Francisco Costa

Assistant Professor of Economics at the University of Delaware 

Invited Researcher at J-PAL (LAC & K-CAI) 

Working Papers

Efficient Forestation in the Brazilian Amazon, with Rafael Araujo and Marcelo Sant'Anna.  [Coverage: VoxDev] R&R at Review of Economic Studies.

This paper estimates the Brazilian Amazon’s carbon-efficient forestation -- i.e. when farmers internalize the social cost of carbon. We propose a dynamic discrete choice land use model and estimate it using a panel of land use and carbon stock of 5.7 billion pixels between 2008 and 2017. Business-as-usual implies an inefficient release of 44 billion tons of CO2 in the long run resulting from the deforestation of an area twice the size of France. We show that targeted mitigation efforts in areas with the largest potential for emission reductions can mitigate up to 13 billion tons of CO2 at a cost under $10/ton. However, the abatement cost curve is convex, making preserving the whole forest an expensive task.

National Borders and the Conservation of Nature, with Robin Burgess and Benjamin Olken. submitted. [new version]

Tropical forests slow climate change, and their conservation is an international priority. We propose using finely grained satellite data at national borders, where one political jurisdiction ends and another begins, to evaluate the net impact of national policies on conservation. Using 30x30 meter satellite data along Brazil’s 12,800 km border in the Amazon, we find dramatic changes in deforestation rates that match changes in Brazilian policies. Between 2001 and 2005, at the tail end of a pro-exploitation period, annual Brazilian deforestation was more than three times the rate across the border. From 2006 to 2013, as Brazil introduced policies to reduce deforestation, these differences at the border disappear. But they re-appear starting in 2014, amid a period of dismantling environmental regulation. National borders offer a means of evaluating the effectiveness of national conservation policies, which are now objects of international interest.

Going Viral: Public Attention and Environmental Action in the Amazon, with Rafael Araujo and Teevrat Garg. submitted.

International agreements to reduce anthropogenic environmental disasters rely on public pressure to drive local action. We study whether focused media and increased public outcry can drive local environmental action, reducing environmental damage. Although an annual affair, forest fires in the Brazilian Amazon went viral in August 2019, receiving unprecedented public scrutiny. Comparing active fires in Brazil versus those in Peru and Bolivia in a difference-in-differences design, we find that increased public attention reduced fires by 22% avoiding 24.8 million tons of CO2 in emissions. Our results highlight the power of public attention to compel local action on pressing environmental issues.

The Deforestation Effects of Trade and Agricultural Productivity in Brazil, with Igor Carreira and Joao Paulo Pessoa.  R&R at Journal of Development Economics. 

This paper quantifies the relative footprint of trade and agricultural productivity on deforestation in Brazil between 2000 and 2017. Using remote-sensing data, we find that these two phenomena have distinct effects on land use. Greater exposure to new genetically engineered soy seeds is associated with faster deforestation through the expansion of cropland. We find no association between exposure to demand from China and deforestation – although, trade induces conversion of cropland to pastureland. Our results suggest that, when taken together, agriculture productivity gains, and not trade, were the main driver of deforestation and the expansion of the agriculture sector.

Advanced Work-in-Progress

The Environmental Costs of Political Interference: Evidence from Hydropower Plants in the Amazon, with Juliano Assunção and Dimitri Szerman. 

This paper estimates the impacts of ten recently built hydroelectric power plants in the Brazilian Amazon on deforestation. Using the inventory of all sites with hydropower potential without a power plant, we employ the synthetic control method to identify the causal impact of each power plant on forest loss. Overall, the construction of the ten plants was responsible for 21 percent of the observed forest loss in areas within a 50-kilometers radius of the construction site. Notably, this effect is solely attributed to the four plants where construction licenses were granted despite technical recommendations to deny them due to environmental risks. The remaining plants, which obtained technical clearance from the environmental agency, have negligible effects. These findings demonstrate the effectiveness of robust environmental regulations and highlight the imperative to address political interference to ensure sustainable energy development while minimizing environmental damage.

Technology Adoption and Structural Transformation: Evidence from the Industrialization of the Sugarcane Sector, with Francisco Luis Lima. 

This paper shows how the adoption of agricultural mechanization can prompt structural transformation in a middle-income economy. We study the fast spread of mechanical harvesting that followed the prohibition of pre-harvest field burning in the sugarcane sector in São Paulo state, Brazil. We combine remote-sensing data on sugarcane production and census data to estimate the impacts of field mechanization on the local labor markets. We find that the adoption of mechanical harvesting led to the industrialization of the field; one standard deviation larger adoption of agricultural mechanization reduces the share of workers employed in the agricultural sector by 2.3 percentage points, and increases the employment share of the manufacturing and services sectors by 1.7 and 1.1 percentage points, respectively.

Too hot to learn? Evidence from grade progression and school dropouts, with Diana Goldemberg. R&R at Economic Inquiry. 

This paper studies the impact of cumulative heat exposure on the dropout and grade promotion of children from 6 to 18 years old in Brazil. As common in other middle-income countries, while primary school enrollment is close to universal, dropout rates are very high, especially for teenagers. We find that extreme heat on school days increases school dropouts of high school students, mostly those in public schools and urban areas. The effects of heat on the grade progression of younger students are small, but estimates are noisy. Our main contribution is demonstrating that, in a setting where school is already an unpleasant experience with low perceived returns, events that worsen the experience on the margin (warmer days) may drive teenagers out of school with long-term implications.

(May, 2023)