Sanitation and Health: Empirical evidence for Brazilian Municipalities (with Enlinson Mattos, Cristine Pinto and Lucas Teixeira) - forthcoming Brazilian Review of Econometrics


Working papers

How does the presence of organized crime affect the intensity of the COVID-19 epidemic? Rio de Janeiro is the perfect laboratory to answer such an overlooked question because two distinct types of criminal organizations operate in its territory. First, there are drug trafficking gangs comprised of slum dwellers. Second, there are the milícias, paramilitary groups with links to the police, usually financed by extortion.  We estimate differences-in-differences specifications that compare the number of cases (deaths) caused by Severe Acute Respiratory Syndrome (SARI) before and after the COVID-19 outbreak in neighborhoods with and without the presence of organized crime. We document two findings. First, neighborhoods controlled by gangs have 43% less SARI deaths and 46% less SARI hospitalizations than areas without any form of organized crime. Second, neighborhoods controlled by milícias have 29% more SARI deaths and 34% more SARI hospitalizations than areas without any form of organized crime. We show that organized crime may take different actions against a public health crisis depending on its economic and political relations with the local communities under their control.


Can entertainment media undermine dictatorship? Evidence from Brazil's novelas (with Alberto Chong, Claudio Ferraz, Frederico Finan and Eliana La Ferrara)

The role played by mass media in the rise and maintenance of non-democratic regimes is a phenomenon well documented in the political economic literature. Anecdotal evidence suggests that Rede Globo, the leading television broadcaster in Brazil served the military regime that ruled Brazil between 1964 and 1985 by distorting the news. The theoretical literature on media capture, however, suggest that media outlets might face a more complicated trade-off: when media outlets bias news to serve autocracies, individuals might disengage from watching TV and advertising revenues decrease. This naturally creates an incentive for media outlets to invest in more appealing entertainment shows.


This paper investigates if entertainment TV can undermine the support for autocratic regimes when dissenting views permeate entertainment shows. This is done by examining whether the entry of Globo TV and its novelas induced changes in the vote share of the ruling party during Brazilian dictatorship in mayoral elections held between 1972 and 1982. We employ a difference-in-difference strategy exploiting variation in the timing of Globo's entry across municipalities. To separately identify the role of novelas and news, we code the content of both novelas and news using machine learning algorithms and exploit changes in content over-time. Our results show that on average, the entry of Globo in a municipality is associated with lower vote share for the authoritarian party during Brazil’s dictatorship, with the effect being larger in localities with more households with TVs. We also document that the reduction in vote share for the authoritarian party increases with the availability of political content in the novelas.


We investigate how powerful non-openly partisan media can change the trajectory of an election by unexpectedly manipulating a single piece of news - the coverage of a political debate during the 1989 Brazilian presidential race. Using geographical distribution of broadcaster-specific TV signals and the timing of election events, we find that the left-wing candidate lost 4.1 p.p. in vote share due to the misleading reporting, potentially costing him the election. We provide evidence that our results are best explained by viewers switching candidates, as their beliefs of who won the debate were swayed by the manipulation.

See it to believe it: the effect of incentives on students' test scores (with Amanda Arabage, Cristine Pinto and Vladimir Ponczek)

In this paper we investigate how students' response to an incentive program varies according to their level of information about the program and to their socioemotional characteristics. We do so by examining the effect of an incentive experiment implemented in Elementary Public Schools in the municipality of Santo André, Brazil. The experiment provided both collective and individual non-financial incentives to students based on their/their classroom's performance on numerical and verbal cognitive tests. We found evidence that incentivized students perform worst than students that did not received incentives, consistent with the theoretical prediction that incentives could undermine students' intrinsic motivation. However, by exploiting the fact that students were exposed to the incentive scheme both in the beginning and the end of the year, we find evidence that students that saw more classmates being rewarded in the first intervention respond more to incentives in the second intervention. We interpret this result as an evidence that students with a higher level of knowledge about the incentive scheme are more likely to respond to it. Finally, we exploit variation from students socioemotional skills and provide evidence that students who are more open to experiences respond more to the incentive scheme.

  Publications in book chapters (in Portuguese)

Sistemas de Accountability nas Escolas Públicas Brasileiras: identificando a eficácia das diferentes experiências. In: Reynaldo Fernandes; André Portela Souza; Luiz Guilherme Scorzafave; Fernando Botelho. (Org.). Políticas Públicas Educacionais e Desempenho Escolar dos Alunos da Rede Pública de Ensino. 1 ed. : FUNPEC, 2014, v. 3, p. 59-80


O Impacto das políticas de utilização de informática na aprendizagem dos alunos. In: Reynaldo Fernandes; André Portela Souza; Luiz Guilherme Scorzafave; Fernando Botelho. (Org.). Políticas Públicas Educacionais e Desempenho Escolar dos Alunos da Rede Pública de Ensino. 1 ed. : FUNPEC, 2014, v. 3, p. 59-80