The Carbon Emissions Reduction Effect of Green Agricultural Subsidy Policy: A Quasi-Natural Experiment

Abstract

There is significant interest among policymakers and academics about whether green agricultural subsidy policy (GASP) has achieved its intended effect of reducing agricultural carbon emissions (ACEs) in China. Based on the panel data of 172 prefecture-level cities from 2010 to 2021, this study evaluates the impact and mechanisms of GASP on ACEs’ reduction effect by employing the DID model. The results demonstrate that GASP does significantly reduce carbon emissions. Mechanism tests illustrate that GASP promotes carbon reduction via two mechanisms: enhancing agricultural technology progress efficiency and increasing scale management efficiency. Further heterogeneity analyses reveal that the carbon reduction effects of GASP are particularly prominent in the main grain-producing regions and in cities with higher levels of carbon emissions. The empirical tests and mechanism analyses in this paper can better clarify the function of GASP, enrich and deepen the existing literature, and provide some useful references for carbon reduction.

1. Introduction

Global climate change, induced by greenhouse gas emissions such as CO2, has significantly impacted human survival and the ecosystem (Heffron and McCauley, 2022) [1]. According to the Global Daily Carbon Emissions Report 2023, global carbon emissions followed a “V” trend from 2019 to 2022, reaching 36.07 billion tons in 2022. In China, the rapid development of agriculture has become an important contributor to the acceleration of global warming. Due to the extensive use of fertilizers and pesticides, as well as the mechanization of farming, China’s agricultural carbon emissions (ACEs) have increased significantly in recent years. The agriculture sector has become the second-largest source of carbon emissions, accounting for approximately 15% of the overall amount. China’s ACEs are in a severe condition, mainly because of the following reasons. First, the environmental awareness of farmers is weak. Farmers rely too much on chemical inputs to increase their yields, instead of adopting sustainable production practices (Xie and Huang, 2021) [2]. Second, the low adoption rate of advanced technologies has caused an overuse of chemical elements (Llewellyn and Brown, 2020) [3]. Third, the fragmentation of cultivated land makes the use of chemical elements inefficient (Wang et al., 2023) [4]. Fourth, the traditional agricultural subsidy system, which is unrelated to environmental preservation, has, to some extent, encouraged farmers to overuse chemical materials such as fertilizers. These problems in China not only lead to serious carbon emission problems, but also deteriorate the environment of cultivated land and hinder sustainable development. Therefore, it has become an urgent challenge for China to develop appropriate ACE reduction policies and reform the traditional agricultural subsidy system based on its own special situations.

In 2015, China began piloting green agricultural subsidy policy (GASP) in five provinces: Anhui, Zhejiang, Hunan, Shandong, and Sichuan. Under the direction of green sustainable agriculture, China has replaced traditional agricultural subsidies with the GASP system. GASP strives to protect cultivated land fertility and promote large-scale management. From the perspective of sustainable development, these objectives may also be seen as promoting the development of environmentally friendly and low-carbon agriculture. Therefore, the purpose of this paper is to analyze whether the implementation of GASP has achieved its environmental protection and sustainable development objectives. Upon summarizing the relevant literature on GASP and ACEs, the majority of studies, however, concentrate on how GASP affects chemical inputs and grain cultivation areas (Fan et al., 2023) [5]. No research has yet explored the relationship between GASP and ACEs’ reduction. Thus, the purpose of this paper is to explore the following questions: (1) Can GASP mitigate ACEs? (2) If GASP does promote ACEs’ reduction, what are the specific mechanisms of this? (3) In which areas does the policy’s abatement effect work better?

Accordingly, this paper uses the IPCC method to measure the ACEs in 172 cities from 2010 to 2021. Considering the implementation of GASP as a quasi-natural experiment, we use the difference-in-difference (DID) method to empirically investigate its impact on reducing ACEs. In order to investigate its mechanisms, this paper utilizes the DEA model to construct indices to measure agricultural technology progress efficiency (ATPE) and scale management efficiency (SME). The test results find that the implementation of GASP led to decreases in the total amount and intensity of ACEs by 20.3% and 21.3%, respectively. The mechanism analyses exhibit that GASP mainly achieves ACEs’ reduction by promoting agricultural technology progress and encouraging moderate-scale management. The heterogeneity tests show that the policy effect is more sensitive considering the main grain-producing areas and in cities with high levels of ACEs.

The possible innovations of this paper are as follows. First, in terms of the research perspective, unlike previous studies focusing on ACEs’ reduction from the provincial perspective, this paper starts from the prefecture level. In this study, the total amount and intensity of the ACEs in each city are measured using carbon source data collected from 172 cities in China. This helps to capture policy effects more clearly. Second, in terms of the mechanism, this paper scientifically reveals the ACEs’ reduction mechanism of GASP. We assume that ACEs will also follow the environmental Kuznets curve theory, that is, ACEs might be influenced by technological progress and large-scale management efficiency. This paper utilizes the DEA model to measure the total factor productivity (TFP) using agricultural input and output data. Indices representing ATPE and SME are deconstructed from the TFP to scientifically construct mechanism variables. The findings of this study provide a theoretical basis for further reducing carbon emissions and achieving sustainable development.

The remainder of this paper is organized as follows. The second part provides an overview of related studies on domestic and international ACEs and ACEs’ reduction. In the third part, the research hypothesis of this paper is put forward after discussing the policy background and theoretical analysis. The data sources and analysis, combined with research design used in this article, are discussed in the fourth part. The fifth part presents and discusses the findings of the empirical investigation. The sixth part provides a summary of the findings and policy recommendations.

2. Literature Review

2.1. Sources and Measurement of ACEs

The sources of ACEs in previous studies are relatively stable. In a broad sense, ACEs come from agricultural inputs, soil tillage, production waste, and livestock waste (Johnson, 2007) [6]. Considering that GASP mainly contributes to farmers’ cultivation activities, this paper mainly focuses on cultivation carbon sources. On the one hand, chemical inputs, especially pesticides, fertilizers, plastic films, and agricultural diesel, are primary cultivation carbon sources (Dyer et al., 2010) [7]. On the other hand, changes in land use during cultivation, such as irrigation, ploughing, and so on, can also lead to ACEs (Elham, 2020) [8]. Therefore, five carbon sources are selected in this paper: fertilizers, pesticides, plastic film, agricultural diesel, and irrigation.

Existing ACE accounting methods mainly include the carbon footprint measurement method, IPCC method, lifecycle assessment (LCA) method, and so on (Peter et al., 2016; Yang et al., 2020; Liu et al., 2024) [9,10,11]. The IPCC method is a basic method, which is highly operational and flexible. The emissions of various agricultural activities can be calculated by multiplying the number of carbon sources by the associated coefficient (Maji et al., 2019) [12]. Researchers from several nations have calculated the ACEs of different nations using this IPCC methodology. Garnier et al. (2019) [13] calculated and estimated the ACEs of France from 1852 to 2014 and concluded that the ACEs of France had more than doubled from 1950 to 1980.

2.2. Influencing Factors of ACEs

In addition to the carbon source perspective, scholars have discussed factors that boost or lower ACEs from the following different perspectives. This will provide useful references for the subsequent control variables selection.

First, government expenditure on environmental protection can reduce ACEs. A government’s financial support is the most effective way to realize the green development of agriculture (Dumortier and Elobeid, 2021) [14]. Second, the industrial structure can affect ACEs. Increased industrialization levels will increase the demand for agricultural products and add to the burden of agricultural production. Industrialization also reduces the cost of agricultural inputs such as fertilizers and pesticides, thus increasing ACEs (Wu et al., 2021; Idowu et al., 2023) [15,16]. Third, the planting structure affects ACEs. Cash crops tend to require more high-carbon inputs than grain crops (Zhang et al., 2020) [17]. Fourth, increased urbanization reduces ACEs. Urbanization can promote the spatial concentration of development factors, contributing to scale effects and the upgrading of the energy structure, both of which are conducive to reducing carbon emissions (Zhang et al., 2023) [18]. Fifth, the size of rural populations increases ACEs. Considering rural population size, the larger the rural population size in a region, the more production and consumption activities related to environmental pollution, causing more ACEs (Han et al., 2021) [19].

2.3. ACE Reduction Strategies

In the exploration of emission reduction pathways, technological progress and policy support are two common approaches. First, technological progress can improve the efficiency of chemical inputs and reduce agricultural waste. Improving the utilization efficiency of nitrogen fertilizer and reducing its absolute application amount are helpful for ACEs’ reduction (Li et al., 2021) [20]. Promoting slow-release and long-acting fertilizer and returning straw to the field also play vital roles in ACEs’ reduction (Chojnacka, 2019) [21]. Second, policy guidance can increase farmers’ incentives to participate in conservation farming. For example, the EU’s CAP includes direct greenhouse gas reduction subsidies. Such carbon reduction subsidies, by promoting farmers’ adoption of environmentally friendly production practices (Gocht, 2017) [22], significantly reduced ACEs by 21% (Himics, 2020) [23]. Ethiopia promotes various climate-smart agricultural practices to enhance crop production and increase farmers’ income while reducing ACEs (Hailemariam, 2019) [24]. In addition to storing carbon and lowering ACEs, land conservation initiatives in the US help farmers by lowering soil erosion, storing soil water, and enhancing soil quality (Wongpiyabovorn, 2020) [25].

In conclusion, the existing literature has developed a comprehensive system of ACEs. There are still some points that need to be enriched and deepened. First, ACEs are mostly measured at the provincial level, and it is difficult to capture their characteristics at a relatively microscopic level. Second, there is a lack of empirical tests on the ACEs’ reduction effect and mechanism of GASP, which aims to develop green and low-carbon agriculture. Therefore, based on data from prefecture-level cities, this paper examines the ACEs’ reduction effect of GASP from a more microscopic perspective at the prefectural level. This paper also explores the policy effects and mechanisms of GASP through empirical tests to assess its policy effectiveness.

3. Policy Context and Conceptual Framework

3.1. Background of GASP

In order to guarantee food security and increase farmers’ income, China has developed a comprehensive subsidy system, which mainly subsidizes for improved seeds, farmers, and agricultural inputs. Most of these subsidies are not linked to environmental protection, creating negative externalities on the environment. Incentivized by traditional subsidies, farmers have over-applied fertilizers and pesticides, exacerbating ACEs (Xing and Yang, 2013) [26]. In order to achieve the “Carbon peaked and Carbon neutral” target, China reformed the traditional subsidy system in 2015. GASP was initially rolled out in five pilot provinces in 2015, namely Anhui, Shandong, Hunan, Sichuan, and Zhejiang, and subsequently implemented across the whole country in 2016.

GASP’s goals have been adjusted to support the protection of cultivated land fertility and the large-scale management of farmers. This cultivated land fertility protection subsidy closely links subsidy funds to the quality of farmland, encouraging farmers to return straw to the fields and strengthening awareness of the protection of the agro-ecological environment. The large-scale management subsidy reflects the principle that “those who produce more grain will receive priority support”. Through policy guidance, subsidies accelerate the growth of large grain farmers, family farms, and farmers’ cooperatives. Large-scale operations allow for the effective use of a variety of production variables, which lowers the number of excess ACEs. The design concept makes it evident that GASP’s characteristics are low-carbon and environmentally friendly. Therefore, there is an urgent need to assess the carbon reduction effects of GASP.

3.2. Theoretical Analysis

3.2.1. GASP and ACEs’ Reduction

GASP may reduce ACEs from the following three perspectives. First, GASP reduces ACEs by imposing behavioral constraints on farmers. The cultivated land productivity protection subsidy of GASP gives subsidies only to farmers whose farmland meets fertility standards. This will constrain their production methods, which are highly dependent on chemical inputs (Liang et al., 2016) [27]. To obtain subsidies, they prefer to use organic farming, mixed cropping, integrated pest management, reduced input systems, and so on to protect their productivity (Koiry and Huang, 2023) [28]. Second, GASP reduces ACEs by providing cost compensation to farmers. GASP allocates 20% of subsidy funds to encouraging farmers to adopt green, low-carbon practices. Farmers who receive subsidies have more funds available to adopt conservation farming patterns, such as replacing chemical fertilizers with organic fertilizers and adopting agro-film recycling projects (He et al., 2022) [29]. Third, GASP reduces ACEs by improving the environmental awareness of farmers. Agricultural subsidy policies are often accompanied by a series of new informational and educational activities to help farmers understand the importance of farmland protection (Mamum et al., 2021) [30]. Subsidized farmers will consciously reduce their use of agro-chemicals and protect their cultivated land (Schaub, 2023) [31]. Increased awareness among farmers about the importance of protecting cultivated land effectively reduces ACEs (Patrick et al., 2021) [32]. Accordingly, Hypothesis 1 is put forward:

Hypothesis 1.

GASP creates a remarkable incentive in a positive way for ACEs’ reduction.

3.2.2. GASP, ATPE, and ACEs’ Reduction

GASP can encourage farmers to adopt new technology. Agricultural technological progress in this paper is referred to in a broad sense. It can be divided into two aspects. chemical technology: (germplasm innovation, fertilizer, pesticide, and machinery, etc.) and learning technology (resource allocation, management, and professional knowledge) (Li et al., 2021) [33].

First, GASP can greatly increase the probability of farmers adopting green agricultural technologies by releasing financial constraints (Bonfiglio, 2020) [34]. Due to the positive externalities and high adoption costs of low-carbon technologies, farmers are seldom actively exposed to innovative technologies (Rychel et al., 2020) [35]. Subsidized funding can reduce the costs of farmers adopting new technologies, thereby reducing the risk of the adoption of new technologies (Mottaleb et al., 2016) [36]. Second, by incorporating new technology such as integrated pest management systems within the subsidy program, GASP increases farmers’ understanding of new technology (Biggs and Justice, 2017) [37]. Third, financial subsidies can greatly motivate farmers to invest in human capital, such as by engaging in more technical training. To some extent, technical training promotes the rational allocation of farmers’ resources (Liu, 2021; Liu et al., 2024) [38,39].

Increasing ATPE is helpful for ACEs’ reduction. First, innovations in chemical inputs help to reduce ACEs. The use of compound fertilizers, slow-release fertilizers, and organic fertilizers can not only reduce ACEs, but also enhance soil fertility and improve yields (Benbi, 2018) [40]. Bio-pesticides and slow-release pesticides that can be recycled can reduce ACEs significantly (Archana, 2022) [41]. Second, innovations in low-carbon services can help farmers to optimize the efficiency of chemical inputs and then reduce their total amount. Integrated fertility management in a timely manner can promote fertilizer reduction, thereby reducing ACEs (Hatirarami, 2018) [42]. Total variable rate nutrient application and pesticide application technologies for precision agriculture can achieve the precise application of chemical elements, thereby reducing the amounts used (Balafoutis, 2017) [43]. Third, the upgrading of agricultural machinery technology can reduce diesel use. Drip, sprinkler, and other similar irrigation technologies not only avoid excessive water and fertilizer consumption, but also reduce ACEs (Islam, 2022) [44]. The use of advanced energy-efficient agricultural machinery also reinforces the positive effect of straw return on ACEs’ reduction (He, 2021) [45]. Accordingly, Hypothesis 2 is put forward:

Hypothesis 2.

GASP realizes ACEs’ reduction by promoting agricultural technological progress efficiency.

3.2.3. GASP, SME, and ACEs’ Reduction

GASP can encourage farmers to expand their planting scale. First, the distribution of GASP is linked to the grain-planting area, prompting farmers to expand their land scale to receive higher subsidies (Zhang et al., 2018) [46]. Second, GASP also restricts farmers from abandoning their farmland or converting it to non-agricultural land (Zhao et al., 2021) [47]. Third, subsidies for large-scale operators, such as family farms and professional farmers’ cooperatives, contribute to the efficient use of existing cultivated land (Duan et al., 2021) [48].

Increasing SME is helpful for ACEs’ reduction. First, moderately expanding the scale of production can make the allocation of land, capital, labor and chemical inputs more reasonable, optimizing the efficiency of land resources and chemical elements (Ju, 2016) [49]. Second, advanced knowledge and the agricultural training of large-scale farmers help to reduce ACEs. Large-scale farmers tend to possess more advanced management skills, which directs them to upgrade their planting modes and optimize their production structures in order to achieve sustainable development (Hu et al., 2022; Guo et al., 2022) [50,51]. Third, the emergence of large-scale farmers helps to achieve agricultural agglomeration. Agricultural agglomeration areas can share advanced agricultural machinery, thereby generating technology spillover effects and reducing ACEs (Liu et al., 2022; Li et al., 2023) [52,53]. The development of rural cooperative production services makes it possible to import advanced agricultural technologies to farmers (Wu et al., 2023) [54]. Larger areas of cultivated land, participation in training, and improved agricultural expertise are associated with farmers’ decisions to reduce the use of agricultural chemicals and use agricultural land in a more environmentally friendly manner, which can help to reduce ACEs. Accordingly, Hypothesis 3 is put forward:

Hypothesis 3.

GASP realizes ACEs’ reduction by increasing scale management efficiency.

4. Research Design

4.1. Variable Definition

4.1.1. Explained Variable

According to previous studies, we believe that ACEs are mainly caused directly or indirectly by the use of agricultural materials, such as fertilizers, pesticides, plastic film, diesel oil, and electricity consumed by agricultural irrigation activities.

Table 1. Each carbon source coefficient.

Carbon Emission Sources	Coefficient	Source
Fertilizer	0.8956 (t/t)	West T. O., Marland G. (2002) [55]
Pesticide	4.9341 (t/t)	Oak Ridge National Laboratory
Diesel oil used in agriculture	0.5927 (t/t)	IPCC (2007) [56]
Agriculture film	5.1800 (t/t)	Institute of Resource, Ecosystem and Environment of Agriculture, Nanjing Agricultural University
Irrigation	25.00 (kg/hm2)	Li et al. (2011) [57]

Note: The carbon emission coefficient of agricultural irrigation was originally 25.0000 (kg/hm²), but considering that only thermal power generation can lead to indirect carbon emission, it was multiplied by thermal power coefficient (that is, the ratio of thermal power generation to China’s total power generation). The final irrigation coefficient is 20.475 (kg/hm²).

displays the carbon emissions of several carbon sources. The ACEs are calculated by multiplying each carbon source by the coefficient and adding them together, see Equation (1):

$$\mathrm{ACEs}={\displaystyle \sum \mathrm{C}{\mathrm{E}}_{\mathrm{i}}}={{\displaystyle \sum {\mathrm{T}}_{\mathrm{i}}\mathrm{\delta }}}_{\mathrm{i}}$$

(1)

where ACEs represents the total carbon emissions from chemical inputs and land use, T_i represents the absolute amount of carbon sources in the first column in Table 1, and δ_i is the coefficient of Ti correspondingly in the second column. To more accurately represent the effectiveness of ACEs’ decrease, we also select the ACEs’ intensity (ACEI) as an explanatory variable, with a calculation method for the carbon emissions per unit planting area, see Equation (2):

$$\mathrm{A}\mathrm{C}\mathrm{E}\mathrm{I}=\frac{\mathrm{A}\mathrm{C}\mathrm{E}\mathrm{s}}{\mathrm{A}\mathrm{r}\mathrm{e}\mathrm{a}}$$

(2)

4.1.2. Core Explanatory Variable

We use the dummy variable to explore the effect of GASP on ACEs’ reduction. Thus, the core explanatory variable is the interaction term between the policy variable and the time dummy variable, treat_i × time_t. GASP has been piloted in five provinces since 2015 in Anhui, Shandong, Sichuan, Hunan, and Zhejiang, and was extended to the whole country in 2016. When a prefecture-level city_i is selected as a pilot area for GASP in year_t of the sample time, then time_t takes the value of 1 in year_t and the following years and takes the value of 0 before year_t.

4.1.3. Control Variables

According to the enlightenment of relevant studies, this paper selects Farmland, Population, Grain, Affected, Industrial, Urban, Education, and Environment as control variables. The variables are defined as shown in

Table 2. The main variable names and definitions.

Symbol	Variable Name	Variable Definition
ACEs	Total amounts of ACEs	Logarithmic value of regional total ACEs
ACEI	Intensity of ACEs	Logarithmic value of regional total ACEI
treati × timet	Policy variable	Dummy variable of GASP
ATPE	Agricultural technological progress efficiency	Measured by the DEA model
SME	Scale management efficiency	Measured by the DEA model
Farmland	Cultivated land scale	Logarithmic value of cultivated land scale
Population	Population size	Logarithmic value of population size
Grain	Planting structure	of grain production in total crop production
Affected	Affected area	Logarithmic value of affected area
Industrial	Secondary industrial structure	Value added of the secondary sector as a share of GDP
Urban	Urbanization rate	Share urban population in total population
Education	Educational expenditure	Logarithmic value of educational expenditure
Environment	Environmental protection expenditure	Logarithmic value of environmental protection expenditure

.

4.1.4. Mechanism Variables

Agricultural technological progress efficiency (ATPE) and scale management efficiency (SME) are selected as the mechanism variables. This paper uses the DEA model to measure efficiency. The output index selected in this paper is the added value of the primary industry after eliminating the price factor. The selected input indicators include the crop planted area, net fertilizer input, total power of agricultural machinery, and number of agricultural employees. Based on the DEA model, the input-oriented Malmquist productivity index method with constant returns to scale is used to decompose the ATPE and SME.

4.1.5. Data Sources

This paper collects data related to the 2011–2022 statistical yearbooks of provinces, cities, and statistical bulletins of prefectural-level cities. The study sample includes 172 prefectural-level cities in China. To reduce the potential impact of extreme values on the overall data, we shrink some of the data. In addition, to reduce heteroskedasticity and statistical bias, we take the logarithm of the absolute data to ensure the accuracy and reliability of the analysis results. The descriptive statistics of the variables are shown in

Table 3. Descriptive statistics.

Variables	Observations	Mean	St.dev	Min	Max
ACEs	2062	12.5666	0.8550	9.9155	19.9953
ACEI	2038	6.4735	0.6609	5.3226	8.7537
treati × timet	2064	0.5276	0.4994	0	1
ATPE	2064	1.0507	0.3549	0.1670	3.0180
SME	2064	1.0090	0.1567	0.5830	1.5670
Farmland	2063	5.1937	0.8916	3.0545	7.2039
Population	2062	5.0972	0.7486	1.7173	7.2123
Grain	2064	0.6544	0.1487	0.3742	0.8641
Affected	2057	6.6709	0.9340	0.4700	8.3485
Industrial	2064	0.4624	0.1230	0.1170	0.9161
Urban	2062	0.5617	0.1338	0.2044	0.9388
Education	2063	13.1873	0.8289	10.9708	16.2559
Environment	2064	11.0551	2.1388	0	14.1598

.

4.2. Regression Model

Based on the pilot of GASP, this paper constructs the following DID model:

$$\mathrm{Emissio}{\mathrm{n}}_{\mathrm{it}}=\mathrm{\alpha }+{\mathrm{\beta }}_1\mathrm{trea}{\mathrm{t}}_{\mathrm{i}}\times \mathrm{tim}{\mathrm{e}}_{\mathrm{t}}+{\mathrm{\beta }}_2\mathrm{Contro}{\mathrm{l}}_{\mathrm{it}}+{\mathrm{\eta }}_{\mathrm{t}}+{\mathrm{\delta }}_{\mathrm{i}}+{\mathrm{\epsilon }}_{\mathrm{it}}$$

(3)

Emission_it represents the ACE or ACEI, and the calculation method is shown in Equation (3). treat_i × time_t is the core explanatory variable. In year_t, if GASP is executed in city_i, then the value of treat_i × time_t is 1, if not, it is 0. Control_it embodies the control variables in this paper, η_t represents the year fixed effect, δ_i is the city fixed effect, and ε_it is the random error term. The coefficient β₁ is the key point which reflects the effectiveness of GASP. This paper’s Hypothesis 1 is supported if 1 is significantly less than 0, which shows that GASP contributes to the decrease in ACEs.

5. The Empirical Results

5.1. Benchmark Regression

According to the theoretical analysis and Hypothesis 1 in part 3, we expect that the implementation of GASP will have a positive effect on reducing ACEs. The results of the regression with Equation (3) are shown in

Table 4. Benchmark regression results.

Variables	(1)	(2)	(3)	(4)
	ACEs	ACEs	ACEI	ACEI
treati × timet	−0.218 ***	−0.203 ***	−0.231 ***	−0.213 ***
	(−3.55)	(−3.32)	(−4.54)	(−4.22)
Farmland		0.136 **		−0.195 ***
		(1.96)		(−3.44)
Population		0.238 *		0.072
		(1.89)		(0.70)
Grain		−0.562 *		−0.583 **
		(−1.67)		(−2.10)
Affected		−0.035 **		−0.047 ***
		(−2.29)		(−3.69)
Industrial		−0.511 *		−0.525 **
		(−1.83)		(−2.29)
Urban		0.014 ***		0.012 ***
		(3.15)		(3.29)
Education		−0.005 *		0.005 **
		(−1.83)		(2.28)
Environment		0.029 **		0.026 ***
		(−2.33)		(−2.58)
Constant	12.682 ***	10.513 ***	6.595 ***	7.043 ***
	(380.01)	(12.27)	(238.40)	(9.99)
City FE	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
Observations	2060	2055	2037	2032
R-squared	0.835	0.840	0.812	0.819

Note: ***, **, and * indicate significance at the statistical levels of 1%, 5%, and 10%, respectively. The t value is in parentheses.

. In the first and third column, the ACEs and ACEI are introduced as explained variables, respectively, and the treat_i × time_t interactions are used as explained variables for regression. The second and fourth columns introduce the control variables. It can be seen that the regression coefficients of the first and third column are significantly negative at the confidence level of 1%, that is, the implementation of GASP can reduce both the total amount and intensity of ACEs.

After introducing the control variables, the results are −0.203 and −0.213, respectively, and they are still credible at the level of 1%. As a result, Hypothesis 1 of this study is proved, namely that the implementation of GASP has a favorable impact on the decrease in ACEs. As analyzed earlier, unlike the traditional subsidy policy aimed at increasing production and income, GASP gives more prominence to the policy objective of “greening”, thus achieving a multi-party balance in the pursuit of economic, social, and ecological benefits. On the one hand, this may be due to the fact that GASP can change traditional production methods, promote innovation in agricultural technology, and motivate farmers to invest in environmental protection. On the other hand, subsidies can be used to promote green production, reduce the use of fertilizers and pesticides, and enhance farmers’ awareness of protecting their farmland. GASP can regulate the production behavior of subsidized farmers, so as to achieve ACEs’ reduction in agriculture.

The control variable regression coefficients are also as expected. Increases in the share of grain cultivation, the share of the secondary sector, the affected area, and public investments in education and environmental protection can reduce ACEs. These results validate the existing literature on the factors influencing ACEs. It is worth noting that an increase in cultivated land significantly increases ACEs, but significantly reduces the ACEI. A possible reason for this is that an increase in the area of cultivated land affects the denominator of the ACEI, which is the area of grain production. In other words, an increase in the cultivated area is faster than an increase in the ACEs, thus we obtain a pair of opposite coefficients for ACEs and ACEI.

5.2. Robustness Test

5.2.1. Parallel Trend Test

The time-varying DID model’s underlying presumption of a parallel trend calls for the ACEs and ACEI in the cities of the control and experimental groups to have a similar evolving tendency prior to the deployment of GASP. The following model was created to run a parallel trend test on the dataset. The specifics are as follows:

$$\mathrm{Emissio}{\mathrm{n}}_{\mathrm{it}}=\mathrm{\alpha }+{\mathrm{\beta }}_{\mathrm{i}}{\displaystyle \sum _{\mathrm{k}=-5}^5{\mathrm{D}}^{\mathrm{k}}_{\mathrm{it}}}+{\mathrm{\beta }}_{\mathrm{\gamma }}\mathrm{Contro}{\mathrm{l}}_{\mathrm{it}}+{\mathrm{\eta }}_{\mathrm{t}}+{\mathrm{\delta }}_{\mathrm{i}}+{\mathrm{\epsilon }}_{\mathrm{it}}$$

(4)

Based on the policy start year, D^k_it represents the intersection of the city_i and year_t. −k means that this year is k years earlier than the pilot year. If the ith city is k years prior than the pilot year, the D^−k_it is given the value 1, and others are given the value of 0. Similarly, if the ith city is k years later than the pilot year, the D^k_it is equal to 1. The meaning of other variables remains the same. We can create parallel trend graphs using the regression findings in Equation (4). Prior to the implementation of the policy, it was predicted that there would be no discernible difference between the treatment and control groups, or that the policy would not have a significant impact on the base year. The range of confidence intervals for the regression findings of D⁻⁵_it-D⁻²_it must contain 0 points in order to satisfy the aforementioned criteria. Correspondingly, there should be significant differences between the experimental group and the control group in the year after the implementation of the policy, which shows that the confidence interval does not include 0. The parallel trend tests for ACEs and ACEI are shown in Figure 1, respectively. The confidence intervals for the coefficients of the dummy variables in the policy year and beyond are far from the 0 point, meeting the parallel trend hypothesis requirement. The confidence intervals for the coefficients of the policy dummy variables D⁻⁵_it-D⁻²_it are around the 0 point.

5.2.2. Placebo Test

Considering that the choice of GASP’s pilot areas is not completely random, the empirical results may be influenced by other factors. Therefore, this paper further utilizes the placebo test to avoid the bias caused by missing variables. This research randomly selects treat_i × time_t from the dataset for benchmark regression to test the impression of this randomly selected variable on ACEs’ reduction.

The findings of 1000 random samplings are displayed in Figure 2. In the regression, the influence of the selected treat_i × time_t on the explained variables should be insignificant, because the dummy variables in the placebo test are chosen at random. Therefore, we draw a nuclear density map from the results of 1000 random samplings, and if the regression coefficient of treat_i × time_t is significantly different from 0, we pass the placebo test. As shown in Figure 2, the regression results display a normal distribution and cluster equally around zero. The benchmark regression test results, represented by the two vertical lines in the picture, deviate significantly from the findings of the placebo test. Thus, we succeeded in the placebo experiment.

5.2.3. Changing the Point of Implementation of the Policy

Considering advancing the pilot of GASP by one year, the regression results are shown in

Table 5. Change the policy point.

Variables	(1)	(2)	(3)	(4)
	ACEs	ACEs	ACEI	ACEI
treati × timet	0.040	0.043	0.051	0.050
	(0.64)	(0.71)	(0.99)	(0.99)
Constant	12.549 ***	10.374 ***	6.451 ***	6.895 ***
	(440.24)	(12.08)	(272.36)	(9.73)
Control variables	YES	YES	YES	YES
City FE	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
Observations	2060	2055	2037	2032
R-squared	0.834	0.839	0.810	0.817

Note: *** indicate significance at the statistical levels of 1%. The t value is in parentheses.

below, showing that the regression coefficients of ACEs and ACEI with the policy one year ahead of schedule are not significant. The results indicate that there is no policy effect when the policy is advanced by one year, suggesting that the timing of the policy is not random, and the regression results are credible.

5.2.4. Excluding Other Policy Effects

Simultaneous policies may also have some impact on ACEs. In order to ensure the robustness of the empirical results, this paper finds policies that are contemporaneous with GASP, incorporates them into the control variables, and utilizes the regression test in Equation (3).

The National Ecological Civilization Pilot Zone (NECPZ) policy was piloted in the Guizhou, Fujian, and Hainan provinces in 2016. All three provinces have requirements for ecological protection and pollutant prevention in the countryside, and, thus, may have an impact on ACEs. Therefore, the policy dummy variable (eco) of NECPZ is included in the range of control variables, and the regression results are shown in

Table 6. Excluding the NECPZ Policy.

Variables	(1)	(2)
	ACEs	ACEI
treati × timet	−0.208 ***	−0.217 ***
	(−3.39)	(−4.30)
eco	0.085	0.094 *
	(1.35)	(1.68)
Constant	10.584 ***	7.101 ***
	(12.34)	(10.06)
Control variables	YES	YES
City FE	YES	YES
Year FE	YES	YES
Observations	2055	2032
R-squared	0.840	0.819

Note: *** and * indicate significance at the statistical levels of 1% and 10%, respectively. The t value is in parentheses.

. After the introduction of the policy dummy variable, the core explanatory variables are still significantly negative, which means that it is proven that the policy of NECPZ does not affect the ACEs’ reduction effect of GASP.

The carbon emissions trading pilot (CETP) policy aims to bring the market mechanism into play to reduce carbon emissions. Agriculture, as an important source of carbon emissions, will likewise be included in this trading system, and, thus, may also have an impact on ACEs. We included the dummy variable representing the CETP policy (trade) in the control variables, and the regression results are shown in

Table 7. Excluding the CETP Policy.

Variables	(1)	(2)
	ACEs	ACEI
treati × timet	−0.216 ***	−0.214 ***
	(−3.55)	(−4.24)
trade	−0.037	−0.052
	(−0.51)	(−0.88)
Constant	10.589 ***	7.067 ***
	(12.49)	(10.01)
Control variables	YES	YES
City FE	YES	YES
Year FE	YES	YES
Observations	1984	2032
R-squared	0.835	0.819

Note: *** indicate significance at the statistical levels of 1%. The t value is in parentheses.

. After the introduction of the policy dummy variable, the core explanatory variables are still significantly negative, which means that it is proven that the policy of CEPT does not affect the ACEs’ reduction effect of GASP.

5.3. Mechanism Analysis

According to the previous theoretical analysis and research hypothesis, it is evident that the ACEs’ reduction effect of GASP is achieved through ATPE and SME. In this paper, ATPE and SME are selected as the mechanism variables, and the model is constructed as follows:

$${\mathrm{M}}_{\mathrm{it}}=\alpha +\omega \mathrm{trea}{\mathrm{t}}_{\mathrm{i}}\times \mathrm{tim}{\mathrm{e}}_{\mathrm{t}}+{\beta }_2\mathrm{Contro}{\mathrm{l}}_{\mathrm{it}}+{\eta }_{\mathrm{t}}+{\delta }_{\mathrm{i}}+{\epsilon }_{\mathrm{it}}$$

(5)

where M_it represents the mechanism variables, representing ATPE and SME, respectively, Control_it represents the control variables, and the rest of the variables are the same as those in the benchmark regression model.

As can be seen from

Table 8. Test of the mechanism effect.

Variables	(1)	(2)
	ATPE	SME
treati × timet	0.244 ***	0.098 ***
	(4.94)	(3.57)
Farmland	−0.064	0.144 ***
	(−1.15)	(4.63)
Population	−0.015	−0.181 ***
	(−0.15)	(−3.19)
Grain	0.823 ***	−0.316 **
	(3.04)	(−2.09)
Affected	−0.020	0.031 ***
	(−1.60)	(4.43)
Industrial	0.446 **	−0.157
	(1.99)	(−1.25)
Urban	0.005	−0.006 ***
	(1.50)	(−3.22)
Education	0.012 ***	−0.005 ***
	(5.41)	(−3.85)
Environment	0.009	0.007
	(0.90)	(1.26)
Constant	0.481	1.694 ***
	(0.70)	(4.40)
City FE	YES	YES
Year FE	YES	YES
Observations	2057	2057
R-squared	0.409	0.327

Note: *** and ** indicate significance at the statistical levels of 1% and 5%, respectively. The t value is in parentheses.

, the first column shows that GASP promotes technological progress at the 1% significance level. This is reflected in the fact that policy implementation significantly promotes the ATPE, with a coefficient of 0.244. That is, GASP can stimulate the progress of agricultural technology. According to the theory of the environmental Kuznets curve, as the efficiency of technological progress increases, the levels of local pollution and carbon emissions will decrease. The reasons for this are as follows: first, an increase in technological progress indicates an improvement in production efficiency using the same capital input. Technological progress can help to reduce the use of chemical elements such as fertilizers and pesticides, thereby reducing carbon emissions. Second, with technological progress, organic fertilizers, low-toxicity pesticides, and organic diesel have been introduced into agricultural production. Low-carbon production strategies such as precision fertilizer application and waste reuse have been widely adopted by farmers, thus reducing ACEs. Therefore, GASP can reduce ACEs and ACEI by promoting technological progress. Hypothesis 2 is supported.

The second column shows that GASP helps farmers to achieve large-scale management at the 1% significance level. This is reflected in the fact that policy implementation significantly promotes SME, with a coefficient of 0.098. First, an improvement in SME means that farmers can rationally allocate resources at a larger scale of operation. A larger scale of management improves the application efficiency of chemical elements and reduces their absolute amount of use. Second, large-scale farmers have a greater awareness of cultivated land protection. They prefer to choose more environmentally friendly production methods and reduce ACEs in the production process. Therefore, GASP can reduce ACEs and ACEI by improving scale management efficiency. Hypothesis 3 is supported.

5.4. Heterogeneity Test

First, the main grain-producing regions are closely related to the positioning of agricultural production policies. Given that ACEs are mainly generated in grain-producing regions, it makes sense that the majority of GASP implementation efforts will likewise concentrate on these areas. Therefore, this paper uses Equation (3) to estimate separate samples for the main and non-main grain-producing regions, and the findings are displayed in

Table 9. Heterogeneity test based on the heterogeneity test of main grain-producing areas.

Variables	Main Grain-Producing Areas		Non-Main Grain-Producing Areas
	ACEs	ACEIs	ACE	ACEI
treati × timet	−0.283 ***	−0.301 ***	−0.032	−0.002
	(−3.00)	(−3.92)	(−0.49)	(−0.04)
Farmland	0.273 **	−0.024	0.135 **	−0.167 ***
	(2.02)	(−0.21)	(2.11)	(−3.34)
Population	0.500 **	0.366 **	0.042	−0.218 *
	(2.52)	(2.26)	(0.26)	(−1.78)
Grain	−0.579	−0.248	−0.170	−0.751 **
	(−1.21)	(−0.64)	(−0.39)	(−2.20)
Affected	−0.033 *	−0.049 ***	−0.002	0.011
	(−1.66)	(−2.99)	(−0.07)	(0.47)
Industrial	−0.549	−0.638 **	−0.165	−0.159
	(−1.45)	(−2.06)	(−0.41)	(−0.51)
Urban	0.016 ***	0.013 ***	0.015 **	0.012 **
	(2.71)	(2.77)	(2.13)	(2.21)
Education	−0.011	0.001	−0.003	0.006 ***
	(−1.49)	(0.23)	(−1.20)	(3.27)
Environment	−0.028 *	−0.024 *	−0.027	−0.027
	(−1.79)	(−1.93)	(−1.24)	(−1.53)
Constant	8.431 ***	4.485 ***	10.911 ***	8.119 ***
	(5.83)	(3.80)	(9.76)	(9.28)
City FE	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
Observations	1230	1231	825	801
R-squared	0.747	0.652	0.936	0.945

Note: ***, **, and * indicate significance at the statistical levels of 1%, 5%, and 10%, respectively. The t value is in parentheses.

. Among them, the coefficients of the effect of policy implementation on ACEs and ACEI are not significant in samples from non-main grain-producing areas. The coefficients of the effect on ACEs and ACEI are −0.283 and −0.301, respectively, in samples from major grain-producing areas, which are significant at the 1% level and larger than the coefficients of the effect in non-main food-producing areas. Therefore, the effect of GASP is more pronounced in the main grain-producing areas.

Second, farmers in regions with high ACE levels could be more conscious of the value of sustainable development for agriculture. Their scale of operation, planting structure, and production techniques are more likely to be influenced by GASP, and, therefore, their behavior may change in the process. This paper uses Equation (3) to make separate estimate for samples above and below the median of the total ACEs, and the results are shown in

Table 10. Heterogeneity test based on whether it is located in a high-emission-level area.

Variables	High-Carbon-Emission City		Low-Carbon-Emission City
	ACEs	ACEI	ACEs	ACEI
treati × timet	−0.224 ***	−0.222 ***	−0.003	0.010
	(−3.52)	(−3.31)	(−0.09)	(0.35)
Farmland	0.118	−0.233 **	0.134 ***	−0.107 ***
	(1.14)	(−2.13)	(4.96)	(−4.01)
Population	0.128	0.101	0.389 ***	0.099 **
	(0.64)	(0.48)	(7.75)	(2.01)
Grain	0.491	0.615	0.342 **	−0.007
	(1.30)	(1.54)	(2.18)	(−0.05)
Affected	−0.029 *	−0.046 ***	0.016 **	0.011
	(−1.77)	(−2.71)	(2.24)	(1.53)
Industrial	−0.336	−0.198	0.363 ***	0.328 **
	(−1.13)	(−0.63)	(2.66)	(2.45)
Urban	0.016 ***	0.013 **	0.010 ***	0.005 **
	(2.88)	(2.24)	(5.13)	(2.58)
Education	−0.002	−0.003	−0.009 ***	0.006 ***
	(−0.42)	(−0.66)	(−7.12)	(5.34)
Environment	−0.061 ***	−0.055 **	0.002	0.004
	(−2.71)	(−2.33)	(0.53)	(0.94)
Constant	11.184 ***	6.870 ***	8.591 ***	5.776 ***
	(8.53)	(4.97)	(23.17)	(15.82)
City FE	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
Observations	1230	1231	825	801
R-squared	0.747	0.652	0.936	0.945

Note: ***, **, and * indicate significance at the statistical levels of 1%, 5%, and 10%, respectively. The t value is in parentheses.

. The coefficients of the effects of policy implementation on ACEs and ACEI are not significant in samples from low-carbon-emission areas. In contrast, the impact coefficients of ACEs and ACEI in samples from areas with high carbon emission levels are −0.224 and −0.222, respectively, and are significant at the 1% level. Therefore, the effect of GASP is more pronounced in relatively high-carbon-emission areas.

6. Conclusions and Policy Implications

Considering the background of the integration of “carbon peak” and “carbon neutrality” into the broader framework of ecological civilization construction, it is crucial to investigate the optimization strategy of ACEs’ reduction. In this paper, the IPCC inventory method is used to calculate the total ACEs, and ACEI is introduced to better reflect the efficiency of ACEs’ reduction. The DID model is used to test the emission reduction effect of GASP. According to the findings of benchmark regression, the pilot of GASP does significantly reduce the total amount and intensity of ACEs, demonstrating that GASP helps to reduce ACEs. The effect of GASP is more significant in the main grain-producing areas and cities with high carbon emission levels. Mechanism tests show that the ATPE and SME of agricultural production are effective mechanisms for GASP to reduce ACEs. The precise and detailed evaluation and examination of the effect of GASP on ACEs in this paper can be a useful supplement to existing research in the area of agricultural subsidies. Sorting out the abatement mechanism of agricultural subsidies can help us to more clearly interpret the mechanisms of financial support policies for agriculture and provide useful references for further optimization.

According to the main conclusions of this paper, we can clearly understand the impact, mechanism, and heterogeneity of GASP on ACEs’ reduction. Based on the above empirical results, this paper puts forward some suggestions on how to optimize GASP.

First, GASP should further clarify the policy intentions of promoting low-carbon and sustainable agricultural development. On the one hand, it is necessary to establish a common assessment standard for the quality of farmland and use it to classify farmland into different grades. Under this standard, the higher the quality of cultivated land, the more subsidy funds farmers will receive from GASP, so as to incentivize farmers to consciously protect their farmland. On the other hand, subsidized funds should be used more for sustainable development. The government can provide more subsidies for organic fertilizers, low-toxicity pesticides, and highly degradable films, thus reducing the adoption cost of low-carbon production behavior among farmers.

Second, the government has to make low-carbon policies more relevant and support the advancement of agricultural technology. In order to establish localized low-carbon agricultural production systems, the government should continuously increase its financial support for the advancement of agricultural technology, especially green and low-carbon production technology, and reduce the adoption cost of farmers by the means of subsidies. Moreover, it is necessary to strengthen the adoption of advanced technologies by farmers. The importation of green technologies should be accompanied by simultaneous professional instruction.

Third, government departments should encourage land transfer, which is important for achieving the optimal agricultural production efficiency and promoting large-scale planting operations. To achieve this large-scale efficiency, farmers are willing to embrace more intense and large-scale production techniques with the help of legislative incentives. The government should also encourage new agricultural operations fields and hire planting specialists to train a number of professional farmers. For farmers who transfer their land, relevant protection policies should be established to form a linkage between urban and rural employment departments to promote non-agricultural employment and ensure farmers’ income.

Fourth, the government should improve the differentiated GASP by adapting to local conditions. Each region actually has a varied agricultural situation. The government should fully analyze the agricultural situations in different regions, understand the regional differences, and formulate different policies according to local conditions. For example, in some areas, the carbon reduction effect of the policy is relatively weak. Therefore, it is necessary to strengthen large-scale subsidies in non-main grain-producing areas. In low-carbon-emission regions, farmers’ knowledge of low-carbon production is weak, so subsidy funds should be used more for education and promotion, so as to ensure the full impact of the incentive effect of GASP.