Socio-economic and environmental trade-offs in Amazonian protected areas and Indigenous territories revealed by assessing competing land uses

Protected areas (PAs) are central to global biodiversity goals, including the 30×30 target, but their expansion has implications for competition with alternative land uses and for rural communities’ livelihoods. While conservation policy increasingly acknowledges socio-economic impacts, rigorous analyses that jointly assess environmental and socio-economic outcomes, and that benchmark protection against specific competing land uses, remain scarce. The Brazilian Legal Amazon (BLA) is a biologically critical and contested region where deforestation is driven largely by agricultural expansion and mining. In this context, the study evaluates how different protection regimes—strict protected areas (SPAs; IUCN I–IV), sustainable-use protected areas (SUPAs; IUCN V–VI) and Indigenous territories (ITs)—perform relative to alternative land uses in terms of deforestation and multiple socio-economic indicators (income, inequality, literacy, sanitation). The primary objective is to contrast environmental and socio-economic outcomes of these protection arrangements with outcomes under alternative land uses (sparsely populated, agriculture differentiated by landholding size, and licensed mining), thereby elucidating trade-offs and synergies to inform conservation and development decision-making in the BLA.

Prior work shows that PA effectiveness varies with governance type and context, with PAs capable of reducing deforestation by restricting alternative land uses (agriculture, logging, mining), but potentially affecting livelihoods by limiting resource access. Conversely, PAs can bolster livelihoods via tourism, ecosystem services that support agriculture, access to development projects, payments for ecosystem services (for example, Bolsa Floresta in Amazonas), and by granting management rights that enable sustainable resource use. Large-scale agriculture is responsible for most global deforestation and often offers limited local poverty reduction, sometimes exacerbating local poverty; small-scale agriculture typically causes less deforestation and retains benefits locally. Mining’s environmental impacts are variable in extent, and socio-economic effects are contested—potentially raising incomes, health, and education through jobs and infrastructure while increasing inequality and undermining other livelihoods. In the Amazon, deforestation peaked in the early 2000s due to soy and cattle expansion, declined after 2004 with policy interventions including PA expansion, but rose again after 2012, remaining below the 2004 peak. Pressures include PADDD events and increased mining inside or near PAs. Evidence suggests governance regime and location influence avoided deforestation success, and recognition of Indigenous land rights can deliver conservation benefits, though Indigenous communities often face structural barriers to socio-economic gains. These contexts motivate evaluation frameworks that benchmark protection outcomes against specific competing land uses to clarify socio-environmental trade-offs.

Study area and unit of analysis: The study covers 5,545 census tracts (CTs; mean ± s.e. 977 ± 44.7 km²) in the Brazilian Legal Amazon (BLA), using 2000 and 2010 population census years. Very small CTs (<50 km²; <1% of BLA) were merged with neighbors to improve detection of land-use impacts on residents. Because some CT boundaries changed between 2000 and 2010, the authors reconstructed 2010 data to 2000 CT boundaries by area-weighted overlays assuming homogeneous population distribution. Agricultural census boundaries (2000 vs 2006) were reconciled similarly. Treatments and controls: Treatments are CTs with ≥10% overlap by protection established between 2000 and 2010: SPAs (IUCN I–IV), SUPAs (subset of Brazilian sustainable-use categories equivalent to IUCN V–VI: FLONA/FLOE, ResEx, RDS), and ITs (inalienable Indigenous territories). CTs overlapping other protection types were excluded from each treatment analysis. Non-protected controls had <1% protected area overlap. Additional control subsets represented specific alternative land uses: sparsely populated (no licensed mines and <10% area in agriculture), agriculture-dominated CTs categorized by dominant property size (very small <10 ha; small 10–50 ha; medium 50–200 ha; large >200 ha) based on number of properties per CT from the 2006 agricultural census, and mining areas (CTs with licensed mines initiated after 2000 per SIGMINE; all CTs lacked mines at baseline 2000; protected CTs with post-2000 mines were excluded). Agricultural controls required ≥10% of CT area settled by agricultural properties; protected CTs with >10% settled land were excluded from those particular comparisons to maintain clear contrasts. Outcomes: Environmental outcome was PRODES-based cumulative deforestation (old-growth forest loss; patches ≥6.25 ha) for 2000–2010. Socio-economic outcomes at CT level were: mean monthly household income (inflation-adjusted via INPC), income inequality (Gini coefficient), literacy (percentage of literate household heads), and sanitation (percentage of households with poor sanitation, i.e., without toilets draining to sewage or septic tank). Confounders and design: As the observational design precludes definitive causal claims, the authors applied statistical matching and regression to reduce confounding. Matching balanced biophysical and socio-economic covariates: baseline poverty measures (income, inequality, literacy, sanitation), CT size, baseline forest cover, agricultural/mining suitability (slope, elevation, flood risk), travel time to cities >50,000 population, population density, state fixed effects, presence of mines before baseline; and, for agriculture comparisons, mines established during the study period; for mining comparisons, proportion of settled land. Post-matching regression models estimated treatment effects; heteroscedasticity-robust standard errors and multiple-testing corrections (Benjamini–Hochberg) were applied. Analyses of variance confirmed similar amounts of settled land and pasture among matched agricultural control groups across landholding sizes. Robustness checks: Nine robustness tests included tighter calipers, higher protection thresholds (≥50% CT overlap), restricting to 2006–2010 deforestation and post-2006 establishments, hidden-bias sensitivity (Oster bounds), spatial autocorrelation checks, controls for settled land and pasture, alternative deforestation data (Hansen et al. 30 m maps), varying spatial buffers for mining influence (up to 100 km), and an alternative fixed-effects panel regression framework. Results were consistent across checks, supporting the main conclusions.

Relative to all non-protected CTs: All protection types reduced deforestation between 2000 and 2010, with ITs showing the largest reductions (SPAs −53.6%, P<0.001; SUPAs −46.7%, P<0.001; ITs −69.0%, P<0.001). Income rose everywhere over time, but increases were 59.3% higher in SPAs versus matched controls (P<0.001); no significant difference for SUPAs; ITs had 24.8% smaller income gains than controls (P<0.001). No overall effects were detected on inequality or sanitation; SUPAs were associated with slightly higher literacy (+5.67%, P=0.01). Relative to agriculture and sparsely populated areas: SPAs and SUPAs reduced deforestation by roughly 40.5–81.6% versus agricultural land uses, except SUPAs vs very small-holder areas (−1.66%, P=0.92, non-significant); SPAs vs small-holder areas showed a marginally non-significant effect (−30.7%, P=0.08). ITs reduced deforestation substantially across all agricultural landholding sizes (−47.7% to −82.8%). SPAs achieved larger income increases than areas dominated by medium (+57.9%, P<0.0001) and large (+26.4%, P=0.03) landholders. ITs had smaller income increases compared with sparsely populated areas (−32.2%, P=0.02), very small (−35.8%, P<0.0001), and large (−29.6%, P=0.01) landholder areas. Inequality: SUPAs reduced inequality vs small-holder areas (−12.3%, P<0.001) but increased it vs medium-holder areas (+59.1%, P<0.001); ITs reduced inequality vs very small (−21.4%, P<0.0001) and small (−14.3%, P<0.001) holder areas; SPAs showed no significant inequality effect. Literacy: SPAs had smaller increases vs sparsely populated areas (−7.59%, P=0.03); SUPAs had smaller increases vs small (−8.49%, P=0.01) and medium (−13.9%, P=0.01) holders but greater increases vs very small (+1.98%, P=0.01); ITs had smaller increases vs small holders (−18.1%, P<0.001). Sanitation: SPAs had lower improvements vs sparsely populated (+3.34%, P=0.03) and large-holder areas (+20.9%, P=0.03) when measured as percent of households with poor sanitation; SUPAs had greater improvements vs very small-holder areas (−2.71%, P<0.001; negative indicates fewer households with poor sanitation). Relative to mining: All protection types reduced deforestation compared to CTs with licensed mines (SPAs −62.4%, P<0.001; SUPAs −76.5%, P<0.001; ITs −62.9%, P<0.001). SUPAs showed lower income gains than mining areas (−9.88%, P=0.04). ITs had higher inequality relative to mining (+9.13%, P=0.01). No significant differences were detected for literacy or sanitation relative to mining. Spatial analyses indicated environmental benefits of protection extend farther (up to 50–75 km) from mining sites than socio-economic trade-offs (typically 5–10 km). Overall, many deforestation reductions—especially for SPAs and SUPAs versus larger-scale agriculture—occurred without adverse socio-economic outcomes, whereas ITs often faced socio-economic trade-offs despite delivering the strongest forest conservation.

The study directly benchmarks protection outcomes against specific competing land uses, revealing heterogeneity in environmental and socio-economic effects. ITs consistently deliver the strongest avoided deforestation relative to agriculture and mining, underscoring conservation gains from recognizing Indigenous land rights. However, ITs tend to experience smaller gains in income and sometimes lower literacy improvements compared with alternatives, indicating socio-economic trade-offs that could be linked to market access and structural barriers. SPAs and SUPAs generally reduce deforestation relative to medium and large-scale agriculture without undermining local socio-economic development, challenging assumptions that large agricultural development necessarily benefits local populations more than protection-focused alternatives. SUPAs appeared ineffective relative to very small-holder areas in reducing deforestation during the study period, though their designation might mitigate future consolidation by large agribusiness. All protection types counteract deforestation from mining, but trade-offs arise for SUPAs (income) and ITs (inequality), with socio-economic effects more localized than environmental benefits. These findings inform policy by highlighting where protection can achieve environmental goals without socio-economic costs, and where complementary development interventions are needed—particularly in ITs—to avoid further disadvantaging Indigenous communities.

This work advances evaluation frameworks by comparing multiple socio-economic and environmental outcomes of distinct protection arrangements against explicit alternative land uses. Key contributions are: (1) demonstrating that ITs deliver the most consistent deforestation reductions, though often with socio-economic trade-offs; (2) showing that SPAs and SUPAs can substantially reduce deforestation relative to larger-scale agriculture without harming socio-economic outcomes; and (3) evidencing that all protection types mitigate deforestation from mining, with limited but notable socio-economic trade-offs for SUPAs and ITs. Policy implications include coupling land-rights recognition for Indigenous peoples with targeted development and improved access to social protection and education to enhance socio-economic outcomes. For 30×30 implementation, avoiding bias toward low-additionality areas and explicitly considering competition with agriculture and mining is critical. Future research should assess displacement (leakage) across regions, dynamic land-use transitions (e.g., smallholder consolidation), mechanisms linking protection to local wages and services, and broader spatial spillovers of both environmental and socio-economic outcomes.

The observational design limits causal inference despite matching and regression adjustments; unmeasured confounding cannot be entirely ruled out (addressed via sensitivity analyses). Deforestation is measured using PRODES, which detects only old-growth forest loss and omits secondary forest dynamics (tested with an alternative 30 m dataset). Agricultural landholding categories rely on 2006 census data applied across 2000–2010, potentially missing temporal changes in farm size structure. Analyses considered only officially licensed mining (SIGMINE); illegal mining and heterogeneity across mine types were not assessed due to small sample sizes and data limits. Protected CTs that also had mines were largely excluded, and differentiation by resource type/scale was not possible. Some CTs were merged, and reconstruction of 2010 data to 2000 boundaries assumes homogeneous within-tract distributions. Certain PA categories (APAs, ARIEs, RPPNs) were excluded due to scope and comparability concerns. Potential leakage of deforestation to other locations and full extents of spillover effects warrant further study.