The Impact of Environmental Regulation on Cultivated Land Use Eco-Efficiency: Evidence from China

Abstract

In the context of tightening resource and environmental constraints, quantitative measurement and influencing factors of cultivated land use eco-efficiency (CLUE) have become hot topics in current academic research. Existing studies primarily focus on the influence of natural, social, and economic factors on CLUE but ignore exploring the impact mechanism and effect of environmental policies on CLUE. Therefore, this study aims to explore the impact of environmental regulations on CLUE. To achieve this objective, a super-efficiency slack-based measure (super-SBM) model is used to calculate the CLUE for 31 provinces (municipalities and autonomous regions) in China from 2000 to 2009. Additionally, the intermediary effect model and the threshold effect model are used to empirically investigate the transmission mechanism and nonlinear characteristics between environmental regulation and the CLUE. The results show that: (1) the temporal dynamics of CLUE exhibit a pattern of initial fluctuating decline followed by gradual growth in China as a whole and across its eastern, central, and western regions. (2) Environmental regulation has a significant negative impact on CLUE, and the effect exhibits a nonlinear characteristic of decreasing marginal effects. (3) Agricultural technological innovation and agricultural industrial structure play a mediating role between environmental regulation and CLUE, reducing the negative impact of environmental regulation on CLUE. This study provides some implications for formulating scientifically sound environmental policies to optimize land use and enhance resource utilization efficiency.

1. Introduction

As one of the main elements of agricultural production activities, cultivated land not only supports the most basic agricultural production function but also supports multiple ecological functions such as water conservation and biological habitat [1,2]. In recent years, with rapid urbanization and exponential population growth, the expansion of urban boundaries has led to a continuous decrease in cultivated land area. According to data from the National Bureau of Statistics, China’s cultivated land area decreased from 2.031 billion mu in 2009 to 2.023 billion mu in 2017, and the national cultivated land area in 2019 was 1.918 billion mu [3]. The proportion of cultivated land area has also declined significantly, with the proportion of cultivated land area to total land area decreasing from 19.62% in 2009 to 15.96% in 2019. This reduction in agricultural land implies a decrease in the planting area of crops, directly affecting the growth of grain production. In addition, China has a large population base but relatively scarce cultivated land resources. In order to increase grain output on limited cultivated land, cultivated land is intensively utilized throughout the year. Furthermore, improper use of fertilizers and pesticides changes the characteristics of water and land resources and causes certain effects on food security and the regional ecological environment [4].

Ecological civilization construction, as an extension of the concept of sustainable development, has become an important part of the agricultural economic development process in 21st-century China. The 18th National Congress of the Communist Party of China (CPC) proposed prioritizing ecological civilization construction and integrating it into all aspects and the entire process of economic, political, cultural, and social development. The 19th National Congress of the CPC provided a comprehensive summary and key deployment strategy for ecological civilization construction and environmental protection. Under the guidance of the requirements of ecological civilization construction and the concept of green agricultural development, the Chinese government issued a series of environmental regulations and policies, aiming to promote the coordinated development of land use and ecological environment protection [5]. However, the following questions remain unanswered: How effective has the development of cultivated land utilization been in China’s agricultural sector? Have environmental regulatory measures effectively promoted the CLUE in China? Therefore, a systematic evaluation of the development status of CLUE and implementation effectiveness of environmental regulations in the agricultural field is of great practical significance for sustainable land use and high-quality green agricultural development in China.

2. Literature Review and Objectives of This Paper

2.1. Literature Review

At present, research on cultivated land use efficiency is relatively abundant. The academic circle mainly conducts exploration and research studies on the selection of measurement methods, the analysis of temporal and spatial evolution, and the identification of influencing factors of cultivated land use efficiency. The measurement methods for cultivated land use efficiency mainly include the stochastic frontier function [6,7], data envelopment analysis (DEA) [8,9], and the Malmquist–Luenberger productivity index [10]. The stochastic frontier function method is a production function-based approach that evaluates land use efficiency by comparing the difference between actual output and optimal output. The DEA method is a linear programming technique that assesses efficiency levels by comparing the input–output ratios of each unit [11]. The Malmquist–Luenberger productivity index measures dynamic changes in land use efficiency by comparing production functions at two different time points [12]. Among these methods, the DEA-SBM model based on traditional DEA is a commonly used approach for evaluating land use efficiency. The DEA-SBM model not only effectively addresses the slack problem of input–output variables but also incorporates the impact of unexpected output factors [13], making it widely applicable in efficiency evaluation. In terms of spatiotemporal evolution analysis, previous studies mainly analyze the spatiotemporal distribution pattern of cultivated land use efficiency in provinces [14], cities [15], and specific regions [16] and summarize the characteristics of cultivated land use efficiency at different spatial scales. Regarding the analysis of influencing factors, scholars have mainly examined them at two levels: micro-level factors, which include individual characteristics of peasant households [17], family characteristics [18], and farmland transfer behavior [19], and macro-level factors, which encompass natural conditions [20], the urbanization level [21], technological inputs [22], and agricultural transfer population [23]. With the deepening of ecological civilization construction and the concept of green development, scholars began to introduce ecology into the evaluation of cultivated land use efficiency. In view of the problems of carbon emissions and non-point source pollution in the process of cultivated land use, some scholars regard CLUE as a dynamic process of “input + expected output + undesired output” [24]. Most of the relevant literature regards carbon emissions and non-point source pollution as unexpected outputs and uses the ecological footprint method [25], DEA [26,27], and SBM model [28,29] to measure CLUE. Some scholars also focused on the analysis of the impact mechanism underlying CLUE and systematically analyzed the internal mechanism of natural, social, economic, and other factors [30]. Among them, urbanization [31] and rural labor transfer [32] are key factors affecting CLUE.

As a direct means to solve the irrational use of input variables and negative environmental externalities, environmental regulation is one of the important factors affecting the utilization of cultivated land resources. Currently, there is a lack of research literature that directly examines the impact of environmental regulation on the utilization of cultivated land resources. Instead, environmental regulations are considered as one of the factors influencing the efficiency of cultivated land utilization. There is no consensus in the existing literature regarding the impact of environmental regulations on the efficiency of land use. One group of studies argues that environmental regulations may inhibit the improvement of CLUE. Liu et al. [33] found that environmental regulations have a significant negative impact on the ecological efficiency of cultivated land use. Kuang et al. [34] believed that regulatory measures in the form of financial subsidies may lead to a dependence of land use entities on government support, lacking the intrinsic motivation to improve production capacity and enhance the efficiency of green cultivated land use. However, there are also studies suggesting that environmental regulation may promote the improvement of CLUE. Lv et al. [35] proposed that government strengthening of land pollution prevention and control with environmental regulations can significantly enhance the efficiency of green cultivated land use. Liu and Li [36] pointed out that in provinces where land use efficiency is at a strong or moderate level, the positive incentive effect of environmental regulations gradually weakens. The third viewpoint is that the impact of environmental regulations on CLUE is not significant. Gao and Li [37] conducted research using household surveys in Shaanxi Province, and the results showed that incentive-based environmental regulations had no significant effect on land use efficiency. Huang and Wang [38] obtained similar conclusions when studying the ecological efficiency of cultivated land use in Jiangxi Province. In addition, some scholars are concerned about the impact of a specific policy on cultivated land utilization, such as the impact of land tax and fee policies on the effect of cultivated land protection [39], the impact of heterogeneous policy tools on the amount of cultivated land [40], etc.

The above relevant research has laid a solid theoretical foundation for this paper, but there is still room for further research on three aspects. First, most studies that assess CLUE focus solely on carbon emissions in the process of cultivated land use while ignoring the crucial function of cultivated land in sequestering carbon. Second, the implementation of environmental regulation directly affects CLUE, but few studies integrate the two concepts into a single analytical framework. Third, most studies use a single linear regression model to examine the relationship between environmental regulation and cultivated land use but fail to test the intermediary transmission mechanism and nonlinear characteristics between them. Based on this, the innovation of this study is that it explores the impact of environmental regulation on CLUE for the first time. The main contributions of this paper are as follows: (1) considering carbon sinks in the process of cultivated land use, the SSBM model is used to measure CLUE, which is helpful in understanding the development of cultivated land use in China and (2) the impact of environmental regulation on CLUE is analyzed, which fills the research gap in this field and provides a new perspective for further research.

2.2. Objectives of This Paper

This paper aims to address the aforementioned research gaps. This paper analyzes the relationship between environmental regulation and CLUE from both theoretical and empirical perspectives. The objectives are as follows:

• To reveal the impact mechanism of environmental regulation on the CLUE from a theoretical perspective.
• To measure and evaluate the CLUE in China, taking into full consideration the carbon sequestration function of farmland, and analyze its development trend.
• To empirically study the multidimensional relationship between environmental regulations and CLUE.

This article aims to achieve the above-mentioned goals using four key steps. The first step explains the direct impact, indirect impact, and threshold effect of environmental regulation on CLUE and proposes research hypotheses. The second step measures the CLUE using a super-efficiency SBM model based on the construction of an evaluation index system. The third step tests the research hypotheses using the basic model, intermediate model, and threshold model simultaneously. The fourth step analyzes and discusses the empirical results. The research process used in this study is demonstrated in Figure 1.

3. Theoretical Analysis and Research Hypotheses

3.1. Direct Effects of Environmental Regulation on CLUE

Existing studies have focused on the direct effects of environmental regulation on CLUE from both static and dynamic perspectives, but there is no consensus. In a static setting, some scholars think that implementing environmental regulation measures may increase expenses associated with pollution control and squeeze out productive investment, thus inhibiting the enhancement of CLUE. This is commonly referred to as the “cost loss” effect [41]. The “cost loss” effect of environmental regulation is mainly embodied in two aspects. On one hand, intensified enforcement of environmental regulations will increase environmental management costs for farmers, including the costs of agricultural waste disposal, clean production operation, and ecological restoration of farmland. This can, in turn, squeeze out other costs associated with land use, such as pesticides, fertilizers, and human capital training [42], thereby reducing the ecological efficiency of cultivated land use in the short term. On the other hand, the intensification of environmental regulations and technical standards may result in the mandatory suspension of ongoing agricultural investment projects [43]. This may inevitably cause certain economic losses to farmers, increase the sunk costs of agricultural production, and subsequently have a negative impact on their profit margins.

From a dynamic perspective, some scholars believe that environmental regulation may stimulate technological innovation and scientific management activities, thereby optimizing resource allocation, enhancing the quality and profitability of agricultural products, and ultimately improving CLUE. This is known as the “benefit compensation” effect [44]. There are also two sides to the “income compensation” effect of environmental regulation. In the short term, when the intensity of environmental regulation is high, farmers tend to use advanced green production technologies to improve the efficiency of resource utilization, reduce pollution emissions, and enhance the added value of products [45]. This leads to an increase in supernormal profits, which offsets the negative impacts of environmental governance costs and production costs. As environmental regulatory policies release clear market signals [46], they can prompt farmers to adjust their production methods, management approaches, or product structures, thus improving the competitive advantage and economic benefits of agricultural products.

The above analysis of the impact mechanism of environmental regulation on CLUE indicates the interaction between the “cost loss” effect and the “benefit compensation” effect determines whether environmental regulation will promote CLUE. When the “cost loss” effect outweighs the “benefit compensation” effect, environmental regulation will increase CLUE. Conversely, when the “benefit compensation” effect surpasses the “cost loss” effect, environmental regulation will hinder or even undermine CLUE. Thus, the first hypothesis is proposed as follows:

H1: 

The impact of environmental regulation on CLUE is uncertain, and its impact direction depends on the interaction between the “Cost loss” effect and the “benefit compensation” effect.

3.2. Indirect Effects of Environmental Regulation on CLUE

Additionally, environmental regulation may indirectly affect CLUE through two paths: agricultural industrial structure and agricultural technology innovation.

Environmental regulation can adjust agricultural production sectors or projects with regulations and standards constraints, improve industrial specialization and division of labor, promote the optimization and upgrading of agricultural industrial structure, and indirectly affect the output efficiency and ecological quality of cultivated land. First, the scientific design of environmental regulation standards is conducive to the mandatory “screening” of regional agricultural projects and agricultural operators. This screening process drives the “green” adjustment of agricultural industrial structures, leading to optimizing the structure and proportion of agricultural sectors. Second, environmental regulation stimulates the innovation of production technology and the diffusion of green innovative technologies. These advancements effectively foster the refined division of labor and specialized management within the agricultural sector, the innovation of production technology and equipment, and the diffusion of green, achieving the continuous development of the agricultural industrial structure in the upgrading direction.

Environmental regulation can optimize the combination of factors input by stimulating technological innovation behavior among stakeholders involved in cultivated land utilization and unlock the ecological potential of cultivated land resources. First, local governments are inclined to enhance agricultural research funding as they concurrently implement environmental regulatory policies, aiming to maximize economic benefits and social welfare. This proactive approach can promote technological innovation, which serves to mitigate the negative externalities associated with resource utilization and environmental pollution [47]. Second, strict environmental standards and high-tech and clean agriculture industries have comparative advantages in promoting green development, which helps reduce their dependence on resource elements. Therefore, farmers are more likely to adopt green production technology under the constraint of environmental regulation, with the aim of improving product quality and product competitiveness [48]. Thus, the second hypothesis is proposed as follows:

H2: 

Environmental regulation indirectly affects CLUE through agricultural industrial structure and agricultural technological innovation.

3.3. Threshold Effects of Environmental Regulation on CLUE

There may be a nonlinear relationship between environmental regulation and CLUE, which is mainly related to self-changing characteristics in the process of implementing environmental regulation policy. In the early stage of implementing environmental regulation policy, various environmental protection objectives are incorporated into the political assessment of local governments, with mandatory punishment measures and restraint measures often being the preferred choices of local officials [49]. During this period, the cost for farmers to obtain green technology information in order to meet environmental standards is high, so the negative impact of environmental regulations on CLUE is relatively obvious in the short term. As environmental regulation continues to improve and relevant environmental subsidies are implemented, the costs associated with developing and acquiring green agricultural production technologies will decrease. This provides a powerful financial guarantee for farmers to strive for more innovative resources. Therefore, environmental regulation will encourage farmers to apply green production technology in the long run, thereby improving agricultural production efficiency and reducing the negative impact of the marginal cost of pollution control.

In addition, due to China’s unbalanced regional development strategies and corresponding regional tilt policies, there exist significant variations in regional agricultural structures and levels of agricultural technology innovation. Therefore, environmental regulation may have heterogeneous effects on CLUE across different agricultural industrial structures and levels of agricultural technological innovation. Therefore, the third hypothesis is proposed as follows:

H3: 

The impact of environmental regulation on CLUE exhibits a non-linear pattern, which is regulated by the prevailing agricultural industrial structure and the level of agricultural technology innovation.

4. Materials and Methods

4.1. Indicators and Variable Selection

4.1.1. Explained Variable

CLUE aims to realize the largest possible economic and social value and the smallest possible environmental loss with the least consumption of resources. It embodies the harmonious development of resource elements, social economy, and ecological environment. Based on this implication, this paper establishes a CLUE indicator system drawing from previous studies [17,50] that encompasses three dimensions: input, expected output, and non-expected output (see

Table 1. The index system for CLUE.

System	Sub-System	Index	Index Meaning
CLUE	Input	Labor	Number of agricultural labors × (agricultural output value/total output value of agriculture, forestry, animal husbandry, and fisheries)/10,000
		Land	Grain sown area (1000 hm2)
		Capital	Total power of agricultural machinery (10,000 kW)
			Net amount of chemical fertilizer consumed (10,000 t)
			Effective irrigation area (1000 hm2)
			Amount of pesticide usage (10,000 t)
			Amount of agricultural film used (10,000 t)
	Expected output	Economic benefit	Total output value of agriculture (CNY 100 million)
		Social benefit	Total output of grain (10,000 t)
		Environmental benefit	Net carbon sink (10,000 t)
	Unexpected output	Agricultural non-point source pollution emissions	The amount of TN and TP emitted (10,000 t)

below).

Considering the characteristics of agricultural production systems, the input indicators for CLUE should cover three aspects, namely, land, capital, and labor. Among them, the labor input, which involves three indicators, is calculated using a weighting coefficient method. First, the ratio of the agricultural output value to the total output value of agriculture, forestry, animal husbandry, and fisheries is calculated to determine the weighting coefficient. Then, this weighting coefficient is multiplied by the rural population quantity to calculate the final labor input. The land input it is represented by the area of grain sown, thus reflecting ‘’land element transformation’’ in the process of cultivated land green utilization. Capital input generally refers to the actual utilization of physical capital commonly applied in the process of cultivated land utilization, including the total power of agricultural machinery, net amount of chemical fertilizer application, area of effective irrigation, use of pesticides, and amount of agricultural film application.

The expected output indicators reflect the final results of farmland production, management, and services over a specific period, including economic, social, and environmental benefits. Specifically, the total output value of agriculture is used to represent the economic benefit output, while the total output of grain represents the social benefit output. The net carbon sink is used to represent the environmental benefit output, which refers to the difference between carbon absorption and carbon emission in the process of cultivated land utilization. Among them, referring to the research of Tian and Zhang [51], the carbon absorption of various food crops on cultivated land is taken into consideration during the calculation of carbon sequestration. The estimation of carbon emissions primarily involves quantifying carbon emissions from various carbon sources during the process of cultivated land use, which is performed using estimation formulas and calculation coefficients referenced by Ke et al. [52].

Agricultural non-point source pollution is selected as a representative indicator for measuring unexpected output. Based on the research of Yang et al. [53], the loss of total nitrogen (TN) and total phosphorus (TP) resulting from nitrogen, phosphorus, and compound fertilizers is utilized to characterize the pollutant emissions during cultivated land use. Additionally, a unit survey and evaluation method is used for the calculations.

4.1.2. Core Explanatory Variable

In practice, there are diverse types of environmental regulatory tools. Specifically for China’s agricultural sector, the command-and-control type of regulation, mainly based on laws and regulations, occupies a dominant position [54]. Therefore, this paper uses a number of policies related to cultivated land conservation and pollution prevention and control in farmland use from a policy perspective. To account for the time lag effect of public policies, the cumulative number of agricultural-related environmental policies implemented by provinces at the end of the year is used to quantify environmental regulation (ER).

4.1.3. Intermediary Variables

According to the theoretical analysis above, environmental regulation can affect the efficiency of cultivated land use by affecting agricultural industrial structure and agricultural technology innovation. Agricultural industrial structure (AIS) is expressed as the ratio of forestry, animal husbandry, and fisheries output value to the total output value of agriculture, forestry, animal husbandry, and fisheries. The level of agricultural technology innovation (ATI) is measured by the total number of patents in agriculture, forestry, animal husbandry, and fisheries.

4.1.4. Control Variables

Referring to the research of Li [55] and Dhehibi et al. [56], the following indicators are selected as control variables in this study in order to prevent endogenous problems caused by missing variables. The first control variable is disaster degree (DIS), expressed as the ratio of the disaster-affected area of crops to the sown area of crops. The second is the multiple cropping index (MIC), evaluated as the percentage of the total sown area of crops to the cultivated area. The third is the economic development level (GDP), measured as the regional GDP. The fourth is farmer living standard (FLS), assessed as the per capita disposable income of rural residents. The last is the urbanization rate (URI), characterized as a proportion of the urban population in a region’s total population.

4.2. Research Methodology

4.2.1. Super-Efficient Slack Based Model (SSBM)

Data envelopment analysis (DEA), a commonly used method for evaluating land use efficiency, can be applied to analyze the relative efficiency of decision-making units with multiple inputs and outputs. However, the adjustment of input and output relaxation is not effectively addressed, which can lead to measurement deviations in efficiency. To overcome this limitation, Tone [13] proposed a non-radial, non-parametric SBM model based on slack variables. This model directly incorporates relaxation variables into the objective function, allowing for different proportional changes in input and output variables, thereby avoiding errors caused by the selection of slack variables and slack angles. However, regarding the measurement results of an SBM model, there may be situations where the efficiency values of multiple decision-making units are all equal to 1, making it difficult to conduct comparative analyses. In comparison with the SBM model, the super-efficiency SBM (SSBM) model effectively addresses the slackness of input and output variables and the comparison problem when multiple decision-making units are simultaneously effective, so it can more effectively estimate the super-efficiency value of decision-making units. Therefore, this paper uses the SSBM model to measure the ecological efficiency of cultivated land use. The specific model is as follows:

$$\begin{array}{l}\min \rho =\left[1+\frac{1}{m}{\displaystyle \sum _{i=1}^m\frac{s_i^{-}}{x_{ik}}}\right]/\left[1+\frac{1}{m}{\displaystyle \sum _{r=1}^m\frac{s_r^{+}}{y_{rk}}}\right]\\ s_{.}{t}_{.}{\displaystyle \sum _{j=1,j\ne k}^n{x}_{ij}}{\lambda }_j-s_i^{-}\le x_{ik}\\ {\displaystyle \sum _{j=1,j\ne k}^n{y}_{rj}}{\lambda }_j+s_r^{+}\le y_{rk}\\ s_i^{-},s_r^{+},{\lambda }_j\ge 0\end{array}$$

(1)

In the formula, $\rho$ is the CLUE value of the evaluation unit; $x$ and represents the input variable; $y$ represents the output variable; ${\lambda }_{j }$ is the weight vector; and ${ s}_i^{-} \mathrm{a}\mathrm{n}\mathrm{d} s_r^{+}$ represent the relaxation vector, respectively.

4.2.2. Benchmark Regression Model

In order to test the direct impact of environmental regulation on CLUE, a benchmark regression model is constructed using the Equation (2):

$$CLU{E}_{i,t}={\partial }_0+cE{R}_{i,t}+{\partial }_{3}contro{l}_{i,t}+{\mu }_i+v_t+{\epsilon }_{i,t}$$

(2)

where $i$ indicates a province in China; $t$ represents the year; $CLU{E}_{i,t}$ indicates the level of cultivated land use eco-efficiency; $E{R}_{i,t}$ denotes environmental regulation; $contro{l}_{i,t}$ includes a series of control variables; ${\partial }_0$ is a constant term; $c$ is the regression coefficient of $E{R}_{i,t}$ on $CLU{E}_{i,t}$; ${\partial }_3$ is the constant term coefficient of $contro{l}_{i,t}$; ${\mu }_i$ is the regional fixed effect; $v_t$ is the time-fixed effect, and ${\epsilon }_{i,t}$ are random disturbance terms.

4.2.3. Intermediary Effect Model

In addition to the total effect calculated using Equation (2), environmental regulation may have an indirect effect on CLUE through two intermediary mechanisms, i.e., agricultural industrial structure and agricultural technology innovation. Accordingly, the mediating effect model proposed by Wen and Ye [57] is used in this paper for empirical analyses. The model is developed as follows:

$$M{E}_{i,t}={\beta }_0+aE{R}_{i,t}+{\eta }_{4}contro{l}_{i,t}+{\mu }_i+v_t+{\epsilon }_{i,t}$$

(3)

$$CLU{E}_{i,t}=a_0+c^{\prime }E{R}_{i,t}+bM{E}_{i,t}+{\eta }_{3}contro{l}_{i,t}+{\mu }_i+v_t+{\epsilon }_{i,t}$$

(4)

Equation (3) represents the test of environmental regulation against the mediating variables, where $M{E}_{i,t}$ is a set of intermediary variables that represents AIS and ATI and $a$ is the regression coefficient of $E{R}_{i,t}$ on $M{E}_{i,t}$. Equation (4) represents the test of environmental regulation and the mediating variables against CLUE, where $c^{\prime }$ is the regression coefficient of $E{R}_{i,t}$ on $CLU{E}_{i,t}$ and $b$ is the regression coefficient of $E{R}_{i,t}$ on $CLU{E}_{i,t}$. The other variables are defined in the same way as in Equation (2).

4.2.4. Threshold Regression Model

To test for a linear or nonlinear relationship between environmental regulation and CLUE, this paper selects ER, AIS, and ATI as threshold variables and uses Hansen’s threshold model for the estimation test [58]. The threshold model is as follows:

$$CLU{E}_{i,t}={\phi }_0+{\phi }_{1}E{R}_{i,t}\times I(k{j}_{i,t}\le {\theta }_1)+{\phi }_{2}E{R}_{i,t}\times I({\theta }_1<k{j}_{i,t}\le {\theta }_2)+{\phi }_{3}E{R}_{i,t}\times I(k{j}_{i,t}>{\theta }_2)+{\phi }_{4}contro{l}_{i,t}+{\mu }_i+v_t+{\epsilon }_{i,t}$$

(5)

where $k{j}_{i,t}$ denotes the threshold variable, including ER, AIS, and ATI. ${\theta }_1$ and ${\theta }_2$ are the two threshold values. ${\phi }_1$, ${\phi }_2$, and ${\phi }_3$ are the coefficients of the impact of the environmental regulation on CLUE when the threshold variables are in different intervals. $I$ is the indicative function. The other variables are the same as the Equation (2).

After obtaining estimated values, statistical tests can be performed to examine whether the estimated parameters of the two groups of samples, divided by a threshold value, are significantly different. Therefore, the null hypothesis without a threshold value is H0: H1 = H2, and an LM statistic is constructed:

$$F=n\frac{S_0-S_n\left(\tau \right)}{S_n\left(\tau \right)}$$

(6)

where $S_0$ is under the null hypothesis. Since the LM statistic does not follow the standard chi-square distribution, Hansen [58] proposed using the bootstrap method to obtain the asymptotic distribution and calculate the corresponding probability p-value, also known as the bootstrap p-value. The method includes simulating a set of dependent variable sequences and calculating the simulated LM statistic to estimate the probability of the LM test. By repeating this process multiple times, the proportion of the number of times the LM statistic is greater than a given threshold to the total number of simulations is obtained, which is the bootstrap p-value.

4.3. Data Resources and Descriptive Statistics

In this study, 31 Chinese provinces (municipalities and autonomous regions, excluding Hong Kong, Macau, and Taiwan) were selected as the research object. Data from 2000 to 2019 were acquired from the “China Statistical Yearbook’’, “China Rural Statistical Yearbook’’, “China Environmental Yearbook’’, and statistical yearbooks for the provinces (municipalities and autonomous regions). The stock of patented technology in agricultural production of each province is represented by the total number of valid agricultural patents at the time point, which is from the PatSnap patent database.

Table 2. Summary statistics for all variables.

Variable	Obs	Mean	Std. Dev.	Min	Max
lnCLUE	620	−0.0457	0.2668	−0.8370	1.0523
lnER	620	7.3738	1.5226	1.0986	10.7175
AIS	620	0.4761	0.0873	0.2167	0.6983
lnATI	620	5.9057	1.7819	0	11.8056
DIS	620	0.2312	0.1606	0	0.9356
MIC	620	1.4266	1.0810	0.4146	25.3574
lnGDP	620	8.9741	1.2557	4.7661	11.5868
FLS	620	2.9011	0.6239	1.85	5.61
URI	620	0.4992	0.1578	0.1447	0.9349

shows the descriptive statistics for each variable. The logarithm of the average ecological efficiency in agricultural land utilization is −0.0457, with a minimum value of −0.8370 and a maximum value of 1.0523. The logarithm of environmental regulatory policies has an average value of 7.3738 with a standard deviation of 1.5226, indicating significant variations in environmental regulatory policies among different provinces. The average value of agricultural industrial structure is 0.4761, ranging from a minimum of 0.2167 to a maximum of 0.6983, exhibiting a difference of approximately threefold. The logarithm of agricultural technology innovation has an average value of 5.9057 with a standard deviation of 1.5226 and a further standard deviation of 1.7819, suggesting uneven development of agricultural technology across provinces.

5. Results

5.1. The Results of CLUE

This paper uses the SSBM model to measure CLUE in China’s provinces from 2000 to 2019. Figure 2 illustrates the changing trend in CLUE in China while also showing the changing trend and distribution characteristics of CLUE among the three regions of China: the east, the middle, and the west.

At the national level, the average CLUE in China decreased from 1.051 in 2001 to 0.999 in 2019, with an average annual growth rate of −0.262%. In terms of regions, the average level of CLUE in the eastern region is the highest, reaching 1.049, with an average change rate of −0.366%. The second is the western region with an average CLUE of 0.991 and an average change rate of −0.495%. The central region is the lowest, having an average CLUE of 0.915, with an average change rate of 0.185%. As can be seen in Figure 1, taking 2011 as the dividing point, the CLUE in China and the three regions displays a pattern of initial fluctuations followed by a gradual increase. This suggests that relevant policies and measures, such as the construction of “two-oriented agriculture” and the prevention and control of agricultural non-point source pollution advocated by the state during the “12th Five Year Plan” period, as well as the implementation of zero growth action in fertilizer and pesticide use during the “13th Five-Year Plan” period, have played a positive role in improving CLUE.

5.2. Benchmark Regression Results

The estimated results of environmental regulations on CLUE are shown in

Table 3. The regression results of the baseline model.

Variable	(1) Pooled OLS Model	(2) Random-Effects Model	(3) Fixed-Effects Model
lnER	−0.0063 **	−0.0085 ***	−0.0084 ***
	(0.0025)	(0.0024)	(0.0024)
DIS	−0.0055	−0.0067 *	−0.0065 *
	(0.0039)	(0.0038)	(0.0038)
MIC	0.0359	0.0137	0.0146
	(0.0266)	(0.0254)	(0.0261)
lnGDP	−0.0820	−0.0199 *	−0.0139 *
	(0.0115)	(0.0112)	(0.0117)
FLS	0.2129 ***	0.0148 **	0.0333 ***
	(0.0731)	(0.0148)	(0.0539)
URI	0.1031 ***	0.1350 ***	0.1166 ***
	(0.0390)	(0.0373)	(0.0385)
provincial fixed effects	No	No	YES
time fixed effects	No	No	YES
constants	−0.2948	0.2623 *	0.1753 *
	(0.3859)	(0.1588)	(0.1598)
R²	0.8346	0.3412	0.3486
N	620	620	620

Note: Standard errors are reported in parentheses. *, **, and *** represent significance at 10%, 5%, and 1%, respectively.

. The results of the pooled OLS model, random-effects model, and fixed-effects model are presented in columns (1), (2), and (3), respectively. Since the Hausman test results indicate that the fixed-effects model is superior to the random-effects model, the fixed-effects model is selected to estimate parameters. The following analysis is conducted using this model.

As can be seen from Table 3, environmental regulation (ER) has a significant negative impact on CLUE, which means that the “cost loss” effect of environmental regulation on CLUE is greater than the “benefit compensation” effect. On the one hand, this is mainly due to the fact that under the guidance of the concept of green development, local governments have increased investment in ecological environment protection and governance, leading to an increase in governance costs [59]. At the same time, farmers are required to carry out agricultural production activities in accordance with relevant legal requirements and technical standards, resulting in an increase in their agricultural production costs. On the other hand, the “benefit compensation” effect of environmental regulation is relatively small, and the negative impact of governance costs and “compliance costs” has not been eliminated. Environmental regulations suppress the improvement of CLUE, thus validating the existence of the “cost loss” effect proposed in hypothesis 1.

Among the control variables, disaster degree (DIS) has a significantly negative effect on CLUE. The occurrence of natural disasters reduces the yield and quality of grain, cause damage to the ecological environment of cultivated land, thereby reducing the ecological efficiency of cultivated land utilization. The multiple cropping index (MIC) makes a positive contribution to CLUE, but it did not passed the significance test. As the multiple cropping index increases, agricultural output and value also increases. However, this is accompanied by increased inputs of fertilizers, pesticides, and plastic film on the cultivated land, which further enhances the unintended outputs, leading to a less significant positive impact. The effect of economic development level (GDP) on CLUE is significantly negative. Behind rapid economic growth, there is a series of issues affecting the agricultural ecosystem, such as illegal occupation, pollution, and destruction. Farmer living standard (FLS) has a significant promotion effect on the improvement in CLUE. An increase in farmer income and an improvement in farmer living standards will enhance awareness of farmland resource protection and green and efficient production [60], thus encouraging farmers to improve agricultural production conditions and adopt environmentally friendly technologies to improve the ecological efficiency of farmland utilization. The urbanization rate (URI) has a significant positive effect on CLUE. Urbanization plays a crucial role in promoting large-scale and organized agricultural operations by effectively allocating production factors, thereby reducing production costs. Moreover, the labor flow formed by urbanization facilitates the accumulation and deepening of agricultural capital, leading to improved agricultural production conditions [61].

5.3. Robustness Test

The above regressions suggest that environmental regulation can inhibit the improvement of CLUE, but is this finding robust? To address this question, this paper uses three methods to test the robustness of the empirical conclusions.

The first robustness test changes the measurement indicator of environmental regulation. Referring to previous relevant studies, this paper uses the total investment in environmental pollution control instead of the number of agricultural environmental protection policies mentioned above. As is shown in column (1) of

Table 4. The results of the robustness tests.

Variable	(1)	(2)	(3)
	Replace Core Explanatory Variable	Exclude the Sample of Provincial Capital Cities	SYS-GMM
L.lnCLUE	-	-	0.4597 ***
			(0.0649)
lnER	−0.0078 **	−0.0102 ***	−0.0201 ***
	(0.0024)	(0.0027)	(0.0076)
DIS	−0.1681	−0.1708 ***	−0.1414 ***
	(0.0409)	(0. 0441)	(0.0076)
MIC	0.0020	0.0021	0.0009
	(0.0048)	(0.0050)	(0.0126)
lnGDP	−0.0301	−0.0156 *	−0.0512 **
	(0.0124)	(0.0112)	(0.0117)
FLS	0.0247 ***	0.0400 ***	−0.0107
	(0.0160)	(0.0176)	(0.0207)
URI	0.3307 ***	0.2786 ***	0.5517 ***
	(0.0990)	(0.1068)	(0.2109)
provincial fixed effects	YES	YES	-
time fixed effects	YES	YES	-
constants	0.0805	−0.0581 *	0.1355
	(0.1135)	(0.1300)	(0.1220)
R²	0.1885	0.1279
N	620	620	620
AR(1)	-	-	0.0050
AR(2)	-	-	0.1950
Hansen			0.9980

Note: Standard errors are reported in parentheses. *, **, and *** represent significance at 10%, 5%, and 1%, respectively.

, environmental regulation has a significant negative effect on CLUE, indicating that the conclusion of this paper is robust.

The second robustness test excludes the sample of provincial capital cities. Environmental regulations at different government levels may have great differences, resulting in substantial deviations in their role to promote the green use of cultivated land. Therefore, samples from Beijing, Tianjin, Shanghai, and Chongqing were discarded from this study, and the remaining 27 provinces were maintained for statistical analysis. The results are shown in column (2) of Table 4. After the introduction of control variables and using fixed effects, the estimated coefficient for the environmental regulations level remains significantly negative at the 1% statistical level, and the research results are the same as above.

The third robustness test applies SYS-GMM to estimate the model in this study. The specific inspection results are shown in column (3) of Table 4. It can be seen that the negative effect of environmental regulation on CLUE still exists, which is consistent with the previous estimates. Thus, the results of the study are strongly robust.

5.4. Intermediary Effect Regression Results

The basic regression results are presented in column (1) of

Table 5. The regression results of the intermediary effect model.

Variable	(1)	(2)	(3)	(4)	(5)
	lnCLUE	AIS	lnCLUE	lnATI	lnCLUE
lnER	−0.0084 ***	0.0039 **	−0.0080 ***	0.0511 **	0.0079 ***
	(0.0024 )	(0.0016)	(0.0024)	(0.0250)	(0.0024)
AIS	-	-	−0.0890 *	-	−0.0101 **
			(0.0643)		(0.0040)
lnATI	-	-	-	-	-
DIS	−0.0065 *	0.0030	−0.0062 *	0.0390	−0.0061 *
	(0.0038)	(0.0025)	(0.0038)	(0.0397)	(0.0038)
MIC	0.0146	−0.0033	−0.0143 ***	−0.4314	−0.0102 ***
	(0.0261)	(0.0168)	(0.0261)	(0.2704)	(0.0260)
lnGDP	−0.0139 *	0.0300 ***	−0.0112 *	1.7170 ***	0.0034
	(0.0117)	(0.0075)	(0.0118)	(0.1211)	(0.0135)
FLS	0.0333 ***	0.2670 ***	0.0574 ***	−5.7736 ***	−0.0248
	(0.0539)	(0.0347)	(0.0564)	(0.5585)	(0.0583)
URI	0.1166 ***	−0.0490 **	0.1122 ***	−1.4709 ***	0.1018 ***
	(0.0385)	(0.0248)	(0.0386)	(0.3994)	(0.0388)
constant	0.1753 *	−1.3674 ***	0.0524 *	−5.3946 ***	0.2093 ***
	(0.1598)	(−1.3674)	(0.1823)	(1.5160)	(0.1471)
R²	0.3486	0.3952	0.3455	0.7719	0.8259
obs	620	620	620	620	620
bootstrap testing	-	p = 0.0038		p = 0.0946
results	-	partial intermediary effect intermediary effect/total effect = 0.0413		partial intermediary effect intermediary effect/total effect = 0.0615

Note: Standard errors are reported in parentheses. *, **, and *** represent significance at 10%, 5%, and 1%, respectively.

. It can be seen that the impact coefficient $c$ of environmental regulation on CLUE is significantly negative. Column (2) shows that the impact coefficient $a$ of environmental regulation on agricultural industrial structure is significantly positive, indicating that environmental regulation significantly promotes the optimization of agricultural industrial structure. Column (3) shows that after adding the mediating variable (AIS), the impact coefficient of environmental regulation on CLUE decreases, but it is still significantly negative. The bootstrap test results indicate that the mediating effect of agricultural industry structure on the relationship between environmental regulation and CLUE is statistically significant at the 5% level. This suggests that the agricultural industry structure plays a significant positive mediating role between environmental regulation and CLUE with a proportion of 4.13%. In other words, agricultural industrial structure significantly mitigates the negative impact of environmental regulation on CLUE. The rising proportion of forestry, animal husbandry and fisheries indicates that the agricultural industrial structure has been dynamically upgraded from a low-income elastic industry (such as planting industry) to a high-income elastic industry. This dynamic upgrading process can promote the improvement of agricultural economic benefits [62], thus eliminating the negative impact of environmental regulation cost loss.

Column (4) shows that the impact of environmental regulation on CLUE is significantly positive at the 5% significance level. In column (5) of the regression results, both environmental regulation and agricultural technological innovation have significant impacts on CLUE. The bootstrap test further indicates that the intermediary path of agricultural technology innovation is significant at the 10% level of significance, suggesting that a mediating effect of agricultural technology innovation exists, which is partly mediated. The mediating effect accounts for 6.15% of the total effect. This finding implies that although environmental regulation may inhibit the improvement of CLUE, this inhibitory effect is weakened by the influence of agricultural technological innovation. This may be because agricultural technology innovation enhances the “income compensation” effect of environmental regulation by optimizing the allocation efficiency of cultivated land production factors and the consumption of agricultural intermediate inputs and improving agricultural production efficiency. Therefore, agricultural industry structure and technological innovation in agriculture play a mediating role between environmental regulation and CLUE, confirming hypothesis 2.

5.5. Threshold Effect Regression Results

Before conducting the threshold model regression analysis, we used the bootstrap method to test the existence of threshold effects and analyze the results. We initially selected 1000 bootstrap samples and gradually increased the number of samples by 200 until 2000 samples. We found that from the first test with 1000 samples to the last test, the p-value results were basically the same, and the model estimation results were completely consistent. This suggests that the threshold effect statistic calculated using the bootstrap method is stable. Based on this result, we used the bootstrap method to conduct the test 1000 times in order to examine the existence of threshold effects in this paper. The results are shown in

Table 6. Threshold effect test results.

Variable	Threshold Type	F Value	p-Value	Threshold Value	95% Confidence Interval
lnER	Single threshold	35.01	0.0000	4.3438	[4.2340–5.2679]
	Double threshold	22.32	0.0100	4.7362	[4.6151–5.0999]
	Triple threshold	2.28	0.9533	-
AIS	Single threshold	27.85	0.0000	0.4342	[4.4302–4.4343]
	Double threshold	20.52	0.0140	0.6009	[6.0000–6.6031]
	Triple threshold	7.82	0.6300	-
lnATI	Single threshold	32.48	0.0133	5.9915	[5.9606–6.0064]
	Double threshold	15.13	0.2533	-
	Triple threshold	12.21	0.4567	-

. lnER, and AIS are significant at the 1% and 5% levels for the single-threshold and the double-threshold, respectively. This indicates that lnER and AIS have significant double threshold characteristics, making them suitable for constructing double-threshold regression models. lnATI passes the significance test for the single threshold, suggesting that it has a single threshold. Furthermore, the reliability of the threshold estimates is evaluated using the likelihood ratio (LR) statistic. As seen in Table 6, the threshold values for each variable are within the 95% confidence interval, indicating that the threshold values all pass the validity test.

The results of the threshold effect regressions are shown in

Table 7. The regression results of the panel threshold model.

Variable	(1)	(2)	(3)
lnER ≤ 4.3438	−0.0105 ***
	(0.0168)
4.3438 < lnER ≤ 4.7362	−0.0080 ***
	(0.0031)
lnER > 4.7362	−0.0076 ***
	(0.0043)
AIS ≤ 0.4342		−0.0521 ***
		(0.0090)
0.4342 < AIS ≤ 0.6009		−0.0082 ***
		(0.0023)
AIS > 0.6009		−0.0009
		(0.0038)
lnATI ≤ 5.9915			−0.0097 ***
			(0.0023)
lnATI > 5.9915			0.0018 **
			(0.0028)
control variables	YES	YES	YES
constant	0.0867 *	−0.0257 *	0.2919 **
	(0.1134)	(0.1141)	(0.1170)
R²	0.2446	0.2861	0.3556
obs	620	620	620

Note: Standard errors are reported in parentheses. *, **, and *** represent significant at 10%, 5%, and 1%, respectively.

. In column (1), when the level of environmental regulation is less than 4.3438, the coefficient of environmental regulation on CLUE is −0.0105. When the environmental regulation level is in the range of [4.3438, 4.7362], its estimation coefficient changes to −0.0080. When the environmental regulation level exceeds 4.7362, its estimation coefficient further changes to −0.0076. It can be seen that although the coefficient of environmental regulation in each interval is significantly negative under the dual-threshold model, this negative impact on CLUE is weakening. This reflects that the “benefit compensation” effect of environmental regulation on CLUE is rapidly increasing. In reality, while implementing environmental regulation measures to reduce carbon emissions from cultivated land and prevent and control agricultural non-point source pollution, the Chinese government has also been encouraging research, development, and application of ecological technologies for cultivated land to promote green and high-quality agricultural development. With the continuous improvement in the quality of cultivated land ecosystems, the competitiveness of agricultural quality and efficiency is enhanced, and the “benefit compensation” effect of environmental regulation will also become increasingly apparent.

Column (2) of Table 7 shows the estimated results with the agricultural industrial structure as the threshold variable. The threshold values of agricultural industrial structures are 0.4342 and 0.6009, which can be divided into three threshold intervals. In these three threshold intervals, the effect of environmental regulation on CLUE is significantly negative, and the impact coefficient is characterized by marginal decline. Within this interval, upgrading the agricultural industrial structure can reduce the suppression of environmental regulation on CLUE. China should continuously improve the agricultural production system, industrial system, and management system following the coordinated development of various agricultural departments, so as to continuously promote the optimization and adjustment of agricultural industrial structure.

In terms of technological effects, in column (3), when the lnATI index is less than the threshold value, the marginal effect of environmental regulation is −0.0097. The positive effect of environmental regulation on CLUE appears with a marginal effect of 0.0018 when the lnATI index crosses the threshold value. This shows that China should vigorously promote the strategy of revitalizing agriculture using science and education and put agricultural science and technology in a prominent position. Specifically, the government should increase investments in agricultural science and technology innovation and improve the ability of agricultural technology extension to play the innovative compensation effect of environmental regulation. Such innovative compensation effects of environmental regulation can improve the efficiency of land production, thus making up for the negative impact of pollution control costs. Therefore, there is a significant dual threshold effect of environmental regulations on CLUE, and agricultural industrial structure and technological innovation in agriculture are the moderating variables that affect the non-linear relationship between environmental regulation and CLUE. Thus, hypothesis 3 is validated.

6. Discussion

The empirical results of this study added to the growing literature concerning ecological security and enhanced cultivated land use in China. This article innovatively integrated environmental regulation and CLUE into the same analytical framework for theoretical and empirical analysis. The results showed that environmental regulation has a significant inhibitory effect on CLUE, but the mediating mechanism of agricultural industry structure and technological innovation weakened this inhibitory effect. Our results were consistent with the studies of Liu et al. [33] and Kuang et al. [34]. Ma and Tan [59] also found that environmental regulation has a significant negative inhibitory effect on agricultural green total factor productivity. They believed that environmental regulation reduces the profitability of agricultural production by crowding out agricultural technological investment. Based on their research on the water environment and water resources in the Yangtze River Basin, Qin and Zhang [63] found that the presence of environmental regulations leads to an increase in agricultural production costs, resulting in crowding out effects on production and profitability-oriented investments, thereby significantly suppressing total factor productivity in agriculture. However, we also noted that many previous studies have opposite conclusions to our research [35,36]. The main reason for the inconsistency between our conclusions and those of other scholars was the different measurement methods for environmental regulation variables. Other scholars selected environmental regulation measurement indicators from the perspective of government governance input, such as using investment in pollution control projects as proxy variables for environmental regulation, and believed that environmental regulation would promote agricultural technological innovation and thus improve total factor productivity [64,65].

In addition, the threshold effect model results of this paper showed that the inhibitory effect of environmental regulations on the ecological efficiency of land use is weakening. This was similar to Becker’s research inference, who believed that the dynamic impact of environmental regulations would not hinder the green development of agriculture [66]. Ye et al. [67] also found that with the increasing attention of the government on environmental protection issues, green financial subsidy measures were constantly being introduced, which would reduce the input costs of agricultural producers.

It is undeniable that there are still some limitations in this article. First, this study uses provincial-level panel data. In the future, it would be beneficial to evaluate CLUE from a micro-perspective, such as the city or county level, or from the perspective of individual farmers. Second, due to length limitations, this paper does not include control variables such as water resource management, climate change, and soil quality changes in the empirical analysis of the model and so cannot discuss their relevant conclusions. These issues require further research and analysis. Third, this paper does not consider the spatial externality of environmental regulation. Stimulated by political promotion competitions, the strategic interactive selection of environmental regulations between different regions can lead to different spatial spillover effects on CLUE. Future research can delve into the spatial spillover effects and influencing mechanisms of environmental regulation on CLUE.

7. Conclusions and Policy Implications

Using the panel data from 31 provinces, autonomous regions, and municipalities in China from 2000 to 2019, this paper designs an SSBM model to calculate CLUE in China in order to explore the development status of cultivated land use and ecological security. Additionally, combined with theoretical analysis, the fixed effect model, mediation effect model, and threshold effect model are used to empirically analyze the impact mechanism of environmental regulation on CLUE in this paper. The main conclusions are as follows:

First, the results reveal a consistent trend in CLUE, with an initial decrease followed by a gradual rise across the entire country and its eastern, central, and western regions. Furthermore, notable disparities in CLUE are observed among these regions, with the eastern region exhibiting the highest CLUE, followed by the western region and finally, the central region.

Second, environmental regulation has a significant negative inhibitory effect on CLUE. This indicates that the “cost loss” effect of environmental regulation during the sample period is greater than the “income compensation” effect.

Third, environmental regulation indirectly weakens its negative impact on CLUE through two intermediary transmission pathways: agricultural technology innovation and agricultural industrial structure.

Fourth, environmental regulation has a dual threshold effect on CLUE. As the intensity threshold range of regulation increases, the negative impact of environmental regulation on CLUE gradually diminishes. Within the high threshold range, agricultural industry structure is conducive to reducing the negative impact of environmental regulation on CLUE. When agricultural technological innovation exceeds a single threshold value, the impact of environmental regulation on CLUE changes from negative to positive.

Based on the above conclusions, this paper proposes the following policy recommendations. First, it is important to scientifically formulate environmental regulatory policies and continuously improve the level of environmental regulation. According to the results, the negative effect of environmental regulation on CLUE is diminishing. Therefore, the government should strengthen the enforcement of environmental regulations. On the one hand, the government should enhance the top-level design of environmental regulation policies in the field of land resources at the institutional level and plan specific policies, policy attributes, and policy means from a global perspective. On the other hand, the government should strengthen the integration of different laws and regulations, continuously improve the legal and regulatory system related to agricultural environmental protection, clarify soil environmental standards and land use standards, control pollution source emissions from the source, and take multiple measures to improve the efficiency of cultivated land use to form a basis for green development of cultivated land resources. Second, it is necessary to strengthen agricultural technological innovation to provide new momentum for low-carbon and green utilization of cultivated land. The government should increase investment in the research and development of agricultural ecological technologies, especially for the technical methods related to carbon sequestration, emission reduction, and pollution control of cultivated land. At the same time, it is important to strengthen the integration and innovation of cultivated land governance technology through formulating scientific and reasonable agricultural technology innovation plans, relying on digital information platforms to achieve digital transformation of the entire process of cultivated land resource utilization and management, and promoting the diffusion of agricultural scientific and technological innovations to the main body of cultivated land use. Third, optimizing the agricultural industry structure is essential to promoting green and efficient development in agriculture. The government should continuously optimize the spatial layout of the agricultural industry, adjust the agricultural industrial structure and planting structure, and realize the rational allocation of cultivated land resources and the improvement of output efficiency according to the characteristics of regional cultivated land resource endowment and market demand.