Industrial Co-Agglomeration and Urban Green Total Factor Productivity: Multidimensional Mechanism and Spatial Effect

Abstract

The impact of industrial co-agglomeration (ICA) on green total factor productivity (GTFP) has garnered considerable academic attention. However, there remains a gap in research systematically investigating how ICA affects China’s GTFP within the framework of green development, specifically by analyzing transmission mechanisms, regulatory mechanisms, and spatial spillover effects. To address this gap, this study utilizes panel data from 283 Chinese cities, spanning the years 2006 to 2020, and conducts both theoretical and empirical analyses to examine ICA’s influence on GTFP through these three mechanisms. Our findings indicate that ICA significantly enhances GTFP by alleviating the mismatch of capital and energy factors but does not improve GTFP by addressing labor mismatches. Furthermore, when the intensity of local government competition exceeds a threshold of 14.3825, the positive impact of ICA diminishes, whereas an environmental regulation intensity above 0.4381 strengthens ICA’s positive effect on GTFP. ICA was found to substantially increase local GTFP and generate positive spatial spillover effects on surrounding cities within a 100 km radius. Co-agglomeration of both high-end and low-end producer services with manufacturing boosts local GTFP, while co-agglomeration of low-end producer services with manufacturing also enhances GTFP in adjacent cities. In megacities, ICA positively influences both local and nearby GTFP, whereas in large cities, ICA tends to suppress GTFP in neighboring areas. Additionally, with the exception of the Middle Yangtze River and Pearl River Delta city clusters, ICA in urban clusters enhances local GTFP; ICA in the Middle Yangtze River cluster promotes GTFP in neighboring areas, whereas ICA in the Chengdu–Chongqing cluster inhibits neighboring GTFP.

1. Introduction

Over the past 40 years since reform and opening-up, China has achieved a globally recognized “growth miracle”, with an average annual GDP growth rate of 9.433% from 1978 to 2019. This remarkable growth has primarily relied on an extensive development model driven by factor inputs. However, the costs of this approach have become increasingly evident, with excessive resource consumption, severe environmental pollution, and ongoing ecosystem degradation. As the world’s second-largest economy, China’s environmental governance performance does not align with its economic standing. According to the 2022 Global Environmental Performance Index (EPI) Report, China ranked 160th out of 180 countries, indicating substantial environmental pressure. The 2016 China Environmental Monitoring Report noted that, in 2016, the air quality exceeded standards in 254 cities, with 21.2% of days in 338 prefecture-level and higher cities surpassing air quality limits, including 2464 days of severe pollution and 784 days of extreme pollution. Furthermore, the 2020 Global Energy Transition Report from the World Economic Forum ranked China 78th out of 115 countries on the energy transition index, highlighting its relative disadvantage on the global stage. The unsustainability of the extensive development model has thus become a major obstacle to China’s green transformation and pursuit of high-quality development.

In response to the dual challenges of economic growth and environmental degradation, ecological civilization has emerged as a national strategy. In 2017, the 19th National Congress of the Communist Party of China elevated ecological civilization to the status of a “millennial project”, underscoring that environmental protection and economic development are not mutually exclusive but, rather, constitute essential paths to sustainable development. By 2024, the government further clarified its objectives of “strengthening ecological civilization and promoting green, low-carbon development” and issued the Decision of the Central Committee of the Communist Party of China on Further Comprehensive Reform and Advancing Chinese-style Modernization, which aims to establish mechanisms for green and low-carbon development. These policy initiatives indicate China’s active push toward an economic transformation focused on ecological preservation and low-carbon development, progressing toward high-quality green growth. Total factor productivity (TFP) provides a new lens for transitioning economic growth from an extensive to an intensive model. Unlike traditional TFP, green total factor productivity (GTFP) incorporates environmental protection into considerations of economic growth, serving as a more scientifically grounded metric for assessing the quality and level of economic development [1]. Enhancing GTFP should therefore be a key objective to promote high-quality economic development. By adopting a range of measures to improve GTFP, China can not only increase its resource utilization efficiency but also effectively reduce environmental pollution, fostering a harmonious relationship between economic growth and environmental preservation. Consequently, enhancing GTFP has become an inevitable pathway for China in its pursuit of high-quality development.

Historically, China’s development strategy, centered on manufacturing, has resulted in a persistently high share of manufacturing in GDP. However, this approach has also led to challenges such as industrial homogeneity, resource constraints, and environmental degradation, all of which have posed significant threats to GTFP. In recent years, China has increasingly emphasized the coordinated development of the service and manufacturing sectors, particularly the integration of producer services with manufacturing. In 2015, the added value of producer services as a share of GDP in China surpassed that of manufacturing for the first time, and the two sectors remained nearly equal in subsequent years. In 2019, China issued the Implementation Opinions on Promoting the Deep Integration of Advanced Manufacturing and Modern Service Industries, which explicitly called for deep integration of these sectors to support the high-quality development of manufacturing. This shift indicates that China’s economic structure is transitioning from a single-engine focus on manufacturing to a “dual-engine” model driven by both manufacturing and producer services. Industrial co-agglomeration (ICA) provides a practical spatial platform that facilitates the coordination, development, and integration of industries, playing a crucial role in China’s current economic landscape. In Chinese cities, industrial agglomeration rarely manifests as the specialized clustering of a single industry, nor is it typically dominated by unconnected multi-industry clusters. Instead, it often takes the form of coordinated agglomeration among related industries, particularly the coordinated agglomeration of producer services and manufacturing industries (CAPSMI), which has become a key pathway for optimizing urban industrial layouts and upgrading the economic structure under the “dual-engine” strategy. In the context of green development, an important question arises: How does ICA impact China’s GTFP? If an effect exists, what are the mechanisms through which it operates? Are there spatial spillover effects or heterogeneous impacts across different regions? Investigating these questions can provide a theoretical foundation and empirical evidence for enhancing China’s GTFP from the perspective of industrial integration. Additionally, it holds substantial theoretical and practical significance for fostering green growth and high-quality development in the global economy.

The body of literature most relevant to this study can be categorized into three primary areas: First are the measurement and determinants of GTFP. Chung et al. (1997) were pioneers in incorporating pollutant emissions as undesirable outputs in TFP measurement, introducing a new perspective for evaluating TFP [2]. Aparicio et al. (2017) employed an optimized Malmquist–Luenberger (ML) index to calculate GTFP at the national level [3]. Additionally, scholars have conducted GTFP measurements at various levels within China, including the provincial [4], industry [5,6], city [7], and firm levels [8]. In terms of influencing factors, existing studies have investigated a range of individual variables, including green finance [9], technological innovation [10], fiscal decentralization [11], environmental regulation [12], the digital economy [13], climate change [14], human capital structure [15], Internet development [16], foreign direct investment [17], artificial intelligence [18], the establishment of free-trade zones [19], e-commerce city pilots [20], low-carbon city pilot policies [21], innovative city pilot policies [22], smart city pilot policies [23], and carbon emissions trading pilot policies [24].

The second area of focus is the effect of single-industry agglomeration on GTFP. Some studies have found that manufacturing agglomeration [25] and producer service agglomeration [26] can enhance TFP, while others indicate an inverted U-shaped relationship between agricultural agglomeration and agricultural GTFP [27]. Financial agglomeration has been shown to stimulate GTFP growth in specific cities but significantly reduces GTFP in neighboring areas [28]. Different forms of agglomeration externalities exert distinct influences on urban GTFP [29]. The third area pertains to the effect of ICA on GTFP. ICA can impact GTFP through mechanisms such as technological innovation and technological progress [1,30,31]. Factors such as innovation agglomeration, institutional proximity, and policy advantages can moderate the relationship between ICA and GTFP [1,32]. ICA has been found to exert positive spatial spillover effects on the GTFP of adjacent areas [33]. Additionally, ICA’s spatial spillover effects have heterogeneous impacts on total factor carbon emission efficiency [34], with the coordinated agglomeration of finance and manufacturing significantly enhancing green economic efficiency in both local and neighboring regions [30].

This study offers several key marginal contributions to the existing literature. First, it deeply elucidates the transmission mechanisms through which ICA affects GTFP from the perspective of resource allocation, focusing particularly on the optimization pathways of capital, energy, and labor resources. This approach enriches the mechanistic understanding of ICA’s impact on GTFP. Second, by employing a panel smooth transition regression (PSTR) model, this study examines how varying intensities of local government competition and environmental regulation modulate the effect of ICA on GTFP, thereby advancing the comprehension of these relationships. Third, this study systematically investigates the spatial spillover effects of ICA and conducts a nuanced analysis based on dimensions such as industry type, city size, and urban cluster characteristics. These heterogeneity analyses provide empirical support for industrial policies tailored to cities and regions of various sizes, aiding policymakers in effectively guiding coordinated agglomeration of producer services and manufacturing industries (CAPSMI) to further green and balanced regional development. Additionally, this research delves into the spatial decay boundary of ICA’s influence on GTFP.

The purpose of this study is to systematically explore the impact of ICA on China’s GTFP within the framework of green development by constructing an analytical model encompassing transmission mechanisms, regulatory mechanisms, and spatial spillover effects. By analyzing ICA’s systemic impact on GTFP through the lenses of resource allocation, local government competition, environmental regulation, and spatial spillover effects, this research addresses existing gaps in the literature on the relationship between ICA and GTFP, offering new theoretical insights for future investigations. Moreover, through empirical analysis, this study provides practical policy recommendations to assist the Chinese government in optimizing industrial structure, fostering the integration of producer services with manufacturing, and ultimately promoting green economic transformation and high-quality growth.

2. Theoretical Background and Research Hypotheses

Marshall’s externality theory suggests that the coordinated agglomeration of producer services and manufacturing industries (referred to here as the “two industries”) enables these sectors to share a specialized labor market, thereby improving the match between labor supply and demand. When a particular industry within a region experiences decline, displaced workers often opt to transfer to other local industries rather than relocating to engage in similar work elsewhere [35]. This intra-regional, inter-industry labor mobility is also evident in areas with CAPSMI. For instance, significant labor mobility has been observed between Nokia’s industry and sectors such as R&D, education, and business services [36]. This mobility is even more pronounced in CAPSMI regions with strong input–output linkages, thereby enhancing the labor allocation efficiency within these areas. As ICA intensifies, workers become more willing to accept lower average wages, as these regions or industries offer suitable employment opportunities not only for the workers themselves but also for their family members, thereby increasing their total household income. Additionally, ICA positively impacts regional (urban) wage levels through backward linkages (market access) and forward linkages (supply access). Higher wages, in turn, enhance workers’ motivation, increase their willingness to work, and improve the efficiency of labor resource allocation. Workers with diverse knowledge and skills can typically move freely across regions and between industries, whereas the mobility of single-skilled labor tends to be more restricted. Thus, as the levels of knowledge and skill diversity among workers increase, so too does the speed of labor mobility across regions and industries, which contributes to greater efficiency in labor allocation. Given the strong input–output relationships between producer services and manufacturing, ICA facilitates labor exchange and learning, thereby enhancing the diversity of skills and knowledge among workers in these “two industries”, ultimately leading to a more efficient allocation of labor resources.

According to rent-seeking theory, capital can achieve an optimal spatial configuration across regions through the specialized division of labor. High-value-added and highly competitive producer services tend to cluster in the core areas of cities, while lower-value-added, less competitive manufacturing industries are more likely to agglomerate in peripheral areas surrounding the urban core. Capital agglomeration also encourages firms within the industrial chain to adopt more advanced technology and equipment, thereby improving capital productivity and technical efficiency. Moreover, industrial agglomeration generates financial externalities, facilitating the flow of credit resources to private enterprises and “relationship-intensive” firms that rely heavily on inter-firm networks, thereby enhancing the efficiency of credit resource allocation [37]. With ICA, which entails both “input–output” relationships and close geographic proximity, companies gain greater insight into one another’s commercial credibility and operational status, thereby reducing information asymmetry and optimizing the allocation of financial capital. ICA reduces the costs for financial intermediaries in acquiring investment information, increases the efficiency of reviewing and screening investment projects, and facilitates the redirection of financial capital from inefficient to high-efficiency projects. This process supports the rational allocation of capital within the manufacturing sector [38], aiding in the green transformation and upgrading of manufacturing industries.

According to agglomeration economy theory, ICA aids in the centralized management of energy consumption among manufacturing firms within agglomeration zones, optimizing energy allocation across firms and facilitating energy recycling between them. Through specialization effects, ICA encourages manufacturing firms to shift their production processes toward low-pollution, high-value-added activities, thereby directing energy resources towards environmentally friendly manufacturing enterprises. Additionally, as per the theory of specialized division of labor, ICA motivates manufacturing firms to outsource high-energy-consuming and polluting operations to cleaner, more efficient producer service firms, thereby achieving a rational allocation of energy resources across industries. The competitive dynamics of ICA also eliminate high-energy-consuming, high-pollution, and low-efficiency manufacturing firms, further enhancing energy allocation efficiency within the agglomeration zone. Moreover, ICA encourages manufacturing firms to utilize advanced resources from producer services—such as information technology, R&D design, modern logistics, and energy-saving technologies—replacing high-pollution, non-renewable fossil fuels such as coal and oil with renewable energy sources, which ultimately improves the overall efficiency of energy resource allocation.

From the perspective of resource allocation theory, effective allocation of resources enhances traditional TFP, while misallocation worsens environmental pollution. Spatial misallocation of resources not only directly impedes regional green development but also restricts the optimization of internal resources necessary for achieving green growth within regions. For energy resources, a well-structured energy pricing system is crucial for promoting industrial upgrading, optimizing the energy mix, and disseminating energy-saving technologies. However, in China, prices for resources such as electricity and natural gas remain government-regulated, which may lead to energy misallocation, often causing resources to flow into high-energy-consuming industries [39], ultimately hampering efforts for energy conservation and emissions reduction. Distortions in the capital market also pose significant challenges to environmental improvement in China. Ideally, rational financial capital allocation would direct more resources toward high-efficiency, low-emission industries, supporting real enterprises in implementing material recycling, thereby optimizing the industrial structure and improving energy efficiency. However, the allocation of financial capital is often constrained by factors such as corporate creditworthiness, government intervention, and administrative fragmentation, which tend to channel resources disproportionately toward large cities and major corporations. This results in financing difficulties for smaller cities and private small enterprises. In sectors such as steel, coal, cement, and chemicals, excessive financial capital allocation has contributed to overcapacity, reducing capital returns and allocation efficiency. Consequently, misallocation of financial capital leads to a flow of funds into low-efficiency sectors, overcapacity industries, and heavily polluting fields, or results in credit discrimination against small and medium-sized enterprises, thereby obstructing improvements in corporate and regional GTFP.

Hypothesis 1: 

ICA can enhance GTFP by optimizing the allocation of labor, capital, and energy resources.

China’s decentralization model provides local governments with two key conditions for competition: administrative authority and fiscal autonomy, enabling local governments to intervene in industrial agglomeration. Consequently, local government competition has a dual impact on the effects of industrial collaborative agglomeration. On the one hand, to enhance economic performance, innovation, and social welfare, local governments focus on infrastructure development and public service improvements, which reduce the friction costs of factor mobility as well as the costs associated with ICA. Policies such as financial subsidies and tax incentives to support producer services, along with the construction of essential infrastructure, can help to realize the positive externalities of CAPSMI. On the other hand, although environmental quality and innovation have become increasingly important in local government evaluations, these metrics are difficult to quantify relative to economic performance. The central government still places a strong emphasis on economic outcomes, which may lead to excessive competition among local governments, driving them to attract capital-intensive firms to local industrial parks in order to boost their short-term fiscal revenues, even if this reduces the labor demand. Additionally, local governments tend to protect high-tax-paying and state-owned enterprises, resulting in a concentration of state-owned firms, limited innovation capabilities, and weakened industrial chains, which constrain the positive impacts of industrial collaborative agglomeration. Moreover, local protectionism and market fragmentation exacerbate regional resource allocation distortions, hinder industrial agglomeration and diversification, and undermine the full potential of ICA’s positive effects. Competing for political achievements and promotions, local governments often prioritize the construction of development zones over building industry-supportive facilities and innovation networks tailored to local needs, further impeding the positive effects of collaborative agglomeration [40].

Environmental regulation is a crucial tool for balancing economic growth with environmental sustainability. Increased regulatory intensity may lead to the clustering of polluting firms, creating a “pollution haven” effect to share pollution control costs, but it can also produce congestion effects that degrade environmental quality. At the same time, according to new trade theory, manufacturing firms often increase their use of producer services to improve product quality and technological sophistication. Stricter environmental regulations encourage firms in agglomeration areas to reorganize resource allocation, share pollution control facilities, and adopt green emission-reduction technologies, thereby promoting collaborative innovation and enhancing ICA’s positive impact on GTFP. Thus, raising environmental standards can encourage manufacturing firms to share infrastructure for energy conservation and emissions reduction, as well as to strengthen their cooperation with producer services in environmental technologies.

Hypothesis 2: 

Local government competition can weaken the positive effect of ICA on GTFP, while environmental regulation can enhance the positive effect of ICA on GTFP.

The “two industries” within a region can exert both positive and negative effects on the GTFP of firms and the region as a whole, directly influencing local GTFP. For instance, in the context of the green transformation and upgrading of manufacturing, if high-energy-consuming, high-pollution manufacturing firms in the agglomeration zone fail to improve their factor input structure or enhance their energy efficiency through technological innovation or specialized technical services, they are likely to be gradually phased out. In contrast, efficient, low-pollution firms are more likely to remain within the agglomeration zone due to competitive pressures, thereby enhancing the region’s overall GTFP.

According to the diffusion and siphoning effect theories, ICA can also generate spatial spillover effects on the GTFP of nearby firms and regions. On the one hand, unlike traditional industries, producer services are less constrained by factors such as transportation costs due to their intangible nature, inability to be stored, and synchronous production and consumption [41]. With advances in new technology (particularly in artificial intelligence, the Internet of Things, and big data), producer services can gather and relay relevant information more quickly and at a lower cost, thereby supporting manufacturing firms in surrounding areas—an effect that is particularly evident between central and peripheral cities. On the other hand, during the early stages of industrial agglomeration, market effects, pricing dynamics, and cumulative cycles often cause the coordinated agglomeration of the two industries to attract labor, capital, technology, and other resources from surrounding areas, resulting in a “siphoning effect” that may hinder GTFP growth in neighboring regions. However, as local industrial agglomeration matures, congestion effects, rent-seeking effects, and bidding pressures tend to drive low-value-added, inefficient firms to relocate to surrounding areas, creating a diffusion effect.

In the context of China’s industrial development, local protectionism increases the communication and transaction costs between regions, obstructing interregional division of labor and collaboration, restricting the cross-regional flow of industrial resources, and thereby impeding the integrated development of industries such as producer services and manufacturing between regions and their neighbors. Furthermore, the experiences of regional industrial coordination may be transferred to neighboring areas through a “demonstration effect”. However, due to differing industrial foundations across regions, policy mismatches with local economic conditions often occur, which lead to low-level redundancy in the development of producer services, resulting in misalignments in the growth of the two industries. Currently, the structure of producer services tends to be homogenized across cities within the same province, significantly impeding the coordinated agglomeration of the secondary and tertiary sectors. This convergence weakens the diffusion effect of producer service agglomeration, limiting the potential for efficiency improvements in manufacturing in surrounding regions and thereby constraining GTFP enhancement for neighboring firms and regions.

Hypothesis 3: 

ICA may have a spatial spillover effect on GTFP.

3. Research Design

3.1. Transmission Mechanism Model

This study employs the System Generalized Method of Moments (SGMM) to analyze the transmission mechanisms through which ICA affects GTFP. The SGMM is advantageous because it controls for unobserved individual and time effects while also effectively addressing endogeneity issues within the model. The SGMM-based panel mediation effect model is specified as follows:

$${LNGTFP}_{it}={\beta }_0+{\sum }_{\tau =1}^{\omega }{{\rho }_{\tau }LNGTFP}_{it-\tau }+{\beta }_1{LNCOAGG}_{it}+{\mathsf{\partial }}_i{X}_{it}+a_i+{\mu }_t+{\epsilon }_{it}$$

(1)

$$ME{D}_{it}={\alpha }_0+{\sum }_{\tau =1}^{\phi }{{\delta }_{\tau }LNLMISS}_{it-\tau }{+\alpha }_1{LNCOAGG}_{it}+{\mathsf{\partial }}_i{X}_{it}+a_i+{\mu }_t+{\epsilon }_{it}$$

(2)

$${LNGTFP}_{it}={\gamma }_0+{\sum }_{\tau =1}^{\omega }{{\rho }_{\tau }LNGTFP}_{it-\tau }+{\gamma }_{1}ME{D}_{it}+{\gamma }_2{LNCOAGG}_{it}+{\mathsf{\partial }}_i{X}_{it}+a_i+{\mu }_t+{\epsilon }_{it}$$

(3)

where LNGTFP is the logarithm of GTFP, LNCOAGG is the logarithm of ICA, and X represents a series of control variables that influence LNGTFPL. $a_i$ denotes the individual fixed effects, ${\mu }_t$ denotes the time fixed effects, and ${\epsilon }_{it}$ represents the random error term. MED is the mediating variable, representing the variables in the transmission mechanism: the labor factor mismatch (LFM) index, capital factor misallocation (CFM) index, and energy factor mismatch (NFM) index.

3.2. Regulatory Mechanism Model

This study employs the PSTR model to examine the regulatory mechanism of ICA’s impact on GTFP. The existing literature has extended the traditional panel threshold model, causing the regression coefficients to transition smoothly between regimes rather than changing abruptly. The basic PSTR model’s expression containing only two mechanisms is as follows:

$$y_{it}={\mu }_i+{\beta }_0^{\prime }{x}_{it}+{\beta }_1^{\prime }{x}_{it}g\left(q_{it};\gamma ,c\right)+{\epsilon }_{it}$$

(4)

where $i=\mathrm{1,2},\dots ,N;t=\mathrm{1,2},\dots ,T$; N represents the number of cross-sections, and T represents the length of time. $y_{it}$ is the dependent variable, $x_{it}$ denotes a k-dimensional time-varying vector, ${\epsilon }_{it}$ is the error term, and $g\left(q_{it};\gamma ,c\right)$ is a continuous function of the transition variable $q_{it}$, which varies continuously between 0 and 1. The regression coefficients vary continuously between ${\beta }_0$ and ${\beta }_0+{\beta }_1$; $\gamma$ is the smoothness parameter of the transition function, determining the speed of transition between different regimes; and c is the location parameter where the transition occurs. The expression for a multi-regime PSTR model is as follows:

$$y_{it}={\mu }_i+{\beta }_0^{\prime }{x}_{it}+{\sum }_{j=1}^r{\beta }_j^{\prime }{x}_{it}{g}_j\left(q_{it}^j;\gamma ,c\right)+{\epsilon }_{it}$$

(5)

where $j=1,2,\dots ,\mathrm{L}$, $r$ represents the number of transition functions, and the meanings of the other parameters are the same as in Equation (4). $y_{it}$ represents GTFP, and $x_{it}$ represents LNCOAGG and the control variables. The transition variable $q_{it}$ includes variables such as local government competition and environmental regulation.

3.3. Spatial Durbin Model

Drawing on the existing literature [42], the spatial Durbin model (SDM) is used to examine the spatial spillover effect of ICA on GTFP. The model is constructed as follows:

$${LNGTFP}_{it}=\alpha +\rho {\sum }_{j=1}^N W_{ij}{LNGTFP}_{jt}+\beta {\mathrm{I}\mathrm{C}\mathrm{A}}_{it}+\phi X+\delta {\sum }_{j=1}^N W_{ij}{\mathrm{I}\mathrm{C}\mathrm{A}}_{it}+\gamma {\sum }_{j=1}^N W_{ij}{X}_{ijt}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(6)

where $\alpha$ is the constant term, $\beta \mathrm{a}\mathrm{n}\mathrm{d} \phi$ are the coefficients representing the effects of ICA on local and surrounding GTFP, respectively, ${\mu }_i$ represents individual fixed effects, $v_t$ represents time fixed effects, and ${\epsilon }_{it}$ denotes the random disturbance term. W_ij is the spatial weight matrix. The meanings of the other parameters are the same as in Equation (1).

3.4. Variable Definitions

(1) Dependent variable: The calculation of GTFP involves the input, desired output, and undesired output indicators. The specific indicators selected are as follows: ① Input indicators: Labor input, measured by the total urban employment, which includes employees in urban units, as well as private and self-employed workers. Capital input: Capital stock is estimated using the perpetual inventory method, which accumulates past investments while accounting for depreciation. Energy consumption: At the provincial level, energy consumption is often standardized to coal equivalents for various energy types (e.g., coal, oil, natural gas). However, city-level data for these energy types are often unavailable. Thus, this study uses urban electricity consumption as a proxy for measuring urban energy consumption. ② Output indicators: Desired output, measured by deflating city-level GDP using the provincial GDP deflator, with 2003 as the base year to adjust for inflation and provide a real measure of economic output. Undesired output indicators: These represent negative externalities of production and include urban industrial wastewater discharge, industrial sulfur dioxide emissions, and industrial smoke and dust emissions, which serve as proxies for environmental pollution at the city level.

This study utilizes the global Malmquist–Luenberger (GML) index, which is based on the super-efficiency EBM (epsilon-based measure) model, to measure GTFP. The GML index accounts for both desirable and undesirable outputs, providing a comprehensive measure of productivity that integrates environmental considerations into the assessment of economic performance.

(2) Core explanatory variable: Referring to the existing literature [32], the calculation formula for ICA is as follows:

$${\mathrm{C}\mathrm{O}\mathrm{A}\mathrm{G}\mathrm{G}}_{ijk}=\left(1-{\displaystyle \frac{\left|{\phi }_{ik}-{\phi }_{jk}\right|}{\left|{\phi }_{ik}+{\phi }_{jk}\right|}}\right)+\left({\phi }_{ik}+{\phi }_{jk}\right)$$

(7)

where ${\phi }_{ik}\mathrm{a}\mathrm{n}\mathrm{d}{\phi }_{jk}$ represent the agglomeration levels of producer services i and manufacturing j in city k, respectively, measured by the location quotient. The larger the COAGG index, the higher the level of ICA.

(3) Conduction mechanism variables: Referring to the method for calculating the labor misallocation index from the existing literature [43], this study employs a C-D production function incorporating energy consumption to compute the LFM index, CFM index, and NFM index for cities. The specific formulae for calculating the factor misallocation indices are as follows:

$${\mathrm{L}\mathrm{F}\mathrm{M}}_i={\displaystyle \frac{1}{{\gamma }_{L_i}}}-1$$

(8)

$${\mathrm{C}\mathrm{F}\mathrm{M}}_{it}={\displaystyle \frac{1}{{\gamma }_{K_i}}}-1$$

(9)

$${\mathrm{N}\mathrm{F}\mathrm{M}}_{it}={\displaystyle \frac{1}{{\gamma }_{N_i}}}-1$$

(10)

where ${\gamma }_{L_i},$ ${\gamma }_{K_i}, \mathrm{a}\mathrm{n}\mathrm{d}{\gamma }_{N_i}$ represent the absolute distortion coefficients of the prices for capital, labor, and energy factors, respectively, indicating the markups of these factors in the absence of relative distortions.

(4) Regulatory mechanism variables: This study measures local government competition aimed at attracting foreign investment (COMPETE) using the per capita actual utilization of foreign direct investment; it uses three indicators—the sulfur dioxide removal rate, industrial smoke (dust) removal rate, and comprehensive utilization rate of industrial solid waste—and applies the entropy method to calculate a comprehensive index for environmental regulation (ER).

(5) Control variables: Referring to the existing literature [32], the following control variables that could influence GTFP were chosen: This study measures human capital (HUMAN) using the ratio of students enrolled in general higher education institutions to the labor force in a city and transportation infrastructure (ROAD) using the per capita road area of a city. Three indicators—the sulfur dioxide removal rate, industrial smoke (dust) removal rate, and comprehensive utilization rate of industrial solid waste—are used, and the entropy method is applied to calculate a comprehensive index for environmental regulation (ER).

Urbanization (URB) is assessed by the ratio of the urban population to the year-end permanent resident population. Resource endowment (MINING) is indicated by the proportion of employment in the mining industry relative to the total population at the end of the year. Openness (OPEN) is represented by the ratio of a city’s annual actual foreign investment (converted to RMB using the year’s average exchange rate) to its GDP. Fiscal input in science and technology (GOVKJ) is measured by the proportion of fiscal expenditure on science and technology to the total fiscal expenditure.

3.5. Sample and Data

In the sample selection and data processing phases of this study, we carefully ensured data completeness and accuracy in order to uphold the reliability and scientific rigor of the analysis. Cities with significant data deficiencies, such as Sansha in Hainan Province, Lhasa in the Tibet Autonomous Region, Haidong in Qinghai Province, and Bijie in Guizhou Province, were excluded from the sample due to insufficient statistical data. Additionally, cities established relatively recently or with incomplete data records, as well as cities undergoing administrative adjustments during the study period—such as Danzhou in Hainan Province and Chaohu in Anhui Province—were also excluded. After screening, a total of 283 cities were retained for analysis. For cities with minor data gaps, we employed linear interpolation and moving-average techniques to fill in missing values, thereby maintaining data completeness.

The variables used in this study were primarily derived from the China City Statistical Yearbook, China Urban Construction Statistical Yearbook, China Energy Statistical Yearbook, China Regional Economic Statistical Yearbook, and China Industrial Statistical Yearbook, along with the China Patent Database from the National Intellectual Property Administration, the CNRDS database, statistical yearbooks of certain prefecture-level cities, and the EPS data platform. Through meticulous sample selection and data processing, we constructed a panel dataset covering 283 Chinese cities from 2006 to 2020, providing a robust data foundation for this research.

4. Empirical Analysis Results

4.1. Examination of the Impact Mechanism

4.1.1. Examination of the Conduction Mechanism

The transmission mechanism test results presented in

Table 1. Results of the conduction mechanism examination.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
LNCOAGG	0.2794 ***	0.0477	0.2122 ***	−0.1087 ***	0.2404 ***	−0.2110 ***	0.2102 ***
	(69.207)	(0.900)	(13.475)	(−3.373)	(26.213)	(−3.727)	(16.614)
LFM			−0.0068
			(−1.589)
CFM					−0.0075 ***
					(−3.268)
NFM							−0.0132 ***
							(−4.407)
Controls	YES	YES	YES	YES	YES	YES	YES
Constant	−0.3039 ***	0.9161 ***	−0.3259 ***	−0.2845 *	−0.2060 ***	0.3659 ***	−0.1637 ***
	(−43.377)	(4.724)	(−6.255)	(−1.940)	(−8.111)	(3.159)	(−7.891)
N	3396	3396	3396	3396	3396	3396	3396
P-AR(1)	0.0000	0.0000	0.0000	0.0000	0.0000	0.0002	0.0000
P-AR(2)	0.1003	0.4009	0.2242	0.6857	0.3148	0.2894	0.3082
P-SAR	0.8303	0.1242	0.1518	0.1090	0.3190	0.3914	0.1654

Note: *** and * denote significance levels of 1% and 10%, respectively. The values in parentheses are the z-statistics of the parameter estimates, and GTFP is lagged by three periods.

indicate that, in all P-AR tests, the P-AR(1) test rejected the null hypothesis of no first-order serial correlation, while the P-AR(2) test confirmed the absence of second-order serial correlation in the regression equation. Furthermore, the results from the P-SAR over-identification test indicate that the validity of the instrumental variables cannot be rejected. Thus, the overall model specification is appropriate, and the instrumental variables are valid.

This study employed Stata 17 to conduct baseline model estimation. The results for the baseline SGMM model (1) in Table 1 show that ICA significantly enhances GTFP, further supporting the conclusions drawn in previous studies. In model (2), the impact coefficient of ICA on LFM is 0.0477 and is not statistically significant, while in model (3), the coefficient of LFM on GTFP is −0.0068 and is also not significant. These results indicate that ICA does not enhance GTFP through improvements in LFM. A possible explanation for this is that ICA generates congestion and lock-in effects, leading to an excessive concentration of labor in specific firms, industries, or regions, which weakens labor allocation efficiency. This finding suggests that improving GTFP through adjustments in LFM is not a viable pathway within the study’s sample.

In model (4), the coefficient of ICA on CFM is −0.1087, which is significant at the 1% level, indicating that ICA significantly improves CFM and thereby enhances capital allocation efficiency. In model (5), the coefficient of CFM on GTFP is −0.0075, also significant at the 1% level, indicating that CFM significantly hinders GTFP. Nevertheless, ICA still has a significant positive effect on GTFP. These results suggest that ICA promotes GTFP by enhancing capital allocation efficiency through improvements in CFM.

In model (6), the coefficient of ICA on NFM is −0.2110, which is significant at the 1% level, suggesting that ICA effectively improves NFM and thereby enhances energy allocation. In model (7), the coefficient of NFM on GTFP is −0.0132, also significant at the 1% level, indicating that NFM significantly hinders GTFP. However, ICA continues to exert a significant positive effect on GTFP. These findings suggest that ICA promotes GTFP by enhancing energy allocation efficiency through improvements in NFM.

4.1.2. Examination of the Regulatory Mechanism

(1)

Local Government Competition

This study utilized MATLAB R2018b to examine the moderation mechanism.

Table 2. Linearity test and residual nonlinearity test of the model with local government competition as the transition variable.

Statistics	Linearity Test: H0: r = 0; H1: r = 1	Residual Nonlinearity Test: H0: r = 1; H1: r = 2
LM	25.381 ***	13.084 *
	(0.001)	(0.059)
LMF	3.264 ***	1.809 *
	(0.002)	(0.095)
LRT	25.581 ***	13.281 *
	(0.001)	(0.059)

Note: The null hypothesis H0: r = 0 indicates the absence of nonlinear effects. The LM statistic follows an asymptotic χ² distribution, and LMF follows an F(mK, TN-N-m(K + 1)) distribution, where r is the number of transition functions; *** and * denote significance levels of 1% and 10%, respectively. The values in parentheses are the p-values of the parameter estimates.

displays the results of the linearity test and residual nonlinearity test for the model using local government competition as the transition variable. According to the linearity test, the LM, LMF, and LRT statistics all reject the null hypothesis of linearity at the 1% significance level, indicating that the model exhibits significant nonlinear characteristics when local government competition is used as the transition variable. This finding necessitates a further residual nonlinearity test.

In the residual nonlinearity test, the LMF, LM, and LRT statistics are significant at the 10% level, suggesting that the null hypothesis of r = 1 should theoretically be rejected, requiring an assessment of the r = 2 hypothesis. However, the software only provided model estimation results for r = 1, indicating that r = 1 is the optimal choice, while r = 2 does not show clear evidence of superiority. Consequently, the model with local government competition as the transition variable contains one optimal transition function, with a single transition point.

The model estimates obtained using local government competition as the transition variable are presented in

Table 3. Regression results of the model with local government competition as the transition variable.

	Parameter Estimates	Variable Coefficient	Variable Coefficient
LNCOAGG	Linear Part	${\beta }_0^{\prime }$	0.2381 ***
			(10.249)
	Nonlinear Part	${\beta }_1^{\prime }$	−0.0172
			(−0.3491)
Controls	YES	YES	YES
	Transition Location	c1	14.3825
	Transition Speed	γ1	4329

Note: *** denotes significance levels of 1%. The values in parentheses are the t-statistics of the parameter estimates.

. These results reveal that the impact of ICA on GTFP exhibits a single-threshold characteristic with respect to local government competition. When local government competition is below the threshold value of 14.3825, ICA has a positive and statistically significant effect on GTFP at the 1% level.

However, once local government competition reaches the threshold of 14.3825, the positive effect of ICA on GTFP begins to diminish, with the impact coefficient decreasing to 0.2295 (0.2381 − 0.0172 × 0.5). This positive influence persists until the competition level fully surpasses the threshold, ultimately stabilizing at 0.2209 (0.2381 − 0.0172); however, in this higher regime, the effect is no longer statistically significant. These findings indicate that, as local government competition intensifies, the impact of ICA on GTFP stabilizes across the two regimes, with coefficients of 0.2381 and 0.2209, respectively. This suggests that local government competition acts as a moderating mechanism on the effect of ICA on GTFP and that excessive local government competition may hinder ICA’s positive impact on GTFP.

One possible explanation for this is that local government competition can encourage local governments to improve their public service provision, thereby enhancing the resource allocation efficiency of ICA. However, when local government competition exceeds a certain threshold, it may lead to excessive governmental intervention in industrial development. This can result in low-level clustering of firms, resource waste, industrial homogeneity, and the formation of “pseudo-industrial agglomeration”. This observation also indirectly suggests that local governments have yet to establish a competitive framework focused on “green economic growth”.

(2)

Environmental Regulation

Table 4. Linearity test and residual nonlinearity test of the model with environmental regulation as the transition variable.

Statistics	Linearity Test: H0: r = 0; H1: r = 1	Residual Nonlinearity Test: H0: r = 1; H1: r = 2
LM	25.041 ***	16.229 **
	(0.001)	(0.035)
LMF	3.598 ***	2.209 **
	(0.001)	(0.029)
LRT	25.125 ***	16.349 **
	(0.001)	(0.035)

Note: *** and ** denote significance levels of 1% and 5%, respectively. The values in parentheses are the p-values of the parameter estimates. Others are similar to those in Table 2.

shows the linearity and residual nonlinearity tests for the model using environmental regulation as the transition variable. The linearity test results indicate that the LM, LMF, and LRT statistics all reject the null hypothesis of linearity at the 1% significance level, suggesting that the model exhibits significant nonlinear characteristics when environmental regulation is used as the transition variable. This finding necessitates a further residual nonlinearity test.

In the residual nonlinearity test, the LMF, LM, and LRT statistics are significant at the 5% level, indicating that the null hypothesis of r = 1 should be rejected and that the r = 2 hypothesis should be tested. However, the software only provided model estimation results for r = 1, suggesting that r = 1 is the optimal choice, while evidence for the r = 2 case is inconclusive. Thus, in the model with environmental regulation as the transition variable, there exists one optimal transition function, with a single transition point.

Table 5. Regression results of the model with environmental regulation as the transition variable.

	Parameter Estimates	Variable Coefficient	Variable Coefficient
LNCOAGG	Linear Part	${\beta }_0^{\prime }$	0.1259 ***
			(3.089)
	Nonlinear Part	${\beta }_1^{\prime }$	0.1127 **
			(2.086)
Controls	YES	YES	YES
	Transition Location	c1	0.4381
	Transition Speed	γ1	12.5919

Note: *** and ** denote significance levels of 1% and 5%, respectively. The values in parentheses are the t-statistics of the parameter estimates.

presents the model estimates with environmental regulation as the transition variable, revealing a single-threshold effect of environmental regulation intensity on the impact of ICA on GTFP. When the environmental regulation intensity is below the threshold value of 0.4381, ICA has a positive and statistically significant impact on GTFP at the 1% level. Once the environmental regulation intensity reaches the threshold of 0.4381, the positive impact of ICA on GTFP gradually intensifies, with the impact coefficient rising to 0.1822 (0.1259 + 0.1127 × 0.5). This positive influence continues beyond the threshold and ultimately stabilizes at 0.2386 (0.1259 + 0.1127), which is significant at the 5% level.

These findings suggest that, as the environmental regulation intensity increases, the impact of ICA on GTFP stabilizes across two regimes, with coefficients of 0.1259 and 0.2386, respectively. This implies that environmental regulation serves as a moderating mechanism on the relationship between ICA and GTFP. As the environmental regulation intensity rises, ICA’s effect on GTFP remains positive and gradually strengthens, indicating that stricter environmental regulation enhances ICA’s positive contribution to GTFP.

A likely explanation for this is that stringent and effective environmental regulation incentivizes firms to innovate, compelling manufacturing and producer service firms within ICA regions to focus more on the division of labor and collaboration. This fosters the diffusion of green emission-reduction technologies among firms, thereby reducing production costs, increasing net profits, and amplifying ICA’s positive impact on GTFP.

4.2. Overall Test of Spatial Effects

Before estimating the spatial econometric model, it is crucial to test and determine which model is most appropriate, such as the spatial lag model (SLM), spatial error model (SEM), or spatial Durbin model (SDM). Both the Wald test and LR test reject the null hypotheses $H_0:\delta =0$ and $H_0:\delta +\mathsf{\rho }\mathsf{\beta }=0$, indicating that the SDM cannot be reduced to either the SLM or SEM. When selecting a spatiotemporal double fixed-effects model, the Hausman test rejects the null hypothesis of the random-effects model at a 1% significance level. Thus, this study employed the spatiotemporal double fixed-effects SDM. To enhance the reliability of the results, an economic distance spatial weight matrix was used.

According to the results presented in

Table 6. Regression results of the SDM.

	SDM		SLM	SEM
	Main	Wx	Main	Main
LNCOAGG	0.2327 ***	−0.2087	0.2319 ***	0.2350 ***
	(12.653)	(−1.019)	(12.609)	(12.722)
Controls	YES (1.215)		YES	YES
$\rho /\gamma$	0.8365 ***		0.8825 ***	0.8846 ***
	(20.675)		(29.694)	(30.151)
Sigma2_e	0.0383 ***		0.0391 ***	0.0391 ***
	(44.350)		(44.323)	(44.341)
N	4245		4245	4245
R2	0.0059		0.0440	0.0620
Log-Likelihood	816.6623		772.7831	770.8239

Note: *** denotes significance levels of 1%. The values in parentheses are the z-statistics of the parameter estimates, and the values in square brackets are the p-values. City fixed effects and year fixed effects are controlled.

, the spatial lag coefficients for urban GTFP are significantly positive, demonstrating a notable positive spatial correlation in urban GTFP. This indicates that the GTFP of a city can positively influence the GTFP of neighboring cities. Additionally, the findings show that ICA significantly boosts the GTFP of the local city but does not have a similar effect on the GTFP of adjacent cities.

Building on the SDM estimation, this study further analyzed the direct, indirect, and total effects of ICA on urban GTFP. As illustrated in

Table 7. Decomposition of spatial effects.

	Direct Effect	Indirect Effect	Total Effect
LNCOAGG	0.2335 ***	0.0262	0.2597
	(11.696)	(0.018)	(0.182)
Controls	YES	YES	YES

Note: *** denotes significance levels of 1%. The values in parentheses are the z-statistics of the parameter estimates. City fixed effects and year fixed effects are controlled.

, the direct effect of ICA on GTFP is significantly positive, while the indirect effect is not significant. This suggests that ICA enhances GTFP only in the local city and does not facilitate improvements in the GTFP of neighboring cities through spatial spillover effects. This lack of spillover may be attributed to administrative fragmentation among cities, which limits the flow of production factors. Additionally, the “demonstration effect”, where neighboring cities blindly imitate policies or actions, also plays a key role in the insignificance of spatial spillover.

4.3. Heterogeneity Test of Spatial Effects

4.3.1. Industrial Heterogeneity of Cities

This study, based on the definition of producer services, examined the heterogeneous effects of the coordinated agglomeration of high-end producer services (GD) and low-end producer services (DD) with manufacturing on urban GTFP (see

Table 8. Regression results of the model for industrial heterogeneity of cities.

	Direct Effect	Indirect Effect	Total Effect
LNGDZZ	0.1461 ***	0.4751	0.6256 *
	(15.791)	(1.354)	(1.714)
Controls	YES	YES	YES
LNDDZZ	0.0630 ***	0.4998 *	0.5145 *
	(6.135)	(1.680)	(1.678)
Controls	YES	YES	YES

Note: *** and * denotes significance levels of 1% and 10%, respectively. The values in parentheses are the z-statistics of the parameter estimates. City fixed effects and year fixed effects are controlled.

). The regression results show that the impact coefficients of the coordinated agglomeration of high-end producer services and manufacturing and low-end producer services and manufacturing on local GTFP are 0.1461 and 0.0630, respectively, with total effect coefficients of 0.6256 and 0.5145, respectively. These findings indicate that the coordinated agglomeration of high-end producer services with manufacturing has a stronger positive effect on urban GTFP than that of low-end producer services with manufacturing.

The green economic transformation of industries is highly dependent on the coordinated development of high-end producer services and manufacturing. Through agglomeration with manufacturing, high-end producer services embed new knowledge and technology into the manufacturing process, facilitating technological advancement and internal structural transformation within manufacturing, which drive green manufacturing. In turn, the green transformation of manufacturing stimulates high-end producer services to continuously improve their service quality in response to market demand, thereby enhancing urban green economic development through a cumulative cycle effect.

However, while the coordinated agglomeration of high-end producer services and manufacturing does not boost GTFP in surrounding cities, the coordinated agglomeration of low-end producer services and manufacturing does contribute positively to GTFP in neighboring areas. This finding contrasts with the typical characteristics of high-end producer services—such as high added value of products, larger service radius, and lower transaction frequency—and low-end producer services, which tend to be less knowledge-intensive, with a smaller service radius and higher transaction frequency. Possible explanations include, firstly, that China’s high-end producer service sector is still underdeveloped, with economies of scale constrained by external environmental factors, limiting its capacity to effectively meet the demand of midstream industries. Secondly, China’s manufacturing sector is still largely low-end and specialized, aligning more closely with the current development stage of low-end producer services.

Additionally, limited urban resources and internal competitive pressures tend to filter out low-value-added, low-efficiency industries, which gradually migrate from city centers to nearby second- and third-tier cities. This relocation results in positive spatial spillover effects from the coordinated agglomeration of low-end producer services and manufacturing, thereby enhancing GTFP in surrounding areas.

4.3.2. Heterogeneity of City Size

This study examined the heterogeneous effects of ICA on GTFP across megacities, large cities, medium-sized cities, and small cities (see

Table 9. Regression results of the model for heterogeneity of city size.

	Direct Effect	Indirect Effect	Total Effect
	Megacities
LNCOAGG	0.3005 ***	0.6684 **	0.9024 ***
	(3.582)	(2.264)	(2.940)
Controls	YES	YES	YES
	Large Cities
LNCOAGG	0.2054 ***	−1.0147 **	−0.7919 **
	(8.057)	(−2.597)	(−2.124)
Controls	YES	YES	YES
	Medium-Sized Cities
LNCOAGG	0.2365 ***	−0.8235	−0.5654
	(6.917)	(−1.541)	(−1.261)
Controls	YES	YES	YES
	Small Cities
LNCOAGG	0.2815 ***	−0.6023	−0.2964
	(4.598)	(−1.291)	(−0.722)
Controls	YES	YES	YES

Note: *** and ** denotes significance levels of 1% and 5%, respectively. The values in parentheses are the z-statistics of the parameter estimates. City fixed effects and year fixed effects are controlled.

). The regression results indicate that ICA enhances GTFP across cities of all sizes, although the magnitude of this effect varies. Specifically, the promotion effect is strongest in megacities, while the effect size diminishes as the city size decreases, with small, medium-sized, and large cities exhibiting a negative correlation between city size and the impact of ICA on GTFP.

Regarding the effects on neighboring cities, ICA in small and medium-sized cities has no significant impact on GTFP in surrounding areas. In large cities, ICA significantly suppresses GTFP growth in neighboring cities, while in megacities, ICA substantially boosts GTFP in the surrounding regions.

The potential reasons for these differences are as follows: in smaller cities, agglomeration economies are generally based on the concentration of firms within the same or similar industries, known as “localization economies”. In this context, firms benefit from improved resource allocation efficiency and innovation levels through the sharing of intermediate goods, competition, and mutual learning, thereby enhancing production efficiency in these industries. This effect also applies to the coordinated agglomeration of producer services and manufacturing.

As city size increases, ICA in larger cities promotes vertical linkages between industries and facilitates specialized production, which enhances the utilization efficiency of production factors and reduces pollution emissions. However, as resource and environmental constraints intensify, the congestion costs associated with increased agglomeration become more pronounced. Additionally, population and consumption growth drive up energy usage and pollution emissions, offsetting the efficiency gains from improved resource allocation.

In even larger cities, the “filtering effect” associated with city size becomes more evident. Some manufacturing firms, unable to bear the high congestion costs, relocate to nearby or more distant specialized cities. This shift transforms the local coordination model from “low-end producer services–low-end manufacturing” to “high-end producer services–high-end manufacturing”. Furthermore, firms in large cities generally have higher productivity. Compared to medium-sized and small cities, large cities offer greater economic scale, more advanced infrastructure, and faster information flows, which reduce transaction costs for interrelated industries, create market opportunities, and attract additional resources. This generates a “siphon effect” that draws resources away from neighboring cities, thereby suppressing GTFP in those areas.

In contrast, in megacities, the “filtering effect” and “congestion effect” associated with ICA act through price mechanisms and market competition to encourage certain industries and firms to relocate to surrounding specialized cities. This redistribution mitigates some of the negative externalities of agglomeration, allowing megacities to positively influence GTFP in neighboring areas.

4.3.3. Heterogeneity of Urban Agglomerations

This study further investigated the heterogeneous effects of ICA on GTFP across five major urban clusters (see

Table 10. Regression results of the model for heterogeneity of urban agglomerations.

	Direct Effect	Indirect Effect	Total Effect
	Beijing–Tianjin–Hebei Urban Agglomeration
LNCOAGG	0.4556 ***	−0.1348	0.2999 ***
	(4.291)	(−0.291)	(3.059)
Controls	YES	YES	YES
	Yangtze River Delta Urban Agglomeration
LNCOAGG	0.2894 ***	−0.1703	0.1019
	(5.297)	(−1.294)	(0.842)
	YES	YES	YES
Controls	0.2894 ***	−0.1703	0.1019
	Middle Yangtze River Urban Agglomeration
LNCOAGG	0.0527	0.4273 **	0.4985 **
	(0.704)	(2.001)	(2.001)
Controls	YES	YES	YES
	Guangdong–Hong Kong–Macao Urban Agglomeration
LNCOAGG	0.3171	0.4040	0.5071 *
	(1.520)	(1.601)	(1.726)
Controls	YES	YES	YES
	Chengdu–Chongqing Urban Agglomeration
LNCOAGG	0.4247 ***	−1.5904 ***	−1.0248 ***
	(5.709)	(−2.988)	(−2.001)
Controls	YES	YES	YES

Note: ***, **, and * denotes significance levels of 1%, 5%, and 10%, respectively. The values in parentheses are the z-statistics of the parameter estimates. City fixed effects and year fixed effects are controlled.

). The regression results reveal that, except for the Middle Yangtze River and Pearl River Delta clusters, ICA positively influences GTFP in the local cities of the clusters. The effect size, in descending order, is as follows: Beijing–Tianjin–Hebei > Chengdu–Chongqing > Greater Bay Area (Guangdong–Hong Kong–Macau).

In terms of spillover effects on neighboring cities, ICA in the Middle Yangtze River cluster significantly enhances GTFP in the surrounding cities, whereas ICA in the Chengdu–Chongqing cluster significantly hinders GTFP growth in nearby areas. The ICA effects in the Beijing–Tianjin–Hebei, Yangtze River Delta, and Greater Bay Area clusters do not contribute to GTFP growth in neighboring cities.

These differences may arise because the Beijing–Tianjin–Hebei, Yangtze River Delta, and Chengdu–Chongqing clusters exhibit relatively high levels of internal industrial coordination, allowing the benefits of coordinated agglomeration—such as efficient factor allocation, technological advancement, and energy conservation—to be fully realized within cities. However, industrial coordination across cities within these clusters is constrained by local protectionism, market segmentation, and lingering administrative barriers, which impede factor mobility and prevent coordinated agglomeration from benefiting surrounding cities.

For the Middle Yangtze River cluster, the Blue Book of the Middle Yangtze River Urban Cluster: Evaluation Report on Coordinated Development (2017) indicates that industrial coordination between cities in the cluster surpasses the levels of coordination within individual cities. In the Chengdu–Chongqing cluster—a key focus of the Western Development Strategy—the presence of both Chengdu (a sub-provincial city) and Chongqing (a directly governed municipality) makes it the only “dual-core” urban cluster in Western China. However, the economic foundations of the surrounding small and medium-sized cities are relatively weak, with simpler industrial structures and limited internal industrial linkages, which often lead to a flow of resources from these peripheral cities into the central cities.

4.4. Test of Spatial Spillover Effect Decay

To examine the regional boundaries of ICA’s impact on urban GTFP, this section utilizes a threshold inverse-distance spatial weight matrix to conduct multiple estimations using the spatial Durbin model. The analysis is based on the magnitude, direction, and statistical significance of the spatial spillover coefficients to ensure a reasonable interpretation of the results. The specific setup of the threshold inverse-distance spatial weight matrix is as follows:

$$W_{ij}=\left\{\begin{array}{c}0 (i=j or d_{\mu }>d_{ij} or d_{ij}>d_l )\\ {\displaystyle \frac{1}{d_{ij}}} \left(i\ne j and d_{\mu }<d_{ij}<d_l\right)\end{array}\right.$$

(11)

where $d_{ij}$ represents the distance between regions of city i and city j, $d_{\mu }$ is the lower limit of the spatial threshold distance, and $d_l$ is the upper limit of the spatial threshold distance, which is set arbitrarily. In this study, the spatial threshold distance started at 50 km, and an SDM estimation was conducted for every 50 km increment, increasing up to 800 km. The spatial spillover coefficients of ICA on urban GTFP, along with their t-values, were recorded for threshold distances of 50 km, 100 km, 150 km, 200 km, 250 km, 300 km, 350 km, 400 km, 450 km, 500 km, 550 km, 600 km, 650 km, 700 km, 750 km, and 800 km (see

Table 11. Empirical results of spatial spillover effect decay.

Distance (KM)	Spillover Effect	Z-Value	Distance (KM)	Spillover Effect	Z-Value
50	−0.1213 ***	−3.125	450	−0.0978 *	−1.661
100	0.0177	0.885	500	−0.0878	−1.371
150	0.0601 **	1.986	550	−0.0447	−0.656
200	0.0003	0.014	600	−0.0187	−0.261
250	−0.0719 *	−1.726	650	−0.0131	−0.178
300	−0.0749 *	−1.658	700	−0.0135	−0.165
350	−0.0898 *	−1.768	750	0.0137	0.164
400	−0.1179 **	−2.179	800	0.0273	0.312

Note: ***, **, and * denote significance levels of 1%, 5%, and 10%, respectively.

).

First, the spatial spillover coefficient within a 50 km radius is significantly negative and passes the 1% significance test. As the distance increases, the spatial spillover coefficient reverses between 50 and 150 km, reaching a peak of 0.0476 at 100 km, where it becomes significantly positive. This pattern within the 0–150 km range indicates a “first-suppress, then-rise” attenuation trend in the spatial spillover effect.

Second, from 200 to 800 km, the spatial spillover coefficient remains negative, with the strongest negative spillover effect observed at 450 km.

The likely reason for this pattern is that when the spatial distance threshold is less than 50 km, ICA significantly suppresses GTFP growth in nearby cities, with the “siphon effect” playing a dominant role. A possible explanation for this is that cities within a 50 km radius of a central city experience lower trade and factor mobility costs, allowing the GTFP of the central city to exert a “siphon effect” on surrounding cities. Additionally, the closer the proximity, the more intense the competition is between cities, with local governments competing to implement policies that favor local industrial development, thereby attracting resources from neighboring areas. This results in a suppressive effect on the GTFP of surrounding cities—a phenomenon that is particularly pronounced between large central cities and adjacent smaller cities.

Within this distance range, ICA exerts a negative spatial spillover effect on GTFP, with the strongest negative effect observed at 450 km. This may be due to the influence of interprovincial differences in economic development, jurisdictional boundaries, and market segmentation around the 450 km mark, producing a negative spatial spillover effect. Additionally, under the guidance of local governments, cities in these regions often attempt to emulate the industrial development patterns of provincial central cities, creating a high degree of dependency on these central hubs. Border cities within provinces, in particular, may mimic the central city’s development model, resulting in an industrial structure that is mismatched with their own economic base and supporting infrastructure, thereby hindering GTFP improvement.

5. Discussion

5.1. Comparison of Results

This study leverages panel data from 283 Chinese cities spanning 2006 to 2020 to examine the transmission mechanisms, regulatory mechanisms, and spatial spillover effects of ICA on GTFP through both theoretical and empirical analyses. As the world’s second-largest economy, China’s practices in green transformation hold profound international relevance. Therefore, investigating the relationship between ICA and GTFP in China not only enriches our theoretical understanding but also provides valuable empirical insights and policy recommendations for other nations.

Regarding the transmission mechanism, this study differs from the existing literature, which often explains the impact of ICA on GTFP primarily through single channels such as technological innovation and progress [1,30,31]. Here, we systematically explore the transmission mechanism from the perspective of resource allocation, revealing that ICA significantly boosts GTFP by optimizing the allocation of capital and energy resources, although it does not enhance GTFP through improvements in labor allocation. This approach broadens the understanding of ICA’s mechanisms of influence and offers new solutions to resource misallocation issues, further enriching research on how ICA affects GTFP.

In terms of the regulatory mechanism, this study innovatively uses the PSTR model to examine how local government competition and environmental regulation moderate the ICA–GTFP relationship. This differs from previous studies, which mainly focused on moderators such as innovation agglomeration, institutional distance, and policy advantages [1,32]. Our results indicate that stronger environmental regulation significantly enhances ICA’s positive effect on GTFP, while the positive impact of ICA on GTFP diminishes when local government competition intensity exceeds a certain threshold.

In examining spatial spillover effects, this study diverges from the previous literature [30,33,34] by analyzing ICA’s impact on GTFP from the perspectives of industry type, city size, and urban cluster characteristics. Specifically, coordinated agglomeration of both high-end and low-end producer services with manufacturing raises local GTFP, while coordinated agglomeration of low-end producer services with manufacturing also enhances GTFP in neighboring cities. ICA in megacities positively influences both local and surrounding GTFP, whereas ICA in large cities tends to suppress GTFP in adjacent areas. Additionally, ICA promotes local GTFP in all urban clusters except the Middle Yangtze River and Pearl River Delta clusters; in the Middle Yangtze River cluster, ICA boosts GTFP in surrounding areas, whereas in the Chengdu–Chongqing cluster, ICA suppresses GTFP in neighboring regions. Furthermore, this study examines the spatial decay boundary of ICA’s impact on GTFP, finding that ICA exerts a significant positive spatial spillover effect within a 100 km radius. This discovery expands the scope of research on spatial spillover effects, offering valuable insights into the spatial boundaries of industrial agglomeration impacts and providing practical guidance for regional coordination and urban cluster planning. This suggests that appropriately planned ICA within defined spatial boundaries can effectively enhance green productivity across regions.

This study not only provides empirical support for China’s green transformation but also offers important insights for other countries as they pursue green economic transitions, formulate industrial policies, and plan regional development. Through well-designed industrial policies, countries can direct resources toward efficient, low-pollution industries, facilitating green technology diffusion and optimized resource allocation. In urban cluster planning, countries can design layouts that enable the agglomeration effects of core cities to extend to surrounding areas, fostering a coordinated regional green structure. Additionally, our findings underscore the critical roles of environmental regulation, regional coordination, and moderate competition within ICA, supporting China’s green transformation while offering practical guidance for sustainable development worldwide.

5.2. Policy Recommendations

In the context of green development, ICA represents a key strategy for transitioning the economy from extensive to intensive growth. This study finds that ICA significantly enhances GTFP in urban areas, indicating that a “dual-engine” ICA strategy can play a comprehensive role in improving GTFP in China. Based on these findings, the industrial spatial layout should be strategically planned at the top level to encourage collaborative spatial positioning, remove barriers to the coordinated distribution of producer services and manufacturing, and dismantle institutional obstacles to their deep integration. For example, by strengthening industrial linkages and promoting both vertical and horizontal expansion of the industrial chain within new districts and industrial parks at the national level, the “coordination level” and “coordination quality” can be gradually enhanced, leading to a highly integrated industrial system. Moreover, it is essential to strengthen the collection of micro-data on producer services and manufacturing firms, enabling a comprehensive evaluation of industrial coordination levels based on regional characteristics, thereby avoiding congestion and lock-in effects and preventing ineffective coordination.

To maximize ICA’s positive impact on GTFP, it is essential to first implement joint pollution control efforts within ICA regions and promote energy-saving policy alignment across regions. Second, it is important to steadily advance industry–university–research collaboration in regions where producer services and manufacturing are co-agglomerated, improve innovation systems, help firms align with market demands, and incentivize R&D investments. Lastly, we should actively dismantle administrative and market barriers; promote regional economic integration; enhance the capital, labor, technology, and energy factor markets; and reduce restrictions on factor mobility—particularly by establishing a robust energy rights trading system to promote efficient energy allocation. Furthermore, integrating transportation infrastructure across cities and optimizing urban spatial structures can reduce transport costs, facilitate the movement of resources and products between cities, stimulate producer services in neighboring cities, and reduce congestion and lock-in effects caused by ICA, thereby fully leveraging its positive spatial spillover effects.

Recognizing the heterogeneous impacts of ICA on GTFP, it is advisable to adopt multi-level, differentiated industrial policies to refine ICA’s effectiveness. Given that the coordination of high-end producer services with manufacturing has a stronger impact on urban GTFP than that of low-end services, regions should prioritize developing high-end industrial coordination, focusing on elevating the “quality” and “level” of high-end producer services in order to amplify their spillover effects on neighboring cities. In particular, integrating the financial sector with intelligent and green manufacturing is essential to support R&D, value creation, and green transformation in manufacturing. Meanwhile, enhancing the specialization of low-end producer services is necessary to support cost-efficiency and transaction effectiveness in manufacturing.

For cities of different scales, promoting an industrial coordination model aligned with their urban level and industrialization status is crucial. Megacities should deepen the integration of high-end manufacturing with high-end producer services, establishing a multi-tiered, networked, and advanced producer service system. Large, medium-sized, and small cities should reduce barriers to market entry and administrative approvals for producer service firms, promoting the development of local producer services. All types of cities should actively cultivate specialized services such as high-end R&D and consulting, fostering high-end producer service clusters in megacities, while forming a multi-layered industrial coordination structure through a “large supports small, many points drive the whole” model, fostering a coordinated development framework with distinct comparative advantages and functional specialization.

Research highlights that cities serve as vital platforms for industrial coordination. Within a 50 km radius, ICA may exert a “siphon effect” on GTFP in neighboring cities, potentially leading to congestion effects and diminishing or even negative marginal returns. However, within a 100 km radius, ICA can produce a positive spillover effect on GTFP in surrounding areas. Therefore, it is essential to further promote urbanization, with a focus on supporting the coordinated development of urban clusters. The spatial layout and collaborative positioning of producer services and manufacturing should be scaled over a larger geographic area. Given the high factor mobility within urban clusters, coordinated agglomeration of producer services and manufacturing should be central, using urban clusters as platforms to establish internally coordinated industrial agglomeration and industry–city integrated “green economic circles” that promote green and sustainable regional development.

These policy recommendations will have profound implications for regional economies and green development in practice. First, by rationally planning industrial spatial layouts and strengthening the coordinated agglomeration of producer services and manufacturing, regional economic transformation and upgrading can be advanced. Developing national new districts and industrial parks will guide traditional high-energy, high-pollution industries toward efficient and low-emission models, achieving a shift from resource-intensive to technology-intensive and innovation-driven growth, thereby improving regional economic quality and efficiency. Second, supporting the coordinated development of urban clusters and facilitating cross-city resource flows will enhance balanced development across urban and rural areas. By leveraging the resource and innovation spillover capacities of megacities to benefit surrounding small and medium-sized cities, a “large supports small, many points drive the whole” industrial structure can facilitate resource sharing and complementary development among megacities, core cities, and neighboring cities, narrowing the urban–rural economic gap. Specifically, differentiated policy support will enable megacities to focus on the deep integration of high-end manufacturing and services, while smaller cities concentrate on foundational industries, fostering functional specialization and comparative advantages across cities. This will enable cities to support one another’s development, significantly enhancing regional agglomeration efficiency and competitiveness, and creating a highly integrated industrial system. Finally, centering on urban clusters to create a CAPSMI-encompassing “green economic circle” will drive resource integration and coordinated development in adjacent areas. This will promote green innovation, technology diffusion, and resource sharing across the region, providing a foundation for green growth in surrounding cities and, ultimately, fostering a virtuous cycle of regional development.

5.3. Limitations and Future Research

This study employed data from 283 Chinese cities to explore the transmission mechanisms, regulatory mechanisms, and spatial spillover effects of ICA on urban GTFP, leading to a series of valuable findings. However, several limitations should be noted: (1) The analysis was based exclusively on panel data from 283 Chinese cities and did not include comparable data from other countries or regions, thereby lacking a cross-national or cross-regional comparison of the relationship between ICA and GTFP. This limits this study’s ability to assess whether similar patterns exist in different national or regional contexts. (2) This study primarily relied on macro-level data and lacked an analysis of individual firms at the micro level. Consequently, it does not reveal how specific aspects of production, innovation, and resource allocation at the firm level are influenced by ICA, constraining a more nuanced understanding of the transmission mechanisms involved.

6. Conclusions

Grounded in theoretical analysis, this study utilized panel data from 283 Chinese cities from 2006 to 2020 to empirically investigate the transmission mechanisms, regulatory mechanisms, and spatial spillover effects of ICA on urban GTFP. The key findings are as follows:

(1)

ICA has a positive effect on urban GTFP—a conclusion supported by multiple studies [1,30,31]. Notably, ICA enhances GTFP not by alleviating labor misallocation but by addressing the misallocation of capital and energy. To date, no research has validated this specific pathway, making this finding novel.

(2)

When the intensity of local government competition exceeds a threshold of 14.3825, the positive effect of ICA on urban GTFP diminishes from statistically significant to non-significant. Conversely, when the environmental regulation intensity surpasses a threshold of 0.4381, ICA’s positive impact on GTFP is significantly amplified.

(3)

The coordinated agglomeration of high-end producer services, low-end producer services, and manufacturing enhances GTFP in local cities, while the agglomeration of low-end producer services with manufacturing also improves GTFP in neighboring cities. In terms of city size, ICA promotes GTFP across all city types, with the strongest effect observed in megacities. Only ICA in megacities enhances GTFP in neighboring areas, while ICA in large cities suppresses GTFP growth in adjacent cities. Examining urban clusters, ICA boosts GTFP in local cities across all clusters except the Middle Yangtze River and Greater Bay Area clusters. The impact follows a hierarchy: Beijing–Tianjin–Hebei > Chengdu–Chongqing > Greater Bay Area. In the Middle Yangtze River cluster, ICA significantly promotes GTFP in neighboring cities, whereas in the Chengdu–Chongqing cluster, it significantly suppresses GTFP in surrounding areas. ICA in the Beijing–Tianjin–Hebei, Yangtze River Delta, and Greater Bay Area clusters does not enhance GTFP in neighboring cities.

(4)

ICA significantly promotes GTFP in local cities but does not improve GTFP in surrounding cities over extended distances. However, ICA does enhance GTFP within a 100 km radius of the local city. This finding diverges somewhat from the conclusions of other studies [30,33,34], suggesting that ICA’s positive effect on neighboring cities’ GTFP is subject to specific distance constraints.