Essays in Development Economics and Political Economy

Abstract

This thesis features three closely linked essays in political economy of development. These all share the following aspects: first, they are all set in the Indian state of Bihar: an important, poor, fast-growing state in the north-east of the country; second, they study the role of village-level local government officials in bringing about changes in economically important outcomes; third, they all study the role of caste in mediating change; and fourth, they are econometrically linked in their dependence on regression discontinuity design strategies to establish causal links. Taken together, these essays attempt to shed light on the complex ways in which ethnically marginalized groups negotiate political power. In Chapter 1, we show that ethnic diversity adversely impacts public good provision, disproportionately affecting minority communities. While political representation could mitigate these effects, successful delivery of public goods often depends on how well minority and majority representatives collaborate. Using data from over 100,000 local politicians in India, we show that ethnic differences cause breakdowns in collaboration. We use a regression discontinuity (RD) design to show that delivery of public goods suffers when ethnic minority (low caste) representatives govern under non-minority (non-low caste) representatives. We then study an institutional innovation that can increase collaboration. In our setting, local politicians can issue complaints to the higher bureaucracy under a formal complaints technology. We show that ethnic minority representatives file over twice as many complaints when exogenously paired with non-minority representatives. Does filing complaints improve public good provision? We run a large field experiment involving 1629 minority representatives in which we randomize offers to file complaints on their behalf regarding project implementation. Our intervention increases filing of complaints by 41 p.p and implementation of public works projects by 24\%, accounting for 60\% of the initial gap in provision. Treatment has positive spillovers on neighboring jurisdictions. Our results are consistent with a Nash bargaining model featuring two politicians bargaining over project implementation in a setting with a costly complaints technology. In Chapter 2, we focus on how affirmative action policies affect inequality. We present evidence on the effects of one such policy: mandated ethnic quotas in local government for members from minority groups, also known as ``political reservation". We use two natural experiments and bring to bear multiple datasets from the Indian state of Bihar. These include secondary data on public goods from across 45,000 villages, private assets from over 17 million rural households, political candidacy data from nearly 96,000 local jurisdictions and a primary survey of over 8,000 households. Using a regression discontinuity design, we show that political reservation: (a) lowers inequality in access to public goods (b) lowers inter-group private asset inequality modestly in the short-run and more substantially in the medium-run and (c) increases presence of minority-group members in local government in the medium-run (even in the absence of reservation). However, we also provide suggestive evidence of backlash by majority group members against the policy of reservation: using a second natural experiment, we show that an increase in the quantum of reserved jurisdictions in their neighbourhood makes minorities in non-reserved jurisdictions worse off. In Chapter 3, we ask: How do affirmative action policies change the nature of interactions between majority and minority groups? Set in the Indian state of Bihar, this paper presents preliminary evidence on the network impacts of one type of affirmative action policy: ethnic quotas in favour of low-caste members, called ``political reservation". We show that exposure to low caste village heads worsen stereotypes about caste-groups, with low caste members being seen as less trustworthy and more likely to commit fraud. Using census data from over 40,000 households and network data from over 11,000 households, we find that low castes interact more with their own-type in reserved areas. While low caste households report better access to local public goods in the long run, we find limited evidence of other effects on downstream outcomes.