Participation in voluntary competitions and gender discrimination

2020-12-07

2021-01-30

proportion of redistributors transferring 0 of the winner payoff to the loser of the competition
standardized amount of the winner's payoff redistributed to the loser (“transfer”)

We define “transfer” as the actual amount transferred to the loser (0 <= transfer <=6). The standardized outcome variable is defined as transfer/sigma_(male loser). The outcome variables are directly elicited through the experiment.

We use a spectator game to identify gender discrimination in redistribution after a voluntary decision to compete.

We align our design closely to Cappelen et al. (2019), henceforth CFT. We study whether women are put at a disadvantage when competition is voluntary and the task is male-stereotypical.

Similar to CFT, we will also conduct a worker experiment and a spectator experiment, but we will run the spectator experiment first and the worker experiment second. This is mainly for logistic reasons.

We ask the impartial spectators to redistribute between two workers who have both voluntarily entered competition in a male stereotypical task. We vary the gender of the more/less productive worker, in a mixed-gender setting. The more productive worker is assigned 6 Pounds, the less productive worker 0 Pounds, and the spectators can decide to redistribute any amount to the loser if they like.

Since we expect gender to be a moderator variable, we collect samples large enough to be able to estimate separate effects for female and male spectators. Thus we have two treatments (male loser/female winner and female loser/male winner), and run those two treatments for both male and female spectators.

After the redistribution stage, spectators participate in a short survey covering:
Open text - reasons for their redistribution choice,
Beliefs about performance in the current experiment (incentivized),
Beliefs about performance of men and women in a similar experiment / stereotypical thinking
Questions on social norms about competitive behavior and gender (incentivized; adapted from \citealp{krupka2013identifying}),
Beliefs about effort of workers
Beliefs about social norms about losing and gender and
Demographics (age, gender, education, income, political affiliation, etc.).


Workers in the worker experiment, which will be conducted shortly after the redistributor experiment, have to answer trivia questions on sports and games adapted from Bordalo et a.. (2019). After completing the 20 multiple-choice questions, workers can decide whether they want to be paid a (low) piece-wise rate of 5 pence per correctly solved question, or be randomly matched into pairs and compete for payment. If workers decide to compete they are randomly matched in pairs and the worker who solved fewer questions correctly receives 0 Pounds while the worker who solved more questions correctly receives 6 Pounds. However, they are told that a third party will have the opportunity to redistribute these initially assigned earnings between the two workers in a pair.

Randomization is done by a computer.

individual

No

2400 spectators

2400 (1600 men and 800 women)

1200

Our main hypothesis concerns the comparison between the share of non-distributors between treatments. Based on a power calculation in G-Power, we are able to detect an effect of 0.11 standard deviations using a two-sided Fisher's exact test with a sample of 2400 subjects at a power of 80% and an alpha of 5%. Only considering male spectators we can detect an effect of 0.14 standard deviations using a two-sided Fisher's exact test with a sample of 1600 subjects at a power of 80% and an alpha of 5%. Only considering female spectators we can detect an effect of 0.20 standard deviations using a two-sided Fisher's exact test with a sample of 800 subjects at a power of 80% and an alpha of 5%.