Productivity under Hostility

We plan to conduct a lab experiment to currently enrolled students from various departments of a Canadian public university to evaluate the impact of hostile environment on individual and group productivity.
The recruitment of students will take place by the TEEL laboratory mailing list.
Participants will be randomly assigned to groups with varying gender composition and working environments. They will be asked to answer a cognitive test, to complete quiz questions and to participate to a contribution game. They will also answer a short questionnaire and complete an incentive-compatible choice experiment to measure their willingness-to-pay for game attributes. Participants' compensation will include both individual-level incentives and group-level incentives, as well as a flat completion fee for the survey experiment.

2025-01-15

2025-06-30

Binary outcome indicating individuals’ contribution decisions, continuous variable indicating individuals’ performance, and group performance, average probability of answering a question correctly across all questions seen in a category.

We will first look at the impact of treatments on contribution decisions and individual performance. Contribution decisions correspond to individual’s decisions to nominate their answer to be the group answer. Individual performance corresponds to individual quiz scores, standardized by topic. Our measure of ability in a category (female-typed and male-typed quizzes) for each individual is the average probability of answering a question correctly across all questions seen in the category. We next will measure group performance, as measured the sum of scores of quiz that were selected through nomination game across rounds.

Final financial payoff, hostile messages sent and received by team members, individual’s self-reported stress level, perceptions of working conditions, self-reported retention, and willingness-to-pay for each game attribute.

We will use final financial payoff to measure the gender wage gap. We will measure the level of endogenous hostility proxied by the number of hostile messages sent and received by team members. To measure individual’s self-reported likelihood of participating in the experiment again, we will use a five-item Likert scale. We will measure individual's self-reported stress level with a scale from ``very low'' to ``very high'', and how much you enjoyed the teamwork with a scale from ``not at all'' to ``very much''. In the second part of the experiment (incentive-compatible choice experiment), we will measure individuals' self-reported willingness-to-pay for each game attribute.

Participants will be randomly assigned to groups with varying gender composition and working environments. They will be asked to answer a cognitive test, to complete quiz questions and to participate to a contribution game. They will also answer a short questionnaire and complete an incentive-compatible choice experiment.

Not available

Randomization will be done by a computer.

1. Random group assignment: After completing Part 1—the baseline performance measurement— individuals are randomly assigned to a group of 4. The gender composition of the group varies accordingly to two treatments: 1F3M, 3F1M. In case we can't accommodate these numbers during a session, we will have gender-balanced teams too. Each member picks a gender neutral nickname at the group formation stage. The group stays the same for the remainder of the experiment.
2. Work environment treatment: The quality of the work environment in the experiment is randomized to be either friendly or unfriendly and is randomized at the group level. Participants are able to communicate with their team members and the variation in work environment is introduced by the content of messages. Participants are able to select one or more messages among twelve pre-specified messages. They can also decide not to send any messages. These messages are noisy in the sense that a computer sometimes augments a group member and sends a message on their behalf. Practically, this can be thought of as miscommunication. We implement this with a decision rule where there is a computer rule that may accidentally send one of a message on the participant’s behalf. This can happen whether the participant has selected a message or not, with a 15% probability. If the participant decides to send a message and the computer sends a message too, the participant’s message will be sent with a 50% probability. We implement a manipulation check at the end of the experiment in which we ask how much participants enjoy working with their group members.

Yes

40 groups of 20 participants each.

800 participants

20 groups in friendly, 20 in unfriendly environments.

For power calculations, we use pilot data regarding performance to quiz question. Our design allows us to measure an effect of 0.052 percentage points from a baseline value of 0.530, or 22% of a standard deviation using standardized score by topic. We benchmark this against Bordalo et al. (2019)'s measure of average performance, i.e the average probability of answering a question correctly across all questions seen in the category (mean=0.7, sd=0.1), and to benchmark the effect size, we picked the mean difference in probability of answering correctly across gender in the ten-question bank (0.036).