Essays in Development Economics and Political Economy

Abstract

This dissertation consists of three field experiments that explore how information and the social environment affect accountability and the provision of public goods in developing countries. The first chapter explores a key constraint on politicians' ability to deliver public goods: their inaccurate beliefs about what citizens prefer. Studies on the role of information in political accountability usually ask whether citizens know enough about politicians. In this chapter, I ask instead whether politicians know enough about citizens to adequately represent them. Using original politician and citizen surveys in Pakistan, I show that politicians hold highly inaccurate beliefs about citizen preferences. In collaboration with a large political party, I conduct a field experiment with 653 politicians to understand how politicians respond when they receive information on citizen preferences. I find that politicians who receive information make recommendations to their party leadership that are closer to what citizens prefer. Directly elected politicians are more responsive than indirectly elected ones. Politicians are more responsive to information about women's preferences compared to men's preferences. I interpret my results using a simple model of belief updating and responsiveness. The model suggests that higher responsiveness to women's preferences should be expected if politicians are less confident in their prior beliefs about women, for which I find evidence in the data. This paper shows that politicians' inaccurate beliefs constrain accountability and public good provision in developing democracies. My results point to the need for better channels for the flow of information from citizens to politicians---channels that include those who are currently underrepresented.

Politicians are less confident in their prior beliefs about women because women engage less frequently in politics. The second chapter asks whether constraints on women's participation lie with women themselves, or with the men in their households who act as gatekeepers. We conduct a field experiment in Lahore, Pakistan to test how a canvassing campaign aimed at increasing women's turnout should be targeted within the household. We randomly assign 2500 households to one of four conditions: no canvassing visit, a visit targeted at men, a visit targeted at women, or both. We find large increases in women's turnout when the visit targets only men or both men and women. Targeting women alone is insufficient to improve their turnout. Using a costly behavioral measure of support for women's role in democracy, we find that canvassing men increases their demonstrated support for women's role in democracy two months after the election. Households where both men and women were treated see greater political discussion among men and women, and men in these households are more likely to have provided women logistical support to vote. The results suggest that engaging men is necessary to reduce gender gaps in political participation, particularly in contexts where women do not enjoy full decision-making power over their own participation.

Do voters care about how connected their candidates are? The third chapter of this dissertation investigates this question in the 2015 local government elections in Pakistan combining: (i) data on ties between candidates, higher level politicians, and bureaucrats; (ii) a large-scale field experiment; and (iii) election outcomes. Before the election, voters considered local candidates' connections important and expected local politicians to help them access services provided by other levels of government. Providing voters information on connections increased support for more connected candidates, but information on past party performance did not. More connected candidates received more votes and were more likely to win office, but there was no electoral benefit to past service provision. The results provide novel evidence of the importance of political connections for electoral outcomes and show that forward-looking expectations based on candidate characteristics and an understanding of higher-level political process play an important role in vote choice.