Spatial Correlation Network Characteristics of Comprehensive Transportation Green Efficiency in China

Abstract

Accurately characterizing the structural features of the spatial correlation network of comprehensive transportation green efficiency (CTGE) is essential for achieving balanced regional transportation development and eliminating regional disparities. This study employs the slacks-based measure-data envelopment analysis (SBM-DEA) model to assess the CTGE of China. Furthermore, the standard deviational ellipse (SDE) model and social network analysis (SNA) method are adopted to delineate the spatiotemporal evolution patterns and spatial correlation network characteristics of CTGE, based on input–output data from the transportation industry across 30 provinces (municipalities and autonomous regions) between 2003 and 2020. The findings reveal that China’s CTGE exhibits a fluctuating trend of an initial decline followed by subsequent increase, with a national average of 0.555 and an average of 0.722 in eastern regions, 0.434 in central regions, and 0.478 in western regions. This demonstrates that China’s CTGE maintains an overall low level while showing significant regional disparities. The spatial center of gravity of China’s CTGE has shifted from a southwestern to a northeastern trajectory, with a generally concentrated spatial distribution pattern. Furthermore, China’s CTGE demonstrates a distinct “core-edge” hierarchical structure, with regions occupying varied roles and statuses within the network. The central and western regions are positioned at the network periphery, predominantly receiving spillover effects from other regions, while the eastern region, driven by its strong spillover effect, serves as the network’s “engine”. The most significant contribution of this study lies in developing a more comprehensive CTGE evaluation framework and precisely identifying the structural positions and functional roles of different regions within the network, which holds substantial theoretical and practical value for advancing sustainable development in China’s transportation sector.

1. Introduction

Over the past four decades since the implementation of reform and opening-up policies, China’s economy has been increasingly integrated into the global economic landscape. Transportation has played a pivotal role in promoting social development, spurring rapid urban economic growth, and enhancing national openness [1,2]. In the early years of the People’s Republic of China, scheduled land, sea, and air travel were re-established. Following the initiation of reform and opening-up policies, transportation development was prioritized, significantly contributing to the nation’s progress. Today, China ranks first globally in highway and railway mileage, with eight of the world’s top ten ports by throughput located within its territory. These remarkable achievements have given rise to a multi-tiered transportation network that encompasses express delivery, trunk, and basic networks, thus solidifying China’s standing as a global transportation powerhouse.

Despite these advancements, China ranked 27th in the World Bank’s recent comprehensive review of international logistics performance, reflecting suboptimal overall transportation efficiency. This inefficiency stems from systemic issues, including insufficient horizontal integration among transportation modes, resulting in imbalances in functional structure (e.g., mismatches between supply and demand growth), spatial distribution (e.g., pronounced regional disparities), and resource allocation (e.g., low transportation efficiency) [3,4].

In response to these challenges, the Chinese government proposed the “supply-side structural reform” in 2015, emphasizing the need to “strengthen supply-side structural adjustments while moderately expanding aggregate demand, with a focus on improving the quality and efficiency of the supply system” [5]. Specifically for the transportation sector, structural issues manifest in two key dimensions: mismatch between supply and demand where the existing transportation supply fails to adequately meet evolving public demands and imbalanced modal development where significant disparities persist among different transportation modes (e.g., road, rail, waterway, and air).

Furthermore, since the beginning of the 21st century, humanity has faced mounting challenges such as global warming, environmental degradation, and resource depletion. The traditional economic development model—characterized by high input, high consumption, and high output—has become unsustainable. Consequently, green development has emerged as a critical research priority for governments and scholars worldwide, driving innovation in sustainable practices across industries [6]. In response, the Central Committee of the Communist Party of China and the State Council issued the “Outline for Building China’s Strength in Transportation” in 2019, proposing accelerated development of an advanced, modern, secure, convenient, eco-friendly, and cost-efficient transportation system. Guided by the principle of high-quality development, this initiative aims to position China as a global leader in transportation by 2050 [7].

Transportation efficiency, as a critical metric for assessing regional transportation development, has long been a focal point of global academic research, yielding substantial scholarly achievements [8]. Existing models and methodologies predominantly analyze transportation efficiency from an economic perspective, examining its developmental levels, evolutionary patterns, spatial characteristics, and driving factors. However, research remains limited on the synergistic growth of economic–environmental–social systems within the transportation sector. Moreover, the spatial network structural changes induced by the flow of transportation elements have been overlooked, rendering current studies inadequate in holistically delineating the complex spatial hierarchical structures and nested relationships of comprehensive transportation green efficiency (CTGE) across regions within a networked societal context.

As a sector characterized by intensive land resource utilization, high energy consumption, and significant pollution, transportation must prioritize regional coordination and green-intensive development to align with national strategic imperatives and advance high-quality sectoral growth [9]. A pressing academic challenge lies in transcending the limitations of efficiency measurement frameworks rooted solely in economic performance and instead investigating the spatial correlation network characteristics of CTGE—a scientific inquiry vital for reconciling efficiency optimization with sustainable development goals.

This study aims to explore the development trends of CTGE and the roles of various regions within the spatial correlation network. To achieve this, it constructs a framework for analyzing the spatial correlation network characteristics of CTGE in China, measures comprehensive transportation green efficiency using the slack-based measure (SBM) model, and examines temporal and spatial evolution patterns and network structures using the standard deviation ellipse model and social network analysis. The original contribution of this article includes the following four aspects: It proposes a new evaluation system of CTGE, including economic growth, environmental regulation, and social welfare. It studies the spatial correlation network characteristics of China’s CTGE using “relational data”. It demonstrates that China’s CTGE has a distinct “core-edge” hierarchical structure. It finds that the inter-provincial CTGE spatial correlation network in China exhibits a gradient in the transfer process between blocks. The order of this article is as follows: Section 2 is literature review. The methodology, our data, and its sources are presented in Section 3. Section 4 displays the empirical results. Section 5 offers the conclusions and policy implications.

2. Literature Review

Transportation efficiency encompasses rich connotations and distinct characteristics. Based on these attributes, scholars have employed diverse methodologies to measure and evaluate transportation efficiency, including qualitative analysis, quantitative analysis, or hybrid approaches combining both. Existing literature categorizes evaluation methods for transportation efficiency into two primary paradigms: single-factor evaluation and total-factor evaluation.

Single-factor evaluation methods typically utilize indicator systems or ratio analysis to assess comprehensive transportation efficiency. For instance, Fielding et al. employed metrics such as passenger/freight volume per unit cost and employee operating time per unit to evaluate public transit efficiency [10]. Anderson and Fielding systematically analyzed comprehensive transportation efficiency by screening and integrating over 50 indicators [11]. However, with the widespread adoption of production frontier theory, single-factor approaches have gradually been superseded by total-factor evaluation methods.

Total-factor evaluation commonly employs models such as stochastic frontier analysis (SFA) and data envelopment analysis (DEA). For example, Karlaftis et al. integrated parametric (SFA), nonparametric (DEA), and neural network methods to assess the transportation efficiency across 15 European cities from 1990 to 2000 [12]. A review of the literature reveals a clear methodological shift from single-factor to total-factor evaluation. While single-factor methods offer simplicity and interpretability (e.g., explicit indicator definitions), they suffer from the critical limitations of subjectivity (a heavy reliance on empirical assumptions for indicator weighting, preselection, and normalization) and a narrow focus (an inability to capture the holistic resource allocation efficiency of transportation systems) [13]. In contrast, total-factor methods, particularly DEA, have gained prominence due to their ability to eliminate arbitrary weight assumptions, avoid predefined input–output functional relationships, handle multidimensional datasets without normalization, and efficiently resolve multi-input/multi-output complex system evaluations [14]. These advantages have established DEA as a preferred methodology in transportation efficiency research, offering robust analytical frameworks for both theoretical exploration and policy formulation.

In terms of indicator selection, early scholars often used transportation industry output or turnover volume as the sole measure of transportation system performance. However, this approach fails to reflect the actual production processes of the transportation sector, which generate not only economic benefits but also undesirable outputs such as pollution, noise, and accidents. Daraio et al. observed a growing adoption of data envelopment analysis (DEA) in transportation efficiency studies, yet noted that few publications incorporated undesirable outputs, casting doubt on the accuracy of such estimates [15]. Subsequently, Tone proposed the slacks-based measure (SBM) model, a non-radial, non-angular DEA framework that comprehensively accounts for slack variables [16]. Building on this, Cooper et al. introduced the SBM-undesirable model, explicitly integrating undesirable outputs into DEA. This methodological advancement has spurred recent efforts to evaluate transportation systems’ environmental efficiency by incorporating non-economic externalities [17]. For example, Omrani et al. combined cooperative game theory with cross-efficiency methods, considering energy consumption and carbon emissions, to rank the operational efficiency of transportation sectors across Iranian provinces [18]. Park et al. applied the SBM-DEA model to U.S. state-level data (2004–2012), assessing environmental efficiency and estimating carbon reduction potential across 50 states [19]. Liu et al. developed a DEA model incorporating energy, environmental, safety, and noise factors to evaluate China’s road transportation environmental efficiency [20]. Chen et al. employed a four-stage DEA with non-radial directional distance functions (NDDF) to measure energy efficiency in China’s transportation industry while accounting for undesirable outputs and environmental constraints [21].

To sum up, global interest in transportation efficiency as a measure of regional transportation development has led to numerous academic contributions. Initial studies emphasized the efficiency of single transportation modes and urban public transport but have since shifted to assessing comprehensive transportation systems as related theories have advanced [22,23,24]. Methodologically, single-factor evaluations focusing on specific indicators have gradually been replaced by total-factor methods emphasizing allocation efficiency, driven by the interconnected nature of production factors in the transportation industry [25,26,27]. The most widely used approach for total-factor efficiency measurement is data envelopment analysis (DEA), which relies on nonparametric theory. Meanwhile, indicator systems have evolved from measuring transportation efficiency solely in terms of economic output to incorporating environmental efficiency, reflecting the realities of transportation production [28,29,30,31]. Recent studies, in alignment with sustainable development goals, focus on the coordinated development of economic output, environmental protection, and social welfare [32]. Some researchers have expanded evaluations to include “green efficiency”, integrating economic growth, environmental regulation, and social welfare into the framework of comprehensive transportation efficiency [32].

Despite significant progress in transportation efficiency research, notable gaps remain. Firstly, existing studies have largely prioritized the economic and environmental benefits of transportation systems over social dimensions. Secondly, spatial analyses often consider only geographically “adjacent” relationships, leading to partial conclusions that fail to capture the broader spatial correlation of comprehensive transportation efficiency. Finally, cross-regional flows and interactions of resources have grown increasingly complex, forming a “relational data” matrix that traditional “attribute data”-based measurement models cannot adequately address. These challenges highlight the need to characterize the structural features of China’s comprehensive transportation green efficiency (CTGE) spatial correlation network and to identify the roles and positions of individual regions within the network. Such efforts are critical to reducing regional disparities and achieving balanced transportation development nationwide.

3. Materials and Methods

3.1. Research Framework

Figure 1 illustrates the research framework. Firstly, the SBM-DEA model is employed to compute the CTGE (comprehensive transportation green efficiency) for 30 Chinese regions from 2003 to 2022 by inputting relevant indicators, followed by an analysis of its temporal evolution trends and spatial distribution patterns. Secondly, the standard deviational ellipse (SDE) model is applied to identify the dynamic evolution characteristics of CTGE’s spatial configuration, including the movement trajectory of the center of gravity and the parametric characteristics of the standard deviational ellipse. Finally, spatial linkages between regional CTGE values are identified using the Granger causality test, and a social network analysis (SNA) model is constructed to examine the global structural features, individual nodal characteristics, and inter-block interaction mechanisms within the CTGE spatial correlation network.

3.2. Model

3.2.1. SBM-DEA Model

Data envelopment analysis (DEA) is a system analysis technique based on nonparametric analysis principles [33]. The basic concept of it is to employ mathematical programming to project decision-making units (DMUs) onto the efficiency frontier while keeping their input or output levels unchanged. Subsequently, the distance between the DMUs and the frontier is compared to determine their relative efficiency [34]. However, traditional DEA methods do not consider slack variables or unexpected outputs, which limits their accuracy. To address this, Tone proposed the slack-based measure (SBM) model, which incorporates slack variables directly into the objective function to resolve their influence on efficiency measurements, thus improving accuracy [35]. The measurement of CTGE in this study includes desirable outputs (economic growth and social progress) and undesirable outputs (carbon emissions from the transportation sector), which align with the data requirements of the model. Consequently, the SBM-DEA model is adopted to measure China’s CTGE.

Assume there are H decision-making units, with N inputs, M expected outputs, and I unexpected outputs (represented, respectively, by $x\in R_N^{+}$, $y\in R_M^{+}$, and $b\in R_I^{+}$) for each decision-making unit. $X=\left[x_1,x_2,\dots ,x_n\right]\in R^{N\times H}$, $Y=\left[y_1,y_2,\dots ,y_n\right]\in R^{M\times H}$, and $B=\left[b_1,b_2,\dots ,b_n\right]\in R^{I\times H}$, respectively, are the definitions of X, Y, and B. Next, $P\left(x\right)=\left\{\left(x,y,b\right)\left|:x\ge X\lambda ,y\le Y\lambda ,b\ge B\lambda ,\lambda \ge 0\right.\right\}$ is defined as the DEA production possibility set P(x). The SBM model formula, as per Tone’s research findings, is as follows:

$$\theta =\min {\displaystyle \frac{1-\frac{1}{N}{\displaystyle \sum _{n=1}^N{s}_n^x/x_{k\prime n}^{t\prime }}}{1+\frac{1}{M+I}\left({\displaystyle \sum _{m=1}^M{s}_m^y/y_{k\prime m}^{t\prime }}+{\displaystyle \sum _{i=1}^I{s}_i^b/b_{k\prime i}^{t\prime }}\right)}}$$

(1)

$$s.t.{\displaystyle \sum _{t=1}^T{\displaystyle \sum _{k=1}^K{\lambda }_k^t}}{x}_{kn}^t+s_n^x=x_{k\prime n}^{t\prime },n=1,\dots ,N$$

$${\displaystyle \sum _{t=1}^T{\displaystyle \sum _{k=1}^K{\lambda }_k^t}}{y}_{km}^t-s_m^y=y_{k\prime m}^{t\prime },m=1,\dots ,M$$

$${\displaystyle \sum _{t=1}^T{\displaystyle \sum _{k=1}^K{\lambda }_k^t}}{b}_{ki}^t+s_i^b=b_{k\prime i}^{t\prime },i=1,\dots ,I$$

$${\lambda }_k^t\ge 0,s_n^x\ge 0,s_m^y\ge 0,s_i^b\ge 0,k=1,\dots ,K$$

where $S_n^x$ and $S_i^b$ stand for redundant input and unexpected output, respectively, $S_m^y$ for expected output deficiency, $\left(x_{k\prime n}^{t\prime },y_{k\prime m}^{t\prime },b_{k\prime i}^{t\prime }\right)$ for the input–output value of the k′th production unit in the t′ period, ${\lambda }_k^t$ for decision-making unit weight, and θ for the green efficiency value of comprehensive transportation. When θ = 1, the decision-making unit is fully effective; when θ < 1, the decision-making unit experiences efficiency loss, which renders the DEA invalid. Therefore, input and output optimization is required to increase the green efficiency of comprehensive transportation.

3.2.2. Standard Deviation Elliptic Model

Lefever first introduced the standard deviation ellipse (SDE) as an analytical technique for precisely depicting the spatial distribution characteristics of an element [36]. This method is widely used to characterize spatial properties and patterns, revealing the concentration and dispersion of data [37,38]. For instance, the dynamic shifts in the gravity center reflect the temporal migration of the distribution center for geographical elements, the rotation angle of the standard deviation ellipse indicates directional changes in data distribution, and the ellipse area reflects the overall dispersion range of data. As transportation serves as an essential geographical element in production and daily life, accurately measuring the spatial pattern changes in CTGE can holistically reflect regional disparities and agglomeration effects across different areas, laying the foundation for identifying the network structural characteristics of CTGE in subsequent analyses. The following presents the calculation formula:

$$P(x_j,y_j)=\left({\displaystyle \frac{{\displaystyle \sum _{i=1}^n{w}_i{x}_i}}{{\displaystyle \sum _{i=1}^n{w}_i}}},{\displaystyle \frac{{\displaystyle \sum _{i=1}^n{w}_i{y}_i}}{{\displaystyle \sum _{i=1}^n{w}_i}}}\right)$$

(2)

$$\tan \theta ={\displaystyle \frac{\left({\displaystyle \sum _{i=1}^n{w}_i^2{\tilde{x}}_i^2}-{\displaystyle \sum _{i=1}^n{w}_i^2{\tilde{y}}_i^2}\right)+\sqrt{{\left({\displaystyle \sum _{i=1}^n{w}_i^2{\tilde{x}}_i^2}-{\displaystyle \sum _{i=1}^n{w}_i^2{\tilde{y}}_i^2}\right)}^2+4{\displaystyle \sum _{i=1}^n{w}_i^2{\tilde{x}}_i^2{\tilde{y}}_i^2}}}{2{\displaystyle \sum _{i=1}^n{w}_i^2{\tilde{x}}_i^2{\tilde{y}}_i^2}}}$$

(3)

$${\delta }_x=\sqrt{{\displaystyle \frac{{\displaystyle \sum _{i=1}^n{(w_i{\tilde{x}}_i\cos \theta -w_i{\tilde{y}}_i\sin \theta )}^2}}{{\displaystyle \sum _{i=1}^2{w}_i^2}}}},{\delta }_y=\sqrt{{\displaystyle \frac{{\displaystyle \sum _{i=1}^n{(w_i{\tilde{x}}_i\sin \theta -w_i{\tilde{y}}_i\cos \theta )}^2}}{{\displaystyle \sum _{i=1}^2{w}_i^2}}}}$$

(4)

where W_i is a spatial weight representing the comprehensive transportation green efficiency of the neighboring area i, and P(x_j, y_j) is the coordinate of the region j’s center of gravity. The standard deviation ellipse’s rotation angle is defined as tanθ, which is the angle created by rotating the true north direction clockwise to the ellipse’s long axis; denote area i’s relative coordinates with respect to area j’s center of gravity; δ_x and δ_y stand for the standard deviations along the X and Y axes, respectively.

3.2.3. Social Network Analysis Method

Social network theory is the most widely used model for constructing spatial correlation network structures [39], offering multiple quantitative metrics such as node degree, clustering coefficient, and centrality, which enable researchers to objectively evaluate nodes and relationships within the network. Simultaneously, the graphical representation and analysis of social networks allow researchers to intuitively understand network structures and features. Wellman (1998) defined a social network as “a relatively stable system composed of social relations among individuals”. Since then, research on social networks has gained prominence in sociological studies, attracting scholars from both domestic and international disciplines. It is now extensively applied in a broad range of scientific fields, including politics [40], economics [41], and the social sciences [42], and has gradually developed into a comprehensive research framework [43]. As transportation serves as a carrier for the cross-regional flow of resource elements, the spatial correlations of CTGE across regions exhibit multi-threaded and complex network characteristics [44]. Therefore, accurately identifying the spatial correlation network features of China’s CTGE and uncovering the roles and positions of different regions within the network hold significant guidance for formulating targeted regional transportation policies and addressing regional disparities.

(1) Establishing the “relationship” in space:

The core of network analysis lies in identifying and defining a “relationship” [45]. In this study, the foundation of social network analysis is the spatial correlation of CTGE in China, represented as a network-type correlation progressively formed by the mutual conduction and interaction of various regions in space. This correlation encapsulates the inter-regional relationships of CTGE across diverse regions. In the spatial correlation network, the mutual conduction of CTGE is depicted as “lines”, while regions are represented as “points”. Together, these points and lines form the spatial correlation network of China’s CTGE.

In existing research, methods to determine inter-element relationships primarily include gravity models (spatial distance-based perspective), Granger causality tests within the vector autoregression (VAR) framework (causal relationship perspective), the spatial autoregressive (SAR) model, and the geographically weighted regression (GWR) model. After comparative analysis, we selected the second approach (VAR-based Granger causality) for the following reasons:

Firstly, for the methodological alignment with research objectives. Our goal was to identify directional causal relationships (e.g., whether province A’s CTGE improvements systematically precede or influence province B’s) rather than static spatial spillovers. Granger causality, embedded in the vector autoregression (VAR) framework, is uniquely suited for this purpose as it quantifies lead–lag dynamics without requiring pre-specified spatial weight matrices [46]. In contrast, SAR/GWR models primarily capture contemporaneous spatial correlations (e.g., adjacency effects) but struggle to disentangle temporal causality chains, which are critical for policy sequencing in transportation decarbonization.

Secondly, the limitations of gravity models in efficiency contexts. While gravity models (distance-based) are widely used to measure spatial linkages (e.g., trade flows), they rely on simplified assumptions (e.g., bilateral distance inversely proportional to interaction intensity) that poorly align with the multi-parametric nature of CTGE [42]. As CTGE is a composite metric derived from input–output optimization, its inter-regional linkages involve nonlinear, non-distance-dependent mechanisms (e.g., technology diffusion, policy mimicry) which gravity models cannot adequately capture.

Thirdly, the advantages of VAR-based Granger causality. The VAR framework allows endogenous determination of interaction directions and magnitudes across multiple regions simultaneously, avoiding the subjectivity of defining spatial weights (e.g., adjacency, inverse distance) [43]. By constructing a causality network (rather than a correlation matrix), we derive policy-relevant insights into who drives whom in the CTGE system—key for targeting nodal regions in coordinated governance.

In this context, this study constructs an inter-provincial CTGE spatial correlation network in China using the VAR-based Granger causality test to identify spatial conduction relationships of CTGE across various regions. The calculation formula is as follows:

$$x_t={\alpha }_1+{\displaystyle \sum _{i=1}^m{\beta }_{1,i}}{x}_{t-i}+{\displaystyle \sum _{i=1}^n{\gamma }_{1,i}}{y}_{t-i}+{\epsilon }_{1,t}$$

(5)

$$y_t={\alpha }_2+{\displaystyle \sum _{i=1}^p{\beta }_{2,i}}{x}_{t-i}+{\displaystyle \sum _{i=1}^q{\gamma }_{2,i}}{y}_{t-i}+{\epsilon }_{2,t}$$

(6)

where the time series of comprehensive transportation green efficiency in two regions are represented by the formula’s variables x_t and y_t, respectively, the parameters to be estimated are α_i, β_i, and γ_i (i = 1, 2), while the residual term is ε_i,t (i = 1, 2), and the lag orders of the autoregressive terms are m, n, p, and q. To address potential omitted variable bias and reverse causality, we conducted rigorous tests on data stationarity and model robustness. The specific procedures are as follows: Firstly, stationarity and unit root tests are performed. To ensure data stability and avoid spurious regression, we conducted unit root tests using three methods in EViews 8.0, the Levin–Lin–Chu (LLC) test (common unit root), the Im–Pesaran–Shin (IPS) test (individual unit root), and the ADF-Fisher test. Results show that the first-difference series of CTGE for all regions passed stationarity tests at the 1% significance level, confirming that the data are I(1) (integrated of order 1) and suitable for further analysis. Secondly, optimal lag selection is performed. The lag order for Granger causality tests was determined by minimizing five criteria, the likelihood ratio (LR), the final prediction error (FPE), the Akaike information criterion (AIC), the Schwarz criterion (SC), and the Hannan–Quinn criterion (HQ). Relationships between regions were retained only if the F-statistic was significant at the 10% level, ensuring robustness against overfitting. Finally, cointegration analysis was performed. We employed the Johansen cointegration test to verify long-term equilibrium relationships among regional CTGE values. Trace statistic and max-eigenvalue tests rejected the null hypothesis of no cointegration (p < 0.05), confirming stable interdependencies across regions.

(2) Description of the aspects of the network structure

Four variables are employed in this research to quantify the overall structural characteristics of the CTGE spatial correlation network, which are network density, network correlation, network hierarchy, and network efficiency [47]. Network density gauges the spatial correlation within a network. A higher density value implies closer spatial linkages. The network correlation degree functions as a means to assess the robustness of the spatial correlation network. A correlation degree of 1 signifies that all regions in the network are interconnected, indicating a robust network structure, while a value less than 1 implies a lack of robustness. The hierarchical structure of different network nodes is represented by network hierarchy, with higher hierarchy levels reflecting greater regional disparities. Network efficiency, on the other hand, is determined by the number of related channels connecting various locations in the network. Higher network efficiency facilitates the spatial flow of transportation resources, thereby reducing regional inequalities and enhancing the connectivity and coordination of the transportation system. The following provides the calculation formula:

$$GD={\displaystyle \frac{a}{b(b-1)/2}}$$

(7)

$$GC=1-\left[{\displaystyle \frac{V}{b(b-1)/2}}\right]$$

(8)

$$GH=1-{\displaystyle \frac{S}{\max (S)}}$$

(9)

$$GE=1-{\displaystyle \frac{K}{\max (K)}}$$

(10)

where GD, GC, GH, and GE stand for network density, network correlation, network hierarchy, and network efficiency, respectively, a is the actual relationship number of the spatial relationship matrix, b is the number of provinces, and b(b − 1)/2 is the maximum correlation number between any two regions in theory. The total number of “0” above the reachable matrix’s diagonal is denoted by V. The actual redundant relation number is K. The maximum possible redundant relation number is max(K). S is the logarithm of the symmetric reachable relation in the spatial incidence matrix and the relational number of region X reachable region Y or region Y reachable region X is max(S).

Three variables, namely degree centrality, closeness centrality, and betweenness centrality, are utilized to quantify the individual characteristics of the spatial correlation network of China’s CTGE [48]. Degree centrality reflects a region’s position within the network by indicating the number of connections it has with other regions. A higher degree centrality signifies that the region occupies a more central or core position in the network. Closeness centrality measures the extent to which a region is “not controlled by other regions”. It reflects how close a region is to others in the network and the ease with which it can establish direct connections. A region with higher closeness centrality is better positioned to access or interact with other regions in the network. Betweenness centrality evaluates a region’s capacity to influence or control interactions between other regions. This measure highlights a region’s role as a “bridge” or intermediary, emphasizing its influence in facilitating or mediating relationships within the spatial correlation network. The calculation formula is as follows:

$$C_{RD}(i)=d_i/(b-1)$$

(11)

$$C_i^{-1}={\displaystyle \sum _{j=1}^b{d}_{ij}}$$

(12)

$$C_{Bi}={\displaystyle \frac{2{\displaystyle \sum _j^b{\displaystyle \sum _k^b{n}_{jk}\left(i\right)}}}{b^2-3b+2}},i\ne j\ne k,j\prec k$$

(13)

where the degree centrality, closeness centrality, and betweenness centrality of the i region are represented by the letters C_RD(i), $C_i^{-1}$, and C_Bi in the formula, the number of regions is b, d_i is the total number of connections that directly connect region i to another region where d_ij, or the number of lines in the shortcut, is the shortcut distance between region i and j, and g_jk is the number of shortcuts that connect the j and k areas, while g_jk(i) is the number of shortcuts that cross the i area. A measure of how much the i region influences the relationship between the j and k regions is given by n_jk(i), where n_jk(i) = g_jk(i)/g_jk.

(3) Block model analysis

Block model analysis is often used to identify and analyze the agglomeration phenomena of various elements (nodes) within spatial correlation networks, along with the interactions among elements or nodes both within and across blocks. This method serves as an essential tool for spatial clustering analysis in social network research [49]. Based on findings from previous studies, this research categorizes the blocks within the spatial correlation network of comprehensive transportation green efficiency (CTGE) into the following four types: main inflow block, main outflow block, bidirectional spillover block, and agent block. To gain a comprehensive understanding of the characteristics and implications of these block types, readers are advised to consult the works of Liu and He [46].

3.3. Indicator Selection and Data Source

This study focuses on 30 Chinese provinces (municipalities and autonomous regions), excluding Tibet, Hong Kong, Macau, and Taiwan. The research period covers the years 2003 to 2020. Based on insights from earlier studies [1,32], the input indicators selected include the operating mileage of the transportation network (encompassing roads, railways, waterways, and pipelines), the capital stock of the transportation industry, the number of employees, and the amount of energy consumed. These indicators correspond to facility investment, capital investment, labor input, and energy input, respectively. Capital stock is calculated using the perpetual inventory method, with 2003 as the base year, following the approach outlined by Li and Zhang [50]. Energy input is transformed using the standard coal conversion coefficient provided in the China Energy Statistical Yearbook [51]. Output indicators include the transportation industry’s added value, carbon emissions, and a social development index, which collectively represent the sector’s economic, environmental, and social outputs. Economic output is converted into constant prices based on 2003 values. Carbon emissions are calculated using a “top-down” approach [52]. The social dimension index system comprises the following four components: living standards, degree of urbanization, transportation development, and level of science and education. The entropy method is employed to derive the overall score for the social development index. For additional details, please refer to the works of Ma et al. [1,32]. All data in this study are sourced from the China Statistical Yearbook, China Energy Statistical Yearbook, and statistical bulletins from the transportation industry. A small number of missing data points were obtained using the interpolation method. For the specific indicator composition and descriptive statistics of the data, please refer to

Table 1. Index system and descriptive statistics of data.

Perspective	Primary Index	Secondary Index	Units	Obs	Mean	Std.Dev.	Min	Max
Inputs	Infrastructure	Total mileage of transportation network	10,000 km	540	13.586	8.028	0.9	41.056
	Capital	Capital stock of transportation industry	100 million	540	4488.417	4520.664	169.387	26,590.23
	Labor force	Transportation industry employees	person	540	244,801.8	14,604.5	32,225	83,1000
	Energy consumption	Energy consumption of transportation industry	10,000 tons of standard coal	540	919.1	646.937	27.054	3720.54
Outputs	Expect outputs	Added value of transportation industry	100 million	540	865.613	757.154	27.64	4189.02
		Social development index	--	540	0.438	0.101	0.227	0.715
	Undesired output	CO2 emission from transportation industry	10,000 tons	540	1850.498	1304.055	60.910	6974.148

.

4. Results and Discussion

4.1. Analysis of China’s CTGE Measurement Results Based on the SBM-DEA Model

The MaxDEA system was employed to evaluate the CTGE of 30 regions in China from 2003 to 2020, utilizing the unexpected output SBM model (see

Table 2. Measurement results of comprehensive transport green efficiency.

Regions	2003	2005	2007	2009	2011	2013	2015	2017	2020	Average
Beijing	1.000	1.000	0.562	0.494	0.547	0.558	0.814	1.000	1.000	0.774
Tianjin	1.000	1.000	1.000	1.000	1.000	1.000	1.000	1.000	1.000	1.000
Hebei	1.000	1.000	1.000	1.000	1.000	1.000	1.000	1.000	1.000	1.000
Shanxi	0.590	0.618	0.704	0.401	0.362	0.373	0.415	0.464	1.000	0.565
Inner Mongolia	0.643	0.399	0.394	0.425	0.361	0.430	0.453	0.394	0.466	0.444
Liaoning	0.447	0.372	0.331	0.344	0.368	0.424	0.505	0.537	0.541	0.439
Jilin	1.000	0.559	0.458	0.448	0.423	0.447	0.439	0.448	0.453	0.514
Heilongjiang	0.447	0.444	0.411	0.388	0.306	0.307	0.357	0.362	0.354	0.379
Shanghai	1.000	1.000	1.000	1.000	1.000	1.000	1.000	1.000	1.000	1.000
Jiangsu	0.429	0.442	0.445	0.486	0.530	0.665	1.000	1.000	1.000	0.693
Zhejiang	0.729	0.456	0.458	0.410	0.412	0.435	0.457	0.477	0.492	0.476
Anhui	0.551	0.658	0.443	0.461	0.380	0.356	0.373	0.352	0.356	0.433
Fujian	1.000	1.000	1.000	0.556	0.475	0.547	0.560	0.643	0.659	0.698
Jiangxi	0.436	0.527	0.494	0.459	0.419	0.484	0.466	0.527	0.638	0.508
Shandong	0.636	1.000	1.000	1.000	1.000	0.527	0.577	0.558	0.457	0.782
Henan	0.677	0.681	0.522	0.381	0.300	0.351	0.442	0.501	0.392	0.466
Hubei	0.299	0.245	0.241	0.253	0.254	0.284	0.294	0.277	0.268	0.269
Hunan	0.371	0.352	0.372	0.365	0.296	0.329	0.329	0.355	0.338	0.349
Guangdong	0.368	0.300	0.308	0.301	0.289	0.303	0.389	0.551	0.355	0.358
Guangxi	0.403	0.343	0.310	0.342	0.353	0.403	0.399	0.398	0.421	0.376
Hainan	0.686	1.000	0.764	1.000	0.587	0.637	0.594	0.591	0.570	0.687
Chongqing	0.650	0.474	0.392	0.405	0.344	0.349	0.351	0.358	0.357	0.399
Sichuan	0.289	0.326	0.299	0.227	0.228	0.260	0.275	0.238	0.251	0.266
Guizhou	0.341	0.434	0.521	0.502	0.456	0.553	0.581	0.572	1.000	0.560
Yunnan	0.186	0.185	0.199	0.159	0.140	0.160	0.164	0.165	0.158	0.177
Shaanxi	0.373	0.360	0.337	0.308	0.291	0.336	0.366	0.402	0.392	0.357
Gansu	0.398	0.412	0.425	0.424	0.426	0.329	0.322	0.303	0.294	0.374
Qinghai	1.000	1.000	1.000	1.000	1.000	1.000	1.000	0.498	0.448	0.882
Ningxia	1.000	1.000	1.000	1.000	1.000	1.000	1.000	1.000	1.000	1.000
Xinjiang	0.346	0.313	0.307	0.307	0.269	0.293	0.351	0.385	0.369	0.336
Eastern	0.754	0.779	0.715	0.670	0.655	0.685	0.718	0.760	0.754	0.722
Central	0.546	0.510	0.456	0.394	0.342	0.366	0.389	0.411	0.475	0.434
Western	0.512	0.477	0.471	0.464	0.442	0.465	0.478	0.489	0.489	0.478
National	0.610	0.597	0.557	0.528	0.494	0.505	0.542	0.565	0.568	0.555

). A time trend chart was generated to illustrate the temporal variation in CTGE (see Figure 2). Each region’s CTGE was classified into one of the following five efficiency levels: low efficiency (0.001–0.200), lower efficiency (0.201–0.400), medium efficiency (0.401–0.600), higher efficiency (0.601–0.999), and high efficiency (1). To further visualize the measurement results for the years 2003, 2010, 2015, and 2020, a spatial distribution map of China’s CTGE was developed using ArcGIS 10.2 software (see Figure 3). These results facilitated a comprehensive analysis of the spatial and temporal evolution characteristics of CTGE across China.

According to the average CTGE values of each region (See Table 2), the top ten regions in the comprehensive transportation green efficiency (CTGE) rankings are Tianjin, Hebei, Shanghai, Ningxia, Qinghai, Shandong, Beijing, Fujian, Jiangsu, and Hainan. Spatially, these regions are primarily distributed in the eastern and western parts of China, with Ningxia and Qinghai ranking notably high. The reasons for this may lie in the fact that, compared to the more developed regions in the eastern and central areas, Ningxia and Qinghai have fewer input factors in their transportation sectors. Although their economic outputs are lower, their undesirable outputs (e.g., carbon emissions) are significantly lower than those of regions with advanced transportation industries. Furthermore, the social development indices of Ningxia and Qinghai are comparable to those of most central regions. Their lower transportation resource inputs, minimal undesirable outputs, and relatively high social welfare outputs collectively result in higher CTGE levels than many economically developed regions. This suggests that the level of comprehensive transportation green efficiency is not solely determined by regional economic development but is also influenced by improvements in environmental protection and social development indices.

Overall, as shown in Figure 2, the CTGE across different regions of China exhibited a pattern of initial decline followed by growth between 2003 and 2020. Specifically, CTGE fell and fluctuated between 2003 and 2010, then rose consistently after 2010. The eastern region demonstrated the highest CTGE during the study period, with an average value of 0.722, followed by the western and central regions, with averages of 0.478 and 0.434, respectively. The eastern region’s superior performance can be attributed to its leading role in China’s social and economic development, bolstered by significant advantages in “hard” factors such as location, investment levels, infrastructure, and technological innovation, as well as “soft” factors like policy implementation, economic benefits, environmental awareness, and social welfare. In contrast, the central region, despite its strong economic growth following the implementation of the Central Plains Rising Strategy in 2006, ranked last. This was largely due to an overemphasis on speed and quantity in transportation development, which came at the expense of efficiency and quality, leading to a mismatch between social and economic growth. Consequently, CTGE in the central region fell sharply, even lagging behind the western region. While it displayed a consistent upward trend after 2010, the earlier efficiency losses were not fully offset by subsequent growth.

As Figure 3 illustrates, eight of China’s thirty regions—Beijing, Tianjin, Hebei, Jilin, Shanghai, Fujian, Qinghai, and Ningxia—achieved a CTGE of 1 in 2003. Seven regions fell within the medium efficiency range (0.401–0.600), while Inner Mongolia, Zhejiang, Shandong, Henan, Hainan, and Chongqing fell within the higher efficiency range (0.601–0.999). The remaining regions were distributed between the low (0.001–0.200) and lower (0.201–0.400) efficiency ranges. By 2010, CTGE had decreased significantly compared to 2003, with the number of regions achieving an efficiency value of 1 dropping to six. No regions remained in the higher efficiency range (0.601–0.999), while the number of regions in the lower and medium efficiency ranges increased substantially. In 2015, Beijing advanced to the higher efficiency range (0.601–0.999), while the number of regions with a CTGE of 1 remained unchanged. During this period, the number of regions in the medium efficiency range (0.401–0.600) increased, while the number in the lower efficiency range declined, indicating an overall improvement in CTGE. By 2020, the number of regions with a CTGE of 1 increased to eight, with Jiangxi and Fujian moving from the medium efficiency range to the higher efficiency range. Other regions exhibited little change, suggesting that CTGE in China followed a trend of initial decline followed by subsequent growth during the study period.

4.2. Spatiotemporal Evolution of China’s CTGE Using the SDE Model

Transportation is simultaneously an economic and a social phenomenon. The concept of a gravity center essentially embodies the average spatial distribution center of economic phenomena. In the context of comprehensive transportation green efficiency (CTGE), the gravity center reflects the spatial distribution center of the coordinated development of the economic, social, and environmental systems within the transportation industry. Changes in the gravity center indicate shifts in the spatial coordination of these systems. Based on the CTGE measurement results, ArcGIS 10.2 software was employed to calculate the parameters of the barycenter-standard deviation ellipse and the spatial position transfer path (see

Table 3. Center of gravity moving direction and distance of China’s CTGE.

Year	Barycentric Coordinates	Moving Direction	Moving Distance (km)	East–West Distance (km)	North–South Distance (km)	Moving Speed (km/a)	East–West Speed (km/a)	North–South Speed (km/a)
2003	113.27° E, 34.62° N
2010	112.71° E, 34.45° N	S-W 16.842°	64.87	62.09	18.80	12.97	12.42	3.76
2015	112.59° E, 34.25° N	S-W 59.602°	25.82	13.07	22.27	5.16	2.61	4.45
2020	113.47° E, 34.31° N	N-E 4.123°	97.82	97.57	7.03	19.56	19.51	1.41

and

Table 4. Standard deviation ellipsometric parameters of CTGE in China.

Year	Rotation θ/°	Area/10,000 km2	XStdDist/km	YStdDist/km	Oblateness
2003	21.54	347.930	988.112	1120.876	0.882
2010	24.05	334.411	993.280	1074.925	0.924
2015	24.56	351.987	1016.878	1101.871	0.923
2020	13.38	308.151	928.858	1056.056	0.880

Note: The oblateness is the ratio of the short half-axis to the long half-axis of the ellipse, ranging from 0 to 1.

, and Figure 4), enabling an analysis of the temporal and spatial evolution characteristics of CTGE in China.

4.2.1. Analysis of CTGE Gravity Center Transfer Route

The results indicate that the distribution of the CTGE gravity center in China (Table 3 and Figure 4) ranged between 112.59° E and 113.47° E in longitude and between 34.25° N and 34.62° N in latitude. Compared to the geometric center of China (103.83° E, 36° N), the east–west offset initially decreased from 10.19° in 2003 to 9.51° in 2015, before increasing to 10.39° in 2020. In contrast, the north–south offset increased from 1.38° in 2003 to 1.75° in 2015, before slightly declining to 1.69° in 2020. Regarding the moving path, the CTGE gravity center was located in Zhengzhou, Henan Province, in 2003, shifted to Luoyang in 2010, moved further southwest to Pingdingshan in 2015, and then northeast to Xuchang in 2020. The overall trajectory reflects a shift from southwest (2003–2015) to northeast (2015–2020). The moving distance of the gravity center exhibited a decreasing trend from 2003 to 2015, with the north–south displacement being smaller than the east–west displacement. However, between 2015 and 2020, the moving distance increased, driven predominantly by changes in the east–west direction. The moving speed of the gravity center also followed a dynamic pattern, first decreasing and then increasing. Between 2003 and 2015, the east–west moving speed declined sharply from 12.42 km/a to 2.61 km/a, representing a nearly five-fold decrease, while the north–south moving speed rose from 3.76 km/a to 4.45 km/a. From 2015 to 2020, the moving speeds in the two directions exhibited contrasting trends, with the east–west speed increasing significantly and the north–south speed declining.

4.2.2. Standard Deviation Ellipse Analysis of CTGE

As shown in Table 4, the area of the standard deviation ellipse decreased across the characteristic years, indicating an increasingly concentrated spatial distribution pattern of CTGE in China. Regarding the oblateness of the ellipse, it increased from 2003 to 2015 but declined in 2020, deviating further from a perfect circle. This trend mirrors the increasing disparities in the spatial distribution of CTGE along the east–west and north–south axes.

During the early stage of the study (2003–2010), the long half-axis decreased from 1120.876 km to 1074.925 km, while the short half-axis increased from 988.112 km to 993.280 km, resulting in a more pronounced oblateness. In the middle stage (2010–2015), both axes increased at comparable rates, leading to minimal changes in oblateness. However, in the later stage (2015–2020), both axes shortened, with the short half-axis experiencing a more significant reduction than the long half-axis, causing the oblateness to decline and further deviate from a perfect circle.

In terms of the rotation angle (θ) of the standard deviation ellipse, the range of variation was substantial, fluctuating between 13.38° and 24.56°. Overall, θ initially increased before decreasing. Between 2003 and 2015, the CTGE gravity center shifted southwest, and θ rose from 21.54° to 24.56°, reflecting a northeast–southwest orientation in the spatial distribution pattern. From 2015 to 2020, θ dropped significantly from 24.56° to 13.38°, a more pronounced change than the earlier increase. This significant decline signals a transformation in the spatial distribution pattern of CTGE. It has shifted from an orientation mainly along the northeast–southwest axis to one predominantly following the north–south axis.

4.3. Network Correlation Characteristics of China’s CTGE via the SNA Model

4.3.1. The Overall Features of the Network Structure

The spatial correlation matrix for China’s inter-provincial CTGE was constructed using the VAR-based Granger causality test, and the resulting spatial correlation network for 30 regions was visualized using Gephi software 0.9.5 (see Figure 5). This network highlights the interdependencies and interactions between provinces, offering insights into the structural relationships and the overall connectivity of CTGE across China.

China’s CTGE spatial correlation network consists of 30 provinces as nodes, with the highest correlation coefficient between nodes reaching 870. Based on the Granger causality test, 309 actual relationships (significant at the 10% level) were identified, resulting in an overall network density of 0.355. This suggests substantial spatial correlation in China’s CTGE during the period from 2003 to 2020, albeit with a relatively low degree of connection closeness. The network correlation degree is 1, demonstrating that all provinces are part of the CTGE spatial correlation network, with no “island” phenomena, as all provinces are directly or indirectly connected. The network hierarchy is high, with a hierarchical value of 0.632, reflecting a clear “core-edge” structure and notable disparities in CTGE performance among provinces, highlighting the need for further optimization of the network structure. However, the network efficiency, calculated at 0.310, suggests that the spatial correlation network lacks stability due to limited redundant connections and the absence of a clear superposition effect.

4.3.2. The Individual Features of the Network Structure

The role and position of CTGEs within each province’s spatial correlation network were analyzed by measuring the individual network characteristics of 30 Chinese regions (see

Table 5. Centrality of China’s inter-provincial CTGE spatial correlation network.

Regions	Degree Centrality				Closeness Centrality		Betweenness Centrality
	Out-Degree	In-Degree	Degree	Ranking	Closeness	Ranking	Betweenness	Ranking
Beijing	20	18	68.966	2	74.359	2	17.891	1
Tianjin	19	3	65.517	4	70.732	4	1.156	20
Hebei	20	3	68.966	3	74.359	3	1.616	15
Shanxi	6	7	20.690	24	47.541	25	0.556	26
Inner Mongolia	6	15	20.690	25	50.000	23	2.983	14
Liaoning	8	22	27.586	17	52.727	18	3.482	10
Jilin	10	8	34.483	12	54.717	16	0.896	23
Heilongjiang	9	7	31.034	15	54.717	14	0.319	30
Shanghai	21	1	72.414	1	76.316	1	3.473	11
Jiangsu	17	2	58.621	5	67.442	5	4.394	6
Zhejiang	15	10	51.724	6	61.702	7	5.865	4
Anhui	10	12	34.483	13	54.717	15	1.553	16
Fujian	14	3	48.276	7	59.184	9	0.935	22
Jiangxi	5	20	17.241	27	43.284	29	4.205	7
Shandong	13	2	44.828	8	63.043	6	0.621	25
Henan	8	10	27.586	18	51.786	20	8.157	2
Hubei	7	20	24.138	20	50.877	21	6.572	3
Hunan	6	8	20.690	23	48.333	24	3.742	8
Guangdong	12	6	41.379	9	56.769	12	2.998	13
Guangxi	8	21	27.586	19	53.704	17	0.320	29
Hainan	11	14	37.931	10	59.184	10	3.687	9
Chongqing	10	7	34.483	14	52.727	19	3.334	12
Sichuan	11	6	37.931	11	58.000	11	1.343	17
Guizhou	7	7	24.138	21	46.774	26	0.951	21
Yunnan	5	13	17.241	28	56.769	13	1.331	18
Shaanxi	7	22	24.138	22	50.877	22	4.742	5
Gansu	9	8	31.034	16	60.417	8	1.241	19
Qinghai	6	5	20.690	26	46.032	27	0.385	27
Ningxia	5	6	17.241	29	43.284	30	0.876	24
Xinjiang	4	23	13.793	30	44.615	28	0.364	28
Average value	10.3	10.3	35.517	-	56.099	-	2.951	-

). To further illustrate the individual characteristics of the CTGE spatial correlation network, a spillover and reception relationship diagram for China’s provincial CTGE spatial correlation network was developed (see Figure 6).

(1) Variations in the quantity of connections: As shown in Table 5, 11 provinces had spillover relations exceeding the average of 10.3 during the study period. Shanghai, Beijing, Hebei, Tianjin, and Jiangsu ranked highest, whereas Xinjiang, Ningxia, Qinghai, and Yunnan had the fewest. Regarding receiving relationships, Shanghai, Jiangsu, Shandong, Hebei, and Tianjin received fewer connections than regions such as Xinjiang, Shaanxi, Guangxi, and Hubei. Figure 6 indicates 15 regions with positive net spillover relations, including Shanghai, Hebei, Tianjin, Jiangsu, Fujian, and Shandong, which had more than 10 net spillover relations. Conversely, 14 regions exhibited negative net spillover relations, with Xinjiang, Shaanxi, and Jiangxi having fewer than −10. On the whole, spillover effects were predominantly manifested in the eastern region, whereas net benefit effects were the prevailing factor in the central and western regions.

(2) Degree centrality: The average degree centrality was 35.517, with regions such as Shanghai, Beijing, Hebei, Tianjin, Jiangsu, Shandong, and Zhejiang exceeding this value. Most of these provinces are located in the eastern region, where higher out-degrees and lower in-degrees indicate strong spatial connections and spillover impacts. These areas, characterized by developed economies, advanced transportation systems, and cutting-edge technology, influence less-developed regions by contributing production factors such as capital and technology. For instance, infrastructure investments are crucial for spillovers. In the Yangtze River Delta, large-scale investments create an efficient transportation network [53]. High-speed railways and expressways enhance internal efficiency and connectivity, facilitating the spread of advanced technologies and management. Although cross-regional spillover of public infrastructure investment can be limited, it generally promotes economic growth and resource allocation [54]. Technology diffusion promotes CTGE network spillover. Advanced transportation technologies, such as intelligent traffic systems, can be transferred between regions. For example, developed coastal cities share their intelligent traffic control systems with central and western cities [55], improving traffic management and CTGE in recipient regions. Research on technology infrastructure investment shows it impacts neighboring regions’ innovation, highlighting technology diffusion’s role in enhancing CTGE and spillover effects. In contrast, provinces like Xinjiang, Ningxia, Yunnan, Qinghai, and Jiangxi had centrality degree values below the national average, indicating weaker connections. These regions are primarily located in western China, where limited economic development and geographic isolation contribute to weaker linkages with other provinces. Future efforts should focus on enhancing CTGE spatial spillover pathways by improving their connections to other locations.

(3) Closeness centrality: Thirteen regions, including Shanghai, Beijing, Hebei, Tianjin, Jiangsu, Zhejiang, Gansu, Fujian, Shandong, Hainan, Sichuan, Guangdong, and Yunnan, had closeness centrality values above the national average of 56.099. These regions act as “central actors” in the CTGE spatial correlation network, characterized by their ability to quickly connect to other locations and their relatively short network distances. Most of these regions are clustered in the southeastern coastal areas of China. There, highly efficient transportation systems and the swift circulation of production factors strengthen their links with other provinces. This, in turn, promotes a high level of correlation efficiency and gives rise to a self-reinforcing mechanism for CTGE. Conversely, 17 regions, including Ningxia, Xinjiang, Qinghai, Jiangxi, and Guizhou, had closeness centrality below the national average, indicating their peripheral positions in the network due to geographic constraints.

(4) Betweenness centrality: The average betweenness centrality was 2.951. Thirteen regions, including Beijing, Henan, Hubei, and Zhejiang, had higher-than-average values, signifying their roles as “intermediaries” or “bridges” within the network. These provinces are critical nodes, facilitating connections among other provinces and exerting significant influence in shaping the CTGE spatial correlation network. Their economic development levels are relatively high, and many, such as Beijing, Zhejiang, and Jiangsu, are located along major east–west transportation corridors. These regions occupy pivotal strategic positions in aspects such as logistics, financial services, scientific and technological innovation, and management capabilities. On the other hand, 17 regions, including Heilongjiang, Guangxi, Xinjiang, and Qinghai, exhibited below-average betweenness centrality. These regions are predominantly located in remote areas with lower economic development, limiting their influence and control within the CTGE spatial correlation network.

4.3.3. Block Model Analysis

The block model of China’s inter-provincial CTGE spatial correlation network was analyzed using Ucinet 6 software, based on the spatial relationship matrix of China’s inter-provincial CTGE. This analysis reveals the internal structural features of the spatial correlation network. To further examine the transfer patterns between CTGE blocks, a spatial correlation diagram was constructed, and the density matrix and image matrix of China’s inter-provincial CTGE spatial correlation blocks were generated using principles commonly applied in current research. The results are presented in

Table 6. Spillover effects of the spatial correlation block of CTGE in China.

Block	Accept Relationship Matrix				Receive Relationship		Spillover Relationship		Expected Internal Relationship Ratio/%	Actual Internal Relationship Ration/%
	(I)	(II)	(III)	(IV)	Inside the Block	Outside the Block	Inside the Block	Outside the Block
(I)	19	45	14	35	19	33	19	94	20.68	16.81
(II)	10	32	4	62	32	73	32	76	34.48	29.63
(III)	12	14	4	23	4	24	4	49	10.34	7.55
(IV)	11	14	6	8	8	120	8	31	24.13	20.51

and

Table 7. Density and image matrix of the spatial correlation block of CTGE in China.

Block	Density Matrix				Image Matrix
	(I)	(II)	(III)	(IV)	(I)	(II)	(III)	(IV)
(I)	0.392	0.228	0.593	0.852	1	0	1	1
(II)	0.393	0.435	0.311	0.861	1	1	0	1
(III)	0.222	0.056	0.333	0.667	1	0	0	1
(IV)	0.204	0.194	0.333	0.267	0	0	0	0

and in Figure 7.

The division results reveal that Block (I) includes seven regions, Tianjin, Hebei, Guangdong, Shandong, Fujian, Jiangsu, and Shanghai, while Block (II) comprises eleven regions, including Beijing, Zhejiang, Jilin, Anhui, Hainan, Heilongjiang, Henan, Hunan, Sichuan, Chongqing, and Gansu. Block (III) consists of Qinghai, Ningxia, Guizhou, and Shanxi, and Block (IV) is made up of the following eight regions: Inner Mongolia, Hubei, Shaanxi, Guangxi, Liaoning, Jiangxi, Yunnan, and Xinjiang. In the spatial correlation network of China’s inter-provincial CTGE, as shown in Table 6, there are 246 inter-block correlations, accounting for 79.61% of the total relationships, and 63 intra-block correlations, accounting for 20.39%. These findings suggest that spatial spillover effects exert a substantial influence within the network. Among these, Block (I) comprises regions such as Guangdong, Hebei, and Tianjin, with 19 internal relationships, 33 spillover relationships received from Blocks (II), (III), and (IV), and 94 relationships sent to other blocks. The actual internal relationship ratio is 16.81%, lower than the expected ratio of 20.68%. Due to its significantly higher number of outgoing relationships compared to incoming ones, Block (I) is classified as the “Main Outflow Block”. Block (II) includes 11 provinces and cities, such as Sichuan, Chongqing, Gansu, Beijing, Zhejiang, Jilin, Anhui, Hainan, Heilongjiang, Henan, and Hunan. It has 32 internal relationships, receives 45, 14, and 14 relationships from Blocks (Ⅰ), (III), and (IV), respectively, and sends 10, 4, and 62 relationships to those blocks. The actual internal relationship ratio is 29.63%, lower than the expected ratio of 34.48%. Thus, Block (II) is identified as a “Bidirectional Spillover Block”. Block (III), composed of Qinghai, Ningxia, Guizhou, and Shanxi, has 4 internal relationships, receives 24 spillover relationships from other blocks, and sends 49 relationships to other blocks. As the number of incoming and outgoing relationships is relatively balanced, Block (III) is designated as the “Agent Block”. Block (Ⅳ) comprises eight regions with 8 internal relationships, 120 spillover relationships received from other blocks, and 31 relationships sent to other blocks. The actual internal relationship ratio is 20.51%, lower than the expected ratio of 24.13%. Accordingly, Block (IV) is categorized as the “Main Inflow Block.” In summary, Block (Ⅰ) consists of regions primarily located in eastern China, while Block (Ⅳ) is largely composed of western regions. Blocks (II) and (III) are mostly located in central and western China. The transmission relationships between these blocks exhibit distinct “gradient” features within the CTGE spatial correlation network. Block (Ⅰ) not only maintains internal relationships but also primarily directs spillover effects to Blocks (Ⅱ) and (Ⅳ), acting as the “engine” that transfers CTGE growth momentum. Block (Ⅲ) serves as an “Agent Block”, mediating and transferring CTGE growth momentum from Block (Ⅰ) to Block (Ⅱ) via bidirectional spillover effects and to Block (Ⅳ) via net benefit effects. Block (Ⅳ) primarily receives spillover effects from the other three blocks, functioning as a “follower” within the spatial correlation network.

5. Conclusions and Policy Implications

5.1. Conclusions

This paper makes use of the SBM model, the standard deviation ellipse model, as well as social network analysis methods to conduct an in-depth exploration of the characteristics of the spatial correlation network of CTGE in China spanning from 2003 to 2020. The findings reveal that CTGE across various regions in China initially declined before rising. The overall development level remains low, and the spatial distribution pattern is notably imbalanced, with the eastern region exhibiting the highest CTGE, followed by the western region, while the central region lags behind.

The spatial trajectory of China’s CTGE gravity center shifted from the southwest to the northeast, indicating a dramatic decline in CTGE in the northeastern regions during the early stages, followed by significant growth in the eastern regions in later stages. The standard deviation ellipse analysis shows that the spatial distribution of China’s CTGE is becoming more centralized.

The CTGE spatial correlation network in China demonstrates distinct “core-edge” structural features. The central and western regions are positioned at the network’s edge, primarily receiving spillover effects from other regions, while the eastern region functions as the network’s “engine”, exerting the most substantial spillover effects. The transmission process between blocks in China’s CTGE spatial correlation network exhibits clear gradient characteristics.

5.2. Policy Implications

Based on the above conclusions, we propose targeted recommendations to assist regions in improving their comprehensive transportation green efficiency (CTGE).

Firstly, according to the measurement results of the SBM model, China’s comprehensive transportation green efficiency (CTGE) shows an upward trend, yet the overall level remains relatively low (with an average value of 0.555). This is primarily attributed to the lack of horizontal coordination among various transportation modes, leading to irrational resource allocation and imbalanced development across these modes. This phenomenon resembles the “short-board effect”, where efficiency is constrained by the weakest link. Therefore, optimizing transportation structures and strengthening effective intermodal connectivity are critical to improving CTGE. Specific measures include promoting new energy vehicles (NEVs), where governments should prioritize the adoption of pure electric, hybrid, and hydrogen-powered vehicles to reduce fossil fuel consumption (e.g., gasoline and diesel) and mitigate carbon emissions, and shifting bulk freight to sustainable modes by accelerating the shift of bulk cargo transportation from road to waterway and rail networks. Innovative organizational strategies, such as intermodal transportation and trailer swap systems, should be implemented to optimize resource allocation at the input end, foster economic growth and social progress, and ultimately elevate CTGE.

Secondly, China’s CTGE has a complex network structure with a high degree of spatial correlation, indicating that no region operates in isolation from the network. This intricate web of connections means that the performance of one region can significantly impact others. For provinces located at the periphery of the CTGE network, their current position poses both challenges and opportunities. To enhance the spatial correlation of CTGE and foster coordinated growth among regions, effective strategies and measures must be actively explored. When developing strategies for CTGE improvement, regions should consider both “attribute data” and “relational data” to continuously optimize the spatial correlation network structure of provincial CTGEs, promoting synchronized CTGE development across all regions. For peripheral provinces, infrastructure development should be a top priority. These areas often suffer from underdeveloped transportation networks, which act as a bottleneck for economic growth and CTGE improvement. Governments at all levels, in collaboration with private enterprises, should increase the investment in building and upgrading transportation infrastructure. This includes constructing new highways, railways, and ports, as well as improving existing facilities to enhance their efficiency. For example, in some western provinces, the expansion of railway networks can better connect them to the central and eastern regions, facilitating the flow of goods and people. This improved connectivity will not only increase the region’s access to resources and markets but also enhance its attractiveness to investment, which in turn can drive the development of the transportation industry and improve CTGE.

Thirdly, the block model analysis shows that China’s CTGE can be divided into four distinct blocks, each demanding tailored promotion strategies. Provinces in the “Main Outflow Block” (e.g., Shanghai, Jiangsu, Zhejiang) and the “Bidirectional Spillover Block” (such as Beijing, Guangdong, Shandong) play a crucial “leading” role. To fully utilize their advantages in capital, technology, science, and management, these provinces should take specific actions. Regarding cross-regional cooperation, they can set up special cross-regional cooperation funds. For instance, allocating a portion of their annual financial revenue to support joint transportation projects with central and western regions, like building integrated logistics parks. They can also encourage large enterprises to invest in or cooperate with local firms in less-developed areas. They could offer tax incentives, such as a corporate income tax reduction for transportation-related investments in these regions, to expand spatial spillover channels and boost CTGE growth in the central and western regions. Provinces acting as “intermediaries” or “bridges” (e.g., Henan, Hubei, Anhui) should focus on optimizing their transportation structures and adopting advanced technologies and management from the “Main Outflow Block” and “Bidirectional Spillover Block”. To optimize transportation structures, they can invest in multimodal transportation hubs in major cities, integrating different transport modes. To adopt advanced practices, they can send delegations to learn from developed regions and then establish local technology transfer centers. Additionally, local enterprises could be encouraged to follow the management models of leading firms in developed areas and offer consulting and training services for improvement. Most western provinces fall into the “Main Inflow Block”. To reduce CTGE spatial imbalances, they should take several steps. They should seek financial support from the central government and attract investment from eastern and central regions. For example, the central government can create a special western transportation development fund, and western provinces can offer land-use and tax incentives. In terms of technology and management, they should establish talent-exchange programs with developed regions, in which they invite experienced professionals to the west and send local talent for training. Also, they should improve the business environment by simplifying administrative approval for transportation projects, strengthening investor protection, and enhancing public services. By narrowing the gap with the east, these provinces can shift from “passive beneficiaries” to “active contributors”, promoting balanced CTGE development nationwide.

5.3. Limitations

Data scope and precision: Due to the availability and accessibility of data, all datasets in this study were sourced from the National Bureau of Statistics, resulting in relatively rough calculations. Future research should enhance the comprehensiveness and accuracy of data by incorporating finer-grained or multi-source datasets. For instance, by leveraging the data from the transportation internet of things and applying the network data envelopment analysis (DEA) approach to perform efficiency allocation, a comprehensive comparative analysis could be carried out on the efficiencies of sectors including road, railway, waterway, and aviation. This in-depth analysis is capable of uncovering the disparities among diverse transportation modes. The insights gleaned could offer more profound and nuanced perspectives for the development of the transportation industry in different regions, facilitating more informed decision-making processes and strategic planning.

Research scale and validation: This study focuses on provincial-level units, which represent a larger research scale. The lack of primary data validation may limit the granularity of insights. Future work should refine the analysis to municipal-level units, supported by field surveys or monitoring data, to conduct case-specific evaluations of China’s CTGE, thereby improving the scientific rigor and credibility of findings.

Temporal coverage and policy generalizability: The data used in this study span from 2003 to 2020. While the latest data are still being collected, we plan to publish updated findings shortly. Additionally, although policy recommendations derived from the SBM and SNA models are proposed, their effectiveness may vary depending on regional resource endowments and network positions. Future research will integrate DEA and SNA frameworks to design hybrid policies that account for both efficiency benchmarks and network dynamics.