Sage Journals: Discover world-class research

Abstract

Using the difference-in-differences method combined with the propensity score matching, this study identifies the causal relationship between high-speed rail (HSR) and energy productivity in China. Furthermore, we investigate the mechanism through which HSR affects energy productivity, as well as the heterogeneity of the impact across quantiles and distances. The results show that HSR connection contributes to the improvement of energy productivity. This finding is consolidated after a potential endogeneity problem is addressed using the instrumental variable method and a variety of potential confounders are controlled through a series of robustness checks. On average, the marginal impact of HSR on energy productivity is approximately 9%. Moreover, HSR connection cannot be completely substituted by traditional railway and aviation in improving energy productivity. The heterogeneity analysis suggests that the positive energy productivity effect of HSR gradually decreases with an increasing distance to the nearest HSR station. In addition, HSR network accessibility has a significant positive effect on energy productivity, while technological innovation mediates the relationship between HSR development and energy productivity. We propose that to achieve the long-term improvement of energy productivity, policymakers should comprehensively consider both transit-oriented development and ecology-oriented development modes.

Keywords

High-Speed rail Energy productivity Propensity score matching Difference-in-differences Instrumental variable method China

1. Introduction

The effects of high-speed rail (HSR) on the economy and productivity have received substantial attention from economists. However, no consensus has been reached about whether HSR is conducive to economic and productivity growth, especially in developing countries. Some studies argue that investments in HSR transportation infrastructure have led to a redistribution of economic factors among different regions in Europe (Albalate and Bel, 2012; Puga, 2002; Rodríguez-Pose and Fratesi, 2004; Sánchez-Mateos and Givoni, 2012). Other studies suggest that an expanding HSR network has the potential to improve cities’ economic growth and productivity (Ryder, 2012; Ke et al., 2017).

Rail transport is more energy-efficient and emits less greenhouse gases (GHGs) than other transportation modes. For example, HSR can support environmentally sustainable development because the electric multiple unit needs electricity from a wide range of sources, including renewable energy, and uses less energy than other vehicles. This claim has verified by previous studies. Wee et al. (2005) demonstrated that on average, road transportation emits three times more GHGs than rail transportation. Janic (2011) showed that the substitution of short-haul aviation with HSR services can help to reduce GHG emissions. Lin et al. (2021a) also found that new HSR routes in China led to an annual reduction of 1.33% of GHG emissions in China’s transport sector. HSR operation can improve the electrification level, optimize the energy consumption structure of railway transport, reduce the dependence on oil, and thus mitigate carbon emissions (Jia et al., 2021).

Existing studies examine the economic and productivity effects of HSR, but rarely investigate the impacts of HSR on energy productivity. To the best of our knowledge, the following questions on the effectiveness of HSR operation remain unanswered: Does the Chinese HSR exert a significant effect on energy productivity? What magnitude is such an effect? What are the channels through which HSR affects energy productivity? What role does HSR network accessibility play in improving energy productivity?

To address these questions, the present study focuses on the causal effect that HSR exerts on energy productivity, as measured by the ratio of economic output to energy consumption.¹ Investigating the effect of the Chinese HSR connection on energy productivity is crucial for the following two reasons: On the one hand, ascertaining and quantifying the effect of HSR on energy productivity is vital as China is currently transforming from a coal-fired economy to a greener economy; on the other hand, the empirical results can provide some insight and guidance for future decision making regarding public infrastructure investment.

Under such backgrounds, taking China’s HSR operation as a natural experiment, this paper investigates whether HSR operation improves energy productivity and identifies the potential transmission channel, i.e., the technological innovation effect. Considering that HSR connections are unlikely to be randomly selected by the Chinese central government, an instrumental variable (IV) approach is adopted.

Using the difference-in-differences (DID) method combined with propensity score matching, we find that after a city operates HSR, on average, energy productivity can increase by approximately 9%. To address a potential endogeneity problem, historical information about rail rays in 1990 and courier stations during the Ming dynasty are used to establish IVs. We find that the IV estimates are consistent with baseline results. The heterogeneity analysis suggests that the positive impact of HSR on energy productivity gradually enhance with increasing quantiles. The results also indicate that HSR effectively improves the energy productivity of cities closer to HSR stations; however, significant decay effects emerge with increasing distance to the nearest HSR station. In addition, HSR network accessibility has a significant positive effect on energy productivity, while technological innovation mediates the relationship between HSR development and energy productivity.

Although a lot of previous studies have investigated the role of HSR in both economic growth (Atack, 2010; Chen and Hall, 2012; Qin, 2017; Shao et al., 2017; Donaldson, 2018) and environmental performance (Yang et al., 2018; Cheng et al., 2020; Sun et al., 2020a; Jia et al., 2021), this is the first study to explore the effect of HSR on energy productivity at China’s city level. To the best of our knowledge, the investigation of Chen (2021) is the most closely related to our work, who examined the impact of HSR on Chinese cities’ energy consumption rather than energy productivity using the DID method. However, our analytical framework and focus differ significantly from Chen (2021) and other existing studies at least in four aspects. First, the causal relationship between HSR and energy productivity remains unclear in previous literature. To clarify their relationship, the claim that HSR connection contributes to the improvement of energy productivity in China is verified. Besides, we also fully consider the potential endogenous problem using the IV method. Second, by manually collecting energy intensity data at the city level, energy consumption is measured using a bottom-up approach rather than a top-down approach, which is an essential strategy to calculate energy productivity. In contrast, based on the Defense Meteorological Satellite Program’s Operational Linescan System (DMSP/OLS) night-time light data, Chen (2021) obtained the data of city-level energy intensity by using the linear regression model without intercept to decompose provincial-level energy consumption to city-level one. However, because of the stochastic disturbance of the linear regression model and the precision of the downscaling of night-time light data, the estimation method of energy consumption in this paper is more accurate than that in previous literature. Third, the decay effect that HSR exerts on energy productivity and the substitutability of HSR and other transportation modes are investigated, which has not been addressed by previous studies. The findings of this study confirm that the positive energy productivity effect of HSR gradually decreases with increasing distance to the nearest HSR station from 2.1 km to 815.7 km. Furthermore, the effect of HSR connection on energy productivity cannot be completely substituted by traditional railway and aviation. Fourth, the role of HSR network accessibility measured through the social network analysis method is further investigated, and it is examined whether HSR affects energy productivity through technological innovation.

Information about how and to what extent HSR affects energy productivity is beneficial for both policymakers and practitioners to integrate transit-oriented development and ecology-oriented development modes. Policymakers should also fully use the diffusion effect of HSR to decrease a spatial inequality between cities with and without HSR stations. To achieve green transformation, the development of HSR should not only consider its pollution reduction effects, but also its energy productivity effects. Therefore, this study can provide some empirical references for the improvement of energy productivity via HSR network in other developing countries that are operating HSR.

The paper is structured as follows: Section 2 provides a literature review and pertinent hypotheses; Section 3 presents the methodology and data; Section 4 discusses the empirical results; Section 5 conducts the robustness checks and heterogeneity analysis; Section 6 identifies the potential mechanisms; Section 7 summarizes and concludes the paper.

2. Literature Review and Hypotheses

Many studies have focused on the regional economic effects of HSR in different countries. Prominent examples are the United States of America (Atack, 2010), the United Kingdom (Chen and Hall, 2012; Vickerman, 2018), China (Chen et al., 2016; Jia et al., 2017; Chen and Haynes, 2017; Qin, 2017; Shao et al., 2017; Liu and Zhang, 2018; Jin et al., 2020; Li et al., 2020a; Lin et al., 2021b), Korea (Kim et al., 2018), Japan (Hiramatsu, 2018; Li and Xu, 2018), and India (Donaldson, 2018). Redding and Turner (2015) recently conducted a survey of the theoretical and empirical literature on the relationship between transport infrastructure and the spatial organization of economic activities. Although the economic effects of HSR have been widely investigated before, how HSR affects energy productivity has received little attention. Especially for the case of developing countries, the effect of HSR on energy productivity requires further investigation.

Increasing economic output with developing low-energy and low-pollution transportation modes is important to realize sustainable development goals. This is especially true for HSR, which has less energy consumption compared with cars and aircraft (de Rus Mendoza, 2012). Dalla Chiara et al. (2017) compared the specific energy consumption of air transport with that of HSR transport and suggested that with the increasing length of the route, the energy gap between these two forms of transportation diminished. Cheng et al. (2020) used a hybrid input-output life cycle method to evaluate the multi-footprints of Beijing-Tianjin intercity HSR in China. Sun et al. (2020a) found that implementing HSR significantly improved China’s environmental efficiency.² Jia et al. (2021) confirmed that HSR development significantly reduced CO₂ emissions in China. These studies provide some institutional backgrounds and theoretical supports for examining the effect of HSR on energy productivity.

Existing empirical studies find that wide difference exists in the measures of energy efficiency with the different stages of transportation construction (Feng and Wang, 2018; Lin and Chen, 2019). Besides, to obtain the robust estimates of the production frontier, a large sample of observations is required in productivity and efficiency measurement techniques (Tanaka and Managi, 2013). As an alternative measure of energy efficiency, energy productivity is defined as the ratio of economic output to energy consumption (Patterson, 1996; Dimitropoulos, 2007). Meanwhile, the determinants of energy productivity have been discussed broadly, e.g., research and development (R&D) activities (Fisher-Vanden et al., 2006), and technical changes (Wang and Wei, 2016; Yang et al., 2018; Li et al., 2021; Berhe et al., 2022). For instance, Atalla and Bean (2017) investigated the drivers of energy productivity in 39 countries during 1995–2009 and found that a higher income level and higher energy prices could lead to greater energy productivity improvements.

This paper focuses on the effect of China’s HSR connection on energy productivity. To address the potential mechanisms of how HSR affects energy productivity, Figure 1 illustrates both the direct and indirect channels through which these mechanisms function.

Figure 1

Potential channels through which HSR affects energy productivity

The top section of Figure 1 (a) presents the direct effect of HSR on energy productivity. In general, cooperation becomes closer and more frequent between cities connected by HSR, which saves time (leading to consumer surplus gains for passengers) and decreases transaction costs. Thus, the corresponding mechanism is that HSR stimulates economic growth and improves the efficiency of resource allocation. For example, Diao (2018) found that HSR increased economic growth by facilitating travel and interaction among cities. Li et al. (2020b) also documented that the opening of HSR exerted a significant positive effect on economic efficiency. Thereby, the following hypothesis is proposed:

Hypothesis 1: HSR operation exerts a positive effect on energy productivity.

The middle section of Figure 1 (b) presents the effect of HSR-induced accessibility on energy productivity. HSR connections expand the transportation network among cities and reduce the time costs of the flows of production factors. These connections also make it more convenient for developing cities to promote industrialization, urbanization, and information development through industrial transfer from developed cities. Since cities with HSR can attract more economic activities, which in turn affects energy productivity through industrial agglomeration and scale effects (Wu and Lin, 2021). This further accelerates the agglomeration of homogeneous industries in cities and promotes the professional division of labor. These effects also help cities along the HSR to optimize, reconstruct, and organize spatial structures. The industrial agglomeration can improve the specialized agglomeration of cities and thus effectively promote energy productivity. In addition, the advantages of HSR network accessibility can lead to massive investments in fixed assets (Diao, 2018). Thus, the following hypothesis is proposed:

Hypothesis 2: HSR network accessibility exerts a positive effect on energy productivity.

Finally, the bottom section of Figure 1 (c) illustrates the technological innovation effect of HSR on energy productivity. Birol and Keppler (2000) highlighted the relationship between the improvement of energy efficiency and technological advancement, which has been supported by some existing studies on technical progress and technological innovation (Cagno et al., 2015; Wei et al., 2016). HSR operation also accelerates the accumulation of technological innovation resources in cities with HSR connection to improve technological innovation (Gao and Zheng, 2020). Thus, we infer that HSR can promote energy productivity through a positive technological innovation effect. The underlying logic is that HSR operation expedites the flows of green production factors to reduce the space-time distance between knowledge and technology transmission. Moreover, HSR operation can also further improve the allocation efficiency of technological innovation resources. Regarding the accumulation of human capital, a well-functioning infrastructure is vital for cities to attract talents, improve the efficiency of high-quality human resource allocation, and even achieve the high-quality growth of the local economy.

In addition, HSR operation accelerates the flows of technological innovation factors to drive the formation of technological innovation networks between cities. First, population mobility pattern is essential for the exchange of knowledge and technology between cities in a technological innovation network (Dong et al., 2020; Li et al., 2016). HSR connection is conducive to strengthening the exchange and cooperation of technological innovation activities between cities connected via HSR, and to promoting the technological innovation quality of HSR-connected cities. Second, HSR operation helps to strengthen technological innovation cooperation among cities and enhances their inclination to implement technological innovation behaviors. These behaviors increase the number of patent applications (Lin et al., 2021), especially for energy-saving patents, and strengthen intellectual property protection. Third, HSR operation is beneficial to expanding both the scale and scope of enterprise production to yield a market size effect (Ma et al., 2021) by sharing R&D costs and increasing the return on the assets of R&D activities. Thus, the third hypothesis is proposed:

Hypothesis 3: Technological innovation mediates the relationship between HSR and energy productivity, i.e., HSR operation conduces to technological innovation to positively affect energy productivity.

3. Methodology and Data

3.1. Baseline Model

First, the impact of HSR on energy productivity is estimated using a DID strategy. In other words, two types of variation are combined: the time variation (i.e., before and after the start of HSR operation for an individual city) and the city variation (i.e., whether cities are connected to HSR stations or not). Furthermore, fixed effects are also introduced to control both time-invariant and time-variation observable city attributes that may be related to the implementation of HSR. The general form of the DID model is expressed as follows:

e p_{it} = a_{0} + α_{1} hs r_{it} + α_{2} x_{it} + δ_{i} + γ_{t} + ε_{it}

(1)

where ep_it represents energy productivity in city i in year t; hsr_it denotes the treatment indicator, which is equal to one if city i operates HSR in year t (and for all subsequent years), and otherwise zero; x_it is a vector of time-varying city characteristics, including gross domestic product (GDP) per capita, the share of foreign direct investment (FDI) in real GDP, the share of secondary industry value added in GDP, and environmental regulation; δ_i represents the city fixed effect, which captures unobservable heterogeneity; γ_t consists of the year fixed effect and controls for temporary shocks; ε_it is the error term; a₀, α₁, and α₂ are coefficients to be estimated.

Two possible biases should be addressed by the DID estimator. The first may emerge if energy productivity trends (and levels) vary in both the treatment and control groups prior to the operation of HSR. The second may arise if HSR stations are chosen based on geographic, political, or socio-economic factors. In this regard, the propensity score matching (PSM) technique proposed by Rosenbaum and Rubin (1983) is employed. Based on this approach, treated cities can be combined with controlled cities (that have similar characteristics) to enable the reduction of potential bias. We will use the Mahalanobis distance matching (MDM) (Cochran and Rubin, 1973) to conduct robustness checks.

3.2. Verifying the Parallel Trend

A critical assumption of the DID method is the existence of parallel trends, implying that pre-intervention trends in energy productivity are identical between treatment (HSR cities) and control (non-HSR cities) groups. The DID estimates may be biased if parallel trends are not satisfied. This assumption can be verified according to Beck et al. (2010) and Yang et al. (2018), i.e., both the leads and lags of the initial connection dummy variable of HSR are incorporated. The reference year is 2008, which is the year when the first HSR line started operating,³ and the sample includes only the data of the eight years after 2008. Thus, eight is adopted as the threshold for specifying the dummy variable, and the time window of this study is 16 years, including the eight years after the operation of HSR, and the eight years preceding its open. Thus, the following model is used to verify the parallel trend:

e p_{it} = b_{0} + \sum_{j = - 8}^{8} β_{j} {hsr}_{it}^{j} + β_{9} x_{it} + λ_{i} + σ_{t} + u_{it}

(2)

For HSR cities, the dummy variable hsr^j is equal to one in the j-th year before the introduction of HSR (each j is negative), while hsr^j is equal to one in the j-th year after the introduction of HSR (each j is positive), and otherwise hsr^j is equal to zero. Specifically, assuming that o_i is the initial year of HSR operation for city i, the dummy variable $.^{.} j_{it}$ indicates whether $t - o_{j} = j (j \in [- 8, 8])$ in year t. For example, if the HSR operation year for both Beijing and Tianjin is 2008, then in 2006, hsr^–2 is equal to one for Beijing and Tianjin, and otherwise zero; in 2010, hsr² is equal to one for Beijing and Tianjin, and otherwise zero. Other cities are similar. Variables λ_i and σ_t represent the city fixed effect and the year fixed effect, respectively; u_it is the error term; b₀ and $β_{j} (j = - 8, - 7, \dots, 7, 8)$ are coefficients to be estimated.

3.3. Data and Variables

The HSR data are manually collected from a number of sources, such as the China Railway Yearbook (2004–2017) and the official website of the China Railway Corporation (12306.cn). HSR data are double-checked with a relevant Wikipedia website.⁴ The HSR-opened years for the sample cities are presented in Appendix A.

The economic benefits of HSR operation have been empirically assessed by some existing studies (Zheng and Kahn, 2013; Baum-Snow et al., 2017; Ahlfeldt and Feddersen, 2018; Yao et al., 2019); however, an econometric approach for studying the relationship between HSR and energy productivity has not been developed to date. To estimate the energy productivity effect of HSR, in the baseline regression, we employ a single factor (i.e., GDP per unit energy consumption) rather than a total factor approach to measure energy productivity. This strategy is adopted for two reasons: First, GDP per unit energy consumption is consistent with China’s existing statistical caliber and energy-conservation and emission-reduction targets, and the corresponding results are more comparable and can provide specific policy implications; second, using a single factor approach to measure energy productivity can alleviate the potential endogeneity problem, to some extent (Sahu and Sharma, 2016). In a total factor productivity framework, for instance, labor force and capital stock may depend on unobserved productivity shocks. Therefore, if input choices and productivity are correlated, the obtained results may be biased.

Prefecture-level GDP and GDP index are collected from the China City Statistical Yearbook (2004–2017). GDP statistics are cross-checked against provincial statistical yearbooks to ensure the accuracy and consistency of economic output. GDP is deflated to the 2003 constant price. For robustness checks, the commonly-used DMSP/OLS night-time light data are employed to represent GDP (Chen and Nordhaus, 2011; Henderson et al., 2011; Michalopoulo and Papaioannou, 2018). In this study, the correlation coefficient of GDP and total night-time light data reaches 86.3%.

Since energy consumption matters in productivity growth (Murillo-Zamorano, 2005), its accurate measurement is also very important. Primary energy consumption at the city level is estimated using a bottom-up approach. Huang et al. (2018) were the first to present the data processing procedure in detail. They collected the energy intensity statistics data of 191 Chinese cities from 2003 to 2013 and derived the total primary energy consumption by multiplying GDP with energy intensity. Following Huang et al. (2018), the energy intensity statistics of an additional 57 cities were collected, and the study period is expanded to 2016. This generates a complete set of energy consumption data in China’s 248 prefecture-level cities from 2003 to 2016, which is the sample size in this study.

Furthermore, variables related to city characteristics, including real GDP per capita (pgdp), the proportion of FDI in GDP (sfdi), the proportion of secondary industry value added in GDP (sind), and environmental regulation (er) proxied by the SO₂ removal rate, are obtained from the China City Statistical Yearbook (2004–2017). The reasons for control variables’ selection are as follows. First, economic development levels measured by GDP per capita in different cities generally correspond to heterogenous energy productivities. However, there is no consensus regarding the effect of economic growth on energy productivity (Sadorsky, 2013; Bhattacharya et al., 2020). Second, overseas capital represented by FDI provides some power and support for regional development, not only providing capital and technology for productivity improvements, but also promoting employment and GDP growth. However, conclusions about the impact of FDI on energy productivity are inconsistent. For example, Elliott et al. (2013) found a negative relationship between FDI inflows and energy intensity (reciprocal of energy productivity) using the city-level data of China, indicating that FDI had a positive impact on energy productivity. On the contrary, Adom and Amuakwa-Mensah (2016) suggested that FDI significantly decreased energy productivity. Given that, we introduce FDI into the regression model. Third, the change from low- to high-energy-efficiency industries can result in a corresponding increase in energy productivity (Elliott et al., 2017).

Besides, in order to effectively control environmental pollution, Chinese policymakers have made great efforts to reduce the energy consumption of different industries by optimizing and adjusting the industrial structure. Hence, changes in the industrial structure play an important role on energy intensity (Wu, 2012). Thus, we introduce industrial structure into the regression model. Finally, environmental regulation can stimulate firms to reduce the costs of pollution emissions by using more clean energy, facilitating energy efficiency improvement in the long run (Sun et al., 2020b). Therefore, we expect that environmental regulation exerts a positive impact on energy productivity. Following Huang and Hua (2018), we use the SO₂ removal rate as a proxy for environmental regulation.

Descriptive statistics are summarized in Table 1. The treatment group consists of prefecture-level cities that opened HSR during 2003–2016, and the control group consists of ones without HSR. In the treatment group, the average energy productivity is approximately 1.478 (10000 CNY/tce, where CNY represents the Chinese Yuan, and tce represents tonne of standard coal equivalent). The average energy productivity of the control group is approximately 0.885 (10000 CNY/tce). A two-tailed t-test identifies a statistically significant variation in the mean values of city characteristics and energy productivity. Matching techniques are adopted to avoid the selection bias.

Table 1

Descriptive statistics of various variables

	Total sample			Control group (hsr = 0)			Treatment group (hsr = 1)
Variables	Obs.	Mean	Std. Dev.	Obs.	Mean	Std. Dev.	Obs.	Mean	Std. Dev.
ep	3472	1.021	0.546	2677	0.885***	0.452	795	1.478	0.584
pgdp	3472	2.928	3.197	2677	2.267***	2.376	795	5.156	4.381
sfdi	3472	0.148	0.162	2677	0.135***	0.157	795	0.191	0.170
sind	3472	49.247	10.839	2677	49.312	11.417	795	49.028	8.616
er	3472	0.427	0.284	2677	0.371***	0.276	795	0.615	0.224

Notes: (1) Variable ep is measured in 10000 CNY/tce; (2) Obs. and Std. Dev. denote observations and standard deviation, respectively; (3) *** indicates that the difference of the means is statistically significant at the 1% level between the treatment and control groups, as tested with a two-tailed t-test.

4. Results and Discussion

4.1. Baseline Results

The baseline DID estimates of Equation (1) are shown in Table 2. Without control variables, the coefficient of hsr is 0.093 (in Column (1)), and 0.094 when control variables are considered (in Column (2)). Both are statistically significant at the 1% level, indicating that HSR connection is beneficial to energy productivity improvement. A back-of-the-envelope calculation shows that one standard deviation increase in hsr raises energy productivity by 0.073 standard deviation.

Table 2

Baseline estimation results of the energy productivity effect of HSR

	DID		PSM-DID		MDM-DID
Variables	(1)	(2)	(3)	(4)	(5)	(6)
hsr	0.093***	0.094***	0.090***	0.093***	0.109**	0.093***
	(0.027)	(0.026)	(0.026)	(0.025)	(0.030)	(0.029)
pgdp		0.017*		0.018**		0.003
		(0.009)		(0.009)		(0.010)
sfdi		0.010***		0.011***		0.011***
		(0.002)		(0.002)		(0.002)
sind		–0.186		–0.154		–0.029
		(0.183)		(0.204)		(0.166)
er		0.017		0.025		0.009
		(0.035)		(0.035)		(0.036)
Year fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
City fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Obs.	3472	3472	3207	3207	2943	2943
R-squared	0.884	0.889	0.884	0.890	0.878	0.887

Notes: (1) Standard errors are clustered at the city level; (2) *** p<0.01, ** p<0.05, and * p<0.1.

Furthermore, we find that both economic development (pgdp) and foreign direct investment (sfdi) are significantly and positively associated with energy productivity, indicating that China’s economic development mode can gradually achieve efficient energy use and that FDI plays a significant role in China’s economic development and transition (Jin et al., 2020) to improve energy productivity. The findings are consistent with those of Shao et al. (2019). Environmental regulation (er) also shows a positive correlation with energy productivity, indicating that stricter environmental standards encourage policymakers to adopt energy-saving and emission-reduction policies, leading to energy productivity improvement through energy transition (Jia et al., 2021). In addition, the coefficient of industrial structure (sind) is not significant. A possible explanation for this finding is that the negative effect of an increase in the share of secondary industry in GDP on energy productivity is partially offset by the gradually increasing energy efficiency of China’s industrial sector (Wu et al., 2012).

For comparison, PSM-DID (in Columns (3) and (4)) and MDM-DID (in Columns (5) and (6)) estimates are presented in Table 2. The results are consistent with the DID estimators, indicating that HSR connection significantly improves energy productivity by approximately 9%, ceteris paribus, and our results are not dependent on matching methods. We further find that one standard deviation increase in hsr raises energy productivity by nearly 0.071 standard deviation. The estimation results of control variables are consistent with those of the DID strategy. Thus, Hypothesis 1 is verified, i.e., HSR connection exerts a significant positive impact on energy productivity. More details on how sample selection bias is tested can be found in Appendix B. Estimation results presented in the following sections are based on the PSM-DID method.

To test the parallel trend assumption, Equation (2) is estimated, and the coefficients are plotted in Figure 2. Evidently, HSR operation imposes a significantly positive effect on energy productivity after 2008. Regarding the magnitude of the effect, the positive effect presents a significant intensifying trend (β_–8 < β_–7 < . . . < β₇ < β₈). One may concern that the coefficients of treatments have an increasing trend (even before 0) in Figure 2. The design and planning of HSR network before the operation of the first HSR line (2008) may have contributed to the use of investments and technology flows between cities prior to the operation of the first HSR line, causing the coefficients of the variable ${hsr}_{it}^{j . .} j <$ to be positive, especially for the prior two years (2006 and 2007). Nevertheless, these coefficients are not significantly different from zero, indicating that there is no anticipated effect of HSR operation. We could not reject the null hypothesis of the consistent time-varying trends of energy productivity in both the HSR and non-HSR cities before the initial of the implementation of HSR. Thus, the DID parallel trend assumption is satisfied. Therefore, the implementation of HSR can be considered as a quasi-natural experiment for investigating the effect of HSR on energy productivity. In addition, our findings indicate that a significant increase in energy productivity occurred after the first opening of the HSR, and the effect of HSR on energy productivity gradually increased in the following eight years. However, in the following two years after 2008, the coefficients of ${hsr}_{it}^{j}$ are just statistically significant at the 10% level. This may be explained by the fact that the impacts of HSR operation on energy productivity have a time lag effect.

Figure 2

Dynamic effects of HSR on energy productivity

4.2. Endogeneity Problem Treatment

In addition to simultaneity and reverse causality, the potential endogeneity problem associated with the relationship between implementing HSR and energy productivity may result from measurement errors, omitted variable bias, and unobservable heterogeneity. Hence, the instrumental variable method is applied to overcome the endogeneity problem and ensure the robustness of the empirical results. The instrumental variables used when investigating the effect of transportation infrastructure include geographic slope (Duflo and Pande, 2007), established railway network (Zheng and Kahn, 2013; Baum-Snow et al., 2017), least-cost path-spanning tree networks (Faber, 2014), and historical planning documents and actual plans (Duranton and Turner, 2012; Michaels, 2008).

This study chooses historical information about rail rays in 1990 and courier stations during the Ming dynasty⁵ as HSR’s instrumental variables. An applicable instrumental variable needs to meet the correlation and exogenous conditions. First, the historical information meets the correlation condition as rail rays in 1990 and courier stations during the Ming dynasty generally do not change over time. It is more favorable to construct HSR with more rail rays and courier stations, which is positively correlated with the implementation of HSR and thus satisfies the correlation condition. Second, both the rail rays and courier stations are exogenous historical variables that do not affect energy productivity directly. Thus, the use of rail rays and courier stations as the instrumental variables of HSR is valid. The spatial distribution of courier stations and routes during the Ming dynasty can be found in Appendix C.

As these two instrumental variables are time-invariant, panel data set cannot be used to run the instrumental variable regression. Following Dong et al. (2020), the long-difference equation from 2003 to 2016 can be estimated. The equation is given as follows:

\log (\frac{e p_{i 2016}}{e p_{i 2003}}) = c_{0} + ρ_{1} hs r_{i 2003 - 2016} + ρ_{2} (x_{i 2016} - x_{i 2003}) + ρ_{3} \log (e p_{i 2003}) + v_{i}

(3)

where $\log (\frac{e p_{i 2016}}{e p_{i 2003}})$ measures the growth of energy productivity in city i from 2003 to 2016; hsr_i_2003–2016 is equal to 1 if city i has implemented HSR during the research period; time-varying city attributes (x_i) are also first-differenced; v_i is the error term; c₀, ρ₁, ρ₂, and ρ₃ are coefficients to be estimated. Specifically, the energy productivity is controlled in the initial year (ep_i₂₀₀₃) (Agrawal et al., 2017; Faber, 2014). Therefore, using these two instrumental variables and based on the controls, it can be concluded that HSR is responsible for changes in energy productivity.

Table 3 presents the two-stage least squares (2SLS) regression results. The results, as presented in Panel A of Table 3, are consistent with our baseline results. This implies that implementing HSR does improve energy productivity. Panel B shows the first-stage regression results of HSR on the railways and courier stations, indicating that HSR is significantly and positively correlated to both railways and courier stations. This confirms that our instrument variables are the strong predictors of the HSR connection. The KP F-statistics far exceed 10, indicating that the null hypothesis about the weak instrument variable is rejected according to the rule-of-thumb F statistics for IV method. Thus, the instrument variables are valid.

Table 3

2SLS estimation results of the energy productivity effect of HSR

	(1)	(2)
Panel A: Second-stage estimation
hsr	0.318***	0.212*
	(0.111)	(0.119)
Obs.	248	248
R-squared	0.219	0.269
Panel B: First-stage estimation
rrs	0.134***
	(0.020)
mdcs		0.353***
		(0.635)
KP F-statistics	43.760	30.870

Notes: (1) Robust standard errors are presented in parentheses; (2) *** p<0.01 and * p<0.1; (3) rrs and mdcs denote the rail rays in 1990 and courier stations during the Ming dynasty, respectively.

5. Robustness Checks and Heterogeneity Analysis

5.1. Robustness Checks

5.1.1. Placebo Test

The placebo test is designed to ensure that the baseline regression results are not affected by chance. For this purpose, we randomly assign hsr to sample cities and ensure the robustness of results. After 1000 placebo tests, 1000 coefficients of hsr^false are obtained, and then it is tested whether the mean of these coefficients is equal to zero. If the coefficients are still significantly positive, then it indicates that the original estimation results are biased, and that the changes of the energy productivity are likely to be influenced by other policies or random factors. After repeating the process 1000 times, the distribution of these coefficients is presented in Figure 3, indicating that it is an approximately normal distribution with the mean value of –0.0002 which is close to zero. Since the true energy productivity effects of HSR from the baseline regression far exceed the coefficients of hsr^false, one can believe that energy productivity growth is indeed explained by the implementation of HSR, instead of being affected by other systematic or random shocks. This confirms the credibility of the baseline results.

Figure 3

Placebo test

5.1.2. Concurrent Events

If any other events occurred around 2008 that affected GDP growth and energy consumption across cities, the estimates could be biased. Two such events are the 2008–2009 financial crisis and the 2008 Olympic Games. To investigate the effects of the financial crisis, both years of 2008 and 2009 are removed from the sample. The estimation results of the curtailed sample are shown in Columns (1) and (2) of Table 4. Regarding the 2008 Olympic Games, He et al. (2016) showed that not only Beijing but also Hebei, Tianjin, Shanxi, Inner Mongolia, and Liaoning were affected. To investigate the potential effects of this event, cities in these provinces that were affected by the Olympic Games in years of 2007 and 2008 are excluded (Columns (3) and (4) of Table 4).

Table 4

Tests for concurrent events

	(1)	(2)	(3)	(4)
Variables	2008–2009 financial crisis		2008 Beijing Olympic Games
hsr	0.098***	0.098***	0.078***	0.067**
	(0.026)	(0.026)	(0.027)	(0.027)
Province×Trend	NO	YES	NO	YES
Control variables	YES	YES	YES	YES
Year fixed effects	YES	YES	YES	YES
City fixed effects	YES	YES	YES	YES
Obs.	2735	2735	2370	2370
R-squared	0.891	0.892	0.890	0.893

Notes: (1) Standard errors are clustered at the city level; (2) *** p<0.01 and ** p<0.05; (3) “Trend” denotes the linear time trend, which is equal to corresponding year minus 2002.

Moreover, an interaction term between province and trend is introduced to capture provincial characteristics which cannot be captured at the city level, a very demanding specification which allows for regional trends. For example, the political turnover for a certain province may affect its environmental governance and economic performance (Li and Zhou, 2005), which further affect energy structure (Kong et al., 2021) and energy productivity. In addition, the reasons for the consideration of the interaction term between province and trend are as follows. First, it is important to understand that HSR operation is not random, which is related to the city’s location. Second, regional differences in economic development also affect energy productivity in various ways. Third, these differences may have different impacts on energy productivity over time, which could result in biased estimations. Given that, we further introduce the interaction term between province and trend in the baseline model. The coefficients of hsr in Table 4 remain similar to the baseline results presented in Table 2. This suggests that these two concurrent events do not affect our baseline estimates.

5.1.3. Substitutability between Hsr and other Transportation Modes

The aforementioned results suggest that HSR improves energy productivity. As discussed in Section 2, HSR operation improves energy productivity by strengthening industrial agglomeration, economic activities, and factor redistribution effects. Moreover, technological innovation effect also plays a principal role in the causal relationship between HSR and energy productivity. However, cities connected via HSR stations may also obtain these effects through other transportation modes, such as traditional railway and aviation. Consequently, both traditional railway and aviation could substitute HSR in terms of energy productivity. Such a potential substitutability between HSR and other transportation modes should be further examined based on the following econometric model:

e p_{it} = d_{0} + θ_{1} hs r_{it} + θ_{2} hs r_{it} \times \mod e_{it} + θ_{3} x_{it} + ξ_{i} + ζ_{t} + ϕ_{it}

(4)

where mode represents other two transportation modes, i.e., traditional railway and aviation, which are proxied by railway and aviation passenger volumes, respectively; $ξ_{i}$ and $ζ$ represent the city fixed effect and the year fixed effect, respectively; $ϕ_{it}$ is the error term; d₀, θ₁, θ₂, and θ₃ are coefficients to be estimated. Compared with Equation (1), the interaction term between hsr and mode is introduced, which allows us to explore how multiple transportation modes operate interactively in improving energy productivity. Thus, we have

\frac{\partial e p_{it}}{\partial hs r_{it}} = θ_{1} + θ_{2} \times \mod e_{it}

(5)

Based on Equation (5), we examine whether multiple transportation modes complement or substitute for each other in increasing energy productivity. Further, we use the economic concept of marginal effect to reflect whether HSR and traditional railway (aviation) work as complements or substitutes. From the complementary view, HSR could improve the marginal effect of another transportation mode on energy productivity. Thus, the coupling of transportation modes will improve energy productivity if they serve as complements. On the contrary, from the substitutive view, HSR could decrease the marginal effect of another transportation mode on energy productivity (Oh et al., 2018).

Given that θ₁ > 0, if θ₂ is significantly negative, then cities with higher energy productivity will be less affected by traditional railway or aviation, indicating a substitution effect. Specifically, if there is a substitution effect, a higher level of HSR operation may lower the marginal effects of other transportation modes on energy productivity. Figure 4 shows the estimation results, indicating that the estimates of θ₂ for both traditional railway and aviation are quantitatively small (approximately –0.006) and not statistically significant. This suggests that for improving energy productivity, HSR operation could not be substituted by traditional railway and aviation, i.e., the substitution effects are not significant. This finding differs from those of Zhang et al. (2018), who found that HSR’s substitutions were significant for short and medium-haul air routes, while introducing HSR services in China increased long-distance air travel.

Figure 4

Substitutability of HSR with traditional railway (a) and aviation (b)

5.1.4. Alternative Measures of Energy Productivity

As we discussed in Section 3.3, the DMSP/OLS satellite (night-time light) data is an excellent proxy for economic growth because the data can eliminate the interference of human factors, to a great extent. The night-time light data can more accurately reflect the spatiotemporal dynamics of GDP at a more detailed level over large geographical areas. Hence, we use the night-time light data to proxy GDP while measuring energy productivity and re-estimate Equation (1), and the corresponding results are shown in Column (1) of Table 5. Moreover, considering a single-factor energy productivity does not involve the substitution or complementarity between energy and other input factors (Stern, 2012), we further adopt a total factor energy productivity for robustness check. Specifically, a stochastic frontier analysis approach (more details can be found in Appendix D) is used to measure energy productivity and re-estimate Equation (1), and the corresponding results are reported in Column (2) of Table 5. After replacing energy productivity indicators, the coefficients of HSR still remain significantly positive. This indicates that the baseline estimation results are robust, again.

Table 5

Estimation results based on the alternative measures of energy productivity

Variables	(1)	(2)
hsr	2.054***	0.002*
	(0.001)	(0.001)
Control variables	Yes	Yes
Year fixed effects	Yes	Yes
City fixed effects	Yes	Yes
Obs.	3196	3207
R-squared	0.801	0.865

Notes: (1) Standard errors are clustered at the city level; (2) *** p<0.01 and * p<0.1.

5.2. Heterogeneity Analysis

5.2.1. Heterogeneity across Quantiles

When evaluating the effect of a certain policy, policymakers may be interested in more than the average impact of the policy. Quantile treatment effects are measured as the difference between the quantiles of the treatment group distribution and control group distribution. It is impractical to estimate quantile treatment effects using traditional methods such as the linear regression, which depends on the effect of an intervention being the same across all individuals. Due to the specific characteristics of the sample cities, the effect of a macroeconomic activity on energy productivity may be heterogeneous across quantiles. Consequently, the quantile regression is a feasible method to investigate such heterogeneous effects because the results of the method are robust and less sensitive to the heteroscedasticity and outliers that frequently exist in real-world data (Koenker and Bassett, 1978; Baur, 2013).

In this section, the heterogeneity effects of HSR on energy productivity are examined across different quantiles. The quantile regression is used to assess how hsr affects the energy productivity for cities whose energy productivity tends to be low (e.g., 5th percentile), high (e.g., 95th percentile), or anywhere in between. Table 6 shows the results of the quantile regression. Evidently, the impacts of HSR on energy productivity gradually grow with increasing quantiles. In addition, the coefficients of hsr remain significantly positive, indicating the robustness of the baseline regression results.

Table 6

Quantile regression results

	(1)	(2)	(3)	(4)	(5)
Variables	5th	25th	50th	75th	95th
hsr	0.187***	0.320***	0.389***	0.545***	0.566***
	(0.006)	(0.012)	(0.018)	(0.014)	(0.023)
Control variables	Yes	Yes	Yes	Yes	Yes
Year fixed effects	Yes	Yes	Yes	Yes	Yes
City fixed effects	Yes	Yes	Yes	Yes	Yes
Obs.	3207	3207	3207	3207	3207

Notes: (1) Standard errors in parentheses are obtained using the Markov Chain Monte Carlo algorithm with 2000 replications; (2) *** p<0.01.

5.2.2. Heterogeneity across Distances

Based on Tobler’s First Law of geography (Tobler, 1970), the heterogeneity effects of HSR on energy productivity need to be further investigated from the perspective of distance decay (in km). Specifically, a dummy variable distance is generated, and the interaction term between hsr and distance are introduced into the baseline model to establish the following regression model:

e p_{it} = e_{0} + η_{1} hs r_{it} + η_{2} hs r_{it} \times distance + + η_{3} x_{it} + ϑ_{i} + υ_{t} + ψ_{it}

(6)

where distance indicates a dummy variable, which is equal to 1 if the distance between a city center and the nearest HSR station is less than 2.1 km (the first percentile, 1%), and zero otherwise. Other percentiles are also estimated for comparison, i.e., 5%, 10%, 25%, 50%, 75%, 90%, 95%, and 99%, with the corresponding distances of 3.7 km, 4.7 km, 7.3 km, 23.5 km, 141.2 km, 382.3 km, 472.1 km, and 815.7 km, respectively. Variables $ϑ_{i}$ and $υ_{t}$ represent the city fixed effect and the year fixed effect, respectively; $ψ_{it}$ is the error term; e₀, η₁, η₂, and η₃ are coefficients to be estimated.

As shown in Figure 5, we find that there is a significant decay effect of HSR on energy productivity, i.e., the effect of HSR on energy productivity gradually decreases with increasing distance to the nearest HSR station, and the magnitude of this effect remains at around 0.1. Thus, heterogeneity effects across distances are verified. The empirical findings are consistent with some previous studies (e.g., Zhang et al., 2018) which suggests that HSR’s effect is the greatest in the short distance class (below 500 km), followed by the median distance (501–1000 km).

Figure 5

Results on heterogeneity effects across distances

6. Mechanism Identification

6.1. Role of HSR Network Accessibility

Using the social network analysis method to calculate the node degree (local analysis), closeness (global analysis), and betweenness (global analysis), we can explore the ramifications of HSR network accessibility for improving energy productivity. Following Opsahl et al. (2010), HSR network is defined (cities are treated as nodes), and three indicators, i.e., the degree of centrality for local analysis and the closeness and betweenness for global analysis, are constructed to measure HSR network accessibility, based on the social network analysis method.

The degree of HSR station centrality provides an approximation of the structural surroundings of the station. However, it does not consider the global structure of HSR network. Thus, although HSR station connectivity may be high, the connections may be slow (Opsahl et al., 2010). To overcome this difficulty and capture the global feature, the centrality measures of closeness and betweenness are considered.

These two measures hinge on identifying and obtaining the lengths of the shortest paths between and among HSR stations. If the network is unweighted (i.e., the edge weight can be set to 1), the shortest path between two nodes can be found by minimizing the number of intermediaries; the resulting length is defined via the lowest number of edges that directly or indirectly connect the two nodes.

In this study, the closeness is the inverse of the shortest path from one node to other nodes in the network—the more central the node, the greater its proximity to all other nodes. The betweenness is a measure to identify the number of shortest paths that traverse through a node. A higher betweenness value indicates that the short paths that traverse through other stations to the HSR station are more numerous. Such high-betweenness stations can control over flow.

The estimation results are presented in Table 7. We find that the coefficients of degree, closeness, and betweenness are all significantly positive, indicating that strengthening the centrality of HSR cities is conducive to improving energy productivity, which is in line with baseline results. Thus, Hypothesis 2 is verified, i.e., HSR network accessibility contributes to a significant improvement in energy productivity. The policy implications of these findings are that optimizing HSR network centrality can improve energy productivity, and HSR network is conducive to the sustainable development of both the economy and the energy sector. Therefore, policymakers should still vigorously develop HSR construction to improve energy productivity and even promote energy transition by optimizing the centrality of HSR network.

Table 7

Estimation results of the energy productivity effects of HSR network accessibility

Variables	(1)	(2)	(3)
degree	0.024***
	(0.009)
closeness		0.010***
		(0.003)
betweenness			0.022**
			(0.009)
Control variables	Yes	Yes	Yes
Year fixed effects	Yes	Yes	Yes
City fixed effects	Yes	Yes	Yes
Obs.	3207	3207	3207
R-squared	0.889	0.889	0.889

Notes: (1) Standard errors are clustered at the city level; (2) *** p<0.01 and ** p<0.05; (3) The measures of degree, closeness, and betweenness are based on Opsahl et al. (2010).

6.2. Identifying the Mediating Mechanism

Through the DID, PSM-DID, MDM-DID, and IV estimates, we confirm that HSR improves energy productivity. However, so far, the mechanism behind this outcome has not been identified. As mentioned above, HSR may improve energy productivity through a technological innovation effect. Following Baron and Kenny (1986), a three-step method is employed to verify the technological innovation effect. First, energy productivity and HSR are regressed. A significant HSR coefficient suggests that it affects energy productivity. Second, HSR and technological innovation are regressed. In this case, a significant HSR coefficient implies that HSR is responsible for the technological innovation effect. Third, technological innovation and HSR are regressed with energy productivity. In this case, an insignificant HSR coefficient points to a complete mediating effect. In the inverse case (i.e., significant HSR coefficient), the mediating effect is not complete, but significantly exists. The detailed verification process for the above is as follows:

Step 1: Examining the effect of HSR on energy productivity based on Equation (1).

Step 2: Testing the effect of HSR on technological innovation based on the following model.

m_{it} = f_{0} + τ_{1} hs r_{it} + τ_{2} x_{it} + g_{i} + h_{t} + φ_{it}

(7)

Step 3: Incorporating HSR and technological innovation variable into the following regression equation.

e p_{it} = f'_{0} + τ'_{1} hs r_{it} + τ'_{2} m_{it} + τ'_{3} x_{it} + g'_{i} + h'_{t} + φ'_{it}

(8)

where m_it denotes the technological innovation variable. We use three indicators to measure the variable: First, the city technological innovation index (innovation) from Kou and Liu (2017); second, the total number of annual patent applications authorized in a city (patent) from the official patent website (https://app.patentcloud.com/); third, the proportion of science and technology expenditure in the government’s general budget (techexp) from the China City Statistical Yearbook (2004–2017). In Equation (7), g_i and h_t represent the city fixed effect and the year fixed effect; φ_it is the error term; f₀, τ₁, and τ₂ are coefficients to be estimated. Similarly, in Equation (8), $g'_{i}$ and $h'_{t}$ represent the city fixed effect and the year fixed effect; $φ'_{it}$ is the error term; $f'_{0}, τ'_{1}, τ'_{2},$ and $τ'_{3}$ are coefficients to be estimated.

The results are shown in Table 8. Column (1) of Table 8 presents that the total effect of HSR on energy productivity is 0.093, which is significantly positive at the 1% level. HSR operation is confirmed to significantly improve energy productivity, and the first step of testing the mechanism is verified. Columns (2) and (3) show the regression results of the HSR effect on the technological innovation, and the effects of both HSR and the technological innovation index on energy productivity, respectively. Column (2) of Table 8 shows that the regression coefficient of HSR is significantly positive at the 1% level, indicating that HSR operation strengthens technological innovation. Column (3) of Table 8 shows that the regression coefficients of both HSR and innovation are significantly positive at the 1% level, indicating that technological innovation effectively promotes energy productivity.

Table 8

Results of the mediation effect identification

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Variables	ep	innovation	ep	patent	ep	techexp	ep
hsr	0.093***	0.048***	0.083***	0.224***	0.068***	0.003***	0.077***
	(0.025)	(0.018)	(0.024)	(0.058)	(0.024)	(0.001)	(0.024)
innovation			0.221***
			(0.024)
patent					0.128***
					(0.035)
techexp							4.707***
							(1.694)
Control variables	YES	YES	YES	YES	YES	YES	YES
Year fixed effects	YES	YES	YES	YES	YES	YES	YES
City fixed effects	YES	YES	YES	YES	YES	YES	YES
Indirect effects	—	0.011**		0.029***		0.014**
		(0.004)		(0.008)		(0.041)
Sobel test statistics	—	2.561		2.655		2.039
Proportion of indirect effects in total effects	—	11.332%		29.659%		15.497%
Obs.	3207	3207	3207	3179	3179	3207	3207
R-squared	0.890	0.602	0.899	0.775	0.899	0.727	0.893

Notes: (1) Standard errors are clustered at the city level; (2) *** p<0.01, ** p < 0.05; (3) Sobel test statistics are obtained by an interactive calculation tool for the mediation test (see http://www.quantpsy.org/sobel/sobel.htm).

Based on these results, HSR operation can strengthen city technological innovation to improve energy productivity. The indirect effect is 0.011 $(= τ_{1} \times {τ'}_{2} = 0.048 \times 0.221)$ , accounting for 11.332% (that is $\frac{τ_{1} \times {τ'}_{2}}{τ_{1} \times {τ'}_{2} + {τ'}_{1}} \times 100 % = \frac{0.048 \times 0.221}{0.048 \times 0.221 + 0.083} \times 100 %$ ) of the total effect. The Sobel test statistic confirms the existence of a technological innovation mediating effect at the 5% level. Similar results are obtained in Columns (4) and (5) for patent and in Columns (6) and (7) for techexp. The empirical results are also consistent with those of Chen (2021), who found an increase in both granted invention patents and invention patent applications after the operation of HSR. It can be inferred that the implementation of HSR contributes to the flow of knowledge both within and between cities (Dong et al., 2020) to promote innovation in science and technology (Chen, 2021; Hanley et al., 2022) and further improve energy productivity. To sum up, the mechanism through which HSR improves energy productivity is verified, indicating that Hypothesis 3 holds.

7. Conclusions and Policy Implications

This study uses the data of China’s 248 prefecture-level cities from 2003 to 2016 to examine the effect of HSR on energy productivity. A sound understanding of how HSR affects energy productivity provides valuable insights into the trajectory of China’s energy consumption. The results show that HSR exerts a significant positive effect on energy productivity over the research period. On average, both DID and matching DID estimates show that the marginal impact of HSR on energy productivity is approximately 9%. Considering the potential endogeneity problem, we adopt an instrumental variable method to re-estimate the energy productivity effect of HSR. The results suggest that the baseline results are robust. Furthermore, a series of robustness checks also confirm the causal effects of HSR on energy productivity. Furthermore, the heterogeneity analysis suggests that the positive impacts of HSR on energy productivity gradually grow with increasing quantiles. This further indicates that HSR effectively improves energy productivity in cities closer to HSR stations; however, a significant decay effect emerges with increasing distance to the nearest HSR station.

Besides, HSR connection cannot be completely substituted by traditional railway and aviation in improving energy productivity. Moreover, the mechanism analysis shows that both HSR network accessibility and technological innovation have positive effects on energy productivity, indicating that these two channels strengthen the effect of HSR operation on energy productivity. These findings provide new perspectives for the design of a more balance development strategy in China. Policymakers should consider the role of the existing HSR network and the technological innovation effect, and practitioners should comprehensively consider both transit-oriented development and ecology-oriented development modes.

Pertinent policy implications are summarized as follows. First, HSR can exert a significant positive externality effect on energy productivity. HSR promotes sustainable energy development by optimizing both the structure and accessibility of the intercity network. Currently, HSR construction is a high-tech and high-input project, which has a lagged economic return rate. However, from a long-term perspective, the advantages of HSR outweigh these disadvantages. Therefore, policymakers should vigorously implement HSR and promote a coordinated development of energy sector and other industries.

Second, the shorter the distance to a city with HSR station, the greater the energy productivity effect. Two potential reasons are the unbalanced development of the regional economy and the siphon effect of large cities, which incentivizes enterprises with low energy consumption and high output to move closer to cities with HSR, while most low value-added enterprises move to cities without HSR. This implies that when planning HSR construction, policymakers should adjust measures to suit local conditions, as well as the current situation of regional economy and energy consumption. Moreover, construction policies for transportation facilities that are consistent with the development level of specific regions should be formulated. Cities with higher HSR network centrality generally have higher energy productivity; therefore, governments should continue to develop and improve HSR network to strengthen the network centrality of key cities and thus improve energy productivity.

Third, our finding shows that implementing HSR is conducive to improving technological innovation capacity, which also improves energy productivity. HSR connection can accelerate the long-term flow of technological innovation elements. Thus, policymakers should vigorously improve the comprehensive quality and efficiency of technological innovation and promote the integration of production, education, and research. In addition, enterprises should be encouraged to obtain higher profits through production and service technological innovation, regional industrial upgrading, and structural adjustment. These measures can comprehensively improve energy productivity and lead to high-quality economic development.

Finally, cities that are not connected to HSR should continuously improve their comprehensive competitiveness via economic and social development to obtain HSR investment. Moreover, their interconnection with HSR cities needs to be accelerated, and a seamless connection with HSR cities should be realized by creating various channels, e.g., expressways and accelerated integration into the technological innovation circle of HSR. The flows of production factors and technology and knowledge spillovers also need to be improved to enhance technological innovation capability and energy productivity.

It is noteworthy that more work needs to be done in the future. The present study could not estimate the potential spatial effect of neighboring cities with HSR stations on energy productivity. This limitation may lead to an underestimation of the actual effect of HSR. Thus, a spatial DID approach (e.g., Jia et al., 2021) may be introduced in a future study. In addition, although this paper investigates the benefits of HSR operation, it is also worth to estimate the economic costs of HSR connection in the general context of improved energy productivity.

Supplemental Material

sj-pdf-1-enj-10.5547_01956574.45.1.yayu – Supplemental material for High-Speed Rail and Energy Productivity: Evidence from China

Supplemental material, sj-pdf-1-enj-10.5547_01956574.45.1.yayu for High-Speed Rail and Energy Productivity: Evidence from China by Yantuan Yu and Shuai Shao in The Energy Journal

Footnotes

Appendix A: HSR-Opened Years for Sample Cities

Table A1

HSR-opened years for sample cities during 2008–2016

City	Year	City	Year	City	Year	City	Year
Beijing	2008	Hefei	2008	Weihai	2014	Jiangmen	2011
Tianjin	2008	Wuhu	2015	Dezhou	2011	Zhaoqing	2014
Shijiazhuang	2009	Bengbu	2011	Zhengzhou	2010	Huizhou	2013
Tangshan	2013	Huainan	2012	Kaifeng	2014	Shanwei	2013
Handan	2012	Maanshan	2015	Luoyang	2010	Qingyuan	2009
Baoding	2012	Tongling	2015	Anyang	2012	Dongguan	2011
Cangzhou	2011	Anqing	2015	Hebi	2012	Zhongshan	2011
Langfang	2011	Huangshan	2015	Xinxiang	2012	Chaozliou	2013
Taiyuan	2009	Chuzhou	2008	Jiaozuo	2015	Jieyang	2013
Yangquan	2009	Suzhou	2011	Xuchang	2012	Yunfu	2014
Jinzhong	2009	Lu’an	2009	Luohe	2012	Nanning	2013
Yuncheng	2014	Chizhou	2015	Sanmenxia	2010	Liuzhou	2013
Linfen	2014	Xuancheng	2015	Shangqiu	2016	Guilin	2013
Dalian	2012	Fuzhou	2009	Xinyang	2012	Beihai	2013
Benxi	2015	Xiamen	2010	Zhumadian	2012	Haikou	2010
Changchun	2010	Putian	2010	Wuhan	2009	Sanya	2010
Jilin	2010	Sanming	2008	Huangshi	2014	Chongqing	2013
Siping	2012	Quanzhou	2010	Yichang	2010	Chengdu	2009
Harbin	2012	Zhangzhou	2013	Ezhou	2014	Deyang	2014
Qiqihar	2015	Nanping	2015	Xiaogan	2012	Mianyang	2014
Daqing	2015	Longyan	2015	Jingzhou	2012	Guangyuan	2015
Suihua	2015	Ningde	2009	Huanggang	2009	Suining	2009
Shanghai	2010	Nanchang	2010	Xianning	2009	Neijiang	2015
Nanjing	2008	Pingxiang	2014	Changsha	2009	Leshan	2014
Wuxi	2010	Jiujiang	2010	Zhuzhou	2009	Nanchong	2015
Xuzhou	2011	Xinyu	2014	Xiangtan	2009	Meishan	2014
Changzhou	2010	Yingtan	2014	Hengyang	2009	Guang’an	2015
Suzhou	2010	Ganzhou	2015	Shaoyang	2014	Ziyang	2015
Zhenjiang	2010	Yichun	2014	Yueyang	2009	Guiyang	2014
Hangzhou	2010	Fuzhou	2014	Chenzhou	2009	Liupanshui	2016
Ningbo	2009	Shangrao	2014	Yongzhou	2013	Anshun	2016
Wenzhou	2009	Jinan	2008	Huaihua	2014	Kunming	2016
Jiaxing	2010	Qingdao	2008	Loudi	2014	Qujing	2016
Huzhou	2013	Zibo	2008	Guangzhou	2009	Yuxi	2016
Shaoxing	2013	Zaozhuang	2011	Shaoguan	2009	Xi’an	2010
Jinhua	2014	Yantai	2014	Shenzhen	2011	Baoji	2013
Quzhou	2014	Weifang	2008	Zhuhai	2011	Xianyang	2013
Taizhou	2009	Jining	2011	Shantou	2013	Weinan	2010
Lishui	2015	Tai’an	2011	Foshan	2010	Lanzhou	2015

Appendix B: Tests for the Sample Selection Bias

To address the selection bias from the non-random operation of HSR, this study uses the PSM-DID method to check the robustness of baseline results. The detailed steps are as follows. First, grouping and characteristics variables are used as matching data. Second, the propensity score is estimated using the Logit model. Third, a one-to-one nearest neighbor matching is used with caliper. Matching balance and common value tests are conducted to ensure the liability of matching results. As shown in Figure B1, a better covariate balance is achieved through the one-to-one nearest neighbor matching with the caliper. The post-matching absolute standardized biases for all covariates are under 15%, indicating that a good covariates balance is obtained (Rosenbaum and Rubin, 1985).

The balance test is used to examine the differences in the means of the variables pre- and after- matching; it is also used to verify the distribution of variables between the treatment and control groups over the same two periods. Figure B2 compares the estimation of the kernel density of propensity score pre- and after- matching. After the matching, the distribution trends of both groups are consistent.

Appendix C: Spatial Distribution of Courier Stations and Routes in the Ming Dynasty

Appendix D: Estimating Energy Productivity Based on the Stochastic Frontier Analysis

For robustness checks, we estimate energy productivity based on the Cobb-Douglas production function. Capital, labor, and energy consumption are treated as input variables while the gross domestic product (GDP) is treated as dependent variable. Energy productivity estimation model can be expressed as follows:

(D1)

Y = f (K, L, E)

where K, L, and E denote capital stock, labor force, and energy consumption, respectively; Y represents GDP.

The stochastic frontier analysis used in this study is the “true” fixed effect model suggested by Greene (2005), because the characteristics of heterogeneous groups of cities can be reflected when using the model. The model can be presented as follows:

(D2)

y_{it} = α_{i} + β_{1} k_{it} + β_{2} l_{it} + β_{3} e_{it} + ε_{it}

where i and t denote cross-section and time-series dimensions, respectively; k_it, l_it, and e_it represent capital stock, labor force, and energy consumption for city i and year t; ε_it represents the error term, which can be divided into two components: idiosyncratic error (v_it) and technical inefficiency (u_it). We further assume that the technical inefficiency term follows a half-normal distribution as follows:

(D3)

ε_{it} = v_{it} - u_{it}

(D4)

v_{it} ~ N (0, σ_{v}^{2})

(D5)

u_{it} ~ N^{+} (0, σ_{u}^{2})

Given that, the energy productivity can be estimated as (Du et al., 2020):

(D6)

e p_{it} = \frac{{\hat{y}}^{f}_{it}}{y_{it}} = \exp (- u_{it})

where ep represents energy productivity, and ${\hat{y}}^{f}_{it}$ denotes the predicted dependent variable when technical inefficiency is zero.

Acknowledgements

This paper is supported by the National Natural Science Foundation of China (Nos. 71903068, 71922015, and 71773075), the Chinese National Funding of Social Sciences (No. 21ZDA084), and Shanghai Soft Science Research Program (No. 22692103400).

1.

This method is also called the single or partial factor productivity method. This method is used because (1) it measures productivity, which is easier to relate to specific processes, and (2) it alleviates the endogeneity problem in the empirical analysis, to some extent.

2.

Similarly, found that railway transportation was conducive to improving environmental efficiency.

3.

The Beijing-Tianjin HSR, the first new-generation HSR with a maximum speed of 350 km/h, started operation in August 2008.

4.

See .

5.

The original data of rail rays in 1990 and courier stations during the Ming dynasty are from Baum-Snow et al. (2017) and , respectively.

References

Adom

Philip Kofi

Amuakwa-Mensah

Franklin

(2016). “What Drives the Energy Saving Role of FDI and Industrialization in East Africa?” Renewable and Sustainable Energy Reviews 65: 925–42. https://doi.org/10.1016/j.rser.2016.07.039.

Agrawal

Ajay

Galasso

Alberto

Oettl

Alexander

(2017). “Roads and Innovation.” The Review of Economics and Statistics 99(3): 417–34. https://doi.org/10.1162/REST_a_00619.

Ahlfeldt

Gabriel

Feddersen

Arne

(2018). “From Periphery to Core: Measuring Agglomeration Effects Using High-Speed Rail.” Journal of Economic Geography 18(2): 355–90. https://doi.org/10.1093/jeg/lbx005.

Albalate

Daniel

Bel

Germà

(2012). “High-Speed Rail: Lessons for Policy Makers from Experiences Abroad.” Public Administration Review 72(3): 336–49. https://doi.org/10.1111/j.1540-6210.2011.02492.x.

Atack

Jeremy

Bateman

Fred

Haines

Michael

Margo

Robert A.

(2010). “Did Railroads Induce or Follow Economic Growth? Urbanization and Population Growth in the American Midwest, 1850–1860.” Social Science History 34(2): 171–97. https://doi.org/10.1017/S0145553200011202.

Atalla

Tarek

Bean

Patrick

(2017). “Determinants of Energy Productivity in 39 Countries: An Empirical Investigation.” Energy Economics 62: 217–29. https://doi.org/10.1016/j.eneco.2016.12.003.

Baron

Reuben

Kenny

David

(1986). “The Moderator–Mediator Variable Distinction in Social Psychological Research: Conceptual, Strategic, and Statistical Considerations.” Journal of Personality and Social Psychology 51(6): 1173–82. https://doi.org/10.1037/0022-3514.51.6.1173.

Baum-Snow

Nathaniel

Brandt

Loren

Henderson

Vernon

Turner

Matthew

Zhang

Qinghua

(2017). “Roads, Railroads, and Decentralization of Chinese Cities.” The Review of Economics and Statistics 99(3): 435–48. https://doi.org/10.1162/REST_a_00660.

Baur

Dirk

(2013). “The Structure and Degree of Dependence: A Quantile Regression Approach.” Journal of Banking & Finance 37(3): 786–98. https://doi.org/10.1016/j.jbankfin.2012.10.015.

10.

Beck

Thorsten

Levine

Ross

Levkov

Alexey

(2010). “Big Bad Banks? The Winners and Losers from Bank Deregulation in the United States.” The Journal of Finance 65(5): 1637–67. https://doi.org/10.1111/j.1540-6261.2010.01589.x.

11.

Berhe

Mesfin Welderufael

Huang

Junbing

Wedajo

Abel Dula

(2022). “Effect of China’s Technology Spillovers on Energy Intensity in Africa.” Chinese Journal of Population, Resources and Environment 20(2): 168–76. https://doi.org/10.1016/j.cjpre.2022.06.007.

12.

Berman

Lex

Zhang

Wei

(2017). “V6 Ming Dynasty Courier Routes and Stations.” Harvard Dataverse, V1. https://doi.org/10.7910/DVN/SB8ZTM.

13.

Bhattacharya

Mita

Inekwe

John

Sadorsky

Perry

(2020). “Convergence of Energy Productivity in Australian States and Territories: Determinants and Forecasts.” Energy Economics 85: 104538. https://doi.org/10.1016/j.eneco.2019.104538.

14.

Birol

Fatih

Keppler

Jan Horst

(2000). “Prices, Technology Development and the Rebound Effect.” Energy Policy 28(6–7): 457–69. https://doi.org/10.1016/S0301-4215(00)00020-3.

15.

Cagno

Enrico

Ramirez-Portilla

Andres

Trianni

Andrea

(2015). “Linking Energy Efficiency and Innovation Practices: Empirical Evidence from the Foundry Sector.” Energy Policy 83: 240–56. https://doi.org/10.1016/j.enpol.2015.02.023.

16.

Chen

Chia-Lin

Hall

Peter

(2012). “The Wider Spatial-Economic Impacts of High-Speed Trains: A Comparative Case Study of Manchester and Lille Sub-Regions.” Journal of Transport Geography 24: 89–110. https://doi.org/10.1016/j.jtrangeo.2011.09.002.

17.

Chen

Jun

(2021). “High-Speed Rail and Energy Consumption in China: The Intermediary Roles of Industry and Technology.” Energy 230: 120816. https://doi.org/10.1016/j.energy.2021.120816.

18.

Chen

Nordhaus

William

(2011). “Using Luminosity Data as a Proxy for Economic Statistics.” Proceedings of the National Academy of Sciences 108(21): 8589–94. https://doi.org/10.1073/pnas.1017031108.

19.

Chen

Zhenhua

Haynes

Kingsley

(2017). “Impact of High-Speed Rail on Regional Economic Disparity in China.” Journal of Transport Geography 65: 80–91. https://doi.org/10.1016/j.jtrangeo.2017.08.003.

20.

Chen

Zhenhua

Xue

Junbo

Rose

Adam

Haynes

Kingsley

(2016). “The Impact of High-Speed Rail Investment on Economic and Environmental Change in China: A Dynamic CGE Analysis.” Transportation Research Part A: Policy and Practice 92: 232–45. https://doi.org/10.1016/j.tra.2016.08.006.

21.

Cheng

Shihui

Lin

Jianyi

Wangtu

Yang

Dewei

Liu

Jiahui

Huimei

(2020). “Carbon, Water, Land and Material Footprints of China’s High-Speed Railway Construction.” Transportation Research Part D: Transport and Environment 82: 102314. https://doi.org/10.1016/j.trd.2020.102314.

22.

Cochran

William

Rubin

Donald

(1973). “Controlling Bias in Observational Studies: A Review.” Sankhyā: The Indian Journal of Statistics, Series A 35(4): 417–46. https://doi.org/10.2307/25049893.

23.

Dalla Chiara

Bruno

De Franco

Davide

Coviello

Nicola

Pastrone

Dario

(2017). “Comparative Specific Energy Consumption Between Air Transport and High-Speed Rail Transport: A Practical Assessment.” Transportation Research Part D: Transport and Environment 52: 227–43. https://doi.org/10.1016/j.trd.2017.02.006.

24.

de Rus Mendoza

Ginés

(2012). Economic Analysis of High Speed Rail in Europe. Fundación BBVA.

25.

Diao

(2018). “Does Growth Follow the Rail? The Potential Impact of High-Speed Rail on the Economic Geography of China.” Transportation Research Part A: Policy and Practice 113: 279–90. https://doi.org/10.1016/j.tra.2018.04.024.

26.

Dimitropoulos

John

(2007). “Energy Productivity Improvements and the Rebound Effect: An Overview of the State of Knowledge.” Energy Policy 35(12): 6354–63. https://doi.org/10.1016/j.enpol.2007.07.028.

27.

Donaldson

Dave

(2018). “Railroads of the Raj: Estimating the Impact of Transportation Infrastructure.” American Economic Review 108(4–5): 899–934. https://doi.org/10.1257/aer.20101199.

28.

Dong

Xiaofang

Zheng

Siqi

Kahn

Matthew E.

(2020). “The Role of Transportation Speed in Facilitating High Skilled Teamwork Across Cities.” Journal of Urban Economics 115: 103212. https://doi.org/10.1016/j.jue.2019.103212.

29.

Kerui

Shao

Shuai

Yan

Zheming

(2020). “Urban Residential Energy Demand and Rebound Effect in China: A Stochastic Energy Demand Frontier Approach.” The Energy Journal 42(2): 175–94. https://doi.org/10.5547/01956574.42.4.kdu.

30.

Duflo

Esther

Pande

Rohini

(2007). “Dams.” The Quarterly Journal of Economics 122(2): 601–46. https://doi.org/10.1162/qjec.122.2.601.

31.

Duranton

Gilles

Turner

Matthew

(2012). “Urban Growth and Transportation.” The Review of Economic Studies 79(4): 1407–40. https://doi.org/10.1093/restud/rds010.

32.

Elliott

Robert

Sun

Puyang

Chen

Siyang

(2013). “Energy Intensity and Foreign Direct Investment: A Chinese City-Level Study.” Energy Economics 40: 484–94. https://doi.org/10.1016/j.eneco.2013.08.004.

33.

Elliott

Robert

Sun

Puyang

Zhu

Tong

(2017). “The Direct and Indirect Effect of Urbanization on Energy Intensity: A Province-Level Study for China.” Energy 123: 677–92. https://doi.org/10.1016/j.energy.2017.01.143.

34.

Faber

Benjamin

(2014). “Trade Integration, Market Size, and Industrialization: Evidence from China’s National Trunk Highway System.” The Review of Economics Studies 81(3): 1046–70. https://doi.org/10.1093/restud/rdu010.

35.

Feng

Chao

Wang

Miao

(2018). “Analysis of Energy Efficiency in China’s Transportation Sector.” Renewable and Sustainable Energy Reviews 94: 565–75. https://doi.org/10.1016/j.rser.2018.06.037.

36.

Fisher-Vanden

Karen

Jefferson

Gary

Jingkui

Jianyi

(2006). “Technology Development and Energy Productivity in China.” Energy Economics 28(5–6): 690–705. https://doi.org/10.1016/j.eneco.2006.05.006.

37.

Gao

Yanyan

Zheng

Jianghuai

(2020). “The Impact of High-Speed Rail on Innovation: An Empirical Test of the Companion Innovation Hypothesis of Transportation Improvement with China’s Manufacturing Firms.” World Development 127: 104838. https://doi.org/10.1016/j.worlddev.2019.104838.

38.

Greene

William

(2005). “Fixed and Random Effects in Stochastic Frontier Model.” Journal of Productivity Analysis 23: 7–32. https://doi.org/10.1007/s11123-004-8545-1.

39.

Hanley

Douglas

Jiancheng

Mingqin

(2022). “High-Speed Railways and Collaborative Innovation.” Regional Science and Urban Economics 93: 103717. https://doi.org/10.1016/j.regsciurbeco.2021.103717.

40.

Guojun

Fan

Maoyong

Zhou

Maigeng

(2016). “The Effects of Air Pollution on Mortality in China: Evidence from the 2008 Beijing Olympic Games.” Journal of Environmental Economics and Management 79: 18–39. https://doi.org/10.1016/j.jeem.2016.04.004.

41.

Henderson

Vernon

Storeygard

Adam

Weil

David

(2011). “A Bright Idea for Measuring Economic Growth.” American Economic Review 101(3): 194–99. https://doi.org/10.1257/aer.101.3.194.

42.

Hiramatsu

Tomoru

(2018). “Job and Population Location Choices and Economic Scale as Effects of High Speed Rail: Simulation Analysis of Shinkansen in Kyushu, Japan.” Research in Transportation Economics 72: 15–26. https://doi.org/10.1016/j.retrec.2018.06.007.

43.

Huang

Jianhuan

Hua

Yue

(2018). “Eco-Efficiency Convergence and Green Urban Growth in China.” International Regional Science Review 42(3–4): 307–34. https://doi.org/10.1177/0160017618790032.

44.

Huang

Jianhuan

Yantuan

Chunbo

(2018). “Energy Efficiency Convergence in China: Catch-Up, Lock-In and Regulatory Uniformity.” Environmental and Resource Economics 70(1): 107–30. https://doi.org/10.1007/s10640-017-0112-0.

45.

Janic

Milan

(2011). “Assessing Some Social and Environmental Effects of Transforming an Airport into a Real Multimodal Transport Node.” Transportation Research Part D: Transport and Environment 16(2): 137–49. https://doi.org/10.1016/j.trd.2010.10.002.

46.

Jia

Ruining

Shao

Shuai

Yang

Lili

(2021). “High-Speed Rail and CO2 Emissions in Urban China: A Spatial Difference-in-Differences Approach.” Energy Economics 99: 105271. https://doi.org/10.1016/j.eneco.2021.105271.

47.

Jia

Shanming

Zhou

Chunyu

Qin

Chenglin

(2017). “No Difference in Effect of High-Speed Rail on Regional Economic Growth Based on Match Effect Perspective?” Transportation Research Part A: Policy and Practice 106: 144–57. https://doi.org/10.1016/j.tra.2017.08.011.

48.

Jin

Mengjie

Lin

Kun-Chin

Shi

Wenming

Lee

Paul

Kevin

(2020). “Impacts of High-Speed Railways on Economic Growth and Disparity in China.” Transportation Research Part A: Policy and Practice 138: 158–71. https://doi.org/10.1016/j.tra.2020.05.013.

49.

Xiao

Chen

Haiqiang

Hong

Yongmiao

Hsiao

Cheng

(2017). “Do China’s High-Speed-Rail Projects Promote Local Economy? New Evidence from a Panel Data Approach.” China Economic Review 44: 203–26. https://doi.org/10.1016/j.chieco.2017.02.008.

50.

Kim

Hyojin

Sultana

Selima

Weber

Joe

(2018). “A Geographic Assessment of the Economic Development Impact of Korean High-Speed Rail Stations.” Transport Policy 66: 127–37. https://doi.org/10.1016/j.tranpol.2018.02.008.

51.

Koenker

Roger

Bassett

Gilbert

(1978). “Regression Quantiles.” Econometrica 46(1): 33–50. https://doi.org/10.2307/1913643.

52.

Kong

Gaowen

Mianmian

Guo

Yuemei

(2021). “Political Promotion Events and Energy Conservation Decisions: Evidence from China.” Energy Economics 95: 105135. https://doi.org/10.1016/j.eneco.2021.105135.

53.

Kou

Zonglai

Liu

Xueyue

(2017). “FIND Report on City and Industrial Innovation in China (2017).” Fudan Institute of Industrial Development, School of Economics, Fudan University.

54.

La Ferrara

Eliana

Chong

Alberto

Duryea

Suzanne

(2012). “Soap Operas and Fertility: Evidence from Brazil.” American Economic Journal: Applied Economics 4(4): 1–31. https://doi.org/10.1257/app.4.4.1.

55.

Haipeng

Luo

Zhang

Xiong

Zhang

Junbiao

(2021). “Dynamic Change of Agricultural Energy Efficiency and Its Influencing Factors in China.” Chinese Journal of Population, Resources and Environment 19(4): 311–20. https://doi.org/10.1016/j.cjpre.2022.01.004.

56.

Hongbin

Zhou

Li-An

(2005). “Political Turnover and Economic Performance: The Incentive Role of Personnel Control in China.” Journal of Public Economics 89: 1743–62. https://doi.org/10.1016/j.jpubeco.2004.06.009.

57.

Xijing

Huang

Rongrong

Zhang

Yipei

(2016). “Exploring the Impact of High Speed Railways on the Spatial Redistribution of Economic Activities—Yangtze River Delta Urban Agglomeration as a Case Study.” Journal of Transport Geography 57: 194–206. https://doi.org/10.1016/j.jtrangeo.2016.10.011.

58.

Xiaolong

Zongfa

Zhao

Xingchen

(2020a). “Economic Effect and Its Disparity of High Speed Rail in China: A Study of Mechanism Based on Synthesis Control Method.” Transport Policy 99: 262–74. https://doi.org/10.1016/j.tranpol.2020.09.003.

59.

Yan

Chen

Zhenhua

Wang

Peng

(2020b). “Impact of High-Speed Rail on Urban Economic Efficiency in China.” Transport Policy 97: 220–31. https://doi.org/10.1016/j.tranpol.2020.08.001.

60.

Zhigang

Hangtian

(2018). “High-Speed Railroads and Economic Geography: Evidence from Japan.” Journal of Regional Science 58(4): 705–27. https://doi.org/10.1111/jors.12384.

61.

Lin

Boqiang

Chen

(2019). “Will Economic Infrastructure Development Affect the Energy Intensity of China’s Manufacturing Industry?” Energy Policy 132: 122–31. https://doi.org/10.1016/j.enpol.2019.05.028.

62.

Lin

Yatang

Qin

Jing

Mandi

(2021a). “Impact of High-Speed Rail on Road Traffic and Greenhouse Gas Emissions.” Nature Climate Change 11: 952–57. https://doi.org/10.1038/s41558-021-01190-8.

63.

Lin

Yatang

Qin

Xie

Zhuan

(2021b). “Does Foreign Technology Transfer Spur Domestic Innovation? Evidence from the High-Speed Rail Sector in China.” Journal of Comparative Economics 49(1): 212–29. https://doi.org/10.1016/j.jce.2020.08.004.

64.

Liu

Liwen

Zhang

Ming

(2018). “High-Speed Rail Impacts on Travel Times, Accessibility, and Economic Productivity: A Benchmarking Analysis in City-Cluster Regions of China.” Journal of Transport Geography 73: 25–40. https://doi.org/10.1016/j.jtrangeo.2018.09.013.

65.

Liu

Qing

(2015). “Firm Investment and Exporting: Evidence from China’s Value-Added Tax Reform.” Journal of International Economics 97(2): 392–403. https://doi.org/10.1016/j.jinteco.2015.07.003.

66.

Liya

Niu

Dongxiao

Sun

Weizeng

(2021). “Transportation Infrastructure and Entrepreneurship: Evidence from High-Speed Railway in China.” China Economic Review 65: 101577. https://doi.org/10.1016/j.chieco.2020.101577.

67.

Miao

Chenglin

Fang

Debin

Sun

Liyan

Luo

Qiaoling

(2017). “Natural Resources Utilization Efficiency Under the Influence of Green Technological Innovation.” Resources, Conservation and Recycling 126: 153–61. https://doi.org/10.1016/j.resconrec.2017.07.019.

68.

Michaels

Guy

(2008). “The Effect of Trade on the Demand for Skill: Evidence from the Interstate Highway System.” The Review of Economics and Statistics 90(4): 683–701. https://doi.org/10.1162/rest.90.4.683.

69.

Michalopoulos

Stelios

Papaioannou

Elias

(2018). “Spatial Patterns of Development: A Meso Approach.” Annual Review of Economics 10: 383–410. https://doi.org/10.1146/annurev-economics-080217-053355.

70.

Murillo-Zamorano

Luis R.

(2005). “The Role of Energy in Productivity Growth: A Controversial Issue?” The Energy Journal 26(2): 69–88. https://doi.org/10.5547/ISSN0195-6574-EJ-Vol26-No2-4.

71.

Won-Yong

Chang

Young Kyun

Kim

Tae-Yeol

(2018). “Complementary or Substitutive Effects? Corporate Governance Mechanisms and Corporate Social Responsibility.” Journal of Management 44(7): 2716–39. doi:10.1177/0149206316653804.

72.

Opsahl

Tore

Agneessens

Filip

Skvoretz

John

(2010). “Node Centrality in Weighted Networks: Generalizing Degree and Shortest Paths.” Social Networks 32(3): 245–51. https://doi.org/10.1016/j.socnet.2010.03.006.

73.

Patterson

Murray G.

(1996). “What Is Energy Efficiency? Concepts, Indicators and Methodological Issues.” Energy Policy 24(5): 377–90. https://doi.org/10.1016/0301-4215(96)00017-1.

74.

Puga

Diego

(2002). “European Regional Policies in Light of Recent Location Theories.” Journal of Economic Geography 2(4): 373–406. https://doi.org/10.1093/jeg/2.4.373.

75.

Qin

(2017). “‘No County Left Behind?’ The Distributional Impact of High-Speed Rail Upgrades in China.” Journal of Economic Geography 17(3): 489–520. https://doi.org/10.1093/jeg/lbw013.

76.

Redding

Stephen

Turner

Matthew

(2015). “Transportation Costs and the Spatial Organization of Economic Activity.” Handbook of Urban and Regional Economics 5: 1339-98. https://doi.org/10.1016/B978-0-444-59531-7.00020-X.

77.

Rodríguez-Pose

Andrés

Fratesi

Ugo

(2004). “Between Development and Social Policies: The Impact of European Structural Funds in Objective 1 regions.” Regional Studies 38(1): 97–113. https://doi.org/10.1080/00343400310001632226.

78.

Rosenbaum

Paul

Rubin

Donald

(1983). “The Central Role of the Propensity Score in Observational Studies for Causal Effects.” Biometrika 70(1): 41–55. https://doi.org/10.1093/biomet/70.1.41.

79.

Rosenbaum

Paul

Rubin

Donald

(1985). “Constructing a Control Group Using Multivariate Matched Sampling Methods that Incorporate the Propensity Score.” The American Statistician 39(1): 33–8. https://doi.org/10.1080/00031305.1985.10479383.

80.

Ryder

Andrew

(2012). “High Speed Rail.” Journal of Transport Geography 22: 303–5. https://doi.org/10.1016/j.jtrangeo.2012.03.004.

81.

Sadorsky

Perry

(2013). “Do Urbanization and Industrialization Affect Energy Intensity in Developing Countries?” Energy Economics 37: 52–9. https://doi.org/10.1016/j.eneco.2013.01.009.

82.

Sahu

Santosh Kumar

Sharma

Himani

(2016). “Productivity, Energy Intensity and Output: A Unit Level Analysis of the Indian Manufacturing Sector.” Journal of Quantitative Economics 14(2): 283–300. https://doi.org/10.1007/s40953-016-0034-7.

83.

Sánchez-Mateos

Héctor Martínez

Givoni

Moshe

(2012). “The Accessibility Impact of a New High-Speed Rail Line in the UK—A Preliminary Analysis of Winners and Losers.” Journal of Transport Geography 25: 105–14. https://doi.org/10.1016/j.jtrangeo.2011.09.004.

84.

Shao

Shuai

Tian

Zhihua

Yang

Lili

(2017). “High Speed Rail and Urban Service Industry Agglomeration: Evidence from China’s Yangtze River Delta Region.” Journal of Transport Geography 64: 174–83. https://doi.org/10.1016/j.jtrangeo.2017.08.019.

85.

Shao

Shuai

Yang

Zhenbing

Yang

Lili

Shuang

(2019). “Can China’s Energy Intensity Constraint Policy Promote Total Factor Energy Efficiency? Evidence from the Industrial Sector.” The Energy Journal 40(4): 101–28. https://doi.org/10.5547/01956574.40.4.ssha.

86.

Sun

Xinyu

Yan

Sen

Liu

Tao

Jianhan

(2020a). “High-Speed Rail Development and Urban Environmental Efficiency in China: A City-Level Examination.” Transportation Research Part D: Transport and Environment 86: 102456. https://doi.org/10.1016/j.trd.2020.102456.

87.

Sun

Yuhuan

Juntao

Wang

Shuhong

(2020b). “Environmental Regulations, Enterprise Productivity, and Green Technological Progress: Large-Scale Data Analysis in China.” Annals of Operations Research 290: 369–84. https://doi.org/10.1007/s10479-019-03249-4.

88.

Song

Malin

Zhang

Guijun

Zeng

Wuxiong

Liu

Jianhua

Fang

Kuangnan

(2016). “Railway Transportation and Environmental Efficiency in China.” Transportation Research Part D: Transport and Environment 48: 488–98. https://doi.org/10.1016/j.trd.2015.07.003.

89.

Stern

David

(2012). “Modeling International Trends in Energy Efficiency.” Energy Economics 34(6): 2200–08. https://doi.org/10.1016/j.eneco.2012.03.009.

90.

Tanaka

Kenta

Managi

Shunsuke

(2013). “Measuring Productivity Gains from Deregulation of the Japanese Urban Gas Industry.” The Energy Journal 34(4), 181–98. https://doi.org/10.5547/01956574.34.4.9.

91.

Tobler

Waldo

(1970). “A Computer Movie Simulating Urban Growth in the Detroit Region.” Economic Geography 46: 234–40. https://doi.org/10.2307/143141.

92.

Vickerman

Roger

(2018). “Can High-Speed Rail Have a Transformative Effect on the Economy?” Transport Policy 62: 31–7. https://doi.org/10.1016/j.tranpol.2017.03.008.

93.

Wang

Yiming

Wei

(2016). “Sources of Energy Productivity Change in China During 1997–2012: A Decomposition Analysis Based on the Luenberger Productivity Indicator.” Energy Economics 54: 50–9. https://doi.org/10.1016/j.eneco.2015.11.013.

94.

Wee

Bert Van

Janse

Pieter

van Den Brink

Robert

(2005). “Comparing Energy Use and Environmental Performance of Land Transport Modes.” Transport Reviews 25(1): 3–24. https://doi.org/10.1080/014416410001676861.

95.

Fei

Fan

Liwen

Zhou

Peng

Zhou

Dequn

(2012). “Industrial Energy Efficiency with CO2 Emissions in China: A Nonparametric Analysis.” Energy Policy 49: 164–72. https://doi.org/10.1016/j.enpol.2012.05.035.

96.

Rongxin

Lin

Boqiang

(2021). “Does Industrial Agglomeration Improve Effective Energy Service: An Empirical Study of China’s Iron and Steel Industry.” Applied Energy 295: 117066. https://doi.org/10.1016/j.apenergy.2021.117066.

97.

Yanrui

(2012). “Energy Intensity and Its Determinants in China’s Regional Economies.” Energy Policy 41: 703–11. https://doi.org/10.1016/j.enpol.2011.11.034.

98.

Yang

Zhenbing

Shao

Shuai

Yang

Lili

Miao

Zhuang

(2018). “Improvement Pathway of Energy Consumption Structure in China’s Industrial Sector: From the Perspective of Directed Technical Change.” Energy Economics 72: 166–76. https://doi.org/10.1016/j.eneco.2018.04.003.

99.

Yao

Shujie

Zhang

Fan

Wang

Feng

Jinghua

(2019). “Regional Economic Growth and the Role of High-Speed Rail in China.” Applied Economics 51(32): 3465–79. https://doi.org/10.1080/00036846.2019.1581910.

100.

Zhang

Fangni

Graham

Daniel J.

Wong

Mark Siu Chun

(2018). “Quantifying the Substitutability and Complementarity Between High-Speed Rail and Air Transport.” Transportation Research Part A: Policy and Practice 118: 191–215. https://doi.org/10.1016/j.tra.2018.08.004.

101.

Zheng

Siqi

Kahn

Matthew E.

(2013). “China’s Bullet Trains Facilitate Market Integration and Mitigate the Cost of Megacity Growth.” Proceedings of the National Academy of Sciences 110(14): E1248–53. https://doi.org/10.1073/pnas.1209247110.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.10 MB