The three essays conforming this thesis are representative pieces of my approach to analyzing the causes and consequences of economic underdevelopment. The overaching topic that ties together these essays is role that institutions and culture play in affecting specific behaviors that undermine development. The approach to the questions addressed in each essay is empirical, using data from Per {u} and Sierra Leone, and relies on economic theory to provide a general framework and deepen our understanding of the observed behaviors. Below, I provide a more detailed summary of the main findings of each chapter in this thesis:

In Chapter 1, ''Turnout, Political Preferences, and Information: Evidence from Perú'', I explore the role of electoral institutions that encourage citizens to vote on voter behavior. These institutions are widely used around the world, and yet little is known about the effects of such institutions on voter participation and the composition of the electorate. In this paper, I combine a field experiment with a change in Peruvian voting laws to identify the effect of fines for abstention on voting. Using the random variation in the fine for abstention and an objective measure of turnout at the individual level, I estimate the elasticity of voting with respect to cost to be -0.21. Consistent with the theoretical model presented in this essay, the reduction in turnout is driven by voters who (i) are in the center of the political spectrum, (ii) are less interested in politics, and (iii) hold less political information. However, voters who respond to changes in the cost of abstention do not have different preferences for policies than those who vote regardless of the cost. Further, involvement in politics, as measured by the decision to acquire political information, seems to be independent of the level of the fine. Additional results indicate that the reduction in the fine reduces the incidence of vote buying and increases the price paid for a vote.

Chapter 2, ''Civil Conflict and Human Capital Accumulation: The Long Term Consequences of Political Violence in Perú'', analyzes the consequences of a long lasting civil conflict on human capital accumulation. In this chapter, I provide empirical evidence of the

long- and short-term effects of exposure to political violence on human capital accumulation. Using a novel data set that registers all the violent acts and fatalities during the Peruvian civil conflict, I exploit the variation in conflict location and birth cohorts to identify the effect of the civil war on educational attainment. Conditional on being exposed to violence, the average person accumulates 0.31 less years of education as an adult. In the short-term, the effects are stronger than in the long run; these results hold when comparing children within the same household. Further, children are able to catch up if they experience violence once they have already started their schooling cycle, while if they are affected earlier in life the effect persists in the long run. I explore the potential causal mechanisms, finding that supply shocks delay entrance to school but don't cause lower educational achievement in the long-run. On the demand side, suggestive evidence shows that the effect on mother's health status and the subsequent effect on child health is what drives the long-run results.

In the third and final chapter of this dissertation, "Transportation Choices, Fatalism, and the Value of Statistical Life in Africa", joint work with Edward Miguel, we take a look at the role culture plays in determining the willigness to pay to avoid life thretening situations. Specifically, we exploit a unique transportation setting to estimate the value of a statistical life (VSL) in Africa. We observe choices made by travelers to and from the airport in Freetown, Sierra Leone (which is separated from the city by a body of water) among transport options -namely, ferry, helicopter, speed boat, and hovercraft - each with differential historical mortality risk and monetary and time costs, and estimate the trade-offs individuals are willing to make using a discrete choice model. These revealed preference VSL estimates also exploit exogenous variation in travel risk generated by daily weather shocks, e.g. rainfall. We find that African travelers have very low willingness to pay for marginal reductions in mortality risk, with an estimated average VSL close to zero. Our sample of African airport travelers report high incomes (close to average U.S. levels), and likely have relatively long remaining life expectancy, ruling out the two most obvious explanations for the low value of life. Alternative explanations, such as those based on cultural factors, including the well-documented fatalism found in many West African societies, appear more promising.

One of the main goals of most governments and international agencies is to improve education, but this task has proven to be elusive. Evidence shows that just increasing resources is not enough to improve outcomes. Therefore, institutional reforms have been proposed to improve the delivery and financing of education. School Based Management (SBM) is one such institutional reform where decision making is transferred to the school level. Funds are transferred directly to the school and parents, along with teachers and the principal, allocate and oversee the use of program funds. By increasing parental and community involvement in schools, SBM is expected to improve education because it is expected to increase accountability and monitoring of school personnel, align incentives and match resource allocation to school needs. Additionally, SBM seeks to increase school autonomy and implement assessments and evaluations. With these ideas in mind, the Mexican Education Ministry has implemented SBM programs in over 30% of elementary public schools in the country.

Even though SBM programs are popular, evidence on the effects of such programs in the country show mixed results. Moreover, there has been no attempt in the literature to explore the underlying mechanisms at work. This dissertation is an effort to learn from the Mexican experience with two SBM programs: The largest SBM program in Mexico and a randomly allocated program in the state of Veracruz. The Quality Schools Program (Programa Escuelas de Calidad , PEC) was first implemented in 2001 and provides an opportunity to learn how SBM has evolved with time. The Veracruz experiment provides short run information on participation by parents on school related activities which is not available for PEC and overcomes the potential self selection problem that has plagued SBM evaluations so far.

The mechanisms I explore are five. First, I analyze how the programs change the cash transfers that parents voluntarily make to the school. If parents substitute voluntary contributions with program funds then the program causes crowding out. In this case there would be no change in resources available to the school and therefore no effect from improved conditions on the quality of education. Second, I analyze whether the matching scheme of PEC creates incentives for parents to increase voluntary contributions and if so, whether the incentive is different for rich and poor schools. Third, I analyze how the program caused parents to change their time investment. Parents may use program funds to either complement or substitute their participation on school related activities. Fourth, I make use of improved data availability to make a more accurate assessment of the effects of the programs on the quality of education as measured by standardized test scores and failure rates. Finally, I analyze if the effects of PEC change over time. Evidence in the United States shows that it may take five to eight years to see improvements in education because school constituents need to learn and adjust to the new institutional framework. On average, Mexican elementary schools have fewer resources than those in the United States and therefore they may face higher returns. As a result, the program may have an effect in the short run.

This dissertation is organized as follows: Chapter I is an analysis of the effects of PEC on voluntary contributions, standardized test scores, failure rates and dropout rates over time, chapter II is a study on the effects of PEC on voluntary contributions as a function of school income and chapter III analyzes the effect of the randomized Veracruz experiment on voluntary contributions, parent participation, standardized test scores and failure rates. Voluntary contributions are cash transfers that parents choose to make to the school to cover salient needs. The contribution amount is set at the beginning of the school year by the parent's association and once it is set, all parents are expected to pay it. Payment is socially enforced. I find no evidence of any of the two programs causing crowding out. Moreover, I find that the matching scheme of PEC creates an incentive for parents in the poorest schools to increase voluntary contributions more. Therefore matching works as a targeting device because more government resources go to the schools with greater needs. More specifically, schools located in localities with medium, high and very high marginality increase voluntary contributions by $0.22.- while those in low or very low marginality areas increase them in $0.11.-. Both increases are small compared to a typical voluntary contribution of $10.72.- and therefore there is no increase in dropout rates as a result of larger voluntary contributions. On the opposite, PEC decreases the probability of a student moving to a different school in urban areas. Therefore parents may be more satisfied with school services. There is no effect on voluntary contributions on the Veracruz experiment which lacks the matching scheme in PEC. In this program, funds and institutional improvements in the school act as a way to balance parent participation. I find that parents that spent less than 5 hours per month in school related activities increase participation while those that spent more than 7 decrease it. I find no effects of either of the two programs in the short run on the quality of education as measured by standardized test scores and failure rates. For the Veracruz experiment where data is available, it is observed that most program funds were invested on infrastructure improvements and therefore it is not surprising not to find effects in the short run. Consistent with findings in the US, the effects of PEC are larger as the school gains experience. Moreover, initial adjustments have a slight negative effect. For example, the average math test score decreases in one point the first year but increases 4 points in the fourth year. The average math test score is 487 points on a scale that ranges from 200 to 800, therefore changes are small but increasing.

Future research is needed to better understand the effects of SBM. More detailed data on how parents and other school constituents react to the program is needed. It is important to explore the processes that the program put at work to understand why the effects change with time. More importantly, it is relevant to analyze the changes that bring improvements on the quality of education. More detailed data and more experiments with careful variation in intervention components could shed light on the underlying mechanisms at work. This is relevant given the large investment the country is making on these kinds of programs. If we learn from the experience, future programs can be adjusted and improved to further improve the quality of education in the country more efficiently. This knowledge could not only benefit Mexico but other countries such as Argentina, Bangladesh, Guatemala, Honduras, India and Sri Lanka which are implementing and operating similar programs.

In the first chapter, I discuss a study that uses experimental variation in a door-to-door information campaign to test for polarization in policy choice across the electorate in Turkey. The campaign took place before a landmark referendum that was initiated by the incumbent party in Turkey after the chaotic coup attempt in July 2016. The referendum was on institutional changes to weaken constraints on the executive branch in Turkey. I designed the implementation of the door-to-door campaign as a randomized experiment. In this campaign, the opposition party gave uniform information on poor economic performance and increased terrorist activity under the incumbent's leadership to more than 130,000 voters. I show that voters, despite being exposed to the same campaign, diverged further in their vote choice on aggregate, leading to a significant increase in political polarization. This is a unique result where polarization in vote choice at the aggregate level is driven by differences in reaction to the same door-to-door campaign.

In the second chapter, I investigate electoral competition in an illiberal democracy. To promote their electoral and policy goals, elected officials make investments in a ``home style" -- a strategy to learn about their voters and select their public platform accordingly. I consider a framework of home style grounded in perception: politicians better able to accurately assess voter preferences will more closely tailor their communications to those preferences. I conducted interviews with Members of Parliament establishing that the majority party regularly conducts polls to assess voter preferences while the opposition party does not. I then analyze nearly a million MPs’ tweets and find that the majority party is more likely to communicate on the issues of most importance to their constituents than the opposition. Finally, I conduct a quasi-experiment and find that providing opposition MPs with polling increases their likelihood to communicate about their constituents’ most important issues while there is no effect among majority MPs. This result supports the qualitative evidence that the incumbent invests more in learning about voters than the opposition.

The third chapter, co-authored with Marshall Burke, Felipe Gonzalez, Solomon Hsiang, and Edward Miguel, studies whether economic or non-economic factors better explain the well-established relationship between temperature and violence in a unique context where intergroup killings by drug-trafficking organizations (DTOs) and “normal” interpersonal homicides are separately documented. A constellation of evidence, including the limited influence of a cash transfer program as well as comparison with both non-violent DTO crime and suicides, indicate that economic factors only partially explain the observed relationship between temperature and violence. We argue that noneconomic psychological and physiological factors that are affected by temperature, modeled here as a “taste for violence,” likely play an important role in causing both interpersonal and intergroup violence.

The present work assesses the effectiveness of educational grants at raising schooling attainment of poor children in rural areas. The per grade gains in reducing drop outs cumulate in an additional half a year in total schooling. Progressive impacts are found along three dimensions: degree of poverty, parents’ education and distance to school. The children of uneducated fathers living far from school gain twice as much as their counterparts with an educated father or close to a school. The intervention successfully closes the schooling gap along the wealth dimension but falls short of achieving the same in the other dimensions of parents’ education and school distance.

This paper investigates how child labor and schooling are responsive to opportunities to work, in particular to opportunities provided by children’s own parents. The paper demonstrates that after controlling for household, parental, regional, and child characteristics, children whose parents are self-employed or employers are more likely to work than children of employees, irrespective of the sector of parent activity. Furthermore, the paper also confirms a recent finding that children from areas with high average adult employment rates are more likely to work than children from areas with low average adult employment rates. Finally, since twice as many children of the selfemployed and employers both work and go to school as those of employees, the paper suggests that child labor does not necessarily represent a trade off with schooling as it depends on the occupation of the parents.

This dissertation studies local governance in China. Guided by theoretical insights originated from the political economy and public economics literature, I exploit natural experiments and randomized field experiments to provided causal evidence on the role of bureaucratic incentives in the implementation of important public policies. The findings in this dissertation could deepen our understanding of the core incentives and constraints faced by Chinese local bureaucrats, and shed new lights on the design of more effective public policies.

In Chapter 1 (coauthored with Guojun He and Bing Zhang), we estimate the effect of environmental regulation on firm productivity using a spatial regression discontinuity design implicit in China’s water quality monitoring system. Because water quality readings are important for political evaluations, and the monitoring stations only capture emissions from their upstream regions, local government officials are incentivized to enforce tighter environmental standards on firms immediately upstream of a monitoring station, rather than those immediately downstream. Exploiting this discontinuity in regulation stringency with novel firm-level geocoded emission and production datasets, we find that upstream polluting firms face a 27% reduction in Total Factor Productivity (TFP), and a 48% reduction in emission intensity, as compared to their downstream counterparts. We find that the discontinuity in TFP does not exist in non-polluting industries, only emerged after the government explicitly linked political promotion to water quality readings, and was entirely driven by prefecture cities with career-driven leaders. Linking the TFP estimate with the emission estimate, a back of the envelope calculation indicates that China’s current water-pollution abatement target leads to an annual economic loss of more than 30 billion dollars.

Chapter 2 intends to provide causal evidence on fiscal competition and coordination between local governments. We first introduce a unique empirical setting created by the combination of two national programs in China, the “Rural School Consolidation (RSC)” program and the “Special Economic Zones (SEZs)” program, in which spatially adjacent townships have discontinuous incentives to either compete or coordinate with a neighboring SEZ. Exploiting such quasi-experimental variation with a spatial-discontinuity design, we find a series of empirical evidence on the existence of competition and coordination between local governments, which can be formally rationalized by a simple model where asymmetric jurisdictions compete for mobile labor. Our results suggest that the strategic interactions among local governments play important roles in public goods provision and labor migration: during the RSC program (2001-2011), townships that coordinate with their neighboring SEZs closed an extra 15% of local primary schools, and lost an extra 7% of local population, as compared to townships that compete with their neighboring SEZs.

Chapter 3 is joint with Alain de Janvry, Guojun He, Elisabeth Sadoulet, and Qiong Zhang. This paper studies the personnel economics of Chinese grassroots bureaucrats. We notice that subjective performance evaluation is widely used by firms and governments to provide work incentives. However, delegating evaluation power to senior leadership could cause influence activities: agents might devote much efforts to please their supervisors, rather than focusing on productive tasks that benefit their organizations. We conduct a large-scale randomized field experiment among Chinese grassroots state employees and provide the first rigorous empirical evidence on the existence and implications of influence activities. We find that state employees are able to impose evaluator-specific influence to affect evaluation outcomes, and this process could be partly observed by their co-workers. Furthermore, introducing uncertainty in the identity of the evaluator, which discourages evaluator-specific influence activities, can significantly improve the work performance of state employees.

This dissertation presents a three-part study in modern empirical development economics. In each part, I tackle an important economics issue that affects the developing world—agricultural risk, principle-agent problems in firms, and poor public service delivery. I use randomized control trails (RCT) to uncover the causal impact of interventions designed to relieve constraints and improve outcomes for stakeholders.

The first chapter evaluates a novel microfinance product that is designed to help small-holder farmers manage uninsured agricultural risk. I run a large-scale RCT involving 300,000 microfinance clients in Bangladesh with one of the country’s largest microcredit institutions. Microfinance clients were randomly given pre-approval to take a loan if they experienced flooding in their local area. I show that this unique type of microcredit improves household welfare through two channels: an ex-ante insurance effect, where households increase investment in risky production, and an ex-post effect, where households are better able to maintain consumption and asset levels after a shock. I also document that households value this product, taking costly action to preserve their guaranteed access. Importantly, the extension of this additional credit improves loan repayment rates and MFI profitability, suggesting that this product can be sustainably extended to households already connected to microcredit networks.

The second chapter examines the problem of moral hazard in employer–employee contracting and how this may be an important barrier to firm efficiency and growth in the developing world. To do so, we run an RCT with a fleet of 255 minibuses

(matatus) in Nairobi, Kenya, where we introduce monitoring devices that track real-time vehicle location, daily productivity, and safety statistics. We randomize whether minibus owners have access to these monitoring data using a novel mobile app. This information allows owners in the treatment group to observe a more precise signal of driver effort, the amount of revenue drivers collected in fares, and the extent to which the driver engages in reckless driving. We find that treated vehicle owners modify the terms of the contract by decreasing the rental price they demand. Drivers respond by working more hours, decreasing behaviors that damage the vehicle, and under-reporting revenue by less. These changes improve firm profits and reduce management costs, thereby helping treated firms grow. The device also improves owners’ trust in their drivers, which drivers say makes their job easier. Finally, we investigate whether these gains to the company come at the expense of passenger safety, in an environment where accidents are common. While we do not find any evidence that conditions deteriorate, offering detailed information on driving behavior also does not improve safety. Only by incentivizing drivers through an additional cash treatment do we detect safety improvements.

Finally, the third chapter investigates the problem of poor road safety that plagues the developing world. Road traffic accidents are rapidly increasing, claim- ing more than 1.35 million lives annually and causing up to 50 million injuries. However, little is known about effective ways to improve road traffic safety in con- texts where policing and government capacity are limited. To address this problem, we use an RCT to examine whether dissemination of safety information to passengers of public transit vehicles can reduce unsafe driving. We find that passengers are not more likely to choose a safe bus after receiving safety information, but are will- ing to switch their bus choice in response to monetary incentives. However, there is some evidence that passengers who care about safety are more likely to choose a bus they already believed was safe ex-ante, suggesting that the failure to respond to safety information may be due to skepticism about the quality of safety information provided. This emphasizes the importance of disseminating information through channels that every-day passengers are familiar with and trust.

Jointly, the three parts of this dissertation aim to apply careful research strategies, in conjunction with economic theory, to provide policy-relevant evidence to- wards improving outcomes in the developing world.