Double-edged sword: China's free trade agreements reinforces embodied greenhouse gas transfers in agricultural products

The study addresses how China's free trade agreements (FTAs) affect greenhouse gas (GHG) emissions embodied in China's agricultural imports. Agriculture is pivotal for food security but accounts for roughly one-third of global anthropogenic GHG emissions. Expanding international agricultural trade relocates emissions across borders, and production-based accounting under UNFCCC does not capture trade-embodied emissions. China’s agricultural imports grew from $19.5 billion (2000) to $219.8 billion (2021), making it the top global importer and a significant driver of embodied GHG transfers. With China having signed 19 FTAs with 26 partners since 2002 and expecting rising agricultural demand due to income growth, urbanization, and dietary shifts, quantifying the FTAs’ effects on embodied GHG transfers is a pressing empirical question. This study compiles embodied GHG emissions for 367 products from 119 countries (2000–2015) and employs a PSM-progressive DID design to causally estimate the impact of China’s FTAs on embodied GHG emissions from agricultural imports, providing evidence for environmental and trade policy.

Prior research has extensively quantified embodied GHG emissions in international trade using environmentally extended multi-regional input–output (EE-MRIO) models for global, regional, and country cases (e.g., Zhao et al., Liang et al., Mangir and Şahin). Some studies examine single gases or products. However, the influence of national policy instruments on trade-embodied emissions is less studied. Evidence on trade liberalization’s environmental effects is mixed: FTAs and regional arrangements (EFTA, ASEAN, Mercosur) can reduce emissions in some contexts, yet liberalization may weaken global mitigation by shifting pollution-intensive production to regions with laxer regulations and by lowering the efficiency of global mitigation. Policy instruments like coupled subsidies can increase emissions, while border measures may reduce them. China, as the largest agricultural importer and second-largest exporter, has influential agricultural trade policies (subsidies, tariffs, trade barriers), but the specific effect of China’s FTAs on embodied GHG transfers in agricultural trade remains underexplored. Given roughly 35% of China’s total trade involves FTA partners, assessing FTAs’ impact on embodied GHG in agricultural imports is critical for policy-making.

Scope and accounting: The study measures consumption-based embodied GHG emissions (CO2, CH4, N2O) in China’s agricultural imports using carbon emission factors (CEF) along full production chains. It adopts an improved, product-specific, time-varying CEF set by Hawkins et al. (2016), preferred over FAO’s activity-based factors. Agricultural activities are split into crops and livestock following FAO classifications. Feed grains are treated as inputs to livestock to avoid double counting.

GHG accounting for crops: Total GHG equals the sum over CO2 from inputs and energy use, CH4 from rice cultivation and combustion, and N2O from manure, residues, and combustion, computed as activity levels multiplied by corresponding emission factors.

GHG accounting for livestock: Emissions include inputs, enteric fermentation (CH4), and manure management (CH4, N2O). Formulas aggregate emissions across animal categories and sources via activity data and CEFs.

Empirical identification: A PSM-progressive DID model is used to estimate the causal impact of FTAs on log embodied GHG emissions (LnAGHG_it) from agricultural imports. Treatment equals 1 for partner countries in years when the FTA is implemented; otherwise 0. Progressive DID accommodates staggered FTA timing with country and year fixed effects. To mitigate selection bias, propensity score matching (1:1 nearest neighbor; robustness with kernel and radius caliper) is performed using covariates: GDP, bilateral distance (DTE), per capita arable land (PAL), trade openness (OPD), and agricultural mechanization level (AML). Parallel trends are examined via event-study; placebo tests randomize treatment timing and assignment.

Data: Panel for 119 major source countries (balanced; 2000–2015). China signed/implemented FTAs with 19 countries (ASEAN members, Pakistan, Chile, New Zealand, Singapore, Peru, Costa Rica, Iceland, Switzerland, among others); South Korea and Australia are excluded from treatment due to short post-implementation windows within the sample period. Agricultural imports data come from FAO and China Customs Yearbook; macro controls from World Bank WDI. Products: 367 items grouped into 15 categories (beef, pork, sheep, poultry, milk, eggs, maize, rice, wheat, pulses, oilcrops, sugar, fruits, vegetables, roots).

Descriptive patterns:

• Embodied GHG in China’s agricultural imports increased markedly after 2005, reaching 107.70 Mt CO2-eq in 2015 (an increase of 85.69 Mt vs. 2005), with an annual growth rate of 11.16% post-2005; 2000–2005 growth was modest (+0.51 Mt). Total rose from 22.01 Mt CO2-eq (2000) to 107.70 Mt (2015).
• In 2000, 10 countries (US, Argentina, Canada, Australia, Brazil, France, New Zealand, Thailand, Denmark, Philippines) accounted for 92.55% of embodied GHG. Product-wise, oilcrops, wheat, poultry, and pork contributed ~89.38% in 2000. By 2015, beef (7.35 Mt) and pork (4.38 Mt) surpassed oilcrops in embodied GHG shares; oilcrops remained large contributors along with beef and pork (>55% combined).
• Country dynamics: The US remained the largest single source (7.36–31.18 Mt over time). Brazil’s embodied GHG to China rose by 895.75% (2.12 to 21.11 Mt, 2000–2015), becoming second. Australia’s embodied GHG rose by 11.64 Mt.
• FTA partners: Embodied GHG from FTA partners grew from 1.86 Mt (2000) to 16.15 Mt (2015), a 14.30 Mt increase and 14.34% average annual growth; period increases were 72.65% (2000–2005), 30.26% (2005–2010), and 286.54% (2010–2015).

Causal impacts (PSM-progressive DID):

• The FTA × Time coefficient is positive and statistically significant across specifications. With PSM-progressive DID and controls, coefficient ≈ 0.530 (log points), implying China’s FTAs increased embodied GHG from agricultural imports by about 53.0% relative to non-FTA partners.
• Controls: GDP and per-capita arable land of partner countries significantly increase embodied GHG (≈5% significance). Trade openness raises embodied GHG (≈1% significance). Distance (DTE) and mechanization (AML) reduce embodied GHG transfers (≈1% and 10% significance, respectively).

Robustness:

• Balance tests after PSM show %bias <5% and no significant differences across covariates between treated and control groups; propensity score distributions align well.
• Parallel trends: Pre-treatment coefficients are insignificant; post-implementation effects grow with time, indicating persistent, non-lagged impacts.
• Placebo tests (1000 simulations) yield coefficients centered near zero with p-values mostly >0.1, supporting causal identification.
• Alternative matching (kernel, radius caliper) yields similar sign, size, and significance.
• Excluding high-emission partners (US, Brazil, Argentina, Australia) preserves positive, significant treatment effects.

Heterogeneity:

• By income: For low- and middle-income partners, FTA × Time ≈ 0.8509 (≈85.09% increase, significant at 5%); for high-income partners, coefficient is negative and insignificant.
• By product groups: Significant positive effects for grain and oil exporters (coefficients ≈ 1.53–1.63, p<0.01) and for livestock exporters (positive, significant at 10%); effects for fruit and vegetable exporters are insignificant.

Overall trade effect: China’s agricultural imports experienced an average annual growth rate of 12.22% during the study period, reflecting FTAs’ role in deepening economic integration while simultaneously amplifying embodied GHG transfers.

The analysis demonstrates that China’s FTAs function as a double-edged sword: they facilitate agricultural import growth and economic integration but significantly increase the GHG emissions embodied in those imports. The causal evidence indicates a sizable 53% rise in embodied GHG attributable to FTAs, particularly driven by imports from low- and middle-income partners and concentrated in grain/oil and livestock products. These outcomes align with theories of pollution relocation under trade liberalization, where production expands in regions with comparative advantages and, potentially, less stringent environmental regulation. The negative association of distance suggests higher transport costs and quality selection reduce trade volumes and embodied emissions. Increased mechanization in partner countries correlates with lower embodied GHG, likely via efficiency gains and input reductions. These findings highlight a need to reconcile trade policy with climate objectives: standards and incentives within FTAs, targeted toward cleaner production, could mitigate the environmental externalities while preserving the economic benefits of integration. The results support designing FTAs and complementary policies that encourage low-emission supply chains, promote mechanization and modernization, and adjust import composition toward lower-GHG products.

This study compiles a product-level dataset (367 products; 119 countries; 2000–2015) and applies a PSM-progressive DID framework to causally estimate the effect of China’s FTAs on GHG emissions embodied in agricultural imports. It finds that FTAs increased embodied GHG by about 53%, with impacts concentrated in low- and middle-income partner countries and in grain/oil and livestock product flows. Distance and mechanization in partner countries mitigate embodied emissions, while income, arable land, and openness amplify them. Policy implications include: expanding and deepening FTAs particularly with high-income partners; embedding higher import quality standards and differentiated tariff preferences to curb high-GHG imports; diversifying import sources and products; and leveraging FTAs to diffuse advanced, greener agricultural technologies and practices (reduced fertilizer/pesticide use, organic/ecological agriculture). Future research should use updated post-2015 data, refine product-level emission factors, and broaden supply chain boundaries to capture all embodied sources.

• Product aggregation: Over 300 products aggregated into 15 categories due to data constraints; intra-category variability in emission factors introduces uncertainty.
• Temporal scope: Analysis limited to 2000–2015 by data availability; effects of newer FTAs and evolving trade policies are not captured.
• Accounting boundaries: Estimates cover defined production-related sources (CO2, CH4, N2O) but do not encompass the full agricultural supply chain, potentially understating embodied emissions.
• Treatment timing and coverage: Some FTAs (e.g., South Korea, Australia) were excluded from treatment due to short exposure within the window, which may affect generalizability.