Experimental Design Details
Stage 1 (Hiring): We post a job ad (for a real job) with a general description of the position and a link to request more information about the job. Individuals who request more information are then randomly assigned to one of the four Hiring Stage Treatments, emailed the corresponding treatment script/materials, and told to submit a brief application and short resume in order to apply. To implement the diversity signaling treatments, we will use short messages signaling workplace diversity in bold at the top of a page of information. Which treatment a job-seeker receives will be randomly determined, randomization is at the individual-level, and is between-subject. In addition to the signals about workplace diversity, we have a high-wage and low-wage version of the job, giving us a 2x2 experiment design for the first stage of treatments. Note that people will actually be hired for each of the 4 conditions, to work in a real job. To measure how the treatments impact selection and distribution of worker productivity characteristics (speed, accuracy, labor supply, cooperation), and whether treatment impacts vary by race or gender, we use data generated during the work-stage of the experiment.
Stage 2 (Working): Individuals will be hired into short-term (approx. 10-day) positions from each of the hiring-stage treatments. The work consists of multiple iterations of a data-entry task performed in teams of two and that allows us to gather many observations for each individual on production speed, production accuracy, labor supply, and cooperation. Workers will perform a data-entry task remotely (online) in which they record data for images as a joint production of 2-person teams. We will have a video-chat in the online work platform through which the workers can communicate with each other, and we will tell the workers they are required to agree on and enter the same information for each question. This will require dialog between the two workers to confirm they are entering the same data – especially for qualitative questions. While described as a “joint” submission, each worker will individually submit the data through their own interface with the online work platform. The data-entry system will intentionally allow them to enter differing information. While the pictures and accompanying data-entry sheets will appear in the same order for each employee in the pair, the work system will not prevent one of the workers from moving on faster than the other. To measure how experiencing workplace diversity impacts worker behavior on the job, we will exogenously vary the working conditions faced by the data-entrants we hire into the position. To test the effects of working in a diverse work setting on employee behavior, the work-stage treatments consist of a homogeneous condition (working in a team with someone of the same gender and race), and diversity conditions (working in a team with someone of a different gender, different race, or both). Randomization is at the shift-level (projected to be approx. 1-hour), generating within-subject variation.