Invisible Threads, Tangible Impacts: Industrial Networks and Land Use Efficiency in Chinese Cities

Abstract

Efficient urban land use is a cornerstone of sustainable city development, yet the drivers of such efficiency are increasingly complex in an era of spatial transformation. As industrial specialization and collaboration deepen, cities are becoming interconnected through complex networks. These “invisible threads” are redefining the dynamics of land use and spatial efficiency. This study examines the influence of intercity industrial networks on urban land use efficiency by constructing urban networks from multi-regional input–output data and evaluating city performance using a super-SBM model. We employed Tobit regression and mediation analysis to identify the mechanisms. Results indicate that both the quantity and quality of urban network connections significantly enhance land use efficiency, with notable differences across city types. The positive effect of industrial network centrality is most pronounced in large cities. In growing cities, both the number and quality of industrial linkages promote efficiency, whereas in shrinking cities, connection quality is more critical than quantity. Mechanism analysis reveals that industrial networks improve land use efficiency primarily by expanding intermediate goods markets and fostering technological innovation.

1. Introduction

Cities are considered the prime economic engines of growth [1]. With the rapid development of transportation infrastructure and information technology, inter-urban connections have become increasingly intensive and interdependent [2]. Urban networks refer to topological structures formed by cities through both tangible and intangible connections [3]. Tangible connections encompass hard infrastructure—such as railways, canals, aviation links, and telecommunications facilities—while intangible connections involve soft ties such as investment relationships, scientific collaboration, population mobility, and governmental cooperation [4,5,6,7,8,9]. Stemming from multifaceted cooperative relationships among individuals, enterprises, and institutions at the micro level [10], such linkages interweave cities into a dynamic network that enables the circulation of resources and production factors, thereby reshaping physical urban spaces and exerting significant impacts on urban land use [11].

The utilization of land resources lays the foundation for economic growth and the enhancement of urban residents’ living standards and well-being [12]. Urban areas accommodate more than half of the world’s population and generate over 75% of the global GDP [13]. As population growth accelerates and urbanization intensifies, efficient land utilization is essential for the sustainable development of urban areas [14,15]. Urban land use efficiency (ULUE), serving as a critical indicator of land use performance, is influenced by factors such as the rate of urbanization, economic growth patterns, and institutional capacity for policy implementation [16]. As urban morphology evolves from monocentric structures to polycentric and networked configurations [17], factors of production increasingly circulate and agglomerate through these urban networks, significantly impacting urban land productivity and sustainable urban development. Therefore, examining ULUE through the lens of urban networks offers a more nuanced understanding of the dynamics inherent in rapid urban development [18] and can inform strategies for optimizing land resource allocation.

Previous studies have constructed urban networks using the data of enterprise investment and railway connectivity, validating the positive impact of urban networks on economic growth [2,19]. However, limited research has examined the influence of the industrial urban network on ULUE. Networks constructed using the multi-regional input–output (MRIO) table capture the characteristics of industrial division and cooperative relationships between cities by tracing interregional economic flows among sectors [20]. From the industrial chain perspective, land resource allocation is shaped by urban networks through collaborative industrial development. In order to address the implications of China’s industrial networks for urban land use efficiency, this study investigates the question through the following three steps: First, it constructs an urban network using MRIO tables to reveal industrial linkages among cities. Second, it characterizes the spatial features of this network. Third, it investigates the relationship between the urban network and urban land use efficiency and identifies the underlying mechanisms.

The following sections are organized as follows: The second section reviews existing research on ULUE, urban networks, and MRIO tables. The third section analyzes the theoretical mechanisms of urban networks’ impact on ULUE. The fourth section outlines methods and data sources. The fifth section presents empirical results, including the characteristics and typical cities of the industrial urban network, the impact of the urban network on ULUE, heterogeneity analysis, and mechanism analysis. The last section consists of conclusions, discussions, and policy implications.

2. Literature Review

2.1. Urban Land Use Efficiency

Urban land use refers to the process by which urban land is allocated to different sectors and projects to meet human needs [21]. ULUE measures the effectiveness of land utilization. In economics, an economic activity is considered efficient when it can no longer improve anyone’s economic welfare without harming others [22]. Similarly, urban land use is considered efficient when it can improve overall urban efficiency without detrimentally affecting any specific area.

The research on ULUE primarily focuses on four aspects: indicator selection, method selection, spatiotemporal pattern analysis, and influencing factor analysis. In terms of indicator selection, ULUE typically includes indicators across economic, social, and ecological dimensions [23,24]. For method selection, commonly used approaches include the single-output method, stochastic frontier analysis (SFA), data envelopment analysis (DEA), and the super-SBM model with undesirable outputs [13,25,26,27,28,29]. Regarding spatiotemporal patterns, the existing literature analyzes the spatiotemporal evolution of ULUE at various scales, ranging from urban and national levels to the continental scale [16]. As for influencing factors, urbanization [30], industrial transformation [31], industrial agglomeration [32], public policies [33], and urban form [34] are found to have impacts on ULUE.

2.2. Urban Networks

In the context of globalization and informatization, the frequent flow of elements such as capital, human resources, materials, and information has driven the spatial concept to evolve from traditional fixed, static physical space to dynamic, fluid virtual space, forming a continuous, real-time community space across different regions [35]. Taylor (1997) proposed analyzing flowing space by relational data, examining urban systems from a network-oriented perspective [36]. Within this framework, cities emerge not only as nodes woven into the fabric of urban networks, but also as living entities shaped and transformed by the dynamic flows that course through these interconnections.

The urban connections encompass multi-dimensional interactions including information flows, capital flows, human flows, and material flows [37]. The relationships between cities constitute the “second nature of cities,” where urban networks and urban development evolve together, making intercity connectivity a key element in supporting sustainable urban development [11]. These intercity relationships generate economic benefits that individual cities could not achieve alone. Such benefits, arising from interactions within urban networks, can be referred to as the externalities of urban networks [2,38].

The externality effects of urban networks are primarily realized through three mechanisms: intercity synergy effects, network integration effects, and borrowed size effects. Unlike traditional agglomeration externalities, network externalities are no longer confined to geographical space, but rather emphasize the coordination and complementarity of urban industries [39]. Intercity synergy effects manifest as horizontal synergy between homogeneous cities and vertical synergy between heterogeneous cities, leading to economies of scale and economies of scope, respectively [40]. Network integration effects are reflected in multiple dimensions including spatial, functional, institutional, and cultural aspects, enhancing the systemic cohesion of urban networks [41]. The borrowed size effect enables smaller cities to achieve higher economic performance by leveraging the agglomeration economy advantages of larger cities [42]. However, small cities are simultaneously affected by the dual influences of borrowed scale and agglomeration shadow [43].

2.3. MRIO Tables

MRIO tables reflect the economic interdependence between different regions [20]. MRIO tables have been widely used to analyze spatial correlations of land use, water resources, energy, and carbon emissions [44]. Recently, some research has combined MRIO with complex network analysis to reflect cities’ positions in the industrial urban network and their impact on carbon emissions, agricultural land use, and international trade [17,45,46,47]. However, there is currently no research that combines MRIO tables with complex network analysis to examine the relationship between cities’ industrial positions and ULUE. With the evolution of supply chains and production networks [48], the influence of urban networks on land use potentially transcends the boundaries of geographic proximity.

Based on intercity industrial linkages derived from MRIO tables, this study constructs urban networks and investigates how these networks influence ULUE. The main contributions of this study are threefold. Firstly, this study constructs an intercity industrial network across China at the prefectural level utilizing MRIO tables, and subsequently examines its structural characteristics through complex network analysis methodologies. Secondly, the investigation incorporates both quantitative and qualitative dimensions of intercity industrial linkages when analyzing the influence of urban networks on ULUE. Thirdly, this study explores the underlying mechanisms through which urban networks affect ULUE, thereby providing insights into how industrial flows influence the allocation of land resources.

3. Theoretical Framework and Hypotheses

Industrial cooperation involves the complementary and collaborative use of resources, information, and technology among enterprises and industries across different geographical areas to achieve higher production efficiency [49]. According to Marshall’s theory of external economies, the geographic proximity of enterprises generates positive externalities by forming specialized labor pools, fostering the development of intermediate goods and specialized services, all of which contribute to lowering production costs for individual firms [50]. In the context of urban networks, the concept of spatial carriers extends from ‘industrial districts’ to ‘urban networks,’ emphasizing that urban network externalities are no longer constrained by geographical space [39]. Therefore, economies of scale and economies of scope can be achieved through cooperative and complementary economic activities in urban networks [51]. Specifically, economies of scale arise when similar firms share labor, infrastructures, or specialized production facilities to reduce average costs, while economies of scope emerge when different types of industries co-produce, providing complementary inputs and knowledge for multiple kinds of products. In addition, urban networks serve as pipelines for global and local knowledge exchange, thereby enhancing the knowledge accumulation within city nodes [52]. With lower production costs and higher innovative capacity, the economic outputs and urban land use efficiency are enhanced [53], as shown in Figure 1. Based on this, the following hypothesis is proposed.

H1: 

The higher the connectivity of nodes within an urban network, the higher the ULUE.

Urban networks can play a pivotal role in enhancing land use efficiency by the expansion of intermediate goods market. The ULUE is quantified as the ratio of input to output. Intermediate goods constitute an important component of production inputs. As urban networks facilitate a greater scale and diversity of intermediate goods, which is known as economies of scale and economies of scope [54], the effective use of intermediate goods can bolster productivity in three main ways. Firstly, the expansion in the scale of intermediate goods contributes to reducing production costs per unit. The multiplier effect inherent in linkages formed via intermediate goods can lead to substantial productivity gains [55]. Secondly, an expansion in the scale of the intermediate goods market facilitates a grater diversity of available products, which in turn enhances matching efficiency between suppliers and users [56]. Thirdly, by incorporating intermediate goods, firms can access new technologies and innovative production methods effectively [57]. Based on this, the following hypothesis is proposed.

H2: 

Industrial correlation promotes ULUE by the expansion of the intermediate goods market.

In the knowledge economy era, technological innovation has become the key driver of economic growth [58]. Knowledge spillover is the process by which knowledge is unintentionally leaked and diffused [59]. Urban networks, serving as pipelines of resource flows, enhance urban innovation capabilities by enabling knowledge spillovers [60,61]. Compared to single cities, urban networks feature higher levels of industrial diversification [62], which promotes intercity technological cooperation and knowledge spillovers [40] and accelerates technological progress. According to the endogenous growth theory, technological advancement could result in higher economic output and thereby improve ULUE [18]. Technological innovation drives industrial upgrading and steers industrial development toward high-value-added, low-environmental-impact trajectories. It improves economic benefits, reduces negative environmental externalities, and thereby increases ULUE. Based on this, the following hypothesis is proposed.

H3: 

Industrial correlation promotes ULUE by promoting technological innovation.

4. Methods and Data Sources

4.1. Study Area

In this study, we focus on prefecture-level cities in China and exclude those with major data gaps or undergoing administrative boundary changes during the study period, resulting in a final sample of 269 cities. The urban network is constructed based on prefecture-level MRIO tables; due to data availability constraints, the study period covers only three years: 2012, 2015, and 2017. In accordance with the “Notice of the State Council on Adjusting the Standards for Categorizing City Sizes” and to ensure a balanced distribution of cities across categories, cities are classified by resident population into four types: Small and Medium Cities (<1 million), Large Cities (1.00–4.99 million), Extra-Large Cities (5.00–9.99 million), and Mega Cities (≥10 million). Additionally, the classification of growing and shrinking cities is based on the population growth rate: cities with a population growth rate greater than 0 are classified as growing cities, while others are classified as shrinking cities [63]. The study area is shown in Figure 2.

4.2. Construction of Industrial Urban Networks

4.2.1. The MRIO Table

The MRIO table divides the economic system into multiple cities and industrial sectors, depicting the flow of intermediate goods both within and between cities, the demand of each city for final products, and the trade of goods and services across cities. This study aggregates the MRIO table at the city level to construct the industrial urban network.

Assume that there are N cities and S industries within the region (N = 1, 2, …, 269, S = 1, 2, …, 42), with each industry producing one type of product. The production in each industry across different cities requires both local production factors (labor and capital) and intermediate inputs (produced internally within the city and/or imported). Both the producing city and other cities can use the output either as intermediate inputs or as final products. The balance relationships can be expressed in matrix form, as follows [64]:

$$\left[\begin{array}{c}{\mathrm{X}}_1\\ {\mathrm{X}}_2\\ ⋮\\ {\mathrm{X}}_{\mathrm{N}}\end{array}\right]=\left[\begin{array}{cccc}{\mathrm{Z}}_{11}& {\mathrm{Z}}_{12}& \cdots & {\mathrm{Z}}_{1\mathrm{N}}\\ {\mathrm{Z}}_{21}& {\mathrm{Z}}_{22}& \cdots & {\mathrm{Z}}_{2\mathrm{N}}\\ ⋮& ⋮& \ddots & ⋮\\ {\mathrm{Z}}_{\mathrm{N}1}& {\mathrm{Z}}_{\mathrm{N}2}& \cdots & {\mathrm{Z}}_{\mathrm{N}\mathrm{N}}\end{array}\right]\left[\begin{array}{c}{\mathrm{X}}_1\\ {\mathrm{X}}_2\\ ⋮\\ {\mathrm{X}}_{\mathrm{N}}\end{array}\right]+\left[\begin{array}{c}{\mathrm{F}}_{11}+{\mathrm{F}}_{12}+\cdots +{\mathrm{F}}_{1\mathrm{N}}\\ {\mathrm{F}}_{21}+{\mathrm{F}}_{22}+\cdots +{\mathrm{F}}_{2\mathrm{N}}\\ ⋮\\ {\mathrm{F}}_{\mathrm{N}1}+{\mathrm{F}}_{\mathrm{N}2}+\cdots +{\mathrm{F}}_{\mathrm{N}\mathrm{N}}\end{array}\right]$$

(1)

where ${\mathrm{X}}_{\mathrm{i}}$ is an $\mathrm{S}\times 1$ matrix representing the total output of city $\mathrm{i}$; ${\mathrm{Z}}_{\mathrm{i}\mathrm{j}}$ is an $\mathrm{S}\times \mathrm{S}$ input–output coefficient matrix, also known as a direct consumption coefficient matrix, whose elements ${\mathrm{z}}_{\mathrm{i}\mathrm{j}}\left(\mathrm{s},\mathrm{t}\right)={\mathrm{m}}_{\mathrm{i}\mathrm{j}}\left(\mathrm{s},\mathrm{t}\right)/{\mathrm{x}}_{\mathrm{j}}\left(\mathrm{t}\right)$ represent the intermediate inputs from industry $\mathrm{s}$ in city $\mathrm{i}$ required for one unit of output in industry $\mathrm{t}$ in city $\mathrm{j}$; ${\mathrm{F}}_{\mathrm{i}\mathrm{j}}$ is an $\mathrm{S}\times 1$ matrix representing city $\mathrm{j}$’s final demand for products from city $\mathrm{i}$.

Based on the direct consumption coefficient matrix Z of the MRIO tables, which represents the direct consumption coefficients across different industries and cities, a city-level aggregation is performed to obtain a new matrix Z’. This matrix quantifies the direct consumption relationships between cities.

4.2.2. The Modified Floyd Algorithm

The Floyd algorithm is used to calculate the shortest paths between all pairs of nodes in a network [65,66]. Building on this idea, this study uses the modified Floyd algorithm to identify the strongest associations between cities based on the direct consumption coefficient matrix of cities [67]. A higher value in the direct consumption coefficient matrix indicates a greater level of input or consumption between the two cities.

For any pair of cities, such as city i and city j, the algorithm compares their direct connection strength with the strength of all possible indirect connections, such as the connection from city i to city k and then from city k to city j. At each iteration, the maximum value between the direct connection and all calculated indirect strengths is retained. By repeating this process, the algorithm ultimately captures the strongest possible relationship between every pair of cities. The modified Floyd algorithm is as follows:

$${\mathrm{d}}_{\mathrm{i}\mathrm{j}}^{\left(\mathrm{k}\right)}=\underset{\mathrm{i},\mathrm{j},\mathrm{k}\in \{\mathrm{1,2},\dots 42\}}{\mathrm{m}\mathrm{a}\mathrm{x}}\left\{{\mathrm{d}}_{\mathrm{i}\mathrm{j}}^{\left(\mathrm{k}-1\right)},{\displaystyle \frac{{\mathrm{d}}_{\mathrm{i}\mathrm{k}}^{\left(\mathrm{k}-1\right)}{\mathrm{d}}_{\mathrm{k}\mathrm{j}}^{\left(\mathrm{k}-1\right)}}{{\mathrm{d}}_{\mathrm{i}\mathrm{k}}^{\left(\mathrm{k}-1\right)}+{\mathrm{d}}_{\mathrm{k}\mathrm{j}}^{\left(\mathrm{k}-1\right)}}}\right\}$$

(2)

$${\tilde{\mathrm{z}}}_{\mathrm{i}\mathrm{j}}=\left\{\begin{array}{c}{\mathrm{z}}_{\mathrm{i}\mathrm{j}},\mathrm{l}\mathrm{p}{\mathrm{l}}_{\mathrm{i}\mathrm{j}}={\mathrm{z}}_{\mathrm{i}\mathrm{j}},\mathrm{i}\ne \mathrm{j}\\ 0,\hspace{1em}\mathrm{e}\mathrm{l}\mathrm{s}\mathrm{e}\end{array}\right.$$

(3)

The result of the modified Floyd algorithm is the strong association matrix $\tilde{\mathrm{Z}}$, which represents the strongest paths in the matrix Z’. In addition, we excluded the relationships within the city and retained only connections greater than the mean value as effective relationships.

4.3. Complex Network Analysis

Consider a complex network containing N nodes. The set A = {${\mathrm{a}}_{\mathrm{i}\mathrm{j}}$}, where ${\mathrm{a}}_{\mathrm{i}\mathrm{j}}$ represents the edge between node i and node j. This study employs two algorithms from complex network analysis to measure city network centrality. Degree centrality reflects the quantity of node connections, specifically the total number of direct connections a city has with other cities [68]. As shown in

Table 1. Index of complex network analysis.

Index	Definition	Formulation
Degree Centrality	DegCen reflects the quantity of node connections.	${\mathrm{D}\mathrm{e}\mathrm{g}\mathrm{C}\mathrm{e}\mathrm{n}}_{\mathrm{i}}={\sum }_{\mathrm{j}}{\mathrm{a}}_{\mathrm{i}\mathrm{j}}$
Eigenvector Centrality	EigenCen reflects both the quantity and quality of node connections.	${\mathrm{E}\mathrm{i}\mathrm{g}\mathrm{e}\mathrm{n}\mathrm{C}\mathrm{e}\mathrm{n}}_{\mathrm{i}}={\mathrm{x}}_{\mathrm{i}}=\tau \sum _{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{a}}_{\mathrm{i}\mathrm{j}}{\mathrm{x}}_{\mathrm{j}}$, $\mathrm{X}=\tau \mathrm{A}\mathrm{x}$ τ is a constant, ${\mathrm{x}}_{\mathrm{i}}$ and ${\mathrm{x}}_{\mathrm{j}}$ represent the importance of city i and city j.

, $a_{ij}$ measures the connection between cities, taking the value 1 when a connection exists and 0 otherwise. Cities with high degree centrality play “hub” roles in the network.

Eigenvector centrality considers both the quantity and quality of connections [69]. As defined in Table 1, τ is a constant, while ${\mathrm{x}}_{\mathrm{i}}$ and ${\mathrm{x}}_{\mathrm{j}}$ represent the importance of city i and city j, respectively. This algorithm uses iterative calculation, making a city’s importance dependent not only on its number of connections but also on the importance of the cities it connects to, thus more comprehensively reflecting the city’s strategic position in the network.

4.4. Measure of ULUE

4.4.1. Super-SBM Model

The Super-SBM model has been provided by Tone [70,71]. This model requires no pre-specified production frontier function and can effectively handle multiple inputs and outputs. It also addresses slack variables and incorporates undesirable outputs like environmental pollution into urban land use efficiency measurement, better aligning with sustainable development principles [72]. To reflect temporal trends in ULUE, cities from different periods are treated as distinct evaluation units in constructing the super-SBM model. The formula is as follows:

$$\begin{array}{ll}\rho & =\mathrm{m}\mathrm{i}\mathrm{n}{\displaystyle \frac{1-\frac{1}{\mathrm{m}}\sum _{\mathrm{i}=1}^{\mathrm{m}}\frac{{\mathrm{s}}_{\mathrm{i}}^{-}}{{\mathrm{x}}_{\mathrm{i}0}}}{1+\frac{1}{{\mathrm{s}}_1+{\mathrm{s}}_2}\left(\sum _{\mathrm{r}=1}^{{\mathrm{s}}_1}\frac{{\mathrm{s}}_{\mathrm{r}}^{\mathrm{g}}}{{\mathrm{y}}_{\mathrm{r}0}^{\mathrm{g}}}+\sum _{\mathrm{r}=1}^{{\mathrm{s}}_2}\frac{{\mathrm{s}}_{\mathrm{r}}^{\mathrm{b}}}{{\mathrm{y}}_{\mathrm{r}0}^{\mathrm{b}}}\right)}}\end{array}$$

(4)

$$\mathrm{s}. \mathrm{t}.{\mathrm{x}}_0=\mathrm{X}\lambda +{\mathrm{s}}^{-}$$

$${\mathrm{y}}_0^{\mathrm{g}}={\mathrm{Y}}^{\mathrm{g}}\lambda -{\mathrm{s}}^{\mathrm{g}}$$

$${\mathrm{y}}_0^{\mathrm{b}}={\mathrm{Y}}^{\mathrm{b}}\lambda +{\mathrm{s}}^{\mathrm{b}}$$

$${\mathrm{s}}^{-}\ge 0,{\mathrm{s}}^{\mathrm{g}}\ge 0,{\mathrm{s}}^{\mathrm{b}}\ge 0,\lambda \ge 0$$

In the above equations, $\rho$ represents the efficiency value of the evaluated unit, where a higher $\rho$ value indicates greater efficiency; X is the input, ${\mathrm{Y}}^{\mathrm{g}}$ is the expected output, and ${\mathrm{Y}}^{\mathrm{b}}$ is the undesirable output. ${\mathrm{s}}^{-}$, ${\mathrm{s}}^{\mathrm{g}}$, and ${\mathrm{s}}^{\mathrm{b}}$ are the slack variables for inputs, expected outputs, and undesirable outputs, respectively, while $\lambda$ denotes the weight.

4.4.2. Indicators of ULUE Measure

Since land is a complex system, the indicators of ULUE Measure should consider input and output of the economic, social and environmental aspects [73]. We have selected the positive input and output indicators across the economic, social, and environmental dimensions, as well as the negative output indicators from the environmental dimension [74], as shown in

Table 2. Indicator system for urban land use efficiency assessment.

Indicator Type	Evaluation Dimension	Indicator	Unit
Input Indicators	Land	Built-up Area	km2
	Capital	Fixed Capital Stock	CNY 104
	Labor	Employment in Secondary and Tertiary Industries	104 persons
Output Indicators	Desirable Output	Value-added of Secondary and Tertiary Industries	CNY 100 million
		Urban Residents’ Disposable Income	CNY/person
		Green Coverage Area in Built-up Areas	Km2
	Undesirable Output	Carbon Emissions from Construction Land	104 tons
		Industrial Wastewater Discharge	104 tons

. In terms of input dimension, the built-up area, fixed capital stock, and employment in secondary and tertiary industries are selected to reflect the input of land, capital, and labor [75,76]. Desirable outputs consider the positive outputs in urban land use, including the value-added of secondary and tertiary industries as economic benefits, the urban residents’ disposable income as social dimension output, and the green coverage area in built-up areas as environmental dimension output [75,77]. Meanwhile, undesirable outputs consist of carbon emissions from construction land and the industrial wastewater discharge [72,73,76,78].

We calculated carbon emissions from construction land using the indirect carbon emission coefficient method, which estimates emissions based on energy consumption from human activities [79]. Based on data availability, this study selects energy sources including liquefied petroleum gas, natural gas, and electricity. The calculation formula is as follows:

$$E_p=\sum e_i=\sum C_i\cdot {\theta }_i\cdot f_i$$

(5)

In the equation, ${\mathrm{E}}_{\mathrm{p}}$ represents the total carbon emissions from the construction land;${\mathrm{e}}_{\mathrm{i}}$ denotes the carbon emissions produced by the consumption of various types of energy; ${\mathrm{C}}_{\mathrm{i}}$ refers to the consumption of different types of energy; ${\theta }_{\mathrm{i}}$ is the conversion coefficient of each type of energy to standard coal; and ${\mathrm{f}}_{\mathrm{i}}$ is the carbon emission coefficient of each type of energy.

4.5. The Influence of Industrial Urban Network on ULUE

4.5.1. Tobit Regression Model

The regression model is applied to examine the impact of the industrial urban network on ULUE. As the value of ULUE is truncated to be above 0, there is an obvious truncation phenomenon. The use of ordinary least squares method could produce large biases and inconsistencies when performing parametric regression. Therefore, a Tobit regression model has been introduced [18].

$$\mathrm{U}\mathrm{L}\mathrm{U}{\mathrm{E}}_{\mathrm{i}\mathrm{t}}=\left\{\begin{array}{ll}{\mathrm{U}\mathrm{L}\mathrm{U}\mathrm{E}}_{\mathrm{i}\mathrm{t}}^{\ast }={\beta }_1\mathrm{U}{\mathrm{N}}_{\mathrm{i}\mathrm{t}}+{\beta }_2{\mathrm{X}}_{\mathrm{i}\mathrm{t}}+{\lambda }_{\mathrm{t}}+{\mu }_{\mathrm{i}}+{\epsilon }_{\mathrm{i}\mathrm{t}}& \left({\mathrm{Y}}_{\mathrm{i}\mathrm{t}}^{\ast }>0\right)\\ 0& \left({\mathrm{Y}}_{\mathrm{i}\mathrm{t}}^{\ast }\le 0\right)\end{array}\right.$$

(6)

i = 1, 2, …, 269, t = 2012, 2015, 2017

In the equation, i refers to the i-th DMU, t refers to the time period. ${\mathrm{U}\mathrm{L}\mathrm{U}\mathrm{E}}_{\mathrm{i}\mathrm{t}}^{\ast }$ is the dependent variable; $\mathrm{U}{\mathrm{N}}_{\mathrm{i}\mathrm{t}}$ is the matrix of the independent variable; ${\mathrm{X}}_{\mathrm{i}\mathrm{t}}$ is the matrix of the control variable; ${\beta }_1$ reflects the effect of the industrial urban network; ${\lambda }_{\mathrm{t}},{\mu }_{\mathrm{i}},{\epsilon }_{\mathrm{i}\mathrm{t}}$ refer to the year fixed effect, city fixed effect, and stochastic error, respectively.

4.5.2. Mediation Analysis

Mediation analysis is a statistical framework designed to quantify the mechanism between an independent variable and a dependent variable via intermediary variables [80]. This study employs a mediation effect framework to investigate how industrial urban networks influence ULUE, focusing on the dual pathways of the intermediate goods market and knowledge spillovers.

$${\mathrm{M}}_{\mathrm{i}\mathrm{t}}={\gamma }_1\mathrm{U}{\mathrm{N}}_{\mathrm{i}\mathrm{t}}+\phi {\mathrm{X}}_{\mathrm{i}\mathrm{t}}+{\lambda }_{\mathrm{t}}+{\mu }_{\mathrm{i}}+{\epsilon }_{\mathrm{i}\mathrm{t}}$$

(7)

$$\mathrm{U}\mathrm{L}\mathrm{U}{\mathrm{E}}_{\mathrm{i}\mathrm{t}}={\beta }_2\mathrm{U}{\mathrm{N}}_{\mathrm{i}\mathrm{t}}+\theta {\mathrm{M}}_{\mathrm{i}\mathrm{t}}+\pi {\mathrm{X}}_{\mathrm{i}\mathrm{t}}+{\lambda }_{\mathrm{t}}+{\mu }_{\mathrm{i}}+{\epsilon }_{\mathrm{i}\mathrm{t}}$$

(8)

where ${\mathrm{M}}_{\mathrm{i}\mathrm{t}}$ represents the effect of the independent variable on the mediator, $\theta$ reflects the effect of the mediator on the outcome, ${\beta }_2$ is the direct effect of the independent variable on the dependent variable after accounting for the mediator, and ${\mu }_{\mathrm{i}\mathrm{t}}$ are error terms.

4.5.3. Variables

ULUE is measured using the super-SBM model. Degree centrality and eigenvector centrality are computed via complex network analysis. The size of intermediate goods is calculated from the direct consumption coefficient matrix to obtain each city’s intermediate goods value. Technological innovation is proxied by the city-level count of invention patent grants. Control variables are reported in

Table 3. Definition of variables.

Category	Definition	Abbreviation (Unit)
Dependent variable
ULUE	The value of super-SBM of city	ULUE
Independent variable
Degree centrality	The degree centrality of city	DegCen
Eigenvector centrality	The eigenvector centrality of city	EigenCen
Mediation variable
Size of intermediate goods	The input value of intermediate goods of city	InterSize
Technological innovation	Invention patent grants of city	TechInov
Control Variable
Borrowing size effect	The population size of the national central city divided by the distance between the city and the national central city	BorSize (person/km2)
Agglomeration effect	Total urban population within a 200 km radius	Agglo (person)
Terrain ruggedness index	Difference between the highest and lowest elevations within the city	TerRug (m)
Slope gradient	Average slope gradient of the city	Slope (%)

. We account for the effects of borrowing size and agglomeration to isolate the impacts of megacity scale spillovers and local network density on urban land use efficiency within the urban network. We also control for cities’ terrain ruggedness and slope gradient to account for the influence of geomorphology on ULUE.

4.6. Data Sources

The data sources are as follows: (1) The MRIO tables at the prefecture city level are sourced from the China Emissions Accounts and Datasets (CEADs, https://www.ceads.net.cn/ accessed on 19 August 2025), covering the years 2012, 2015, and 2017. The input–output tables of each city include 42 sectors, consistent with the statistical caliber of the National Bureau of Statistics’ input–output tables for 2012. (2) Socioeconomic and energy consumption data are derived from the 2013, 2016, and 2018 editions of the China City Statistical Yearbook, which report data for the preceding year (https://data.oversea.cnki.net/ accessed on 19 August 2025). (3) The population data are sourced from the national census datasets from 2010 and 2020. (4) The data on invention patent grants come from the China National Intellectual Property Administration (CNIPA, https://www.cnipa.gov.cn/ accessed on 19 August 2025). (5) The list of coastal ports comes from the National Coastal Port Layout Planning (2006).

5. Results

5.1. Characteristics of the Industrial Urban Network

China’s provincial capitals play the role of “urban engines,” forging dense networks with their neighboring prefecture-level cities, as shown in Figure 3. For example, Chengdu links arms with an impressive 17 other cities across Sichuan, effectively weaving the province into a tight-knit economic fabric. Similarly, Wuhan radiates its influence to 7 cities in Hubei, Lanzhou connects with 10 cities across the stretches of Gansu, while Harbin serves as an economic hub for 5 cities dotting the vast lands of Heilongjiang. In these local constellations, provincial capitals act as economic “output” cities, channeling industrial flows and transmitting resources from their regional peers—the “input” cities.

Zooming out to the national scale, Shanghai and Beijing emerge as real “national dispatchers” in the industrial network. Shanghai alone establishes connections with a staggering 190 cities, and Beijing with 120, crisscrossing the national landscape with their influence. Yet, interestingly, these links often fall into the third- and fourth-level tiers—signaling a broad but subtly layered, multi-level urban network where the giants frame the flow but local bridges do much of the day-to-day connecting.

5.2. Typical Cities in the Industrial Urban Network

Based on the degree centrality and eigenvector centrality, cities can be classified into four roles, as shown in Figure 4: core hub cities, connecting support cities, regional node cities, and peripheral cities. Core hub cities score high on both metrics, positioning them at the network center with extensive connections and significant driving and radiating effects. Connecting support cities have high eigenvector centrality but medium or low degree centrality, serving as important bridges between core cities and other network cities, functioning as conduits in resource flows. Regional node cities display high degree centrality but low eigenvector centrality, indicating active local or surrounding connections but limited overall influence, primarily serving their own regions. Peripheral cities score low on both metrics, demonstrating weak connections and influence, and typically occupy marginal positions in the network.

Beijing, as a core hub city, exhibits exceptionally high degree centrality and eigenvector centrality. As the capital and a leading center of politics, economy, and technology, its industrial linkages extend nationwide, forming input–output relationships with 124 cities—120 of which serve as input cities to Beijing. Yan’an, functioning as a connecting support city in the network, has relatively low degree centrality but high eigenvector centrality. It maintains input–output relationships with both core cities such as Beijing and Xi’an, and peripheral cities such as Baoji, facilitating the integration of industrial networks across different regions. Jingdezhen, as a representative regional node city, exhibits relatively high degree centrality but low eigenvector centrality. Renowned for its ceramic heritage, the city maintains strong industrial connections with neighboring cities in the provinces of Hunan and Jiangxi. This positions Jingdezhen as a vital local connector, fostering the exchange of goods and resources within the region. Longnan exemplifies a peripheral city, as it exhibits relatively low degree centrality and eigenvector centrality, occupying a marginal position in the industrial urban network. Its industrial connections are minimal, with ties to only two cities—Lanzhou and Jiayuguan—reflecting its limited ability to influence resource flows within the industrial network.

5.3. The Impact of Industrial Urban Network on ULUE

5.3.1. Baseline Regression

Columns (1)–(4) in

Table 4. Results of baseline regression.

	(1)	(2)	(3)	(4)
Variables	ULUE	ULUE	ULUE	ULUE
DegCen	0.560 ***	0.533 ***
	(4.70)	(4.42)
EigenCen			0.448 ***	0.429 ***
			(3.18)	(2.98)
BorSize		0.002 **		0.003 ***
		(2.27)		(2.69)
Agglo		−0.007		−0.006
		(−0.78)		(−0.56)
TerRug		−0.274 **		−0.234 *
		(−2.21)		(−1.83)
Slope		0.001		0.001
		(0.58)		(0.39)
Constant	0.436 ***	0.575 ***	0.465 ***	0.564 ***
	(38.53)	(3.50)	(53.02)	(3.31)
Observations	807	807	807	807
Number of Cities	269	269	269	269
Wald Test	22.13	35.09	10.09	23.47
LR Test	185.9	172.5	210.3	193.0

Notes: t statistics in parentheses. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

measure the impact of degree centrality and eigenvector centrality of the industrial urban network on ULUE. Under ceteris paribus conditions, the impact coefficients of degree centrality and eigenvector centrality of the industrial urban network on ULUE are 0.533 and 0.429, respectively, both passing the significance test at the 1% level. This indicates that both the quantity and quality of external connections significantly and positively affect ULUE. This result supports Hypothesis 1. Ceteris paribus, a one-unit increase in degree centrality raises ULUE by 0.533, while a one-unit increase in eigenvector centrality raises ULUE by 0.429. In addition, the coefficient on BorSize is significantly positive, indicating that proximity to more populous national central cities helps improve ULUE. The coefficient on terrain ruggedness (TerRug) is significantly negative, suggesting that steeper terrain increases development costs and reduces land use efficiency.

5.3.2. Addressing Endogeneity Issues

(1)

Lagged variable method

The urban network demonstrates a preferential selection mechanism in which cities with larger economic scales are more likely to occupy central positions [81], thus introducing an endogeneity issue. To address the endogeneity issue arising from potential reverse causality, lagged first-order degree centrality and eigenvector centrality are used as explanatory variables in regression analysis, since past centrality measures cannot be affected by current economic efficiency. As shown in

Table 5. Results of the lagged variable method.

Variables	(1)	(2)
L.DegCen	0.441 ***
	(4.62)
L.EigenCen		0.198 *
		(1.73)
Control Variables	Yes	Yes
Observations	538	538
Number of Cities	269	269
Wald Test	53.73	32.80
LR Test	209.6	236.0

Notes: t statistics in parentheses. * and *** denote significance at the 10% and 1% levels, respectively.

, the impact coefficients of both centrality measures remain consistent with the baseline regression, demonstrating that the conclusions remain robust after mitigating the endogeneity issue caused by reverse causality. Relative to the baseline regression, the estimated coefficients for lagged first-order degree centrality and eigenvector centrality decrease. This evidence suggests that replacing contemporaneous with lagged variables can partially attenuate the upward bias induced by reverse causality.

(2)

Instrumental Variable Method

Instrumental variables help address endogeneity from omitted variables [82]. Network centrality and ULUE may be simultaneously influenced by unobserved variables such as regional policies and local industrial structures, leading to omitted variable bias. This study applies a two-step IV-Tobit model using the minimum distance to coastal ports as the instrument [83]. This choice is justified because the geographical distance between cities and coastal ports is exogenous, unaffected by ULUE, yet it can influence a city’s network position through trade costs and economic activities. This method effectively isolates the causal relationship between network centrality and ULUE, thereby eliminating potential estimation bias arising from contemporaneous omitted variables.

Results show that in columns (1) and (3) of

Table 6. Results of the two-step IV-Tobit model.

	(1)	(2)	(3)	(4)
Variables	DegCen	ULUE	EigenCen	ULUE
Port	−0.011 ***		−0.007 ***
	(−4.83)		(−3.16)
DegCen		2.707 ***
		(3.68)
EigenCen				4.196 ***
				(2.70)
Control Variables	Yes	Yes	Yes	Yes
Observations	807	807	807	807
Number of Cities	269	269	269	269
AR Test	20.87 **		20.88 ***
Wald Test	13.72 ***		7.31 ***

Notes: t statistics in parentheses. ** and *** denote significance at the 5% and 1% levels.

, the regression coefficients for the instrumental variables are all significant, and both the AR and Wald statistics for the weak instrument test are significant at the 1% level, indicating that the instrument meets the relevance requirement. In columns (2) and (4), the fitted coefficients for degree centrality and eigenvector centrality are significantly positive at the 1% level, demonstrating that even after addressing potential endogeneity from reverse causality, the centrality of the industrial urban network still has a significant positive effect on ULUE. In the first stage, port distance exhibits a significant negative correlation with both degree centrality and eigenvector centrality, aligning with the theoretical expectation that cities closer to ports occupy more central positions in urban networks. In the second stage, the estimated coefficients of both degree centrality and eigenvector centrality on ULUE are notably higher than in the baseline regression, suggesting that the baseline regression may have underestimated their relationship due to the influence of omitted variables.

5.3.3. Robustness Test

In the robustness test, we exclude municipalities to avoid potential bias introduced by their unique administrative status in political standing, resource allocation, and administrative structure. The results show that the estimated coefficients of degree centrality and eigenvector centrality remain consistent with the baseline regression, confirming the robustness of the baseline regression results, as shown in

Table 7. Results of robustness test after removing municipalities.

	(1)	(2)
Variables	ULUE	ULUE
DegCen	0.738 ***
	(4.16)
EigenCen		0.386 ***
		(2.60)
Control Variables	Yes	Yes
Observations	807	807
Number of Cities	269	269
Wald Test	32.12	20.49
LR Test	157.8	188.6

Notes: t statistics in parentheses. *** denotes significance at the 1% levels.

. Relative to the baseline regression, the estimated coefficient of degree centrality on ULUE exhibits an increase following the exclusion of municipalities. It means that ordinary cities rely more heavily on network connections for efficiency improvement, while municipalities benefit from their privileged administrative and economic status. Additionally, the exclusion of municipalities yields a reduction in the estimated coefficient of eigenvector centrality on ULUE. Since municipalities function as “critical nodes” in the network, their removal diminishes the quality of urban network connections, consequently attenuating the impact of network linkages on ULUE.

Key information in complex networks often lies in crucial relationships, and using top networks to simplify such networks is a common practice [84]. This study used the top 40% of industrial ties to construct a top network for robustness testing. The results show that the direction and significance of the coefficients of the explanatory variables are largely consistent with those of the baseline regression results, verifying the robustness of the main conclusion, as shown in

Table 8. Results of robustness test of top networks.

	(1)	(2)
Variables	ULUE	ULUE
DegCen	1.503 ***
	(4.22)
EigenCen		0.099
		(0.75)
Control Variables	Yes	Yes
Observations	807	807
Number of Cities	269	269
Wald Test	32.73	14.35
LR Test	162.0	186.8

Notes: t statistics in parentheses. *** denote significance at the 1% levels.

. Relative to the baseline regression, the estimated coefficient of degree centrality on ULUE increases in top networks, indicating that direct intercity connections exert a stronger positive influence on land use efficiency. By contrast, the coefficient on eigenvector centrality is smaller and statistically insignificant at the 10% level, indicating that it has little impact on ULUE in top networks due to relatively homogeneous connection quality.

5.4. Heterogeneity Analysis

(1)

Cities with different population scales

In accordance with the criteria of urban classification, the empirical results reveal distinctive patterns of network effects that vary by city size. As shown in

Table 9. Results of heterogeneity analysis based on population scale.

	Small and Medium Cities (≤1 Million)		Large Cities (1–4.99 Million)		Extra-Large Cities (5–9.99 Million)		Mega Cities (≥10 Million)
Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
DegCen	−1.099		1.122 ***		0.164		0.259 *
	(−1.13)		(3.66)		(0.93)		(1.72)
EigenCen		3.143		0.641 **		0.257 *		0.302
		(1.22)		(2.40)		(1.72)		(0.87)
Control Variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	27	27	526	526	210	210	44	44
Number of Cities	11	11	180	180	77	77	18	18
Wald Test	7.49	7.69	16.70	8.77	24.62	27.11	18.37	14.86
LR Test	1.13	1.39	72.20	88.21	94.82	97.83	26.52	26.11

Notes: t statistics in parentheses. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

, both degree centrality and eigenvector centrality exert significant positive effects on ULUE in large cities, and their coefficients in large cities exceed those in the baseline regression. Large cities, with their moderate economic scale and diversified industrial structure, can fully leverage the industrial urban network to enhance ULUE. In contrast, neither centrality measure demonstrates significant impact in small and medium cities or mega cities. Small and medium cities are constrained by their homogeneous industrial composition, while mega cities suffer from industrial crowding-out effects due to excessively high land prices, both weakening the optimization benefits of the urban network.

(2)

Growing cities and shrinking cities

Based on dynamic population trends, cities were classified into “growing cities” and “shrinking cities”. As shown in

Table 10. Development heterogeneity based on the industrial urban network.

	Shrinking Cities		Growing Cities
Variables	(1)	(2)	(3)	(4)
DegCen	0.403		0.608 ***
	(1.42)		(2.93)
EigenCen		5.346 ***		0.434 *
		(3.65)		(1.90)
Control Variables	Yes	Yes	Yes	Yes
Observations	246	246	351	351
Number of Cities	82	82	117	117
Wald Test	5.57	17.02	17.37	11.81
LR Test	61.51	62.76	63.50	71.04

Notes: t statistics in parentheses. * and *** denote significance at the 10% and 1% levels, respectively.

, degree centrality significantly improves ULUE in growing cities, but has no significant effect in shrinking cities. Eigenvector centrality has a significant positive impact on ULUE in both growing and shrinking cities. Compared with baseline models, the effect of degree centrality on ULUE is stronger in growing cities, while eigenvector centrality demonstrates greater influence in shrinking cities. This suggests that growing cities can enhance land use efficiency by establishing more diversified industrial connections, effectively integrating various industrial resources. In contrast, shrinking cities rely more on high-quality connections with core cities, maintaining land use efficiency by embedding in high-value industrial networks. This reflects their adaptive strategies under limited resources. In summary, ULUE in growing cities is significantly affected by both the number and quality of industrial input–output connections, whereas in shrinking cities, it is primarily driven by the quality of these connections.

5.5. Mechanism Analysis

5.5.1. Intermediate Goods Effect

The market size of intermediate goods is measured based on the input value of intermediate goods. The results indicate that the centrality of cities within the urban network significantly enhances ULUE by expanding the market size of intermediate goods and promoting firms’ access to external economies. As presented in

Table 11. Results of mediation effect test for the intermediate goods effect.

Variables	(1)	(2)	(3)
	ULUE	InterSize	ULUE
DegCen	0.533 ***	3.995 ***	0.243 *
	(4.42)	(8.51)	(1.81)
InterSize			0.044 ***
			(4.55)
The Proportion of Mediation Effect to Total Effect	0.385
Control Variables	Yes	Yes	Yes
Observations	807	807	807
Number of Cities	269	269	269
Wald Test	35.09	263.7	59.10
LR Test	172.5	613.3	148.9

Notes: t statistics in parentheses. * and *** denote significance at the 10% and 1% levels, respectively.

, Models (1) and (3) demonstrate that degree centrality has a direct positive effect on ULUE with a coefficient of 0.533, which remains significant at 0.243 when controlling for the influence of the market size of intermediate goods. Model (2) shows that the degree centrality also has significantly positive effect on the market size of intermediate goods. The proportion of the mediating effect to the total effect is 0.385, indicating that the intermediate goods market size accounts for 38.5% of the impact of degree centrality on ULUE. The result supports Hypothesis 2, confirming the significant mediating role of intermediate goods market size in the relationship between urban network and ULUE.

5.5.2. Technological Innovation Effect

Using patent licensing as a measure of technological innovation effect, the results indicate that degree centrality in the urban network significantly improves ULUE by promoting technological innovation. As detailed in

Table 12. Results of mediation effect test for the technological innovation effect.

Variables	(1)	(2)	(3)
	ULUE	TechInov	ULUE
DegCen	0.535 ***	4.439 ***	0.360 ***
	(4.42)	(4.97)	(2.76)
TechInov			0.017 ***
			(3.40)
The Proportion of Mediation Effect to Total Effect	0.180
Control Variables	Yes	Yes	Yes
Observations	807	807	807
Number of Cities	269	269	269
Wald Test	35.21	112.5	47.52
LR Test	169.9	675.7	170.1

Notes: t statistics in parentheses. *** denotes significance at the 1% levels.

, Models (1) and (3) demonstrate that the degree centrality has a direct positive effect on ULUE with a coefficient of 0.535, which remains statistically significant at 0.360 when controlling for the impact of patent licensing. Model (2) shows that the degree centrality also has significantly positive effect on technological innovation. The proportion of the mediating effect to the total effect is 0.180, indicating that technological innovation explains 18% of the total impact of urban centrality on ULUE. These results support Hypothesis 3, confirming the important mediating role of technological innovation in the centrality–efficiency relationship.

6. Conclusions, Discussions, and Policy Implications

6.1. Conclusions

This study utilizes MRIO tables to construct an industrial urban network for Chinese cities. The super-SBM model is employed to measure the ULUE. Furthermore, the Tobit model and mediation model are used to investigate the influence of the industrial urban network on ULUE. The key findings are as follows. First, both the quantity and quality of industrial linkages demonstrate significant positive effects on ULUE. The heterogeneity analysis indicates that the network position significantly enhances ULUE in large cities. Moreover, in growing cities, both the quantity and quality of industrial connections significantly affect ULUE, whereas in shrinking cities, it is mainly influenced by connection quality. Second, the market size of intermediate goods serves as a crucial mediation mechanism through which urban network centrality affects ULUE. Empirical analysis demonstrates that urban networks enhance ULUE by expanding the market size of intermediate goods, thereby promoting economies of scale and scope. Third, the technological innovation mechanism serves as another mediation mechanism. Cities occupying central positions in networks benefit from knowledge spillovers that stimulate technological advancement, which in turn drives industrial upgrading toward higher value-added activities.

6.2. Discussions

In terms of methods, this study examines the impact of urban networks on land use efficiency from perspectives both on the quantity and quality of network connections. In addition, we conduct a heterogeneity analysis for cities of different sizes and developmental stages, providing comparative insights into how network effects manifest differently among diverse urban typologies. Furthermore, this study examines network externality effects by analyzing mechanisms related to intermediate goods market size and knowledge spillovers.

Regarding the models, this study constructed a nationwide urban network based on MRIO tables and the modified Floyd algorithm. This approach extends land use research based on MRIO tables. While previous research explored land use issues from an embodied land perspective, this study investigates the problem through a complex network approach. Chen et al. (2019) discussed the interdependence of land use with the MRIO tables [44]. Chuai et al. (2023) investigated embodied land use and industry efficiency [64].

Regarding the results, this study reveals that both the quantity and quality of intercity industrial connections have significant positive impacts on ULUE, aligning with findings from previous research [85,86]. Ding et al. (2024) utilized corporate investment data to construct industrial urban networks and applied the MGWR model, demonstrating the critical role industrial urban networks play in fostering urban development [85]. Shi et al. (2024) discovered that urban innovation network connections enhance the utilization rate and allocation efficiency of construction land [86].

In addition, heterogeneity analysis indicates that network positions have varying effects on ULUE across various types of cities. In large cities, diversified industrial structures enable industrial urban network connections to effectively match economic supply and demand [87], thereby enhancing resource utilization efficiency. In contrast, small and medium-sized cities face challenges such as population loss and sluggish economic development [88], which weaken their network influence and make it difficult to fully utilize the external benefits of networks. Meanwhile, in mega-cities, characterized by excessive concentration of economic activities and nearly saturated land development [89], improvements in ULUE driven by industrial connections are also limited.

Regarding urban population dynamics, the ULUE of growing cities is influenced by both the quantity and quality of their network connections, whereas in shrinking cities, efficiency depends more on qualitative connections. Growing cities typically serve as core hubs within the industrial urban network, exerting strong influence on resource aggregation and distribution [88]. In contrast, shrinking cities, with diminished competitive advantages, have been increasingly marginalized within the urban network and have fewer network connections [90]. For these cities, improvement in ULUE largely relies on the quality of their limited industrial connections.

In addition, several limitations should be noted. To begin with, the analysis of urban network structure is conducted from a general industrial perspective, without incorporating other data sources such as transportation infrastructure or population mobility. This might not fully capture the comprehensive intercity connections. Future research could leverage multiple data sources to analyze the multi-level network linkages and characteristics among cities. Additionally, this study focuses on the broader industrial linkage network. Subsequent studies may examine industrial networks within specific sectors. Furthermore, the analysis is primarily based on a “node-centric” perspective, without exploring deeper relationships such as network hierarchy and community structure. Future research could incorporate perspectives on network stratification and community structure to further enrich the understanding of the mechanisms linking urban networks and ULUE.

6.3. Policy Implications

In light of the above analysis, the policy implications are summarized as follows. First, strengthening intercity industrial linkages is essential for enhancing ULUE. Each category of city should clearly articulate its functional role and fully harness its comparative advantages to foster deeper industrial collaboration. Integration policies that support the flow of resources, information, and technology across cities should be targeted. In addition, the establishment of intercity coordination mechanisms is essential for facilitating collaborative governance across administrative boundaries.

Second, it is important to optimize the city scale. A moderate scale is found to be more conducive to leveraging the positive effects of these networks. For megacities, the redistribution of industries and urban functions can help alleviate the pressures of excessive population concentration. For small and medium-sized cities, cultivating specialized industrial clusters and improving infrastructure and public services can help integrate them effectively into the regional industrial urban network, thus contributing to improvements in ULUE.

Third, promoting collaborative synergies among cities and technological innovation is fundamental for sustainable regional development. Mechanism analysis reveals that intermediate product markets play a pivotal role in regional resource allocation. This entails optimizing circulation environments, enhancing infrastructure and public services to facilitate efficient flows of intermediate products. Furthermore, establishing resource-sharing platforms can effectively amplify regional technological spillover effects and accelerate overall technological innovation progress.