Experimental Design Details
Employment stage
In this stage, workers recruited on MTurk are instructed to come up with slogans for two advertisement campaigns. To generate the first slogan (Task 1), all workers work individually. To generate the second slogan (Task 2), workers are randomized to work either individually (the “Individual work” treatment) or in teams of two (the “Team work” treatment, in which workers can communicate through a chat window). Workers are randomized across four different product pairs:
1. Task 1: Sports motorcycle, Task 2: Drill
2. Task 1: Truck, Task 2: Package of sports TV channels
3. Task 1: Diet shake, Task 2: Shampoo
4. Task 1: Sewing machine, Task 2: Sun screen for kids
The first two product pairs represent stereotypically male products, while the last two represent stereotypically female products. These products were selected based on a pilot study conducted on MTurk in fall 2018.1
The tasks are incentivized: In addition to the show-up fee of $2, workers earn a bonus if their slogan receives a high evaluation from a set of independent evaluators. For each product, workers receive a bonus of $1 if their slogan is ranked, on average, in the top third by the evaluators.
Under the “Team work” treatment, the two workers in the same team can communicate through a chat window. The team members are only eligible for the bonus if they agree on one slogan.2 Thus, while we allow workers to not reach an agreement on the slogan to propose, we provide incentives for the team to produce a joint output.
Evaluation stage
Evaluators are recruited on MTurk and instructed to evaluate slogans according to their perceived quality. We show evaluators groups of 9 slogans, and we ask them first to rank the slogans and then to assign to each slogan a quality score between 0 and 10.4 The groups of 9 slogans are formed randomly, grouping together slogans for the same product from both working conditions (“Individual work” or “Team work”).
Each evaluator is shown one group of slogans from each of four randomly chosen and randomly ordered products. The task is not incentivized, but we stress that the compensation of the MTurk participants in the Work stage is linked to this evaluation. We reward evaluators with a $2 show-up fee. When eliciting quality scores, we include in the instructions: "If you judge a slogan to be unusable (e.g. it is blank, unrelated to the product, or poor English) please give it a score of 0." We then use these “zero scores” to screen out unusable slogans, analogous to a very basic pre-screening of job applications.
Employment stage
Employers are shown information about two candidates (one female and one male) and asked to indicate which candidate they want to hire to come up with a slogan individually. The information includes the candidate’s fictional name, education level, work experience, and the slogan they produced (either individually or in a team) for Task 2 of the Work stage. The gender of the candidate (and, in the “Team work” treatment, their co-worker) is conveyed through virtual names (see below) and an avatar showing a silhouette of a man or a woman. We tell the participants that we do not use workers’ real name (see details about the names below). Education is displayed in three categories: “High School or less”, “Some College education”, and “College degree or more”. Experience is also shown in three categories: “Up to 1 year”, “1 to 3 years” and “3 years or more”. Since we do not mention gender explicitly, and, since we show information about education and work experience, we do not believe that gender is too salient for the employers when evaluating candidates.
Employers are shown information about two candidates (one female and one male) and asked to indicate which candidate they want to hire to come up with an advertisement slogan individually. The information includes the candidate’s fictional name, education level, work experience, and the slogan they produced (either individually or in a team). The gender of the candidate, and, in case of a team work evaluation, their co-worker's gender, is conveyed through virtual names and an avatar showing a silhouette of a man or a woman. Education is measured in three categories: “High School or less”, “Some College education”, and “College degree or more”. Experience is also measured in three categories: “Up to 1 year”, “1 to 3 years” and “3 years or more”. We tell the participants that we do not use workers’ real name.
We randomize:
1. Whether the slogan is produced individually or in a team. (Individual vs. Team treatment)
2. The gender of the co-workers of the candidates in the “Team work” treatment. In particular, we vary whether the employer chooses between (i) a male candidate with a female co-worker and a female candidate with a male co-worker, (ii) a male candidate with a male co-worker and a female candidate with a female co-worker, (iii) a male candidate with a male co-worker and a female candidate with a male co worker, and (iv) a male candidate with a female co-worker and a female candidate with a female co-worker.
3. Which information/instructions the employer observes (a “Baseline”-, a “Hidden gender”-, and a “Cognition” treatment). In the “Hidden gender” treatment, all information about the gender of the candidates (and, under the “Team work” treatment, their co-workers) is removed from the resume. In the “Cognition” treatment, before making their hiring decision participants are forced to reflect on the productivity weight they give to the candidate they hire and as well as their co-worker.
4. Whether the products are stereotypically male or stereotypically female.
To reduce the total number of treatments, and increase statistical power, the ``Hidden gender''- and ``Cognition'' treatments are only implemented under the ``Team work'' treatment for stereotypically male products. Further, all four combinations of candidate/co-worker gender are only implemented under the ``Baseline'' treatment for stereotypically male products.