The Impact of Environmental Regulation on Greenization Level of Manufacturing Industrial Chains: A Dual Perspective of Direct Effects and Spatial Spillovers

Abstract

Amid escalating global climate change and the urgent international demand for low-carbon development, enhancing the greenization level of manufacturing industrial chains has emerged as a critical policy priority. This study investigates the impact of environmental regulation on the greenization of China’s manufacturing industrial chains using provincial panel data from 30 regions (2010–2022), employing two-way fixed effects and spatial Durbin models. The results demonstrate that environmental regulation significantly promotes industrial chain greenization through three pathways: industrial structure rationalization, green technology innovation, and industrialization advancement. Heterogeneity analysis reveals stronger regulatory effects in regions characterized by coal-dependent energy structures, low shares of energy-intensive industries, and underdeveloped digital economies, while negligible impacts are observed in areas with cleaner energy mixes, high energy-intensive industrial concentrations, or advanced digitalization. Spatial econometric results confirm positive spatial spillovers, indicating that environmental regulation in one region enhances neighboring areas’ greenization through policy coordination and technology diffusion. Based on these findings, this study proposes tailored policy recommendations, including strengthening regulation in coal-reliant regions, optimizing industrial structures in energy-intensive hubs, and fostering cross-regional governance synergy to mitigate pollution haven effects. The research provides novel insights into achieving sustainable manufacturing transitions under the “dual carbon” framework.

1. Introduction

Amid intensifying global climate change, the international community faces mounting imperatives for low-carbon development, with the reconciliation of ecological preservation and economic growth emerging as a pivotal policy challenge for contemporary nations [1]. China’s manufacturing value-added accounted for 30% of the global total in 2023, consolidating its position as the “world’s manufacturing hub”. This dominance, however, amplifies the pressures confronting China in achieving its “dual carbon” (carbon peak and carbon neutrality) goals. The Report to the 20th National Congress of the Communist Party of China (CPC) framed this challenge through the lens of harmonious coexistence between humanity and nature, underscoring the imperative to greenify economic development and positioning decarbonization as a cornerstone of high-quality growth. Subsequent deliberations at the Third Plenary Session of the 20th CPC Central Committee advanced a comprehensive policy framework encompassing fiscal, financial, investment, pricing, and standardization mechanisms to institutionalize green, low-carbon, and circular economic systems. These developments affirm China’s strategic commitment to leveraging “green development” as the linchpin for reconciling its manufacturing supremacy with climate obligations, wherein the ecological upgrading of industrial chains constitutes both a prerequisite for high-quality development and a critical pathway toward dual carbon objectives.

In recent years, China’s environmental governance framework has undergone systematic refinement, exemplified by the establishment of carbon emission trading markets and the proliferation of green supply chain management practices, with multi-dimensional policy instruments accelerating the “greening” of industrial and supply chains. Existing scholarship on environmental regulation has yielded theoretically and pragmatically significant insights, primarily focusing on its impacts on green technological innovation [2,3], green transition [4,5], and industrial structure upgrade [6,7]. From the above literature, we understand that against the dual backdrop of urgent global climate governance and China’s role as the “world’s manufacturing hub,” China is promoting the green transformation of its manufacturing industry chain through systematic environmental policies and market mechanisms to achieve synergy between its “dual carbon” goals and high-quality economic development. However, despite the increasingly refined policy framework, current academic research still exhibits significant limitations: it overly focuses on the impact of environmental regulations on macro-level green innovation or industrial upgrading, while neglecting in-depth exploration of the specific mechanisms and micro-level dynamics within the green transformation of industry chains.

Theoretically, environmental regulation may not only compel technological and process innovations through coercive constraints but also reconfigure cross-industrial collaboration patterns via market incentives, thereby enhancing systemic green performance across value chains. Yet critical questions persist: What precise effects does environmental regulation exert on the greenization of China’s manufacturing industrial chains? Through what mechanisms do these effects materialize? Do heterogeneous impacts exist across regions or sectors? Is there spatial interdependence in regulatory outcomes?

Addressing these knowledge gaps, this study employs provincial panel data and advanced econometric methods—including two-way fixed effects models and spatial Durbin models—to investigate the relationship. Our contributions are threefold:

(1)

Methodological: We develop a novel provincial-level measurement framework for manufacturing industrial chain greenization and empirically quantify environmental regulation’s impacts.

(2)

Heterogeneous effects: This study further examines the heterogeneous effects of environmental regulation on the greenization level of manufacturing industrial chains through two novel analytical perspectives: energy consumption structure and the market value share of energy-intensive industries.

(3)

Spatial Analysis: We reveal significant spatial spillover effects, demonstrating how environmental regulation in one region positively influences neighboring areas’ industrial chain greenization.

2. Literature Review and Research Hypotheses

Environmental regulation serves as a policy tool that internalizes pollution costs to reduce the negative externalities imposed by environmental pollution on the welfare of socioeconomic entities, aiming to maximize public welfare through green and sustainable development.

Types of environmental regulation include command-and-control, market-incentive, and public-participation regulations; ex ante incentive and ex post punitive regulations; cost-based and investment-oriented regulations; and formal versus informal regulations [8]. This study adopts government-led environmental governance policies as its analytical lens to explore their direct effects, mechanistic influences, and spatial spillover effects on the greenization level of manufacturing industrial chains.

2.1. Impact of Environmental Regulation on the Greenization Level of Manufacturing Industrial Chains

According to externality theory, pollution emissions from manufacturing activities exhibit negative externalities [9]. After internalizing pollution costs, firms’ marginal private costs rise to align with social costs. The resulting compression of profit margins forces enterprises to reduce pollution-intensive production and invest in green technology innovation, thereby mitigating negative externalities. In practice, environmental regulation operates through two pathways. First, as a policy instrument, it internalizes pollution costs to adjust the cost constraints on manufacturing firms. To alleviate cost pressures, manufacturing industrial chains improve energy efficiency at the production source, transition to clean energy, and enhance the recycling and monitoring of industrial emissions at the downstream end. These adjustments driven by policy pressure and cost constraints elevate the greenization level across both upstream and downstream segments of industrial chains. Second, environmental regulation restructures regional industrial configurations through cost constraints [7], imposing fiscal burdens on energy-intensive and high-emission manufacturing firms to compel their green transition.

Thus, this study proposes:

Hypothesis H1.

Environmental regulation significantly enhances the greenization level of manufacturing industrial chains.

2.2. Mechanistic Analysis of Environmental Regulation on Industrial Chain Greenization

2.2.1. Mechanism of Industrial Structure Rationalization

Drawing on industrial organization theory [10], environmental regulation raises entry barriers for manufacturing sectors by adjusting cost constraints, modifies corporate profit margins and survival conditions, and reshapes market competition dynamics. This process eliminates outdated, energy-intensive, and inefficient production capacities while accelerating the exit of small and medium-sized polluting enterprises. The consequent increase in market concentration rationalizes industrial structures, allowing surviving manufacturing firms to leverage economies of scale to reduce per-unit pollution emissions and costs, thereby advancing industrial chain greenization.

This leads to:

Hypothesis H2.

Environmental regulation promotes the greenization level of manufacturing industrial chains by enhancing industrial structure rationalization.

2.2.2. Mechanism of Green Technology Innovation

When confronting environmental regulation, manufacturing firms often resort to the creation and application of green patents, which may inherently elevate industrial chain greenization. However, academic consensus remains divided on whether environmental regulation genuinely stimulates green innovation.

Scholars evaluate this issue by comparing the “cost offset effect” and “innovation compensation effect” [11]. Excessively stringent environmental regulation may push pollution control costs beyond firms’ tolerable thresholds, where the “innovation compensation effect” fails to counterbalance additional expenses. The resulting compliance costs crowd out normal production activities and stifle green innovation, manifesting a “cost burden effect”. Conversely, under moderate regulatory intensity, the “innovation compensation effect” outweighs cost pressures. Firms adopt green innovations to reduce emissions, secure competitive advantages, and offset R&D expenditures, aligning with the Porter Hypothesis [12]. Given these contradictory perspectives, this study posits competing hypotheses:

Hypothesis H3a.

Environmental regulation enhances green patent innovation, thereby improving industrial chain greenization.

Hypothesis H3b.

Environmental regulation suppresses green patent innovation, thereby hindering industrial chain greenization.

2.2.3. Mechanism of Industrialization Level

Romer’s endogenous growth theory [13] emphasizes knowledge capital accumulation as a driver of economic growth. Environmental regulation similarly exerts directional pressures, channeling resources toward green technology development and fostering positive externalities through technological diffusion. This facilitates green and efficiency-oriented transitions, thereby elevating industrialization levels. Furthermore, environmental regulation increases pollution emission costs for manufacturers. Within industrial chains, this compels firms to adopt energy-saving practices, phase out inefficient capacities, and optimize resource allocation. Across industries, it triggers the decline of high-pollution, low-value-added sectors and the expansion of high-tech green industries, upgrading industrialization structures toward high-value-added and low-pollution paradigms. Such advancements in industrialization levels further enhance industrial chain greenization through optimized resource allocation and industrial restructuring.

Consequently, this study proposes Hypothesis H4: Environmental regulation elevates industrialization levels, thereby exerting a positive influence on the greenization of manufacturing industrial chains.

3. Data and Methods

3.1. Model Specification

3.1.1. Benchmark Regression Model

To test Hypothesis H1, this study establishes a two-way fixed effects regression model as follows (Equation (1)):

$${\mathbf{I}\mathbf{C}\mathbf{G}}_{\mathbf{i}\mathbf{t}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathbf{E}\mathbf{R}}_{\mathbf{i}\mathbf{t}}+{\sum }_{\mathbf{m}=1}^{\mathbf{n}}{\mathsf{\theta }}_{\mathbf{m}}{\mathbf{C}\mathbf{o}\mathbf{n}\mathbf{t}\mathbf{r}\mathbf{o}\mathbf{l}}_{\mathbf{i}\mathbf{t}}+{\mathsf{\mu }}_{\mathbf{i}}+{\mathsf{\delta }}_{\mathbf{t}}+{\mathsf{\epsilon }}_{\mathbf{i}\mathbf{t}}$$

(1)

In this equation, ICG represents the dependent variable, namely the greenization level of manufacturing industrial chains; ER denotes the core explanatory variable, environmental regulation; Control refers to control variables related to the dependent variable; β signifies the coefficient capturing the impact of environmental regulation on industrial chain greenization; μi and λt represent individual and time fixed effects, respectively; and ϵit is the random error term.

3.1.2. Spatial Autocorrelation Model

The premise of spatial econometric analysis lies in the presence of spatial autocorrelation among variables. Spatial autocorrelation refers to the tendency of adjacent geographic regions to exhibit similar attribute values. Positive spatial autocorrelation occurs when high-value regions cluster near other high-value regions or low-value regions cluster near other low-value regions. Conversely, negative spatial autocorrelation arises when high-value regions neighbor low-value regions. A random spatial distribution of high and low values indicates no autocorrelation. This study employs Moran’s I index to measure spatial autocorrelation, calculated as

$${\mathbf{M}\mathbf{o}\mathbf{r}\mathbf{a}\mathbf{n}}^{\mathbf{’}}\mathbf{s} \mathbf{I}={\displaystyle \frac{\sum _{\mathbf{i}=1}^{\mathbf{n}}\sum _{\mathbf{j}=1}^{\mathbf{n}}{\mathbf{w}}_{\mathbf{i}\mathbf{j}}\left({\mathbf{x}}_{\mathbf{i}}-\overline{\mathbf{x}}\right)\left({\mathbf{x}}_{\mathbf{j}}-\overline{\mathbf{x}}\right)}{{\mathbf{S}}^2\sum _{\mathbf{i}=1}^{\mathbf{n}}\sum _{\mathbf{j}=1}^{\mathbf{n}}{\mathbf{w}}_{\mathbf{i}\mathbf{j}}}}$$

Here, ${\mathbf{S}}^2$ denotes the sample variance, $\overline{\mathbf{x}}$ is the mean of nn observations, n represents the total number of provinces, ${\mathbf{x}}_{\mathbf{i}}$ is the observed value for the ii-th province, and w is the spatial weight matrix. Moran’s I ranges between −1 and 1, with larger absolute values indicating stronger spatial correlations.

3.1.3. Spatial Econometric Model

To investigate spatial interdependencies among variables, this study constructs spatial econometric models, including the Spatial Lag Model (SAR), Spatial Error Model (SEM), and Spatial Durbin Model (SDM). The optimal model is selected based on Lagrange Multiplier (LM) tests, Wald tests, and Likelihood Ratio (LR) tests. Building on the previous equation, spatial interaction terms are incorporated to formulate the SDM, SAR, and SEM as follows:

Spatial Durbin Model (SDM):

$$\begin{array}{cc}{\mathit{I}\mathit{C}\mathit{G}}_{\mathit{i}\mathit{t}}={\mathsf{\gamma }}_{\mathsf{\alpha }0}\hfill & +{\mathsf{\rho }}_1{\displaystyle \sum _{\mathit{i}=1}^{\mathit{n}}}{\mathit{W}}_{\mathit{i}\mathit{j}}{\mathit{I}\mathit{C}\mathit{G}}_{\mathit{i}\mathit{t}}+{\mathsf{\gamma }}_{\mathsf{\alpha }1}{\mathit{E}\mathit{R}}_{\mathit{i}\mathit{t}}+{\mathsf{\rho }}_2{\displaystyle \sum _{\mathit{i}=1}^{\mathit{n}}}{\mathit{W}}_{\mathit{i}\mathit{j}}{\mathit{E}\mathit{R}}_{\mathit{i}\mathit{t}}+{\mathsf{\gamma }}_{\mathsf{\alpha }2}{\mathit{C}\mathit{o}\mathit{n}\mathit{t}\mathit{r}\mathit{o}\mathit{l}}_{\mathit{i}\mathit{t}}+{\mathsf{\mu }}_{\mathit{i}}+{\mathsf{\lambda }}_{\mathit{t}}\\ & +{\mathsf{\epsilon }}_{\mathit{i}\mathit{t}}\hfill \end{array}$$

Spatial Lag Model (SAR):

$${\mathit{I}\mathit{C}\mathit{G}}_{\mathit{i}\mathit{t}}={\mathsf{\gamma }}_{\mathsf{\beta }0}+{\mathsf{\rho }}_1\sum _{\mathit{i}=1}^{\mathit{n}}{\mathit{W}}_{\mathit{i}\mathit{j}}{\mathit{E}\mathit{R}}_{\mathit{i}\mathit{t}}+{\mathsf{\gamma }}_{\mathsf{\beta }1}{\mathit{E}\mathit{R}}_{\mathit{i}\mathit{t}}+{\mathsf{\gamma }}_{\mathsf{\beta }2}{\mathit{C}\mathit{o}\mathit{n}\mathit{t}\mathit{r}\mathit{o}\mathit{l}}_{\mathit{i}\mathit{t}}+{\mathsf{\mu }}_{\mathit{i}}+{\mathsf{\lambda }}_{\mathit{t}}+{\mathsf{\epsilon }}_{\mathit{i}\mathit{t}}$$

Spatial Error Model (SEM):

$${\mathit{I}\mathit{C}\mathit{G}}_{\mathit{i}\mathit{t}}={\mathsf{\gamma }}_{\mathsf{\sigma }0}+{\mathsf{\rho }}_1\sum _{\mathit{i}=1}^{\mathit{n}}{\mathit{W}}_{\mathit{i}\mathit{j}}{\mathsf{\varphi }}_{\mathit{i}\mathit{t}}+{\mathsf{\gamma }}_{\mathsf{\sigma }1}{\mathit{E}\mathit{R}}_{\mathit{i}\mathit{t}}+{\mathsf{\gamma }}_{\mathsf{\sigma }2}{\mathit{C}\mathit{o}\mathit{n}\mathit{t}\mathit{r}\mathit{o}\mathit{l}}_{\mathit{i}\mathit{t}}+{\mathsf{\mu }}_{\mathit{i}}+{\mathsf{\lambda }}_{\mathit{t}}+{\mathsf{\epsilon }}_{\mathit{i}\mathit{t}}$$

In these equations, γ denotes the estimated coefficients, while ${\rho }_1$ represent the spatial autoregressive coefficients. $W_{ij}$ stands for the spatial weight matrix, and φ refers to the spatial disturbance term. All other variables align with those defined in Equation (1). Furthermore, this study employs a geographic inverse-distance-squared weight matrix, constructed by first calculating the squared distances between provincial centroids and then taking their reciprocals.

3.2. Variable Selection

3.2.1. Dependent Variable

The dependent variable in this study is the greenization level of manufacturing industrial chains (ICG). The concept of industrial chain greenization embodies the principles of sustainable development and contributes significantly to its advancement. This study constructs an evaluation framework focusing on three key aspects: the implementation of energy-saving measures, the control of pollutant emissions, and the establishment of green management systems. More specifically, the greenization level of industrial chains is assessed from two dimensions, end-of-pipe treatment and source reduction, with the latter encompassing clean production and circular economy practices. In the context of achieving carbon peak and carbon neutrality goals, promoting the greenization of China’s manufacturing industrial chains hinges on critical initiatives such as reducing industrial waste emissions, advancing clean production and circular economy models, and strengthening source-oriented environmental pollution control. Drawing on the measurement frameworks proposed by [14], this study evaluates ICG through two dimensions: source governance and end governance. The comprehensive provincial-level ICG index is calculated using the entropy method based on the raw data collected for these indicators (

Table 1. Construction of ICG.

Tier 1 Indicator	Tier 2 Indicator	Tier 3 Indicator	Polarity
Industrial Chain Greenization	Source Governance	Utilization rate of general industrial solid waste	Positive
		Logarithm of industrial pollution control investment	Positive
		Energy consumption intensity (per unit GDP)	Negative
	End Governance	Wastewater emissions per unit of industrial added value	Negative
		Exhaust gas emissions per unit of industrial added value	Negative
		Smoke and dust emissions per unit of industrial added value	Negative
		Energy consumption per unit of industrial added value	Negative

).

It is important to clarify that a key distinction between our construction of the ICG indicator and that of Wang et al. [14] lies in our use of the logarithm of industrial pollution control investment, rather than the logarithm of total environmental pollution control investment. This decision is grounded in the fact that the core dependent variable of this study is the greenization level of the manufacturing industrial chain. The end-of-pipe governance dimension of this variable must accurately reflect pollution abatement efforts within the industrial sector.

Total environmental pollution control investment is a comprehensive macro-level indicator that includes expenditures from agriculture, domestic pollution sources, and industry. Its use would introduce significant noise unrelated to the industrial sector, thereby distorting the actual greening efforts of manufacturing. In contrast, industrial pollution control investment is precisely aligned with the research focus of this study, as it directly captures investments in pollution treatment specifically within the manufacturing sector.

3.2.2. Core Explanatory Variable

The core explanatory variable of this paper is Environmental Regulation (ER). The understanding of environmental regulation in academia has undergone a continuously deepening process. Initially, scholars defined the meaning of environmental regulation as direct government regulation, i.e., the government’s direct intervention in the utilization of environmental resources primarily through administrative commands. As academia’s understanding of information asymmetry theory continuously deepened, Kathuria (2007) pointed out that besides formal environmental regulation [15], there are many other informal regulatory instruments that can influence pollution control behavior of polluting firms because polluting firms are very sensitive to their social reputation and potential future cost increases due to pollution incidents.

Pargal & Wheeler (1996) believed that when formal environmental regulations implemented by the government are absent or weak, many groups will negotiate or consult with local polluting firms to promote the achievement of pollution reduction [16]. This phenomenon is called “informal regulation”, namely the behavior of social groups pursuing higher environmental quality based on their own interests.

Currently, the main methods for measuring formal environmental regulation variables are as follows: (1) Using the proportion of pollution control investment to total enterprise costs or total output value as a proxy variable to measure the degree of compliance with formal environmental regulation by economic entities [17]. (2) Using operating costs of pollution control facilities, or per capita operating costs, to measure [18]. (3) Using changes in pollution emissions or pollution emission intensity per unit output value under formal environmental regulation to measure [19].

Methods for measuring informal environmental regulation variables also include (1) using media exposure rate of pollution incidents to measure informal regulation intensity [15] and (2) using voter turnout in parliamentary elections and growth rate of education level as proxy variables for informal regulation intensity [20]. However, since government management systems, education levels, laws, media coverage rates, employment situations, places of residence, etc., are all important factors affecting the level of public environmental awareness, they are not only complex but also difficult to express quantitatively. Therefore, using a single indicator to measure informal regulation intensity is overly one-sided and inaccurate.

Given the research theme of this paper, the environmental regulation variable construction has a strong correlation with the measurement indicators for the greenization level of the manufacturing industry chain, exhibiting obvious endogeneity issues. Furthermore, environmental regulation takes various forms such as administrative commands and economic constraints, and the above indicators may be unable to fully reflect comprehensive government governance policies. Therefore, this paper draws on the method of Chen et al. [21], utilizing Python 3.13 to perform word segmentation processing on government work reports. It separately counts the word frequency of keywords related to environmental regulation in provincial government work reports in China. Identified vocabulary includes environmental protection, environmental protection, energy consumption, emission reduction, pollutant discharge, ecology, green, low-carbon, air, chemical oxygen demand, sulfur dioxide, PM10, PM2.5. The ratio of the total word frequency of these words to the total word count of the provincial government work report is used as the proxy variable for environmental regulation. This constructed proxy variable can comprehensively reflect the intensity of environmental governance by provincial governments. On the other hand, since local government work reports are generally released at the beginning of the year, the economic activities they plan will be implemented throughout the year, thereby alleviating endogeneity issues to a certain extent [22].

3.2.3. Control Variables

To mitigate the potential impact of omitted explanatory variables on the dependent variable [22,23], this study selects the following control variables that influence the greenization level of manufacturing industrial chains: (1) Population Density (PD), measured as the ratio of the resident population to the administrative area, with logarithmic transformation applied; (2) Human Capital Level (Hum), quantified by the proportion of undergraduate and college students to the regional resident population; (3) Openness to Globalization (Openness), calculated as the ratio of total import-export volume to regional GDP; (4) Research and Development Intensity (RDI), defined as the ratio of internal R&D expenditures to regional GDP; and (5) Informationization Level (Inf), measured as the ratio of total postal service volume to regional GDP. The detailed variable definitions are summarized in

Table 2. Construction of ER.

Variable Type	Variable Name	Variable Definition	Measurement Method
Dependent Variable	ICG	Greenization level of manufacturing industrial chains	Composite index integrating source governance (e.g., solid waste utilization) and end governance (e.g., emission intensity)
Core Explanatory Variable	ER	Environmental regulation	Ratio of environmental policy-related term frequency to total word count in provincial government work reports (textual analysis)
Control Variable	PD	Population density	Logarithm of resident population per unit administrative area
	Hum	Human capital level	Ratio of undergraduate/college enrollments to regional resident population
	Openness	Openness to globalization	Ratio of total import-export volume to regional GDP
	RDI	Research and development intensity	Ratio of internal R&D expenditures to regional GDP
	Inf	Informationization level	Ratio of postal service volume to regional GDP (proxy for digital infrastructure)

.

3.3. Data Sources and Descriptive Statistics

Based on the variable definitions, this study utilizes panel data from 30 Chinese provinces (excluding Tibet, Hong Kong, Macao, and Taiwan) spanning 2010–2022. The data are sourced from authoritative national statistical yearbooks, including the China Industrial Statistical Yearbook, China Labor Statistical Yearbook, China Energy Statistical Yearbook, and China Statistical Yearbook. Missing values were addressed using interpolation methods. A VIF value of 1.88 indicates that there is no serious multicollinearity problem in the model. Descriptive statistics for all variables are presented in

Table 3. Descriptive statistics.

Variable	Obs	Mean	Std. dev	Min	Max
ICG	390	0.651	0.167	0.192	0.959
ER	390	0.003	0.001	0.001	0.006
lnPD	390	5.463	1.285	2.053	8.282
Hum	390	0.021	0.014	0.006	0.277
Openness	390	0.276	0.291	0.008	1.464
RDI	390	0.018	0.011	0.003	0.068
Inf	390	0.06	0.052	0.015	0.29

.

3.4. Data Analysis

To comprehensively depict the temporal evolution characteristics of environmental regulation in China, this study employs the kernel density estimation method and constructs a three-dimensional kernel density estimation graph based on annual data. As shown in Figure 1, the distribution of environmental regulation in China has undergone profound changes over the past decade. The kernel density curve overall exhibits a fluctuating upward trend. It evolved from a low-level, decentralized state in the early 2010s to a high-level, highly concentrated state in the mid-term, demonstrating strong policy effectiveness and implementation convergence. However, around 2020, the distribution began to show new characteristics: a slight decrease in peak height, a broadening morphology, and initial signs of divergence, which may indicate that environmental regulation is entering a new phase, transitioning from a unimodal to a bimodal distribution.

From 2010 to 2014, which marks the early stage of environmental regulation in China, the overall pattern showed a rightward shift in the peak, with the morphology transitioning from steep to flat, while the height remained largely unchanged. This suggests that during this period, China’s environmental regulation was in a phase of steady development, with little change in regional disparities. Notably, between 2010 and 2012, the peak shifted significantly to the right, likely due to the initiation of China’s Ambient Air Quality Standards in 2011 and its formal promulgation in 2012. The new standards included PM2.5 concentration limits for the first time, compelling local governments to intensify air pollution control efforts and significantly enhance the stringency of environmental regulation. From 2014 to 2018, the peak continued to shift rightward amid fluctuations, while the morphology became noticeably flatter, even showing a preliminary bimodal pattern in 2018. This indicates that although environmental regulation in China continued to improve overall, interprovincial disparities increased significantly. Finally, from 2018 to 2022, the peak shifted leftward in this stage, suggesting a temporary relaxation regulatory intensity, while the right tail showed signs of expanding again, possibly indicating the emergence of a new divergence trend. This may be due to differentiated policies, economic pressures, or new development priorities, leading some regions to move toward higher standards.

Overall, the distribution of environmental regulation intensity in China has experienced profound evolution over the past decade. It transitioned from a low-level, decentralized state in the early 2010s to a high-level, highly concentrated state in the mid-term, reflecting strong policy effectiveness and implementation convergence. However, after 2020, the distribution began to exhibit new features: a slight decrease in peak height, a broadening morphology, and initial signs of divergence. This may signal that environmental policy is entering a new phase, shifting from the forceful implementation of “one-size-fits-all” measures to a more complex and potentially more differentiated new framework.

4. Regression Results and Analysis

4.1. Analysis of Baseline Regression Results

This study first conducted Ordinary Least Squares (OLS) regression, with results reported in Column (1) of

Table 4. Baseline regression results.

	OLS	Two-Way Fixed Effects
	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Variables	ICG	ICG	ICG	ICG	ICG	ICG	ICG
ER	11.33 **	9.474 *	10.32 **	9.759 **	9.303 **	8.834 *	8.249 *
	(5.353)	(4.850)	(4.864)	(4.800)	(4.717)	(4.749)	(4.696)
lnPD	0.0883 ***		0.295 ***	0.290 ***	0.337 ***	0.366 ***	0.344 ***
	(0.00512)		(0.0940)	(0.0949)	(0.0916)	(0.102)	(0.103)
Hum	−0.516			−0.638 ***	−0.549 ***	−0.535 ***	−0.451 ***
	(0.414)			(0.124)	(0.130)	(0.126)	(0.119)
Openness	0.0799 ***				0.0951 **	0.0923 **	0.0792 *
	(0.0280)				(0.0412)	(0.0404)	(0.0404)
RDI	0.716					−1.327	−1.233
	(0.898)					(1.348)	(1.343)
Inf	−0.226 **						0.422 **
	(0.102)						(0.165)
Individual-FE	NO	YES	YES	YES	YES	YES	YES
Time-FE	NO	YES	YES	YES	YES	YES	YES
Constant	0.123 ***	0.621 ***	−0.995 *	−0.949 *	−1.234 **	−1.369 **	−1.271 **
	(0.0301)	(0.0153)	(0.518)	(0.523)	(0.506)	(0.549)	(0.554)
Observation	390	390	390	390	390	390	390
R-squared	0.649	0.914	0.917	0.919	0.921	0.921	0.923

Note: t-statistics in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

. Subsequently, a two-way fixed effects model incorporating control variables was applied through stepwise regression, with outcomes documented in Columns (2) to (7).

As shown in Table 4, environmental regulation exhibits a statistically significant and positive impact on the greenization level of manufacturing industrial chains across all specifications—whether control variables are included or excluded, and regardless of the estimation method. These results robustly confirm Hypothesis H1.

4.2. Analysis of Control Variables

Population Density (lnPD) remains significant at the 1% level throughout stepwise regressions. In Column (7), its coefficient of 0.344 indicates a positive association with industrial chain greenization. This may be attributed to the concentration of market demand in high-density regions, where economies of scale reduce per-unit resource consumption and pollution emissions. Additionally, densely populated areas prioritize environmental quality, driving investments in wastewater treatment and clean energy infrastructure, thereby enhancing greenization.

Human Capital Level (Hum) displays significant negative coefficients at the 1% level in the two-way fixed effects model. The coefficient of −0.344 in Column (7) suggests a suppressive effect. Two potential mechanisms explain this finding: (1) firms tend to allocate skilled labor to short-term profit-driven activities rather than long-cycle green technology R&D, potentially engaging in “policy arbitrage” rather than substantive innovation; (2) higher wage costs associated with skilled labor may crowd out investments in green technology or equipment upgrades, generating a negative “crowding-out effect”.

Openness to Globalization (Openness): The correlation coefficient is 0.0792 and is statistically significant at the 10% level. Consistently positive and significant coefficients imply that international engagement facilitates industrial chain greenization. This likely stems from technology transfer and compliance with stringent global environmental standards, which incentivize export-oriented firms to adopt greener practices.

R&D Intensity (RDI) is statistically insignificant across specifications. This may reflect firms prioritizing productivity-enhancing or product-oriented R&D over green technology development, thereby diluting the impact of R&D intensity on greenization.

Informationization Level (Inf): Positive significance at the 5% level and the correlation coefficient of 0.422 highlight its catalytic role. Enhanced digital infrastructure enables cross-regional diffusion of green technologies and improves real-time monitoring of carbon emissions and industrial waste during production.

4.3. Robustness Test

To ensure the reliability of findings, this study conducts four robustness checks: (1) Winsorization: The dataset was winsorized at the 1% level to mitigate the influence of extreme values. (2) Model Replacement: Given the continuous nature of the dependent variable, a Tobit model was employed as an alternative estimation strategy. (3) Additional Control Variable: Financial development level was incorporated into the model to address potential omitted variable bias. (4) The core explanatory variable was redefined as ERnew, calculated as the ratio of total environmental keyword frequency to the overall word frequency (rather than total word count) in provincial government work reports.

As demonstrated in

Table 5. Robustness test.

	Winsorization	Model Replacement	Additional Control Variable	Proxy Variable Replacement
VARIABLES	ICG	ICG	ICG	ICG
ER	10.25 **	8.249 **	8.497 *
	(4.496)	(4.012)	(4.676)
ERnew				2.794 *
				(1.588)
lnPD	0.332 ***	0.344 ***	0.369 ***	0.343 ***
	(0.102)	(0.0862)	(0.115)	(0.102)
Hum	−0.306	−0.451 **	−0.448 ***	−0.450 ***
	(0.859)	(0.201)	(0.119)	(0.119)
Openness	0.0934 **	0.0792 **	0.0822 **	0.0798 **
	(0.0426)	(0.0340)	(0.0413)	(0.0405)
RDI	−1.063	−1.233	−1.417	−1.246
	(1.285)	(1.218)	(1.448)	(1.345)
Inf	0.460 **	0.422 ***	0.414 **	0.422 **
	(0.182)	(0.141)	(0.166)	(0.165)
var(e.ICG)		0.00216 ***
		(0.000155)
FD			0.00687
			(0.0126)
Constant	−1.224 **	−1.769 ***	−1.430 **	−1.262 **
	(0.554)	(0.606)	(0.639)	(0.553)
Observations	390	390	390	390

Note: t-statistics in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

, the results of these robustness tests align with the baseline regression outcomes. Environmental regulation consistently exhibits a statistically significant positive effect on the greenization level of manufacturing industrial chains, confirming the robustness of the model. The direction and significance of coefficients for control variables remain stable across all specifications, further validating the reliability of the findings.

4.4. Endogeneity Test

To thoroughly mitigate potential endogeneity issues arising from reverse causality between environmental regulation and industrial chain greenization, this study replaces the core explanatory variable with its first-order lagged term for regression. Column (2) of

Table 6. Endogeneity test.

VARIABLES	ICG	ICG
L.ER	13.42 ***	12.14 **
	(4.944)	(4.691)
lnPD		0.314 ***
		(0.105)
Hum		−0.536 ***
		(0.129)
Openness		0.0745 *
		(0.0449)
RDI		−0.352
		(1.277)
Inf		0.361 **
		(0.158)
Constant	0.611 ***	−1.123 *
	(0.0157)	(0.571)
Observations	360	360
R-squared	0.920	0.927

Note: t-statistics in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

reports the estimation results. It can be observed that the direction of the lagged term of environmental regulation remains consistent with the original regression, whether control variables are included or not, and it is statistically significant at the 1% and 5% levels, respectively. This result indicates that the positive promoting effect of environmental regulation on industrial chain greenization remains robust even after controlling for the possibility of reverse causality. Furthermore, the signs and significance of other control variables are largely consistent with the baseline model, further demonstrating the reliability of the findings of this study.

To mitigate potential reverse causality, we employ a one-period lag of the core explanatory variable. A series of robustness checks are implemented, including Winsorizing continuous variables at the 1% level, substituting estimation models, incorporating additional control variables, and replacing the measurement of the core independent variable. Furthermore, the level of financial development is included in the model to control for potential omitted variable bias.

4.5. Heterogeneity Analysis

4.5.1. Heterogeneity in Energy Consumption Structure

Environmental regulation may exert heterogeneous effects on the greenization level of manufacturing industrial chains across regions with varying energy consumption structures [23]. This study measures energy consumption structure as the ratio of coal consumption to total regional energy consumption. A higher value indicates greater reliance on coal, a high-carbon-emission energy source. Using the 2010–2022 full-sample data, provinces were categorized into two groups based on their average coal dependency relative to the national mean: Group (1) coal-dependent regions and Group (2) less coal-dependent regions [24,25].

As shown in

Table 7. Heterogeneity in energy consumption structure.

	(1)		(2)
VARIABLES	ICG	ICG	ICG	ICG
ER	27.93 ***	18.73 **	4.775	0.449
	(9.957)	(7.690)	(5.562)	(5.463)
lnPD		0.564 ***		−0.193
		(0.117)		(0.152)
Hum		2.302 *		−0.540 ***
		(1.372)		(0.125)
Openness		0.0819		0.0547
		(0.144)		(0.0462)
RDI		2.931		−2.016
		(2.390)		(1.663)
Inf		0.250		0.446 **
		(0.428)		(0.178)
Constant	0.511 ***	−2.364 ***	0.669 ***	1.799 **
	(0.0290)	(0.576)	(0.0185)	(0.862)
Observations	143	143	247	247
R-squared	0.908	0.930	0.913	0.924

Note: t-statistics in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

, environmental regulation significantly enhances manufacturing industrial chain greenization in coal-dependent regions (Group 1), whereas this effect is statistically insignificant in less coal-dependent regions (Group 2). This divergence arises because coal-reliant regions face more severe environmental challenges, such as high carbon emissions and air pollution. In these areas, environmental regulation directly increases corporate operating costs, compelling manufacturers to adopt cleaner energy sources to mitigate cost pressures, thereby driving systemic greenization. In contrast, regions with lower coal dependency already exhibit broader adoption of clean energy, reducing the marginal impact of environmental regulation on greenization outcomes.

4.5.2. Heterogeneity in Market Value Share of Energy-Intensive Industries

The market value share of energy-intensive industries reflects regional industrial structures to some extent. This study categorizes provinces into two subsamples based on their average market value share of energy-intensive industries relative to the full-sample mean: high-share regions and low-share regions. Regression results in

Table 8. Heterogeneity in market value share of energy-intensive industries.

	High-Share Regions		Low-Share Regions
VARIABLES	ICG	ICG	ICG	ICG
ER	−2.511	−5.412	14.07 **	13.28 **
	(6.518)	(6.334)	(6.988)	(6.149)
lnPD		−0.177		0.472 ***
		(0.166)		(0.115)
Hum		−3.829 **		−0.413 ***
		(1.762)		(0.132)
Openness		−0.0323		0.0765 *
		(0.175)		(0.0443)
RDI		−2.947		−0.786
		(2.413)		(1.666)
Inf		0.362 *		0.688 ***
		(0.208)		(0.237)
Constant	0.533 ***	1.404 *	0.671 ***	−2.198 ***
	(0.0220)	(0.738)	(0.0215)	(0.672)
Observations	130	130	260	260
R-squared	0.924	0.931	0.869	0.890

Note: t-statistics in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

reveal that environmental regulation significantly enhances the greenization level of manufacturing industrial chains in low-share regions, while its effect is statistically insignificant in high-share regions.

This divergence may stem from two factors. In low-share regions, industrial structures are predominantly characterized by light industries or high-tech sectors, where rapid green technology innovation and application amplify the positive impact of environmental regulation. Conversely, high-share regions likely face dual challenges of developmental path dependency and industrial structure rigidity, where entrenched production paradigms and institutional inertia diminish the efficacy of regulatory interventions.

4.5.3. Heterogeneity in Digital Economy Development Level

As an emerging technological and economic paradigm, the digital economy may moderate the impact of environmental regulation on manufacturing industrial chain greenization depending on its developmental maturity. Drawing on the measurement framework by He et al. [22], specific indicator construction methods are detailed in the Appendix A

Table A1. Evaluation index system for the development level of the digital economy.

Level 1 Indicator	Level 2 Indicator	Level 3 Indicator	Unit
Digital Infrastructure	Hardware Facilities	Long-distance Optical Cable Line Length	10,000 km
		Internet Broadband Access Ports	10,000 units
		Mobile Phone Base Stations	10,000 units
	Software Facilities	Number of Internet Domain Names	10,000 units
		Number of IPv4 Addresses	10,000 units
		Number of Internet Websites	10,000 units
Digital Industry Development	Digital Industrialization	Software Business Revenue	100 million yuan
		Telecom Business Volume	100 million yuan
		Number of Electronic Information Manufacturing Enterprises	unit
	Industrial Digitalization	Number of Websites per 100 Enterprises	unit
		Proportion of Enterprises with E-commerce Transaction Activities	%
		E-commerce Sales Volume	100 million yuan
		Number of Computers Used per 100 People	unit
Digital Economy Environment	Application Environment	Mobile Internet Users	10,000 households
		Mobile Phone Users	10,000 households
		Digital Telephone Users	10,000 households
	Talent Environment	Proportion of Information-related Employees in Total Employment	%
		Number of Undergraduate Graduates	person
	Innovation Environment	R&D Personnel (Full-time equivalent)	person-year
		Number of R&D Institutions	unit
		Patents Granted	unit

, this study evaluates digital economy development through three dimensions—digital infrastructure, digital industry growth, and digital economic environment—and categorizes provinces into high-level and low-level subsamples [26,27].

As shown in

Table 9. Heterogeneity in digital economy development level.

	High Digital Economy		Low Digital Economy
VARIABLES	ICG	ICG	ICG	ICG
ER	2.099	0.910	13.20 **	10.75 *
	(8.121)	(7.826)	(6.259)	(5.846)
lnPD		−0.335		0.507 ***
		(0.227)		(0.120)
Hum		−0.595 ***		−0.358
		(0.147)		(0.891)
Openness		0.0136		5.29 × 10−5
		(0.0702)		(0.0676)
RDI		2.491		−1.566
		(3.139)		(1.540)
Inf		0.876 *		0.324
		(0.446)		(0.216)
Constant	0.756 ***	2.783 *	0.535 ***	−1.921 ***
	(0.0254)	(1.446)	(0.0199)	(0.579)
Observations	156	156	234	234
R-squared	0.854	0.873	0.898	0.912

Note: t-statistics in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

, environmental regulation significantly enhances industrial chain greenization in regions with low digital economic development, while its effect is statistically insignificant in high-level regions [28]. This divergence arises because low-development regions have yet to fully integrate advanced digital technologies into green governance. In these areas, regulatory pressures and cost increases compel firms to adopt digital solutions for emission reduction and efficiency gains, amplifying the marginal benefits of environmental regulation. Conversely, in highly developed regions, digital technologies are already deeply embedded in green governance systems. Further regulatory constraints face diminishing returns due to saturation effects and escalating transition costs, thereby attenuating their impact on greenization.

4.6. Mechanism Tests

To validate Hypotheses H2, H3a, H3b, and H4, this study employs a mediation effect model. The model specification is as follows:

$${\mathrm{M}\mathrm{e}\mathrm{d}\mathrm{i}\mathrm{a}\mathrm{t}\mathrm{e}}_{\mathrm{i}\mathrm{t}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1{\mathrm{E}\mathrm{R}}_{\mathrm{i}\mathrm{t}}+{\sum }_{\mathrm{m}=1}^{\mathrm{n}}{\mathsf{\theta }}_{\mathrm{m}}{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\delta }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}}$$

Hypothesis H2 on Industrial Structure Rationalization: The industrial structure rationalization index TL is constructed using the Theil index, which quantifies sectoral deviations in output and employment while weighing their economic significance. Industrial structure rationalization refers to the aggregation quality between industries. On one hand, it reflects the degree of coordination between industries; on the other hand, it should also reflect the degree of effective utilization of resources. It is a measure of the coupling degree between factor input structure and output structure. Regarding this coupling, researchers generally use the structural deviation degree to measure industrial structure rationalization. Its formula is

$$E=\sum _{i=1}^n \left| {\displaystyle \frac{\raisebox{1ex}{$Y_i$}\!\left/ \!\raisebox{-1ex}{$L_i$}\right.}{\raisebox{1ex}{$Y$}\!\left/ \!\raisebox{-1ex}{$L$}\right.}}-1 \right|=\sum _{i=1}^n \left| {\displaystyle \frac{\raisebox{1ex}{$Y_i$}\!\left/ \!\raisebox{-1ex}{$Y$}\right.}{\raisebox{1ex}{$L_i$}\!\left/ \!\raisebox{-1ex}{$L$}\right.}}-1 \right|$$

In the formula, E represents the structural deviation degree, Y represents output value, L represents employment, i represents industry, and n represents the number of industrial sectors. According to classical economic assumptions, the economy ultimately reaches an equilibrium state where productivity levels are identical across all industrial sectors. Y/L represents productivity; therefore, when the economy is in equilibrium, Yᵢ/Lᵢ = Y/L, and thus E = 0. Simultaneously, Yᵢ/Y represents the output structure, and Lᵢ/L represents the employment structure; therefore, E also reflects the coupling between the output structure and employment structure. A larger E value indicates a greater deviation of the economy from the equilibrium state and a more irrational industrial structure. Since economic disequilibrium is a normal phenomenon, particularly more pronounced in developing countries, it is impossible for the E value to be 0.

However, the structural deviation degree indicator treats all industries “equally,” ignoring the relative importance of each industry within the economy. Additionally, the calculation using absolute values creates inconvenience for research. For this reason, Gan et al. (2011) introduced the Theil index and redefined it [23]. Its calculation formula is as follows:

$$TL=\sum _{i=1}^n\left( {\displaystyle \frac{Y_i}{Y}} \right) \ln \left( {\displaystyle \raisebox{1ex}{$\frac{Y_i}{L_i} $}\!\left/ \!\raisebox{-1ex}{$ \frac{Y}{L}$}\right.} \right)$$

Similarly, if the economy is in an equilibrium state, there is also TL = 0. Moreover, this index considers the relative importance of industries and avoids the calculation of absolute values. Simultaneously, it retains the theoretical foundation and economic meaning of the structural deviation degree. Therefore, it is a better measure of industrial structure rationalization. A Theil index not equal to 0 indicates that the industrial structure deviates from the equilibrium state, signifying that the industrial

The Theil index is inversely coded, where a negative coefficient signifies improved rationalization. Columns (1) and (2) in

Table 10. Mechanism tests.

	(1)	(2)	(3)	(4)	(5)	(6)
VARIABLES	TL	TL	GP	GP	ID	ID
ER	−10.29 *	−8.459 *	44.14 **	38.63 **	8.582 ***	8.334 ***
	(5.413)	(5.074)	(18.90)	(18.44)	(2.069)	(1.924)
lnPD		−0.292 **		2.711 ***		0.225 ***
		(0.125)		(0.434)		(0.0642)
Hum		−0.263 **		−0.176		0.143 ***
		(0.106)		(0.341)		(0.0479)
Openness		−0.220 ***		0.509 ***		−0.0109
		(0.0388)		(0.135)		(0.0142)
RDI		0.659		−15.17 ***		−2.571 ***
		(1.242)		(5.555)		(0.656)
Inf		−0.614 ***		1.645 ***		−0.0279
		(0.178)		(0.570)		(0.0594)
Constant	0.209 ***	1.891 ***	7.291 ***	−7.468 ***	0.311 ***	−0.870 **
	(0.0174)	(0.686)	(0.0606)	(2.367)	(0.00632)	(0.350)
Individual FE	YES	YES	YES	YES	YES	YES
Time FE	YES	YES	YES	YES	YES	YES
Observations	390	390	390	390	390	390
R-squared	0.820	0.844	0.981	0.984	0.926	0.934

Note: t-statistics in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

show that environmental regulation significantly reduces TL with coefficients of −10.29 and −8.459, both statistically significant at the 10% level. This confirms that environmental regulation enhances industrial rationalization by phasing out energy-intensive sectors and reallocating resources to greener industries.

Competing Hypotheses H3a and H3b on Green Technology Innovation: Green technology innovation GP is measured as the logarithm of green patent grants. Columns 3 and 4 reveal that environmental regulation significantly increases GP with coefficients of 44.14 and 38.63, significant at the 5% level. These results support Hypothesis H3a, indicating that internalizing pollution costs drives firms to adopt green technologies, consistent with the Porter Hypothesis.

Hypothesis H4 on Industrialization Level: Industrialization level ID is measured as the ratio of industrial added value to regional GDP. Columns 5 and 6 demonstrate that environmental regulation significantly boosts ID with coefficients of 8.582 and 8.334, both significant at the 1% level. This validates Hypothesis H4, as regulation promotes structural shifts toward high-value green industries and improves resource efficiency.

5. Analysis of Spatial Spillover Effects

5.1. Spatial Correlation Test

To examine the spatial interdependence between environmental regulation and the greenization level of manufacturing industrial chains, we conducted spatial autocorrelation tests on provincial industrial chain greenization (ICG).

Table 11. Global Moran’s I test.

Year	I	E(I)	Sd(I)	Z	p-Value
2010	0.240	−0.035	0.076	3.633	0.000
2011	0.213	−0.035	0.076	3.249	0.001
2012	0.180	−0.035	0.076	2.816	0.005
2013	0.178	−0.035	0.076	2.786	0.005
2014	0.182	−0.035	0.077	2.826	0.005
2015	0.235	−0.035	0.076	3.532	0.000
2016	0.215	−0.035	0.076	3.279	0.001
2017	0.292	−0.035	0.076	4.281	0.000
2018	0.320	−0.035	0.076	4.685	0.000
2019	0.278	−0.035	0.075	4.144	0.000
2020	0.260	−0.035	0.076	3.853	0.000
2021	0.277	−0.035	0.076	4.064	0.000
2022	0.263	−0.035	0.077	3.865	0.000

presents the global Moran’s I test results for 30 provinces from 2010 to 2022.

The Moran’s I values are consistently positive and statistically significant at the 1% level throughout the observation period, indicating robust positive spatial autocorrelation in manufacturing industrial chain greenization under the inverse distance-squared spatial weight matrix.

To delve into the spatial correlation of ICG across provinces, based on a significant Global Moran’s I, this study utilized ArcMap 10.8 to plot LISA cluster maps for the years 2010, 2014, 2018, and 2022. As illustrated in the Figure 2, H-H represents high–high clustering, L-H denotes low–high clustering, L-L signifies low–low clustering, H-L indicates high–low clustering, the yellow areas represent regions where the local Moran’s I is not significant, and the blank area corresponds to Tibet, which was excluded from the study sample due to data unavailability.

From the figures, it can be observed that on one hand, under the condition of a significant Global Moran’s I, the local Moran’s I is mostly insignificant, with only the eastern and western regions showing high and low clustering characteristics. This implies that two strong clusters drive the significance of the global index [29,30,31]. On the other hand, from 2010 to 2022, the LISA cluster maps did not undergo significant changes, while a relatively clear “club convergence” phenomenon emerged. The high–high clustering areas are primarily distributed in the eastern coastal regions, where economic development levels are higher, technological resources are abundant, and spatial spillover effects may exist, making them key drivers leading the green transformation in eastern China [32,33]. The low–low clustering areas are mainly concentrated in the western regions, where industrial transformation faces difficulties, funding is relatively scarce, and technological resources need further enhancement, resulting in a trend of low-value regional aggregation. The above analysis indicates that spatial effects spontaneously lead to club convergence where the strong become stronger and the weak become weaker, thereby resulting in regional imbalances in green development [34,35,36].

5.2. Spatial Weight Matrix and Econometric Model Selection

The preceding analysis reveals that the interaction between environmental regulation and the greenization level of manufacturing industrial chains is not randomly distributed spatially but exhibits significant positive spatial interdependence. Consequently, spatial factors must be integrated into the research framework to comprehensively capture their influence pathways and effect magnitudes [37,38]. Given that environmental regulation may drive firms to relocate to neighboring regions through the “pollution have effect” or trigger regional policy synergies, the inverse distance-squared spatial weight matrix is adopted to accurately characterize such short-distance spillover effects. Furthermore, the upstream and downstream linkages of manufacturing industrial chains and green technology spillovers are typically concentrated in adjacent regions [39,40]. The inverse distance-squared matrix, which rapidly attenuates spatial weights as distance increases through the squared term, aligns more closely with the practical characteristics of manufacturing industrial chains. Therefore, this study employs the inverse distance-squared matrix and selects the appropriate spatial econometric model through diagnostic tests.

The test results in

Table 12. Examination of spatial econometric model.

Type	LM-Error	Robust LM-Error	LM-Lag	Robust LM-Lag	Wald Spatial Error	Wald Spatial Lag	LR: SDM vs. SAR	LR: SDM vs. SEM
Statistic	22.375	31.113	19.907	28.644	30.38	30.63	29.17	29.49
p-value	0.000	0.000	0.000	0.000	0.0000	0.0000	0.0001	0.0000

indicate that the inverse distance-squared matrix passes both the Lagrange Multiplier (LM) and Robust LM tests. Further Likelihood Ratio (LR) and Wald tests demonstrate that all test statistics reject the null hypothesis at the 1% significance level, confirming that the Spatial Durbin Model (SDM) does not degenerate into the Spatial Autoregressive (SAR) or Spatial Error Model (SEM). Consequently, the SDM is selected. As the Hausman and LR tests indicate the superiority of the fixed effects model, this study employs a two-way fixed effects spatial model for estimation.

5.3. Spatial Econometric Results

Based on the selected inverse distance-squared spatial weight matrix and Spatial Durbin Model (SDM), this study conducts spatial econometric regression, with results presented in

Table 13. Spatial econometric results.

	Main	Wx
VARIABLES	ICG	ICG
ER	7.641 *	38.84 ***
	(3.938)	(13.50)
lnPD	0.259 **	0.959 ***
	(0.101)	(0.356)
Hum	−0.406 **	−0.413
	(0.201)	(0.335)
Openness	0.0518	−0.171 **
	(0.0355)	(0.0665)
RDI	−2.081 *	−7.757 *
	(1.227)	(4.177)
Inf	0.563 ***	−1.003 **
	(0.162)	(0.422)
Individual FE	YES	YES
Time FE	YES	YES
Observations	390	390
R-squared	0.567	0.567

Note: t-statistics in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

. The regression results reveal that the coefficient of environmental regulation (ER) on the greenization level of manufacturing industrial chains (ICG) remains positive, reaffirming Hypothesis H1. Under the inverse distance-squared matrix, the spatial lag coefficient of environmental regulation is significantly positive at the 1% level, indicating that environmental regulation exerts a robust positive spatial spillover effect on the greenization of manufacturing industrial chains.

To further investigate spatial spillover effects, this study decomposes the total effects into direct and indirect components. As shown in

Table 14. Total effects into direct and indirect.

	LR_Direct	LR_Indirect	LR_Total
VARIABLES	ICG	ICG	ICG
ER	7.192 *	33.97 ***	41.16 ***
	(4.082)	(12.74)	(13.70)
lnPD	0.241 **	0.816 **	1.057 ***
	(0.102)	(0.318)	(0.275)
Hum	−0.380 **	−0.291	−0.671 *
	(0.192)	(0.291)	(0.355)
Openness	0.0545	−0.155 **	−0.101
	(0.0347)	(0.0608)	(0.0622)
RDI	−1.945 *	−6.577 *	−8.523 **
	(1.178)	(3.610)	(3.836)
Inf	0.591 ***	−0.960 **	−0.369
	(0.163)	(0.395)	(0.337)
Individual FE	YES	YES	YES
Time FE	YES	YES	YES
Observations	390	390	390
R-squared	0.567	0.567	0.567

Note: t-statistics in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

, the direct effect of environmental regulation on manufacturing industrial chain greenization is significantly positive at the 10% level (coefficient: 7.192), while the indirect effect (spatial spillover) is significantly positive at the 1% level (coefficient: 33.97). These results confirm that environmental regulation generates substantial spatial spillover effects on industrial chain greenization.

This phenomenon can be attributed to two mechanisms. First, policy synergy effects counteract the “pollution haven effect.” Firms influenced by local environmental regulations may intend to relocate pollutive activities to neighboring regions [41]. However, coordinated environmental policies across regions compel these firms to abandon relocation strategies, thereby avoiding negative impacts on the greenization of external industrial chains [42]. Second, environmental regulation internalizes pollution costs, incentivizing local firms to develop green technologies. These technologies diffuse to adjacent regions through patent licensing and knowledge spillovers, enhancing the greenization of neighboring manufacturing industrial chains [43,44].

6. Discussion and Conclusions

6.1. Conclusions and Recommendations

The baseline regression confirms that environmental regulation significantly enhances the greenization level of manufacturing industrial chains. Heterogeneity analysis further reveals three key patterns: (1) Environmental regulation drives greenization in coal-dependent regions but shows no significant effect in regions with diversified energy structures. (2) Regions with low market shares of energy-intensive industries experience regulatory benefits, while high-share regions exhibit muted responses. (3) Regulatory impacts are significant in areas with underdeveloped digital economies but negligible in digitally advanced regions. Mechanism tests validate that environmental regulation promotes greenization through industrial structure rationalization, green technology innovation, and industrialization upgrades. Spatial analysis further identifies positive spillover effects, underscoring cross-regional synergies in green transitions.

Based on these findings, this study proposes three policy recommendations: First, intensify environmental regulation in regions with coal-dominated energy structures to accelerate green transformation [45,46]. Second, refine regulatory approaches in areas reliant on energy-intensive industries, balancing economic growth with dual carbon objectives through structural adjustments [47]. Third, enhance interregional policy coordination to mitigate pollution haven effects and ensure comprehensive regulatory coverage, thereby holistically elevating the greenization of manufacturing industrial chains. Future research should incorporate firm-level data, diverse regulatory instruments, and exogenous policy variables to deepen understanding of these dynamics [48,49,50].

The findings of this study also offer insights at both political and economic levels. Politically, the research confirms that strengthening environmental regulation is an effective policy tool for promoting the greening of the manufacturing industrial chain, providing a solid empirical foundation and a clear implementation pathway for the concept of sustainable development [51,52]. It demonstrates, from a micro-level mechanism, the synergistic feasibility of China’s dual-carbon strategy and high-quality development goals, helping to enhance international confidence in China’s green governance capabilities and strengthening its bargaining power in global climate negotiations [53,54]. Economically, the transmission mechanisms revealed by the study offer precise entry points for policymakers: by designing an incentive-compatible policy mix, it is possible not only to effectively avoid the short-term pains of “environmental protection constraining economic growth” but also to cultivate new quality productive forces centered on technological innovation and green supply chain management [55,56]. This is not only crucial for the sustainable upgrading of China’s industries but also provides valuable lessons for green industrial transformation in other countries.

6.2. Research Limitations

While this study advances understanding of environmental regulation’s role in manufacturing industrial chain greenization, several limitations warrant acknowledgment for future research. First, the reliance on provincial-level data precludes micro-level analysis of firm- or facility-specific regulatory impacts. Second, the analysis focuses on a single proxy for environmental regulation, neglecting potential variations across regulatory types. Third, exogenous policy shocks, such as national carbon trading initiatives or sudden regulatory reforms, are not fully accounted for, which may introduce omitted variable bias. Addressing these gaps could enhance the granularity and generalizability of findings. Fourth, in reality, inventive patents typically represent higher-quality innovative outcomes and substantive technological breakthroughs, whereas utility model patents often reflect incremental improvements, design adaptations, or functional optimizations. Relying solely on aggregate patent statistics may introduce measurement inaccuracies. However, due to constraints in data availability and reliability, it is currently not feasible to separately analyze these two types of patents in the regression model.