Can international railway express improve green total factor productivity?

Abstract

The establishment of China-Europe Railway Express (CR Express) within the framework of the “One Belt and One Road” initiative has provided a new avenue for economic openness. However, this development has also led to industrial agglomeration along the railway, which may impose potential environmental pressures on the cities along its route. Utilizing a panel data covering prefecture-level cities from 2008 to 2019, this research employs the staggered difference-in-differences (DID) method to examine the effects of CR Express on urban green total factor productivity (GTFP) and the underlying mechanisms. The findings indicate that the opening of CR Express significantly enhances the GTFP of node cities. Specifically, the introduction of freight trains on average increases the GTFP of node cities by 3.3%. However, the impact varies across different channels, regions, and coverage areas. Regarding the influencing mechanisms, it is found that CR Express improves the GTFP of node cities through the facilitation of green innovation and industrial agglomeration. Given these results, we emphasize the importance of continuously promoting, supporting, and guiding more inland cities to actively participate in the Economic Belt along the CR Express, either directly or indirectly. Additionally, node cities government should encourage green innovation and foster a favorable business environment to further enhance the green innovation capabilities of cities and promote industrial agglomeration, ultimately advancing the high-level economic openness of cities.

Introduction

Since the opening of the first “Chongqing-Xinjiang-Europe” line from Chongqing to Duisburg, Germany in 2011, the CR Express have received extensive praise and active participation from countries along the route, and the coverage of the routes continues to expand and extend. As of the end of 2021, the CR Express have made a total of 49,000 trips, transporting 4.432 million TEUs of goods, and covering 180 cities in 23 countries, with a logistics service network covering the entire Eurasian continent. It has become a popular international public product. The operation of the ‘green skin’ carriages, which shuttle closely between the continents of Asia and Europe, has indeed promoted the industrial efficiency of node cities (Zhang and Gong 2021; Fang and Xie 2022). However, while industrial efficiency has been improved, it has also brought about environmental pollution problems for cities. For example, the annual average air quality in Chengdu declined after the opening of the CR Express. The PM2.5 index increased from 55.59 μg/m³ in 2012 to 63.66 μg/m³ in 2013, with an increase of approximately 14%. Similarly, Hefei, Chongqing, and Jinan saw increases of 7.0%, 5.3%, and 1.8%, respectively, in PM2.5 annual average values in the year the CR Express was opened. Similar situations have also occurred in other node cities. The CR Express seems to have caused serious urban environmental problems while assisting local industrial agglomeration. The output of both positive and negative aspects has grown simultaneously, making people think about the impact of the CR Express on the overall socioeconomic value. Based on this concern, we adopt the Data Envelopment Analysis (DEA) model based on the Slacks-Based Measure (SBM) method. to examine output and pollution emissions in the same system, effectively evaluating the true value brought by the train to urban industries. The core questions of this research are: as a long-distance freight transport that crosses regions and borders, could the operation of the CR Express improve the region’s green production efficiency? What is the core mechanism of its impact? The exploration of these questions has important practical significance for promoting high-quality opening-up in China.

This paper is based on data from prefecture-level cities in China and employs the overlapping Difference-in-Differences (DID) method to investigate the impact of the opening of the CR Express on urban green total factor productivity (GTFP) and its underlying mechanism. The aim is to provide a more comprehensive understanding of urban high-quality development within the context of the Belt and Road Initiative. The empirical findings indicate that the opening of the CR Express has significantly enhanced the GTFP of node cities during the sample period. However, the promotion effect on urban GTFP is significant in the East-West and Central-Western corridors and regions. Additionally, the number of foreign node cities that domestic node cities can access positively influences the GTFP of domestic node cities. Furthermore, the CR Express contributes to the improvement of the GTFP of node cities through green innovation and industrial agglomeration.

The main contributions of this paper are as follows: Firstly, this paper systematically analyzes and examines the impact of the opening of the CR Express on urban GTFP from the perspective of foreign trade routes. It provides valuable insights for promoting the high-quality development of the Belt and Road Initiative and enriches the existing research on urban high-level opening-up. Secondly, we explored the impact of green innovation and industrial agglomeration resulting from the CR Express on the GTFP of node cities, which differs from previous studies in terms of the underlying mechanism. Thirdly, this research further enriches the theory of transportation economics. The CR Express serves as a long-distance international trade tool and emerging freight transportation mode that transcends regions and borders. This paper quantitatively analyzes the GTFP of node cities along the CR Express route, shedding light on the impact and mechanism of the train on urban GTFP. This contributes to a deeper understanding of transportation economics. Lastly, we combine the latest theoretical and empirical literature, formally applies the decomposition and the heterogeneity treatment effect tests in research practice, and to some extent, mitigates the biased estimates produced by traditional difference-in-differences (DID) methods.

The remainder of the paper is structured as follows: the second part provides literature review. Part 3 provides theoretical analysis on the mechanism between CR Express, green innovation level, and GTFP, and put forward the related hypothesis. Part 4 describes empirical model and data. Part 5 presents the empirical analysis and robustness tests, and part 6 summarizes the paper.

Literature review

Research on the relationship between transportation infrastructure and industrial efficiency has a long history, and scholars have conducted extensive discussions on this topic. The majority of studies believe that transportation infrastructure generally has a positive impact on total factor productivity (TFP). The main mechanisms include reducing transportation costs (Aitken and Harrison 1999), attracting more domestic and foreign investments (Bustos 2011; Liu and Lu 2015), promoting innovation (Guo and Bai 2019), optimizing the efficiency of capital or labor allocation (Xu 2020), and enhancing market accessibility (Liu et al. 2020). However, the impact varies significantly among regions and different transportation facilities. For example, Huang and Sun (2019) suggest that high-speed rail has a positive impact on the productivity of enterprises in the eastern and central regions of China, while it has a suppressive effect on enterprises in the western region due to the suction effect. Li et al. (2022) found that the impact of China’s high-speed rail on urban TFP is mainly evident in coastal hub cities, inland hub cities, and inland cities near the hubs, but the impact on coastal cities near the hubs is not significant. Zhao and Wang (2020) argue that the driving effect of high-speed rail construction on regional TFP is stronger than that of conventional railways, mainly because high-speed rail has a faster speed, accelerating the flow of capital, labor, and technology between regions.

As to the impact of transportation infrastructure on regional environmental quality, scholars hold contrasting views. Some researchers believed that the construction of transportation infrastructure contributes to regional environmental degradation. This degradation primarily stems from the infrastructure construction process and the energy consumption associated with economic growth (Chen 2020). Using prefecture-level city data, Luo et al. (2019) found that the high-speed rail network has, to some extent, exacerbated the imbalance of ecological efficiency in cities. On the other hand, some studies argued that transportation infrastructure construction leads to improved environmental quality. Song et al. (2016) affirmed the positive impact of railways on the environment in China, asserting that railways are the most efficient and environmentally friendly mode of transportation. Yang et al. (2019) also observed a significant reduction of 7.35 percentage points of the polluting gases emissions in China due to high-speed rail, which attributed to technological, allocation, and substitution mechanisms. Some studies argued that treating the impact of transportation construction on the environment as solely promoting or harming is inadequate. This impact exhibits considerable heterogeneity, depending on factors such as the level of economic development. For instance, Sun and Zhang (2021) highlighted that the opening of high-speed rail accelerated the transfer of low value-added industries with high pollution emissions from affluent cities to relatively impoverished cities. As a result, pollution emissions decreased in affluent cities but increased in impoverished cities. Similarly, Cai et al. (2021) suggested that the opening of high-speed rail made it more challenging for pollution-intensive enterprises to enter developed node cities, while underdeveloped cities attracted more of these enterprises, thereby exacerbating environmental pollution in underdeveloped cities.

The CR Express could be classified within the domain of trade, as it facilitates trade development and openness, thereby impacting the environment of trading countries. This area of research remains a subject of ongoing investigation. Current studies suggest that the specific mechanisms through which trade influences environmental quality primarily manifest in three dimensions. Firstly, technology plays a crucial role, particularly in terms of the enhancement of environmental technology in host countries due to foreign direct investment (FDI). Relevant research also indicates that international trade offers developing countries opportunities to adopt new technologies, thereby reducing emissions of sulfur dioxide, nitrogen dioxide (Frankel and Rose 2005), and carbon (Perkins 2012) generated during the production process, and thus preventing the emergence of “pollution havens.” Secondly, in order to attract investment, latecomer regions engage in a competition to lower environmental standards, exacerbating environmental issues in developing countries and creating “pollution havens” (Levinson and Taylor 2008). Research conducted by Chang et al. (2022) on countries along the Belt and Road Initiative also confirms the “pollution haven” effect arising from variations in environmental regulatory levels on carbon emissions caused by FDI. Countries with inadequate regulatory frameworks are likely to become havens for energy-intensive enterprises. Thirdly, trade structure plays a significant role. If a country’s foreign trade exhibits a structure characterized by “high-energy consumption, high pollution, and resource-based” exports, its exported manufactured products will have low value-added, and the growth of foreign trade will further deplete non-renewable resources and contribute to increased environmental pollution, thereby turning the country into a “pollution haven” (Tang and Zhou 2017). However, Wang and Zhang (2020) argue that optimizing trade structure allows companies to accumulate innovative resources, seize more technological opportunities, promote green innovation activities, and ultimately improve the environment.

Existing research on the CR Express pertaining to the theme of this research primarily focuses on assessing its impact on industrial efficiency. Zhang and Gong (2021) discovered that the establishment of the CR Express has led to a noteworthy enhancement in the TFP of node cities and their surrounding areas. Fang and Xie (2022) identified a significant positive influence of the CR Express on the TFP of the central and western regions of China, as well as large cities and their peripheries. However, its influence on the eastern region and smaller cities was found to be less significant. Furthermore, Gu and Zhang (2022), utilizing data from manufacturing listed companies, demonstrated that the CR Express has significantly bolstered the TFP of companies located in node cities by stimulating technological innovation and improving resource allocation efficiency.

In summary, while previous research has provided important theoretical and practical insights into the development of the CR Express, there is still room for further expansion. Firstly, although extensive research has been conducted on the relationship between transportation infrastructure and industrial efficiency, transportation infrastructure and pollution, and trade and environmental quality, the conclusions of these studies still remain controversial. As railways are the largest capacity and the most cost-effective mode of land transportation, they have unique impacts on socioeconomic factors. It is necessary to specifically research the impact of railway trade and the opening of the CR Express with a focus on foreign trade on urban GTFP from the perspective of city socioeconomic development. This will enable us to analyze and summarize the regularities and provide valuable insights. Secondly, current research on the efficiency of the CR Express mainly focuses on productivity and lacks discussions that incorporate environmental factors into the analysis of industrial efficiency. In fact, studying this issue can further explore the real impact of the CR Express in facilitating high-quality urban development under the Belt and Road Initiative. Based on the previous research, this research explores the impact of the opening of the CR Express on the GTFP of node cities. By analyzing its mechanisms, conclusions and policy implications are drawn, which will support the implementation of the country’s high-quality opening-up policy.

Mechanisms and research hypotheses

The direct effect

The establishment and operation of the CR Express enhance trade convenience and reduce trade costs for cities situated along its route. This, in turn, attracts increased foreign direct investment (FDI) and facilitates the influx of high-end industries into node cities, thereby optimizing the allocation of capital, labor, and technological resources. Building upon the approach adopted by Aschauer (1989) and Barro (1990), this research employs the Cobb-Douglas production function to examine the specific direct effect of the CR Express on the GTFP of node cities. In this analysis, the output Y is determined by the production technology A, labor input L, energy input E, material input M, transportation infrastructure (represented by the CR Express) T, and capital input K (excluding transportation infrastructure), as demonstrated by Eq. (1).

1

$Y=A^{\epsilon }{L}^{\alpha }{K}^{\beta }{E}^{\delta }{M}^{\theta }{T}^{\sigma }$Where 0 < α, β, δ, ε, θ, σ < 1, representing the output elasticities of technology, labor, capital, energy, materials, and transportation infrastructure (CR Express), respectively.

Furthermore, the emission of pollutants P in the production process is proportional to the inputs of energy E and materials M, as energy and materials consumption generate pollutants. The emission of pollutants P is inversely proportional to the technological level A, as higher production or pollution control technologies reduce pollutant emissions. This relationship is shown in Eq. (2).

Additionally, the relationship between pollutant emissions P and total cost TC can be represented by Eq. (3).

2

$P=E^{\phi }{M}^{\omega }/A^{\rho }$

3

$TC=A{P}_A+L{P}_L+K{P}_K+E{P}_E+M{P}_M+T{P}_T+P{P}_P$Where in Eq. (3), P_A, P_L, P_K, P_E, P_M, P_T, and P_P represent the prices of technology, labor, capital, energy, materials, transportation infrastructure, and the cost per unit of pollution control, respectively. It should be noted that since production technology and pollution control technology are often closely related, we set ε = ρ. Additionally, as energy and materials consumption is primarily used for production before generating pollution, we have δ > φ and θ > ω. Therefore, the objective is to maximize net profits:

4

$\max \left\{Y-TC\right\}=\max \left\{A^{\epsilon }{L}^{\alpha }{K}^{\beta }{E}^{\delta }{M}^{\theta }{T}^{\sigma }-\left(A{P}_A+L{P}_L+K{P}_K+E{P}_E+M{P}_M+T{P}_T+\frac{E^{\phi }{M}^{\omega }}{A^{\rho }}{P}_P\right)\right\}$

According to the first-order conditions, we have Eq. (5) and Eq. (6):

5

$Y=\frac{P_{L}L}{\alpha };Y=\frac{P_{K}K}{\beta };Y=\frac{P_{T}T}{\sigma }$

6

$E=\frac{\delta Y-\phi P{P}_P}{P_E};M=\frac{\theta Y-\omega P{P}_P}{P_M};A=\frac{\epsilon Y+\rho P{P}_P}{P_A}$

By substituting Eq. (5) into Eq. (3) and letting ε + α + β + δ + θ + σ = a, ω + φ - ρ = b, we can derive Eq. (7) considering the scale effect on output and the decreasing effect on pollutant emissions due to technological progress:

7

$TC=\left(\epsilon +\alpha +\beta +\delta +\theta +\sigma \right)Y+\left(1+\rho -\omega -\phi \right)P_{P}P=aY+\left(1-b\right)P_{P}P$

According to Eq. (6), we can determine the optimal output and optimal pollutant emissions represented in terms of energy inputs and technology:

8

$\left\{\begin{array}{c}\hfill Y^{*}=\frac{\rho E{P}_E+\phi A{P}_A}{\rho \delta +\epsilon \phi }\hfill \\ \hfill P^{*}=\frac{\delta A{P}_A-\epsilon E{P}_E}{\left(\rho \delta +\epsilon \phi \right)P_P}\hfill \end{array}\right)$

Since GTFP measures TFP considering environmental constraints, GTFP could be understood as the net output efficiency of unit cost. In this paper, the net output efficiency of unit cost (e*) is expressed as:

9

$e^{*}=\frac{Y^{*}-P^{*}}{T{C}^{*}}=\frac{Y^{*}-P^{*}}{a{Y}^{*}+\left(1-b\right)P_P{P}^{*}}$

Since the prices of each factor are exogenously given, we could standardize them as P_P = P_E = P_A = 1. Then, using Eqs. (8) and (9), we can obtain:

10

$e^{*}=\frac{\left(\rho +\epsilon \right)E+\left(\phi -\delta \right)A}{\left[a\rho -\left(1-b\right)\epsilon \right]E+\left[a\phi +\left(1-b\right)\delta \right]A}=\frac{\left(\rho +\epsilon \right)\frac{E}{A}+\left(\phi -\delta \right)}{\left[a\rho -\left(1-b\right)\epsilon \right]\frac{E}{A}+\left[a\phi +\left(1-b\right)\delta \right]}$

Given the equations aρ-(1-b)ε = W and aφ + (1-b)δ = V, it follows that V > W. Taking the derivative of e* with respect to E/A yields:

11

$\frac{d\left(e^{*}\right)}{d\left(\frac{E}{A}\right)}=\frac{\left(\rho +\epsilon \right)V-\left(\phi -\delta \right)W}{{\left(W\frac{E}{A}+V\right)}^2}>0$

From Eq. (11), it can be inferred that d(e*)/d(E/A) > 0. Therefore, an increase in E/A will promote the net output efficiency of unit cost. The progress in production technology brought about by the CR Express will increase the energy carried by each unit of technology, thereby improving the net output efficiency of unit cost. As a result, the opening of the CR Express will enhance the GTFP of the node cities.

In addition, from Eqs. (6) or (8), we can obtain:

12

$P^{*}=\frac{\delta Y^{*}-E}{\phi }=\frac{A-\epsilon Y^{*}}{\rho }$

By combining Eqs. (9) and (12), we can obtain Eqs. (13) and (14):

13

$e^{*}=\frac{\left(\phi -\delta \right)Y^{*}+E}{\left[a\phi +\delta \left(1-b\right)\right]Y^{*}-\left(1-b\right)E}$

14

$e^{*}=\frac{\left(\rho +\epsilon \right)Y^{*}-A}{\left[a\rho +\epsilon \left(1-b\right)\right]Y^{*}+\left(1-b\right)A}$

Furthermore, taking the partial derivative of Eq. (14) with respect to T, we have:

15

$\frac{\partial e^{*}}{\partial T}=E\frac{\partial A}{\partial T}\left(a+b-1\right)\left(\delta \rho -\phi \epsilon \right)$

As mentioned earlier, a > 1, 0 < b < 1, therefore a + b > 1. ε = ρ, δ > φ, thus ρδ > φε. The opening of the CR Express can attract foreign investment (Xiao and Ye 2022), and foreign investment has a technology spillover effect. In fact, the opening of the CR Express has promoted regional innovation level and innovation efficiency improvement (Li et al. 2020). Therefore, ∂A/∂T > 0. Finally, this paper concludes that ∂e*/∂T > 0, indicating that the opening of the CR Express promotes the efficiency of unit net output e*.

Similarly, taking the partial derivatives of Eq. (14) with respect to L and K, we have:

16

$\left\{\begin{array}{c}\hfill \frac{\partial e^{*}}{\partial L}=E\frac{\partial A}{\partial L}\left(a+b-1\right)\left(\delta \rho -\phi \epsilon \right)\hfill \\ \hfill \frac{\partial e^{*}}{\partial K}=E\frac{\partial A}{\partial K}\left(a+b-1\right)\left(\delta \rho -\phi \epsilon \right)\hfill \end{array}\right)$

Capital and labor optimization and iteration will promote technological progress, therefore ∂A/∂L > 0, ∂A/∂K > 0, so ∂e*/∂L > 0, ∂e*/∂K > 0. The opening of the CR Express brings resource optimization and will also promote the efficiency of unit net output e*. In conclusion, this paper proposes Hypothesis 1.

Hypothesis 1: The opening of the CR Express is beneficial to improving urban GTFP.

CR Express, green innovation, and GTFP of node cities

The opening of the CR Express contributes to the enhancement of the green innovation level in node cities, as supported by the research conducted by Grossman and Krueger (1995). The concept of green technology, introduced by Brawn and Wield (1994), is considered an innovative model for promoting sustainable economic development. The CR Express plays a significant role in promoting the green innovation level of cities in two main aspects. Firstly, it improves the innovation environment of node cities. The establishment of the CR Express leads to improvements in foreign policies, infrastructure, and marketization levels of node cities, providing a favorable environment for innovation development, as noted by Xiao and Ye (2022). Additionally, the “2021 CR Express Development Report” emphasizes the establishment of international platforms for green technology cooperation and exchange, such as green technology exchange and transfer bases, technology demonstration and promotion bases, and science and technology parks, which are continuously supported by the CR Express. These platforms and bases greatly enhance the innovation environment of node cities. Secondly, the CR Express facilitates the flow of innovation factors. In terms of human resource flow, the operation of the CR Express leads to increased connectivity between node cities and cities along the route (Liu et al. 2022), thereby driving regional economic development. Cities with improved infrastructure and convenient transportation tend to attract high-quality human capital, thus promoting the innovation level of node cities, as indicated by Debrezion et al. (2011). Ji and Yang (2020) also found a significant increase in the number of patent authorizations and applications for companies along the route after the opening of high-speed rail, primarily due to the increase in the proportion of employees with undergraduate and above degrees and technical staff in the labor force of node city enterprises. Similar studies have shown that high-speed rail promotes regional innovation levels through the migration of human capital (Wang et al. 2020). In terms of capital flow, the opening of the CR Express attracts more foreign investment to node cities (Zhang et al. 2019). In addition to promoting the flow of technological factors, foreign investment also provides financial support for green innovation research and development. Li et al. (2020) further demonstrated that the CR Express breaks down barriers between cities along the route, facilitates the flow of innovation factors, and enhances the technological innovation level of node cities. The improvement in the innovation level not only promotes economic development but also helps suppress environmental pollution, particularly through the enhancement of the green innovation level. In summary, the CR Express enhances the green innovation level of node cities by improving the innovation environment and promoting the flow of innovation factors, thus enhancing the GTFP of node cities. Therefore, this paper proposes Hypothesis 2.

Hypothesis 2: The CR Express promotes the GTFP of node cities by improving their level of green innovation.

CR Express, industrial agglomeration, and GTFP of node cities

The characteristics of the European market and the transportation facilitated by the CR Express contribute to the concentration of high-end industries in node cities along the route. These industries often exhibit “high output, low pollution” characteristics, leading to an increase in the GTFP of these cities. For instance, the CR Express has attracted high-end enterprises such as HP, Asus, Bosch, Porsche, LG, and Danone to Chongqing, while Xi’an has attracted companies like Konka and Guanjie in the high-end manufacturing sector. According to Dunning’s (1977) theory of international production trade-offs, companies investing internationally require three fundamental elements: ownership advantage, internalization advantage, and location advantage. Node cities along the CR Express route possess abundant resource elements, efficient logistics networks, preferential tax policies, relatively low trade costs, and favorable business environments, making them attractive for industry agglomeration. Moreover, as the CR Express primarily serves major cities in developed European countries, consumers in these destinations exhibit strong purchasing power and a preference for high-quality, high-end products. In terms of transportation costs, rail transport is approximately 1/5 the price of air transport and takes about 1/4 of the time compared to sea transport. Consequently, the CR Express is highly appealing for transporting high-end products that are less sensitive to natural conditions, require timely delivery, have a certain volume, and are less affected by transportation costs. With the rapid development of new service formats like cross-border e-commerce trains, postal trains, and “train + supermarket,” the range of goods transported to Europe via the train is constantly expanding to include mobile phones, home appliances, computers, vehicles, machinery, electronic products, high-end fashion, furniture, and epidemic prevention materials. These goods predominantly originate from high-end industries. It is worth noting that the concentration of high-end industries may also lead to industrial substitution effects, where low value-added industries are gradually replaced by high value-added ones. When the level of industrial agglomeration surpasses a certain threshold, it becomes conducive to improving environmental pollution (Yang 2015). Furthermore, industries concentrated in industrial parks along the CR Express route can achieve centralized pollution control, which is more effective and efficient compared to decentralized control, while also benefiting from economies of scale in pollution prevention and control. The “2021 CR Express Development Report” also underscores the establishment of green industrial cooperation demonstration bases at transportation nodes of the CR Express and the exploration of the “train + industrial park” development model. Thus, industrial agglomeration leads to a greater increase in output than in pollution emissions, ultimately boosting the GTFP of node cities. Additionally, the scale effect and technology spillover effect resulting from increased levels of industrial agglomeration make the contribution of transportation infrastructure to GTFP more prominent (Tan et al. 2022). Based on these findings, this paper proposes Hypothesis 3.

Hypothesis 3: The CR Express promotes the GTFP of node cities through industrial agglomeration.

The basic framework of theoretical analysis in this paper is illustrated in Fig. 1.

Fig. 1 [Images not available. See PDF.]

Analysis framework.

Model construction and data description

Model specification

This paper focuses on Chinese prefecture-level cities as the research object. Cities with a large amount of missing data are deleted, resulting in a final sample of 280 cities and 3360 observations from 2008 to 2019. Among them, 58 cities that have CR Express services by the end of 2019 are selected as the experimental group, while the remaining 222 cities are the “control group”. The final model specification is as follows:

17

$GTF{P}_{it}={\alpha }_0+{\beta }_{1}trea{t}_i\times pos{t}_{it}+{\beta }_2\sum X_{it}+{\mu }_i+{\phi }_t+{\epsilon }_{it}$GTFP_it represents the GTFP of city i in year t. treat_i is a dummy variable for the experimental group, taking the value of 1 if the city has opened the CR Express service, and 0 otherwise. post_it is a dummy variable for the opening time, taking the value of 1 for the year and subsequent years if the city opens the service in the first half of the year, and taking the value of 1 for the next year and subsequent years if the city opens the service in the second half of the year. The years before the opening year are assigned a value of 0. treat_i×post_it is the interaction term, and the coefficient β₁ of this term represents the impact of CR Express service on the GTFP of cities before and after its opening. The significance and direction of this coefficient are the main focus of this paper. X_it represents a series of control variables, which will be specifically explained in the following section. μ_i and φ_t represent city fixed effects and time fixed effects, respectively, and ε_it is the error term.

Variable selection

Dependent variable

The dependent variable GTFP_it is calculated using the SBM model, following the approach of Lin and Meng (2021). The fixed asset stock of cities is calculated using the perpetual inventory method (adjusting the fixed asset investment price index of the province where the city is located in 2003 as the base year). Energy input and labor input are represented by the total electricity consumption and the number of employed persons at the end of the year, respectively, from the “China City Statistical Yearbook”. GDP is considered as the expected output (adjusted using the GDP price index of the province where the city is located in 2003 as the base year). Industrial sulfur dioxide emissions, wastewater emissions, and particulate matter emissions are considered as non-desired outputs.

This paper treats each city as a decision-making unit to construct the production frontier. It is assumed that each decision-making unit uses N types of inputs x = (x₁, …, x_N) ∈ R⁺_N to produce M types of desired outputs y = (y₁, …, y_M) ∈ R⁺_N, while accompanying I types of undesired outputs b = (b₁, …, b_I) ∈ R⁺_I. In each period t = 1, …, T, the kth = 1, …, K city’s production possibility set is (x^k,t, y^k,t, b^k,t), which satisfies the free disposability, zero combination axiom, and weak disposability of outputs hypothesis. Data Envelopment Analysis (DEA) can be used for model processing:

18

$\begin{array}{c}{P}^t\left(x^t\right)=\left\{\left(y^t,b^t\right):\sum _{k=1}^K{\lambda }_k^t{y}_{km}^t\ge y_{km}^t,\forall m;\sum _{k=1}^K{\lambda }_k^t{b}_{ki}^t\right)\hfill \\ \left(=b_{ki}^t,\forall i;;\sum _{k=1}^K{\lambda }_k^t{x}_{kn}^t\le x_{kn}^t,\forall n;\sum _{k=1}^K{\lambda }_k^t=1,{\lambda }_k^t\ge 0,\forall k\right\}\hfill \end{array}$Here, ${\lambda }_k^t$ represents the weight of each cross-sectional observation, ∑^K_k = ₁λ^t_k = 1, λ‘ ≥ 0, ∀k implies that the production technology is variable returns to scale (VRS); if this constraint is removed, it indicates constant returns to scale (CRS).

According to Fukuyama and Weber (2009), the SBM directional distance function considering the energy environment is defined as:

19

$\begin{array}{c}{S}_V^t\left(x^{t,k},y^{t,k^{\prime }},b^{t,k},g^x,g^y,g^b\right)=\underset{s^x,s^y,s^b}{\max }\left[\frac{1}{N}\sum _{n=1}^N\frac{s_n^x}{g_n^x}+\frac{1}{M+I}\left(\sum _{m=1}^M\frac{s_m^y}{g_m^y}+\sum _{i=1}^I\frac{s_i^b}{g_i^b}\right)\right]/2\hfill \\ \mathrm{s}\mathrm{.t}\mathrm{.}\sum _{k=1}^K{\lambda }_k^t{x}_{kn}^t+s_n^x=x_{k^{\prime }n}^t,\forall n;\sum _{k=1}^K{\lambda }_k^t{y}_{km}^t-s_m^y\hfill \\ =y_{k\prime m}^t,\forall m;\sum _{k=1}^K{\lambda }_k^t{b}_{ki}^t+s_i^b=b_{k\prime i}^t,\forall i;\hfill \\ \sum _{k=1}^K{\lambda }_k^t=1,{\lambda }_k^t\ge 0,\forall k;s_n^x\ge 0,\forall n;s_m^y\ge 0,\forall m;s_i^b\ge 0,\forall i\hfill \end{array}$$S_v^t$ represents the directional distance function under VRS, and the directional distance function under CRS is denoted by $S_c^t$; (x^t,k, y^t,k, b^t,k), (g^x, g^y, g^b), and (s^x_n, s^y_m, s^b_i) represent the input and output vectors, direction vectors, and slack vectors of city K’, respectively. (s^x_n, s^y_m, s^b_i) being greater than zero indicates that actual inputs and pollution exceed the boundary inputs and pollution, while actual outputs are less than the boundary outputs. Therefore, (s^x_n, s^y_m, s^b_i) represents the amount of overuse of inputs, excessive emission of pollution, and insufficient production of desired outputs.

Chambers et al. (1996) proposed the Luenberger productivity index, which can consider both the reduction of inputs and the increase of outputs without choosing a measurement perspective, thus being more general than the Malmquist productivity index and ML productivity index. This paper will specifically calculate the GTFP, and to emphasize the environmental variables added, the Luenberger index is marked as GTFP. The expression for GTFP between period t and t + 1, and its sources, is as follows:

20

$\begin{array}{c}GTF{P}_t^{t+1}=\frac{1}{2}\left\{\left[S_C^t\left(x^t,y^t,b^t;g\right)-S_C^t\left(x^{t+1},y^{t+1},b^{t+1};g\right)\right]\right)\hfill \\ \left(+\left[S_C^{t+1}\left(x^t,y^t,b^t;g\right)-S_C^{t+1}\left(x^{t+1},y^{t+1},b^{t+1};g\right)\right]\right\}\hfill \end{array}$

21

$Effe=S_C^t\left(x^t,y^t,b^t;g\right)-S_C^{t+1}\left(x^{t+1},y^{t+1},b^{t+1};g\right)$

22

$\begin{array}{c}Tech=\frac{1}{2}\left\{\left[S_C^{t+1}\left(x^{t+1},y^{t+1},b^{t+1};g\right)-S_C^t\left(x^{t+1},y^{t+1},b^{t+1};g\right)\right]\right)\hfill \\ \left(+\left[S_C^{t+1}\left(x^t,y^t,b^t;g\right)-S_C^t\left(x^t,y^t,b^t;g\right)\right]\right\}\hfill \end{array}$Effe and Tech represent the changes in technical efficiency and technological progress from period t to t + 1, respectively.

Core explanatory variable

The core explanatory variable in this paper is the interaction term treat_i×post_t, which reflects the changes before and after the opening of CR Express service. If the coefficient is significantly negative, it indicates that the opening of the CR Express service has a suppressing effect on the dependent variable GTFP_it. If the coefficient is significantly positive, it suggests that the opening of the CR Express service has a promoting effect on GTFP_it.

Control variables

We incorporate the following control variables into the regression Eq. (17), based on relevant researches:

Per capita GDP (lnpgdp): This indicator reflects the level of economic development in a region. Energy usage is closely related to the level of economic development, as regions with higher economic development tend to have lower consumption of polluting energy sources and higher energy efficiency. Therefore, higher per capita GDP is expected to be positively associated with GTFP (Wu and Dong 2016); (2) Fiscal concentration (gov): This variable measures the proportion of government fiscal expenditure to GDP, indicating the degree of fiscal concentration. A higher fiscal budget allows the government to intervene in the economy and allocate more resources to environmental pollution control. Thus, higher fiscal concentration is expected to have a positive impact on GTFP (Wu and Dong 2016); (3) Population density (lnpop): Population density is closely related to the environment. Due to economies of scale and technological spillover effects, higher population density can significantly improve energy and environmental efficiency. Therefore, higher population density is expected to be positively associated with GTFP (Liu and Guo 2022); (4) Environmental regulation (env): This variable includes environmental protection policies and regulations. Environmental regulations are expected to reduce pollution emissions in cities (Wu et al. 2020). The frequency of vocabulary related to “environmental protection” in local government work reports, as a proportion of the total number of words in the reports, is used as a proxy variable for environmental regulation, following the method of Chen and Chen (2018); (5) Urbanization rate (urb): The increase in urbanization can lead to higher energy consumption in cities through investment promotion, which may have a negative impact on urban energy efficiency (Zhang et al. 2021). The urbanization rate is measured as the ratio of urban permanent population to total permanent population; (6) Foreign Direct Investment (lnfdi): Foreign direct investment is able to bring in technology and advanced management experience, which has a significant positive impact on GTFP (Wang et al. 2010).

Data sources

Due to the fact that the first CR Express started operating in 2011, and the years after 2020 were greatly impacted by the COVID-19 pandemic, this research uses data from 2008 to 2019 for cities at the prefecture level and above in China as the research sample. The data on cities where the CR Express is operating, the timing of operation, and the number of routes opened were obtained through sources such as the “Belt and Road” website, the freight train website, official websites of local railway bureaus, and news reports. A total of 58 cities with operational CR Express services were found within the sample period.1GTFP and TFP are calculated using the SBM model, with the former including undesirable outputs and the latter excluding them. City GDP, fiscal expenditure, population, sulfur dioxide emissions, city-wide electricity consumption, and other data are sourced from the “China City Statistical Yearbook” and development bulletins of various provinces and cities. Fixed asset investment data is sourced from the “China Fixed Asset Investment Statistical Yearbook.” After 2017, fixed asset investment is only reported as growth rate, so the fixed asset investment amounts for 2017 and beyond are calculated based on the growth rate information published in the statistical bulletins of various provinces and cities. The initial sample data is processed as follows: (1) Exclusion of cities that were revoked or underwent major administrative changes during the sample period (such as Chaohu, Laiwu, Bijie, Tongren, and Haidong, etc.), as well as cities with significant data gaps (such as Jiayuguan). Finally, a total of 280 cities at the prefecture level and above are selected as the sample; (2) For a small amount of missing data in some prefecture-level cities, the calculation method of moving averages is used to fill in the gaps. For example, for the sulfur dioxide emissions and city-wide electricity consumption data in some provinces and cities in 2013. This research obtained a total of 3360 valid observations from 2008 to 2019. The data sample also shows that approximately 6% of the observations were in the state of having operational CR Express services.

Empirical results and analysis

Baseline regression analysis

In order to examine the impact of the CR Express on the GTFP of node cities, this research uses ordinary least squares to estimate Eq. (17) (fixed effects model). Table 1 presents the baseline regression results of the impact of the CR Express opening on the GTFP of cities. The interaction term (core explanatory variable) in columns (1) and (2) of the table is significant and positive, indicating a positive association between the opening of the CR Express and the GTFP of node cities. This verifies the hypothesis 1 of this study. In column (1), the coefficient of the core explanatory variable (interaction term) is 0.319, while in column (2) with the inclusion of control variables, the coefficient decreases to 0.018. This suggests that the opening of the CR Express leads to an average increase of 1.8 percentage points in the GTFP of node cities, which translates to a 0.01 increase in the GTFP based on its mean value. Additionally, the control variables in column (2) such as per capita GDP, population density, and foreign investment all significantly promote the GTFP of node cities, which is consistent with the expectations. Urbanization rate and government expenditure are negatively correlated with the GTFP of cities, with the former aligning with expectations. The negative relationship between government expenditure and GTFP may be attributed to the fact that government intervention, as indicated by government expenditure, does not necessarily promote the GTFP of cities. Distortions in resource allocation caused by enhanced local government intervention can inhibit the improvement of energy efficiency in industrial agglomeration (Shi and Shen 2013). Finally, the coefficient of environmental regulation intensity is negative but not significant, and the direction of the coefficient is inconsistent with expectations. This may be due to the lack of rationality in environmental regulation policies during the sample period, which does not align with the characteristics of local economic endowments. Governments should formulate environmental regulation policies based on different regional endowment characteristics, pay attention to the heterogeneity of different regions, and achieve the goal of economic development and environmental improvement by improving the scientific and rational nature of environmental regulation policies (Shen and Liu 2012).

Table 1. Baseline regression results.

	GTFP		TFP		Lnwater
	(1)	(2)	(3)	(4)	(5)	(6)
Treat×post	0.319*** (3.60)	0.018** (2.11)	0.049*** (4.41)	0.025** (2.40)	0.154*** (4.27)	0.105*** (2.90)
lnpgdp		0.040*** (4.73)		0.066*** (6.56)		−0.030 (−0.84)
gov		−0.142*** (−12.06)		−0.307*** (−22.07)		−0.171*** (−3.47)
lnpop		0.111*** (3.34)		0.122*** (3.11)		0.802*** (5.80)
urb		−0.002*** (−3.87)		−0.003*** (−4.54)		−0.005** (−2.06)
env		−0.294 (−0.30)		−0.970 (−0.83)		−11.185*** (−2.70)
lnfdi		0.006*** (2.63)		0.003 (1.17)		0.003 (0.28)
Intercept	0.282*** (48.83)	−0.688*** (−3.23)	0.549*** (76.42)	−0.644** (−2.56)	8.465*** (360.56)	4.364*** (4.92)
Year fixed effect	YES	YES	YES	YES	YES	YES
City fixed effect	YES	YES	YES	YES	YES	YES
Observations	3360	3360	3360	3360	3360	3360
R2	0.298	0.345	0.215	0.344	0.374	0.388

The values in parentheses are t-values, * indicates significance at the 10% level, ** indicates significance at the 5% level, and *** indicates significance at the 1% level. The same applies to the following table.

The last four columns of Table 1 examine the impact of the CR Express on TFP and GTFP calculated after incorporating pollution emissions (represented by industrial wastewater) into the analysis. In column (4), the CR Express has a significant positive relationship with the TFP of node cities, which is consistent with the findings of Zhang and Gong (2021). The estimated coefficient of 0.025 is higher than the 0.018 in column (2), indicating that the promotion effect of the CR Express on TFP is significantly stronger than its effect on GTFP. This implies that if the negative effects of pollution emissions in the production process are not considered, empirical tests will overestimate the impact of the freight train opening on social and ecological effects. The results in column (6) also show that the opening of the freight train significantly promotes the discharge of industrial wastewater (Lnwater) in cities, confirming the mentioned cases earlier that the opening of the CR Express does increase the emission of pollutants to some extent in node cities.

Robustness checks

Identification assumption test

The use of the Difference-in-Differences (DID) method requires the assumption of common trends. If the opening and operation of the CR Express service in a city is purely an exogenous shock, the trends in GTFP for the treatment group (cities with the train service) and the control group (cities without the train service) should be consistent before the train service and show significant changes after its opening. We adopt the event analysis approach proposed by Jacobson and Sullivan (1993) to assess the impact of the CR Express service. The equation is as follows:

23

$GTF{P}_{it}=\beta +\sum _{-2\le f\le 2}{\alpha }_{f}GE{R}_{it+f}+\lambda \sum X_{it}+{\mu }_i+{\phi }_y+{\epsilon }_{it}$Where GER_it+f represents a dummy variable for the year f after the opening of the train service in city i. When the treatment group cities are in the t + f year after the opening of the train service, this variable takes the value of 1; otherwise, it is 0. We examine the dynamic effects of the policy implementation for two years before and after the opening of the train service, with the year two years before the opening as the baseline. Since the opening dates of the train service differ across cities, t + f represents different years for different cities. The regression results are reported in Table 2, with a focus on the coefficients for α_-2 to α₂. We can see that the coefficients for the two years before the opening are not significant, indicating that the treatment and control groups satisfy the parallel trend assumption before the opening of the train service. Furthermore, all the coefficients for α₀ to α₂ are significantly positive, indicating a significant improvement in GTFP after the opening of the train service.

Table 2. Dynamic effects of the opening of the CR express service on city GTFP - parallel trend assumption test regression results.

	α−2	α−1	α0	α1	α2
GTFP	−0.008 (−0.52)	−0.005 (−0.34)	0.028* (1.80)	0.028* (1.84)	0.024* (1.85)
Intercept	−1.755*** (−3.39)	−1.764*** (−3.41)	−1.682*** (−3.76)	−1.765*** (−3.42)	−3.470*** (−3.07)
Control variables	YES	YES	YES	YES	YES
Year fixed effect	YES	YES	YES	YES	YES
City fixed effect	YES	YES	YES	YES	YES
Observations	516	516	516	516	516
R2	0.312	0.311	0.306	0.316	0.339

The values in parentheses are t-values, *, **, *** represent significance levels of 10%, 5%, and 1% respectively.

Heterogeneous treatment effects test

Next, following the study by de Chaisemartin and D’Haultfoeuille (2020), the heterogeneity of treatment effects across groups and over time is an important factor that can lead to bias in the two-way fixed effects model. Since, we diagnose the potential heterogeneity of treatment effects in the baseline regression using the multiple period and multiple entity doubling model (DIDM) and the corresponding estimators proposed by de Chaisemartin and D’Haultfoeuille (2020, 2022). We identify individuals whose treatment status changes before and after the policy implementation as the treatment group, and those whose treatment status remains unchanged as the control group. By comparing the actual outcomes of the treatment group with their counterfactual outcomes, we obtain the treatment effects. After weighting, we obtain an unbiased estimate of the policy switching effect. Figure 2 presents the event study graph for the six periods before and after the opening of the CR Express service, estimating the dynamic treatment effects for each period. Before the opening, the effect on GTFP (GTFP) is not significant. However, after the opening of the train service, the effect gradually emerges and reaches a higher level in the second year after the opening. These results are consistent with the sign, magnitude, and trend of the treatment effects in the baseline regression. This indicates that the results of the baseline regression are robust through the test of heterogeneity of treatment effects.

Fig. 2 [Images not available. See PDF.]

Event study by de Chaisemartin and D’Haultfoeuille (2022).

Inconsistent policy timing

Goodman-Bacon (2021) argues that there can be bias in the estimation of the DID when policy timing is inconsistent. This research refers to their method to examine whether there is bias in the estimated overlapping DID. Firstly, the control group is divided into three groups based on the actual situation in this research: the first group consists of cities that have never opened the CR Express during the sample period (Treatment vs. Never Treated), the second group consists of cities that opened the CR Express later as the control group (Earlier Group Treatment vs. Later Group Comparison), and the third group consists of cities that opened the CR Express earlier during the sample period as the control group (Later Group Treatment vs. Earlier Group Comparison). According to Goodman-Bacon (2021), the estimates from the first and second groups are unbiased, but the estimates from the third group are prone to bias. This is because the third group, which uses cities that opened the China-Europe Railway earlier as the control group, includes not only the time trend but also the treatment effect in the estimation of the outcome variable. The treatment effect varies over time, which can lead to biased estimates. Additionally, the estimate of the overlapping DID is a weighted average of the estimates from these three groups (Goodman-Bacon 2021). Therefore, if the third group has a larger weight, it will have a greater impact on the final estimate. To assess whether there is bias in the overall estimate, the research conducts a bacon decomposition, as shown in Fig. 3. The diamond represents the first group, the hollow circle represents the second group, and the cross represents the third group (the group that introduces bias). The focus of this research is on the third group. If its weight is small, it will not have a significant impact on the overall estimate. In Fig. 3, the third group is shifted to the left, indicating that it has a smaller weight, while the first group is shifted to the right and has the largest weight. Table 3 also reports the specific weights, with the first group accounting for 92.4% and the third group accounting for only 2.3%, which is the smallest among the three groups. This suggests that it does not have a significant negative impact on the overall estimate of this study. In conclusion, the overlapping DID estimation results in this research are robust.

Fig. 3 [Images not available. See PDF.]

Bacon decomposition.

Table 3. Specific results of bacon decomposition.

Estimate	Weight	Estimate
First group	0.924	0.033
Second group	0.053	0.023
Third group	0.023	0.003
Weighted DID estimate	0.0319

The values in parentheses are t-values, *, **, *** represent significance levels of 10%, 5%, and 1% respectively.

Replacing variables

In order to further verify the credibility of the regression results in this research, the GTFP is replaced with carbon productivity, which is the ratio of city’s GDP to carbon dioxide (CO₂) emissions. This indicator, proposed by Beinhocker et al. (2008), reflects the economic benefits generated by unit carbon emissions. A higher value of this indicator indicates higher output, less pollution, and higher economic efficiency. To smooth the data, the logarithm of this ratio is taken. It should be noted that this research refers to the method of calculating carbon emissions in cities as proposed by Wu and Guo (2016). Table 4 reports the results after replacing the variables. Even after replacing the variable, the estimated coefficient of the opening of the CR Express on carbon productivity remains significantly positive, which confirms the reliability of the conclusions in this research.

Table 4. Results of regression with replaced variables.

	Ln(GDP/CO2)
	(1)	(2)
Treat×post	0.065** (2.22)	0.069** (2.35)
Cons	1.316* (1.83)	5.425*** (282.62)
Control variables	YES	NO
City fixed effect	YES	YES
Year fixed effect	YES	YES
Observations	3360	3360
R2	0.292	0.269

The values in parentheses are t-values, *, **, *** represent significance levels of 10%, 5%, and 1% respectively.

Time placebo

To avoid the possibility that the difference in GTFP between the treatment group and the control group is due to time changes, this research advances the opening time of the CR Express by 1, 2, and 3 years, respectively. Since Chongqing was the earliest to open the freight train in 2011 and the sample period is from 2008 to 2019, the time interval for the pre and post placement is set to three years in order to include all the treatment group cities. Table 5 presents the results. The estimated coefficients in columns (1), (2), and (3) are not statistically significant, and the coefficients and t-values of the interaction terms decrease as the advance time increases. This indicates that there is no systematic difference in the time trend between the treatment group cities and the control group cities. Thus, it can be concluded that the opening of the CR Express has a positive impact on the GTFP of cities.

Table 5. Results of time placebo regression.

	One year in advance	Two year in advance	Tree year in advance
	(1)	(2)	(3)
	GTFP	GTFP	GTFP
Treat×post	0.014 (0.96)	0.010 (0.72)	0.003 (0.17)
Cons	−0.691* (−1.71)	−0.696* (−1.71)	−0.708* (−1.73)
Control variables	YES	YES	YES
City fixed effect	YES	YES	YES
Year fixed effect	YES	YES	YES
Observations	3360	3360	3360
R2	0.345	0.344	0.345

The values in parentheses are t-values, *, **, *** represent significance levels of 10%, 5%, and 1% respectively.

City placebo test

Furthermore, to address the concern that the baseline regression results may be influenced by unobservable omitted variables, this research follows the approach of Cai et al. (2016) and conducts a city placebo test by randomly replacing the treatment group cities. The specific steps are as follows: 1. Randomly select 58 cities from the sample as the placebo treatment group, and assign the remaining 222 cities as the placebo control group. 2. Perform regression analysis using Eq. (17) to obtain the corresponding coefficients and p-values for the interaction terms. 3. Repeat steps 1 and 2 for 500 iterations to obtain 500 sets of interaction term coefficients and p-values. 4. Plot the kernel density distribution of the 500 interaction term coefficients and p-values. As shown in Fig. 4, the randomly selected interaction term coefficients are centered around zero and follow a normal distribution, with most regression results being statistically insignificant. The estimated interaction term coefficient in the baseline regression is 0.02 (indicated by the vertical dashed line in the graph), and it does not overlap with the distribution of the placebo regression interaction term coefficients. Therefore, this research rules out the possibility that the baseline estimation results are driven by unobservable factors.

Fig. 4 [Images not available. See PDF.]

Kernel density distribution of placebo test interaction term coefficients and p-value.

Heterogeneity tests

Western development strategy

As most node cities of the CR Express are located in the western region, it is natural to consider the implementation of the Western Development Strategy. The main approach of this strategy is tax incentives, which may lead to the concentration of industries in the western region, thus improving the GTFP of western node cities and ultimately driving the overall GTFP of node cities. Specifically, the following reasons can be considered: First, from 2011 to 2020, enterprises located in the western region and whose main business income accounted for more than 70% of the total income, as specified in the “Catalog of Encouraged Industries in the Western Region,” were eligible for a 15% reduced corporate income tax rate. The CR Express was opened in 2011, coinciding with the second round of tax reduction policies under the Western Development Strategy. Secondly, tax incentives may attract domestic and foreign investment, leading to industrial agglomeration and promoting the GTFP of cities. Therefore, the increase in GTFP in node cities may not be attributed to the opening of the CR Express, but rather to the industry clustering caused by the tax reduction policies implemented from 2011 to 2020. It is necessary to test this possibility.

Considering that the Western Development Strategy had two rounds, with the first round from 2000 to 2010 and the second round from 2011 to 2020, and based on the “Notice on Tax Policy Issues regarding the Western Development Strategy” in 2001 and the “Notice on Tax Policy Issues regarding the Implementation of the Western Development Strategy” in 2011, the tax incentives of the second round continued the policies of the first round. That is, both rounds provided a 15% preferential tax rate for corporate income tax for domestic and foreign-funded enterprises located in the western region that fell within the encouraged industry projects and whose main business income accounted for more than 70% of the total income. Therefore, this research attempts to compare the average growth rates of fixed asset investment and FDI between the western node cities and the eastern and central node cities from 2000 to 2010, in order to observe whether the tax incentives under the Western Development Strategy significantly attracted industries and FDI during this period. If the average annual growth rates of fixed asset investment and foreign direct investment in the western node cities do not show significant differences compared to the eastern and central node cities, it indicates that the tax incentives of the Western Development Strategy from 2000 to 2010 did not significantly improve the industrial agglomeration of the western node cities. Therefore, there is no reason to believe that it would have a significant impact in the second round (2011–2020). This indirectly eliminates the possibility that the increase in GTFP in node cities from 2011 to 2020 is caused by the tax incentives of the Western Development Strategy. As shown in Fig. 5, during the period from 2000 to 2010, the average annual growth rate of fixed asset investment in the western node cities was 24.60%, the lowest among the three regions, and the average annual growth rate of foreign direct investment was 48.18%, higher than that of the eastern node cities but lower than that of the central node cities. In summary, the tax incentives under the Western Development Strategy from 2000 to 2010 did not significantly differentiate the fixed asset investment and foreign direct investment in the western node cities from those in the eastern and central regions. Therefore, this research has reason to exclude the influence of tax incentives under the Western Development Strategy from 2010 to 2020 on the GTFP of western node cities.

Fig. 5 [Images not available. See PDF.]

Average growth rate of fixed asset investment and FDI by region.

Low-carbon pilot cities

Furthermore, considering that the National Development and Reform Commission announced the first, second, and third batches of low-carbon pilot cities in October 2010, December 2012, and January 2017 respectively, these pilot cities have started developing low-carbon industries, building low-carbon cities, and advocating low-carbon lifestyles. Therefore, there may be interference in the research of the impact of the opening of the CR Express on the GTFP of nodal cities due to the coincidence of the timing of low-carbon policies and the opening of the CR Express. Thus, this research attempts to avoid the interference of low-carbon pilot policies by excluding nodal cities that have overlapping timing. Following the approach of Zhang et al. (2021), if a city starts the pilot in the first half of the year, the timing is set as the same year, and if a city starts the pilot in the second half of the year, the timing is set as the next year. For example, if the pilot starts in July 2012, it is considered as 2013. 2The regression is then performed according to Eq. (17), where treat×post* refers to the interaction term between the new experimental group dummy variable formed after excluding the experimental group cities with overlapping timing of low-carbon pilot policies and the time dummy variable. The regression results are shown in the first column of Table 6, and the regression coefficient of treat×post* is significantly positive, indicating that even after excluding the impact of low-carbon policies, the results remain robust, further validating Hypothesis 1.

Table 6. Regression results after excluding interference.

	GTFP
	(1)	(2)
Treat×post*	0.026*** (2.81)
Treat×post**		0.022** (2.50)
Cons	−0.685*** (−3.19)	−0.680*** (−3.19)
Control variables	YES	YES
City fixed effect	YES	YES
Year fixed effect	YES	YES
Observations	3288	3324
R2	0.346	0.346

The values in parentheses are t-values, *, **, *** represent significance levels of 10%, 5%, and 1% respectively.

Domestic high-speed rail

Relevant studies have also pointed out that the opening of high-speed rail will also promote the GTFP of the opening cities (Wang et al. 2021). Therefore, it is necessary to exclude the impact of high-speed rail opening. The approach of this study to exclude the interference of high-speed rail opening is consistent with the method of excluding the interference of low-carbon pilot cities on the results. That is, after excluding the experimental group cities that coincide with the timing of high-speed rail opening, regression is performed again.3 In this case, treat×post** refers to the interaction term between the new experimental group dummy variable formed after excluding the experimental group cities with overlapping timing of high-speed rail opening and the time dummy variable. The regression results are shown in the second column of Table 6, and the coefficient of the interaction term (treat×post**) is also significantly positive, indicating that even after excluding the interference of high-speed rail opening, the conclusion of Hypothesis 1 remains robust.

East, Central, and West Route

The current routes of the CR Express are primarily divided into the East Route (exiting from Manzhouli), the Central Route (exiting from Erlianhaote), and the West Route (exiting from Khorgos). There are significant differences in the impact of different route cities on the GTFP of cities. Table 7 reports the results of the heterogeneity test for outbound routes. The CR Express on the East Route and the West Route, on average, increased the GTFP of node cities by 2.8% and 5.1% respectively. However, the CR Express on the Central Route had a inhibitory effect on the GTFP of node cities, although the result was not significant, which is similar to the findings of Xiao and Ye (2022). The possible reason for this difference lies in the regions served by different routes and the differences in the number of freight trains. The East and West Route mainly serve Europe, with a wider market coverage and more freight trains. As of November 2022, the cumulative number of freight trains passing through the East Route has exceeded 20,000, and the West Route has accumulated nearly 30,000, while the Central Route mainly serves trade with Mongolia and has fewer freight trains.

Table 7. Outbound route heterogeneity test results.

	East route		Central route		West route
	(1)	(2)	(3)	(4)	(5)	(6)
	GTFP		GTFP		GTFP
Treat×post	0.028* (1.67)	0.064*** (3.78)	−0.009 (−0.68)	−0.002 (−0.11)	0.051*** (3.21)	0.059*** (3.71)
Cons	0.556 (0.81)	0.278*** (22.25)	−1.166** (−2.43)	0.296*** (32.42)	−0.862*** (−2.94)	0.269*** (29.99)
Control variables	YES	NO	YES	NO	YES	NO
City fixed effect	YES	YES	YES	YES	YES	YES
Year fixed effect	YES	YES	YES	YES	YES	YES
Observations	588	588	1404	1404	1368	1368
R2	0.396	0.307	0.351	0.293	0.374	0.345

The values in parentheses are t-values, *, **, *** represent significance levels of 10%, 5%, and 1% respectively.

Reginal heterogeneity

Table 8 reports the results of regional heterogeneity test. This study found that the opening of the CR Express promotes GTFP by 2.5% and 3.7% for cities in the central and western regions respectively. There is also a positive relationship with eastern cities, but the result is not significant. This may be because the eastern cities are mostly located near the coast and have already formed mature industry clusters with a focus on maritime transportation since the reform and opening up. These industry clusters are not significantly affected by the opening of the CR Express. Therefore, the opening of the CR Express does not have a significant impact on the GTFP of eastern cities. However, as inland regions, the central and western regions cannot rely on maritime transportation to form relatively mature industry clusters. After the opening of the CR Express, they will form “CR Express industry clusters.” These clusters are mainly composed of high-value-added industries that are not sensitive to railway transportation prices and are oriented towards the European market. They have the characteristics of “high output, low pollution.” Therefore, it can be seen that the new trade channels have a positive impact on the GTFP of cities in the central and western regions. In addition, the GTFP of western cities is about 1.2% higher than that of central cities. This may be because compared to the central region, the western region is closer to the European market, and high-value-added enterprises are more willing to invest in the western region.

Table 8. Regional heterogeneity test results.

	East region		Central region		West region
	(1)	(2)	(3)	(4)	(5)	(6)
	GTFP		GTFP		GTFP
Treat×post	0.001 (0.06)	0.019 (1.29)	0.025* (1.94)	0.033** (2.40)	0.037* (1.96)	0.042** (2.24)
Cons	−1.476*** (−2.96)	0.301*** (29.82)	−0.371 (−1.37)	0.271*** (32.21)	−0.785 (−1.15)	0.270*** (23.40)
Control variables	YES	NO	YES	NO	YES	NO
City fixed effect	YES	YES	YES	YES	YES	YES
Year fixed effect	YES	YES	YES	YES	YES	YES
Observations	1200	1200	1212	1212	948	948
R2	0.320	0.267	0.468	0.401	0.277	0.273

The values in parentheses are t-values, *, **, *** represent significance levels of 10%, 5%, and 1% respectively.

Operation range heterogeneity

To measure the impact of the operation range of the CR Express on cities’ GTFP (GTFP), this study constructs an interaction term (treat_i×add_t) representing the coverage of the CR Express to foreign node cities and its impact on the GTFP of domestic node cities. Equation (19) introduces a time dummy variable add_t, which takes a value of 1 if the number of foreign node cities that a city can reach exceeds 3 in a certain year, and 0 if the number is less than or equal to 3 (3 being the median number of foreign node cities that all domestic node cities can reach from 2008 to 2019). Other variables are consistent with the previous discussion. Table 9 reports the results of the heterogeneity test for the operation range of the CR Express. Taking the first column as an example, the coefficient of the interaction term is significantly positive at the 5% level, indicating that the number of foreign node cities that domestic node cities can reach has a positive impact on their GTFP.

24

$GTF{P}_{it}={\alpha }_0+{\beta }_{1}trea{t}_i\times ad{d}_t+{\beta }_2\sum X_{it}+{\mu }_i+{\phi }_t+{\epsilon }_{it}$

Table 9. Operation range heterogeneity test results.

	(1)	(2)
	GTFP	GTFP
Treat×add	0.026* (1.90)	0.025* (1.83)
Cons	−0.132 (−1.58)	0.282*** (48.76)
Control variables	YES	NO
City fixed effect	YES	YES
Year fixed effect	YES	YES
Observations	3360	3360
R2	0.301	0.296

The values in parentheses are t-values, *, **, *** represent significance levels of 10%, 5%, and 1% respectively.

Firm level analysis

It is difficult to avoid the interference caused by unobservable factors due to the significant heterogeneity in city characteristics. Therefore, we conduct a firm level analysis using listed companies in various cities. Specifically, we examine whether the GTFP of enterprises located in cities along the CR Express routes increases more significantly compared to enterprises in other regions after the launch of the railway service. This approach is more reasonable in controlling for omitted variable endogeneity and avoids the issue of reverse causality. In total, we matched 15,734 observations for 2432 firms during the sample period.

Regarding the measurement of enterprise GTFP, we refer to the studies by Cui and Lin (2019) and Zhang et al. (2024), and incorporate firm environmental pollution into the evaluation. We use the non-radial SBM (Slacks-Based Measure) index to measure enterprise GTFP. The inputs and outputs for enterprise GTFP are measured as follows: (1) Input factors: Labor input is proxied by the number of employees; capital input is proxied by the net value of fixed assets; energy input is first obtained from the standard coal consumption of the enterprise’s region in the China Energy Statistical Yearbook. Subsequently, a regional standard coal consumption adjustment coefficient is calculated to obtain the weighted adjusted regional standard coal consumption. Finally, based on the proportion of firm revenue to the total industrial output value of the region, combined with the adjusted regional standard coal consumption, the standard coal consumption of industrial enterprises is derived. (2) Desired output: Firm revenue is used as the proxy for desired output. (3) Undesired output: The proportion of firm employees to the total urban employment in the city is used to calculate the emissions of industrial wastes, namely industrial sulfur dioxide, industrial wastewater, and industrial dust emissions, which serve as proxies for undesired outputs.

Table 10 presents the estimation results using the overlapping DID (Difference-in-Differences) method. In this process, we include several control variables reflecting firm production characteristics, such as firm size (number of employees), age, ownership structure (whether state-owned), profit margin, debt-to-asset ratio, and industry fixed effects. The estimation results confirm that the enterprise GTFP in cities along the railway has been positively promoted, with an average increase of 0.43 in productivity levels (Column (2) of Table 10). However, it should be noted that we could not determine whether the companies had strong trade links with European countries, i.e., whether they relied on the CR Express for foreign trade activities. Thus, this test could be served as a supplementary test.

Table 10. Test results for firm level analysis.

	GTFP
	(1)	(2)
Treat×post	1.0038*** (0.2034)	0.4341** (0.2118)
Control variables	NO	YES
Industry fixed effect	YES	YES
City fixed effect	YES	YES
Year fixed effect	YES	YES
Observations	15734	15734
R2	0.806	0.777

The values in parentheses are t-values, *, **, *** represent significance levels of 10%, 5%, and 1% respectively.

Endogeneity test

In the sample interval, there may be endogeneity issues with the opening of the CR Express, meaning that the opening of the train is not random but influenced by other omitted factors that also affect the selection of nodal cities and their GTFP. This can lead to bias in the results of the baseline regression. To address this endogeneity problem, We use cities along the ancient Silk Road as instrumental variables (silk×post).4 The rationale for using these cities as instrumental variables is as follows: firstly, cities along the ancient Silk Road are more likely to become nodal cities for the CR Express because the cultural characteristics, business environment, resource factors, and geographical conditions make trade with Central Asia and Europe easier, thus making them more likely to be selected as nodal cities for the train. Secondly, cities along the ancient Silk Road meet the exogeneity conditions. The first concern regarding the use of cities along the ancient Silk Road as instrumental variables is that these cities may have better economic development, thus having a positive impact on GTFP. In fact, cities along the ancient Silk Road have both economically developed and relatively poor cities during the sample period, so they do not have a significant impact on the economy. Table 11 reports the results of testing the exogeneity of the instrumental variable. The results in column (1) show that the effect of cities along the ancient Silk Road on economic development is not significant, thus rejecting this concern. Another concern is that the cities along the ancient Silk Road may be associated with environmental pollution, and industries that transfer from coastal to inland regions are more likely to be high-polluting industries. This study attempts to examine the relationship between cities along the ancient Silk Road and pollution emissions. First, following the approach of Wang and Sun (2017), this study uses the entropy weight method to calculate the Pollution Index, which is a dimensionless treatment of industrial sulfur dioxide, wastewater, and smoke emissions. The weighted coefficients for each year are calculated using the entropy method and then summed to obtain the pollution index for each city. The regression results in column (2) show that there is no significant relationship between cities along the ancient Silk Road and pollution emissions. In addition, the regression results in column (3) also show that cities along the ancient Silk Road do not have a significant effect on the GTFP of cities. These test results indicate that the instrumental variable satisfies the exogeneity condition, and therefore, this study combines two-stage least squares and instrumental variables for testing.

Table 11. Testing the Exogeneity of Instrumental Variable.

	(1)	(2)	(3)
	lngdp	Pollution	GTFP
Silk	−0.042 (−0.57)	0.030 (1.27)	−0.086 (−1.29)
Cons	8.084*** (24.61)	−0.101 (−0.97)	−0.575** (−2.03)
Control variables	YES	YES	YES
City fixed effect	YES	YES	YES
Year fixed effect	YES	YES	YES
Observations	3360	3360	3360
R2	0.979	0.736	0.648

The values in parentheses are t-values, *, **, *** represent significance levels of 10%, 5%, and 1% respectively.

Table 12 reports the estimation results of the 2SLS. From column (1), it can be seen that cities along the ancient Silk Road have a significant positive relationship with the opening of the CR Express, with an F-value of 37.39 (greater than 10), indicating that the instrumental variable is not weak. Furthermore, from the results of the second stage regression, we observe that the coefficient of the interaction term is significantly positive. It is worth noting that we find the regression coefficient of the interaction term (0.033) is larger than the baseline coefficient (0.018) from the baseline regression, indicating that endogeneity does indeed lead to an underestimation of the regression results. However, overall, we can conclude that the opening of the CR Express has a reliable positive impact on the GTFP of cities.

Table 12. Endogeneity test regression results.

	Instrumental variable method
	(1)	(2)
	First stage	Second stage
	treat×post	GTFP
Silk×post	0.869*** (54.15)
Treat×post		0.033* (1.66)
Cons	−1.714*** (−3.33)	−0.530** (−2.00)
Control variables	YES	YES
City fixed effect	YES	YES
Year fixed effect	YES	YES
Observations	3360	3360
R2	0.522	0.648
F-stat	37.39	—

The values in parentheses are t-values, *, **, *** represent significance levels of 10%, 5%, and 1% respectively.

Mechanism test

The theoretical analysis in the previous sections suggests that the CR Express improves the GTFP of node cities by promoting green innovation levels and industrial agglomeration. To verify the hypotheses proposed earlier, this study introduces a mediation model to identify the mechanisms through which the opening of the CR Express promotes the GTFP of node cities.

25

$Me{d}_{it}={\delta }_0+{\delta }_{1}trea{t}_i\times pos{t}_t+{\delta }_2\sum X_{it}+{\mu }_i+{\phi }_t+{\epsilon }_{it}$

26

$\begin{array}{c}GTF{P}_{it}={\omega }_0+{\omega }_{1}trea{t}_i\times pos{t}_t+{\omega }_{2}Me{d}_{it}+{\omega }_3\sum X_{it}\hfill \\ +{\mu }_i+{\phi }_t+{\epsilon }_{it}\hfill \end{array}$

Among them, Med is the mediating variable, and other variables are set consistent with the baseline regression. According to the mediation model, if the coefficient of the interaction term is significantly positive in Eq. (17), and the coefficient of the interaction term δ₁ is also significantly positive in Eq. (20), then the mediating variable is included in Eq. (21). If the coefficients of the interaction term ω1 and the Med variable ω₂ are both significantly positive, it indicates the presence of partial mediation effect. That is, the opening and operation of the CR Express improves the GTFP of cities through the mediating variables (green innovation and industrial agglomeration). If ω₁ is not significant but ω₂ is significant, it indicates the presence of complete mediation effect. If ω₂ is not significant, it indicates the absence of mediation effect.

In this research, the green innovation level (Inn) of cities is measured by the ratio of the number of green invention patent applications to the total number of annual patent applications in the city. This measure is adopted to effectively eliminate the interference of other unobservable factors that promote the green innovation level of cities outside the CR Express. Additionally, the ratio of patent applications is chosen instead of the ratio of granted patents due to the lag in patent granting process. Patent applications are considered more reliable and timely compared to granted patents (Jin et al. 2022). Industrial agglomeration is represented by the annual number of newly registered enterprises in cities, which is logarithmically transformed. A higher number of newly registered enterprises indicates a higher degree of industrial agglomeration in the city. If the opening of the CR Express attracts industrial agglomeration, the number of newly registered enterprises in the city will increase. At the same time, the “scale effect” and “technological spillover effect” will lead to a much larger increase in the output of these agglomerated enterprises than the increase in pollution emissions, thereby improving the GTFP of node cities.

The baseline regression results in the previous sections indicate that the opening of the CR Express has a significant promoting effect on the GTFP of node cities. In addition, as shown in the mechanism test results in Table 13, in column (1), the opening of the freight train significantly increases the city’s green innovation level by about 8.9 percentage points. In column (2), both the Inn variable and the treat×post coefficient are significantly positive, and the regression coefficient of the interaction term (0.014) shows a significant decrease compared to the baseline regression coefficient of 0.018 in Table 1. This indicates that the green innovation level is indeed one of the important mechanisms through which the CR Express promotes the GTFP of node cities. Through calculations, this research finds that the mediating effect of green innovation accounts for approximately 5.0% of the total impact on the GTFP of node cities, which supports hypothesis 2. Similarly, in column (3), the opening of the freight train significantly increases the level of industrial agglomeration by 6.9%. In column (6), the treat×post coefficient is significantly positive at the 5% level and its value decreases. Additionally, the coefficient of Lnindus is also significantly positive, indicating that industrial agglomeration is also an important mechanism, and the mediating effect of industrial agglomeration accounts for approximately 7.7% of the total impact on the GTFP of node cities, confirming hypothesis 3.

Table 13. Mechanism test.

	Green innovation		Industrial agglomeration.
	(1)	(2)	(3)	(4)
	Inn	GTFP	Lnindus	GTFP
Treat×post	0.893* (1.88)	0.014** (1.69)	0.069*** (3.80)	0.017* (1.95)
Inn		0.001* (1.66)
Lnindus				0.020** (2.37)
Cons	12.72 (1.10)	−0.716*** (−4.47)	4.070*** (9.09)	−0.771*** (−3.58)
Control variables	YES	YES	YES	YES
City fixed effect	YES	YES	YES	YES
Year fixed effect	YES	YES	YES	YES
Observations	3360	3360	3360	3360
R2	0.293	0.345	0.839	0.347

The values in parentheses are t-values, *, **, *** represent significance levels of 10%, 5%, and 1% respectively.

Conclusion and policy recommendations

This research focuses on the CR Express and systematically analyzes the impact and mechanisms of the opening of the CR Express on the GTFP of cities from 2008 to 2019. Heterogeneity tests were also conducted, and the following conclusions were drawn:

Firstly, the opening of the CR Express did not lower the green TFP of node cities. On the contrary, it significantly improved the GTFP of node cities. After conducting a series of robustness tests, the estimated results remained robust, with an average increase of 3.3 percentage points in the GTFP of node cities. This indicates that the “green” brought by the green trains is indeed effective. Secondly, in terms of rail routes, the eastbound and westbound routes increased the GTFP of node cities by 2.8% and 5.1% respectively, while the impact of the central route was not significant. In terms of city regions, the CR Express promoted the GTFP of node cities in the central and western regions by 2.5% and 3.7% respectively, while there was a positive but insignificant relationship with the eastern node cities. In terms of coverage, the number of foreign node cities that domestic node cities can reach has a positive impact on their GTFP. Thirdly, the mechanisms through which the CR Express improve the GTFP of cities are the improvement of green innovation level and the agglomeration of industries in node cities, with mediation effects accounting for 5.0% and 7.7% respectively.

Based on the conclusions of this research, the following policy recommendations are proposed: Firstly, it is recommended to encourage, support, and guide more cities (especially inland cities) to participate directly or indirectly in the “CR Express economic belt”. By relying on the CR Express, cities can leverage their abundant factor resources, establish efficient logistics networks, and create a favorable business environment, thereby promoting the high-quality development of regional economies and narrowing the regional economic development gap in China.

Secondly, the eastbound and westbound routes play a crucial role in improving the GTFP. Therefore, these two routes should be the focus of land transportation infrastructure construction. In addition, the CR Express have a significant impact on the GTFP of cities in the central and western regions, so the construction of node cities should focus on the inland central and western regions, further highlighting the role of the CR Express in promoting the high-quality development of cities.

Finally, to promote green innovation and industrial agglomeration, node cities should be encouraged and supported in green innovation. For example, leading the establishment of a China-Europe green innovation exchange platform, building more green industry cooperation demonstration bases and green technology exchange and transfer bases relying on the CR Express transportation nodes. Meanwhile, efforts should be made to create a favorable business environment, improve relevant infrastructure, and attract higher-end industrial agglomeration, such as establishing corresponding industrial parks and implementing tax incentives, simplifying commercial processes.

The limitation of this research lies in the source of data. Due to the lack of overlap between the industrial enterprises data and the opening time of CR Express, unlike some research focused on enterprise GTFP (Wu et al. 2022; Gao et al. 2024), the representativeness of using listed companies as research subjects may be insufficient. Therefore, this paper has conducted corresponding discussions from the perspective of prefecture-level cities. The subsequent studies will continue to refine the sample analysis and further examine the impact of the opening of CR Express on high-quality development.

Acknowledgements

The study is funded by Jiangxi Provincial Social Science Fund (No: SZ222027).

Author contributions

Author Contributions: (Hai Yin, Lingyu Zhang, Changlin Wan, Ting Xiao, Liang Ding) Hai Yin, Changlin Wan, Ting Xiao conceived and designed the research question.Lingyu Zhang, Ting Xiao, Liang Ding constructed and analyzed the models.Hai Yin, Changlin Wan, Ting Xiao, Liang Ding wrote the paper.All authors reviewed and edited the manuscript. All authors read and approved the manuscript.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Competing interests

The authors declare no competing interests.

Ethical approval

The study does not involve human participants or their data.

Informed consent

The study does not involve human participants or their data.

1

These 58 cities include: Baotou, Baoding, Chengdu, Dalian, Datong, Dongguan, Fuzhou, Ganzhou, Guangzhou, Guiyang, Harbin, Handan, Hefei, Huaihua, Ji’an, Jinan, Kunming, Lanzhou, Lianyungang, Linyi, Nanchang, Nanjing, Panjin, Qinzhou, Qingdao, Rizhao, Xiamen, Shanghai, Shangrao, Shenzhen, Shenyang, Shijiazhuang, Suzhou, Taiyuan, Tangshan, Tianjin, Weihai, Ulanqab, Urumqi, Wuhan, Wuwei, Xi’an, Xining, Xuzhou, Yantai, Jinhua, Yinchuan, Yingtan, Yingkou, Changchun, Changsha, Zhengzhou, Chongqing, Zhuzhou, Jingdezhen, Nanning, Bayan Nur, Linfen.

2

The excluded experimental group cities include: Chongqing, Suzhou, Fuzhou, Xi’an, Yantai, Yinchuan.

3

The excluded experimental group cities include: Datong, Rizhao, Jinhua.

4

Main cities along the land-based Silk Road include: Pingliang, Guyuan, Baiyin, Wuwei, Baoji, Tianshui, Dingxi, Xining, Zhangye, Lanzhou, Jiuquan, Yinchuan, Urumqi, Xianyang, Sanmenxia, Qingyang, Luoyang, Karamay. Main cities along the maritime Silk Road include: Guangzhou, Ningbo, Quanzhou.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

Aitken, BJ; Harrison, AE. Do domestic firms benefit from foreign direct investment? Evidence from Venezuela. Am Econ Rev; 1999; 89, pp. 605-618. [DOI: https://dx.doi.org/10.1257/aer.89.3.605]

Aschauer, D. Is public expenditure productive?. J Monetary Econ; 1989; 23, 89 pp. 177-200. [DOI: https://dx.doi.org/10.1016/0304-3932(89)90047-0]

Barro, RJ. Government spending in a simple model of endogenous growth. J Political Econ; 1990; 98, 5 pp. 103-125. [DOI: https://dx.doi.org/10.1086/261726]

Beinhocker E, Oppenheim J, Irons B, Lahti M, Farrell D, Nyquist S, Remes J, Nauclér T, Enkvist P-A (2008) The carbon productivity challenge: Curbing climate change and sustaining economic growth. McKinsey Global Institute

Brawn E, Wield D (1994) Regulation as a means for the social control of technology. Technol Anal Strat Manag 3:497–505

Bustos, P. Trade liberalization exports and technology upgrading: Evidence on the impact of MERCOSUR on Argentinian firms. Am Econ Rev; 2011; 101, 1 pp. 304-340. [DOI: https://dx.doi.org/10.1257/aer.101.1.304]

Cai, H; Zhong, C; Han, J. Upgrading of transportation infrastructure and location of polluting enterprises. China Ind Econ; 2021; 403, 10 pp. 136-155.

Cai, X; Lu, Y; Wu, M; Yu, L. Does environmental regulation drive away inbound foreign direct investment? Evidence from a quasi-natural experiment in China. J Dev Econ; 2016; 123, pp. 73-85. [DOI: https://dx.doi.org/10.1016/j.jdeveco.2016.08.003]

Chambers, RG; Fāure, R; Grosskopf, S. Productivity growth in APEC countries. Pac Econ Rev; 1996; 1, 3 pp. 181-190. [DOI: https://dx.doi.org/10.1111/j.1468-0106.1996.tb00184.x]

Chang, D; Miao, Q; Yuan, J; Tao, Z; Dong, Z; Bai, Y. Carbon emissions of countries along the “Belt and Road”: Analysis of the joint influence of foreign direct investment and development factors. Environ Sci Res; 2022; 35, 07 pp. 1556-1563.

Chen, J. Transportation infrastructure construction environmental pollution and regional economic growth. East China Econ Manag; 2020; 34, 09 pp. 72-79.

Chen, S; Chen, D. Haze pollution government governance and high - quality economic development. Econ Res; 2018; 53, 02 pp. 20-34.

Cui, X; Lin, M. How does FDI affect corporate green total factor productivity? An empirical analysis based on the Malmquist-Luenberger index and PSM-DID. Econ Manag; 2019; 41, 03 pp. 38-55.

de Chaisemartin, C; D’Haultfoeuille, X. Two - way fixed effects estimators with heterogeneous treatment effects. Am Econ Rev; 2020; 110, 9 pp. 2964-2996. [DOI: https://dx.doi.org/10.1257/aer.20181169]

de Chaisemartin C, D’Hautfoeuille (2022) Difference - in - differences estimators of intertemporal treatment effects. NBER Working Paper, No 29873

Debrezion, G; Pels, E; Rietveld, P. The impact of rail transport on real estate prices: An empirical analysis of the Dutch housing market. Urban Stud; 2011; 48, 5 pp. 997-1015. [DOI: https://dx.doi.org/10.1177/0042098010371395]

Dunning JH (1977) Trade, Location of Economic Activity and the MNE: A Search for an Eclectic Approach. In: Ohlin B, Hesselborn PO, Wijkman PM (eds) The International Allocation of Economic Activity. Palgrave Macmillan, London

Fang, H; Xie, H. The impact of China-Europe block trains on urban total factor productivity. Econ Manag Rev; 2022; 38, 03 pp. 38-52.

Frankel, JA; Rose, AK. Is trade good or bad for the environment? Sorting out the causality. Rev Econ Stat; 2005; 87, 1 pp. 85-91. [DOI: https://dx.doi.org/10.1162/0034653053327577]

Fukuyama, H; Weber, WL. A directional slacks-based measure of technical inefficiency. Socio Econ Plan Sci; 2009; 43, 4 pp. 274-287. [DOI: https://dx.doi.org/10.1016/j.seps.2008.12.001]

Gao, D; Zhou, X; Mo, X; Liu, X. Unlocking sustainable growth: exploring the catalytic role of green finance in firms’ green total factor productivity. Environ Sci Pollut Res; 2024; 31, pp. 14762-14774. [DOI: https://dx.doi.org/10.1007/s11356-024-32106-6]

Goodman-Bacon, A. Difference-in-differences with variation in treatment timing. J Econ; 2021; 225, 2 pp. 254-277.432864207414291[DOI: https://dx.doi.org/10.1016/j.jeconom.2021.03.014]

Grossman, GM; Krueger, AB. Economic growth and the environment. Q J Econ; 1995; 110, 2 pp. 353-377.0831.90040[DOI: https://dx.doi.org/10.2307/2118443]

Gu, N; Zhang, R. Can the opening of China - Europe block trains promote the improvement of enterprise total factor productivity? Quasi-Natural Experiment Based on Data of Chinese Manufacturing Listed Companies. Guangdong Soc Sci; 2022; 03, pp. 5-14+286.

Guo, J; Bai, J. How does high-speed rail construction promote the innovation and development of enterprises? Empirical test based on the face-to-face theory. Econ Theory Econ Manag; 2019; 41, 05 pp. 60-74.2019ECM..189..60G

Huang, K; Sun, G. ow does the opening of high-speed rail affect enterprise total factor productivity? A study based on Chinese listed manufacturing companies. J China Univ Geosci; 2019; 19, 01 pp. 144-157.

Jacobson, L; Sullivan, L. Earnings losses of displaced worker. Am Econ Rev; 1993; 83, 4 pp. 685-709.

Ji, Y; Yang, Q. Can high-speed rail opening promote corporate innovation: A study based on quasi-natural experiment. World Econ; 2020; 43, 02 pp. 147-166.

Jin, H; Yu, L; Xu, Y. Green industry policies and green technological innovation in manufacturing industry. China Popul Resour Environ; 2022; 32, 06 pp. 136-146.

Levinson, M; Taylor, S. Unmasking the pollution heaven effect. Int Econ Rev; 2008; 49, 1 pp. 223-254. [DOI: https://dx.doi.org/10.1111/j.1468-2354.2008.00478.x]

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 Econ Plan Sci; 2022; 80, [DOI: https://dx.doi.org/10.1016/j.seps.2022.101269] 101269.

Li, J; Min, Y; Wang, X. The impact of the opening of China-Europe freight train on urban innovation: On the innovative effects of China-Europe freight train under policy dilemma. World Econ Res; 2020; 321, 11 pp. 57-74+136.

Lin, P; Meng, N. Spatial and temporal differentiation and dynamic convergence of green total factor productivity growth. Quant Tech Econ Res; 2021; 38, 08 pp. 104-124.

Liu, B; Li, Q; Li, C. Does cross-border railway transportation accelerate China’s opening to the west?—Empirical evidence based on the city-product level. Manag World; 2022; 38, 08 pp. 101-118.2022ddwc.book...L

Liu, C; Wu, Q; Liu, Q. Transportation infrastructure, market accessibility, and enterprise productivity: Perspective of competition and resource allocation. Econ Res; 2020; 55, 07 pp. 140-158.

Liu, H; Guo, W. Carbon emissions trading pilot policies and energy-environment efficiency—Empirical evidence from 287 prefecture-level cities in China. J Xi’ Jiaotong Univ; 2022; 42, 05 pp. 72-86.

Liu, Q; Lu, Y. Firm investment and exporting evidence from China’s value-added tax reform. J Int Econ; 2015; 97, pp. 392-403. [DOI: https://dx.doi.org/10.1016/j.jinteco.2015.07.003]

Luo, N; Tian, M; Yang, J; Li, J; Yl, Wang. The impact of high-speed rail network on urban ecological efficiency: A spatial econometric study based on 277 prefecture-level cities in China. China Population. Resour Environ; 2019; 29, 11 pp. 1-10.2019RSEnv.220..1L

Perkins, R. Do recipient country characteristics affect international spillovers of co-efficiency via trade and foreign direct investment?. Clim Change; 2012; 112, 2 pp. 469-491.2012ClCh.112.469P[DOI: https://dx.doi.org/10.1007/s10584-011-0204-8]

Shen, N; Liu, F. Can high-intensity environmental regulation really promote technological innovation?—Re-examination based on the “Porter hypothesis”. Chin Soft Sci; 2012; 256, 04 pp. 49-59.

Shi, B; Shen, K. Government intervention, economic agglomeration and energy efficiency. Manag World; 2013; 10, 6 18.

Song, M; Zhang, G; Zeng, W; Liu, J; Fang, K. Railway transportation and environmental efficiency in China. Transp Res Part D Transp Environ; 2016; 48, pp. 488-498. [DOI: https://dx.doi.org/10.1016/j.trd.2015.07.003]

Sun, X; Zhang, G. High-speed rail opening, environmental pollution, and urban economic development. Soft Sci; 2021; 35, 06 pp. 103-108.

Tan, R; Pan, L; Xu, M; He, X. Transportation infrastructure, economic agglomeration and non-linearities of green total factor productivity growth in China: Evidence from partially linear functional coefficient model. Transp Policy; 2022; 129, pp. 1-13. [DOI: https://dx.doi.org/10.1016/j.tranpol.2022.09.027]

Tang, J; Zhou, X. Analysis of the comprehensive effects of China’s foreign trade on the environment China. Popul Resour Environ; 2017; 27, 04 pp. 87-94.

Wang, B; Wu, Y; Yan, P. Regional environmental efficiency and environmental total factor productivity growth in China. Econ Res; 2010; 45, 05 pp. 95-109.

Wang, C; Lan, Z; Zhang, C; Hou, X. High-speed rail construction labor capital migration and regional innovation China. Ind Econ; 2020; 393, 12 pp. 102-120.

Wang, H; Zhang, Y. Upgrading of trade structure environmental regulations and regional green technological innovation in China. China Soft Sci; 2020; 350, 02 pp. 174-181.

Wang, K; Pang, S; Zhang, F. Can the opening of high-speed rail improve urban green total factor productivity?. Ind Econ Res; 2021; 112, 03 pp. 112-127.

Wang, Y; Sun, X. The mechanism of government subsidies driving industrial transformation and upgrading China. Ind Econ; 2017; 10, pp. 99-117.

Wu, C; Dong, X. Research on total factor energy efficiency of the Yangtze River economic belt under environmental constraints. Soft Sci; 2016; 303, 03 pp. 3-83.

Wu, J; Xia, Q; Li, Z. Green innovation and enterprise green total factor productivity at a micro level: A perspective of technical distance. J Clean Prod; 2022; 344, [DOI: https://dx.doi.org/10.1016/j.jclepro.2022.131070] 131070.

Wu, J; Guo, Z. Analysis of convergence of China’s carbon emissions based on continuity dynamic distribution method. Stat Res; 2016; 33, 01 pp. 54-60.

Wu, L; Jia, X; Wu, C; Peng, J. The impact of heterogeneous environmental regulation on green total factor productivity in China. China Popul Resour Environ; 2020; 30, 10 pp. 82-92.

Xiao, T; Ye, H. Can the China-Europe freight train bring real money—Empirical test on the attractiveness of the China-Europe freight train to foreign investment. Int Trade Issues; 2022; 08, pp. 69-84.

Xu, Q. High-speed rail opening and enterprise total factor productivity: A study based on the data of Chinese industrial enterprises database. Mod Econ Res; 2020; 01, pp. 71-78.

Yang, R. Can industrial agglomeration improve China’s environmental pollution. China Popul Resour Environ; 2015; 25, 02 pp. 23-29.

Yang, X; Lin, S; Li, Y; He, M. Can high-speed rail reduce environmental pollution? Evidence from China. J Clean Prod; 2019; 239, [DOI: https://dx.doi.org/10.1016/j.jclepro.2019.118135] 118135.

Zhang, B; Zhou, J; Yan, Z. Can low-carbon city pilot policies promote total factor energy efficiency?—Evidence from quasi-natural experiments of three batches of pilot policy implementation. Econ Rev; 2021; 231, 05 pp. 32-49.

Zhang, J; Gong, E. Research on the impact of China-Europe block trains on urban total factor productivity. World Econ Res; 2021; 11, pp. 106-119+137.

Zhang, Z; Wang, L; Song, Y. Can corporate social responsibility disclosure improve green total factor productivity?. Audit Econ Res; 2024; 39, 04 pp. 103-115.

Zhang, X; Li, Y; Zhao, X. The impact of China - Europe freight train on trade growth in inland regions. Res Financial Econ Issues; 2019; 45, 11 pp. 97-111.

Zhao, X; Wang, L. Heterogeneous traffic network density labor mobility and total factor productivity. Chin J Circ Econ; 2020; 34, 05 pp. 95-107.