Experimental Design
Program slots were allocated to applicants through a randomly sorted list. The City provided the research team with application information for all program applicants. The research team first cleaned the list of applications and removed duplicate applications in the instances where the same individual applied for the program multiple times. In total, there were 11,698 eligible and unique applicants.
The research team then identified household groups within the application data. Applicants provided names and ages of other household members in the application. Applicants were grouped into households based on address information and matching names across listed household members. Each individual on the list of unique applicants was assigned a random number and the list was sorted from smallest to largest numbers. In total, there were 11,590 household groups among the 11,698 applicants. During each cohort’s enrollment period, the research team provided contact information to BCFF working down the randomly sorted list until all program slots were filled. If an individual was selected from a household group with more than one applicant, all subsequent individuals in that household group would be skipped even if they had a sufficiently low random number to be selected. Thus, the number of applications in a household group slightly affected an applicant’s treatment probability. In total, 203 and 207 individuals were contacted to enroll the 175 and 176 program participants in cohorts 1 and 2, respectively.