EN中文

Journal of Geographical Sciences >

2024 , Vol. 34 >Issue 4: 779 - 803

DOI: https://doi.org/10.1007/s11442-024-2227-x

Research Articles

Network and mechanism of China’s new energy vehicle industry from the perspective of value chain

  • HE Yao , 1 ,
  • YANG Yongchun , 1, 2, * ,
  • WANG Shaobo 3
Expand
  • 1. College of Earth and Environmental Sciences, Lanzhou University, Lanzhou 730000, China
  • 2. Key Laboratory of Western China's Environmental Systems, Ministry of Education of the People’s Republic of China, Lanzhou University, Lanzhou 730000, China
  • 3. Institute of County Economic Development, Lanzhou University, Lanzhou 730000, China
* Yang Yongchun, Professor, E-mail:

This paper is initially published in Acta Geographica Sinica (Chinese edition), 2023, 78(12): 3018‒3036.

He Yao, PhD Candidate, specialized in urban and regional development planning. E-mail:

Received date: 2023-12-09

  Accepted date: 2024-01-20

  Online published: 2024-04-24

Supported by

National Natural Science Foundation of China(41971198)

The Second Tibetan Plateau Scientific Expedition and Research Program(2019QZKK1005)

Fold

Abstract

Based on the data of listed companies in the core industry chain of China's new energy vehicles in 2015 and 2021, this paper constructs their industrial network from the perspective of the value chain, and uses methods such as social network and negative binomial regression model to study the characteristics, evolution, differences, and formation mechanisms of different value chain networks. The results show that: (1) R&D-oriented, production-oriented, and service-oriented networks share several common features: These networks are simultaneously expanding in scale and transitioning towards more efficient “small world” network; The degree distribution in these networks follows a power-law distribution, indicating a scale-free network structure; There is a decrease in the power-law exponent of network’s degree distribution, indicating an increase in network heterogeneity. Furthermore, there is a significant positive correlation between the degrees of nodes in networks with diverse value chains, suggesting that the same node holds a similar level of significance across different networks. (2) The number of power-prestige, power and prestige nodes increases in the networks of all value chain segments, except in the service-oriented network, where there are no power nodes. In each value chain network, these nodes have different agglomeration directions: In R&D-oriented network, the nodes tend to cluster around headquarters and high-level cities. In contrast, service-oriented network shows a concentration of nodes in municipalities, sub-provincial and provincial capitals. Similarly, production-oriented network demonstrates a clustering of nodes in traditional production bases. (3) Different value-added segments of industry form different types of agglomeration in pursuit of different factor endowments and agglomeration effect, and form the spatial structure of the strongest connection industrial network with different characteristics. The R&D-oriented networks have always been an integrated and closely connected multiple core-periphery structure community with the influence of social, technological and geographical proximities; Transformation of service-oriented network from an integrated and closely connected multiple core-periphery structure community to a multiple core-semi-periphery-periphery structure community with the influence of social, geographical and institutional proximities; Transformation of production-oriented network from the partially integrated and localized core-periphery structure community to the more decentralized multiple independent core-periphery structure community with the influence of the social, institutional of administrative boundaries and geographical proximities.

Key words： value chain; new energy vehicle industry; listed company; industrial network; core-periphery structure

Cite this article

HE Yao , YANG Yongchun , WANG Shaobo . Network and mechanism of China’s new energy vehicle industry from the perspective of value chain[J]. Journal of Geographical Sciences, 2024 , 34(4) : 779 -803 . DOI: 10.1007/s11442-024-2227-x

1 Introduction

The study of industrial networks involves utilizing network methods to investigate various relationships between industries, either inter-industry or intra-industry (Huang and Li, 2006). Network construction varies based on research needs, primarily using spatial units like cities, provinces, or countries as nodes. Complex industrial relationships, such as supply chains and intra-industry enterprises, are used to build these networks (Taylor et al., 2002; Tang et al., 2019; Jin, 2020; Li et al., 2022; Wang et al., 2022a; Zhan and Gu, 2022). The value chain represents the complete combination of processes in which enterprises create value (Zhou, 2021). Owing to varying proportions of input factors at different stages, regions with diverse factor endowments gain locational cost advantages in various production stages, leading to the segmentation of value chains across regions. Each region functions as a spatial carrier for enterprises engaged in distinct value-added tasks (Ren, 2011; Liu et al., 2021). Therefore, from the perspective of the value chain, under labor spatial division, different value-added segments within an industry will choose locations based on their respective adaptability to production geography. For similar value-added segments (such as research and development (R&D), production, or services), there is a typical mapping relationship through the organization and layout of production networks in similar regions (Liu et al., 2021). This results in the formation of an industrial network structure under the value chain system. Existing studies indicate that the global division of labor in high-added manufacturing industries is generally higher in developed countries than in developing countries (Liang and Liang, 2011). If regions within a country can follow endowment advantages for division of labor, nurturing, establishing, and integrating high-quality value chains can facilitate their ascent in the global value chain (Yuan et al., 2019). It becomes crucial to identify regional divisions of labor and improve the domestic value chain. The production segmentation of the value chain can facilitate the partial transfer of industries, enabling peripheral or underdeveloped regions to participate in value chain tasks (Zhang and Li, 2009). This promotes economic development. Simultaneously, the value-added segments in each region undergo changes with the development of industries within the region. The functional, value, and usage aspects of their spatial locations also evolve (Zhou, 2021). This evolutionary process represents the continuous internalization of external economies. Industries and enterprises discover, change, and reshape spatial value-driven industrial network structures through spatial clustering.
As China deepens its integration into international division of labor, the phased characteristics of regional division of labor positions in the Chinese automotive industry production system become apparent. From the Northeast Changchun First Automobile Manufacturing Plant, which initially introduced Soviet technology before the 1978 reform and opening up, to allow the establishment of joint ventures in key eastern cities like Beijing Jeep Automobile Co., Ltd., SAIC Volkswagen Automobile Co., Ltd., and Guangzhou Honda Automobile Co., Ltd., and then to the development of the automotive manufacturing industry in third and fourth-tier cities prompted by rigid constraints on the localization rate of components and the reduction of joint venture policy restrictions (Fei, 2019), the regions participating in the division of labor in the automotive manufacturing industry have gradually increased. Along with the maturation of the domestic automotive market, emphasis has been placed on research and sales. High-level cities where headquarters are located, such as Shanghai and Beijing, have established research and sales centers to achieve value upgrades. Due to considerations of environmental and energy security, China has initiated the development of the new energy vehicle industry. New energy vehicles, as a crucial direction for electrification and decarbonization of automobiles, hold profound significance for enhancing industry competitiveness, improving future energy structures, and developing low-carbon transportation. With the evolution of new energy vehicle technology, the Chinese automotive industry continuously explores new development spaces. Domestic brands such as BYD and NIO are gradually maturing, propelling cities like Xi’an and Hefei to hold crucial positions in the division of labor within the new energy vehicle industry. The demand for new products, such as lithium batteries, has also positioned cities with battery production capabilities and raw materials like Ningde and Ganzhou at important junctures, achieving value upgrades. Major developed countries and regions worldwide consider new energy vehicles a crucial strategic direction for future development, accelerating their industrial layout. The wave of new energy vehicles is propelling a rapid restructuring of the automotive industry’s value chain. On the other hand, the automotive manufacturing industry exhibits strong positive clustering effects (Beaudry and Swann, 2001). Under industrial spatial clustering and large-scale production, individual automotive industry clusters are formed. In the 1980s, the Chinese automotive industry tended to be decentralized, concentrated in the 1990s. Decentralized again in the early 21st century, this is mainly due to the decisive and pioneering influence of the level of clustering in the component and accessory manufacturing industry on the automotive manufacturing industry (Zhao et al., 2014). With the shift in the automotive industry’s production methods from “Fordism” to “platform-based” and further development into “modularization,” the degree of local clustering has decreased. However, the organizational structure of reconfigured regional production networks has created a competitive advantage that transcends local boundaries (Zhao et al., 2021). With the shift in the automotive industry’s production methods from “Fordism” to “platform-based” and further development into “modularization,” the degree of local clustering has decreased. However, the organizational structure of reconfigured regional production networks has created a competitive advantage that transcends local boundaries (Zhao et al., 2021). The restructuring of the industrial value chain and changes in production methods will reshape the domestic automotive industry’s network structure and layout, leading to regional population migration and industrial restructuring, driving regional economic development and spatial structural evolution (Han and Pan, 2005).
Some scholars believe that in the contemporary economy, the most significant value creation and acquisition primarily come from the production of intangible goods rather than tangible goods and standardized services. In the value chain, knowledge-intensive activities in the upstream and downstream, such as research and development, brand management, distribution, and after-sales services, create and acquire value significantly higher than production activities (Mudambi, 2008). Enterprise investment activities and the “segmentation” of value chains shift various functions to potentially more profitable locations (Gereffi et al., 2005; Dicken, 2011). Economic geographers and economists have demonstrated that high-value-added, knowledge-intensive activities and corporate control often concentrate in more developed core regions, while low-value-added production networks tend to concentrate in less developed peripheral areas (Massey, 1979; Hymer, 1992; Dicken, 2011). The situation in the automotive industry remains the same. For example, Sturgeon et al. based on the relationships between car manufacturers and their major suppliers, outlined the spatial structure of automotive production networks. They argued that to ensure timely delivery and leverage economies of scale and low labor costs, production processes tend to regionalize, while research and development concentrate in a few centers (Sturgeon et al., 2008) In a case study of the BMW Group, Coe et al. (2004) mentioned that its production network in the European Union and ASEAN regions reveals regional connections between suppliers involved in production, manufacturing, and assembly in Bavaria and Rayong Province, Thailand. However, these studies primarily focus on a few leading enterprises and their suppliers, emphasizing the suppliers’ headquarters and overlooking branch studies. The true spatial significance of production layout lies in the branches of each supplier. This is also true in studies of the Chinese automotive production network. For example, Zhao et al. (2022) analyzed the case of FAW-Volkswagen, examining the cross-domain associations and influencing factors of China’s automotive production network from a “global-local” perspective, as well as the spatial reorganization of China’s automotive industry clusters under modular production. Meanwhile, domestic research (Wang and Wang, 2015; Cong et al., 2021; Zhao et al., 2021; Lin and Sun, 2022; Zhao et al., 2022) has overlooked the value segmentation in automotive production networks, whereas the spatial distribution of economic activities with different potentials for value creation and acquisition is crucial for regional development (Pavlínek and Ženka, 2016). Secondly, in situations where foreign government intervention is low, leading companies control the development trajectory of their domestic automotive industries. For example, in the Czech Republic (Natsuda et al., 2022) and Turkey (Özatağan, 2011), component suppliers play a significant role. However, under China’s regulatory policies, leading companies have no alternative (Godfrey, 2019; Schwabe, 2020), limiting their control over the development of the domestic automotive industry. In such situations, the analysis focusing only on individual leading companies exhibits significant limitations. Finally, current studies on the layout of the automotive industry production network often stay at the analysis of traditional automotive patterns. For example, Pavlínek (2020; 2021) drew inspiration from Harvey’s concepts of uneven development and spatial fix to conceptualize the geographical expansion and restructuring of the European automotive industry. Additionally, using EU automotive trade data, he depicted the core-semi-peripheral-peripheral structure of the European automotive industry from 2003 to 2017 based on production network and value chain analyses. Kuroiwa et al. (2022) utilized new data from companies in the Thai automotive industry to analyze the localization of automotive parts suppliers and assembly companies in Thailand. Currently, there is relatively limited research on the formation mechanisms and differences in the geographical connections within the layout of the new energy vehicle industry and across different value chain segments.
Therefore, based on the value chain perspective, this paper constructs the organizational network of China’s new energy vehicle core industry with cities as nodes through A-share and Hong Kong-listed companies. It employs methods such as social network analysis and negative binomial regression models to reveal the evolutionary characteristics and differences in the industrial network across different value chain segments of China’s new energy vehicle industry. Additionally, this paper analyzes the formation mechanisms of the network structure from the perspectives of agglomeration and multidimensional proximity. In terms of mechanism exploration, agglomeration theory and multidimensional proximity theory provide a rich theoretical foundation for explaining the mechanisms of agglomeration and geographical connections across different value chain segments. This allows for a comprehensive analysis of the spatial agglomeration and spatial connections in the formation mechanism of the new energy vehicle industry. Regarding data, the analysis includes the headquarters and branches of all domestically listed companies in the A-share and Hong Kong-listed companies of the new energy vehicle core industry. This approach goes beyond a limited number of leading automobile manufacturers and their supplier headquarters, aiming to encompass more domestic cities with production branches. It provides a clear depiction of the overall layout and evolution of China’s new energy vehicle industry. In terms of practical implications, the new energy vehicle industry is one of the directions for high-quality development in cities and a focal point of competition for industrial development among various cities. Analyzing the network structure of China’s new energy vehicle industry from the perspective of the value chain enhances the understanding of the regional development of the industry. It identifies the division of labor and advantages of different cities in various value networks, analyzes the formation mechanism of the network structure, and promotes the value upgrading of relevant cities as well as the coordinated development among cities. This provides a scientific basis for the rational configuration of the new energy vehicle industry chain and the improvement of the domestic new energy vehicle value chain.

2 Research methods and data sources

2.1 Data collection and processing

The data collection and processing in this paper involve the following steps: (1) Based on the Wind database, Guosen Securities, and Changjiang Securities, the relevant data and information of the core industry chain of new energy vehicles are collected. As of the end of 2021, the total stock of new energy vehicles nationwide reached 7.84 million, with pure electric vehicles accounting for 6.4 million, representing 81.63% of the total. Therefore, the analysis focuses on the core industry chain of new energy electric vehicles with the three core components as the core, including processes such as lithium mining, rare earths, other metals, electrolytes, positive electrode materials, negative electrode materials, separators, permanent magnet materials, batteries, motors, electronic controls, thermal management, lightweighting, wiring harnesses, inverters, transmissions, relays, semiconductors, vehicle production, charging stations and other links. (2) Based on the Wind database, a list of A-share and Hong Kong-listed companies is obtained. Reference is made to the company list in the core industry chain of new energy vehicles provided by the Tianfeng Securities automotive team, as well as the enterprise introductions, business scopes, product types, and production product lists from the official websites of each company (for products that have production capacity and are used in new energy electric vehicles). Representative listed companies in the core segments were collected, with 190 companies in 2015 and 270 in 2021. (3) According to Qichacha, information on the controlling subsidiaries of listed companies is collected, including their business scope, industry, and location. After filtering out subsidiaries that are deregistered, relocated, or have completely unrelated business scopes, a final dataset was obtained, comprising 2415 subsidiary companies related to the relevant listed companies in 2015 and 6157 subsidiary companies in 2021. The data does not currently include the Hong Kong, Macao, and Taiwan regions. The required data for urban permanent population and GDP are sourced from Zhongjing Data and the 2021 National Economic and Social Development Statistics Bulletin for each city. City patent data is obtained from the 2021 China Patent Announcement published by the National Intellectual Property Administration.

2.2 Research methods

(1) Construction of Industry Network: This paper constructs an industrial network with cities as nodes based on the organizational relationships among enterprises across regions. The network organizational relationships between enterprises include internal organizational connections and business connections between enterprises. Due to the confidentiality of business connection data between enterprises, which is difficult to obtain, and the ease of access and relative stability of internal organizational connection data, most scholars analyze based on the connections between enterprises internally (Zhao et al., 2019). Therefore, this paper, based on the correspondence quantity relationship between the headquarter and branch locations of listed companies, borrows from the chain-lock model to construct a directed weighted network. Combining Defever’s proposed functional classification standards for multinational corporations’ branch functions, Gawc’s assignment method for functional importance, and the industry and scope of operations of each branch, the listed companies and their branches in the core industry chain of new energy vehicles are classified and assigned values as follows: Group headquarters-type function (5 points), R&D-type function (4 points), Service-type function (3 points), Professional materials and components production (3 points), Complete vehicle production (3 points), General components production (2 points), Raw material acquisition (2 points). The service-type branches mainly include productive service branches such as logistics, software development, finance, leasing, business services, as well as wholesale and retail branches.
The connectivity strength Tij between city nodes reflects the network connectivity strength established by the headquarters-branch relationships of all enterprises between two cities. The formula is as follows:
${{T}_{ij}}=\underset{n=1}{\overset{m}{\mathop \sum }}\,{{V}_{if}}\times {{V}_{jf}}$
where m represents the number of enterprises simultaneously establishing headquarters and branches in cities i and j, and Vif and Vjf respectively represent the scores of enterprise f in cities i and j.
(2) Network Indicators.
Degree Centrality: Centrality indicators can be applied to identify key nodes in a network. Node degree is the most direct measure of node importance; the higher the degree, the higher the centrality, indicating more partners. Out-degree is the number of correspondences between the headquarters location of listed companies and the location of their branches, while in-degree is the number of correspondences between the location of various branches and the headquarters location.
${{K}_{i}}=K_{i}^{in}+K_{i}^{out}$
where Kiin is the in-degree of node i, Kiout is the out-degree of node i, and Ki represents the degree of the node i, the number of all nodes connected to node i.
Average Path Length: The average path length refers to the average of the shortest path lengths between any two nodes in the network, portraying the diffusion ability of elements in the network. A lower average path length indicates faster exchange of information and goods among network nodes, facilitating inter-city learning and promoting regional coordinated development (Guo et al., 2019).
$L=\frac{1}{1/2n\left( n+1 \right)}\underset{i\ge j}{\mathop \sum }\,{{d}_{ij}}$
where L represents the average path length of the network; n is the number of nodes; dij is the shortest path from node i to node j.
Clustering Coefficient: The average clustering coefficient is the average of the local clustering coefficients of all nodes in the network. A higher coefficient indicates that the connections between nodes in the network are more closely knit (Guo et al., 2019).
${{C}_{i}}=\frac{2{{E}_{i}}}{{{K}_{i}}\left( {{K}_{i}}-1 \right)}$
$C=\frac{1}{n}\underset{i=1}{\overset{n}{\mathop \sum }}\,{{C}_{i}}$
where Ki is the degree of node i, the number of nodes connected to node i; Ei is the actual number of edges between nodes connected to node i.
(3) Correlation and Module Division. To analyze the correlation between the degrees of nodes in different value segments, this study utilized the correlation analysis module in SPSS to conduct Pearson correlation analysis between the degrees of nodes in service, production, and R&D networks. Additionally, to identify and analyze the network structure and community division in industrial networks of different value segments, this study employed the modularity measure and classification in GEPHI.
(4) Negative Binomial Regression Model. Regression analysis can clarify the impact and relationship of variables on the dependent variable (Zhang et al., 2019). The strength of connections between nodes in the network is a count variable, with values being non-negative integers. By calculating variance much greater than the mean and exhibiting clustering, in such cases, the negative binomial regression model demonstrates good fitting performance (Liu et al., 2022). Thus, this study constructs a negative binomial regression model to explore the mechanisms of multiple factors on the strength of connections between nodes. The model is as follows:
${{D}_{ij}}=\alpha +\beta \mathop{\sum }^{}proximit{{y}_{ij}}+\mathop{\sum }^{}controlva{{r}_{ij}}+{{\varepsilon }_{ij}}$
where the dependent variable Dij represents the total strength of connections between nodes in different value chain stages of two cities. The independent variable proximityij includes multidimensional proximities, namely geographic proximity, social proximity, technological proximity, and institutional proximity. controlvarij is the control variable, and this study selects the difference in GDP between cities, the product of patent quantities, or the product of the resident population quantity as control variables.
Geographic proximity refers to the geographical distance between actors. On the one hand, geographic proximity can reduce transportation and communication costs between enterprises, thereby promoting the layout of production. On the other hand, it enables enterprises to jointly address external risks and uncertainties (Bathelt et al., 2004) and pursue the exchange of tacit knowledge, thereby enhancing innovation efficiency. The commonly used measurement indicator is the distance between cities. In this study, it is calculated and standardized through latitude and longitude.
$Ge{{o}_{ij}}=1-ln({{{y}_{ij}}}/{max\ {{y}_{ij}}}\;)$
where yij represents the geographical distance between city i and city j; max yij is the maximum distance between nodes in the research sample. The result is a continuous variable between 0 and 1.
Social proximity originates from embeddedness research and refers to the degree of common relationships between actors (Granovetter, 1985). It represents the closeness of relationships between actors (Zhang and Gu, 2022) and can also indicate other relationships involving trust between actors (Dai et al., 2022). These social relationships can be kinship, geographical, or friendship (Boschma, 2005), encouraging mutual understanding between actors, thus determining their ability to exchange information (Boschma and Frenken, 2009). Relationships established based on friendship, trust, and frequent interaction make it easier for familiar actors to cooperate, facilitating knowledge dissemination between different industries and increasing the likelihood of industrial connections (He and Yu, 2022). Therefore, this study refers to existing research, choosing the degree of overlap of partners between two nodes to measure social proximity and using the Jaccard index for calculation:
$So{{c}_{ij}}=\frac{{{q}_{ij}}}{{{r}_{i}}+{{s}_{j}}-{{q}_{ij}}}$
where qij is the number of shared partners between city nodes i and j, ri is the number of partners for city node i, and sj is the number of partners for city node j.
Technological proximity reflects the similarity in cognitive understanding and knowledge endowment distribution when cities communicate, providing a technical and knowledge foundation for the generation of industrial connections. When actors have similar knowledge cognitive foundations, their opportunities for mutual learning through communication increase (Cohen and Levinthal, 1990). Referring to the studies of scholars (Zhang and Gu, 2019), this study categorizes enterprises in different value chain stages into various types, which uses the similarity of enterprise types between cities to measure technological proximity. The calculation formula and categorization are as follows:
$Te{{c}_{ij}}=\frac{\mathop{\sum }_{s=1}^{t}\left( {{x}_{is}}{{x}_{js}} \right)}{\sqrt{(\mathop{\sum }_{s=1}^{t}x_{is}^{2})\left( \mathop{\sum }_{s=1}^{t}x_{js}^{2} \right)}}$
where xis and xjs represent the quantities of types enterprises in city nodes i and j, respectively; t is the total number of enterprise types. R&D enterprises are further categorized based on business operations into various types such as component R&D, specialty material R&D, charging technology R&D, electronic R&D, testing technology R&D, recycling R&D, and other technical R&D. Production enterprises are categorized into various types such as specialty material production, specialty component production, whole vehicle production, and raw material acquisition. Service enterprises are categorized into various types such as financial services, technology promotion services, software development services, charging and discharging operation services, supply chain services, recycling services, wholesale and retail services, and other services.
Institutional proximity is employed to explain the fact that the connections between entities are influenced, shaped, and constrained by institutional environments (Gertler, 2005). According to Coase’s theory, institutional proximity essentially boils down to the convenience or similarity in institutional design. That is, one party in a collaborative relationship has a convenient policy environment or favorable policy support, or both collaborating parties share similar policy environments. This reduces the uncertainty of cooperation between both parties, and leads to cost reduction (Zhou et al., 2021). It is also a crucial foundation for establishing industrial connections between regions (He and Yu, 2022). Considering, on one hand, the strong organizational role of the Chinese government in industrial development and the typical administrative hierarchy in city industrial development policies, and on the other hand, research indicating that cross-boundary institutional differences can hinder knowledge spillover between industries, thus affecting the formation of industrial linkages (He and Yu, 2022). Therefore, this study uses the administrative hierarchy of dominant cities and administrative boundaries to measure the convenience and similarity of institutions between cities, thereby completing the measurement of institutional proximity. Referring to studies by scholars (Zhan and Gu, 2022; Dai et al., 2023) and the classification of administrative levels in Chinese cities, the institutional proximity under administrative hierarchy InsijI is constructed as follows: the connection between directly-controlled municipalities, vice-provincial cities, and provincial capital cities is assigned a value of 3, the connection between these cities and ordinary prefecture-level cities is assigned a value of 1, and the connection between ordinary prefecture-level cities is assigned a value of 0, thus constructing institutional proximity Ⅰ between all pairs of cities; the institutional proximity under administrative boundaries InsijII is constructed as follows: the connection between cities within the same provincial domain is assigned a value of 1, and the connection between cities in different provincial domains is assigned a value of 0, thus constructing institutional proximity Ⅱ between all pairs of cities.
Control Variables: Based on production network theory, the study primarily controls for the social-economic characteristics and industry features of regions. The differences in economic scale and development level between cities may attract listed companies to establish branches through local market effects. Therefore, this model needs to control for the GDP difference between cities, considering the GDP difference between cities as one of the control variables. Considering the differences in value chains, the population size of cities to some extent represents the consumption capacity and labor capital of cities, which significantly influences both services and production in cities. Thus, the product of the resident population sizes of two cities is measured as one of the control variables for production-oriented and service-oriented networks. The scale of patented authorizations in a city represents the city’s innovation capability, significantly influencing the research and development of cities. Therefore, the product of the number of patents in a city is measured as one of the control variables for R&D-oriented city networks.

3 Characteristics, evolution, and differences of networks in different value chain segments

Based on the R&D functions, production functions (specialized materials and components, general components, complete vehicles, etc.), and service functions of enterprises, this paper constructs R&D-oriented, production-oriented, and service-oriented networks in the new energy vehicle industry, with cities as nodes. And this paper compares and analyzes their characteristics, evolution, and differences.

3.1 Characteristics, differences, and correlations of the three types of networks

(1) Synchronized expansion in scale, synchronized increase in connection strength. In comparison to 2015, the number of nodes in the R&D -oriented network increased from 73 to 128 in 2021, the number of connections between nodes increased from 144 to 512, and the average connection strength increased from 1.973 to 3.556. For the service-oriented network, the number of nodes increased from 146 to 241, the number of connections between nodes increased from 570 to 2193, and the average connection strength increased from 3.904 to 9.100. In the production-oriented network, the number of nodes increased from 205 to 240, the number of connections between nodes increased from 824 to 1662, and the average connection strength increased from 4.100 to 6.925. The three types of networks expanded synchronously, with an increase in both the number of network nodes and connections. As of 2021, the scale of city networks, from largest to smallest, is in the order of service-oriented, production-oriented, and R&D-oriented. The scale of the service-oriented network surpassed that of the production-oriented network, mainly due to the increased demand for new energy vehicles and the specialization of service types.
(2) Synchronized reduction in average path length, synchronized increase in clustering coefficient. In comparison to 2015, the average path length of the R&D-oriented, service-oriented, and production-oriented networks decreased from 3.391, 2.854, and 3.101 to 3.071, 2.537, and 2.952 in 2021. The average path length of the three types of networks decreased, indicating a reduction in the number of connections and a more convenient network. The clustering coefficients of the R&D-oriented, service-oriented, and production-oriented networks increased from 0.106, 0.216, and 0.138 to 0.17, 0.408, and 0.186, respectively. The clustering coefficients of all types of networks increased, indicating an increase in the degree of network clustering. Overall, the average path length of the three types of networks decreased, and the clustering coefficients increased. This indicates that the three types of networks are evolving towards a more efficient “small-world” network. The service-oriented network consistently has the smallest average path length and the largest clustering coefficient, suggesting that, relative to the other networks, the service-oriented network has more convenient connections and higher network clustering efficiency.
(3) Scale-free network distribution, enhanced network heterogeneity. Compared to 2015, in 2021, the average degrees of R&D-oriented, service-oriented, and production-oriented networks increased from 1.726, 2.233, and 2.473 to 2.672, 4.278, and 3.554 respectively, with the contact range of nodes in each city expanding. The degree distribution of the three types of networks follows a power-law distribution (Figure 1), with a fitting degree greater than 0.92, indicating a consistent scale-free network with a pronounced Matthew effect. Most city nodes have relatively low degrees, while there are a few nodes with relatively high degrees. The power-law index of the three types of networks shows a decreasing trend, indicating an increase in network heterogeneity and a widening overall gap in functional connections. Compared to R&D-oriented and production-oriented networks, the power-law index of the service-oriented network consistently remains the smallest, indicating a relatively greater overall gap in functional connections for the service-oriented network.
Figure 1 Degree distribution of China’s new energy vehicle industry network from the perspective of value chain in 2015 and 2021
(4) Significant positive correlations exist between the degrees of nodes in different value chain segments, and the importance of the same city in different segments is similar. Compared to 2015, in 2021, the Pearson coefficients of degrees for nodes in R&D-oriented and service-oriented networks show significant positive correlations (R22015=0.900, R22021=0.845), as do the degrees in R&D-oriented and production-oriented networks (R22015=0.832, R22021=0.918), and in service-oriented and production-oriented networks (R22015=0.888, R22021=0.883). All the mentioned significance test values are less than 0.01. The degree of nodes indicates their importance in the network, and a significant positive correlation implies that the importance of nodes in the same city is largely similar across different networks. This suggests the possibility of entering the industry and upgrading from lower-value chain segments. Therefore, for cities that have not yet entered the core industry, actively introducing production in lower-value chain segments to integrate into the new energy vehicle production network can lead to further upgrades.

3.2 Classification of node types in different value chain segments

In 2021, the in-degrees and out-degrees of city nodes in the three types of networks were classified into two levels (high-level and low-level) using natural breakpoints in ArcGIS. Using the same classification criteria, the 2015 networks for the three types were also categorized. Nodes in the network that exhibit high-level out-degrees and in-degrees have a large radiation range and strong attractiveness. These nodes are both divergent locations for foreign investments and receiving locations for attracting investments, and can be referred to as Power-Prestige nodes. Nodes with high out-degrees and low in-degrees have a large radiation range and weak attractiveness, typically serving as the headquarters of companies, and can be termed Power nodes. Nodes with low out-degrees and high in-degrees have strong attractiveness and a small radiation range, attracting branch enterprises from various places, and can be referred to as Prestige nodes. It should be noted that the quantity of nodes with both out-degrees and in-degrees in the low-level, and their radiation range and attractiveness are relatively small, so they are not listed in Table 1.
Table 1 Evolution of node types in China’s new energy vehicle industry network from the perspective of value chain
Year
Type		 2015 2021
R&D-
oriented		 Service-
oriented		 Production-
oriented		 R&D-
oriented		 Service-
oriented		 Production-
oriented
Power-Prestige Shenzhen Shanghai, Beijing, Shenzhen,
Hangzhou		 Shanghai, Shenzhen,
Beijing, Hangzhou		 Shanghai, Shenzhen,
Beijing,
Guangzhou,
Hangzhou,
Qingdao, Hefei		 Shenzhen, Beijing,
Shanghai, Hangzhou,
Guangzhou
Power Shenzhen Hefei, Ningbo,
Dongguan,
Qingdao, Xi’an,
Ningde		 Xiamen
Prestige Shanghai, Beijing, Tianjin Shanghai, Beijing, Tianjin, Chengdu Guangzhou, Tianjin, Suzhou, Wuhan, Hefei, Ningbo, Chengdu, Chongqing, Wuhu, Changsha, Nanjing, Dongguan Tianjin, Chengdu,
Suzhou, Nanjing,
Changzhou, Wuxi,
Changsha		 Chengdu, Suzhou,
Haikou, Wuhan,
Chongqing, Tianjin, Nanjing, Xi’an,
Xiamen, Ningbo,
Sanya, Zhengzhou,
Wuxi, Taiyuan,
Shenyang,
Changsha, Zhuhai,
Dongguan, Harbin,
Lasa, Nanchang		 Nanjing, Dongguan,
Hefei, Chengdu, Ningbo,
Xi’an, Chongqing,
Suzhou, Tianjin,
Wuhan，Changsha,
Wuhu, Yichun, Jiaxing,
Nantong, Wuxi,
Chuzhou, Changchun,
Liuzhou, Zhenjiang,
Changzhou, Qingdao,
Baotou, Taizhou,
Yangzhou, Ganzhou,
Shiyan, Ningde, Zhuhai

Note: Classification criteria for levels: Research-oriented networks, high-level in-degree (8-36), low-level in-degree (0-7), high-level out-degree (10-40), low-level out-degree (0-9); Service-oriented networks, high-level in-degree (10-40), low-level in-degree (0-9), high-level out-degree (33-124), low-level out-degree (0-32); Production-oriented networks, high-level in-degree (7-18), low-level in-degree (0-6), high-level out-degree (28-90), low-level out-degree (0-27).

In the R&D-oriented network, there were no Power-Prestige nodes in 2015; only Shenzhen was a Power node, serving as a hub for numerous corporate headquarters with a relatively wide range of influence. Prestige nodes included Shanghai, Beijing, and Tianjin, attracting enterprises from various regions to establish research branches. In 2021, Shanghai, Shenzhen, Beijing, and Hangzhou have risen to become Power-Prestige nodes, with increased influence and attractiveness. Hefei, Ningbo, Dongguan, Qingdao, Xi’an, and Ningde have become Power nodes, reflecting the growing number of listed companies and emphasis on research in these areas. Notably, Hefei experienced the fastest growth. In addition to Tianjin, Chengdu, Suzhou, Nanjing, Changzhou, Wuxi, and Changsha have become Prestige nodes, with increased attractiveness. Suzhou has shown the fastest growth, attracting numerous companies, particularly in the research and development of charging and charging station technologies. These nodes also serve as locations for many corporate headquarters.
In the service-oriented network, only Shenzhen was a Power-Prestige node in 2015, while Shanghai, Beijing, Tianjin, and Chengdu were Prestige nodes. In 2021, in addition to Shenzhen, Shanghai, Beijing, Guangzhou, Hangzhou, Qingdao, and Hefei have risen to become Power-Prestige nodes. Chengdu, Suzhou, Haikou, Wuhan, Chongqing, and 21 other cities have become Prestige nodes. Notably, 82.14% of these nodes are directly governed cities, deputy-provincial cities, and provincial capitals.
In the production-oriented network, in 2015, only Shanghai, Beijing, Shenzhen, and Hangzhou were Power-Prestige nodes, while Guangzhou, Tianjin, Suzhou, and 12 other cities were Prestige nodes. In 2021, in addition to Shanghai, Beijing, Shenzhen, and Hangzhou, Guangzhou has risen to become a Power-Prestige node, and Xiamen has become a Power node. Although it has headquarters enterprises and a large radiation range, its attractiveness is relatively weak. Apart from Tianjin, Suzhou, Wuhan, Hefei, and 11 other cities remaining as Prestige nodes, Xi’an, Yichun, Jiaxing, Nantong, Wuxi, Changchun, and 18 other cities have become Prestige nodes. Traditional automotive parts production cities still hold advantages, attracting core enterprises of the new energy vehicle industry to establish branches, and the demand for new materials has led companies to establish production branches in cities with resource advantages, such as Baotou for magnetic materials and Yichun and Ganzhou for lithium battery materials.
Overall, with the increasing number of listed companies and the growth in the number of branches in research, production, and service, the quantity of various node types in R&D-oriented, production-oriented, and service-oriented networks has increased. From Figure 2, it can be observed that in 2021, the city nodes of R&D-oriented, service-oriented, and production-oriented networks have expanded from the eastern region to the central and western regions compared to 2015, but they still concentrate more in the eastern region beyond the “Hu Huanyong Line”. Additionally, Power-Prestige nodes in each segment consistently concentrate in the eastern coastal areas. In 2021, Shanghai, Beijing, Shenzhen, and Hangzhou are Power-Prestige nodes in different value chain networks, demonstrating strong attractiveness and influence in various segments of the value chain, showcasing significant leadership qualities. Tianjin has consistently been a Prestige node in R&D, service, and production in 2015 and 2021, with a relatively small increase in radiation capacity. This is primarily due to the limited number of large listed companies establishing headquarters in Tianjin, limiting its radiation capacity. In all three types of networks, the number of nodes with out-degrees of 0 far exceeds the number of nodes with in-degrees of 0, indicating that corporate headquarters continue to concentrate in a few cities.
Figure 2 Distribution and evolution of node types in China’s new energy vehicle industry network in 2015 and 2021

Note: Based on the standard map production with the approval number GS (2019) 1815 on the standard map service website of the Ministry of Natural Resources, the boundary of the base map has not been modified

3.3 Evolution of industrial network structure based on strongest connections

By analyzing the strongest connections between nodes using the GEPHI modular clustering algorithm, we analyzed the core-periphery structure and community divisions in networks across different value chains. The strongest connections between nodes indicate nodes that are closely, conveniently, and profoundly linked, further fostering trust and close collaboration among organizations within the industry across cities (Figure 3).
Figure 3 The evolution of China’s new energy vehicle industry network structure based on the strongest connection in 2015 and 2021
The structure of the R&D-oriented network based on the strongest connections consistently exhibits an integrated, closely connected multiple core-periphery structure community.
In 2015, the strongest connection existed between Shenzhen and Shanghai, forming the largest community, followed by the community centered around Beijing, with Beijing having the strongest connection with Shanghai. Besides, each community had fewer members with relatively lower connection intensity, and community differentiation was not very distinct. In 2021, Shanghai and Shenzhen remained the strongest connection pair, but with the increase and enhancement of R&D-oriented nodes, they formed core communities centered around Shenzhen and Shanghai, followed by communities centered around Beijing, Qingdao-Tianjin, Hefei, and Dongguan. The number of communities and nodes within communities increased simultaneously. Overall, Shenzhen and Shanghai still formed the strongest connection and were the strongest connections in the entire network. Beijing’s strongest connection was also with Shanghai, and there were strongest connections between core nodes of each community, leading to the nationwide integrated development of R&D-oriented enterprises.
The structure of the service-oriented network based on the strongest connections has transitioned from closely connected multiple core-periphery community to multiple core-semi- periphery-periphery structure communities. In 2015, the strongest connection pair was formed between Shanghai and Beijing, becoming the largest community, followed by the community centered around Shenzhen, with Shenzhen having the strongest connection with Beijing. In 2021, the connection between Beijing and Shenzhen strengthened, forming the strongest connection pair and becoming the largest community. Additionally, communities centered around Shanghai, Guangzhou, Chengdu, etc., grew and developed. Shanghai had the strongest connection with Beijing, Guangzhou with Shenzhen, and Chengdu with Shenzhen. New communities centered around Xi’an, Hangzhou, etc., also formed. Overall, the service-oriented network structure formed communities centered around Shanghai, Beijing, Shenzhen, and Guangzhou as core nodes and secondary-level communities centered around Xi’an, Chengdu, Hangzhou, etc. There are strongest connections between core nodes of each community, promoting the integrated development of nationwide service-oriented enterprises. Enterprises in the high value-added R&D and service sectors still focus on connections between high-level core nodes and diffuse to other secondary-level cities.
The structure of the production-oriented network based on the strongest connections has shifted from partial integration and localization core-periphery structure community to the more decentralized multiple independent core-periphery structure community. In 2015, communities were formed with core nodes in Shanghai, Beijing, Hangzhou, Shenzhen, Xiamen, Guangzhou, etc., where the strongest connection for Shenzhen was with Shanghai, establishing the strongest connection between these two communities. In 2021, communities centered around a specific city became more differentiated, and the strongest connections among core nodes were within their respective communities, forming multiple independent core-periphery structure communities. Each core node had the strongest connections within its community, resulting in multiple strongest connection pairs such as Shanghai-Nanjing, Shenzhen-Huizhou, Beijing-Changsha, Hangzhou-Jinhua, and Dongguan-Shaoguan. The production process in low value-added sectors involved direct connections between core nodes and secondary-level nodes, primarily diffusing to other levels of cities through nearby diffusion, with a noticeable regionalization process.

4 Formation mechanism of the network structure in different value chain segments

Different agglomeration types are formed in regional space as various value chain segments pursue different factor endowments and agglomeration effects, leading to distinctive network spatial structures (Figure 4). Multidimensional proximity exerts both common and heterogeneous influences on the evolution of different value chain segments. Among them, both geographic and social proximities passed significance tests, while technical proximity in the service segment, two institutional proximities in the R&D segment, technical proximity in the production segment, and institutional proximity under administrative levels did not pass significance tests (Table 2).
Figure 4 The Formation mechanism of the network structure of China’s new energy vehicle industry from the perspective of value chain
Table 2 Regression results of multidimensional proximity mechanism in China’s new energy vehicle industry network from the perspective of value chain
Variable Model 1
R&D		 Model 2
Production		 Model 3
Service		 Variable Model 1
R&D		 Model 2
Production		 Model 3
Service
Geoij 0.0851016** 0.1028103*** 0.0610812** Gdpij 0.0014708*** 0.0008793** 0.0008893***
(0.0365204) (0.0332531) (0.0309952) (0.0005832) (0.0004873) (0.0004504)
Socij 1.329344*** 1.699106*** 1.744836*** Patij/Popij 0.0000146** 0.0004679*** 0.0005694***
(0.4147179) (0.3668167) (0.2370462) (0.00000463) (0.000085) (0.0000679)
Tecij 0.2277579** -0.05579 0.121154 Cons 2.513486*** 2.292235*** 1.999777***
(0.1062781) (0.1089859) (0.1586508) (0.1472295) (0.1332833) (0.1844383)
InsijI 0.013834 0.046011 0.1607244*** α 0.21133 0.377051 0.372476
(0.0347458) (0.0322685) (0.0266826) Log
likelihood		 -1256.93 -3193.46 -3817.34
InsijII -0.04883 0.186526** 0.4449469***
(0.1109177) (0.0885989) (0.0865728)

Note: The variance inflation factors (VIF) for Models 1-3 are all below the critical value of 10, indicating no multicollinearity among explanatory variables. Values in parentheses represent standard errors. *, *, and *** denote p<0.10, p<0.05, and p<0.01, respectively.

In the R&D network, with the increasing contribution of knowledge as a production factor and its growing significance in economic and social development, research enterprises, universities, and research institutions have become primary innovators and sources of innovation in the new era. Companies tend to concentrate their R&D activities in cities rich in these innovative resources (Dolores and Alfonso, 2021), pursuing positive feedback from technical externalities and learning effects rooted in the local environment and industrial atmosphere (Liu et al., 2021). Moreover, in larger cities with higher levels of human capital, labor is more likely to generate learning effects (Glaeser, 1999; 2003). Given the characteristics of R&D activities, such as high investment, long cycles, and high risks, they tend to be distributed in places with initially higher capital stocks. Therefore, nodes in the R&D network have high-level or headquarters orientations. High-value R&D enterprises in these nodes continuously reinforce themselves under the preferential support of regional government policies, technological spillover localization, and agglomeration inertia, often forming specialized agglomerations in regional space based on the research of specific knowledge in the industrial chain.
In the regression results of the proximity mechanism, geographic proximity passed the significance test, indicating that spatial proximity is conducive to cooperation between R&D cities, with a positive effect. The main actors in R&D activities seek to form agglomerations in spatial proximity to increase face-to-face communication and learning opportunities, achieve lower innovation costs, and promote risk sharing and cooperative innovation (Sheng, 2012). Technical proximity also has a significant positive effect on the connections between cities, indicating a higher likelihood of cooperation between cities with similar R&D types. Similar knowledge backgrounds and common research levels are essential foundations for collaboration (Boschma, 2005), promoting the absorption and understanding of R&D knowledge between collaborating entities. Social proximity has the largest regression coefficient and passes the significance test at the 1% level, indicating a higher probability of cooperation between nodes with common partners. The more common friends, the easier it is to establish higher trust, thereby reducing risks in transactions and cooperation processes, enhancing cooperation efficiency (Boschma, 2009). Both institutional proximities did not pass the significance test. On the one hand, there are fewer cities in the R&D network, most of which are high-level cities. The institutional proximity values between these high-level cities are similar but vary greatly under the influence of geographic, social, and technical proximity. On the other hand, there are some cities in the R&D nodes that are not high-level but are headquarters gathering places, such as Changzhou and Dongguan; therefore, institutional proximity regression is not significant. In summary, agglomeration and multidimensional proximity make the R&D network under the strongest connection always show an integrated core-periphery structure. The high-intensity connections between core cities, Beijing, Shenzhen, and Shanghai, in each community enhance the efficiency of connections between peripheral cities, facilitating the acquisition of technical knowledge spillover from core nodes and increasing the possibility of receiving R&D industry transfers and enhancing technical knowledge spillover localization. This network pattern improves the overall efficiency and stability of R&D nationwide.
In the service network, the service sector under the core industry of new energy vehicles exhibits high value-added characteristics, can afford higher land rents in large cities, and has a higher degree of knowledge demand. Usually, larger cities can meet this demand (Smętkowski et al., 2021). As productive services become more specialized, listed companies need more specialized service branches to improve and match the quality and quantity of related service segments. Therefore, listed companies often establish multiple service branches of different types in a certain region, forming diverse agglomerations in regional space.
In the regression results of the proximity mechanism, geographic proximity passed the significance test, indicating that spatial proximity contributes to the conduct of services between cities, with a positive effect. The impact mechanism of geographic proximity on service cooperation between cities mainly manifests as reducing transaction time and space costs, especially within nodes and their connections to core nodes in each community. Social proximity also passed the significance test, tending to cooperate with nodes that have common partners. This is similar to the R&D segment, where higher trust can enhance cooperation efficiency between nodes. Technical proximity did not pass the significance test, further indicating that service cities pursue diversified agglomeration, and the similarity of service types does not promote increased cooperation between cities. Both administrative level and institutional proximity under administrative boundaries passed the test, and the superiority of one party’s policies and the same institutional environment facilitate communication and cooperation between entities. On the one hand, higher-level city nodes have stronger and more diverse service categories, thus forming stronger cooperation. This also enables the national service information to be interconnected, helping service enterprises in nodes of small and medium cities connect with a broader national service network. This mainly manifests as connections between communities. On the other hand, provincial administrative center cities have higher resource control capabilities and play a crucial hub role in external connections within the province. Simultaneously, under time and space costs, services have a certain service range. Cities closer to the core of the network face strong competitive effects and are prone to agglomeration shadows. Therefore, in peripheral cities farther from the core of productive service network functions, their productive service functions are stronger, that is, under “borrowing of scale,” they have acquired some productive service functions that they did not originally possess (Wang et al., 2022b). Based on these agglomerations and multidimensional proximities, the service network under the strongest connection has transformed from an integrated core-periphery structure in 2015 to an integrated core-semi-periphery-periphery structure in 2021. With the improvement of the service network, Xi’an, Chengdu, Hangzhou, and other semi-core nodes have been formed.
In the production network, companies, aiming to share indivisible goods, facilities, production services, and bear risks collectively, tend to cluster in cities with complete production factors to enjoy the benefits of agglomeration (Sheng, 2012). Listed companies, seeking urbanized economies for cost savings resulting from the concentration of various production enterprises (Ge et al., 2005), often form diverse agglomerations in regional space. With the application of modular production, the increased adaptability of automotive components enhances economies of scale for both whole vehicle manufacturers in procurement and component production enterprises in production (Zhao et al., 2021), thus promoting close connections between regions.
In the regression results of the proximity mechanism, geographical proximity passed the significance test, indicating that the smaller the distance between urban nodes, the greater the likelihood of production cooperation. This is mainly due to higher transportation and land costs in the production process compared to R&D and service processes. To reduce transportation costs using increasing returns to scale and access lower-rent production land, listed companies often arrange production processes in cities at a lower hierarchical level in geographic proximity to their headquarters or hinterlands. Social proximity passed the significance test. The higher the overlap in the “circle of friends,” the stronger the connection intensity. This is similar to R&D and service processes, making it easier to establish deep trust and improve cooperation efficiency between urban nodes. Technical proximity did not pass the significance test. This further indicates that production-type nodes pursue diverse agglomerations, and the similarity of production types does not promote increased cooperation between city nodes. In institutional proximity, the institutional proximity under administrative levels did not pass the significance test. The increase in production costs for higher-level cities reduces production connections. Institutional proximity across administrative boundaries passed the significance test. On the one hand, the similarity of institutions within the same province can reduce the uncertainty of urban production cooperation and lower transaction costs. On the other hand, the production process in the value chain has the lowest added value and is the most easily contested value link for marginal cities. It generally has strong preferential policies, and within the same province, there is often a top-down urban cooperation mechanism to prevent the outflow of this link. Thus, based on the strongest connections, the production network in 2015 exhibited a partially integrated and localized core-periphery structure community. By 2021, it has evolved into multiple independent core-periphery structure communities.

5 Conclusions and discussion

5.1 Conclusions

(1) Compared to 2015, in 2021, China’s new energy vehicle industry networks, including R&D, production, and service, expanded simultaneously. The number of network nodes increased, and the connections between nodes grew. Among them, the service network showed the fastest growth in both nodes and connections. By 2021, the R&D network had the smallest scale. High-value-added segments in R&D network still concentrated in a few cities. The degree distribution of the three types of networks followed a power-law distribution, indicating that they consistently exhibit a scale-free network with a clear Matthew effect. While the majority of nodes have relatively low degrees, there are a few nodes with relatively high degrees. The power-law indices of the degree distribution for the three types of networks all showed a decreasing trend, indicating an increased heterogeneity of nodes in these networks, leading to a growing overall gap in node degrees among the three types of networks. There was a significant positive correlation between the degrees of nodes in different value segments, suggesting that the importance of nodes in the same city is similar across different networks. This also implies the possibility of upgrading from low-value segments to industries within the same city.
(2) In the industrial network, the classification of nodes into power-prestige, power-type, and prestige-type showed an increase in the number of nodes for all types except the service network, which consistently had no power-type nodes. The number of power nodes in each type of network was relatively small compared to prestige nodes. Nodes in different value segments had different agglomeration directions: R&D nodes tended to have headquarters or high-level city orientation, service nodes were oriented towards sub-provincial and provincial capitals, and production nodes were directed towards traditional production bases. In 2021, Shanghai, Beijing, Shenzhen, and Hangzhou were identified as power-prestige nodes in all three types of networks, indicating strong overall capabilities. The number of nodes with an out-degree of 0 far exceeded the number of nodes with an in-degree of 0, indicating that corporate headquarters continued to concentrate in a few cities.
(3) In the new energy vehicle industry, different value chain segments form different agglomeration types in regional space under the pursuit of different factor endowments and agglomeration effects. Simultaneously, under the influence of multidimensional proximity, they create a network spatial structure with distinctive characteristics. To pursue learning effects and localized technological spillovers, entities engaging in R&D activities often form specialized agglomerations in cities based on specific expertise. Under the influence of social, technological, and geographical proximity, the R&D-oriented strongest connection network consistently exhibits an integrated structure with closely connected multiple core-periphery structure communities. The specialization of service types prompts publicly traded enterprises to establish more specialized service branches, enhancing the quality and quantity of relevant service segments from a professional perspective. However, as market demand increases, it is also affected by the scale borrowing and agglomeration shadows. Simultaneously, under the influence of social, geographical, and institutional proximity, the strongest connection network in the service sector shifts from an integrated structure with closely connected multiple core-periphery structure communities to multiple core-semi-periphery-periphery structure communities. To reduce transportation costs, pursue shared effects, economies of scale, and urbanization, entities engaging in production activities often form agglomerations with diverse production types in a specific city. Meanwhile, under the influence of social, geographical, and institutional proximity, the strongest connection network in the production sector manifests as partially integrated and localized core-periphery structure communities evolving into decentralized multiple independent multiple core-periphery communities.

5.2 Discussion

This paper examines the spatial distribution and formation mechanism of China’s new energy vehicle industry from the perspective of the value chain, enhancing the understanding of the regional development of the new energy vehicle industry. It also provides insights for improving the competition of different cities in the domestic development of the new energy vehicle industry. The following revelations are provided: (1) For cities that have not yet integrated into the core of the new energy vehicle industry, on the one hand, they should emphasize the role of geographical proximity and actively engage in relevant cooperation with neighboring cities to take advantage of the benefits brought by geographical proximity. On the other hand, they should actively undertake the industrial transfer from core cities, integrate into the industry by entering at the low-added value chain, and then achieve value upgrading. Meanwhile, with numerous partners in core cities, under the influence of social proximity, the credibility and number of partners in the industry can be increased. (2) For cities that have integrated into the core of the new energy vehicle industry, on the one hand, when developing in a specific value chain segment of the new energy vehicle industry, they should clarify their position and advantages while enhancing their attractiveness and radiation power to grasp industry leadership. On the other hand, during the ascent of the value chain, it is necessary to pay attention to the heterogeneity of the mechanisms in different segments, seize the opportunity to fill the gap in the province’s specialized service types, continuously deploy diversified service types, enhance specialization in technology R&D, and improve the R&D capabilities of the city.
Combining the value chain and production network, this paper analyzes the national-scale network of the automotive industry with cities as nodes based on the organizational network of listed companies in China’s new energy vehicle industry. It reveals the specific geographical distribution patterns and spatial structures in China. Based on a large amount of headquarters-branch data from enterprises, it scientifically validates the distribution of production networks under the value chain system, which differs from the traditional approach that focuses on a few leading enterprises and headquarters as cases. The paper also analyzes the evolution and formation mechanisms of the new energy vehicle industry network structure from the perspective of the value chain, providing a scientific basis for cities to seize the new energy vehicle industry, as well as for value upgrading. However, there are still shortcomings: the influence of Chinese policies on the development of the new energy vehicle industry is strong. In this study, due to the difficulty in quantifying policy data in the measurement of institutional proximity, administrative levels and administrative boundaries are approximated. Some scholars argue that political initiatives in China have altered the original competitive landscape in the electric vehicle market and supply chain and the conditions of acquiring core technological components for well-established foreign automotive companies (Godfrey, 2019). Therefore, a more detailed exploration is needed regarding the impact of new energy vehicle policies on industrial development. On the other hand, with the increasing global trade frictions, European and American countries intervene in transnational dominant enterprises through administrative means, thereby altering the operation of the global production network. Joint ventures, however, face drawbacks such as technological dependence and path locking, especially considering the risks posed by the technological closure of foreign companies to joint ventures and even the domestic automotive manufacturing industry. Therefore, it is imperative to anticipate the vulnerability and resilience of the domestic automotive industry in the post-technological closure era, addressing the development strategies and path choices for relevant local automotive industry enterprises.

References

Publishing order | Descend order by publishing year | Descend order by cited within
[1]
Bathelt H, Malmberg A, Maskell P, 2004. Clusters and knowledge: Local buzz, global pipelines and the process of knowledge creation. Progress in Human Geography, 28(1): 31-56, https://doi.org/10.1191/0309132504ph 469oa.

DOI

[2]
Beaudry C, Swann P, 2001. Growth in industrial cluster: A bird’s eye view of the United Kingdom. SIEPR Discussion Paper, 1-38, https://doi.org/10.1.1.200.7347&rep=rep1&type=pdf.

[3]
Boschma R, 2005. Proximity and innovation: A critical assessment. Regional Studies, 39(1): 61-74, https:// doi.org/10.1080/0034340052000320887.

DOI

[4]
Boschma R, Frenken K, 2010. The spatial evolution of innovation networks:A proximity perspective. In:Boschma R, Martin R. The Handbook of Evolutionary Economic Geography. Cheltenham: Edward Elgar Publishing, 120-135, https://doi.org/10.4337/9781849806497.00012.

[5]
Chen Le, 2022. Theoretical basis and empirical studies of agglomeration economy influencing urban economic growth: Literature review and prospect. Progress in Geography, 41(7): 1325-1337, https://doi.org/10.18306/ dlkxjz.2022.07.015. (in Chinese)

DOI

[6]
Chen Xiaofei, Miao Changhong, Pan Shaoqi et al., 2018. Characteristics and construction mechanism of enterprise networks in “Hub-and-Spoke” cluster: Empirical evidence from Chery cluster in 2014, China. Geographical Research, 37(2): 353-365, https://doi.org/10.11821/dlyj201802009. (in Chinese)

DOI

[7]
Coe N M, Hess M, Yeung, H W-C et al., 2004. ‘Globalizing’ regional development: A global production networks perspective. Transactions of the Institute of British Geographers, 29: 468-484, https://doi.org/0.1111/j.0020-2754.2004.00142.x.

DOI

[8]
Cohen W M, Levinthal D A, 1990. Absorptive capacity: A new perspective on learning and innovation. Administrative Science Quarterly, 35(1): 128-152, https://doi.org/10.1016/B978-0-7506-7223-8.50005-8.

DOI

[9]
Cong Haibin, Zou Deling, Gao Bo et al., 2021. Network patterns and influence factors of new energy vehicle trade along the countries of the Belt and Road. Economic Geography, 41(7): 109-118, https://doi.org/10. 15957/j.cnki.jjdl.2021.07.012. (in Chinese)

[10]
Dai Liang, Cao Zhan, Ma Haitao et al., 2023. The influencing mechanisms of evolving structures of China’s intercity knowledge collaboration networks. Acta Geographica Sinica, 78(2): 334-350, https://doi.org/10. 11821/dlxb202302005. (in Chinese)

DOI

[11]
Dai Liang, Liu Chengliang, Wang Song et al., 2022. Evolutionary characteristics and proximity mechanism of intercity knowledge innovation networks in China. Geographical Research, 41(9): 2499-2515, https://doi.org/ 10.11821/dlyj020211014. (in Chinese)

DOI

[12]
Dicken P, 2011. Global Shift: Mapping the Changing Contours of the World Economy. London: Sage.

[13]
Dolores H, Alfonso D M, 2021. Do universities matter for the location of foreign R&D? BRQ Business Research Quarterly, 234094442110423, https://doi.org/10.1177/2340944421104238.

[14]
Fei Jin, Chen Luan, Xie Zhiying, 2019. Development course and embedded evolution of China’s automobile industry. Productivity Research, 321(4): 82-87, 128, https://doi.org/10.19374/j.cnki.14-1145/f.2019.04.014. (in Chinese)

[15]
Ge Ying, Yao Shimou, Pu Yingxia et al., 2005. Application of spatial autocorrelation for the spatial patterns of urbanization and localization economy. Human Geography, (3): 21-25, https://doi.org/10.13959/j.issn.1003-2398.2005.03.005. (in Chinese)

[16]
Gereffi G, Humphrey J, Sturgeon T, 2005. The governance of global value chains. Review of International Political Economy, 12(1): 78-104, https://doi.org/10.1080/09692290500049805.

DOI

[17]
Gertler M S, 2005. Tacit knowledge, path dependency and local trajectories of growth. In: FuchsG, ShapiraP. RethinkingRegional Innovation and Change. New York: Springer, 23-41.

[18]
Glaeser E L, 1999. Learning in cities. Journal of Urban Economics, 46(2): 254-277, https://doi.org/10.1006/juec. 1998.2121.

DOI

[19]
Glaeser E L, Albert S, 2003. The Rise of the Skilled City. Cambridge: National Bureau of Economic Research, https://doi.org/10.3386/w10191.

[20]
Godfrey Yeung, 2019. ‘Made in China 2025’: The development of a new energy vehicle industry in China. Area Development and Policy, 4(1): 39-59. https://doi.org/10.1080/23792949.2018.1505433.

DOI

[21]
Granovetter M, 1985. Economic action and social structure: The problem of embeddedness. American Journal of Sociology, 91(3): 481-510, https://www.jstor.org/stable/2780199.

DOI

[22]
Guo Jianke, He Yao, Wang Shaobo et al., 2019. Rank-size distribution changes and transportation network connections of the coastal container port system in Chinese mainland since 1985. Geographical Research, 38(4): 869-883, https://doi.org/10.11821/dlyj020181050.

DOI

[23]
Han Ying, Pan Zhigang, 2005. Analysis of the driving effectiveness of automobile industry to its correlative industry. China Soft Science, (6): 139-143. (in Chinese)

[24]
He Canfei, Yu Changda, 2022. Multi-dimensional proximity, trade barriers and the dynamic evolution of industrial linkages between China and the world market. Acta Geographica Sinica, 77(2): 275-294, https://doi.org/10.11821/dlxb202202002. (in Chinese)

DOI

[25]
Huang Shoukun, Li Wenbin, 2005. Analysis of industrial networks and its evolution patterns. China Industrial Economics, (4): 53-60, https://doi.org/10.19581/j.cnki.ciejournal.2005.04.007. (in Chinese)

[26]
Hymer S, 1992. The multinational corporation and the law of uneven development. International Economic Policies and Their Theoretical Foundations. 2nd ed. California: Academic Press, 436-463, https://doi.org/10.1016 /B978-0-12-444281-8.50025-8.

[27]
Jin Ying, 2020. The evolution of the spatial pattern and the resilience of Liaoning province’s production network[D]. Changchun: Northeast Normal University. https://doi.org/10.27011/d.cnki.gdbsu.2020.000568.

[28]
Kuroiwa I, Techakanont K, Keola S, 2022. Evolution of production networks and the localisation of firms: Evidence from the Thai automotive industry. Journal of the Asia Pacific Economy, https://doi.org/10.1080/ 13547860.2021.2024361.

[29]
Li Y Q, Xiong J, Ma W Q et al., 2022. Environmental decoupling, eco-efficiency improvement, and industrial network optimization: Insights from 44 sectors in China. Journal of Cleaner Production, 376(20): 134374, https://doi.org/10.1016/j.jclepro.2022.134374.

DOI

[30]
Liang L, Liang Y W, 2021. An empirical study on the national heterogeneity of high-end manufacturing technology innovation and GVC’s division of labor status. Mathematical Problems in Engineering, 2021: 7464384, https://doi.org/10.1155/2021/7464384.

[31]
Lin Bingquan, Sun Bindong, 2022. The impact of inter-cluster networks on firms’ TFP: A case study of China’s automobile manufacturing industry. Geographical Research, 41(9): 2385-2403, https://doi.org/10.11821/ dlyj020220080. (in Chinese)

DOI

[32]
Liu Lin, Liu Huiting, Chen Jianguo et al., 2022. The impact of “Thunder Anti-drug” operation on drug dealing crime: A case study of the main urban area of ZG city. Acta Geographica Sinica, 77(6): 1461-1474, https://doi.org/10.11821/dlxb202206011. (in Chinese)

DOI

[33]
Liu Qing, Yang Yongchun, Jiang Xiaorong et al., 2021. An analysis of the multidimensional globalizing city networks based on global value chain: A case study of iPhone suppliers. Acta Geographica Sinica, 76(4): 870- 887, https://doi.org/10.11821/dlxb202104007. (in Chinese)

DOI

[34]
Massey D, 1979. In what sense a regional problem? Regional Studies, 13(2): 233-243, https://doi.org/10. 1080/00343400701232181.

DOI

[35]
Mudambi R, 2008. Location, control and innovation in knowledge-intensive industries. Journal of Economic Geography, 8(5): 699-725, https://doi.org/10.1093/jeg/lbn024.

DOI

[36]
Natsuda K, Thoburn J, Blažek J et al., 2022. Industrial policy and automotive development: A comparative study of Thailand and Czechia. Eurasian Geography and Economics, 63(2): 212-238, https://doi.org/10.1080/ 15387216.2020.1836983.

DOI

[37]
Özatağan G, 2011. Shifts in value chain governance and upgrading in the European periphery of automotive production: Evidence from Bursa, Turkey. Environment and Planning A, 43: 885-903, https://doi.org/10.1068/ a43428.

DOI

[38]
Pavlinek P, 2020. Restructuring and internationalization of the European automotive industry. Journal of Economic Geography, 20(2): 509-541, https://doi.org/10.1093/jeg/lby070.

[39]
Pavlínek P, 2021. Relative positions of countries in the core-periphery structure of the European automotive industry. European Urban and Regional Studies, 29(1): 59-84, https://doi.org/10.1177/09697764211021882.

DOI

[40]
Pavlínek P, Ženka J, 2016. Value creation and value capture in the automotive industry: Empirical evidence from Czechia. Environment and Planning A: Economy and Space, 48(5): 937-959, https://doi.org/10.1111/j.0020-2754.2004. 00142.x.

DOI

[41]
Ren Jinling, 2011. Production fragmentation of value chain, incomplete transfer and industrial transfer in China. Forum on Science and Technology in China, 186(10): 37-41, https://doi.org/10.13580/j.cnki.fstc.2011.10.019. (in Chinese)

[42]
Schwabe J, 2020. From “obligated embeddedness” to “obligated Chineseness”? Bargaining processes and evolution of international automotive firms in China’s new energy vehicle sector. Growth and Change, 51: 1102-1123, https://doi.org/10.1111/grow.12393.

DOI

[43]
Sheng Lei, 2012. Spatial Agglomeration and Knowledge Spillovers of Foreign R&D in China. Shanghai: East China Normal University Press.

[44]
Smętkowski M, Celińska-Janowicz D, Wojnar K, 2021. Location patterns of advanced producer service firms in Warsaw: A tale of agglomeration in the era of creativity. Cities, 108: 102937, https://doi.org/10.1016/j.cities. 2020.102937.

DOI

[45]
Sturgeon T, Van Biesebroeck J, Gereffi G, 2008. Value chains, networks and clusters: Reframing the global automotive industry. Journal of Economic Geography, 8(3): 297-321, https://doi.org/10.1093/jeg/lbn007.

DOI

[46]
Tang M H, Hong J K, Liu G W et al., 2019. Exploring energy flows embodied in China’s economy from the regional and sectoral perspectives via combination of multi-regional input-output analysis and a complex network approach. Energy, 170: 1191-2101, https://doi.org/10.1016/j.energy.2018.12.164.

DOI

[47]
Taylor P J, Catalano G, Walker D R F, 2002. Measurement of the world city network. Urban Studies, 39(13): 2367-2376, https://doi.org/10.1080/00420980220080011.

DOI

[48]
Wang Cheng, Wang Maojun, Chai Qing, 2015. The relationship between centrality and power in the city network. Acta Geographica Sinica, 70(12): 1953-1972, https://doi.org/10.11821/dlxb201512008. (in Chinese)

DOI

[49]
Wang C W, Wang M, Lu M M, 2022a. Evolution of the Chinese industrial structure: A social network perspective. Technological Forecasting and Social Change, 184: 121972, https://doi.org/10.1016/j.techfore.2022.121972.

DOI

[50]
Wang Yixiao, Sun Bindong, Zhang Tinglin, 2022b. Urban producer service functions and geographical distance in Chinese city clusters: A perspective from network externalities. Geographical Research, 41(9): 2418-2432, https://doi.org/10.11821/dlyj020211066. (in Chinese)

[51]
Wu Zhengzheng, Wu Dianting, Yuan Jun et al., 2008. Study on the geographical concentration and contributing factors of Chinese automative industry. China Population, Resources and Environment, 18(1): 116-121. (in Chinese)

[52]
Yuan Kaihua, Peng Shuijun, Chen Hongwen, 2019. The facts and explanations of the domestic value chain driving the export value of China’s manufacturing industry. Economist, 249(9): 93-103, https://doi.org/10.16158/j. cnki.51-1312/f.2019.09.010. (in Chinese)

[53]
Zhan Yarong, Gu Renxu, 2022. Characteristics and multi-dimensional proximity mechanism of the online game industry cooperative network in China. Progress in Geography, 41(7): 1145-1155, https://doi.org/10.18306/ dlkxjz.2022.07.001. (in Chinese)

DOI

[54]
Zhang Kaihuang, Qian Qinglan, 2021. Characteristics and proximities mechanism of China’s new energy vehicle industry innovation network. Geographical Research, 40(8): 2170-2187, https://doi.org/10.11821/dlyj020 201065. (in Chinese)

DOI

[55]
Zhang Shaojun, Li Dongfang, 2009. Industrial transference region in a mode of global value chain: From the perspective of business cost and learning curve. Economic Review, 156(2): 65-72, https://doi.org/10.19361/j.er. 2009.02.010.

[56]
Zhang Y O, Gu Renxu, Ma Shang, 2019. Spatial characteristics and proximity mechanism of technology transfer among cities in China. Progress in Geography, 38(3): 370-382, https://doi.org/10.18306/dlkxjz.2019.03.007. (in Chinese)

DOI

[57]
Zhao Junzhu, Sun Tieshan, Li Guoping, 2014. Agglomeration and firm location choice of China’s automobile manufacturing industry. Acta Geographica Sinica, 69(6): 850-862, https://doi.org/10.11821/dlxb201406011. (in Chinese)

DOI

[58]
Zhao Xinzheng, Li Qiuping, Rui Shang et al., 2019. The characteristics of urban network of China: A study based on the Chinese companies in the Fortune Global 500 List. Acta Geographica Sinica, 74(4): 694-709, https://doi.org/10.11821/dlxb201904006. (in Chinese)

DOI

[59]
Zhao Ziyu, Wang Shijun, Chen Xiaofei, 2021. Beyond locality in restructuring the spatial organization of China’s automobile industry clusters under modular production: A case study of FAW-Volkswagen. Acta Geographica Sinica, 76(8): 1848-1864, https://doi.org/10.11821/dlxb202108003. (in Chinese)

DOI

[60]
Zhao Ziyu, Wang Shijun, Chen Xiaofei et al., 2022. Cross-regional relatedness and influencing factors of China’s automobile production network from the perspective of ‘global-local’: A case study of FAW-Volkswagen. Progress in Geography, 41(5): 741-754, https://doi.org/10.18306/dlkxjz.2022.05.001. (in Chinese)

DOI

[61]
Zhou Ruibo, Qiu Yifeng, Hu Yaozong, 2021. Characteristics, evolution and mechanism of inter-city innovation network in China: From a perspective of multi-dimensional proximity. Economic Geography, 41(5): 1-10, https://doi.org/10.15957/j.cnki.jjdl.2021.05.001. (in Chinese)

DOI

[62]
Zhou Tao, 2021. Research on Urban Spatial Evolution from the Perspective of Value Chain. Beijing: China Social Sciences Press.

Options
Outlines

/

〈 〉