High-speed Rail Network and Changing Industrial Dynamics in Chinese Regions

Abstract

High-speed railway (HSR) network can significantly reduce the transport cost of people and facilitate interregional knowledge spillovers. It may thus affect regional industrial dynamics. By employing the industrial relatedness indicator, this article shows that regional industrial dynamics is path dependent in China. It further adopts several classical accessibility indicators to capture the network characteristics of transport infrastructure and the accessibility of Chinese cities in the HSR network. In response to the endogeneity issue, we design an instrumental variable based on historic transport network. Another econometric strategy is to include only two groups of cities in the sample: cities with existing HSR stations and cities with planned HSR stations. The empirical results suggest that high accessibility in the HSR network not only pushes forward new industry creation but also enables regions to be more pathbreaking and diversify into less related industries.

Keywords

regional industrial dynamics technological relatedness high-speed rail network path dependence pathbreaking transport infrastructure

Transport cost is one key factor in regional economic development and industrial dynamics (Percoco 2016). Firms and regions benefit from the decline of transport costs and the subsequent enhanced access to economic markets, due to interregional trade, knowledge spillover, and other types of agglomeration economies (Ahlfeldt and Feddersen 2018). Massive public expenditures have been invested into transport infrastructure, with the strong belief that economic activities could be facilitated by an ease of interaction. There are already plenty of studies exploring whether the development of transport infrastructure can contribute to the improvement of various economic indices such as industrial output (Crescenzi and Rodríguez-Pose 2012; Faber 2014; Melo, Graham, and Brage-Ardao 2013), GDP and per capita GDP (Ahlfeldt and Feddersen 2018; Faber 2014; Qin 2017), income (Cosci and Mirra 2018), employment (Duranton and Turner 2012; Iacono and Levinson 2016; Percoco 2016), fixed asset investment (Qin 2017), productivity (Gerritse and Arribas-Bel 2017; Li, Wu, and Chen 2017), and trade (Duranton, Morrow, and Turner 2014). There are also heated debates on the distributional consequences of infrastructures and whether the development of infrastructure network and the subsequent decline of trade costs enable peripheral regions to catch up (Baum-Snow 2007; Baum-Snow et al. 2017) or favor the core and deteriorate regional disparity (Faber 2014; Qin 2017).

This article extends earlier work in several directions. First, it focuses on one type of transport infrastructure: high-speed rail (HSR). HSR is an expensive but increasingly popular transport mode in both developed and developing countries. The European Union, South Korea, Japan, and China have all invested heavily in HSR services due to their social, economic, and environmental benefits (Ryder 2012; Shaw et al. 2014). China is an interesting case to examine the effect of HSR on industrial dynamics, since it has the largest scale of HSR and plans to build more in the near future. Furthermore, Chinese government’s willingness to channel massive public expenditures into the development of HSR is bolstered by the hope that HSR can have some positive effects on regional economic development (Cheng, Loo, and Vickerman 2015). China has a plan to connect all its main cities with a population larger than 50,000 via the HSR network (Shaw et al. 2014). The establishment of such a large-scale HSR network is anticipated to have fundamental effects on regional industrial dynamics (Chen 2012), through enhancing or weakening certain regions’ accessibility (Cao et al. 2013; Shaw et al. 2014).

While the impacts of infrastructures on regional economic development and on the spatial (re)organization of economic activities are relatively well understood, one aspect that has been left largely underexplored is that better connections also enhance knowledge spillovers between regions, and such extra-regional linkages may further affect regional industrial dynamics (Bathelt, Malmberg, and Maskell 2004; Zhu, He, and Zhou 2017). Unlike many other types of transport infrastructure (e.g., roads), HSR only affects the passenger rail service in China, allowing us to distinguish the transport cost of people from that of products. Specifically, some types of infrastructure like roads mainly affect the material logistics and reduce the transport cost of products. Low transport costs can therefore affect regional industrial dynamics via the pro-competition effects as regions face greater levels of competition due to increasing trade with other regions and the scale economies effects due to better exploitation of economies of scale driven by increased trade exposure (Behrens and Murata 2012). In contrast, HSR mainly reduces the transport cost of people rather than products. It thus facilitates interregional knowledge spillovers, by enabling people to travel around much more easily and bring back new knowledge, ideas, and information from outside.

A natural question to ask is how enhanced knowledge spillovers between regions contribute to regional industrial restructuring and new path creation. Shedding light on this issue not only allows us to better understand the impact of infrastructure on regional economic development. It is also an opportunity to relate the literature on infrastructure and transport costs in economic geography and urban economics to another strand of literature on evolutionary economic geography (EEG), which focuses on the idea of path dependence and new path creation in regional industrial dynamics (Boschma and Frenken 2006; Martin and Sunley 2006).

Second, we test the impact of transport infrastructure on regional industrial dynamics by paying particular attention to the network characteristics of transport infrastructure. Works similar to this one often either adopt a dummy variable to capture if a region has access to infrastructure networks or employ the length of roads, highways, or railroads as proxies that fail to reflect the quality and network features of transport infrastructure (Cosci and Mirra 2018; Crescenzi and Rodríguez-Pose 2012). Third, estimation of the casual effect of transport infrastructure on regional industrial dynamics faces serious empirical challenges. Such infrastructures are often designed to meet perceived need rather than randomly allocated to all regions. Regions with good or poor access to infrastructures may thus differ on many preexisting dimensions (Ahlfeldt and Feddersen 2018; Faber 2014; Heuermann and Schmieder 2018). The issue of endogeneity should be dealt with carefully.

We adopt the relatedness indicator developed by Hidalgo et al. (2007) to measure the extent to which regional industrial diversification is pathbreaking or path dependent. We also bring in several classical accessibility indicators to capture the network characteristics of transport infrastructure and the accessibility of Chinese cities in the HSR network. In response to the endogeneity issue, we design an instrumental variable based on historic transport networks. The second econometric strategy is to include only two groups of cities in the sample: cities that already had HSR stations and cities that did not have HSR stations but would build stations in the near future. The empirical results are consistent: high levels of accessibility in the HSR network not only push forward new industry creation but also enable regions to be more pathbreaking and diversify into less related industries. The next section reviews the literature. The third section describes our study area, followed by an introduction of our research design in the fourth section. The empirical results are reported in the fifth section. The last section concludes and provides some discussion.

Regional Industrial Dynamics and the Network of Transport Infrastructure

Regional Industrial Dynamics: From Geographical Proximity to Cognitive Proximity

Regional industrial dynamics can be seen as a complex process, consisting of both qualitative and quantitative changes. Even though quantitative changes like regional growth in terms of employment, value added, and output may reflect the evolution of regional economies and industrial restructuring, these changes often take place as a result of qualitative changes in regional industrial structures (Neffke, Henning, and Boschma 2011). Regions are shaped by a never-ending process of creative destruction, relying not only on the capacities of local firms to invent new products or technologies that can replace old, traditional ones in the short term but also on the regions’ capabilities to create and entice new firms and industries to offset the destruction due to firm exit and industrial decline in the long term (Schumpeter 1939, 1942).

The idea of regional economic growth and how new products, technologies, and industries emerge has been widely researched in economic geography and regional science (Martin 2010; Martin and Sunley 2006; Scott 1988; Storper and Walker 1989). Some studies have stressed the key role of geographical proximity and differentiate two ways in which regional industrial dynamics could be affected by local source of competitiveness. For Marshall ([1890] 1920), agglomeration externalities emerge as specialization, leading to thick and dense local labor market, access to specialized suppliers, and promoting local knowledge spillovers. Jacobs (1969) was more generally concerned with externalities derived from diverse regional industrial structures. In such a regional structure, diversity is the catalyst of interaction, recombination, and modification of knowledge and technologies from various industrial sectors. The gist of those two strands of literature is that geographical proximity is of central importance for knowledge spillovers to occur (Jaffe, Trajtenberg, and Henderson 1993). Other localized mechanisms, such as labor mobility, firm diversification, social networking, and entrepreneurial spinoffs, are also important to induce local knowledge spillovers, resulting in localized capabilities (Boschma and Frenken 2011; Maskell and Malmberg 1999).

Yet, recent studies in EEG literature argue that, apart from geographical proximity, cognitive proximity between firms is also important for fruitful knowledge creation and interactive innovation process to occur (Boschma 2005; Nooteboom 2000). Knowledge spillovers that are useful for innovation are most likely to take place when firms are technologically and cognitively neither so far that they cannot understand one another nor so close that there is nothing new to be learned (Frenken, Van Oort, and Verburg 2007; Nooteboom 2000). Knowledge spillovers that spur regional economic development are not expected between all industrial sectors, since some complementarities and relatedness are required or at least beneficial for knowledge spillovers (Boschma and Iammarino 2009; Boschma, Minondo, and Navarro 2012, 2013; Frenken, Van Oort, and Verburg 2007; Neffke, Henning, and Boschma 2011).

In addition to pushing forward firm performance and the growth of existing industries through knowledge spillovers, relatedness contributes to new industry creation (Boschma and Capone 2015; Boschma, Minondo, and Navarro 2013; Delgado, Porter, and Stern 2016; Neffke, Henning, and Boschma 2011). New industries do not emerge from scratch randomly; rather, new industry creation is heavily shaped by regional preexisting industrial structures via knowledge spillovers across related industrial sectors (Boschma and Frenken 2011; Boschma and Martin 2007). Regions often diversify into new industries through a process of branching, where, driven by technological relatedness, a new industry originate either from related ones (Klepper and Simons 2000) or from the interaction, recombination, and modification of competencies from a number of related industries (Klepper 2002; Tanner 2016). It is further pointed out that regions tend to develop new industrial sectors that are related to existing industrial profiles by a large number of empirical studies (Binz, Truffer, and Coenen 2016; Colombelli, Krafft, and Quatraro 2014; Simmie 2012; Tanner 2014).

The Role of Transport Infrastructure

Despite the transition from geographical proximity to cognitive proximity, the emphasis of the EEG literature is still on localized capabilities in general and on endogenous industrial diversification in particular (Grillitsch and Trippl 2014; Tanner 2014). It has a problem of regional fetishism (Binz, Truffer, and Coenen 2016; Martin and Sunley 2006) and tends to treat regions as self-contained entities (Boschma, Martín, and Minondo 2017). However, regions are often interconnected with each other through extra-regional linkages, which may bring in new knowledge from outside and lead to the emergence of new products, technologies, and industries (Bathelt and Cohendet 2014; Bathelt, Malmberg, and Maskell 2004; Maskell 2014; Maskell, Bathelt, and Malmberg 2006). Knowledge diffusion via extra-regional linkages is also referred to as “pipelines,” which are often seen as crucial for sustainable regional economic development (Bathelt, Malmberg, and Maskell 2004; Boschma and Iammarino 2009).

Such extra-regional linkages could be reinforced by the network of transport infrastructure. The improvements of transport infrastructure will definitely lead to the “collapse” or “compression” of the time and space and changes in certain regions’ accessibility, via the reduction of travel time and transport costs (Hou and Li 2011; Wang et al. 2016). Transport infrastructure thus contributes to economic interactions and integration between regions and are often seen as a catalyst for economic development and regional industrial dynamics (Cheng, Loo, and Vickerman 2015; Percoco 2016). In our case, unlike other types of transport infrastructure, HSR has been designed exclusively for passenger transportation (Ahlfeldt and Feddersen 2018). It thus allows us to isolate the impacts of facilitated human interactions from the impacts of the reduction in transport costs of products, providing an ideal case to examine the effects of enhanced knowledge diffusion across regions on regional industrial dynamics (Ahlfeldt and Feddersen 2018; Heuermann and Schmieder 2018). With the help of knowledge diffusion via the HSR network, regions may be able to import new knowledge from even remote areas through more frequent human interactions and expand the diversity of their own knowledge bases.

In this article, we differentiate two types of new industry creation: pathbreaking and path dependent (Zhu, He, and Zhou 2017). The first one takes place when firms introduce new sectors that are closely related to regional preexisting industrial profile. This type of new industry creation, also known as “industrial change” (Neffke et al. 2018), is path dependent and heavily reliant on technological relatedness. New industries can also be created by firms that are able to transcend the reliance on technological relatedness. If regions manage to develop some less related or even unrelated industries (Boschma, Balland, and Kogler 2015), regional development could be more pathbreaking, resulting in unrelated diversification or “structural change” (Neffke et al. 2018). Here, we examine whether HSR has enabled Chinese regions to diversify their industrial structures in a more pathbreaking way, due to the ease of communication between regions. Extra-regional linkages may be formulated or enhanced via the HSR network and bring in fresh knowledge and ideas from outside, which are less related or even unrelated to regional preexisting industrial profile, resulting in unrelated diversification and pathbreaking new industry creation. In doing so, our research not only allows us to better understand the effect of infrastructure on regional economic development but also brings the EEG literature and research on transport infrastructure into dialogue.

In addition, most studies on infrastructures tend to measure the stock of infrastructure and overlook the network character of transport infrastructure (Cheng, Loo, and Vickerman 2015; Faber 2014; Li, Wu, and Chen 2017; Percoco 2016; Qin 2017). However, variables on the length of roads and railroads and whether a region has access to infrastructure networks cannot reflect the variegated quality and conditions of the networks (Crescenzi and Rodríguez-Pose 2012). The increase in the kilometers of roads and railroads is the only one aspect of the services offered by new transport infrastructure (Cosci and Mirra 2018). This article is among the first to study the effects of the network character of transport infrastructure on regional industrial dynamics and focusing specifically on a region’s accessibility in the network that is dependent on relevant infrastructure bottlenecks, travel time, and the number of links. Furthermore, in our case, accessibility in the HSR network should be also influenced by its integration with other types of transport infrastructure. This article thus includes the conventional rail (CR) network as a complement to the HSR network. Regions that are not linked to the HSR network may still be influenced by HSR indirectly, as they may be linked to the HSR network via the CR network.

Finally, estimating the impact of infrastructures on regional industrial dynamics could be empirically challenging. The density of economic activities and the network character of transport infrastructure in a region could be interdependent or simultaneously dependent on some preexisting locational conditions such as economic development and good institutions (Ahlfeldt and Feddersen 2018; Faber 2014; Heuermann and Schmieder 2018). In other words, transport infrastructure is often not randomly distributed, thus resulting in additional endogeneity problem. The existing literature has often employed planned or historic transport networks (Duranton, Morrow, and Turner 2014; Duranton and Turner 2012) or least-cost networks (Faber 2014) as instrumental variables. This article builds on these works and employs several strategies to deal with the possibility that transport infrastructures are allocated to regions based on some unobserved determinants of economic activities.

China’s HSR Development

In 2003, China’s first HSR route from Shenyang to Qinhuangdao was opened. Its HSR network quickly expanded thereafter and become the largest in the world in 2013, with a total length of 9,760 km accounting for 46 percent of the world total (UIC 2018). Indeed, the Ministry of Railway (MOR) planned to build an HSR line between Beijing and Shanghai as early as 1990. The plan was not approved by the National People’s Congress due to the massive investment required and the lack of necessary technologies. Nonetheless, in the face of the increasing competition with high road and air transportation, the MOR accelerated the speed of existing railway lines six times during 1997–2007. Four rounds of speed acceleration took place before 2003 and increased the average speed to 60 km/hour, and two were implemented during 2003–2007 in order to upgrade existing railway to HSR, with speeds above 200 km/hour. After 2007, the main goal of the MOR changed gradually from speed acceleration on existing rail lines to building new HSR lines. The first new HSR line was constructed connecting China’s two main cities—Beijing and Tianjin—with a speed of 350 km/hour, signaling a new wave of rapid HSR development. The ambition of the MOR was also supported by Chinese central government, which sought to stimulate or retain economic growth after the outbreak of the 2008 global financial crisis (Cheng, Loo, and Vickerman 2015). Although the massive construction of HSR was forced to slowdown after the train accident in Wenzhou in 2011, it rebounded in late 2012 (Shaw et al. 2014).

China HSR construction was mainly guided by the “Mid- and Long-term Railway Development Plan” released in 2004 and revised in 2008 by the MOR. The plan aims to increase investment in China’s HSR and build up an HSR network consisting mainly of passenger dedicated lines (PDLs; 12,000 km), intercity HSR lines (4,000 km), and upgraded HSR lines (40,000 km) by the end of 2020. The planned HSR network is also known as the “four verticals” (north–south lines) and “four horizontals” (east–west lines). The former is made up by Beijing–Shanghai line, Hangzhou–Ningbo–Fuzhou–Shenzhen line, Beijing–Shenyang–Harbin line, and Beijing–Wuhan–Guangzhou–Shenzhen line. The latter includes Lanzhou–Zhengzhou–Xuzhou line, Changsha–Nanchang–Hangzhou line, Taiyuan–Shijiazhuang–Qingdao line, and Chengdu–Chongqing–Wuhan–Nanjing line (Figure 1). The planned HSR network connects major cities in China’s developed coastal region and less developed inland China.

Figure 1.

China’s national high-speed railway plan.

There are now seven types of railway lines in China: general trains (100 km/hour), fast trains with a prefix of K (120 km/hour), express trains with a prefix of T (140 km/hour), direct express trains with a prefix of Z (160 km/hour), upgraded preexisting railway lines with a prefix of D (200 km/hour), intercity HSR lines with a prefix of C (250–300 km/hour), and new PDLs with a prefix of G (300–350 km/hour). Here, by HSR, we are referring to new lines with average speeds above 250 km/hour and upgraded preexisting railway lines with speeds around 200 km/hour. In other words, here the HSR lines include lines for all G-, C- and D-type trains.

Data and Research Design

Data

We use China’s export data collected by the Chinese Customs Trade Statistics (2002–2011), which provides information on all merchandise transactions such as information on exporters, export value and quantity, and export destination. The data include some intermediary firms that do not actually manufacture and export products but act as traders. Their export behaviors are likely to be distinct from those of manufacturing firms. We remove intermediary firms as our results may be distorted by these trading agents’ business networks. We follow Bernhofen, Upward, and Wang (2017) and use a list of key words that are typically used by various types of intermediary firms in their names in Chinese (e.g., “importer,” “exporter,” and “trading”). Such firms represent around 4 percent of our observations. Furthermore, processing exports may be peculiar since their export patterns could be affected by their partners and sometimes do not have a say on their export activities. Excluding processing exports leads to the drop of one-third of the sample.

Relatedness Indicator

To examine whether the HSR network has enabled Chinese regions to diversify their industrial profiles in a pathbreaking/path-dependent way and develop new, related/unrelated industries, we need an index to measure the extent to which industrial sectors are related to one another. We adopt the co-occurrence method proposed by Hidalgo et al. (2007), who consider two industrial sectors as being related with one another if regions often have a revealed comparative advantage (RCA) in both of them. If the proportion of an industry’s exports in the regional total is greater than the proportion of the industry’s exports in the national total, this region is defined as having an RCA in this industry. The relatedness (ϕ) between industry i and j can be calculated as:

ϕ_{i, j} = min {P (R C A_{c, i} > 1 | R C A_{c, j} > 1), P (R C A_{c, j} > 1 | R C A_{c, i} > 1)},

where,

R C A_{c, i} = \frac{E x p o r t V a l u e_{c, i} / \sum_{i} E x p o r t V a l u e_{c, i}}{(\sum_{c} E x p o r t V a l u e_{c, i} / \sum_{c, i} E x p o r t V a l u e_{c, i})} .

RCA_c,i denotes the RCA of industry i in city c. This index calculates the relatedness between two industries by computing the conditional probability of two industries being exported by the same city. P(RCA_c,i > 1|RCA_c,j > 1) represents the conditional probability of city c having an RCA in industry i, provided that it also specializes in industry j. The relatedness between industry i and j is thus the minimum value of two conditional probabilities: P(RCA_c,i > 1|RCA_c,j > 1) and P(RCA_c,j > 1|RCA_c,i > 1). The idea is that if the relatedness between two industries is high, they probably demand similar input factors, resources, knowledge, and technologies and tend to be manufactured and exported together. Based on these two equations, we can calculate the relatedness index for 1,080 four-digit industries at China’s prefectural city level.¹

Furthermore, we compute the density index also developed by Hidalgo et al. (2007), which is the average relatedness between industry i and all industries in which city c has an RCA. This index thus measures the extent to which industry i is related to city c’s industrial profile. Recent empirical studies suggest that the density of industry i in city c should be high if city c has an RCA in many industries that are related to industry i. In this scenario, the likelihood of city c diversifying into industry i should also be large. The formula to compute the density index is as follows:

d_{i, c} = \frac{\sum_{j} x_{j, c} ϕ_{i, j}}{\sum_{j} ϕ_{i, j}},

where x_j,c is a dummy, taking the value of 1 if city c has an RCA in industry j and 0 otherwise. This index should be large if a city has an RCA in most of the industries related to the industry under consideration.

Model Specifications

We estimate the following model to examine whether HSR has enabled Chinese regions to diversify their industrial structures in a more pathbreaking or path-dependent way, due to the ease of communication between regions.

E n t r y_{i, c, t 2} = β_{0} + β_{1} D e n s i t y_{i, c, t 1} + β_{2} H S R_{c, t 1} + β_{3} D e n s i t y_{i, c, t 1} \times H S R_{c, t 1} + β_{4} X + γ Y + ε_{i, c, t},

where t2 is 2011 and t1 is 2009.² The research focuses on 1,080 four-digit industrial sectors and their entry into China’s 337 prefecture-level cities. The dependent variable is industry entry. If the RCA indicator of industry i in city c is below 1 in year t1 but above 1 in year t2, industry i is defined as an industry entry in city c in year t2. In this case, the dependent variable takes the value of 1. Entry is equal to 0, if the RCA indicator of industry i in city c is below 1 in both year t1 and t2. HSR_c,t ₁ represents indicators capturing the accessibility of city c in the HSR network in year t1. X represents control variables. Specifically, PGDP is added to control the impact of regional economic development on new industry entry. Some other localized capabilities may also affect regional industrial dynamics. HCAP, defined as the share of the population with more than secondary schooling in city c, is included to control the impact of human capital. We also control the research and development (R&D) intensity (R&D), by computing the ratio of R&D investments of industry i to its output in city c. Industrial dynamics in a city may be also shaped by the city’s accessibility in the network of other types of transport infrastructure. Hence, we include the density of highway in a city (Highway), measured as the length of highways over land area in city c to control the effect of the highway network. Data on these variables are derived from China’s City Statistical Yearbooks. Y denotes a vector of region dummy variables at China’s provincial level and industry dummy variables at the two-digit level to control regional and industry characteristics, respectively.

High level of Density_i,c,t means that industry i is closely related to city c’s preexisting industrial profile. If the impact of density is positive, it indicates that regions tend to diversify into related industries and regional industrial evolution is a path-dependent process. However, the accessibility of a region in the HSR network may facilitate knowledge diffusion via extra-regional linkages and thus change the region’s capability to diversify as well as its reliance on relatedness between industries in the process of regional industrial diversification. Therefore, the extent to which regional industrial diversification is shaped by density should vary across space, since different regions often have different levels of accessibility in the HSR network. To test our hypothesis, several interaction terms between HSR variables and the density index are included in our model. If the parameter of an interaction term is positive and significant, it is fair to argue that specific HSR variables strengthen regions’ reliance on relatedness in the process of industrial evolution. On the contrary, a negative and positive parameter of the interaction term implies that industrial relatedness has a weaker impact on regional industrial diversification. In other words, regions’ accessibility in the HSR network reduces their reliance on relatedness and allows them to diversify to less related industries, resulting in more pathbreaking regional industrial diversification. Finally, if the parameter is not significant, it shows that regions’ accessibility in the HSR network cannot change their capability to diversify as well as their reliance on relatedness in regional industrial diversification. The effect of density thus does not vary across space due to regions’ different levels of accessibility in the HSR network.

HSR Accessibility Indicators

The first HSR accessibility indicator employed in this article is a dummy, HSR1, taking the value of 1 if city c has a HSR station in year t1. This indicator is, however, too crude and cannot reflect the network character of transport infrastructure. We thus bring in three classical accessibility indicators: weighted average travel time (WATT), daily accessibility (DA), and potential accessibility (PA; Cao et al. 2013; Wang et al. 2016). WATT captures the average travel time from city c to all others. The size of other cities, measured by GDP and population, is taken into account in the calculation of WATT.

W A T T_{c} = \frac{\sum_{c^{'} = 1}^{n} (T_{c c^{'}} \times M_{c^{'}})}{\sum_{c^{'} = 1}^{n} M_{c^{'}}},

where WATT_c is the WATT of city c. T_cc _′ is the minimum travel time between city c and c′. M_c _′ represents the mass of city c′, calculated as the square root of the product of the city’s population and GDP.

M_{c^{'}} = \sqrt{P_{c^{'}} \times G_{c^{'}}},

where P_c _′ is the resident population in city c′ and G_c _′ is the GDP of city c′. The DA indicator computes the size of population or economic activities that can be reached from city c within a certain amount of time. It is DA, since the time limit is often set at three hours, making a daily round trip possible (Cao et al. 2013; Wang et al. 2016). DA, as a measurement of the population and economy, which can be reached from a certain region in a certain amount of time, can be calculated as follows:

D A_{c} = \sum_{c^{'} = 1}^{m} P_{c^{'}} δ_{c c^{'}},

where DA_c is the DA of city c. The term δ _cc’ is a dummy, taking the value of 1 if T_cc _′ is below three hours and otherwise 0.

The PA indicator is gravity based, capturing the closeness of potential economic activities to city c. It is computed as the weighted sum of other cities’ mass, and the weight is the inverse of the distance between the city under consideration and other cities. The PA of city c is calculated as follows:

P A_{c} = \sum_{c^{'} = 1}^{n - 1} \frac{M_{c^{'}}}{T_{c c^{'}}},

where M_c _′ and T_cc _′ are the same as defined in equation (5). Note that the larger the DA and PA indicators, the greater the accessibility level of the city. In contrast, a high level of the WATT indicator indicates poor accessibility.

Data on GDP and population can be easily extracted from China’s City Statistical Yearbooks. T_cc _′ is calculated using data from China’s official railway timetable and the website of China’s HSR (http://www.nra.gov.cn/ztzl/hyjc/gstl_/). It is easy to obtain the minimum travel time between two cities with direct railway service. For those pairs of cities that are not directly connected on the railway network, Dijkstra algorithm is adopted to compute the shortest travel time between them (Dijkstra 1959). The transfer time is defined as two hours for each transfer. A travel time matrix can be calculated for China’s city pairs. Note that the CR and HSR networks are combined together in the calculation of the minimum travel time between a pair of cities. Accessibility in the HSR network should be also influenced by its integration with the CR network. Cities that have no access to the HSR network may still be affected indirectly by the HSR network, as they may be linked to the HSR network via the CR network.

Econometric Strategy

To estimate causal effects of transport infrastructure on regional industrial dynamics, we need to cope with the endogeneity issue, as railways may be assigned to regions based on some unobserved determinants of economic activities. For instance, infrastructures like railways may be targeted to developed regions where there is perceived need, and such regions may be more capable to bring in new industries and more likely to develop in a pathbreaking way. In other words, industrial dynamics and the network character of transport infrastructure in a region could be simultaneously determined by the region’s preexisting economic development (Ahlfeldt and Feddersen 2018; Faber 2014; Heuermann and Schmieder 2018). This is a conventional endogeneity problem, and here we propose three strategies to deal with it. First, we add city c’s PGDP in year t1 in equation (4) to control the influence of preexisting regional economic development.

Second, we take advantage of quasi-random variation in the 1970 railway network. China’s peculiar historical context enables us to develop plausibly exogenous instrumental variables for transport networks based on China’s railway network in 1970. In 1970, before the economic reform, China was still characterized by the planning economy. Railroads were mostly built out of political, military, and institutional concerns, in order to move raw materials, resources, and products among various cities according to the dictates of national and provincial five-year plans. We use the 1970 railway network to formulate an instrumental variable (IV), which is a dummy taking the value of 1 if city c had a railway station back in 1970 and otherwise 0. The rationale for this instrument is that 1970 railroads were established for some other reasons but were upgradable to modern railways at relatively lower cost. In other words, the 1970 railway network influences the current regional industrial dynamics in Chinese regions only through the current railway network. The data for the 1970 railway network are derived from the Maps of China’s Railways 1876–1980 published by the Cartography Press in 1984. In addition, to test the validity of the instrumental variable, we include the PGDP in the IV regressions. If the 1970 railway network could affect the current regional industrial dynamics not only through the current railway network but also through the current economic development of Chinese regions, the coefficients of key variables (HSR accessibility indicators) would change drastically and become statistically insignificant when we include PGDP in the model. Finally, the correlation coefficient between PGDP and HSR1 in 2009 is .19, while that between PGDP and our IV is only .14. This implies that we may face endogeneity issue while estimating the casual effect of the accessibility indicators on regional industrial dynamics, whereas the issue is less serious if we develop the IV based on the 1970 railway network.

Our last strategy is to include only two groups of cities in the sample: cities that already had HSR stations in 2009 and cities that did not have HSR stations in 2009 but would build stations in the near future on the basis of the Mid- and Long-term Railway Development Plan released in 2004 and revised in 2008 (Figure 2). The rationale for this treatment is that the second group of cities must share something in common with the first group of cities in terms of economic development, since the former was picked to build new HSR stations. However, the influence of the planned HSR stations on the second group of cities had not been felt, as knowledge cannot diffuse via the planned HSR network. Such a comparison thus enables us to better capture the influence of transport infrastructure on regional industrial dynamics.

Figure 2.

China’s existing (top) and planned (bottom) high-speed railway network in 2009.

Empirical Results

In the estimated equation, the logarithm of PGDP is adopted. Correlation analysis in Online Appendix Table A1 suggests that there is no serious issue of multicollinearity. The LOGIT model has been used to estimate equation (4), since the dependent variable is a dummy. Due to the multilevel structure of our model, we cluster standard errors at the two-digit industry and city level. Empirical results in Table 1 show a positive relationship between density and new industry creation and that the effects are highly statistically significant (model 1). This resonates with the findings of other studies, confirming that regional industrial evolution is path dependent (Boschma, Minondo, and Navarro 2013; Colombelli, Krafft, and Quatraro 2014; Neffke, Henning, and Boschma 2011). The parameter of HSR1 is also positive and significant, indicating that having HSR stations raises the probability of the city entering a new industry. With the help of knowledge diffusion via the HSR network, regions could expand the diversity of their knowledge bases and resources and become more capable of bringing in new industries. Three control variables, HCAP, R&D, and Highway, all have significant and positive effects, suggesting that regions with higher levels of human capital, R&D intensity, and accessibility in the highway network tend to be more capable of developing a comparative advantage in a new industry. Surprisingly, PGDP, however, has a negative impact on new industry creation. One possible explanation is that developed regions, particularly those in Chinese coastal region, have already developed most industries. It is thus relatively difficult for such regions to create new industries, as there are only a limited number of new industries available for them to diversify into.

Table 1.

Estimation Results: All Regions (LOGIT).

	HSR1			WATT		DA		PA
	1	2	3	4	5	6	7	8	9
Density	7.799*** (0.417)	9.884*** (0.629)	9.896*** (0.635)	162.1*** (15.36)	164.2*** (15.03)	7.544*** (0.615)	7.594*** (0.613)	−162.8*** (16.10)	−169.8*** (16.14)
lnPGDP	−0.240*** (0.0376)		−0.257*** (0.0379)		−0.403*** (0.0385)		−0.103*** (0.0325)		−0.299*** (0.0357)
Highway	0.249*** (0.0648)	0.0465 (0.0652)	0.109 (0.0664)	0.495*** (0.0754)	0.572*** (0.0773)	0.389*** (0.0744)	0.404*** (0.0746)	0.706*** (0.0858)	0.758*** (0.0876)
R&D	17.13*** (2.496)	15.20*** (2.544)	16.61*** (2.587)	15.70*** (2.560)	17.99*** (2.625)	15.33*** (2.490)	15.76*** (2.483)	17.53*** (2.608)	19.24*** (2.664)
HCAP	2.271*** (0.435)	3.467*** (0.590)	2.075*** (0.525)	3.339*** (0.424)	1.123*** (0.411)	2.531*** (0.421)	1.987*** (0.461)	3.778*** (0.464)	2.235*** (0.441)
HSR	0.234*** (0.0450)	−0.0187 (0.0765)	−0.0345 (0.0769)	5.458*** (0.380)	5.628*** (0.369)	−0.0700*** (0.00693)	−0.0685*** (0.00692)	−3.570*** (0.276)	−3.622*** (0.272)
HSR × density		1.537** (0.763)	1.635** (0.769)	−21.06*** (2.135)	−21.37*** (2.090)	0.564*** (0.0738)	0.557*** (0.0735)	17.54*** (1.632)	18.26*** (1.636)
Constant	−1.763*** (0.542)	−4.012*** (0.338)	−1.986*** (0.505)	−43.82*** (2.699)	−41.77*** (2.469)	−4.592*** (0.288)	−3.765*** (0.386)	30.32*** (2.761)	33.22*** (2.855)
Region dummies	Y	Y	Y	Y	Y	Y	Y	Y	Y
Industry dummies	Y	Y	Y	Y	Y	Y	Y	Y	Y
Pseudo R ²	.0937	.105	.1064	.1375	.1405	.1198	.12	.1327	.1347
Wald χ²	5,530.76	5,593.23	5,552.86	5,243.73	5,278.1	5,194.24	5,202.62	5,228.45	5,222.84
Probability > χ²	0	0	0	0	0	0	0	0	0
N	214,565	214,565	214,565	214,565	214,565	214,565	214,565	214,565	214,565

Note: HSR represents HSR1 in models 1–3, WATT in models 4 and 5, DA in models 6 and 7, and PA in models 8 and 9. Standard errors are in parentheses. HSR = high-speed railway; WATT = weighted average travel time; DA = daily accessibility; PA = potential accessibility; PGDP = per capita GDP.

*p < .1.

**p < .05.

***p < .01.

In models 2 and 3, the interaction terms between HSR1 and the density index present a positive and significant sign, implying that HSR failed to bring in pathbreaking knowledge and did not result in unrelated industrial diversification in Chinese regions. This is not consistent with our theoretical prediction. We further replace HSR1 with some other accessibility indicators: WATT, DA, and PA (models 4–9). However, the interaction terms between density and the DA and PA indicators have positive and statistically significant signs, while that between density and WATT has a negative and statistically significant sign. This further confirms that extra-regional linkages facilitated by transport infrastructure do not bring into fresh knowledge and enable regions to diversify into unrelated industries. On the contrary, better accessibility forces regions to rely more on relatedness and develop in more path-dependent ways. This finding does not change with or without the inclusion of PGDP in our model. One possible reason for this counterintuitive finding is the endogeneity issue.

Table 2 reports the econometric results of IV regressions. The estimated coefficients of variables in model 1 are unaltered. The coefficients of HSR1 × density and DA × density become statistically significant when we include PGDP in the model (models 2 and 3 and 6 and 7). Furthermore, the parameter of PA × density is negative and statistically significant in both models 8 and 9. These results imply that regions with higher levels of accessibility rely less on relatedness and are more likely to diversify their industrial structures in pathbreaking ways. Similarly, the parameter of the interaction term between WATT and density is positive and significant in both models 4 and 5, indicating regions with lower levels of accessibility rely more on relatedness and tend to be more path dependent. Overall, accessibility has a quite consistent effect in these models, and regions do benefit from knowledge diffusion via the HSR network. The latter pumps new know-how, resources, capabilities that are unrelated to regional preexisting industrial structures into regions, even from distant locations, enabling regions to jump into unrelated, new industries and achieve pathbreaking industrial diversification.

Table 2.

Estimation Results: All Regions (IV).

		HSR1		WATT		DA		PA
	1	2	3	4	5	6	7	8	9
Density	8.517*** (0.428)	15.33*** (3.432)	20.11*** (3.419)	−164.1* (98.34)	−155.8 (99.80)	21.73 (46.32)	96.06** (44.31)	1,422*** (238.6)	1,392*** (240.3)
lnPGDP	−0.385*** (0.0577)		−0.443*** (0.0565)		0.0816 (0.232)		−0.565*** (0.0735)		−0.272*** (0.0385)
Highway	0.296*** (0.0638)	0.0602 (0.0639)	0.110* (0.0654)	0.133** (0.0659)	0.131** (0.0658)	0.0727 (0.0642)	0.114* (0.0655)	0.0318 (0.0642)	0.0925 (0.0654)
R&D	17.88*** (2.469)	14.66*** (2.449)	15.01*** (2.454)	16.63*** (2.540)	16.52*** (2.529)	15.21*** (2.473)	15.04*** (2.448)	13.71*** (2.411)	14.98*** (2.453)
HCAP	2.044*** (0.494)	3.544*** (0.676)	2.195*** (0.562)	2.169*** (0.542)	2.238*** (0.601)	3.276*** (0.665)	2.072*** (0.551)	3.611*** (0.617)	2.160*** (0.544)
HSR	2.130*** (0.503)	1.065** (0.416)	3.222*** (0.520)	−6.671*** (1.077)	−8.379* (5.036)	0.411 (0.566)	5.292*** (0.830)	19.83*** (2.891)	20.22*** (2.869)
HSR × density		−5.162 (3.753)	−10.38*** (3.711)	25.31* (14.24)	24.11* (14.45)	−1.377 (5.756)	−10.60* (5.502)	−137.9*** (23.31)	−135.0*** (23.49)
Constant	−1.908*** (0.519)	−4.734*** (0.437)	−2.579*** (0.539)	41.43*** (7.494)	52.39 (32.51)	−7.261 (4.483)	−41.07*** (5.967)	−206.5*** (29.40)	−208.3*** (29.08)
Region dummies	Y	Y	Y	Y	Y	Y	Y	Y	Y
Industry dummies	Y	Y	Y	Y	Y	Y	Y	Y	Y
Pseudo R ²	.0939	.1048	.1074	.1062	.1062	.1046	.1076	.1066	.108
Wald χ²	5,525.65	5,566	5,483.11	5,552.89	5,552.43	5,583.93	5,486.05	5,505.76	5,468.31
Probability > χ²	0	0	0	0	0	0	0	0	0
N	214,565	214,565	214,565	214,565	214,565	214,565	214,565	214,565	214,565

*p < .1.

**p < .05.

***p < .01.

According to the Mid- and Long-term Railway Development Plan, some cities did not have HSR stations in 2009 but would build stations in the future. Such cities were picked by the state, and they should be similar to cities with existing HSR stations in terms of economic development. However, the impact of the planned HSR station had not been felt, since knowledge cannot diffuse via the planned infrastructure network. The empirical results are shown in Table 3. First, the parameter of HSR1 × density is negative and significant in models 2 and 3, suggesting that the HSR network has helped regions reduce their reliance on relatedness in regional industrial evolution, enabling regions to absorb fresh, new ideas and less related technologies from outside and subsequently to achieve pathbreaking economic development. Again, in models 4–9, we bring in three accessibility indicators. The coefficient of the interaction term between WATT and density is positive and significant in both models 4 and 5, confirming that cities with lower levels of accessibility are more likely to rely on relatedness and develop in a more path-dependent way. In contrast, the coefficient of PA × density is negative and significant in models 8 and 9, implying that higher levels of accessibility may decrease cities’ dependence on relatedness and enable the latter to develop unrelated industries. The econometric results in Table 3 are thus quite consistent. Accessibility in the HSR network plays a role in pathbreaking regional industrial evolution. High levels of accessibility tend to allow regions to accomplish more radical innovation and develop less related industries, which is consistent with the theoretical prediction.

Table 3.

Estimation Results: Regions with Existing and Planned HSR Stations (LOGIT).

		HSR1		WATT		DA		PA
	1	2	3	4	5	6	7	8	9
Density	8.941*** (0.385)	10.99*** (0.694)	11.10*** (0.691)	−32.04** (12.72)	−29.22** (12.65)	9.647*** (0.695)	9.710*** (0.696)	56.69*** (13.47)	53.16*** (13.45)
lnPGDP	−0.0904 (0.0742)		−0.179** (0.0747)		−0.168** (0.0743)		−0.179** (0.0756)		−0.162** (0.0759)
Highway	0.288*** (0.0823)	0.0884 (0.0830)	0.120 (0.0840)	0.0867 (0.0864)	0.120 (0.0872)	0.150* (0.0845)	0.175** (0.0853)	0.0466 (0.0915)	0.0668 (0.0917)
R&D	2.987 (3.212)	1.788 (3.025)	3.321 (3.085)	4.225 (2.898)	5.574* (2.952)	4.090 (2.909)	5.559* (2.961)	2.200 (2.937)	3.455 (2.973)
HCAP	4.246*** (0.450)	0.724* (0.425)	0.0121 (0.536)	0.720 (0.438)	0.0301 (0.549)	0.895** (0.429)	0.210 (0.530)	0.844* (0.433)	0.216 (0.535)
HSR	0.0465 (0.0514)	0.224** (0.0898)	0.215** (0.0905)	−0.414 (0.342)	−0.363 (0.342)	−0.00753 (0.00767)	−0.00880 (0.00770)	0.631*** (0.235)	0.638*** (0.237)
HSR × density		−2.324*** (0.746)	−2.262*** (0.747)	5.829*** (1.777)	5.451*** (1.766)	0.000622 (0.0643)	0.0220 (0.0647)	−4.676*** (1.341)	−4.304*** (1.341)
Constant	−4.213*** (0.682)	−4.874*** (0.258)	−3.359*** (0.677)	−1.714 (2.459)	−0.669 (2.500)	−4.871*** (0.260)	−3.336*** (0.690)	−11.04*** (2.336)	−9.749*** (2.390)
Region dummies	Y	Y	Y	Y	Y	Y	Y	Y	Y
Industry dummies	Y	Y	Y	Y	Y	Y	Y	Y	Y
Pseudo R ²	.0767	.0887	.089	.0888	.089	.0885	.0887	.0888	.089
Wald χ²	3,740.07	3,879.6	3,872.2	3,880.61	3,873.22	3,896.07	3,884.01	3,879.65	3,870.43
Probability > χ²	0	0	0	0	0	0	0	0	0
N	113,520	113,520	113,520	113,520	113,520	113,520	113,520	113,520	113,520

*p < .1.

**p < .05.

***p < .01.

As a robustness check, we reestimate our models by employing different criteria (0.9 and 1.1) to define an RCA (see Online Appendix Tables A2–A7). All models are also estimated at the two-digit industry × city level. The four-digit-level industry data at the city level tend to be affected by some unobservable factors, while data at the two-digit industry × city level are likely to be more stable. The results are shown in Online Appendix Tables A8–A10. Generally speaking, these changes do not generate significant effects.

Conclusion and Discussion

The reduction of transport costs is at the heart of major policy intervention, as it is anticipated to have some positive impacts on various aspects of regional economic development. In this article, we have investigated the construction of China’s HSR network as well as its influence on regional industrial dynamics. Specifically, we have focused on the network characteristics of transport infrastructure and the accessibility of Chinese cities in the network as a policy treatment and conducted an econometric analysis in which regional industrial diversification has been considered to be the main outputs. One main concern of the recent literature on the impact of transport infrastructure on regional economic development is that transport infrastructure is often built to meet existing or expected demand rather than randomly allocated across space. In response to the endogeneity issue, we have adopted several strategies, for instance, relying on planned and historic transport networks. Our empirical results indicate that access to the network of transport infrastructure not only has a positive impact on new industry creation but also enables regions to be more pathbreaking and jump further into less related industries.

Some theoretical studies in EEG have stressed that extra-regional linkages are able to bring in new ideas, resources, and technologies from outside, which are less related or even unrelated to region’s preexisting industrial profile and thus may result in more pathbreaking industrial diversification. This view is also supported by empirical studies in EEG. This article seeks to complement existing theoretical and qualitative literature by developing a quantitative research. Furthermore, this article picks one specific type of extra-regional linkages formulated or reinforced by the HSR network and examines whether the latter has allowed Chinese regions to accomplish unrelated diversification and pathbreaking new industry creation. In doing so, this article not only enables us to better capture the role of infrastructure in economic development but also brings two strands of literature into dialogue: one on EEG and the other on transport infrastructure.

We also stress the importance of the network characteristics of transport infrastructure. Our empirical results confirm that indicators such as whether a region has access to infrastructure networks, which have been widely used in empirical studies, can only reflect one aspect of the quality and conditions of the infrastructure networks. A region’s accessibility in the network that is dependent on relevant infrastructure bottlenecks, travel time, and the number of links should also be taken into account.

Empirically, our article also hopes to show developing regions/countries a more promising future by highlighting that they can break the seemingly dominant path-dependent development model and catch up. One way to do so is to invest in the formulation of extra-regional linkages by developing a network of transport infrastructure that links both developed and underdeveloped regions. Interregional knowledge spillover facilitated by the network of infrastructure may bring new knowledge, resources, and capabilities that are unrelated to regional preexisting industrial structures, enabling laggard regions to jump further into high-end, new industries and catch up. In this sense, regional disparity between and within countries could be narrowed. It seems that the massive construction of the HSR network in China has been also conducted in order to boost economic development in main cities and more recently to alleviate regional disparity. On the one hand, it is argued that HSR services in China tend to favor large cities and developed regions and may hurt regions that are less politically favored or economically prosperous (Qin 2017). On the other hand, such inequality is expected to decline with further development of the HSR network to inland China at the national level (Jiao et al. 2014) and to peripheral areas in each province (Wang et al. 2016) by 2020.

Supplemental Material

Supplemental Material, IRSR_appendix - High-speed Rail Network and Changing Industrial Dynamics in Chinese Regions

Supplemental Material, IRSR_appendix for High-speed Rail Network and Changing Industrial Dynamics in Chinese Regions by Shengjun Zhu, Chong Wang and Canfei He in International Regional Science Review

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Natural Science Foundation of China [grant number 41731278, 41701115, 41425001].

Supplemental Material

Supplemental material for this article is available online.

Notes

References

Ahlfeldt

G. M.

Feddersen

. 2018. “From Periphery to Core: Measuring Agglomeration Effects Using High-speed Rail.” Journal of Economic Geography 18:355–90.

Bathelt

Cohendet

. 2014. “The Creation of Knowledge: Local Building, Global Accessing and Economic Development—Toward an Agenda.” Journal of Economic Geography 14:869–82.

Bathelt

Malmberg

Maskell

. 2004. “Clusters and Knowledge: Local Buzz, Global Pipelines and the Process of Knowledge Creation.” Progress in Human Geography 28:31–56.

Baum-Snow

2007. “Did Highways Cause Suburbanization?*” The Quarterly Journal of Economics 122:775–805.

Baum-Snow

Brandt

Henderson

J. V.

Turner

M. A.

Zhang

. 2017. “Roads, Railroads, and Decentralization of Chinese Cities.” The Review of Economics and Statistics 99:435–48.

Behrens

Murata

. 2012. “Trade, Competition, and Efficiency.” Journal of International Economics 87:1–17.

Bernhofen

Upward

Wang

. 2017. “Quantity Restrictions and Price Adjustment of Chinese Textile Exports to the US.” The World Economy 41(11):2983–3000.

Binz

Truffer

Coenen

. 2016. “Path Creation as a Process of Resource Alignment and Anchoring: Industry Formation for On-site Water Recycling in Beijing.” Economic Geography 92:172–200.

Boschma

2005. “Proximity and Innovation: A Critical Assessment.” Regional Studies 39:61–74.

10.

Boschma

Balland

P.-A.

Kogler

D. F.

. 2015. “Relatedness and Technological Change in Cities: The Rise and Fall of Technological Knowledge in US Metropolitan Areas from 1981 to 2010.” Industrial and Corporate Change 24:223–50.

11.

Boschma

Capone

. 2015. “Institutions and Diversification: Related Versus Unrelated Diversification in a Varieties of Capitalism Framework.” Research Policy 44:1902–14.

12.

Boschma

Frenken

. 2006. “Why Is Economic Geography Not an Evolutionary Science? Towards an Evolutionary Economic Geography.” Journal of Economic Geography 6:273–302.

13.

Boschma

Frenken

. 2011. “Technological Relatedness and Regional Branching.” In Beyond Territory: Dynamic Geographies of Knowledge Creation, Diffusion and Innovation, edited by Bathelt

Feldman

Kogler

D. F.

, 64–81. London, UK: Routledge.

14.

Boschma

Iammarino

. 2009. “Related Variety, Trade Linkages, and Regional Growth in Italy.” Economic Geography 85:289–311.

15.

Boschma

Martin

. 2007. “Editorial: Constructing an Evolutionary Economic Geography.” Journal of Economic Geography 7:537–48.

16.

Boschma

Martín

Minondo

. 2017. “Neighbour Regions as the Source of New Industries.” Papers in Regional Science 96:227–45.

17.

Boschma

Minondo

Navarro

. 2012. “Related Variety and Regional Growth in Spain.” Papers in Regional Science 91:241–56.

18.

Boschma

Minondo

Navarro

. 2013. The Emergence of New Industries at the Regional Level in Spain: A Proximity Approach Based on Product Relatedness.” Economic Geography 89:29–51.

19.

Cao

Liu

X. C.

Wang

. 2013. “Accessibility Impacts of China’s High-speed Rail Network.” Journal of Transport Geography 28:12–21.

20.

Chen

C.-L.

2012. “Reshaping Chinese Space-economy through High-speed Trains: Opportunities and Challenges.” Journal of Transport Geography 22:312–16.

21.

Cheng

Y.-S.

Loo

B. P. Y.

Vickerman

. 2015. “High-speed Rail Networks, Economic Integration and Regional Specialisation in China and Europe.” Travel Behaviour and Society 2:1–14.

22.

Colombelli

Krafft

Quatraro

. 2014. “The Emergence of New Technology-based Sectors in European Regions: A Proximity-based Analysis of Nanotechnology.” Research Policy 43:1681–96.

23.

Cosci

Mirra

. 2018. “A Spatial Analysis of Growth and Convergence in Italian Provinces: The Role of Road Infrastructure.” Regional Studies 52:516–27.

24.

Crescenzi

Rodríguez-Pose

. 2012. “Infrastructure and Regional Growth in the European Union.” Papers in Regional Science 91:487–513.

25.

Delgado

Porter

M. E.

Stern

. 2016. “Defining Clusters of Related Industries.” Journal of Economic Geography 16:1–38.

26.

Dijkstra

E. W.

1959. “A Note on Two Problems in Connexion with Graphs.” Numerische Mathematik 1:269–71.

27.

Duranton

Morrow

P. M.

Turner

M. A.

. 2014. “Roads and Trade: Evidence from the US.” The Review of Economic Studies 81:681–724.

28.

Duranton

Turner

M. A.

. 2012. “Urban Growth and Transportation.” The Review of Economic Studies 79:1407–40.

29.

Faber

2014. “Trade Integration, Market Size, and Industrialization: Evidence from China’s National Trunk Highway System.” The Review of Economic Studies 81:1046–70.

30.

Frenken

Van Oort

Verburg

. 2007. “Related Variety, Unrelated Variety and Regional Economic Growth.” Regional Studies 41:685–97.

31.

Gerritse

Arribas-Bel

. 2017. “Concrete Agglomeration Benefits: Do Roads Improve Urban Connections or Just Attract More People?” Regional Studies 52(8):1134–1149.

32.

Grillitsch

Trippl

. 2014. “Combining Knowledge from Different Sources, Channels and Geographical Scales.” European Planning Studies 22:2305–25.

33.

Heuermann

D. F.

Schmieder

J. F.

. 2018. “The Effect of Infrastructure on Worker Mobility: Evidence from High-speed Rail Expansion in Germany.” Journal of Economic Geography.

34.

Hidalgo

C. A.

Klinger

Barabási

A.-L.

Hausmann

. 2007. “The Product Space Conditions the Development of Nations.” Science 317:482–87.

35.

Hou

S.-M.

. 2011. “Transport Infrastructure Development and Changing Spatial Accessibility in the Greater Pearl River Delta, China, 1990–2020.” Journal of Transport Geography 19:1350–60.

36.

Iacono

Levinson

. 2016. “Mutual Causality in Road Network Growth and Economic Development.” Transport Policy 45:209–17.

37.

Jacobs

1969. The Economy of Cities. New York: Vintage.

38.

Jaffe

A. B.

Trajtenberg

Henderson

. 1993. “Geographic Localization of Knowledge Spillovers as Evidenced by Patent Citations.” The Quarterly Journal of Economics 108:577–98.

39.

Jiao

Wang

Jin

Dunford

. 2014. “Impacts on Accessibility of China’s Present and Future HSR Network.” Journal of Transport Geography 40:123–32.

40.

Klepper

2002. “The Capabilities of New Firms and the Evolution of the US Automobile Industry.” Industrial and Corporate Change 11:645–66.

41.

Klepper

Simons

K. L.

. 2000. “Dominance by Birthright: Entry of Prior Radio Producers and Competitive Ramifications in the U.S. Television Receiver Industry.” Strategic Management Journal 21:997–1016.

42.

Chen

B. R.

. 2017. “Is Road Infrastructure Investment in China Excessive? Evidence from Productivity of Firms.” Regional Science and Urban Economics 65:116–26.

43.

Marshall

[1890] 1920. Principles of Economics, 8th ed. London, UK: Macmillan.

44.

Martin

2010. “Roepke Lecture in Economic Geography—Rethinking Regional Path Dependence: Beyond Lock-in to Evolution.” Economic Geography 86:1–27.

45.

Martin

Sunley

. 2006. “Path Dependence and Regional Economic Evolution.” Journal of Economic Geography 6:395–437.

46.

Maskell

2014. “Accessing Remote Knowledge—The Roles of Trade Fairs, Pipelines, Crowdsourcing and Listening Posts.” Journal of Economic Geography 14:883–902.

47.

Maskell

Bathelt

Malmberg

. 2006. “Building Global Knowledge Pipelines: The Role of Temporary Clusters.” European Planning Studies 14:997–1013.

48.

Maskell

Malmberg

. 1999. “The Competitiveness of Firms and Regions: ‘Ubiquitification’ and the Importance of Localized Learning.” European Urban and Regional Studies 6:9–25.

49.

Melo

P. C.

Graham

D. J.

Brage-Ardao

. 2013. “The Productivity of Transport Infrastructure Investment: A Meta-analysis of Empirical Evidence.” Regional Science and Urban Economics 43:695–706.

50.

Neffke

Hartog

Boschma

Henning

. 2018. “Agents of Structural Change: The Role of Firms and Entrepreneurs in Regional Diversification.” Economic Geography 94:23–48.

51.

Neffke

Henning

Boschma

. 2011. “How Do Regions Diversify over Time? Industry Relatedness and the Development of New Growth Paths in Regions.” Economic Geography 87:237–65.

52.

Nooteboom

2000. Learning and Innovation in Organizations and Economies. Oxford, UK: Oxford University Press.

53.

Percoco

2016. “Highways, Local Economic Structure and Urban Development.” Journal of Economic Geography 16:1035–54.

54.

Qin

2017. “‘No County Left Behind?’ The Distributional Impact of High-speed Rail Upgrades in China.” Journal of Economic Geography 17:489–520.

55.

Ryder

2012. “High Speed Rail.” Journal of Transport Geography 22:303–5.

56.

Schumpeter

J. A.

1939. Business Cycles: A Theoretical, Historical, and Statistical Analysis of the Capitalist Process. New York: McGraw-Hill.

57.

Schumpeter

J. A.

1942. Capitalism, Socialism and Democracy. New York: Harper & Row.

58.

Scott

1988. New Industrial Spaces: Flexible Production Organization and Regional Development in North America and Western Europe. London, UK: Pion.

59.

Shaw

S.-L.

Fang

Tao

. 2014. “Impacts of High Speed Rail on Railroad Network Accessibility in China.” Journal of Transport Geography 40:112–22.

60.

Simmie

2012. “Path Dependence and New Technological Path Creation in the Danish Wind Power Industry.” European Planning Studies 20:753–72.

61.

Storper

Walker

. 1989. The Capitalist Imperative: Territory, Technology and Industrial Growth. New York: Basil Blackwell.

62.

Tanner

A. N.

2014. “Regional Branching Reconsidered: Emergence of the Fuel Cell Industry in European Regions.” Economic Geography 90:403–27.

63.

Tanner

A. N.

2016. “The Emergence of New Technology-based Industries: The Case of Fuel Cells and Its Technological Relatedness to Regional Knowledge Bases.” Journal of Economic Geography 16:611–35.

64.

UIC. 2018. High Speed Line in the World. Accessed May 5, 2018. https://uic.org/highspeed.

65.

Wang

Liu

Sun

Liu

. 2016. “Accessibility Impact of the Present and Future High-speed Rail Network: A Case Study of Jiangsu Province, China.” Journal of Transport Geography 54:161–72.

66.

Zhu

Zhou

. 2017. “How to Jump Further and Catch Up? Path-breaking in an Uneven Industry Space.” Journal of Economic Geography 17:521–45.

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

0.26 MB