Experimental Design Details
Descriptive Analysis
In our study, we use our survey to collect unique data on bystanders' beliefs and support the effect of information treatments on various outcomes related to individuals' willingness to help victims of sexual harassment. The descriptive methodology involves a simple OLS regression, where the outcome variable for individual i (Yi) is regressed on the various beliefs about sexual harassment prevalence, costs and good practices, using only the control group. Beliefs will be assessed by presenting three sets of questions to the respondent regarding sexual harassment in the workplace. Firstly, they will be shown a definition of sexual harassment and will be asked estimate its prevalence in the workplace: variable “Percentage of Victims”, coded from 1 to 100. We will also construct a variable the deviation from the true share, from -30 to 70 (true share being approx. 30%). Secondly, they will quantify who suffers the most in harassment cases, whether it is the victim or the perpetrator, by indicating the percentage of cases where each party experiences the most significant consequences: variables “Costs”, coded similarly from 1 to 100, and the deviation as well. Finally, respondents will be asked to assess their knowledge about good practices in cases of sexual harassment. This will take the form of a test on the dos and don’ts when faced to a sexual harassment case: variable “Test Score” coded from 1 to 6.
Yi = B1*Percentage of Victims1i + B2* Costs 2i + B3*Test Score3i + µs (1)
The beliefs variables will allow first to draw descriptive conclusion about the type of beliefs that are important to foster to create a supporting environment for victims.
Causal Analysis
In our study, we use a Randomized Control Trial (RCT) to estimate the effect of information treatments on various outcomes related to individuals' willingness to help victims of sexual harassment. The empirical methodology involves a simple OLS regression, where the outcome variable for individual i (Yi) is regressed on the treatment dummies (Treatment1i, Treatment2i, Treatment3i) and other control variables (Xi), with the omission of the control groups as the reference group.
Yi = B1*𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡1 i + B2*𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡2 i + B3*𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡3 i + µs (1)
The treatment dummies (Treatment1i, Treatment2i, Treatment3i) take a value of 1 if the respondent receives a specific information treatment and 0 otherwise. Treatment 1 is giving the right information about the prevalence of sexual harassment will change their perception of the pervasiveness of sexual harassment, influencing individuals' willingness to help victims of sexual harassment.
With Treatment 2, we can challenge the belief that the cost ration between victims and perpetrator favour the victims with quantitative information about the percent of victim who felt that they suffered more consequences will affect individual behaviour when faced with a situation of harassment. Finally, providing a set of simple and easily applicable actions in case they are confronted with a harassment situation can strengthen bystanders’ knowledge and give them the confidence to support victims.
Finally, treatment 3 incentives individual to act regardless of others’ behaviour and to be aware of pluralistic ignorance, and to acknowledge that
We expect certain characteristics, such as gender, age, gender norms, and previous experience of sexual harassment, to play a significant role in influencing individuals' responses to the treatments and their willingness to support victims. By examining these heterogeneity factors, we can gain valuable insights into how different subgroups react to the information treatments, contributing to a comprehensive understanding of the study's findings.
Experiment Design
The survey methodology consists of several steps, as depicted in Figure 1. Initially, respondents answer questions about their individual characteristics, such as age, education, household composition, gender, employment type, and the nature of their workplace. This "Baseline" part of the survey also includes questions about gender norms from the World Value Survey and experiences of harassment and workplace discrimination.
Next, respondents are presented with three sets of questions that capture their "prior beliefs" regarding the extent of sexual harassment, the costs for victims and abusers, and good practices for addressing such situations.
Afterward, respondents are randomly assigned to three treatment groups and two control groups. The control groups will initially be pooled together for the analysis, and later separated for robustness checks, ensuring that any effects observed are not solely due to information provision itself (Haarland et al., 2022). This step helps to verify that the results are not influenced by respondents spending more time on the survey due to information exposure.
Finally, all respondents will be asked the same set of questions following randomization. First, they will be asked about working conditions broadly, ensuring that their responses to the primary outcomes are not direct reactions to the treatment and helping to address potential sociability desirability bias. Lastly, they will be presented with the questions described in the "primary outcome" section, which are focused on their willingness to allocate personal resources to support causes related to workplace violence.
By following this comprehensive survey paired with an RCT methodology, we aim to obtain valuable insights into how information treatments impact individuals' beliefs and behaviours regarding sexual harassment in the workplace and their willingness to help potential victims.