A sample of married college-educated Saudi men will be recruited to participate in an experiment in Riyadh, Saudi Arabia. Participants will be transported to a hotel conference center to take an online survey in rooms of 30 participants each. All 30 participants in the room will take the survey concurrently on their personal smartphone devices using the platform Qualtrics.
While participants are asked to provide their name, phone number and email on a sign-in sheet at the beginning of each session, an important feature of our experimental design is that participants are never asked to provide their full names and phone numbers during the survey itself to maintain a condition of anonymity with regard to individual survey responses. Instead, we ask participants to provide the last three digits of their phone number which we use to assign them to either treatment or control conditions during the randomization we perform, so that we can later reference a respondent's phone number using information provided on the sign-in sheet to determine their treatment status without ever linking their contact information to their individual survey response.
The survey begins by collecting basic demographic information. Participants are then presented with three statements regarding the labor market in general: “In my opinion, the minimum wage for Saudis (SAR 3000) should be kept at its current level”; “In my opinion, the current unemployment insurance system (Haafez) is good for the economy”; “In my opinion, Saudi nationals should receive privileged access to job vacancies before expatriate workers.” This is followed by two statements regarding participation of women in the labor force: “In my opinion, women should be allowed to work outside of the home”; “In my opinion, a woman should have the right to work in semi-segregated environments”.
For each statement, participants are asked whether they agree or disagree with the statement (elicitation of private beliefs). For the two statements regarding participation of women in the labor force as well as the statement about keeping the minimum wage the same (which will serve as a point of comparison for the female labor force participation questions), participants are also asked to estimate how many of the other 29 participants they expect would agree with the statement. The two other general labor market questions are only included in the elicitation of private beliefs since their primary purpose is to ensure that participants treat the survey as a general labor market survey during the elicitation of beliefs. To incentivize participants to estimate the beliefs of others accurately, participants are told that the respondent who guesses most accurately will receive a $20 Amazon gift card (to be distributed by email after the conclusion of the survey).
Randomization: After all participants have completed this section of the survey, half of the respondents are given feedback about the responses of the other participants in the room to the two statements about female labor participation shown (treatment group). For each statement, participants in the treatment group are shown a chart displaying the proportion of respondents who agreed and disagreed with the statement. The other half are not given this feedback (control). Randomization will be at the individual level, assigned randomly based on the last three digits of their phone number.
Both treatment and control participants are then asked to read a short passage about a startup which provides an online platform aiming to connect job-seeking Saudi women with employers. The passage contains basic information about the online platform as well as the company’s outreach and mentorship initiatives.
Outcome variables: After reading the passage, participants are asked to choose between receiving a $5 Amazon gift card or the opportunity to sign up their wives for access to the company's platform and services. If participants choose to sign up, they are asked to provide the contact information of their wife. This information will be sent to the company and wives who have been signed up by participants in our survey will be contacted directly by the company's representatives. Those who do not provide their wife's name and number (even if initially choosing the sign-up option) will be treated as having chosen not to sign up. Taking into consideration our expectation that some participants may instead give their own contact information instead of their wives', we treat the decision to sign-up using a phone-number that does not match the three trailing digits of the respondents own number as our outcome of interest.
If beliefs that others are more conservative are inhibiting otherwise progressively-minded respondents from signing up their wives for the job matching service, we expect participants in the treatment group (those receiving feedback on the private beliefs of others) to exhibit a higher sign-up rate than participants in the control group (who receive no feedback).
At the end of our survey, we ask participants to estimate the percent of private sector firms in Saudi Arabia that have semi-segregated work environments. If there is a positive effect of treatment on expected percent of firms with semi-segregated environments, this provides evidence consistent with female labor demand being a function of social norms regarding female labor force participation.
Our main outcome, to be used in all the specifications that follow, is a dummy indicating whether the subject signed up the wife and provided her contact details, for the company's services. A supplementary outcome, to probe for mechanisms, is subjective beliefs on the prevalence of mixed-gender work places across all private sector firms in the country, as a proxy for perceived labor demand (thickness) for females.
Our main treatment variable, randomized at the individual level, is whether the subject received information feedback on the aggregate responses to the survey questions (first order beliefs) on the subject's view on female labor force participation (FLP).
First specification: We will first test whether treatment had an average effect on subjects, regressing signup on the treatment dummy. Our main specification includes session fixed effect, and additional specifications will include covariates for demographics and beliefs at baseline. Conditional on finding a significant wedge between average 2-order beliefs and aggregate 1-order beliefs of other participants, our primary hypothesis is that treatment has a positive effect on signup. However, if the distribution of 2-order beliefs is approximately symmetric around aggregate 1-order beliefs, then the average effect may be zero, masked by positive and negative treatment effects across individuals. We will therefore also investigate heterogeneous treatment effects, per below.
Second specification: Our hypothesized mechanism is that the treatment effect is heterogeneous, due to heterogeneity in individual beliefs. Specifically, the treatment effect may be more positive (negative) when the wedge between a subject's 2-order beliefs and the aggregate 1-order beliefs is more negative (positive), all else equal. Therefore, we will measure the sign and size of the wedge for each individual, and test for the interaction between the individual wedge and treatment. Our prediction is that the treatment effect is weakly monotonically decreasing in the wedge, all else equal.
Third specification: A final possibility is that the treatment effect depends on heterogeneity in private 1-order FLP beliefs. The willingness to sign up for the company's services may depend on whether the husband is privately in favor or against women working. Therefore, we will use an interaction specification to estimate whether the effect is increasing or decreasing in the private 1-order beliefs, holding constant the wedge. Our theoretical prediction is ambiguous; on the one hand, the marginal individuals may be those who are the strongest pro-FLP because they are the most willing to signup the wife for labor market search services; on the other hand, they may signup their wives even in the absence of treatment and so the marginal individuals may be those at intermediate levels of pro-FLP beliefs. Moreover, if social learning/persuasion is relevant in our setting, then those with both negative opinions about FLP and low 2-order FLP beliefs might also respond strongly to the intervention.