Impact of Digital Economy Industrial Agglomeration on Carbon Emissions: A Case Study of the Four City Clusters Along the Eastern Seaboard of China

Abstract

Digital economy industrial agglomeration is significant for economic development and the realization of “dual carbon” goals. Based on the point of interest (POI) data of digital enterprises, this study uses kernel density estimation, a fixed-effect model, a spatial Durbin model, and other methods to analyze the spatiotemporal characteristics of digital industrial agglomeration in the four major urban clusters along the east coast of China and examines their corresponding influence on carbon emissions, including spatial spillover effects. The key conclusions are as follows: First, digital industry development and the degree of agglomeration display increasing trends. The degree of agglomeration is high in the east and low in the west, with high-value areas characterized by core prominence and orderly expansion. Second, a negative relationship is observed between digital industrial agglomeration and carbon emissions, with specialized agglomeration significantly reducing carbon emissions, while diversified agglomeration has a weaker effect. Third, the influence of digital industrial agglomeration on carbon emissions exhibits spatial spillover effects with heterogeneity. These findings provide a theoretical basis for the development of regional digital industry agglomeration and have significance as a reference for the formulation of energy conservation and carbon-reduction policies.

1. Introduction

Climate change has emerged as a pressing global challenge, driving nations to collaborate on sustainable development goals while implementing proactive measures to curb carbon emission growth [1]. In September 2020, China proposed the “dual carbon” goal, which is not only China’s emission-reduction commitment to actively engage in global environmental governance but also an inevitable requirement for low-carbon transformation. The “Action Plan for Carbon Peaking before 2030” further proposed that all regions across China should advance carbon peaking through an orderly progression. At the forefront of efforts to reduce carbon emissions, representatives from almost 200 countries congregated at the 27th session of the Conference of the Parties (COP27) in 2022 to discuss measures for carbon reduction and for limiting increases in temperature [2]. According to the China Carbon Accounting Database, China’s cumulative carbon emissions reached 11 billion tons in 2022, accounting for approximately 28.87% of global carbon emissions. Critically, research indicates that about 85% of carbon emissions stem from cities in China [3], reflecting a concentrated pattern of carbon emissions [4]. Therefore, reducing urban emissions is a key factor in achieving carbon neutrality.

At present, China is facing dual challenges of high-quality economic transformation and coping with climate change. The digital economy, which integrates digital technologies with societal systems under green economic principles characterized by low energy consumption, minimal pollution, and reduced emissions, serves as a critical nexus for reconciling these challenges. While restructuring industrial value chains and injecting new kinetic energy into the economy, this transformation has also produced complex environmental effects; still, the digital economy has aided in coordinating economic growth and environmental preservation [5]. Advancing the pace of digital economy development and building digital industry clusters with national competitiveness have become important paths to mitigate pollution and reduce carbon emissions. China’s digital industry clusters are located mainly in the eastern coastal region, of which the Beijing–Tianjin–Hebei, Yangtze River Delta, Pearl River Delta, and Shandong Peninsula city clusters are core areas. They are not only the drivers of economic development but also the keys to realizing carbon reduction and sustainable progression regionally [6]. The scale-economy effect of digital industry agglomeration and the rebound effect of digital technology affect carbon dioxide (CO₂) emissions. Consequently, analyzing the spatiotemporal characteristics of digital industrial agglomeration in typical regions and exploring its profound influence on carbon emissions as well as spatial spillover are of important scientific significance and application value for realizing regional green transformation.

Given this background, we explore the following questions (RQs): RQ1: How did the digital industry evolve in eastern Chinese coastal city agglomerations? RQ2: What has been the impact of digital economy industrial agglomeration on carbon emissions, and are there spatial spillover effects? On the basis of panel data from four city clusters, a two-way fixed-effects model and a spatial Durbin model are used to analyze the relationship between digital industrial agglomeration and carbon emissions. Finally, we propose several policy recommendations aimed at promoting the digital industry and facilitating the low-carbon transformation of the four city clusters along China’s east coast.

The rest of this study is organized as follows: Section 2 presents a literature review, Section 3 proposes the theoretical mechanisms and research hypotheses, Section 4 is the methodology and data sources, Section 5 presents the empirical results, Section 6 presents the discussion, and Section 7 summarizes the conclusions.

2. Literature Review

Industrial agglomeration, defined as a spatial organization pattern emerging from the evolution of the industrial division of labor, is characterized by the geographic concentration of industrial capital elements. Through this process, interconnected industries establish competitive–cooperative relationships that yield external scale economies, optimize business environments, and amplify innovation capacities [7]. Extant scholarship has systematically investigated the multidimensional impacts of industrial agglomeration, with particular attention to its role in green transformation [8], land-use efficiency [9], urban–rural income disparity [10], and energy efficiency [11]. Industrial agglomeration holds a vital position in stimulating economic development [12,13], creating scaled economies, and reducing production costs and has an optimistic impact on green development [14]. With respect to the resource and environmental effects generated by industrial agglomeration, there are three prevalent views: First, industrial agglomeration offers significant environmental benefits by saving energy and minimizing resource consumption [15]. Ma and Yao [16] reported that expanding the scale of urban agglomerations and promoting industry diversification strengthen regional specialization and cross-jurisdictional collaboration and, ultimately, enhance carbon emission efficiency. Chen et al. [17] also confirmed that technological cooperation and knowledge spillover among industrial agglomerations serve as effective mechanisms for pollution mitigation. Second, a nonlinear relationship may exist. Li et al. [18] argued that industrial agglomeration can effectively contribute to decreasing the carbon intensity with reasonable resource allocation, but if the resource allocation surpasses a given threshold, the agglomeration effect will be transformed into a crowding effect, increasing the carbon intensity. Wu et al. [19] also emphasized that industrial agglomerations have to develop to a certain threshold before positive effects are observed. Third, an inhibitory effect may exist. For example, it has been reported that the agglomeration of the energy industry [20], thermal power industry [21], and manufacturing industry [22] increases energy utilization and environmental pollution.

Compared with ordinary industries, the digital economy industry led by data resources is characterized by high permeability, with the potential to extend the geospatial agglomeration of the real economy into the virtual space, promote industrial linkages, and facilitate spatial relations. China’s digital economy has been developing rapidly. During this process, industrial centers have generally converged, gradually forming a “4 + N” development situation dominated by the YRD, PRD, BTH, and Chengdu–Chongqing urban agglomerations and showing a decreasing spatial distribution from the coast to inland areas [23]. There are significant regional differences in the development of the digital industry, and the distribution of enterprises is highly consistent with the economic level, city size, and development degree of urban agglomerations, which reflects the geographical location and economic and social preferences for the distribution of digital enterprises [24,25]. Digital industry development positively affects the digital transformation of enterprises by increasing innovation investment and reducing operational costs, and it plays an essential role in enhancing the employment structure, improving human capital, and other economic and social effects [26,27,28].

The digital industry is becoming increasingly dominant in the overall economy [29], producing not only notable economic effects but also potential resource and environmental effects. A related study indicated that the digital economy can foster industrial agglomeration, which has an indirect impact on air pollution management [30]. With respect to the relationship between digital industry development and carbon emissions, some studies have suggested that the digital economy is environmentally friendly and that the digital industry can promote green transformation through breakthroughs in production technology, business models, and industrial institutions [31,32,33]. Digital innovation is a force for urban development [34,35]. The effect of digital industrial agglomeration on innovation exhibits a gradient enhancement accompanied by positive spatial spillover effects, which are beneficial for green development and the achievement of the “dual carbon” goal [36]. For example, Lu et al. show that digital industrial agglomeration is conducive to urban innovation [37]. Digital industrial agglomeration can promote technological cooperation, accelerate the digital transformation of traditional enterprises, and enhance production and operation efficiencies, thus promoting energy conservation and environmental protection. Yang et al. verified the inverse U-shaped relationship between digital industry agglomeration and technological innovation of manufacturing enterprises and showed that technological innovation inhibited carbon emissions by these enterprises [38]. However, some researchers have stated the opposite opinion, arguing that the high demand for electricity due to digital industrial agglomeration and the increased energy use of digital technologies have exacerbated resource consumption and carbon emissions [39,40]. Therefore, the impact of digital industry agglomeration on carbon emissions still needs to be discussed.

In summary, while existing studies provide foundational theoretical frameworks and methodological guidance for research on the impact of digital industrial agglomeration on carbon emissions, there is still room for expansion. First, current methodologies inadequately leverage enterprise big data to quantify the digital industry’s developmental scale, geospatial clustering patterns, and regional heterogeneity. Second, the existing studies focused on the national and provinces levels, neglecting comparative examinations across distinct urban agglomerations. Third, the impact of digital industries on carbon emissions has not been sufficiently considered in terms of agglomeration externalities. To address these limitations, we examined the spatiotemporal evolution of digital industrial agglomeration on the basis of enterprise data from four major city clusters along China’s eastern coast and explored the impacts of the digital industry on carbon emissions and their spatial spillover effects from the perspectives of the specialization and diversification of agglomerations to supplement the existing research.

3. Theoretical Mechanisms and Research Hypotheses

3.1. The Impact of Digital Industrial Agglomeration on Carbon Emissions

With the characteristics of innovation, openness, and sharing, the digital industry has gradually formed a large-scale industrial cluster. The externalities arising from the geospatial agglomeration of industries include two main types: One type is the Mar externality, which emphasizes specialized industrial agglomeration [41]. In this case, the clustering of single or similar industries facilitates the sharing of infrastructure and comparative advantages. The other is the Jacobs externality, which emphasizes diversified industrial agglomeration [42]. Notably, the agglomeration of different types of industries can lead to a division of labor and cooperation, creating complementary effects and forming industrial networks [43,44]. The externalities of industrial agglomeration emphasized by the new economic geography include knowledge and technology spillover, specialized labor markets, infrastructure sharing, etc. [45,46]. Digital industries are talent- and technology-intensive, and their clustering can enhance regional competitiveness, attract high-quality professionals, and form a labor force “reservoir”. Industrial agglomeration improves the resource utilization efficiency, and enterprises can conveniently access external resources to accelerate the dissemination of R&D knowledge. Additionally, green innovation technology can be enhanced, and an “innovation compensation effect” can be achieved to reduce technological overlap and waste. In addition, environmental pressure and competition in similar industries can reduce resource consumption and promote regional carbon reductions. However, if the agglomeration level exceeds a specific threshold, the “congestion effect” will outweigh the “scale effect”, which is not beneficial for resource savings and carbon reduction [47]. Our first hypothesis is therefore as follows:

H1. 

Digital industrial agglomeration can effectively reduce carbon emissions.

3.2. Spatial Spillover Effects of Digital Industrial Agglomeration on Carbon Emissions

Digital industrial agglomeration strengthens the structure of the network economy. Due to the intangibility of digital elements and the effectiveness of information transmission, digital industries transcend spatiotemporal constraints inherent to traditional production factors, thereby enabling cross-regional technology sharing, knowledge diffusion, and resource synergies. This convergence not merely diminishes local carbon emissions but also exerts a spatial spillover effect on carbon reductions in surrounding regions. On the one hand, emission-reduction policies, technological innovations, and capital and talent pools in core areas have a “demonstration effect” and a “spillover effect” on surrounding areas [48]. Through emulation and knowledge transfer, traditional industries have also given rise to intelligent manufacturing, the Internet of Things, platform economies, and other new business forms. Through the integration of data elements with the traditional elements of labor, capital, and technology, productivity and energy efficiency can be improved, and the utilization of highly polluting and energy-consuming resources can be curbed [31]. On the other hand, when agglomeration in a core area achieves a certain scale, the digital industry will expand to neighboring areas through industry transfer, headquarters and branches, and the establishment of supply chain networks that connect, and thus promote, digital development in the surrounding regions. In addition, digital industrial agglomeration may produce negative spillover effects, and the agglomeration area may have a “siphon effect” that attracts laborers and enterprises, resulting in local convergence. Blind agglomeration will result in energy consumption and carbon emissions [49]. Therefore, we propose a second hypothesis.

H2. 

Digital industrial agglomeration has spatial spillover effects on carbon reductions in the four major city clusters, and the effects in different city clusters are heterogeneous.

4. Methodology and Data

4.1. Methods and Models

4.1.1. Kernel Density Estimation

The kernel density estimation (KDE) model is commonly used to estimate the characteristics of overall distributions based on sample point data, and we used the KDE model to explore the levels of spatial agglomeration and disaggregation of enterprises [50,51]. The formula is as follows:

$$f\left(x\right)={\displaystyle \frac{1}{nh}}{\sum }_{i=1}^{n}K\left({\displaystyle \frac{x_i-x}{h}}\right)$$

(1)

where K(·) is the kernel function, h is the bandwidth, n is the sample number, and $x_i-x$ is the distance from the estimated point to the sample point.

4.1.2. Spatial Autocorrelation

Moran’s I was used to assess the spatial autocorrelation of carbon emissions in the four urban agglomerations along China’s eastern coast to determine whether there are any aggregation characteristics. The formula is as follows:

$$I={\displaystyle \frac{\sum _{i=1}^n\sum _{j=1}^n{w}_{ij}\left(\left(x_i-\overline{x}\right)\left(x_j-\overline{x}\right)\right)}{S^2\sum _{i=1}^n\sum _{j=1}^n{w}_{ij}}}$$

(2)

where I ranges from −1 to 1. When I > 0, a positive spatial correlation exists; when I < 0, a negative spatial correlation exists; and when I = 0, this reveals no spatial correlation. $x_i$ and $x_j$ are the attribute values of the spatial units $i$ and $j$. $\overline{x}$ is the mean of the attribute values, $S^2$ is the sample variance, and $w_{ij}$ is the spatial weight matrix.

4.1.3. Basic Model

The following model was constructed to regressively analyze the impact of digital industrial agglomeration on carbon emissions in the four major city clusters. The model expression is

$${CE}_{i,t}={\alpha }_0+{\alpha }_1{AGG}_{i,t}+{\alpha }_2{X}_{i,t}+{{\mu }_i+\lambda }_t+{\epsilon }_{i,t}$$

(3)

where $i$ represents an individual, $t$ represents time, ${CE}_{i,t}$ and ${AGG}_{i,t}$ represent the carbon emission and digital economy industrial agglomeration levels, $X_{i,t}$ denotes the control variables, ${\mu }_i$ and ${\lambda }_t$ represent individual and time fixed effects, and ${\epsilon }_{i,t}$ is the random error term.

4.1.4. Spatial Econometric Model

To examine the spatial effect, the SDM was established for regression, and the results were further decomposed into direct and indirect effects. The model expression is

$${CE}_{i,t}=\alpha +\rho {\sum }_{j=1}^n{W}_{ij}{CE}_{j,t}+\beta X_{i,t}+\theta {\sum }_{j=1}^n{W}_{ij}{X}_{j,t}+{\epsilon }_{i,t}+{{\mu }_i+\lambda }_t$$

(4)

where $W_{ij}$ is the spatial weight matrix, and $\rho ,\beta ,\mathrm{a}\mathrm{n}\mathrm{d}\theta$ are the coefficient equivalents. The other variables are the same as those in Equation (3). When $\theta =0$, the SDM can be reduced to the SLM, and when $\theta =-\rho \beta$, the SDM can be reduced to the SEM.

4.2. Variables and Data

4.2.1. Explained Variable: Carbon Emissions (CE)

According to the carbon emission coefficients of different energy types provided in the 2006 IPCC Guidelines and the energy consumption of each city provided in the Statistical Yearbook, the total carbon emissions from 2011 to 2022 from the four major city clusters along the eastern seaboard of China were calculated. The calculation method was based on that of Wu et al. [52].

4.2.2. Explanatory Variable: Level of Digital Economy Industrial Agglomeration

In most industrial agglomeration measurement methods, the location entropy (LQ), industry concentration (CRn), Herfindahl index (HHI), EG index, and similar indices are used [53,54]. In this study, the number of digital economy enterprises was used to calculate the specialized agglomeration index and the diversified agglomeration index [55]. We analyzed the impact of different agglomeration types on carbon emissions from the perspectives of industrial competition and industrial synergy.

(1) The specialized agglomeration index (RZI): Specialized agglomeration reflects the degree of agglomeration for the same type of industry and the level of competition in the same industry. The calculation is as follows:

$$RZI=(t_{it}/T_t)/(f_{it}/F_t)$$

(5)

where $t_{it}$ and $T_t$ are the number of digital economy enterprises in city i and the total area in year t, respectively. $f_{it}$ and $F_t$ are the total number of enterprises in the corresponding regions. If RZI > 1, the industry has a specialization advantage, and the larger the RZI is, the greater the comparative advantage of this type of industry.

(2) The diversified agglomeration index (RDI): Diversified agglomeration reflects the horizontal agglomeration of firms across industries, forming scale economies through industrial linkages. The calculation is as follows:

$$RDI=1/\sum _j\left|S_{ij}-S_j\right|$$

(6)

where $S_{ij}$ is the share of enterprises in industry j to the total number of firms in city i, and $S_j$ is the ratio of the number of enterprises in industry j to that in the whole region.

4.2.3. Control Variables

Considering the various factors that affect carbon emissions, we used the industrial structure (IS), economic development level (PGDP), openness to the outside world (FDI), and population density (POP) as control variables, which are represented by the ratio of secondary industry output to GDP, per capita GDP, FDI as a ratio of GDP, and population per unit area, respectively.

4.2.4. Data Sources and Processing

The digital industry data were sourced from an industrial and commercial enterprise registration data platform (https://www.qcc.com, accessed on 31 December 2024. We constructed a digital industry database for the four major city clusters along China’s east coast. The detailed process was as follows: First, referring to [56,57], we categorized the digital industries into industrial digitization and digital industrialization classes based on the Statistical Classification of the Digital Economy and Its Core Industries (2021), as shown in

Table 1. Classifications for the digital economy industry.

Classification	Digital Economy Industry
Digital industrialization	Computer communications and other electronic equipment manufacturing
	Telecommunications, radio, television, and satellite transmission services
	Internet-related services
	Software and information technology services
Industrial digitalization	Smart agriculture
	Smart manufacturing
	Smart logistics
	Digital finance and trade
	Digital society and related industries
	Other digital industries

. Second, we screened the relevant enterprises to retain those with a registered capital of at least CNY 1 million as well as surviving and active enterprises while excluding individual business households, obtaining a final total of 1,478,570 valid records. Third, the latitude and longitude of each enterprise’s registered address were extracted through the Gaode API platform interface and transformed into WGS1984 coordinates. Finally, we utilized ArcGIS 10.7 to transform these points into point vector data.

The estimation of carbon emissions was based on the carbon emission factors for different energy types provided in the 2006 IPCC Guidelines and the consumption of various energy sources in each city reported in the China Urban Statistical Yearbook and the China Electricity Statistical Yearbook. Data for the control variables were sourced from the China Urban Statistical Yearbook and the statistical yearbooks from each province and city. To reduce the effect of heteroskedasticity, the variables were log-transformed. All the variables are presented in

Table 2. Descriptive statistics of the variables.

Variables	Obs	Mean	S.D.	Min	Max
lnCE	869	6.650	1.044	4.039	9.052
lnRZI	869	−0.531	0.481	−2.051	0.890
lnRDI	869	3.292	0.393	2.097	4.832
lnIS	869	3.820	0.195	2.762	4.314
lnPGDP	869	11.053	0.577	9.219	13.056
lnFDI	869	0.538	0.850	−2.620	2.438
lnPOP	869	6.366	0.544	4.550	8.058

.

5. Empirical Results

5.1. Spatiotemporal Evolution of Digital Industrial Agglomerations

5.1.1. Temporal Evolution

The number of digital economy enterprises in the four major city clusters along China’s eastern seaboard has grown significantly, rising from 126,000 in 2011 to 1.478 million in 2024, representing an 11.7-fold increase (Figure 1). During the pre-2016 period, the accelerated growth phase saw annually increasing growth rates. After 2016, a quality-improvement phase emerged, where although the growth rate moderated, the absolute annual growth volume continued to rise, driving continual expansion of the digital industry. Recent advancements in China’s digital infrastructure and smart city development, coupled with national policies including tax incentives, talent cultivation, and financial support, had fostered a favorable environment for digital economy enterprises. Among the four city clusters, the YRD led with 609,100 enterprises, followed by the PRD (483,900), BTH (196,100), and SP (189,400). By of 2024, these clusters collectively accounted for 50.81% of China’s total digital economy enterprises. The synergistic effect and industrial complementarity within the city clusters provided strong support for the industrial chain, supply chain and innovation chain, propelling China’s digital industry advancement.

In terms of the agglomeration indices, both the specialized agglomeration index and the diversified agglomeration index demonstrated upward trends (Figure 2). The average value of the RZI increased from 0.62 in 2011 to 0.74 in 2024, while the average RDI rose from 24.57 in 2011 to 36.15 in 2024. Notably, the RDI transitioned from a low-value concentrated distribution to a high-value dispersed pattern, whereas the RZI exhibited convergence tendencies. Both the RZI and RDI indices displayed right-skewed distributions, indicating that the levels of specialization and diversification of digital industries in most regions remained below the overall average degree.

5.1.2. Spatial Evolution

There are differences in the degree of digital industrial agglomeration among the four major city clusters (Figure 3). Both the RZI and RDI displayed spatial distributions characterized by higher levels in the east and lower levels in the west, with digital industrial agglomeration in coastal cities generally exceeding that in inland cities. In terms of specialized agglomeration, while the number of low-value cities decreased markedly over time and the medium-value cities increased in number, the high-value areas remained stable. In 2011, the RZI in most regions was less than 0.6, whereas those in core cities exceeded 1.0, such as in Beijing, Tianjin, Jinan, Nanjing, Guangzhou, and Shenzhen, indicating their digital industries’ specialized agglomeration advantage. Shenzhen held a relatively leading position, with a specialization agglomeration index of 2.19. Over time, the RZI across cities became more balanced, and low-value cities disappeared. In terms of diversified agglomeration, the total number of low-value cities decreased, evolving into a decentralized distribution, and the number of high-value cities increased significantly. This transition resulted in 30 cities achieving RDI values exceeding 30.01, representing approximately 38% of all the surveyed cities. The high-value areas of diversified agglomeration gradually extended from coastal cities to inland cities; for example, in the YRD, this expansion progressively extended westward and southward, with Shanghai as the starting point.

The digital industries in the four major city clusters displayed a spatial evolution trend of core prominence and orderly expansion (Figure 4). Notably, the number of high-kernel-density zones increased, and the highest values rose. Digital economy enterprises exhibited a certain “polarization effect”, which indicated that new enterprises preferred locations in existing industrial clusters, and the layout of enterprises displayed significant path dependence. In 2011, the high-kernel-density areas in BTH were located mainly in Beijing, Tianjin, and Shijiazhuang, and numerous smaller high-density areas gradually emerged in other core cities. The Shandong Peninsula evolved from an obvious Jinan–Qingdao double-core structure to that of a multicore structure. The YRD formed a V-shaped high-value zone, with Shanghai as the apex and two axes extending toward Nanjing and Hangzhou. The high-kernel-density areas in the PRD were mainly concentrated in Guangzhou and Shenzhen. The scope of these areas decreased slightly, but the number of enterprises rose sharply, and the concentration of digital economy enterprises increased.

5.2. Baseline Estimation Results

Aiming to investigate the impact of digital industrial agglomeration on CE in China’s four major eastern coastal city clusters, we applied individual fixed-effect (Ind), time fixed-effect (Time), and two-way fixed-effect (Both) models for regression in conjunction with the constructed panel data regression model (

Table 3. Baseline regression results.

Variables	(1)	(2)	(3)	(4)	(5)	(6)
	Ind	Time	Both	Ind	Time	Both
lnRZI	−0.004	−0.217 ***	−0.940 ***
	(−0.23)	(−5.32)	(−2.98)
lnRDI				0.027	−0.132 **	−0.049
				(1.56)	(−2.20)	(−0.64)
lnPOP	0.278 ***	0.103 ***	−0.495 ***	0.279 ***	0.475 ***	−0.735 ***
	(6.23)	(4.11)	(−2.03)	(6.09)	(11.19)	(−2.83)
lnIS	−0.223 ***	−1.641 ***	1.074 ***	−0.399 ***	−0.889 ***	0.300 *
	(−3.89)	(−20.43)	(5.11)	(−14.01)	(−6.94)	(1.72)
lnPGDP	0.112 ***	0.481 ***	−0.178 **	0.177 ***	1.069 ***	−0.011
	(7.64)	(13.97)	(−2.33)	(14.59)	(24.70)	(−1.34)
lnFDI	0.001	−0.053 ***	0.065 ***	0.002	−0.093 ***	0.041
	(0.13)	(−3.43)	(2.59)	(0.28)	(−3.63)	(1.50)
Cons	4.899 ***	8.692 ***	1.282	5.976 ***	−4.364 ***	12.550 ***
	(9.44)	(20.70)	(0.57)	(16.41)	(−6.82)	(7.15)
City fixed effect	YES	NO	YES	YES	NO	YES
Year fixed effect	NO	YES	YES	NO	YES	YES
R2	0.2385	0.4388	0.9013	0.2501	0.4294	0.9154

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

). Columns (1)–(3) show that the specialized agglomeration of digital industries negatively affects carbon emissions, whereas the regression results for diversified agglomeration are different (Columns (4)–(5)). Because the data were subject to the influence of both spatial and temporal changes, the estimation results of the two-way FE model were prioritized for analysis.

The degree of digital industrial agglomeration exhibits a statistically significant negative correlation with carbon emissions, thereby confirming H1. As shown in Column (3), specialized agglomeration has a significant inhibitory influence on CE, with a correlation coefficient of −0.940 and with the significance test passed at the 1% level. This outcome suggests that the externality of specialized agglomeration in the digital industry drives the agglomeration of talent, technology, capital, information, and other innovation factors, thereby generating scale effect; consequently, resources and energy can be utilized efficiently and reasonably. In contrast, the correlation coefficient between diversified agglomeration and carbon emissions is −0.049, but this fails the significance test, implying limited decarbonization potential from this agglomeration mode. This discrepancy likely stems from the unbalanced development of diversification in the digital industry, which makes it difficult to form effective multifactor flows and is not favorable for realizing the agglomeration effect.

5.3. Robustness Check

To validate the robustness of our regression findings, we conducted three distinct tests (

Table 4. Robustness test results.

Variables	(1)	(2)	(3)	(4)	(5)
lnRZI	−0.117 *		−0.469 **
	(−1.67)		(−2.00)
lnRDI		−0.225 ***		−0.063
		(−4.72)		(−0.76)
lnNUM					−0.016 *
					(−1.67)
Cons	−12.186 ***	−11.549 ***	7.187 **	11.992 ***	11.316 ***
	(−14.75)	(−18.68)	(2.54)	(6.38)	(47.04)
Control	YES	YES	YES	YES	YES
City fixed effect	NO	NO	YES	YES	YES
Year fixed effect	NO	NO	YES	YES	YES
R2	0.7540	0.7702	0.9195	0.9225	0.6882

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

). First, we added control variables and replaced the regression models. We used the OLS model and added two control variables to the original data: the college-educated population per 10,000 residents and the prevalence of households with internet connectivity. Columns (1) and (2) show that the inhibitory effects of the RZI and RDI of digital industries on CE remain stable, with regression coefficients of −0.117 and −0.225. Second, we lagged the explanatory variables by one period, as shown in Columns (3) and (4), with coefficients of −0.469 and −0.063. Third, the explanatory variable was replaced with the number of digital economy firms. Column (5) shows that improvements in the digital industry are negatively correlated with CE at the 10% significance level, with a correlation coefficient of −0.016. In summary, the test results of the three methods support that digital industrial agglomeration in the four major urban agglomerations has notably reduced regional carbon emissions, and the findings are confirmed to be robust and reliable.

5.4. Comparative Analysis of City Clusters

To further assess the regional heterogeneity of digital industrial agglomeration’s impact on CE, we applied a two-way FE model to estimate the relations across different city clusters (

Table 5. Regression results for the four urban agglomerations.

Variables	BTH		SP		YRD		PRD
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
lnRZI	−0.505 **		−0.181		−0.327 ***		0.774 ***
	(−2.00)		(−0.55)		(−2.60)		(2.91)
lnRDI		−0.393 **		−0.418		−0.441 *		0.278 **
		(−2.28)		(−1.21)		(−1.67)		(2.42)
Cons	−16.770 ***	−14.994 ***	31.319 **	34.249 ***	−2.137	10.493 ***	16.254 ***	18.197 ***
	(−4.05)	(−3.41)	(2.45)	(2.66)	(−0.51)	(2.96)	(5.78)	(6.45)
Control	YES	YES	YES	YES	YES	YES	YES	YES
City and year fixed effect	YES	YES	YES	YES	YES	YES	YES	YES
R2	0.9607	0.9651	0.8261	0.8275	0.9366	0.9326	0.9563	0.9549

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

). The regression coefficients of the RZI and RDI in BTH are statistically significant at the 5% level, measuring −0.505 and −0.393, respectively. Similarly, in the YRD, both agglomeration types show significantly negative coefficients of −0.327 and −0.441. The government has actively promoted digital development in the BTH and the YRD. Over time, the radiating influence of core cities such as Beijing, Shanghai and Nanjing has increased significantly. The use of digital technologies in BTH and the YRD fostered efficient circulation and rational allocation within city clusters, continuously improving resource utilization efficiency while transforming and upgrading industrial structures, thereby reducing CE. In contrast, the negative impact in SP remains statistically insignificant. This may stem from SP’s lack of a conducive ecosystem for digital economy development, effective industrial networks, and stringent environmental regulations. These factors collectively resulted in limited emission reductions. Both RZI and RDI in the PRD are positively correlated with CE, potentially attributable to the excessive number and concentration of digital industries, where a “crowding-out effect” emerges, triggering overcapacity, reducing resource efficiency, and, consequently, elevating CE.

5.5. Comparative Analysis of Different Industries

We used the number of enterprises to empirically explore the impact of different digital industries on CE (

Table 6. Regression results for the different types of industries.

Variables	(1)	(2)	(3)	(4)	(5)
lnCOMP	0.0252 ***
	(4.91)
lnTELE		−0.345 ***
		(−2.74)
lnINTER			−0.069 ***
			(−4.15)
lnSOFT				−0.134 ***
				(−6.28)
lnDIG					−0.296 ***
					(−6.98)
Cons	8.167 ***	4.932 **	2.642	2.985	5.346 **
	(3.86)	(2.24)	(1.14)	(1.39)	(2.58)
City and year fixed effect	YES	YES	YES	YES	YES
Control	YES	YES	YES	YES	YES
R2	0.9244	0.9228	0.9237	0.9258	0.9267

Note: ** p < 0.05, and *** p < 0.01; t statistics in parentheses.

). Columns (1)–(4) are the regression results of computer communications and other electronic equipment manufacturing (COMP); telecommunications, radio, television, and satellite transmission services (TELE); internet-related services (INTER), and software and information technology services (SOFT), respectively. Column (5) is the result of the digitization efficiency enhancement industry (DIG) and represents all industrial digitalization. The production process of COMP requires a large amount of energy and raw materials, which may form resource competition with other traditional industries, resulting in uneven resource distribution and a positive correlation with CE. As technology-intensive industries, TELE, INTER, and SOFT significantly reduced CE in the four major urban agglomerations, with impact coefficients of −0.345, −0.069, and −0.134, respectively. These three types of industries can innovate green technology, improve the level of information technology, improve production efficiency, and reduce energy consumption. The DIG also has a significant negative impact. The digital industry promotes the integration of digital technology and traditional industries and the digital development of enterprises, which is conducive to reducing CE.

5.6. Spatial Spillover Effect

Prior to analyzing spillover effects, we assessed the spatial autocorrelation trend (

Table 7. The Moran’s I results.

Year	Moran’s I	p Value	Year	Moran’s I	p Value
2011	0.167	0.002	2017	0.239	0.000
2012	0.176	0.001	2018	0.225	0.000
2013	0.174	0.001	2019	0.231	0.000
2014	0.187	0.000	2020	0.251	0.000
2015	0.178	0.001	2021	0.262	0.000
2016	0.128	0.019	2022	0.278	0.000

). The Moran’s I values of CE ranged from 0.128 to 0.278, which were all statistically significant at the 1% level. These results demonstrate a persistent and intensifying positive spatial dependence in CE distributions over time, thereby validating the appropriateness of spatial econometric approaches for our analysis.

Based on the geographic distance matrix, we explored the spatial spillover effects of both RZI and RDI of the digital industry on CE across the four major city clusters along China’s eastern coast. The degradation of the spatial econometric model into a general econometric model was rejected based on the LM test, and the LR test indicates that the SDM cannot be simplified to the SLM or SEM; therefore, we selected the SDM for regression analysis (

Table 8. Selection test of the spatial econometric model.

Spatial Autocorrelation Test	BTH		SP		YRD		PRD
	RZI	RDI	RZI	RDI	RZI	RDI	RZI	RDI
LM (lag)	12.761 ***	10.047 ***	24.510 ***	19.563 ***	20.878 ***	30.633 ***	4.609 **	15.110 ***
Robust LM (lag)	18.170 ***	14.566 ***	18.951 ***	10.794 ***	11.003 ***	20.334 ***	5.844 **	10.650 ***
LM (error)	11.516 ***	4.247 **	13.483 ***	32.480 ***	104.109 ***	64.725 ***	25.928 ***	12.629 ***
Robust LM (error)	16.925 ***	8.766 ***	7.924 ***	23.711 ***	94.234 ***	54.427 ***	27.162 ***	8.170 ***
LR (lag)	12.780 ***	22.290 ***	15.950 ***	31.130 ***	14.780 ***	31.710 ***	45.870 ***	38.760 ***
LR (error)	19.620 ***	21.260 ***	12.020 **	25.780 ***	27.040 ***	35.970 ***	47.800 ***	40.350 ***

Note: ** p < 0.05, and *** p < 0.01.

).

Table 9. Results of spatial spillover effects.

	lnRZI			lnRDI
	Direct	Indirect	Total	Direct	Indirect	Total
BTH	−0.538 *	−4.365 ***	−4.903 ***	−0.129	−1.058 **	−1.186 **
	(−1.74)	(−3.28)	(−3.23)	(−0.71)	(−2.10)	(−2.48)
SP	−1.595 ***	0.883	−0.706	−0.732 ***	0.318	−0.414
	(−2.60)	(0.489)	(0.476)	(−2.87)	(0.68)	(−1.21)
YRD	−0.315 **	0.738 ***	0.422 **	−0.078	0.558 **	0.478 **
	(−2.45)	(3.01)	(2.44)	(−0.91)	(2.47)	(2.24)
PRD	1.117 ***	0.907	2.02	0.619 ***	2.249 ***	2.869 ***
	(3.48)	(0.63)	(1.22)	(3.84)	(2.72)	(3.01)
Control	YES	YES	YES	YES	YES	YES
City and year fixed effect	YES	YES	YES	YES	YES	YES

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

presents the SDM estimation results. Both the direct and indirect effects of digital industry specialization agglomeration in the BTH are negative, indicating that digital industry agglomeration significantly reduces local emissions, and generates a potential spatial spillover effect, facilitating carbon reductions in adjacent areas. As a center of scientific and technological innovation, the digital enterprises gathered in Beijing have developed low-carbon technologies through collaborative innovation and have spread these to neighboring regions through industrial chain division. For the SP cluster, the direct effect coefficient is negative at the 1% confidence level, whereas the indirect effect remains statistically non-significant, which indicates that high RZI and RDI values are associated with low CE without a “demonstration effect” on neighboring regions. SP mainly comprises hardware manufacturing and lacks software service output capacity. The RZI in the YRD significantly decreases local CE, while the direct effect of the RDI is insignificant, and the indirect effects show positive values. The “siphoning effect” of the central cities concentrates digital and innovation resources locally, potentially creating negative spillover effects on CE in neighboring regions due to resource competition. Both direct and indirect effects in the PRD are positive, suggesting increased CE locally and in the surrounding areas. This is most likely due to the “congestion effect” caused by excessive digital industrial agglomeration, which intensifies competition and increases energy consumption, and the surrounding areas need to undertake industrial transfer, which hinders efforts to reduce CE. Collectively, the above results support H2.

6. Discussion

Clustering represents an inevitable trend and a significant mode in modern industrial development, and the digital economy has become a key pillar of the national economy [58]. In the context of pressing climate challenges and carbon emission-control policies, it is vital to understand the level of digital industrial agglomeration and its impact on regional carbon emissions. China’s eastern coastal city clusters have actively engaged in international digital economy cooperation, gradually enhancing the competitiveness and influence of their own digital industries. The four major city clusters exhibit distinct advantages in terms of digital development, such as in leading economic development, providing high-quality talent, and accessing innovative resources. Moreover, the degrees of specialization and diversified agglomeration of digital industries are increasing, resulting in the establishment of mature digital industrial ecosystems.

There is a negative relationship between digital industry development and carbon emissions, and the studies of Chang et al. [59] and Liu et al. [60] yielded findings similar to those in this study. Chang et al. proved that the macro-level digital economy reduced carbon emission intensity by promoting the upgrading of industrial structures, while our study, based on the perspective of medium industry, proves the role of digital industry in carbon emission reduction. Compared with Liu et al.’s study, we divided the specialization and diversified agglomerations of digital industry from the perspective of agglomeration externality and discussed their respective impact on carbon emissions. The impact mechanisms are shown in Figure 5. Specialized agglomeration in the digital industries drives continuous acceleration of technological turnover, enhances information transparency, optimizes production processes and resource allocation efficiency, reduces resource waste, and reduces the marginal energy consumption of digital enterprises. In addition, digital technology promotes efficient cooperation among enterprises. By relying on innovation networks, digital industries can catalyze traditional enterprises’ digital transformation, break down sectoral barriers, and accelerate the transition of high-polluting firms toward environmentally sustainable practices. In contrast, the relatively weak carbon-reduction effect of diversified agglomeration likely stems from serious competition among homogeneous industries within city clusters, and a reasonable structure of complementary industries has not been formed, which is not conducive to factor sharing.

From the perspective of city cluster comparisons, the effects of digital industrial agglomeration on carbon emissions in different city clusters are heterogeneous. The digital industries in BTH and the YRD exhibit a clustering effect, which mitigates carbon emissions. With the exception of core cities such as Jinan and Qingdao, the majority of cities in the Shandong Peninsula remain entrenched in traditional processing and manufacturing sectors, and the intensity of environmental regulations is weak. Excessive agglomeration in the PRD has paradoxically resulted in a failure to achieve optimal economies of scale while inducing “crowding effects”, triggering overcapacity and inefficient resource utilization. The digital industrial agglomeration in each city cluster is in different stages and states and, thus, has different impacts. Regional innovation theory suggests that areas characterized by geographically similar levels of technology are more inclined toward economic imitation and creating spatial spillover effects [61]. Digital industrial agglomeration serves dual functions in fostering cross-industry labor division, collaboration, and labor-market pooling while establishing network-based platforms for inter-firm information exchange that enable knowledge and technology sharing. Both specialized and diversified agglomeration generate industrial linkages and complementarities, which further strengthen the network economy and facilitate carbon reduction through the “diffusion effect” and “learning effect”, resulting in a positive spillover. However, when the “siphon effect” exceeds the “diffusion effect”, resources continue to shift to the central nodes of the network. Alternatively, development of the local digital industry may have a squeezing effect on high-carbon enterprises, causing them to move to neighboring cities, thus exacerbating carbon emissions in the surrounding areas.

This paper has certain limitations associated with the data sources and research period. In future research, more comprehensive indicators should be considered, the time scale of analyses should be extended, and the rationale for the evaluation of digital industrial agglomeration should be further explored. In addition, the research area should be broadened to other key regions to examine the development of the digital industry and its economic and environmental effects while enhancing analyses of the corresponding influence mechanisms and paths.

7. Conclusions

In this study, the agglomeration of digital industries in four major city clusters along China’s east coast is explored using enterprise data. A two-way FE model and an SDM were constructed to empirically assess the impacts of the specialization and diversification of digital industrial agglomeration on carbon emissions and spatial spillover effects from 2011 to 2022, and the following conclusions were obtained:

(1)

The number of digital economy enterprises is exhibiting a rising trend, accompanied by a consistent enhancement in industrial clustering. However, a pronounced spatial imbalance was observed, with an uneven distribution pattern of high concentrations in the east and low in the west. Provincial capital cities and core cities displayed obvious agglomeration, with certain polarization effects. The number of zones with high nuclear density values gradually increased, and the highest values constantly increased and expanded from the center to the periphery of the city clusters. Thus, RQ1 has been addressed.

(2)

Digital economy industrial agglomeration exerts a negative inhibitory effect on carbon emissions, and specialized agglomeration demonstrates a significant effect due to the advantages of industrial association effects and economies of scale; diversified agglomeration results in a comparatively weaker carbon reduction because of the inability to form a complete industrial chain or because of the presence of homogeneous competition. The impact of each city cluster is heterogeneous, and spatial spillover effects can be either positive or negative. Notably, there is a positive spillover effect in BTH and negative spillover effects in the YRD and the PRD; however, the spillover effect in SP is not significant. Thus, RQ2 has been answered.

Based on the above conclusions, we propose a number of targeted policy recommendations. First, for the problem of regional differences in digital industry clusters, administrative barriers need to be broken to promote the gradient transfer of innovation elements. It is best to establish a unified data-factor market and computing-power trading platform and to improve the cross-regional benefit compensation mechanism in order to promote knowledge diffusion and industrial collaboration and to narrow the digital industry development gap [60]. Second, due to the weak carbon emission-reduction effect of diversified agglomeration, the government needs to provide a good market and institutional environment in which to establish a multi-symbiotic ecosystem for the digital industry. When developing suitable industries, all localities should consider the complementary effect of industries and avoid homogenized competition [62]. Third, aiming at the difference of the impact of digital industry agglomeration on carbon emissions in various urban agglomerations, governments should formulate differentiated development strategies on the basis of the advantages of the regional resource endowments and industrial foundations [55]. For example, in Beijing–Tianjin–Hebei, investments in basic science and digital frontier innovation should be enhanced. In the Yangtze River Delta, focus should be placed on cross-regional integration. In the Pearl River Delta, innovation, openness, and sharing should be prioritized and complemented by strengthened international exchange and cooperation. Moreover, in the Shandong Peninsula region, a supportive digital environment should be constructed to attract more digital industries.