Education Production and Peer Networks Among Out-of-School Children in India

The PI proposes investigating the education production function in the context of an informal community based model of instruction targeted at out-of-school children in India. Using the planned expansion of the program, the PI will conduct a three part randomization that will allow us to distinguish the effects of student, teacher, classmate, and non-class peers on student achievement while also generally evaluating the effectiveness of community based class model. The evaluation design comprises three randomizations. First, from one hundred communities we will randomly choose sixty-six in which to provide the intervention. In the treatment communities, the PI will randomly choose from the out-of-school children who indicate an interest in the program a subset to receive the treatment. To generate variation in the coverage of students peer networks, the PI will also varying the fraction of out of school children that will be treated in these communities. Finally, the PI will randomly assign students to classes and teachers to classes to allow measurement of the effects of teacher and classmate characteristics. The PI will track interested children in 100 communities (6,000 children) for two years. Student performance will be measured through tests, attendance rates, and subsequent enrollment rates.
The study will expand the understanding of the education production function in three ways. First, the PI will contribute to the growing literature that employs randomized evaluations to estimate the causal effects of educational interventions in the context of developing countries. The PI's estimates taken in combination with the existing literature will expand understanding of how individual institutional structures shape students experiences. For example, comparing the change in student performance due untrained teachers through the Bridge Course program to the change in student performance due to similar teachers operating in conjunction with normal school teachers (Banerjee, Cole, Duflo and Linden, 2005) allows improved understanding of the relative effectiveness of such teachers operating in different environments and with different populations of students. To complement these global estimates, the PI also seeks to conduct micro-level analyses in which the PI estimates more fundamental relationships underlying the education production function. Drawing on the classic theoretical specifications of the education production function and the newer models of peer effects, the PI directly estimates the effects of teacher, classmate, and non-class peer characteristics on student learning. Finally, the PI investigates a richer set of characteristics than have been previously investigated in the developing context. For example, the PI investigates the influence of different types of peers by testing the effects of peers that have different relationships to a child by surveying described levels of closeness, shared activities, etc.
The intervention the PI proposes to evaluate is extremely cheap (about $1.50 per student per month), and operates completely outside of traditional administrative boundaries (schools, religious institutions, etc.). If effective, it could provide a viable mechanism for educating the over 100 million primary-aged children in the developing world who are currently out of school. In addition, the entire data set resulting from the proposed study will be made available to other researchers once the PI has finished working with the data.