The Road to Low Carbon: Can the Opening of High-Speed Railway Reduce the Level of Urban Carbon Emissions?

Abstract

Adhering to green and low-carbon development is the only way to achieve sustainable economic development in China. High-speed railway (HSR) has the advantages of low energy consumption and low pollution, which is representative of the high-quality development of China’s transportation industry. Based on the panel data of 285 cities in China from 2004 to 2017, this paper explores the impact of HSR development on CO₂ emissions using a difference-in-difference model. The results show that the opening of HSR can effectively reduce the level of urban CO₂ emissions. After the endogeneity test and the robustness test, this core conclusion is still valid. The mechanism test confirms that the opening of HSR can reduce the level of urban carbon emissions by promoting the upgrading of industrial structures and improving the level of green technology innovation. Based on empirical results, this paper proposes that we should make rational use of the positive externalities of the HSR network to promote green development, intensive development, and innovative development.

1. Introduction

With the continuous improvement of urbanization and industrialization, the intensity of development and utilization of various types of energy is also increasing. The use of energy has laid a foundation for the life of residents and the production of enterprises, and has effectively promoted rapid economic growth. However, it has also caused environmental problems such as greenhouse gas emissions and air pollution. In particular, China’s energy consumption structure is mainly based on fossil energy. The excessive consumption of polluting fossil energy generates a large amount of carbon dioxide, making China the world’s largest carbon emitter. Based on the requirements of green development, China is striving to change the mode of production, constantly promote the adjustment of industrial structures, and make its mode of economic development more environmentally friendly. However, due to the obvious path dependence of air pollutants, it is difficult to significantly reduce carbon dioxide emissions in a short time. As a responsible country, China will take more effective measures to reach peak CO₂ emissions before 2030 and strive to be carbon neutral before 2060 [1]. The report of the 19th National Congress of the Communist Party of China also pointed out that the construction of ecological civilization is a millennium plan for the sustainable development of the Chinese nation. It is necessary to establish and practice the concept that green water and mountains are golden and silver mountains, and adhere to the basic national policy of resource conservation and environmental protection [2]. Although China is facing enormous pressure on carbon emission reduction, adhering to green and low-carbon development has become the only way to achieve high-quality development.

The transportation industry is a major contributor to carbon emissions, accounting for more than 10% of China’s total carbon emissions. The transportation industry not only has a direct impact on carbon emissions, but also indirectly affects the regional carbon emission level through the agglomeration effect of transport infrastructure. Therefore, the internal structural adjustment and high-quality development of the transportation industry have a significant impact on the economic development model and pollutant emissions. Relevant studies have shown that the technological upgrading of transportation tools and the green transformation of transportation industry can effectively reduce pollutant emissions and promote sustainable development. For example, Chen and Whalley [3] found that the operation of Taipei Metro reduced the usage of motor vehicles, thus effectively reducing CO emissions. Liang and Xi [4] comprehensively used various policy assessment methods to confirm that cities with rail transit have better air quality. According to the research of Zhang and Feng [5], the opening of high-speed railway (HSR) can reduce haze pollution through scale, structural, and technical effect, and the opening of HSR can also reduce haze pollution by improving the status of urban transport network [6]. In view of this, we infer that efficient and intensive transportation can promote carbon emission reduction by affecting the production and living mode of economic subjects. Thus, a natural question arises: as a green, environmentally friendly, and efficient railway transit infrastructure, can the construction and development of HSR reduce urban carbon emissions?

HSR is a typical representative of high-quality development of China’s transportation industry. China’s HSR construction has entered a stage of rapid development since 2008, when the Beijing–Shanghai and Beijing–Guangzhou HSR and other special passenger lines were started. In 2018, a state-planned “four vertical and four horizontal” operation network was formed. Compared with traditional railway transport and road transport, HSR, as a new mode of transportation, can not only optimize the internal structure of the transport industry and improve the efficiency of transport services, but also promote the sustainable development of the economy and society through the spatial–temporal compression effect and the factor flow effect [7]. In theory, the above two aspects are conducive to indirectly reduce carbon emissions. Identifying the net effect of HSR network on regional carbon emissions and analyzing its internal impact mechanism will not only help to understand the internal relationship between HSR development and carbon emissions, but also be of great significance to exploring the environmental benefits brought by the construction of HSR. At present, the green environmental effect of HSR has gradually emerged, but the impact mechanism of transportation infrastructure on regional carbon emissions is relatively complex, and there are relatively few empirical studies on CO₂ emissions caused by the opening of HSR. Among them, Sun and Ge [8] used the difference-in-difference (DID) model to prove that the opening of HSR can reduce urban industrial carbon emissions by affecting technological upgrading and production costs of enterprises. However, we think that the measurement method of carbon emissions may not be authoritative and accurate enough. Lin et al. [9], based on the annual passenger and freight flow monitoring information of China’s expressways and ordinary national highways from 2009 to 2016, as well as the information about HSR stations and mileage, used the DID model to confirm the CO₂ emission reduction effect of HSR in terms of traffic substitution for expressways, but ignored the impact of the indirect economic effect behind the opening of HSR on carbon emissions. In addition, Jia et al. [10] focused on analyzing the impact of HSR service intensity on carbon emission reduction from the perspective of spatial spillover, providing a future research direction for the author.

In view of this, based on the panel data of 285 cities in China from 2004 to 2017, this paper mainly uses the time-varying DID model to empirically analyze the effect of HSR development on urban carbon emissions. The possible contribution of this study mainly includes the following aspects. First, most of the studies focus on the growth effect and structural effect of HSR, while few studies focus on the environmental effect of HSR. Under the background of “double carbon”, this paper also enriches the research on HSR and carbon emissions. Second, this paper uses the carbon emissions data calculated based on night light data. These data are more reliable than the existing algorithm, which is more conducive to identifying the role of HSR development on CO₂ emission reduction. Third, from the perspective of system theory, this paper uses the mediation effect model to explore the indirect CO₂ emission reduction effect of the opening of HSR. The above has certain guiding significance for local governments to achieve regional low-carbon development through high-quality transportation infrastructure.

2. Literature Review and Research Hypothesis

The whole cycle of HSR from construction, opening, operation to maintenance involves different energy consumption, which is closely related to regional carbon emission. During the construction stage, the construction of HSR project needs to consume a lot of energy and raw materials, and the use of cement and steel is the main source of carbon emissions in this stage. According to the China High-speed Railway Development Report issued by the World Bank, CO₂ emissions generated by HSR construction are about 25,000 to 30,000 tons per kilometer. During the operation stage, HSR is not “green” because it is driven by electricity. At present, coal-fired power plays a dominant role, and this power production mode emits a large amount of greenhouse gases. However, compared with other transportation modes, railway transportation has the advantages of large carrying capacity and long transport mileage, and its unit carbon emission is still significantly lower than that of other modes of transportation. According to the relevant statistics of the Ministry of Transport in the current carbon emission structure of China’s transportation field, roads are the main body, accounting for 87%; shipping and aviation are about 6%; and railways have the lowest proportion, 0.68%. Moreover, in the process of HSR construction, the mode of leading the road by bridge is widely used, and the construction of the green belt is strengthened along the way, which also makes the HSR transportation network have an additional carbon reduction effect.

From the perspective of the internal structure of the transportation industry, HSR, as a new public transportation tool, plays an important role in replacing the traditional mode of transportation. The study of Lin et al. [9] confirms that after the cities were connected by HSR, the traffic flow of passengers and goods on expressways decreased significantly. Moreover, passengers switch from the traditional railway to HSR, which also releases the freight transport capacity of the traditional railway, thus promoting the transfer of road freight transport to the more environmentally friendly traditional railway. Under the same transportation volume, the energy consumption of HSR is far lower than that of highway and aviation, and the carbon dioxide emission is less than one tenth of that of aircraft. It is reported that from 2008 to 2016, the operation of HSR has helped China reduce greenhouse gas emissions equivalent to 11.18 million tons of CO₂ each year, accounting for about 1.33% of the total greenhouse gas emissions of China’s transportation industry [9]. Dalkic et al. [11] also confirmed that HSR consumes less fossil fuels in passenger and freight transportation, thus contributing to carbon emission reduction. Moreover, the railway department has also actively promoted the electronic ticketing service to reduce paper consumption and make railway development more low-carbon and environmentally friendly. Therefore, the substitution effect caused by the opening of HSR can effectively reduce CO₂ emissions. In addition, the relevant literatures also point out that the traditional transportation mode has a high level of energy consumption, and the carbon emission level of the transport industry should be reduced by using advanced transportation equipment, providing high-quality transportation services, and adjusting the transportation structure [12]. In the future, with the technological transformation of the transportation industry, positive environmental externalities caused by this substitution effect will continue to appear. Given this, we propose the following hypothesis:

Hypothesis 1.

The development of HSR can effectively reduce the level of urban CO₂ emissions.

Regarding the influencing factors of carbon emissions, some studies are analyzed based on the IPAT equation (I = P × A × T, where I, P, A, and T, respectively, represent Impact, Population, Affluence and Technology). The STIRPAT (Stochastic Impacts by Region on Population, Affluence, and Technology) model optimizes the IPAT equation and allows economic, environmental, social, and other factors to be incorporated into the driving factors, expanding the scope and core of influencing factors [13,14]. Some scholars also pointed out that the structure of transportation industry and transportation volume are important factors affecting regional carbon emissions through the research framework of ASIF (A, S, I, F represent Activity, Structure, Intensity, and Fuels, respectively) [15]. On this basis, Chinese scholars have conducted research on China’s carbon emissions. Relevant literature shows that regional energy structure, urbanization process, technological innovation, industrial structure, and industrial production scale are all key factors affecting CO₂ emissions [16,17].

Considering the system theory, the social system is a typical complex system composed of many subsystems such as population, resources, capital, technology, and the environment. As a subsystem, the impact of transportation infrastructure on regional carbon emissions is produced under the joint action of other influencing factors in the economic and social systems. As the key infrastructure for high-quality development of the transportation industry, HSR not only represents the technological upgrading of the railway transportation industry, but also deeply affects the industry, innovation, and other subsystems in the economic and social systems. For example, the opening of HSR can affect economic growth [18], capital flow [19], regional economic integration [20], industrial structure upgrading [21], and enterprise innovation [22]. The above studies confirm that the positive external spillover effect of HSR development is obvious. This means that the opening of HSR may affect the regional green and low-carbon development in many aspects, thus forming the occurrence mechanism of environmental pollution change [23].

First of all, from the perspective of industrial structure, the opening of HSR may reduce carbon emissions by promoting the adjustment of industrial structure. The development model of high input, high pollution, low output, and unreasonable industrial layout of the secondary industry will significantly increase carbon emissions. Compared with industry, the tertiary industry has the characteristics of knowledge intensive and low pollution intensity. Therefore, under the pressure of industrial structural contradictions and green development, the upgrading of industrial structure is an effective way to stabilize growth and promote emission reduction [6]. The opening of HSR has weakened the spatial flow barriers of factors and promoted the optimal allocation of production factors on a larger scale [24]. As the HSR mainly focuses on passenger transport, it will have a more obvious impact on the service industry with strong mobility of production factors, which will help strengthen the agglomeration development of the service industry in cities along the route [25]. The agglomeration of service industry is closely related to the adjustment of the industrial structure, which not only helps to increase the proportion of tertiary industry, but also has a crowding-out effect on industrial production factors [23]. The research of Liu and Li [26] has shown that the flow of resources brought by the opening of HSR can promote the development of the tertiary industry while leading to the decline of the proportion of the secondary industry. Deng et al. [21] further pointed out that the opening of HSR can realize the optimization and upgrading of industrial structures through improving market potential, promoting technological innovation, and optimizing resource allocation. The upgrading of industrial structure reduces the dependence of economic development on energy and resources and promotes the flow of factors to sectors with higher efficiency, thus helping to improve the efficiency of energy use and build a green production system. Yu [27] also pointed out that the optimization and upgrading of industrial structures has a significant emission reduction effect, which is an important part of green and low-carbon development. Given this, we propose the following hypothesis:

Hypothesis 2a.

The opening of HSR can reduce the level of urban carbon emissions by optimizing the industrial structure.

Secondly, this can also be studied from the perspective of technological innovation. On the one hand, the opening of HSR shortens commuting time, improves accessibility, effectively alleviates the obstacles to the flow of elements [28], reduces the unnecessary loss of knowledge and technology in the dissemination, and then helps to improve innovation ability [29]. On the other hand, due to the advantages of high transportation efficiency, safety, and comfort, HSR has become the first choice for time-sensitive technical talents, who are an important carrier of factor flow and technological innovation [30]. Moreover, the opening of HSR will not only make innovation elements flow efficiently among regions, but also cause a knowledge spillover effect, which will generate a learning effect, an imitation effect, and an incentive effect among innovation subjects. Such effects will effectively promote the generation and development of intellectual property rights and the innovation mode, and ultimately help to improve the innovation level of cities and enterprises [22,31]. Regarding the impact of technological innovation on energy conservation and emission reduction, Grossman and Krueger [32] earlier put forward the conclusion that technological innovation is helpful to reduce the negative environmental output brought about by economic development, thus improving the urban environmental quality. As an important part of technological innovation, green technology innovation was first proposed by Braun and Wield [33]. They believe that green technology is different from previous technologies, which can not only increase the total economic output, but also effectively promote sustainable development. As an important way to achieve low-carbon development [34], green technology innovation can not only promote clean production of enterprises and improve energy efficiency, but also promote the transformation and development of traditional industries to green low-carbon industries, thus reducing energy consumption from the production side and achieving a source control of carbon emissions. Yuan et al. [35] also confirmed that green technology innovation has a significant effect on carbon emission reduction. Therefore, we propose the following hypothesis:

Hypothesis 2b.

The opening of HSR can reduce urban carbon emissions by promoting green technology innovation.

3. Methodology and Data

3.1. Model

STIRPAT model is often used as a basic model for analyzing the driving factors of environmental change, which not only allows the coefficients to be estimated as parameters, but also allows the appropriate decomposition and supplementation of influencing factors [13]. On this basis, many literatures have made corresponding improvements to carry out various empirical studies according to their own research characteristics.

I = aP^β1A^β2T^β3e

(1)

In the above formula, a is the model coefficient, I refers to environmental pressure, P, A, and T represent demographic, economic, and technological factors, respectively, and e is error term.

This paper constructs the econometric model based on the STIRPAT model and expands and improves it accordingly. According to the theoretical analysis above, this paper includes HSR, industrialization development, and urbanization into the research framework of the driving factors of environmental change, and all continuous variables except dummy variables (whether to open HSR) are logarithmically treated. At the same time, considering the research topic and the applicability of the method, we used the DID model to explore the impact of the opening of HSR on CO₂ emissions.

DID model is one of the methods of policy evaluation, which can effectively eliminate the heterogeneity and increments that do not change with time before and after the implementation of the policy to identify the net effect of policy shock on individuals. At present, China is constructing a HSR transportation network of “eight vertical and eight horizontal”. This large-scale construction of high-quality transportation infrastructure can be studied as a quasi-natural experiment. However, considering the differences in the opening time of HSR in various cities, this paper draws on the ideas of Sun and Ge [8] and Shao et al. [25], adopting the time-varying DID model as shown below.

lnC_it = α₀ + α₁HSR_it + α₂lnX_it + μ_i + v_t + ε_it

(2)

In this equation, lnC_it is the dependent variable, which represents the level of CO₂ emissions of city i in year t. The core independent variable is HSR_it. If city i opens HSR in year i, the value of HSR_it is “1” in or after the year t, and “0” otherwise. The coefficient α₁ measures the net impact of the opening of HSR on urban carbon emissions. If α₁ is significantly negative, it indicates that the opening of HSR reduces carbon emissions. X_it is a set of control variables, which have been logarithmically treated in the empirical analysis. Variables μ_i and v_t represent the city fixed effect and year fixed effect, respectively. The variable ε_it is a random error term. On the one hand, the treatment effect (whether HSR is opened or not) occurs at the city level. The same city is correlated in different years, and the cities in different provinces are also correlated. On the other hand, biased standard error and wrong inference caused by too few cluster numbers should be avoided [36]. Therefore, this paper uses the robust standard error clustered to the city level to solve potential heteroscedasticity, sequence correlation, and other problems.

3.2. Variable Definition

3.2.1. Dependent Variable

The dependent variable lnC is measured by the logarithm of urban CO₂ emission. Different from previous studies, the CO₂ emission data used in this paper are from the public data of the latest study by Chen et al. [37]. The data are processed as follows. First, the particle swarm optimization-back propagation (PSO-BP) algorithm is used to match and unify the DMSP/OLS and NPP/VIIRS data with time spans of 1992–2013 and 2012–2017, respectively, and then obtain 26 years of carbon emission data in China, which provide a basis for the long-term study of carbon emission drivers. Second, the PSO-BP algorithm was adopted to build the relationship between night light data and CO₂ emission data from fossil energy consumption of Chinese provinces and cities. At last, the CO₂ emissions from fossil energy consumption in China were inversely calculated by taking the total light brightness of each county as the weight. Compared with the traditional CO₂ accounting methods, these data have the advantages of long research period, unified measurement standards, strong reliability, and so on. Based on the research scope, we add the CO₂ data at the county level to the city level and perform logarithmic processing.

3.2.2. Independent Variable

Most studies describe the opening of HSR by setting dummy variables 1 and 0, but some scholars point out that only using dummy variables may not be able to fully identify the differences in the development of HSR in various cities [25]. In view of this, this paper uses three variables: the opening of HSR, the number of HSR stations and the number of HSR lines to represent the construction of HSR in each city. Specifically, HSR_it is a dummy variable for whether city i opened HSR in year t; Station_it is the cumulative number of HSR stations operated by city i in year t; Route_it is the cumulative number of HSR lines operated by city i in year t. In the empirical analysis, we have logarithmized the last two. This paper holds that if HSR in city i is opened on or before June 30 of a year, it will be regarded as opening in that year; if it is opened on or after July 1, it will be deemed to be opened in the next year [6], and the statistics of HSR stations and lines also follow this principle. In addition, in the basic regression part, this paper uses the above three variables at the same time, while in the follow-up test, it focuses on the HSR_it.

3.2.3. Control Variables

This paper mainly selects the following control variables. Specifically, it includes: (1) GDP per capita (lnperGDP) and its square term “(lnperGDP)²”. Regional economic development level is closely related to environmental change, and carbon emissions will increase with economic growth. However, the theory of environmental economics indicates that the environmental effect of economic growth is not necessarily linear, so the square term of GDP per capita is also used in this paper. (2) Population density (lnPD). The population size directly affects the regional carbon emission level. In this paper, the population density index is introduced to control its impact on carbon emissions, which is expressed by the ratio of the registered population to the land area of each city. (3) According to the relationship between economic development and carbon emissions, we also selected the urban built-up area(lnArea) and industrial electricity consumption(lnEC) to control the impact of urbanization and industrialization development on carbon emissions. Moreover, the technical factor is an important factor emphasized in the STIRPAT model, so this paper also tried to use lnRD (logarithm of science and technology expenditure) to measure, However, this variable is not significant in regression. The author conducted a multicollinearity test on the variables, and the results (see Appendix A for details) further proved that the reason why lnRD is not significant is not multicollinearity, but lack of explanatory capacity. Under the suggestion of the reviewer, the author conducted regression analysis again after removing lnRD.

3.3. Data Sources and Descriptive Statistics

In addition to the CO₂ emission data published by Chen et al. [37], the other data used in this paper mainly form the following two aspects. The first aspect of the data is mainly from the China City Statistical Yearbook from 2004 to 2017, the China Research Data Service Platform, and the China Economic Database. Specifically, it includes data on economic and social development indicators such as GDP per capita, registered population, built-up area, and industrial electricity consumption. The HSR data in the second aspect mainly come from the public data provided by the State Railway Administration, the 12306 website and Baibu Encyclopedia. In view of this, we sorted out the HSR database at the city level. It should be noted that since the State Council implemented the Medium and Long-Term Railway Network Planning in 2004, the national HSR construction entered the substantive construction stage, so this paper began to study from 2004. At the same time, cities with severe data loss and major administrative division adjustments during the study period, such as Chaohu, Sansha, and Bijie, were excluded from the sample processing. Finally, this paper uses the unbalanced panel data of 285 cities across the country from 2004 to 2017 for empirical testing. By the end of 2017, 189 cities in the sample cities had opened HSR. Figure 1 shows the distribution map of cities with or without HSR in China. The descriptive statistics of the variables are shown in

Table 1. Descriptive statistics of variables.

Variable	Definition	Obs	Mean	SD
lnC	Logarithm of CO2 emissions (million tons)	3990	2.946	0.841
HSR	Whether to open HSR (1–0 dummy variable)	3990	0.247	0.431
lnStation	Logarithm of the number of HSR stations	3990	0.306	0.583
lnRoute	Logarithm of the number of HSR lines	3990	0.214	0.400
lnPerGDP	Logarithm of GDP per capita (yuan/person)	3987	10.193	0.774
lnEC	Logarithm of industrial electricity consumption (10,000 kWh)	3625	12.301	1.418
lnPD	Logarithm of population density (people/km2)	3922	5.729	0.912
lnArea	Logarithm of urban built-up area (km2)	3966	4.332	0.871

.

4. Results

4.1. Main Results

Table 2. Basic regression results.

Variable	Dependent Variable: lnC
	(1)	(2)	(3)	(4)	(5)	(6)
HSR	−0.0389 ***	−0.0205 *
	(0.0122)	(0.0112)
lnStation			−0.0314 ***	−0.0173 *
			(0.0093)	(0.0091)
lnRoute					−0.0465 ***	−0.0281 **
					(0.0150)	(0.0141)
lnPerGDP		0.4901 **		0.4749 **		0.4667 **
		(0.1993)		(0.2039)		(0.2041)
(lnPerGDP)2		−0.0151		−0.0143		−0.0139
		(0.0105)		(0.0107)		(0.0107)
lnEC		0.0227 ***		0.0225 ***		0.0227 ***
		(0.0081)		(0.0081)		(0.0080)
lnPD		0.2560 *		0.2571 *		0.2625 *
		(0.1542)		(0.1541)		(0.1547)
lnArea		0.0124 **		0.0122 **		0.0119 **
		(0.0060)		(0.0060)		(0.0059)
Constant	2.4150 ***	−2.5932 **	2.4150 ***	−2.5218 **	2.4150 ***	−2.5130 **
	(0.0078)	(1.0374)	(0.0078)	(1.0472)	(0.0078)	(1.0449)
City FE	YES	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES	YES
Observations	3990	3543	3990	3543	3990	3543
Adj-R2	0.8558	0.9104	0.8861	0.9105	0.8861	0.9106

Note: the values in brackets are robust standard errors clustered to the city level; *** p < 0.01, ** p < 0.05, * p < 0.10.

shows the basic regression results of the impact of HSR construction on urban CO₂ emissions. The independent variables in columns (1) and (2) are dummy variables of whether HSR is opened or not, while columns (3) and (4), as well as columns (5) and (6), respectively, estimate the results of the impact of the number of HSR stations and the number of HSR lines on carbon emissions. In addition, columns (1), (3) and (5) do not include any control variables but control for the city and year fixed effects, while columns (2), (4) and (6) report the results of adding all control variables and conducting fixed effects. From the above regression results, the development of HSR can effectively reduce the level of urban CO₂ emissions, that is, it has a significant carbon emission reduction effect. Thus, Hypothesis 1 was verified.

Moreover, the coefficient of lnperGDP is significantly positive, while the coefficient of square term “(lnperGDP)²” is not significant, indicating that there is no “inverted U-shaped” relationship between economic growth and environmental pollution mentioned by Kuznets Curve during the study period, which is similar to the conclusion of Zhou and Wang [38]. By comparing the GDP per capita, urbanization rate and industrial structure of European and the United States when they achieved peak CO₂ emissions, we can see that China has not yet reached the development level of such countries (see Appendix A for details). In addition, the speech made by the General Secretary of China also confirmed that China has not yet reached the peak CO₂ emissions [1]. To sum up, China is still at the stage where CO₂ emissions increase with economic growth, and has not yet reached the stage where economic growth is decoupled from CO₂ emissions to achieve carbon neutrality. The coefficient of lnEC is significantly positive at the 1% level, which once again confirms that industrial electricity consumption is an important source of CO₂ emissions. The coefficient of lnPD is significantly positive, indicating that population agglomeration can increase carbon emissions [39]. The coefficient of lnArea is also significantly positive, which indicates that with the improvement of the level of urbanization, the carbon emission problem is further aggravated.

4.2. Endogenous Problem

The construction and site selection of HSR are mainly based on regional central cities. These cities have large populations, many industrial enterprises and good economic foundations, but they also face serious air pollution and greenhouse gas emissions. This may make it difficult for the sample to meet the requirements of random distribution. In addition, if the carbon emission level of cities with HSR is higher than that of regions without HSR, then the fixed-effects model will underestimate the impact of the opening of HSR on CO₂ emissions. Thus, this paper further solves the potential endogenous problem of the model through the instrumental variable (IV) method. Drawing on relevant literatures and considering the actual situation of HSR construction in China, this paper constructs an IV for whether to open HSR based on the average slope of each city [40,41]. Specifically, from the perspective of correlation, the greater the average slope of a city, the more difficult it is to build transportation infrastructure such as HSR; from the perspective of exogeneity, slope, as a geographical variable, is an established objective fact, which is usually has no correlation with economic and social development indicators. In addition, since the slope does not change with time, this paper uses the product of the average slope of each city and the year dummy variable as an IV. One of the selection principles of IV is the high correlation between IV and endogenous explanatory variables. The IV method includes two stages. A common method to judge weak IV is to observe the F value in the first stage regression. Specifically, the first stage is the regression of the endogenous variable to the IV, mainly to judge whether the coefficient of the IV is 0.

Table 3. Results of IV regression.

Variable	First Stage	Second Stage
	(1)	(2)
HSR		−0.5367 **
		(0.2266)
HSRiv	−0.0710 ***
	(0.0247)
Control variables	YES	YES
City FE	YES	YES
Year FE	YES	YES
Observations	3543	3543
Adj-R2	0.5436	0.9661

*** p < 0.01, ** p < 0.05.

shows that in the first stage regression, there is a significant negative correlation between the opening of HSR and the IV (HSRiv) at the level of 1%, that is, the greater the average slope of a city, the more difficult it is to build HSR. In addition, the F statistic in the first stage regression is 41.04. According to the analysis of Stock and Yogo [42], the F value of the first stage regression in this paper is greater than the critical value (16.38) under the 10% error level, which indicates that there is no weak IV problem. The results of the second stage show that the coefficient of HSR is significantly negative at the 5% level, indicating that the conclusion of the basic regression is reliable.

4.3. Robustness Test

To ensure the reliability of the results, robustness tests are carried out in this section, including parallel trend test, eliminating key cities, eliminating outliers, changing the research period and controlling the impact of other policies. Details are shown in Figure 2 and

Table 4. Robustness test.

Variable	Dependent Variable: lnC
	(1)	(2)	(3)	(4)
HSR	−0.0228 *	−0.0202 **	−0.0186 *	−0.0219 *
	(0.0117)	(0.0101)	(0.0111)	(0.0112)
Control variables	YES	YES	YES	YES
City FE	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
Observations	3121	3187	3006	3486
Adj-R2	0.9154	0.9112	0.8908	0.9109

** p < 0.05, * p < 0.10.

.

4.3.1. Parallel Trend Test

The prerequisite of the use of the DID model is that the treatment and the control group must have the same change trend before the policy is implemented, that is, the parallel trend assumption needs to be satisfied. In order to test whether this hypothesis is valid, this paper draws lessons from Beck et al. [43] and uses the event study method to test. The corresponding model settings are as follows.

$$ln{C}_{it}=\alpha +{\displaystyle \sum }_{k\ge -10}^8{\beta }_k{D}_{it}^k+\gamma ln{X}_{it}+{\mu }_i+{\nu }_t+{\epsilon }_{it}$$

(3)

where $D_{it}^{k<0}(D_{it}^{k>0}$) indicates a value of 1 if the sample is “the city where the HSR was opened” and “the year before (after) the opening of the HSR”, and 0 otherwise; $D_{it}^{k=0}$ implies a value of 1 if the sample is “the city where the HSR was opened” and “the year after the HSR was opened”, otherwise the value is 0. The settings of other variables are the same as above. Based on the scope of the sample, we measure the trend of CO₂ emissions for 10 years before and 7 years after the opening of HSR. In particular, we use the year when HSR was opened as the base year, that is, the dummy variable when k = 0 is not included in the actual regression. Figure 2 shows that the regression coefficients change significantly before and after the opening of HSR. The estimated coefficients before the opening of HSR are not significant, while the estimated coefficients after the opening of HSR are significantly negative. This result not only proves that the study meets the hypothesis of parallel trend, but also proves that the opening of HSR can effectively reduce urban CO₂ emissions.

4.3.2. Excluding Key Cities

As the level of transportation infrastructure construction in municipalities, provincial capitals, and sub-provincial cities is already better, and their economic foundation, urban planning, and environmental protection policies are also obviously different from those of other cities. To ensure the comparability of the research objects, this paper conducts regression after excluding these cities from the total sample. Column (1) of Table 4 shows that the coefficient of HSR is significantly negative, which indicates the conclusion that “the opening of the HSR has significantly reduced the urban CO₂ emissions” is robust.

4.3.3. Eliminating Outliers

Since the CO₂ emissions used in this paper are derived from remote sensing inversion data, there is a high possibility that some observed values deviate from the mean value. To eliminate the interference of outliers, the regression was conducted again based on the 5–95% quantiles of carbon emission data. Column (2) of Table 4 shows that HSR is significantly negative at the 5% level, which again supports the core conclusion of this paper.

4.3.4. Changing the Research Period

Due to the data limitation and the background of the Medium and Long-term Railway Network Planning, the basic regression is mainly based on 2004–2017. To test the robustness of the above results again, we select different time periods for regression according to Zhang et al. [44] The empirical results can be seen in column (3) of Table 4 (only the results from 2005 to 2015 are listed here). Although the coefficient value of HSR has decreased, the conclusion is still robust.

4.3.5. Controlling the Low-Carbon City Policy

When studying the effect of HSR development on carbon emissions, it is inevitable to be interfered with by other environmental protection policies. Some studies have also shown that low-carbon city policy can significantly and continuously reduce urban CO₂ emissions [45,46]. To exclude the impact of this policy, this paper sets up the policy dummy variable of low-carbon cities based on Formula (2) according to the pilot list of low-carbon cities announced by the National Development and Reform Commission. Column (4) of Table 4 points out that the carbon reduction effect of HSR is still significant, which can once again confirm that the core conclusion is relatively robust.

4.4. Mediating Effect Analysis

The above theoretical analysis points out that the opening of HSR can achieve the low-carbon development of cities by promoting the upgrading of industrial structure and improving the level of green technology innovation. In order to identify whether these two effects exist, we refer to the method of Wen and Ye [47], and set up the following model based on Formula (2) to test the possible mediating effect of the opening of HSR.

$$M_{it}={\beta }_0+{\beta }_{1}HS{R}_{it}+{\beta }_{2}ln{X}_{it}+{\mu }_i+{\nu }_t+{\epsilon }_{it}$$

(4)

$$ln{C}_{it}={\gamma }_0+{\gamma }_{1}HS{R}_{it}+{\gamma }_2{M}_{it}+{\gamma }_{3}ln{X}_{it}+{\mu }_i+{\nu }_t+{\epsilon }_{it}$$

(5)

In Formula (4), M represents the mediating variable, which is represented by industrial structure (IND) and green technology innovation (lnGP), respectively. Among them, IND is expressed by the urban tertiary industry’s value added to the secondary industry’s value added, which can also indirectly measure the advanced level of the regional industry; lnGP is represented by the logarithm of the number of urban green invention patents authorized. The patent data are obtained from the China Intellectual Property Office and screened according to the International Patent Classification Standard for Green Patents. The basic regression has confirmed that the opening of HSR is beneficial to urban carbon emission reduction. On this basis, Formula (4) is regressed to test the effect of the opening of HSR on the mediating variable. Then, both HSR and the mediating variable are included in the model for regression, that is, the regression of Formula (5). On the premise that ${\beta }_1,{\gamma }_1$ and ${\gamma }_2$ are all significant, if ${\beta }_1\times {\gamma }_2$ and ${\gamma }_1$ have the same sign, it means that the opening of HSR can reduce the urban CO₂ emissions by affecting the industrial structure and green technology innovation. If the sign of ${\beta }_1\times {\gamma }_2$ is opposite to that of ${\gamma }_1$, it means that there is a masking effect.

Column (1) in

Table 5. Mediating effect test.

Variable	IND	lnC	lnGP	lnC
	(1)	(2)	(3)	(4)
HSR	0.0444 **	−0.0189 *	0.2034 ***	−0.0197 *
	(0.0202)	(0.0097)	(0.0545)	(0.0103)
Mediating variable		−0.0830 **		−0.0058 *
		(0.0400)		(0.0033)
Control variables	YES	YES	YES	YES
City FE	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
Observations	3472	3472	3530	3530
Adj-R2	0.2258	0.9103	0.6958	0.9108

*** p < 0.01, ** p < 0.05, * p < 0.10.

confirms that when the IND is taken as the mediating variable, HSR is significantly positive at the level of 5%, indicating that the opening of HSR can promote the upgrading of urban industrial structure. The results in column (2) show that the coefficient of IND on carbon emissions is significantly negative at the level of 5%, that is, the upgrading of industrial structure is helpful to reduce the urban CO₂ emissions. From the perspective of coefficient sign, the product of the coefficients of HSR in column (1) and IND in column (2) is negative, which is consistent with the sign of HSR in column (3). Moreover, the coefficient of HSR in column (3) is significantly negative and the absolute value is lower than that of column (2) of the basic regression. Combined with the test principle of the mediation effect, the industrial structure is a partial mediating variable, that is, the opening of HSR can play a role in reducing carbon emissions by optimizing urban industrial structure. Similarly, when lnGP is taken as the mediating variable, the regression results in column (3) show that HSR is significantly positive at the level of 1%, that is, the opening of HSR is helpful to improve the level of urban green technology innovation. By observing the sign of the estimated coefficient and comparing HSR estimation results in the basic regression and column (4), green technology innovation also plays a partial mediating role in the carbon reduction effect of the opening of HSR. To sum up, the opening of HSR can reduce urban carbon emissions by optimizing the industrial structure and improving the level of green technology innovation. Therefore, Hypothesis 2 is verified.

5. Conclusions and Policy Implications

5.1. Conclusions

The transportation industry is a major contributor to CO₂ emissions. As a green and efficient new railway transit infrastructure, HSR not only affects the economy, structure, and technology, but also gradually shows its indirect environmental dividends. Under the background of China’s persistent implementation of the new development concept, in-depth promotion of ecological civilization construction and realization of the “double carbon” goal, this paper regards the opening of HSR as a quasi-natural experiment, based on the panel data of 285 cities in China from 2004 to 2017, the time-varying DID model is used to explore whether the opening of HSR can reduce the level of urban CO₂ emissions and the mediation effect model is used to analyze its internal mechanism. The research results show that the construction and development of HSR reduces urban carbon emissions, and this core conclusion still holds after endogeneity treatment and multiple robustness tests. Additionally, the mechanism test points out that the opening of HSR can reduce CO₂ emissions by promoting the upgrading of the industrial structure and improving the level of green technology innovation.

5.2. Policy Implications

The research results of this paper have some policy implications for China’s HSR construction and low-carbon development. Firstly, build a more extensive HSR network to promote low-carbon and green development of the regional economy. High-quality transportation infrastructure is conductive to accelerating the flow of factors and promoting regional integration. In this process, the rational allocation of factors will promote the transformation and upgrading of regional industries, and then achieve high-quality economic development. In the future, the expansion of the HSR network will optimize the industrial structure through agglomeration effects and learning effects, reduce industrial energy consumption, improve enterprise production efficiency and energy utilization, and ultimately contribute to the realization of regional sustainable development. Secondly, make rational use of the positive externalities of the HSR network to drive the city to take the road of intensive and efficient development. On the one hand, the opening of HSR can adjust the internal transport structure of the city through the spatial–temporal compression effect, promote the development of public transportation, guide residents to green travel, and then indirectly reduce regional CO₂ emissions. On the other hand, the construction of HSR can optimize the spatial layout, guide cities to make rational use of resources, improve the energy utilization rate, and transform production and lifestyle through an intensive development model, thus reducing carbon emissions. Finally, give full play to the leading role of HSR in technological innovation, vigorously promote urban technological innovation, and improve the efficiency of energy utilization. HSR represents a technological change in the transportation industry, and technological innovation itself also reflects a high-quality development model that optimizes the transportation structure and improves energy efficiency. In the context of the transformation in energy consumption mode, the railway sector should adhere to the innovation-driven development strategy, promote low-carbon development through technological change, and further play the green environmental effect of HSR network by expanding the coverage of HSR.