Social norms and career decisions: Experimental evidence on gender and parenthood biases in STEM and non-STEM jobs

The intervention consists of an online survey experiment conducted with a representative sample of 2,500 Italian adults. Each participant is randomly assigned to one of eight vignette scenarios that vary along three dimensions: the gender of the protagonist (male/female), parental status (has children/does not), and occupational sector (STEM/non-STEM). In the vignette, the protagonist is offered a high-level promotion that comes with higher pay and prestige but also longer hours and frequent travel.
Participants are asked two key questions: (1) whether they would personally recommend the protagonist accept the offer (first-order beliefs), and (2) what percentage of similar people (in age, gender, and region) they believe would do so (second-order beliefs). The second-order belief question is incentivized monetarily for accuracy. This design allows the study to identify biases rooted in gender and parental status stereotypes, and to explore the role of social norms and perceived consensus in shaping decision-making.

2025-04-28

2025-06-30

Our primary outcomes capture participants’ beliefs about the social advisability of accepting a career promotion, as described in the vignette. Specifically, we measure:
• Second-order belief (SOB): the respondent’s estimate of the percentage of people similar to them (in terms of age, gender, and geographic region) who would recommend accepting the promotion.
• First-order belief (FOB): the respondent’s own likelihood of recommending acceptance of the promotion.
Both outcomes are recorded on a continuous scale from 0 to 100.
In addition to analyzing each outcome separately, we will examine the gap between SOB and FOB. We hypothesize that, on average, participants will perceive others as more conservative than they actually are—i.e., second-order beliefs will tend to reflect greater hesitation toward the promotion than participants’ own recommendations—suggesting the presence of pluralistic ignorance.

We are collaborating with a professional survey company in Italy to conduct an online survey in which respondents are asked what they think others would do (second-order beliefs) in hypothetical workplace scenarios involving a high-level career promotion and then what they would do (first-order beliefs) in the same situation.
Each respondent is randomly assigned to one of 8 vignette scenarios, varying the protagonist’s:
i. gender (male or female);
ii. parental status (has children or does not have children);
iii. occupational sector (STEM or non-STEM).
This results in 8 (2x2x2) total scenarios, structured as follows:

STEM Non-STEM
No children Scenario 1-F Scenario 2-F
Has children Scenario 3-F Scenario 4-F

STEM Non-STEM
No children Scenario 1-M Scenario 2-M
Has children Scenario 3-M Scenario 4-M

In each vignette, the respondent reads a short scenario in which the protagonist, a worker in a dual-earner household, is offered a high-level promotion. The promotion offers a significant salary increase and higher prestige, but it also involves more responsibility, longer working hours, and frequent business travel.
For example, in the scenario where the protagonist is a woman with no children working in a STEM field, the vignette reads:
“Paolo and Elisa are both 35 years old, have graduated with honors in Computer Engineering, and work for an IT consulting firm. They have no children, no one to help them with household chores, work about the same number of hours, and earn about the same salary.
One day, Elisa is offered the opportunity to become a partner in the company. This choice would guarantee her a leadership role, greater prestige, and a significant salary increase. At the same time, it would also mean more responsibility, a significant increase in working hours, and frequent business trips.
Elisa can choose between:
• Option 1: Accept the offer and embark on a brilliant career as a partner.
• Option 2: Decline the offer and decide to maintain her current work-life balance.
We asked people similar to you in age, gender, and geographic area of residence what advice they would give Elisa.”
Participants are then asked:
• Second-order beliefs: What percentage of people similar to them (in age, gender, and region) they think would recommend that the protagonist accept (or decline) the offer.
• First-order beliefs: How likely it is that they would personally recommend to the protagonist to accept/decline.
Responses to the second-order belief question are incentivized with a small monetary reward if the estimate is sufficiently close to the true average response.
To make each vignette more realistic to the respondent and facilitate identification with the protagonist, we also match the protagonist’s education level to that of the respondent (i.e., with or without degree). Prior research shows that such adjustments increase the plausibility of the scenario and therefore incentivize more realistic responses.

Each participants is randomly assigned to one of the 8 vignette conditions (2x2x2 design). Randomization is performed at the respondent level within the survey platform.

individual respondent

No

We do not have cluster, we have 2500 individuals to be randomly assigned to one of the eight treatments

2500

about 312 participants per treatments