Essays on the Political Economy of Development

Abstract

This dissertation focuses on understanding four phenomena that affect political accountability in democratic regimes and the capacity of governments to produce public goods in many developing countries: misinformation, corruption, violence, and coordination between (subnational) governments. Each of these factors challenges political accountability by distorting the way voters and representatives relate to each other, how officials self-select into public office, how public goods are produced —and therefore how officials are evaluated by voters—, and how malpractice is confronted. While I use the Mexican context to provide empirical support to the theoretical arguments presented, throughout the dissertation I also discuss how my findings travel to other contexts, particularly developing democracies.

The first chapter of the dissertation presents experimental evidence from a massive online anti-corruption campaign. My co-authors and I evaluate the effect of increasing the share of the electorate receiving the same piece of political information on voter behavior. More specifically, we evaluate a massive, non-partisan campaign that distributed information about municipal spending irregularities, which often constitute corruption, to 2.7 million Facebook users ahead of the 2018 elections in Mexico. We find that the vote shares of incumbent parties with little or no spending irregularities increased by six to seven percentage points (~0.5 SDs) in targeted electoral precincts (direct effect). Also, we find that within treated municipalities, vote shares increased by three percentage points (~0.25 SDs) in untargeted precincts (spillover effects). Based on a design that exploits variation in the share of the electorate that is targeted by the ad campaign (0%, 20%, or 80% of Facebook users in the municipality), we estimate that most of the effects of the Facebook ad campaign are driven by high-saturation information campaigns (80% saturation), where online and offline interactions amplify effects. Overall, this project generates evidence that (1) non-partisan campaigns can be effective, (2) offline interactions produce indirect effects that compare, in size, to direct effects, and (3) the effect of an information campaign is increasing in the share of voters targeted in the constituency as community discussion and voter coordination arise in response to the information intervention.

In the second chapter of the dissertation, I advance our understanding of how criminal organizations attempt to obtain political influence. More specifically, I provide an explanation of why criminal groups use violence against politicians during electoral times. I argue that criminal groups use violence when other channels of influence, namely, bribes, are no longer available. The evidence in favor of this argument is twofold. First, relying on an entry model with asymmetric information, I show that the criminal organizations' use of bribes decreases, and the use of violence increases, when the efficacy of bribes decreases relative to its costs. Second, I empirically test the predictions of the model in the context of a national-level reform that increased the cost of bribing politicians in Mexico. I use confidential data on suspicious financial transaction reports to measure bribes and an original dataset of violent attacks on politicians to measure violence. My findings suggest that the bribe-reducing reform decreased bribes but increased the number of attacks against politicians. Consistent with the model, I also find that this effect is more prevalent in places where the marginal benefit of bribes is higher (i.e. places where politicians are more financially constrained by narrow campaign spending limits), and places where politicians have little information about the criminal(s) group(s) (i.e. places where a criminal group has not had a hegemonic presence). Taken together, this chapter documents the unintended consequences of a bribe-reducing reform and calls for attention to consider the objectives and constraints of both illegal and legal actors when designing, implementing, and evaluating transparency-enhancing reforms.

Finally, in the last chapter of the dissertation, I analyze how policies may generate spillovers in multi-level government settings, which present a threat to the identification of causal effects, and how the lack of coordination between subnational governments may produce unintended consequences. I focus on two cases: (1) social distancing policies at the outset of the COVID-19 pandemic in 2020, and (2) the introduction of an automated speed ticket program by a local authority that did not have the capacity to penalize out-of-the-state drivers. In the first part of the chapter, my coauthors and I provide evidence of spillovers across social distancing policies enacted by municipal governments in Mexico during the COVID-19 pandemic. Using fine-grained mobility data from Facebook and state-of-the-art machine learning techniques, we show that evaluating policies in isolation may lead to wrong conclusions as there are (unquantified) spillovers arising from the lack of coordination between local governments. In the second part of the chapter, my co-author and I study the introduction of an automated speeding ticket system in Mexico City. We find that the introduction of speed-detecting cameras led out-of-the-state drivers to engage in riskier driving behavior, which is reflected in more tickets (both automated and non-automated) and accidents. In general, this chapter provides evidence that coordination between subnational governments is essential when designing and implementing optimal public policy packages.