Experimental Design
Participants will be asked to form beliefs about a group of US participants (henceforth workers) on Prolific, who worked on up to 25 decoding tasks under three compensation rules:
1) Fixed: Workers received a payment of $2, regardless of how many decoding tasks they solved correctly.
2) Comparative (top 10%): Workers received a fixed payment of $1, regardless of how many decoding tasks they solved correctly. In addition, they received an additional bonus of $10 if their performance was among the top 2 in a comparison group of 20 participants (including them) who worked on the same tasks.
3) Comparative (top 50%): Workers received a fixed payment of $1, regardless of how many decoding tasks they solve correctly. In addition, they received an additional bonus of $2 if their performance was among the top 10 in a comparison group of 20 participants (including them) who worked on the same tasks.
Participants will be asked to form probabilistic beliefs about the relative performance of different gender and race for each of the three compensation rules. We incentivize the belief reports with the binarized scoring rule without providing detailed information about the incentives (as in Danz et al., 2022). Participants know that they have the chance to win a $2 bonus, and the chance of winning this bonus increases with the accuracy of their belief reports.
In the gender comparison, participants will be asked to report beliefs about the relative likelihood that the average performance level is higher among men or women for each of the 3 compensation rules. In the race comparison, participants will be asked to report beliefs about the relative likelihood that the average performance level is higher among Asian or Hispanic/Black (randomized in a between-subjects design) workers for each of the 3 compensation rules. In a between-subjects design, we will randomize the order of asking for gender or race comparisons, as well as the labeling the belief elicitation scale (e.g., men---women vs. women--men). The randomization is stratified across participant gender and Part A payoffs.