Team-Up to Move-Up: Team Interactions and Social Mobility

Participants will complete a series of individual and team tasks on Play Together, a platform designed to assess and enhance teamwork and leadership skills among secondary and tertiary education students through performance-based tasks. The platform was developed by one of the co-PIs. We will assess teamwork abilities and study team interactions with the platform with students at flagship university in Colombia.

The tasks on the platform were selected following the existing literature on teamwork in organizational studies (Woolley et al., 2010), psychology (Hall & Watson, 1970), and economics (Weidmann & Deming, 2021). Specific examples of the tasks include the Guilford test of alternative uses, where participants come up with ideas about how to use an object; reproducing art, where participants reproduce the pattern of a color grid; and Raven's Progressive Matrices, where participants identify the missing element that completes a pattern. Other tasks are specifically associated with the measurement of teamwork, including "Lost at Sea" and "Lost on the Moon", where participants first individually and then in teams need to rank fifteen objects in order of importance to survive. To ensure the reliability and validity of the measurement of teamwork and leadership skills, we will complement these tasks with academic tests compatible with the coursework of the participants in our project. These tests will cover various subjects, including microeconomics and macroeconomics. In general, while previous research has employed some of these tasks and tests to assess individual abilities, our methodology aligns with recent studies that employ them to investigate contributions within a team dynamic beyond individual abilities (Weidman & Deming, 2021).

The platform has two different stages for most of the tasks. First, participants solve the tasks individually, allowing us to measure individual abilities. At this stage, we also collect other measures of non-cognitive skills and social preferences (personality, trust, social perceptiveness). In the second stage, participants solve the tasks in teams. For most of the tasks, participants need to agree on their answers to proceed on the platform. For other tasks, such as “Reproducing Art”, participants need to coordinate their actions to make progress as a team. During the group stage, participants can communicate with their teammates via chat boxes. The use of chat boxes to assess collaborative problem-solving skills has been validated in other contexts—such as the PISA assessment of collaborative problem-solving abilities (OECD, 2017)—and mimics current workplace practices where chats are an important tool for coworkers to communicate.

The main interventions will take place at the team stage. There are two levels of randomization in the experimental design:

1. Team composition: Participants will be randomly assigned to either single-income-level teams or mixed-income-level teams, conditional on their socioeconomic status. For example, low-income individuals can be randomized either to an all-low-income team or a mixed-income team. In this case, there would be three types of groups: all-low-income teams, all-high-income teams, and income-mixed teams.
2. Information: The second randomization is the information about team members that is revealed or not to participants. Here, we will have three treatment arms:
a. Anonymous: all activities will be anonymous.
b. High school revealed: we will reveal the graduating high school as a signal of income in Latin America.
c. Names: we will reveal names that signal both income and other information about participants.

2024-01-01

2024-12-31

1. Leadership nominations
2. Performance metrics

1. Leadership nominations: the number of teammates who consider the participant a leader during the team activities.
2. Team performance metrics: performance measures across the different tasks, such as the number of correct answers or the time participants take to complete a task.

1. Preferences for teammates
2. Ratings of communication and teamwork
3. Disagreement within a team
4. Chat interactions

1. Preferences for teammates: participants will indicate which of their teammates they would like to continue for another round of team activities.
2. Ratings of communication and teamwork: participants will rate their teammates on various metrics, including their communication and teamwork.
3. Disagreement within a team: we will measure the suggested answers by participants of the same team by calculating the percentage of differing responses to measure disagreement.
4. Chat interactions: we will explore the information in the chats to characterize the type of interactions within a team.

The experimental design has two levels of randomization:

1. Team composition: Depending on their income level, high or low, participants can be assigned to a single-income level or to a mixed-income level team.
2. Information: a team can be randomly assigned to three levels of information: (a) anonymous, (b) high school revealed, (c) names (identities) revealed.

Not available

Randomization done in office by a computer

The randomization unit is the participant for the type of team and it is the team for the information treatment.

No

1. 720 participants for the type of team randomization: 300 low-income and 300 high-income. This derives into the following types of teams:
a. 60 high-income teams.
b. 120 mixed-income teams.
c. 60 low-income teams.

2. 300 teams for the information randomization stratifying by the type of team:
a. Anonymous: 100 teams with 50 of each type of income level.
a. High school: 100 teams with 50 of each type of income level.
a. Names: 100 teams with 50 of each type of income level.

For outcomes at the individual level: 600 observations. For outcomes at the team level: 200 observations. For outcomes at the dyad level: 2160 observations. For outcomes at the task level: 1200 observations with each team completing 5 tasks.

1. 720 Participants, with 360 being low-income and 360 high-income: 180 participants of each group are assigned to a single-income level team, and 180 to a mixed-income team.

2. 240 teams for the information treatment: 80 teams are assigned to the anonymous treatment arm, 80 to the high school revealed arm, and 80 to the names revealed arm. This randomization is conducted by stratifying the type of team, and we expect 20 segregated teams of each income type and 40 mixed-income teams to be assigned to each information treatment arm.

We present power calculations for our primary outcome, top leadership nominations. The outcome ranges from 0 to 3 for teams of three participants. The mean of this outcome in previous pilots is 0.8, with a standard deviation of 0.85. The intra-cluster correlation within a team is 0.056. All our power calculations are performed conditional on the participant's income level, as we will conduct our empirical exercises by splitting the sample by this variable. We present three types of power calculations: 1. Type of team (single-income vs. mixed-income teams): the type of team is randomized at the participant level. As we condition on the income level, we consider a sample of 360 participants per income level, 180 assigned to a single-income level team and 180 to a mixed-income level team. The MDE of the type of team on top leadership nominations is 0.257. 2. Information treatment arm: Following the experimental design, 20 low-income, 20 high-income, and 40 mixed-income teams are assigned to each information treatment arm. Conditional on the participant's income level, this implies 60 teams assigned to each information arm. As for the cluster size, for single-income teams, the size per cluster is 3 participants, while for mixed-income teams, the size per cluster is 1.5 participants, with a weighted average of 2 participants per cluster. The MDE of comparing each information treatment arm conditioning on the participant's income level is then 0.316. 3. Interaction between the type of team and information treatment arm: for the arms of the combination between type of team and information arm, we will have 20 single-income teams and 40 mixed-income teams, with single-income teams having 3 participants and mixed-income teams having 1.5 participants. The MDE for this interaction effect conditional on income level is 0.449.