Does Urban–Rural Integrated Development Promote Eco-Environmental Quality? Evidence from China

Abstract

The coordinated improvement of urban–rural integrated development (URID) and eco-environmental quality (EEQ) is a core strategic objective for pursuing sustainable development. However, existing studies have predominantly relied on qualitative discussions or single-region analyses, with insufficient empirical attention to multi-pathway mechanisms and spatial spillover effects. This study aims to examine the direct and heterogeneous effects of URID on EEQ, identify the dual mediating pathways, and quantify the spatial spillover effects across regions. Using panel data from 284 prefecture-level cities in China (2011–2023), this study employs panel regression, mediation analysis, and spatial econometric methods to investigate how URID affects EEQ. The results indicate that URID significantly promotes EEQ, with resource allocation efficiency and environmental regulation intensity serving as dual mediating pathways. This promoting effect varies across regions and policy stages. Moreover, URID exerts positive spatial spillover effects on neighboring regions. By providing national-scale evidence on mediating mechanisms and spatial externalities, this study extends prior research and offers implications for policy-making aimed at advancing the Sustainable Development Goals. The findings should be interpreted in light of the macro-level indicators employed and the observational research design.

1. Introduction

The coordinated development between urban–rural integrated development (URID) and eco-environmental quality (EEQ) constitutes a crucial academic and practical issue in overcoming the urban–rural dual structure and strengthening the basis for sustainable development [1,2]. In recent years, China’s urban–rural relationship has entered a stage of profound transformation. A slew of relevant policies, including household registration system reform, the Rural Revitalization Strategy, and the establishment of urban–rural integrated development pilot zones, have been rolled out in succession [3,4]. Developed regions including the Yangtze River Delta and the Pearl River Delta have taken the lead in forming integrated urban–rural infrastructure and collaborative ecological governance [5]. However, the dualistic nature of urban–rural ecological governance has not been fundamentally resolved. In rural areas, problems such as excessive application of chemical fertilizers and pesticides and unregulated discharge of domestic sewage remain prominent, while in urban areas, resource consumption and pollution emissions caused by industrial expansion have not been effectively alleviated [6,7]. China’s unique urban–rural dual structure, significant regional development differences, and rapidly advancing urbanization process have made the interaction mechanism between URID and EEQ more complex in the Chinese context, which cannot be directly explained by foreign paradigms alone [8,9,10,11]. Therefore, systematically analyzing how URID affects EEQ is both a pressing practical need and an important academic inquiry.

URID differs from conventional urbanization. It is characterized by the two-way flow of urban–rural factors (i.e., labor, capital, technology, and land), industrial coordination, and ecological co-governance [1,10], with its advancement driven by diverse institutional and market mechanisms including the optimization of factor allocation mediated by rural collective economic organizations [12], green finance [13], and the digital economy [14]. EEQ refers to a comprehensive evaluation of the health status of ecosystems and the quality of environmental elements within a specific region, covering multiple dimensions including pollution control and ecological service provision [15,16]. The governance of EEQ needs to address practical dilemmas such as the absence of governance subjects and insufficient institutional adaptability, so as to achieve a dynamic balance between ecological protection and socio-economic development [4,6,7].

The academic community has conducted explorations into the relationship between URID and EEQ at three progressive levels of analytical depth. At the first level, studies have examined the direct effects of URID on EEQ, yielding findings that diverge depending on contextual conditions. In regions where institutional barriers to factor flow are relatively low and bidirectional urban–rural interaction is well-established, URID tends to generate positive ecological outcomes. For instance, comprehensive territory consolidation promotes urban–rural spatial restructuring and ecological restoration [17,18,19]; urban–rural ecological co-governance mechanisms mitigate the fragmentation of environmental governance [20]; and in the Beijing–Tianjin–Hebei region, URID has facilitated the coordinated treatment of rural domestic sewage and the establishment of systematic pollution prevention and control systems [17]. Conversely, in regions where factor flow remains predominantly unidirectional or where institutional reform lags behind, URID may impose ecological pressure. The one-way flow of urban–rural factors can lead to overexploitation of rural ecological resources; industrial expansion in peri-urban areas results in cultivated land encroachment and environmental pollution [21]; and improper guidance of population agglomeration along the urban–rural gradient may trigger systemic problems such as reduced vegetation coverage and intensified heat island effects [22,23]. A study on urban agglomerations in the Yellow River Basin confirmed this conditionality, showing that some regions still face contradictory tension between development and protection due to insufficient integration depth [24]. These divergent findings suggest that the ecological effect of URID is not uniformly positive but is contingent upon the quality and directionality of factor flows and institutional arrangements—an observation that motivates the heterogeneity analysis in this study.

At the second research level, scholars have shifted from analyzing unilateral influence to exploring the coupling and coordinated linkage between URID and EEQ, which are regarded as mutually interactive systems. Relevant empirical investigations have been conducted in major regions, including the urban agglomeration in the middle reaches of the Yangtze River, the Yellow River Basin, and the Yangtze River Delta, as well as at the national prefecture-level city scale [5,18,24,25,26]. Scholars generally agree that coupling coordination between URID and EEQ has been increasing nationally [16], though regional differentiation remains distinct [25], with core driving factors including per capita GDP, urbanization level, industrial structure, and urban–rural factor circulation [5,18]. While these coupling studies have made valuable contributions in describing the co-evolution patterns of URID and EEQ, they are inherently limited as descriptive rather than explanatory tools: coupling coordination models characterize the degree of synchronization between two systems but do not identify the causal direction, transmission mechanisms, or spatial externalities of the URID–EEQ relationship.

This critical assessment of existing approaches at both levels points to a clear trajectory for advancing the field—moving from descriptive coupling analysis toward causal mechanism identification and spatial effect decomposition—which directly motivates the following specific research gaps. First, existing empirical studies have predominantly employed coupling coordination models to describe the co-evolution of URID and EEQ, but have not systematically examined the causal direction and transmission mechanisms from URID to EEQ using regression-based approaches. In particular, the mediating pathways through which URID affects EEQ—such as resource allocation efficiency and environmental regulation intensity—remain empirically unverified. Second, while spatial autocorrelation between URID and EEQ has been documented, the decomposition of direct versus indirect (spillover) effects of URID on EEQ through spatial econometric models remains largely absent, leaving the cross-regional dimension of this relationship underexplored. Third, existing national-scale studies have primarily focused on measuring integration levels or coupling degrees rather than testing multi-pathway impact mechanisms, and insufficient attention has been paid to special regions such as the areas on both sides of the Hu Line.

Against such research insufficiencies, this study centers on three key research issues:

(1)

Does URID have a significant positive promotional impact on EEQ, and does this effect exhibit heterogeneous features across various regions and different development stages?

(2)

Do resource allocation efficiency and environmental regulation intensity play a dual mediating role between URID and EEQ?

(3)

Does URID exert an obvious positive spatial spillover impact on the eco-environmental quality in neighboring regions?

By addressing these questions, this study contributes to the literature by providing regression-based evidence on the relationship between URID and EEQ, identifying the dual mediating pathways of resource allocation efficiency and environmental regulation intensity, and quantifying the spatial spillover effects through direct-indirect effect decomposition.

2. Theoretical Analysis and Research Hypotheses

Building on the literature assessment in Section 1, this section develops four testable hypotheses through theoretical analysis. The overall analytical framework is illustrated in Figure 1. URID is hypothesized to improve EEQ through three channels: a direct promoting effect (H1), indirect effects mediated by resource allocation efficiency (H2) and environmental regulation intensity (H3), and positive spatial spillover effects on neighboring regions (H4). The following subsections elaborate on each pathway in turn.

2.1. Analysis of the Main Effect of URID on EEQ

Previous studies have demonstrated that URID constitutes a progressive process that breaks down the urban–rural dual structure, encourages the free flow of factors between urban and rural areas, advances the equalization of public services, and fosters collaborative ecological governance [1,10]. Characterized by optimized factor flow, coordinated ecological co-governance, and spatial structure restructuring [12,19], URID restructures the interactive relationship within the urban–rural regional system and the mechanism of human–land coordination, thereby promoting the improvement of EEQ.

First, the optimized factor flow mechanism provides core impetus for EEQ improvement. The urban–rural dual structure restricts the cross-regional flow of factors such as labor, capital, technology, and land, leading to inefficient and resource-wasteful production and ecological degradation [12,27]. By breaking institutional barriers, URID enables the bidirectional flow of these factors—labor, capital, technology, and land—between urban and rural areas, thereby alleviating rural ecological carrying pressure, promoting agricultural green transformation, and curbing ecological damage from unordered urban expansion [19,28,29,30,31]. Second, the coordinated ecological co-governance mechanism comprehensively improves the efficiency of ecological environment governance. The urban–rural dual structure has given rise to an urban-oriented development pattern. In this context, rural regions are plagued by a lack of governance entities and inadequate resource input, while the spillover effects of urban governance cannot extend to rural areas [6,32]. URID breaks this segmentation by promoting public service equalization, establishing cross-urban–rural monitoring networks and joint prevention and control mechanisms, and shifting governance from “end-of-pipe treatment” to “source prevention and control” [3,4,6,33,34,35]. Third, the spatial structure restructuring mechanism significantly enhances ecosystem service functions. Under the urban–rural dual structure, single spatial functions—industrial agglomeration in urban areas and agricultural production in rural areas—weaken ecosystem services [10,20]. URID reconstructs spatial patterns to coordinate “production-living-ecological spaces” across multiple scales: optimizing industrial layout at the urban agglomeration level, constructing ecological corridors at the regional level, and improving rural human settlements at the local level [9,11,30,36,37].

Hypothesis 1.

Urban–rural integrated development has a significant positive effect on promoting eco-environmental quality.

2.2. Analysis of the Mediating Effect of URID on EEQ

2.2.1. The Mediating Role of Resource Allocation Efficiency

Under the urban–rural dual structure, barriers to factor flow lead to resource misallocation, which is mainly manifested in the overinvestment of capital in urban industrial sectors, driving excessive expansion of industrial production and a corresponding surge in pollution emissions. Meanwhile, rural agricultural sectors suffer from capital and technology shortages and severe land fragmentation, and the excessive application of chemical fertilizers and pesticides gives rise to non-point source pollution [27]. In addition, the inequality of urban–rural public services leads to insufficient accumulation of rural human capital, which further exacerbates the vicious cycle of “resource misallocation—inefficient and resource-wasteful production—ecological degradation” [27,38].

URID can optimize resource allocation efficiency by breaking down barriers to factor flow, thereby indirectly improving EEQ. Specifically, the bidirectional flow of capital, labor, land, and technology between urban and rural areas reduces factor misallocation—redirecting capital toward green industries, enabling labor reallocation to more productive sectors, optimizing land-use structure, and diffusing environmental protection technologies to rural areas [12,14,19,31,34,39].

The theory of factor misallocation holds that the improvement of resource allocation efficiency can reduce the inefficient consumption of production factors and pollution emissions [27,40]. The principles of ecological economics point out that the efficient allocation of resources can realize the coordination of economic development and ecological protection [3,4]. Therefore, by optimizing resource allocation efficiency, URID reduces the ecological damage caused by inefficient and resource-wasteful production and indirectly improves EEQ.

Hypothesis 2.

Urban–rural integrated development can indirectly improve eco-environmental quality by optimizing resource allocation efficiency.

2.2.2. The Mediating Role of Environmental Regulation Intensity

Against the backdrop of the urban–rural dual structure, environmental regulation exhibits typical features of urban–rural differentiation. Urban areas are equipped with a well-established environmental regulation system with high enforcement intensity, whereas rural areas suffer from the absence of regulatory subjects, imperfect institutions, and weak enforcement, thus becoming environmental protection depressions [6,32]. Such a dualistic framework has triggered the relocation of pollution-intensive enterprises from urban to rural regions. Coupled with the inadequate supervision of rural pollution emissions, this phenomenon has further expanded the ecological gap between urban and rural regions [33,37].

URID can strengthen environmental regulation intensity through both demand-side and supply-side channels, thereby indirectly improving EEQ. On the demand side, the extension of public services to rural areas raises residents’ environmental awareness, generating policy pressure for improved governance [6,32,35,41]. On the supply side, the expansion of green industries to rural areas prompts unified environmental standards, while the transfer of urban environmental technologies and professional resources enhances rural enforcement capacity [4,6,13,17]. These demand-pull and supply-push forces jointly elevate overall environmental regulation intensity. The enhancement of environmental regulation intensity constitutes a crucial pathway for EEQ improvement. Stronger regulation restricts polluting behaviors, incentivizes green innovation consistent with the Porter Hypothesis [13], and promotes collaborative governance across urban–rural boundaries [6,20,32,33].

Hypothesis 3.

Urban–rural integrated development can indirectly improve eco-environmental quality by strengthening environmental regulation intensity.

2.3. Analysis of the Spatial Spillover Effect of URID on EEQ

The ecological environment has significant public goods attributes and spatial correlation, and its quality improvement depends not only on local governance but also is deeply affected by adjacent regions [24,42]. By breaking the constraints of administrative boundaries, URID generates significant spatial spillover effects, a logic that is highly consistent with spatial dependence theory, ecological network theory, and regional coordinated development theory.

The spatial spillover effect is mainly manifested as positive spillover, achieved through three interconnected mechanisms. First, the technology diffusion mechanism: green production and ecological governance technologies spread from regions with higher URID levels to adjacent regions through cross-regional exchanges and enterprise cooperation [9,12,14,36,43]. Second, the factor flow mechanism: green capital, environmental protection talents, and ecological products flow across administrative boundaries, constituting positive spillover driving forces [3,34,39]. Third, the policy coordination mechanism: URID promotes unified environmental standards and cross-regional joint prevention and control mechanisms, avoiding “pollution transfer” while amplifying ecological co-governance effects through ecological compensation agreements and corridor co-construction [9,11,17,44].

It should be noted that the spatial spillover effect of URID may also have potential negative impacts. Some regions may trigger a “siphon effect” due to excessive factor agglomeration, leading to the loss of environmental protection resources and the weakening of ecological governance capacity in surrounding areas, or the emergence of a “pollution haven” phenomenon due to industrial transfer. However, from the perspective of theoretical logic and existing empirical evidence, the positive spillover effects brought by technology diffusion, optimal factor allocation, and policy coordination are more significant.

Hypothesis 4.

Urban–rural integrated development has a positive spatial spillover effect on the improvement of eco-environmental quality in related regions.

3. Research Design

3.1. Study Area Overview

This study focuses on 284 prefecture-level cities in the Chinese mainland, excluding the Hong Kong Special Administrative Region, Macao Special Administrative Region, Taiwan region, and districts/counties under municipalities directly under the Central Government, with a research period spanning 2011–2023 (Figure 2). The selection of this research scope is based on three primary considerations: First, as core administrative units for URID and EEQ governance in China, prefecture-level cities serve not only as key carriers for policy implementation but also as important spatial scales for factor mobility and industrial layout, enabling accurate reflection of the practical logic of urban–rural-ecological interactions. Second, the 284 prefecture-level cities cover three major geographical regions (Eastern, Central, and Western China) and include diverse city types such as core cities in urban agglomerations, medium and small-sized cities, and cities in ecologically fragile areas. They encompass both densely populated regions southeast of the Hu Line and ecologically sensitive areas northwest of the line, as well as pilot and non-pilot cities for URID. The comprehensiveness and heterogeneity of the sample can fully support the testing of core research issues such as regional differences and spatial spillover effects. Third, during the research period, China witnessed the intensive implementation of URID policies (e.g., the Rural Revitalization Strategy, household registration system reform) and ecological governance policies (e.g., the tough battle against pollution, regular environmental inspections). The development practices of the 284 prefecture-level cities can fully capture the dynamic correlation between policy evolution and the transformation of urban–rural-ecological relations, providing a rich empirical basis for mechanism analysis and policy implications. In terms of regional basic characteristics, the research area exhibits significant disparities in physical geography and development levels: prefecture-level cities in Eastern China boast a solid economic foundation, low barriers to urban–rural factor mobility, and relatively advanced ecological governance technologies and facilities; Central China, as a hub for industrial transfer and URID, faces dual demands of urban industrial upgrading and rural ecological protection; Prefecture-level cities in Western China have fragile ecological backgrounds, prominent contradictions between economic development and ecological protection, and a relatively backward level of URID. Such regional heterogeneity not only provides a natural research scenario for exploring the differentiated laws of the ecological effects of URID but also enhances the universality and practical guiding value of the research conclusions.

3.2. Model Specification

3.2.1. Benchmark Regression Model

Given that the panel data cover 284 prefecture-level cities over 13 years, a two-way fixed-effects specification is adopted to control for unobserved city-specific heterogeneity (e.g., time-invariant geographical and institutional characteristics) and common year-specific shocks (e.g., nationwide policy changes). The appropriateness of fixed effects over random effects is verified by the Hausman test (see Section 4.2). The benchmark regression model is specified as follows:

$${EEQ}_{it}={\beta }_0+{\beta }_1{URID}_{it}+\gamma {controls}_{it}+{\lambda }_i+{\upsilon }_t+{\epsilon }_{it}$$

(1)

Among them, $i$ denotes the city, t represents the time period, ${controls}_{it}$ denotes a set of control variables, ${\lambda }_i$ stands for the city fixed effect, ${\upsilon }_t$ denotes the time fixed effect, and ${\epsilon }_{it}$ is the random error term.

3.2.2. Mediating Effect Model

Since this research primarily explores the indirect impacts of URID on EEQ from the perspectives of resource allocation efficiency and environmental regulation intensity, a mediating effect model is established to verify the mechanism underlying the influence of URID on EEQ. The specific models are as follows:

$${Med}_{it}={\alpha }_0+{\alpha }_1{URID}_{it}+\gamma {controls}_{it}+{\lambda }_i+{\upsilon }_t+{\epsilon }_{it}$$

(2)

In the formulas, ${Med}_{it}$ represents the mediating variables, namely resource allocation efficiency and environmental regulation intensity. Following Jiang’s suggestion [45], on the basis that Model (1) has fully identified the causal relationship between URID and EEQ, this study combines the theoretical and logical analysis of the relationship between resource allocation efficiency/environmental regulation intensity and EEQ, and adopts Model (2) to identify the causal relationship between URID and resource allocation efficiency/environmental regulation intensity, thereby enabling the test of the mediating effect.

3.2.3. Spatial Spillover Effect Model

To examine the spatial spillover effects of URID on EEQ (Research Question 3) [46], a spatial panel model is employed. Among the commonly used specifications, the Spatial Autoregressive Model (SAR) captures spatial dependence in the dependent variable, the Spatial Error Model (SEM) accounts for spatial dependence in the error term, and the Spatial Durbin Model (SDM) simultaneously incorporates spatial lags of both the dependent and independent variables, nesting SAR and SEM as restricted cases. The SDM is adopted as the preferred specification for two reasons: first, it avoids the potential bias arising from omitting spatially lagged independent variables; second, it enables the decomposition of total effects into direct and indirect (spillover) components, which is essential for quantifying the cross-regional ecological externalities of URID. The appropriateness of the SDM is empirically verified through a sequential diagnostic procedure—including LM tests, LR tests, Wald tests, and the Hausman test—with detailed results reported in Section 4.6.2. The SDM is specified as follows:

$$EE{Q}_{it}={\beta }_0+{\beta }_{1}URI{D}_{it}+\rho \sum _{j=1}^n{W}_{ij}EE{Q}_{jt}+\theta \sum _{j=1}^n{W}_{ij}URI{D}_{jt}+\gamma Control{s}_{it}+{\lambda }_i+v_t+{\epsilon }_{it}$$

(3)

In the model, $W_{ij}$ denotes the element of the geographical distance weight matrix capturing the spatial relationship between cities $i$ and $j$; $\rho$ is the spatial autoregressive coefficient reflecting the influence of EEQ in neighboring cities on local EEQ; $\theta$ is the coefficient of the spatially lagged explanatory variable capturing the impact of URID in neighboring cities on local EEQ; ${\beta }_1$ and $\gamma$ retain the same meanings as in Equation (1). All other variables are defined as before.

3.3. Variable Selection

3.3.1. Explained Variable

EEQ. Based on the research methodology proposed by Xu et al. [47], the measurement system for EEQ is optimized and refined by incorporating the multidimensional considerations of synergistic improvement in ecology and the environment, with simultaneous emphasis on data accuracy and measurement comprehensiveness.

First, data preprocessing: Landsat imagery is chosen as the fundamental data source and undergoes systematic preprocessing steps, including cloud elimination, image mosaicking, stripe correction, clipping, synthesis, and resampling (with the resolution adjusted to 1 km), so as to guarantee the completeness and standardization of the dataset.

Second, calculation of core indexes: Using the preprocessed annual mean Landsat surface reflectance dataset, four core indexes are calculated separately: the Normalized Difference Vegetation Index (NDVI, greenness index), the Normalized Difference Built-up and Soil Index (NDBSI, dryness index), the Land Surface Temperature (LST, heat index), and the Wetness Index (WET, humidity index). Meanwhile, MODIS data are selected to calculate the above four indexes, which are used to fill in the missing values in the pixels of Landsat data and improve the integrity of the index data. In addition, combined with Land Use/Land Cover (LULC) data and with reference to the Technical Standards for Ecosystem State Assessment, the Land Cover Type Abundance Index (LCTAI) is calculated as an additional indicator.

Third, index standardization and comprehensive measurement: The five indicators mentioned above—NDVI, NDBSI, LST, WET and LCTAI—are successively processed through standardization, indicator combination and principal component analysis. Ultimately, the EEQ index is obtained through comprehensive measurement, which serves as the measurement result of eco-environmental quality. This methodology enables a scientific and comprehensive evaluation of EEQ by means of a modified pixel-level eco-environmental quality model.

3.3.2. Core Explanatory Variable

URID: Drawing on the research frameworks of Su et al. [48] and Zhou et al. [49], this study abides by the indicator design principles proposed by Zhou et al. [42], which stipulate that indicator selection should take into account three core dimensions: contrast indicators reflecting the incentives for URID, state indicators characterizing the outcomes of URID, and driving force indicators embodying the advancement process of URID. On this basis, an evaluation system for URID comprising 22 secondary indicators is constructed from four dimensions (

Table 1. Evaluation index system of URID.

First-Level Indicators	Second-Level Indicators	Indicator Explanation	Attribute	Indicator Role
Urban–rural Economic Integration	Urban–Rural Economic Aggregate	Logarithm of per capita GDP	Positive	State
	Ratio of Per Capita Income between Urban and Rural Residents	Per capita disposable income of urban residents/Per capita disposable income of rural residents	Negative	Contrast
	Ratio of Per Capita Consumption between Urban and Rural Residents	Per capita consumption expenditure of urban residents/Per capita consumption expenditure of rural residents	Negative	Contrast
	Urban–Rural Wage Income	Logarithm of average employee wage level	Positive	Driving Force
	Dual Contrast Coefficient	(Primary industry output value/Primary industry employment)/(Secondary and tertiary industry output value/Secondary and tertiary industry employment)	Positive	Contrast
	Urban–Rural Employment Structure	Employment in secondary and tertiary industries/Employment in primary industry	Positive	Contrast
	Scientific and Technological Innovation Capacity	Science and technology expenditure/Fiscal expenditure	Positive	Driving Force
Urban–rural Social Integration	Urban–Rural Basic Education	Public education expenditure/Fiscal expenditure	Positive	Driving Force
	Urban–Rural Medical Service Level	Number of licensed assistant physicians	Positive	Driving Force
	Urban–Rural Endowment Insurance Coverage Coefficient	Number of participants in basic endowment insurance for employees/Permanent population	Positive	Driving Force
	Urban–Rural Unemployment Insurance Coverage	Number of participants in unemployment insurance/Permanent population	Positive	Driving Force
	Urban–Rural Infrastructure Construction	Fixed asset investment in urban–rural construction/GDP	Positive	Driving Force
	Urban–Rural Population Density	Permanent population/Administrative area	Positive	State
Urban–rural Spatial Integration	Urban–Rural Transportation Network	Per capita road area	Positive	Driving Force
	Population Urbanization Rate	Permanent population urbanization rate	Positive	State
	Built-Up Area	Area of built-up district	Positive	Driving Force
	Per Capita Communication Expenditure	Per capita postal business income	Positive	Driving Force
	Built-Up Area Ratio	Area of built-up district/Total administrative area	Positive	State
Urban–rural Ecological Integration	Green Coverage Rate	Green coverage rate of built-up district	Positive	State
	Park Green Space Area	Per capita park green space area	Positive	State
	Domestic Waste Treatment	Domestic waste treatment rate	Positive	Driving Force
	Sewage Treatment	Sewage treatment rate	Positive	Driving Force

Note: Indicator roles are classified following Zhou et al. [42]: “Contrast” captures urban–rural disparities, “State” reflects URID outcomes, and “Driving Force” represents factors advancing the URID process.

). The entropy weight-TOPSIS method is applied to measure the URID level of prefecture-level cities from 2011 to 2023. It is acknowledged that moderate correlations among indicators within the same dimension are expected, as they collectively characterize a coherent aspect of URID. The entropy weighting method assigns weights based on the degree of cross-city variation (information entropy) in each indicator rather than the covariance structure among indicators, which makes it less sensitive to inter-indicator correlations compared to methods reliant on correlation matrices such as principal component analysis. Additionally, the robustness of the URID measurement is indirectly supported by the alternative variable test in Section 4.3.2, where the urban–rural income gap (Theil index) is employed as a substitute proxy and yields consistent core conclusions.

3.3.3. Control Variables

This research incorporates a series of control variables with reference to relevant existing studies [50,51]. The specific definitions and measurement methods of each variable are as follows: (1) Economic growth: Measured by the GDP growth rate, this variable reflects the dynamic characteristics of regional economic growth, as well as the dynamic changes in resource consumption, pollution emissions and ecological governance investment during the economic growth process. (2) Industrial structure: Comprehensively measured by the entropy method using two indicators, namely the advanced level of industrial structure (the ratio of the added value of the secondary industry to that of the tertiary industry) and the rationalization of industrial structure (Theil index). The adjustment and optimization of industrial structure are directly related to resource utilization efficiency and pollution emission intensity, and the improvement of its advanced and rationalized levels exerts a profound structural impact on EEQ. (3) Technological progress: Measured by the natural logarithm of R&D expenditure, this variable reflects the strength of regional scientific and technological innovation capacity. Technological progress can improve EEQ by enhancing the efficiency of resource recycling, promoting the application of clean production technologies, and optimizing pollution control methods. (4) Government intervention: Measured by the proportion of local general budgetary fiscal expenditure in GDP, this variable reflects the government’s regulatory capacity and investment intensity in regional economic and social development as well as ecological environment governance. (5) Opening-up: Measured by the proportion of actually utilized foreign direct investment in GDP, where FDI is converted into RMB at the annual actual RMB to US dollar exchange rate, this indicator reflects the potential effects of foreign capital inflows on technology spillovers, industrial upgrading and ecological governance. (6) Infrastructure: Measured by the internet penetration rate, this variable captures the level of digital infrastructure development, which may influence EEQ through environmental information dissemination, public environmental awareness, and digital environmental monitoring.

3.3.4. Mediating Variables

(1)

Resource Allocation Efficiency

Efficient resource allocation represents an allocation scenario where the free mobility of production factors under market forces leads to the maximization of social output. Market distortion is generally recognized as the primary cause of resource misallocation, and factor market distortion is the core driver of low resource allocation efficiency. On this basis, drawing on existing measurement methods [52,53], this study adopts the classic Cobb–Douglas (C-D) production function and synthesizes the distortions of two factors (capital and labor) to obtain the overall degree of market distortion. The resource misallocation degree of each city is measured by the relative difference between the market distortion degree of the city and the maximum value among all cities in the same year, which is used as the proxy variable for resource allocation efficiency. A larger value indicates a higher degree of resource misallocation and thus a lower level of resource allocation efficiency.

(2)

Environmental Regulation Intensity

Government work reports serve as the guiding framework for promoting law-based governance and executing relevant decisions while also acting as a programmatic blueprint for governmental operations. Consequently, the frequency of environmental governance-related expressions in these reports can effectively mirror the government’s commitment to environmental management and the overall orientation of environmental regulatory policies [54]. Although the proportion of environmental protection-related word frequency can measure the intensity of regional environmental regulation, the sole use of this indicator neglects semantic and sentential meanings, which is prone to measurement bias. In this research, environmental regulation intensity is quantified according to the proportion of words related to environmental protection keywords in municipal government work reports, relative to the total word count [55,56]. This method can more precisely reflect the governmental commitment to environmental governance.

3.4. Data Sources and Processing

The empirical data adopted in this research are primarily obtained from multiple official databases and statistical publications, including the China Statistical Yearbook, China Regional Economic Statistical Yearbook, China City Statistical Yearbook, China County Statistical Yearbook, China Environmental Yearbook, and China Urban Construction Statistical Yearbook. Supplementary data are also collected from provincial and municipal statistical yearbooks, the Patent Database of the China National Intellectual Property Administration, local economic and social development bulletins, the National Earth System Science Data Center (www.geodata.cn (accessed on 15 January 2026)), and the China Research Data Service Platform (CNRDS). The linear interpolation technique is employed to fill in individual missing values, and the descriptive statistics for all variables are summarized in

Table 2. Descriptive statistics of variables.

Variable	Symbol	Mean	Standard Deviation	Minimum	Maximum	Sample Size
Ecological Environment Quality	EEQ	0.3988	0.0906	0.0915	0.5945	3692
Urban–Rural Integrated Development	URID	0.0283	0.0280	0.0039	0.2636	3692
Economic Growth	EG	0.0742	0.0377	−0.2101	0.2200	3692
Industrial Structure	IS	0.7165	0.3093	0.0885	4.4460	3692
Technological Progress	TP	0.0145	0.0323	1.73 × 10−7	0.9358	3692
Government Intervention	GI	0.2096	0.1061	0.0438	0.9155	3692
Opening-Up	OPEN	0.0140	0.0159	1.34 × 10−6	0.1843	3692
Infrastructure	INF	0.2640	0.1433	0.0265	1.2187	3692
Resource Allocation Efficiency	RAE	0.3266	0.1674	0.0039	1.0000	3692
Environmental Regulation Intensity	ERI	0.9759	0.3364	0.0000	8.9430	3692

.

4. Results

4.1. Spatio-Temporal Pattern Characteristics of URID and EEQ

4.1.1. Temporal Evolution Characteristics

Figure 3a depicts the temporal dynamic changes of URID across China during the period 2011–2023. Overall, URID displayed a continuous annual upward tendency, rising steadily from 0.0221 in 2011 to 0.0377 in 2023, representing an average annual growth rate of 4.42%. In terms of temporal change, URID experienced two main stages: an initial exploration stage (2011–2017), during which growth was slow and fluctuating as the policy focus remained on rural infrastructure without a systematic framework for bidirectional factor flow; and an accelerated improvement stage (2018–2023), during which URID rose rapidly from 0.0307 to 0.0377 (an increase of 22.80%), driven by the Rural Revitalization Strategy and related institutional reforms. A detailed discussion of the policy context underlying each stage is provided in Appendix A.

From 2011 to 2023, China’s EEQ exhibited a fluctuating upward trend, with 2019 serving as the turning point (Figure 3b). It rose moderately from 0.4682 in 2011 to 0.5078 in 2023, with an average annual growth rate of approximately 0.65. The temporal evolution can be divided into three stages: a slow improvement stage (2011–2018), a short-term correction in 2019 (EEQ declined from 0.5006 to 0.4891, possibly reflecting phased pains in industrial green transformation), and a steady optimization stage (2020–2023) driven by the deepening of the ecological civilization strategy. A detailed analysis of each stage is provided in Appendix A.

4.1.2. Spatial Evolution Characteristics

To explore the spatial distribution features and dynamic evolution rules of URID and EEQ across China, this research adopts the natural breaks classification method in ArcGIS 10.8 (Esri, Redlands, CA, USA) to categorize the research subjects into five grades: low, relatively low, medium, relatively high, and high.

The spatial distribution of URID evolved from point-like agglomeration in core eastern cities (2011) to block-like agglomeration across urban agglomerations (2023), with the radiation-driven role of core cities continuously strengthening and overall spatial polarization gradually decreasing (Figure 4). This evolution is highly consistent with the national URID policies, the regional coordinated development strategy, and the laws of market factor flow. Detailed descriptions of each time cross-section are provided in Appendix A.

The spatial distribution of EEQ exhibited a pattern of “strong in the south and weak in the north” that gradually converged over the study period (Figure 5). High-value areas expanded from discrete patches in South China and Southwest China to cross-regional contiguous belts, while low-value areas in North China contracted significantly, reflecting the coupled synergy between ecological governance and regional development under the ecological civilization strategy. Detailed cross-sectional analyses are provided in Appendix A.

4.2. Benchmark Regression Analysis

Following the Hausman test, this study employs a two-way fixed-effects model to examine the impact of URID on EEQ empirically, with the baseline regression results presented in

Table 3. Benchmark regression results.

Variables	FE (without Controls)	FE (with Controls)
	(1)	(2)
URID	0.4217 *** (0.0403)	0.1889 *** (0.0032)
EG		−0.0354 *** (0.0097)
IS		0.0039 ** (0.0018)
TP		−0.0818 * (0.0476)
GI		−0.0316 *** (0.0072)
OPEN		0.0671 ** (0.0290)
INF		0.0190 *** (0.0029)
Constant	0.3924 *** (0.0010)	0.3997 *** (0.0027)
Time Fixed Effects	Yes	Yes
City Fixed Effects	Yes	Yes
N	3692	3692
R-squared	0.311	0.592

Notes: *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. Clustered robust standard errors are reported in parentheses. Model (1) excludes control variables to show the baseline association. The same applies to the following tables.

. Model (1) represents the baseline regression without control variables, whereas Model (2) includes the key control variables. The regression coefficients of the core explanatory variable URID are significantly positive in both specifications and are statistically significant at the 1% level. In Model (1), the coefficient of URID is 0.4217, indicating that without considering other interfering factors, URID has a significant positive promoting effect on EEQ. After adding control variables in Model (2), the coefficient decreases to 0.1889 but remains highly significant, suggesting that even after excluding the impacts of confounding factors, the improvement effect of URID on EEQ still exists stably. This result provides direct empirical support for the main effect pathway illustrated in Figure 1 (H1). Consistent with the theoretical analysis in Section 2.1, the significantly positive coefficient confirms that URID promotes EEQ through the combined mechanisms of optimized factor flow, coordinated ecological co-governance, and spatial structure restructuring, with the specific transmission channels further examined through the mediating effect analysis in Section 4.5. Thus, Hypothesis 1 is positively verified.

Regarding the control variables, economic growth (−0.0354, p < 0.01) and government intervention (−0.0316, p < 0.01) show significant negative effects, suggesting that extensive development paths and the tendency of “valuing growth over environmental protection” in some regions still exert short-term ecological pressure. Industrial structure (0.0039, p < 0.05), opening-up (0.0671, p < 0.05), and infrastructure (0.0190, p < 0.01) exhibit significant positive effects, reflecting the ecological dividends of structural upgrading, technology spillovers from foreign investment, and digital infrastructure improvement, respectively. Technological progress shows a significant negative coefficient (−0.0818, p < 0.10), possibly attributable to the short-term “rebound effect” of expanded production scale.

4.3. Robustness Analysis

4.3.1. Endogeneity Test

In addition to the factors controlled in the benchmark regression, there may exist unobservable variables that exert an impact on EEQ, thereby resulting in omitted variable bias. Meanwhile, a significant two-way causal relationship may exist between URID and EEQ, which further gives rise to endogeneity bias in the model. To resolve this problem, this study constructs instrumental variables based on topographic features and exogenous policy shocks and employs the two-stage least squares (IV-2SLS) approach to address endogeneity issues.

Referring to the study of Wang et al. [57], due to the good exogeneity of natural geographical factors, this paper further uses urban topographic relief to construct an instrumental variable. However, since this indicator is time-invariant, to avoid the problem of difficulty in measurement in the fixed-effects model, drawing on the study of Nunn and Qian [58], a time-varying variable is multiplied by a historically time-invariant variable to characterize the time variability of the instrumental variable. Here, referring to the processing methods of Zheng et al. [59] and Wang et al. [60], the interaction term between topographic relief and the URID index of the previous year is used as the first instrumental variable (IV1). A smaller topographic relief is more conducive to building a convenient urban–rural transportation network; reducing factor flow costs; promoting the diffusion of urban industries, capital, and technology to rural areas; and at the same time facilitating large-scale agricultural operation and efficient utilization of rural ecological resources, thereby significantly affecting the level of URID. However, as a natural geographical feature, relief is an exogenously given variable and has no correlation with EEQ. In addition, the “Reform Plan for National URID Pilot Zones” issued in 2019 announced 11 national URID pilot zones. Referring to the study of Peng et al. [61], exogenous policy shocks are used as instrumental variables to re-estimate the model. Thus, this paper sets a dummy variable for national urban–rural integrated development pilot zones, assigning a value of 1 to cities where urban–rural integrated development pilot zones were established after 2020, and 0 to other samples. The dummy variable for national URID pilot zones (IV2) is a policy variable overall planned by the central government and steadily promoted by local governments, with strong exogeneity. In addition, the core task of these pilot zones is to break down barriers to urban–rural factor flow and deepen URID through institutional innovation, and their pilot policies can directly empower the improvement of local URID levels, which meets the relevance requirement of instrumental variables.

Table 4. Endogeneity test: instrumental variable regression results.

Variables	First-Stage Results	Second-Stage Results	First-Stage Results	Second-Stage Results
	(1)	(2)	(3)	(4)
	URID	EEQ	URID	EEQ
Instrumental Variable	−0.0079 *** (0.0068)		0.0271 *** (0.0097)
URID		0.2137 ***(0.1293)		0.0779 ** (0.0322)
Kleibergen–Paap rk LM Statistic	19.42 [0.0000]		6.16 [0.0031]
Kleibergen–Paap Wald rk F Statistic	10.46 {8.96}		10.42 {8.96}
Control Variables	Controlled	Controlled	Controlled	Controlled
Time Fixed Effects	Yes	Yes	Yes	Yes
City Fixed Effects	Yes	Yes	Yes	Yes
N	3692	3692	3692	3692
F Statistic	13.36		12.04

Notes: **, and *** indicate statistical significance at the 5% and 1% levels, respectively. Kleibergen–Paap rk LM statistic is used for the identification test of instrumental variables, and Kleibergen–Paap Wald rk F statistic is for the weak instrumental variable test. Values in [ ] are the p-values of the LM test, and values in { } are the Stock–Yogo 10% critical values.

reports the estimation results obtained using instrumental variable methods. Columns (1) and (3) show the first-stage regression outcomes, indicating that all first-stage F-statistics are greater than 10. This result confirms a strong correlation between the selected instrumental variables and URID. The underidentification test indicates that the Kleibergen–Paap rk LM statistics are significant at the 1% level, supporting the validity of the instrumental variable identification. Furthermore, the Kleibergen-Paap rk Wald F-statistics all exceed the 10% Stock–Yogo critical values, thus eliminating the weak instrument problem. In the second-stage regression of the 2SLS model (Columns (2) and (4)), the effect coefficients of URID on EEQ are 0.2137 and 0.0779, which are significant at the 1% and 5% levels, respectively. This indicates that URID continues to exert a positive influence on EEQ, which is in line with the benchmark regression results obtained earlier.

4.3.2. Robustness Analysis

This paper conducts a series of robustness checks by replacing the core explanatory variable, shortening the research period, and excluding samples of key cities, with the corresponding results displayed in

Table 5. Robustness tests.

Variables	EEQ
	(1)	(2)	(3)	(4)
URID	−0.1424 *** (0.0171)	0.2373 *** (0.0635)	0.1909 *** (0.5239)	0.1824 *** (0.0555)
Control Variables	Controlled	Controlled	Controlled	Controlled
Constant	0.4159 *** (0.0030)	0.4098 *** (0.0034)	0.4000 *** (0.0027)	0.4047 *** (0.0028)
Time Fixed Effects	Yes	Yes	Yes	Yes
City Fixed Effects	Yes	Yes	Yes	Yes
N	3692	2840	3640	3354
R-squared	0.745	0.761	0.578	0.600

Notes: *** indicates statistical significance at the 1% level. Column (1) represents the replacement of the core explanatory variable, Column (2) denotes the shortening of the time window, Column (3) indicates the exclusion of municipalities directly under the Central Government, and Column (4) stands for the exclusion of provincial capital cities.

. First, the core explanatory variable is substituted. In general, a narrower urban–rural income gap (measured by the Theil index) corresponds to lower obstacles in factor mobility between urban and rural sectors, more balanced allocation of resources and public services, and a higher actual level of urban–rural integration development. Accordingly, this study adopts the urban–rural income gap as a proxy indicator for URID to perform robustness testing [62], with the results reported in Column (1). The coefficient of the urban–rural income gap is −0.1424 and passes the 1% significance test, which logically echoes the conclusion of positive significance in the benchmark regression. It indicates that the narrowing of the urban–rural income gap (improvement of integration level) will significantly improve EEQ, which is consistent with theoretical expectations. Second, shortening the time window: After the outbreak of the COVID-19 pandemic in 2020, the flow of urban–rural factors and the rhythm of economic activities were significantly impacted, interfering with the conventional correlation between the two. Here, data during the pandemic period (2020–2022) are excluded, and the results are shown in Column (2). The coefficient of URID is 0.2373, which is still significantly positive at the 1% level, and the absolute value of the coefficient is slightly higher than that of the benchmark regression (0.1889). This reflects that the correlation between the two is more direct under the scenario of conventional development, further confirming the stability of the core conclusion. Third, excluding key cities: Considering that key cities are significantly different from other ordinary prefecture-level cities in terms of economic foundation, policy support, and resource agglomeration, which may interfere with the identification of the core relationship between URID and EEQ, four municipalities and 26 provincial capital cities are excluded, and the results are shown in Columns (3) and (4). The coefficients of URID are 0.1909 and 0.1824, respectively, both passing the 1% significance test, and the coefficient sizes are highly close to that of the benchmark regression (0.1889). This indicates that the particularity of key cities does not interfere with the core correlation, and the estimation results are reliable.

4.4. Heterogeneous Effects Analysis

This section conducts grouped regressions based on geographical location, policy background, and population scale. Meanwhile, the Chow test and Fisher combination test are adopted to perform inter-group coefficient difference tests on the regression results.

4.4.1. Geographical Location Heterogeneity

In this study, 284 prefecture-level cities are categorized into three sub-samples: 100 in the eastern zone, 100 in the central zone, and 84 in the western zone. The corresponding estimation outcomes are reported in Columns (1)–(3) of

Table 6. Heterogeneity analysis.

Variables	EEQ
	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	Eastern Region	Central Region	Western Region	2011–2014	2015–2023	Southeast of the Hu Line	Northwest of the Hu Line
URID	0.2958 *** (0.0662)	0.3608 ** (0.1134)	0.1679 (0.1075)	0.3526 *** (0.1331)	0.2469 *** (0.0582)	0.1636 *** (0.0568)	0.0658 (0.0995)
Control Variables	Controlled	Controlled	Controlled	Controlled	Controlled	Controlled	Controlled
Constant	0.3676 *** (0.0047)	0.4151 *** (0.0048)	0.4097 *** (0.0048)	0.4321 *** (0.0104)	0.3884 *** (0.0028)	0.4076 *** (0.0030)	0.3239 *** (0.0054)
Time Fixed Effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes
City Fixed Effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	1300	1300	1092	1136	2556	3341	351
R-squared	0.611	0.709	0.651	0.628	0.395	0.637	0.101

Notes: ** and *** indicate statistical significance at the 5% and 1% levels, respectively.

. URID generates a significantly positive boosting effect on EEQ in both the eastern (0.2958, p < 0.01) and central (0.3608, p < 0.05) regions, whereas its favorable influence in the western region is not statistically significant. The coefficient differences across regions are consistent with the conditional analysis in the Introduction: the promoting effect is significant in the eastern and central regions where institutional barriers to factor flow are relatively low and bidirectional urban–rural interaction is well-established, while the effect is insignificant in the western region where ecological fragility, weak economic foundations, and insufficient factor flow efficiency constrain the ecological benefits of URID.

4.4.2. Policy Background Heterogeneity

Using 2014 as the cutoff year, this study separates the full sample into two periods: 2011–2014 and 2015–2023. The corresponding empirical results are presented in Columns (4) and (5) of Table 6. The impact coefficients of URID on EEQ in both stages pass the 1% significance test, but the coefficient in 2011–2014 (0.3526) is slightly higher than that in 2015–2023 (0.2469). During the earlier stage, URID was promoted in core regions with targeted ecological governance, resulting in a more pronounced effect in targeted areas. After 2015, the broader scope of policy implementation introduced short-term adjustment costs in some regions, slightly attenuating the coefficient while maintaining high significance.

4.4.3. Population Scale Heterogeneity

This study categorizes 284 prefecture-level cities into two major regions according to the Hu Line, with the corresponding outcomes presented in Columns (6) and (7) of Table 6. On the southeast side, the effect coefficient of URID on EEQ is significantly positive (0.1636, p < 0.01), whereas the positive influence on the northwest side does not achieve statistical significance. In areas with a large population, close urban–rural spatial connection, and frequent factor mobility, the improvement of EEQ driven by URID is more evident. By contrast, this facilitating role has not been effectively realized in regions with sparse population and distinctive ecological conditions.

Taken together, these heterogeneity results confirm the observation raised in the Introduction that the ecological effect of URID is contingent upon the quality and directionality of factor flows and institutional arrangements, providing nuanced evidence that complements the average effect reported in the benchmark regression.

4.5. Mechanism Analysis

To further explore the intrinsic transmission mechanism through which URID influences EEQ, this study adopts resource allocation efficiency and environmental regulation intensity as mediating variables and establishes a mediating effect model for empirical examination. Taking into account potential endogeneity concerns within the model, this paper focuses on verifying the influence channels of URID on the mediating variables, with the corresponding regression outcomes reported in

Table 7. Mediating effect regression results.

Variables	RAE	ERI
	(1)	(2)
URID	−0.8424 *** (0.1639)	0.8995 *** (0.2923)
Control Variables	Controlled	Controlled
Constant	0.4603 *** (0.0183)	0.8652 *** (0.0326)
Time Fixed Effects	Yes	Yes
City Fixed Effects	Yes	Yes
N	3692	3692
R-squared	0.155	0.368

Notes: *** indicates statistical significance at the 1% level.

.

Table 7, Column (1) presents the regression outcomes of URID’s impact on resource allocation efficiency. The estimated coefficient of URID is −0.8424, which is statistically significant at the 1% level. This result suggests that URID can notably mitigate the extent of resource misallocation, thereby enhancing the efficiency of resource allocation. By breaking barriers to urban–rural factor flow, URID promotes the reallocation of capital, labor, and technology from low-efficiency to high-efficiency areas, reducing resource waste and inefficient pollution emissions. This result empirically confirms the mediating pathway via resource allocation efficiency illustrated in Figure 1 (H2): URID reduces factor misallocation, which in turn alleviates inefficient and resource-wasteful production and ecological degradation. Thus, Hypothesis 2 is positively verified.

Column (2) of Table 7 displays the estimation results regarding the influence of URID on environmental regulation intensity. The coefficient of URID is 0.8995, which is significantly positive at the 1% level, suggesting that URID can effectively strengthen the level of regional environmental regulation. URID raises rural residents’ environmental awareness through public service equalization (demand-side pressure) and prompts unified environmental standards through industrial coordination (supply-side empowerment), jointly elevating regulation intensity. This result confirms the mediating pathway via environmental regulation intensity illustrated in Figure 1 (H3): URID strengthens governance capacity, which restricts polluting behaviors and promotes collaborative governance. In accordance with the suggestions of Jiang [45], the mediating effect of environmental regulation intensity is verified, thus positively confirming Hypothesis 3.

4.6. Spatial Spillover Effects

4.6.1. Spatial Correlation Test

The global Moran’s I values of both URID and EEQ are consistently positive and significant at the 1% level throughout the study period (detailed results in

Table A1. Global Moran’s I of URID and EEQ.

Year	URID Moran’s I	p-Value	EEQ Moran’s I	p-Value
2011	0.1354	0.0000	0.1845	0.0000
2012	0.1440	0.0000	0.1757	0.0000
2013	0.1436	0.0000	0.1928	0.0000
2014	0.1353	0.0000	0.1910	0.0000
2015	0.1457	0.0000	0.1959	0.0000
2016	0.1455	0.0000	0.2052	0.0000
2017	0.1443	0.0000	0.2081	0.0000
2018	0.1468	0.0000	0.2132	0.0000
2019	0.1445	0.0000	0.2114	0.0000
2020	0.1485	0.0000	0.1900	0.0000
2021	0.1547	0.0000	0.2044	0.0000
2022	0.1565	0.0000	0.2030	0.0000
2023	0.1549	0.0000	0.2146	0.0000

of Appendix A), confirming significant positive spatial autocorrelation. The Moran’s I of EEQ (ranging from 0.1757 to 0.2146) is generally higher than that of URID (0.1353 to 0.1565), indicating stronger spatial dependence in ecological outcomes. These findings provide a reasonable basis for adopting a spatial econometric model to further explore the spatial spillover effects of URID on EEQ.

4.6.2. Regression Results of the SDM

In accordance with the sequential test procedure described in Section 3.2.3, the LM statistics verify the existence of both spatial lag and spatial error components. The LR and Wald tests reject the simplification of the SDM into the SAR or SEM model, while the LR test supports the inclusion of two-way fixed effects (detailed statistics are shown in

Table A2. Relevant tests for the spatial effect model.

Suitability Test	Test Name and Null Hypothesis	Statistic	p-Value
Model Selection	LM-Error	752.881	0.000
	LM-Error-Robust	313.114	0.000
	LM-Lag	628.749	0.000
	LM-Lag-Robust	188.981	0.000
Hausman Test	H0: Random effects model is superior	−107.39
Model Robustness (LR Test)	H0: Individual fixed effects are better	77.23	0.000
	H0: Time fixed effects are better	658.44	0.000
	H0: SDM can be simplified to SAR	212.70	0.000
	H0: SDM can be simplified to SEM	242.42	0.000
Model Robustness (Wald Test)	H0: SDM can be simplified to SAR	218.50	0.000
	H0: SDM can be simplified to SEM	249.75	0.000

in Appendix A). Therefore, the two-way fixed-effects SDM is employed in the analysis.

The results in Column (1) of

Table 8. Regression results of the Spatial Durbin Model.

SDM Coefficients	EEQ	Direct Effect	Indirect Effect	Total Effect
	(1)	(2)	(3)	(4)
URID	0.0585 *** (0.0224)	0.0896 ** (0.0400)	0.0680 *** (0.0221)	0.0590 *** (0.0213)
EG	−0.0070 (0.0077)	−0.0128 (0.0078)	−0.0245 * (0.0140)	0.0519 *** (0.0111)
IS	0.0050 *** (0.0011)	0.0082 *** (0.0016)	0.0871 ** (0.0374)	0.0880 ** (0.0375)
TP	−0.0166 *** (0.0062)	0.0843 ** (0.0360)	−0.0761 ** (0.0323)	−0.0676 ** (0.0311)
GI	0.0086 (0.0057)	−0.0075 (0.0056)	−0.0547 ** (0.0269)	−0.0623 ** (0.0292)
OPEN	0.1457 *** (0.0453)	0.0661 *** (0.0229)	−0.1529 (0.0976)	−0.0867 (0.0960)
INF	0.0614 *** (0.0092)	−0.0031 (0.0031)	0.0323 *** (0.0102)	0.0292 *** (0.0106)
ρ	2.6797 *** (0.0305)
sigma2_e	0.0001
N	3692	3692	3692	3692

Notes: *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

show that the coefficient ρ of the spatial lag term of EEQ is 2.6797, which is significant at the 1% level, further verifying the spatial autocorrelation and confirming the rationality of the spatial econometric model specification. On this basis, this paper decomposes the spatial effects into direct effects, indirect effects, and total effects by means of the partial differential method. The results in Columns (2)–(4) of Table 8 are as follows: The direct effect coefficient of URID on EEQ is 0.0896 (p < 0.05), indicating that local URID can significantly improve its own EEQ, which echoes the benchmark regression and further verifies Hypothesis 1. The indirect effect coefficient of URID is 0.0680 (p < 0.01), confirming that local URID exerts a significant positive spillover effect on the EEQ of neighboring regions. This indirect effect corresponds to the spatial spillover pathway in Figure 1 (H4), confirming that technology diffusion, factor flow, and policy coordination across administrative boundaries generate positive ecological externalities. Thus, Hypothesis 4 is positively verified. The total effect coefficient is 0.0590 (p < 0.01), confirming that the positive impact of URID on EEQ is holistic and stable at the level of “local + adjacent regions”.

For control variables, the direct, indirect, and total effects of industrial structure and infrastructure are mostly significantly positive, reflecting the technology diffusion effect of green industrial upgrading and the factor flow empowerment effect of infrastructure interconnection. The direct effect of technological progress is significantly positive (0.0843, p < 0.05), while its indirect effect is significantly negative (−0.0761, p < 0.05), indicating the coexistence of local technological agglomeration and the siphon effect on adjacent regions. The indirect effect of government intervention is significantly negative (−0.0547, p < 0.05), implying potential policy fragmentation or regional protectionism in local environmental governance.

4.6.3. Robustness Test of Spatial Spillover Effects

The robustness of the SDM results is further confirmed using an alternative spatial weight matrix (0–1 adjacency matrix) and an alternative dependent variable (urban ecological resilience), with the coefficients of URID remaining significantly positive in all specifications (detailed results in

Table A3. Robustness test results of the SDM.

Variables	Model (1)			Model (2)
	Direct Effect	Indirect Effect	Total Effect	Direct Effect	Indirect Effect	Total Effect
URID	0.0744 ** (0.0311)	0.0614 ** (0.0289)	0.0852 *** (0.0078)	0.1508 *** (0.0340)	0.0421 *** (0.0690)	0.0071 *** (0.0014)
ρ	0.8025 *** (0.0096)			0.9393 *** (0.0162)
sigma2_e	0.0000			0.0001
N	3692			3692

Notes: ** and *** indicate statistical significance at the 5% and 1% levels, respectively. Model (1) replaces the geographical distance weight matrix with a 0–1 adjacency matrix. A small number of geographically isolated cities (e.g., island-type cities such as Haikou and Zhoushan) lack physical contiguity with other sample cities under this matrix; they are retained and assigned to their nearest functionally connected prefectures, as they maintain substantive economic and policy linkages through maritime transportation and cross-jurisdictional governance despite the absence of land borders. This treatment does not affect the baseline SDM, which uses a continuous distance weight matrix. Model (2) replaces the dependent variable with urban ecological resilience.

of Appendix A).

5. Discussion

5.1. Hypothesis Validation and Theoretical Implications

This section consolidates the empirical evidence for each hypothesis, interprets the findings within the theoretical framework (Figure 1), and notes the caveats inherent in the observational research design.

H1: Direct promoting effect of URID on EEQ: The two-way fixed-effects model confirms a significant positive effect of URID on EEQ (β = 0.1889, p < 0.01; Table 3, Column 2), which remains robust across instrumental variable estimation (Table 4), alternative proxy variables, shortened time windows, and exclusion of key cities (Table 5). The SDM further corroborates this through a significant direct effect (0.0896, p < 0.05; Table 8, Column 2). To provide an interpretable sense of magnitude, a one-standard-deviation increase in URID (0.0280) is associated with an increase in EEQ of approximately 0.0053 (0.0280 × 0.1889), which corresponds to roughly 5.9% of the standard deviation of EEQ (0.0906). While statistically significant, this modest effect size is consistent with the nature of URID as a gradual institutional process whose ecological benefits accumulate over time. It should be noted that, despite the instrumental variable approach, the observational design cannot fully eliminate the possibility of omitted variable bias, and the estimated coefficient should be interpreted as an association rather than a definitive causal effect.

The heterogeneity analysis reveals important boundary conditions. The promoting effect is significant in the eastern (0.2958, p < 0.01) and central (0.3608, p < 0.05) regions but insignificant in the western region (Table 6, Columns 1–3), and significant on the southeast side of the Hu Line (0.1636, p < 0.01) but insignificant on the northwest side (Table 6, Columns 6–7). These patterns confirm that the ecological effect of URID is contingent upon factor flow intensity and institutional readiness, as theorized in the Introduction. The policy stage heterogeneity—a higher coefficient in 2011–2014 (0.3526) than in 2015–2023 (0.2469)—suggests that broader policy diffusion may introduce short-term adjustment costs that temporarily attenuate effectiveness.

H2–H3: Dual mediating pathways: URID significantly reduces resource misallocation (β = −0.8424, p < 0.01; Table 7, Column 1) and strengthens environmental regulation intensity (β = 0.8995, p < 0.01; Table 7, Column 2), confirming both mediating pathways illustrated in Figure 1. The two mediators operate through distinct channels: resource allocation efficiency works primarily through factor reallocation that reduces inefficient and resource-wasteful production and associated pollution, while environmental regulation intensity operates through governance capacity enhancement that shifts regulation from end-of-pipe treatment to source prevention. It is important to note that this study adopts the approach recommended by Jiang [45], verifying the URID-to-mediator relationship rather than estimating the full mediation chain, which means the relative quantitative weight of each mediating pathway cannot be precisely determined. Future research employing structural equation modeling could further decompose the relative contributions.

H4: Spatial spillover effects: The SDM confirms significant positive spatial spillovers, with an indirect effect coefficient of 0.0680 (p < 0.01; Table 8, Column 3). To provide an interpretable sense of magnitude, a one-standard-deviation increase in neighboring cities’ URID (0.0280) is associated with an EEQ improvement in the focal city of approximately 0.0019 (0.0280 × 0.0680), or about 2.1% of the EEQ standard deviation—modest but non-negligible given the cross-boundary nature of this effect. This finding supports the theoretical prediction that technology diffusion, cross-boundary factor flow, and policy coordination generate positive ecological externalities across administrative boundaries. The robustness of the spatial results is confirmed using an alternative weight matrix (0–1 adjacency matrix) and an alternative dependent variable (urban ecological resilience), with URID coefficients remaining significantly positive in all specifications (Table A3). However, the spatial spillover results are dependent on the specification of the weight matrix, and the theoretical analysis in Section 2.3 acknowledges that negative spillovers—such as the siphon effect or pollution haven phenomenon—may coexist with positive spillovers in specific regional contexts. The consistently positive indirect effect observed here suggests that, at the national scale, positive spillover mechanisms dominate, but this aggregate finding does not preclude localized negative spillovers that warrant further investigation with finer spatial resolution.

5.2. Comparison with Existing Literature

The significant positive effect of URID on EEQ aligns with the core findings of Dong et al. [12] and Zhou et al. [16], who documented that urban–rural factor flow and collaborative ecological governance jointly improve environmental outcomes. However, while these prior studies relied on coupling coordination models to describe co-evolution patterns, the present study advances the analysis by establishing a regression-based causal framework that identifies specific transmission mechanisms. This methodological progression explains why the present findings are directionally consistent with prior work while offering substantially deeper mechanistic insight.

The regional heterogeneity findings are consistent with Yang et al. [18], who observed that differences in factor endowments lead to differentiated ecological effects. The present study extends this observation by identifying the Hu Line as a critical spatial boundary—a dimension not examined in prior research—and by demonstrating that factor interaction intensity, rather than factor endowment alone, is the key determinant of effect heterogeneity. The insignificant effect in the western region and northwest of the Hu Line diverges from the universally positive narrative implied by some coupling studies [5,24], likely because regression-based approaches with fixed effects impose stricter identification requirements than coupling coordination models, filtering out spurious associations driven by common trends.

The dual mediation finding represents a departure from the single-mechanism focus prevalent in the existing literature. Prior studies have separately examined resource allocation [27] or environmental regulation [6] in isolation, without testing their simultaneous mediating roles. By identifying both pathways within a unified framework, this study demonstrates that the URID–EEQ relationship operates through complementary rather than competing channels—a finding that has direct implications for policy design, as interventions targeting only one pathway would capture incomplete benefits.

The spatial spillover findings extend beyond the spatial autocorrelation documented by Zhou et al. [42] by decomposing total effects into direct and indirect components. The significant indirect effect (0.0680, p < 0.01) highlights the substantial role of inter-city ecological externalities that prior studies, focused on within-city effects, could not quantify.

5.3. Research Limitations and Future Directions

When interpreting the empirical results obtained in this research, certain limitations need to be recognized. First, regarding research design, this study is observational in nature. Although the instrumental variable approach (topographic relief and URID pilot zone policy shocks) addresses endogeneity concerns to some extent, it cannot fully eliminate the possibility of omitted variable bias or reverse causality. The estimated effects should therefore be interpreted as robust associations rather than definitive causal effects.

Second, regarding the URID index construction, the composite index comprises 22 indicators weighted by the entropy-TOPSIS method. While this approach has the advantage of objectivity (weights are data-driven rather than subjectively assigned), the results are sensitive to indicator selection and weighting decisions. The robustness check using the Theil index as an alternative proxy (Table 5, Column 1) partially mitigates this concern, but future research could strengthen the findings through cross-validation with alternative index construction methods (e.g., principal component analysis or equal weighting) to assess sensitivity.

Third, regarding spatial model specification, the SDM results are dependent on the choice of the spatial weight matrix. Although robustness is confirmed with an alternative adjacency matrix (Table A3), the analysis does not explore alternative distance thresholds or economic-based weight matrices, which could yield different spatial interaction patterns. Future research should conduct more comprehensive spatial robustness tests with multiple matrix specifications and distance radii.

Fourth, the assessment of URID and EEQ is centered on macro-level statistical data, lacking micro-level subjective perception indicators (such as residents’ environmental satisfaction), which limits the comprehensiveness of the evaluation.

Fifth, despite verifying the existence of dual mediating effects, the present research does not explore the moderating functions of factors including the digital economy and social capital. Furthermore, the interactive transmission dynamics among different mediating variables are not examined in this study.

Sixth, focusing on the city-level analysis, this study does not refine to the urban–rural micro scale, and the lack of an international comparative perspective hinders the extraction of universally applicable laws.

In view of the above deficiencies, future studies can be further extended in the following aspects: integrating macro statistical data with micro survey data, along with subjective perception indicators; adopting threshold regression and moderating effect models to explore boundary conditions; conducting cross-validation of the URID index with alternative construction methods; expanding spatial robustness tests with multiple weight matrices and distance radii; and extending the analysis to the urban–rural micro scale and international comparative studies.

6. Conclusions and Policy Implications

6.1. Conclusions

(1)

URID significantly promotes EEQ (β = 0.1889, p < 0.01; Table 3). A one-standard-deviation increase in URID is associated with an EEQ improvement of approximately 5.9% of its standard deviation. This effect is robust to instrumental variable estimation, alternative proxies, and sample adjustments (Table 4 and Table 5).

(2)

The promoting effect exhibits significant heterogeneity: it is stronger in the central and eastern regions than in the western region, more pronounced during the early policy stage (2011–2014) than the later stage (2015–2023), and significant only on the southeast side of the Hu Line (Table 6). These patterns underscore the contingent role of factor flow intensity and institutional readiness.

(3)

URID indirectly improves EEQ through dual mediating pathways, reducing resource misallocation (β = −0.8424, p < 0.01) and strengthening environmental regulation intensity (β = 0.8995, p < 0.01) (Table 7).

(4)

URID generates significant positive spatial spillovers on neighboring regions’ EEQ (indirect effect = 0.0680, p < 0.01; Table 8), operating through technology diffusion, cross-boundary factor flow, and policy coordination. This conclusion is robust across alternative spatial weight matrices and dependent variables (Table A3).

6.2. Policy Implications

The findings carry several broad implications for policy-making aligned with the Sustainable Development Goals, though it should be noted that the specific policy designs would require further investigation beyond the scope of this study.

First, the significant regional heterogeneity suggests that uniform national policies may not be equally effective across all regions. The promoting effect of URID on EEQ is contingent upon factor flow intensity and institutional readiness, implying that policy-makers should adopt differentiated approaches that account for regional differences in economic foundations, ecological conditions, and integration maturity. Regions where the effect is insignificant (e.g., western China and northwest of the Hu Line) may require foundational investments before the ecological benefits of URID can materialize.

Second, the identification of resource allocation efficiency and environmental regulation intensity as dual mediating pathways suggests that policies targeting only one dimension may yield incomplete ecological benefits. While this study confirms that both pathways matter, the specific institutional reforms needed to strengthen each pathway warrant further empirical investigation.

Third, the significant positive spatial spillover effects indicate that the ecological impact of URID extends beyond administrative boundaries. This finding highlights the importance of intermunicipal coordination in ecological governance, as purely localized approaches may underestimate the broader regional benefits of URID policies.

Fourth, the higher coefficient during the early policy stage (2011–2014) compared to the later stage (2015–2023) suggests that the timing and sequencing of policy implementation may influence the magnitude of ecological effects, and that attention to implementation rhythm could help sustain policy effectiveness during broader diffusion. Nevertheless, the specific mechanisms behind this temporal pattern remain an unresolved issue that requires further investigation in future studies.