Pol. J. Environ. Stud. Vol. 33, No.

2 (2024), 1767-1779
DOI: 10.15244/pjoes/174511 ONLINE PUBLICATION DATE: 2024-01-24

Original Research
Structural Evolution and Policy Orientation
of China’s Rare Earth Innovation Network:
A Social Network Analysis Based on
Collaborative Patents

Feng Hu1, Xiaojiao Shi2#, Shaobin Wei3, Liping Qiu4*, Hao Hu5**,
Haiyan Zhou6***, Bingnan Guo7****

Institute of International Business and Economics Innovation and Governance, Shanghai University of International
1

Business and Economics, Shanghai, China

2
School of Business Administration, Zhejiang Gongshang University, Hangzhou, China
3
China Center for Economic Research, East China Normal University, Shanghai, China
4
CEEC Economic and Trade Cooperation Institute, Ningbo University, Ningbo, China
5
School of Economics, Shanghai University, Shanghai, China
6
Graduate School, Nueva Ecija University of Science and Technology, Cabanatuan, Philippines
7
School of Humanities and Social Sciences, Jiangsu University of Science and Technology, Zhenjiang, China

Received: 3 September 2023

Accepted: 25 October 2023

Abstract

To clearly describe the structural characteristics of China’s rare earth innovation network (REIN),
this study used data on China’s rare earth patents since 2001 in the Incopat global patent database to
analyze the structural characteristics of China’s REIN from 2001 to 2020 at the provincial and municipal
levels using social network analysis. The study finds the following: (1) The overall characteristics of the
network show that the connectivity and diversification of China’s REIN is increasing. (2) The network
spatial pattern analysis shows that the REIN exhibits a radial spatial pattern centered on Beijing
at both the provincial and municipal levels. (3) Research on network nodes shows that the network
is dense in eastern regions and sparse in western regions, showing distinct characteristics of coastal
and resource agglomeration. Not all regions with high levels of economic development play leading roles
in the REIN. Small and medium-sized cities with abundant rare earth resources also play important

#
Co-first Author: Xiaojiao Shi contributed equally
to this work and share first authorship.
*e-mail: livy.qiu@hotmail.com
**e-mail: huhao@staff.shu.edu.cn
***e-mail: zhouhaiyansh@hotmail.com
****e-mail: 200600002509@just.edu.cn
1768 Hu F., et al.

leading roles due to their high levels of rare earth innovation. At the provincial and municipal levels,
the REIN consists of four subgroups, among which there is not significant factional diversity.
(4) In terms of influencing factors, economic development level, technology absorptive capacity,
innovation output level, infrastructure, and market leadership have significant impacts at the provincial
and municipal levels. In addition, government leadership, industrial structure, and education level have
significant impacts at the municipal level. The study clarified China’s technological accumulation in the
field of rare earths and provided practical guidance for the innovative development of China’s rare earth
industry.

Keywords: rare earth, innovation network, geographical detector, social network analysis

Introduction coefficient, network centralization, and modularity have

been used in studies [14]. In research on the position of
China is the world’s major rare earth country [1], innovation networks, most scholars have used indicators
accounting for half of the global rare earth reserves such as closeness centrality, betweenness centrality,
[2]. Since the 1980s, China has gradually replaced the and constraint to explore the position of nodes in
United States and become an important international innovation networks [15]. In the study of the relationships
supplier of rare earths [3]. After more than a decade in innovation networks, scholars have determined the
of development, China’s rare earth industry has made characteristics of knowledge flow between innovation
significant progress. However, due to the lack of network nodes based on the one-way or two-way
awareness of patent protection and rare earth innovation characteristics of the relationships between these nodes
in the early stages of China’s rare earth industry [4], and analyze the relationship characteristics within the
coupled with Japan’s rare earth patent barriers, China’s network based on collaborative strength [16].
overall position is in the middle-to-low end of the global As dynamic open systems, innovation networks are
rare earth value chain [5]. During the National People’s constantly evolving and optimized due to the changing
Congress and the Chinese People’s Political Consultative internal and external environments of an innovation
Conference in 2023, there was once again an emphasis network. Accordingly, the study of the dynamic
on “enhancing the scientific and technological evolution of networks has gradually become a research
innovation capabilities of the rare earth industry”. hotspot among scholars. The structural characteristics of
Protecting rare earth resources and promoting China’s networks have been studied mostly from organizational
transition from a major rare earth country to a rare earth dimensions (based on node types such as companies,
powerhouse has attracted widespread attention from the universities, and research institutes) [17], technological
Chinese government [6]. dimensions (including information technology,
Regarding the research on the rare earth industry, communication technology, new energy technology,
scholars have mainly analyzed, from the perspective and other technological categories) [18, 19], and spatial
of rare earth demanders, problems in global rare dimensions (including cross-national, cross-provincial,
earth supply [7], China’s role in the rare earth supply cross-municipal, and other cross-regional collaboration
and China’s rare earth policies [4, 8], the industry of and innovation) [20]. Regarding the research on factors
future global rare earth demand [9], the economic influencing innovation networks, scholars have mainly
and social issues caused by rare earth mining, and the discussed two aspects: the internal structural factors
importance of rare earth reserves [10]. However, the rare of the network and the external environment. In terms
earth industry seems to have not been widely used in of the influence of the internal network structure,
innovation network research. attributes such as the centrality and structural holes of
A review of the relevant literature finds a wealth of network nodes have a large influence on the connections
research on innovation networks, resulting in abundant between the nodes. The connection and even the
findings. Theories such as regional innovation system whole network will have a greater impact, and core
theory, complex network theory, and collaborative nodes in particular play a crucial role in driving the
innovation theory have provided a solid theoretical development of the entire network [21, 22]. As for the
foundation for the study of innovation networks [11]. factors related to the external environmental influence
Co-authored papers and co-invented patents serve as on networks, moderate proximity, such as cognitive
the primary sources of quantitative data for innovation proximity and social proximity, has an important impact
network research [12, 13]. Research on the structural on innovation networks [23]. In addition, the economic
characteristics of innovation networks mainly focuses and social environment where the nodes are located also
on the overall characteristics of the network, network significantly influences innovation networks [24].
positions, and network relationships. In the research Based on a review of previous studies, there is a lack
on the overall characteristics of innovation networks, of research on innovation networks from the provincial
indicators such as network density, average clustering perspective. Furthermore, research on innovation
Structural Evolution and Policy Orientation... 1769

networks has mostly focused on individual levels, e.g., as objectively as possible and to avoid differences
country, province, urban agglomeration, urban area, in some special years, a more scientific approach
or rural area [16], with few analyses of the structural network was adopted to demonstrate innovation
characteristics of China’s rare earth innovation development, that is, the study period was divided into
network (REIN) from a multi-level perspective. four-year intervals: 2001-2005, 2006-2010, 2011-2015,
Therefore, using China’s rare earth patent collaboration and 2016-2020.
data to explore the spatial dynamics of China’s rare
earth collaboration, problems can be identified from Research Methods
a novel perspective, thus helping to clarify China’s
current technological accumulation in the rare earth Social Network Analysis for the Structural Evolution
sector. In addition, an in-depth analysis of the pattern of China’s REIN
evolution of the REIN at two different spatial scales
(provincial and municipal levels) contributes to Social networks were used to analyze the changes
promoting the innovation and development of China’s in indicators (e.g., centrality, network density, average
rare earth industry and provides strategic guidance for clustering coefficient, and centralization) of the REIN to
its advancement. reveal its characteristics [25]. In particular, we evaluate
This study provides the following marginal node symmetry which was calculated based on the
contributions. First, this study enriches the practical concept of effective flow rate [26]:
connection between social network analysis methods
and innovation geography theory through two spatial
scales, provincial and municipal, in China’s REIN.
Secondly, the study of China’s REIN enriches the study (1)
of innovation network theory; finally, the analysis of
the influencing factors affecting China’s REIN provides where Ai is the effective flow rate of a province or city,
strategic basis for technological innovation by the Ii is the number of rare earth innovation collaborations
government and relevant functional departments of the in which province or city i participated, and Qi is the
rare earth industry. number of rare earth innovation collaborations led by
The structure of the other sections of this paper is province or city i. When Ai is -1, province or city i is a
as follows. Section 2 describes the research data and main participating region for rare earth innovation, and
methods, Section 3 resents the research results and when Ai is 1, province or city i is a core leading region
discussion, and Section 4 gives the research conclusions for rare earth innovation.
and policy suggestions.
Geographical Detector Analysis for the Influence
Factors of China’s REIN
Research Data and Methods
The geographical detector method is a spatial
analysis model used to explore the relationship between
Research Data
a certain geographical attribute and its explanatory
factors. It is widely applied to explain the influencing
The data in this study were derived from the
factors of various phenomena. The advantage of this
Incopat global patent database. The patent abstracts
method is that it is less constrained by prerequisites
in the database were searched for keywords related
when analyzing multiple types of data [27, 28].
to rare earths. First, patents that were not related
The equation is as follows:
to rare earths were excluded; second, patents with
natural persons as patent applicants, owned by foreign
companies, or with only one applicant were eliminated.
A total of 6435 patents from 2001 to 2020 met the (2)
requirements. In the analysis of patent collaborations,
a distinction was made between the first and non-first where q is the detection value of the influencing
patent applicants. Collaborations were divided into factor on the centrality of China’s REIN; h = 1, ..., L
two categories: collaborations led by the first patent denotes the classification of each factor of the variable;
applicant and collaborations involving non-first patent σ2 is the total variance in the centrality of China’s
applicants, forming a directed data matrix representing REIN at the provincial and municipal levels; σh2 is the
the collaborative innovation among patent applicants1. variance in the centrality of China’s REIN at provincial
To present the evolution characteristics of the REIN and municipal levels, respectively; N is the number of
provinces or cities in China; and Nh is the number of
types of the influencing factor X. The value range of q
is [0, 1], and a larger value indicates a greater influence
1
For example, if a patent was applied for by four applicants
of the factor on China’s REIN at the provincial and
A, B, C, and D, collaborations between A and B, between A
and C, and between A and D were each counted once.
municipal levels.
1770 Hu F., et al.

Results and Discussion to 2005, connections existed only at first to third tiers,
and from 2011 to 2015, connections were missing at the
Results third tier. Based on the spatial distribution, there existed
only a few high-tier (third and fourth tier) connections,
Analysis of the Overall Characteristics which were mainly concentrated in four provinces
of the Network (Beijing, Shanghai, Liaoning, and Hebei) in eastern
China. The second-tier connections only expanded
As seen in Table 1, the connection and diversification from coastal provinces in the east to central and
of China’s REIN are gradually increasing. Both at the western provinces in the 2016-2020 period, resulting
provincial or municipal levels, the number of nodes and in a generally sparse network in western regions.
connections in the network has increased substantially, The network density increased from 0.087 to 0.18,
indicating clear growth in the number of provinces indicating further improvement in the development
and cities participating in rare earth innovation. of the network. Specifically, the connection between
The outdegree centralization and indegree centralization Beijing and Shanghai was outstanding in all four
of the networks have been decreasing year by year, stages, with the connection strength increasing from
with the outdegree centralization being greater than 48 in the 2001-2005 period to 158 in the 2016-2020
the indegree centralization, a finding that indicates that period, far surpassing the strength of the second-ranked
the leading nodes in the REIN are more concentrated connection. In the 2016-2020 period, the connections
in a few provinces and prefecture-level cities, reflecting (strength) in the third tier were Beijing→Hebei (69)
a greater influence of the leading nodes on the REIN. and Beijing→Liaoning (46), and more than 60% of
The average clustering coefficient of the networks the connections in the second tier linkage lines were
shows an upward trend, indicating closer collaborative connected to Beijing, highlighting the importance of
innovation among the nodes in the network. At the Beijing in the REIN.
provincial level, the network modularity and average At the municipal level, the spatial pattern was
distance decreased from 0.272 and 2.336 in the 2001- basically similar to that at the provincial level. In
2005 period to 0.231 and 2.033 in the 2016-2020 terms of spatial distribution, the second tier expanded
period, indicating that in provincial-level collaboration from being only located in the eastern coastal cities
networks, the degree of group differentiation within to the central cities in the 2006-2010 period and
the network increased significantly over time, and the extended to the western cities between 2011 and 2015.
number of edges passed to complete collaboration Specifically, Beijing→Shanghai ranked as fourth tier
between provinces decreased, improving collaboration in all four stages. From 2011 to 2015, Beijing→Fushun
efficiency; in contrast, the municipal-level collaboration was also in this tier; from 2016 to 2020, the connections
networks exhibited the opposite trend, a finding that (strength) in the third tier included Xiamen→Longyan
could be related to factors such as administrative (81), Longyan→Xiamen (44), and Beijing→Langfang
boundaries and proximity. (3). In terms of administrative jurisdiction and
proximity, except for Xiamen and Longyan, the rest
Analysis of the Spatial Pattern of the city pairs are not in the same province and not
of the Networks bordering cities, indicating that geographic proximity
is not the main influencing factor affecting the REIN.
Using data from 2016 to 2020 as a baseline, the Comparing the network spatial pattern at the two levels,
REIN was divided into four tiers based on strength of except for Longyan-Xiamen, other city pairs were
connections using the natural breakpoint method in composed of cities in different provinces or composed
ArcGIS; then, the results were visualized. of non-bordering cities, suggesting that geographical
At the provincial level, the REIN basically formed proximity is not the major influencing factor for
a radial spatial pattern centered on Beijing. From 2001 the REIN.

Table 1. Overall characteristics of the REIN from 2001 to 2020.

Provincial level Municipal level
Indicator
2001-2005 2006-2010 2011-2015 2016-2020 2001-2005
Node 16 26 30 30 35
Edge 21 60 127 157 34
Modularity 0.272 0.249 0.262 0.231 0.457
Average clustering coefficient 0.168 0.282 0.369 0.38 0.019
Average path length 2.336 2.320 2.238 2.033 2.176
Out-degree centralization 12.278 9.131 8.098 8.919 5.448
In-degree centralization 6.204 3.931 3.599 3.794 2.862
Structural Evolution and Policy Orientation... 1771

Comparing the network spatial patterns at the two which ranked second. At the provincial level, Shanghai
levels, except for Longyan-Xiamen, the node pairs ranked in the second tier for centrality; 11 provinces,
with higher rare earth innovation connection strength including Guangdong, Jiangsu, Hebei, and Shandong,
are less geographically adjacent, indicating that the which are mainly located in the central and eastern
REIN overcomes the limitation of spatial geographical regions of China, ranked in the third tier for centrality;
distance. and the remaining provinces had relatively low centrality
and ranked in the fourth tier. At the municipal level,
Analysis of Network Node Characteristics Shanghai, Xiamen, and Longyan ranked in the second
tier for centrality; 14 cities, including Tianjin, Baotou,
To demonstrate the spatial characteristics of the Xi’an, Guangzhou, and Ganzhou, ranked in the third
most recent nodes in China’s REIN and due to the tier for centrality; and the remaining cities had relatively
space limitation of the article, the following sections low centrality and ranked in the fourth tier.
only analyze the period from 2016 to 2020 to provide Combining the REIN at the two levels, the REIN
constructive suggestions based on the latest node was clearly characterized by coastal and resource
characteristics. agglomeration. Specifically, provinces or cities in the
(1) Analysis of network node centrality first and second tiers for centrality were located in the
Using the natural breakpoint method, the centrality eastern coastal areas of China, and those in the third
of REIN nodes was divided into four tiers, as shown tier were located in the central and eastern regions of
in Fig. 3. As seen in Fig. 3 and Table 2, at both the China. Unlike previous studies, our study found that not
provincial and municipal levels, the centrality of Beijing all provinces or cities with high centrality in the REIN
ranked first and was much higher than that of Shanghai, had a high level of economic development. For example,

Fig. 1. Connection map of China’s REIN at the provincial level from 2001 to 2020.
1772 Hu F., et al.

Fig. 2. Connection map of China’s REIN at the municipal level from 2001 to 2020.

provinces such as Jiangxi and Inner Mongolia as well nodes in the REIN, and the net inflow nodes are the
as cities such as Longyan, Baotou, and Ganzhou are participating nodes. At the provincial level, there were
regions with abundant rare earth resources and relatively 14 provinces with a net outflow of rare earth innovation
high levels of rare earth research and technology, and 16 provinces with a net inflow. Among the net
including the Baotou Research Institute of Rare Earths outflow provinces, Ningxia, Yunnan, Beijing, Jiangxi,
in Baotou, Inner Mongolia, the Fujian Changting Golden Fujian and Hunan had high net outflows and played
Dragon Rare Earth Co., Ltd. in Longyan, Fujian, and the leading roles in rare earth patent collaboration and,
Institute of Rare Earths, Chinese Academy of Sciences hence, in the innovation network, and Jilin, Qinghai,
in Ganzhou, Jiangxi, which are authoritative rare earth Zhejiang, and Jiangsu had moderate net outflows.
research institutions in China. In addition, although Among the net inflow provinces, Hebei, Hainan, Shanxi,
Fujian ranked in the fourth tier for centrality among Xinjiang, and Guangxi had high net inflows, indicating
provinces, Longyan and Xiamen were in the second tier that these provinces played participatory roles in patent
for centrality, and other cities were in the fourth tier for collaboration, and Guangdong, Shanghai, and Liaoning
centrality, reflecting substantial polarization in the level had moderate net inflows, with Shanghai’s centrality
of rare earth innovation collaboration among cities in ranking in the second tier. At the municipal level,
Fujian. there were 73 cities with a net outflow of rare earth
(2) Node symmetry analysis of the network innovation and 94 cities with a net inflow. Cities with
The natural breakpoint method was used to divide high net outflows and cites with high net inflows were
the node symmetry of the REIN into four tiers, as relatively scattered in spatial distribution. Cities with
shown in Fig. 4. The net outflow nodes are the leading high net outflows included Zhuhai, Fuzhou, Quanzhou
Structural Evolution and Policy Orientation... 1773

Fig. 3. Centrality of nodes in China’s REIN from 2016 to 2020.

Table 2. Centrality of nodes in China’s REIN from 2016 to 2020 (Top 30).
Provincial level Municipal level
Province Centrality Province Centrality City Centrality City Centrality
Beijing 578 Hubei 30 Beijing 578 Foshan 41
Shanghai 267 Sichuan 25 Shanghai 267 Langfang 39
Guangdong 121 Henan 24 Xiamen 140 Dalian 38
Jiangsu 117 Yunnan 23 Longyan 132 Jinan 31
Hebei 99 Jilin 15 Tianjin 71 Shenyang 30
Shandong 98 Gansu 14 Baotou 70 Shaoguan 30
Zhejiang 80 Guangxi 11 Xi’an 58 Jinhua 25
Liaoning 71 Chongqing 10 Guangzhou 57 Wuhan 25
Tianjin 71 Shanxi 7 Ganzhou 56 Baoding 23
Inner
70 Heilongjiang 7 Shenzhen 56 Hefei 23
Mongolia
Shaanxi 70 Qinghai 5 Suzhou 55 Kunming 22
Jiangxi 64 Guizhou 2 Nanjing 53 Xiangtan 21
Anhui 63 Hainan 1 Ningbo 48 Nantong 21
Hunan 61 Xinjiang 1 Changsha 47 Wuxi 20
Fujian 32 Ningxia 1 Hangzhou 45 Nanchang 20

and other cities with relatively high levels of economic (3) Cohesive subgroups of network nodes
development as well as cities with average levels of A cohesive subgroup analysis of China’s REIN from
economic development such as Baiyin and Qiannan. 2016 to 2020 was conducted using UCINET software.
Cities with high net inflows included 74 cities, such as As seen in Fig. 5, the REIN had four types of subgroups
Liangshan, Jilin, Changzhou, and Baoding. at both the provincial and municipal levels. The macro-
In general, not all provinces or nodes with high structure of the network and the E-I index analysis
levels of economic development played leading roles in results are shown in Table 4.
the REIN. Small and medium-sized cities with abundant At the provincial level, the difference in the number
earth resources also played important leading roles due of provinces among subgroups increased. More than
to their high level of rare earth innovation. 53% of the provinces were concentrated in the second
1774 Hu F., et al.

Fig. 4. Node symmetry of China’s REIN from 2016 to 2020.

subgroup, with Beijing as the core. This subgroup Baotou. The first, third, and second subgroups accounted
radiated to drive innovative collaboration in rare earths for 27%, 22%, and 20%, respectively, with Longyan,
with provinces such as Zhejiang and Shandong. Notably, Xiamen, and Beijing as their respective cores. There
among the provinces that radiated, six were among was no close connections within each subgroup; among
the top ten in terms of centrality, and there were close the subgroups, close connections existed only between
internal connections within the subgroup but sparse subgroups 4 and 2 as well as between subgroups 1
connections outside of the subgroup. The first subgroup and 3.
was mainly centered around Shanghai and spread to Combining the cohesive subgroups of the REIN at
provinces such as Inner Mongolia, accounting for 20% the two levels, strong subgroups (provincial subgroup 2
of the total. Both internal and external connections and municipal subgroup 4) formed between the “strong”
within this subgroup were close. The third and fourth nodes in the REIN. At the provincial level, subgroup 2
subgroups were centered around Guangdong and had close internal and external connections, but at the
Jiangsu, respectively, with each accounting for 13.3% of municipal level, subgroup 4 was only closely connected
the total. to subgroup 2, and its internal connections were
At the municipal level, the differences in the number relatively weak.
of cities among subgroups were smaller than those at As seen in Table 3, at the provincial level, the E-I
the provincial level. More than 30% of the cities were indices of subgroups 1, 3, and 4 and the overall network
concentrated in the fourth subgroup, with Shanghai were all positive, the nodes in these subgroups had
as its core, and radiated to drive rare earth innovation significantly more external connections than external
connections in the top ten cities in centrality, such as connections, with no obvious factional diversity, and
Shenzhen, Guangzhou, Ganzhou, Tianjin, Xi’an, and they had great potential for development. Compared

Table 3. Density matrix and image matrix of cohesive subgroups of nodes in China’s REIN from. 2016 to 2020.
Density matrix Image matrix
Level Subgroup
Subgroup 1 Subgroup 2 Subgroup 3 Subgroup 4 Subgroup 1 Subgroup 2 Subgroup 3 Subgroup 4
Subgroup 1 5.80 3.45 4.67 5.50 1 0 1 1

Provincial Subgroup 2 11.30 7.64 4.77 2.60 1 1 1 0

level Subgroup 3 6.00 4.00 0.00 4.00 1 0 0 0
Subgroup 4 6.25 3.00 2.00 0.00 1 0 0 0
Subgroup 1 0.60 0.44 4.67 1.23 0 0 1 0

Municipal Subgroup 2 0.83 1.27 1.15 7.08 0 0 0 1

level Subgroup 3 8.67 2.43 0.86 1.64 1 0 0 0
Subgroup 4 1.51 3.18 1.64 2.60 0 1 0 0
Structural Evolution and Policy Orientation... 1775

Fig. 5. Cohesive subgroups of nodes in China’s REIN from 2016 to 2020.

Note: Only the names of the top ten nodes in terms of centrality are shown in the figure

with nodes in other subgroups, the nodes in subgroup 2 per capita (to characterize the economic development
had the strongest internal and external innovation level), number of industrial enterprises above designated
connections, but their internal connections were slightly size (to characterize the technology absorptive capacity),
weaker than their external connections, indicating local scientific and technological financial resources (to
a trend toward the development of small groups. At the characterize government leadership), number of patents
municipal level, all subgroups had positive E-I indices, granted (to characterize the innovation output level),
and there was no formation of small groups within the number of books in public libraries (to characterize
subgroups. To avoid information barriers and factions the infrastructure), total retail sales of consumer goods
caused by the excessive formation of small groups, (to characterize the market leadership), proportion of
which may hinder the overall development of subgroups, the tertiary industry output value (to characterize the
subgroup 2 at the provincial level should strengthen industrial structure), and number of college and junior
exchanges with external nodes and diversify the college graduates (to characterize the higher education
development of the patent collaboration network. level). The geographical detector method was used
to analyze and calculate the correlation coefficients
between the provincial-level and municipal-level
Discussion network centrality and its influencing factors. Before
the analysis using the geographical detector method,
Selection of Influencing Factors each variable was divided into five levels using natural
breakpoints in ArcGIS. Due to potential endogeneity
Following the principles of comprehensiveness and issues, the data for all these explanatory variables
data availability and drawing on the research views of were from the end year, namely 2020. The data for the
previous scholars [21, 22, 29, 30], the centrality of the influencing factors were all obtained from the China
provincial-level and municipal-level network from 2016 Statistical Yearbook and China City Statistical Yearbook
to 2020 was selected as the explained variable, and the for each year.
following were selected as explanatory variables: GDP

Table 4. Results of the E-I faction analysis of cohesive subgroups of nodes in China’s REIN from 2016 to 2020.
Provincial level Subgroup 1 Subgroup 2 Subgroup 3 Subgroup 4 Entire network
Number of internal connections 6 68 0 0 74
Number of external connections 38 67 27 30 162
EI index 0.727 -0.007 1 1 0.373
Municipal level Subgroup 1 Subgroup 2 Subgroup 3 Subgroup 4 The entire network
Number of internal connections 6 10 20 124 160
Number of external connections 63 121 104 162 450
E-I index 0.826 0.847 0.677 0.133 0.475
1776 Hu F., et al.

Analysis of the Results the number of industrial enterprises in a region also

reflects the technology absorptive capacity of the region
Taking the top three decisive factors as the core to some extent, with a stronger technology absorptive
influencing factors, as seen in Table 5, at the provincial capacity in a region leading to a higher level of rare
level, the core influencing factors included innovation earth innovation. Although indicators such as local
output level, infrastructure, and economic development government science and technology expenditure had
level, and the secondary influencing factors included explanatory powers greater than 0.3, they did not pass
technology absorptive capacity and market leadership; the significance test, indicating that some influencing
at the municipal level, the core influencing factors factors, such as government leadership, have little
included government leadership, innovation output importance.
level, and infrastructure, and the secondary influencing At the municipal level, all indicators passed the
factors were other influencing factors. significance test at the 1% level, indicating that these
At the provincial level, the number of patents influencing factors have a significant impact on
granted and the number of books in public libraries had China’s REIN. Similar to those at the provincial level,
the most explanatory power and both were significant infrastructure, market leadership, and innovation output
at the 1% level, indicating that innovation output level had important influences. There were differences
level and infrastructure are the two most important between the municipal level and the provincial level.
factors in the REIN at the provincial level. A higher The influence of GDP per capita and number of large-
level of local innovation output leads to a stronger sized industrial enterprises was quite different between
local technological innovation capability, and a well- the municipal level and the provincial level, and the
developed infrastructure and a favorable innovation spatial consistency of the economic development
environment provide strong impetus for the generation level and technology absorptive capacity with the
and dissemination of rare earth innovation. GDP per centrality of REIN nodes differed significantly at
capita, total retail sales of consumer goods, and number different levels. In contrast, the differences in economic
of large-sized enterprises had explanatory powers of development level and technology absorptive capacity
0.605, 0.582 and 0.452, respectively, significant at the at the municipal level were small, which is conducive
5% level, indicating that economic development level, to reducing the centrality gap in the REIN at the
market leadership, and technology absorptive capacity municipal level in the future. The influence of local
are important to the provincial-level REIN. Regions science and technology expenditure, the proportion of
with higher levels of economic development can provide the tertiary industry output value, and the number of
a more abundant material basis for relevant innovative college and junior college graduates at the municipal
activities in rare earth research. Innovation also relies level was more significant than those at the provincial
on market demand, and rare earth elements as industrial level, indicating substantial differences in government
vitamins play an irreplaceable role in the development leadership, industrial structure, and educational level
of high-tech industries. Industrial enterprises are the within the municipal level, further widening the gap
main players for patent absorption and application, and in the centrality of the REIN at the municipal level in

Table 5. Results of the regression using the geographical detector method.

Provincial level Municipal level
Influencing factors Indicator Explanatory Significance Explanatory Significance
power level power level
Economic development
GDP per capita 0.605 0.016 0.361 0.000
level
Technology absorptive Number of industrial enterprises above
0.452 0.080 0.353 0.000
capacity designated size
Local science and technology
Government leadership 0.514 0.158 0.551 0.000
expenditure
Innovation output level Number of patents granted 0.656 0.002 0.576 0.000
Infrastructure Number of public library books 0.672 0.000 0.615 0.000
Total retail sales of social consumer
Market leadership 0.582 0.010 0.549 0.000
goods
The proportion of the tertiary industry
Industrial structure 0.322 0.479 0.275 0.000
output value
The number of college and junior
Education level 0.326 0.174 0.358 0.000
college graduate
Structural Evolution and Policy Orientation... 1777

the future. First, the uneven distribution of rare earths output level, infrastructure, and economic development
has led to significant variations in the formulation of level, and the secondary influencing factors include
relevant policies and research priorities on rare earths in technology absorptive capacity and market leadership.
different cities. For example, the city of Baotou boasts At the municipal level, the core influencing factors
of a number of rare earth innovation platforms, such as include government leadership, innovation output
the National Rare Earth Functional Material Innovation level, and infrastructure, and the secondary influencing
Center, Baotou Research Institute of Rare Earths, and factors consist of other influencing factors.
Baotou Rare Earth R&D Center, the Chinese Academy
of Sciences. The tertiary industry is more attractive than Policy Implications
the primary and secondary industries for the generation
and cultivation of rare earth innovations due to its (1) Market leadership and government leadership
accumulation of more highly educated talent. The same should be combined. Government-led and market-led
is true for areas with high education levels. approaches play important roles in the development of
rare earth innovation. However, it is necessary to be
cautious of excessive government intervention and free
Conclusions and Policy Implications market leadership. Only with proper regulation and
balance between the two can efficient and sustainable
Conclusions collaboration in rare earth innovation be ensured.
(2) It is necessary to build a cross-city rare earth
To clearly describe the structural characteristics of innovation platform. From the overall characteristics
China’s REIN, this study used data on China’s rare earth of the network, it can be found that the overall
patents in the Incopat global patent database since 2001 connection density of the REIN is low at the municipal
to analyze the structural characteristics of China’s REIN level, necessitating urgent improvement in the overall
between 2001 and 2020 from both the provincial and connection level of the network. At the national level,
municipal perspectives using social network analysis. it is necessary to establish a cross-city platform for
The degree of connection and diversification of the REIN, set up cross-city collaboration funds,
China’s REIN is gradually increasing. A few provinces improve policy support for cross-city collaboration,
and cities lead the REIN. The group differentiation strengthen institutional supply, reduce the cost of rare
of the network at the provincial level has decreased earth innovation collaboration, and thereby lower the
significantly, resulting in improved collaboration barriers to collaboration and innovation in the rare earth
efficiency. However, the opposite trend is observed at industry.
the municipal level. The REIN exhibits a radial spatial (3) It is necessary to improve regional technology
pattern with Beijing as the center at both the provincial absorptive capacity and education level. Technology
and municipal levels. Overall, the network is dense in absorptive capacity and education level are key
the eastern regions and sparse in the western regions, capabilities that nodes in the REIN need to improve.
indicating that the REIN has overcome the limitations Each region can achieve this by attracting talent to
of spatial and geographical distances. the rare earth sector, optimizing training mechanisms,
The REIN exhibits distinct characteristics of increasing financial investment in education, and
coastal and resource agglomeration. Specifically, the enhancing training efforts, particularly in higher
provinces or cities in the first and second tiers for education, which will subsequently enhance the region’s
centrality are located in the eastern coastal areas of innovation and construction capabilities in the rare
China, and those in the third tier are located in the earth knowledge network. By strengthening innovation
central and eastern regions of China. Overall, not exchanges and collaboration with core regions such as
all provinces or nodes with high levels of economic Beijing, Shanghai, Xiamen, and Longyan, a region can
development play leading roles in the REIN. Small enhance its capacity to absorb and transform rare earth
and medium-sized cities with abundant rare earth knowledge.
resources also play important leading roles due to their
high level of rare earth innovation. The REIN has four Theoretical Contributions
subgroups at both the provincial and municipal levels,
and there is no significant factional diversity within (1) This study can be seen as a deepening of the
these subgroups. study by [5]. This study investigates China’s REIN at
In terms of influencing factors, economic two spatial scales, namely the provincial and municipal
development level, technology absorptive capacity, levels, and thus adds a new dimension to the previous
innovation output level, infrastructure, and market innovation network research methods. The results
leadership have significant impacts at the provincial and reveal the pattern characteristics of the innovation
municipal levels. In addition, government leadership, network in different dimensions, break the limitation of
industrial structure, and education level have significant traditional research on innovation networks from single
impacts at the municipal level. At the provincial level, dimensional perspectives, and enrich the theoretical
the core influencing factors include the innovation system of innovation geography.
1778 Hu F., et al.

(2) This study integrates the content of the REIN into References
the existing regional innovation system theory, complex
network theory, and collaborative innovation theor and 1. SHUAI J., ZHAO Y., SHUAI C., WANG J., YI T., CHENG
examines the spatial patterns and influencing factors of J. Assessing the international co-opetition dynamics of
China’s REIN from a dynamic perspective, enriching rare earth resources between China, USA, Japan and the
the relevant theories of the REIN and contributing to the EU: An ecological niche approach. Resources Policy. 82,
103446, 2023.
theoretical innovation and practice of REIN research.
2. WÜBBEKE J. Rare earth elements in China: Policies and
(3) This study explores the formation mechanism narratives of reinventing an industry. Resources Policy. 38,
of the REIN at two spatial (provincial and municipal) (3), 384, 2013.
levels. Using the geographical detector method, this 3. HU X., SUN B., WANG C., LIM M. K., WANG P., GENG
study investigates various aspects such as economic X., YAO C., CHEN W.-Q. Impacts of China’s exports
development level, technology absorptive capacity, decline in rare earth primary materials from a trade
and government leadership, providing new ideas and network-based perspective. Resources Policy. 81, 103321,
approaches for innovation network research, and is of 2023.
great reference significance for promoting the high- 4. SHUAI J., PENG X., ZHAO Y., WANG Y., XU W.,
CHENG J., LU Y., WANG J. A dynamic evaluation
quality development of China’s rare earth industry
on the international competitiveness of China’s rare
[8, 21]. earth products: An industrial chain and tech-innovation
perspective. Resources Policy. 75, 102444, 2022.
Research Limitations and Prospects 5. LENG Z., SUN H., CHENG J., WANG H., YAO Z. China’s
rare earth industry technological innovation structure
(1) In terms of research data, this study only uses and driving factors: A social network analysis based on
patent collaboration data to illustrate the REIN, and patents. Resources Policy. 73, 102233, 2021.
the data source is relatively limited. In the future, 6. SHU W., LI F., ZHANG Q., LI Z., QIAO Y., AUDET
comprehensive research can be conducted using sources J., CHEN G. Pollution caused by mining reshaped
the structure and function of bacterial communities
such as co-authored papers, patent transfers, and parent-
in China’s largest ion-adsorption rare earth mine
subsidiary company data to improve the scientific rigor watershed. Journal of Hazardous Materials. 451, 131221,
and accuracy of the research. 2023.
(2) Regarding spatial scales, the characteristics of 7. THIBEAULT A., RYDER M., TOMOMEWO O., MANN
innovation networks differ across spatial scales due to M. A review of competitive advantage theory applied to
the differences in the economic and social environments the global rare earth industry transition. Resources Policy.
at different levels. This paper only examines the related 85, 103795, 2023.
issues of the REIN at the provincial and municipal 8. GUO Q., YOU W. A comprehensive evaluation of the
levels. In the future, REINs across a broader range of international competitiveness of strategic minerals in
China, Australia, Russia and India: The case of rare earths.
spatial scales, such as global-local, city clusters, and
Resources Policy. 85, 103821, 2023.
county levels, should be analyzed. 9. WANG J., GUO M., LIU M., WEI X. Long-term outlook
(3) Research methods need to be further improved. for global rare earth production. Resources Policy. 65,
Firstly, this study uses social network analysis to 101569, 2020.
investigate China’s REIN using rare earth patent 10. SALIM H., SAHIN O., ELSAWAH S., TURAN H.,
collaboration data. In a follow-up study, case studies and STEWART R. A. A critical review on tackling complex
questionnaire interviews will be conducted to further rare earth supply security problem. Resources Policy. 77,
examine rare earth innovation research. Secondly, 102697, 2022.
this study uses the geographical detector method to 11. FREEMAN C. Networks of innovators: a synthesis of
research issues. Research policy. 20 (5), 499, 1991.
analyze the factors which affecting the China’s REIN,
12. GEORGE S., LATHABAI H. H., PRABHAKARAN T.,
and econometric method can be adopted in future CHANGAT M. A framework for inventor collaboration
research to discuss the influencing mechanism of REIN recommendation system based on network approach.
in China. Expert Systems with Applications. 176, 114833, 2021.
13. TSAY M.-Y., LIU Z. Analysis of the patent cooperation
network in global artificial intelligence technologies based
Acknowledgments on the assignees. World Patent Information. 63, 102000,
2020.
This work was supported by the National Natural 14. ITO H., HANAOKA S., SUGISHITA K. Structural
changes in the cruise network by ship size in Northeast
Science Foundation of China (Grant number 72373135),
Asia. The Asian Journal of Shipping and Logistics. 38, (4),
the Humanity and Social Science Foundation of Ministry 207, 2022.
of Education of China (Grant number 22YJAZH027). 15. BENI S. A., SHEIKH-EL-ESLAMI M.-K. Market power
assessment in electricity markets based on social network
analysis. Computers & Electrical Engineering. 94, 107302,
Conflict of Interest 2021.
16. HU F., QIU L., WEI S., ZHOU H., BATHUURE I.A.,
The authors declare there is no conflict. HU H. The spatiotemporal evolution of global innovation
Structural Evolution and Policy Orientation... 1779

networks and the changing position of China: a social 24. LIU Y., SHAO X., TANG M., LAN H. Spatio-
network analysis based on cooperative patents. R&D temporal evolution of green innovation network and its
Management. 2023. multidimensional proximity analysis: Empirical evidence
17. ZHAO Y., WEN S., ZHOU T., LIU W., YU H., XU H. from China. Journal of Cleaner Production. 283, 124649,
Development and innovation of enterprise knowledge 2021.
management strategies using big data neural networks 25. KIM J.-H., HASTAK M. Social network analysis:
technology. Journal of Innovation & Knowledge. 7 (4), Characteristics of online social networks after a disaster.
100273, 2022. Int. J. Inf. Manag. 38, 86, 2018.
18. LI F., LIU W., BI K. Exploring and visualizing spatial- 26. STILLWELL J., HUSSAIN S. Exploring the Ethnic
temporal evolution of patent collaboration networks: A Dimension of Internal Migration in Great Britain using
case of China’s intelligent manufacturing equipment Migration Effectiveness and Spatial Connectivity. Journal
industry. Technology in Society. 64, 101483, 2021. of Ethnic and Migration Studies. 36, 1381 2010.
19. HU F., QIU L., XIANG Y., WEI S., SUN H., HU H., 27. LI S.S., LIU C., SUN P.-P., NI T. Response of
WENG X., MAO L., ZENG M. Spatial network and cyanobacterial bloom risk to nitrogen and phosphorus
driving factors of low-carbon patent applications in China concentrations in large shallow lakes determined through
from a public health perspective. Frontiers in Public geographical detector: A case study of Taihu Lake, China.
Health. 11, 1121860, 2023. Science of The Total Environment. 2021.
20. YE Q., XU X. Determining factors of cities’ centrality 28. SONG Y., WANG J., GE Y., XU C. An optimal
in the interregional innovation networks of China’s parameters-based geographical detector model enhances
biomedical industry. Scientometrics. 126, 2801 2021. geographic characteristics of explanatory variables for
21. GU W., LUO J.-D., LIU J. Exploring Small-World Network spatial heterogeneity analysis: cases with different types of
with an Elite-Clique: Bringing Embeddedness Theory spatial data. GIScience & Remote Sensing. 57, 593 2020.
into the Dynamic Evolution of a Venture Capital Network. 29. QIU L., YU R., HU F., ZHOU H., HU H. How can
ArXiv. abs/1811.07471, 2018. China’s medical manufacturing listed firms improve their
22. HENNING M., MCKELVEY M. Knowledge, technological innovation efficiency? An analysis based
entrepreneurship and regional transformation: contributing on a three-stage DEA model and corporate governance
to the Schumpeterian and evolutionary perspective on the configurations. Technological Forecasting and Social
relationships between them. Small Business Economics. Change. 194, 122684, 2023.
54, 495, 2020. 30. WANG K., TAO S. Why Do Chinese Private Enterprises
23. BALLAND P.-A. Proximity and the Evolution of Seek Outward Foreign Direct Investment? China & World
Collaboration Networks: Evidence from Research and Economy. 31 (4), 200, 2023.
Development Projects within the Global Navigation
Satellite System (GNSS) Industry. Regional Studies. 46,
741, 2012.