Measuring the Determinants of School Completion in Pakistan: Analysis of Censoring and Selection Bias

This paper explores the demand for child schooling in Pakistan, using the Pakistan Integrated Household Survey (1991). There have been few such studies for Pakistan, a country with relatively low enrollment rates and education levels, high illiteracy, and large disparity between male and female education. Additionally, this study focuses on two potential sources of bias in the estimation of the demand for schooling. First, studies which do not distinguish between currently enrolled children and those who have completed their schooling subject their estimates to a form of censoring bias. Second, studies which exclude children who have left the household from their samples may introduce sample selection bias if the decisions to leave home and to attend school are related. This study finds evidence of both “censoring” and “sample selection” bias in the demand for child schooling in Pakistan. JEL classification: I2, C24 I would like to thank T. Paul Schultz, Michael Boozer, Jennifer Hunt, Joel Waldfogel, John Maluccio, Benoit Perron, Sofronis Clerides and participants in the Labor and Population workshop for helpful guidance and numerous suggestions. Bill Greene has provided enormous help by adapting LIMDEP to suit the empirical needs of this dissertation and giving much technical support. All remaining errors are my own. I also gratefully acknowledge the World Bank and Guilherme Sedlacek for providing the data and the Rockefeller Foundation for financial support.