The Spatial Spillover Effects of Transportation Infrastructure on Regional Economic Growth—An Empirical Study at the Provincial Level in China

Abstract

This study examines the spatial spillover effects of transportation infrastructure on regional economic growth, utilizing panel data from 31 provincial-level administrative divisions in China from 2003 to 2022. Using the spatial Durbin model (SDM) and three distinct spatial weight matrices—0–1 adjacency, spatial economic–geographical nested, and GDP-based economic distance matrices—this study comprehensively analyzes the multifaceted impacts of transportation infrastructure. The results show that transportation infrastructure significantly promotes economic growth in both local and neighboring regions across all spatial weight matrices. The total effect is most pronounced in geographically proximate regions, with a coefficient of 7.845 (p < 0.01). Regions with similar economic development levels also show strong collaborative effects, with a coefficient of 2.074 (p < 0.01), although the marginal effect of transportation infrastructure diminishes. Furthermore, adjustments in industrial structure and innovation inputs demonstrate a short-term inhibitory effect on economic growth, highlighting the need for synchronized development of transportation infrastructure alongside industrial and innovation policies. This study incorporates multi-dimensional spatial weight matrices to systematically reveal the direct and indirect impacts of transportation infrastructure on regional economies, providing essential empirical support for regional coordination and infrastructure investment policies. The findings offer valuable insights for infrastructure planning in other regions, particularly in formulating policies that promote cross-regional economic cooperation and optimize resource allocation.

1. Introduction

Transportation infrastructure serves as a crucial driver of regional economic growth by reducing transportation costs, improving regional accessibility, and shortening travel times, directly fostering the expansion of economic activities and the flow of resources. With the acceleration of globalization and regional economic integration, the network characteristics, externalities, and diffusion effects of transportation infrastructure increasingly became core driving forces for regional economic growth. These features not only directly enhance economic activities by reducing transportation costs and shortening spatial distances, but also indirectly influence regional economic structures, the formation of industrial clusters, and the redistribution of production factors by promoting the flow of capital, labor, and goods between regions. Simultaneously, transportation infrastructure is not only a key factor in improving productivity and promoting regional economic growth, but also serves as a fundamental element in facilitating the flow of factors, industrial agglomeration, and economic structural adjustment. However, despite the extensive literature exploring the impact of infrastructure investment on economic growth, most studies still focus primarily on the direct effects of transportation infrastructure.

In recent years, the spatial spillover effects of transportation infrastructure on regional economic growth garnered increasing attention. These effects encompass not only the direct economic benefits brought about by infrastructure development, but also the radiating and spillover effects on neighboring regions. Spatial spillover is a significant feature of transportation infrastructure, where not only does it produce direct effects through economic activities within a region, but it also generates indirect impacts through interactions between neighboring regions. Existing research shows that the spatial spillover effects of transportation infrastructure on regional economic growth can vary significantly depending on factors such as the level of regional economic development, industrial structure, and geographic location. For example, developed regions, with well-established infrastructure and dense transportation networks, tend to generate stronger positive spillover effects, boosting economic vitality in neighboring areas. However, in some cases, uneven infrastructure provision may lead to imbalanced regional economic development and even negative spillover effects, exacerbating resource competition and economic disparities between regions.

Thus, exploring the mechanisms through which transportation infrastructure operates in different economic regions and understanding its multidimensional effects on regional economic growth became key issues in both theoretical research and policy formulation. Although numerous studies examined the relationship between transportation infrastructure and regional economic growth, most existing literature remains limited by short time spans and single economic indicators, such as GDP growth or infrastructure investment levels. These studies fail to fully capture the long-term impacts of transportation infrastructure across different economic cycles and overlook the multidimensional effects of infrastructure investment on regional economic interactions and coordinated development. Particularly in China, a country characterized by significant regional economic disparities, the varied levels of transportation infrastructure development across regions make the study of spatial spillover effects highly relevant in both theoretical and practical contexts.

In response to this, the present study conducts a systematic analysis of the spatial spillover effects of transportation infrastructure and its impact on regional economic growth, utilizing long-term panel data (2003–2022) from 31 provincial-level regions in China. By employing the spatial Durbin model (SDM) and multi-dimensional spatial weight matrices, and incorporating key indicators such as transportation network density, labor force levels, and trade openness, this study not only investigates the mechanisms by which transportation infrastructure promotes regional economic growth, but also reveals the differential effects across regions. The findings aim to provide empirical support for promoting regional coordinated development and optimizing infrastructure investment strategies.

2. Literature Review

As a pivotal component of economic infrastructure, transportation infrastructure plays a crucial role in driving economic growth, characterized by its network properties, externalities, and diffusion effects. Not only does it directly stimulate economic activities, but it also facilitates the inter-regional flow and distribution of capital, labor, and goods, thereby influencing the regional economic structure and growth dynamics. Furthermore, transportation infrastructure is instrumental in reshaping regional spatial patterns, generating complex spatial spillover effects on regional economic growth. These effects can be both positive and negative, reflecting its multidimensional impact on regional development. Extensive literature, both domestic and international, explored the impact of infrastructure investment on economic growth and its mechanisms.

Since Aschauer’s (1989) seminal analysis of the elasticity of economic growth to public infrastructure investment in the United States, the international academic community employed various methods and data types to thoroughly investigate the relationship between infrastructure and economic growth worldwide. The nexus between transportation infrastructure development and economic growth is a focal point of research [1]. According to Barro and Sala-I-Martin (1992) and Egert et al. (2009), the principle of diminishing marginal returns suggests that the degree of public good congestion is a determinant factor in the impact of new infrastructure on private production efficiency [2,3]. Hulten (1996) and Bougheas et al. (2000) argue that the match between infrastructure and the economy is reflected in their scale and structural adaptability [4,5]. Munnell (1992) identified mechanisms through which infrastructure facilitates economic growth, including direct enhancement of total output, stimulation of private investment, and job creation [6]. De La Fuente (2000) posited that infrastructure primarily fosters economic growth by alleviating developmental bottlenecks [7]. Seung and Kraybill (2001) emphasized that infrastructure investment, through interaction with other investment elements, effectively enhances productivity [8]. Boscá et al. (2002), using data from Spain, found significant effects of infrastructure investment on improving private sector productivity and reducing production costs. Meanwhile, international research presents a complex and varied perspective on the relationship between infrastructure and economic growth [9]. Shirley and Winston (2004), utilizing data from U.S. firms, demonstrated how highway infrastructure promotes economic growth by reducing corporate inventory costs [10]. Studies by Hulten and Schwab (1991) and Holtz-Eakin (1994), controlling for hard-to-observe regional variables and using U.S. state-level data, found no significant impact of infrastructure investment on economic growth [11,12]. Garcia-Mila et al. (1996)’s analysis, focusing on highways, water supply, and sewage systems, indicated no significant positive effect on economic growth [13]. Employing a vector error correction model, Ghali (1998) explored the long-term effects of infrastructure investment on economic growth, revealing significant negative impacts on private investment and economic growth [14]. Research by JinSuk et al. highlighted the paramount importance of maritime transport over air and land transport in promoting economic growth among OECD member countries [15]. Conversely, Pelayo Arbués et al.’s study on Spanish transport modes showed that only road transport had a positive effect on economic growth [16]. In terms of model construction, Boarnet incorporated transportation infrastructure investment into the Douglas production function for analysis [17]. Through Anselin’s LM test criterion and LeSage’s partial differential approach, which decomposes the coefficients of explanatory variables into direct, indirect, and total effects, a plethora of scholars embarked on a multidimensional exploration of the impact of transportation construction investment or the physical volume of transportation infrastructure on economic growth [18,19].

Research on infrastructure investment in China predominantly supports its role in promoting economic growth. Hu Angang and Liu Shenglong (2009) confirmed the positive impact of transportation infrastructure on economic growth [20]; similarly, Guo Xiaoli (2014) identified a positive spatial spillover effect of China’s transportation infrastructure that becomes more pronounced over time [21]. Empirical analysis based on national and provincial panel data, represented by Zhang Xueliang (2012) and Wang Xiaodong (2014), selected inter-provincial panel data to comprehensively analyze the spatial spillover effects of transportation infrastructure and other variables on regional economic growth [22,23]. Hu Yu (2015) and Hu Yan (2015), utilizing data from China’s eastern, central, and western regions, conducted a classified comparative empirical analysis of the role of transportation infrastructure in regional disparities [24,25]. In empirical analyses based on specific categories of transportation infrastructure, Liu Yong (2010) used panel data from Chinese provinces from 1978 to 2008 to list the positive impacts of road and water transport fixed capital stock on economic development [26]; Huang Suping (2017) selected panel data from 26 cities in the Yangtze River Delta over ten years to verify the impact of road and railway infrastructure on the economic growth of the Yangtze River Delta urban agglomeration [27].With the development of spatial econometric models, Zhang Zhi (2012) and Zhou Hao (2012) incorporated three types of economic significance spatial weight matrices into their models to study the impact of spillover effects, finding that the spatial spillover effects of China’s transportation infrastructure are closely related to the patterns and stages of economic development [28]. Zhang Qiang (2016) and Zhang Yingqin (2017) selected provincial panel data from the five northwestern provinces of the “Silk Road Economic Belt” and constructed a spatial spillover effects model influenced by multiple variables, empirically analyzing the spatial spillover effects of transportation infrastructure stock and flow on economic growth [29,30]. Wang Lei (2018), based on data from the Yangtze River Economic Belt from 2005 to 2014, established a spatial Durbin model to empirically study the impact of transportation infrastructure on economic growth [31]. Liu Qihong (2017) empirically analyzed the impact of transportation infrastructure on regional economic growth from a quantitative perspective [32]; Hou Zhiqiang (2018) focused on railways and roads as transportation infrastructure variables, constructing spatial econometric models for the effects on regional tourism economic growth, with roads categorized into three levels [33]. Liao Maolin (2018) used provincial panel data from China from 1994 to 2016 to investigate the facilitative role of infrastructure in economic growth during different growth phases [34]. Wang Jian (2023) utilized the entropy weight method to construct a comprehensive index reflecting the transportation infrastructure of 41 prefecture-level cities in the Yangtze River Delta region and analyzed the spatial distribution pattern of transportation infrastructure [35]. Yang (2024) employed data from the Beijing–Tianjin–Hebei urban agglomeration to reveal the dual effects of transport infrastructure on regional economic coordination, highlighting inhibitory effects in certain areas and a polarization effect that surpasses the trickle-down impact within the urban cluster [36]. Guo (2024) identified significant spatial autocorrelation in the influence of transport infrastructure on economic growth in Hunan Province, with variations contingent on different spatial weight matrices [37]. Chen (2024) emphasized, through a quantitative model, the critical role of synergy between data elements and transport infrastructure in advancing industrial structure upgrading [38]. Yu (2024), using a multi-sector, multi-regional spatial economic model, demonstrated that transport infrastructure effectively enhances economic growth efficiency by boosting regional trade flows [39]. Wang (2023) illustrated that transport infrastructure in the Yangtze River Delta exerts a significant positive spillover effect on economic growth [40]. Wang (2022) further elucidated the pronounced spatial spillover effects of transport infrastructure, noting divergent mechanisms between highways and railways [41]. Luo (2024) examined the spatial spillover effects of market integration in the Guangdong–Hong Kong–Macao Greater Bay Area, confirming transport infrastructure as a pivotal intermediary driving economic growth [42].

Existing literature primarily focuses on the relationship between transportation infrastructure and regional economic growth at the national level and in key economic regions [43,44]. At the same time, recent studies highlighted the crucial role of intelligent transportation infrastructure and energy-saving solutions in promoting regional economic development. Particularly in the context of smart cities, optimizing transportation systems plays a key role in fostering regional coordination and improving energy efficiency [42]. However, these studies often rely on shorter time spans and single indicators, such as the total scale of transportation infrastructure or GDP growth, making it challenging to fully capture the long-term effects, multidimensional impacts, and spatial spillover effects across the entire country. This study aims to advance the understanding of this topic through several innovative approaches: First, we utilize a 20-year panel dataset from 2003 to 2022, covering 31 provincial-level administrative regions in China, which allows us to capture the long-term impacts of transportation infrastructure on regional economies and to analyze how economic fluctuations over different periods influence the results. Second, we incorporate multiple key economic and transportation indicators, such as transportation network density, labor force levels, trade openness, and R&D investment, enabling a multifactor analysis that reveals the interactions between these variables and their comprehensive impact on regional economic growth. Finally, by employing the spatial Durbin model (SDM), we systematically analyze the spatial spillover effects of transportation infrastructure, with a particular focus on the regional disparities across the country. Through these methodologies, this study not only deepens the understanding of the relationship between transportation infrastructure and regional economic growth, but also provides policymakers with multilayered, multidimensional empirical support, contributing to the coordinated and sustainable development of China’s regional economies.

Although extensive literature explored the impact of transportation infrastructure on regional economic growth, particularly the relationship between transportation network density and economic performance, most existing studies remain confined to specific regions or economically developed areas, lacking differentiated analysis across regions with varying levels of economic development, geographical conditions, and policy contexts. Moreover, the majority of research focuses on a single time span or a few core indicators, such as the scale of transportation infrastructure or GDP growth, limiting the ability to fully capture the long-term spillover effects and multidimensional impacts of transportation infrastructure across different economic cycles. Some studies suggest that transportation infrastructure generates significant spillover effects by enhancing economic mobility and promoting regional trade, yet these studies are often restricted to developed regions and fail to consider the heterogeneity across different regions. Therefore, this paper employs the spatial Durbin model (SDM) and multi-dimensional spatial weight matrices, utilizing long-term panel data from 2003 to 2022 and multiple key indicators, to systematically analyze transportation infrastructure across 31 provincial-level regions in China. Through this analysis, the study delves into the long-term spatial effects and regional heterogeneity of transportation infrastructure, providing multilayered and multidimensional empirical support for regional economic coordination and infrastructure investment decisions.

3. Research Hypotheses and Model Specification

Under the assumption that the construction of transportation infrastructure in both local and neighboring areas influences local economic growth, this theoretical foundation is predicated on the existence of spatial autocorrelation. To validate this hypothesis, the initial step is to conduct a test for spatial autocorrelation. Based on the outcomes of this test, further guidance will be provided for selecting the appropriate econometric methods to accurately assess the empirical model of the impact of transportation infrastructure on economic growth.

3.1. Spatial Correlation Test

“Spatial autocorrelation” refers to the phenomenon where regions in close proximity exhibit similar values of a variable. There are many methods available to test for spatial autocorrelation, among which the “Moran’s I” index is the most commonly used:

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

(1)

In Equation (1), $S^2={\displaystyle \frac{{\displaystyle {\sum }_{i=1}^n(x_i-\overline{x})}}{n}}$ represents the sample variance, and $W_{ij}$ is the $\left(i,j\right)$ element of the spatial weight matrix. The Moran’s I index, also known as the “Global Moran’s I”, tests for spatial clustering within the entire spatial sequence ${\left\{x_i\right\}}_{i=1}^n$. The value of Moran’s I typically ranges between −1 and 1. A value greater than 0 indicates positive spatial autocorrelation of the economic variables in the region, while a value less than 0 indicates negative spatial autocorrelation. If the index is close to 0, it suggests the absence of spatial correlation.

3.2. Spatial Weight Matrices

In the study of the spatial spillover effects of transportation infrastructure, constructing spatial econometric models is a common analytical approach. This method involves incorporating spatial weight matrices into the model to conduct a comprehensive analysis. The core premise of spatial econometric model analysis lies in the precise quantification of the “distance” between regions. To this end, this study constructs three types of spatial weight matrices based on selected indicator data to capture and analyze the multidimensional spatial dynamics of the impact of transportation infrastructure on regional economies.

(1) 0–1 spatial weight matrix

The 0–1 matrix is defined based on adjacency relationships. For two geographical units, if they are geographically adjacent, the matrix assigns a value of 1; otherwise, it assigns a value of 0. This matrix reflects the direct adjacency relationship between spatial units and is commonly used to capture geographical proximity effects.

$$W_{ij}=\left\{\begin{array}{l}1\hspace{1em}\hspace{1em}\hspace{1em}if region i and j are adjacent\\ 0\hspace{1em}\hspace{1em}\hspace{1em}if region i and j are not adjacent\end{array}\right.$$

(2)

(2) Spatial economic–geographical nested matrix

This matrix simultaneously considers both geographical distance and economic similarity between regions. The elements of the matrix are typically based on the geographical location (latitude and longitude) and economic characteristics (such as GDP) of the regions. This nested matrix reflects the spatial dependence of economic activities, taking into account both the spatial proximity and the differences in economic development levels between regions.

$$W_{ij}={\displaystyle \frac{1}{d_{ij}}}\times {\displaystyle \frac{\left|GD{P}_i-GD{P}_j\right|}{GD{P}_{avg}}}$$

(3)

where $d_{ij}$ represents the geographical distance between regions $i$ and $j$ (usually calculated using latitude and longitude), $\left|GD{P}_i-GD{P}_j\right|$ represents the GDP difference between regions $i$ and $j$, and $GD{P}_{avg}$ represents the average GDP (gross domestic product) of all regions.

(3) GDP-based economic distance matrix

This matrix is defined based on the inverse square of economic distance, reflecting the degree of economic disparity (such as GDP) between regions. Unlike the spatial economic–geographical nested matrix, this matrix considers only economic factors and is typically used to study the spillover effects of economic activities.

$$W_{ij}={\displaystyle \frac{1}{{(GD{P}_i-GD{P}_j)}^2}}$$

(4)

where $GD{P}_i$ and $GD{P}_j$ represent the economic levels (usually GDP) of regions $i$ and $j$. This matrix indicates that regions with smaller economic differences tend to have stronger economic connections, whereas regions with larger economic disparities have weaker connections.

This study employs three distinct spatial weight matrices: the geographic contiguity matrix, the matrix combining geographic distance and economic development levels (such as GDP), and the economic distance matrix based on GDP differences. The selection of these matrices is grounded in their ability to comprehensively capture the spatial spillover effects of transportation infrastructure from different perspectives. First, the geographic contiguity matrix is designed to capture spatial dependencies between neighboring regions, reflecting the direct interactions and influences between geographically proximate areas, making it particularly suitable for analyzing regions with direct economic and resource linkages. Second, the matrix that combines geographic distance and economic development levels not only considers the geographic proximity of regions, but also incorporates the similarity in their economic development, offering a more accurate measure of the spatial interactions in economic activities, especially when analyzing regional economic synergies. Lastly, the economic distance matrix based on the inverse square of GDP differences is utilized to depict the spatial spillover effects of economic activities and the strength of economic linkages between regions, making it particularly useful for uncovering economic interactions between regions that are economically similar but not necessarily geographically proximate. These three matrices not only independently capture different aspects of spatial dependencies and interactions, but also complement each other, providing a more comprehensive perspective on the impact of transportation infrastructure on regional economic growth.

3.3. Model Specification

Considering the impact of spatial correlation, this study adopts the spatial Durbin model (SDM), which is generally expressed as follows:

$$y=\alpha t_n+\lambda W_y+X\beta +WX\delta +\epsilon ,\epsilon ~N(0,{\sigma }^2{I}_N)$$

(5)

wherein, $\alpha$ represents the constant term, $t_n$ is a unit vector of order $N\times 1$, $W$ and $X$ are sequentially the spatial weight matrix and the data matrix, $\lambda$, $\beta$, and $\delta$ are the corresponding coefficient vectors, and $W_y$ and $WX$, respectively, denote the impact of spatial lag on the dependent and independent variables. The spatial Durbin model goes beyond merely examining the effect of independent variables on dependent variables, emphasizing instead the reflection of its spatial effects.

Equation (5) can be rewritten as:

$$y={(I-\lambda W)}^{-1}\alpha t_n+{(I-\lambda W)}^{-1}(X\beta +WX\delta )+{(I-\lambda W)}^{-1}\epsilon .$$

(6)

From Equation (6), it can be derived that:

$$(I-\lambda W{)}^{-1}=I+\lambda W+{\lambda }^2{W}^2+{\lambda }^3{W}^3+\cdots$$

(7)

Assuming that, in X, the explanatory variable $x_n=(x_{1n}{x}_{2n}\cdots x_{nn})$ has the following characteristics:

$$(X\beta +WX\delta )=(x_1\cdots x_n)({\beta }_1+w{\delta }_1,\cdots ,{\beta }_n+W{\delta }_n)={\displaystyle \sum _{n=1}^N({\beta }_n+W{\delta }_n)}{x}_n.$$

(8)

Therefore, Equation (6) can be written as:

$$\begin{array}{l}y={\displaystyle \sum _{n=1}^N({\beta }_n+W{\delta }_n){(I-\lambda W)}^{-1}{x}_n+{(I-\lambda W)}^{-1}\alpha t_n+{(I-\lambda W)}^{-1}\epsilon }\\ ={\displaystyle \sum _{n=1}^N{S}_r(W)x_n+{(I-\lambda W)}^{-1}\alpha t_n+{(I-\lambda W)}^{-1}\epsilon }\end{array}$$

(9)

wherein, $S_r(W)=({\beta }_n+W{\delta }_n){(I-\lambda W)}^{-1}$, and by expanding Equation (9), we obtain Equation (10):

$$\left(\begin{array}{l}{y}_1\\ y_2\\ ⋮\\ y_n\end{array}\right)=\left(\begin{array}{ccc}{s}_r{(W)}_{11}& s_r{(W)}_{12}& s_r{(W)}_{1n}\\ s_r{(W)}_{21}& \cdots & s_r{(W)}_{2n}\\ ⋮& ⋮& ⋮\\ s_r{(W)}_{n1}& \cdots & s_r{(W)}_{nn}\end{array}\right)\left(\begin{array}{l}{x}_{1n}\\ x_{2n}\\ ⋮\\ x_{nn}\end{array}\right)+{(I-\lambda W)}^{-1}\alpha t_n+{(I-\lambda W)}^{-1}\epsilon$$

(10)

Based on Equation (10), $S_r{(W)}_{ij}={\displaystyle \frac{y_i}{x_{jn}}}$. It can be observed that the variable in region $j$ can potentially affect the dependent variable in any region $i$. This forms the crucial link in spatial econometric modeling. Thus, the total effect of variable $X_r$ on all regions consists of the direct effect on the region itself, the indirect effects on other regions, and the total effects across all regions. Specifically: the matrix $S_r\left(W\right)$ has the average of the diagonal elements representing the direct effect of variable $X_r$; the sum of the elements in the $i$-th row of matrix $S_r\left(W\right)$, denoted as ${\displaystyle \sum _{j=1}^n{S}_r{(W)}_{ij}}$, represents the total effect of the explanatory variable $y_i$ on region $i$.

4. Variable Data and Spatial Autocorrelation Testing

4.1. Data Sources and Variable Descriptions

This study is based on panel data from 31 provincial-level administrative regions in China, covering the period from 2003 to 2022. The data were sourced from the China Statistical Yearbook, China Transportation Yearbook, and the statistical yearbooks of individual provinces. The variables are defined as follows:

(1) Transportation infrastructure density (infrastructure density)

This variable is measured by the total mileage of roads, railways, and inland waterways divided by the land area of the province, yielding the transportation network density. This indicator reflects the level of coverage and development of transportation infrastructure within a region.

(2) Economic growth (per capita GDP)

This study uses per capita GDP from the 31 provinces as an indicator of regional economic growth, providing a more intuitive measure of the differences in economic development levels across regions.

(3) Control variables (X)

Labor force: measured by the natural logarithm of the number of employed persons at the end of each year, representing the impact of labor on economic growth.

Trade openness: the level of regional openness and international trade impact is measured by dividing the total import and export trade by the regional gross domestic product (GDP) and multiplying it by the USD/CNY (United States Dollar/Chinese Yuan) exchange rate.

R&D intensity (innovation input level): represented by the ratio of R&D expenditure to GDP in each region, reflecting the impact of innovation input on economic growth.

Fiscal expenditure ratio (government intervention): measured by the ratio of local government expenditure to GDP, indicating the extent of government intervention in economic activities.

Industrial structure (industrial structure ratio): measured by the ratio of the output value of the tertiary sector to the secondary sector, reflecting the upgrading and optimization of the region’s industrial structure.

ICT intensity (information and communications technology level): measured by the proportion of total postal and telecommunications services to regional GDP, it reflects the penetration level of information technology in the regional economy.

Urbanization rate: the ratio of urban population to total population is used to measure the impact of regional urbanization on economic development.

In this study, the selection of indicators and control variables is grounded in their widespread application in existing literature and their profound impact on regional economic growth. First, transportation infrastructure density is widely recognized as a critical factor in driving regional economic development, directly influencing transportation costs, regional accessibility, and market size, making it the primary focus of our analysis. Per capita GDP is used as a standard measure of economic growth, clearly reflecting regional differences in economic development and providing a consistent basis for comparison. In selecting control variables, we incorporated labor force levels, trade openness, R&D intensity, fiscal expenditure ratio, industrial structure, ICT intensity, and urbanization rate, all of which were demonstrated in numerous studies to significantly impact economic growth. Labor force levels are fundamental drivers of production capacity; trade openness reflects the influence of international trade on regional economies; R&D intensity is crucial for long-term economic vitality; fiscal expenditure ratio is closely related to the regulation of economic activities; industrial structure optimization is essential for sustainable growth; ICT intensity became a new engine driving modern economic development; and urbanization affects labor allocation and consumer demand. By comprehensively considering these factors, this study provides a more complete and accurate analysis of the impact of transportation infrastructure on regional economic growth and its spatial spillover effects.

4.2. Spatial Autocorrelation Testing

To verify the spatial dependence between transportation infrastructure and economic growth, this study conducts spatial autocorrelation tests on panel data from 31 provinces during the period from 2003 to 2022. Using Stata 18.0 software, the Moran’s I statistics for per capita GDP and transportation network density were calculated for each province, assessing the clustering effects in their spatial distribution.

The data in

Table 1. Moran’s I statistics for per capita GDP and transportation network density.

Year	0–1 Spatial Weight Matrix (Moran’s I) (${\mathit{W}}_1$)		Economic–Geographical Nested Matrix (Moran’s I) (${\mathit{W}}_2$)		GDP-Based Economic Distance Matrix (Moran’s I) (${\mathit{W}}_3$)
	Per Capita GDP	Infrastructure Density	Per Capita GDP	Infrastructure Density	Per Capita GDP	Infrastructure Density
2003	0.138 *** (5.018)	0.139 *** (5.030)	0.194 *** (4.433)	0.101 *** (2.599)	0.379 *** (4.561)	0.341 *** (4.125)
2004	0.160 *** (5.454)	0.139 *** (5.030)	0.075 ** (2.028)	0.103 *** (2.634)	0.357 *** (4.152)	0.339 *** (4.100)
2005	0.147 *** (5.263)	0.137 *** (4.962)	0.205 *** (4.628)	0.103 *** (2.640)	0.393 *** (4.701)	0.334 *** (4.041)
2006	0.145 *** (5.216)	0.124 *** (4.656)	0.204 *** (4.601)	0.093 ** (2.471)	0.389 *** (4.646)	0.282 *** (3.513)
2007	0.144 *** (5.170)	0.123 *** (4.629)	0.205 *** (4.623)	0.096 ** (2.525)	0.383 *** (4.581)	0.282 *** (3.505)
2008	0.143 *** (5.123)	0.123 *** (4.630)	0.208 *** (4.660)	0.098 *** (2.576)	0.379 *** (4.517)	0.283 *** (3.520)
2009	0.143 *** (5.118)	0.121 *** (4.555)	0.212 *** (4.738)	0.100 *** (2.607)	0.379 *** (4.511)	0.279 *** (3.483)
2010	0.141 *** (5.074)	0.121 *** (4.545)	0.211 *** (4.708)	0.101 *** (2.631)	0.378 *** (4.497)	0.280 *** (3.487)
2011	0.140 *** (5.029)	0.119 *** (4.490)	0.210 *** (4.688)	0.100 *** (2.609)	0.375 *** (4.457)	0.277 *** (3.448)
2012	0.137 *** (4.952)	0.118 *** (4.463)	0.209 *** (4.668)	0.098 *** (2.566)	0.372 *** (4.421)	0.274 *** (3.423)
2013	0.132 *** (4.786)	0.118 *** (4.468)	0.205 *** (4.586)	0.097 *** (2.559)	0.364 *** (4.340)	0.275 *** (3.431)
2014	0.125 *** (4.582)	0.116 *** (4.405)	0.200 *** (4.488)	0.096 ** (2.532)	0.353 *** (4.223)	0.271 *** (3.385)
2015	0.119 *** (4.420)	0.115 *** (4.368)	0.197 *** (4.434)	0.094 ** (2.491)	0.343 *** (4.103)	0.269 *** (3.361)
2016	0.114 *** (4.288)	0.116 *** (4.407)	0.191 *** (4.308)	0.094 ** (2.493)	0.334 *** (4.015)	0.272 *** (3.388)
2017	0.114 *** (4.277)	0.117 *** (4.415)	0.186 *** (4.231)	0.093 ** (2.473)	0.339 *** (4.070)	0.274 *** (3.404)
2018	0.115 *** (4.313)	0.116 *** (4.379)	0.183 *** (4.167)	0.090 ** (2.402)	0.348 *** (4.180)	0.272 *** (3.386)
2019	0.114 *** (4.294)	0.115 *** (4.370)	0.181 *** (4.134)	0.086 ** (2.316)	0.349 *** (4.182)	0.271 *** (3.364)
2020	0.112 *** (4.234)	0.115 *** (4.351)	0.175 *** (4.019)	0.083 ** (2.258)	0.344 *** (4.142)	0.274 *** (3.396)
2021	0.111 *** (4.192)	0.116 *** (4.380)	0.174 *** (4.003)	0.082 ** (2.248)	0.339 *** (4.075)	0.276 *** (3.425)
2022	0.107 *** (4.079)	0.115 *** (4.367)	0.172 *** (3.942)	0.081 ** (2.220)	0.332 *** (3.991)	0.276 *** (3.423)

Note: *** p < 0.01, ** p < 0.05, and * p < 0.10. Values in parentheses are t-statistics or Z-statistics.

reveal that under the 0–1 spatial weight matrix, the Moran’s I statistic for per capita GDP was 0.138 in 2003, while that for transportation network density was 0.139, both significant at the 1% level (p < 0.01). This result indicates significant spatial clustering of per capita GDP and transportation network density among geographically adjacent provinces. This positive spatial correlation suggests that improvements in transportation infrastructure not only directly promote local economic growth, but also drive economic development in neighboring regions through spatial spillover effects. Over time, this clustering effect further intensified, with the Moran’s I statistics for per capita GDP and transportation network density in 2022 remaining high at 0.107 and 0.115, respectively. This persistent positive spatial correlation reflects the long-term stability of the spatial spillover effects of transportation infrastructure on regional economic growth. The Moran’s I statistics for per capita GDP and transportation network density are presented in Table 1.

In the economic–geographical nested matrix, the Moran’s I statistic for per capita GDP was 0.194 in 2003, significantly higher than the result under the 0–1 spatial weight matrix, while the Moran’s I statistic for transportation network density was 0.101. By 2022, these figures decreased to 0.172 and 0.081, respectively, but remained at relatively high levels. This indicates a tighter spatial association between transportation infrastructure and economic growth in provinces with similar levels of economic development. According to new economic geography theory, transportation infrastructure development in these provinces enhances the flow of factors of production and reduces “economic distance”, thereby promoting coordinated economic growth within the region. This effect is especially pronounced in regions with similar levels of economic development, underscoring the critical role of transportation infrastructure.

Most notably, the results from the GDP-based economic distance matrix indicate that the Moran’s I statistic for per capita GDP remained consistently high throughout the study period. In 2003, the Moran’s I statistic for per capita GDP reached 0.379, while that for transportation network density was 0.341; by 2022, these figures decreased to 0.332 and 0.276, respectively, but were still significantly higher than the corresponding values under other spatial weight matrices. This indicates that the spatial dependence between transportation infrastructure and economic growth is strongest in regions with similar GDP levels. This high degree of spatial correlation aligns with the predictions of neoclassical growth theory, which suggests that improvements in transportation infrastructure facilitate the flow of factors of production and market integration between regions, significantly enhancing regional economic cohesion and growth potential. In provinces with similar GDP levels, transportation infrastructure plays a particularly crucial role in optimizing resource allocation and driving regional coordination and development.

In summary, the analysis in Table 1 clearly reveals that under different spatial weight matrices, the Moran’s I statistics for per capita GDP and transportation network density are all significantly positive, highlighting the substantial spatial spillover effects of transportation infrastructure in promoting regional economic growth. Moreover, the upward trend in Moran’s I statistics over time reflects the growing role of transportation infrastructure in regional economic growth as regional transportation linkages strengthen. These results not only provide a solid theoretical and empirical basis for constructing spatial econometric models that analyze the relationship between transportation infrastructure and regional economic growth, but also offer important empirical support for the formulation of regional coordination and development policies.

5. Empirical Results Analysis

5.1. Ordinary Econometric Regression Testing

In examining the relationship between per capita GDP and transportation infrastructure density, this study employs ordinary least squares (OLS), fixed effects models, and two-way fixed effects models to evaluate the impact of transportation infrastructure and control variables on economic growth.

Specifically, the regression models used in this study can be expressed as follows:

(1) Ordinary least squares (OLS) regression model:

$$ln(PerGD{P}_{it})=\alpha +{\beta }_{1}ln(Transpor{t}_{it})+{\beta }_{2}ln(Labo{r}_{it})+{\beta }_{3}ln(Ope{n}_{it})+\cdots +{ϵ}_{it}$$

(11)

where $PerGD{P}_{it}$ represents the per capita GDP of region $i$ at time $t$, $Transpor{t}_{it}$ represents transportation infrastructure density, $Labo{r}_{it}$ represents the level of labor, $Ope{n}_{it}$ represents trade openness, and other control variables such as innovation input, industrial structure, and information technology levels are similarly included in the model. $ϵit$ is the random error term.

(2) Fixed effects model:

To control for unobservable heterogeneity across different regions, a fixed effects model is used, which can be formulated as:

$$ln(PerGD{P}_{it})=\alpha +{\beta }_{1}ln(Transpor{t}_{it})+\cdots +{\mu }_{it}+{ϵ}_{it}$$

(12)

where ${\mu }_{it}$ represents individual fixed effects, capturing time-invariant characteristics specific to each region.

(3) Two-way fixed effects model:

To further control for time effects, such as macroeconomic fluctuations and policy changes, the study employs a two-way fixed effects model, formulated as:

$$ln(PerGD{P}_{it})=\alpha +{\beta }_{1}ln(Transpor{t}_{it})+\cdots +{\mu }_{it}+{\lambda }_t+{ϵ}_{it}$$

(13)

where ${\lambda }_t$ represents time fixed effects, controlling for time-varying factors that uniformly affect all regions, such as nationwide policy changes or economic conditions.

To initially explore the direct impact of transportation infrastructure on regional economic growth, this study first applies the OLS, fixed effects, and two-way fixed effects models to conduct empirical tests on the relationship between per capita GDP and transportation infrastructure density. The analysis is based on panel data from 31 Chinese provinces spanning from 2003 to 2022, providing an assessment of the role of transportation infrastructure in regional economic growth and its performance under different control variables. The regression results for the Ordinary Least Squares (OLS), fixed effects, and two-way fixed effects models are presented in

Table 2. Regression Results from OLS, Fixed Effects, and Two-Way Fixed Effects Models.

Year	OLS	Fixed Effects	Two-Way Fixed Effects
Infrastructure density	0.219 *** (4.77)	0.583 *** (5.77)	0.231 ** (2.28)
Labor force	0.166 *** (3.05)	−0.0488 (−0.23)	0.0642 (0.43)
Trade openness	0.0251 (0.69)	−0.156 *** (−2.87)	0.0400 (1.02)
R&D intensity	0.0198 (0.53)	0.319 *** (3.25)	−0.198 *** (−2.84)
Fiscal expenditure ratio	0.467 *** (5.64)	0.224 (1.56)	−0.134 (−0.97)
Industrial structure ratio	0.392 *** (5.04)	0.555 *** (5.70)	−0.321 *** (−3.78)
ICT intensity	−0.0996 *** (−3.22)	−0.0866 *** (−2.92)	−0.0212 (−0.57)
Urbanization rate	1.259 *** (11.04)	0.538 *** (3.85)	0.0844 (0.86)
Constant	10.66 *** (23.02)	12.62 *** (7.20)	7.870 *** (6.58)
N	620	620	620
Adjusted R-squared	0.5542	0.5614	0.8219

Note: *** p < 0.01, ** p < 0.05, and * p < 0.10. Values in parentheses are t-statistics or Z-statistics.

.

(i) Ordinary least squares (OLS) regression model:

The OLS model results indicate that the coefficient for transportation infrastructure density is 0.219, and it is significant at the 1% level (t-value = 4.77). This finding suggests that the expansion of transportation infrastructure has a significant positive impact on regional economic growth. The OLS model further validates the theoretical expectation that transportation network construction can effectively enhance transportation efficiency within a region, reduce transaction costs, and thereby stimulate economic activities. However, it is important to note that the OLS model does not account for potential heterogeneity and time effects across provinces, which may introduce bias into the estimation results.

(ii) Fixed effects model:

To control for unobservable heterogeneity across provinces, the fixed effects model is employed. The results show that the coefficient for transportation infrastructure density increases to 0.583 and remains significant at the 1% level (t-value = 5.77). This indicates that even after accounting for fixed characteristics across provinces, transportation infrastructure continues to have a significant and stronger positive impact on regional economic growth. The fixed effects model better controls for province-specific factors, such as geographical location and climate conditions, which may have long-term effects on economic growth. These results further emphasize the importance of transportation infrastructure as a critical driver of regional economic development.

(iii) Two-way fixed effects model:

To further control for time effects, such as macroeconomic fluctuations and policy changes, the two-way fixed effects model is employed. The results show that the coefficient for transportation infrastructure decreases to 0.231 but remains significant at the 5% level (t-value = 2.28). This result indicates that, even after controlling for both time and individual effects, transportation infrastructure continues to exert a positive impact on regional economic growth. This suggests that the role of transportation infrastructure extends beyond static effects within a region, playing a sustained role in dynamic economic environments.

(iv) Influence of control variables:

The results across all regression models also shed light on the impact of control variables on regional economic growth. First, labor force levels have a significant positive impact on per capita GDP in the OLS model (coefficient = 0.166, p < 0.01), consistent with the traditional economic theory that labor is a key factor in driving economic growth. However, in the fixed effects and two-way fixed effects models, the impact of labor force levels is no longer significant, possibly due to the reduced marginal contribution of labor after controlling for fixed characteristics across provinces. Additionally, trade openness shows a significant negative effect in the fixed effects model (coefficient = −0.156, p < 0.01), possibly reflecting the negative impact of international market fluctuations on economic growth in provinces overly reliant on foreign trade.

Innovation input levels (R&D intensity) show a positive and significant impact on economic growth in the fixed effects model (coefficient = 0.319, p < 0.01), indicating that enhanced innovation capacity contributes to regional economic growth. However, in the two-way fixed effects model, the impact of R&D intensity turns negative (coefficient = −0.198, p < 0.01), suggesting that the benefits of innovation may have a delayed effect or that the high short-term costs of R&D spending may not immediately translate into economic benefits. Additionally, ICT intensity shows a negative impact in all models, especially in the OLS model, where it is significantly negative (coefficient = −0.0996, p < 0.01). This may indicate that the short-term improvement in ICT intensity does not immediately lead to economic growth and may require better alignment with other factors to realize its potential.

(v) Discussion on model applicability:

By comparing the different regression models, it is evident that the significance and influence of transportation infrastructure vary across models. The OLS model results may be biased due to the omission of heterogeneity across provinces and time effects. The fixed effects and two-way fixed effects models, by controlling for these factors, provide more robust results. Therefore, in analyzing regional economic growth, it is preferable to use models that account for individual heterogeneity and time effects to obtain more accurate estimates.

In summary, the results from ordinary econometric regression analysis indicate that transportation infrastructure has a significant positive impact on regional economic growth, and this conclusion remains robust even after controlling for individual and time effects. The results for control variables provide additional insights into key factors influencing regional economic growth. The ordinary econometric regression analysis lays a solid foundation for the subsequent construction of spatial econometric models and further validates the critical role of transportation infrastructure in regional economic development.

5.2. Spatial Econometric Model Specification

Given the presence of spatial autocorrelation, it is reasonable to infer that the development of transportation infrastructure in neighboring regions significantly impacts local economic growth. This necessitates the use of spatial econometric models for analysis. To comprehensively evaluate the spatial spillover effects of transportation infrastructure on economic growth, as well as the spatial dynamics of economic growth itself, this study employs the spatial Durbin model (SDM). By setting up the spatial econometric regression model, this study aims to delve into the spatial interactions between transportation infrastructure and economic growth, and how these interactions propagate across regions, influencing the spatial distribution of economic growth.

$$\begin{array}{l}PerGD{P}_{it}=\rho {\displaystyle \sum _{j=1}^n{w}_{ij}PerGD{P}_{jt}}+{\alpha }_0{t}_i+{\alpha }_{1}Transpor{t}_{it}+{\alpha }_{2}Labo{r}_{it}+\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}{\alpha }_{3}Ope{n}_{it}+{\alpha }_{4}Innovat{e}_{it}+{\alpha }_{5}GovEx{p}_{it}+{\alpha }_{6}Instru{c}_{it}+\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}{\alpha }_{7}IC{T}_{it}+{\alpha }_{8}Urba{n}_{it}+{\beta }_1{\displaystyle \sum _{j=1}^n{w}_{ij}Transpor{t}_{jt}+}{\beta }_2{\displaystyle \sum _{j=1}^n{w}_{ij}Labo{r}_{jt}}+\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}{\beta }_3{\displaystyle \sum _{j=1}^n{w}_{ij}Ope{n}_{jt}+}{\beta }_4{\displaystyle \sum _{j=1}^n{w}_{ij}Innovat{e}_{jt}}+{\beta }_5{\displaystyle \sum _{j=1}^n{w}_{ij}GovEx{p}_{jt}+}\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}{\beta }_6{\displaystyle \sum _{j=1}^n{w}_{ij}Instru{c}_{jt}}+{\beta }_7{\displaystyle \sum _{j=1}^n{w}_{ij}IC{T}_{jt}+{\beta }_8{\displaystyle \sum _{j=1}^n{w}_{ij}Urba{n}_{jt}+}}{\mu }_i+{\lambda }_t+{\epsilon }_{it}\end{array}$$

(14)

In the equation, $i$ and $t$, respectively, represent regions and years. The spatial lag term of the dependent variable is denoted as ${\displaystyle \sum _{j=1}^n{w}_{ij}PerGD{P}_{jt}}$, and the regression coefficient of the spatial lag term is denoted as $\rho$. ${\displaystyle \sum _{j=1}^n{w}_{ij}Transpor{t}_{jt}}$, ${\displaystyle \sum _{j=1}^n{w}_{ij}Labo{r}_{jt}}$, ${\displaystyle \sum _{j=1}^n{w}_{ij}Ope{n}_{jt}}$, ${\displaystyle \sum _{j=1}^n{w}_{ij}Innovat{e}_{jt}}$, ${\displaystyle \sum _{j=1}^n{w}_{ij}GovEx{p}_{jt}}$, ${\displaystyle \sum _{j=1}^n{w}_{ij}Instru{c}_{jt}}$, ${\displaystyle \sum _{j=1}^n{w}_{ij}IC{T}_{jt}}$, and ${\displaystyle \sum _{j=1}^n{w}_{ij}Urba{n}_{jt}}$ represent the spatial lag terms of the explanatory variables, while α and β are n-dimensional column vectors. ${\mu }_i$, ${\lambda }_t$, and $W$ correspond to spatial effects, time effects, and the spatial weight matrix, respectively (note: in the model, the variable perGDP represents economic growth; Labor represents labor force; Open represents trade openness; Innovate represents R&D intensity; GovExp represents fiscal expenditure ratio; Instruc represents industrial structure; ICT represents ICT intensity; and Urban represents urbanization rate).

Based on the model specification in Equation (14), we proceed to estimate the SDM using Stata 18.0 software. The estimation results are summarized in

Table 3. Spatial Durbin model (SDM) estimation results.

Explanatory Variable	0–1 Spatial Weight Matrix (Fixed Effects)	Spatial Economic–Geographical Nested Matrix (Fixed Effects)	GDP-Based Economic Distance Matrix (Fixed Effects)
Infrastructure density	0.0689 (0.76)	0.157 (1.63)	0.0131 (0.15)
Labor force	0.0725 (0.56)	0.00709 (0.05)	0.0741 (0.61)
Trade openness	0.00210 (0.06)	0.0176 (0.47)	−0.0312 (−0.94)
R&D intensity	−0.153 ** (−2.31)	−0.134 * (−1.92)	−0.191 *** (−3.11)
Fiscal expenditure ratio	−0.182 (−1.45)	−0.250 * (−1.90)	−0.120 (−1.01)
Industrial structure ratio	−0.198 ** (−2.36)	−0.348 *** (−4.23)	−0.153 ** (−2.07)
ICT intensity	−0.0185 (−0.57)	−0.0189 (−0.56)	−0.0231 (−0.79)
Urbanization rate	−0.0786 (−0.86)	−0.0945 (−0.98)	−0.0449 (−0.55)
W (infrastructure density)	2.541 *** (3.75)	1.550 *** (2.78)	0.662 *** (2.69)
W (labor force)	−0.478 (−0.47)	0.632 (0.69)	−0.134 (−0.41)
W (trade openness)	0.763 *** (2.84)	0.360 * (1.93)	0.267 *** (2.82)
W (R&D intensity)	−0.143 (−0.33)	0.464 (1.35)	−0.0295 (−0.19)
W (fiscal expenditure ratio)	0.421 (0.49)	−0.0409 (−0.07)	−0.0820 (−0.27)
W (industrial structure ratio)	−0.884 (−1.51)	−1.606 *** (−3.65)	−0.0194 (−0.10)
W (ICT intensity)	0.0583 (0.23)	−0.00488 (−0.05)	−0.00444 (−0.05)
W (urbanization rate)	−0.835 (−1.36)	−1.992 *** (−3.80)	0.205 (0.99)
Spatial rho	0.644 *** (8.80)	0.0419 (0.38)	0.664 *** (14.01)
Variance sigma2_e	0.0538 ***	0.0616 ***	0.0456 ***

Note: *** p < 0.01, ** p < 0.05, and * p < 0.10. Values in parentheses are t-statistics or Z-statistics.

. The results indicate that the coefficient for the spatial lag of per capita GDP is significant across all three spatial weight matrices: 0.644 for the 0–1 matrix, 0.0419 for the spatial economic–geographical nested matrix, and 0.664 for the GDP-based economic distance matrix. These findings suggest that there is a strong spatial dependence in regional economic growth. Using the Lagrange multiplier (LM) test and the Wald test, we reject the null hypothesis that the spatial Durbin model can be simplified to a spatial lag model (SLM) or a spatial error model (SEM) at the 5% significance level, indicating that the SDM is more appropriate for the context of this study. Additionally, the Hausman test results support the choice of a two-way fixed effects SDM, which allows for a more accurate identification of the spatial spillover effects of transportation infrastructure on economic growth. The estimation results for the Spatial Durbin Model (SDM) are presented in Table 3.

(i) Direct effects of transportation infrastructure:

In the 0–1 spatial weight matrix, the direct effect coefficient of transportation infrastructure density on per capita GDP is 0.0689 (not statistically significant), and in the other matrices, the coefficients are similarly insignificant. This suggests that, after controlling for spatial dependence, the direct contribution of transportation infrastructure to local economic growth may be limited. However, this does not imply that transportation infrastructure lacks importance. Rather, it may indicate that its impact primarily operates through indirect effects, specifically spatial spillover effects, rather than direct contributions to the local economy.

(ii) Spatial spillover effects:

In all spatial weight matrices, the spatial lag coefficient of transportation infrastructure is significantly positive, particularly in the 0–1 spatial weight matrix, where the coefficient is 2.541 (p < 0.01). This finding demonstrates that transportation infrastructure exerts a substantial positive spatial spillover effect on neighboring regions’ economic growth. In other words, improvements in transportation infrastructure not only boost the economic performance of the region where the infrastructure is located, but also stimulate economic activity in adjacent regions by enhancing regional connectivity and reducing transportation costs.

(iii) Spatial effects of control variables:

The spatial effects of other control variables, such as labor force, trade openness, and R&D intensity, vary across the different matrices. For example, the spatial lag of trade openness is significantly positive in all three matrices, particularly in the 0–1 spatial weight matrix, where the coefficient is 0.763 (p < 0.01). This indicates that trade activities positively affect not only the local economy, but also neighboring regions through spillover effects. However, in the spatial economic–geographical nested matrix, the spatial lag coefficients for the industrial structure ratio and urbanization rate show negative effects. This might suggest competitive or substitution effects among neighboring regions. For example, an upgrade in industrial structure in one region may lead to increased competition for resources in adjacent areas, while accelerated urbanization in one region may suppress labor markets in nearby regions. The spatial lag coefficient ρ is significant in both the 0–1 spatial weight matrix and the GDP-based economic distance matrix, with values of 0.644 and 0.664 (p < 0.01), respectively. This further underscores the strong spatial dependence in regional economic growth, indicating that such dependence is influenced not only by geographical proximity, but also by similarities in economic development levels.

The results of the spatial Durbin model in Table 3 further confirm that the spatial spillover effects of transportation infrastructure on regional economic growth are significant, especially in regions with strong geographical proximity. This finding provides a theoretical basis for investment in transportation infrastructure, suggesting that regional economic development policies should fully consider the role of transportation infrastructure in promoting economic coordination across regions. Additionally, attention should be given to the coordination of industrial structures among regions to avoid potential competitive effects.

5.3. Decomposition of Effects in the Spatial Durbin Model

In the preceding analysis, we established the direct and spillover effects of transportation infrastructure on regional economic growth. However, to gain a deeper understanding of the spatial impact of transportation infrastructure on regional economies, this study further decomposes the effects of the spatial Durbin model (SDM) into direct effects, indirect effects, and total effects. This allows for a more comprehensive depiction of the economic impact pathways of transportation infrastructure under different spatial weight matrices. The direct effect reflects the impact of transportation infrastructure on the economic growth of the region itself. The indirect effect captures the spillover effect, where transportation infrastructure influences neighboring regions. The total effect, which is the sum of the direct and indirect effects, represents the overall impact of transportation infrastructure on regional economic growth.

(1) Direct Effect Analysis

(i) Estimation results of direct effects

Table 4. Decomposition of the Dubin model effect (direct effect).

Explanatory Variable	0–1 Spatial Weight Matrix (Fixed Effects)	Spatial Economic–Geographical Nested Matrix (Fixed Effects)	GDP-Based Economic Distance Matrix (Fixed Effects)
Infrastructure density	0.271 ** (2.36)	0.164 * (1.70)	0.0131 (0.15)
Labor force	0.0220 (0.15)	−0.00433 (−0.04)	0.0437 (0.36)
Trade openness	0.0633 (1.31)	0.0208 (0.53)	0.0133 (0.33)
R&D intensity	−0.168 ** (−2.09)	−0.127 * (−1.66)	−0.214 *** (−2.96)
Fiscal expenditure ratio	−0.193 (−1.11)	−0.277 * (−1.78)	−0.178 (−1.18)
Industrial structure ratio	−0.265 ** (−2.36)	−0.349 *** (−4.23)	−0.164 * (−1.75)
ICT intensity	−0.0145 (−0.34)	−0.0186 (−0.55)	−0.0261 (−0.72)
Urbanization rate	−0.166 (−1.57)	−0.123 (−1.36)	−0.0306 (−0.34)
N	620	620	620
Adjusted R-squared	0.1503	0.0894	0.1651

Note: *** p < 0.01, ** p < 0.05, and * p < 0.10. Values in parentheses are t-statistics or Z-statistics.

presents the estimated direct effects of transportation infrastructure on per capita GDP under the three types of spatial weight matrices. Specifically, under the 0–1 spatial weight matrix, the direct effect coefficient of transportation infrastructure density on per capita GDP is 0.271 (p < 0.05), indicating a significant positive effect of transportation infrastructure on local economic growth. This result is consistent with theoretical expectations that the construction of transportation networks can effectively improve local transportation conditions, enhance production efficiency, and thereby drive economic growth. Under the spatial economic–geographical nested matrix, the direct effect coefficient is 0.164 (p < 0.1), showing a positive effect, albeit with lower significance. This suggests that even when considering the geographical and economic similarities between regions, transportation infrastructure still plays a role in promoting economic growth. This finding aligns with the perspective in regional economic theory that the construction of transportation infrastructure helps to strengthen economic linkages within a region, facilitate the flow of production factors, and thus stimulate regional economic growth. Under the GDP-based economic distance matrix, the direct effect coefficient is 0.0131 and does not pass the significance test, indicating that in regions with similar GDP levels, the direct contribution of transportation infrastructure to local economic growth is relatively limited. This could be because in regions with similar economic levels, the marginal effect of transportation infrastructure diminishes, and regional economic growth relies more on cooperation and interaction rather than solely on infrastructure improvements. The decomposition of the effects of the Durbin Model (direct effects) is presented in Table 4.

(ii) Interpretation of direct effect results

The direct effects of transportation infrastructure on regional economic growth exhibit some variation across different spatial weight matrices. In particular, the direct effect is more pronounced in regions with strong geographical proximity. This result underscores the fundamental role of transportation infrastructure in enhancing regional economic efficiency, reducing transportation costs, and promoting economic growth. However, it is worth noting that the direct effects are not significant under the GDP-based economic distance matrix, which suggests that in regions with high economic similarity, the standalone impact of transportation infrastructure may be relatively limited. This could be related to the convergence of economic levels between regions. In these areas, the promotion of economic growth by transportation infrastructure may be realized more through spillover effects than direct effects. This finding suggests that policymakers, in regions with high economic similarity, should not only focus on investing in transportation infrastructure, but also emphasize regional cooperation and resource sharing to achieve higher economic growth.

(iii) Direct effects of other variables

When analyzing the control variables, R&D intensity shows a significant negative direct effect across all three spatial weight matrices, particularly in the 0–1 spatial weight matrix, where the coefficient is −0.168 (p < 0.05). This indicates that, in the short term, the high costs and long payback periods associated with innovation inputs may exert a certain inhibitory effect on economic growth. While innovation is crucial for long-term economic growth, its high risk and uncertainty in the short term may lead to a temporary slowdown in economic expansion. Additionally, the industrial structure ratio exhibits a negative direct effect in both the 0–1 spatial weight matrix and the spatial economic–geographical nested matrix. This may reflect the high short-term costs associated with industrial restructuring during regional economic transformation, which may negatively impact economic growth. These results suggest that, although industrial structure optimization is a critical direction for economic growth, its implementation needs to be carefully managed to balance short-term pain with long-term benefits.

(iv) Policy implications of direct effects

Based on the above analysis, transportation infrastructure has significant direct effects on regional economic growth, particularly in regions with strong geographical proximity. Therefore, policymakers should prioritize the construction of transportation infrastructure in these regions to maximize its economic benefits. Furthermore, in regions with similar levels of economic development, transportation infrastructure investment should be accompanied by an emphasis on regional cooperation and collaboration mechanisms to fully leverage the potential economic effects of transportation infrastructure. Regarding the negative direct effects of R&D intensity and industrial restructuring, policymakers should adopt supportive measures to mitigate innovation risks, encourage long-term investment, and provide necessary transitional support during industrial restructuring to alleviate the short-term negative impacts and ensure sustainable regional economic growth.

(2) Indirect Effect Analysis

(i) Estimation results of indirect effects

Table 5. Decomposition of the Dubin model effect (indirect effect).

Explanatory Variable	0–1 Spatial Weight Matrix (Fixed Effects)	Spatial Economic–Geographical Nested Matrix (Fixed Effects)	GDP-Based Economic Distance Matrix (Fixed Effects)
Infrastructure density	7.573 *** (3.23)	1.655 *** (2.96)	1.944 *** (2.88)
Labor force	−1.419 (−0.46)	0.609 (0.59)	−0.306 (−0.31)
Trade openness	2.248 ** (2.21)	0.394 * (1.70)	0.717 ** (2.29)
R&D intensity	−0.760 (−0.54)	0.488 (1.25)	−0.466 (−0.94)
Fiscal expenditure ratio	0.553 (0.22)	−0.138 (−0.22)	−0.569 (−0.69)
Industrial structure ratio	−2.742 (−1.54)	−1.680 *** (−3.62)	−0.279 (−0.47)
ICT intensity	0.137 (0.18)	−0.00454 (−0.04)	−0.0543 (−0.19)
Urbanization rate	−2.612 (−1.59)	−2.143 *** (−3.71)	0.487 (0.86)
N	620	620	620
Adjusted R-squared	0.1503	0.0894	0.1651

Note: *** p < 0.01, ** p < 0.05, and * p < 0.10. Values in parentheses are t-statistics or Z-statistics.

presents the estimated indirect effects of transportation infrastructure on per capita GDP under the three spatial weight matrices. The results indicate that transportation infrastructure has a significant positive spillover effect on neighboring regions’ economic growth across all matrices. Specifically, in the 0–1 spatial weight matrix, the indirect effect coefficient is 7.573 (p < 0.01), which is substantially higher than the direct effect coefficient. This result suggests that the primary impact of transportation infrastructure is realized through its spillover effects, which extend beyond the boundaries of the region where the infrastructure is located. The decomposition of the effects of the Durbin Model (indirect effects) is presented in Table 5.

In the spatial economic–geographical nested matrix, the indirect effect coefficient is 1.655 (p < 0.01), still significant, but lower than that in the 0–1 spatial weight matrix. This indicates that while transportation infrastructure continues to promote economic growth in neighboring regions, the magnitude of the spillover effect is influenced by both geographical proximity and economic similarity between regions.

In the GDP-based economic distance matrix, the indirect effect coefficient is 1.944 (p < 0.01), demonstrating that regions with similar levels of economic development benefit from each other’s transportation infrastructure improvements. The positive spillover effects suggest that infrastructure investment in one region can lead to enhanced economic collaboration and integration in economically similar neighboring regions, thereby fostering overall regional economic growth.

(ii) Interpretation of indirect effect results

The results clearly indicate that the indirect effects of transportation infrastructure are not only significant, but also much larger than the direct effects, particularly in the 0–1 spatial weight matrix. This finding underscores the importance of considering the broader regional implications of infrastructure investments. The strong positive spillover effects suggest that transportation infrastructure plays a critical role in enhancing regional economic interdependence and fostering spatial economic integration. The varying magnitudes of the indirect effects across different spatial weight matrices highlight the role of both geographical proximity and economic similarity in shaping the extent of these spillover effects.

Regions that are geographically proximate or economically similar are more likely to experience positive spillover effects from each other’s infrastructure investments. This insight is crucial for policymakers, as it suggests that coordinated infrastructure investment strategies across regions can lead to greater overall economic benefits. Policymakers should therefore consider not only the direct benefits of infrastructure investments, but also the potential for significant indirect benefits that accrue to neighboring regions.

(iii) Policy implications of indirect effects

Given the significant indirect effects identified in this analysis, regional planning and development strategies should place a strong emphasis on collaborative infrastructure investment. Policymakers should prioritize projects that enhance connectivity between regions, particularly those that are geographically close or economically similar. Such investments can lead to broader economic benefits by facilitating the flow of goods, services, and labor across regions, thereby promoting regional economic integration.

Moreover, the substantial indirect effects suggest that infrastructure investments in one region can have far-reaching benefits, reinforcing the need for inter-regional cooperation and coordination in planning and executing such projects. By recognizing and leveraging these spillover effects, policymakers can maximize the returns on infrastructure investments, ensuring that the economic benefits are widely distributed across regions.

In summary, the indirect effects of transportation infrastructure on regional economic growth are both significant and substantial. These effects highlight the importance of considering the broader spatial dynamics of infrastructure investments and underscore the need for coordinated regional development strategies that take full advantage of the positive spillover effects identified in this study.

(3) Total Effect Analysis

(i) Estimation results of total effects

Table 6. Decomposition of the Dubin model effect (total effect).

Explanatory Variable	0–1 Spatial Weight Matrix (Fixed Effects)	Spatial Economic–Geographical Nested Matrix (Fixed Effects)	GDP-Based Economic Distance Matrix (Fixed Effects)
Infrastructure density	7.845 *** (3.24)	1.819 *** (3.29)	2.074 *** (2.89)
Labor force	−1.397 (−0.44)	0.605 (0.58)	−0.262 (−0.25)
Trade openness	2.311 ** (2.20)	0.414 * (1.79)	0.730 ** (2.18)
R&D intensity	−0.928 (−0.64)	0.361 (0.93)	−0.680 (−1.29)
Fiscal expenditure ratio	0.360 (0.14)	−0.415 (−0.61)	−0.747 (−0.83)
Industrial structure ratio	−3.007 (−1.61)	−2.029 *** (−4.19)	−0.443 (−0.67)
ICT intensity	0.123 (0.15)	−0.0231 (−0.21)	−0.0804 (−0.26)
Urbanization rate	−2.778 (−1.62)	−2.266 *** (−3.78)	0.456 (0.73)
N	620	620	620
Adjusted R-squared	0.1503	0.0894	0.1651

Note: *** p < 0.01, ** p < 0.05, and * p < 0.10. Values in parentheses are t-statistics or Z-statistics.

presents the estimated total effects of transportation infrastructure on per capita GDP under the three types of spatial weight matrices. In the 0–1 spatial weight matrix, the total effect coefficient of transportation infrastructure is 7.845 (p < 0.01), indicating a highly significant overall impact on regional economic growth. Compared to the direct effects, the indirect effects account for a larger portion of the total effects, highlighting the extensive reach and spillover potential of transportation infrastructure as a driver of regional economic development. Specifically, improvements in transportation networks not only directly enhance the efficiency of economic activities within a region, but also reduce cross-regional transaction costs, strengthen inter-regional economic linkages, and, in turn, stimulate broader regional economic growth. The decomposition of the effects of the Durbin Model (total effects) is presented in Table 6.

Under the spatial economic–geographical nested matrix, the total effect coefficient is 1.819 (p < 0.01), also showing a significant positive impact, though lower than that of the 0–1 spatial weight matrix. This suggests that even when accounting for geographical and economic similarities, the economic influence of transportation infrastructure remains significant. This effect reflects the critical role of transportation infrastructure in enhancing regional economic integration and fostering closer economic cooperation and resource flows within the region.

In the GDP-based economic distance matrix, the total effect coefficient is 2.074 (p < 0.01), indicating that the total effect of transportation infrastructure remains significant in regions with similar levels of economic development. This result suggests that in these regions, transportation infrastructure significantly promotes economic growth by improving the flow of factors of production and facilitating market integration.

(ii) Interpretation of total effect results

The results of the total effect analysis demonstrate the comprehensive impact of transportation infrastructure on economic growth across different regional contexts. In the 0–1 spatial weight matrix, the total effect is significantly higher than in other matrices, indicating that in regions with strong geographical proximity, the construction of transportation infrastructure not only exerts direct and significant positive impacts on local economic growth, but also substantially drives economic growth in neighboring regions through spatial spillover effects. This further reinforces the role of transportation infrastructure as a key engine of regional economic growth.

In the spatial economic–geographical nested matrix and the GDP-based economic distance matrix, although the total effects are lower, they still exhibit significant positive impacts. This means that even when considering the heterogeneity in geographical location and economic development levels, the overall economic impact of transportation infrastructure persists. However, the effectiveness may depend more on economic cooperation and collaborative development among specific regions. Particularly in regions with similar levels of economic development, the role of transportation infrastructure is not only reflected in physical connectivity but also in the economic integration process.

(iii) Total effects of other variables

The total effect analysis also reveals the comprehensive influence of control variables under different spatial weight matrices. For instance, the total effect of trade openness is significantly positive across all matrices, particularly in the 0–1 spatial weight matrix, where the coefficient is 2.311 (p < 0.05). This indicates that open trade policies not only directly stimulate local economic growth, but also drive economic development in surrounding areas through international trade and cross-regional economic activities, further validating the importance of regional cooperation under globalization.

On the other hand, the total effect of the industrial structure ratio shows a significant negative impact in the spatial economic–geographical nested matrix (coefficient = −2.029, p < 0.01), suggesting the possibility of competitive effects arising during the process of industrial restructuring. This implies that when significant changes occur in the economic structure of neighboring regions, it may negatively impact adjacent areas, particularly in regions with similar economic structures where such effects are more pronounced. Therefore, it is crucial to focus on regional coordination during industrial restructuring to avoid resource misallocation and uneven economic development due to regional competition.

(iv) Policy implications of total effects

Based on the results of the total effect analysis, policymakers should recognize the comprehensive role of transportation infrastructure in promoting regional economic growth. In regions with strong geographical proximity, priority should be given to the construction of transportation infrastructure to maximize its economic benefits. For regions with similar levels of economic development, transportation infrastructure investment should be integrated with regional integration strategies to fully leverage its potential in promoting regional cooperation and resource sharing.

Furthermore, policymakers should also pay attention to the interaction between industrial restructuring and trade openness among regions. By coordinating regional development strategies, it is essential to ensure that transportation infrastructure investments can exert their influence over a broader scope, preventing imbalanced development due to regional competition. This approach would contribute to the overall enhancement and long-term development of regional economies.

6. Conclusions and Policy Recommendations

This study conducts a spatial econometric analysis of panel data from 31 provincial-level administrative regions in China, revealing the spatial spillover effects of transportation infrastructure on regional economic growth. Utilizing the spatial Durbin model (SDM), the study explores the data across different spatial weight matrices. Based on these analyses, the study draws the following three major conclusions and policy recommendations, aiming to provide more targeted empirical support and strategic guidance for regional economic development.

6.1. The Significant Spatial Spillover Effects of Transportation Infrastructure Highlight the Need for Regional Coordination

The study finds that the total effect coefficient of transportation infrastructure on per capita GDP is as high as 7.845 (p < 0.01) under the 0–1 spatial weight matrix, and it also shows significant positive effects under other matrices. This indicates that improvements in transportation infrastructure not only directly stimulate local economic growth, but also drive economic development in neighboring regions through significant spatial spillover effects. However, these spillover effects are more pronounced in regions with strong geographical proximity, while the effects remain significant but relatively smaller under the spatial economic–geographical nested matrix and the GDP-based economic distance matrix (with coefficients of 1.819 and 2.074, respectively; p < 0.01). This finding emphasizes the importance of regional coordination: policymakers should focus on overall regional coordination in transportation infrastructure investment planning, avoiding the “siphon effect”, where over-investment in infrastructure in one area attracts substantial resources, thereby weakening the economic vitality of surrounding regions. Therefore, it is recommended that regional development planning prioritize the construction of transportation infrastructure that fosters regional integration, interconnectivity, and resource sharing, to maximize the comprehensive economic impact of transportation infrastructure across the entire region.

6.2. Transportation Infrastructure Construction in Economically Similar Regions Should Focus on Synergistic Effects and Integration Strategies

In the GDP-based economic distance matrix, the total effect of transportation infrastructure is 2.074 (p < 0.01), indicating that infrastructure construction in regions with similar GDP levels can significantly promote regional economic linkage and market integration. This means that in these regions, improvements in transportation infrastructure not only enhance local economic growth, but also effectively drive the coordinated development of neighboring regions, forming “growth poles” of economic development. However, the study also shows that in the spatial economic–geographical nested matrix, the indirect effect of transportation infrastructure is somewhat diminished (with an indirect effect coefficient of 1.655, p < 0.01), which may be due to increased competitive effects arising from economic similarity between regions. Therefore, when advancing transportation infrastructure construction in economically similar regions, policymakers should focus on implementing integration strategies that enhance regional economic collaboration and resource integration, reducing destructive competition between regions. Through regional coordination mechanisms, it is essential to ensure that transportation infrastructure investment promotes coordinated development across the entire region, avoiding the suppression of development potential in other areas due to the competitive advantage of a single region.

6.3. Industrial Structure Adjustment and Innovation Investment Need to Be Coordinated with Transportation Infrastructure Construction to Form Comprehensive Policy Support

The results show that under the 0–1 spatial weight matrix, the direct effect coefficient of innovation investment (R&D intensity) is −0.168 (p < 0.05), and the direct effect of industrial structure adjustment (industrial structure ratio) also shows a negative impact under the spatial economic–geographical nested matrix (with a coefficient of −2.029, p < 0.01). These results suggest that although improvements in transportation infrastructure have a significant positive effect on regional economies, if not accompanied by corresponding industrial structure upgrades and innovation investment policies, there may be short-term negative effects on economic growth, especially in regions with similar levels of economic development. Therefore, policymakers should focus on developing and implementing complementary industrial and innovation support policies alongside transportation infrastructure construction. By promoting the diversification of industrial structures and enhancing innovation capabilities within regions, the competitive effects brought about by transportation infrastructure improvements can be mitigated, ensuring that infrastructure investments are aligned with industrial structure and innovation capacity improvements, thereby fostering sustainable regional economic development.