Regional differences of agricultural total factor carbon efficiency in China

The paper addresses how China can improve agricultural total factor carbon efficiency (ATFCE) to support a transition to low-carbon agriculture while maintaining economic growth and food security. Although industry is the main source of national emissions, agriculture contributes about 17% of China’s greenhouse gas emissions, especially methane and nitrous oxide. Given China’s commitments to peak carbon by 2030 and achieve carbon neutrality by 2060, agriculture has substantial mitigation potential. However, farmers in developing contexts face trade-offs: mandatory emission reductions may constrain production and income. ATFCE integrates economic output and carbon reduction under a unified framework, aligning efficiency with environmental performance. The study asks: What is the level, temporal trend, and spatial distribution of ATFCE in China? Are there significant regional differences across agricultural functional zones (AFZs) defined by grain production-consumption balance—grain-producing (GPZ), grain-balance (GBZ), and grain-selling (GSZ)? By quantifying ATFCE and decomposing disparities, the work informs targeted agricultural and environmental policies.

Prior research has extensively measured environmental and carbon efficiency in China for sectors such as industry, power, and transportation, but agriculture has received comparatively less focus. Existing agricultural studies report generally low carbon efficiency, with regional heterogeneity and occasional high performance in ecological protection areas. A key challenge for comparability is the diverse scope and precision in calculating agricultural carbon emissions (ACEs), ranging from inputs-only approaches to inclusion of livestock and other sources. Methodologically, stochastic frontier analysis (SFA) requires assumed functional forms and parameterization, making results sensitive to specification, whereas data envelopment analysis (DEA) is nonparametric but sensitive to outliers; modified DEA procedures can mitigate this. Gaps identified include limited analysis of ATFCE by AFZs (as opposed to geographic regions like east/central/west), and inconsistent ACE source coverage and precision. This study contributes by applying a biennial weight modified Russell model (BWMRM) and constructing a more comprehensive and heterogeneous ACE inventory (including often-neglected sources like straw burning, and differentiating rice varieties and provincial conditions).

Study scope and data: Panel of 31 Chinese provinces, 1999–2018. Data from China Rural Statistical Yearbook, China Agricultural Yearbook, and China Animal Husbandry and Veterinary Yearbook. Provinces are grouped into three AFZs: GPZ, GBZ, and GSZ. Inputs and outputs: Inputs include labor (number of agricultural practitioners), land (sown area), machinery (agricultural electricity consumption), fertilizer (chemical fertilizer applied), pesticide (chemical pesticide consumption), agricultural film (use amount), water (actual irrigation area), and farm animals (number of draft animals). Desirable output is actual agricultural production; undesirable output is agricultural carbon emissions (ACEs). Agricultural carbon emission sources (ACES): Five categories—(1) agricultural materials (fertilizer, pesticides, plastic sheeting, diesel oil, irrigation energy), (2) rice cultivation (early, late, in-season rice), (3) soil surface emissions for crops (paddy rice, winter/spring wheat, soybean, corn, vegetables), (4) livestock and poultry (cow, buffalo, cattle, mule, camel, donkey, horse, pig, sheep, goat, rabbit, poultry), and (5) straw burning (rice, wheat, corn, rape, soybean, cotton straw). Emissions are computed via source-specific conversion and emission coefficients following Huang et al. (2019). For livestock and poultry, quantities are adjusted for breeding cycles (e.g., pigs 200 days, rabbits 105 days, poultry 55 days); other adjustments use year-end inventories across t and t-1. Efficiency model: The biennial weight modified Russell model (BWMRM) combines biennial environmental production technology (BEPT) with a weighted Russell directional distance model (WRDDM). BEPT constructs the production frontier using observations from two adjacent periods, avoiding reconstruction as the sample window changes and reducing infeasibility. WRDDM allows decomposition of inefficiency contributions across inputs and outputs with direction vectors and weights. To enhance robustness, two sets of weights are applied in the objective function and averaged: external weights at the category level—inputs, desirable outputs, undesirable outputs—set to (1/3, 1/3, 1/3), and when combining desirable and undesirable outputs into a single “output” category, external weights set to (1/2, 1/2). Internal weights are reciprocals of the number of factors in each category: (1/M, 1/Q, 1/H) and (1/M, 1/(Q+H), 1/(Q+H)). ATFCE is computed from the ratio of optimal to actual undesirable outputs (ACEs), using the geometric mean of biennial measures analogous to a two-period window analysis. Distributional and disparity analyses: Kernel density estimation characterizes the temporal-spatial distribution of ATFCE (omitting GSZ’s plot due to extreme concentration). Dagum Gini decomposition partitions total inequality into within-region (DWS), between-region (DBS), and transvariation (IT) components, and computes inter-regional Gini ratios for AFZ pairs (GPZ–GBZ, GPZ–GSZ, GBZ–GSZ).

• National level: Mean ATFCE in China during 1999–2018 was 0.761. It declined by about 25%, from 0.825 (1999) to 0.6983 (2018). Kernel density shows bimodality and growing dispersion nationally. - AFZ ranking: GSZ had the highest ATFCE (0.9865), followed by GBZ (0.7201) and GPZ (0.6666). GSZ remained around 1 with no clear trend; GPZ and GBZ showed larger fluctuations and overall downward trends (inverted U-shapes over different subperiods). - Provincial extremes: Highest ATFCE provinces: Tibet (0.9997), Hainan (0.9981), Shanghai (0.997), Beijing (0.9937), Jiangsu (0.9924). Lowest: Hubei (0.4743), Yunnan (0.4645), Hunan (0.441), Anhui (0.4295), Heilongjiang (0.4130), Jiangxi (0.3354). Some provinces (e.g., Sichuan, Hebei) trended up; others (e.g., Yunnan, Hubei, Anhui, Heilongjiang, Jilin, Inner Mongolia) declined. - Inequality dynamics within AFZs and nationally: Gini within China rose from 0.1399 (1999) to 0.2204 (2018) (+57%). Gini within GPZ increased from 0.1648 to 0.2736 (>60%). GBZ’s within inequality rose overall but declined after 2007. GSZ showed little change. - Inter-AFZ differences: All three interregional Gini ratios increased. Largest gap was GPZ–GBZ (0.164 in 1999 to 0.2492 in 2018), followed by GPZ–GSZ (0.1532 to 0.2412). GBZ–GSZ difference doubled (0.1077 to 0.2246) with high volatility. - Decomposition of total difference: Difference between subregions (DBS) was the largest contributor (mean share 43.66%), followed by difference within subregions (DWS, 30.04%) and intensity of transvariation (IT, 25.94%). Around 2007, DBS spiked while IT dropped, with minimal change in DWS. - Contextual explanations: Despite expectations that GPZ would perform best, its ATFCE was lowest, attributed to lower per capita incomes and difficulty attracting skilled labor, terrain limiting mechanization, pressure to meet grain demand leading to input-intensive practices, more traditional techniques, and potentially stricter environmental management in GBZ/GSZ. The 2007 financial crisis induced labor flows back to agriculture, increasing input use and emissions in GPZ/GBZ, widening disparities.

The study integrates economic production and environmental performance to quantify ATFCE across China and AFZs, revealing high overall efficiency yet notable decline and widening disparities. Findings counter expectations that grain-producing provinces would be most carbon-efficient; instead, structural, geographic, and policy factors depress GPZ’s ATFCE relative to GBZ/GSZ. Urbanized, high-income regions (e.g., Beijing, Shanghai, Jiangsu) show very high ATFCE, consistent with stronger environmental regulation, technological adoption, and favorable production conditions. The Dagum decomposition clarifies that inter-subregion differences dominate total inequality, especially around exogenous shocks (2007), when labor shifts back to agriculture amplified emissions-intensive practices in GPZ/GBZ. These results underscore the need for zone-specific strategies: improving technology adoption and input structure in GPZ; sustaining modern, efficient practices in GBZ/GSZ; and addressing intra-zone imbalances to reduce national disparities while recognizing that some inter-zone inequality reflects structural differences in agricultural endowments.

Using the biennial weight modified Russell model with a comprehensive ACE inventory, the study measures China’s ATFCE (1999–2018) and uncovers substantial regional heterogeneity by AFZ. Main conclusions: (1) National ATFCE averaged 0.761 and declined by about a quarter over the period; GSZ maintained near-unit efficiency, while GPZ and GBZ fell and fluctuated. (2) Provincial leaders in ATFCE were Tibet, Hainan, Shanghai, Beijing, and Jiangsu; laggards included Hubei, Yunnan, Hunan, Anhui, Heilongjiang, and Jiangxi. (3) Inequalities widened nationally and across AFZs, with the largest gap between GPZ and GBZ. (4) Inter-subregion differences (DBS) were the largest source of total inequality, followed by within-subregion differences and transvariation. Policy implications: - Promote advanced agricultural technologies and management, strengthen information services, and enhance resource-use efficiency and human capital accumulation. - Tailor AFZ-specific strategies: GBZ/GSZ should continue optimizing input allocations and improving development quality; GPZ requires strengthened macro-level regulation, appropriate production inputs, improved infrastructure and services, optimized agricultural structures, and mitigation of livestock-related pollution in fragile ecosystems. - Address disparities within and among AFZs; allow some inter-AFZ inequality reflecting structural endowments. Anticipated benefits include higher productivity and incomes, ecological protection, balanced regional development, enhanced international competitiveness, and export growth; potential unintended effects include overreliance on technology, environmental pressures from scaling up, rural labor shifts, and transition risks for some farmers.

The article does not explicitly list study limitations. Data sharing is restricted due to an ongoing project, and the kernel density plot for GSZ was omitted due to extreme concentration.