Research on the Impact of Command-and-Control Environmental Regulations on Green Innovation of Agricultural-Related Enterprises

Abstract

With the intensification of global environmental challenges and the growing demand for sustainable agricultural transformation, understanding how environmental regulation shapes enterprise innovation has become increasingly important. This study examines the impact of command-and-control environmental regulation on green innovation in agricultural enterprises using panel data from agriculture-related enterprises listed on the Shanghai and Shenzhen A-share exchanges. The analysis focuses on the period 2012–2021, which is characterised by relatively stable environmental regulation and reliable data, providing a consistent empirical context for assessing the effects of command-and-control environmental regulation. By analyzing the characteristics of command-and-control environmental regulation and green innovation in agricultural enterprises, this research constructs and estimates a two-way fixed effects model, a moderating effects model, a mediating effects model, and a spatial Durbin model to explore both direct and spillover effects. The empirical results show that the following findings: (1) Command-and-control environmental regulation significantly promotes green innovation in agricultural enterprises, and this effect remains robust across alternative measurements and model specifications. (2) Heterogeneity analysis indicates that the direct effect of command-and-control environmental regulation is most pronounced in eastern regions, non-state-owned enterprises, and enterprises with weaker environmental, social, and governance performance. (3) Moderation analysis shows that agricultural industrial coordination and executive green cognition significantly strengthen the positive relationship between command-and-control environmental regulation and green innovation in agricultural enterprises. (4) Mediation analysis demonstrates that green management costs serve as a partial mediator in this relationship. (5) Spatial analysis reveals that spatial correlation patterns are evolving over time, with significant positive spillover effects observed among geographically and economically adjacent regions. The findings provide theoretical and empirical evidence to inform the design of coordinated environmental regulation frameworks that effectively stimulate green innovation and foster sustainable agricultural development.

1. Introduction

With the intensification of global environmental challenges, promoting green innovation has become a strategic imperative for achieving sustainable economic development. Governments worldwide have adopted diverse environmental regulatory instruments to curb pollution and stimulate green innovation, among which command-and-control environmental regulations (CCERs)—including emission standards, pollution discharge limits, and mandatory technological requirements—remain the most direct and widely applied tools [1]. These regulations play a crucial role in enforcing environmental accountability, especially in developing countries where market-based regulatory systems are still evolving [2]. However, whether and how CCERs encourage or inhibit green innovation remains a central question in environmental economics and policy design.

In agriculture, environmental degradation both drives and results from farming [3], with diffuse non-point source pollution threatening soil and water quality and hindering sustainable development in China and beyond. Environmental regulation is thus essential for controlling pollution and promoting green transformation, although its effectiveness remains debated [4]. Evidence from Chinese listed enterprises shows that regional non-point source pollution can strongly drive green innovation via regulatory and subsidy mechanisms, highlighting the role of environmental policy in shaping enterprise-level innovation [5]. As agricultural modernisation accelerates, agriculture-related enterprises(AEs)—including agro-processing enterprises, input suppliers, and agri-tech enterprises—have emerged as key actors in linking production, environmental management, and technological progress [6]. These enterprises are tasked with improving resource efficiency and reducing pollution, and adopting cleaner production technologies to support sustainable agricultural transformation. Understanding how CCER influences green innovation in agricultural enterprises (AEGI) is therefore crucial for designing regulations that balance environmental protection with industrial competitiveness.

The ongoing discourse on the relationship between environmental regulation and innovation is grounded in the Porter Hypothesis, which posits that appropriately designed environmental regulation can stimulate innovation by improving organisational efficiency and competitiveness, in contrast to the conventional view that regulation mainly increases compliance costs and reduces profitability [6,7]. Empirical evidence remains mixed: many studies support the weak form of the Porter Hypothesis [8], showing that environmental regulation can stimulate innovation at both enterprises and regional levels in China [9,10], whereas other work highlights inhibitory or non-linear effects depending on regulatory design and economic context [5]. In China’s developing economy—where market mechanisms are less mature and CCER remains predominant—research suggests that such regulation can act as a significant driver of adaptive and incremental green innovation under certain conditions [4,8,11].

Existing studies have largely focused on manufacturing and energy-intensive industries, where both pollution control and technological change can be more precisely measured [12]. Agricultural enterprises differ across multiple dimensions: innovation activities are predominantly incremental rather than radical; environmental effects are geographically dispersed; and innovation returns are highly contingent on policy uncertainty and ecological risk [13]. Consequently, findings from manufacturing cannot be simply extrapolated to agriculture [14]. In addition, much of the literature aggregates regulatory instruments into a single “environmental regulation index” and fails to distinguish among command-and-control, market-based and voluntary mechanisms [15]. In developing economies such as China, CCERs still dominate environmental governance, yet their transmission mechanisms and spatial effects remain empirically under-explored. Furthermore, little attention has been paid to how CCERs interact with internal firm-level factors (e.g., managerial environmental cognition) or external industrial structures to shape green innovation outcomes [16].

To address deficiencies in existing research, this study investigates the effects of CCERs on AEGI in China. Panel data from A-share listed agricultural enterprises spanning 2012 to 2021 are analyzed using a two-way fixed-effects model, a moderating-effects model, a mediating-effects model, and a spatial Durbin model, enabling the identification of direct, indirect, and spatial effects of CCERs. Specifically, our analysis addresses four key areas: (1) The study extends the scope of environmental regulation research to agriculture-related enterprises (AEs), a sector that is both critical and underexplored in the context of the green transition. (2) By distinguishing among regulatory types, the study empirically tests the Porter Hypothesis in the context of agricultural enterprises. (3) The study identifies green management costs (GMCs) as a key mediating mechanism through which compliance pressure is transformed into incentives for innovation. (4) Spatial analysis demonstrates that CCER-induced innovation generates positive spillover effects across regions through technology diffusion and policy learning. Collectively, the findings offer theoretical and empirical insights for the design of adaptive environmental regulatory frameworks capable of effectively stimulating green innovation, enhancing inter-regional collaboration, and advancing sustainable transformation in the agricultural sector.

2. Theoretical Analysis and Research Hypotheses

2.1. Direct Effect of Command-and-Control Environmental Regulations on Green Innovation in Agricultural Enterprises

In accordance with institutional and legitimacy theories, innovation behavior in enterprises is shaped not only by internal resources but also by external institutional pressures. As a core regulatory instrument, CCERs compel AEs to comply with environmental standards through enforcement and penalties, thereby influencing their innovation strategies [13]. Within the framework of compliance cost pressures, the imposition of more stringent CCERs has been shown to increase adherence costs [17]. Consequently, many enterprises pursue technological innovation to enhance resource efficiency and reduce regulatory risks when the expected long-term benefits exceed the associated short-term costs [18]. Conversely, enterprises that achieve compliance through legitimate incentives gain access to policy support measures such as green subsidies, tax incentives, and preferential financing, while also enhancing their social reputation through environmental certification. CCERs serve as a catalyst for AEGI through the dual mechanisms of compliance cost pressure and legitimacy acquisition. This regulatory framework consequently encourages enterprises to increase investment in green innovation as a strategic response to strengthen their sustainable competitiveness [1].

However, the innovation response of agricultural enterprises to CCERs is constrained by pronounced sector-specific characteristics. Agricultural production is often marked by technological rigidity and strong path dependence, particularly in pollution-intensive activities such as fertiliser application and livestock breeding, where clean or end-of-pipe abatement technologies are difficult to substitute and innovation-related costs remain relatively high. Moreover, agricultural innovation is typically characterised by long development cycles and substantial uncertainty, with returns highly exposed to natural risks and market volatility, thereby weakening enterprises’ incentives and capabilities to engage in independent R&D. In addition, most agricultural enterprises are small or medium-sized and face persistent financing constraints, limiting their ability to absorb the short-term compliance costs imposed by stringent regulation. Against this backdrop, CCERs exert a dual effect on agricultural enterprises: regulatory pressure may crowd out innovation in the short term, whereas the combination of policy incentives and legitimacy-based rewards can stimulate adaptive innovation behaviour. Consequently, agricultural enterprises tend to adopt selective green innovation strategies, prioritising innovations that simultaneously reduce compliance costs and improve access to policy support, thereby alleviating regulatory risks and fostering long-term sustainable development.

It should be noted that CCER in this study is measured at the provincial level and does not directly determine enterprises-level innovation decisions. Rather, province-level CCER reflects the overall regulatory environment faced by agricultural enterprises. Through differentiated enforcement intensity, policy incentives, and local government implementation practices, regional environmental regulations are transmitted to enterprises and shape their perceived compliance pressure and innovation incentives. Accordingly, CCER is treated as a proxy for the institutional environment under which firms make green R&D investment decisions, rather than assuming homogeneous exposure across enterprises.

H1. 

CCER has a positive effect on AEGI.

2.2. Moderating Effects of Agricultural Industry Coordination and Executive Green Cognition

The level of Agricultural Industry Coordination (AC) reflects the degree of diversification and balance within the agricultural sector. Higher coordination indicates a more even distribution across farming, forestry, animal husbandry, and fishery subsectors, thereby enhancing resource complementarities and inter-industry synergies. Such coordination shapes the influence of CCER on innovation investment in AEs by strengthening cross-sectoral linkages and improving resource allocation efficiency. Specifically, AC moderates the impact of CCER on enterprise innovation input through two primary mechanisms, which are discussed here.

Regions with higher levels of AC exhibit stronger complementarities across subsectors, thereby reducing the cost burden of AEGI. For example, livestock enterprises and crop producers can share resources and technologies, such as converting livestock manure into organic fertiliser for crop cultivation. Such cooperative arrangements enhance resource-use efficiency and simultaneously reduce the environmental compliance costs borne by individual enterprises [19]. Moreover, a well-coordinated industrial structure enhances the effectiveness of policy transmission. In diversified agricultural systems, local governments can utilise mechanisms such as the “chain-chief system” to tailor policy instruments to the specific needs of enterprises, effectively converting regulatory pressures into targeted support. For instance, in Shandong’s livestock and poultry waste utilisation initiative, governments identified and linked key nodes along the agricultural value chain, providing customised assistance to livestock producers and helping them better respond to the tightening of environmental regulations.

Furthermore, from a risk dispersion perspective, diversified industrial structures contribute to the stabilization of regional income streams and enhance enterprise resilience to shocks arising from regulatory compliance costs [20]. When one agricultural sub-sector faces elevated innovation expenditure under CCERs, stable returns from complementary industries, such as agricultural services, logistics or processing, can offset financial pressure. This alleviates the short-term ‘crowding-out’ effect and supports the continuity of long-term innovation investment. Overall, greater link by strengthening the collaborative and adaptive capacities of agricultural enterprises under environmental constraints.

At the micro level, Executive Green Cognition (EGC) constitutes a critical determinant shaping enterprise-level interpretation and behavioural responses to CCERs [21]. Drawing on upper echelons theory, the present analysis contends that EGC orientation constitutes a fundamental determinant shaping executive perception of institutional pressure and its transformation into strategic action. Executives who demonstrate a high level of environmental awareness are more likely to interpret CCERs as opportunities for strategic transformation rather than as compliance burdens [22]. Enterprises with a forward-looking cognitive orientation engage in the design of innovation roadmaps, allocate R&D resources to eco-efficient technologies, and integrate sustainability objectives into corporate strategies [23]. This orientation enhances sensitivity to policy signals, strengthens the capacity to anticipate regulatory trends, and sustains the continuity of AEGI. Conversely, executives with limited environmental awareness tend to adopt a reactive, cost-minimisation mindset, focusing narrowly on immediate compliance requirements while neglecting long-term technological planning. This weakens the potential innovation-inducing effect of CCERs [24].

H2a. 

The degree of AC positively moderates the relationship between CCERs and AEGI.

H2b. 

EGC positively moderates the relationship between CCERs and AEGI.

2.3. Mediating Role of Green Management Costs Between Command-and-Control Environmental Regulation and Green Innovation in Agricultural Enterprises

CCERs exerts an indirect influence on the AECI, operating through its impact on GMC. From the resource-based view, stringent CCER requirements implemented through mandatory technical standards and penalty mechanisms compel enterprises to increase investments in pollution control and environment-specific assets, leading to a progressive increase in GMC. As environmental regulations become more stringent over time, enterprises must continually increase their GMC to maintain compliance with increasingly rigorous emission standards.

It is noteworthy that, in the short term, increased GMC may lead to a temporary reduction in R&D budgets. However, owing to their high sensitivity to policy incentives, agricultural enterprises typically alleviate financial constraints through external financing or other channels to maintain sustained investment in innovation. In circumstances where the financial implications of environmental management exceed a pivotal limit, the incremental expense of pollution control may potentially exceed that of technological enhancement. At this juncture, enterprises strategically adjust their cost structures, reallocating resources from end-of-pipe treatments to front-end R&D activities. This reallocation has been shown to engender an innovation compensation effect, wherein firms are able to reduce long-term environmental management costs through technological advancement [25]. Within this mechanism, GMC act as a “cost converter”, effectively transmitting regulatory pressure into technological innovation dynamics.

H3. 

GMC mediate the relationship between CCER and AEGI.

2.4. Spatial Spillover Effect of Command-and-Control Environmental Regulations on Green Innovation in Agricultural Enterprises

The spatial influence of CCER on AEGI manifests through two interrelated mechanisms: the demonstration effect and the resource siphoning effect. The demonstration effect is characterized by the emergence of advanced technological solutions and governance models in regions implementing stringent environmental regulations. Models diffuse across spatial boundaries through industrial linkages, knowledge transfer, and labour mobility. Diffusion processes establish technological benchmarks for neighboring regions, reduce innovation adoption costs, and generate competitive pressure for technological upgrading. Conversely, the phenomenon of resource siphoning occurs when regulatory stringency endows pioneering regions with disproportionate advantages in attracting financial capital, skilled labour, and policy support, thereby draining innovation resources from adjacent regions. The net spatial spillover effect of CCER is contingent on the relative strength of these opposing forces, which is subject to variation due to regional characteristics such as industrial complementarity, technological capability, and institutional maturity [26]. The specific mechanisms of action are illustrated in Figure 1.

H4. 

CCER exhibits spatial spillover effects on AEGI.

3. Research Design

3.1. Variable Description

3.1.1. Dependent Variable

AEGI serves as the dependent variable in this study and is measured along two dimensions: innovation input and innovation output. AEGI is strongly influenced by crop growth cycles and climatic conditions, which require sustained financial support. Seasonal growth dynamics and climatic variability influence input allocation, crop yields, and production stability, necessitating sustained investment in agronomic management and adaptive technologies to maintain productivity and foster environmentally sustainable innovation. AEGI is operationalized using innovation input measured as the ratio of R&D expenditure to operating revenue (unit: %), capturing enterprise-level capacity for long-term innovation [27].

3.1.2. Explanatory Variable

CCER, the earliest form of ecological governance adopted in China, is primarily implemented through laws, regulations, and mandatory emission-related policies. Existing studies typically assess CCER using indicators such as the number of environmental protection regulations, the incidence of administrative penalty cases, and pollutant compliance rates. However, environmental regulation in the agricultural sector remains relatively weak, and data limitations make it difficult to consistently measure regulatory enforcement or supervision intensity, resulting in the absence of a unified measurement standard. In this context, this study emphasizes the policy density dimension of CCER rather than regulatory enforcement strength and uses the number of agriculture-related pollution prevention and environmental protection policies (measured in hundreds) across 31 provinces (autonomous regions and municipalities) as a proxy for CCER intensity [28]. As the density of environmental regulation policies reflects the frequency and scope of policy supply and institutional attention at the regional level, this approach has been widely applied in environmental economics research [29]. Relevant policy documents are systematically retrieved from the PKULaw database by province and year, using “local regulations” as the legal type and “environmental protection” as the regulation category, to construct a province-level indicator of agricultural CCER.

3.1.3. Moderating Variable

The first moderating variable, AC, represents agricultural industrial structure adjustment and is employed as a feasible empirical proxy for coordination within the agricultural industry, given that coordination in a strict sense is difficult to measure directly. This study measures AC as one minus the share of crop production in the total output of the agriculture, forestry, animal husbandry, and fishery sectors [30]. A higher level of AC indicates a more diversified agricultural structure and a lower dependence on crop production, reflecting the outcome of structural adjustment across agricultural sub-sectors. Such structural diversification indirectly captures the reallocation of resources and factor mobility among crop production, forestry, animal husbandry, and fisheries, which is often associated with improved production complementarity and more efficient resource use. Consequently, a higher AC is expected to enhance the adaptive capacity of AEs to environmental regulations, thereby supporting green innovation responses, while the moderating effect of EGC further reinforces this relationship. EGC shapes AEGI decisions in response to environmental regulation. This study employs a text-mining approach to extract relevant keywords from annual reports of listed enterprises and constructs an EGC index comprising three dimensions: green competitive advantage, corporate social responsibility awareness, and perceived external environmental pressure. The frequency of these keywords in annual reports is used to quantify the level of executives’ environmental cognition.

3.1.4. Mediating Variable

GMC serves as the mediating variable. Corporate environmental expenditures can be divided into two categories: front-end green investments and end-of-pipe green costs. Front-end investments represent proactive expenditures directed toward pollution prevention and process optimisation, whereas end-of-pipe costs denote reactive expenditures associated with pollution treatment and environmental compensation, reflecting a passive form of governance. Measurement of GMC follows established methodologies in prior research, calculated as the ratio of environment-related administrative expenses to total operating revenue. Environment-related administrative expenses encompass greening fees, environmental impact assessment fees, and resource compensation fees.

3.1.5. Control Variable

In order to control for other factors that may affect the AEGI and reduce the homogeneity problem caused by omitted variables, a selection of control variables was made. Firm size (SIZE), measured as the natural logarithm of total assets, reflects resource capacity and managerial complexity; larger firms have stronger innovation potential but lower flexibility. The fixed asset ratio (FAR), defined as the ratio of net fixed assets to total assets, is a metric that captures asset structure. It has been demonstrated that an excessive fixation on this ratio can potentially constrain the liquidity required for R&D activities. Ownership concentration (SHR), measured by the largest shareholder’s equity share, is indicative of governance control. Moderate concentration supports innovation, whereas excessive dominance may induce conservatism. Leverage (LEV), defined as the ratio of total liabilities to total assets, is a measure of financial pressure; high leverage has been shown to restrict innovation investment. The quick ratio (QR), calculated as (current assets—inventory)/current liabilities, has been shown to reflect short-term solvency and financial flexibility. The Independent Director Ratio (IDR) is a metric used to assess governance, transparency and long-term orientation within an organization. It calculates the proportion of independent directors on the board and provides a quantitative measure of the board’s composition. At the regional level, the agricultural economic level (AGDP), measured by the ratio of agricultural output to GDP, has been shown to represent agricultural development intensity. The institutional environment (MI), measured by Fan Gang’s Marketization Index, has been demonstrated to capture the regional policy and market environment for innovation.

Definitions of the variables and descriptive statistical results are reported in

Table 1. Descriptive statistics of variables.

Variable Type	Variable Symbol	Variable Declaration	Average	Standard Deviation	Minimum	Maximum
Explained Variable	AEGI	The ratio of R&D investment to operating revenue	1.757	1.732	0.000	8.814
Core explanatory variable	CCER	The number of pollution prevention and environmental protection policies related to agriculture	0.538	0.304	0.080	1.420
Moderating Variables	AC	1-(Agricultural output value/Total output value of agriculture, forestry, animal husbandry and fishery)	0.491	0.076	0.255	0.636
	EGC	The frequency of keywords related to executive green cognition measurement dimensions appearing in annual reports of listed companies	2.862	3.533	0.000	17.000
mediating variable	GMC	The percentage (%) of total environmental protection-related expenses in the management cost breakdown relative to operating revenue	0.058	0.171	0.000	0.993
Control Variables	SIZE	Natural logarithm of year-end total assets	22.036	1.028	19.939	25.181
	FAR	Net fixed assets/total assets	0.265	0.136	0.028	0.610
	SHR	Number of shares held by the largest shareholder/total number of shares	0.350	0.143	0.093	0.730
	LEV	Year end total liabilities/Year end total assets	0.387	0.196	0.047	0.966
	QR	(Current Assets—Inventory)/Current Liabilities	1.787	2.164	0.163	14.695
	IDR	Independent directors/number of directors	0.380	0.061	0.308	0.600
	AGDP	Total output value of agriculture, forestry, animal husbandry and fishery/GDP	0.146	0.091	0.007	0.435
	MI	Market level index of Fan Gang by province	9.111	1.927	3.580	12.014

Source: Authors’ own creation.

. Please refer to

Table A1. Multi-Collinearity test of variables.

Variable	VIF	1/VIF
CCER	1.193	0.838
AC	1.397	0.716
EGC	1.156	0.865
SIZE	1.168	0.856
FAR	1.137	0.880
SHR	1.065	0.939
LEV	1.678	0.596
QR	1.707	0.586
IDR	1.022	0.978
AGDP	2.716	0.368
MI	3.240	0.309
Mean VIF	1.542	0.742

Note: The average VIF of each variable is 1.542, and the maximum value is less than 4, indicating that there is no serious col-linearity problem between variables. Source: Authors’ own creation.

in the Appendix A for the results of the multicollinearity test of the variables.

3.2. Model Setting

To examine the impact of CCER on AEGI we construct a fixed-effects model

$$AEGI=a_0+a_{1}CCE{R}_{it}+\sum a_{2}Contro{l}_{it}+u_i+v_t+{\epsilon }_{it}$$

(1)

$a_0$ denotes the constant term, $i$ and $t$ signify AEs and time periods, respectively. ${\mathit{AEGI}}_{\mathit{it}}$ denotes the ratio of R&D investment to operating revenue for enterprise i in year t, while ${\mathit{CCER}}_{\mathit{it}}$ denotes the level of command-and-control environmental regulation in the province where agriculture-related enterprise i is registered in year t, with $a_1$ denoting its regression coefficient. ${\mathit{Control}}_{\mathit{it}}$ and $\gamma$ encompass sets of control variables and their corresponding coefficients, ${\mu }_i$ captures the enterprise fixed effect, $v_t$ accounts for the time fixed effect, and ${\epsilon }_{\mathit{it}}$ denotes the random error term.

To examine the moderating effects, we constructed the following model:

$${AEGI}_{it}={\beta }_0+{\beta }_{1}CCE{R}_{it}+{\beta }_{2}A{C}_{it}+{\beta }_{3}CCE{R}_{it}\times A{C}_{it}+\sum {\beta }_{4}Contro{l}_{it}+u_i+v_t+{\epsilon }_{it}$$

(2)

$${AEGI}_{it}={\pi }_0+{\pi }_{1}CCE{R}_{it}+{\pi }_{2}EG{C}_{it}+{\pi }_{3}CCE{R}_{it}\times EG{C}_{it}+\sum {\pi }_{4}Contro{l}_{it}+u_i+v_t+{\epsilon }_{it}$$

(3)

${\mathit{AC}}_{\mathit{it}}$ denotes the Agricultural Industry Coordination degree and ${\beta }_2$ represents its regression coefficient, and $CCE{R}_{it}\times A{C}_{it}$ signifies the interaction term between CCER and AC degree with ${\beta }_3$ as its associated coefficient. Similarly, $EG{C}_{it}$ denotes the executive green cognition and ${\pi }_2$ represents its regression coefficient, while $CCE{R}_{it}\times EG{C}_{it}$ signifies the interaction term between CCER and executive green cognition with ${\pi }_3$ as its associated coefficient. The other variables are defined in the same manner as in Equation (1).

In order to examine the mediating role of GMC in the relationship between CCER and innovation investment in AEs, this study employs the three-step testing approach. The initial step involves estimating Equation (4) to assess the impact of CCER on innovation investment. The second step in the process involves the examination of the effect of CCER on GMC, utilising the following specification:

$${\mathit{GMC}}_{\mathit{it}} = {\lambda }_0 + {\lambda }_1{\mathit{CCER}}_{\mathit{it}} + \sum {\lambda }_2{\mathit{Control}}_{\mathit{it}} + {\mu }_i + v_t{ + \epsilon }_{\mathit{it}}$$

(4)

where rep ${\lambda }_0$ resents the constant term; ${GMC}_{it}$ denotes green management costs; ${\lambda }_1$ indicates the coefficient of CCER; ${\lambda }_2$ represents the coefficients of control variables; and other symbols retain their meanings as defined in the fixed effects model.

The third step of the research process involved the evaluation of the mediating effect of GMC in the relationship between CCER and innovation investment. This was achieved through the following equation:

$${AEGI}_{it}={\psi }_0+{\psi }_{1}CCE{R}_{it}+{\psi }_{2}GM{C}_{it}+\sum {\psi }_{3}Contro{l}_{it}+u_i+v_t+{\epsilon }_{it}$$

(5)

where ${\psi }_0$ is the constant term; ${\psi }_1$ represents the coefficient of CCER; ${\psi }_2$ indicates the coefficient of GMC; ${\psi }_3$ denotes the coefficients of control variables; and other symbols maintain their previously defined meanings.

Innovation investment by AEs may exhibit spatial interdependence and spillover effects across provincial regions. In order to assess this spatial correlation, the Global Moran’s I statistic is computed in order to evaluate the degree of geographical clustering or dispersion of innovation inputs. The formula is specified as follows:

$$\mathrm{Global} \mathrm{Moran}\prime \mathrm{s} I={\displaystyle \frac{n}{{\displaystyle \sum _{i=1}^n{\displaystyle \sum _{j=1}^n{W}_{ij}}}}}\times {\displaystyle \frac{{\displaystyle \sum _{i=1}^n{\displaystyle \sum _{j=1}^n{W}_{ij}(AEG{I}_i-\overline{AEGI})(AEG{I}_j-\overline{AEGI})}}}{{\displaystyle \sum _{i=1}^n{(AEG{I}_i-\overline{AEGI})}^2}}}$$

(6)

where Global Moran’s I represents the global Moran’s index; n denotes the number of provincial-level regions; i and j indicate the provinces where the AEs are registered; AEGI_i refers to the innovation level in the province where the enterprises are registered; $\overline{AEGI}$ represents the mean value of innovation levels across all provinces; and W_ij stands for the spatial weight matrix.

To examine the spatial spillover effect of CCER on AEGI across provincial boundaries, a spatial econometric model was developed for the analysis. Spatial econometric models primarily encompass spatial error, spatial lag, and spatial Durbin models. Subject to specific constraints, the spatial Durbin model can be converted into either a spatial error model or a spatial lag model. The general form of the spatial Durbin model is as follows:

$$AEGI{P}_{it}={\theta }_0+{\theta }_1{\sum }_{j=1}^n{W}_{ij}{AEGIP}_{jt}+{\theta }_2{CCER}_{it}+{\theta }_3{\sum }_{j=1}^n{W}_{ij}{CCER}_{jt}+\sum {\theta }_{4}Contro{l}_{it}+u_i+v_t+{\epsilon }_{it}$$

(7)

${\theta }_0$ denotes the constant term, ${\theta }_1$ represents the spatial auto-regressive coefficient of R&D investment, ${\theta }_2$ signifies the general regression coefficient of regulations ${\theta }_3$, denotes the spatial lag regression coefficient of regulations, $u_i$ captures the province fixed effect,$v_t$ accounts for the time fixed effect, and ${\epsilon }_{it}$ represents the random error term.

3.3. Data Source

The present study has selected agricultural enterprises listed on the Shanghai and Shenzhen A-share enterprises from 2012 to 2021 as the research sample. The data pertaining to CCER was retrieved from the Peking University Law database, while the indicators measuring AEGI were obtained from the CSMAR database. The data on control variables at the enterprise level is sourced from the CSMAR database, the Wind database, and the annual reports of listed companies. Provincial-level control variables are extracted from the China Statistical Yearbook. In order to guarantee the validity of the data and to mitigate the impact of outliers on the research findings, the following data were processed: Initially, the analysis excluded both ST and *ST companies, as well as companies for which incomplete data was available. Secondly, all continuous variables were subjected to shrinkage, with the parameters set at 1% and 99%. In conclusion, a total of 315 AEs, encompassing 2296 observations, were obtained for the study.

4. Results and Discussions

4.1. Regional Disparities in Command-And-Control Environmental Regulations Development and Policymaking in China

In this study, CCER is operationalised as the number of agricultural pollution control and environmental protection policies (per hundred documents) issued by 31 provincial-level regions, providing a consistent and policy-based measure of regulatory stringency. Widely adopted in environmental economics [31], this metric captures regulatory stringency through the intensity of policy issuance and offers a coherent and comparable measure across regions, which is particularly valuable given the limited data availability and the relatively weak regulatory framework in the agricultural sector. Relevant provincial regulations were systematically compiled from the PKU Law Database under the categories of “local regulations” and “environmental protection.” To analyse regional heterogeneity in CCER from 2012 to 2021, all provinces were classified into eastern, central, and western regions according to the standard of the National Bureau of Statistics of China, reflecting systematic differences in economic development, industrial structure, and ecological vulnerability.

From 2012 to 2017, the CCER level was highest in the central region, followed by the eastern region, and lowest in the western region (see Figure 2). The central region, being the primary grain-producing area in China, is characterised by extensive cultivation practices, intensive use of fertilisers and pesticides, and substantial manure emissions. Consequently, local governments in this region have implemented stricter regulatory measures. Following 2017, the eastern region overtook the central region to achieve the highest level of environmental regulation, while the western region retained its status as the least regulated region.

The increase in regulatory intensity in the eastern region can be attributed to three key factors. Firstly, the region boasts a strong economic base. Secondly, there is a higher demand for green agricultural development. Thirdly, the region already has well-established regulatory frameworks in place. Furthermore, as a policy pilot region for innovative environmental governance and green development, the eastern region has been proactive in promoting the implementation of higher environmental standards. Conversely, the western region, distinguished by its sparse population density, small-scale agricultural practices, and a minimal degree of agricultural intensification, experiences comparatively negligible pollution pressure. This results in a reduced necessity for stringent environmental regulation when compared to the central and eastern regions.

4.2. Regional Disparities in Green Innovation in Agricultural Enterprises Development and Policymaking in China

In defining AEs, this study follows the Industry Classification Guidelines for Listed Companies issued by the China Securities Regulatory Commission in 2012. Based on this standard, twelve agriculture-related industries are included in the sample: agriculture; forestry; animal husbandry; fishery; agricultural, forestry, animal husbandry and fishery services; agricultural and sideline food processing; food manufacturing; liquor, beverage and refined tea manufacturing; textiles; leather, fur and related products; papermaking and paper products; and wood processing and products made from bamboo, rattan, palm and grass (see Figure 3).

In 2012, the predominant concentration of industries with high innovation inputs was observed to be in agriculture-related services and light manufacturing. The focus in these sectors was on enhancing production efficiency, optimizing processes and materials, and upgrading traditional industries. It is anticipated that by 2021, the industries with the highest levels of innovation input will be those of “paper and paper products”, “agriculture” and “textiles”.

While high-level innovation industries in China have traditionally prioritized agriculture and light manufacturing, a distinct structural transition toward green development, intelligent manufacturing, and high value-added production has become increasingly evident. The rapid advancement of smart agriculture has further intensified this transformation by markedly amplifying innovation investment within the agricultural sector, reflecting the deepening integration of digital technologies, sustainability objectives, and modern production systems. Innovation within the agricultural sector now extends across cutting-edge domains including biological breeding, agricultural automation, and digital management. Advancements in these fields have substantially improved the efficiency and sustainability of agricultural production, reinforcing the broader trajectory of industrial transformation, structural upgrading, and technological progress.

Dividing AEs into 31 provinces according to their place of registration, Figure 4 and Figure 5 show the number of AEs and the level of AEGI in each province in 2012 and 2021.

Overall, the spatial distribution of AEs is relatively decentralized, and there are considerable variations in the number of enterprises and AEGI between different provinces. In terms of the spatial distribution of innovation input, provinces such as Hubei, Hebei, Zhejiang, Guizhou, Shanghai, and Guangdong demonstrated relatively higher levels of AEGI in 2012, with values of 4.352%, 2.526%, 2.450%, 2.390%, and 2.041%, respectively. In contrast, Tianjin, Jilin, Yunnan, Tibet, and Shaanxi exhibited AEGI levels below 0.02%. In 2021, the provinces of Hubei, Qinghai, Zhejiang, Heilongjiang, and Tibet exhibited the highest AEGI levels (4.883%, 4.732%, 3.073%, 2.907%, and 2.843%, respectively). Despite increases in innovation input in Shanxi, Guizhou, Inner Mongolia, and Yunnan, AEGI levels in these provinces remained below 0.500%, indicating comparatively low investment.

It is evident that an increase in the number of enterprises does not necessarily result in an increase in the level of AEGI. Despite the prevalence of AEs in Shandong Province, the level of AEGI remains comparatively low. This suggests that enterprises in this region predominantly adhere to conventional production methods and invest only a limited amount in technological research and development. Conversely, regions such as Hubei, despite having a smaller number of AEs, have experienced similar growth patterns. This finding indicates that the mere expansion of AEs is inadequate for enhancing overall innovation capacity. Instead, there is an urgent need for policy guidance and financial support to drive technological R&D and innovation.

4.3. Benchmark Regression Results

Table 2. Benchmark regression results.

Variable	(1)	(2)
	RD	RD
CCER	0.285 *** (0.104)	0.280 *** (0.103)
SIZE		−0.222 *** (0.066)
FAR		1.424 *** (0.311)
SHR		−0.693 ** (0.352)
LEV		−0.736 *** (0.242)
QR		0.090 *** (0.028)
IDR		−0.508 (0.432)
AGDP		1.472 (1.693)
MI		−0.015 (0.047)
Constant	1.596 *** (0.057)	6.605 *** (1.648)
Enterprise	YES	YES
Year	YES	YES
R2	0.823	0.836

Note: The standard error for robustness is in parentheses; **, ***, respectively, indicate significant at the 5%, and 1% levels, the same applies below. Source: Authors’ own creation.

provides benchmark regression results for the impact of CCER on the innovation input of AEGI by agricultural enterprises. Column (1) shows the results of the benchmark regression excluding control variables. The CCER coefficient was 0.258, which is significant at the 1% level. In column (2), including the control variables, increases the CCER coefficient to 0.280, which remains statistically significant at the 1% level. Regardless of variable control, CCER is found to significantly promote AEGI, thus corroborating the “compliance pressure” mechanism of institutional isomorphism theory. Thus, Hypothesis 1 was validated.

With regard to the regression results of the control variables in column (2), the coefficient of SIZE is significantly negative at the 1% level. This suggests that larger firm scale may lead to more complex management structures, thereby reducing innovation flexibility and decision-making efficiency. The FAR demonstrates a remarkably favourable coefficient at the 1% level, signifying that enterprises with elevated fixed asset investments possess considerable physical capital, thereby providing a fundamental material basis for innovation endeavours. The SHR coefficient demonstrates a significantly negative relationship at the 5% level, suggesting that highly concentrated ownership may result in excessive decision-making centralisation, favouring short-term profit returns over long-term technological innovation. The LEV demonstrates a considerably negative coefficient at the 1% level, indicating that elevated leverage has the potential to impede innovation investment due to heightened debt servicing pressures, which are further compounded by the seasonal and uncertain nature of agricultural production. The QR coefficient demonstrates a statistically significant positive relationship at the 1% level. This finding suggests that optimising short-term liquidity positions can enhance cash flow management and fund allocation capacity, thereby facilitating innovation investments. The coefficients for IDR, AGDP, and MI are statistically insignificant, potentially due to constraints from family-controlled governance structures, the indirect nature of agricultural economic conditions, and the delayed impact of institutional improvements on innovation investment.

4.4. Heterogeneity Analysis

This section examines the heterogeneous effects of CCER on innovation investment across three dimensions: regional distribution, ownership structure, and ESG performance. The results are presented in

Table 3. Heterogeneity Analysis result.

Variable	(1) East	(2) Midwest	(3) State-Owned Enterprises	(4) Non-State-Owned	(5) High ESG Rating	(6) Low ESG Rating
CCER	0.367 ** (0.150)	0.136 (0.136)	0.174 (0.196)	0.337 *** (0.126)	−0.0732 (0.146)	0.395 ** (0.165)
Control variable	Yes	Yes	Yes	Yes	Yes	Yes
Constant	10.51 *** (2.530)	3.187 * (1.683)	7.340 *** (2.381)	7.101 *** (2.273)	21.97 *** (3.336)	1.809 (2.100)
Firm	Yes	Yes	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes	Yes	Yes
R2	0.831	0.843	0.837	0.840	0.844	0.861

Note: The standard error for robustness is in parentheses; *, **, ***, respectively, indicate significant at the 10%, 5%, and 1% levels, the same applies below. Source: Authors’ own creation.

.

(1)

Regional Heterogeneity

In accordance with the classification system employed by the National Bureau of Statistics, enterprises are to be divided into two distinct regions, namely the eastern and central-western regions, on the basis of their registration location. Column (1) displays a highly significant positive CCER coefficient of 0.367 (at the 5% level) for eastern regions, while column (2) presents an insignificant coefficient for central-western regions. This regional disparity suggests that well-developed markets and stronger regulatory pressure in eastern regions motivate innovation for compliance advantages, whereas weaker institutional support in central-western regions diminishes regulatory effectiveness.

(2)

Ownership Heterogeneity

Enterprises are divided into two broad categories: state-owned enterprises (SOEs) and non-SOEs. Column (3) indicates an insignificant CCER coefficient for SOEs, while column (4) demonstrates a significantly positive coefficient of 0.337 (1% level) for non-SOEs. It is evident that Non-SOEs exhibit a higher degree of innovation in their responsiveness to environmental regulation, potentially seeking policy recognition and market advantages through innovation. Conversely, the innovation decisions of SOEs appear to be constrained by bureaucratic processes and governmental interventions.

(3)

ESG Heterogeneity

Utilising ESG ratings from the HuaZheng Database, firms are categorised based on their sample mean ESG scores. Column (5) demonstrates that the CCER coefficient for high-ESG firms is not statistically significant. Conversely, column (6) reveals that the coefficient for low-ESG firms is 0.395, which is statistically significant at the 5% level. Lower ESG-rated firms, facing greater regulatory pressure due to weaker environmental compliance, demonstrate stronger innovation responses. Higher ESG-rated enterprises, with well-established environmental practices, demonstrate minimal additional innovation stimulation from command-and-control regulations.

4.5. Robustness Test Results

This paper employs multiple testing methods, including replacing the explanatory variable, replacing the dependent variable, accounting for lag effects, and substituting with a Tobit model. The results are presented in

Table 4. Robustness test results.

Variable	(1) Replace Explanatory Variables	(2) Replace the Explained Variable	(3) Explanatory Variable Lags Behind by One Period	(4) Replace with Tobit Model
CCER	0.224 *** (0.084)	0.141 ** (0.070)	0.281 *** (0.109)	0.650 *** (0.218)
SIZE	−0.231 *** (0.066)	−0.328 *** (0.047)	−0.201 *** (0.0729)	−0.082 (0.094)
FAR	1.451 *** (0.308)	0.859 *** (0.202)	1.243 *** (0.355)	1.573 ** (0.664)
SHR	−0.746 ** (0.354)	0.651 ** (0.265)	−0.420 (0.383)	−1.044 (0.654)
LEV	−0.696 *** (0.243)	−0.601 *** (0.121)	−0.600 ** (0.263)	−1.335 *** (0.470)
QR	0.093 *** (0.028)	−0.010 (0.009)	0.0945 ** (0.0378)	0.107 ** (0.044)
IDR	−0.520 (0.430)	−0.498 (0.308)	−0.310 (0.439)	−0.827 (1.233)
AGDP	1.233 (1.714)	0.089 (0.816)	1.410 (1.764)	1.366 (1.739)
MI	−0.045 (0.048)	−0.049 * (0.030)	−0.0286 (0.0508)	0.245 *** (0.066)
Constant	7.203 *** (1.632)	8.669 *** (1.101)	6.098 *** (1.814)	1.279 (2.045)
Enterprise	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes
R2	0.836	0.816	0.844

Note: The standard error for robustness is in parentheses; *, **, ***, respectively, indicate significant at the 10%, 5%, and 1% levels, the same applies below. Source: Authors’ own creation.

.

The regression coefficient from the replacement test is shown in column (1) to be 0.224, which is significantly positive at the 1% level and consistent with the benchmark regression results, thus confirming the robustness of the conclusions. In order to more intuitively quantify corporate innovation investment levels, the present study replaces the dependent variable with R&D expenditure as an alternative measure. Column (2) of the table reports regression results using R&D expenditure as the dependent variable, with the coefficient for CCER estimated at 0.141 and statistically significant at the 5% level. This further validates the robustness of the conclusions that were previously drawn. Considering that CCERs may exert a lagged effect on AEGI, with the impact of current regulations potentially materialising in subsequent periods, a one-period lag is introduced into the core explanatory variable to address potential endogeneity. Column (3) of the table reports the regression coefficient for CCER, estimated at 0.281 and statistically significant at the 1% level, confirming the robustness of the benchmark regression results. It is notable that certain agribusinesses have reported zero innovation investments, which has led to the observation of zero-value clustering in the dependent variable. This, in turn, has the potential to introduce bias into fixed-effects regression models. Consequently, this study replaces the two-way fixed-effects model with a Tobit model in order to more effectively address the issue of dependent variable truncation. Column (4) of the table reports the regression coefficient for CCER under the Tobit model, estimated at 0.650 and statistically significant at the 1% level, further validating the robustness of the benchmark regression results.

4.6. Moderating Effects Regression Results

We examined the moderating effects on the relationship between CCER and AEGI. Column (1) in

Table 5. The moderation effect analysis results.

Variable	(1)	(2)	(3)
CCER	0.280 *** (0.103)	0.287 *** (0.104)	0.247 ** (0.105)
AC		0.766 (0.624)
ER*AC		1.901 ** (0.939)
EGC			0.001 (0.007)
CCER × EGC			0.035 ** (0.017)
SIZE	−0.222 *** (0.066)	−0.229 *** (0.066)	−0.220 *** (0.067)
FAR	1.424 *** (0.311)	1.437 *** (0.312)	1.432 *** (0.311)
SHR	−0.693 ** (0.352)	−0.708 ** (0.354)	−0.666 * (0.353)
LEV	−0.736 *** (0.242)	−0.722 *** (0.243)	−0.722 *** (0.243)
QR	0.090 *** (0.028)	0.090 *** (0.028)	0.090 *** (0.028)
IDR	−0.508 (0.432)	−0.531 (0.431)	−0.480 (0.432)
AGDP	1.472 (1.693)	1.376 (1.752)	1.591 (1.696)
MI	−0.015 (0.047)	−0.014 (0.048)	−0.012 (0.047)
Constant	6.605 *** (1.648)	6.375 *** (1.669)	6.491 *** (1.652)
Enterprise	YES	YES	YES
Year	YES	YES	YES
R2	0.836	0.836	0.836

Note: The standard error for robustness is in parentheses; *, **, ***, respectively, indicate significant at the 10%, 5%, and 1% levels, the same applies below. Source: Authors’ own creation.

presents the results.

The moderating effect of the AC degree is reported in column (2). While the regression coefficient of the AC degree is statistically insignificant, the coefficient of its interaction term and CCER is 1.901, significant at the 5% level. This finding suggests that the degree of AC positively moderates the effect of CCER on AEGI, thereby validating Hypothesis 2a. Regions with higher levels of AC exhibit more balanced subsector structures and greater efficiency in factor allocation. By facilitating shared pollution-control facilities, by-product recycling, and cross-sector risk sharing, such coordination lowers unit innovation costs and alleviates short-term investment crowding-out effects associated with CCER. Consequently, a well-coordinated agricultural industrial structure strengthens agricultural enterprises’ capacity to respond to CCERs and stimulates higher levels of innovation investment.

Column (3) reports the moderating effect of executive environmental cognizance on innovation investment. Although the regression coefficient of EGC is statistically insignificant, the coefficient of the interaction term between EGC and CCER is 0.035, significant at the 5% level. This result demonstrates a positive moderating effect of EGC on the relationship between CCER and AEGI, providing empirical support for Hypothesis 2b. Specifically, executive teams with strong environmental awareness are more likely to perceive CCER as an opportunity rather than merely a cost burden, for example by leveraging government subsidies and developing green brands. They tend to adopt long-term environmental strategies, proactively increase R&D investment to optimise production processes, and invest early in environmental technology innovation, thereby avoiding regulatory penalties and reputational losses while achieving a balance between compliance and cost efficiency. In contrast, executives with weaker environmental awareness often adopt short-term, compliance-oriented responses, investing only the minimum required resources and lacking systematic innovation planning.

4.7. Regression Results of the Mediation Effect

This section employs the three-step regression approach to examine whether CCER influences innovation investment in agricultural enterprises through GMC. The results of the study are presented in

Table 6. Mediation effect analysis results of firms’ green costs.

Variable	(1) RD	(2) GMC	(3) RD
CCER	0.280 *** (0.103)	0.026 ** (0.0132)	0.269 *** (0.103)
GMC			0.409 ** (0.205)
Control variable	YES	YES	YES
Constant	6.605 *** (1.648)	0.001 (0.008)	6.546 *** (1.667)
Individual	YES	YES	YES
Year	YES	YES	YES
R2	0.836	0.648	0.836

Note: The standard error for robustness is in parentheses; **, ***, respectively, indicate significant at the 5%, and 1% levels, the same applies below. Source: Authors’ own creation.

. Column (1) replicates the baseline regression from Table 6, thereby confirming the direct positive effect of CCER on innovation investment.

Column (2) presents the regression results of CCER on green management costs (GMC), with a coefficient of 0.026 that is statistically significant at the 5% level. The result suggests that stringent environmental regulations incentivize agricultural enterprises to increase end-of-pipe pollution control expenditures. Column (3) demonstrates that after incorporating GMC into the baseline regression, the coefficient of GMC is 0.409 (significant at 5% level), confirming that increased GMC effectively promote AEGI. These results verify the mediating role of GMC in the relationship between CCER and innovation investment, thus supporting Hypothesis H3. The direct effect of CCER decreases from 0.280 to 0.269 while remaining significant at the 1% level, indicating partial mediation.

During the initial phase of regulatory stringency, there is a possibility that compliance requirements, such as investments in pollution treatment equipment, may temporarily result in a reduction in R&D funding. However, agricultural enterprises have the capacity to alleviate financial constraints through external financing channels. The documentation of environmental expenditures has been shown to facilitate access to government subsidies or green credit, thereby transforming compliance costs into financial support. As GMC accumulates, enterprises are motivated to increase innovation investment to achieve long-term cost reduction and resource efficiency improvements, ultimately realizing innovation compensation effects.

4.8. Space Autocorrelation Test

Recognising that interregional spatial dependence is shaped not only by geographic proximity or economic linkages but also by their interaction, this study employs an economic–geographic nested weight matrix to more comprehensively examine the spatial effects of CCER on AEGI. In order to examine the spatial correlation of AEGI, the Global Moran’s I index was calculated for the period 2012–2021 (

Table 7. Global Moran’s Index of green innovation in provincial agriculture-related enterprises, 2012–2021.

Year	I	z	p
2012	0.063	2.462	0.007
2013	0.074	2.781	0.003
2014	0.088	3.126	0.001
2015	0.029	1.554	0.06
2016	0.002	0.692	0.074
2017	0.004	0.961	0.168
2018	−0.003	0.784	0.217
2019	−0.003	0.784	0.217
2020	−0.017	0.449	0.327
2021	−0.045	−0.296	0.384

). The index demonstrated a significant positive trend from 2012 to 2015, but subsequently exhibited an annual decline, indicating a strong yet weakening spatial dependence. This was characterised by the formation of policy- and geography-driven clusters, with high- and low-investment provinces. The 2012 “No.1 Central Document” advocated for the promotion of agricultural innovation. Consequently, the eastern provinces established innovation hubs, while the central and western regions experienced a lag due to their comparatively weaker foundations. As infrastructure and technology diffusion improved, disparities narrowed, and by 2016–2017, Moran’s I lost significance, reflecting a trend towards spatial equilibrium. From 2018 to 2021, the index underwent a negative and insignificant shift, indicating the emergence of spatial heterogeneity driven by technological barriers and regional competition. Provinces such as Zhejiang have been able to consolidate their digital agriculture advantages, whereas their neighboring region of Anhui has remained constrained by traditional practices. Despite the weakening of spatial autocorrelation over time, spatial factors continue to exert an influence on the relationship between CCER and AEGI. Consequently, the decision was taken to conduct further testing of the Structural Equation Model (SEM), the Spatial Autoregressive Model (SAR), and the Spatial Error Model (SEM), in order to ascertain the most appropriate specification for subsequent analysis.

4.9. Spatial Effect Regression Results

The identified spatial spillover effects should be interpreted at the regional level. They reflect interregional linkages arising from regulatory coordination, policy learning, technology diffusion, and shared innovation environments, rather than direct interactions or imitation behaviors among individual agricultural enterprises.

This study evaluates alternative spatial econometric specifications (

Table A2. Spatial econometric model test results of the impact of CCER.

Test Method	Statistical Value	p Value	Test Method	Statistical Value	p Value
LM-error	6.369	0.012	Wald-SDM/SEM	37.91	0.000
Robust LM-error	2.951	0.086	LR-SDM/SAR	54.34	0.000
LM-lag	3.543	0.060	LR-SDM/SEM	62.68	0.000
Robust LM-lag	0.125	0.724	LR-both/ind	20.05	0.029
Hausman	36.79	0.000	LR-both/time	268.72	0.000
Wald-SDM/SAR	40.33	0.000

Note: The spatial dependence of both explanatory variables and interpreted variables can be captured simultaneously by the Spatial Durbin model, which exhibits superior model inclusiveness. Source: Authors’ own creation.

in the Appendix A) and selects the Spatial Durbin Model (SDM) based on the LM, Wald, and LR test results. The SDM allows for the simultaneous identification of spatial dependence in green innovation outcomes and the spatial transmission of regulatory intensity, thereby providing a comprehensive framework for analysing interregional interactions.

The estimated spatial autoregressive coefficient (p) is −0.483 and statistically significant at the 1% level, indicating negative spatial dependence in AEGI across provinces. This implies that higher green innovation investment in one province is associated with relatively lower innovation input in neighboring provinces, reflecting spatial competition rather than pure technological spillovers. Such competition likely arises from the concentration of scarce innovation resources—such as green finance, skilled labour, and policy support—in provinces with stronger innovation performance, which may crowd out adjacent regions with weaker absorptive capacity.

After controlling for spatial dependence, the direct effect of CCER on provincial AEGI remains significantly positive at the 1% level, confirming that stricter CCER effectively stimulates green innovation investment within a province. Given that spatially lagged coefficients in the SDM cannot be directly interpreted as spillover effects when p is significant, this study applies the partial differentiation approach to decompose the total impact of CCER into direct, indirect, and total effects (see

Table 8. Regression results of spatial Durbin model.

Variable	Main	Wx	Direct Effect	Indirect Effect	Total Effect
CCER	0.414 ** (0.202)	3.149 ** (1.469)	0.350 * (0.203)	2.141 ** (1.011)	2.490 ** (1.055)
Control	YES	YES	YES	YES	YES
Province	YES	YES	YES	YES	YES
Year	YES	YES	YES	YES	YES
ρ	−0.483 ** (0.209)
R2	0.199

Note: The standard error for robustness is in parentheses; *, ** respectively, indicate significant at the 10%, 5%, levels, the same applies below. Source: Authors’ own creation.

).

The results show a significantly positive indirect effect of CCER, indicating that environmental regulation in one province promotes green innovation in neighboring regions. This spillover effect is mainly driven by regulatory demonstration and policy learning mechanisms. Provinces with stricter CCER often serve as reference points for surrounding regions, reducing information and adjustment costs associated with green technology adoption. Improvements in transportation infrastructure and digital connectivity further facilitate cross-regional diffusion of regulatory practices and environmental technologies.

The coexistence of a negative spatial autoregressive coefficient and a positive regulatory spillover effect suggests that innovation competition and policy diffusion operate simultaneously. Overall, the diffusion and demonstration effects dominate, resulting in a significantly positive net spillover effect of CCER on AEGI, thereby supporting Hypothesis 4.

These spatial spillover effects are interpreted at the regional level, capturing interregional linkages in regulatory practices, policy learning, and technology diffusion rather than direct imitation or interaction among individual agricultural enterprises.

5. Conclusions and Recommendation

Using panel data on Chinese A-share listed agricultural enterprises from 2012 to 2021, this study finds that CCER significantly promotes AEGI, a result robust to multiple specifications and estimation methods. The effect exhibits clear heterogeneity, being stronger in eastern regions, non-state-owned enterprises, and enterprises with lower ESG ratings. Mechanism analysis shows that regional industrial coordination and EGC amplify the innovation effects of CCER, while GMC partially mediate this relationship as compliance inputs are transformed into innovation incentives. Spatial analysis further reveals significant positive spillover effects, indicating that technology diffusion and policy learning outweigh potential crowding-out effects. Compared with existing studies that focus mainly on manufacturing sectors or single mechanisms, this study provides a more comprehensive agriculture-oriented perspective by integrating heterogeneity, mechanism, and spatial spillover analyses.

It is evident that, in consideration of the findings, a number of policy recommendations can be made. (1) Enhancing the legal framework for agricultural environmental protection is essential to ensuring policy coherence and institutional stability. Achieving this objective requires the establishment of a flexible regulatory framework aligned with the developmental stages of enterprises, thereby fostering the integration of environmental and innovation objectives. (2) As CCER exerts a stronger stimulative effect on innovation in the eastern region than in the central and western regions, environmental policies should be adapted to regional conditions. Targeted support such as subsidies or tax incentives can help enterprises in less-developed regions mitigate innovation costs. Moreover, since non-state-owned enterprises respond more actively to regulation than state-owned enterprises, joint R&D and shared green-technology initiatives should be encouraged to enhance technology diffusion and strengthen innovation incentives within state-owned enterprises. (3) Agricultural enterprises should be encouraged to increase investment in innovation, adopt green technologies to enhance long-term competitiveness, and diversify financing channels through appropriate government support mechanisms. (4) Cultivating stronger EGC is essential for embedding sustainability as a core element of strategic decision-making. Promoting cross-industry collaboration and resource sharing is essential for accelerating technological diffusion and advancing the green transformation of the industrial system. (5) In consideration of the spatial spillover effects of CCER, there is a compelling argument for the enhancement of region-specific regulatory mechanisms and the promotion of interregional cooperation. Such measures are considered essential for the optimisation of policy effectiveness and the advancement of sustainable agricultural innovation.

6. Research Deficiencies and Prospects

Although this study provides robust evidence on the effects of CCER on AEGI several limitations remain.

(1)

The measurement of CCER relies on policy counts due to data constraints and the historically peripheral status of agricultural pollution control; however, this proxy may not capture regulatory intensity at the enterprise level. Future research should construct more granular, enterprise-specific indicators as data availability improves.

(2)

The analysis spans 2012–2021, a period of relatively stable regulation and reliable data, but the limited timeframe restricts the assessment of long-term policy effects. Extending the study to include periods of major regulatory shifts would improve the robustness and generalisability of the findings.

(3)

Information disclosure by AEs remains limited, and policy incentives may induce overreporting of innovation investment. Future studies should combine survey data with third-party sources to enhance data accuracy and reinforce empirical reliability.