Promoting or Inhibiting? The Nonlinear Impact of Urban–Rural Integration on Carbon Emission Efficiency: Evidence from 283 Chinese Cities

Abstract

In the context of global climate governance and China’s ‘Dual Carbon’ strategy, enhancing carbon emission efficiency (CEE) is a critical pathway toward high-quality development. Urban–rural integration (URI), reshaping urban–rural structures and resource allocation, has significant environmental implications. However, the mechanisms through which URI influences city-level CEE remain underexplored. Using panel data from 283 Chinese prefecture-level cities (2005–2022), we employ a Spatial Durbin Model to investigate URI’s direct and spatial spillover effects. First, spatiotemporally, URI demonstrates an imbalanced pattern, with higher levels in eastern coastal regions and lower levels in central and western areas. Conversely, CEE exhibits a north–south divide, with higher efficiency in the south. URI advancement has been sluggish with persisting imbalances, whereas CEE has demonstrated a consistent upward trend. Second, the relationship between URI and CEE is characterized by nonlinearity and spatial dependence. The direct effect follows a U-shaped curve, initially inhibiting but later promoting local CEE once a threshold is surpassed (URI = 0.103). The spatial spillover effect follows an inverted U-shaped trajectory (threshold URI = 0.179), suggesting that inter-regional dynamics evolve from synergistic promotion to potential competition. These findings underscore the necessity of phased, adaptive policies to unlock the potential between URI and CEE, providing a scientific basis for coordinating urban–rural development with carbon neutrality objectives.

1. Introduction

Confronting global climate change constitutes a central pillar of the 21st century’s sustainable development agenda [1]. Consequently, reducing anthropogenic greenhouse gas (GHG) emissions, particularly CO₂, has become a key metric for evaluating national and regional sustainability transitions [2]. Enhancing carbon emission efficiency (CEE)—maximizing economic output per unit of carbon emissions—is therefore a crucial indicator of the transition towards green, low-carbon, and high-quality development, extending beyond a simple measure of environmental performance [3,4]. As the world’s largest carbon emitter and a pivotal actor in global climate governance, China has announced its ambitious ‘dual carbon’ goals: peaking their carbon emissions before 2030 and achieving carbon neutrality before 2060 [5,6]. Achieving this goal requires transcending a singular focus on emissions control and instead exploring new pathways that decouple economic growth from carbon output, particularly amid intensifying resource and environmental constraints. In this context, enhancing CEE is an urgent priority, and its implementation pathway is deeply embedded in China’s unprecedented urban–rural structural transformation, forming an intrinsic link between the two.

In developing nations, a persistent urban–rural imbalance has triggered unidirectional factor flows, inequitable public resource allocation, and disparate development opportunities, thereby continuously widening the multidimensional urban–rural gap [7,8,9,10]. This structural imbalance has transcended the socioeconomic sphere and now poses a critical impediment to the green and low-carbon transition. Indeed, a growing body of evidence indicates that urban–rural gaps in income, technology, and industrial structure significantly impede urban carbon mitigation efforts and can intensify environmental degradation [11,12,13]. For instance, Xu et al. (2023) [14] find that urban carbon efficiency generally surpasses that of rural areas, a gap widened by the spatially imbalanced distribution of technology and innovation. Complementing this, Wang and Zhang (2021) [15] demonstrated that narrowing the urban–rural divide promotes both carbon abatement and pollution control. Therefore, overcoming the impasse of urban–rural development imbalance and fostering systematic urban–rural coordination is not merely essential for social equity and coordinated regional development but also provides a critical breakthrough for advancing carbon neutrality goals in developing nations.

Systemic policy adjustments to China’s urban–rural relations, initiated in 2005, have driven a profound evolution in the urban–rural integration policy agenda: from the initial phases of the ‘New Socialist Countryside’ and “urban supporting the rural” [16] to the formal establishment of ‘Urban-Rural Integrated Development’ as a core national strategy in 2017. This shift marked a transition from superficial coordination to a concerted effort to dismantle the entrenched urban–rural dual system [17]. Theoretically, China’s URI approach is rooted in the ‘Human-Land Relationship System’ theory, conceptualizing urban and rural areas as a single, symbiotic organism [18,19]. Its core mechanism involves promoting the bidirectional flow and efficient allocation of key factors to achieve economic interconnectivity, social cohesion, spatial integration, and ecological harmony. This process enhances region-wide resource efficiency and green productivity [20,21], thereby contributing to carbon mitigation objectives. Preliminary evidence suggests that URI substantially influences key drivers of CEE, including population mobility, energy consumption patterns, public service provision, industrial upgrading, and land-use efficiency [22,23,24,25,26]. However, while the socioeconomic value of URI is well-documented, its environmental ramifications—particularly its systemic impact on city-level CEE—remain significantly underexamined. Existing research has primarily examined the effects of URI within discrete sectors, including rural household energy consumption (Peng et al., 2025), agricultural carbon efficiency (Xie and Wu, 2023), and land-use-related emissions (Li et al., 2025) [27,28,29]. Specifically, research on rural household energy consumption has revealed a complex “inverted N-shaped” relationship between URI and carbon emissions, indicating that emissions fluctuate alongside the integration process [27]; meanwhile, investigations into agricultural carbon efficiency have identified a clear threshold effect, where the environmental benefits of URI only manifest after regional economic development surpasses a specific tipping point [28]. Consequently, although these studies have gradually elucidated the specific impact mechanisms of URI on carbon emissions within particular rural production and consumption sectors and have provided important references for understanding overall city-level carbon emission efficiency, the systemic mechanism through which URI—as a multidimensional transformation—influences holistic urban CEE remains elusive, necessitating in-depth research.

Moreover, a rigorous examination of the interplay between urban–rural integration (URI) and carbon emission efficiency necessitates the integration of inherent spatial attributes [30]. Guided by Tobler’s First Law of Geography—which posits that “everything is related to everything else, but near things are more related than distant things” [31]—extant research demonstrates that both URI and carbon emission efficiency exhibit significant spatial dependence [32,33]. Their geographical distributions typically manifest as distinct clusters rather than stochastic patterns. Neglecting such spatial autocorrelation would induce model specification bias, thereby undermining the reliability of resulting policy implications [34]. Consequently, this study adopts a spatial econometric framework to quantify inter-city spillover effects. Among the alternative specifications—namely, the Spatial Autoregressive Model (SAR), Spatial Error Model (SEM), and Spatial Durbin Model (SDM)—the SDM is selected as the optimal framework. By incorporating spatial lags of both dependent and independent variables, the SDM provides a more sophisticated characterization of spatial externalities, rendering it uniquely capable of capturing the complexities inherent in this inquiry [35].

Building on these arguments, this study develops an integrated analytical framework (Figure 1). Using a Spatial Durbin Model (SDM) and panel data from 283 Chinese prefecture-level cities (2005–2022), we systematically investigate the impact of URI on CEE, examining its direct local effects, spatial spillover effects, and underlying mechanisms. The core research questions are as follows: (1) does URI significantly affect city-level CEE? (2) Does URI affect the CEE of neighboring cities through spatial spillovers? Theoretically, this study contributes to the literature in two ways. First, it deepens the understanding of the nexus between urban–rural integration and carbon emission efficiency. Second, it enriches the cross-disciplinary research framework by incorporating a spatial spillover perspective. Practically, our findings offer spatially explicit evidence to inform more targeted URI policymaking. This can foster synergy between high-quality development and China’s ‘dual carbon’ targets, realize potentials for energy conservation and emission reduction, and mitigate environmental risks.

2. Theoretical Framework and Hypotheses Development

2.1. Local Impact Mechanisms of URI on Carbon Emission Efficiency

The Environmental Kuznets Curve (EKC) framework posits that the environmental impact of economic structural transformation is fundamentally driven by the triad of scale, composition, and technique effects [36]. This paradigm has been extended to urban–rural studies to elucidate the nonlinear trajectories through which urban–rural interactions influence carbon performance [37,38]. Since energy intensity and resource allocation efficiency vary across developmental stages, the impact of urban–rural integration (URI) on carbon emission efficiency (CEE) is expected to exhibit a nonlinear trajectory. Drawing upon these theoretical foundations, this study hypothesizes that the URI-CEE nexus follows a U-shaped trajectory, transitioning from a phase dominated by “scale expansion costs” to one defined by “structural and technical dividends.”

Specifically, in the nascent stages of URI, dismantling entrenched urban–rural dualities incurs significant institutional and spatial restructuring costs, primarily through government-led infrastructure connectivity and peripheral sprawl [39]. As Seto et al. (2016) argue, capital-intensive investments in high-carbon physical infrastructure risk triggering pervasive “carbon lock-in” effects [40]. Consequently, the negative externalities arising from the rapid expansion of factor flows may temporarily suppress CEE gains [41]. However, as URI matures, the urban–rural relationship evolves from physical connectivity toward deep functional coupling. Optimized factor networks accelerate the diffusion of green technologies and low-carbon management paradigms from urban cores to rural hinterlands [42]. Concurrently, the synergistic development of urban–rural industries catalyzes a transition from carbon-intensive production to integrated, low-carbon intensification [43]. In this advanced phase, the convergence of technical dividends and structural optimization drives sustained CEE enhancement. Accordingly, we propose the following hypothesis:

H1: 

There exists a U-shaped nonlinear relationship between URI and CEE, whereby URI initially inhibits and subsequently promotes carbon emission efficiency.

2.2. Spatial Spillover Effects of URI on Carbon Emission Efficiency

Theoretical and empirical literature consistently underscore the pronounced spatial dependence inherent in both URI and CEE [32,33], necessitating a spatial perspective to examine their nexus. The essence of URI lies in the dynamic reallocation of labor, capital, and technology across urban and rural spaces—a process that transcends administrative boundaries and fosters connectivity between localized development and neighboring jurisdictions [44]. Simultaneously, CEE is shaped by regional environmental governance and technological spillovers, both of which exhibit significant spatial externalities [45]. Thus, the spatial attributes of both variables demand an investigation into inter-regional interactions and spillover mechanisms.

Mechanistically, the spatial impact of URI on CEE operates through two divergent pathways: “positive spillovers” and “negative siphoning.” URI facilitates the proliferation of green technologies, the relocation of low-carbon industries, and the diffusion of innovative management models [44,46]. Through geographic proximity, these processes generate positive spatial spillovers that bolster the CEE of adjacent regions. For instance, Li et al. (2025) utilized a Spatial Durbin Model (SDM) to demonstrate that URI, through the facilitation of cross-regional factor mobility, significantly mitigates land-use carbon intensity in neighboring cities [29]. Conversely, according to the ‘Growth Pole’ theory, regions with advanced URI may exert a “siphoning effect” on green resources—such as environmental investments and high-caliber talent—from their neighbors [47], thereby impeding the latter’s CEE improvements. Such competitive dynamics are well-documented in studies on regional energy efficiency and environmental regulation [48]. This suggests that a region’s URI level may not only dictate its internal CEE but could also exert heterogeneous spatial spillover effects on neighboring peers. Based on these considerations, we propose the following hypothesis:

H2: 

Urban–rural integration exerts significant spatial spillover effects on carbon emission efficiency.

3. Materials and Methods

3.1. Data Sources

This study utilizes a panel dataset for 283 Chinese prefecture-level cities from 2005 to 2022. Data for the dependent variable, carbon emission efficiency (CEE), the explanatory variable, urban–rural integration (URI), and a suite of control variables were primarily sourced from the China Statistical Yearbook, the China Urban Statistical Yearbook, data officially released by the National Bureau of Statistics, various local statistical yearbooks, and the EPS database. Notably, CO₂ emissions data required to calculate CEE were obtained from the Emissions Database for Global Atmospheric Research (EDGAR). Any missing values in the panel data were imputed using linear interpolation and regression-based methods. The administrative boundary basemap for China was obtained from the National Platform for Common Geospatial Information Services.

While most variables were directly available from the sources above, two indicators for the ecological dimension of URI required specialized calculation. First, the landscape connectivity index was calculated in Fragstats 4.2 using land-use/land-cover (LULC) data (30 m resolution) from the Resource and Environment Science and Data Center, Chinese Academy of Sciences (RESDC), based on the methodology proposed by Nazar Neghad et al. (2020) [49]. Second, leveraging the same LULC dataset, the total value of ecosystem services was estimated in ArcGIS 10.8 based on the equivalent factor method proposed by Li et al. [50].

3.1.1. Explanatory Variable: Urban–Rural Integration (URI) Index

The core explanatory variable in this study is the urban–rural integration (URI) index. To measure this, we constructed a comprehensive evaluation index system by synthesizing and refining insights from existing measurement studies [51,52,53], while ensuring close alignment with national strategies such as the 2019 ‘Opinions on Integrated Urban-Rural Development’. The system encompasses four dimensions: economic integration (factor mobility and industrial synergy), social–demographic integration (equalization of public services and social security convergence), spatial integration (infrastructure connectivity and planning coordination), and ecological integration (joint environmental governance and ecological value sharing).

Based on this framework, we selected 18 specific indicators, categorized as comparative indicators of urban–rural disparity, drivers of urban–rural interaction, and comprehensive measures of development level (

Table 1. Indicator system for measuring urban-rural integration.

Sub-Dimension	Serial Number	Indicator	Calculation	Unit	Expected Impact
A. Economic Integration	A1	Urban–Rural Income Gap	Ratio of per capita disposable income of urban residents to that of rural residents	Ratio	−
	A2	Economic Duality Index	Ratio of labor productivity between primary and non-primary (secondary and tertiary) industries	%	+
	A3	Urban–Rural Consumption Gap	Ratio of per capita consumption expenditure of urban residents to that of rural residents	Ratio	−
	A4	Agricultural Modernization Level	Total power of agricultural machinery (kW) per unit of cultivated land area (ha)	kW/ha	+
	A5	Rural–Urban Fixed Asset Investment Ratio	Ratio of fixed asset investment in rural areas to that in urban areas	%	+
B. Social–Demographic Integration	B1	Urbanization Rate	Proportion of urban population in the total population	%	+
	B2	Non-agricultural Employment Rate	Proportion of the labor force in secondary and tertiary industries in the total labor force	%	+
	B3	Overall Population Density	Total population per unit of total administrative area	Capita/ km2	+
	B4	Urban–Rural Infrastructure Disparity	Ratio of per capita expenditure on transport and communication by urban residents to that by rural residents	Ratio	−
	B5	Ratio of Rural to Urban Education Investment	Ratio of total investment in rural basic education to that in urban basic education	%	+
C. Spatial Integration	C1	Cultivated Land Proportion	Proportion of cultivated land in the total administrative area	%	−
	C2	Public Transport Provision	Number of public transport vehicles per 10,000 persons	Vehicles/10 k pop.	+
	C3	Road Network Density	Total length of roads and railways per unit of administrative area	km/km2	+
	C4	Information Accessibility	Proportion of internet users in the total population	%	+
D. Ecological Integration	D1	Ecosystem Service Value to GDP Ratio	Ratio of the monetized value of ecosystem services to the Gross Domestic Product (GDP)	%	+
	D2	Landscape Connectivity	Index of landscape connectivity (e.g., calculated via landscape pattern analysis)	Index	+
	D3	Urban Greening Rate	Proportion of green space coverage within the built-up area	%	+
	D4	Environmental Governance Investment	Proportion of total investment in environmental pollution control in GDP	%	+

Note: + signifies positive indicators; − signifies negative indicators.

). We employed the Entropy Weight-TOPSIS method for objective weighting and to construct the composite index [54,55]. The calculation procedure comprises several steps: standardization of positive and negative indicators, weight determination, construction of a weighted matrix, calculation of positive and negative ideal solutions, calculation of relative distances, and calculation of the TOPSIS evaluation value. This process yields the final URI index for each of the 283 prefecture-level cities from 2005 to 2022. (See Figure 2 for the detailed calculation process.)

3.1.2. Dependent Variable: Carbon Emission Efficiency (CEE) Measurement

Carbon emission efficiency (CEE) is a comprehensive indicator that reflects the equilibrium between economic growth and environmental sustainability. Unlike carbon intensity—a unidimensional metric (CO₂/GDP) that often overlooks the complex interdependencies among production factors—CEE is conceptualized as a total-factor productivity (TFP) measure [56]. It quantifies a system’s ability to maximize economic output while minimizing both carbon emissions and resource consumption, given specific inputs of capital, labor, and energy. By integrating resource allocation efficiency and technological substitution effects, this multidimensional framework provides a more robust assessment of green development than conventional intensity ratios, thereby accurately capturing the intricate socio-economic dynamics underlying emissions [57].

To rigorously quantify CEE, this study employs the Super-Efficiency Epsilon-Based Measure (EBM) model. Proposed by Tone and Tsutsui (2010) [58], the EBM model is a non-parametric tool within the Data Envelopment Analysis (DEA) framework designed for multi-input and multi-output systems [59]. A key advantage of the EBM model lies in its hybrid integration of radial and non-radial distance measures; it facilitates proportional input reductions while simultaneously addressing specific factor slacks, significantly enhancing measurement precision. As an advanced DEA iteration, it evaluates the nexus between energy consumption, economic growth, and carbon emissions. By treating carbon emissions as an “undesirable output,” the model aligns with the core principles of green economic efficiency. By establishing a “production frontier” and calculating the relative proximity of each Decision-Making Unit (DMU) to this benchmark, the EBM model yields a composite CEE score [60]. The inclusion of a super-efficiency mechanism overcomes the limitations of traditional DEA models, enabling the effective differentiation and ranking of multiple efficient DMUs on the frontier [61].

Drawing on relevant research, we constructed a comprehensive evaluation system for the CEE comprising three inputs (capital, labor, and energy), one desirable output (Gross Domestic Product), and one undesirable output (CO₂ emissions) (

Table 2. Input and output indicators for carbon emission efficiency analysis.

Variable	Indicator	Calculation	Unit
Inputs	Capital	Real capital stock estimated via the Perpetual Inventory Method (PIM)	104 CNY
	Labor	Annual average number of employees	Persons
	Energy	Total energy consumption, converted into standard coal equivalent	tce
Desirable Output	GDP	Gross Domestic Product at the city level (at constant prices)	104 CNY
Undesirable Output	CO2 Emissions	The 0.1° × 0.1° gridded data of global carbon emissions provided by the EDGAR database (https://edgar.jrc.ec.europa.eu/; accessed on 3 June 2025) is processed by ArcGIS 10.8.	104 tons

Note: Capital stock is estimated via the Perpetual Inventory Method (PIM) and formulated as ${\mathrm{K}}_{\mathrm{t}}={\mathrm{K}}_{\mathrm{t}-1}(1-\mathsf{\delta })+{\mathrm{I}}_{\mathrm{t}}$. Following the methodology of Lin et al. (2025), the depreciation rate $\left(\mathsf{\delta }\right)$ is specified as $9.6\mathrm{\%}$ [63]. The initial capital stock for the base year $2005\left({\mathrm{K}}_{2005}\right)$ is derived as ${\mathrm{I}}_{2005}/(\mathrm{g}+\mathsf{\delta })$, where $\mathrm{g}$ denotes the geometric mean growth rate of fixed asset investment during the 2000–2005 period. To ensure temporal comparability and adjust for inflationary effects, nominal capital and GDP data are deflated to 2005’s constant prices using the fixed asset investment price index and the GDP deflator, respectively.

) [56,62]. The core computational procedure involves three steps. First, a production frontier is constructed using linear programming based on the input and output data of all DMUs, establishing a production possibility boundary that represents optimal performance. Second, standard efficiency scores are calculated. For each DMU, the model determines a composite efficiency score (γ∗) that simultaneously accounts for both radial adjustments and non-radial slacks. Third, super-efficiency scores are determined for all efficient units. This is achieved by re-evaluating each efficient DMU against a frontier constructed from all other units. This procedure allows its efficiency score (γsuper) to exceed 1, enabling a complete and differentiated ranking of all units. (See Figure 3 for the detailed calculation process.)

3.1.3. Control Variables

To mitigate potential omitted variable bias and thereby more accurately isolate the net effect of urban–rural integration (URI) on carbon emission efficiency (CEE), we include a set of control variables selected based on the extant literature [64,65]. These variables are energy consumption intensity, level of industrialization, land urbanization, technological innovation, and economic development level. The measurement methods and descriptive statistics for all control variables are presented in

Table 3. Descriptive statistics of control variables.

Variable	Calculation	Unit	Obs.	Max.	Min.	Mean.	Std. Deviation
ECI	Energy consumption per unit of GDP	tce/104 CNY	5094	8.98	0.06	1.01	0.71
ILV	Share of secondary industry value added in GDP	%	5094	88.73	1.04	39.92	12.42
LUL	Ratio of built-up area to total administrative area	%	5094	55	0.21	5.5	6.47
TIL	Per capita expenditure on Research and Development (R&D)	102 CNY/capita	5094	258.94	0.23	36.97	73.99
EDL	Gross Domestic Product (GDP) per capita	104 CNY/capita	5094	24.77	2.4	8.59	3.48

Note: ECI = energy consumption intensity; ILV = level of industrialization; LUL = Level of Land Urbanization; TIL = Level of Technological Innovation; and EDL = Level of Economic Development.

.

3.2. Methods

3.2.1. Spatial Autocorrelation Analysis

Spatial econometric modeling first requires testing for spatial autocorrelation. Accordingly, this study assesses the degree of spatial dependence using the Global Moran’s I statistic. The presence of spatial dependence violates the core assumption of observation independence in classical regression models [66]; ignoring this property can lead to correlated error terms and, consequently, biased and inconsistent estimations. To assess the degree of global spatial clustering for the dependent variable (carbon emission efficiency) [67], this study employs the Global Moran’s I, defined as follows:

$$Moran\prime s I={\displaystyle \frac{\sum _{i=1}^n\sum _{j=1}^n{w}_{ij}(x_i-\overline{x})(x_j-\overline{x})}{S^2\sum _{i=1}^n\sum _{j=1}^n{w}_{ij}}}, where S^2={\displaystyle \frac{1}{n}}\sum _{i=1}^n(x_i-\overline{x}{)}^2$$

(1)

where $n$ represents the number of cities, ${\mathrm{x}}_{\mathrm{i}}$ and ${\mathrm{x}}_{\mathrm{j}}$ denote the observed CEE values for cities $\mathrm{i}$ and $\mathrm{j}$, $\overline{\mathrm{x}}$ is the sample mean, and ${\mathrm{w}}_{\mathrm{i}\mathrm{j}}$ signifies an element of the spatial weight matrix $\mathrm{W}$. To accurately capture the spatial dependencies catalyzed by economic convergence, this research adopts an economic distance spatial weight matrix (${\mathrm{W}}^{\mathrm{e}\mathrm{c}\mathrm{o}}$) [68]. While geographic proximity is a traditional cornerstone of spatial analysis, ${\mathrm{W}}^{\mathrm{e}\mathrm{c}\mathrm{o}}$ is prioritized here to better reflect the intrinsic spatial attributes of CEE. The rationale is that Carbon Emission Efficiency (CEE) is not merely a function of geospatial attributes; rather, it is intrinsically contingent upon the regional industrial architecture and its specific developmental trajectory that ${\mathrm{W}}^{\mathrm{eco} }$ is uniquely equipped to operationalize [56,69]. Within the context of intensifying urban–rural integration, the mobility of production factors—including green innovation, high-quality labor, and capital—frequently transcends physical boundaries, gravitating instead toward patterns of economic complementarity [44]. Consequently, ${\mathrm{W}}^{\mathrm{e}\mathrm{c}\mathrm{o}}$ provides a robust analytical framework for uncovering the spatial spillovers and factor-flow interdependencies essential to understanding CEE configurations. The elements of the economic distance matrix are formulated as follows:

$$W_{ij}^{eco}=\{\begin{array}{ll}{\displaystyle \frac{1}{{\left(\overline{(\overline{gdp}{p}_i-\overline{pgdp}{p}_j}\right)}^2}},& i\ne j\\ 0,& i=j\end{array}$$

(2)

Specifically, each element ${\mathrm{W}}_{\mathrm{i}\mathrm{j}}^{\mathrm{e}\mathrm{c}\mathrm{o}}$ is defined as the reciprocal of the squared difference in mean per capita GDP between city $\mathrm{i}$ and city $\mathrm{j}$. Furthermore, to delineate the local spatial distribution and identify “hot spots” and “cold spots,” this study employs the Local Moran’s Index (LISA), following the methodological framework of Wei (2025) [70]:

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

(3)

where ${\mathrm{x}}_{\mathrm{i}}$ is the CEE observation for city $\mathrm{i};\overline{\mathrm{x}}$ is the sample mean; ${\mathrm{S}}^2$ denotes the sample variance; and ${\mathrm{w}}_{\mathrm{i}\mathrm{j}}$ represents the elements of the row-standardized spatial weight matrix.

3.2.2. Spatial Durbin Model (SDM) Construction

The confirmation of spatial autocorrelation renders traditional panel models inadequate for model specification [71]. Consequently, a spatial regression model must be employed. Standard spatial models include the Spatial Lag Model (SLM), the Spatial Error Model (SEM), and the Spatial Durbin Model (SDM). This study adopts the SDM as the baseline model. By simultaneously incorporating spatial lags of both the dependent and explanatory variables, the SDM effectively mitigates potential specification bias from omitted spatially correlated variables and represents a more generalized form of the SLM and SEM [35,68]. The panel SDM with spatial and time fixed effects is specified as follows:

$$Y_{it}=\rho W_{ij}{Y}_t+X_{it}\beta +W_{ij}{X}_t\theta +{\delta }_i+{\gamma }_t+{ϵ}_{it}$$

(4)

where $Y_{it}$ is the dependent variable $; X_{it}$ is the vector of explanatory variables; $W_{ij}$ is the spatial weight matrix; $\rho$ is the spatial autoregressive coefficient; $\beta$ and $\theta$ are the vectors of coefficients for the explanatory variables and their spatial lags, respectively; ${\delta }_i$ and ${\gamma }_t$ represent spatial and time fixed effects; and ${ϵ}_{it}$ is the stochastic error term.

A key feature of the SDM is its capacity to capture feedback effects, whereby a change in an explanatory variable in one region spills over to neighboring regions, which in turn affects the region of origin. Consequently, the model coefficients, specifically β and θ, do not directly represent the marginal effects of the explanatory variables. To address this, LeSage and Pace (2009) [72] developed a partial differentiation method to decompose the total impact into direct and indirect effects [73]. This methodology was later extended by Elhorst (2013) to the context of spatial panel data [74].

First, Equation (4) is expressed in matrix form for all N cross-sectional units at time t:

$$(I-\rho W_{ij})Y_t=X_t\beta +W_{ij}{X}_t\theta +\delta +{\gamma }_t{\iota }_N+{ϵ}_t$$

(5)

where $Y_t$, $X_t$, $\delta$, and ${ϵ}_t$ are the $N\times 1$ or $N\times K$ vectors (or matrices) of the dependent variable, explanatory variables, spatial fixed effects, and the error term, respectively. $I$ is the $N\times N$ identity matrix, and ${\iota }_N$ is an $N\times 1$ vector of ones. Solving Equation (5) for $Y_t$ yields

$$Y_t=(I-\rho W_{ij}{)}^{-1}(X_t\beta +W_{ij}{X}_t\theta )+(I-\rho W_{ij}{)}^{-1}(\delta +{\gamma }_t{\iota }_N+{ϵ}_t)$$

(6)

From Equation (6), the matrix of partial derivatives of $Y$ with respect to the $k$-th explanatory variable, $X_k$, is an $N\times N$ matrix given by

$$\left[{\displaystyle \frac{\partial Y}{\partial X_k}}\right]=\left[\begin{array}{ccc}{\displaystyle \frac{\partial Y_1}{\partial X_{1k}}}& \cdots & {\displaystyle \frac{\partial Y_1}{\partial X_{Nk}}}\\ ⋮& \ddots & ⋮\\ {\displaystyle \frac{\partial Y_N}{\partial X_{1k}}}& \cdots & {\displaystyle \frac{\partial Y_N}{\partial X_{Nk}}}\end{array}\right]=(I-\rho W_{ij}{)}^{-1}\left[\begin{array}{cccc}{\beta }_k& w_{12}{\theta }_k& \cdots & w_{1N}{\theta }_k\\ w_{21}{\theta }_k& {\beta }_k& \cdots & w_{2N}{\theta }_k\\ ⋮& ⋮& \ddots & ⋮\\ w_{N1}{\theta }_k& w_{N2}{\theta }_k& \cdots & {\beta }_k\end{array}\right]$$

(7)

The average of the diagonal elements of this matrix represents the direct effect, while the average of the off-diagonal elements constitutes the indirect effect.

4. Results

4.1. Spatio-Temporal Pattern of Urban–Rural Integration

The development of urban–rural integration (URI) in China exhibits significant spatio-temporal heterogeneity (Figure 4a–c). Spatially, the overall pattern shows a marked imbalance. High-URI areas are predominantly clustered in the eastern coastal agglomerations, with National Center Cities such as Shanghai, Shenzhen, Beijing, and Guangzhou serving as growth poles that exert strong spillover and leading effects. Conversely, most cities in the central and western regions have long remained at a low level of integration, especially non-capital and non-hub cities in central provinces like Hunan and Hubei and in western provinces such as Guizhou, Yunnan, Qinghai, Ningxia, and Guangxi. This creates a stark gap with the eastern core cities, clearly illustrating the “East- West” gradient in coordinated urban–rural development.

Temporally, the data revealed a positive trend between 2005 and 2022. The national average URI increased from 0.050 to 0.067 (Figure 4d,e), the number of high-URI cities grew, and the spatial extent of low-URI cities contracted, suggesting that national urban–rural coordination policies have achieved initial success. The indices for all sub-dimensions also rose steadily, with the social–demographic dimension showing notably rapid growth after 2013. This is likely attributable to the household registration (hukou) system reforms associated with the National New-Type Urbanization Strategy, which spurred the urbanization of the rural migrant population. Despite this progress, the overall URI increase was modest and has not been sufficient to bridge the significant pre-existing regional disparities. The lagging integration process in the majority of cities indicates that deep-seated structural barriers to dismantling the urban–rural dual system persist. Consequently, future policies must target cities at different development levels more precisely, particularly less-developed areas in the central and western regions.

4.2. Spatio-Temporal Pattern of Carbon Emission Efficiency

China’s city-level carbon emission efficiency (CEE) also exhibits significant spatio-temporal heterogeneity (Figure 5a–c). Spatially, the overall pattern shows a “south-high, north-low” regional differentiation. Specifically, the eastern and central urban agglomerations constitute the high-efficiency zones, with some cities in the Yangtze River Delta, the Pearl River Delta, and the middle reaches of the Yangtze River demonstrating a particularly high performance. In contrast, cities in the northwest and northeast—notably, in provinces like Heilongjiang, Liaoning, Qinghai, and Ningxia—generally demonstrate lower levels of CEE. This spatial distribution pattern highlights the significant disparities in CEE between regions and is closely correlated with imbalances in regional economic development and green transition progress.

Temporally, the national average CEE increased significantly, from 0.20 in 2005 to 0.54 in 2022. The distribution of CEE values also shifted upwards, as evidenced by the rising trend across its quartiles (Figure 5d,e). This upward trend indicates that China’s ‘dual carbon’ strategy has been effective in enhancing overall carbon emission efficiency.

4.3. Analysis of Spatial Effect Mechanisms Based on SDM

4.3.1. Spatial Autocorrelation Test

The application of spatial econometric models requires the presence of spatial autocorrelation in the dependent variable. Accordingly, we first test for such autocorrelation in carbon emission efficiency (CEE). To account for the interactive effects of economic activities, we employ a spatial economic distance weight matrix (W) for this test. The test results show that the Global Moran’s I index for CEE (

Table 4. Global Moran’s I index for carbon emission efficiency (CEE).

Year	Moran’s I	Year	Moran’s I
2005	0.097 ***	2014	0.142 ***
2006	0.123 ***	2015	0.152 ***
2007	0.146 ***	2016	0.175 ***
2008	0.153 ***	2017	0.242 ***
2009	0.197 ***	2018	0.264 ***
2010	0.187 ***	2019	0.236 ***
2011	0.171 ***	2020	0.217 ***
2012	0.173 ***	2021	0.210 ***
2013	0.202 ***	2022	0.291 ***

Note: *** denotes significance at the 1% level.

), calculated using this matrix, is positive and statistically significant at the 5% level for all years, indicating the presence of significant positive spatial autocorrelation. Furthermore, its Moran’s I scatter plot (Figure 6) reveals that the vast majority of cities are distributed in the first (High–High) and third (Low–Low) quadrants, exhibiting significant spatial homogeneous clustering.

Drawing upon the methodology of Wei (2025) [70], High–High (HH) clusters represent “hot spots,” indicating that cities with high carbon emission efficiency (CEE) are spatially contiguous with high-efficiency neighbors. Conversely, Low–Low (LL) clusters denote “cold spots,” reflecting the spatial aggregation of low-efficiency cities. As illustrated by the LISA cluster map in Figure 7, significant HH clusters are primarily concentrated in the eastern coastal region—specifically the Yangtze River Delta, Pearl River Delta, and Beijing–Tianjin–Hebei urban agglomerations—as well as Ordos in Inner Mongolia. In contrast, LL clusters are predominantly distributed across the northwestern and northeastern regions (e.g., Ningxia, Gansu, and Heilongjiang) and Hubei Province in Central China. This observed spatial configuration confirms that spatial disparities in CEE are driven by pronounced regional clustering and spatial path dependence. These findings further underscore the necessity of accounting for spatial effects within empirical models to rigorously evaluate the impact of urban–rural integration (URI).

4.3.2. Spatial Model Specification Tests

To prevent model specification bias and accurately identify spatial effects, we performed a series of rigorous diagnostic tests. The selection process proceeded as follows. First, the Hausman test result (p < 0.01) rejected the random effects model in favor of a fixed effects specification. Second, significant results from the Lagrange Multiplier (LM) tests confirmed the presence of spatial autocorrelation, necessitating the use of a spatial econometric approach. Crucially, both the Wald and Likelihood Ratio (LR) tests strongly rejected the null hypotheses that the Spatial Durbin Model (SDM) could be simplified to either a Spatial Lag Model (SLM) or a Spatial Error Model (SEM). This sequence of tests provides compelling evidence that the SDM, which captures spatial interactions from the dependent variable, the explanatory variables, and the error term, is a more comprehensive and appropriate specification than the more restrictive SLM or SEM. The detailed results of these specification tests are presented in

Table 5. Results of spatial econometric model specification tests.

Test	Statistic	p-Value	Test	Statistic	p-Value
Tests for Spatial Dependence			Tests for Model Simplification (SDM vs. SAR/SEM)
LM test (lag)	429.209	<0.001 ***	Wald test (lag)	13.99	0.016 **
LM test (error)	315.531	<0.001 ***	Wald test (error)	28.9	<0.001 ***
Robust LM test (lag)	130.094	<0.001 ***	LR test (lag)	19.15	<0.001 ***
Robust LM test (error)	16.416	<0.001 ***	LR test (error)	28.8	<0.001 ***
Test for Fixed vs. Random Effects
Hausman test	45.073

Note: *** and ** denote significance at the 1% and 5% levels, respectively.

. Therefore, the fixed effects Spatial Durbin Model is selected as the baseline model for our analysis.

4.3.3. Model Regression Results and Analysis

To demonstrate the necessity of a spatial econometric approach, this section compares the results of the Spatial Durbin Model (SDM) with two-way fixed effects against those of a standard Ordinary Least Squares (OLS) model. As shown in

Table 6. The results of model regression.

Variables	OLS Model (Fixed Effects)	Spatial Durbin Model (Fixed Effects)
Main Variables
URI	−0.485 ** (−2.03)	−0.554 *** (−6.57)
URI 2	1.226 *** (2.96)	1.748 *** (10.52)
Control Variables
ECI	−0.023 *** (−3.50)	−0.065 *** (−14.39)
ILV	0.026 (0.55)	0.036 (0.65)
LUL	−0.006 * (−1.74)	−0.004 ** (−2.21)
TIL	0.101 * (1.93)	0.157 *** (5.32)
EDL	0.002 *** (6.62)	0.001 *** (19.5)
Spatial Lag Terms
W × URI		−0.662 ** (−2.52)
W × URI 2		0.866 * (1.86)
W × ECI		−0.058 *** (−4.66)
W × ILV		0.088 (0.65)
W × LUL		−0.010 ** (−2.21)
W × TIL		0.415 *** (2.72)
W × EDL		0.002 *** (7.40)
Model Diagnostics
Constant	0.282 *** (6.55)
sigma2_e	0.003 *** (66.39)	0.003 *** (50.44)
Observations	5094	5094
R2	0.3742	0.4342
Log-likelihood		7823.8307
City FE	YES	YES
Year FE	YES	YES

Note: W represents the spatial weight matrix, and W× denotes the spatially lagged term of the respective variable. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively; t-values are in parentheses.

, the SDM outperforms the OLS model on key goodness-of-fit metrics. The SDM’s R² (0.4342) is markedly higher than the OLS model’s (0.3742), and its higher log-likelihood value (7823.83) further indicates a superior statistical fit. Crucially, the coefficients of the spatially lagged terms, W × URI (coeff. = −0.662, t = −2.52) and W × URI² (coeff. = 0.866, t = 1.86), are statistically significant. While both models identify a U-shaped nonlinear effect of urban–rural integration (URI), the distinct advantage of the SDM is its ability to capture the spatial spillover mechanisms that the OLS model overlooks, thereby uncovering how URI development in one region affects outcomes in its neighbors.

4.3.4. Analysis of Decomposed Spatial Effects

Following the decomposition approach of LeSage and Pace (2009) [73], we separated the total effect of urban–rural integration (URI) on carbon emission efficiency (CEE) into direct and indirect effects. The direct effect captures the impact on the local region (including feedback loops), while the indirect effect, defined as the spatial spillover, reflects the impact on neighboring regions. The total effect, the sum of these two, represents the variable’s overall marginal impact.

The decomposition results (

Table 7. Decomposition of spatial effects: direct, indirect, and total effects.

Variable	Direct Effects	Indirect Effects	Total Effects
URI	−0.506 *** (−6.26)	0.456 ** (2.31)	−0.05 (−0.34)
URI2	2.447 *** (10.20)	−1.267 *** (−3.71)	1.180 * (1.80)
ECI	−0.065 *** (−15.20)	−0.070 *** (−5.75)	−0.135 *** (−10.59)
ILV	0.044 (0.99)	0.094 (0.62)	0.138 ** (2.32)
LUL	−0.003 ** (−2.20)	0.011 (1.20)	0.008 (1.45)
TIL	0.359 *** (5.41)	0.497 *** (3.12)	0.856 *** (4.53)
EDL	0.000 *** (18.69)	0.001 *** (6.76)	0.001 (0.35)

Note: ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively; t-values are in parentheses.

) indicate that the effect of URI on CEE is significantly nonlinear. For the direct effect, URI exhibits a distinct U-shaped relationship with local CEE, with statistically significant linear (coefficient = −0.506) and quadratic (coefficient = 2.447) terms. This suggests that, in the initial stages of integration (i.e., when URI is below the inflection point of 0.103), further development inhibits local CEE. However, beyond this critical threshold, URI begins to exert a significant positive influence, driving improvements in local efficiency (Figure 8). Conversely, the indirect effect, or spatial spillover, follows an inverted U-shaped path (linear coefficient = 0.456; quadratic coefficient = −1.267). This implies that, while URI initially generates positive, synergistic spillovers for neighboring regions, its effect may reverse after crossing a higher threshold (0.1799). This reversal is potentially driven by mechanisms such as resource siphoning and inter-regional competition, which create negative spatial externalities.

Decomposition of the control variables reveals that various drivers influence carbon emission efficiency (CEE) through distinct direct and spatial spillover channels. Within the urbanization and industrialization dimensions, land urbanization (LUL) and industrialization (ILV) exhibit divergent impact trajectories. The direct effect of LUL is significantly negative, primarily because rapid urban expansion drives the conversion of ecological land with a high carbon sequestration capacity into built-up areas. Furthermore, intensive short-term infrastructure development escalates energy consumption, thereby undermining local CEE [75]. Notably, its spatial spillover effect is insignificant, suggesting that land resource allocation—predominantly steered by local governments—lacks cross-regional coordination, which constrains the systematic diffusion of environmental impacts to adjacent regions [76]. Conversely, while the direct and indirect effects of ILV lack individual significance, the total effect is significantly positive. This suggests that the impact of industrialization on CEE is cumulative, with embedded technological advancements and industrial restructuring yielding net positive contributions to the broader regional context.

In terms of energy, technology, and economic dimensions, these factors exhibit pronounced spatial linkages. The direct and indirect effects of energy consumption intensity (ECI) are both significantly negative, implying that excessive local energy intensity not only hampers internal carbon efficiency [77] but also exerts negative externalities on neighboring regions through industrial interdependencies, reflecting a deep-seated path dependency on regional energy consumption. In contrast, technological innovation (TIL) and economic development (EDL) demonstrate robust positive synergies. For TIL, both direct and spatial spillover effects are significantly positive, suggesting that local R&D investment catalyzes internal low-carbon transitions [78], while simultaneously benefiting adjacent areas through knowledge diffusion and technological collaboration [79]. Similarly, EDL exhibits significantly positive direct and indirect effects, underscoring that economic prosperity provides the requisite financial, institutional, and market-driven support for green transitions [80], while fostering regional improvements through collaborative governance and policy demonstration effects [81].

4.4. Endogeneity Analysis

Although the benchmark regression includes a comprehensive set of control variables and two-way fixed effects, the relationship between urban–rural integration (URI) and carbon emission efficiency (CEE) may remain susceptible to bidirectional causality. Specifically, while URI promotes CEE through optimized resource allocation, cities with higher CEE levels may simultaneously leverage their green technological and industrial advantages to attract high-quality factor inflows, thereby further accelerating urban–rural integration. To mitigate potential endogeneity bias, following Hu et al. (2023) [82], this study adopts the one-period lagged values of URI and its quadratic term $(\mathrm{L}.\mathrm{U}\mathrm{R}\mathrm{I}$ and $(\mathrm{L}.\mathrm{U}\mathrm{R}\mathrm{I}{)}^2$) as instrumental variables within a two-stage least squares (IV-2SLS) framework. This lagging strategy effectively addresses reverse causality by ensuring that the explanatory variables maintain temporal precedence over the dependent variables.

The results reported in

Table 8. Endogeneity test results.

Variables	(1) URI	(2) URI2	(3) CEE
L.URI	0.856 *** (20.15)	0.156 ** (2.21)
(L.URI)2	0.102 ** (2.34)	0.765 *** (15.40)
URI			−0.318 ** (−2.24)
URI2			1.825 ** (2.08)
Control Variables	YES	YES	YES
Constant	YES	YES	YES
Observations	4811	4811	4811
R2	0.885	0.842	0.450
First-stage F-statistic	125.40	110.15
Kleibergen–Paap rk LM	40.390 ***
Kleibergen–Paap rk Wald F	45.180 {16.38}

Note: *** and ** denote significance at the 1% and 5% levels, respectively.

indicate that the instrumental variables exhibit strong relevance in the first-stage regressions (Columns 1–2), with coefficients significant at the $1\mathrm{\%}$ or $5\mathrm{\%}$ levels. The reported $\mathrm{F}$-statistics (125.40 and 110.15) substantially exceed the conventional threshold of 10. Furthermore, the Kleibergen–Paap rk Wald F-statistic significantly surpasses the Stock–Yogo critical values, thereby rejecting the null hypothesis of weak instruments. In the second-stage regression (Column 3), after accounting for endogeneity, the coefficients for the linear and quadratic terms of URI are −0.318 and 1.825, respectively, both significant at the $5\mathrm{\%}$ level. These findings confirm that the U-shaped relationship between URI and CEE remains robust, reinforcing the validity and reliability of the benchmark conclusions.

4.5. Robustness Analysis

To validate the robustness of our core conclusions, we conducted two sets of robustness checks. First, we tested the sensitivity of our findings to the spatial weight matrix specification by replacing it with an economic–geographical nested matrix $\left({\mathrm{W}}_{\mathrm{i}\mathrm{j}}^{\mathrm{n}\mathrm{e}\mathrm{s}}\right)$ (Model 1). Drawing on Lei et al. (2024) [69], this matrix integrates the combined influences of geographic distance and economic interconnectedness, defined as follows:

$${\mathrm{W}}^{\mathrm{n}\mathrm{e}\mathrm{s}}={\mathrm{W}}^{\mathrm{d}}\times \mathrm{d}\mathrm{i}\mathrm{a}\mathrm{g}\left({\displaystyle \frac{{\overline{\mathrm{p}\mathrm{g}\mathrm{d}\mathrm{p}}}_1}{\overline{\mathrm{p}\mathrm{g}\mathrm{d}\mathrm{p}}}},{\displaystyle \frac{{\overline{\mathrm{p}\mathrm{g}\mathrm{d}\mathrm{p}}}_2}{\overline{\mathrm{p}\mathrm{g}\mathrm{d}\mathrm{p}}}},\cdots ,{\displaystyle \frac{{\overline{\mathrm{p}\mathrm{g}\mathrm{d}\mathrm{p}}}_{\mathrm{n}}}{\overline{\mathrm{p}\mathrm{g}\mathrm{d}\mathrm{p}}}}\right)$$

(8)

where ${\mathrm{W}}^{\mathrm{d}}$ denotes the inverse distance weight matrix, with elements ${\mathrm{W}}_{\mathrm{i}\mathrm{j}}^{\mathrm{d}}=1/{\mathrm{d}}_{\mathrm{i}\mathrm{j}}$ for $\mathrm{i}\ne \mathrm{j}$ and ${\mathrm{W}}_{\mathrm{i}\mathrm{j}}^{\mathrm{d}}=0$ for $\mathrm{i}=\mathrm{j}$. Here, ${\mathrm{d}}_{\mathrm{i}\mathrm{j}}$ represents the great-circle distance between cities $\mathrm{i}$ and $\mathrm{j}$ derived from geographic coordinates. ${\overline{\mathrm{p}\mathrm{g}\mathrm{d}\mathrm{p}}}_{\mathrm{i}}$ and $\overline{\mathrm{p}\mathrm{g}\mathrm{d}\mathrm{p}}$ represent the average per capita GDP of city $\mathrm{i}$ and the entire sample during the study period, respectively. The matrix is row-standardized prior to estimation.

Second, we addressed the potential influence of extreme values by re-estimating the model after excluding municipalities and provincial capitals, which may act as outliers (Model 2). The SDM estimation results under both scenarios (

Table 9. Robustness test results.

	Model (1)			Model (2)
Variable	Direct Effects	Indirect Effects	Total Effects	Direct Effects	Indirect Effects	Total Effects
URI	−0.447 *** (−4.74)	0.316 * (1.76)	−0.131 ** (−2.41)	−0.463 *** (−5.24)	0.130 * (1.93)	−0.333 *** (−6.24)
URI2	1.938 *** (5.43)	−1.594 *** (−3.00)	0.344 *** (5.79)	2.542 *** (8.78)	−0.802 ** (−1.96)	1.740 *** (7.78)
Control Variables	YES			YES
R2	0.4212			0.3832
Observations	5094			4536

Note: ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively; t-values are in parentheses.

) show that the coefficients for the explanatory variable (URI) and its quadratic term remain consistent with our baseline findings in terms of their signs, significance levels, and turning points. This process confirms the robustness of our core finding: the relationship between URI and CEE is characterized by a U-shaped direct effect and an inverted U-shaped spatial spillover.

4.6. Heterogeneity Analysis

4.6.1. Regional Heterogeneity Results

To account for China’s pronounced regional disparities, we conducted a heterogeneity analysis by dividing the sample into eastern, central, and western regions. The results (

Table 10. Results of regional heterogeneity analysis.

	Eastern Region			Central Region			Western Region
Variable	Direct Effects	Indirect Effects	Total Effects	Direct Effects	Indirect Effects	Total Effects	Direct Effects	Indirect Effects	Total Effects
URI	−0.460 *** (−2.96)	1.071 ** (2.44)	0.611 (1.32)	−0.174 * (−1.80)	−1.310 * (−1.68)	−1.484 (−1.42)	−4.839 *** (−5.51)	3.104 *** (2.79)	−1.735 * (−1.73)
URI2	1.617 *** (7.05)	−1.465 ** (−2.46)	0.152 (0.34)	0.336 (0.56)	−3.013 (−0.75)	−2.677 (−0.72)	6.397 *** (7.13)	−4.990 *** (−3.71)	1.407 (1.23)
Control Variables	YES			YES			YES
Observations	1782			1800			1512
R2	0.3745			0.2932			0.3541

Note: ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively; t-values are in parentheses.

) reveal significant regional heterogeneity in the impact of urban–rural integration (URI) on carbon emission efficiency (CEE). In the eastern region, the effects closely mirror the national-level findings, exhibiting a baseline pattern of a U-shaped direct effect and an inverted U-shaped indirect effect. The western region exhibits a more pronounced nonlinear relationship. While the U-shaped and inverted U-shaped paths are consistent, the steeper slopes of the curves indicate that the URI induces a greater marginal change in CEE in this region. The central region, by contrast, displays a distinct pattern where the U-shaped direct effect is not statistically supported due to an insignificant quadratic term. Instead, the spatial spillover manifests as a significant linear negative effect. This spatial differentiation—a baseline pattern in the east, a more pronounced effect in the west, and a deviating pattern in the center—underscores the significant spatial imbalance in the synergistic development of URI and CEE across China.

4.6.2. Dimensional Decomposition Results

To further explore the underlying mechanisms, we analyzed the respective impacts of the four constituent dimensions of the aggregate URI index. Beginning with the direct effects, the analysis (

Table 11. Results of heterogeneity analysis for URI dimensions.

Variable	Direct Effects	Indirect (Spillover) Effects	Total Effects
A. Economic Integration
ECON	−0.559 *** (−3.60)	1.263 *** (2.68)	0.705 ** (1.42)
ECON2	1.982 *** (8.44)	−1.581 *** (−2.59)	0.401 (0.61)
B. Social-Demographic Integration
SOC	−0.052 * (−1.72)	−0.103 *** (−4.01)	−0.155 *** (−5.80)
SOC2	0.006 * (1.66)	0.016 *** (3.70)	0.022 *** (4.99)
C. Spatial Integration
SPA	−0.250 * (−1.88)	−0.491 * (−1.82)	−0.741 ** (−2.54)
SPA2	0.390 ** (2.52)	0.593 * (1.78)	0.983 *** (2.72)
D. Ecological Integration
ECO	−0.010 (−1.43)	0.033 *** (4.71)	0.023 *** (3.01)
ECO2	−0.000 (−1.42)	−0.000 * (−1.73)	−0.000 *** (−4.88)
Control Variables	YES
Observations	5094
R2	0.2974

Note: ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively; t-values are in parentheses.

) shows that economic, social–demographic, and spatial integration each follow a U-shaped impact trajectory on local CEE, mirroring the aggregate URI effect of initial inhibition followed by promotion. Ecological integration, however, deviates from this pattern; its direct effect is linear and negative, though statistically insignificant.

In terms of indirect (spillover) effects, two distinct patterns emerge. The spillover effects of economic and ecological integration align with the aggregate model, following an inverted U-shaped path (initial promotion, later inhibition). In contrast, the spillovers from social–demographic and spatial integration diverge, exhibiting a U-shaped trajectory. Notably, across all dimensions, economic integration provides the most significant contribution, with both its direct and indirect effect coefficients significant at the 1% level. This finding underscores that economic integration acts as the primary driver, representing the dominant mechanism through which URI influences CEE. Spatial and social–demographic integration subsequently emerge as critical secondary drivers.

5. Discussion

This study reveals a complex, nonlinear relationship between urban–rural integration (URI) and carbon emission efficiency (CEE), governed by a dynamic balance between promotional and inhibitory mechanisms. This interplay manifests as two distinct spatial effects: a U-shaped direct effect on local CEE and an inverted U-shaped spillover effect on neighboring regions.

5.1. The “U-Shaped” Direct Impact of URI on Local CEE

The U-shaped trajectory of urban–rural integration (URI) impacts on local carbon emission efficiency (CEE) extends the Environmental Kuznets Curve (EKC) framework [36] into the realm of urban–rural transformation. This finding refines the linear positive relationship hypothesized in the previous literature. While scholars like Wang & Zhang (2021) argue that narrowing the urban–rural gap linearly facilitates carbon mitigation [15], and provincial-level evidence from China suggests that URI enhances CEE via technological innovation and factor allocation optimization [83], our results reveal that URI is not a monotonic driver. Instead, it follows a bifurcated evolutionary path, characterized by an initial “inhibition phase” that transitions into a “promotion phase” only after crossing a critical threshold.

In the nascent stages where URI remains below the threshold (URI = 0.103), inhibitory effects prevail. Approximately 82% of the sampled cities over the past five years fall within this category, primarily comprising “catching-up” cities in central, western, and northeastern China, alongside 36% of “near-threshold” cities in the eastern and central regions. Regarding city hierarchy, this inhibition is most pronounced in medium-sized cities—specifically secondary hubs within major clusters (e.g., Xiangyang, Yichang, and Jiujiang in the Yangtze River Middle Reaches) and transforming industrial bases (e.g., Tangshan and Xuzhou). These cities are navigating a critical juncture of rapid spatial expansion and functional restructuring, subjecting CEE to significant downward pressure. This inhibitory phase is attributable to several mechanisms: (1) large-scale infrastructure and new town construction driven by spatial integration often entail high-carbon material consumption, risking a “carbon lock-in” effect [40,84]. For example, high-speed rail construction in the Yangtze River Middle Reaches urban agglomeration has induced large-scale urban–rural land development [85], while Tianjin Binhai New Area witnessed a 225% surge in total emissions during its peak construction phase [86]. (2) Transitional phases of economic and social–demographic integration may be characterized by structural friction and resource misallocation due to the inertia of the urban–rural dual system [87]. Fragmented “semi-urban, semi-rural” zones in the Pearl River Delta (PRD) exemplify how institutional rigidities lead to inefficient land use and heightened commuting energy intensity [88]. (3) A path dependency on traditional growth models by local governments can temporarily depress CEE [89], as seen in heavy-industrial hubs like Hohhot and Baotou, where industrial reliance creates a high-carbon-intensity dilemma during URI [90].

As URI matures and surpasses the inflection point (URI = 0.103), its facilitative role in enhancing CEE significantly strengthens. Roughly 18% of the cities—predominantly in developed coastal clusters and national strategic pilot zones—have entered this stage. This promotional effect is unlocked through the multidimensional synergy of economic, spatial, and socio-demographic integration. Economic integration facilitates the optimized, bidirectional allocation of production factors and fosters an industrial low-carbon transition [91,92,93,94]. For instance, Xie & Wu (2023) demonstrated that, in developed regions, URI improves agricultural CEE by boosting technical efficiency [28]. Furthermore, spatial and social–demographic integration enhance CEE by improving infrastructure network effects and scale efficiency in public service delivery [95,96]. Notably, unlike provincial-level findings by Wang et al. (2025), which suggest that all dimensions (including ecological) significantly improve carbon performance [83], our municipal-scale analysis reveals no significant direct effect from ecological integration. This discrepancy likely stems from the following: first, the enhancement of carbon sequestration potential through ecological restoration and landscape connectivity—driven by ecological integration—is a protracted and incremental process; consequently, its marginal contribution remains negligible in the short term [97]. Second, during the nascent and intermediate phases of integration, ecological investments are primarily channeled into policy-driven “end-of-pipe” solutions [98,99] rather than source-level abatement strategies aimed at the fundamental transformation of industrial structures and energy systems. In small and medium-sized cities facing intense developmental pressures, policy priorities often gravitate toward economic expansion [100]. This shift results in a deficit of sustained capital investment and technical support for synergistic pollution and carbon reduction, ultimately creating a decoupling of policy objectives from actual developmental practices. These findings suggest that the emission reduction effects of ecological integration exhibit a pronounced time lag, necessitating indirect realization through the strengthened green orientation of both economic and spatial integration.

Since URI’s elevation to a national strategy in 2017 [17], the 2019 overarching institutional framework has served as a pivotal catalyst for this transition. During the 2017–2022 period, the cohort of cities surpassing the URI threshold (0.103) expanded steadily, exhibiting heterogenous implementation pathways. In “near-threshold” pilot zones, such as the Guangzhou–Qingyuan hinterland and western Chengdu, practices are anchored in ecological value realization and the equalization of basic public services. Regions like Jiangjin (Sichuan) and Anji (Zhejiang) have successfully operationalized the conversion of ecological assets into green productivity via carbon trading and eco-agriculture [101,102], while Jinan (Shandong) has mitigated energy dissipation in infrastructure through integrated utility planning. Conversely, in “beyond-threshold” hubs—notably the core Yangtze River Delta and Pearl River Delta—practices prioritize the multi-directional flux of green capital and technology. These regions consolidate their low-carbon competitive advantages through institutional innovations such as trans-boundary ecological compensation and industrial synergy. Specifically, Shanghai, the Suzhou–Wuxi–Changzhou cluster, and Zhejiang have pioneered “digital countryside” initiatives, leveraging green technologies and digital infrastructure (e.g., 5 G and IoT) to optimize factor allocation efficiency while minimizing energy footprints [103]. Furthermore, in the Jiaxing–Huzhou region, comprehensive land consolidation has been utilized to channel green capital into rural sectors [104], while the Guangzhou–Foshan and Shenzhen–Dongguan clusters have effectively promoted the decoupling of economic growth from emissions by reducing carbon intensity through integrated spatial planning and unified rail transit networks.

5.2. The Inverted U-Shaped Spatial Spillover Effect of URI on Neighboring Regions

The impact of urban–rural integration (URI) on the carbon emission efficiency of neighboring regions exhibits a distinct inverted U-shaped spatial spillover pattern. Specifically, when URI remains below the critical threshold (URI = 0.179), positive spillovers predominate; however, surpassing this inflection point may precipitate a shift toward negative spillovers. Empirical evidence suggests that approximately 88% of the sampled cities have traversed the ascending phase of this curve over the past five years, characterized by beneficial spatial externalities. In this stage, core cities function as growth poles, facilitating the synergistic flow of capital, expertise, and managerial experience toward rural peripheries and urban–rural interfaces. This process enhances regional resource and energy efficiency through knowledge spillovers, the diffusion of low-carbon technologies, and urban–rural industrial economies of scale [27,105], thereby bolstering the carbon efficiency of adjacent areas. Most observed cities—including Jiaxing and Huzhou in Zhejiang, the metropolitan peripheries of Wuhan and Changsha, and sectors of the Chengdu-Chongqing region—reside within this phase. For instance, Wuhan and Changsha have leveraged industrial chain extensions to channel low-carbon technologies and management paradigms into peripheral county-level industrial parks, catalyzing the green transition of rural industries [106]. Similarly, Yiwu and Cixi have utilized e-commerce networks to integrate supply chains, driving the intensification and decarbonization of rural cottage industries and fostering synergistic emission reductions.

Conversely, negative spillovers may emerge once core cities cross the URI inflection point (URI = 0.179). This cohort represents approximately 12% of the sample, primarily comprising Tier 1 hubs like Beijing, Shanghai, Guangzhou, and Shenzhen, alongside major provincial capitals in central and western China. To enhance local carbon efficiency and comply with stringent environmental mandates, these cities often externalize carbon-intensive production stages via industrial restructuring. This dynamic triggers “carbon leakage” risks that potentially suppress the carbon emission efficiency of neighboring regions [107,108]. For instance, leading metropolises such as Beijing and Chongqing have strategically relocated carbon-intensive industries from their urban cores during the deepening of urban–rural integration, guided by their functional positioning and spatial blueprints to optimize industrial configurations [109]. Similarly, the core zone of the Yangtze River Delta has reallocated traditional manufacturing to peripheral regions, such as Northern Jiangsu and Northern Anhui, during its transition toward a service-led economy. These trends underscore that, in the advanced stages of urban–rural integration, vigilance is imperative regarding potential regional disparities and environmental negative externalities that may arise from localized improvements in carbon efficiency.

6. Conclusions

6.1. Main Research Conclusions

This study finds that the spatio-temporal evolution of urban–rural integration (URI) and carbon emission efficiency (CEE) is characterized by two key features: a consistent upward trend over time and a pronounced non-equilibrium in spatial patterns. Specifically, (1) temporally, both exhibit consistent improvement, though CEE has advanced more rapidly than the more gradual URI process. Among URI’s sub-dimensions, social–demographic integration shows the most significant progress. (2) Spatially, URI exhibits a clear “East-West” gradient, with eastern coastal areas significantly outpacing the central and western regions, perpetuating a significant regional gap. CEE, meanwhile, follows a “south-high, north-low” pattern, with a higher efficiency concentrated in the south. Notably, the high-value clusters for both URI and CEE demonstrate a strong spatial coupling within national-level urban agglomerations like the Yangtze River and Pearl River Deltas, highlighting the deep-seated nature of China’s regional disparities.

The Spatial Durbin Model (SDM) reveals the nonlinear effects of URI on CEE: (1) the direct effect follows a “U-shaped” trajectory, evolving from inhibition at low-URI levels to promotion at high levels. (2) The spatial spillover effect displays an inverted “U-shaped” curve, reflecting a dynamic shift in regional interactions from synergistic enhancement to intensified competition. (3) The effects exhibit significant regional heterogeneity: the impact is strongest in the western region and weakest in the central region, relative to the eastern region. (4) An analysis of sub-dimensions indicates that economic integration is the primary driver within the URI framework. These findings underscore the complex, stage-dependent, and spatially heterogeneous nature of URI’s impact on CEE. Consequently, policy interventions must move beyond a “one-size-fits-all” approach and instead be dynamically adaptive, regionally differentiated, and proactive.

6.2. Policy Implications

To foster synergy between urban–rural integration (URI) and carbon emission efficiency (CEE), the following policy directions are recommended based on our findings:

(1)

Design dynamic, differentiated policy frameworks to navigate the nonlinear threshold effects of URI. In regions positioned below the critical threshold (nascent integration stages), policymakers should prioritize “preemptive greening” to avert long-term carbon lock-in. It is imperative for local governments to catalyze green infrastructure investment through targeted fiscal incentives—for instance, by incorporating photovoltaic arrays into rural transport networks and deploying distributed renewable energy microgrids [110]. Such measures facilitate sustainable cost-sharing mechanisms, preventing nascent developments from being tethered to high-carbon trajectories while streamlining the bidirectional flow of green production factors across the urban–rural interface.

Conversely, for regions that have surpassed the threshold (advanced integration stages, e.g., the Yangtze River Delta), the governance paradigm must shift toward market-driven collaborative frameworks [111]. The primary objective is to internalize the environmental externalities of industrial relocation by implementing regional carbon emission trading schemes (ETS) and robust ecological compensation mechanisms. Establishing a unified carbon market alongside a framework for rural carbon sequestration credits (carbon sinks) would incentivize industrial upgrading and low-carbon innovation, transitioning from administrative interventions to market-based instruments. This shift will optimize marginal abatement costs, safeguard environmental regulatory standards against dilution, and fully harness the decarbonization potential of the urban–rural continuum.

(2)

Develop differentiated carbon-reduction pathways based on regional heterogeneity. For the pioneer zones of URI in Eastern China, policy should prioritize institutional innovation and technological spillovers, channeling high-end factors into the agricultural and rural sectors to promote the green transformation of urban–rural industrial chains, thereby enhancing overall regional carbon efficiency [47,112]. For regions with high URI potential, such as those in the northeast, central, and west of China, strategies should involve the deployment of new green infrastructure, the development of distinctive low-carbon industries based on local resource endowments (e.g., smart agriculture, ecotourism), and targeted policy interventions to support areas that are lagging [113,114,115].

(3)

Strengthen regional collaborative governance to mitigate negative spatial spillovers. This requires establishing urban–rural technology-sharing platforms to reduce barriers to green innovation in rural areas and accelerate the diffusion of advanced technologies. It also involves fostering cross-regional industrial collaboration networks to optimize overall abatement costs through economies of scale and scope. Crucially, a harmonized environmental regulatory framework must be institutionalized across the urban–rural divide to thwart the leakage of pollution-intensive industries and prevent the formation of “pollution havens.”

6.3. Research Limitations and Future Outlook

Despite its contributions, this study is subject to several limitations that warrant further exploration. First, regarding causality and endogeneity: although we employed one-period lagged variables and IV-2SLS estimation, the recursive interaction between URI and CEE remains complex. Regions with superior CEE may naturally attract high-quality factors due to optimized structures, potentially accelerating URI. Future research should consider the dynamic panel Generalized Method of Moments (GMM) or quasi-natural experiments to better disentangle these causalities. Second, concerning data granularity, our indicators rely primarily on conventional statistical yearbooks. Future studies could leverage high-resolution geographic grids or satellite-derived remote sensing data to enhance spatiotemporal resolution, thereby capturing the fine-grained spatial dynamics between URI and carbon emissions with greater precision. Third, the transmission mechanisms underlying the URI-CEE nexus warrant further empirical scrutiny. While this study establishes a preliminary theoretical framework, future efforts could employ multi-step mediation or moderated mediation models to systematically quantify these pathways and unveil the specific channels through which URI influences CEE. Finally, subsequent research should consider refining the spatial scale by focusing on county-level units or specific trans-regional urban agglomerations. By integrating quantitative modeling with qualitative approaches—such as in-depth case studies and field investigations—future work can generate more tailored, context-specific policy insights that account for diverse geographical and developmental backgrounds.