How Can Cooperatives Drive Small-Scale Farmers to Achieve a “Carbon Reduction Effect” in the Planting Industry: Evidence from China

Abstract

China is vigorously promoting agricultural energy conservation and carbon reduction and accelerating the transformation of traditional agriculture towards green development, which is a key measure adopted by the Chinese government to advance agricultural modernization. Based on the panel data of 30 provinces in China spanning 2006–2023, this paper systematically studies the impact of agricultural cooperatives driving small-scale farmers on carbon emissions in the planting industry by comprehensively applying linear regression, mediating effect, threshold effect, and spatial econometric models. Studies show that cooperatives have significantly reduced carbon emissions for small-scale farmers, with a stable “carbon reduction effect”, and this effect is most obvious in the eastern region, presenting a regional gradient characteristic of “east > central > west”. The differences between major grain-producing areas and non-major grain-producing areas are relatively small, indicating that their emission reduction effect has wide applicability. Mechanism analysis indicates that improvements in agricultural technology and rural land transfers are key pathways to achieving emissions reductions. Further findings reveal that exemplary cooperatives have a dual threshold effect: they may initially experience a short-term “carbon increase effect”, but as the organization matures, it turns into a significant “emission reduction”. In addition, the development of cooperatives in this region has a positive spillover effect on the carbon emissions of the planting industry in the surrounding areas. This study makes up for the deficiency of the existing literature in the mechanism of “organization-driven individual” promoting agricultural green transformation; it reveals the path of cooperatives promoting low-carbon agriculture through technological promotion and land integration, enriches the theoretical system of agricultural green transformation, and provides replicable practical references for developing countries to promote energy conservation and carbon reduction in agriculture.

1. Introduction

Against the backdrop of increasingly severe global climate change, agricultural carbon emissions have attracted widespread attention. As a fundamental sector of the national economy, reducing agricultural carbon emissions is critical for China to achieve high-quality development and green and low-carbon transformation in agriculture and rural areas [1]. Goal 13 “Climate Action” and Goal 2 “Zero Hunger” in the United Nations Sustainable Development Goals particularly emphasize the need to promote climate resilience and low-carbon agricultural development while ensuring food security. The European Green Deal also explicitly puts forward the “from farm to Table” strategy, which is committed to reducing pesticide use by 50% and chemical fertilizer use by 20% by 2030 and promoting the proportion of organic agriculture to reach 25%, providing an important reference path for global agricultural emission reduction. China has been continuously promoting carbon reduction in agriculture for many years. Data from the “2023 Report on Low-Carbon Development of China’s Agriculture and Rural Areas”, released by the Chinese Academy of Agricultural Sciences, shows that 2023 agricultural production in China emitted 828 million tons of CO₂ equivalent, accounting for 6.7% of national total carbon emissions while contributing 9.5% to GDP—generating nearly one-tenth of GDP with only one-sixteenth of total emissions. Agricultural greenhouse gas emissions, as a significant portion of total emissions, exert a non-negligible impact on climate warming [2]. Against the strategic backdrop of China’s carbon peaking (2030) and carbon neutrality (2060) goals, society has actively explored agricultural carbon sequestration and emission reduction measures. After years of effort, some results have been achieved. According to the “2024 China Rural Low-Carbon Development Report” released by the Chinese Academy of Agricultural Sciences in May 2024, with the in-depth implementation of the rural revitalization strategy and the comprehensive advancement of building a strong agricultural country, China’s low-carbon development in agriculture and rural areas has achieved remarkable results, with both the total amount and intensity of agricultural carbon emissions decreasing.

While achieving some initial results, it is undeniable that current agricultural carbon reduction efforts still face numerous challenges and issues. Due to China’s national and agricultural conditions of “a large country with small-scale farmers”, small-scale farming households form the foundational institutional arrangement for agricultural production and operations in China [3]. Small-scale farmers, as the basic units of agricultural production, typically refer to small-scale agricultural production and management units organized around the family. This type of production entity, due to its small scale and characteristics, such as dispersion and vulnerability [4], engages in agricultural production activities with minimal reliance on employment relationships [5]. Agricultural production methods remain predominantly extensive, and compared to cooperatives, there is a noticeable lag in the adoption of green technologies and efficient resource utilization, which undoubtedly has a significant impact on carbon emissions. Although modern agricultural transformation in some provinces has entered a new developmental phase, a complete decoupling between non-point source pollution and agricultural development has not yet been achieved [6].

In the process of integrating into modern agriculture, small-scale farmers are constrained by insufficient cognition, risk-averse tendencies, cost–benefit considerations, and loose interest connections with cooperatives. They have not yet formed a community of “risk sharing and benefit sharing”, which has restricted their enthusiasm for participating in low-carbon production. Although the Chinese government attaches great importance to cooperatives as a key carrier connecting small-scale farmers with the large market—for instance, the Central “No. 1 Document” in 2013 clearly defined their functional positioning, and in 2019, eleven departments jointly issued a document to strengthen their service functions [7]—problems such as policy implementation, subject transformation, and element guarantee still exist. However, the academic circle also lacks in-depth exploration of the influence of micro-subject behavior and organizational paths. In particular, there is still a significant research gap in the internal mechanism of how cooperatives can drive small-scale farmers to achieve low-carbon transformation through organizational integration, service supply, and technology diffusion. The actual emission reduction efficiency, action path, and contextual heterogeneity of cooperatives still lack systematic empirical support.

Based on this, this study aims to address this gap by adopting a theoretical and empirical research approach. Through constructing a theoretical framework encompassing “direct effects—mediating effects—nonlinear relationships—spatial spillovers”, it reveals the operational mechanisms through which cooperatives mitigate information asymmetry, overcome resource constraints, promote the diffusion of green technologies, and advance the development of appropriately scaled land management. This research not only enriches the theoretical framework for low-carbon agricultural transformation but also provides evidence for formulating targeted policies, thereby supporting China’s agricultural sector in integrating with the global sustainable development agenda.

2. Literature Review

2.1. Research on Carbon Emissions in Agriculture

The academic community has conducted extensive research on agricultural carbon emissions and achieved fruitful results. Through induction, the existing research mainly focuses on two aspects. The first is to discuss the root causes and evolving trends of agricultural carbon emissions, and the second is to study the influencing factors of agricultural carbon emissions.

In terms of the measurement methods of agricultural carbon emissions and their evolving trends, the academic community has formed a relatively systematic theoretical framework. The current mainstream methods for calculating carbon emissions include the emission coefficient method, model simulation method, and field measurement method [8]. The emission factor method (formula factor method) has been widely applied by many scholars due to its simplicity of operation and easy access to data. Model simulation rules conduct dynamic simulations of agricultural ecosystems by means of system dynamics or process models, thereby reflecting the spatio-temporal variation characteristics of agricultural carbon emissions. The field measurement method directly determines the greenhouse gas emission flux through flux observation stations or box methods, which has high accuracy, but it is costly and difficult to promote widely. By applying these methods, scholars have calculated the carbon emissions of China’s agriculture and analyzed the current situation and evolution patterns of agricultural carbon emissions [9]. Scholars point out that as the growth mode of agriculture shifts from extensive to intensive and the transformation process from traditional agriculture to modern agriculture accelerates, the carbon emission intensity of China’s agriculture shows a downward trend [10].

In the research on the influencing factors of agricultural carbon emissions, through comprehensive induction, many scholars have mainly conducted in-depth discussions around the following dimensions.

In the dimension of production methods, enhancing agricultural production efficiency is an important means to reduce carbon emissions. By optimizing the planting density of crops, the land utilization rate be improved, and the carbon emission intensity per unit output can also be effectively reduced [11]. In addition, improving the application of nitrogen fertilizers, reducing the redundancy of agricultural production resources, and enhancing the utilization efficiency of agricultural resources are also key measures to slow down agricultural greenhouse gas emissions [12].

In the dimension of industrial organization, industrial agglomeration, rural industrial integration, and the digital economy are the focuses of academic research on agricultural carbon emissions. Scholars point out that the agglomeration of agricultural industries is conducive to achieving resource sharing and recycling, thereby reducing the overall carbon footprint [13]. Some scholars also emphasize that the integration of rural industries can help reduce agricultural carbon emissions through channels such as the outflow of the rural labor force, large-scale land operation, and agricultural technological progress [1]. The enabling effect of the digital economy on the social and economic development has always been a key focus of the academic community. Regarding the digital economy, some scholars believe that it can optimize resource allocation and enhance the intelligent level of agricultural production, thereby reducing unnecessary environmental burdens [14].

In the dimension of production factors, the academic circle mainly conducts research on land, capital, labor, technology, and data factors. In terms of land elements, some scholars have pointed out that the transfer of agricultural land helps to curb agricultural carbon emissions [15]. In terms of capital elements, digital inclusive finance provides support for the green transformation of agriculture by improving resource allocation and enhancing the risk-resistance capacity of entities [16]. In terms of the labor factor, it is believed that the outflow of the rural labor force can curb agricultural carbon emissions by promoting large-scale agricultural operations [17]. In terms of technical elements, it is believed that although the promotion of agricultural mechanization may increase energy consumption in the short term, it is conducive to improving labor productivity and resource utilization efficiency in the long run [18]. In terms of data elements, scholars have noted that the emergence of Internet of Things (iot) technology has stimulated the multiplier effect of data elements, enabling on-demand fertilization and irrigation in agricultural production, thereby reducing the input of agricultural production factors [19].

In the dimension of policy mechanisms, measures such as environmental regulations and carbon tax systems have been widely discussed. Scholars believe that carbon taxes can not only regulate farmers’ behaviors but also contribute to the green development of agriculture [20]. Environmental regulations for agricultural emission reduction can also exert constraints and guidance on different types of business entities [21]. Some scholars have also studied the positive role of policy-based agricultural insurance in encouraging low-carbon production [22].

2.2. Research on the Aspect of Cooperatives Driving Farmers

The issue of small-scale farmers has long been not only a focus of political attention but has also been extensively studied by many scholars in the academic circle. Through summarization and induction, the existing research results mainly focus on two aspects: increasing agricultural efficiency and raising farmers’ income.

At the level of agricultural efficiency improvement, the coordinated development of new agricultural business entities mainly composed of cooperatives and small-scale farmers is the result of the autonomous choice of the entities under the drive of marketization and policies [23]. With the digital transformation of cooperatives, the exploitation of economic interests in the middle layer of agricultural product circulation has been eliminated [24], the disadvantaged position of small-scale farmers in the agricultural industrial chain has been changed, and the interest connection mechanism between cooperatives and small-scale farmers has been strengthened, enabling efficient connection between small-scale farmers and modern agriculture [25]. This driving effect not only reduces the intermediate consumption in small-scale farmers’ agricultural production, solving the problems of low income and low production efficiency for small-scale farmers, but also promotes the transformation and upgrading of traditional agriculture to green agriculture [26].

At the level of increasing farmers’ income, cooperatives have significantly promoted the growth of small-scale farmers’ income, which in turn helps narrow the income gap between urban and rural residents [27]. The main income-increasing paths are through unified purchase and sale of products, cooperation of production factors, and socialized services, guiding small-scale farmers to participate in cooperative management and enhancing their comprehensive capabilities [28]. However, there are certain limitations for cooperatives in increasing the income of small-scale farmers, such as information asymmetry [29], financial fund constraints, shortage of cooperative management talents, unreasonable personnel structure, and many other problems, like “shell cooperatives” and “zombie cooperatives” [30]. In addition, some scholars have also noted the impact of cooperatives on agricultural carbon emissions, pointing out that new types of agricultural business entities mainly composed of cooperatives have achieved a “carbon reduction effect” on agricultural production through channels such as large-scale land operation, rural human capital, and agricultural science and technology levels [31].

Overall, the academic circles in China have conducted in-depth research on these two topics, providing a useful research basis for this article. However, through summary and synthesis, it was found that there are still several areas worthy of further exploration in existing research. Firstly, from a research perspective, while some scholars have explored the relationship between cooperatives and agricultural carbon emissions, few have delved into the specific role of cooperatives’ driving effect on small-scale farmers. Secondly, in terms of research content, demonstration cooperatives, as high-quality groups within cooperatives, have significant advantages in resource integration, technology promotion, and incentive mechanism design. This selection mechanism may have a stronger positive driving effect on small-scale farmers, thereby demonstrating a more prominent inhibitory effect on agricultural carbon emissions. However, existing research still pays insufficient attention to such impacts. Against this backdrop, this paper aims to adopt quantitative research methods and construct a multivariate econometric regression model to clarify how cooperatives’ efforts to drive small-scale farmers specifically affect agricultural carbon emissions and to unravel the underlying mechanisms. This research not only deepens theoretical understanding of the drivers of agricultural green development, but also provides solid empirical evidence and policy insights for other countries and regions in formulating low-carbon agricultural strategies.

3. Theoretical Analysis and Research Hypotheses

This paper conducts a theoretical analysis on the mechanism by which “cooperatives drive small-scale farmers” to affect carbon emissions in the planting industry, and systematically explores its direct impact, indirect impact, nonlinear characteristics, and spatial spillover effects. As shown in the mechanism path diagram of Figure 1, it reveals the specific realization path of the “carbon reduction effect” of “cooperatives driving small-scale farmers” in their planting industry, providing theoretical support for understanding the low-carbon transformation of agriculture.

3.1. Analysis of the Direct Impact Effect of Cooperatives on Small-Scale Farmers

The cooperative drives small-scale farmers to overcome challenges such as limited operational scale, weak bargaining power, and insufficient risk resilience [32]. As an important organizational form linking agricultural modernization, cooperatives primarily achieve “carbon reduction effects” on agricultural carbon emissions through three dimensions: resource integration, professional management, and information promotion. Firstly, based on transaction cost theory, cooperatives effectively reduce transaction costs for smallholder farmers and enhance the efficiency of agricultural resource allocation by collectively “internalizing” external market activities—which were previously fragmented and costly—into organizational management and coordination [33]. Therefore, in the dimension of resource integration, cooperatives enhance the bargaining power of individual farmers in the large market by uniformly purchasing agricultural supplies, enabling them to obtain stable-quality agricultural supplies at a lower cost [34], thereby guiding farmers to adopt low-carbon fertilizers and energy-saving agricultural machinery, improving the utilization efficiency of agricultural supplies, and curbing carbon emissions in the planting industry. This not only reduces the production costs of small-scale farmers, but more importantly, optimizes the utilization efficiency of chemical fertilizers, pesticides, and agricultural films, thereby exerting a “carbon reduction effect” on the planting industry. Secondly, in the dimension of professional management, cooperatives drive small-scale farmers to achieve coordination and upgrading in agricultural production at the technical, management and operational levels, and promote the transformation of business models from decentralized to intensive and large-scale. For instance, soil testing and formula fertilization, unified prevention and control of pests and diseases, intelligent irrigation, and unified procurement of agricultural materials have effectively reduced repetitive operations and resource waste, thereby helping to lower the carbon emission intensity in the production process of the planting industry. Thirdly, in the dimension of information promotion, cooperatives leverage their organizational advantages to accelerate the spread of low-carbon technologies and management practices, enhancing small-scale farmers’ understanding and willingness to adopt existing green production models or technologies, thereby providing a guarantee for the sustainable development of the planting industry. Based on this, this paper proposes the following hypothesis:

Hypothesis H1.

Cooperatives have driven small-scale farmers to curb carbon emissions from agriculture.

3.2. Analysis of the Mediating Effect of Cooperatives in Driving Small-Scale Farmers

3.2.1. Analysis of the Mediating Effect of Agricultural Science and Technology Level

The cooperative drives small-scale farmers to achieve “carbon reduction effects” in the planting industry by improving agricultural technology. According to the “risk-sharing theory” in economics, by diversifying risks such as technology adoption and market uncertainty among multiple entities, cooperatives have significantly reduced the operational risks faced by individual farmers [35], thereby creating conditions for the promotion of new technologies and high-return agricultural models. Specifically, on the one hand, cooperatives have effectively reduced the marginal cost of agricultural products at the production stage through centralized procurement, unified production standards, and shared services [36]. When individual small-scale farmers independently face technological changes, they often encounter high information search and learning costs. However, cooperatives have driven small-scale farmers to enhance the diffusion efficiency of green technologies through centralized procurement, unified training, and shared services [37]. On the other hand, the leading role of cooperatives enhanced the risk-resistance capacity of small-scale farmers, making them more willing to adopt low-carbon technologies such as precision agriculture and water-saving irrigation—which require higher initial investments but yield significant long-term returns—thereby contributing to the advancement of agricultural science and technology. The improvement of agricultural science and technology has curbed carbon emissions from the planting industry. Firstly, under the theoretical framework of resource allocation efficiency, investment in agricultural science and technology has significantly enhanced the efficiency of factor utilization. For instance, the application of technologies such as precise fertilization and intelligent irrigation has enabled the allocation of resources on demand, reducing the waste of chemical fertilizers, pesticides and water resources [38], thereby lowering the carbon emissions of the planting industry. Secondly, technology-driven forces have promoted the optimization of agricultural production structure, achieving a transformation from extensive to intensive. Through the integration of biotechnology, information technology, and equipment technology, agricultural production has gradually achieved digital and intelligent management [39], which not only increases the land output rate but also reduces the environmental burden. Based on this, this paper proposes the following hypothesis:

Hypothesis H2.

Cooperatives drive small-scale farmers to curb agricultural carbon emissions by enhancing agricultural science and technology levels.

3.2.2. Analysis of the Mediating Effect of Rural Land Transfer

The cooperative drives small-scale farmers to achieve a “carbon reduction effect” on the planting industry by accelerating the transfer of rural land. Firstly, cooperatives integrate the land resources of small-scale farmers to enhance the professionalization and scale of agricultural production, thereby accelerating the process of rural land transfer. According to the theory of factor substitution, under the premise that production factors are substitutable, operators who pursue cost minimization will substitute with production factors that are relatively cheaper. During this process, cooperatives drive small-scale farmers to invest a large amount of agricultural machinery and equipment in their production and operation, creating a labor substitution effect, releasing more rural surplus labor, and prompting some rural surplus labor to enter non-agricultural industries, thereby enhancing the efficiency of labor allocation [40] and providing a realistic basis for the transfer of idle rural land [41]. Secondly, through contractual arrangements and benefit connection mechanisms, cooperatives have enhanced small-scale farmers’ acceptance and willingness to participate in land transfer. In traditional small-scale farming economies, due to risk aversion and the uncertainty of property rights, small-scale farmers tend to maintain a decentralized business operation state. Cooperatives have significantly reduced the transaction uncertainty and opportunistic behavior faced by small-scale farmers by providing stable income expectations and risk-sharing mechanisms and have alleviated the psychological lock-in effect of small-scale farmers’ reliance on land, thereby enhancing the liquidity of the land factor market [42]. Rural land transfer has an inhibitory effect on agricultural carbon emissions. On the one hand, rural land transfers promote the intensification and standardization of agricultural production, improving the efficiency of agricultural resource utilization. Large-scale business entities are more inclined to adopt efficient and energy-saving agricultural technologies and equipment, such as precise fertilization and water-saving irrigation, thereby reducing the carbon emission intensity per unit output [43]. On the other hand, rural land transfers help promote the adjustment of the agricultural industrial structure, optimize the crop layout and rotation system, and reduce the proportion of high-carbon emission crops. Additionally, contiguous land are more conducive to the promotion of ecological agriculture and circular agriculture models. This not only reduces non-point source pollution in agricultural production and enhances the carbon sequestration capacity of agricultural systems but also enables the realization of “green premiums”. Based on this, this paper proposes the following hypothesis:

Hypothesis H3.

Cooperatives drive small-scale farmers to curb agricultural carbon emissions by accelerating rural land transfers.

3.3. Analysis of the Threshold Effect on the Development Level of Demonstration Cooperatives

According to the notice issued by the Ministry of Agriculture and Rural Affairs of China in 2017 on the “Interim Measures for the Evaluation and Monitoring of National Model Farmers’ Professional Cooperatives”, specific requirements have been established for the economic strength, service effectiveness, and social reputation of cooperatives participating in the selection process for demonstration cooperatives. The aspect of social reputation emphasizes the leading role. This indicates that the economic strength of a cooperative is not the only criterion; it should also have a strong exemplary and leading role. Therefore, under the influence of policy incentives and market elimination mechanisms, some cooperatives often adopt a capital-intensive development path in the early stages to enhance scale efficiency and market competitiveness, which refers to investing in more means of production and labor on the same land, carrying out intensive farming, and increasing the total amount of products by increasing the output per unit area [44]. As a result, cooperatives continuously increase the input intensity of agricultural materials such as agricultural machinery and equipment, fertilizers, and pesticides [45]. This kind of intensive management mode with blindly expanding material input as the core can improve the per unit yield and labor productivity, but it easily leads to excessive resource consumption and environmental pollution, which may increase the carbon emissions of the planting industry in the short term, and ultimately affect the sustainable development [46]. At the same time, during this stage, cooperatives focused more on maximizing their own economic benefits, which weakened the leading role of small-scale farmers, caused the level of benefit linkage to remain low, and failed to effectively achieve the goal of green development. With the improvement of the selection method of demonstration cooperatives by the state, the selection standards and requirements pay more attention to social benefits, and the social responsibility of small farmers driven by cooperatives is gradually strengthened. At this stage, the cooperative explored various paths of coordinated development, such as “benefit linkage + green transformation”; cooperatives continuously established interest connection with small farmers, guided small farmers to adjust agricultural operation mode, and reduced resource consumption, thus producing an obvious “carbon emission reduction effect”. The longer this incentive and guidance approach lasts, the stronger the “carbon reduction effect” of the cooperative driving small-scale farmers in the planting industry will be. Based on this, this paper proposes the following hypothesis:

Hypothesis H4.

The development of demonstration cooperatives has strengthened the “carbon reduction effect” of cooperatives driving small-scale farmers to reduce carbon emissions from the planting industry.

3.4. Analysis of the Spatial Spillover Effect of Carbon Emissions in the Planting Industry

As a core component of agricultural production, the carbon emissions of crop cultivation are not only the direct result of local agricultural production activities, but also have significant spatial spillover effects on adjacent areas through various geographical and economic mechanisms. Firstly, from the perspective of geographical proximity, carbon emissions from the planting industry have a significant spatial autocorrelation. Adjacent regions share similarities in terms of climatic conditions, soil types, and farming systems, resulting in regional agglomeration characteristics of carbon emission behaviors. Additionally, farmland management measures (such as fertilization methods and irrigation intensity) have spread rapidly among regions, further strengthening the spatial linkage effect of carbon emissions. Secondly, factor flow is an important mechanism driving the spatial spillover of carbon emissions [47]. The interregional flow of production factors such as labor, capital, and agricultural machinery and equipment will change the input structure and technology path between regions, thereby affecting the carbon emission pattern of the planting industry. For instance, high-carbon emission technologies or practices may shift across regions due to disparities in environmental regulation intensity, potentially flowing from strictly regulated areas to those with laxer regulations. This creates a “pollution refuge” effect, triggering cross-regional redistribution of carbon emissions and generating negative spatial spillover effects. Secondly, technology diffusion also has a significant impact on the spatial spillover of carbon emissions [48]. The dissemination of advanced low-carbon agricultural technologies (such as precision agriculture and conservation tillage) among regions can effectively reduce the carbon emission intensity of local planting industries. At the same time, through the demonstration effect, it can drive the imitation and adoption in surrounding areas, creating positive technology spillover. Conversely, if high-carbon technologies are promoted, negative spillover effects may occur. Based on this, this paper puts forward the following statement:

Hypothesis H5.

There is a spatial correlation between carbon emissions from agriculture in this region and surrounding regions, and carbon emissions from agriculture in this region will lead to an increase in surrounding regions.

4. Materials and Methods

In the model design for exploring the impact of “cooperatives driving small-scale farmers” on carbon emissions in the planting industry, it is necessary to take into account not only the robustness and endogeneity of the model itself, but also the diversity of model settings to meet the needs of different path analyses. This paper mainly considers models and methods such as ordinary least squares, ordinary least squares, generalized moment estimation, the instrumental variable method, and spatial autoregression. Among them, the ordinary least squares method (OLS) is used as the benchmark model to preliminarily estimate the linear relationship between variables, but it is prone to endogenous biases caused by omitted variables and reverse causality, underestimating or overestimating the actual emission reduction effect of cooperatives. To alleviate the endogeneity problem of the model, it is necessary to adopt the instrumental variable method (IV) for two-stage least squares regression (2SLS). By introducing exogenous instrumental variables, the endogeneity problem of omitted variables can be alleviated. However, this method requires that the instrumental variables meet the requirements of exclusivity and correlation, but it has relatively high requirements for the validity of the instrumental variables. Generalized moment estimation (GMM) can handle the endogeneity problem caused by reverse causality. This method is suitable for model setting verification and efficiency improvement in long panel scenarios, but it is sensitive to sample size and moment condition setting. In addition, carbon emissions from the planting industry have a significant spatial spillover effect. Spatial autoregressive models (SAR) or spatial Dubin models (SDM) can capture the spatial dependence brought about by the diffusion of cooperatives and technology spillovers in neighboring areas, thereby improving the estimation accuracy. However, the spatial model has a strong assumption dependence on the setting of the weight matrix. If the spatial structure is set incorrectly, it may lead to inference bias. The specific model is shown as follows:

4.1. Model Design

4.1.1. Benchmark Regression Model

Based on theoretical analysis and hypothesis H1, in order to identify the direct relationship between “cooperatives driving small farmers” and carbon emission intensity of the planting industry, this paper builds the ordinary least square method to test the benchmark relationship and builds the panel model for the empirical test, as shown below:

$$y_{i,t}={\alpha }_0+{\alpha }_1{X}_{i,t}+{\alpha }_2{C}_{i,t}+{\mu }_i+{\tau }_t+{\epsilon }_{i,t}$$

(1)

Among them, y_i,t represents the carbon emission intensity of the planting industry; X_i,t drives small-scale farmers for the cooperative; C_i,t represents all control variables; μ_i represents the individual fixed effect; τ_t represents the time-fixed effect; ε_i,t is the random disturbance term; α₀ is a constant term; α₁ and α₂ are the model estimation coefficients.

4.1.2. Mediating Effect Model

Based on theoretical analysis and Hypotheses H2 and H3, drawing on the research approach of existing scholars [49], we used stepwise regression to test the mediating effect, as shown below:

$$M_{i,t}={\beta }_0+{\beta }_1{X}_{i,t}+{\beta }_2{C}_{i,t}+{\mu }_i+{\tau }_t+{\epsilon }_{i,t}$$

(2)

$$y_{i,t}={\gamma }_0+{\gamma }_1{X}_{i,t}+{\gamma }_2{M}_{i,t}+{\gamma }_3{C}_{i,t}+{\mu }_i+{\tau }_t+{\epsilon }_{i,t}$$

(3)

Among them, M_i,t is the mediating variable, representing the level of agricultural science and technology and the transfer of rural land, respectively. The other variables are explained in the same way as in Equation (1).

4.1.3. Threshold Effect Model

Based on theoretical analysis and hypothesis H4, the threshold effect of the development level of demonstration cooperatives is tested, and the panel threshold model is constructed by referring to the existing literature [50], as shown below:

$$y_{i,t}={\kappa }_0+{\kappa }_1{C}_{i,t}+{\eta }_1{X}_{i,t}{I(N}_{i,t}\le {\nu }_1)+{\eta }_2{X}_{i,t}{I({\nu }_1<N}_{i,t}\le {\nu }_2)+{\eta }_3{X}_{i,t}{I(N}_{i,t}>{\nu }_2)+{\mu }_i+{\tau }_t+{\epsilon }_{i,t}$$

(4)

Among them, N_i,t is a threshold variable, representing the development level of the demonstration cooperatives; ν₁ and ν₂ represent the first and second threshold values; I(·) is the characteristic function. The other variables are explained in the same way as in Equation (1).

4.1.4. Spatial Lag Model

Based on theoretical analysis and hypothesis H5, in order to further capture the spatial influence relationship of economic activities among regions, this paper discusses the spatial influence of the “cooperative driving small-scale farmers” in this region and carbon emissions of the planting industry in surrounding areas. Since the spatial autoregressive model (SAR) is mainly used to analyze how a spatial unit is affected not only by explanatory variables in its own region but also by neighboring regions, it can meet the needs of testing spatial spillover effects. Therefore, this paper draws on existing research to construct a spatial lag model [51], as shown below:

$$y_{i,t}={\lambda }_0+{\rho W}_{i,t}{y}_{i,t}+{\lambda }_1{X}_{i,t}+{\lambda }_2{C}_{i,t}+{\mu }_i+{\tau }_t+{\epsilon }_{i,t}$$

(5)

Among them, ρ represents the autoregressive coefficient of the explained variable, indicating the relationship between the explained variable and the spatial spillover influence of the surrounding areas. There are three situations for ρ: greater than 0, less than 0, and equal to 0. A value greater than 0 indicates a positive space overflow, a value less than 0 indicates negative spatial overflow, and a value equal to 0 indicates no spatial overflow effect. The other variables are explained in the same way as in Equation (1). W is the spatial weight matrix. This paper mainly considers the adjacency (0, 1) weight matrix.

4.2. Variable Settings

4.2.1. Explained Variable

The explained variable in this paper is the carbon emission intensity of the planting industry (pl). Due to the significant differences in the scale of agricultural planting in different regions, there may be disputes regarding the measurement based on the carbon emissions of the planting industry [52]. Therefore, the ratio of the carbon emissions of the planting industry to the total output value of the planting industry is used for measurement. Among them, the total output value of the planting industry adopts the price reduction with the base period of 2000. At present, China’s agricultural carbon emissions primarily originate from four sources: agricultural land use, farmland soil, rice fields, and livestock breeding. Since this article focuses on the planting industry, livestock farming carbon emissions are not considered. Furthermore, based on the literature review, the formula factor method is currently the most commonly used approach in the academic circle; so this paper will continue to employ this method. The carbon emission coefficient can be found in the research results of Li et al., (2011) [53].

The formula for calculating carbon emissions in the planting industry is as follows:

$$C_{\mathit{it}}=\sum C_{\mathit{nit}}=\sum T_{\mathit{nit}}\cdot {\sigma }_n$$

(6)

Among them, C_it represents the total carbon emissions from the planting industry in the t year of the province; C_nit represents the total carbon emissions of the n type of carbon source; T_nit represents the amount of each carbon emission source; σ_n represents the carbon emission coefficients of each carbon emission source.

The formula for calculating carbon emission intensity in the planting industry is as follows:

$${CO}_{2it}={\displaystyle \frac{C_{it}}{{GCP}_{it}}}$$

(7)

Among them, i represents the province, and t represents the time; C_it represents the total carbon emissions from the planting industry; CO_2it represents the carbon emission intensity of the planting industry; GCP_it represents the total output value of the planting industry at a constant price.

4.2.2. Explanatory Variable

The explanatory variable in this paper is the cooperative driving small-scale farmers (cooper). Drawing on existing literature, the ratio of the cooperative driving small-scale farmers to the total number of farmers is used as the proxy variable [54]. The main reason is that in reality, there are significant obstacles to obtaining the economic benefit data generated by cooperatives driving small-scale farmers. There is a positive correlation between this proportion and the driving effect of cooperatives. That is, the higher the proportion, the greater the influence of cooperatives on small-scale farmers in terms of agricultural production organization and resource allocation, indicating a stronger driving effect.

4.2.3. Mediating Variable

The mediating variables of this article are agricultural science and technology levels (sci-tec) and rural land transfer (transf). Among them, agricultural technology and technology level is proxied by the natural logarithm of the number of patent applications granted. The main consideration is that the number of applications for authorization in science and technology directly reflects the actual achievements of agricultural technological innovation in a region or period. Compared with the prerequisite indicators such as R&D investment or the number of scientific researchers, it can better demonstrate the effectiveness and maturity of technological output. More importantly, the authorized quantity has been reviewed and confirmed by relevant national departments, possessing authority and comparability. This can eliminate the interference of low-quality applications that have not passed the review, ensuring the authenticity and reliability of the data. Rural land transfer is proxied by the ratio of total land transfer area to total farmland area.

4.2.4. Threshold Variable

The threshold variable of this paper is the development level of demonstration cooperatives (demon). Based on the availability of data, this paper uses the absolute number of demonstration cooperatives as the natural logarithm to represent the development level of demonstration cooperatives.

4.2.5. Control Variables

To minimize the endogeneity effects caused by omitted variables on the model, this paper mainly controls the influencing factors of the planting structure (stru), the degree of agricultural disaster (disa), the level of agricultural openness (openness), the degree of rural electricity consumption (consum), and the per capita added value of agriculture (per). Among them, the planting industry structure is characterized by the ratio of the sown area of grain to the sown area of crops. The planting structure directly affects the efficiency of agricultural resource allocation and the output pattern. The proportion of the sown area of grain reflects the primary and secondary directions of agricultural production and has an important impact on the performance of the agricultural economy. Therefore, controlling this variable helps to eliminate the interference caused by the differences in planting structure. The degree of agricultural disaster is characterized by taking the logarithm of the natural number of the affected area of crops affected by disasters. This is a key exogenous factor affecting the stability of agricultural production. Measuring with the logarithm of the affected area of crops can effectively reflect the impact of natural disasters on agricultural output and avoid missing endogenous problems caused by major environmental disturbances. The level of agricultural openness is characterized by the ratio of the total volume of agricultural product import and export trade to GDP. This variable reflects the influence of the external market on agricultural development. The ratio of the total volume of agricultural product import and export trade to GDP can quantify the degree of agricultural internationalization, which is in line with the current trend of agriculture integrating into the global value chain. The degree of rural electricity consumption is represented by taking the natural logarithm of rural residents’ electricity consumption. As an important indicator of the development level of rural infrastructure, rural electricity consumption reflects the degree of modernization of production conditions. The per capita added value of agriculture is characterized by the ratio of agricultural added value to the total population. This metric directly reflects agricultural production efficiency. By using this ratio, the population scale effect is effectively isolated, enabling a more accurate representation of actual agricultural output levels.

4.3. Data Sources

This paper uses data from 30 provincial-level administrative regions in China from 2006 to 2023 as the observation sample. Due to the absence of data from Xizang and the Hong Kong, Macao, and Taiwan regions, these areas are not included in the analysis. Since some data has not been updated to 2024, the panel data can only be retained up to 2023. For provinces with missing data, linear interpolation methods were used to fill in the gaps, and the panel data was subjected to 1% tail trimming to reduce the impact of outliers. The research data primarily comes from the China Agricultural Yearbook, China Statistical Yearbook, China Population and Employment Statistical Yearbook, China Rural Statistical Yearbook, China Rural Operations and Management Statistical Annual Report, and China Rural Cooperative Economy Statistical Annual Report, as well as the National Bureau of Statistics and local statistical yearbooks. The descriptive statistics of the variables are shown in

Table 1. Descriptive statistics of variables.

Variable Name	N	Mean	Std	Min	Max	VIF
pl	540	0.3080	0.1440	0.0827	0.7370	2.2300
cooper	540	0.0843	0.0513	0.0077	0.2460	1.3000
sci-tec	540	4.2890	1.3020	1.0990	7.0710	2.9100
transf	540	15.5900	1.3370	11.8600	17.9900	3.4900
demon	540	2.2410	2.1910	0.0479	11.5900	1.3400
stru	540	0.6530	0.1490	0.3660	1.0240	1.5500
disa	540	6.1100	1.5470	1.2810	8.3320	2.6100
openness	540	0.0153	0.0141	0.0007	0.0535	2.2300
consum	540	4.7590	1.3350	1.4040	7.5430	2.6300
per	540	1.6430	1.1340	0.2340	6.1500	1.9900

.

5. Analysis of Empirical Results

5.1. Analysis of Benchmark Test Results

Before the model regression, through the collinearity test, the mean of the variance inflation factor (VIF) was 2.2300 < 10, indicating that the model does not have a serious collinearity problem.

Table 2. Benchmark test results.

Variable Name	Carbon Emission Intensity of the Planting Industry
	OLS	System GMM		IV
cooper	−0.2349 *** (0.0703)	−0.2427 *** (0.0709)	−0.1904 * (0.0759)	−0.4750 *** (0.1269)
pl(T−1)			0.5249 *** (0.1205)
stru		−0.1449 * (0.0664)	0.5420 ** (0.1723)	−0.1426 * (0.0640)
disa		0.0014 (0.0029)	−0.0404 ** (0.0141)	0.0009 (0.0028)
openness		−0.0343 (0.5353)	−8.9107 *** (2.7078)	0.3904 (0.4855)
consum		−0.0107 (0.0099)	0.0301 ** (0.0106)	−0.0193 * (0.0092)
per		−0.0041 (0.0048)	−0.0192 (0.0118)	−0.0072 (0.0049)
individual fixed effects	YES	YES	YES	YES
time fixed effects	YES	YES	YES	YES
intercept term	0.4470 *** (0.0186)	0.5733 *** (0.0730)	−0.0091 (0.0397)	0.3931 *** (0.0742)
Obs	540	540	510	510

Note: *, ** and ***, respectively, indicate significance at the 10%, 5% and 1% levels. According to STATA 16 calculations, the numbers in parentheses represent robust standard errors. The same is the case below.

presents the benchmark test results, reporting the ordinary least squares (OLS) estimation, the system GMM estimation, and the instrumental variable (IV) regression results, respectively, with all models controlling for dual fixed effects. Firstly, without controlling for variables, the estimated coefficient of carbon emissions from the planting industry driven by cooperatives driving small-scale farmers is −0.2349, which is significantly negative at the 1% level. The results initially indicate that cooperatives driving small-scale farmers can help curb carbon emissions from the planting industry. After incorporating control variables, the estimated coefficient is −0.2427, which is significantly negative at the 1% level, and the degree of influence further increased. The results indicate that adding control variables to the model effectively controls for some confounding factors, thereby revealing the impact effect of cooperatives driving small-scale farmers.

Secondly, to enhance the robustness of the estimation results and avoid the endogeneity issues caused by the reverse causality of variables and the omission of variables in the model, the model requires endogeneity testing. First, it is necessary to address the reverse causality issue existing in the core variables. That is, through technology promotion, resource integration, and large-scale operation, cooperatives can help enhance production efficiency and optimize the allocation of input factors, thereby curbing the excessive growth of carbon emissions. Conversely, the changes in carbon emission levels may also reciprocally affect the cooperatives and their ability to drive small-scale farmers. For instance, high carbon emissions are often accompanied by intense agricultural inputs, such as excessive use of chemical fertilizers, pesticides, and irrigation, which may lead to stricter environmental regulations, enhanced policy constraints, and raised market access thresholds, thereby weakening the operational efficiency of cooperatives and their ability to absorb small-scale farmers. Therefore, when analyzing the relationship between the two, endogenous treatment methods must be used to identify causal directions and avoid estimation biases caused by reverse causality. Second, it is necessary to address the endogeneity problem caused by omitted variables. Although the model has controlled some interfering factors and released the influence effect of cooperatives driving small-scale farmers, it cannot comprehensively control those unobtrusive factors that are both related to cooperatives driving small-scale farmers and can affect the integration of urban and rural areas. This means the model must overcome endogeneity issues caused by omitted variables. Based on this, this paper employs system GMM estimation to address reverse causality issues and instrumental variable regression to address omitted variable issues.

By adopting the systematic GMM estimation, AR(1) is 0.0400 and AR(2) is 0.3930, indicating that the model exhibits first-order autocorrelation. The Sargan statistic is 0.5940 and the Hansen statistic is 1.0000, passing the over-identification test. After GMM regression, it is found that the estimated coefficient of the cooperative driving small-scale farmers is 0.1904, which is significantly negative at the 10% level. This indicates that after the model alleviates the reverse causality problem, it can still confirm that the cooperative driving small-scale farmers has an inhibitory effect on carbon emissions from the planting industry. Furthermore, drawing on the existing literature, this study uses the lag of one period of the explanatory variable as the instrumental variable (pl T−1) to minimize the endogeneity influence of the model as much as possible [55], and it also selects the interaction term between terrain undulation and the cooperatives driving small-scale farmers as a potential instrumental variable. This approach is mainly to meet the dual requirements of exogeneity and the correlation of instrumental variables. Terrain undulation, as a natural geographical feature, has strong exogeneity and is not easily affected by the reverse causality of economic activities. Additionally, it is related to the agricultural production and management patterns, potentially influencing the agricultural production scale, organizational forms, and resource acquisition capabilities, thereby moderating the cooperative driving small-scale farmers. Through two-stage regression, it was found that the estimated coefficient is 0.4750, which is significantly negative at the 1% level. This sufficiently demonstrates that the cooperative driving small-scale farmers has a significant “carbon reduction effect” on the planting industry, thereby validating hypothesis H1.

5.2. Analysis of Robustness Test Results

To enhance the reliability of the model estimation results, the robustness test mainly considers the substitution of variables, the addition of interaction control between individuals and time, and the lag period for robustness testing.

Firstly, in the method of replacing variables, on the one hand it is necessary to consider that the carbon emission sources of the planting industry should not only include the carbon emissions caused by the input of agricultural production materials (fertilizers, pesticides, and agricultural films), but also fully take into account the carbon emissions caused by rice cultivation. On the other hand, explanatory variables must be replaced to explore whether the absolute quantity of small-scale farmers driven by cooperatives can also produce a “carbon reduction effect”. Secondly, the model adds the interaction effect between individuals and time because the individual effect reflects the inherent differences among different cross-sectional units that do not change over time, while the time effect depicts the changes in the time dimension that all individuals experience together. The introduction of the interaction term between the two aims to more precisely identify the impact of differentiated exogenous shocks or policy interventions on individuals at different time points and to enhance the model’s fitting ability for complex real-world situations. Thirdly, by lagging the explanatory variable and the explained variable by one period, respectively, the main purpose is to explore whether this influence effect still exists after the explanatory variables lag by a certain number of periods. Finally, control variables are supplemented. In the model, the influences of rural human capital and agricultural policy support factors are controlled. Among them, the level of rural human capital is represented by the average years of education in rural areas, while agricultural policy support is represented by the ratio of expenditure on agriculture, forestry, animal husbandry, and fishery and water conservancy affairs to the total amount of general fiscal expenditure.

Table 3. Robustness test results.

Variable Name	Increase Rice Cultivation	Replace Variable	High-Dimensional Fixation	Lag Period		Add Control Variables
	pl	pl	pl	pl	L. pl	pl
cooper	−5.2867 ** (2.0365)		−0.2877 * (0.1358)		−0.2060 ** (0.0702)	−0.2475 *** (0.0715)
cooper (1)		−0.0164 ** (0.0055)
L. cooper				−0.2314 *** (0.0669)
control variables	YES	YES	YES	YES	YES	YES
individual fixed effects	YES	YES	YES	YES	YES	YES
time fixed effects	YES	YES	YES	YES	YES	YES
individual # time	NO	NO	YES	NO	NO	NO
intercept term	5.9527 *** (1.6456)	0.6104 *** (0.0766)	0.6464 (0.3885)	0.5167 *** (0.0736)	0.5739 *** (0.0728)	0.6339 *** (0.1148)
Obs	540	540	540	510	510	540

Note: *, ** and ***, respectively, indicate significance at the 10%, 5% and 1% levels. L. represents a lag period of 1, and cooperative driving small-scale farmers (1) is a replacement explanatory variable. # represents the intersection of individual and time.

presents the results of the robustness test and reports on various robustness test methods to explore the “carbon reduction effect” of cooperatives driving small-scale farmers in the planting industry. From the results of the robustness test, first, by replacing the explained variable method, the estimated coefficient of the cooperative driving small-scale farmers is 5.2867, which is significantly negative at the 5% level. After replacing the explained variable, the estimated coefficient of the cooperative driving small-scale farmers is 0.0164, which is significantly negative at the 1% level. Second, by adopting a high-dimensional fixed method, after the model controlled the interaction term between individuals and time, the estimated coefficient of the cooperative driving small-scale farmers is 0.2877, which is significantly negative at the 10% level. Third, by using the lag period method, after the cooperative driving small-scale farmers lags for one period, the estimated coefficient is 0.2314, which is significantly negative at the 1% level. However, after the explained variable lags for one period, the estimated coefficient of the explanatory variable is 0.2060, which is significant at the 5% level. It indicates that there is indeed a certain lag in the carbon emissions of the planting industry when cooperatives drive small-scale farmers. Based on the above methods, it can be fully proved that the cooperative driving small-scale farmers has a significant “carbon reduction effect” on the carbon emissions of the planting industry, meaning that the original conclusion has strong reliability. Finally, after controlling the level of rural human capital and agricultural policy support, the estimated coefficient of cooperatives driving small-scale farmers was 0.2475, which was significantly negative at the 1% level, indicating that the “carbon reduction effect” of cooperatives driving small-scale farmers has strong robustness.

5.3. Analysis of Heterogeneity Test Results

The results of the heterogeneity test mainly discussed the differences in the impact of cooperatives driving small-scale farmers on carbon emissions from the planting industry in the three major regions and grain functional areas. Mainly considering the uneven distribution characteristics of these regions in terms of economic development level, resource endowment, policy support intensity, and planting structure, which have varying degrees of impact on cooperatives driving small-scale farmers and subsequently have different degrees of effect on the “carbon reduction effect” of the planting industry, it is necessary to conduct an in-depth analysis.

Table 4. Heterogeneity test results.

Variable Name	Carbon Emission Intensity of the Planting Industry
	Eastern Region	Central Region	Western Region	Major Grain-Producing Areas	Non-Grain Production Areas
cooper	−0.2209 *** (0.0511)	−0.2819 * (0.1264)	−0.0829 (0.1101)	−0.2699 ** (0.1005)	−0.2769 ** (0.1050)
control variables	YES	YES	YES	YES	YES
individual fixed effects	YES	YES	YES	YES	YES
time fixed effects	YES	YES	YES	YES	YES
intercept term	0.5768 *** (0.0446)	1.6565 *** (0.2001)	0.0140 (0.1938)	1.2644 *** (0.1452)	0.3763 *** (0.0927)
Goodness of Fit	0.9649	0.9572	0.9328	0.9435	0.9329
Obs	198	144	198	234	306

Note: *, ** and ***, respectively, indicate significance at the 10%, 5% and 1% levels.

presents the results of the heterogeneity test. Firstly, analyzing the three major economic regions, the impact degrees of cooperatives driving small-scale farmers on the carbon emissions of the planting industry in the three regions are 0.2209, 0.2819, and 0.0829, respectively. The impact is significant at the 1% level in the eastern region, significant at the 10% level in the central region, and not significant in the western region. The degree of effect shows a gradient distribution pattern of “east > central > west”. The reason for this difference might be that the eastern region has well-developed agricultural infrastructure, a sound science and technology promotion system, and strong cooperative driving capabilities, which can effectively guide small-scale farmers to adopt low-carbon production technologies, thereby significantly reducing carbon emissions. The central region, though in a catch-up phase, has seen increasing policy support, and many provinces within it are traditional agricultural powerhouses with good agricultural resource endowments. The radiation and driving effect of cooperatives on small-scale farmers are beginning to emerge. In contrast, the western region is constrained by harsh natural conditions, low agricultural production efficiency, and a low degree of marketization. The development of cooperatives remains underdeveloped, making it difficult to form an effective driving force for low-carbon transformation, resulting in a negative and insignificant impact.

Secondly, from the perspective of grain functional areas, although the estimated carbon emission coefficients of the planting industry in major grain-producing areas and non-grain-producing areas are both significantly negative at the 5% level, the estimated coefficients of the two are relatively similar and the differences are small, reflecting that there are certain commonalities in the agricultural emission reduction mechanisms of the two types of regions. Nevertheless, subtle differences in coefficients can still reveal the impact of different policy orientations, resource endowments, and degrees of organization among regions on carbon emission behaviors. On the one hand, major grain-producing areas, as they undertake the main task of grain supply, have higher levels of agricultural intensification and more developed technology promotion systems. Cooperatives are more efficient in promoting the application of green production technologies, thereby generating more significant carbon emission reduction effects. On the other hand, the agricultural production structure in non-grain production areas is more diversified. Cooperatives can also effectively reduce carbon emissions per unit output by optimizing resource allocation and enhancing factor utilization efficiency. Therefore, although there are differences in functional positioning between the two types of regions, the cooperative organizational form has played a positive role in enhancing the sustainable development capacity of agriculture.

5.4. Analysis of Mechanism Path Verification Results

Based on theoretical analysis and hypothesis H2, this paper further examines the mediating effect between agricultural science and technology level and rural land transfer by using the stepwise regression method.

Table 5. Mechanism path verification results.

Variable Name	(1)	(2)	(3)	(4)
	sci-tec	pl	transf	pl
cooper	2.4150 *** (0.6947)	−0.2039 ** (0.0693)	1.3444 *** (0.3798)	−0.2074 ** (0.0717)
sci-tec		−0.0161 ** (0.0052)
transf				−0.0263 *** (0.0078)
control variables	YES	YES	YES	YES
individual fixed effects	YES	YES	YES	YES
time fixed effects	YES	YES	YES	YES
intercept term	2.3686 ** (0.8056)	0.6114 *** (0.0713)	13.2429 *** (0.7510)	0.9219 *** (0.1306)
Goodness of Fit	0.9158	0.9290	0.9558	0.9299
Obs	540	540	540	540

Note: ** and ***, respectively, indicate significance at the 5% and 1% levels.

shows the results of the mechanism path verification. From the results, it can be seen that the prerequisite for conducting the mediating effect test is that the cooperative driving small-scale farmers must have a significant impact on the carbon emissions of the planting industry, that is, the total effect must be valid. This conclusion has been confirmed in the previous text and will not be elaborated on here. The focus is on discussing the verification of the conduction path. Columns (1) and (3) report, respectively, the impact of cooperatives driving small-scale farmers on the level of agricultural science and technology and the transfer of rural land. Through regression, the estimated coefficients are 2.4150 and 1.3444, respectively, both significantly positive at the 1% level, indicating that cooperatives driving small-scale farmers are conducive to improving the level of agricultural science and technology and accelerating the transfer of rural land. Next, the mediating variables are added to the model to observe the estimated coefficients of the mediator variables and explanatory variables. The agricultural science and technology level and the rural land transfer show, respectively, the regression of the carbon emissions of the planting industry. The estimation coefficients are 0.0161 and 0.0263, respectively, which are significantly negative at the 5% and 1% levels, respectively. The results show that the improvement of the agricultural science and technology level and the degree of rural land transfer are conducive to curbing carbon emissions from the planting industry. At this point, the estimated coefficients of the cooperative driving small-scale farmers are 0.2039 and 0.2074, respectively, both significantly negative at the 5% level. The “carbon reduction effect” of the cooperative driving small-scale farmers has not changed, which indicates that the cooperative driving small-scale farmers has an inhibitory impact on the carbon emissions of the planting industry through two paths: improving the level of agricultural science and technology and accelerating the transfer of rural land. Thus, hypotheses H2 and H3 are supported.

5.5. Analysis of the Threshold Effect Test Results

The previous section discussed the impact effect and mechanism path of cooperatives driving small-scale farmers on carbon emissions in the planting industry. Based on theoretical analysis and hypothesis H3, to further verify the threshold effect of the development level of demonstration cooperatives, this study adopted the bootstrap method for repeated sampling 200 times.

Table 6. The result of the threshold value test.

Variable Name	Threshold Inspection	Threshold Value	F Value	p Value	Critical Value
					10%	5%	1%
demon	the first threshold	0.8058	83.58	0.0000	29.8013	34.8860	40.6194
	the second threshold	5.5118	25.14	0.0750	23.3678	27.6723	33.0669
	the third threshold	10.3702	19.26	0.3400	30.0693	38.0956	59.0664

shows the results of the threshold value test. From the results, the first threshold value is 0.8058, which is significantly positive at the 5% level; the second threshold value is 5.5118, which is significant at the 10% level, while the third threshold value fails the test, indicating that the development level of the demonstration cooperatives has a dual threshold effect.

After the above tests, it was found that the development level of the demonstration cooperatives has two thresholds. By establishing a panel threshold model for regression, we further discuss the evolution trend of the threshold effect.

Table 7. The test results of the threshold effect.

Variable Name	Cooper	Std	95% Confidence Interval
demon < 0.8058	0.5184 ***	(0.1212)	0.2802	0.7565
0.8058 < demon ≤ 5.5118	−0.2806 ***	(0.0667)	−0.4117	−0.1495
demon > 5.5118	−0.7925 ***	(0.1132)	−1.0148	−0.5702
control variables	YES	YES	YES	YES
intercept term	0.9285 ***	(0.0814)	−3.3943	−1.7199
Goodness of Fit	0.8216
Obs	540

Note: ***, indicate significance at the 1% levels.

presents the test results of the threshold effect; when the development level of the demonstration cooperative is lower than 0.8058, the estimated coefficient of the explanatory variable for the carbon emissions of the planting industry is 0.5184, which is significantly positive at the 1% level. The results indicate that under the influence of the demonstration cooperative, this period will lead to an increase in agricultural carbon emissions. When the development level of the demonstration cooperative is between 0.8058 and 5.5118, the estimated coefficient of the explanatory variable is 0.2806, which is significantly negative at the 1% level, with the coefficient sign changing, indicating the emergence of a “carbon reduction effect”. When the development level of the demonstration cooperative continues to rise and crosses the threshold of 5.5118, the estimated coefficient of the explanatory variable is 0.7925, and the “carbon reduction effect” continues to strengthen, with the overall impact showing a phased upward trend. The reason for this is that in the early stage of the green transformation of agriculture, cooperatives often introduced industrialized production methods such as mechanization, increased input of chemical fertilizers and pesticides, and expansion of facilities to enhance production efficiency and organization. Although this increased output in the short term, it exacerbated carbon emissions, creating a “transformation pain”. In the initial stage, it even led to an increase in environmental costs. Meanwhile, after joining cooperatives, small-scale farmers, influenced by technological path dependence and incentive mechanisms, tend to expand high-investment models in pursuit of returns, further intensifying the pressure on carbon emissions. Nevertheless, as the diversified development model of demonstration cooperatives has taken shape, emphasizing the position of ecological benefits in agricultural production, agricultural production has shifted from traditional resource-dependent to efficiency-driven production, thereby revealing the “carbon emission reduction effect”. Furthermore, as the development level of demonstration cooperatives continues to improve, the demand for green technologies expands, and through methods such as technology diffusion, model replication, and organizational promotion, the ability to guide and support small-scale farmers is enhanced, ultimately promoting the evolution of the planting industry towards low carbonization and sustainability, thereby validating hypothesis H4.

6. Extended Analysis

6.1. Analysis of Global Autocorrelation Test Results

Before conducting the test for spatial spillover effects, it is necessary to identify whether the carbon emissions from the planting industry have spatial correlation. The conventional practice in academia is to conduct global autocorrelation tests and use the Moran index as the criterion for assessing spatial correlation.

Table 8. Global autocorrelation test results.

Year	Carbon Emission Intensity of the Planting Industry
	Moran Value	Z Value	p Value
2006	0.1780	1.7620	0.0390
2007	0.1630	1.6120	0.0530
2008	0.0910	1.0350	0.1500
2009	0.1450	1.4710	0.0710
2010	0.1150	1.2490	0.1060
2011	0.1000	1.1690	0.1210
2012	0.1170	1.3130	0.0950
2013	0.0960	1.1390	0.1270
2014	0.1030	1.1850	0.1180
2015	0.1380	1.4720	0.0710
2016	0.1640	1.6700	0.0470
2017	0.1620	1.6210	0.0520
2018	0.1510	1.5340	0.0630
2019	0.1750	1.7250	0.0420
2020	0.1990	1.9180	0.0280
2021	0.2430	2.2870	0.0110
2022	0.2240	2.1500	0.0160
2023	0.1580	1.5920	0.0560

presents the results of the global autocorrelation test and reports the global spatial autocorrelation test of carbon emissions in the planting industry. The results indicate that the Moran index of carbon emissions from the planting industry was positive during the period from 2006 to 2023, and it shows certain spatial agglomeration characteristics, indicating that an increase in carbon emissions from the planting industry in a certain area often leads to a synchronous rise in carbon emissions in surrounding areas.

6.2. Analysis of the Test Results of Spatial Spillover Effects

Table 9. Test results of spatial spillover effect.

Variable Name	Carbon Emission Intensity of the Planting Industry
	SAR	OLS
cooper	−0.2229 *** (0.0542)	−0.2427 *** (0.0709)
pl	−0.1611 ** (0.0497)	−0.1449 * (0.0664)
disa	0.0010 (0.0029)	0.0014 (0.0029)
openness	−0.1467 (0.5361)	−0.0343 (0.5353)
consum	−0.0086 (0.0088)	−0.0107 (0.0099)
per	−0.0040 (0.0038)	−0.0041 (0.0048)
L. pl	0.1668 ** (0.0574)
Diff in individual effect characteristics	0.0015 *** (0.0001)
Obs	540	540

Note: *, ** and ***, respectively, indicate significance at the 10%, 5% and 1% levels. L. represents a lag period of 1.

presents the test results of the spatial spillover effect, reporting the use of the lag model to test the spatial spillover effect of cooperatives driving small-scale farmers on carbon emissions from the planting industry. Firstly, the result coefficient of the lag term of the explained variable is 0.1668, which is significantly positive at the 5% level. This indicates that there is a significant positive spatial spillover effect between the increase in carbon emissions from the planting industry in this region and the surrounding areas. The main reasons lie in the geographical proximity of agricultural production and the diffusivity of environmental impacts. On the one hand, agricultural production activities have a high degree of spatial agglomeration characteristics. Carbon emission behaviors such as fertilization, irrigation, and mechanized operations have cross-regional impacts through atmospheric transport, hydrological cycles, and soil migration among regions. On the other hand, factors such as the promotion of agricultural technology, policy intervention, and the imitation of farmers’ behaviors have led to a convergence of carbon emission patterns in space, creating a “demonstration effect” and a “learning effect”. Secondly, after introducing spatial factors, the estimated coefficient of cooperatives driving small-scale farmers is 0.2229, which is significantly negative at the 1% level. This indicates that even considering spatial factors, cooperatives driving small-scale farmers can still produce a “carbon reduction effect”. When comparing with the ordinary least squares estimation results, it is found that the estimation coefficient of the spatial autoregressive model is slightly lower than that of the ordinary least squares estimation. This is mainly because when the explained variable has a significant spatial spillover effect, ignoring the spatial effect will cause the ordinary least squares estimation result to be higher, thereby generating a “local amplification effect”. In contrast, the spatial autoregressive model effectively identifies and controls estimation biases caused by spatial dependence compared to ordinary least squares estimation. By correcting for spatial spillover effects, the estimated coefficients are reduced. Based on the above tests, hypothesis H5 is confirmed to be valid.

7. Research Findings, Recommendations, and Limitations

7.1. Research Findings

Against the backdrop of global efforts to combat climate change and advance sustainable development, achieving the United Nations Sustainable Development Goals, promoting sustainable resource management and efficient utilization, fostering green production and consumption patterns (SDG 12), reducing environmental impacts and actively addressing the effects of climate change (SDG 13), and enhancing resilience and adaptive capacity to climate disasters have become shared concerns of the international community. As a fundamental industry, the green and low-carbon transformation of agriculture not only concerns the realization of a country’s emission reduction targets, but also holds profound significance for the construction of a global sustainable food system. Based on the panel data of 30 provincial regions in China from 2006 to 2023, this article explores the impact of cooperatives driving small-scale farmers on carbon emissions in the planting industry. Research has found that “cooperatives driving small-scale farmers” has a significant “carbon reduction effect”, which is mainly achieved through two paths: enhancing agricultural science and technology levels and accelerating land transfer. Meanwhile, demonstration cooperatives exhibit dual threshold characteristics: in the early stage of development, they may lead to a temporary increase in carbon emissions, but as the standardization and scale of the cooperatives improve, they gradually show sustained emission reduction effects. Furthermore, this study identifies a significant spatial spillover effect in planting industry carbon emissions, meaning that cooperative development in one region also drives carbon reduction in surrounding areas. This research achievement not only reveals the characteristics of the green transformation path of China’s agriculture, but also provides a reference that is both universal and special for other developing countries. Compared with India, its cooperatives have a long history and are more mature in certain fields (such as dairy and sugar production), but the overall coverage is limited, and the green transformation process is relatively slow. Brazil promotes the development of cooperatives through family farming support programs, but it is facing carbon emission pressure in the expansion of tropical agriculture. Vietnam has attempted to enhance its organizational level through the model of “company + cooperative + farmers”, but it is still constrained by the application of technology and the land transfer mechanism. In contrast, China has achieved the systematic role of cooperatives in carbon reduction by relying on policy guidance and the construction of exemplary cooperatives. Particularly in the eastern regions, a scalable technological and management innovation system has been formed, demonstrating stronger institutional integration capabilities and development coordination. This provides empirical evidence and practical paths for the implementation of SDG 12 and SDG 13.

7.2. Research Recommendations

First, strengthen the fiscal support mechanism. In addition to setting up a special green fund, a “central-local” coordinated investment mechanism should be established to ensure that funds are inclined towards ecologically fragile areas and economically underdeveloped regions. At the same time, a project performance evaluation system should be established, linking the distribution of subsidies with the application effects of technologies, to promote the implementation and effectiveness of technologies such as energy-saving agricultural machinery, organic fertilizer substitution, and comprehensive utilization of straw. Encourage financial institutions to develop “Green Agricultural Benefit Loan” products, providing low-interest loans to business entities implementing low-carbon renovations, and form a pattern of coordinated support from both fiscal and financial sectors.

Second, deepen the capacity building for carbon emission reduction. On the basis of regular publicity and education, a technical service team composed of agricultural technology experts, environmental protection organizations, and local experts is formed to carry out household guidance and field demonstrations. Build an online training course library through digital platforms, promote new communication methods such as short videos and live streaming, and enhance small-scale farmers’ awareness of climate change and their willingness to adopt technologies. Promote the establishment of internal carbon management teams in cooperatives to achieve autonomous and standardized low-carbon production.

Third, optimize the tax and market incentive mechanisms. Expand the coverage of tax preferential policies, not limited to income tax and value-added tax reductions and exemptions, and explore the implementation of consumption tax refunds for green inputs. Support enterprise cooperatives in participating in the national carbon emission trading market and encourage them to develop agricultural carbon sink projects and obtain additional income. Establish a green production points system, where points can be used to exchange for agricultural supplies or for priority access to project support.

7.3. Research Limitations

Although this paper has made certain research progress in exploring the impact of cooperatives driving small-scale farmers on carbon emissions in the planting industry, there are still several limitations and areas for further expansion.

Firstly, although the provincial panel data adopted in this paper covers a relatively long period of time, it is difficult to capture the heterogeneity of the behaviors of micro-subjects. Future research can combine the micro-survey data at the farmer level to further identify the differentiated response mechanisms of carbon emission behaviors of different types of small-scale farmers during their participation in cooperatives.

Secondly, although the mechanism analysis has verified the mediating role of agricultural science and technology level and rural land transfer, it has not fully explored other potential paths, such as the efficiency of production factor allocation and the improvement of farmers’ environmental awareness. Subsequent research can construct a more systematic mediating effect model to deepen it.

Thirdly, although the panel threshold model reveals the nonlinear characteristics of the demonstration cooperatives’ development, there are still certain challenges in handling the endogeneity of threshold variables. In the future, the robustness of the estimation can be enhanced by using the instrumental variable method or the dynamic panel method. Furthermore, spatial econometric analysis only reveals the spatial spillover effect of carbon emissions, but the specific transmission mechanisms of spatial dependence, such as technology diffusion, policy imitation, or factor flow, still need to be further deconstructed.

Finally, this paper focuses on carbon emissions from the planting industry and does not take the carbon sink function of agriculture into comprehensive consideration. Future research can construct a comprehensive evaluation system for the green and low-carbon development of agriculture from the dual perspectives of “carbon source—carbon sink”.

Overall, as the green transformation of agriculture progresses in depth, the ecological functions of cooperatives in achieving the organic connection between small-scale farmers and modern agriculture deserve continuous attention. The optimization of their emission reduction paths and the policy coordination mechanism still need to be further explored.