Unveiling the environmental impact of the Amazon Soy Moratorium on soy-driven deforestation in MATOPIBA, Brazil; A causal estimation of the policy effect

By Isaura Menezes de Oliveira Guido

Abstract
Achieving sustainable food production while ensuring food security remains a critical global challenge. As a key agricultural commodity, soybean contributes to economic growth and global food security. However, its expansion often raises environmental concerns, particularly in ecologically sensitive regions. The Amazon Soy Moratorium (SoyM) represents the first major private-sector zero-deforestation commitment, and it has been credited with reducing deforestation in the Amazon. However, the policy’s narrow geographical focus and emphasis on a single commodity raise concerns it may induce leakage—the displacement of agricultural activities to less regulated regions in response to stricter policies. While there is evidence of leakage within the Amazon, its impact on MATOPIBA—Brazil’s latest agricultural frontier and home to extensive remnant Cerrado vegetation—remains unexplored. Additionally, most land use and land cover (LULC) change studies rely on predictive or econometric models, which often fail to explicitly capture the causal mechanisms driving these changes, preventing causal understanding and systematic confounder selection to reduce bias. This study addressed these gaps through a literature review, counterfactual scenario analysis, and heterogeneity analysis. It identified four potential leakage types in Brazil—activity leakage, land-market leakage, commodity-market leakage, and supply-chain leakage— and developed the first known causal diagram describing leakage mechanisms in MATOPIBA. The counterfactual analysis reveals that, between 2007 and 2011, in MATOPIBA overall deforestation rate at the municipal level increased by 1.07 Km2 yr−1, while soy-driven deforestation rose by 2.78 km2 yr−1 compared to the counterfactual, reinforcing concerns about leakage in the region. The heterogeneity analysis highlights that Western Bahia, Southern Maranhão, and parts of Piauí were most affected, aligning with soybean expansion and revenue hotspots. Environmental policies, climatic conditions, and soybean profitability were significantly associated with deforestation in MATOPIBA, though their causal role remains uncertain. While further empirical research is needed to confirm specific leakage mechanisms and refine soy-driven deforestation models, this study provides quantitative evidence of MATOPIBA’s vulnerability to leakage, highlighting the need for a more holistic, spatially adaptive approach. Future research should refine the proposed causal model and expand the use of causal inference frameworks to better understand policy-driven LULC dynamics, both within and beyond their intended scope, and support more adaptive and efficient policymaking.