Are women less demanding? And if so, why?

The intervention is that subjects can directly determine who gets the larger share of the joint output (as opposed to determining it in a stylized negotation).

(Initially, we wanted to study this intervention jointly with the knowledge whether they performed better or worse than their partner before they decide on the share of the output. We then wanted to separate out the effect of both interventions in case that we found one. When we saw the low registration numbers for our experiment, we decided to only focus on one intervention because we were afraid that we would later not be able to have enough subjects to separate both interventions.)

2022-10-24

2022-10-28

Frequency with which higher share of joint output is demanded

The frequency can be directly observed.

Intensity with which higher share of joint output is demanded

The intensity can be directly observed.

In pairs, subjects engage anonymously in a real effort task. After this task, they can either claim 60% or 40% of the team output for themselves. The design is stripped of any exterior motive. In particular, subjects do not know the gender of their team partner.

Subjects are asked to correctly solve as many as possible of 10 quiz questions. They are paired and the product of correct answers counts as their team output. The team with the highest output gets the largest prize. After the task but before knowing the prize, subjects decide in an incentive compatible way whether they want 60% or 40% of the prize. They can also specify, how much they insist on their proposal. We also elicit what subjects think about their performance in an incentive compatible way.

coin flip

Randomization Unit is one experimental session.

No

1

12 Sessions of about 30 subjects, leading to a total of about 360 subjects. We are trying to ensure that half of the participants are male and the other half female by stratifying and reserving half of the places for each gender.

180 subjects control
180 subjects joint intervention
(only if interventions turn out to jointly have an effect, we will extend the study and run a treatment with the first intervention only in order to separate out both effects)

if we are able to fill all our sessions and half of the subjects are male, we have about 90 male and 90 female subjects in each treatment. For a gender gap to be detected with a power of 95% and a significance level of 5% using a one-sided Mann-Whitney-U test (when the original is a normal distribution), the minimal detectable effect size would have to be about 0.5. Our interest is, of course, also whether this gender gap changes across treatments. As a lower benchmark, we may look at the power of the treatment effect within each gender. With roughly 90 observations in each treatment using once more a Mann-Whitney-U test and requiring a power of 95% and a significance level of 5%, we get the same detectable effect size of roughly 0.5.