Gender Discrimination, Competition, Negotiation, and Self-Promotion

We conduct a large-scale online experiment to examine how gender discrimination causally affects women’s behavior in three important work situations: competition, negotiation, and self-promotion.

We designed an online experiment featuring 3,825 women residing in the UK to study how gender discrimination causally changes their behavior in three work situations: competition, negotiation, and self-promotion. Participants start by taking a test for which they receive a piece rate per correct answer. Afterwards, we disclose the piece rate of another participant—which is (i) equal, (ii) unequal without a stated reason, or (iii) unequal because of gender. Participants then take part in one of three work situations: competition, negotiation, and self-promotion. The analysis plan in the supplementary material provides a detailed description of the intervention.

2023-12-10

2024-03-01

Two measures for each of three work situation (competition, negotiation, and self-promotion). Participants’ choices in the three situations are first measured before they learn about their performance in a test, and then a second time after they learn about their performance.

In the competition situation, participants choose whether a competition with another individual determines their payment. In the negotiation situation, participants submit a wage request to another individual. In the self-promotion situation, participants assess their own performance for another individual. The analysis plan in the supplementary material provides a detailed description. The analysis plan in the supplementary material provides a detailed description of each primary outcome.

Secondary outcomes are factors potentially underlying the effect of discrimination. Factor 1 is confidence, factor 2 is anger, factor 3 is risk aversion, factor 4 is deservingness, factor 5 is gender norms, and factor 6 is motivation.

The analysis plan in the supplementary material provides a detailed description of each factor.

See analysis plan in the supplementary material.

We conduct an experiment on the online platform Prolific to measure the impact of gender discrimination on women’s behavior. All participants in the study are women. We also use men and women in separate studies. We randomize participants in one of three treatments—EQUALITY, INEQUALITY, and DISCRIMINATION—and one of three work situations—COMPETITION, NEGOTIATION, and SELF-PROMOTION. Participants complete eight stages and are matched in some stages with individuals from separate studies.

The experiment features eight stages.
1. Participants provide socio-demographic information.
2. They learn that they earn a piece rate of 15 pence per correct answer in a 10-question math and science test and then complete the test (incentivized).
3. We inform participants of the piece rate per correct answer that we pay another individual who took the same test. We mention that both their piece rate and the piece rate of the other individual were determined without looking at test performances. Each participant is randomized into one of three treatments: EQUALITY, INEQUALITY, and DISCRIMINATION. In EQUALITY, participants learn that we pay the other individual the same piece rate as they receive. In INEQUALITY, participants learn that we pay the other individual a higher piece rate of 25 pence. We provide neither a reason for inequality nor any information about gender. In DISCRIMINATION, we inform participants that we pay the other individual a higher piece rate of 25 pence because he is a man.
4. We elicit three confidence measures: participants' absolute and relative confidence about their test performance and their beliefs about women's performance (all incentivized). The relative measure is relative to a UK sample (resembling the UK adult population) and the women measure is about women in this UK sample.
5. We randomize each participant in one of three work situations (all incentivized): COMPETITION (task four of Niederle and Vesterlund, 2007), NEGOTIATION (building on the design of the online study in Exley et al., 2020, and the study in Hernandez-Arenaz and Iriberri, 2023), and SELF-PROMOTION (Exley and Kessler, 2022).
6. We inform participants about their absolute and relative performance on the test.
7. Participants take part in the same situation as in stage five again (incentivized), but this time they know their test performance. This factors out the effect of confidence about one’s test performance.
8. We measure other potential factors underlying the effect of discrimination—anger, risk aversion (incentivized for one of two measures), deservingness, gender norms (incentivized), and motivation—and ask additional survey questions.

The analysis plan in the supplementary material provides a detailed description of the design.

Randomization done by Qualtrics software.

Individuals are randomly assigned to one of three treatments in one of three situations (nine possibilities).

No

NA

3,825 women. ***Update: we started data collection on 19/12/2023 and only around 2,800 participants completed the study before the initial end date of 31/12/2023. We continue data collection until reaching 3,825 women or we stop it if we have not reached this number of participants on 01/03/2024.

425 women in each of three treatments in each of three situations (nine possibilities).

We estimate that we have a statistical power of 80% to detect a difference of 0.19 standard deviations between two treatments in our main analysis (see analysis plan in the supplementary material for details).