Exploring the Impact of Transportation Infrastructure on Regional Economy in China based on GMM and Threshold Effects

doi:10.21203/rs.3.rs-3576851/v1

Download PDF

Research Article

Exploring the Impact of Transportation Infrastructure on Regional Economy in China based on GMM and Threshold Effects

https://doi.org/10.21203/rs.3.rs-3576851/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

With the continuous implementation of transportation strategies, the relationship between transportation infrastructure and the regional economy has been receiving extensive attention. However, in the face of the economic downturn, whether the investment in transportation infrastructure can obtain the expected return is worthy of further discussion. Firstly, this study uses random forests to downscale the provincial transportation indicators and constructs a comprehensive indicator reflecting the transportation infrastructure. Secondly, to control the influence of lagged economy, this research uses multiple linear regression model to determine the number of lagged periods of the economy. Subsequently, system Generalized Method of Moments (GMM) and difference GMM estimation are used to explore the relationship between transportation infrastructure and regional economy. Finally, this study explores the nonlinear relationship between the two through the threshold effect. The results show that in the context of economic downturn, transportation infrastructure is still positively correlated with the economy in general. In addition, the employment rate, population, fixed asset investment all have a driving effect on the economy. Industrial structure, on the other hand, currently harms the economy. Further, from a non-linear perspective, transportation infrastructure has an inverted U-shaped effect on the economy. Below the first threshold, transportation infrastructure is negatively correlated with the regional economy; when it lies between the first and second thresholds, the construction of transportation infrastructure has a high driving effect on the regional economy; when it is higher than the third threshold, this driving effect will be weakened.

Transport Infrastructure

Regional Economy

Generalized Method of Moments

Threshold model

From the early expansion of domestic demand to the deployment of a country with great transportation strength made by the 19th National Congress to the present "Belt and Road", the construction of transportation infrastructure has continued to be mentioned as a priority area for development (Meng et al., 2021; Lin, 2020). Since 2004, the mileage of highways, railroads, and expressways has increased significantly, as it shown in Fig. 1. During this period, the economy has also maintained a high growth rate, and the total economic volume has been rising (Li et al., 2022). As a forerunner of economic development, transportation infrastructure has externalities that can accelerate the flow of factors and the integration of resources among cities along its routes, effectively promoting complementarity of advantages among regions (Shen and Pan, 2022; Guan et al., 2023).

Nowadays, there are still some problems in exploring the relationship between transportation infrastructure and the economy (Alam et al., 2021). On the one hand, there is no doubt about the effectiveness of improving transportation as a policy to boost the economy in the early days (Kong et al., 2021; Li et al., 2022). It remains to prove whether continued investment in it will continue to pay off as it has done in the past, with the global economic decline presenting China with enormous challenges and opportunities. On the other hand, under the conditions of a market economy, various factors within a region interact with each other. The impact of transportation infrastructure on the regional economy is complex (Lu et al., 2022). It is not sufficient to consider their relationship only from a static point of view. Therefore, it is necessary to consider more factors and to sort out the relationship between transportation infrastructure and the regional economy again from the perspective of dynamic and non-linear relationships.

There has been a large amount of literature examining the economic impact of transportation infrastructure development at the macro level. The earliest research on this topic can be traced back to the 18th century when Adam Smith theorized that the transportation environment has a positive effect on the economy and society. Since then, many subsequent studies have confirmed the conjecture (Kong et al., 2022; Onifade et al., 2022; Zhu et al., 2020). There is a long-run equilibrium relationship between transportation infrastructure and economic growth, carbon emissions, and income inequality (Zhang and Zhang, 2021). Transportation infrastructure can enhance the quality of urban economic development (Kong et al., 2022; Onifade et al., 2022). In the background of high-speed railways (HSR) becoming the mainstream mode of transportation in China, exploring the impact of HSR on urban structure reveals that it has a positive impact on urban expansion (Zhang et al., 2022; Wong et al., 2022). The emergence of HSR accelerates the economic development of the region (Zhu et al., 2020). At the same time, HSR can increase the efficiency of the investment (Wu et al., 2022; Lin, 2023). However, the empirical results are not as Adam Smith's conjecture at that time, the development of transportation infrastructure does not always positively promote the regional economy. Transportation infrastructure inequality undermines the growth of the economy (Chen, 2021). The contribution of transport infrastructure to economic growth varies across regions, as confirmed by the time series approach. In the long run, the positive impact of investment disappears due to the lack of maintenance of transportation (Magazzino and Mele, 2021).

In summary, the impact of transportation infrastructure on the regional economy has been the subject of several studies. Nevertheless, based on the diversity of geographic scales used (Alotaibi, 2022), the question of what exactly the impacts are between them is still a controversial subject. The topic remains shortcomings in the following areas. Firstly, regarding the measurement of transportation infrastructure, many researchers use rail density, highway density, and transportation investment to portray transportation infrastructure. Some other studies measure it by measuring accessibility. Both of them have limitations, as the single indicator is too partial, and the accessibility indicator focuses only on the micro impact of transportation infrastructure on the regional economy. Secondly, considering dynamic panels with only short-term dynamic impacts is common. The empirical part is based on the empirical experience of the predecessors. The dependent variable with one period lag is directly put into the regression, which lacks certain relevant support and does not consider the longer-term dynamic impact. Finally, the impact of transportation infrastructure on the regional economy is complex. A simple linear relationship cannot completely portray the impact, but only roughly show its positive or negative effect. Numerous researches have focused on the linear relationship based on panel data, without considering the possibility of a non-linear relationship between the transportation infrastructure and the regional economy.

This paper aims to answer the following questions. (1) Whether there exists a way to construct transport infrastructure more comprehensively and objectively? (2) What is the impact of transport infrastructure on the regional economy after controlling for relatively long-term lagged economies? (3) How the combined linear and non-linear relationship portraying the relationship between transport infrastructure and the regional economy is consistent?

To make up for the aforementioned shortcomings, this research firstly adopts Random Forest for the comprehensive measurement of physical indicators related to transportation infrastructure. This method makes the indicator measurement more comprehensive and objective. Secondly, this study considers the dynamic impact in the long term instead of sticking to the short term. Specifically, the multiple linear regression model is used to determine the lag order before the Generalized Method of Moments (GMM) regression, so that the empirical process is more justified. Thirdly, after dealing with the dynamic panel data using GMM, further consideration was provided to whether the impact of transportation infrastructure on the regional economy is a simple linear relationship. In addition, it is crucial to further explore whether there is a threshold in between. The empirical section continues with the threshold model.

Therefore, this research bridges the previous gap by improving the following three aspects. (1) Provide a new perspective to comprehensively consider the relevant indicators of transportation infrastructure through an effective dimensionality reduction approach. (2) Determine the lag order of the GMM regression through a more optimal method to control the relatively long-term dynamic impact of the lagged economy. (3) Integrate the linear and nonlinear relationships between transportation infrastructure and regional economy to enrich the existing research.

The rest of this paper is organized as follows. The second section provides an overview of previous related studies. The third and fourth sections present the reasons for variable selection as well as the data processing methods, and the main models used in this research, respectively. The fifth section analyzes the results, which mainly contain two parts, the GMM results and the threshold effect results. Finally, section seven presents the conclusions of this study.

2.1 Relationship between transportation infrastructure and the economy

With the continuing development of transportation infrastructure, a variety of approaches have been used to explore its economic impact, but the conclusions have been inconsistent.

Some studies have argued that transportation infrastructure can significantly increase the economic aggregate of the regions along the route. It can effectively promote the upgrading of the industrial structure while boosting the urbanization rate (Wang et al., 2023; Zhang et al., 2020). In the meantime, the size of the economy has an important role in exploring the impact of transportation infrastructure on green total factor productivity. The higher the degree of economic agglomeration, the more significant the impact of transportation on green total factor productivity (Tan et al., 2022). For some particular economies, such as resource-oriented Saudi Arabia, and agriculture-oriented Middle East and North African countries, transportation infrastructure development contributes positively to the growth of the economy (Alotaibi, 2022). Given that transportation infrastructure generates carbon dioxide emissions, this emission interacts constructively with the economic development of the region (Rehman et al., 2023). To enhance the role of transport in supporting the economy, it is also crucial to clarify the pathways through which the transportation infrastructure affects the economy. Using structural equation modeling, it finds that transportation can improve economic development performance through investment, industrial growth, and employment growth in related sectors (Lin, 2020).

Nevertheless, there is a segment of empirical evidence that draws the opposite conclusion. This negative effect still exists for HSR. The presence of HSR has certainly led to a new transportation pattern in China. It turns out that HSR connectivity has hindered economic development to some extent, especially in peripheral areas. This hindrance is caused by population mobility and industrial restructuring (Gao et al., 2020).

There is also a small body of empirical evidence that shifts the research direction not to the macro but to micro indicators. For example, it is explored that the construction of transportation infrastructure can affect urban livability (Zhao et al., 2022; Yosritzal et al., 2023), and influence urban agglomeration (Tang et al., 2021). It can affect income as well as income distribution (Nugraha et al., 2020), and sway labor mobility (Heuermann and Schmieder 2019; Feng et al., 2023).

2.2 Indicator construction for transportation infrastructure

The present methodology for assessing transportation infrastructure is not standardized. A part of the research directly uses some indicators related to transportation infrastructure, such as transportation investment (Sun et al., 2019), public investment, highway mileage, and railroad density (He et al., 2019). Ke et al. (2020) use highway and railroad mileage to represent transportation development, confirming that the quality of highway and railroad improvement and structural upgrading of infrastructure contribute significantly to Chinese regional economic growth.

It is evident that if one or two indicators are used alone, then these indicators can only depict a partial view of the transport infrastructure but not the whole picture. At this stage, accessibility can serve as a measure of the development of transport infrastructure. It generally appears as a comprehensive indicator to evaluate whether the regional transportation network is of high quality and whether it can achieve balance and coordination. Its evaluation methods have been very rich. The common ones include using the weighted average time model to calculate the accessibility (Wang et al., 2023), combining the average shortest travel time (ASTT) and weighted average travel time (WATT) to portray accessibility (Gao and Li, 2022; Chen et al., 2020). Adopting the average travel time, the economic potential of travel, and the economic linkage of tourism can also portray the accessibility (Chen et al., 2020). The latter improves on this by using a gravity model to capture the accessibility of the transportation network for travelers and goods to further measure the development of transportation infrastructure (Alotaibi et al., 2022; González et al., 2019; Cascetta et al., 2020). This reflects that the representation of accessibility is flexible and well developed, and can be continuously improved depending on the focus. However, for macroeconomic analysis, accessibility no longer seems to fully capture the changes brought about by transportation infrastructure.

2.3 Relevant empirical methods

The choice of economies to focus on and the timing of the samples are the main reasons for the different results on the impact of transport infrastructure on the regional economy (Zhang and Cheng, 2023). To address this relationship, most studies have used the vector error correction model (VECM), the difference in difference (DID), and GMM methods.

Using an autoregressive distributive lag (ARDL) and VECM, it can be found that transportation infrastructure has a long-term positive impact on the development of the economy (Alam et al., 2021). The combination of the VECM with the vector autoregressive (VAR) explores the existence of a long-term equilibrium relationship between logistics infrastructure and the economy (Wang et al., 2021). The DID method compares the difference between the period before the policy and the period after the policy. The HSR has an obvious effect on the agglomeration effect of population (Wang et al., 2022). However, based on the Shanghai-Kunming HSR line, its pull effect on the economy is limited (Wang et al., 2023). The fact that transportation infrastructure is built through close selection rather than complete randomization is a major challenge for DID. The use of GMM to describe dynamic panels is a more comprehensive step to consider as opposed to the more traditional static panel models. Through GMM, it can be confirmed that transport has an inverse relationship with CO2 emissions (Ghannouchi et al., 2023). Transportation infrastructure that enhances the economy also accelerates the generation of carbon emissions (Acheampong et al., 2023). In addition, there is a paucity of literature that uses threshold modeling to explore the impact of transportation infrastructure on economic development. It is also essential to examine this topic using this methodology, which provides a completely different perspective to explore whether the relationship of transportation infrastructure on the regional economy may not be linear.

2.4 Summary

There are three shortcomings in the existing literature. The first is that many direct indicators cause problems in model fitting. Reducing the number of variables cannot fully demonstrate the development of transportation infrastructure. Moreover, adopting accessibility to portray transportation infrastructure is not comprehensive enough. Secondly, the majority of studies using GMM to study dynamic panels have only considered shorter-term dynamic effects, which is no exact basis to support this practice. Thirdly, most researchers use a methodology to explore this relationship. It is more reasonable to consider dynamic impacts than static panels. However, the economic impact of transportation infrastructure may have more than a simple linear relationship.

The following measures could amend some of the problems. At first, the transportation infrastructure indicators are comprehensively processed using Random Forest. This method provides a new way of thinking and perspective for the subsequent construction of this indicator. Furthermore, in response to the previous dynamic panel model selecting a short lag period, this study determines the number of lags through the multiple linear regression model. Therefore, the lagged economy that controls and observes a relatively long period is chosen when there is a basis to do so. Finally, neither the static nor the dynamic model can illustrate the impact of transportation infrastructure on the regional economy in as many dimensions as possible. The choice to explore the impact of transportation infrastructure on the regional economy from both linear and nonlinear perspectives is reasonable and relatively comprehensive.

3.1 Variable selection

The regional transportation infrastructure and economic development level are reflected in several aspects and contain many elements at the same time. Therefore, the selection of appropriate variables to portray the impact of transportation infrastructure on the regional economy must truly reflect their relationship. Due to the availability of data and operability, this research chooses to deal with the following indicators to objectively reflect the state of transportation infrastructure and economic development. The variable definitions are shown in Table 1.

Table 1

Variable definitions
Variable type	Variable name	Measurement method
dependent variable	lnGDP	Logarithm of gross domestic product
Independent variable	rf	The variable to measure traffic input is obtained by reducing the dimensionality of the mileage of highways, railroads, and expressway through random forest
control variables	employment	Employment rate
	assetrate	Growth rate of investment in fixed assets
	lnpopu	The logarithm of the permanent population at the end of the year
	structure	Ratio of value added of tertiary industry to gross regional product
	urban	The ratio of town population to total population
	lnforeign	Logarithm of total foreign investment in imports and exports

(1) Dependent variable

This research uses the gross domestic product (lnGDP) as an indicator of economic development (Bian and Yeh, 2020; Alam et al., 2021). GDP is the core indicator of the national economic accounting system, and its accounting originates from the market economy, which can better reflect the characteristics of the market economy.

(2) Independent variable

Because the construction of transportation infrastructure is multifaceted, a single indicator measurement appears to be quite weak. Thus, this research does not use directly available indicators to measure transportation infrastructure but instead conducts a multi-dimensional comprehensive evaluation of physical indicators related to transportation infrastructure through the random forest model. This led to both the use of more reliable physical indicators and the synthesis of indicators. Taking into account the availability of data, this composite indicator (rf) is downscaled from highway mileage (Xue et al., 2023), railroads mileage, and expressway mileage (Wang et al., 2023; Lu et al., 2023).

(3) Control variables

Economic growth depends on numerous factors and it is crucial to incorporate a range of control variables to avoid omitted variables.

Employment (employment) affects the growth of the economy by influencing consumption investment (Wang et al., 2023). A good employment rate will promote consumption to drive the economy.

Fixed asset investment (assetrate) can increase consumption and promote economic development, but its intervention may also affect the efficiency of economic operations and hinder economic development (Chen et al., 2020; Wang et al., 2023; Zhao et al., 2022).

Population (lnpopu) is measured by the residential population. Population in the Douglas function can have an indirect effect on economic growth (Zhao et al., 2022). Transportation enhances the location advantage of small and medium-sized cities, and more population means a larger market size (Zhang and Xue, 2023).

Industrial structure (structure) is determined by the proportion of GDP contributed by the tertiary industry to the total GDP (Guan and Hu, 2023). Currently, Chinese tertiary industry plays the largest role in economic development. Rationalization of industrial structure is an important factor in promoting the economy.

The degree of urbanization (urban) is measured using the ratio of urban population to total population (Wang et al., 2023; Ke et al., 2020). Urbanization is a manifestation of the national transition from an agrarian country to a predominantly industrial, service-oriented country. China’s current urbanization rate is still not high compared to developed countries, and the lack of urbanization brings scale effects and insufficient impetus to the economy. Therefore, the urbanization process is an important variable affecting the regional economy.

Foreign investment (lnforeign) is represented by total foreign investment in imports and exports (Chen et al., 2020; Zhao et al., 2022). Learning the advanced technology and management methods already existing in developed countries is one of the important means to promote the economic development of developing countries. The influx of foreign investment can intensify healthy competition in the industry, squeeze the profits of lagging enterprises, and promote technological innovation.

3.2 Data sources

Based on the completeness and availability of data, this study selects data from 30 provinces in mainland China, except Tibet, for the period 2004–2021. The data are mainly obtained from the China Bureau of Statistics, and local statistical bulletins. Missing data for individual cities are filled in by interpolation. The measures of each variable are shown in Table 2.

Table 2

Descriptive statistics
Variable	Obs	Mean	Std.Dev.	Min	Max
lnGDP	540	9.404	1.052	6.095	11.734
rf	540	0.33	0.214	0.02	1.1
employment	540	96.545	0.692	93.5	98.8
assetrate	540	15.142	11.738	-56.6	41.8
lnpopu	540	8.182	0.749	6.29	9.448
structure	540	0.466	0.093	0.298	0.837
urban	540	0.552	0.143	0.263	0.896
lnforeign	540	15.782	2.498	8.33	20.199

4.1 Random Forest

Random Forest (RF) is an integrated algorithm consisting of multiple decision trees with no association between different decision trees, which can be used to solve classification or regression problems. Constructing RF mainly includes the following four steps.

(1) Using the Bootstrap idea to draw N samples from N samples with putbacks N times, the resulting N samples are used to train a decision tree as the samples at the root node of the decision tree.

(2) When each node of the decision tree needs to be split, $\text{m} (\text{m}<<\text{M})$ features are randomly selected from M features. Select 1 feature from the m features using information gain as the splitting attribute of that node.

(3) Each node in the decision tree formation process splits according to step (2) until it is not able to split anymore.

(4) Build p decision trees according to steps (1)-(3) to form a random forest. Combine the classification results of p decision trees to get the final classification result.

4.2 Generalized method of moments

To investigate the impact of transportation infrastructure development on economic growth, this study incorporates the traditional static panel models, which include mixed regression model (OLS), fixed effects model (FE), and random effects model (RE). Considering that the economy is subject to its inertia in the process of change i.e., the previous economy will have an impact on the current economy, the lagged dependent variable is included in the model Alotaibi (2022). The dynamic panel model is constructed as Eq. (1).

$${lnGDP}_{it}=\alpha +\sum _{j=1}^{n}{K}_{j}{lnGDP}_{it-j}+{rf}_{it}+\sum _{j=1}^{6}{\beta }_{j}{Control}_{it}+{{\lambda }_{i}+\epsilon }_{it}$$

where ln denotes the natural logarithm, n is the number of lag periods, and $i$ and $t$ represent provinces and years, respectively. ${K}_{j}{, \beta }_{j}$are proxy estimation parameters. ${rf}_{it}$is the transportation infrastructure for the $i$ year $t$ area. ${Control}_{it}$ are a series of control variables including employment, lnpopu, structure, urban, assertrate, lnforeign. ${\lambda }_{i}$ are individual fixed effects, and ${\epsilon }_{it}$ are independent and identically distributed error terms. However, in Eq. (1), considering that this study is a short panel data, the easy inclusion of lags of the dependent variable in the regression will inevitably lead to endogeneity problems in the model thus biasing the estimation results. GMM is the best estimation method that can be used in the case of short panel data with limitations and endogeneity (Arellano and Bond, 1991). Therefore, the latter researches mainly use GMM to conduct dynamic panel model analysis. GMM has been widely applied in the process of studying the impacts of transportation infrastructure (Chakrabarti, 2018). Numerous literatures have defaulted to directly include the lagged period of the explanatory variables in the regression in practice. This study uses multiple linear regression to determine the magnitude of the lagged order n (Wintoki, 2012) as Eq. (2).

$${lnGDP}_{it}=\alpha +\sum _{j=1}^{n}{K}_{j}{lnGDP}_{it-j}+\sum _{j=1}^{6}{\beta }_{j}{Control}_{it}+{{\lambda }_{i}+\epsilon }_{it}$$

The final model for this research is derived as Eq. (3).

$${lnGDP}_{it}=\alpha +\sum _{j=1}^{4}{K}_{j}{lnGDP}_{it-j}+{rf}_{it}+\sum _{j=1}^{6}{\beta }_{j}{Control}_{it}+{{\lambda }_{i}+\epsilon }_{it}$$

GMM employs either further lags of the dependent variable or the difference of the dependent variable as instrumental variables for the dependent variable in a dynamic panel. Arellano and Bond (1991) proposed a two-stage approach to GMM. Later, system GMM was further proposed. The system GMM is theoretically more efficient and reliable than the difference GMM, which can observe not only the level change but also its differential change.

4.3 Threshold models

Whether the relationship between the construction of transportation infrastructure and economic growth is purely linear or whether there is a threshold effect needs to be further explored. This study assumes that different scales of transportation infrastructure development have different impacts on regional economic development, which implies that there is a threshold effect of transportation infrastructure construction on economic development. The threshold effect formula is as Eq. (4).

$${lnGDP}_{it}=\alpha +{\beta }_{1}{rf}_{it}I({rf}_{it}<{\gamma }_{1})+{\beta }_{2}{rf}_{it}I({\gamma }_{1}\le {rf}_{it}<{\gamma }_{2})+{\beta }_{3}{rf}_{it}I({rf}_{it}\ge {\gamma }_{2})+{\beta }_{4}{Control}_{it}+{\epsilon }_{it}$$

where ${lnGDP}_{it}$denotes that the dependent variable consists of the logarithm of GDP. ${rf}_{it}$is the core independent variable as well as the threshold variable, which is a composite indicator downgraded by Random Forest.$i$ is a city $i$and $t$ is time$t$. $I$ is an indicator function that is 1 when the corresponding condition is reached and 0 when the opposite is true. ${\gamma }_{1}, {\gamma }_{2}$ is the threshold to be assessed. ${Control}_{it}$is the control variables, including employment, lnpopu, structure, urban, assertrate.

This section is divided into three main parts, the first section is to carry out the test of multicollinearity and correlation before empirical evidence. In the second part, before using GMM to analyze the relationship between transportation and the regional economy, this research completes the traditional static panel. Then GMM regression is performed after determining the lag order of the dynamic panel. Subsequently, the GMM is tested for robustness and the results are more reliable. The third section is to make a test of the nonlinear relationship between transportation infrastructure and regional economy using the threshold model.

5.1 Correlation test

Before proceeding to the specific regression, testing for correlation between variables is the first step. The result, as Table 3 shows, is that the linear correlation coefficients between the variables are all below 0.8. In addition, Table 4 shows an overall VIF of 3.33, indicating that there is no serious overall multicollinearity between the variables. The core explanatory variable rf is significant at a 1% level of significance for the economy, and the coefficient is positive, which indicates that transportation infrastructure has a significant positive impact on the economy. The correlation coefficients between the other control variables are significant, indicating that the variables are interacting with each other.

Table 3

Correlation test
	lnGDP	rf	employment	assertrate	lnpopu	structure	urban	lnforeign
lnGDP	1.000
rf	0.717***	1.000
employment	0.325***	0.247***	1.000
assetrate	-0.328***	-0.346***	-0.300***	1.000
lnpopu	0.742***	0.589***	0.003	0.078*	1.000
structure	0.228***	-0.034	0.493***	-0.483***	-0.291***	1.000
urban	0.457***	0.071*	0.347***	-0.506***	-0.150***	0.718***	1.000
lnforeign	0.730***	0.239***	0.202***	-0.143***	0.491***	0.253***	0.613***	1.000
* p < 0.05, p < 0.01, * p < 0.001

Table 4

Multiple covariance test
Variable	VIF	1/VIF
urban	5.93	0.168513
lnforeign	4.70	0.212752
lnpopu	4.09	0.244597
structure	2.90	0.344413
rf	2.36	0.424326
assetrate	1.84	0.542618
employment	1.48	0.675788
Mean VIF	3.33

5.2 GMM regression analysis

5.2.1 OLS-FE-RE analysis

After confirming that there is no multicollinearity, this section discusses the static panel model in Table 5. The regression results for ordinary least squares (OLS), fixed effects (FE), and random effects (RE) are given in columns (1), (2), and (3) of the static panel model, respectively.

From the OLS results in column (1), it is clear that transportation infrastructure, employment rate, fixed asset growth rate, urbanization rate, and foreign investment as the important factors of regional economic development all play a positive role in driving the economy at the 1% significance level. Among them, the core explanatory variables rf and urbanization rate urban have the strongest pulling effect on the economy, and every unit increase in transportation infrastructure and urbanization rate will raise lnGDP by 1.475% and 2.895% respectively. This is followed by a 0.85-unit and 0.937-unit boost to the economy from population and economic structure, respectively.

Column (2) gives the results for FE. In this case, transportation infrastructure, population, urbanization rate, and foreign investment are at a 1% level. Employment rate and fixed asset growth rate at 5% and 10% levels respectively have an impact on the economy and all of them are positively driving the economy. Among them, the greater economic impetus is the population and urbanization. In contrast, the role of employment rate is minimal. It is worth noting that the coefficients of the main explanatory variable, rf, differ significantly between the OLS and FE estimating methods. OLS tends to overestimate the coefficients, while FE yields lower coefficients.

Column (3) gives the results of RE. In this model, the constant is common, and the error term has a component. Its significance is more consistent with that shown by OLS. It is clear that the core explanatory variables, as well as the urbanization rate and population, have a greater positive push on the economy under the results presented by the RE model. For every 1 unit rise in transportation infrastructure, the economy subsequently rises by 1.034 units.

From the aforementioned three static panel estimation methods, transportation infrastructure, populations and urbanization rate are the three factors that promote economic development the most. The coefficients of transportation infrastructure in the three models are positive and significant, indicating that transportation infrastructure improves the reliability of economic development. Transportation infrastructure accelerates interregional factor flow and reduces transportation costs, thus playing the role of economic improvement (Han et al.,2023). At the same time, the construction of transportation infrastructure is itself an economic production activity, which requires a large amount of human and material resources (Dou et al.,2023). Its development will inevitably produce a certain economic value, which will lead to an increase in the income of residents in the process.

Table 5

Estimation results for the static models of regional GDP
	(1)	(2)	(3)
	OLS	FE	RE
rf	1.475***	0.480*	1.034***
	(0.000)	(0.028)	(0.000)
employment	0.081***	0.117**	0.139***
	(0.000)	(0.009)	(0.000)
assetrate	-0.000	0.003*	0.002
	(0.805)	(0.048)	(0.096)
lnpopu	0.850***	2.106***	1.032***
	(0.000)	(0.000)	(0.000)
structure	0.937***	0.483	1.060***
	(0.000)	(0.502)	(0.000)
urban	2.895***	5.473***	4.066***
	(0.000)	(0.000)	(0.000)
lnforeign	0.037***	0.089**	0.047***
	(0.000)	(0.002)	(0.000)
_cons	-8.451***	-23.940***	-16.270***
	(0.000)	(0.000)	(0.000)
N	540	540	540
adj. R2	0.945	0.937
p-values in parentheses
* p < 0.05, p < 0.01, * p < 0.001

5.2.2 System GMM and Difference GMM

In contrast to static models, the first step in considering dynamic effects is to determine the lag order. This step conducts multiple linear regression based on what is described in the introduction of the model in the previous section, referring to Wintoki (2012). Multiple linear regression includes the dependent variable lagged from the first to the fifth order as the independent variable in the regression, and the results are shown in Table 6.

In the previous section, OLS does not consider the individual heterogeneity effect, and FE ignores the dynamic impact of the economy in the previous period on the economy in the later period when considering the individual heterogeneity. Thus, the estimation results of the static panel model may have a large bias in the results. Taking into account that the economy is a dynamic process of change, there may be a certain lag effect in the changes in the economy. As a result, this research adds the lag term of the dependent variable based on the static panel data model and analyses it through the dynamic panel model. As seen in the model, lnGDP from lag one to lag four is significantly correlated at the 1% level. Whereas lnGDP in lag five is not significant, indicating that the economy from the fifth order does not have a significant effect on the current period. Therefore, this study determines the number of lagged periods of the dependent variable at four in the subsequent dynamic panel model.

Table 6

Multiple linear regression
	lnGDP1		lnGDP1
L.lnGDP1	1.103***	employment1	0.006*
	(0.056)		(0.004)
L2.lnGDP1	-0.365***	assertrate1	0.002***
	(0.091)		(0.000)
L3.lnGDP1	0.511***	structure1	-0.040
	(0.090)		(0.031)
L4.lnGDP1	-0.238***	lnforeign1	0.002*
	(0.087)		(0.001)
L5.lnGDP1	-0.021	_cons	-0.481
	(0.051)		(0.350)
N	390
adj. R2	0.998
z-statistics in parentheses
* p < 0.01, p < 0.05, * p < 0.1

After determining the lag order to be four periods, this research uses the column (1)-(2) system GMM and the column (3)-(4) two-stage difference GMM estimation respectively to solve the endogeneity problem that will be caused by adding the lag term, and the results are shown in Table 7.

The lagged fourth order of the regional economy is all significant at the 1% level, indicating that the regional economy has development inertia, i.e., there is a certain degree of time stickiness. Consequently, the use of GMMs for the adjustment of the dynamic panel is a reasonable option. From the results of the serial autocorrelation test, none of the perturbation terms has a higher-order autocorrelation of order 2, and i.e., AR (2) is not significant, which is in line with the premise assumption of GMM regression. In addition, the results of Sargan's over-identification test also accept the original hypothesis of "all instrumental variables are valid", indicating that the model suitability test is passed. The estimation results are valid.

Columns (1)-(2) show the estimation results of the system GMM. After adding various control variables, transportation infrastructure is significant at the 1% level and contributes positively to the regional economy. This result supports the results from OLS, FE, and RE shown in Table 5. By incorporating the economy into the model with a lag of several periods, the parameter estimates produced by the dynamic model and the static model are somewhat inconsistent. It appears that the coefficients on transportation infrastructure in the dynamic model are generally larger than those estimated in the static model. The positive impact of transportation infrastructure on the economy is greater when the lagged effects of the economy are taken into account.

Column (1) indicates that for every 1 unit rise in transportation infrastructure, the corresponding lnGDP rises by 0.416 units. This coincides with the findings of previous research, suggesting that transportation infrastructure has been a trailblazer in modernization that continues to drive the economy (Wang et al., 2023; Kong et al., 2022). For every 1 unit increase in employment rate, the economy improves by 0.052 units. The higher the employment rate, the more people have stable employment, and the stable income can drive consumption to some extent, increasing demand for goods and services and promoting economic development. The pull of the population on the economy is strong, which is the situation in China today. It is undeniable that part of the sustained and astonishing GDP growth in China can be attributed to the demographic dividend. A large labor force of school-age provides favorable demographic conditions for economic development. However, the pulling effect of the growth rate of fixed assets on the economy, though significant, is very weak, with the economy rising by 0.002 units for every unit of its rise. The coefficients of industrial structure and urbanization process are positive but not significant.

In column (2), the research first eliminates the two control variables mentioned aforementioned which are not significant. The fit of the model should be better and the noise of the model is reduced to some extent. It can be observed that after removing the insignificant control variables, the coefficient of transportation infrastructure decreases slightly, and each unit increase in transportation infrastructure can pull the economy by 0.396 units. Meanwhile, the driving effect of the population on the economy decreases slightly, but the driving effect of foreign investment on the economy increases. It is the same as Wang et al. (2023) who concluded that the labor force has an enhancing effect on GDP. The coefficient of employment rate in the model is 0.054 and so is foreign investment. It can be explained that foreign investment can promote many aspects of development, the most important being that it requires a large number of human and material resources, which can bring a large number of jobs and provide more employment opportunities. Therefore, foreign investment is closely related to employment. In Column (2), even after removing the insignificant control variables, fixed investment is still consistent with Column (1) and does not change much, so it can be judged that its impact on the economy is weak after reducing the noise.

Columns (3)-(4) report the estimation results derived from the difference GMM after adding different control variables. The lagged lnGDP is statistically significant, indicating that the model is reasonable and that there are indeed dynamic changes. Transportation infrastructure is the most important and interesting variable in this study, but this does not mean that the control variables can be completely disregarded. Therefore, it is necessary to investigate the effect of different control variables on the results in the same model. Transportation infrastructure is still very significant in the two model columns (3) and (4) of the difference GMM. Unlike the system GMM, the control variable industrial structure and urbanization rate are significant here at the 1% level. For every unit increase in the urbanization rate, there is a corresponding 1.55 unit increase in the economy. Urbanization and economic development are generally considered to be interrelated (Maparu and Mazumder, 2017; Zhao, 2023). Urbanization leads to a more concentrated use of infrastructure, which allows for more efficient industrial development and the creation of more jobs. In the meantime, urbanization leads to a more socialized family life, allowing more farmers to free themselves from agriculture, accelerating the development of the service sector, and boosting the economy. Industrial structure hurts on the economy. For every unit of improvement in industrial structure, the economy will decrease by 0.374 units. The rise of industrial structure means that the proportion of tertiary industry increases, and the economy is transformed from an industry-driven economy to a service-driven economy, which tends to slow down the economic growth rate. But in the long run, the continuous optimization and adjustment of the industrial structure is conducive to the sustainable development of the economy. Column (4) excludes the less influential fixed asset investment ratio in column (3), and the significance level of the total variable does not change significantly and the sign of the coefficients does not change.

Table 7

Estimation results for the dynamic models of regional GDP
	SYS-GMM		DIFF-GMM
	(1)	(2)	(3)	(4)
VARIABLES	lnGDP	lnGDP	lnGDP	lnGDP
L.lnGDP	0.762***	0.730***	0.635***	0.732***
	(19.29)	(18.03)	(14.70)	(27.48)
L2.lnGDP	-0.390***	-0.363***	-0.265***	-0.307***
	(-12.67)	(-12.31)	(-12.70)	(-15.96)
L3.lnGDP	0.481***	0.481***	0.343***	0.337***
	(20.52)	(27.49)	(15.02)	(29.54)
L4.lnGDP	-0.100***	-0.089***	-0.022	-0.059***
	(-3.53)	(-3.89)	(-1.03)	(-3.92)
rf	0.416***	0.396***	0.108***	0.110***
	(7.51)	(8.04)	(4.11)	(6.26)
employment1	0.052***	0.054***	0.026***	0.027***
	(5.87)	(7.56)	(6.31)	(6.11)
assertrate	0.002***	0.002***	0.002***
	(6.50)	(7.65)	(6.87)
lnpopu	0.287***	0.242***	0.467***	0.426***
	(3.32)	(3.33)	(3.40)	(3.44)
structure	-0.101		-0.376***	-0.510***
	(-0.79)		(-3.75)	(-9.18)
urban	0.155		1.550***	1.466***
	(0.67)		(6.82)	(10.27)
lnforeign	0.049***	0.054***	0.035***	0.033***
	(7.87)	(8.31)	(7.58)	(8.40)
Constant	-5.860***	-5.788***	-4.604***	-4.278***
	(-7.44)	(-9.34)	(-3.84)	(-3.64)
AR(1) AR(2) Sargan	0.001 0.824 1.000	0.002 0.922 1.000	0.001 0.210 0.999	0.008 0.635 0.999
Observations	420	420	390	390
Number of province	30	30	30	30
wald	46231	34705	178089	162246
z-statistics in parentheses
* p < 0.01, p < 0.05, * p < 0.1

5.2.3 Robustness Tests

To further test the robustness of the empirical results, this research chooses the three methods of replacing the dependent variable, replacing the independent variable, and shortening the sample years. The results of the robustness test are presented in Table 8.

Column (1) is the replacement of the original economic development expressed by lnGDP with the logarithm of GNP per capita lnpergdp used by Banerjee et al. (2020). In column (1) this study can prove that the core dependent variable is highly significant at the 1% level, and AR (2) strictly rejects the original hypothesis indicating that there is no higher-order autocorrelation, where the Sargan test also rejects the original hypothesis indicating that the instrumental variables are valid.

Column (2) replaces the original independent variable rf with the core variable pca, which is obtained by Principal Component Analysis. Principal Component Analysis (PCA) can be used as a dimensionality reduction method to measure transportation infrastructure (Zhang and Cheng, 2023). The new features not only reconstruct the original features but also eliminate the redundant information of the original features, making them more distinguishable. Essentially both PCA and random forest are a method of dimensionality reduction, thus it is used here as the independent variable for replacement. Similarly, it turns out that the absence of higher-order autocorrelation also applies to the instrumental variables, and the core independent variable is significant at the 1% level.

Column (3) performs a robustness test by shortening the sample years. Our original sample is selected to handle the relevant data from 2004–2021. As we all know, Corona Virus Disease 2019 started to ravage in previous years. To control the further development of the epidemic, there was a long period when China did not carry out normal construction activities, and the transportation infrastructure was thus halted for some time. In column (3) the data for 2019–2021 will be excluded, it can be found that the core variables are significant and there is no higher-order autocorrelation. Sargan rejects the original hypothesis at a 1% level of significance. Thus, in summary, the GMM model in this research yields robust and reliable results.

Table 8

Robustness check
Replacement of dependent variable		Replacement of independent variables		Shorter sample years
Variable	(1) lnpergdp	Variable	(2) lnGDP	Variable	(3) lnGDP
L.lnpergdp	0.841***	L.lnGDP	0.781***	L.lnGDP	0.996***
	(28.61)		(18.35)		(28.13)
L2.lnpergdp	-0.450***	L2.lnGDP	-0.425***	L2.lnGDP	-0.543***
	(-22.12)		(-16.64)		(-20.81)
L3.lnpergdp	0.490***	L3.lnGDP	0.483***	L3.lnGDP	0.546***
	(24.22)		(28.76)		(26.88)
L4.lnpergdp	-0.127***	L4.lnGDP	-0.072***	L4.lnGDP	-0.224***
	(-8.67)		(-3.51)		(-10.25)
rf	0.285***	f1	0.123***	rf	0.324***
	(6.45)		(8.43)		(6.71)
employment	0.042***	employment	0.052***	employment	0.033**
	(6.60)		(6.19)		(2.54)
assertrate	0.001***	assertrate	0.002***	assertrate	0.001***
	(7.33)		(6.57)		(8.48)
lnpopu	0.247***	lnpopu	0.353***	lnpopu	0.380***
	(3.51)		(4.68)		(6.20)
structure	-0.188	structure	-0.173	structure	0.047
	(-1.50)		(-1.22)		(0.30)
urban	0.590**	urban	-0.046	urban	0.174
	(2.50)		(-0.19)		(0.59)
lnforeign	0.027***	lnforeign	0.046***	lnforeign	0.050***
	(6.51)		(7.02)		(7.86)
Constant	-4.163***	Constant	-6.271***	Constant	-5.146***
	(-5.15)		(-9.78)		(-3.91)
AR(1) AR(2) Sargan	0.001 0.645 1.000	AR(1) AR(2) Sargan	0.001 0.901 1.000	AR(1) AR(2) Sargan	0.002 0.014 0.999
Observations	420	Observations	420	Observations	330
Number of province	30	Number of province	30	Number of province	30
wald	91438	wald	53562	wald	58617
z-statistics in parentheses
* p < 0.01, p < 0.05, * p < 0.1

5.3 Threshold effect analysis

To investigate whether the impact of transportation infrastructure on the regional economy is a simple linear relationship. This research tests whether there is a threshold effect between them. First of all, as shown in Table 9, the triple threshold test is conducted, and the single threshold and double threshold are passed at a 1% significance level, but the p-value of the triple threshold is as high as 0.417, indicating that there is no triple threshold effect.

Table 9

Threshold effect test
Threshold	Fstat	Prob	Crit10	Crit5	Crit1
Single	149.50	0.000	53.060	58.9733	88.434
Double	89.29	0.000	30.144	33.981	53.220
Triple	31.89	0.417	103.512	123.878	163.214

Therefore, after passing the test of threshold effect, this section further analyzes the double threshold estimates. Table 10 presents the corresponding estimates of the double threshold and the corresponding confidence intervals.

Table 10

Threshold estimates and confidence intervals
Variables	Threshold	Threshold values	Prob	Crit10	Crit5	Crit1	95% conf. Interval
rf	Single	0.07	0.000	51.755	63.111	82.172	(0.060,0.080)
rf	Double	0.52	0.000	29.5679	34.319	44.851	(0.500,0.530)

From Table 11, the impact of transportation infrastructure on the regional economy is indeed not a simple linear relationship when the transportation infrastructure is in different intervals, and there is a significant double-threshold feature between the two. This is consistent with the conclusion reached by Deng et al. (2014). However, the difference is that when the transportation infrastructure rf is lower than the first threshold of 0.07, that is, the level of transportation infrastructure development is low, the coefficient of the transportation infrastructure front at this time is -4.013, which is significant at the 1% level. This shows that when transportation infrastructure is not fully built and transport investment is low, transportation infrastructure rf does not have a positive pull effect on the economy. On the contrary, it will have a deterrent effect on the growth of the regional economy at low levels of development. If the transportation infrastructure rf does not reach the second threshold and is between the first and second thresholds, $0.07\le \text{r}\text{f}\le 0.52$, its coefficient estimate improves from just − 4.013 to 1.062. It represents that when the transportation infrastructure is developing and the investment is increasing, it contributes to the regional economy. At the time of this interval, for every unit of transportation infrastructure enhancement, the corresponding regional economy is enhanced by 1.062 units. Once the transportation infrastructure reaches and exceeds the second threshold, i.e., $\text{r}\text{f}>0.52$. The impact of the transportation infrastructure on the regional economy decreases from 1.062 to 0.621, indicating that at a certain point, the transportation infrastructure does not provide a higher return. The relationship between them is an inverted U-shaped change. The early development level is low and cannot promote economic development, later increased investment in transportation infrastructure makes the return rate greatly enhanced. But this enhancement is not infinitely superimposed. After reaching the second threshold, although still has a role in promoting the elasticity of this time is reduced.

Table 11

Threshold model estimation results
Variables	lnGDP	Variables	lnGDP
employment	0.0561	lnforeign	0.0921***
	(0.035)		(0.017)
assetrate	0.00238*	rf1(rf < 0.07)	-4.013***
	(0.001)		(1.272)
lnpopu	1.744***	rf2(0.07 ≤ rf ≤ 0.52)	1.062***
	(0.282)		(0.186)
structure	0.471	rf3(rf > 0.52)	0.621***
	(0.531)		(0.150)
urban	5.326***	Constant	-15.20***
	(0.514)		(3.732)
Observations	540
Number of provinces	30
R-squared	0.957
Robust standard errors in parentheses
* p < 0.01, p < 0.05, * p < 0.1

6.1 Main results

The findings contribute to the understanding of the impact of transportation infrastructure on the regional economy. Based on panel data from thirty provinces in mainland China excluding Tibet from 2004 to 2021, the impact of transportation infrastructure on regional economy is explored from two main aspects. The first aspect is to consider the dynamic effect, using multiple linear regression to determine the lag order, using system GMM and difference GMM for estimation to effectively avoid the endogeneity problem, and comparing with the traditional OLS, FE, and RE. The empirical results prove that the transportation infrastructure still exerts a more positive impact on the regional economy after considering the long-term impact on the economy. The second aspect is to consider whether this impact may be more than a simple linear relationship, which is tested using a threshold model. Transportation infrastructure has an inverted U-shaped impact on the regional economy. When the transportation infrastructure is below the first threshold, it is not able to promote economic development quickly. The highest economic return is obtained when the development of transportation infrastructure is controlled between the first threshold and the second threshold. When it exceeds the second threshold, it still provides a boost to the economy, but not as much as when it is between the first and second thresholds.

6.2 Implications

6.2.1 Theoretical implications

Theoretically, it makes the following contributions to subsequent research. Firstly, it provides a new perspective on the construction of transportation infrastructure as an indicator. It is difficult to describe all of the development of transportation infrastructure from only a few dimensions. The dimensionality reduction method of Random Forests can incorporate as many relevant transportation variables as possible into this indicator, which makes the construction of transportation infrastructure more comprehensive. Secondly, it inspires future researches to pay attention to the role of the dependent variable in the relatively distant future when considering the dynamic impact. The multiple regression model can determine the number of lags of the dependent variable to be included in the model.

6.2.2 Practical implications

Because of the aforementioned series of conclusions, this study can provide certain suggestions for the formulation of transport construction policies for countries with the same national conditions. It should be necessary to review the layout of transport construction and rationally plan the construction of transportation infrastructure.

The first step is to vigorously develop transportation infrastructure that has not yet reached the first threshold. Transportation infrastructure below the first threshold in attracting the regional economy has not yet been demonstrated. In this zone, imperfect transportation infrastructure can hinder economic development. Second, the construction and maintenance of transportation infrastructure are maintained between the first and second thresholds. From the threshold model, investment between the first and second thresholds will yield the highest return. Finally, under the current situation of unbalanced transportation infrastructure construction throughout the country, the construction of transportation infrastructure in cities that have already crossed the second threshold can be slightly slowed down.

6.3 Limitations

Although efforts to bridge the research gaps, there are still many areas in this study that deserve further improvement. First, based on data availability, the current random forest includes fewer types of data. When measuring transportation infrastructure, more factors can be considered, such as inland waterways, air transportation, etc., to portray the quality of transportation services in multiple dimensions. Moreover, some of the data are not publicly released, thus the sample in this research contains incomplete information and there is room for further improvement. Finally, in model selection, as technology develops, more advanced models can be chosen to explore the effects of both.

Declarations of interest

There is no competing financial interests in this paper.

Author Contribution

Yunlin Wei: manuscript writing, data collection and analysis.Ke Lu: model construction, manuscript revision.

Acknowledgement

The study in this paper was jointly supported by research grants from National Nature Science Foundation of China (Grant NO.72003098), the Philosophy and Social Science Fund of Education Department of Jiangsu Province (Grant NO.2019SJA0155), Jiangsu Provincial Double-Innovation Doctor Program (Grant NO.202030599).

Availability of data and materials:

The data will not be shared since further research will be carried out.

Acheampong, A.O., Dzator, J., Dzator, M., Salim, R.: Unveiling the effect of transport infrastructure and technological innovation on economic growth, energy consumption and CO2 emissions. Technol. Forecast. Soc. Chang. 182, 121843doi (2022). 10.1016/j.techfore.2022.121843
Alam, K.M., Li, X., Baig, S., Ghanem, O., Hanif, S.: Causality between transportation infrastructure and economic development in Pakistan: An ARDL analysis. Res. Transp. Econ. 88, 100974 (2021). 10.1016/j.retrec.2020.100974
Alam, K.M., Li, X., Baig, S., Ghanem, O., Hanif, S.: Causality between transportation infrastructure and economic development in Pakistan: An ARDL analysis. Res. Transp. Econ. 88, 100974 (2021). 10.1016/j.retrec.2020.100974
Alotaibi, S., Quddus, M., Morton, C., Imprialou, M.: Transport investment, railway accessibility and their dynamic impacts on regional economic growth. Res. Transp. Bus. Manage. 43, 100702 (2022). 10.1016/j.rtbm.2021.100702
Arellano, M., Bond, S.: Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. Rev. Econ. Stud. 58(2), 277–297 (1991). 10.2307/2297968
Banerjee, A., Duflo, E., Qian, N.: On the road: Access to transportation infrastructure and economic growth in China. J. Dev. Econ. 145, 102442 (2020). 10.1016/j.jdeveco.2020.102442
Bian, F., Yeh, A.G.: Spatial–economic impact of missing national highway links on China’s regional economy. Transp. Res. Part D: Transp. Environ. 84, 102377 (2020). 10.1016/j.trd.2020.102377
Cascetta, E., Cartenì, A., Henke, I., Pagliara, F.: Economic growth, transport accessibility and regional equity impacts of high-speed railways in Italy: Ten years ex post evaluation and future perspectives. Transp. Res. Part A: Policy Pract. 139, 412–428 (2020). 10.1016/j.tra.2020.07.008
Chakrabarti, S.: Can highway development promote employment growth in India? Transp. Policy. 69, 1–9 (2018). 10.1016/j.tranpol.2018.05.009
Chen, A., Li, Y., Ye, K., Nie, T., Liu, R.: Does transport infrastructure inequality matter for economic growth? Evidence from China. Land. 10(8), 1–21 (2021). 10.3390/land10080874
Chen, Z., Li, Y., Wang, P.: Transportation accessibility and regional growth in the Greater Bay Area of China. Transp. Res. Part D: Transp. Environ. 86, 102453 (2020). 10.1016/j.trd.2020.102453
Deng, T., Shao, S., Yang, L., Zhang, X.: Has the transport-led economic growth effect reached a peak in China? A panel threshold regression approach. Transportation. 41, 567–587 (2014). 10.1007/s11116-013-9503-4
Dou, S., Xu, D., Keenan, R.J.: Effect of income, industry structure and environmental regulation on the ecological impacts of mining: An analysis for Guangxi Province in China. J. Clean. Prod. 400, 136654 (2023). 10.1016/j.jclepro.2023.136654
Feng, Q., Chen, Z., Cheng, C., Chang, H.: Impact of high-speed rail on high-skilled labor mobility in China. Transp. Policy. 133, 64–74 (2023). 10.1016/j.tranpol.2023.01.006
Gao, D., Li, S.: Spatiotemporal impact of railway network in the Qinghai-Tibet Plateau on accessibility and economic linkages during 1984–2030. J. Transp. Geogr. 100, 103332 (2022). 10.1016/j.jtrangeo.2022.103332
Gao, Y., Song, S., Sun, J., Zang, L.: Does high-speed rail connection really promote local economy? Evidence from China’s Yangtze River Delta. Rev. Dev. Econ. 24(1), 316–338 (2020). 10.1111/rode.12642
Ghannouchi, I., Ouni, F., Aloulou, F.: Investigating the impact of transportation system and economic growth on carbon emissions: Application of GMM System for 33 european countries. Environ. Sci. Pollut. Res. 30(39), 90656–90674 (2023). 10.1007/s11356-023-28595-6
González-González, E., Nogués, S., Stead, D.: Automated vehicles and the city of tomorrow: A backcasting approach. Cities. 94, 153–160 (2019). 10.1016/j.cities.2019.05.034
Guan, C., Hu, Q.: Does high-speed railway impact urban logistics industry agglomeration? Empirical evidence from China's prefecture-level cities. Socio-Economic Plann. Sci. 87, 101557 (2023). 10.1016/j.seps.2023.101557
Guan, C., Chen, L., Li, D.: Does the Opening of High-Speed Railway Improve High-Quality Economic Development in the Yangtze River Delta, China? Land. 12(8), 1629 (2023). 10.3390/land12081629
Han, D., Attipoe, S.G., Han, D., Cao, J.: Does transportation infrastructure construction promote population agglomeration? Evidence from 1838 Chinese county-level administrative units. Cities. 140, 104409 (2023). 10.1016/j.cities.2023.104409
He, D., Yin, Q., Zheng, M., Gao, P.: Transport and regional economic integration: Evidence from the Chang-Zhu-Tan region in China. Transport Policy, 79,193–203. doi: 193-203.10.1016/j.tranpol.2019.04.015. (2019)
Heuermann, D.F., Schmieder, J.F.: The effect of infrastructure on worker mobility: evidence from high-speed rail expansion in Germany. J. Econ. Geogr. 19(2), 335–372 (2019). 10.1093/jeg/lby019
Ke, X., Lin, J.Y., Fu, C., Wang, Y.: Transport infrastructure development and economic growth in China: Recent evidence from dynamic panel system. -GMM Anal. Sustain. 12(14), 5618 (2020). 10.3390/su12145618
Kong, Q., Li, R., Jiang, X., Sun, P., Peng, D.: Has transportation infrastructure development improved the quality of economic growth in China’s cities? A quasi-natural experiment based on the introduction of high-speed rail. Res. Int. Bus. Finance. 62, 101726 (2022). 10.1016/j.ribaf.2022.101726
Krugman, P.: Increasing returns and economic geography. J. Polit. Econ. 99(3), 483–499 (1991). 10.1086/261763
Li, C., Zhang, J., Lyu, Y.: Does the opening of China railway express promote urban total factor productivity? New evidence based on SDID and SDDD model. Socio-Economic Plann. Sci. 80, 101269 (2022). 10.1016/j.seps.2022.101269
Li, G., Ma, X., Song, Y.: Green building efficiency and influencing factors of transportation infrastructure in china: Based on three-stage super-efficiency sbm–dea and tobit models. Buildings. 12(5), 623 (2022). 10.3390/buildings12050623
Li, X., Ullah, S.: Caring for the environment: how CO2 emissions respond to human capital in BRICS economies? Environ. Sci. Pollut. Res. 29(12), 18036–18046 (2022). 10.1007/s11356-021-17025-0
Lin, X.: Multiple pathways of transportation investment to promote economic growth in China: a structural equation modeling perspective. Transp. Lett. 12(7), 471–482 (2020). 10.1080/19427867.2019.1635780
Lin, X.: Mechanism of the Fundamental Role of High-Speed Railway on Regional Development. In High-Speed Railways and New Structure of Socio-economic Development in China. Singapore: Springer Nature Singapore. 39–57 (2023). 10.1007/978-981-19-6387-2_2
Lu, H., Zhao, P., Hu, H., Zeng, L., Wu, K.S., Lv, D.: Transport infrastructure and urban-rural income disparity: A municipal-level analysis in China. J. Transp. Geogr. 99, 103292 (2022). 10.1016/j.jtrangeo.2022.103292
Lu, Y., Zhang, T., Qiu, S., Liu, X., Yu, X., Yao, H.: Carbon emissions reduction potentiality for railroad transportation based on life cycle assessment. High-speed Railway. (2023). 10.1016/j.hspr.2023.08.004
Magazzino, C., Mele, M.: On the relationship between transportation infrastructure and economic development in China. Res. Transp. Econ. 88, 100947 (2021). 10.1016/j.retrec.2020.100947
Maparu, T.S., Mazumder, T.N.: Transport infrastructure, economic development and urbanization in India (1990–2011): Is there any causal relationship? Transp. Res. Part A: Policy Pract. 100, 319–336 (2017). 10.1016/j.tra.2017.04.033
Meng, X., Gu, A., Wu, X., Zhou, L., Zhou, J., Liu, B., Mao, Z.: Status quo of China hydrogen strategy in the field of transportation and international comparisons. Int. J. Hydrog. Energy. 46(57), 28887–28899 (2021). 10.1016/j.ijhydene.2020.11.049
Nugraha, A.T., Prayitno, G., Situmorang, M.E., Nasution, A.: The Role Of Infrastructure In Economic Gro\Tth And Income Inequality In Indonesia. Econ. Sociol. 13(1), 102–115 (2020). 10.14254/2071-789X.2020/13-1/7
Okamoto, C., Sato, Y.: Impacts of high-speed rail construction on land prices in urban agglomerations: Evidence from Kyushu in Japan. J. Asian Econ. 76, 101364 (2021). 10.1016/j.asieco.2021.101364
Onifade, S.T., Gyamfi, B.A., Bekun, F.V., Altuntaş, M.: Significance of air transport to tourism-induced growth hypothesis in E7 economies: exploring the implications for environmental quality. Tourism: An. International Interdisciplinary Journal. 70(3), 339–353 (2022). 10.37741/t.70.3.1
Rehman, A., Ma, H., Ahmad, M., Işık, C., Ozturk, I.: Estimating interlinks of carbon emissions from transportation, industrialization, and solid/liquid fuels with economic progress: evidence from Pakistan. Int. J. Environ. Sci. Technol. 20(2), 1981–1996 (2023). 10.1007/s13762-022-04111-0
Romer, P.M.: Increasing returns and long-run growth. J. Polit. Econ. 94(5), 1002–1037 (1986). 10.1086/261420
Romer, P.M.: Endogenous technological change. Journal of Political Economy, 98(5, Part 2), S71-S102. (1990). 10.1086/261725
Shao, S., Tian, Z., Yang, L.: High speed rail and urban service industry agglomeration: Evidence from China's Yangtze River Delta region. J. Transp. Geogr. 64, 174–183 (2017). 10.1016/j.jtrangeo.2017.08.019
Shen, Q., Pan, Y.: Explaining and modeling the impact of industrial co-agglomeration on regional economic growth in China: integrated a quality concern of night-time light perspective. Environ. Sci. Pollut. Res. 1–26 (2023). 10.1007/s11356-023-28709-0
Solow, R.M.: Technical change and the aggregate production function. Rev. Econ. Stat. 39(3), 312–320 (1957). 10.23:07/1926047
Sun, C., Zhang, W., Fang, X., Gao, X., Xu, M.: Urban public transport and air quality: Empirical study of China cities. Energy Policy. 135, 110998 (2019). :/10.1016/j.enpol.2019.110998
Wang, C., Chen, J., Li, B., Chen, N., Wang, W.: Impact of high-speed railway construction on spatial patterns of regional economic development along the route: A case study of the Shanghai–Kunming high-speed railway. Socio-Economic Plann. Sci. 87, 101583 (2023). 10.1016/j.seps.2023.101583
Wang, C., Chen, J., Li, B., Chen, N., Wang, W.: Impact of high-speed railway construction on spatial patterns of regional economic development along the route: A case study of the Shanghai–Kunming high-speed railway. Socio-Economic Plann. Sci. 87, 101583 (2023). 10.1016/j.seps.2023.101583
Wang, C., Kim, Y.S., Kim, C.Y.: Causality between logistics infrastructure and economic development in China. Transp. Policy. 100, 49–58 (2021). 10.1016/j.tranpol.2020.10.005
Wang, F., Liu, Z., Xue, P., Dang, A.: High-speed railway development and its impact on urban economy and population: A case study of nine provinces along the Yellow river, China. Sustainable Cities and Society. 87, 104172 (2022). 10.1016/j.scs.2022.104172
Wang, N., Zhu, Y., Yang, T.: The impact of transportation infrastructure and industrial agglomeration on energy efficiency: Evidence from China’s industrial sectors. J. Clean. Prod. 244, 118708 (2020). 10.1016/j.jclepro.2019.118708
Wang, Y., Wang, J., Chen, J., Liu, K.: Optimal Location of Emergency Facility Sites for Railway Dangerous Goods Transportation under Uncertain Conditions. Appl. Sci. 13(11), 6608 (2023). 10.3390/app13116608
Wong, Z., Chen, A., Shen, C., Wu, D.: Fiscal policy and the development of green transportation infrastructure: the case of China's high-speed railways. Economic Change and Restructuring. 55(4), 2179–2213 (2022). 10.1007/s10644-021-09381-1
Wu, Y., Lee, C.C., Lee, C.C., Peng, D.: Geographic proximity and corporate investment efficiency: evidence from high-speed rail construction in China. J. Banking Finance. 140, 10651 (2022). 10.1016/j.jbankfin.2022.106510
Xue, X., Jia, Y., Tang, Y.: National Economic Benefits of Railway Transportation Analyzed Using Data Mining. Transp. Res. Rec. 2677(5), 870–892 (2023). 10.1177/0361198122113759
Yosritzal, Y., Aziz, R., Noer, M., Putri, D.O., Sani, A.: Determinant factors in increasing the livability of the city of Padang from the perspective of transportation. SINERGI. 27(1), 73–80 (2023). 10.22441/sinergi.2023.1.009
Zhang, H., Xie, T.: A key to urban economic growth or an unnecessary burden? Opening airports in small and medium-sized cities. Cities. 133, 104105 (2023). 10.1016/j.cities.2022.104105
Zhang, J., Zhang, Y.: The relationship between China's income inequality and transport infrastructure, economic growth, and carbon emissions. Growth and Change. 52(1), 243–264 (2021). 10.1111/grow.12472
Zhang, X., Hu, Y., Lin, Y.: The influence of highway on local economy: Evidence from China's Yangtze River Delta region. J. Transp. Geogr. 82, 102600 (2020). 10.1016/j.jtrangeo.2019.102600
Zhang, Y., Cheng, L.: The role of transport infrastructure in economic growth: Empirical evidence in the UK. Transp. Policy. 133, 223–233 (2023). 10.1016/j.tranpol.2023.01.017
Zhang, Y., Liu, J., Wang, B.: The impact of High-Speed Rails on urban expansion: An investigation using an SDID with dynamic effects method. Socio-Economic Plann. Sci. 82, 101294 (2022). 10.1016/j.scs.2022.104172
Zhao, J., Yan, J., Ran, Q., Yang, X., Su, X., Shen, J.: Does the opening of high-speed railways improve urban livability? Evidence from a quasi-natural experiment in China. Socio-Economic Plann. Sci. 82, 101275 (2022). 10.1016/j.seps.2022.101275
Zhao, L.: The Economic and Environmental Impacts of Urbanization in China. Highlights in Business, Economics and Management. 5, 142–146 (2023). 10.54097/hbem.v5i.5041
Zhu, X., Qian, T., Wei, Y.: Do high-speed railways accelerate urban land expansion in China? A study based on the multi-stage difference-in-differences model. Socio-Economic Plann. Sci. 71, 100846 (2020). 10.1016/j.seps.2020.100846

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

Exploring the Impact of Transportation Infrastructure on Regional Economy in China based on GMM and Threshold Effects

Status:

Version 1

Abstract

Figures

1 Introduction

2 Literature review

2.1 Relationship between transportation infrastructure and the economy

2.2 Indicator construction for transportation infrastructure

2.3 Relevant empirical methods

2.4 Summary

3 Variable selection and data sources

3.1 Variable selection

3.2 Data sources

4 Methodology

4.1 Random Forest

4.2 Generalized method of moments

4.3 Threshold models

5 Results

5.1 Correlation test

5.2 GMM regression analysis

5.2.1 OLS-FE-RE analysis

5.2.2 System GMM and Difference GMM

5.2.3 Robustness Tests

5.3 Threshold effect analysis

6 Conclusion

6.1 Main results

6.2 Implications

6.2.1 Theoretical implications

6.2.2 Practical implications

6.3 Limitations

Declarations

Declarations of interest

Author Contribution

Acknowledgement

Availability of data and materials:

References

Additional Declarations

Status:

Version 1