Impact of high-speed railway on regional
economic development in China
Msc International Economics and Business
Master thesis
2018-2019
Abstract
China has its first line in 2005. Since then, it has experienced a major growth of HSR. Now it becomes the country with the longest HSR mileage and the highest transport density. Given the increasing amount of HSR and the continuing huge amount of investment, a natural question is can HSR promote economic growth as expected? If it can promote economic growth, what are channels for HSR to influence the economy? To answer these questions, I use panel data, from 2005 to 2016, to examine the impact of HSR on economic development. I also use steps provided by Wen et al (2014) to examine the mediation effect of FDI and knowledge spillover. The result shows that HSR has a positive impact on economic development generally. FDI and knowledge spillover have partial mediation effect between HSR and economic development.
Acknowledgements
Hereby, I would like to thank my thesis supervisor dr. D.H.M Akkermans sincerely for supporting me throughout the process of writing my master thesis. He is very patient and always gives me valuable feedback. When I started writing my thesis, I was not sure about my topic. He encouraged me and said that every idea is good. I can always get good inspirations and new challenges from the discussion. I really enjoyed the time we discussed together. I also would like to thank my all teachers in master program. Their lectures and tutorials gave me more ideas. Finally, I would like to thank my family members and friends. When I finished my draft, they read it and talked about their feelings to make my thesis easier to follow.
Table of Contents
1. Introduction ... 1
2. Literature review and hypothesis development ... 4
2.1 HSR ... 4
2.2 HSR and Economic Development ... 5
2.3 Knowledge Spillover ... 8
2.4 HSR, Knowledge Spillover and Economic Development ... 8
2.5 FDI ... 9
2.6 HSR, FDI and Economic Development ... 9
3. Data and Methodology ... 12
3.1 Sample ... 12
3.2 Data ... 13
3.3 Methodology ... 17
4. Empirical Test ... 21
4.1 HSR and regional economic development ... 21
4.2 Channels ... 23
5. Conclusion and Limitation ... 26
5.1 Conclusion and Recommendations ... 26
5.2 Limitation and Further Research ... 27
6. References ... 28
1. Introduction
Since Japan built the first HSR: Tokaido Shinkansen in 1964, many countries have begun to pay attention to high-speed railway (HSR). For example, France has TGV since 1983 and Spain has AVE since 1992. In China, the first HSR was built in 2003 and this is the only one until 2008. Though the opening time is relatively late, the development is very fast. By the end of 2017, the HSR operating mileage is 25,000 kilometers, and the annual passenger traffic is 1.2 billion in China. We can easily see the fast development from Figure 1, which shows the HSR traffic in several countries.
Figure 1: High-speed railway traffic in several countries, passenger-km (billions) Source: Data comes from UIC (The Worldwide Railway Organization)
Figure 2: HSR map planned in 2030
Given the increasing amount of HSR and the continuing huge amount of investment, one question rises naturally: does HSR really influence the economy as expected? There are many discussions about the relation between transportation and economy. From theoretical perspective, Marshall discussed the relation between the improvement of transportation and distribution of industries. In economic geography, transportation cost plays an important role in agglomerations and dispersion. From empirical perspective, there is no consistent conclusion about HSR’s impact. People study the impact from different points, use different methods and draw different conclusions. On one hand, some researchers proved that HSR promote the disparities of economic activities and benefit economy. On the other hand, others think there is also competition between big cities and small cities and HSR may make peripheral cities more peripheral. Given the inconsistent conclusions in previous literature, I will firstly use city-level data to study the impact of HSR.
paper attempts to find potential ways through which HSR influence the economy.
To address above questions, panel data over 12 years will be used. The result shows that HSR has substantially positive impact on regional economic development. However, when making comparison between big cities and small ones, it seems that big cities benefit more, indicating the inequality problem may rise with the building of HSR. However, this result is insignificant in statistics. After examining the overall impact, I further discuss channels through which HSR influence the economy. I select two mediators: knowledge spillover and FDI and find the mediation effect of FDI and Knowledge spillover are both significant. This indicates that HSR can influence the economic development indirectly.
2. Literature review and hypothesis development
Before reviewing the previous literature, it is better to show the conceptual framework of the paper. Figure 3 lists variables used in this thesis and their relationship. High-speed railway is an independent variable and economic development is a dependent variable. Besides the direct impact of HSR on economic development, I think HSR can also influence the economy indirectly. Knowledge spillover and FDI are channels between HSR and economic development. Control variables are used in order to make the relationship between HSR and economic development more accurate. I have 5 hypothesizes in total.
Figure 3: Conceptual Framework
2.1 HSR
the new line which is built at a speed of 250-300 kilometers per hour. In 1985, The Geneva Agreement gave it new regulations: HSR are the new passenger-cargo collinear railways with a speed of more than 250 kilometers per hour and the new passenger-only railways with a speed of more than 350 kilometers per hour. In China, HSR is defined as the line with a speed of 250 kilometers per hour or above. At the same time, some track lines with a speed of 200 kilometers per hour are also included in the category of China's high-speed railway network. I adopt this definition and include the trains numbered by letter C and letter G into the sample.
2.2 HSR and Economic Development
During the 1940s, Rosenstein-Rodan proposed the theory of the big push. He thought one prerequisite for promoting national economic development is investment in infrastructure. Since then, more and more papers (Esfahan and Ramı́rez, 2006; Hulten et al, 2006; Boopen, 2006) discussed the relationship between infrastructure and economy. HSR, as an advanced transportation infrastructure, also has attracted much attention since its birth in 1964. However, there is no unified conclusion about its impact from empirical perspective.
Shinkansen network expansion leads the dispersion from big cities to some degree. However, because HSR also benefits the developed cities, it cannot solve the problem of excessive agglomeration totally. On the other hand, some people argue that HSR brings some problems. Hall (2009) thought the HSR mainly serves areas with profitable areas and ignores smaller cities. The HSR connecting large centers may make peripheral regions more remote. And this will lead to the polarization effect. López, et al (2013) analyzed the HSR in Spain from efficiency and equity. They found that though the territorial equity improved, new kinds of inequity raised. HSR does not change the dominant status of some cities.
From previous literature, I can find that researchers have different conclusions about HSR’s impact because they have different concerns. If researchers consider the overall level of economic development, then the impact of HSR is generally positive. If researchers consider the size of cities connected by HSR and the distribution of HSR, then the problem of inequality will be highlighted.
The conclusion about HSR’s impact on China is relatively consistent. Many researchers in China thought HSR can promote regional development. Wen and Han (2017) used the gravity model, regional superiority potential model, and Moran index to study the impact of the high-speed rails in China. They found that high-speed railways improve the accessibility of connected cities. The growth rate of economic relation intensity in second-tier cities is higher than that in first-tier cities1. The high-speed railways bring
diffusion effects and the gap between regions has decreased. You and Zheng (2018) used data of 12 middle-size cities in Hunan province to study the impact of HST and find HSR has a significant positive impact on these cities in terms of GDP. Besides, HSR has a greater prositive effect on cities with lower initial GDP, and there is no significant difference between cities with different size.
1
Recently, some authors try to use econometric models to find the impact of HSR in China. However, most articles are still from the perspective of explaining mechanism and the number of papers using econometrics models is relatively small. Many papers explain the effect of HSR in terms of accessibility (Shaw, et al 2014). In these papers, authors usually use GIS or calculate manually to measure accessibility. Then they compare accessibility before and after the openness of HSR to show HSR reduces relative distance and expand the access distance. The improved accessibility may indicate that HSR benefits the economy. However, the conclusion of this method is quite obvious and simple. The direct impact of HSR is reducing travel time and make cities closer. Besides, if there is no direct HSR between two cities, some researchers neglected the transfer time in the journey. This is not in line with reality and will lead to a less precise result. HSR may experience some adjustments in the course of its operations. This brings additional problems to the calculation of accessibility. Therefore, I plan to use econometric models instead of comparing accessibility to study the impact of HSR. Meanwhile, most studies only focus on one line of HSR or one specific area (Luo, et al 2004; Wang, et al 2016). The number of cities on a single HSR line or in a specific area is relatively small. Therefore, I plan to include all cities in China to study the impact of HSR on economic development. Now, based on previous literature, I have my first hypothesis:
H 1: High-speed railway has a positive impact on economic development
2.3 Knowledge Spillover
Knowledge spillover is a complex concept. Generally speaking, if people or firms benefit from others without paying, knowledge spillover occurs. The main characteristic of it is the externality. In view of knowledge spillover developed by Marshall, Arrow and Romer, the closer the firms are to each other, the greater the spillover. In new growth theory, knowledge and human capital are important for growth and they have a spillover effect. This means one firm’s capital accumulation will also benefit other firms. In new economic geography, sharing, matching and learning are three mechanisms of agglomeration economies (Duranton and Puga 2004). Among them, learning includes face-to-face contact and spillover.
2.4 HSR, Knowledge Spillover and Economic Development
With the development of the Internet, knowledge can be shared without distance limitation. However, knowledge is not always passed by this way (Schilling, 2000). Sometimes tacit knowledge, such as the ability of someone, is difficult to codify and learned by other people (Cowan and Foray, 1997). Fallah and Ibrahim (2004) discussed knowledge in terms of accessibility and thought transmitting non-codified knowledge needs direct interaction. Therefore, face-to-face communication is still an important source of transmitting knowledge and generating spillover. Because HSR in China is mainly used for passengers, it naturally can promote the mobility of people and make face-to-face communication more convenient. Besides, knowledge spillover is often geographically limited (Fisher et al,2009). FU (2009) found the technology spillover often happened within 800 km or between adjacent provinces. Good transportation such as HSR can reduce time-space distance, change the relative location of firms and workers and make them closer to each other. In other words, HSR expands the scope of knowledge spillover.
important sources of economic growth. Varga and Schalk (2014) find that localized knowledge spillover plays an important role in promoting technological change, which is related to macroeconomic growth. Duguet (2006) found that spillover between firms is important for radical innovation, which makes contribution to TFP growth. Therefore, based on previous literature, I have my next hypothesizes:
H 2 (a): HSR has a positive impact on promoting knowledge spillover.
H 2 (b): Knowledge spillover has a positive impact on economic development
Combined H 2 (a) and H 2 (b), I assume that the impact of HSR on the economy is mediated by knowledge spillover.
2.5 FDI
FDI is an investment made by a firm or individual in foreign countries. Usually, it has three forms: Greenfield, Joint Venture and cross-border M&A. It also can be distinguished between vertical and horizontal FDI. Under this condition, vertical FDI is related to lower costs while horizontal FDI is related to market access. According to OECD, FDI can promote international economic integration and it may improve the international competitiveness of both host and home countries.
2.6 HSR, FDI and Economic Development
FDI and economic growth in Latin America. They found FDI has a positive and significant impact on economic growth in all models. Using panel data from 12 Asian countries, Wang (2009) discussed FDI using in different sectors and found that FDI inflowed to the manufacturing sector has an important impact on economic growth. Pegkas (2015) used full modeled OLS and dynamic OLS models to study the impact of FDI. He found that FDI stock has a positive impact on economic growth in Eurozone countries. Fu (2004) used the difference of GDP per capita between coastal and inland provinces to measure the disparities between provinces. He found that one percent increase in the average FDI flow gap between these regions will lead to 0.02 percent of the income gap.
Though FDI can promote economy, it has some requirements for host countries. In other words, FDI is not sufficient condition for economic growth. The impact of HSR is influenced by host country’s characteristics such as financial market, trade openness, human capital and infrastructure (Borensztein et al, 1998; Alfaro et al, 2004). Among them, infrastructure is an important factor. Good infrastructure can reduce the cost of FDI and make countries more attractive. Kinoshita and Lu (2006) used infrastructure to represent the absorptive capacity of FDI host countries and studied how it determined the effect of FDI on economic development. They found that the infrastructure with better quality can make host countries benefit more from FDI spillovers and achieve economic growth. Nourzad, et al (2014) found that the infrastructure level of host countries influences the effect of FDI. The better infrastructure leads to the higher marginal effect of FDI on real income. HSR, as convenient and advanced transportation, can improve the regional infrastructure obviously. It reduces the production cost and improves returns to capital, making cities more attractive. Therefore, I have the following hypothesizes:
H 3 (a): HSR has a positive impact on attracting FDI.
Combined H 3 (a) and H 3 (b), I think the impact of HSR on the economy is mediated by FDI.
3. Data and Methodology
To assess the effect of HSR on the economy, I collect data of the opening time of HSR, regional economic growth rate and other city-level variables which can influence the result. In this section, I will discuss the data and econometric methods.
3.1 Sample
I collected prefecture-level city-level data over 12 years, from 2005 to 2016. The classification of cities is based on the Chinese administrative boundary. The prefecture-level city is the second prefecture-level of administrative area. Such cities are important for development and many of them are the core cities of clusters. Some cities only occurred in several years, so I dropped them. I also dropped city San Sha, because it is set up for politic reasons and there is no economic activity. Hence, based on data availability and consistency, I finally keep 280 cities in China. In total, I have data for 280 cities over 12 years, so 3360 observations are my sample for regression. Figure 4 shows the distribution of these cities. The areas with light pink are prefecture-level cities. Besides, all original data except the opening time of HSR comes from ‘CHINA CITY STATISTICAL YEARBOOK’. Data regarding openness time is collected from all kinds of news and website and entered manually.
Figure 4: Distribution of prefecture-level cities in China (light pink)
3.2 Data
Independent variable:
When measuring the impact of HSR, accessibility is a popular indicator (Kim and Sultana, 2015; Shaw, et al 2014). It is related to a city’s relative location and characteristic. Usually, it is the travel time weighted by GDP of each economic activity center. A higher value means worse accessibility. When using this indicator, people usually choose two periods of data before and after the opening of HSR to reflect how HSR cuts the space-time distance. Because accessibility is calculated by real economic and travel data, it is also continuous and can influence growth rate gradually.
However, I set up a dummy variable to capture the impact of HSR on regional economic development. Here are the reasons for using this instead of accessibility. Firstly, I include 280 cities across the country in my sample and this is a large scale. It is difficult to decide the cut-off distance within which cities have close connections with each other. If there is a long distance between cities, they cannot have a big influence on each other. Secondly, if using accessibility as the indicator, I have to calculate it in each year for all cities. But during this period, China adjusted the speed of railways several times. It is hard to collect travel time between cities each year. Compared with it, the opening time of HSR, which is being reported by many media, is relatively clear and easy to find. One problem related to this choice is that this variable only has two values: 0 or 1. The obvious change between these two values will lead to a strong change in the result, and this change is one-off. However, based on data availability, I still choose this dummy variable to represent HSR. If city i has HSR in year j, the value of this variable is 1. Otherwise, it is 0.
Dependent variable:
macro-economy. It measures the speed of economic growth. Secondly, economic development is a dynamic process and growth rate also means a dynamic change. Compared with GDP per capita, GDP growth rate can better reflect the development and reduce the probability of reverse causal problems.
Mediator variables：
FDI：
FDI can reflect the attractiveness of a region to foreign firms. To capture its impact on regional economic growth, I use the actual amount of foreign capital to represent the mediator FDI. It is measured by 10000 dollars.
Knowledge spillover:
Just as mentioned above, knowledge spillover is a complex concept, so measuring it is more difficult. Because of different research purposes and data in hand, people use different methods, such as technology flow approach (Wolff and Nadiri, 1993) and the production function approach (Feldman, 1994), to measure spillover. The patent is a popular choice to represent spillover (Acs, et al 1992; Thompson, et al, 2005). It can reflect the output of knowledge and innovation. Therefore, consistent with other researchers, I also use the number of patents to represent knowledge spillover. Because the effect of patents on economic development may be lagging, I choose one-stage lag variable of patents. Another reasonable indicator to measure knowledge spillover is the number of papers published by scholars in important journals. HSR makes face-to-face communication and provides more chances for scholars to exchange ideas and cooperate. However, due to time limitation, it is difficult to collect the data.
Control variables:
people. Because in the previous section, I think the openness of HSR will reduce the travel time and promote the mobility of people. A closer connection between people and cities will provide more opportunities for individuals to communicate and further promote development. Other transports will have similar functions as HSR, so I want to control the effect of them. Besides, people can also communicate via the internet. The internet can free people from distance. For example, sometimes people can use video phone to have a meeting or discuss problems. This means the internet can replace face-to-face communication to some extent and reduce the use of transportations. So, when examining the impact of HSR, I also control the effect of the internet. Water passenger traffic is not included because seldom paper discusses the relationship between water passenger traffic and economy. Based on data availability, I cannot distinguish HSR passengers from normal train passengers. Therefore, I do not use normal rail passenger traffic as control variable.
Secondly, industrial structure is used. Industrial structure refers to the composition of each industry and the relationship between various industries. In the process of economic development, the division of labor becomes more and more detailed, and more and more production departments have emerged. Every sector has its own characteristics and makes different contributions to economic development. In China, the economy is divided into three departments: primary industry, secondary industry and tertiary industry2. Among these industries, the secondary industry which includes
manufacturing sections and tertiary industry which includes service sections play a more important role for cities. Their contribution to economic growth may be different and HSR may also have a different impact on different industries. Therefore, I use the proportion of the secondary industry and the tertiary industry to represent the industrial structure and use it as the control variable.
2
Then, according to the growth theory, capital and labor are important factors that will influence the economy. Capital can be divided into human and physical capital. Therefore, I want to use the growth rate of fixed asset investment to control for physical capital and use the ratio of government expenditure on education to GDP to measure human capital. I use the growth rate of the population to control for labor. The ratio of government expenditure on science and technology to GDP represents the total factor productivity. If I can find the expenditure of firms on R&D, the result may be much better. The growth rate of GDP may also be influenced by the level of GDP. For example, cities with higher GDP level may have a lower growth rate of GDP. Thus, I use GDP per capita as the control variable and take the logarithmic form of it. Logarithmic transformations are often used for monetary variables, which have distributions with a long tail to the right. Here is the list of variables:
Figure 5: List of variables
Variable Definition
Rate GDP annual growth rate of every city
HSR The opening year of HSR
Road Highway passenger traffic
Air Civil aviation passenger traffic
Internet The number of subscribers of Internet services
FDI Actual amount of foreign capital used in every year
Patent The number of patents each year (to represent knowledge
spillover)
Structure Industrial structure presented by the ratio of the
secondary industry to the tertiary industry
GDP per capita GDP per capita calculated by annual GDP and population
R&D The ratio of government’s R&D expenditure to GDP (to
represent technology level)
Education The ratio of government’s education expenditure to GDP
(to represent human capital)
Asset growth rate Annual growth rate of fixed asset investment
3.3 Methodology
I use fixed effect model to measure the relation between HSR and economy. Here is the regression set-up:
Yij = β0 + β1Dij +β2Xij + Tj + Ci + eij (1)
Because the variable of the interest is a dummy variable, it is noticeable that this equation can also be regarded as an extension of basic DID (difference-in-differences) model. DID model is an important way to estimate the influence of policy. It compares differences within treatment and control groups and the difference between them. It is also a choice to solve the endogenous problem. Basic DID model has a unified policy implementation time and includes three dummy variables: time, treat and the interaction of them. The time variable equals 0 before the implementation of policy and equals 1 after implementation. The treat variable equals 0 if the individual is not affected by the policy and equals 1 if it is affected. The interaction equals 1 only when the individual is really affected by policy. This variable captures the effect of the policy. However, the policy implementation time in my sample is inconsistent, the basic form cannot be applied directly. I drop time and treatment variable and only keep the interaction. The coefficient of the interaction variable is in my interest. One requirement of the DID model is the parallel trend, which means the treatment group and control group should have similar trend without policy. The sample passed this test and the result can be found in the Appendix Table A.
Now back to equation (1). In equation (1), Yij measures the GDP growth rate in city i
in year j, Ci and Tj are dummy variables that capture the city-fixed and year-fixed effects.
Ci is used to control for time-invariant, unobserved individual characteristics that
influence the economy across cities. In this thesis, individuals are geographical regions, it is reasonable to introduce this term. Tj is used to control for national-wide shocks,
is a set of control variables which are used to make the result more reliable and eij is the
error term. I choose them according to economic growth theory. Dij is the variable I am
interested in. It equals 1 when city i in year j has HSR and equals 0 otherwise. The coefficient indicates the impact of HSR on regional economic development. A significant and positive coefficient means HSR has a positive impact on economic development, while a significant but negative coefficient means HSR does not benefit economic development.
After examination and drawing residual plots, heteroskedastic exists in the model and makes the results less reliable. To address this problem, I estimate equation (1) allowing for robust standard errors, which are valid in large samples for both heteroskedastic and homoscedastic errors.
I further divided the sample into two parts: the sample including big cities and the sample including small cities. Some people argue that HSR may benefit big cities more and attract resources from small cities. This may enlarge the gap between different cities. Therefore, I distinguish four big cities (Beijing, Shanghai, Guangzhou and Shenzhen) from other cities and study the impact of HSR separately. These cities are most developed cities in China, and I assume HSR may have a different impact on them.
When studying the channels, mediators FDI and knowledge spillover are used. There are many ways to examine the mediation effect, and the procedure introduced by Baron
and (1986) is a popular method（Saeidi et al, 2015）.This method includes three steps
type-Ⅱ error increased. If the mediation effect is not strong, the result may indicate the mediation effect is not significant. Therefore, based on this process, I adopt the examination process provided by Wen et al (2004). They compared several methods about examining mediation variables (Judd and Kenny, 1981; Baron and Kenny, 1986; Sobel, 1982; Mackinnon, et al 1998; Freedman and Schatzkin, 1992) and provided a new process to examine it. This method has a detailed discussion of every step. This method can reduce the probability of making Type I error and Type II error and it can examine both partial mediation effect and complete mediation effect. It is also easy to follow. The method is showed in Figure 6.
Figure 6: Examining process of mediation effect
GDP growth rate= cHSR+Xit+e1 (2)
FDI/Knowledge spillover= aHSR+Xit+e2 (3)
GDP growth rate = c’HSR+b FDI/Knowledge spillover +Xit+e3 (4)
When examining the mediation effect of FDI, I follow these steps. If the relationship between HSR and dependent variable in equation (2) is significant, I will go to the next step. If it is insignificant, I will stop the test of mediation effect because it does not meet the requirements to test the mediating effect. Then, I will examine the equation (3) and (4) separately. If the coefficient of HSR in equation (3) and coefficient of FDI in equation (4) are both significant, I will look at the coefficient of HSR in equation (4). If it is significant, there is a partial mediation effect. If it is insignificant, there is a full mediation effect. If at least one of the coefficients of HSR in equation (3) and coefficient of FDI in equation (4) is insignificant, I will do the Sobel test. If the result is significant, the mediation effect is significant. Otherwise, the mediation effect is insignificant. The same process will be used to examine the mediation effect of knowledge spillover between HSR and GDP growth rate.
4. Empirical Test
Before implementing regressions, I did two tests to detect the multicollinearity problem and to decide the model. Though individual in this thesis is city and it is clearly “fixed”, I still do Hausman Test to prove it. The results, which are shown in Appendix Table B and C indicate that the model does not have the multicollinearity problem and I should use the fixed model. Here is the descriptive statistics before regression.
Table 1: Descriptive Statistics
Variable Obs Mean Std.Dev. Min Max
HSR 3360 .242 .428 0 1
GDP growth rate 3356 11.853 4.198 -19.38 37
Asset growth rate 3357 .268 .879 -.929 29.694
Population growth rate 3360 .008 .065 -.743 3.051
R&D 3358 .002 .002 0 .063 Education 3357 .03 .018 0 .159 Lngdppc 3358 10.194 .822 7.802 13.135 FDI 3186 73340.65 180000 12 3080000 Patent 3051 .263 .839 0 10.836 Structure 3357 1.455 .691 .052 7.776 Internet 3347 59.718 133.441 0 5174 Road 3352 8629.552 14567.53 82 287000 Air 3360 166.741 853.984 0 29301
4.1 HSR and regional economic development
Table 2: Regression results (1) Full Sample (2) Full Sample (3) Big cities (4) Big cities (5) Small cities (6) Small cities rate rate rate rate rate rate
HSR 0.267 0.328 0.261 HSR_1 (0.275) 0.489* (0.268) (0.369) 0.742 (0.494) (0.279) 0.483* (0.273) Structure 0.926** 0.924** 2.775 3.309 0.911** 0.908** (0.364) (0.366) (1.858) (1.587) (0.366) (0.368)
Asset growth rate 0.314*** 0.312*** 1.933** 2.005** 0.313*** 0.311***
(0.092) (0.091) (0.449) (0.496) (0.092) (0.091)
Population growth rate 3.600*** 3.619*** -8.656* -4.364 3.615*** 3.631***
(0.590) (0.589) (2.841) (4.723) (0.596) (0.595) R&D 129.904*** 130.393*** -654.187** -621.512** 131.759*** 132.624*** (30.921) (30.741) (171.979) (143.417) (31.760) (31.665) Education 92.517*** 93.457*** -58.262 -86.410 93.143*** 94.092*** (15.599) (15.626) (77.135) (95.442) (15.656) (15.686) Lngdppc 6.259*** 6.329*** -1.227 -0.568 6.288*** 6.356*** (1.212) (1.214) (0.876) (1.232) (1.229) (1.231) _cons -48.061*** -48.727*** 24.306* 17.379 -48.205*** -48.857*** (11.350) (11.366) (8.494) (12.025) (11.490) (11.507) Obs. 3351 3351 48 48 3303 3303 R-squared 0.567 0.568 0.886 0.888 0.567 0.568
Standard errors are in parenthesis *** p<0.01, ** p<0.05, * p<0.1
In Table 2, column 1 and 2 are results for the full sample. Column 3 and 4 show the results for big cities and the last two columns show results for smaller cities. In terms of HSR, all coefficients are positive but insignificant. This result is in line with expectations. Just as mentioned above, many cities operated HSR at the end of the year. For example, HSR between Beijing and Zhengzhou opened in December. HSR operated at the end of the year is difficult to affect the economy of the year.
very large. HSR can make cities closer and can better reallocate resources. The convenient transportation also gives employers and employees more chances and reduces the research cost. The coefficient of HSR_1 is also positive and significant in the sample with small cities. This indicates that the growth rate of cities with HSR is 48.3% higher than the growth rate of cities without HSR. HSR has an important and significant impact on promoting small cities’ development. However, the same coefficient in the sample with big cities is not significant. There are several reasons for this result. Firstly, other transportation such as aviation is very developed in big cities and HSR does not have advantages in terms of time and cost. For example, there are over 30 flights per day from Beijing to Shanghai and the travel time is 2.5 hours. Usually it takes 350-600 Yuan. Compared with it, HSR takes around 5 hours and the cheapest seat costs around 550 Yuan. Secondly, there are more important factors, such as human capital and advanced technology, affecting the economy in big cities. Transportation may have little impact on these cities.
Though the coefficient in big cities is insignificant, it is much higher than that of smaller cities, indicating that HSR may promote factor flow from small cities to big ones and enlarge the gap between them. The difference between HSR and HSR_1 shows HSR can affect the economy through factor flow and this needs time to accumulate.
Looking at coefficients of control variables, all of them are positive and significant at 5% in the full sample. The same result can be found from models with small cities. However, things look different when only considering big cities. Only two control variables: asset growth rate and R&D expenditure are significant. This further indicates the complicated situation about big cities.
4.2 Channels
channels of the impact. Based on previous literature, I choose two mediator variables: FDI and Knowledge spillover. Table 3 and Table 4 are results for the mediation effect examination.
Table 3: Mediation regression results for FDI
(a) Dependent Variable: rate
Model Coefficient Std Error t sig
HSR_1 0.521 0.183 2.84 ***
(b) Dependent Variable: FDI
Model Coefficient Std Error t sig
HSR_1 4.993 0.566 8.83 ***
(c) Dependent Variable: rate
Model Coefficient Std Error t sig
HSR_1 FDI 0.437 0.017 0.186 0.006 2.35 2.80 ** ***
Table 4 (b): Mediation regression results for Knowledge spillover
(a) Dependent Variable: rate
Model Coefficient Std Error t sig
HSR_1 0.532 0.186 2.86 ***
(b) Dependent Variable: Patent
Model Coefficient Std Error t sig
HSR_1 0.280 0.030 9.22 ***
(c) Dependent Variable: rate
Model Coefficient Std Error t sig
HSR_1 Patent 0.328 0.609 0.188 0.117 1.74 5.22 * ***
knowledge spillover. According to the steps mentioned in section 3, the impact of HSR can be partially mediated by FDI and Knowledge spillover. In Table 3, there is a significant relationship between HSR and rate, which satisfied the first step. There is a significant relationship between HSR as the independent variable and FDI as the dependent variable, satisfying the second step. FDI is also related to the GDP growth rate when controlling HSR and both coefficients are significant. Therefore, according to the method, the partial mediation effect is proved and there is no need to do other tests. In Table 4, the same result can be found for knowledge spillover. Controlling other variables, HSR as the independent variable is related to GDP growth rate and it is also related to knowledge spillover. Knowledge spillover is also related to the GDP growth rate when controlling HSR.
5. Conclusion and Limitation
5.1 Conclusion and Recommendations
Researchers have different conclusions about HSR’s impact and winners and losers from the construction of HSR. While some argue that HSR can reduce time-space distance, improve the accessibility of regions and promote regional economic development, others hold that considering the different size of cities and distribution of HSR lines, HSR may enlarge the gap between cities and make peripheral cities more peripheral.
Using city-level data over 12 years, I find that HSR can promote economic growth when considering all cities. However, the impact of HSR should be achieved through the accumulation of factors. The accumulation needs time. Besides, HSR has possibility to expand the gap between smaller cities and big ones. However, the result is insignificant in statistics.
I find evidence that HSR can influence economic growth indirectly by attracting more FDI. HSR, as advanced transportation, can change the relative location of cities, reduce production cost and increase returns to capital. Construction of HSR can attract more FDI and FDI can provide capital and advanced knowledge for economic growth. There is also evidence that knowledge spillover is another channel. HSR can reduce travel time and make it easier for face-to-face communication. Frequent face-to-face communication leads to more knowledge spillover and expands the spillover’s scale. The spillover between individuals can benefit economic growth by promoting innovation and improving total factor productivity.
spillover are two indirect paths. Based on the findings, here is the recommendations. The government can use HSR as a way to promote regional economic development. The impact of HSR is especially significant on small cities. However, the government should consider cities with different size when drawing HSR map. Large cities may attractive resources from small cities. The government should pay attention to potential inequity problem in the process of constructing HSR.
5.2 Limitation and Further Research
There are still limitations to discuss in the future. The first limitation is related to data selection. Because I study the effect of HSR using city-level data, it is more difficult to collect them. Just as mentioned in section 3, I only use air traffic, highway traffic and internet data as control variables. I cannot find data and distinguish HSR passenger traffic from railway passenger traffic. If I can distinguish them, the result for the mediation effect may be more precise.
The second problem is related to how to represent HSR. Also as mentioned in section 3, I use a dummy variable to represent HSR which will equal one since its first open. If I can find other continuous variables to represent it, I may find HSR will influence the economy gradually.
6. References
Acs, Zoltan J., David B. Audretsch, and Maryann P. Feldman. "Real effects of academic research: comment." The American Economic Review 82.1 (1992): 363-367.
Baron, Reuben M., and David A. Kenny. "The moderator–mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations." Journal of personality and social psychology 51.6 (1986): 1173.
Bengoa, Marta, and Blanca Sanchez-Robles. "Foreign direct investment, economic freedom and growth: new evidence from Latin America." European journal of political economy19.3 (2003): 529-545.
Clogg, Clifford C., Eva Petkova, and Edward S. Shihadeh. "Statistical methods for analyzing collapsibility in regression models." Journal of Educational Statistics 17.1 (1992): 51-74.
Ciruelos, Alejandro, and Miao Wang. "International technology diffusion: Effects of trade and FDI." Atlantic Economic Journal 33.4 (2005): 437-449.
Cowan, Robin, and Dominique Foray. "The economics of codification and the diffusion of knowledge." Industrial and corporate change 6.3 (1997): 595-622.
Duguet, Emmanuel. "Innovation height, spillovers and TFP growth at the firm level: Evidence from French manufacturing." Economics of Innovation and New technology 15.4-5 (2006): 415-442.
Freedman, Laurence S., and Arthur Schatzkin. "Sample size for studying intermediate endpoints within intervention trials or observational studies." American Journal of Epidemiology 136.9 (1992): 1148-1159.
Fischer, Manfred M., Thomas Scherngell, and Martin Reismann. "Knowledge spillovers and total factor productivity: evidence using a spatial panel data model." Geographical Analysis 41.2 (2009): 204-220.
Fu, Xiaolan. "Limited linkages from growth engines and regional disparities in China." Journal of Comparative Economics 32.1 (2004): 148-164.
Hejazi, Walid, and A. Edward Safarian. "Trade, foreign direct investment, and R&D spillovers." Journal of International Business Studies 30.3 (1999): 491-511.
treatment evaluations." Evaluation review 5.5 (1981): 602-619.
Kinoshita, Yuko, and Chia-Hui Lu. "On the role of absorptive capacity: FDI matters to growth." (2006).
LUO, Peng-fei, Yi-lun XU, and Nan-nan ZHANG. "STUDY ON THE IMPACTS OF REGIONAL ACCESSIBILITY OF HIGH SPEED RAIL——A CASE STUDY OF NANJING TO SHANGHAI REGION [J]." Economic Geography 3 (2004).
MacKinnon, David P., Chondra M. Lockwood, and J. Hoffman. "A new method to test for mediation." annual meeting of the Society for Prevention Research, Park City, UT. 1998.
Monzón, Andrés, Emilio Ortega, and Elena López. "Efficiency and spatial equity impacts of high-speed rail extensions in urban areas." Cities 30 (2013): 18-30.
Nourzad, Farrokh, David N. Greenwold, and Rui Yang. "The interaction between FDI and infrastructure capital in the development process." International Advances in Economic Research 20.2 (2014): 203-212.
Peterman, David R., John Frittelli, and William J. Mallett. "High speed rail (HSR) in the United States." LIBRARY OF CONGRESS WASHINGTON DC CONGRESSIONAL RESEARCH SERVICE, 2009.
Pegkas, Panagiotis. "The impact of FDI on economic growth in Eurozone countries." The Journal of Economic Asymmetries 12.2 (2015): 124-132.
Shaw, Shih-Lung, et al. "Impacts of high speed rail on railroad network accessibility in China." Journal of Transport Geography 40 (2014): 112-122.
Jiao, Jingjuan, et al. "Impacts on accessibility of China’s present and future HSR network." Journal of Transport Geography 40 (2014): 123-132.
Sasaki, Komei, Tadahiro Ohashi, and Asao Ando. "High-speed rail transit impact on regional systems: does the Shinkansen contribute to dispersion?" The annals of regional science 31.1 (1997): 77-98.
Schilling, Melissa A. "Toward a general modular systems theory and its application to interfirm product modularity." Academy of management review 25.2 (2000): 312-334.
Sobel, Michael E. "Asymptotic confidence intervals for indirect effects in structural equation models." Sociological methodology 13 (1982): 290-312.
Thompson, Peter, and Melanie Fox-Kean. "Patent citations and the geography of knowledge spillovers: A reassessment." American Economic Review 95.1 (2005): 450-460.
Wang, Lvhua, et al. "Accessibility impact of the present and future high-speed rail network: A case study of Jiangsu Province, China." Journal of Transport Geography 54 (2016): 161-17
Wang, Miao, and M. C. Sunny Wong. "What drives economic growth? The case of cross‐border M&A and greenfield FDI activities." Kyklos 62.2 (2009): 316-330. Wang, Miao. "Manufacturing FDI and economic growth: evidence from Asian economies." Applied Economics 41.8 (2009): 991-1002.
游士兵, and 郑良辰. "高铁对沿线中型城市的经济拉动效应评估." 改革 10 (2018): 15.
Appendix
Table A (1) Parallel Trend when independent variable=HSR
Table A (2) Parallel Trend when independent variable=HSR_1
Table B Result for multicollinearity test
Table C The result for Hausman test
VIF 1/VIF 1.400 0.715 1.310 0.766 2.260 0.443 3.000 0.333 1.160 0.862 1.050 0.954 1.010 0.991 1.860 0.538 1.940 0.514 2.080 0.482 2.120 0.473 2.210 0.452 2.410 0.415 2.700 0.371 2.700 0.370 2.730 0.367 2.940 0.340 3.040 0.329 2.110
