In our experiment, we introduce color groups by giving each participant a wristband in a specific color. Interacting participants might have the same or different wristband color.
Furthermore, participants are assigned a role. They either are employers or employees. Participants are sorted into labor markets such that each labor market includes 3 employers and 3 employees. Each employer is paired with an employee of her labor market. This employee is her current employee.
Employees have a randomly assigned ability (high, middle, low), which determines how valuable their work is for the employer. Depending on the employee's ability employers make a decision whether or not to promote their employee. The other employers in the labor market do not observe the ability of employees who they don't employ currently. The promotion decision on the other hand is visible.
Also, employers can make a wage offer to their own and the other two employees in their labor market. For their own employees they can make the wage offer dependent on their ability. For the other employees they can make their wage offer only dependent on whether or not the employee is promoted by her original employer. Subsequently, a set of rules decides for whom the employees will work.
The rules of the experiment are designed in a way that theoretically, tow equilibria are possible. In one equilibrium, only high ability employees are promoted. In the other equilibrium, employees with high and middle ability are promoted.
In the experimental treatments we vary the color group composition of employers and employees. We expect first, that employers are more likely to promote their employees with middle ability if they are of the same group and second, that employers expect other employers to more likely promote their middle ability employees if they are from the same group.