Exploring the Effects of Gender Inequality on Labor Market Decisions and Cognitive Outcomes using Virtual Reality

We use Virtual Reality (VR), a reality‑enhancing technology that provides a unique opportunity to study how inequality affects the outcomes of interest by recreating safe yet realistic environments. Specifically, we will recruit bachelor’s and master’s students at Maastricht University to participate in a virtual reality experience in which they attend the open day of a fictional company. Participants will be exposed to one of the following conditions: (i) the inequality condition, in which the company’s leadership team is fully composed of men; (ii) the equality condition, in which the company’s leadership team is gender-balanced. The VR environment presents participants with information about the company’s structure, employees, and decision-making processes. The content, narrative, and tasks are identical across conditions, with the only difference being the gender composition of the leadership team.
This manipulation is designed to vary participants’ exposure to gender inequality in a controlled and immersive setting.

2026-04-20

2026-12-18

Primary outcomes for STUDY 1 include three behavioral measures: application decisions, information‑seeking behavior, and voice/feedback behavior.
Primary outcomes for STUDY 2 include two cognitive measures: memory and creative problem solving.

STUDY 1:
Behavioral Response. Our primary outcomes capture students’ behavioral responses following exposure to information on gender inequality. We measure these responses using three binary indicators derived from participants’ choices within the survey:
(i) Application decision: an indicator equal to 1 if the respondent chooses to submit an application to the company, and 0 otherwise.
(ii) Information-seeking behavior: an indicator equal to 1 if the respondent chooses to ask questions to the HR team, and 0 otherwise.
(iii) Voice/feedback behavior: an indicator equal to 1 if the respondent chooses to express questions or concerns about the company’s structure to the management team, and 0 otherwise.

STUDY 2:
(i) Memory. We will measure memory through the recall accuracy of the reminder and recall group. We will establish accuracy by comparing the score they originally gave to the idea in the evaluation task and the new score they arrive at in the memory task. We expect that participants exposed to gender inequality recall accuracy will be lower.
(ii) Creativity. To measure creativity we will use two tasks, one open and one closed task. Both creativity tasks have been used previously and are considered reliable measures of creative thinking (Charness and Grieco, 2019).
- Open creativity task. Participants will be asked to reflect on the following question and write their answer in a maximum time of two minutes: “If you had a talent to invent things just by thinking of them, what would you create and why?”. Participants will be informed that their responses will be evaluated based on their level of creativity, including dimensions such as originality, innovativeness, and elaboration. Responses to the open-ended task will be assessed by at least four independent human assessors. Each response will be rated on a scale from 0 (not creative at all) to 10 (highly creative), based on originality, innovativeness, and elaboration. The final creativity score will be constructed as the simple average of ratings across graders.
- Closed creativity task. Participants will be asked to choose a combination of numerical operations to start with the number 27 and reach the number 6. They can perform different types of operations such as addition, subtraction, multiplication, division, exponential, factorial, logarithm and had a maximum of two minutes to write as many numerical operations as they could. Also in this case, evaluation will be conducted by human assessors. Responses will be rated by independent assessors based on standardized criteria of originality and innovativeness.
We will create an index based on the average performance of participants in these two tasks. For the sake of completeness, we also report results for each individual indicator.

Secondary outcomes for STUDY 1 include fluid intelligence, projected self-efficacy, projected belonging, perceptions of fairness
Secondary outcomes for STUDY 2 include motivation, aspirations, and affective responses

STUDY 1:
Fluid intelligence. We measure cognitive performance using a short version of Raven’s Progressive Matrices, a widely used test of fluid intelligence developed by Bilker et al. (2012). The outcome variable is the number of correctly solved items (0–9).

Projected self-efficacy. We measure participants’ perceived organizational self-efficacy in the company using an adapted version of the Occupational Self-Efficacy Scale (Rigotti et al., 2008). Example items include statements such as “If I am confronted with a problem at this company, I could find several solutions” or “I feel prepared for most of the demands in this company.” Responses are recorded on a scale of 1 (strongly disagree) to 5 (strongly agree). We construct a standardized index of projected self-efficacy by averaging responses across all items, after coding them such that higher values indicate higher perceived self-efficacy. The index will be standardized to have mean zero and standard deviation one in the control group.

Projected sense of belonging. Using a modified version of Jena and Pradhan (2018)’s work belongingness scale, we aim to quantify the extent to which participants can see themselves integrating into the company environment they were exposed to in the VR scenario. Specific items include sentences like “I would be able to work in this company without sacrificing my principles” or “Being a part of this organisation would inspire me to do more than what is expected”. Responses a recorded on a scale of 1 (strongly disagree) to 5 (strongly agree). We construct an index of projected sense of belonging by averaging responses across all items, after coding them such that higher values indicate a stronger sense of belonging. The index will be standardized to have mean zero and standard deviation one in the control group.

Fairness. We measure participants’ perceptions of fairness and gender inequality using a set of survey items capturing both general views and company-specific assessments.
General fairness and inequality attitudes. Participants will be asked to indicate their level of agreement with the following statements on a 5-point Likert scale (1 = strongly disagree, 5 = strongly agree):
i) Differences in opportunities between men and women in the Netherlands are too large;
ii) Opportunity gaps between men and women in the Netherlands are inevitable;
iii) The government should implement policies to reduce differences in women and men's opportunities.
Company-specific fairness perceptions. Participants will be asked two additional questions about fairness and their views about inequalities in the company:
iv) How fair do you think this company is? Answers are recoded on a scale from 1 (very unfair) to 5 (very fair);
v) Do you think differences in opportunities between women and men are a serious problem in the company? Answers are recoded on a scale from 1 (Definitely not) to 5 (Definitely yes);
All items will be coded such that higher values consistently reflect greater concern about inequality and lower perceived fairness. In particular, items (ii) and (iv) will be reverse-coded. We construct a standardized fairness index by averaging responses across all items after harmonizing their direction. The index will be standardized to have mean zero and standard deviation one in the control group.

STUDY 2:
Emotions. We measure participants` emotional response to the VR scenario using the Scale of Positive and Negative Experience (SPANE) (Diener et al, 2009). Specifically, we ask participants: “During the virtual reality experience, to what extent did you feel the following emotions: positive; negative; good; bad; pleasant; unpleasant; happy; sad; afraid; joyful; angry; contented”. Answers to each statement can range on a 1-5 Likert scale (not at all; a little; moderately; a lot; a great deal). We construct two indices by summing the respective items: a positive emotions index and a negative emotions index. Each index ranges from 6 to 30, with higher values indicating stronger positive or negative emotional responses, respectively. In addition, we construct a balance score, defined as the difference between the positive and negative emotion indices, as a summary measure of overall affect.
Aspirations. Using the achievement aspirations subscale of the Career Aspiration Scale (Gregor & O'Brien, 2016), we aim to capture how important career development is for participants. Participants are presented with 6 items (e.g. I want to be among the very best in my field) and are asked to indicate from 0 (not very true of me) to 4 (very true of me) how much they resonate with each item.
Motivation. Participants will be presented with 10 lines of text containing multiple zeros and ones. They will be instructed to count how many zeros are present in each line, with a time limit of 40 seconds. The task has been widely used as an approximation of individuals’ willingness to exert effort (e.g., Abeler et al., 2011). The outcome variable is defined as the total number of correctly counted lines (continuous measure ranging from 0 to 10), which captures variation in effort provision.
In this case as well, all indices will be standardized to have a mean of zero and a standard deviation of one in the control group.

We use Virtual Reality (VR), a reality‑enhancing technology that provides a unique opportunity to study how inequality affects the outcomes of interest by recreating safe yet realistic environments. Specifically, we will recruit bachelor’s and master’s students at Maastricht University to participate in a virtual reality experience in which they attend the open day of a fictional company. Participants will be exposed to one of the following conditions: (i) the inequality condition, in which the company’s leadership team is fully composed of men; (ii) the equality condition, in which the company’s leadership team is gender-balanced. After completing the VR experience, students will be asked to respond to a series of questions and complete several tasks aligned with our outcomes of interest. A detailed mapping of procedures to outcomes for Study 1 and Study 2 is provided separately in the two analysis plans.

Not available

Participants will be randomly assigned to either the inequality or equality condition using a computerized randomization procedure ensuring balanced assignment across conditions. Specifically, assignment is implemented such that participants are allocated to the condition with the fewest participants at the time of sign-up; in case of equal numbers, assignment is random with equal probability (p = 0.5).

Individual survey participants

No

250 participants, treatment is not clustered

250

STUDY 1: The study will be conducted within a total sample of approximately 250 participants, with approximately 120 participants per condition in the inequality and equality conditions, respectively.
STUDY 2: The study will be conducted within the same total sample of approximately 250 participants, with a minimum of approximately 60 participants per condition across the four experimental conditions (Inequality × Recall, Inequality × Reminder, Equality × Recall, Equality × Reminder).

Study 1 and Study 2 are conducted within the same overall sample of approximately 250 participants. The two studies differ in their primary outcome measures and task structures but are implemented within the same experimental sessions and share the same randomization procedure. The unit of analysis is the individual participant. The total sample size is constrained by laboratory capacity (N ≈ 250). Assuming a balanced design and under standard assumptions of a two-sided test with α = 0.05 and power = 0.80, the minimum detectable effect size for pairwise comparisons between conditions is approximately Cohen’s d = 0.50–0.55 in Study 1 and Cohen’s d = 0.45–0.50 in Study 2. No clustering at the session level is assumed in the baseline power calculations. These power calculations are based on the main outcomes of interest in each study. Drawing on Murphy et al. (2007), who detected a significant effect of gender inequality on participants’ decision to join a scientific conference (d ≈ 0.7), our sample should be enough to detect the expected effects in Study 1. For Study 2, the power calculations are based on evaluation outcomes derived from participants’ assessments of candidate profiles. Prior evidence suggests that memory constraints can generate substantial gender biases in evaluation contexts. For example, Miserocchi (2023) shows that, when decision-makers rely on memory, female candidates are significantly less likely to be recommended compared to observationally identical male candidates, with large effect sizes. While the present study is powered to detect moderate effects (Cohen’s d ≈ 0.45–0.50), these estimates are conservative relative to the magnitude of effects observed in related settings.