Identifying class size e ects in developing countries: Evidence from rural schools in Bolivia

Although great interest surrounds class size as a policy instrument, inferences on its e ects are controversial. Recent work highlights a particular way to consider the endogeneity issues that a ect this variable: class size is often correlated with enrollment, which may in turn be related to socioeconomic status. This paper shows that these correlations are signi cant in the case of Bolivia. Identifying the e ects of class size therefore requires either: i) isolating variation in it that is not caused by enrollment, or ii) nding situations in which, at a minimum, enrollment and socioeconomic status are not related. Building from institutional arrangements that determine pupil/teacher ratios in the rural area, the paper implements two research designs that seek to satisfy these conditions. The rst uses a teacher allocation pattern, which implicitly limits class size, as an instrumental variable. The second focuses on remote schools that have a single class per grade, and a monopoly in their area of in uence. Both approaches suggest class size negatively a ects test scores.