Document Type

Discussion Paper

Date of Paper



When agricultural productivity improves, farmers may react by expanding farming and further encroach on forest lands, or they may choose to intensify and produce more output with less land. We specify the conditions under which agricultural productivity can have such ambiguous effects on deforestation. We then examine the predictions of that model using county-level data from five waves of the Brazilian Census of Agriculture and satellite-based measures of land use. We identify productivity shocks using exogenous variation in rural electrification in Brazil during 1960-2000. We show that locations suitable for hydropower generation experienced improvements in crop yields, and that credit-constrained farmers subsequently shifted away from land-intensive cattle-grazing and into cropping. As a result, agricultural land use declines, more native vegetation is protected, and these effects persist 25 years later in both census and satellite data. Brazil’s deforestation rate would have been almost twice as large between 1970 and 2000 without that increase in agricultural productivity. That makes the conservation benefits of productivity improvements comparable to the most prominent conservation packages ever implemented in Brazil.


We thank Arthur Bragança, Francisco Costa, Taryn Dinkelman, Torfinn Harding, Seema Jayachandran, Dana Kassem, Melanie Morten, Richard Hornbeck, Alex Pfaff, Eustáquio Reis, Mark Rosenzweig, Edson Severnini, Jillian Stallman, Joe Shapiro, Gabriel Ulyssea, and seminar participants at Stanford University, the NBER Summer Institute (Energy and Environmental Economics), Yale University, Pontifical Catholic University of Rio de Janeiro (PUC-Rio), Duke University, Pontifical Catholic University of Chile (PUC-Chile), Universidade de São Paulo (USP), Fundação Getúlio Vargas Rio de Janeiro (FGV/EPGE) and São Paulo (FGV/EESP), Universidade Federal do Rio de Janeiro (UFRJ), UC–Santa Barbara, University of Zurich, the Econometric Society Summer European Meeting, Association of Environmental and Resource Economics (AERE) Annual Meetings, CESIfo Energy and Climate Area Conference, NEUDC for valuable comments and discussions. We also thank Steven Helfand for kindly sharing data from the Brazilian Census of Agriculture.