Experimental Design Details
Specifically, we will present each subject with a sequence of 20 individual profiles and ask them to make decisions about each profile. Of the 20 profiles, 2 will represent Asian individuals, 8 will represent Black individuals, 2 will represent Latino/a individuals, and the remaining 8 will represent white individuals. Within each race category, half of the profiles will represent females and the other half will represent males. To manipulate subjects’ perception of the race and gender of each individual, we use portraits from the Chicago Face Database (Ma et al., 2015) and names from a list of common names by race and gender maintained by the Georgia Department of Public Health (https://oasis.state.ga.us/oasis/babynames/).
In each profile, we will reveal to subjects an initial signal—either a C- (lowest), C, C+, B-, B, B+, A-, A, or an A+ (highest)—from 32 potential signals associated with each individual. The mix of signals in each set of potential signals depends on an individual’s type—A, B, or C—which is unobservable to subjects. An individual of any type can receive any grade, but A types are more likely to have A signals in their set of potential signals, B types are more likely to have B signals, and C types are more likely to have C signals. The distribution of signals is roughly triangular for each type, and, with positive weight on all possible signals, all three distributions overlap.
To ensure that initial signals overlap across race and gender within each subject, we use the following assignment mechanism to reveal initial signals:
- Among the 4 Black female profiles, we assign one B-, one B, one B+, and one randomly selected non-B signal.
- Among the 4 Black male profiles, we assign one B-, one B, one B+, and one randomly selected non-B signal.
- Among the 4 white female profiles, we assign one B-, one B, one B+, and one randomly selected non-B signal.
- Among the 4 white male profiles, we assign one B-, one B, one B+, and one randomly selected non-B signal.
- Among the 4 Asian and Latino/a profiles, we assign signals at random.
Profiles will then appear sequentially in random order.
For each profile, we will ask subjects to decide whether to 1) enter the individual into a comparison with a B type or 2) pass on the individual. In the comparison, a signal will be randomly drawn from the individual and from the B type and whoever draws a higher signal will win the comparison. If the individual wins, then the subject will earn up to $1.00. If the individual loses, then the subject will earn up to $0.15. Alternatively, if the subject decides to pass on the individual, then they will earn up to $0.50. At the end of the experiment, we will randomly select an individual and pay each subject based on the outcomes associated with their decisions about that individual.
Before making a decision, subjects will have the option to purchase up to 15 additional signals for $0.01 per signal. The cost of the total purchase for an individual will be deducted from the payoffs described above.