Beyond Local Effects: Spatial Spillovers of Transportation Infrastructure and Deforestation in Brazil

Abstract

This paper examines the relationship between transportation infrastructure and deforestation in Brazil. We combine exploratory spatial data analysis with standard econometric, spatial econometric, and machine learning using spatially disaggregated data covering all Brazilian biomes. Exploratory analysis reveals a strong spatial concentration of both deforestation and transportation networks, particularly in the Centro-Sul and Northeast regions. Conventional econometric results indicate a positive conditional association between transportation infrastructure and deforestation after controlling for a broad set of structural and institutional characteristics. However, explicitly accounting for spatial effects highlights the central role of spatial spillovers, interactions, and spatially correlated confounders, suggesting that the environmental impacts of transportation infrastructure extend beyond local boundaries. Finally, results from the machine learning evaluation show that incorporating spatial structure substantially improves out ofsample predictive accuracy. Overall, the findings underscore the importance of spatially informed empirical approaches for understanding, anticipating, and monitoring deforestation patterns associated with infrastructure expansion.