Anticipated Discrimination and Gender Stereotypes in Job Applications

We will run the following treatments in the worker experiment. For each treatment,
we have four alphabetical indicators which are in order:
1. (Math/Emot) - Describes the real-effort task. Math=Math memory task and
Emot=emotional recognition memory task.
2. (U/K) - Describes whether the workers’ genders are Unknown or Known to the
manager/recruiter.
3. (N/M) - Describes whether the manager’s gender is Neutral or known to be Male
to workers/recruiter.
4. (Man/Rec) - Describes whether there is just a Manager or if there is a Recruiter
and a Manager.
We run the following treatments.
T1: (Math-U-N-Man) In treatment T1, workers perform the math-memory task,
workers and managers are blind to each others’ genders and only a manager is present.
2
T2: (Math-K-N-Man) In treatment T2, workers perform the math-memory task,
worker gender is known to the manager but not vice-versa, and only a manager is
present.
T3: (Emot-U-N-Man) In treatment T3, workers perform the emotion-recognition
task, workers and managers are blind to each others’ genders and only a manager is
present.
T4: (Emot-K-N-Man) In treatment T4, workers perform the emotion-recognition
task, worker gender is known to the manager and only a manager is present.
T5: (Math-K-M-Rec) In treatment T5, workers perform the math-memory task,
worker gender is known to the manager, the manager’s gender is revealed to be male
and a recruiter is included who makes the hiring decision on behalf of the manager.

2025-12-01

2025-12-19

The primary outcome of interest is worker’s tournament entry decision in the third
(out of 3) round where a hiring manager or recruiter selects a winner based on a noisy
signal of performance.

The hypotheses are specified in the attached document.

Workers' first-order beliefs on managers' selection decisions

Each worker is asked to report their belief on how many managers selected the worker with the second-highest noisy signal when the score difference is less than three points between the best and the second-best worker. For half of the workers, the best worker is male and the second-best is female; for the other half, the best worker is female and the second-best is male.

The detailed hypotheses are specified in the attached document.

Participants in our experiment will work in either of three roles: workers, managers, or recruiters.
Workers have to complete several rounds of real-effort task under varying payment schemes and difficulty. They then have to choose one of two payment schemes for a final round of play. Depending on their choice, they receive a fixed payment per point to score, or a higher payment per score conditional on another participant's decision.

Managers see several sets of workers and ability signals of theirs. For each set of workers separately, they then have to select one worker that they hire. If they hire the best-performing worker, they receive a bonus payment.

Recruiters see the same sets of workers and their ability signals as managers. For each such set, they then have to anticipate the Manager's decision on whom to select. If they are successful in this, they receive a bonus payment.

We run three experiments: the workers experiment, the managers experiment and the
recruiters experiment.
1
In the workers experiment, workers complete three rounds of a real-effort task,
namely finding matching pairs of cards in decks of 16 cards. Workers are paid their
earnings from one of the three rounds chosen at random after the experiment. The
rounds vary in task difficulty and how workers are paid (piece-wise or tournament).
In the manager experiment, each manager evaluates 5 randomly matched worker
groups. In each worker group, the manager observes the round 2 performance scores
of the four workers. Depending on the treatment the manager might also be informed
about the gender of the workers (through the display of gendered rather than neutral
avatars). The manager chooses to hire one of the four workers in each group. The
manager receives a bonus of £1 only if the selected worker has the highest Round 3
performance out of all workers in the worker group. Out of the 5 worker groups, one
manager decision is randomly implemented for payout of the manager.
In the recruiter experiment, each recruiter evaluates 5 worker groups, where the
composition of the groups are identical to the groups the managers saw in the same
treatment. As in the manager experiment, in each worker group, the recruiter observes
the round 2 performance scores and the gender of the four workers and the gender of
the male manager (through gendered avatars). The recruiter selects one of the four
workers in each group. The recruiter receives a bonus of £1 if their selection matches
the manager’s selection for this group. Out of the 5 worker groups, one recruiter
decision is randomly implemented for payout of the recruiter.

Random assignment to treatment by a computer.

Individual level, with blocking by gender.

No

9540 individuals

9540 UK-based participants on Prolific (as we don't have clusters).

900 men and 900 women as workers, and 90 managers in each of our five treatments.
Treatment 5 will additionally feature 90 recruiters.

Unit: percentage of workers choosing the competitive payment scheme (separately for women and men) Standard deviation: 0.5 Percentage: 6.3pp (MDES) Baseline entry rate: 35.5% (women), 51.2% (men) Percentage: