UC Irvine

This dissertation studies the human capital formation and the partisan provision of public goods in the context of developing countries. In the first chapter, we use a natural experiment to study the impact of a remediation program that provided customized instruction to the bottom 3 to 7 students in a class from grades 3 to 8 in Pakistan. Using a Fuzzy Regression Discontinuity Design, we show that students in these remediation classes gained 0.288s.d and 0.217s.d in English and Math, but not in Urdu. Importantly, we find that remediation classes were equally beneficial at the primary and the secondary grades levels and across genders and class sizes. We investigate the mechanisms driving these findings, including the null result for Urdu, using a survey of school principals in our sample. The survey revealed that some schools assigned non-subject specialist teachers to Urdu remediation classes. In cases where subject-specialist teachers led these classes, we find positive effects of remediation on Urdu test scores. We conclude that remediation classes can be effective in mitigating the within-classroom learning disparities prevalent across the developing world. However, their impact is sensitive to program design and implementation. In the second chapter, I estimate the effect of the incumbent politician's partisanship on local political violence in Pakistan. Sub-national political violence poses a serious threat to democracies in the developing world. It is consequently vital that we understand the factors that contribute to such violence. Using a close-election regression discontinuity framework for identification, I find that constituencies that narrowly elect a ruling party politician experience significantly less subsequent violence. I show that these "ruling party effects" effects are not associated with a particular ruling party’s ideology, but that they are concentrated in areas with high police penetration and areas with low media exposure. I conclude that these "ruling party effects" operate through control of the state machinery and are mitigated by fear of electoral penalties. The third chapter studies the causal effects of the gender composition of siblings on the education of children in Pakistan, a country that suffers from a significant male bias in educational investments. Using the plausibly exogenous gender of the second-born child, I find that eldest sons with a younger sister fare better on a range of educational outcomes than eldest sons with a younger brother. I carry out several robustness checks to validate my estimates, including instrumenting fertility by twin births.