Working Papers
(with Martin Mattsson, Nathalia Sales )
November 2025
Abstract (click to expand): We extend Brazilian municipal corruption data using a Large Language Model (LLM) to analyze 2,197 audit reports, totaling 185,000 pages. LLM-generated measures correlate nearly as strongly with manual classifications as different manual classifications correlate with each other, highlighting both the relative reliability of the LLM classification and the inherent subjectivity involved in quantifying corruption from textual sources. Reexamining reelection incentives' role in limiting corruption, we find effects that are consistent with prior studies, but smaller and more precisely estimated, with statistical significance for only one of three corruption measures. We document substantial temporal heterogeneity in effects across electoral terms, with reelection incentives reducing corruption from 2003 to 2004, the period analyzed by the previous literature, and from 2013 to 2015, but not from 2005 to 2012. We explore potential explanations for this pattern, including politician composition changes and evolving legal enforcement patterns.
(with Iván de las Heras, Santiago Saavedra )
Resubmitted at Journal of the Association of Environmental and Resource Economists
October 2025
Abstract (click to expand): While deforestation may yield short-term economic gains, it has lasting impacts on climate change, biodiversity, and water systems. The age of political leaders is a key determinant of how these tradeoffs are managed. We study closely contested elections in Brazil and show that when a young politician is in power there is less deforestation (p-value 0.000), lower greenhouse gas emissions (p-value 0.061), but higher local income (p-value 0.023). Results are partly explained by young mayors turning over the bureaucracy with younger public servants (p-value 0.043).
SocArXiv
(with Thiago Scot, Bernardo Ricca, Lucas Nascimento, Nathalia Sales )
Resubmitted at World Bank Economic Review
November 2025
Abstract (click to expand): This paper introduces a new disaggregated and harmonized dataset on public procurement and budget execution by Brazilian subnational entities, which covers half of Brazilian municipalities and spans the years 2003-21. This dataset provides key information that was previously unavailable from aggregate data, such as the identities of suppliers, details on purchases of goods and services, and granular information on the life cycle of each expenditure action. It then uses these data to provide new stylized facts about local public finance. First, it shows that about one-quarter of government purchases are locally procured and discusses implications for efficiency. Second, it demonstrates that close to 15 percent of payments exceed the 30-day threshold and that payment timeliness is systematically correlated with the income level of the municipality. Finally, it shows that municipalities where mayors have reelection incentives systematically employ more non-competitive tenders and buy more from local suppliers.
Project's website
World Bank Policy Research working paper WPS 10598
Dataset at Data Basis (Base dos Dados)
(with Luiz Bines, Juliano Assunção )
July 2025
Abstract (click to expand): This study documents how electoral behavior changes based on the salience of national security threats. Leveraging novel data on siren timing and location in Israel, we estimate a difference-in-differences design that compares localities brought into rocket range just before the 2015 election with otherwise similar areas. While almost all rockets from that period were intercepted by the Iron Dome defense system, the analysis shows that Red Alerts on the days immediately before the election boosted Likud’s vote share by 2.6 percentage points, while earlier alerts had no effect. Polarization increases as the effects are larger where Likud support was already higher. Analysis of Google search data shows that alerts increased interest in security-related topics but not in politics or parties, consistent with a salience rather than ideological mechanism.
Black-White Income Inequality During Jim Crow: Evidence from 'Passing' for White
(with Ludovica Mosillo, Emily Nix, Nancy Qian )
May 2025
Featured at: The Weeds (40:45), Kellogg Insight
Abstract (click to expand): This paper documents several new facts about racial discrimination from 1910-1940 in the United States. Exposure to racial discrimination was positively associated with some Black men taking on White identities, resulting in large income gains and economic mobility. However, this imposed high social costs; beyond the loss of one’s Black identity, passing for White was associated with higher separation rates from a person’s community and family. Using the income gains from passing, we estimate that removing contemporaneous labor market discrimination would have increased the Black-to-White income ratio by at least 16 percentage points.
(with Bernardo Ricca, Thiago Scot )
April 2025
Abstract (click to expand): We provide evidence of a new channel through which politicians can exchange favors with campaign donors: earlier payment in procurement contracts. We exploit an electoral reform in Brazil that bans corporate contributions and partially breaks down the relationship between donors and politicians. Using a within-firm difference-in-differences identification strategy, we find that connected firms experience longer payment terms post-reform. The effect is larger in municipalities with low liquidity, where payment delays are more common, and for contracts awarded through a competitive tendering process. Our results point to the importance of designing rules that curb discretion over payment dates in government purchases.
SSRN
(with João Pedro Vieira, Juliano Assunção )
August 2023
Abstract (click to expand): We study how environmental sanctions and spillovers improve forest conservation in the Brazilian Amazon. Using a difference-in-differences framework and novel farm-level data, we show that sanctions curbed deforestation and promoted reforestation among punished farmers and their neighbors. Heterogeneity analysis reveals that even sanctions with limited incapacitation potential elicited relevant behavioral changes. In particular, farmers’ responsiveness to sanctions coincided with the government’s commitment to enforcement. We do not find substantial evidence of spatial displacement or monitoring evasion. Overall, sanctions prevented 1.6 billion tons of CO2 emissions between 2006 and 2019, equivalent to 31% of US emissions in 2021.
SocArXiv
(with Pedro Bessone, Lisa Ho )
April 2021
Abstract (click to expand): What is the impact of mobile broadband internet on children's test scores? We compare standardized test scores before and after the staggered entry of 3G into Brazil's 5,570 municipalities using a heterogeneity-robust event-study design. We find no effects of mobile internet on test scores for 5th or 9th grade students, and can reject effect sizes of 0.04 standard deviations in both math and Portuguese. Taken together, our results indicate that the arrival of high-speed mobile internet is not suf- ficient to improve educational outcomes either through direct take-up by individuals or through broader changes to the economy.
Publications
(with Christiane Szerman )
Forthcoming at AEJ Applied Economics
May 2025
Featured at: BBC Brasil
Abstract (click to expand): Changes in political boundaries aimed at devolving power to local governments are common in many countries. We examine the economic consequences of redistricting through the creation of smaller government units. Exploiting reforms that led to sharp variations in the number of government units in Brazil, we show that voluntary redistricting increases the size of the public sector, public services delivery, and economic activity in new local governments over the long-term. These benefits are not offset by losses elsewhere and are stronger in peripheral and remote backward areas that are neglected by their parent governments. We provide evidence that the decentralization of decision-making power boosts local development in disadvantaged areas beyond simply gains in fiscal revenues.
(with Laura Schiavon, Thiago Scot )
Review of Economics and Statistics Volume 107 Issue 2
March 2025
Abstract (click to expand): In the absence of strong incentives, public service delivery crucially depends on bureaucrat selection. Despite wide adoption by governments, it is unclear whether civil service examinations reliably select for job performance. We investigate this question focusing on state judges in Brazil. Exploring monthly data on judicial output and cross-court movement, we estimate that judges account for at least 23% of the observed variation in number of cases disposed. With novel data on admission examinations, we show that judges with higher grades perform better than lower-ranked peers. Our results suggest competitive examinations can be an effective way to screen candidates.
(with Diego Cardoso )
Economic Letters, Volume 237, April 2024, 111673
March 2024
Abstract (click to expand): Cost-benefit analyses in public health typically calculate the benefits of mortality reduction interventions by multiplying the Value of a Statistical Life (VSL) and the expected decrease in fatalities. This procedure approximates the benefits of small mortality changes but is inaccurate for large risk changes because it holds constant the VSL---an estimate of a marginal rate of substitution. Building on the theoretical framework of the VSL, we outline a practical approach to calculate the benefits of non-marginal mortality reductions with empirically-calibrated compensating variations. We derive closed-form expressions that are easy to calculate and only require widely-available mortality statistics.
(with Rafael H. M. Pereira, Renato S. Vieira, Fernando Bizzarro, Rogério J. Barbosa, Daniel Travassos Ferreira )
Electoral Studies, Volume 86, December 2023, 102690
September 2023
Abstract (click to expand): Transportation costs are an under-studied barrier to political participation. In many elections worldwide, subsidies to voter transportation are already provided or are under discussion. However, these types of incentives have not been rigorously evaluated. Here we examine possibly the world’s largest-ever intervention to lower these costs, the adoption of a fare-free public transit policy in Brazil during the 2022 national election, when about half of Brazilian voters were granted the right to use public transit for free on election days. However, while some cities adopted the benefit for both rounds of the election, others adopted it only for the second round. Using an event study design, we exploit this difference in adoption timing to examine the policy’s causal impact on voter turnout rates and human mobility levels. We find that fare-free transit increased ridership between 7.2% and 17.5% on election days, however, we estimate a precise and robust null effect of the policy on voter turnout (Coef. 0.03p.p. with standard error of 0.22p.p.). Our results illustrate that monetary transport costs may not always be a critical factor behind non-participation. Although reducing transportation costs improves access to polling places, we show that even a full transit subsidy may not be sufficient to increase voter turnout.
(with Arthur Bragança )
Journal of Public Economics, Volume 215, 104753
September 2022
Abstract (click to expand): Government policies may impact economic outcomes directly but also indirectly through effects on political incentives. This paper examines the effects of the PPCDAm -- a centralized set of environmental policies that effectively raised the expected cost of illegal deforestation -- on the behavior of a powerful special-interest group operating in the Brazilian Amazon: farmers. Using different identification strategies, we document that municipalities governed by farmer politicians experienced larger declines in deforestation, greenhouse gases (GHG) emissions, and violence than municipalities governed by other politicians after this set of policies were implemented. Our findings are consistent with the PPCDAm changing political incentives in a persistent way that amplifies its impact on environmental and social outcomes.
Work in Progress
Housing Microfinance and Climate Resilience in the Philippines
(with Teevrat Garg, Russell Toth )
Abstract (click to expand): Low-income households in the Philippines lack sufficient access to affordable and resilient housing to protect them from the physical and economic damages brought by increasingly frequent extreme weather events, such as typhoons and floods. We partner with a large microfinance NGO and an NGO focused on housing for low-income households, to evaluate a new housing microfinance product. We will run a 2x2 randomized evaluation that cross-randomizes the education component (versus no education) and the new housing loan terms (versus being offered the old housing loan terms), against a control group. The main outcomes are short-term impacts on perceptions of climate risk and resilience, knowledge and plans for house construction, and other measures of resilience. The study will create new knowledge about how to support low-income households in obtaining housing that is resilient to extreme weather events, which is relevant across a number of developing countries that are vulnerable to climate change.
AEA RCT Registry 9124
IPA summary
J-PAL summary
Killing the Forest: Assassinations and Deforestation in Brazil
(with Mariana Carvalho )
