Recent Papers

Does BRAC Provide Highly Effective Schooling in Developing Countries?

John C. Ham and Saima Khan

Abstract: BRAC, a pioneer of non-formal primary education, has been operating in Bangladesh for the past 30 years often filling the gap that the formal sector failed to address. Since the 2000s BRAC has taken its schooling model to international locations such as Afghanistan, Philippines, Uganda etc. Despite its long tenure and wide coverage, there are few independent rigorous evaluations of BRAC schools. Thus, in this paper, we compare BRAC to Government, JAAGO (an innovative new school-type) and other NGO schools in two slums of Dhaka city. We examine school-type impact on student achievement in mathematics, using high quality data that we collected ourselves. We find that in terms of fluid intelligence (IQ) and parents’ education, ’weaker’ students sort into BRAC schools as compared to Government, JAAGO and other NGO schools. We control for this selection using a rich set of conditioning variables. We use propensity score matching in the presence of choice-based sampling to estimate the treatment effects. We find that controlling for fluid intelligence (IQ) has, by far, the biggest effect on reducing se- lection bias, and hence indicates the importance of collecting such data when using non-experimental methods. We find that once we control for selection, BRAC students are worse off than their counterparts in Government and JAAGO schools; we find no difference between comparable students at BRAC versus NGO schools. We disaggregate by gender to find that the BRAC versus Government effect is being driven by boys and the BRAC versus JAAGO effect is being driven by girls.

Keywords: Marginalized Communities, Gender Heterogeneity, Selection, Propen- sity Score Matching, Choice Based Sampling.

Selection, Gender and the Impact of Schooling Type in the Dhaka Slums: Much Better Data Matters!

John C. Ham and Saima Khan

Abstract: We study three school-types in two urban slums of Dhaka - JAAGO, Government and NGO schools; our study is the first to investigate education in urban Bangladesh. We examine the impact of the 3 school-types on student achievement in mathematics, using high quality data that we collected. We find that in terms of fluid intelligence (IQ), parents education and family income, better students sort into JAAGO and Government schools. We control for this selection bias using a rich set of conditioning variables using two matching estimators. We find that controlling for fluid intelligence (IQ) has, by far, the biggest effect on reducing selection bias, and hence indicates the importance of collecting such data when using nonexperimental methods. We find that girls perform better at JAAGO schools, but there is no significant difference between comparable girls at NGO and Government schools. On the other hand, there is no significant difference between comparable boys at JAAGO and Government schools; boys perform worse at NGO schools. When we conduct within school-type comparisons, we find that boys outperform girls in Government schools; however there is no sognificant difference between comparable boys and girls within JAAGO and NGO schools.

Keywords: School-type, Gender Heterogeneity, Selection, Propensity Score Matching, Choice Based Sampling.