<<

The Route of Development in intra-regional Income Equality via High-Speed Rail: Evidence from

Wenjing YU, China Academy for Rural Development, University [email protected] Yansang YAO, China Academy for Rural Development, Zhejiang University [email protected]

Selected Paper prepared for presentation at the 2019 Agricultural & Applied Economics Association Annual Meeting, Atlanta, GA, July 21 – July 23

Copyright 2019 by [Wenjing Yu, Yansang Yao]. All rights reserved. Readers may make verbatim copies of this document for non-commercial purposes by any means, provided that this copyright notice appears on all such copies. The Route of Development in intra-regional Income Equality via High-Speed Rail: Evidence from China

Abstract This paper mainly studies how the bullet trains, a new generation of vehicles, are associated with income inequality in China for the years between 2008 and 2018. Gini coefficients are used to measure the income inequality from county level, and within urban and rural areas of China. A staggered Difference-in-Difference (DID) approach is taken to identify the causal effect of high-speed rail on intra-regional income inequality. Factors other than transportation are also considered in our regression model, including a few social variables and major economic indicators. It is found that the Gini coefficient of reginal economy would rise by 0.0327 in average when new high-speed railway stations are opened, which means that the intra-regional income inequality is being exacerbated. We also find that the treatment drawn by high-speed rail system is not uniform in different regions, and the greatest impact was set on the western region.

Key words: high-speed rail, income inequality, county level

Introduction With rapid economic growth following the opening-up policy, economic disparity has become a major challenge in China (Jian et al., 1996). Income inequality problem is not only reflected on the income gap among different regions, but also manifested in the income inequality within certain areas, which has a more direct influence on the harmony and stability of local society. Most literatures(Kaiyuen, 1998; Adler & Schmid,2013; Hui, 2008) summarize the causes of income inequality into two aspects, namely the endowments differences and the frequency of production factor flows. The former mainly refers to human capital, such as education level and health conditions; the latter is closely related to the transportation network. Transportation system and the economy are unavoidably linked. Traffic itself set no value, but it is considered as a way to bridge the spatial gap between regions, which can respond to the derivative needs of human activities. Beginning with “New Economic Geography” (Krugman, 1991), it has been proved that under the background of the imperfectly competitive world, changes in transportation costs and accessibility can generate a profound impact on the location and agglomeration of activities. One proposed solution to the issue of income inequality is putting investments in public transportation. Most governments, especially in developing countries, prefer to invest in transportation infrastructure to stimulate the economy, and the bullet train, one of the most advanced ground transportation modes, has commanded attention in recent years. (Amos et al., 2013; Ashish et al., 2013; Ke et al., 2017). In current China, the development of traditional railways is far from meeting the needs of integrated transportation system for other industries. Rather than upgrading traditional railway station, Chinese government chose to brand tens of thousands of new high-speed rail stations, especially in suburban areas, with the purpose to stimulate new towns’ development and accelerate regional urbanization. Most researchers (Zhang & Zou, 2012; Zheng & Kahn, 2013) focused on the impact of HSR on economic productivity and competitiveness, arguing that economies can benefit from lower generalized costs of transport. A few scholars (Chen & Haynes, 2017; Yang et al., 2018) pointed out that the HSR projects have also shown the positive effects on regional imbalance during the past few decades, supporting that HSR investment could have a transformative impact on the economy as a whole, rather than just on directly affected local areas. However, little attention has been paid to income inequity within regions and the economic theory is unclear on how investments in transportation system should affect income inequality. In principle, the opening of high-speed rail can accelerate the speed of inter-regional elements flowing, especially of the labor resources. As a result, the degree of market information asymmetry between regions is weakened, as well as the structure of production factors within regions changing, which is presented as the structure reorganization of labor force and three main industries. The income distribution within regions will be rewritten due to the resource reallocation. Empirical analysis of the effect of public transportation investment on income inequality has been mixed.Yan Li and Maria N. DaCosta (2013) studies the relationship between various types of transportation modes and income inequality in China for the years between 1978 and 2007, finding that most transportation modes are negatively associated with income inequality in urban areas while the coefficients are positive for rural areas. Each of the previous results were established using cross-city data, and examine the impart of high-speed rail system on economic growth or inter-regional inequality. However, there is growing evidence that public transportation does affect income equality of intra-regions. Investing in high-speed rail is on the front line of action to revitalize the railways. The ultimate objective is to create new generation of passenger transport for the sake of reducing congestion, accidents and environmental externalities. High-speed rail investment is seen as a feasible measure, with the aim of drawing the benefit from railways which is associated with lower total travel time, higher comfort and reliability, a reduction in the probability of accident and, in some cases, the release of extra capacity which helps to enlarge the frequency of information exchange among regions. Last but not least, it has been argued that high-speed rail investment weakens the differences of absolute economic advantages among regions and boosts regional development. In fact,the main element transported by high-speed railway is human beings and the impact of high-speed rail stations set on the cross-regional mobility depend on the characteristics of target passengers, because the ticket price of bullet trains is several times higher than that of ordinary transport modes, and residents' demand for high-speed rail is related to their economic affordability. Furthermore, each station is equipped with a complete public system, including commercial facilities and social Infrastructure, which increase the employment opportunities for labors within a certain radius. No matter in which region, individual’s income and labor market structure are both probably shocked by high-speed rail project. Our current work provides evidence adding to the body of empirical literature examining the effects of high-speed rail investment on reginal development. As income inequality is a serious issue in China, the primary question in this paper is to examine how high-speed rail stations openings in China impact intra-regional income inequality, which would be measured by Gini index. In principle, one might calculate the effect of public transportation on income inequality by regressing the Gini coefficient, an authoritative proxy variable of income inequality based on the Lorenz curve, which plots the share of population against the share of income received. However, this approach is likely to be flawed because of other confounding factors driving both variables. For example, economic structure is closely related to the regional quadratic assignment strategy, such as tax policy formulated by government. To be specific, we regress the degree of intra-regional income inequality on a dummy variable for new high-speed railway stations’ openings in a difference-in-difference framework where time is the running variable. Our key identifying assumption is that the intra-regional income disparity of full samples keeps parallel trend which means all other factors influencing Gini index are smooth except the new rail stations themselves. Changes to other contributors to income equality, such as the local population and economy, do not threaten our identifying assumption as long as they evolve parallelly among all the counties. As a result, the treatment effect in our outcomes of interest at the time of the high-speed stations’ openings can be attributed to the change in income disparity of local residents. We compare the Gini index per month—between counties that made their traffic system more generous and inclusive and other counties that did not—before and after the opening of high- speed rail stations (difference-in-difference) to estimate the difference made by changes in the high-speed rail system. We find that high-speed rail stations’ openings cause large ups in intra-reginal income inequality. In our primary specification, the major finding is the significant impact of the high- speed rail project on income inequality within regions, as the Gini index changed by 0.0327, which means intra-regional income disparity has generally increased since the development of high-speed rail. We test whether our findings are robust to alternative explanations. The results are satisfactory since the coefficients are robust to various specifications, including different control variable sets, different forms of error variance, and different alternative levels of fixed effect. Other robust checks are also considered, for example, replacing the independent variable with the proportion of high-income people or dropping the sample of counties equipped with departure or terminal stations. The debate on the merit of high-speed rail investment has received increasing attention worldwide in recent years, but it is still uncertain. Our study contributes to a growing literature on transportation investment in developing countries, estimating the impact of large transportation projects on income disparity, as well as the economic and social value of the railway system. The remainder of the paper is organized as follows. Section 2 provides an overview of China’s bullet railway development. Section 3 explains using a simple theoretical model how income inequality might respond to new high-speed rail stations’ opening. Section 4 lays out the empirical framework and data. Section 5 presents our findings and discusses robustness checks for these results. Section 6 concludes.

Background Investing in high-speed rail is a central planning decision. The Chinese government decides the introduction of a new rail technology which allows trains running at a speed of 300-350 kilometers per hour (although the average commercial speed is substantially below the technically feasible speed). Since the early 2000s, an ambitious strategy, the ‘‘Mid- and Long- Term Railway Network Plan”, has been carried out by Chinese government to develop national railway infrastructure system, outlining the detailed objectives of rail network expansion, rolling stock promotion and rail facilities improvement. One of the initiative features was to develop a national passenger high-speed rail system with a total track length of over 16,000 km by 2020. As shown in Fig.1, the high-speed rail stations are geographically dispersed. This railway technology is particularly popular in the China. High-Speed Rail investment projects are financially supported by the Chinese government. Revitalizing the railways is the new motto in China’s transport policy, meaning both introducing competition in the railway industry and giving priority to public investment in the rail network. For the process of Railway System expansion in China is tightly managed, the opening year for railway lines and stations can be perfectly controlled. According to the requirements of the National Development and Reform Commission(NDRC) in China, “the trans-provincial (regions and municipalities) railroads or those of 100 kilometers or longer are subject to the approval of the investment authority of the State Council while others are subject to the approvals of the industry authorities of the State Council or provincial government investment departments according to the affiliation of the railroads”. The application should be submitted in advance by nearly 5 years to NDRC, if the local officials intended to update the traffic network of their county. Once the NDRC publishes their approval notification on the official website, local government, according to the permission, could add the construction of the new railway station to its local development plan. Generally, large construction projects are subject to unexpected obstacles, but local governments prefer to dedicate themselves to opening new stations and expanding new train lines on schedule because these development plans are essential elements in evaluations of government officials. Our staggered difference-in-difference (DID) strategy resolves these problems by leveraging the causal effect in reginal economy generated by new high-speed rail stations openings. China has engaged in an unprecedented expansion of its railway system since 2003, but the high- speed rail stations have not been put into use until after 2008. These data are summarized in Table 1, the number of new high-speed railway stations opened in China each year increased rapidly in the first three years and reached its first peak in 2010, accompanied by a small decline in next year. Then it kept increasing during the 12th Five-Year Plan period, especially in 2015. In addition, as shown in the Fig. 2 and 3, the investment of high-speed railway stations in the three regions of China is not simultaneous. On the early stage, the construction of high-speed railway stations mainly concentrated in the eastern region. In recent years, the investment in the middle and western regions began to increase. Considering the availability of other economic variables, we examine how the separate railway stations openings affect regional income inequality between the period from 2008 to 2015. Andthe heterogeneity of the three regions is also analyzed.

Table 1: The Number Of China’s New High-Speed Rail Stations Put Into Operation Per Year 2003 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 9 1 16 53 81 48 51 79 138 153 106 52 2 Railway map of People's Republic of China Mohe

Colored lines showing CRH and other Mangui Mordaoga Jiagedaqi Fuyuan high speed rail services Yi tuli he Hailar Yak es hi Last update: 2018-01-15 Beian Qianjin Nancha Dongfanghong Tarqi Linkou Beitun

Ala Shankou Da'an Jinghe Taipingchuan Khorgos Urümqui Siping Tür pan Chaoyang Ji'an Akshi Bayan Obo Benhong Ceke Changdian Liugou Yi ngkou Linhe Huairou Ejin Jilantai Shenmu Jiayuguan Yul in Wei hai Yi nchuan Wuwei Rongcheng Gantang Lüliang Lines capable for Wei fang Zhongchuan Yan'an speed above 300 km/h Taian Newly built lines capable for 200-299 km/h Xinyi Upgraded lines and other Huashan Xi'an lines with CRH service Huai'an Haian Conventional lines with Guangy uan Nanyang no CRH service Nagqu Dujiangyan Wanz hou Yi chang Enshi Huangshan Suining Quz hou Guang'an Lichuan Taizhou Lhasa Emeishan Shimen Yi ngtan Wenz hou Tongr en Yi bin Ji'an Kaiyang Yongtai Ganz hou Dongchuan Dali Longchuan Baoxiu Guangz hou Kaiyuan Wuz hou Hekou Yul in Lianjiang Xuwen

Dongfang Wanni ng

Sanya

Fig. 1. Map of Railway Lines in China

160 140 120 100 80 60 40 20 0 2008 2009 2010 2011 2012 2013 2014 2015

Eastern Middle Western

Fig. 2 Number of New High-Speed Rail Stations Put Into Operation Within Three Regions

300

250 200

150 100 50 0 2008 2009 2010 2011 2012 2013 2014 2015

Eastern Middle Western

Fig. 3 Total Number of High-Speed Rail Stations Put Into Operation Within Three Regions Empirical framework Firstly, we provide a brief description about how a new high-speed rail station is likely to affect regional economic disparity. Consider a representative worker deciding whether to take outgoing work or not. This worker's decision is likely to depend on several factors such as the outgoing distance, travel time and the monetary cost of the trip. The opening of high-speed rail, the latest and efficient transportation mode, creates the impact on workers' decision-making of whether go out or not, which even draw the influence on individuals’ employment opportunities and income. Now consider the opening of a new high-speed rail station. Suppose that the passenger can always take the same kind of transportation as he did before the nearest railway station opening, a new high-speed rail station will expand the set of options for the passenger and will either leave the traveler meaningless or better off. Considering the cost of time, monetary expense and convenience, it is doubtless more likely for workers to choose high-speed rail instead of other forms of transportation when they take interregional travel. However, this result is conditional on the traveler taking a given trip, and the schedule to work out may change in general equilibrium if the costs of travel decrease with the new high- speed station opening, which is an empirical matter to investigate.

Difference-in-difference approach In principle, one might estimate the effect of increased railway traffic on income distribution by performing Ordinary Least Squares regressions with Gini coefficient as the dependent variable and opening status of high-speed railway as the independent variable. However, this approach is likely to be biased because there are often observable and unobservable characteristics, such as regional policies. To address these endogeneity concerns, we estimate as our primary specification based on ordinary least squares model. In order to identify the effect of high-speed rail stations on income disparity within regions, we implement our empirical strategy as leveraging the causal effect in reginal economy generated by new high- speed rail stations openings Specifically, we utilize the following staggered Difference-in-Difference (DID) design:

!"# = %& + %()*+"# + %,-"# + ." + /# + 0"# (1) Where !"# denotes the intra-regional income disparity, in other words, the inequality within county 1 in year 2. )*+"# is a dummy variable indicating whether the high-speed rail station closed to county 1 is open or not in year 2. The variable /# is a running variable representing the time trend: the difference between the station opening in year 2 and the variable

." represents the fixed effect of county 1. And 0"# is the corresponding regression error. We also include -"#, a vector of other control variables that may affect intra-regional income inequity for county 2 in year 1. In our regressions, we include eight types of additional controls: GDP, GDP per capita, the proportion of primary industry, the proportion of secondary industry, fiscal revenue and expenditure, dummy for whether there is an airport, total vehicle passenger volume. We reason that the first four control variables can impact patterns of economy. The financial indicators are used to reflect local government strategies on income redistribution. The last two variables are designed to exclude the interference from other transportation modes,mainly for vehicles that can achieve cross-regional communication.

The variable of interest is %(, the local average treatment effect of the new high-speed rail stations opening on intra-regional income inequality, the size and direction of which depends on depending on how regional economy responds to new transportation modes. The key assumption, maintained throughout the paper, is that the opening of high-speed railway stations may impact the local industrial structure and employment structure, which could even affect the income level of local labor force. The flow of production factors will have a significant influence on income inequality in a certain region, especially the transfer of high- income or low-income groups. The direction of treatment impact depends on whether the regional production factor flows in or out. We implement this approach using annual observations before and after high-speed rail stations openings. Considering that China's high-speed rail system has been operating since 2008, and Chinese government has built many high-speed rail stations in different areas in the next five years, this paper estimates the data collected during the period from 2008 to 2015.

Threats to identification Our identifying assumption is that the income distribution structure within the region only shows a stable trend when the new high-speed railway station is not opened. This assumption is reasonable as long as there is no significant change in the socio-economic environment within certain regions when the new high-speed railway is opened. The control variables and fictitious variable factors in our regressions account for eliminating interference and controlling the fixed effect. The high-speed railway stations are opening up one after another, which pose a threat to our identification strategy. If the high-speed rails openings occur at the same time as events that cause changes in local economy, it might destroy the original steady state of reginal income distribution. For example, if government officials strategically raise the income tax of high- income people in order to ease the income gap among local residents, there will be errors in our estimation of %(. The second concern is the presence of alternative policies that came into effect during our sample period. The most prominent change in these policies is the income redistribution policy, taking the tax revenue as an example, and the government subsidies level in the three regions is also quite different. In another word, each region has a fixed feature at each time point. The last concern is that the construction activity associated with opening new high-speed rail stations creates regional economic activities. For example, the update of the high-speed rail system will inevitably be accompanied by the improvement of supporting facilities, and the basic living facilities such as bus, subway, catering, accommodation will gradually be improved, which will also have an impact on residents' income. Such basis in our regression would be manifested as overestimation or underestimation of coefficients.

Description of data We are trying to investigate the relationship between transportation and income inequality, especially for the intra-regional income distribution. Thus, the estimation of income inequality is one of the key issues to be considered, which is often complex and controversial. Given the various approaches, each measurement may yield slightly different findings, providing a slightly different perspective of income distribution evolution in China (Gustafsson et al., 2008; Benjamin et al., 2008). In this paper, we use Gini coefficient as the proxy variable of income inequality,which is one of the first few contributions to this field. Unfortunately, the government seldom discloses the data of income distribution at the county level, making us have to find a new method to converge framework. We draw two micro-survey databases to calculate the Gini index, namely the China Family Panel Studies (CFPS) and the Chinese Household Income Project survey (CHIPs). Because the information provided by these two databases can be traced back on the county level, as well as the respondents are randomly sampled, which is in line with the calculation requirements. We finally integrated Gini coefficient data of some counties in China from 2008 to 2015. In addition, the data of control variables such as GDP and finical index are obtained from the China Statistical Yearbook (2008–2015).

Table 2: Summary Year 2009 2011 2013 2015 Obs 160 158 157 155

Gini 0.4283 0.4867 0.2046 0.4535 (0.0692) (0.0804) (0.072) (0.1073)

HSR 0.0313 0.0949 0.1465 0.1871 (0.1745) (0.2941) (0.3547) (0.3913)

Population 6.0582 6.2402 6.2608 6.2183 [# unit =million person ] (4.7379) (4.8379) (4.8952) (4.8964)

GDP 287.3829 391.067 460.1228 501.4716 [# unit =billion yuan ] (445.8071) (574.6602) (656.131) (737.348)

Mean_GDP 31753.05 41315.37 48864.79 53154.45 [# unit = yuan ] (24123.65) (26322.67) (29649.6) (31961.68)

First_GDP 12.8267 16.8385 19.6375 21.1802 [# unit =billion yuan ] (9.3379) (12.3208) (14.7451) (16.3328)

Second_GDP 125.8294 178.2259 196.771 191.8507 [# unit =billion yuan ] (177.0326) (236.5157) (245.635) 240.3656

Fiscal Revenue 35.7654 52.2499 65.2743 76.7709 [# unit =billion yuan ] (75.9616) (103.7279) (124.638) (159.3632)

Fiscal Expenditure 47.7518 67.3322 82.9663 101.6737 [# unit =billion yuan ] (87.857) (116.8235) (135.559) (177.0727)

Airport 0.5313 0.5 0.5478 0.5742 (0.5006) (0.5016) (0.4993) (0.4961) Note: Our data interval is 2 years, because the databases we choose are surveyed every two years. Standard errors are listed in parentheses.

Table 3: Staggered Difference-in-Difference-Based Results. (1) (2) (3) (4) (5) (6) Gini Index OLS DID DID DID DID DID -0.0123 -0.0078 -0.0481** 0.0058 0.0001 0.0327*** HSR (0.0162) (0.0176) (0.0222) (0.0142) (0.0154) (0.0116)

County FE NO NO YES NO YES YES

Year FE NO NO NO YES YES YES

Control Var YES NO NO NO NO YES

(N=630) for all regressions

Notes: This table shows the regression results for the Staggered DID Eq.(1). The column(1) is the basic result of OLS and the column(2)-(6) is estimated with DID strategy. The difference lies in these regressions is whether to control the fixed effect or whether to add other control variables. Standard errors are listed in parentheses and clustered at province level. *** p<0.01, ** p<0.05, * p<0.1.

Table 4: Robust Results.

1 2 3 (4) 5

0.0327** 0.0303*** 0.0353*** 0.0215* 0.0330* HSR (0.0146) (0.0104) (0.0090) (0.0106) (0.0173)

County FE YES YES YES YES YES

Year FE YES YES NO YES YES

Pro#Year FE NO NO YES NO NO

Control Var YES YES YES YES YES

N 440 440 440 440 440

Notes: This table shows the results of robust check for the Staggered DID Eq.(1). The mainly difference lies in these regressions is the control variables. Standard errors are listed in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

Empirical results In this section, a staggered Difference-in-Difference (DID) approach is applied to explore the intra-regional income inequality in rural China, urban areas, the three-region classification as well as the whole country. Various explanatory factors are employed, including policy level, economic indicators, and particularly transportation factors.

Staggered Difference-in-Difference (DID) results We now estimate the treatment effect of new high-speed rail stations openings on the patterns of intra-income inequality in China using Eq. (1). The regression results illustrated in Table 3 say that the impact of using efficient transport mode on income distribution measured by Gini index, after controlling for other factors. Each entry in this table represents the result of a separate regression, with the dependent variable in the column headings and the functional form of the regression in the row headings. The coefficient reported in each cell is %(, the local average treatment effect of the high-speed rail station opening. In first regression (column 1), we use OLS to estimate the causal effect of high-speed rail station opening with control variables, including GDP, GDP per capita, the proportion of primary industry, the proportion of secondary industry, fiscal revenue and expenditure, dummy for whether there is an airport, total vehicle passenger volume. The results show that the opening of high-speed rail stations has no significant impact on income inequality within the region. In the last five regressions (Column 2-6), we implicate the staggered difference-in- difference method to modify the results by adopting different fixed effect control strategies. The result of column 2 and 3 are estimated by a new discriminant method without controlling other variables. The results show that the update of high-speed rail system does not put the significant impact on income inequality in its region. But the results are slightly different in regression 3. The value of %( sharply drops to -0.0481, indicating that the coefficient in regression 2 is underestimated. When county fixed effect is controlled, we find that high-speed rail station openings have a large and statistically significant impact on income disparity levels in China. That is to say, excluding the endowment characteristics of the county itself, the opening of high-speed railway stations in the region can effectively alleviate its intra-income inequality. In the last regression, we try to control the time-fixed effect, the results of which show that the time trend does exist. Comparing the numbers of column 3 with column 5, it was found that the significance of %( disappeared. On this basis, we further control the socio-economic factors and get the results of column 6. It is found that the interest coefficient turn to be significantly positive after correction, increasing to 0.0327, which means that the opening of high-speed railway stations will draw a significant shock to the regional income structure with the absolute income gap of residents be widen, leading the regional income inequality issue become more serious. In addition, we test the robustness of our results (Table 4) from diversified perspectives. Our first approach is to use robust setting instead of clustering, making more strict assumptions about the variance of residual terms. It is found that the coefficient %( that we are concerned with is stable. Then, we changed the variable information in the control variable set as replacing the economic structure ratio with the absolute output value of economic activities. The result is shown in the Table 4 that the direction of %( keeps significantly positive and the value decreases slightly. In our third regression, we add multiple-fixed effect factor, the interaction dummy variables of province and year. Fortunately, the work is consistent with the previous ones. We liberated the definition of the location of high-speed rail stations in the fourth case and count the variable depend on whether there are high-speed rail stations in city that county locates in. It is as expected that the treatment caused by new high-speed rail stations is still significantly positive. Considering that there may be lag in economic activities, the lag term of the high-speed railway station opening time is also used as an independent variable in our regression, the result of which is still in line with our expectations. Overall, our results are credible that the emergence of new high-speed rail stations will indeed exacerbate the inequity of intra-reginal income distribution.

Heterogeneity Analysis In order to deeply investigate the differences in the influence caused by high-speed rail station opening in regions on their Gini coefficient, we classify the samples into diversified regions. One classification method is to divide the samples into eastern, central and western regions according to geographical location and economic indicators. The other method to separate the data to confirm the sample’s administrative power. Specifically, the sample would be defined as a as long as its administrative power belongs to a city. Otherwise, the sample would be labelled as a county. Table 5 present the regression result of first method. Columns 1, 2 and 3 are estimated for eastern, central and Western samples, respectively. By comparison, we can clearly find that the impact of the high-speed rail station openings in the eastern region on the Gini index is slightly higher than the average. The influence in the western region is also significantly positive with large magnitude, while the coefficient of the middles samples is not significant. We believe that the economy in the eastern region is more developed and the demand for efficient modes of transportation is greater. The opening of high-speed rail station provides a convenient transportation option for the local residents, promoting the flow of people. In the past, traffic in the western region was very backward. With the improvement of the transportation system in recent years, especially the investment of high-speed rail system, the potential of local economic factors have been activated, and its economic structure has been affected much moredrastically than that of the eastern and middle regions. A large number of financial and technical resources have the possibility to transport into the western region. For example, the increasing number of tourists has revitalized the local tourism industry, which has greatly affected income inequality in the western region.

Table 5: Heterogeneity Result of Three Regions 1 2 3 Eastern Middle Western HSR 0.0384** -0.0106 0.2819*** (0.0125) (0.0413) (0.0361)

Individual FE Yes Yes Yes Year FE Yes Yes Yes Control variables yes yes yes Observation 288 184 158 Notes: This table shows the results of heterogeneity analysis with first classification method for the Staggered DID Eq.(1). The mainly difference lies in these regressions is the different sample. Standard errors are listed in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

Table 6 shows the estimated results obtained using the second classification method. Column (3) is the estimated by full sample. The sample selected for the first column are managed by the city, and those listed in second column are administrated by themselves. In addition, the variable HSR*County is also used to illustrate the difference between district and county. Comparing the first two columns, it is not difficult to find that the coefficient of district samples is much higher than that of county samples. The former is significance positive while the coefficient of county is very small and not significant. The possible reason is that district, as the center of city, has obvious advantages in economy, politics and infrastructure. When the high-speed railway is opened, human resources, capital and technology prefer to accelerate flowing between districts. For example, high-level talents prefer to work in districts, and high-tech industries are more likely to land in districts. In addition, the traffic construction of districts is generally considered earlier than county since it is generally closer to the political center. The operation of the high-speed rail in district will affect the county elements flowing simultaneously and county economy could enjoy the spillover effect, which weaken the impact of high-speed rail station openings in county.

Table 6: Heterogeneity Result of County and Districts (1) (2) (3) Power in City Power in County Total HSR 0.0457*** 0.0077 0.0407*** (0.0118) (0.0235) (0.0096)

HSR*County -0.0187 (0.0193)

Individual FE Yes Yes Yes Year FE Yes Yes Yes Control variables yes yes yes Observation 238 392 630 Notes: This table shows the results of heterogeneity analysis with second classification method for the Staggered DID Eq.(1). The mainly difference lies in these regressions is the different sample. Standard errors are listed in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

Conclusion High-speed rail has attracted an increasing attention worldwide in recent years given that inter-regional travel has been greatly improved both in terms of time efficient and onboard amenity. This paper has attempted to shed some light about how the high-speed rail openings in China affect intra-regional income inequality. The issue was investigated in three dimensions. First, a theoretical framework was established to evaluate the impact of high-speed rail on economic disparity. Second, we analyses the development of high-speed railway construction in China in the past ten years, following the design of a staggered difference-in-difference strategy to calculate the causal effect of high-speed railway station opening on regional income inequality. Third, the linkages between intra-regional income disparity and high-speed rail development was examined empirically. In our main estimation, we find that the Gini coefficient will rise by 0.0327 in average if a new high-speed railway station is opened in certain county, which means that the local income inequality will increase. This result is robust across a broad set of specifications and potential alternative explanations. We also find that the treatment drawn by high-speed rail system is not uniform in different regions. The impact of East and West regions is significant positive, and the coefficient estimated by the eastern region samples is slighter than that of the western region, while the impact of the central region is negligible. . Specifically, we provide contributions to the study of high-speed rail impacts in China with the purpose to settle the linkage between transportation and inequities issue. Our results are consistent with the growing evidence that there is significant social shock in China, differing from earlier studies in examining the effectiveness of HSR projects in intra-regional income distribution. However, China's high-speed rail is expanding, and the existing research excluded the cases of the upcoming openings, which still need to be studied in the future.

Appendix A: List of Provinces Opening High-speed Railway Stations In section 2, we mentioned that China's high-speed railway stations are opening in succession within three major regions. The specific information is as follows: Year Number1 Location2 2003 9 ,

2007 1 Zhejiang

2008 16 Beijing, Tianjin, ,

2009 53 Anhui, Shandong, Zhejiang, , , , , , 2010 81 Sichuan, Shandong, Guangdong, Zhejiang, Hubei, Fujian, Shanghai, , , , , Shannxi 2011 48 Sichuan, Tianjin, Anhui, Shandong, Guangdong, Jiangsu, Hebei, Shannxi, Jilin 2012 51 Beijing, Jilin, Anhui, Guangdong, Hebei, Henan, Hubei, Fujian, Liaoning, 2013 79 Sichuan, Tianjin, Guangdong, Jiangsu, Jiangxi, Hebei, Zhejiang, Hubei, Hunan, Fujian, Liaoning, Shannxi, Chongqing, 2014 138 Sichuan, Anhui, Shandong, Shanxi, Guangdong, Guangxi, Jiangxi, Henan, Zhejiang, Hubei, Hunan, Shannxi, , , , 2015 153 Jilin, Sichuan, Tianjin, Anhui, Guangdong, Guangxi, Jiangsu, Jiangxi, Hebei, Henan, Zhejiang, Hainan, Fujian, Guizhou, Liaoning, Chongqing, Heilongjiang 2016 106 Anhui, Shandong, Guangdong, Jiangsu, Henan, Hubei, Hunan, Guizhou, Chongqing, 2017 52 Sichuan, Anhui, Shandong, Guangdong, Jiangxi, Hebei, Zhejiang, Hubei, Hunan, Gansu, Shannxi, 2018 2 Shandong, Liaoning

1. The number of new high-speed railway stations per year 2. The provinces or regions where the new high-speed railway stations are located

References Adler M , Schmid K D . Factor Shares and Income Inequality. Empiral Evidence from Germany 2002 – 2008[J]. Schmollers Jahrbuch : Journal of Applied Social Science Studies / Zeitschrift für Wirtschafts- und Sozialwissenschaften, 2013, 133. Amos, P., Bullock, D., & Sondhi, J. (2013). High-speed rail: the fast track to economic development?. High Speed Rail. Ashish Verma, H S Sudhira, Sujaya Rathie, Robin King, Nibedita Dash. (2013). Sustainable urbanization using high speed rail (hsr) in karnataka, india. Research in Transportation Economics, 38(1), 67-77. Cheng Y S , Loo B P Y , Vickerman R . High-speed rail networks, economic integration and regional specialisation in China and Europe[J]. Travel Behaviour and Society, 2015, 2(1):1-14. Chen, Z., & Haynes, K. E. (2017). Impact of high-speed rail on regional economic disparity in china. Journal of Transport Geography, 65, 80-91. Hui L . Factor Decomposition of Rural Regional Income Inequality Changes in China[J]. Acta Geographica Sinica, 2008. ITF Round Tables Competitive Interaction between Airports, Airlines and High‐Speed Rail: (Complete Edition ‐ ISBN 9789282102466)[J]. Sourceoecd Transport, 2009, volume 2009:i- 212(213). Kaiyuen T . Factor Decomposition of Chinese Rural Income Inequality: New Methodology, Empirical Findings, and Policy Implications[J]. Journal of Comparative Economics, 1998, 26(3):502-528. Ke, X., Chen, H., Hong, Y., & Hsiao, C. .(2017). Do china\"s high-speed-rail projects promote local economy?—new evidence from a panel data approach. China Economic Review, 44, 203-226. Kuznets S . Economic Growth and Income Inequality[J]. American Economic Review, 1955, 45(1). Li Y , Dacosta M N . Transportation and income inequality in China: 1978–2007[J]. Transportation Research Part A: Policy and Practice, 2013, 55:56-71. Yang, J. , Yajun, B. , Zhang, Y. , Li, X. , & Quansheng, G. . (2018). Impact of accessibility on housing prices in dalian city of china based on a geographically weighted regression model. Chinese Geographical Science(6), 1-11. Zhang, Q., & Zou, H.. (2011). Regional Inequality in Contemporary China. China Economics and Management Academy, Central University of Finance and Economics. Zheng, S. , & Kahn, M. E. . (2013). Chinas bullet trains facilitate market integration and mitigate the cost of mega city growth. Science Foundation in China, 110(1), E1248-E1253.