Essays on Brazilian Political Economy

Abstract

The article 1 addresses the role of political parties and studies whether popularity shocks are crucial for electoral accountability beyond their effects on voter behavior. It specifically examines the impact of the revelation of information about a government's conduct on the types of candidates who stand for election. The empirical test focuses on the Brazilian city council elections in 2004 and 2008. The identification approach exploits the randomness of the timing of the release of audit reports on the use of federal funds by municipal governments. The study finds that when the audit reveals a high level of corruption (i.e., when it represents a negative popularity shock), the parties supporting the incumbent select more educated candidates. On the contrary, parties pick, on average, less educated candidates when the audit finds low levels of corruption (i.e., when it represents a positive popularity shock). These effects are stronger in municipalities that have easier access to local media. The evidence confirms that parties are strategic players that consider specific features of the electoral competition when making decisions and that their decisions are affected by shocks that influence the electoral race.

The article 2 is devoted to examining aspects of the voter preference assumptions. Citizen assessment of government performance is a cornerstone of successful democratic functioning. However, accountability is a double-edged sword. When voters misunderstand the stakes, and provide the wrong incentives to elected officials, political accountability leads to an implementation of suboptimal welfare policies. This paper reveals that an electorate can demand clientelism. To address this question, I study the behavior of voters in a context of vote-buying in Brazilian politics known as the drought industry. The data cover the Brazilian democratic elections from 1998 to 2012, and as empirical strategies I implement both fixed-effects models with panel data and a regression discontinuity design with heterogeneous treatment effects. I find evidence that after a drought, voters increase the vote share of local incumbent parties that are politically aligned with the central government to ensure the inflow of partisan government aid relief. Such behavior reinforces the central government's incentives to bias policies in favor of politically aligned municipalities to influence elections. Consequently, the cycle of distortion of aid relief allocation is perpetuated.

In connection with the findings indicating that the incidence of droughts and the Brazilian political economy are directly linked, the article 3 investigates the behavior of the local governments regarding the level of corruption. The analysis studies whether the allocation of aid relief policies increases the level of corruption in the context of natural disasters. More specifically, the study investigates the number of federal emergency declarations against droughts, as a proxy for aid relief, and the number of irregularities in the local governments' expenditures found by auditors in Brazilian cities during a full mayor's term. The study implements an instrumental variable approach exploiting the quasi-random nature of the cycle component of a municipality's aridity relative to its trend. The findings show that an additional recognition of the state of emergency leads to an increase in corruption per capita for an entire term of a mayor.