Does Environmental Regulation Affect China’s Agricultural Green Total Factor Productivity? Considering the Role of Technological Innovation

Abstract

Agricultural green total factor productivity (AGTFP) is vital to sustainable agricultural development (SAD), which plays a pivotal role in achieving high-quality economic growth in China. The current research quantified China’s AGTFP from 2007 to 2019 with the Super-SBM model and the GML index. Subsequently, it examined the impact of environmental regulation (ER) on AGTFP and its heterogeneity. Finally, this study developed a mediating effect model and a panel threshold model to investigate the role of technological innovation (TI) in affecting environmental regulation (ER) on AGTFP. The findings indicate that the following: (1) The average annual growth rate of AGTFP is 7.84%, which is mostly driven by green technological innovation progress. (2) ER has a significant positive impact on AGTFP with noticeable regional differences. The eastern and central regions experience a more substantial promotion effect compared to the western region. Additionally, the prominent grain-producing areas and main grain-marketing areas have a more significant promotion effect compared to the grain-balanced areas. The promotion effect of heterogeneous ER on AGTFP varies, with the effects of command-and-control environmental regulation (ERC), market-based incentives for environmental regulation (ERM), and public participation regulation (ERP) decreasing in magnitude. (3) The mechanism analysis reveals that promoting TI is a crucial way to enhance AGTFP through ER. There exists a notable dual threshold for TI in ER, encompassing both ERC and ERM. Moreover, AGTFP becomes increasingly pronounced. This study presents a novel perspective for promoting SAD, with a focus on the rise in AGTFP and the path to achieving it.

1. Introduction

In the past 75 years, China’s agricultural economy has made extraordinary success, with the gross value of agricultural output growing from CNY 0.13 trillion in 1979 to CNY 8.44 trillion in 2022, at an average annual growth rate of more than 5%, which has forcefully guaranteed China’s food security. However, with the acceleration of industrialization and urbanization, the agricultural sector in China is facing severe environmental challenges. Due to pollution from industrial waste such as the “three wastes” (waste gas, waste water, and waste solids), the rate of soil contamination exceeding standard levels in China has reached 19.4%. The area of arable land affected by heavy metal pollution exceeds 20 million hectares, accounting for one-sixth of the country’s total arable land. Furthermore, the spatial and temporal distribution of water resources is highly uneven. The Huaihe River Basin and regions to the north, which cover 62% of the country’s arable land, have less than 20% of the total water resources. Hebei Province has become the world’s largest groundwater funnel area [1]. Simultaneously, the traditional agricultural development model that heavily depends on inputting numerous agricultural production factors has caused a significant strain on resources and the environment. This has resulted in various agroecological issues, including a sharp increase in agricultural carbon emissions, the severe pollution of agricultural water bodies, and intensified pollution on agricultural surfaces. Consequently, this has greatly hindered the progress of constructing an ecological civilization in China. The Twentieth National Congress of the Communist Party of China and other significant national plans have repeatedly emphasized the importance of speeding the green transformation of the development mode. China is currently undergoing a critical phase of economic transformation, focusing on optimizing its economic structure and transitioning to new growth drivers. This shift aims to cultivate sustainable development momentum over the long term. In this context, transforming agricultural development practices is pivotal to advancing sustainability and ensuring the country’s ecological and economic resilience.

Driven by the principles of sustainable development, agriculture—fundamental to societal well-being—has increasingly underscored the importance of its own sustainability. Sustainable development, by definition, advocates for a balance between economic growth, social equity, and environmental stewardship [2]. Agriculture not only constitutes a vital component of global ecosystems but also serves as a cornerstone for ensuring food security and fostering rural economic development. In light of the dual constraints posed by resource scarcity and environmental degradation, scholars and policymakers are reassessing the trajectory of agricultural development. Against this backdrop, the concept of AGTFP has emerged. Traditional measures of agricultural total factor productivity often neglect resource and environmental limitations, thereby impeding accurate evaluations of agricultural performance [3,4]. To address this gap, some scholars have incorporated environmental and resource factors into the TFP framework, enabling the assessment of the resource and environmental costs incurred during agricultural growth and offering a more comprehensive measure of agricultural sustainability. AGTFP extends beyond conventional TFP by integrating negative externalities such as environmental pollution into its calculations. It emphasizes the pursuit of an integrated approach, wherein technological innovation and optimal resource allocation achieve synergies among economic, social, and environmental outcomes. Viewed through the lens of classical political economy, improvements in AGTFP reflect a deeper understanding of the interplay between productivity and natural resources, while also signifying a proactive response to environmental constraints in modern agricultural development [4,5,6,7]. This theoretical framework not only builds on classical economics’ emphasis on productivity efficiency but also incorporates contemporary ideals of environmental protection and social equity, providing both a robust analytical tool and practical guidance for SAD in the 21st century [8].

At present, extensive research was conducted on the factors influencing AGTFP, with scholars examining the roles of agricultural digitalization [9,10], industrial structure [11], and mechanization [12]. Among these factors, ER is the most efficient way to manage excessive factor consumption and negative environmental externalities among the various factors affecting AGTFP [13,14]. The “National Agricultural Green Development Plan for the 14th Five-Year Plan”, released in collaboration with the Ministry of Agriculture and Rural Development and other ministries and commissions, emphasizes that the nation should share the same objective of creating a government–market–society participatory model to promote SAD and rural areas jointly. The Chinese government has recently implemented a number of ERs for agricultural output under the auspices of the new development concept in an effort to guarantee a “win-win” scenario for both environmental preservation and agricultural economic growth. So, how effective are these kinds of ERs? And have they contributed to the improvement of AGTFP? It is essential to properly assess the development of AGTFP under the implementation of ER, and the key lies in realizing the core driving role of TI on AGTFP [15]. Based on that, the current paper examines the effect of ER on AGTFP and its mechanism in China.

Compared with the existing literature, the study has the potential to make contributions in the following aspects: First, it is recognized that both agricultural carbon emissions and surface contamination are undesirable outcomes. The “resource-energy-environment-agricultural economy” theoretical evaluation system was developed based on this. Furthermore, in order to support policy development, the variations in the effects of various regulatory types on AGTFP and their geographical heterogeneity are examined. Lastly, a thorough examination of the TI pathways of ER impacting AGTFP is conducted. The current state of regional TI levels is also examined, along with the TI threshold features of various ER types pertaining to AGTFP.

The remainder of this manuscript is ordered as follows: Section 2 covers a review of the related literature and theoretical hypotheses; Section 3 details the methodology, variables, and data description; Section 4 presents the empirical results; and Section 5 provides conclusions and policy recommendations.

2. Literature Review and Hypothesis

2.1. Impact of ER on AGTFP

When discussing the effect of ER on AGTFP, the foremost thing is to clarify the difference between green agricultural growth and conventional agricultural growth. The rapid pace of ecological civilization construction and the overuse of agricultural resources need the achievement of SAD through the improvement of AGTFP [4,12]. Hence, it is imperative to examine the influence of ER on SAD from the input–output perspective when measuring AGTFP. AGTFP differs from the conventional agricultural total factor productivity, where the latter only considers the positive impact of inputs and the former takes into account the limitations imposed by environmental resources. AGTFP involves assessing the combined changes in input factors such as fertilizers, pesticides, agricultural films, and machinery, as well as the desired agricultural output value and the undesirable pollutants [13,14,15].

Some argue that laws, regulations, administrative orders, market mechanisms, etc. can be employed to influence agricultural entities to enhance AGTFP. These regulations can be used to drive agricultural entities to change the scale structure of input factors, improve resource utilization, and thus promote SAD [16,17]. In other words, minimizing the “bad output” that is harmful to the natural ecology and increasing the “good output” under specific inputs is necessary [13]. Numerous research findings indicate that the primary driver of AGTFP growth is agricultural green technological progress. Furthermore, it was observed that ER can effectively stimulate the growth of AGTFP by aligning the direction of agricultural green technological progress with the changes in the scale structure of agricultural input factors and agricultural production methods.

An increased intensity of ER can limit undesirable output. Because the government will increase its support for resource conservation and environmental protection when formulating ER. Thus, the effective use of fertilizers, pesticides, and agricultural films on the input side and the reduction in undesirable outputs will promote the increase in AGTFP. Although in the short term, the increased intensity of ER will bring about part of the “Compliance cost”, higher environmental costs crowd out profits and R&D expenses, thus making ER a disincentive to AGTFP [18]. However, in the long term, “Porter’s hypothesis” emphasizes that strict and appropriate ER can “force” agricultural entities to carry out TI, giving rise to more reasonable and efficient factor combinations, more intelligent and precise agricultural equipment, and cleaner and friendlier agricultural production methods, which can not only eliminate the problem of pollution but also reduce the profitability of agricultural entities. This not only eliminates the “Compliance cost” effect but also improves the competitiveness of agricultural entities and the comprehensive capacity of the agricultural sector through the “Innovation offset” effect and ultimately promotes the improvement of AGTFP [19,20].

China’s ER could be divided into three types, namely ERC, ERM, and ERP. Their role in the AGTFP path and effect are not the same due to the differences in subjects and forms of implementation [21]. The ERC type relies on government authority, which is a crucial tool in China for preventing and controlling environmental pollution. It involves the government implementing laws and regulations at the national level to restrain and punish agricultural entities that fail to comply with emission standards and technical specifications [22]. For example, the Chinese government issued regulations in 2017 to reduce agricultural inputs, promote the efficient use of straws, and prevent and control agricultural surface source pollution. In response to immediate ER, agricultural entities will either proactively expand their investments in environmental pollution prevention and control, undertake corrective measures, or cease their agricultural operations. SAD may be impeded as a consequence. In the long term, agricultural entities will allocate the funds toward end-of-pipe pollution prevention and control, fostering TI under this kind of ER constraint. Adopting a more rational factor combination structure and production mode can also significantly boost AGTFP [23].

Nevertheless, differences in enforcement efforts and expenses among regions undermine the efficiency of ERC. ERM prioritizes harnessing the power of the market economy, thereby maximizing the incentive effect [24,25,26]. It regulates the prices of production means, mainly utilizing sewage charges and subsidies, which incentivize agricultural operators to take the initiative to reduce environmental damage, thus influencing their production decisions and modes of production [27,28]. Examples of such charges are fees for sewage disposal and taxes imposed on agricultural entities that are highly polluted. ERP, which relies on the public’s environmental awareness, is a necessary addition to the first two. This type of regulation makes full use of the citizens’ rights to environmental supervision and litigation, promoting the environmental protection actions of agricultural entities through the social awareness of environmental protection [29], self-media, letters, and petition centers, etc., such as the newly emerged environmental hearings in recent years.

In particular, China has vast regional variations, and the three categories of regulations have varying effects on the AGTFP of different regions. For instance, a study conducted in 2022 by Guo and Li [30] discovered that ERC exhibited a U-shaped nonlinear relationship with the AGTFP of the region and its neighboring regions and that ERP will contribute to the promotion of the AGTFP of the region and its neighboring regions, while ERM has no discernible effect. Local factors must be taken into account while developing and enforcing ER. Accordingly, this paper proposes the following:

H1. 

ER is conducive to promoting AGTFP and exhibits notable regional variation, meaning that ER has diverse impacts on AGTFP across different regions.

H2. 

Different types of ERs have varying effects on AGTFP impacts.

2.2. Mediating and Threshold Effects of TI

Based on the above analysis, China’s ER has an impact on AGTFP. Meanwhile, in order to make up for the “Compliance cost”, the government will implement the innovation-driven development strategy, leading the agricultural entities to adopt new technologies, new equipment, and new processes, which will promote the improvement of AGTFP. The existing literature has shown that ER can promote SAD through TI [31,32,33]. On the one hand, the development of modern technology has led to an accelerated process of modernization and industrialization of agriculture. Advanced agricultural machinery and equipment and high-quality varieties of agricultural products have improved food productivity and increased desirable output [34]. On the other hand, agricultural entities have adopted a more reasonable combination of factors to promote the factors’ utilization efficiencies, such as agricultural diesel fuel, chemical fertilizers, and pesticides, reducing undesirable outputs. At the same time, this approach achieves improved agricultural pollution defense and treatment technologies in the process of green transformation as a way to promote AGTFP.

Furthermore, as environmental regulation becomes more stringent, agricultural entities need to address undesirable byproducts. Taking into account financial limitations and the amount of technological advancement, they will have to make decisions about whether to focus on treating the byproducts at the end of the production process or the source. Terminal management mitigates the negative externalities resulting from agricultural production’s undesirable outputs. However, it falls short of fundamentally reducing these undesirable outputs and has a limited impact on improving AGTFP. When reasonable ERs act as “Innovation offset”, they encourage agricultural entities to address environmental pollution at its source and utilize scientific TIs to adjust the agricultural industry’s structure, energy usage, and production methods. This approach significantly enhances AGTFP, which was supported by several scholars [16,35]. Hence, the impact of ER is intricately linked to the extent of TI. In other words, the advancement of technical innovation will determine the influence of ER on AGTFP. Consequently, this paper presents the following hypotheses:

H3. 

TI exhibits a significant role in the transmission path effect of ER on AGTFP.

H4. 

ER is influenced by TI and exhibits a threshold effect on AGTFP, and the threshold varies with the type of ER.

The relationships and mechanisms among ER, TI, and AGTFP are shown in Figure 1 below.

3. Research Methodology, Variables, and Data

3.1. Research Methodology

Measurement of AGTFP. When selecting measurement methodologies, the two primary options are Stochastic Frontier Analysis (SFA) and Data Envelopment Analysis (DEA). The use of DEA is prevalent in efficiency evaluation due to its ability to consider numerous input–output units and avoid the inherent assumption bias of SFA, which assumes a normal distribution for the model setup and stochastic disturbance term. Referring to Tone’s (2003) [36] study, this paper utilizes the Super-SBM model under Variable Return to Scale (VRS) to quantify AGTFP, incorporating non-expected outcomes. The measurement is conducted using the following equation:

$${\rho }^{\ast }=\min \frac{1+\frac{1}{m}{\sum }_{i=1}^m\frac{s_i^{-}}{x_{ik}}}{1-\frac{1}{q_1+q_2}}\left({\displaystyle \sum }_{r=1}^{q_1}\frac{s_r^{+}}{y_{rk}}+{\displaystyle \sum }_{r=1}^{q_2}\frac{s_r^{u-}}{u_{rk}}\right)$$

$$\mathrm{s.t.}\{\begin{array}{l}{\displaystyle \sum _{j=1,j\ne k}^n}{x}_{ij}{\lambda }_j-s_i^{-}\le x_{ik}\\ {\displaystyle \sum _{j=1,j\ne k}^n}{y}_{rj}{\lambda }_j-s_r^{+}\ge y_{rk}\\ {\displaystyle \sum _{j=1,j\ne k}^n}{u}_{rj}{\lambda }_j-s_r^{u-}\le u_{rk}\\ {\displaystyle \sum _{j=1,j\ne k}^n}{\lambda }_j=1\\ {\lambda }_j,s_i^{-},s_r^{+}\ge 0\\ j=1,2, \mathrm{L},n(j\ne k);i=1,2, \mathrm{L},m;r=1,2, \mathrm{L},q\end{array}$$

(1)

In Equation (1), ${\rho }^{\ast }$ is the efficiency of decision-making units (DMUs), where the higher the value, the higher the agricultural greening. Agricultural production is a continuous and ongoing process; however, the efficiency above is measured by static analysis. The Malmquist index enables dynamic analysis of the efficiency level of DMUs and, together with a directional distance function that includes undesirable outputs, constitutes the Malmquist–Luenberger (ML) index. To address the issue of linear insolvability in the ML index, the current paper draws upon the research findings of Oh (2010) [37]. It utilizes the Global Malmquist–Luenberger (GML) index, conducting an in-depth analysis of the growth and the source of AGTFP.

$$GM{L}^{t,t+1}\left(x^t,y^t,b^t,x^{t+1},y^{t+1},b^{t+1}\right)=\left(1+{\mathrm{D}}_G^T\left(x^t,y^t,b^t\right)/\left(1+{\mathrm{D}}_G^T\left(x^{t+1},y^{t+1},b^{t+1}\right)\right.\right.$$

(2)

In Equation (2), if $GMLt,t+1$ is equal to 1, it means that AGTFP did not change from period $t$ to period $t+1$; if $GMLt,t+1$ is greater than 1, it means that AGTFP increased; and vice versa, it decreased. The $GML$ index is further decomposed as follows:

$$\begin{gathered} GM{L}^{t,t+1}\left(x^t,y^t,b^t,x^{t+1},y^{t+1},b^{t+1}\right)={\displaystyle \frac{1+{\mathrm{D}}_G^T\left(x^t,y^t,b^t\right)}{1+{\mathrm{D}}_G^T\left(x^{t+1},y^{t+1},b^{t+1}\right)}} \\ ={\displaystyle \frac{1+{\mathrm{D}}_C^t\left(x^t,y^t,b^t\right)}{1+{\mathrm{D}}_C^{t+1}\left(x^{t+1},y^{t+1},b^{t+1}\right)}}\times \left[{\displaystyle \frac{1+{\mathrm{D}}_G^T\left(x^t,y^t,b^t\right)/1+{\mathrm{D}}_C^t\left(x^t,y^t,b^t\right)}{1+{\mathrm{D}}_G^T\left(x^{t+1},y^{t+1},b^{t+1}\right)/1+{\mathrm{D}}_C^{t+1}\left(x^{t+1},y^{t+1},b^{t+1}\right)}}\right] \\ ={\displaystyle \frac{GT{E}^{t+1}}{GT{E}^t}}\times \left[{\displaystyle \frac{\mathrm{P}{G}_{t+1}^{t+1}}{\mathrm{P}{G}_t^{t,t+1}}}\right] \\ =GTE{C}^{t,t+1}\times GT{C}^{t,t+1} \end{gathered}$$

(3)

In Equation (3), $GTC$ and $GTEC$ represent the green technology progress index and green technology efficiency index, respectively. When the values of $GTC$ and $GTEC$ exceed 1, it signifies advancements in green technology progress and green technology efficiency, respectively. The converse also holds true.

3.1.1. Benchmark Regression Model

The following model was used to assess the impact of ER on AGTFP:

$$y_{it}={\beta }_0+x_{it}{\beta }_1+k_{it}{\beta }_2+u_i+{\lambda }_t+{\epsilon }_{it}$$

(4)

In Equation (4), $y_{it}$, $x_{it}$ is AGTFP and the ER in province $i$ of period $t$, respectively. Kit represents control variables, ${\mu }_i$ and ${\lambda }_t$ denote individual and time effects, respectively. ${\epsilon }_{it}$ is a random perturbation term, ${\mu }_i$ and ${\epsilon }_{it}$ are assumed to be random variables, and if their distributions are not correlated with $x_{it}$ and ${\epsilon }_{it}$, the model is called a random effects model; and if ${\mu }_i$(${\lambda }_t$) is correlated with the explanatory variable $X_{it}$, it is called a fixed effects model.

3.1.2. Mediated Effects Model

To verify H3, i.e., to test whether ER promotes AGTFP through TI, the paper refers to the method put forward by Baron and Kenny (1986) [38] and Xu et al. (2021) [39] to construct the mediating effect model.

$$y_{it}={\alpha }_0+{\alpha }_1{X}_{it}+\Sigma {\gamma }_i{Z}_{it}+u_i+{\epsilon }_{it}$$

(5)

$$M_{it}={\beta }_0+{\beta }_1{X}_{it}+\Sigma {\gamma }_i{Z}_{it}+u_i+{\epsilon }_{it}$$

(6)

$$y_{it}={\delta }_0+{\delta }_1{X}_{it}+{\delta }_2{M}_{it}+\Sigma {\gamma }_i{Z}_{it}+u_i+{\epsilon }_{it}$$

(7)

In Equations (5)–(7), $y_{it}$ is the dependent variable; $X_{it}$t is the core independent variable; M_it is the mediating variable; and $Z_{it}$ is the control variable, which is the random disturbance term.

Equations (5)–(7) are tested sequentially. If the coefficients ${\delta }_1$, ${\beta }_1$, and ${\delta }_2$ are significant, the mediation effect exists; if ${\delta }_1$ is significant, but at least one of ${\beta }_1$ and ${\delta }_2$ is not significant, the Bootstrap method should be used to test it; if the coefficients ${\delta }_1$ are significant and the sign is the same as that of ${\beta }_1$ and ${\delta }_2$ in the presence of the mediation effect, then it is considered to be a partially mediated effect, otherwise it is a complete mediation or a masking effect.

3.1.3. Panel Threshold Model

According to the mechanism above, the promotion effect of ER on AGTFP is affected by the level of agricultural TI. There may be a certain threshold in the level of agricultural TI. When it reaches the threshold point, the promotion effect of ER on AGTFP jumps or even deflects. Therefore, based on the benchmark regression, this paper refers to Hansen (1999) [40] to construct a panel threshold model to test whether there is a TI threshold effect in the process of ER on AGTFP.

$$y_{it}={\beta }_0+{\mu }_i+{\lambda }_t+x_{it}^{\prime }{\beta }_{11}\cdot I\left(q_{it}\le {\gamma }_1\right)+x_{it}^{\prime }{\beta }_{12}\cdot I\left({\gamma }_1\le q_{it}\le {\gamma }_2\right)+$$

(8)

$$\cdots +x_{it}^{\prime }{\beta }_{1n+1}\cdot I\left(q_{it}>{\gamma }_n\right)+k_{it}^{\prime }{\beta }_2+{\epsilon }_{it}$$

In Equation (8), $I\left(\cdot \right)$, $q_{it}$, and ${\gamma }_1$ are the indicator function, threshold variable, and threshold value, respectively. The coefficient of ER on AGTFP changes when the threshold variable agricultural TI is located in different intervals.

3.2. Variable Selection and Data Sources

3.2.1. The Dependent Variable: AGTFP

The Super SBM-GML model was used to perform the measurement. Land, which is determined by the total area sown with crops; labor, which is the product of the number of people working in the primary industry and the percentage of agricultural added value in the total value added of agriculture, forestry, animal husbandry, and fisheries; capital, which is expressed as the total power of agricultural machinery, the number of large livestock, the discounted amount of fertilizer application, pesticide usage, and the quantity of agricultural film used, respectively; and irrigation, which is measured by the actual effective irrigated area, are the main input variables.

The desirable output is the gross agricultural output, while the undesirable outputs pertain to pollution originating from agricultural sources. This pollution encompasses five specific components: loss of fertilizers, livestock and poultry husbandry, agricultural solid waste, aquaculture, and rural life (Zou et al., 2020) [41]. The second undesirable output is agricultural carbon emissions. Li et al. (2011) [42] conducted a study that specifically quantified the carbon emissions from six essential agricultural production and management activities. The GML index measures the growth rate of agricultural total factor productivity, not the productivity itself. This paper adopts the approach of Cai and Zhou (2017) [43], which assumes that the AGTFP in 2007 is 1. The level of AGFTFP in 2008 is then calculated by multiplying the efficiency in 2007 by the GML index in 2008, and so forth. The AGTFP in the provincial area from 2008 to 2019 is the dependent variable.

3.2.2. The Core Independent Variable: ER

This study utilizes the entropy approach to generate the environmental regulatory index, which is then employed as the core independent variable. The ER can be divided into three types: ERC, ERM, and ERP. According to Guo and Li (2022) [30], ERC is characterized by the number of policies related to agricultural pollution prevention and control and environmental protection regulation. The levying of sewage tax incorporates the expenses associated with environmental pollution into the adjustments of prices for resources and goods, thereby influencing the distribution of resources and the production methods employed by agricultural enterprises. Thus, this study employs sewage charges and environmental taxes to classify ERM. The oversight authority at the societal level promotes the enhancement of environmental protection awareness among agricultural management entities. This study categorizes the form of ERP based on the number of recommendations from the National People’s Congress. Lastly, the entropy method is employed to assess the ER quantitatively.

3.2.3. The Control Variables

Agricultural structure (INS), the indicator of INS, is determined by calculating the proportion of value added by the planting industry in relation to the total value added from agriculture, forestry, livestock, and fisheries. A higher proportion of cultivation leads to increased scale, which promotes the intensive production of agricultural entities, which is projected to have a beneficial impact on AGTFP. Affected disaster rate (ADR) is the proportion of the entire area seeded with crops affected by this damage. The deepening of ADR restricts AGTFP. Concerning rural economic development (AE), the degree of AE is assessed based on the proportion of the primary sector’s contribution to the overall GDP. The rise in rural economic development leads to a significant increase in output growth and results in higher emissions of pollutants throughout the agricultural production process. Consequently, the direction of its impact remains uncertain. Agricultural machinery intensity (MAC) is expressed as the total power of agricultural machinery as a percentage of the total sown area of crops. The intensive use of modern agricultural machinery has a scale effect on agricultural production but inevitably increases carbon emissions; thus, the impact of the increase in the density of agricultural machinery on AGTFP needs to be further tested. Trade dependency (TRA), is expressed as the total amount of regional agricultural exports and imports as a percentage of total agricultural output. Different regions have different levels of agricultural development and the structure of people’s food demand, and the unfolding of agricultural trade affects the relative prices of agricultural production factors and products, which in turn affects agricultural production decisions and production methods (Minten et al., 2010 [44]), with the direction of the impact unknown.

3.2.4. Mechanism Variables: Technological Innovation (TI)

TI functions as an intermediary and threshold variable. At present, the primary indices for assessing TI include the allocation of funds toward scientific research, the recruitment of scientific research workers, and the number of patent applications. Considering that the first two indicators are difficult to obtain and the number of agricultural patent applications is more accurate, this paper employs the number of agricultural patent applications as a metric for measuring TI.

This study utilizes panel data from 30 provinces in China from 2007 to 2019. Notably, Hong Kong, Macao, Taiwan Province of China, and Tibet are excluded due to significant missing data on indicators. The primary data sources include the China Agricultural Yearbook, China Agricultural Products Trade Development Report, China Statistical Yearbook, China Environmental Statistical Yearbook, and the National Bureau of Statistics (NBS) website (http://www.stats.gov.cn/, accessed on 7 January 2024). The producer price index (2007 = 100) was utilized to transform the indicators that required monetary calculations. Simultaneously, certain variables, such as the independent and core independent variables, were transformed into logarithmic forms to enhance comparability and mitigate heteroscedasticity while preserving the time-varying nature of the original variables. The descriptive statistics are as follows (

Table 1. Descriptive statistics of variables.

Variables	Variable Description	Sample Size	Average	Std.	Min	Max
AGTFP	Agricultural green total factor productivity	360	0.484	0.382	−0.502	1.702
ER	Environmental regulation	360	7.146	0.734	4.928	8.634
ERC	Command-and-control ER	360	5.023	0.778	2.639	7.265
ERM	Market-based incentives ER	360	10.687	0.939	7.955	12.791
ERP	Public participation ER	360	5.857	0.967	3.219	8.673
INS	Agricultural structure	360	0.493	0.084	0.314	0.703
ADR	Affected disaster rate	360	0.181	0.135	0.006	0.695
TI	Technological innovation	360	1.823	2.628	0	16.780
AE	Agricultural economic development	360	0.102	0.054	0.003	0.300
MAC	Agricultural machinery intensity	360	0.618	0.243	0.245	1.386
TRA	Trade dependence	360	0.507	1.361	0.001	9.842

).

4. Results and Analysis

4.1. Estimation Results of AGTFP

The measurements of AGTFP in China are displayed in

Table 2. The overall AGTFP and its decomposition in China, 2007–2019.

Year	GTFPC	GTEC	GTC	Year	GTFPC	GTEC	GTC
2007—2008	1.0445	1.0154	1.0287	2016—2017	1.0920	1.0580	1.0321
2008—2009	1.0742	1.0299	1.0430	2017—2018	1.1529	1.0316	1.1175
2009—2010	1.1762	0.9266	1.2693	2018—2019	1.1384	1.0120	1.1249
Average of 2007 to 2010	1.0969	0.9896	1.1085	Average of 2015 to 2019	1.0707	0.9911	1.0803
2010—2011	1.0624	0.9834	1.0802	Average of the eastern region	1.0776	1.0070	1.0701
2011—2012	1.1034	1.0098	1.0926	Average of the central region	1.0718	0.9813	1.0922
2012—2013	1.1007	0.9974	1.1035	Average of the western region	1.0837	0.9799	1.1059
2013—2014	1.0495	0.9727	1.0650	Average of main grain-producing areas	1.0779	0.9738	1.1069
2014—2015	1.0534	0.9727	1.0830	Average of main grain-marketing area	1.0763	1.0278	1.0472
Average of 2010 to 2015	1.0737	0.9897	1.0848	Average of grain-balancing area	1.0782	0.9873	1.0974
2015—2016	0.9170	0.8735	1.0497	National average	1.0784	0.9901	1.0892

Note: The table’s mean values are geometric, as shown below.

, where GTFPC, GTEC, and GTC represent overall green total factor productivity, green technology efficiency, and green technology progress, respectively. China’s AGTFP experienced an average growth rate of 7.84% from 2007 to 2019. During the period from 2007 to 2010, the growth rate reached a high of 9.69%, but it subsequently decreased. This is because changes in AGTFP are closely linked to AGTFP policy formulation and implementation. In 2007, China carried out its inaugural nationwide survey of pollution sources, which established the groundwork for the execution of agricultural and rural environmental conservation initiatives. A special conference on rural environmental protection was held in 2008, highlighting the need to prioritize rural environmental conservation as a crucial strategic element during the urbanization and industrialization process. In 2009, documentation pertaining to pollution mitigation methods for cattle and poultry production, as well as regulation for water discharges, were released for public input. Additionally, the implementation of satellite technology to monitor straw burning has begun. The rapid implementation of these agricultural environmental protection policies has led to a significant increase in GTFPC in the short term. However, there has been a delay in R&D, dissemination, and adoption of TIs, which has resulted in a decrease in the growth rate of GTFPC after 2010. The decomposition analysis reveals that GTEC experienced an average annual growth rate of −0.99% during the period of 2007–2019, while GTC had a growth rate of 8.92%. Based on that, the growth of China’s AGTFP primarily originates from green technological progress.

The average annual growth rate of GTFPC is higher in the western region compared to the eastern and central regions, indicating differences in spatial distribution. Furthermore, compared to the primary grain-marketing region and the main grain-producing area, the average yearly growth rate of GTFPC is higher in the grain-balancing area. This tendency is explained by the relatively low level of agricultural green growth in the western provinces, especially in the large western region that contains the grain-balancing area. The adoption of cutting-edge agricultural technologies from the eastern region has led to a higher growth rate of AGTFP. In addition, the major grain-marketing region in the east and the eastern region (which are mostly on the southeast coast) have made tremendous strides in GTC and GTEC, which has contributed to a noteworthy increase in overall development. An average yearly growth rate of 0.7 percent was attributed to the region’s higher degree of economic development, as well as the population’s generally stronger environmental awareness and greater capacity for allocating, organizing, and managing green resources.

4.2. Benchmark Regression Results

Table 3. Impacts of ER on AGTFP.

	AGTFP
	REM (1)	FEM (2)	REM (3)	FEM (4)
ER	0.325 ***	0.428 ***	0.227 ***	0.274 ***
	(0.040)	(0.072)	(0.036)	(0.073)
INS			1.560 ***	4.817 *
			(0.440)	(2.717)
ADR			−0.964 ***	−0.663 ***
			(0.179)	(0.206)
AE			−0.040	−2.639
			(1.310)	(2.868)
MAC			0.383 **	0.636 *
			(0.160)	(0.321)
TRA			−0.029	−0.050 **
			(0.026)	(0.022)
Constant variable	−1.839 ***	−2.570 ***	−1.947 ***	−3.823 **
	(0.290)	(0.515)	(0.397)	(1.765)
Fixed Effects	NO	YES	NO	YES
Hausman Test	8.67		44.16
	[0.013]		[0.000]
Sample size	360	360	360	360
R2	0.178	0.178	0.375	0.421

Note: ***, **, and * represent 1%, 5%, and 10% significance levels, respectively; standard errors are in parentheses; p-values are in parentheses.

displays the regression outcomes of the general panel model, examining the impact of ER on AGTFP. Models (1) and (3) are random effect models (REMs), while models (2) and (4) are fixed effect models (FEMs). The Hausman test strongly rejects the initial hypothesis of using a REM. Thus, this paper primarily relies on the regression results of FEMs. In models (1) and (2), the introduction of ER without any control variables has a significant positive impact on the level of AGTFP. However, after introducing a series of control variables such as INS, ADR, AE, MAC, TRA, etc., the regression coefficient of ER on AGTFP decreases to 0.274. This decrease is statistically significant at the 1% level.

Based on the previous theoretical analysis, ER will result in the occurrence of “Compliance cost”. As a result of regulation pressure, the main entities responsible for agricultural management have started to adopt a passive approach toward addressing agricultural surface pollution at the production stage. This includes reducing the use of pesticides, agricultural film, and fertilizers to minimize undesirable agricultural outputs, resulting in the squeeze of TI funds and increased pollution prevention and control costs. However, the cost of the agricultural sector is not as significant as that of the industrial sector. The intensification of ER has led to the emergence of the “Innovation offset” effect, where agricultural entities actively embrace new technology, equipment, and industries. This promotes the industrialization, modernization, scale, and efficiency of agriculture, ultimately improving AGTFP. In summary, the impact of ER on AGTFP depends on the interplay between these two forces.

The benchmark regression results indicate that the impact of ER on “Innovation offset” is more significant than its impact on “Compliance cost”. This has a significant positive effect on China’s AGTFP, which is closely linked to the objectives outlined in Central Document No. 1. This is strongly linked to the focus on rural revitalization in Central Document No. 1 during the past few years, which highlights that AGTFP provides the means to expedite the industrialization of agriculture and establish a robust, modernized agricultural nation. With the deepening of the concept of green development in the agricultural sector, regions are enhancing their agricultural infrastructure and promoting techniques to reduce input quantities and increase efficiency under the pressure of ER, leading to an increase in AGTFP. The benchmark regression results validate H1, indicating that ER promotes AGTFP.

4.3. Endogeneity Treatment and Robustness Tests

4.3.1. Endogenous Processing

ER and AGTFP can be jointly influenced by a variety of factors, such as government policymaking and the level of economic development, and there may be reverse causality. Therefore, this paper draws on the data processing put forward by Peng et al. (2013) [45] by introducing two instrumental variables (IVs), i.e., lagging the core independent variables by one period (z1) and two periods (z2), respectively, so as to perform 2SLS to address the potential endogeneity problem.

As seen in

Table 4. Regression results of instrumental variables.

	ER	AGTFP	ER	AGTFP
	(1)	(2)	(3)	(4)
Z1	0.928 ***
	(0.020)
Z2			0.913 ***
			(0.025)
ER		0.249 ***		0.265 ***
		(0.027)		(0.032)
INS	0.068	1.233 ***	0.080	1.243 ***
	(0.173)	(0.179)	(0.216)	(0.180)
ADR	−0.113	−0.980 ***	0.246 *	−0.961 ***
	(0.110)	(0.150)	(0.144)	(0.174)
AE	0.453	1.916 ***	0.747 *	2.243 ***
	(0.326)	(0.459)	(0.443)	(0.481)
MAC	0.138 **	0.229 ***	0.268 ***	0.220 ***
	(0.063)	(0.062)	(0.079)	(0.065)
TRA	0.012	0.029	0.030 **	0.034
	(0.011)	(0.021)	(0.013)	(0.022)
Constant variable	0.397 **	−2.052 ***	0.347 **	−2.177 ***
	(0.183)	(0.249)	(0.233)	(0.281)
Fixed Effects	YES	YES	YES	YES
KP rk LM-statistic	105.599		94.885
LM p-value	0.000		0.000
KP rk wald F-statistic	2125.512		1338.657
Sample size	360	360	360	360
R2	0.178	0.178	0.375	0.421

Note: ***, **, and * represent 1%, 5%, and 10% significance levels, respectively; standard errors are in parentheses; p-values are in parentheses.

, the IVs are significant at the 1% level in model (1), indicating that the IVs are highly correlated with the endogenous variables. The p-value of the LM test statistic is 0.000, which rejects the original hypothesis and indicates that the selection of the IVs is reasonable. The KP-F statistic is much larger than the empirical statistic value of 10, indicating that weak instrumental variables do not exist. Thus, it can be considered that the “ER lagged by 1 and 2 periods” meets the prerequisites for selecting IVs. Further, when the two IVs are introduced into the model as core independent variables, they are positively correlated with the dependent variable at the 1% significance level, which confirms the robustness of the baseline regression results.

4.3.2. Robustness Test

Four approaches are selected for enhancing the benchmark regression’s robustness, as shown in

Table 5. Robustness test results.

	Replacing the Measurement of AGTFP	Replacing the Measurement of ER	Dynamic Effect	Classifying Types of ER
	(1)	(2)	(3)	(4)
ER	0.140 ***	0.098 ***	0.042 *
	(0.038)	(0.026)	(0.023)
ERC				0.139 ***
				(0.042)
ERM					0.102 *
					(0.051)
ERP						0.085 **
Ln.AGTFP			0.780 ***
			(0.072)			(0.032)
INS	2.661 *	1.662 ***	−0.680 **	3.710	5.420 *	4.769
	(1.423)	(0.487)	(0.309)	(2.846)	(2.775)	(4.043)
ADR	−0.409 ***	−1.080 ***	−0.188 *	−0.746 ***	−0.774 ***	−0.713 ***
	(0.131)	(0.200)	(0.104)	(0.215)	(0.218)	(0.212)
AE	−1.573	−0.417	0.447	−2.986	−3.718	−3.412
	(1.559)	(1.250)	(0.380)	(2.792)	(2.918)	(2.948)
MAC	0.331 *	0.377 *	0.051 *	0.619 *	0.665 *	0.741 *
	(0.181)	(0.200)	(0.065)	(0.337)	(0.356)	(0.366)
TRA	−0.017 *	−0.041	−0.013	−0.042 *	−0.044	−0.045 *
	(0.009)	(0.028)	(0.008) *	(0.023)	(0.026)	(0.025)
Constant variable	−1.979 *	−0.308	0.174	−1.962	−3.148 *	−2.321
	(0.981)	(0.354)	(0.159)	(1.570)	(1.657)	(1.780)
Fixed Effects	YES	YES	YES	YES	YES	YES
AR(1)			0.017
AR(2)			0.775
Hansen			0.827
Sample size	360	360	360	360	360	360
R2	0.408	0.331		0.402	0.372	0.378

Note: ***, **, and * represent 1%, 5%, and 10% significance levels, respectively; standard errors are in parentheses; p-values are in parentheses.

.

Model (1) substitutes the SBM-GML method with the EBM-GML method for measuring AGTFP, which accounts for both radial and non-radial efficiency, providing a more comprehensive evaluation of performance. Model (2) employs Principal Component Analysis (PCA) instead of the entropy method to measure environmental regulation intensity, effectively addressing correlations between indicators and offering a more objective weighting approach. Model (3) incorporates the system GMM estimation method to construct a dynamic panel model, which controls for potential endogeneity and captures the dynamic effects of environmental regulation. Model (4) subdivided environmental regulatory instruments into ERC, ERM, and ERP and substituted their alternative environmental regulatory indices into the benchmark regression model to examine the direction of influence and significance of the core independent variables. The results of the above robustness tests indicate that the core conclusion that ER promotes AGTFP is robust, and the coefficients of the three types of regulations on AGTFP are decreasing in the order of significance.

4.3.3. Heterogeneity Analysis

This study further develops the heterogeneity analysis of the impact of ER on AGTFP. The regions are divided into eastern, center, and western, as well as the main grain-producing area, the main grain-marketing area, and the grain-balancing area, in order to carry out the sub-sample test, and the specific regression results are shown in

Table 6. Results of effect of ER on regional heterogeneity of AGTFP.

	AGTFP
	Eastern (1)	Central (2)	Western (3)	Main Grain-Producing Area (4)	Main Grain- Marketing Area (5)	Grain-Balancing Area (6)
ER	0.318 ***	0.419 **	0.079	0.336 ***	0.303 **	0.104
	(0.097)	(0.168)	(0.120)	(0.096)	(0.115)	(0.010)
INS	3.997	−6.039	5.916 *	3.169	0.930	6.245
	(2.970)	(4.627)	(2.880)	(4.553)	(3.944)	(5.141)
ADR	0.063	−0.697 **	−1.111 ***	−0.455 *	−0.079	−1.2873 ***
	(0.095)	(0.235)	(0.254)	(0.236)	(0.134)	(0.160)
AE	−8.835 *	0.679	−3.986	−1.166	−12.186 *	−2.592
	(4.115)	(3.595)	(3.826)	(3.939)	(6.159)	(3.016)
MAC	0.567	0.100	0.932	0.847	−0.224	0.594 **
	(0.401)	(0.513)	(0.895)	(0.633)	(0.458)	(0.248)
TRA	−0.052 **	−5.474 **	−1.791 ***	−0.853 **	−0.0436 **	−1.885 **
	(0.022)	(2.208)	(0.351)	(0.327)	(0.017)	(0.589)
Constant variable	−3.323	0.701	−2.938	−3.681	−1.009	−3.421
	(1.869)	(2.705)	(2.374)	(2.438)	(2.308)	(3.293)
Fixed Effects	YES	YES	YES	YES	YES	YES
Sample size	132	96	132	156	84	120
R2	0.558	0.412	0.627	0.419	0.554	0.589
Number of num	11	8	11	13	7	10

Note: ***, **, and * represent 1%, 5%, and 10% significance levels, respectively; standard errors are in parentheses; p-values are in parentheses.

.

ER passed the 5% significance test in both the eastern and central regions, with impact coefficients of 0.318 and 0.419, respectively. In contrast, the coefficients in the western region did not pass the significance test. ER in the main grain-producing and grain-marketing regions also passed the 5% significance test, with impact coefficients of 0.336 and 0.303. In contrast, the coefficients in the grain-balancing region did not pass the significance level test. It may be due to the fact that the main grain-producing areas are of great significance for maintaining China’s food security, and these regions are facing a more severe task of increasing grain production and improving efficiency, which brings tremendous pressure on resources and the environment. And to stabilize grain-producing and promote SAD, the main grain-producing areas are receiving more policy support. The main grain-marketing areas are mainly concentrated in the eastern coastal areas, which have highly educated and more environmentally conscious people, ensuring the implementation of ER. It can be seen that there are differences in the promotion effect of ER on AGTFP in different regions, confirming H2. Specifically, the eastern and central regions are better than the western region, possibly due to the higher level of economic development and TI. Those conditions accelerate the process of agricultural modernization and industrialization and create the same intensity and type of ER policy in the eastern region, which faces a better external environment, which has a more significant promotion effect on AGTFP.

4.4. Mechanism Analysis

4.4.1. Analysis of Mediating Effect of TI

According to the previous mechanism of action, ER can promote SAD by strengthening agricultural TI.

Table 7. Mechanistic test of ER on AGTFP.

Variables	AGTFP	TI	AGTDP
	(1)	(2)	(3)
ER	0.274 ***	1.094 **	0.235 ***
	(0.073)	(0.528)	(0.070)
TI			0.035 **
			(0.015)
INS	4.817 *	13.21	4.354
	(2.717)	(8.584)	(2.579)
ADR	−0.663 ***	−0.775	−0.636 ***
	(0.206)	(0.479)	(0.198)
AE	−2.639	−26.93 ***	−1.696
	(2.868)	(9.043)	(2.862)
MAC	0.636 *	1.852 *	0.571 *
	(0.321)	(1.068)	(0.309)
TRA	−0.050 **	−0.051	−0.048 **
	(0.022)	(0.064)	(0.021)
Constant variable	−3.823 **	−10.74 **	−3.447 **
	(1.765)	(5.017)	(1.680)
Fixed Effect	Yes	Yes	Yes
Sample size	360	360	360
R2	0.421	0.207	0.451
[95% Conf. Interval]	[0.065, 0.142]

Note: ***, **, and * represent 1%, 5%, and 10% significance levels, respectively; standard errors are in parentheses; p-values are in parentheses.

presents the estimation results of the path-conduction effect of technological progress in the impact of ER on AGTFP. Columns (1), (2), and (3) show the test results of models (6), (7), and (8), respectively. It can be found that the coefficient of ER on AGTFP is 0.274 (p < 0.01) when the TI variable is not introduced, i.e., for every 1% increase in ER, AGTFP will be increased by at least 0.274%. At the same time, it was proved that a test basis exists for the path-transferring effect of ER on AGTFP. The coefficients of the core independent variables in model (2) passed the 5% significance level test, i.e., ER significantly contributes to the increase in the level of agricultural TI. The results of model (3) further show that after introducing TI, ER still passes the 1% significance test, but the coefficient of influence changes from 0.274 to 0.235. This indicates that TI plays a partial mediating role in the relationship between ER and AGTFP.

4.4.2. Analysis of Threshold Effect of TI

(1)

Threshold Existence Test

In order to test whether there is a threshold effect of TI in the impact of ER on AGTFP, this paper takes ER as a threshold-dependent variable and TI as a threshold variable. It uses the Bootstrap method to determine the number of thresholds by repeated sampling 1000 times, and the test results are shown in

Table 8. Results of threshold test.

Threshold Variables	Threshold Model	F-Value	p-Value	10% Threshold Level	5% Threshold Level	1% Threshold Level
ER	Single	72.61	0.000	23.09	27.49	35.32
	Double	19.73	0.093	19.13	22.62	30.53
	Triple	19.58	0.793	49.05	54.87	65.64
ERC	Single	65.76	0.000	25.41	30.72	40.77
	Double	19.22	0.089	18.29	21.93	30.38
	Triple	21.59	0.659	44.75	55.75	61.93
ERM	Single	76.45	0.001	26.29	31.41	39.40
	Double	20.16	0.084	19.22	23.46	32.81
	Triple	17.87	0.744	43.40	49.97	63.09
ERP	Single	74.72	0.000	25.09	29.12	38.55
	Double	17.46	0.161	20.18	24.64	32.79
	Triple	21.09	0.744	53.24	59.81	72.17

. It can be seen that the p-value of the double threshold of ER, ERC, and ERM is less than 0.01, and the p-value of the triple threshold does not pass the significance test, so the threshold number is set to 2. Similarly, the threshold number of ERP is set to 1.

(2)

Threshold value estimation

Table 9. Results of threshold estimation.

Type of Regulation	Number of Thresholds	Threshold Value	95% Conf. Interval (Lower Limit)	95% Conf. Interval (Upper Limit)
ER	The First Threshold	0.530	0.500	0.540
	The Second Threshold	1.650	1.615	1.680
ERC	The First Threshold	0.120	0.100	0.130
	The Second Threshold	0.530	0.500	0.540
ERM	The First Threshold	0.530	0.500	0.540
	The Second Threshold	1.650	1.580	1.680
ERP	The First Threshold	0.530	0.500	0.540

displays the results of threshold estimation, the red dotted line represents the 5% critical value, while the blue line shows the likelihood ratio estimator. It reveals that the first and second thresholds for TI are identical for both ER and ERM, with values of 0.530 and 1.650. For ERC, the first and second thresholds for TI are 0.120 and 0.530, respectively. As for ERP, the first threshold for TI is 0.530, while the second threshold value is yet to be determined. To further assess the significance of these threshold values, this study utilizes the Likelihood Ratio (LR) to clearly illustrate the threshold values and confidence intervals of TI, as shown in Figure 2. The findings presented in the graph align with those in Table 8, thus confirming the accuracy of the threshold values.

(3)

Results and analysis of threshold regression

The results of the above threshold of TI and different types of regulations for the panel threshold model regression are shown in

Table 10. Regression of TI threshold for ER on AGTFP.

Variables		AGTFP
		(1)	(2)	(3)	(4)
ER	(TI ≤ 0.530)	0.170 ***
		(0.061)
	(0.530 < TI ≤ 1.650)	0.210 ***
		(0.061)
	(TI > 1.65)	0.235 ***
		(0.061)
ERC	(TI ≤ 0.120)		−0.007
			(0.041)
	(0.120 < TI ≤ 0.530)		0.050
			(0.034)
	(TI > 0.53)		0.108 ***
			(0.033)
ERM	(TI ≤ 0.530)			0.058
				(0.043)
	(0.530 < TI ≤ 1.650)			0.086 *
				(0.042)
	(TI > 1.65)			0.104 **
				(0.042)
ERP	(TI ≤ 0.530)				0.038
					(0.024)
	(TI > 0.53)				0.091 ***
					(0.026)
TI		0.018	0.033 **	0.024 *	0.036 **
		(0.013)	(0.014)	(0.013)	(0.014)
INS		3.230	2.683	3.611	3.578
		(2.213)	(2.443)	(2.209)	(2.551)
ADR		−0.463 ***	−0.472 ***	−0.515 ***	−0.486 ***
		(0.158)	(0.149)	(0.160)	(0.160)
AE		0.965	−0.337	0.438	−0.330
		(2.146)	(2.237)	(2.219)	(2.268)
MAC		0.594 **	0.567 **	0.603 **	0.673 **
		(0.253)	(0.255)	(0.262)	(0.253)
TRA		−0.038 *	−0.042 *	−0.032	−0.045 *
		(0.022)	(0.023)	(0.025)	(0.024)
Constant variable		−2.965 **	−1.535	−2.524 *	−2.047
		(1.307)	(1.253)	(1.278)	(1.383)
Fixed Effect		YES	YES	YES	YES
Sample size		360	360	360	360
R2		0.569	0.549	0.548	0.524

Note: ***, **, and * represent 1%, 5%, and 10% significance levels, respectively; standard errors are in parentheses; p-values are in parentheses.

. It could be seen from model (1) when the TI is lower than the first threshold value of 0.530, the coefficient of influence of ER on AGTFP is 0.170. With the increase in the level of TI, the first threshold value of 0.530 and the second threshold value of 1.650, the effect of ER on the promotion of AGTFP significantly increased; the impact coefficient is 0.210 and 0.235, respectively. When TI reaches a high threshold range, on the one hand, it contributes to the process of informatization, industrialization, scaling, and intensification in the agricultural field. It greatly improves the output and economic effect of the agricultural field. On the other hand, the spillover of TI and the diffusion effect of learning prompted agricultural producers to utilize new technologies, equipment, and processes actively. It makes the agricultural entities shift from mainly targeting the terminal treatment of pollution to focusing on the treatment of the source, i.e., through the use of more effective combining factors and more environment-friendly ways to improve resource utilization, land productivity, labor productivity, etc. Through the above paths, TI could increase desirable outputs and reduce undesirable outputs under regulations, thus vigorously promoting AGTFP.

Models (2)–(4) demonstrate the threshold impacts of the three types of regulations on TI. The impact coefficients of all three types of regulations exhibit a notably greater strength once the TI threshold is surpassed. Notably, the impact of ERC on AGTFP changes from negative to positive to significantly positive as the level of TI surpasses the first and second thresholds. Similarly, the impact coefficient of ERM and ER is significantly enhanced after surpassing the first and second thresholds of TI. Additionally, the impact of ERP on AGTFP has changed from non-significant to significantly promoted.

ERC is highly obligatory, requiring agricultural entities to adhere to laws and regulations and to address agricultural pollution. However, the limited dissemination of relevant knowledge in the initial stages has resulted in minimal impact and crowded R&D funds, thereby impeding the progress of sustainable development in agriculture. The “Innovation offset” impact of regulation can only be realized when TI reaches a specific level. ERP relies on the voluntary and independent involvement of the public. The advancements in technology and emergence of various monitoring channels motivate agricultural entities to adopt and implement green technology proactively. Through an analysis of ER and the threshold effect of TI across three different types of regulations, it was determined that TI plays a significant role in the threshold effect of ER on AGTFP. Furthermore, the characteristics of the threshold effect of TI differ among the various types of regulations, thus confirming H4.

According to the different TI threshold intervals faced by the three types of ER, this paper further divides the TI level of 30 provinces and cities in China into four stages in 2008 and 2019, and the results are shown in

Table 11. Types of TI and regional distribution.

Year	Low-Level Technological Innovation Regions (TI ≤ 0.120)	Low- to Medium-Level TI Region (0.120 < TI ≤ 0.530)	Medium- to High-Level TI Region (0.530 < TI ≤ 1.650)	High-Level TI Regions (TI > 1.650)
2008	Inner Mongolia, Ningxia, Qinghai, Hainan, Gansu, Guizhou, Jilin, Xinjiang	Jiangxi, Chongqing, Anhui, Shanxi, Guangxi, Yunnan, Heilongjiang, Fujian, Sichuan, Hebei, Henan, Tianjin, Liaoning	Hubei, Hunan, Shanghai, Zhejiang, Shaanxi, Guangdong, Shandong	Jiangsu, Beijing
2019		Qinghai, Ningxia, Jilin	Inner Mongolia, Hainan, Xinjiang, Shanxi, Chongqing, Heilongjiang, Liaoning, Tianjin, Gansu	Yunnan, Guizhou, Hebei, Guangxi, Shaanxi, Jiangxi, Fujian, Hubei, Hunan, Sichuan, Shanghai, Henan, Beijing, Anhui, Zhejiang, Guangdong, Shandong, Jiangsu

.

As shown in Figure 3, 21 provinces and cities, such as Inner Mongolia, Ningxia, Qinghai, etc., were located in the low-level TI region and the medium–low-level TI region in 2008. By the end of the sample period, all provinces and cities had passed the low-level TI region. The rest of the provinces and cities had entered the medium- to high-level TI region. The pattern of TI development had gradually spread from the east to the central and western parts of the country. This indicates that the eastern region produced noticeable technology spillover effects during the sample period, and the level of agricultural TI in China significantly improved, which created favorable conditions for the development of ER on AGTFP.

5. Conclusions and Recommendations

5.1. Research Conclusions

The conclusions of this study are as follows: First, China’s AGTFP experienced an average annual growth rate of 7.84% from 2007 to 2019, which can be attributed primarily to the progress made by green technology. Furthermore, the average annual growth rate in the western region surpassed that of the eastern and central regions. Additionally, the main grain-balancing area exhibited a higher growth rate compared to the main grain-producing area and the main grain-marketing area. ER plays a significant role in promoting green growth in agriculture in both the overall sample and the sub-sample of the Benchmark regression. Furthermore, ER contributes to AGTFP better in the east and center than in the western region and better in the main grain-producing and main marketing regions than in the balancing regions. During the analysis of the heterogeneity of regulation types, it was found that the ERC, ERM, and ERP all showed significant results on AGTFP, with decreasing influence coefficients. Technological innovation acts as a crucial threshold factor in the effects of ER on AGTFP. As TI moves from the low threshold range to the high threshold range, the promotion effect of ER on AGTFP increases significantly. Among them, there are double thresholds for ER, ERC, and ERM; the first threshold of ERC is low, and the coefficient of influence on AGTFP turns from negative to significantly positive after crossing the threshold; and there is only a single threshold for ERP, which prompts an upward jump of AGTFP after crossing the threshold value of 0.530.

5.2. Recommendations

The United Nations Sustainable Development Goals (SDGs) outline a range of global development targets, including the eradication of hunger, combating climate change, and promoting responsible production and consumption. Drawing on the findings of this study, this study proposes the following policy recommendations to advance the sustainable transformation of agriculture, and contribute to the achievement of these global sustainability goals through a Chinese approach:

First, we will focus on the new agricultural model of AGTFP growth. Considering the low contribution of green technology efficiency to the growth of AGTFP, it is possible to improve the efficiency of fertilizer, pesticide, and agricultural machinery use; agricultural businesses can enhance the application of agricultural technology through publicity campaigns, utilizing network platforms, and rural night schools. Furthermore, it is imperative to enhance the active surveillance of agricultural pollution, establish a systematic mechanism for preventing and predicting pollution, and effectively implement a system of penalties and incentives. This will encourage agricultural entities to prioritize the improvement of the overall AGTFP. In the present political climate of China, it is vital to include relevant indicators in the local performance evaluation to facilitate the transition of agricultural development methods and effectively advance China’s AGTFP. These measures not only enhance AGTFP but also play a pivotal role in supporting the achievement of global goals such as ending hunger and promoting responsible production and consumption.

Second, ER must be enforced in accordance with the specific circumstances of the region. In areas with significant agricultural surface pollution and carbon emissions, such as major grain-producing regions, it is essential to integrate ERC with ERM effectively. The government will utilize legal frameworks, regulations, and administrative directives to enhance the oversight of agricultural entities, compelling them to prioritize resource conservation and environmental protection throughout the agricultural production process. At the same time, in order to reduce the disadvantages of excessive government intervention brought about by strong regulations, it should be coordinated with ERM to stabilize the prices of agricultural products through acts such as taxes and subsidies and thus stimulate the incentives of agricultural entities to protect the environment and engage in food production. In regions where agricultural pollution is relatively light and human capital is relatively abundant, in addition to the implementation of the above two types of regulations, ERP should be given more space to play. Entities can make good use of the masses, self-media, and other public opinion supervision roles in creating a regulation system “government-market-economic subject”. The government effectively promotes the competitiveness of the agricultural entities to improve the competitiveness of the total welfare of the community and promote China’s agricultural development and sustainable transformation. These measures not only boost agricultural productivity but also play a crucial role in advancing global goals, such as eradicating hunger and promoting responsible production and consumption.

Third, enhance advancements in agricultural technology and facilitate the implementation of a novel approach to reduce emissions in the agricultural sector. In order to maximize the effect of ER on AGTFP, the government should take the following measures: encourage agricultural research and business organizations to develop high-efficiency agricultural technologies; improve the rate at which green technology advancements are adopted and implemented, while also enhancing land pollution control and resource recycling efforts; promote biological breeding and the modernization of agricultural machinery and equipment, as well as the widespread use of precision irrigation technology; utilize information technology to automate and optimize agricultural production processes; and establish a collaborative agricultural science and technology innovation system, while providing support for the construction of laboratories and demonstration areas.

5.3. Theoretical Contributions

(1)

Expanding the Theoretical Framework of AGTFP Analysis. The existing literature on agricultural total factor productivity (TFP) has largely neglected the undesirable outputs inherent in agricultural production, such as resource, energy, and environmental constraints. Additionally, the selection of undesirable outputs has often been narrow. Failure to fully account for environmental issues like agricultural non-point source pollution and carbon emissions during production may lead to biased assessments of agricultural performance. This paper addresses this gap by incorporating both non-point source pollution and carbon emissions as undesirable outputs. It constructs a unified theoretical framework that integrates resources, energy, environment, and agricultural economy. This approach offers distinct advantages in accurately assessing agricultural TFP under environmental regulation, while better reflecting the current state of China’s AGTFP.

(2)

Enriching Research on Environmental Governance and Agricultural Sustainability. Most existing studies have focused on the impact of environmental regulation on the broader economy and industrial sectors. However, the agricultural sector, as a fundamental industry, also deserves more attention. In the limited literature addressing environmental regulation and agricultural sustainability, few scholars have considered the heterogeneous effects of environmental regulations on AGTFP. This oversight limits the ability to evaluate the differential impacts of various regulatory approaches. This study seeks to fill this gap by exploring how environmental regulation influences AGTFP through the lens of technological innovation. It aims to provide a scientifically rigorous measure of AGTFP while exploring the interrelationships and mechanisms among environmental regulation, technological innovation, and AGTFP.

(3)

Constructing a Unified Analytical Framework for Environmental Regulation, Technological Innovation, and AGTFP. The existing literature typically examines the relationships among “environmental regulation—AGTFP”, “environmental regulation—technological innovation”, and “technological innovation—AGTFP” in isolation. Few studies have integrated all three variables into a single framework. This paper, therefore, adopts a technological innovation perspective to explore the interactions and impact mechanisms among environmental regulation, technological innovation, and AGTFP. It identifies technological innovation pathways within AGTFP under environmental regulation and, by considering the current state of regional technological innovation, discusses the innovation thresholds for different types of environmental regulation in promoting agricultural green development. This contribution offers a valuable addition to the existing research on environmental regulation and AGTFP and provides a more comprehensive research foundation for implementing environmental policies, fostering agricultural technological innovation, and advancing SAD in China.

5.4. Research Limitations

This study has several limitations. First, due to the availability of input–output indicators for agricultural production only at the provincial level, AGTFP can currently only be measured at this scale. As data availability improves, future work will allow for more granular measurements of AGTFP. Second, agriculture is a unique sector that both contributes to greenhouse gas emissions and serves as a significant carbon sink. However, comprehensive data on this aspect is currently lacking, which suggests that future research should aim to develop a more robust AGTFP evaluation framework. Third, agricultural production is closely linked to factors such as climate conditions and resource endowments. Future studies could focus on key food-producing regions to better understand these dynamics. Finally, environmental regulation tools in agriculture remain underdeveloped compared to the industrial sector. As carbon neutrality policies advance, agricultural regulations and the selection of relevant indicators will become more refined and comprehensive.