Randomization is done in the office by a computer. For randomization, we asked for a big list of schools from Agastya after meeting all their requirements, from which we could randomize schools into treatment and control groups. Agastya provided us with 151 schools in the 4 districts (Ghazipur, Gonda, Lucknow, and Varanasi) of Uttar Pradesh. The list of schools was created based on donors' preferences for geographic location and Agastya's operational requirements - like schools closer to their base location and optimal utilization of resources. Essentially, it meant: (i) ensuring that each geographic block should have treatment schools (ii) catering to a maximum number of students while keeping a reasonable class size: 30-40 students per instructor in each session. In treatment 1, the minimum strength in each school had to be 120-160, while the total strength in a given district should not exceed 2000-2200. In treatment 2, the combined minimum strength in each school had to be 60-80, while the total strength in a given district should not exceed 1000-1200. Agastya pegged the strength based on previous years' enrollment numbers. Out of 151 schools, we dropped 14 schools - first removing 8 outliers (5 very big schools, 1 with low strength, 2 schools with English as the medium of instruction), and finally, 6 randomly selected schools to keep the total treated students within the operational limit of Agastya. We were left with a list of 137 schools before randomization.
The randomization was done using the Unified District Information System for Education Plus (UDISE and UDISE+) data. UDISE is one of the largest Management Information Systems initiated by the Department of School Education and Literacy, Ministry of Education, Government of India, covering more than 14.89 lacs of schools, 95 lacs of teachers, and 26.5 crores of children. UDISE+ is an updated version of UDISE that was in use from 2005-2017.
Using the unique ID of schools (11-digit UDISE code) we merged the school list with the 2017 and 2020 UDISE+ data on school/teacher characteristics such as the caste composition of the school, enrollment in schools by class, the facilities (such as number of classrooms, the condition of classrooms, available textbooks and uniform, concrete structure of schools, playground, etc.) available at the school, teachers characteristics (such as training, gender, number of teachers) at the school and randomized the sample into the two treatment groups, and a control group.