Informing Large-Scale Policies in Education

Abstract

This dissertation consists of three essays that inform large-scale, national-level policies in education. The first essay estimates the effects of a federal grant to improve access and quality in state pre-K programs in the U.S. Using a difference-in-differences framework, the study finds federal support increased public pre-K enrollment by 51% in states with the smallest prior state-supported programs. In states with larger prior state-supported pre-K programs, there are no discernible effects on access or enrollment but modest effects on teacher availability. For a subset of districts with available data, cognitive achievement at Kindergarten entry improved. The second essay examines the effects of a conditional cash transfer pilot program in Tanzania. Using the random assignment of the program across villages, the study finds that cash transfers increased the probability that children ever attended school by 4 to 5 percentage points. It finds no effects on primary school completion. Girls and boys benefited equally, and there is suggestive evidence that children with stronger initial educational performance have higher primary completion rates. Together, these two studies provide causal evidence to guide policymakers in the implementation of large-scale policies. The third essay examines the measurement properties of survey-based measures of classroom instructional quality. Drawing on data from 100 teachers and 4,000 students across 23 schools in Colombia, Brazil, and Botswana, it finds that student survey measures generally show strong internal consistency and reliability, while teacher survey measures do not. Factor analysis provides support for a three-factor solution to student survey measures intended to capture teachers’ stimulation of students’ reading engagement, student-teacher relations, and instructional scaffolding, but teacher survey measures related to instructional differentiation and promotion of enhanced instructional activities are not reliable measures. At the same time, some of these measures show positive, statistically significant relationships to measures of PIRLS reading comprehension across study countries. This study suggests that classroom instruction, increasingly the focus of policy efforts to improve educational quality, can be measured through survey-based approaches.