The Cost of Rejection: Understanding Gender Dynamics in Leadership Aspirations

2025-12-07

2026-04-30

Leadership application: binary variable (applied = 1, did not apply = 0), assessed across rounds
Willingness to lead: 0-10 points scale measure, assessed across rounds
Leadership status (candidates only): accepted / rejected, assessed across rounds
Beliefs about outcome attribution and number of candidates: perceptions of the reasons for acceptance or rejection and number of candidates, assessed across rounds (incentivized)
Confidence beliefs: both relative and global confidence measures (incentivized)
Individual performance (incentivized)

Hypothetical Leadership Status: For participants who did not initially apply for the leadership role, I provide a hypothetical leadership outcome, that is, information on whether they would have been accepted or rejected had they applied. This manipulation aims to test whether receiving such a counterfactual signal influences subsequent decisions to apply for leadership positions
Explicit gender biases
Risk attitudes
Previous knowledge about the task
Previous leadership experience
Demographic information - age, gender, occupation, field of study for students

I build on the framed “survival” task used by Born et al. (2022) to study willingness to lead. Specifically, my experiment employs a modified version of this paradigm to measure individual performance and confidence in two preliminary rounds, followed by group-based tasks in which a leader is appointed. Participants are informed that the leader's appointment is determined through a lottery weighted by individual performance. Over multiple rounds, they have repeated opportunities to put themselves forward for the leadership role and, if selected, to determine their group’s outcome. Observing how individuals modify their willingness to lead after acceptance or rejection allows for an analysis of whether and why men and women react differently to these experiences.

Part 1 - Individual Phase
Part 1a: Participants are presented with 5 pairs of items from the ”Lost at Sea” task and must select the item they believe is most critical for survival. We also ask them to guess their score, and rank compared to a reference sample of other participants.
Part 1b: Participants complete the whole task again with new pairs of items and a different survival scenario. We also ask them to guess their score, and rank compared to a reference sample of other participants.
This individual phase establishes baseline measures of ability and confidence.

Part 2 - Group-based Phase
Part 2a
Group Formation: Participants are placed in groups, their anonymity is maintained. They are informed that a leader will be selected to represent the group during a second round of the task. The leader’s answers alone will determine the group’s final payoff for that round. Participants are informed that leader selection will be based on the outcome of a lottery in which each candidate’s chance of being selected depends on their performance scores from the two initial tasks (Part 1a and 1b). Participants with higher performance scores will have higher chances of being selected as leader, but no one is guaranteed to be chosen. If no one applies, all group members will automatically be included in the lottery using the same rule—so even if you choose not to apply, you may still be selected as leader if no one else applies.
Willingness to Lead elicitation: Each participant is asked to indicate whether they wish to apply for the leadership position + their WTL on a 0-10 points scale. We also ask a guess about the number of candidates.

Group Task: While we process leaders’ selection, all participants complete the task again with new item pairs. Only the elected leader’s answers determine the group’s final payout. They do not get any feedback on outcomes for this task until the end of the experiment.

Leader Revelation: Participants learn about whether they have been appointed as leader or not for this round.
Among those who applied: “Your application has been rejected/accepted for this round”
Among those who did not apply: “You were not appointed as leader for this round” + hypothetical feedback “Had you applied you would have been accepted/rejected”. (this feedback is obtained by adding the participant to the existing candidate pool and running the election again)

Attribution Question: Participants are asked about their beliefs on how much of the outcome was due to luck vs. performance. They are also asked to guess again the number of candidates in the round.

Confidence Elicitation: We elicit global and relative confidence the same way as we did in Part 1.

Part 2b
Same overall structure as Part 2a. Participants are told that they are going to complete the same task again, but that a new leader will be appointed for this task. Again, they are asked whether they want to apply, how willing they are to lead on the scale and their guess about the number of candidates. The same rule applies for the leader’s selection and revelation.

Group Task: They complete the task with a new set of items. Participants are asked again about their beliefs on how much of the outcome was due to luck vs. performance, number of candidates and confidence.

(Hypothetical) Part 3c
Participants are told to imagine that they could complete the same task again. Again, they are asked whether they would apply for the leader’s role, and how willing they would be to lead on the scale.

Post-Experiment Survey Participants complete a survey measuring risk preferences, the belief in good luck scale, explicit gender biases, previous knowledge about the task, previous leadership experience, demographics and some open-ended questions about their decisions during the experiment.

Feedback and Payment
The leaders’ answers are revealed, and the group’s overall outcomes are shared. They also have the opportunity to view the experts' solutions to better understand their scores.

Final payoffs consist of:
A £5 Prolific participation fee.
A bonus equal to the performance in one randomly selected question from one randomly selected Part (Part 1a, Part 1b, Part 2a or Part 2b. If Part 2a or 2b are selected, only the leader’s answers determine the payout. They receive £2 if their answer to this question is correct.
£0.5 bonuses for on the beliefs questions (14 in total: 6 confidence questions, 2 attributions, 4 number of candidates)
A £0.5 for being the leader (up to £1 if someone is selected twice)

Anonymity is maintained throughout the experiment. This ensures that gender remains private information, ruling out additional barriers to women’s willingness like concerns about backlash for deviating from gender expectations.

Not available

Randomization is implemented by computer within the experimental software. All random assignments (including leader selection lotteries, and feedback conditions) are determined algorithmically using a reproducible random seed.

While group-level outcomes exist, all random assignments (grouping, leader selection, hypothetical feedback, bonus draws) are implemented at the individual level through computer randomization.

No

450 participants

450 participants

450 participants

Based on budget constraints, I plan to start by recruiting approximately 450 participants (≈75 groups of 6). Based on estimates of application rates (≈62% of women and ≈78% of men applied in an earlier online study using a similar design) I expect ≈311 applicants, and ≈236 rejected applicants (≈118 per arm). The average applicants per group is ≈4.15, so the design effect for rejected applicants is DE = 1 + 2.1×ICC. Under α= .05 and 80% power, this yields an achievable MDE of ≈18–19 percentage points gender gap in re-application and ≈1.17–1.21 points gap on the 0–10 WTL scale. Given that baseline rates of application are unequal (≈62% of women and ≈78% of men applied in an earlier online study using a similar design), we will invite more women than men to achieve a balanced number of applicants, we will invite more women than men, aiming for around 150 female and 150 male candidates (therefore inviting 250 women and 200 men). We will monitor application rates during piloting and adjust recruitment targets if needed. Based on prior data, willingness to lead (WTL) has a standard deviation of roughly 2.5 points on a 0–10 scale. With 150 rejected applicants per gender, this sample provides 80% power (α = 0.05) to detect gender differences in rejection effects of about 0.9 points (Cohen’s d ≈ 0.36). For binary re-application outcomes, the design is powered to detect gender gaps of about 18–20 percentage points.