/12/2025: This design has now been modified so that all subjects will complete the experience task first.

2024-12-17

2025-06-01

1. Normalized recorded reports for sexual assault victimization – rape (women only)
2. Normalized recorded reports for sexual assault victimization – other nonconsensual sexual contact (women only)
3. Normalized recorded reports for sexual assault perpetration – rape (men only)
4. Normalized recorded reports for sexual assault perpetration – other nonconsensual sexual contact (men only)

The victimization and perpetration questions are binary. In this experiment, the rate of garbling is set to 20% for hard-garbling treatments; this rate is 0% for direct elicitation treatments. For individuals in hard-garbling treatments, I observe a garbled response rather than the intended response. I will define normalized recorded reports as: (recorded response – the garbling rate) / (1 – the garbling rate).

1. Misperception of women’s agreement with victim-blaming
2. Misperception of men’s agreement with victim-blaming
3. Misperception of women’s agreement with lying
4. Misperception of men’s agreement with lying

Actual social norms will be defined as the share of respondents with agree with the social norm statements. An individual’s misperception of the norm will be calculated as the difference between their perceived norm and the actual norm.

The survey experiment will involve individuals aged 18-27 that indicate being either current undergraduate students or recent college graduates. These individuals are recruited via Prolific.

The survey consists of two primary tasks on social norms and experience. The social norm task contains questions on agreement with social norm statements and predictions of what other people think. The predictions will be incentivized, and subjects will be paid $0.25 for each correct guess. The experience task contains binary questions on sexual assault experiences. Note that all men will be asked about sexual assault perpetration, and all women will be asked about victimization. The basic experiment is a 2x2 design in which I vary which survey technique is used and what order subjects complete the tasks.

I randomly assign participants into one of four treatment conditions:
• T1: Less Salient Social Norms, Direct Elicitation
• T2: Less Salient Social Norms, Hard-Garbling
• T3: Salient Social Norms, Direct Elicitation
• T4: Salient Social Norms, Hard-Garbling

Edited on 2/12/2025: This design has now been modified to only include T1 and T2. Preliminary data collected in December 2024 from 802 women showed that changing the ordering of the social norm task fails to make social norms more or less salient in the experiment, and introduces noise. The data collection will therefore proceed with a fixed order, i.e., the social norm task will come after the experience task.

In the social norm exercise, subjects are asked separately about men and women’s beliefs and are asked to predict men and women’s agreement with two separate statements:

1. If a woman experiences unwanted sexual contact, she may be at least somewhat responsible for what happened.
2. Women who say they were sexually assaulted sometimes lie. For instance, they may just regret otherwise consensual encounters.

I randomize the order in which subjects are asked about men and women’s agreement and which statement they see first. Subjects also answer a third question regarding men and women’s perceptions of the prevalence of sexual assault. This question is always asked third:

3. We told these men (women) that sexual assault is considered to be any kind of nonconsensual or unwanted sexual contact, including situations like unwanted kissing or groping. We then asked these men (women) what percentage of women they think are sexually assaulted while in their undergraduate programs. What do you think the average answer of these men (women) was? You will receive $0.25 if your guess is within 2 points

On November 29, 2024, I conducted a small-scale pilot study of approximately 200 individuals for the purpose of incentivizing the social norm questions. In this pilot, subjects were asked to indicate their agreement with the victim-blaming and lying statements and were asked what percent of women they think are sexually assaulted while in their undergraduate programs.

Subjects in the full study will be paid $0.25 for each correct (within 2 percentage point) guess regarding men and women’s answers from the pilot study, for a maximum bonus payment of $1.50.

The rate of garbling in this experiment is set at 20% (i.e., 1 of every 5 responses is automatically recorded as a “Yes” response) for individuals in hard-garbling treatments. This rate is 0% for individuals in direct elicitation treatments. Individuals in hard-garbling treatments are asked 2 incentivized comprehension questions. Subjects will be paid $0.15 for each correct answer on the comprehension questions, for an additional bonus payment of up to $0.30.

In the experience task, subjects will answer two binary questions on either sexual assault victimization or perpetration. The first binary question asks about experience with nonconsensual sexual penetration or oral sex. The second binary question asks about experience with any other type of nonconsensual sexual contact, including situations like unwanted kissing or sexual touching. In the second binary question, subjects are asked to answer only regarding situations that did not involve unwanted sexual penetration or oral sex.

All subjects complete a variety of demographic questions and a social desirability scale at the start of the survey. All subjects complete questions regarding their level of trust in Prolific at the end of the survey.

I plan to target a sample size of 1,600 men and 1,600 women, for a total of 400 subjects in each treatment arm. The study is only available to individuals between the ages of 18-27 that indicate being either current undergraduate students or college graduates. However, obtaining this sample size relies on the availability and responsiveness of the sample in Prolific. Final sample sizes may be smaller due to this restriction.

Edited on 2/12/2025: Changes to the design from a 2x2 to a 2x1 experiment implies that the sample will now be made of 800 women and 800 men.

Randomization is done by the software system.

Individual

No

3,200 individuals

Edited on 2/12/2025: 1,600 individuals

3,200 individuals Edited on 2/12/2025: 1,600 individuals

1,600 men and 1,600 women; 400 men and 400 women in each treatment arm:
• T1: Less Salient Social Norms, Direct Elicitation
• T2: Less Salient Social Norms, Hard-Garbling
• T3: Salient Social Norms, Direct Elicitation
• T4: Salient Social Norms, Hard-Garbling

Edited on 2/12/2025: The revised design only comprises T1 and T2, i.e., the social norm task comes after the experience task for all study participants. Changes in the design from a 2x2 experiment to a 2x1 experiment implies that the sample will now be made of 800 women and 800 men total.