Task divisions in teams and the (in)efficiency of gender stereotypes: The impact of communication

Subjects are paired after they have performed the two tasks described in the abstract. They are not informed about their performance. Each pair consists of one male and one female (mixed-gender pairs). Both members of a team vote independently whether the male’s number of correct answers in the rotation task and the number of the female’s correct answers in the emotion recognition determines the sum of their correct answers, or vice versa. If both vote for the efficient task division that maximizes the sum of their correct answers, both of them receive a bonus. If exactly one of them votes for the efficient task division, both of them receive the bonus with 50% probability. If none of them votes for the efficient task division, they do not get the bonus. If both task divisions yield the same sum of correct answers, both divisions are treated as efficient.

There are two treatments:
In treatment N, participants vote directly after they are paired.
Treatment C (Communication): Identical to Treatment N, except that pair members can communicate fvia a chat window before voting.
Thus, the intervention varies whether communication is allowed prior to the individual votes on the task division that determines whether subjects receive the bonus or not.

We have three main hypotheses. In all three hypotheses we control for efficiency gains, which we define as the difference in the number of correct answers with the efficient and the inefficient allocation of task performances.

H1. Stereotype-consistent efficient task divisions are more likely to be chosen than counter-stereotypical efficient task divisions.
H2. Communication increases the probability of choosing efficient task divisions.
H3 (if we find support for H1 and H2). Communication reduces the impact of stereotypes on efficient task divisions.

In addition, we have the following secondary hypotheses:
H4. The probability of choosing efficient task divisions increases with efficiency gains.
H5. Communication increases the accuracy of estimating the partner’s performance.

2026-07-02

2026-08-31

The primary outcome (dependent variable in our regressions) for Hypotheses H1-H4 is the probability with which efficient task divisions are chosen, depending on gender stereotypes and treatments. The primary outcome for H5 is the accuracy of the estimation of the partner’s performance.

In addition to H1-H5 we explore whether there are gender differences
• in choosing efficient task divisions,
• in the impact of stereotypes on choosing efficient task divisions,
• in the impact of efficiency gains on choosing efficient task divisions,
• in the impact of communication on choosing efficient task divisions,
• in the impact of communication on performance estimates

We have no hypotheses for these gender differences.
We also document actual gender differences in performance and use non-parametric tests for significance.

For the communication treatment, we descriptively summarize whether participants used the chat for task-relevant communication. This analysis is descriptive and is not part of our preregistered hypothesis tests.

The online experiment follows a between-subject design structured in two parts. In Part I all participants perform the same two tasks, one task stereotyped to males (identifying rotated pictures) and one to females (recognizing emotions). Participants are not informed about their performance. After each task, participants answer four self-evaluation questions about their own performance and one question on performance differences of men and women at the task.

In Part II, participants are paired and these pairs are randomly assigned to one of the two treatments explained above. At the end of the instructions for Part II, we ask two comprehension questions. Participants who answer more than one question incorrectly, cannot proceed to Part II.

After the actual experiment, we elicit the participants’ beliefs about their own and their partner’s performance in both tasks.

Not available

Participants are assigned to treatments and pairs based on their entry to the experiment. Participants who self-report that they do not identify as male or female, or who prefer not to disclose their gender, can participate in Part I, receive the same compensation for Part 1 as other participants, but cannot proceed to Part II.

Individuals are randomly assigned to pairs.

No

no clustering, see number of observations

The experiment is run on Prolific. Our target sample consists of 1,200 participants who complete both parts of the experiment. Because some participants may complete only Part I (because they failed more than one comprehension question; are not successfully matched with a partner; do not identify as male or female, or prefer not to disclose their gender), the final sample of participants completing both parts may be smaller. Data collection ends as soon as 1,200 participants completed both parts or 1,400 participants completed Part I.

600 participants in treatment N and 600 participants in treatment C (but may be lower; see the previous point).