Agricultural subsidies and global greenhouse gas emissions

The paper investigates how current agricultural support policies influence global GHG emissions from agricultural production. Despite large public support to agriculture (about US$553–600 billion annually), there has been no rigorous quantification of its impact on emissions. The study addresses this by examining: (i) average support rates; (ii) differences across support types (coupled subsidies versus border measures); (iii) heterogeneity across commodities and countries given wide differences in emission intensities; and (iv) how support affects production methods and input use. Because emissions are highly concentrated in a few commodities (beef, dairy, rice) and support often targets these sectors, the net effect is ambiguous and requires a general equilibrium, multi-commodity, multi-country assessment. The purpose is to quantify how current subsidies and trade measures affect output, location of production, and emission sources, and to explore the potential of repurposing support toward technologies that reduce emission intensities.

The study situates itself within work on agricultural support and environmental impacts. OECD monitoring documents substantial support and its composition across countries. A related analysis (Mamun et al.) suggests output of individual commodities is more responsive to differential support than aggregate output is to average support; it also finds that, over 1993–2015, average support rates for high-emission commodities were often below those for lower-emission commodities. Literature on emission intensities shows strong heterogeneity across countries and products and links higher productivity to lower emission intensities (Tubiello). Evidence on technologies suggests feasible emission reductions via dietary supplements for ruminants and alternate wetting and drying (AWD) in rice (Mernit; Chidthaisong). High returns to agricultural R&D and its strong poverty reduction impacts (Alston et al.; Ivanic & Martin) support repurposing subsidies toward innovation. Rebound effects from productivity gains remain limited for food due to low demand elasticities (Valin et al.).

The authors construct a global assessment using IFPRI’s MIRAGRODEP computable general equilibrium (CGE) model, an extension of MIRAGE. The model is multi-region, multi-sector, with a CES-LES demand system calibrated to estimated income and price elasticities; Armington CES nests allocate demand between domestic and imported varieties. Production uses Leontief intermediates and a CES value-added nest (unskilled labor and a composite of skilled labor-capital), with land and natural resources in agriculture and mining. The underlying database is GTAP v10 (pre-release 3) for 2014 with 141 regions and 65 products, updated to a 2020 baseline using UN demographics and IMF growth outlooks. Agricultural support data incorporate OECD measures of domestic support and bilateral protection, adjusted for major trade preferences. An emissions database is created by reverse-engineering FAOSTAT emissions to a full matrix by country, commodity, production stage, and emission source (e.g., enteric fermentation, manure management, rice cultivation, synthetic fertilizers, crop residues, burning, pesticides). Emissions of CH4 and N2O are converted to CO2e using 21 and 310, respectively. Missing activity data for synthetic nitrogen fertilizer use are reconstructed by combining FAOSTAT totals with IFA nutrient-by-crop data for 54 countries and scaling to all countries. Emissions from livestock are allocated between joint products (meat, milk, wool) by product value. The final database provides emission intensities by commodity and location. A post-solution module links MIRAGRODEP outputs (by commodity and inputs) to the emissions database to compute CO2e by source. Policy experiments compare observed outcomes with a counterfactual without support, focusing on two instruments: (1) coupled subsidies tied to output and input use; and (2) border measures (market price support via trade protection). The model is run in static mode with macro assumptions designed to isolate policy impacts: fixed total employment, constant aggregate real public expenditures with a consumption tax ensuring a fixed government budget share of GDP, constant land use (to focus on production emissions), and no dynamic investment effects. Decoupled payments are assumed not to alter output unless accompanied by environmental conditions; general services support (GSS/GSSE) is acknowledged but not directly simulated except in technology scenarios. Technology repurposing scenarios illustrate potential gains: (i) a 30% reduction in emission intensities for covered production with minimal cost reduction (to enable adoption) and no productivity increase; and (ii) a 30% reduction in emission intensities coupled with a 30% reduction in input needs (productivity increase), capturing potential rebound effects through lower prices and higher demand.

• Emission intensities vary widely: for bovine meat, from 12.1 kg CO2e/kg in the United States to 108.3 kg CO2e/kg in India, with developing countries generally exhibiting higher intensities due to lower productivity.
• Coupled subsidies increase global agricultural output by 0.9% and raise agricultural GHG emissions by 34,420 kt CO2e (+0.6%) relative to no-subsidy baseline. Roughly one-third of the emissions increase is due to higher synthetic fertilizer use. Effects are similar in magnitude across developed and developing countries.
• Border measures have minimal impact on global output (+0.1% globally; +0.6% in developed, −0.1% in developing countries) but reduce global agricultural emissions by 127,635 kt CO2e (about −2.1%). Protection raises consumer prices in protecting countries, lowering demand and shifting production toward lower-emission-intensity locations; negative protection in some developing countries reduces high-intensity bovine meat output.
• Combined, current coupled subsidies and border measures increase global farm output by about 1.1% but reduce global agricultural emissions by 102,071 kt CO2e (≈ −1.7% vs. no-support counterfactual). Emission reductions arise mainly from compositional and locational shifts in production toward lower-intensity producers, notably reduced bovine meat output in Brazil (−18%), India (−32%), and Australia (−31%), partially offset by increases in the EU and China.
• Support patterns are heterogeneous: some high-income countries exhibit high border support (e.g., Japan ~57%, Norway ~63% of output value), while India has negative market price support (−12%) with sizable coupled subsidies (~7%). Globally, 2017–2019 average rates were ~5.5% (coupled subsidies) and ~5.7% (market price support).
• Technology repurposing scenarios: a 30% reduction in emission intensities (for covered production, ~two-thirds of global agricultural emissions) without productivity gains lowers global emissions by nearly 20%. When the same 30% emission-intensity reduction is paired with a 30% productivity increase, net global emission reduction is just under 10%, reflecting a modest rebound effect due to low food demand elasticities.
• Results exclude emissions from land-use change; including them would likely increase the estimated benefits of productivity-enhancing, emission-saving innovations due to reduced land footprint.

The research quantifies how current agricultural support influences global GHG emissions from production by contrasting the status quo with a no-support counterfactual. It finds that production-stimulating coupled subsidies modestly raise both output and emissions, while border measures reduce emissions primarily through demand-side effects in protecting countries and by shifting production toward lower-emission-intensity regions. On balance, the current mix of policies slightly increases output but modestly reduces emissions, implying that simply abolishing support could slightly increase emissions. Given the concentration of emissions in ruminant products and rice, and large cross-country intensity differences, the main emission effects operate via changes in product mix and geography of production. Policy relevance: The modest net effect of current support on emissions suggests that substantial agricultural emission reductions require repurposing support toward instruments that directly target emissions—e.g., pricing emissions, incentivizing low-emission consumption, and especially funding R&D to reduce emission intensities and support adoption of climate-smart practices. Because food demand is relatively inelastic, productivity-enhancing, emission-reducing innovations are likely to yield sustained net emission reductions, potentially augmented when considering land-use change and carbon sequestration benefits.

This study contributes a rigorous, model-based quantification of the impacts of current agricultural support on global agricultural GHG emissions, integrating a newly constructed emissions-by-commodity-and-source database with a global CGE model. It shows that coupled subsidies modestly increase emissions, border measures reduce them, and the combined effect is a small net reduction relative to a no-support world. The results indicate that simply eliminating current support would not meaningfully reduce emissions and could increase them slightly. Instead, meaningful mitigation requires repurposing existing support—shifting resources toward R&D to lower emission intensities, facilitating adoption of climate-smart practices, and potentially implementing targeted instruments such as GHG pricing in agriculture and demand-side measures. Future work should integrate land-use change and carbon sequestration dynamics, explore granular policy mixes across countries and commodities, and evaluate welfare, food security, and distributional outcomes alongside emission targets.

• Emissions from land-use change and carbon sequestration in soils and forests are excluded; land area is held constant, likely understating benefits of productivity improvements.
• Static modeling framework omits dynamic investment and technology diffusion pathways; results reflect comparative statics rather than long-run equilibria.
• Emissions database relies on FAOSTAT/IPCC Tier 1 emission factors and constructed activity data (e.g., fertilizer use by crop), which may introduce measurement error and aggregation bias.
• Decoupled support and GSSE are not directly modeled for emission impacts beyond illustrative technology scenarios; environmental conditionalities in existing programs are only qualitatively considered.
• Coverage of support measures and emission intensities is partial (roughly two-thirds of emissions fully represented), affecting the scale of scenario-driven reductions (e.g., 30% intensity cuts map to ~20% global reductions). Adoption constraints for new technologies are stylized in scenarios.