Has the Digital Economy Facilitated Regional Collaborative Carbon Reduction? A Complex Network Approach Toward Sustainable Development Goals

Abstract

The digital economy (DE) serves as a crucial engine for breaking through technological stagnation at the low end and achieving carbon neutrality. However, existing studies predominantly explore the impact of the DE on local carbon reduction based on “attribute data”, with less focus on regional carbon collaborative reduction. This study employs a directed-weighted complex network analysis, using provincial panel data from China spanning 2012 to 2022, to characterize the evolutionary features of China’s Inter-regional Collaborative Carbon Reduction Governance Network (ICCGN). Using the Exponential Random Graph Model (ERGM) as an empirical test, the study explores how the DE facilitates collaborative carbon reduction. The results indicate the following: (1) The ICCGN demonstrates transitive triadic linkages, accompanied by increasingly blurred governance boundaries. The Eastern coastal areas have the highest network centrality, and the network core areas, including Guangdong, Chongqing, Gansu, and Qinghai, are gradually expanding, leading to further weakening of governance boundaries. The network’s spatial clustering structure presents four distinct blocks, with network spillover relationships concentrated in the first, third, and fourth blocks. The Eastern coastal areas play a “hub” role in undertaking carbon collaborative reduction, radiating and driving the central and western provinces. (2) From the perspective of the induced effect, the DE enables carbon collaborative reduction, exhibiting isotropic characteristics. (3) Heterogeneity tests show that regions with well-developed digital infrastructure and those with free trade zone constructions promote better effects, with a positive feedback effect in network status: betweenness centrality > degree centrality > closeness centrality. (4) Regarding the enabling mechanism, the DE drives carbon collaborative governance by enhancing technological innovation, promoting industrial structure upgrades, nurturing scientific talents, and reducing educational disparities.

1. Introduction

Intensifying global climate pressures have made limiting greenhouse gases, especially CO₂, a shared objective worldwide [1]. Statistics for 2023 indicate that energy-related CO₂ emissions in China constituted nearly a third of global figures [2]. Given its position as the world’s primary CO₂ contributor, China’s mitigation measures are pivotal to international climate governance. To this end, China has actively deployed a series of policy tools, including the construction of a national carbon market, enhancement of energy efficiency in key industries, and low-carbon city pilot programs [3], achieving significant localized emission reductions [4]. However, the space for emission reductions relying solely on end-of-pipe technology and administrative orders is narrowing [5], and there is an urgent need to explore new reduction potentials from a systemic and collaborative perspective. The emergence and swift advancement of next-generation digital technologies—including big data, AI, and IoT—are transforming patterns of economic growth and modes of social operation [6]. The digital economy (DE), with its strong permeability, connectivity, and network characteristics [7], is expected to generate significant synergistic emission reduction effects among regions by optimizing resource allocation efficiency, promoting the diffusion of green technologies, and empowering the transformation of traditional industries [8]. The interconnectivity of digital infrastructure, cross-domain flow of data elements, and regional linkages of digital industries mean that carbon emission behaviors across different regions may no longer be isolated but are interconnected and interdependent through complex digital ties. In this context, a key scientific question urgently needs to be addressed: Does the booming DE drive the synergistic decrease in regional carbon emissions, or does it exacerbate the differentiation and imbalance in regional carbon reduction? Unveiling how the DE dynamically influences inter-regional carbon collaborative governance (ICCG) is essential for refining the nation’s “dual carbon” strategies and fostering harmonized low-carbon progress across different areas.

In recent years, research has intensively examined how the digital economy influences carbon emissions from various perspectives. Powered by innovations in digital technologies and interconnected through data flows, the DE allows enterprises to transcend spatial boundaries, collaborate via online platforms, and establish an integrated ecosystem spanning technology, commerce, and markets [9]. At the micro level, digital technology optimizes resource allocation, alleviates innovation resource constraints, and promotes green technological innovation through knowledge spillovers and economies of scale, thereby improving corporate carbon efficiency [10]. At the macro level, the digital economy strengthens technological interaction between regions, enhancing overall innovation efficiency [11], which not only helps improve local carbon neutrality performance but also drives a broader green and low-carbon transformation through spatial spillover effects [12]. Therefore, some scholars suggest that constructing a multi-level, multi-polar, and networked regional digital collaborative system is a key path to improving urban carbon neutrality performance [13]. It‘s noteworthy that while the digital economy reshapes the regional economic geography and industrial networks [14], it may also give rise to new carbon emission spatial patterns and interregional dependencies. On the one hand, regional differences in the level of digital economic development may either widen or narrow the gap in carbon reduction effectiveness between regions by influencing technological progress and industrial upgrading [15]. On the other hand, close digital economic linkages may facilitate the flow of low-carbon knowledge, technologies, and capital, which could influence carbon pricing, potentially leading to “carbon leakage” or exacerbating regional development imbalances [16]. Therefore, there is an urgent need to integrate the spatial pattern of carbon emissions with the networked structure of digital economy relationships into a unified framework for systematic analysis.

In terms of research methodology, Social Network Analysis (SNA) focuses on “relational data” and provides a powerful tool for understanding the network mechanisms of implicit carbon flows and green technology diffusion between regions. Existing studies have used SNA to reveal the complexity of carbon emission spatial relationships. For example, Li et al. found that interprovincial carbon emissions in China exhibit a multi-thread network structure, with developed eastern provinces at the center of the network, while underdeveloped western regions are mainly the initiators of carbon output [17]. Wyckoff et al. and Gao et al., based on global input–output data, demonstrated that global value chain (GVC) division of labor and positions significantly promote the formation of carbon emission transfer relationships between countries [18,19]. In the area of green technology diffusion, some studies, based on green patent data, constructed cooperative networks and found that these networks exhibit a “core–periphery” structure. Node centrality measures, including degree and closeness indices, show an inverse association with carbon emissions, whereas structural hole metrics exhibit heterogeneous effects [20]. Other studies have focused on the multi-layered network structure of green technology transfer, pointing out that factors such as knowledge diversity and geographical distance play different roles in technology absorption and output [21].

Overall, although prior research offers substantial insights into how the digital economy relates to carbon emission from an attribute-based perspective, it remains limited in applying relational-data-oriented network analysis, particularly in clarifying how the DE facilitates the development of ICCG. Therefore, we utilized provincial-level panel data for China covering 2012–2022 and applied a directed, weighted complex network approach to characterize the evolution of the inter-regional collaborative carbon reduction governance network (ICCGN) structure. In addition, an Exponential Random Graph Model (ERGM) was employed to verify the pathways through which the DE promotes joint efforts in carbon emission reduction. The core contributions of this work are outlined in Figure 1. First, it shifts the research perspective from single regions to ICCG. Moving from past “attribute data” to “relational data”, it depicts the network structure characteristics of multi-directional connections in ICCG. Second, it constructs a three-dimensional analysis framework driven by innovation, supported by industry, and led by talent, which addresses the lack of discussion on the mechanisms through which the DE empowers ICCG. This further refines the “point-to-point” relationships within the network, delving deeper into the convergence (or divergence) patterns of the DE nodes’ attributes in the formation of ICCGN. Third, the use of the ERGM effectively overcomes the limitations of traditional econometric models in handling endogeneity and selecting control variables. It incorporates the network’s endogenous structure, exogenous network, and attribute nodes into a unified analytical framework. This comprehensive approach considers both endogenous and exogenous factors that may influence the model during the DE’s mechanism process, enriching the related theories on the genesis of ICCGN.

The structure of the study is as follows: theoretical foundations and hypothesis development are covered in Section 2; methodology and variable definitions appear in Section 3; Section 4 contains the empirical results; and the final section synthesizes conclusions together with policy-oriented insights.

2. Theoretical Mechanisms and Research Hypotheses

2.1. Digital Economy and Carbon Collaborative Governance: Inducement Effect

Collaborative governance theory emphasizes that ICCG is achieved through inter-regional collaboration in space and a shift from “pollute first, manage later” to “prevention and control at the source” in time. Various measures are utilized to shorten the management cycle and reduce the costs of governance. As an advanced productive force, the DE promotes innovation in traditional production models, effectively reducing energy consumption and carbon emissions, thereby empowering ICCG [22]. The inducement effect of the DE on regional carbon collaborative reduction manifests in several dimensions, as follows: First, the DE lowers the costs of carbon reduction. Leveraging big data to identify, select, and filter, the DE integrates, stores, and shares previously non-standardized, dispersed environmental policy documents in a convenient, standardized, and low-cost form. This integration embeds dispersed environmental systems into the ICCGN, enhancing the speed of environmental governance information dissemination, increasing network density and interactivity [23], and reducing the costs of ICCG. Second, the DE enhances carbon emission efficiency. Open innovation platforms, such as “Internet Plus” and digital finance, provide strong support for the exchange, sharing, and spatial spillover of technological innovation elements, leading to enhanced economic output and efficiency [24]. Third, it shortens the carbon governance cycle. In the DE era, digital tools and platforms also provide real-time environmental monitoring services, enabling more convenient and timely access to and monitoring of carbon emission issues [25], reducing the uncertainty in the carbon supervision process, and shortening the carbon governance cycle.

The driving role of the DE in ICCG may manifest as a complex network structure in practice. Theoretically, the convergence of DE development models aims to break down information barriers and reduce collaboration costs through technology standardization, policy coordination, and industrial linkages, thus creating the foundational conditions for ICCG. However, regional connections in reality are often constrained by resource endowments and stages of development. This means that, even in the presence of an overall trend towards the convergence of development models, actual inter-regional connections are more likely to follow the “compatibility” principle. That is, regions with similar levels of DE development, due to stronger technological compatibility, similar institutional environments, and tighter industrial linkages, are more likely to establish efficient ICCG relationships. In contrast, pronounced differences in development levels can hinder interregional cooperation, largely due to technological disparities, misaligned policies, and conflicting interests, despite the presence of potential synergies. Based on this reasoning, we posit: H1—DE progress stimulates ICCG; H2—similarity in DE maturity supports the establishment of ICCGN.

2.2. Digital Economy and Carbon Collaborative Governance: Heterogeneous Impacts

Currently, the development levels and innovation efficiencies of the DE vary across different regions in China, leading to significant disparities in empowering ICCG [26]. These differences are not only reflected in traditional economic indicators such as economic development levels and industrial structure upgrades but are more profoundly manifested in the capacity and effectiveness of utilizing the DE to promote ICCG [27]. Existing literature on the heterogeneous empowerment of carbon reduction by the DE mainly revolves around digital infrastructure, policy and institutional ecosystems, and cyberspace. More concretely, enhancing digital infrastructure, including the spatial arrangement of big-data hubs and the establishment of smart sensing systems, is pivotal for improving governance efficiency in carbon reduction [28]. These technologies enable real-time data capture, swift communication, and automated analytics of carbon emissions, which are essential for precise and prompt environmental-policy responses. Second, the policy and institutional environment also play an undeniable role in the application of the DE to carbon control [29], covering aspects from national-level strategic planning and legislation to the execution and incentive mechanism design of local governments. An open and inclusive policy environment that encourages innovation can accelerate the pilot promotion of new technologies and promote the iterative upgrading of governance models. Additionally, cyberspace, as a new domain in the information age, increasingly shows its potential in promoting the sharing of environmental governance resources, public participation, and transnational cooperation. The establishment of network platforms, the enhancement of information transparency, and the innovative application of digital governance tools help break information silos and enhance the synergy and comprehensiveness of carbon reduction. Based on this, this paper proposes Hypothesis 3: the DE-enabled ICCGN may exhibit heterogeneity in digital infrastructure, free trade zone policy construction, and network status.

2.3. Theoretical Mechanism of Digital Economy Enabling Carbon Collaborative Governance

2.3.1. The Digital Economy Enhances Technological Innovation and Drives ICCG

Based on endogenous growth theory, the DE has driven the innovation of environmental protection technologies, such as green recycling technology and waste recycling systems, promoting the synergy of technological innovation and intelligent applications. This broadens the boundaries of technological innovation within the DE, facilitating technological innovation cooperation across multiple sectors and industries, and achieving coordinated national carbon governance [30]. At the same time, the development of the DE is also a process of green technology innovation and the diffusion and promotion of green technology. This process integrates into the innovation chain, effectively breaking away from resource consumption, labor-intensive, and capital-intensive driven extensive growth, and achieving “intensive” growth driven by technological innovation. This forms a green, low-carbon circular economic system, enhancing the quality and efficiency of clean production across multiple fields [31]. Furthermore, the DE leverages the organizational integration function of technological innovation to effectively achieve cross-sector and cross-boundary cooperation of internal and external resources, promoting the formation of a carbon governance system involving multi-stakeholder participation [32]. Therefore, this paper proposes Hypothesis 4: the DE empowers ICCG through innovation-driven mechanisms.

2.3.2. The Digital Economy Promotes Industrial Structure Upgrading and Facilitates Carbon Collaborative Governance

The key to facilitating ICCG lies in the transformation and upgrading of industrial structures [33]. The DE aids ICCG through industrial structure upgrading, primarily in two aspects: First, the DE leverages the advantage of industrial coordination and linkage to accelerate the reshaping of ICCG patterns [34]. The advancement of industrialization accelerates resource extraction, leading to issues such as resource scarcity and environmental pollution. Currently, most industries in China are characterized by high consumption, high pollution, low added value, and low technological content, with severe challenges such as high-end industry technology lock-in, lack of coordination, and instability barriers. The DE breaks down barriers to the flow of technology, information, and capital between regions, enabling coordinated linkage between industries [35,36], providing the possibility to overcome the “Solow Paradox” and stimulate the vitality of ICCG. Meanwhile, as the industrial structure advances towards technology-intensive and knowledge-intensive levels, the government encourages the development of new quality productivity to reduce corporate pollution emissions, promote green production and sustainable development, and accelerate the reshaping of the ICCG pattern. Second, according to Porter’s “competitive advantage” theory, the DE stimulates structural changes and technological upgrades in industries, enhancing industrial competitiveness and improving ICCG performance. On one hand, the DE can motivate traditional industrial sectors to leverage economies of scale and scope, maximizing the efficiency of carbon collaborative reduction [37]. On the other hand, the DE capitalizes on the effect of technological innovation, spurring industries to transition from low-end to “servitization” and “high-end”, thereby enhancing the performance of ICCG [38]. Therefore, this paper proposes Hypothesis 5: the DE empowers ICCG through industrial structure upgrading.

2.3.3. The Digital Economy Cultivates Human Resources and Empowers Carbon Collaborative Governance

In the neoclassical economic growth theory, technology is regarded as an exogenous variable, while human resources and capital are considered independent variables, forming the Cobb-Douglas production function to explain economic growth phenomena. Thus, human resources can be seen as the most decisive and dynamic element in the DE. The accumulation of human resources mainly occurs through two modes: the talent chain and the education chain. The ongoing development of the digital economy, driven by disruptive technological innovations and breakthroughs in key technologies, has fostered the emergence of numerous leading scientific talents, innovative teams, and young leaders. This has established a precise docking platform between green technology innovation talents and industry needs, promoting ICCG among governments. Simultaneously, the development of the DE also provides a digital platform support for high-quality talents, enabling collaboration and sharing through intelligent platforms, resulting in multi-platform knowledge and technology spillover effects [39], thereby empowering carbon collaborative governance. Within the framework of endogenous growth theory, education and training are also important means of human capital accumulation. By using educational resources and systems to cultivate high-quality talents with practical and innovative abilities, the process expands the capacity of innovative human resources across the stages of “knowledge and skills, discipline cultivation, results transformation, and employment output,” further empowering ICCG. Therefore, this paper proposes Hypothesis 6: the DE empowers ICCG through the cultivation of human resources.

In a word, this paper has constructed the theoretical analysis framework as shown in Figure 2.

3. Materials and Methods

3.1. Network Construction

ICCGN is a network collective characterized by interconnections, mutual constraints, and collaborative governance among regions. Therefore, when exploring ICCG from the perspective of complex networks, it is first necessary to measure the performance of carbon reduction ($T_i$). Building on the approaches of Cui et al. [40] and Fang et al. [41], this study applies the non-radial, non-desired output Slack-Based Measure (SE-SBM) model to evaluate carbon emission performance, using MATLAB R2016b. Drawing from prior work on efficiency indicator quantification [42], three types of inputs are considered: labor, capital, and energy. Capital is estimated via the perpetual inventory method, with a depreciation rate of 0.96. Labor is captured as the year-end number of employed persons in each province or municipality. Energy use is converted into standard coal equivalents based on total consumption. Real GDP serves as the desirable output, whereas carbon emissions are taken as the undesirable output. The specific calculation process is detailed in [43].

Gravitational models have become a common approach for understanding spatial relationships and accurately capturing dynamic trends in networks. The model can be applied to various types of relationships, incorporating factors such as economic and geographical elements to develop an adjusted gravity model. This model, when studying CO₂ emission connections, accounts for the carbon transfer resulting from factor flows, offering a more precise representation of carbon linkage networks compared to the traditional gravity model. Building on the methods of Zhao [44], this study utilized adjusted gravity models to assess the gravity intensity of CO₂ emissions across provinces. Meanwhile, this study uses population size (P) and GDP total (E) as representations of regional scale attraction to reflect the demand foundation for collaborative governance. Additionally, carbon reduction performance (T) is introduced to characterize the complementary learning dynamics between regions, which constitutes the core of collaborative governance. Furthermore, economic distance is used as a substitute for geographical distance, aiming to measure the homogeneity of development levels and institutional environments between regions, which better aligns with the inherent requirements of collaborative governance for policy compatibility and implementation capacity. The specific formula is shown in Equation (1).

$$C_{ij}=a_{ij}{\displaystyle \frac{\sqrt[3]{P_i{T}_i{E}_i}\sqrt[3]{P_j{T}_j{E}_j}}{{\left(\frac{D_{ij}}{e_i-e_j}\right)}^2}},\hspace{1em}{a}_{ij}={\displaystyle \frac{T_i}{T_i+T_j}}$$

(1)

In the equation, $C_{ij}$, $a_{ij}$, and $D_{ij}$ represent the ICCG link, the gravity coefficient, and the geographical distance between region i and region j, respectively. The geometric center distance $D_{ij}$ between the capital cities of provinces i and j is measured using spherical distance in ArcGIS 10.8. P, T, E, and e correspond to population size, carbon reduction performance, total GDP, and GDP per capita, respectively; $D_{ij}/(e_i-e_j)$ represents the economic distance between cities. After generating the initial gravity matrix using Formula (1), the row mean threshold method is used to binarize it into a [0, 1] matrix. The calculation method is as follows: if the value is above the threshold, it is recorded as “1”, indicating the presence of carbon collaborative governance between the two regions; if the value is below the threshold, it is recorded as “0”, indicating the absence of collaborative governance.

3.2. Research Methods

3.2.1. Social Network Analysis

Conventional spatial analytical methods often fail to fully capture the complex, nonlinear network patterns arising from cross-regional resource movements. Social network analysis (SNA), grounded in graph theory, constructs relational models that convert discrete spatial units into intuitive network representations. Using gravity-matrix calculations to assess the strength and direction of these flows, SNA exposes the evolving interaction structures among region agglomerations and clarifies potential collaborative links. It also enables examination of both micro-level actors and aggregate entities in spatial datasets, revealing underlying connection patterns. Accordingly, this study applies SNA to investigate collaborative carbon-reduction linkages between regions in China.

3.2.2. Exponential Random Graph Model

The Exponential Random Graph Model (ERGM) integrates both internal network features and external node attributes to explore driving mechanisms. Parameter estimation is conducted via Markov Chain Monte Carlo Maximum Likelihood Estimation (MCMCMLE), accompanied by goodness-of-fit evaluation. In contrast to conventional regression and simulation approaches, ERGM relaxes the assumption of independent ties, enabling the inclusion of endogenous structures—such as edge counts and reciprocity—and exogenous factors, including attribute data, to provide richer insights into structural dynamics and node influence. This study applies ERGM for estimation, with the formulation shown below.

$$P(Y=y)={\displaystyle \frac{1}{k}}\exp \left[\sum _{\theta }{\eta }_{\theta }{g}_{\theta }(y)\right]$$

(2)

In Equation (2), $P(Y=y)$ represents the probability of network structure realization; $Y$ represents a random variable of a network connection; $y$ represents the network simulated by the model; $\theta$ represents different network configurations (including three types: endogenous network structure variables, attribute variables, and exogenous network covariates); $g_{\theta }(y)$ represents the statistics corresponding to different network configurations $\theta$; ${\eta }_{\theta }$ represents the coefficients of different network configurations $g_{\theta }(y)$; and $k$ represents the normalization factor. The statistics selected in the ERGM model are explained as follows (

Table 1. Description and interpretation of ERGM model statistics.

Variable	Name	Diagram	Statistic	Statistical Significance
$edges$	Number of Network Edges		${\displaystyle \sum _{i,j}}{y}_{ij}$	Like a constant in a model, generally not explained.
$mutual$	Reciprocity		${\displaystyle \sum _{i,j}}{y}_{ij}{y}_{ji}$	Test whether it is easier to form reciprocal relationships for carbon collaborative governance between regions.
$gwesp$	Geometrically Weighted Edgewise Shared Partners (GWESPs)		$AT{K}_{\lambda }(y)$	Test whether the ICCGN exhibits transitivity.
$nodematch$	Homophily		${\displaystyle \sum _{i,j}}{y}_{ij}{\delta }_i{\delta }_j$	Test whether regions with the same or similar attribute ($\delta$) are more likely to establish ICCGN.
$absdiff$	Heterophily		${\displaystyle \sum _{i,j}}{y}_{ij}{\delta }_i{\delta }_j$	Test whether regions with significantly different attributes are more likely to establish ICCGN.
$edgecov$	Exogenous Network Covariates		${\displaystyle \sum _{i,j}}{y}_{ij}{g_{ij}}^{\prime }$	Test whether regions are more likely to form ICCGN within other existing exogenous networks.

).

This study employs panel data from 30 provinces in mainland China covering the period 2012–2022, with data from Taiwan, Hong Kong, Macau, and Tibet currently unavailable. The dataset is compiled from multiple sources, including the China Industrial Economy Statistical Yearbook, China Environmental Statistics Yearbook, China Environment Yearbook, China Science and Technology Statistical Yearbook, China Statistical Yearbook, the China Emissions Accounts and Datasets (CEADs), and the database of the National Bureau of Statistics of China. Interpolation techniques are applied to address a limited number of missing observations.

3.3. Variable Explanation

3.3.1. Core Explanatory Variable

DE Development Index (nodecov.DE). Based on the theoretical foundation of the DE, this paper considers constructing a “factors of production-technology innovation-industrial development” framework to measure the level of DE development, thereby analyzing the channels through which the DE enables ICCG. Relying on a multi-attribute comprehensive evaluation system, a newly improved entropy-weighted TOPSIS method is used to comprehensively calculate the DE development index. The evaluation index system is shown in

Table 2. DE evaluation index system.

Primary Indicator	Secondary Indicator	Variable Selection	Unit	Weight
Digital Production Factors	Digital Literacy	Total Number of Employees in Strategic Emerging Industries and Future Industries Listed Companies/Total employed population	%	0.253
		Average Years of Education per Capita	Year	0.051
	Digital Infrastructure	Robot Penetration Rate	%	0.057
		Internet Penetration Rate	%	0.022
		Number of 4G Base Stations per 10,000 People	Units per 10,000 People	0.071
Digital Technology Innovation	Digital Technology Application	Number of AI Enterprises	Units	0.062
		Labor Productivity of Industrial Enterprises Above Designated Size	CNY10,000/Person	0.056
	Digital Innovation Capability	Revenue from High-tech Industries	CNY10,000	0.073
		R&D Expenditure of Industrial Enterprises Above Designated Size	CNY10,000	0.031
		Number of Domestic Patents Granted	Units	0.063
Digital Industry Development	Industry Digitalization	Software Business Revenue	CNY10,000	0.069
		Length of Optical Cable Lines/Area of the Region	Meters/Square Kilometer	0.025
	Digital Industrialization	E-commerce Sales	CNY10,000	0.034
		Number of Internet Broadband Access Ports	Units	0.048
		Total Telecom Business Volume	CNY10,000	0.066
		Integrated Circuit Output	10,000 Pieces	0.019

. In the indicator system, the “average years of education per person” and the “percentage of total employees in listed companies in strategic emerging industries and future industries relative to the total employed population” are key indicators of social development level and can effectively reflect digital literacy. The former reflects society’s reliance on high-tech industries, while the latter represents the educational background and skill reserves of the labor force. Together, these indicators reveal the knowledge structure and technological adaptability of the labor market, thereby reflecting the overall digital literacy level of society, providing data support for analyzing the relationship between DE and ICCG.

To further explore whether the DE enabling the ICCGN exhibits a convergence effect, we intend to explore the effect of the DE from the perspective of node province assortativity. The DE index is ranked from high to low and specifically divided into three categories: high, medium, and low. The top 33% is classified as the high-level area (nodematch.high-DE), the 33–66% range as the medium-level area (nodematch.mid-DE), and the bottom 33% as the low-level area (nodematch.low-DE).

3.3.2. Mechanism Variables

(1)

Technological Innovation (nodecov.Tech₁ and nodecov.Tech₂)

Two indicators are selected to represent the level of technological innovation: Internet Development (nodecov.Tech₁) and Digital Inclusive Finance (nodecov.Tech₂). The Internet Development indicator is represented by the internet penetration rate, while the Digital Inclusive financial indicator is derived from the China Digital Inclusive Finance Index, compiled through a joint effort by Peking University’s Digital Finance Research Center and Ant Financial Services Group.

(2)

Industrial Structure Upgrading (nodecov.Stru)

Industrial structure upgrading reflects the process of production factors adjusting from a lower efficiency stage to a higher efficiency stage, with the continuous evolution and replacement of leading industries. The structural level coefficient is used to represent the overall level of industrial structure upgrading. The calculation formula is: $f_1\times 1+f_2\times 2+f_3\times 3$, where $f_i$ represents the output value proportion of the $i$-th industry.

(3)

Human Resources (nodecov.Inter₁ and absdiff.Inter₂):

Technological talent and differences in educational investment are two important factors influencing human resource accumulation. Technological talent is quantified by the proportion of graduates from higher education who hold bachelor’s degrees or above. Educational investment is expressed as local government education expenditure divided by overall general budget expenditure.

3.3.3. Control Variables

(1)

Endogenous network structural variables:

To minimize the risk of network degeneration, three typical endogenous structural variables are selected as control variables: the number of edges (edges), mutual connections (mutual), and geometrically weighted edgewise shared partners (gwesp).

(2)

Exogenous network covariates:

Since exogenous network disturbances can also bias estimation results, two exogenous network covariates are introduced: the urbanization gap network (edgecov.Urban) and the marketization gap network (edgecov.Mar). These variables are used to test whether disparities in urbanization and marketization across regions increase the likelihood of ICCG. The urbanization gap network and marketization gap network are measured using matrices of interregional differences in urbanization levels and marketization indices, respectively.

(3)

Other control variables:

Beyond endogenous structural variables, we control for nodecov. Pgdp—per-capita economic development; nodecov. Edu—the science and education environment, defined as science and education expenditure over total fiscal expenditure; and nodecov. Gov—government intervention, defined as public fiscal expenditure divided by the region’s economic development level.

4. Results and Discussion

4.1. Analysis of the Evolutionary Characteristics of the ICCGN Structure

Using NetDraw software (version 2.084), the evolutionary characteristics of ICCGN topology from 2012 to 2022 were depicted (Figure 3). The nodes, representing provincial regions, both emit and receive carbon reduction relations. These have transcended traditional spatial geographic proximity attributes, forming a clearly directed collaborative governance spatial association network, with no isolated points within the entire network.

4.1.1. Analysis of the Evolutionary Characteristics of Network Node Importance

Table 3. Top 10 Provincial Regions by Degree Centrality in the ICCGN (2012–2022).

Province	2012	Province	2017	Province	2022
Shanghai	0.089	Shanghai	0.094	Shanghai	0.096
Jiangsu	0.082	Jiangsu	0.094	Beijing	0.093
Beijing	0.079	Beijing	0.087	Jiangsu	0.093
Zhejiang	0.072	Guangdong	0.055	Zhejiang	0.057
Guangdong	0.056	Zhejiang	0.055	Guangdong	0.052
Tianjin	0.056	Tianjin	0.043	Tianjin	0.045
Gansu	0.036	Fujjian	0.04	Gansu	0.042
Chongqing	0.032	Gansu	0.036	Fujjian	0.042
Fujjian	0.032	Chongqing	0.033	Chongqing	0.034
Hubei	0.032	Hebei	0.029	Qinghai	0.031

reports the top 10 provincial regions in terms of degree centrality within the Chinese ICCGN from 2012 to 2022. It can be observed that the status pattern of various provincial ICCGN has a certain degree of stability over the years. The provinces with higher centrality rankings are mainly concentrated in the eastern coastal regions. This suggests that, compared to other areas, the eastern coastal provinces have a greater direct impact on carbon reduction in other regions. Not only can these coastal provinces gradually integrate governance technologies and advantageous resource elements by expanding their carbon collaborative reduction partners, but they can also utilize their network resource advantages. While enhancing participation in ICCG, they can disseminate advanced carbon reduction technologies to closely connected partners, improving the efficiency and competitiveness of collaborative governance. Thus, they hold a central hub position in the ICCGN. Over time, the eastern coastal regions have led to technology diffusion and spatial spillover effects through talent and technology exchanges and resource integration with neighboring provinces and regions, driving collaborative carbon reduction in the central and western provinces. In 2022, Qinghai successfully entered the top 10 in centrality rankings, playing an increasingly pivotal role in the ICCGN.

4.1.2. Analysis of Core–Periphery Structure Evolution Characteristics

Based on the closeness of connections between provincial nodes in the ICCGN, the core–periphery structure index for the years 2012 to 2022 was measured using Ucinet 6.204 software (Figure 4). Referring to the division method by Sun et al. [7], provinces with a coreness greater than the average density value (0.25) are considered core provinces of the network, while those with coreness less than or equal to 0.25 are categorized as semi-peripheral and peripheral provinces. As shown in Figure 4, the core–periphery structure index of the ICCGN exhibits a fluctuating upward trend, indicating an increasing number of provinces participating in the ICCGN. In 2012, Guangdong, Chongqing, Sichuan, and Guizhou occupied core positions in the ICCGN, playing a coordinating role and deeply engaging in regional carbon reduction. Over time, more regions have joined the ICCGN. By 2022, Guangdong, Chongqing, Gansu, and Qinghai emerged as core areas in the ICCGN, taking on more leadership roles in the network, with the overall density of the ICCGN increasing. It was also found that the hierarchical differentiation within the ICCGN is weakening, indicating that more regions are participating in the carbon reduction governance system, and the scope of the core areas in ICCG is gradually expanding, showing a diffusion from the eastern to the western regions, with governance boundaries weakening.

4.1.3. Analysis of Spatial Clustering Evolution Characteristics

As can be seen from the results of the block-model analysis (Figure 5 and

Table 4. Block Structure of the ICCGN.

Project	Block 1	Block 2	Block 3	Block 4	Number of Receiving Relationships	Number of Spillover Relationships	Expected Relationship Proportion/%	Actual Relationship Proportion/%	Characteristic
Block 1	4	2	14	2	87	18	10.34	18.18	“Two-Way Spillover” Block
Block 2	7	0	4	10	33	21	6.90	0.00	“Broker” Block
Block 3	48	8	3	3	20	59	41.38	4.84	“Net Spillover” Block
Block 4	32	23	2	15	15	57	31.03	20.83	“Net Beneficial” Block

), there are 22 internal connections within the blocks and 155 connections between blocks, indicating that the ICCGN has strong spatial relevance and spatial spill-over effects. The first block mainly consists of relatively economically developed provincial regions with high requirements for carbon control. There are 87 carbon governance relationships and 18 external spill-over relationships among the provincial regions in this block. It generates spill-over relationships both within and outside the block, presenting a “two-way spill-over” block characteristic. The members of the second block are mainly economically developed eastern coastal provincial regions. Relying on the advantages of special economic zones and free trade zones, they play an intermediary role in undertaking ICCG, acting as “brokers”. The third block includes 13 provincial regions: Jilin, Inner Mongolia, Hebei, Heilongjiang, Anhui, Liaoning, Ningxia, Shandong, Henan, Hubei, Shanxi, Xinjiang, and Shaanxi. With a relatively large number of members, this block is quite active in the ICCGN. It has 20 receiving relationships and 59 external spill-over relationships, showing the characteristics of a “net spill-over” block. This indicates that more provincial regions are participating in ICCG, presenting a spatial spill-over effect in the network. The members of the fourth block are mainly less-developed central and western provincial regions. Due to their relatively poor technological innovation and economic strength, it is not conducive to the diffusion of low-carbon environmental protection technologies. In the network, they mainly undertake the spatial spill-over of external technological innovation, presenting the characteristics of a “net beneficial” block.

4.2. Digital Economy Empowerment of Carbon Collaborative Governance Analysis

4.2.1. Baseline Regression Results

Table 5. Baseline ERGM Regression Results.

Variable	(1)	(2)	(3)	(4)	(5)	(6)
$edges$	−2.616 *** (0.261)	−2.771 *** (0.235)	−4.307 *** (1.283)	0.851 (1.302)	−1.852 (1.495)	3.232 * (1.597)
$mutual$	1.451 *** (0.262)	0.975 *** (0.282)	0.618 * (0.312)	4.412 *** (0.596)	0.529 * (0.321)	4.332 *** (0.611)
$gwesp$	0.575 ** (0.175)	−0.195 (0.206)	−0.592 ** (0.204)	−0.299 (0.189)	−0.656 ** (0.207)	−0.432 * (0.179)
$nodecov\text{.}DE$	—	2.948 *** (0.475)	9.416 *** (1.247)	5.862 *** −1.149	8.979 *** −1.253	5.882 *** −1.119
$nodecov\text{.}Pgdp$	—	—	−4.474 *** (1.175)	−2.187 * (1.158)	−4.318 *** (1.220)	−2.340 * (1.232)
$nodecov\text{.}Edu$	—	—	6.992 * (3.775)	4.472 −3.630	1.790 −4.186	−0.404 −4.159
$nodecov\text{.}Gov$	—	—	−1.086 (0.975)	−1.125 (1.013)	−1.659 (1.031)	−1.567 (1.039)
$nodematch.high\text{−}DE$	—	—	—	—	0.524 * (0.240)	0.389 * (0.232)
$nodematch\text{.}low\text{−}DE$	—	—	—	—	−0.494 * (0.210)	−0.584 * (0.229)
$edgecov\text{.}Urban$	—	—	—	−5.283 *** (1.118)	—	−4.824 *** (1.151)
$edgecov\text{.}Mar$	—	—	—	−6.810 *** (1.149)	—	−7.184 *** (1.165)
AIC	854.128	778.835	736.635	542.230	727.628	535.348
BIC	868.600	798.164	770.436	585.644	771.144	588.478

Note: Statistical significance thresholds denoted as * p < 0.10, ** p < 0.05, *** p < 0.01, with standard error estimates enclosed in parentheses.

reports the regression results of the DE on the ICCGN. Column (1) presents the estimated results of endogenous structural variables related to the network, including edges, mutual, and gwesp, to depict the distribution characteristics of the network structure and to control for potential disturbances within the network. In the model, the $edges$ acts like the constant term in regression analysis, typically negative, indicating that the ICCGN is not formed randomly and features low network density. The estimated value of mutual is significantly positive at the 1% significance level, indicating that inter-regional environmental information disclosure and cooperation significantly affect the formation of the ICCGN. Carbon governance linkages tend to form between provinces with comparable environmental standards, and reciprocal benefits are essential for sustaining these inter-regional ties. The estimated coefficient of the gwesp is 0.575, indicating that ICCG often occurs via a shared third province, forming transitive triangular relationships. This reflects a strong self-organizing mechanism in the network, which plays a notable role in shaping the inter-provincial governance structure and is consistent with the reality of regional collaboration.

In Table 5, columns (2)–(4) present ERGM regression outcomes that sequentially introduce core explanatory variables, attribute controls, and exogenous network controls. Focusing on column (4), which incorporates all controls, the coefficient for DE (nodecov.DE) is positive and significant at the 1% level. This implies that provinces with more advanced DE development are more inclined to participate in ICCG. Rapid DE expansion facilitates digital regulation, communication, and resource sharing, lowering environmental governance costs, shortening cycles, and enhancing performance. Furthermore, innovation diffusion driven by higher DE levels produces spillovers that stimulate participation in ICCGN, confirming Hypothesis 1.

To assess whether the DE fosters compatibility within ICCGN, column (5) of Table 5 integrates a DE compatibility variable. Results reveal that the nodematch.high-DE coefficient (0.524) is positive and significant at the 10% level, while nodematch.low-DE (−0.494) is significantly negative at the same level. This suggests that provinces with stronger DE development tend to integrate into ICCGN more effectively, revealing a notable compatibility effect. Conversely, less-developed regions encounter greater difficulty connecting with the network. This may be because regions with high levels of development have compatible technological standards, similar policy goals, and strong industrial complementarity, which lowers collaboration costs. In contrast, lower-level regions are constrained by information, technology, and infrastructure gaps, creating a “technological divide” with high-level regions, which hinders the establishment of collaborative relationships. Thus, Hypothesis 2 is verified.

Column (6) shows the ERGM regression results with all variables included, where the signs and significance of the estimated coefficients have not changed significantly, further proving the robustness of Hypotheses 1 and 2. Focusing on the impact of network covariates on the ICCGN, it can be observed that both the urbanization difference network (edgecov.Urban) and the marketization difference network (edgecov.Mar) variables are significantly negative at the 1% level. This implies that widening gaps in urbanization and marketization hinder ICCG, suggesting that exploring regional collaborative governance needs to consider the coordination and sustainability of various dimensions, multiple stakeholders, and multiple objectives, including narrowing the gaps in urbanization and marketization.

4.2.2. Goodness-of-Fit Diagnostics and Robustness Checks

(1)

Goodness-of-Fit Diagnostics

In most ERGM studies, model evaluation commonly relies on AIC and BIC, which perform well for simple, binary-independent structures. For more complex networks—those involving binary or higher-order dependencies—goodness-of-fit (GOF) diagnostics become essential. The core approach is to compare structural traits of the fitted and observed networks. Figure 6a–c illustrate edge-shared partners, geodesic distances, and degree distributions for both the empirical carbon governance network and the network simulated from Model 1. In the plots, black solid lines represent measurements from the observed network, while boxplots denote simulated results within the 95% confidence range. When the observed values (lines) lie within the boxplot range, it indicates that the simulation captures the structural features effectively. Visual inspection of Figure 6 shows Model achieves a strong fit. Additionally, Receiver Operating Characteristic (ROC) and Precision–Recall (PR) curves are applied as non-parametric robustness tests. Curves that approach the upper boundary signify better predictive accuracy, and the outcomes in Figure 6d further validate the model’s reliability.

(2)

Robustness Check

To assess the adaptability of the ERGM estimation under varying threshold criteria, this study adopts the first and third quartiles as alternative cut-off points to test the robustness of the model. Based on these thresholds, if the ICCG gravity between regions i and j exceeds the chosen value, the connecting edge is assigned a weight of 1; otherwise, it is set to 0, thereby generating a binary network. The robustness check results, presented in

Table 6. Robustness test.

Variable	Mean Threshold	First Quartile Threshold	Third Quartile Threshold
$edges$	3.232 * (1.597)	−1.852 (1.495)	0.851 (1.302)
$mutual$	4.332 *** (0.611)	0.529 * (0.321)	4.412 *** (0.596)
$gwesp$	−0.432 * (0.179)	−0.656 ** (0.207)	−0.299 (0.189)
$nodecov\text{.}DE$	5.882 *** −1.119	8.979 *** −1.253	5.862 *** −1.149
Control	YES	YES	YES
AIC	535.348	727.628	542.230
BIC	588.478	771.144	585.644

Note: Statistical significance thresholds denoted as * p < 0.10, ** p < 0.05, *** p < 0.01, with standard error estimates enclosed in parentheses.

, reveal that both the significance of key explanatory variables and the direction of their effects remain consistent, confirming the stability of the empirical findings.

4.2.3. Heterogeneity Analysis

(1)

Digital Infrastructure

While developing the DE using big data and artificial intelligence, networked and digital empowerment for ICCG is indispensable, which relies on the support and guarantee of digital infrastructure. The completeness of digital infrastructure directly influences the quality and effectiveness of DE development, thereby causing differentiated empowerment in ICCG. Therefore, this paper reflects the development level of digital infrastructure from two aspects: the “Broadband China” initiative and the national big data pilot zones (NBDPZs). The average number of cities within a province participating in the “Broadband China” initiative and designated as NBDPZ serves as the standard; provinces above this average are considered to have better “Broadband China” and NBDPZ development. Conversely, provinces below the average are categorized as having average “Broadband China” and NBDPZ. Results in

Table 7. Regression Results on the Heterogeneity of Digital Infrastructure and Free Trade Zone Construction.

Variable	Better “Broadband China”	Average “Broadband China”	Better NBDPZ	Average NBDPZ	FTZs	Non-FTZs
$edges$	2.934 ** (1.360)	2.209 ** (1.248)	0.897 (1.336)	1.486 (1.202)	2.527 * (1.246)	1.053 (1.252)
$mutual$	4.674 *** (0.606)	5.003 *** (0.588)	4.719 *** (0.610)	5.017 *** (0.591)	4.958 *** (0.599)	4.944 *** (0.594)
$gwesp$	−0.138 (0.178)	0.139 (0.178)	−0.105 (0.181)	0.138 (0.177)	0.063 (0.166)	0.134 (0.175)
$nodecov\text{.}DE$	2.503 *** (0.605)	−0.787 (0.576)	2.422 *** (0.560)	−0.747 * (0.442)	2.401 *** (0.713)	1.009 (0.721)
$nodecov\text{.}Pgdp$	2.855 *** (0.966)	2.513 ** (1.094)	−0.689 (1.030)	1.341 ** (0.811)	2.749 ** (0.934)	0.492 (1.162)
$nodecov\text{.}Edu$	−2.727 (3.823)	0.834 (3.456)	2.783 (3.614)	2.802 (3.157)	−0.243 (3.382)	4.797 (3.434)
$nodecov\text{.}Gov$	−1.430 (0.952)	−1.040 (0.883)	1.161 (1.034)	−0.179 (0.889)	−1.645 (0.926)	−0.634 (0.861)
$edgecov\text{.}Urban$	−5.820 *** (1.155)	−6.712 *** (1.110)	−5.826 *** (1.120)	−6.617 *** (1.141)	−6.569 *** (1.144)	−6.552 *** (1.105)
$edgecov\text{.}Mar$	−6.934 *** (1.125)	−7.021 *** (1.124)	−7.281 *** (1.161)	−7.238 *** (1.118)	−7.099 *** (1.126)	−6.982 *** (1.128)
AIC	557.511	579.370	554.171	576.435	569.857	577.042
BIC	601.027	622.785	597.687	619.850	613.272	620.457

Note: Statistical significance thresholds denoted as * p < 0.10, ** p < 0.05, *** p < 0.01, with standard error estimates enclosed in parentheses.

indicate that regions with better “Broadband China” and NBDPZ have a positive impact on ICCG through the DE. However, for regions with average development levels in “Broadband China” and NBDPZ, the effects of DE empowerment on carbon governance are, respectively, insignificant and negatively inhibitory. This is because regions with better development of either “Broadband China” or NBDPZ have more complete digital infrastructure, facilitating government dynamic monitoring and management of carbon emissions, as well as enabling enterprises to monitor pollutants emitted during production processes in real-time, thereby aiding in the DE’s empowerment of ICCG.

(2)

Free Trade Zone Construction

The establishment of Free Trade Zones (FTZs) is a key initiative in China’s reform and opening-up, aimed at advancing green, high-quality development. Under the dual-circulation framework, FTZs leverage domestic and global markets and resources to attract foreign investment and technological inflows, driving the economy’s green transformation and industrial upgrading. China’s first pilot FTZ was launched in Shanghai in September 2013. By 2024, China had established 21 FTZs. Compared to non-FTZ areas, FTZs incubate emerging industries and create platforms for industrial R&D and innovation, leading new growth poles, which may result in differences in how the DE empowers ICCG. To test this heterogeneity, this paper conducts a sub-sample analysis of 9 non-FTZs and 21 FTZs, with regression results shown in Table 7. The results indicate that within FTZs, the DE significantly promotes the formation of ICCGN at a 1% significance level, whereas the impact coefficient of the DE in non-FTZs is not significant. This implies that the construction of FTZs facilitates the DE’s empowerment of ICCG, while non-FTZs lack policy and platform advantages in leading carbon governance. Furthermore, FTZs not only accumulate more policy support and possess complete innovation platforms but also serve as new bases for international competition, targeting the forefront of technology, and exerting a radiating and driving effect on the surrounding areas. Therefore, compared to non-FTZs, FTZs enjoy more digital dividends and technological innovation advantages, leveraging the intelligent DE to empower ICCG.

(3)

Network Position

To further investigate whether the effectiveness of the DE’s empowerment varies among provinces with different network positions, this paper characterizes a province’s status within the ICCGN using degree centrality, betweenness centrality, and closeness centrality indicators. These reflect the actual influence, control, and freedom of the province within the ICCGN, respectively. After standardizing these measures, they are incorporated into the regression model, with results shown in

Table 8. Regression Results on Network Position Heterogeneity.

Variable	Degree Centrality	Intermediate Centrality	Closeness Centrality
$edges$	−0.334 (1.463)	0.825 (1.542)	0.636 (1.287)
$mutual$	4.251 *** (0.590)	4.289 *** (0.584)	4.565 *** (0.600)
$gwesp$	−0.442 ** (0.168)	−0.492 * (0.196)	−0.171 (0.195)
$nodecov\text{.}DE$	0.481 *** (0.077)	0.606 *** (0.094)	0.327 *** (0.077)
$nodecov\text{.}Pgdp$	−1.770 (1.134)	−2.692 * (1.240)	−1.420 (1.077)
$nodecov\text{.}Edu$	8.641 * (4.034)	5.215 (4.131)	5.580 (3.578)
$nodecov\text{.}Gov$	0.369 (1.079)	0.410 (1.107)	−1.187 (0.973)
$edgecov\text{.}Urban$	−4.972 *** (1.160)	−4.819 *** (1.151)	−5.752 *** (1.150)
$edgecov\text{.}Mar$	−6.889 *** (1.166)	−7.086 *** (1.143)	−6.713 *** (1.146)
AIC	526.847	524.823	553.159
BIC	570.262	568.238	596.675

Note: Statistical significance thresholds denoted as * p < 0.10, ** p < 0.05, *** p < 0.01, with standard error estimates enclosed in parentheses.

. It is evident that degree centrality, betweenness centrality, and closeness centrality all significantly and positively influence the ICCGN, with impact coefficients of 0.481, 0.606, and 0.327, respectively. This means that the higher a region’s network position, the stronger the role of the DE in empowering ICCG. Comparing the impact coefficients reveals: betweenness centrality > degree centrality > closeness centrality. This implies that the higher the betweenness centrality of a province, the higher its efficiency in resource allocation and control, and the greater the effect of the DE in empowering carbon collaborative reduction. Additionally, an increase in a region’s actual influence and trade freedom within the network can enhance its international competitiveness, increase trade flow, and level of openness. Therefore, provinces with higher degree centrality and closeness centrality are more conducive to the DE’s empowerment of ICCG.

4.2.4. Mechanism Tests

(1)

Technology Innovation Drive Mechanism

In

Table 9. Results of Mechanism Tests.

Variable	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)
$nodecov\text{.}DE$	6.836 *** (1.130)	4.727 *** (1.141)	4.962 *** (1.141)	10.890 * (4.523)	7.500 *** (1.460)	6.352 *** (1.172)	3.996 ** (1.240)	3.321 * (1.399)	8.242 *** (1.608)	7.396 *** (1.230)
$nodecov\text{.}Tec{h}_1$						0.077 * (0.038)	—	—	—	—
$nodecov\text{.}Tec{h}_2$						—	1.550 ** (0.491)	—	—	—
$nodecov\text{.}Stru$						—	—	2.671 ** (0.981)	—	—
$nodecov\text{.}Inte{r}_1$						—	—	—	0.013 * (0.005)	—
$absdiff\text{.}Inte{r}_2$						—	—	—	—	−2.468 *** (0.677)
Control	YES	YES	YES	YES	YES	YES	YES	YES	YES	YES
AIC	710.005	711.736	601.857	618.005	759.826	540.003	540.003	537.372	539.194	527.657
BIC	782.414	775.455	720.342	676.538	829.192	588.276	588.276	585.644	587.466	575.828

Note: Statistical significance thresholds denoted as * p < 0.10, ** p < 0.05, *** p < 0.01, with standard error estimates enclosed in parentheses.

, concerning the technology innovation drive mechanism, the estimation coefficient for the level of internet development (nodecov.Tech₁) is significant at the 10% level, with a coefficient of 0.077, indicating that every increase in the level of internet development increases the probability of regional environmental collaborative governance by approximately 7.6%. The estimation coefficient of digital inclusive finance (nodecov.Tech₂) is 1.550 and significant at the 5% level, indicating that technological innovation promotes the DE’s role in facilitating ICCG. As a key manifestation of technological strength, the expansion of the internet and digital inclusive finance exerts a notable influence on enhancing this process. The widespread adoption of green technologies in these domains helps dismantle traditional production models characterized by high resource consumption and labor intensity, enabling more efficient and intensive resource utilization. Furthermore, the diffusion of big data and internet applications removes spatial and temporal barriers, integrates resources across regions and sectors, and strengthens cross-border as well as interdepartmental cooperation in environmental governance. These factors jointly improve resource allocation efficiency and reinforce the DE’s capacity to support ICCG, thereby validating Hypothesis 4.

(2)

Industrial Support Mechanism

In Table 8, concerning the industrial support mechanism test, the variable for industrial structure upgrading (nodecov.Stru) is significantly positive at the 5% level, with an estimated coefficient of 2.671, indicating that the DE supports ICCG by promoting industrial structure upgrading. Developing the DE can break through the “Solow Paradox,” advancing the industrial structure towards more technology-intensive and knowledge-intensive sectors, leveraging scale and scope economies, and maximizing the efficiency of environmental pollution management. This confirms Hypothesis 5.

(3)

Talent Leadership Mechanism

In Table 9, the variable for tech talent (nodecov.Inter₁) is significantly positive at the 10% level, with an estimated coefficient of 0.013, meaning that for every increase in the level of tech talent, the probability of establishing ICCG increases by approximately 1.32%. Continuous emergence of technological innovation and cutting-edge science cultivates a large number of leading tech talents, innovative teams, and young leaders, realizing a high-quality supply mode for the DE and promoting structural transformation in carbon reduction models. The variable for differences in educational investment (absdiff.Inter₂) is significantly negative at the 1% level, indicating that a greater educational gap is detrimental to establishing ICCG relationships. Based on new growth theory, human capital generated through education is a key driver for promoting technological innovation. A widening educational gap hinders the transformation of technological achievements and is an intrinsic force driving regional disparities in technological innovation. Thus, the DE empowers ICCG by nurturing tech talent and narrowing educational gaps, further confirming Hypothesis 6.

5. Conclusions

(1)

From the perspective of network evolution patterns, the ICCGN structure from 2012 to 2023 presents a transitive triangular linkage. The network core–periphery structure index has shown an overall fluctuating increase, with the core area of ICCG gradually expanding and diffusing from the eastern to the western regions, indicating a weakening of governance boundaries. According to the spatial clustering structure analysis results, the network spillover relationships are concentrated in the first, third, and fourth sectors. The underdeveloped areas in the central and western regions have become the hinterland of the ICCGN, while the eastern coastal areas play a “hub” role in the network.

(2)

From the perspective of induced effects, the DE empowers ICCG. When considering network endogenous structure variables, it is highly likely that indirect connections will be formed through common third-party trading partners during ICCG. The convergence of DE development models will promote the achievement of ICCG. Compared to regions with underdeveloped digital infrastructure, areas with better digital infrastructure are more conducive to the DE empowering ICCG. Free trade zones, compared to non-free trade zones, enjoy more digital dividends and technological innovation advantages, with the former having greater empowerment advantages. The impact of betweenness centrality, degree centrality, and closeness centrality on the DE empowering ICCG is in the order of betweenness centrality > degree centrality > closeness centrality, with the efficiency of resource allocation and control being key factors affecting the heterogeneity of the DE’s empowerment of ICCG.

From the perspective of empowerment mechanisms, the innovation chain, industrial chain, and talent chain are the three main mechanisms through which the DE empowers ICCG. The DE effectively breaks away from the traditional extensive production methods that rely on resource consumption and labor intensity by enhancing technological innovation, promoting industrial structure upgrading, cultivating technological talents, and narrowing regional educational gaps. This leads to the optimal allocation of resources, driving the DE to empower ICCG.

6. Suggestions

(1)

Optimize and improve the structure of the ICCG network, and stimulate the multi-level linkage and integration of network node advantages and the DE. Relying on the advantages of network nodes, establish a “leading-following” ICCGN. For example, Beijing, Tianjin, Shanghai, and Jiangsu are in the “dual-spillover” sector. When formulating energy-saving and emission-reduction policies, they should focus on leveraging the advantages of regional collaborative governance to promote the complementary advantages of talents, technology, industry, and capital between developed and underdeveloped areas. For sectors like Hunan, Hainan, and Chongqing in the “net beneficiary” category, they are more likely to receive spatial spillover from external technological innovations in carbon collaborative reduction. Therefore, the “net beneficiary” sectors should increase support for the DE by introducing carbon reduction environmental technologies and high-end research talents from developed areas, forming a multi-center support and multi-level linkage ICCG pattern in line with sector attributes.

(2)

Cultivate and develop the regional DE to form an innovation-driven belt that leads transformational development. Relying on regional coordinated development strategies, create leading areas of the DE and continuously deepen the integration between the DE and ICCG. Strengthen the construction of the new generation of digital infrastructure, accelerate the establishment of an extensive ICCGN system, and provide a solid support platform for ICCG. Implement differentiated regional governance strategies to narrow the DE regional gap. For regions with underdeveloped digital economies, leverage their comparative advantages to achieve technological innovation through clustered industrial development, strengthen innovative open cooperation, and cultivate new momentum and upgrade traditional momentum. For regions with developed digital economies, they should play a “benchmark” demonstration role in digital empowerment, create regional technological demonstration highlands, cultivate innovation growth poles, and drive the transformational development of regional industries and economic growth.

(3)

Promote the formation of a multi-chain integration empowerment mechanism, including innovation chains, industrial chains, and talent chains, to reshape the new pattern of ICCG. The government should increase support for high-tech strategic emerging industries and accelerate the transformation of scientific and technological achievements, advance the construction of DE and innovation systems to serve the DE, form a synergistic force that integrates industrial development and technological innovation, accelerate the development of strategic emerging industries, and achieve a green transformation of economic development. Enterprises should adjust their industrial structures, eliminate backward industries with high pollution, high energy consumption, and low added value, and accelerate new paths of digital–physical integration. Leverage the chain-multiplying effect of digital productivity by embedding digital technology into every aspect of industrial structure transformation and upgrading. Simultaneously, the government should increase support for talents and strive to reduce regional educational disparities. Exploring regional collaborative governance requires balancing coordination and sustainability across multiple dimensions, subjects, and objectives, such as urban–rural and market integration, to narrow the gaps in urbanization and marketization.