21
Dynamic patterns of technology collaboration: a case study of the Chinese automobile industry, 1985–2010 Yuandi Wang Jian Li Lutao Ning Deming Zeng Xin Gu Received: 7 October 2013 / Published online: 6 July 2014 Ó Akade ´miai Kiado ´, Budapest, Hungary 2014 Abstract To investigate patterns of technology collaboration within the Chinese auto- mobile industry, this study employs a unique dataset of patent applications that reveal a record of 64,938 collaborative relations in the industry during the period from 1985 to 2010. Our results indicate that over 60 % of the total collaborations were conducted after China entered the WTO. The invention and utility types of patents account for 98 % of the total collaborations throughout the sample period. Using a network analysis method, we find that the key differences between domestic enterprises collaborating with indigenous enterprises (DD collaboration) and with foreign firms (DF collaboration) are in patent types and technology domains. The DF network is also denser and more centralized than the DD network, although the amount of nodes and links of the DD network is greater than that of the DF collaboration network. The analysis and visualization of the collaboration networks and corresponding largest components reveal that a large number of domestic enterprises prefer to collaborate with top global automobile manufacturers. We also find that a number of universities have become key players in the collaborations among industry, universities and research institutes. This study provides a deeper understanding of technology col- laborations from various perspectives and also highlights several avenues for future research. Keywords Collaboration Network Patent Evolution process Automobile industry China Y. Wang X. Gu Business School of Sichuan University, Sichuan University, No. 29, Wangjiang Road, Chengdu 610064, Sichuan, China J. Li (&) D. Zeng School of Business Administration, Hunan University, No.11, Lushan South Road, Changsha 410082, Hunan, China e-mail: [email protected] L. Ning School of Business and Management, Queen Mary, University of London, Francis Bancroft Building, Mile End Road, London E1 4NS, UK 123 Scientometrics (2014) 101:663–683 DOI 10.1007/s11192-014-1366-8

Dynamic patterns of technology collaboration: a case study

  • Upload
    others

  • View
    1

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Dynamic patterns of technology collaboration: a case study

Dynamic patterns of technology collaboration: a casestudy of the Chinese automobile industry, 1985–2010

Yuandi Wang • Jian Li • Lutao Ning • Deming Zeng • Xin Gu

Received: 7 October 2013 / Published online: 6 July 2014� Akademiai Kiado, Budapest, Hungary 2014

Abstract To investigate patterns of technology collaboration within the Chinese auto-

mobile industry, this study employs a unique dataset of patent applications that reveal a

record of 64,938 collaborative relations in the industry during the period from 1985 to

2010. Our results indicate that over 60 % of the total collaborations were conducted after

China entered the WTO. The invention and utility types of patents account for 98 % of the

total collaborations throughout the sample period. Using a network analysis method, we

find that the key differences between domestic enterprises collaborating with indigenous

enterprises (DD collaboration) and with foreign firms (DF collaboration) are in patent types

and technology domains. The DF network is also denser and more centralized than the DD

network, although the amount of nodes and links of the DD network is greater than that of

the DF collaboration network. The analysis and visualization of the collaboration networks

and corresponding largest components reveal that a large number of domestic enterprises

prefer to collaborate with top global automobile manufacturers. We also find that a number

of universities have become key players in the collaborations among industry, universities

and research institutes. This study provides a deeper understanding of technology col-

laborations from various perspectives and also highlights several avenues for future

research.

Keywords Collaboration � Network � Patent � Evolution process � Automobile industry �China

Y. Wang � X. GuBusiness School of Sichuan University, Sichuan University, No. 29, Wangjiang Road,Chengdu 610064, Sichuan, China

J. Li (&) � D. ZengSchool of Business Administration, Hunan University, No.11, Lushan South Road,Changsha 410082, Hunan, Chinae-mail: [email protected]

L. NingSchool of Business and Management, Queen Mary, University of London, Francis Bancroft Building,Mile End Road, London E1 4NS, UK

123

Scientometrics (2014) 101:663–683DOI 10.1007/s11192-014-1366-8

Page 2: Dynamic patterns of technology collaboration: a case study

Introduction

Technology innovation has become a predominant driving force for economic develop-

ment in the knowledge economy era (Schmiele 2012; Yam et al. 2011). How firms choose

an innovation mode that gives them comparative advantages in both local and global

markets has become a key question in recent literature (Cassiman and Veugelers 2006;

Enkel et al. 2009; Von Hippel 2007). Chesbrough (2003) introduced the concept of open

innovation as firms search for external organizations with business models that are better

suited to commercialize a certain technology rather than relying entirely on the firms’

internal R&D efforts. An increasing body of studies have suggested that innovators may

need to link themselves to external actors more than ever to obtain network resources and

opportunities for their own innovative activities (Ahuja 2000; Gulati 1999; Inkpen and

Tsang 2005; Walter et al. 2007). This is because resources, such as human capital, financial

investment, and social capital, have become increasingly important for innovation, and the

lack of such resources may constrain the pace of innovation. Moreover, the uncertainties

and risks associated with in-house R&D produces obstacles for both R&D investment and

innovative output (Becker and Dietz 2004; Howells et al. 2008; Narula 2004).

Previous literature has primarily focused on identifying factors that can influence the

decisions of technology collaboration, such as research interests, knowledge background,

research capability, infrastructure, geopolitical location, cultural relations and languages

(Glanzel 2001; Glanzel and Schubert 2001; Schubert and Glanzel 2006; Zhou et al. 2013).

However, some scholars have indicated that the basis of technology collaboration is, in fact,

collective innovation (Allen 1983; Saxenian 1994; Schrader 1991; Von Hippel 1987). This

collective innovation is usually driven by sharing and exchange of information and knowledge

within collaboration networks that consist of groups of socially connected entities (Dahl and

Pedersen 2004). The characteristics of the collaboration network structure therefore determines

the extent to which information and knowledge can be diffused and exchanged (Cowan and

Jonard 2004; Eslami et al. 2013), and this has a huge implication for firms’ innovation success.

Although the benefits of technology collaboration and the effect of network structure

have been extensively discussed in recent years, the extant literature overwhelmingly

focuses on studying the general features of effective collaboration network structures. It is

often ignored that the heterogeneity and idiosyncratic characteristics of different types of

links between various technology collaborators might exert diverse influences on the

collaboration output (Bierly et al. 2009; Gulati 1999; Li 2011; Uzzi 1996). Therefore, more

efforts are needed to reveal the industry-specific collaboration modes and how they have

changed with the evolution of the industry.

This study aims to extend the strand of general literature on the collaboration network

structure by systematically analyzing the pattern of different types of industry-specific

technology collaboration. For our analysis, we choose the Chinese automobile sector based

on a unique dataset of patent applications in the industry from 1985 to 2010. The devel-

opment of the Chinese automobile industry started in 1953. Indigenous automobile manu-

facturers in that phase focused on assimilation of technology from former Soviet Union until

the implementation of ‘opening up’ policy in 1978, which provided access to technologies

from overseas top automobile manufacturers, such as General Motors, Volkswagen, Toyota,

etc. The National Work Conference on Automobiles held in 1991 confirmed the development

strategy of automobile sector, which switched its focus from motor truck to car manufac-

turing. The biggest joint venture in the Chinese automobile sector was established in the

same year. From then on, new modes of technology improvement, such as collaboration with

overseas partners, began to emerge. Since China entered the WTO in December 11th, 2001,

664 Scientometrics (2014) 101:663–683

123

Page 3: Dynamic patterns of technology collaboration: a case study

the trend of globalization brought huge foreign investment into the automobile sector.

Moreover, the implementation of the Policy of Automobile Sector Development in 2004

encouraged private consumption of cars, which has produced immense market demand in the

automobile market. Many Chinese self-owned brands, e.g., Geely, Chery, BYD, and Great

Wall, reaped valuable strategic development opportunities during this period. Since 2010, the

automobile market has entered into the ‘micro increase era.’ The adjustment of product

structure, technology structure and brand structure have been the top issues for the Chinese

automobile manufacturers. Overall, China is now the largest vehicle-producing country,

having overtaken the EU and US in 2010 (Bloomberg 2010). The automobile industry is

considered to be one of the ‘pillar industries’ because of its extensive links with other

industries and significant contribution to job creation (Wang 2003). Since China opened up

and removed protectionist industrial policies, foreign enterprises have begun to play a critical

role in the growth of the sector (Liu and Dicken 2006; Sun et al. 2010). Relatively little

though is known about the pattern of collaborations and how their collaborative links have

changed over time between various actors, i.e., collaboration between domestic enterprises,

collaboration between domestic and foreign enterprises, and collaboration among enter-

prises, universities and research institutes.

The remainder of this study is organized as follows. In the next section, we describe the

data sources and main indicators for measuring collaboration networks. The descriptive

analysis of the pattern of collaborative relations and the analysis of the evolution of three

types of collaboration network based on social network analysis (SNA) techniques are

presented in the Results section. The Discussion and Conclusions section reviews the main

implications and summarizes the findings of this study. Research limitations and avenues

for future research discussed at the end of this paper.

Data sources and methods

Data

The data used in this study was collected from the Patent Information Services Platform

(PISP) that organized and supported by the State Intellectual Property Office of PRC (SIPO).

The PISP website (http://www.chinaip.com.cn/) provides patent information for ten national

key industries in China, i.e., the automobile, steel, electronic information, logistics, textile,

equipment manufacturing, ferrous metal, light industry, petrochemical, and shipbuilding

industries. The patent data for automobile sector was downloaded in November 2012. As the

joint application is a crucial form for linking different innovators (Motohashi and Yun 2007),

we omit those patents applied for by single applicant. We define that a collaborative relation

existed between applicant A and applicant B if they jointly applied for one patent during the

period of 1985–2010. After tidying up and cleaning the patent data, we finally obtained a total

of 64,938 collaborative relations in the Chinese automobile sector over 26 years.

SNA on collaborative relations

Collaboration between innovators is commonplace for the purpose of innovation. According

to prior studies (Newman 2001), collaborative efforts can be formalized in terms of a network

in which any two nodes are connected if they are involved in a common project. Nodes can be

individuals (Cattani and Ferriani 2008), organizations (Schilling and Phelps 2007), regions

(Gao et al. 2011), or even nations (Dodgson 2009). Innovation networks can be analyzed by

Scientometrics (2014) 101:663–683 665

123

Page 4: Dynamic patterns of technology collaboration: a case study

using the tool of social networks (Borgatti and Cross 2003; Butts 2008; Sorenson 2005).

Social network analysis is a widely used technique for exploring relational data and offers a

number of metrics for understanding network structure and dynamics (Butts 2008). SNA has

been employed widely in studying collaboration networks during the last two decades (Butts

2008; White and Jorion 1996). Typically, an innovation network has two categories of

properties: node-level and network-level properties (Wasserman and Faust 1994). Like actors

in a geographic space, actors in the network have different positions that are measured

through various indicators such as centrality and brokerage (Burt 2004). These indicators are

useful in exploring the roles and influence of actors in a network. For example (Cho and Shih

2011), drawing on patent citation data, constructed centrality indicators to identify core and

emergent industries. Zaheer and Soda (2009) identified structural holes in a co-membership

network in the Italian TV production industry and found that homogeneity rather than

diversity influences performance across structural holes.

Networks that are constructed through SNA

As there is very limited information about the individual applicants in our patent dataset, it

might lead to a serious bias if we constructed a collaboration network based only on the

name of individual applicants. Therefore, this study focused on the three types of col-

laboration networks at organizational level using the SNA technique: (1) co-application

network between domestic enterprises (DD collaboration network), (2) co-application

network between a domestic enterprises and foreign enterprises (DF collaboration net-

work), and (3) co-application network among enterprises and universities and research

institutes (UIR collaboration network). The nodes in the DD collaboration network are the

indigenous enterprises and their domestic collaborators only, and the links represent the co-

application relationships (e.g., joint R&D effort) among these domestic enterprises. The

nodes in the DF collaboration network are the Chinese-owned enterprises and their

overseas collaborators (foreign invested enterprises (FIEs) and overseas enterprises), and

the links represent co-application relationships among them. The nodes in the UIR col-

laboration network are enterprises and their collaborators (e.g., domestic and overseas

universities and research institutes), and the links represent co-application relationships

among one enterprise and one university or research institute.

Network properties that are measured through SNA

To illustrate the dynamic changing of the three aforementioned types of collaboration

networks, several network properties were calculated using the SNA techniques. The node-

level properties were measured via the number of nodes (size of network), total number of

links (i.e., co-application relationships), density, clustering coefficient, and K-core. In

addition, we used average path (i.e., average geodesic distance of the whole network) and

degree of centralization to measure the network-level properties.

The network density is the number of actual links (the links that are realized in a network)

divided by the number of possible links between pairs of nodes present in a network.

Density ¼ 2k=NðN � 1Þ ð1Þ

where k is the amount of actually linked relationships between collaborators. N is the total

number of nodes (i.e., patent applicants). A network with a high density has more available

channels for knowledge diffusion and spillover among collaborators.

666 Scientometrics (2014) 101:663–683

123

Page 5: Dynamic patterns of technology collaboration: a case study

The clustering coefficient is the degree to which nodes in a network tend to cluster together.

It measures how many of a network actor’s friends are connected to each other. This study

employed the weighted overall clustering coefficient measurement method (Newman 2001).

Clusteringw ¼ 3Ta=Tp ð2Þ

where Ta is the number of triangles in a network; Tp is the number of connected triples in

the network. The triangle is composed of three actors, any two of which are linked

together. The connected triple is that at least one of the three actors links the other two

actors. The larger the clustering coefficient of an actor is, the higher its surrounding

clustering level is (Zhu and Guan 2013).

A K-core is a maximal group of actors, all of whom are connected to some number (K) of

other members of the group. The K-core value of an actor is the largest K of the K-core network

it belongs to. The larger the K-core value of an actor is, the more important is in the network

(Zhu and Guan 2013). In other words, K-core indicates the influence range of the actor. The

parameter is usually adopted to identify the most influential and essential players in a network.

The average path measures the average number of intermediaries between any two

reachable actors in their shortest path in a network. The geodesic distances among actors in

a network may be important macro-characteristics of the network as a whole. The geodesic

path is often the ‘optimal’ or most ‘efficient’ connection between two actors.

The network degree centralization measures the extent to which entire nodes are con-

centrated in the center of a network (Wasserman and Faust 1994). It is a proxy of the level

of difference between actors in the whole network. Unlike degree centrality, which mea-

sures an actor’s direct linkage with neighborhood actors, degree centralization indicates the

concentration status at the network level.

CentralizationD ¼P

v2V maxv0 2V C�ðv

0 Þ � C�ðvÞ� �

maxP

v2V maxv0 2V C�ðv0 Þ � C�ðvÞ

� �� � ð3Þ

The numerator in formula (3) denotes the centrality variation measure of nodes that is

measured by Butts (2006), whereas the denominator is the maximum variation in the

degree centrality of a network of the same size. Therefore, degree centralization is a ratio

of variation in degree centrality of all nodes against the highest possible variation in the

degree centrality of a graph with the same size (de Nooy et al. 2005). The higher value of

the degree centralization is, the more concentrated the network is.

Both the node and network properties of the three types of co-applicant networks in this

study were calculated by using UCINET software (Borgatti et al. 2002). The visualization

of the evolution of the networks was performed using an R package (STATNET)

(Handcock et al. 2008).

Results

Descriptive statistics

Figure 1 illustrates the overall trend of the number of collaborative relations in the Chinese

automobile sector from 1985 to 2010. There were 178 collaborations existing in 1985,

which is only 2.25 % of the number of collaborations in 2010 (7,904 collaborations). In

particular, the number of co-applications in the Chinese automobile sector was approxi-

mately 1,500 before year 2002. This figure increased to nearly 3,000 in 2002, and it was

Scientometrics (2014) 101:663–683 667

123

Page 6: Dynamic patterns of technology collaboration: a case study

growing gradually from then on (except in 2007). The total collaborations during the

period from 2002 to 2010 (42,856 collaborations) account for approximately 66 % of all

collaborations in the whole sample period (1985–2010). This indicates that entering the

WTO encouraged Chinese automobile manufacturers to be more open to external partners

than before. This visualizing result is consistent with some recent studies that emphasize

the surge of patenting and collaborations since China joined the WTO (Hong and Su 2013;

Li 2012; Zheng et al. 2013). Details of collaborations in each year can be found in Table 4

in the Appendix.

There are four types of applicants in the Chinese patent system, i.e., individuals,

enterprises, universities, and research institutes. Nearly 70 % of collaborations were

conducted by individual applied patents, whereas collaborations among universities and

research institutes hosting R&D projects account for only 5 %. In addition, a quarter of all

co-application relations involved enterprise patent applications. As Goncalves and

Almeida (2009) noted, a large proportion of individuals in the patenting system is very

common in the developing country context. The main reason there is a high proportion of

individual applications in China is because of the principal of the inventive activities and

the representatives of the firms’ legal person or R&D teams usually registered themselves

as the applicants of the patents rather than using the organization name (Wang et al. 2013).

Moreover, as the assessment criteria for the three types of SIPO patent (i.e., invention,

utility model, external design) are different, both the innovativeness and practical value of

inventions and utility models are higher than that of external designs (Cheung and Lin

2004). Specifically, over half of collaborations were conducted involving utility models,

and 42.52 % of collaborations were for invention patents in the Chinese automobile

industry during the sample period. This portfolio implies that collaborated innovation

outputs usually involve a higher level of innovativeness and useful knowledge (Fabrizio

2009; Hong and Su 2013; Wang and Zhou 2013).

We notice that the distribution of collaborations in the six automobile technology fields

is uneven. These six technology domains can be classified into three groups in terms of the

amount of collaborations. Automobile body and chassis collaborations are classified into

the first group as each of them accounts for over 20 % of all collaborations. Engine,

electrical device and other collaborations belong to the second group, as the share of

collaborations in each of them exceeds 16 %. Fuel collaboration forms the last group,

which accounts for only 3.21 % of all collaborations.

We can also divide the collaborations into several subsets in accordance with the type of

specific applicant, i.e., enterprise, university and research institute, and individual. The

majority of collaborations were conducted by inventors (70.16 %), whereas co-applica-

tions between organizations and individuals (OI collaborations) account for only 3.95 % of

the total collaborative activities. In particular, we further divided collaborations related to

domestic enterprises into two categories, collaborations between domestic firms (DD

collaborations) and collaborations between domestic and foreign invested/owned firms (DF

collaborations).1 A further analysis revealed that the proportions of DD collaborations and

DF collaborations are very close, both of which are approximately 10–12 %. In contrast,

the relatively low level of university–industry–research institutes (UIR) collaborations

(4.22 %) indicates that the bilateral and trilateral interactions among UIR relationships in

the Chinese automobile sector are relatively weak, although the joint UIR projects can

motivate the knowledge-based strategy and speed up the rate of socioeconomic

1 Although the R&D collaborations between foreign invested/owned firms can be an important topic, thefocus of this paper is the DD, DF, and UIR R&D collaborations.

668 Scientometrics (2014) 101:663–683

123

Page 7: Dynamic patterns of technology collaboration: a case study

development (Park and Leydesdorff 2010). Considering the heterogeneities of domestic

and foreign firms in terms of innovativeness (Choi et al. 2011; Li 2011) and also the

decentralizing trend of UIR collaborations during the last decade in China (Hong 2008), we

hereby focus on the collaborations among organizations in the following sections.

As mentioned in the preceding sections, both the novelty and level of technology are

varied according to the patent type. Therefore, we analyzed the trend of development of

collaborations by focusing on the distribution of co-applications in invention, utility model,

and design in each year during 1985–2010. Moreover, considering the heterogeneities and

idiosyncratic features of specific technology domains, we focused on the amount of col-

laborations involving the categories of automobile body, chassis, electrical device, engine,

fuel, and others as well.

The longitudinal trend of DD, DF, and UIR collaborations

Figure 2 shows that there have been far more R&D collaborations in invention and utility

models than in design. The first design collaboration occurred in 2001, which is much later

than that for inventions (in 1986) and utility models (in 1985) collaborations. In addition,

the amounts of invention and utility model collaborations are very similar during the

sample period. A sharp growth of co-applications for invention and design was detected in

the last 5 years in the sample period (2006–2010). This indicates that a vast number of

resources and highlights have been driven by domestic enterprises engaging in joint R&D

projects for obtaining innovative output rather than for external designs.

Figure 3 illustrates that except for collaborations in the technology domain of fuel, DD

collaborations in the other five technology domains display similar development tracks

during the sample period (except for a peak value for the others domain in 1992). Since

2004, collaborations involving the automobile body, chassis, electrical device, engine, and

others domain experienced huge increases from approximately 25–300 or so in 2010. This

indicates that the collaborative focuses of Chinese indigenous enterprises are balanced in

technology domains except for the fuel domain.

Fig. 1 Overall distribution of the number of collaborations in the Chinese automobile sector by year,1985–2010

Scientometrics (2014) 101:663–683 669

123

Page 8: Dynamic patterns of technology collaboration: a case study

Figure 4 depicts that there are huge differences between the number of collaborations

involving inventions and utility models and designs during the sample period. It is sur-

prising to find that there are no DF collaborations for utility models and designs before

1995, and the number of DF collaborations in the utility model and design categories

fluctuate between 10 and 30 (except for 46 utility model collaborations in 2004) from

1996. This phenomenon suggests that the aim of collaborations across national boundaries

is to create advanced knowledge or frontier technology rather than to focus on R&D with

middle- to low-level innovativeness. Moreover, the number of invention DF collaborations

decreased heavily from 1,015 to 467 in 2010, which is even less than the number of DF

collaborations in 2003 (512). A possible reason for the drop of cross-boundary collabo-

rations is that domestic firms accumulated R&D resources in their prior collaborations

(Cohen and Levinthal 1989), and they can easily collaborate with indigenous counterparts

with less communication and absorption costs than with overseas partners.

Fig. 2 The distribution of R&D collaborations between domestic enterprises by patent type

Fig. 3 The distribution of number of collaborations between domestic enterprises by technology domain

670 Scientometrics (2014) 101:663–683

123

Page 9: Dynamic patterns of technology collaboration: a case study

It is obviously that the six technology domains can be classified into three groups in

terms of the amount of DF collaboration (shown in Fig. 5). Specifically, the number of

collaborations in the automobile body and chassis domains is higher than for the other four

technology domains, and the gap between them has increased since 2002. The second

group includes the electrical device, engine, and others domains, and the frequencies of DF

collaboration in these three domains display a similar pattern during the sample period.

Greatly fewer DF collaborations are found for the fuel domain, and this is basically in

agreement with Fig. 3. The longitudinal trend shown in Fig. 5 indicates that domestic

enterprises are more willingly to conduct joint R&D projects with foreign counterparts in

the automobile body and chassis technology domains.

Fig. 4 The distribution of number of collaborations between domestic and foreign enterprises by patenttype

Fig. 5 The distribution of number of collaborations between domestic and foreign enterprises bytechnology

Scientometrics (2014) 101:663–683 671

123

Page 10: Dynamic patterns of technology collaboration: a case study

UIR collaboration is considered to be a facilitator for the production of knowledge,

which is helpful for building innovation systems, and innovation can in turn spur the

economic growth of developing countries (Swar and Khan 2013). Figure 6 shows that the

UIR collaborations in the Chinese automobile sector mainly focused on inventions and

utility models rather than designs during the period from 1985 to 2010. The growth trend of

UIR collaboration in the invention and utility model categories is even much stronger since

2002, as the number of collaborations grows dramatically from 41 and 40 to 405 and 334 in

2009, respectively. This huge increase reflects that UIR collaboration has been highlighted

at the industrial level, and the majority of joint R&D projects conducted by UIR alliances

are interested in innovations with higher novelty.

Comparing Figs. 7 with 3 and 5, we notice that the chassis and others domains are the

two main focuses of UIR collaborations, whereas DD and DF collaborations mainly focus

on the automobile body and chassis domains. This suggests that UIR collaborations are

complementary with collaborations between enterprises as the knowledge bases of uni-

versities and research institutes are not the same as that of enterprises (Schartinger et al.

2002).

Network analysis

As discussed in a previous section, collaborative relations can be analyzed using the SNA

technique. Specifically, three types of R&D collaboration network have been examined in

this paper, i.e., collaboration network of domestic enterprises (DD network), collaboration

network of domestic enterprises and foreign enterprises (DF network), and collaboration

network of enterprises, universities and research institutes (UIR networks). Both the node-

and network-level properties are measured to analyze the structural change of the col-

laboration networks in different phases during the period 1985–2010. We suppose that both

the joint R&D effort and co-application require some time. During this collaboration

period, it is assumed that information exchange occurs to a great extent among the patent

applicants (He and Hosein Fallah 2009). Following prior studies (Baum et al. 2003; Eslami

Fig. 6 The distribution of number of collaborations between UIR by patent type

672 Scientometrics (2014) 101:663–683

123

Page 11: Dynamic patterns of technology collaboration: a case study

et al. 2013; Fleming et al. 2007; He and Hosein Fallah 2009), we have assumed that the life

length of each link in the created collaboration network is 5 years. Therefore, we extracted

the collaborative relationships for five snapshots from 1985 to 2010. This procedure

resulted in five undirected one-mode networks for each type of collaboration network, i.e.,

1985–1990, 1991–1995, 1996–2000, 2001–2005, and 2006–2010.

Evolution of DD collaboration network structure

As shown in Table 1, although the size of the DD collaboration network increased from

1985 to 2010, the growth pace was not even. The number of nodes (applicants) in the DD

network from the period 2001–2005 to the period 2006–2010 increased 1,148, whereas it

increased merely 26 from the period 1985–1990 (184) to the period 1996–2000 (210). This

circumstance is vividly shown in the graph of the whole network of DD collaboration in

each period. The visualization not only depicts the huge increases of the size and links of

DD network but also indicates the features of the evolution process of the network. For

instance, the DD network becomes even less dense in the period 2006–2010 (0.0021) than

the period 1985–1990 (0.0086). This implies that although a huge number of domestic

enterprises participated in the network, the links between domestic enterprises are still very

scarce. Moreover, the increasing average path (average geodesic distance) within the DD

network during the period 1985–1990 (1.257) to the period 2006–2010 (2.021) indicates

that domestic enterprises may have opportunities to collaborate with potential partners in a

wider space of the network. In addition, with the decreasing value of centralization of the

DD network, e.g., the value dropped from 0.0126 (1985–1990) to 0.0016 (2006–2010), the

probability of a node with abundant relationships within the network decreased.

Interestingly, we find that the clustering coefficient of the DD network is much higher in

the period of 1991–1995 (1.103) than in other periods. This situation indicates that local

clustering increased when more nodes were linked with each other in a short distance. With

the huge increase in the amount of new participants into the network in the period

2006–2010, the probability of many of the domestic enterprise’s collaborators being

connected to each other is higher as the clustering coefficient reaches 0.705. Moreover, a

Fig. 7 The distribution of number of UIR collaborations by technology domain

Scientometrics (2014) 101:663–683 673

123

Page 12: Dynamic patterns of technology collaboration: a case study

high K-core value indicates highly influencing power (Zhu and Guan 2013). If a node has a

high K-core value, it can be considered an important player in the network. The maximal

K-core value of the DD collaboration network in the period 1991 to 1995 (K-core = 9) and

the period 2006 to 2010 (K-core = 6) implies that domestic automobile enterprises in these

periods had much wider influence range than others, and they are important players in the

whole network. The bottom row of Table 1 further provides the visualization of the largest

component within the DD collaboration network in each period, which results in various

shapes. For instance, the structure of the largest component in the period of 1985–1990 and

in 2001–2005 is a star shape, which implies the node located in the center is the most

essential player who can influence other linked players by controlling knowledge diffusion

(Wasserman and Faust 1994). Similarly, there is a key player in the largest component in

the DD network in the period 2006–2010. The largest component in the period 1991–1995

is composed of 15 nodes, and the majority of them are linked with each other. This

suggests that there are many channels to obtain knowledge or information from other

collaborators, and no single node can function as a component hub. The size of the largest

component increased from 9 in the period 1985–1990 (accounts for 4.89 % of the whole

network) to 39 in the period of 2006–2010 (accounts for 2.48 % of the whole network).

Evolution of DF collaboration network structure

A huge growth in the average path is shown in Table 2, as it increased Compared with the

DD collaboration network (as shown in Table 1), the size of the DF collaboration network

(see Table 2) is much smaller. For instance, there are only 31 nodes with 38 links within

the DF network in the period 1985–1990, but the DD network in this period has 184 nodes

with 216 links. This indicates that collaboration between domestic automobile enterprises

and foreign enterprises was very scarce at the very beginning of the ‘‘opening up’’ policy.

Table 1 Collaboration networks between domestic enterprises

Indices 1985–1990 1991–1995 1996–2000 2001–2005 2006–2010

Node 184 208 210 426 1574

Link 216 376 260 522 2270

Density 0.0086 0.0111 0.0085 0.0057 0.0021

Average path 1.257 1.126 1.382 1.378 2.021

K-core 2 9 3 2 6

Clustering 0.509 1.103 0.627 0.393 0.705

Centralization 0.0126 0.0120 0.0082 0.0036 0.0016

Whole network

Largest component

Source Compiled by authors

674 Scientometrics (2014) 101:663–683

123

Page 13: Dynamic patterns of technology collaboration: a case study

An increasing number of both indigenous enterprises and foreign enterprises participated in

the R&D collaboration network of the Chinese automobile sector from 1991 to 2010 as the

number of nodes and links reached 933 and 1,924, respectively, in the period 2006–2010.

Although the size and linkages of the DF network are still smaller than that of the DD

network, the average number of links of each node of the DF network (2.06 in 2006–2010)

is higher than that of the DD network (1.44 in 2006–2010). Moreover, the overall trend of

the DF network indicate a decreasing density compared with earlier periods, e.g., the

density of the DF network decreased from 0.0731 in the period of 1985–1990 to 0.0045 in

the period of 2006–2010, and the density of the DF network is greater than that of the DD

network in each period, which provides further evidence that linkages between domestic

automobile enterprises and foreign enterprises are more prosperous than the collaborations

between indigenous enterprises.

A huge growth in the average path is shown in Table 2 as it increased 222.6 % in the

period from 1985 to 1990 (1.552) to the period from 2006 to 2010 (5.007). This fact

indicates that domestic automobile enterprises could increasingly engage in collaborations

with potential overseas partners in a much wider scope than with indigenous partners. A

comparison between the DD network and DF network in terms of the average path reveals

that the collaborations between domestic enterprises are more likely to happen in a scope

within a short distance. This argument is proven by the higher level of clustering coeffi-

cient of the DD network than that of the DF network. In particular, except in the first period

(1985–1990), the DF network has a smaller probability to include a cluster in which a

node’s neighbors are connected with each other than the DD network. In addition, the

smaller K-core value of the DF network compared with the DD network in each period

(except 2001–2005) indicates that the size of the subgroup in which each group member is

connected with other members of the DF network is smaller than that of the DD network.

Therefore, it is not surprising to find that degree of centralization of the DF network is

greater than that of the DD network in each period, which implies that the probability of

Table 2 Collaboration networks between domestic and foreign enterprises

Indices 1985–1990 1991–1995 1996–2000 2001–2005 2006–2010

Node 31 82 218 676 933

Link 38 104 308 1262 1924

Density 0.0731 0.0232 0.0148 0.0070 0.0045

Average path 1.552 1.175 2.432 4.438 5.007

K-core 2 2 2 4 5

Clustering 0.750 0.625 0.267 0.222 0.309

Centralization 0.0404 0.0130 0.0185 0.0096 0.0093

Whole network

Largest component

Source Compiled by authors

Scientometrics (2014) 101:663–683 675

123

Page 14: Dynamic patterns of technology collaboration: a case study

the existence of a hub node within the DF network is higher than that within the DD

network. In other words, it is more likely that an enterprise will be found as the center of

the DF network with abundant relationships with other enterprises.

The visualization of the DF network and its largest component in Table 2 provides

intuitive evidence of the evolution of collaborations between domestic automobile enter-

prises and foreign enterprises. It is obvious that an increasing number of enterprises in the

DF network began to link with others from 1991. The size of the largest component grew

from 6 (accounting for 19.4 % of whole network) in the period 1985–1990 to 454

(accounting for 23.6 % of whole network) in the period 2006–2010. More importantly, we

notice that the size of other components is much smaller than the largest component during

the period of 2001–2010. For example, the second largest component within the DF

network in 2006–2010 includes only seven nodes. This indicates that a technology alliance

that includes the main technology sources of the supply chain may exist in the global

automobile sector, and Chinese indigenous manufacturers are more willing to be a member

of this large alliance to obtain advanced knowledge and valuable opportunities rather than

to ally with domestic enterprises.

Evolution of UIR collaboration network structure

Table 3 illustrates that both the size and links of the UIR network increased during the

period of 1985–2010, which indicates that an increasing number of enterprises began to

realize the benefits of collaborating with universities and research institutes. Compared

with the DD and DF networks, the UIR network is denser but has a lower K-core value

overall. This may be because universities and research institutes are usually the key players

linking enterprises through joint R&D efforts (Hong and Su 2013), whereas those enter-

prises linked with key universities and research institutes may not connect with each other.

Table 3 Collaboration networks between enterprises and universities and research institutes

Indices 1985–1990 1991–1995 1996–2000 2001–2005 2006–2010

Node 95 93 111 238 724

Link 112 106 146 330 1076

Density 0.0161 0.0164 0.0180 0.0101 0.0042

Average path 1.273 1.315 1.448 2.545 4.501

K-core 1 1 2 2 2

Clustering 0.000 0.000 0.323 0.143 0.030

Centralization 0.0162 0.0240 0.0186 0.1110 0.0029

Whole network

Largest component

Source Compiled by authors

676 Scientometrics (2014) 101:663–683

123

Page 15: Dynamic patterns of technology collaboration: a case study

Moreover, the relatively low level of the clustering coefficient implies a low probability of

the UIR network including a cluster in which a node’s neighbors are connected with each

other. A further comparison between the UIR network with DD and DF networks indicates

that the average path of in the UIR network is longer than in the DD network but smaller

than in the DF network, which indicates that the collaboration scope of UIR is even larger

than that of collaboration between domestic enterprises but is still narrow relative to

collaboration between indigenous automobile enterprises and foreign enterprises. In

addition, the relatively high level of centralization of the UIR network may be due to the

predominant role played by some first-tier universities and the limited number of UIR

collaborations. In addition, the large decrease in the centralization of the UIR network from

0.1110 in the period 2001–2005 to 0.0029 in the period 2006–2010 may have been caused

by the huge increase of nodes and links within the network, which decreases the probability

of a single node being the center of the UIR network.

The visualization of both the whole UIR network and its largest component shows that

the network structure evolved gradually from 1985 to 2010. Furthermore, the largest

component of the UIR network is a bit more complicated than that of the DD network and

includes much fewer participants than the DF network. Specifically, the size of the largest

component increased from 3 in the period of 1985–1990 to 113 in the period of 2006–2010,

which accounts for 9.67 and 12.11 % of the whole network, respectively. A deep analysis

of the degree centrality in the last two periods (2001–2005 and 2006–2010) identified key

players in the UIR network, including Tsinghua University, Zhejiang University and Jilin

University, among others. All of these universities are first-tier higher education institu-

tions in China.

Discussion and conclusions

Although the benefits and importance of the R&D collaboration with external partners

have been investigated and highlighted by recent literature (Doloreux and Shearmur 2012;

Zheng et al. 2013; Zhou et al. 2013), we still understand very little about the pattern and

evolution of industry-specific collaborative activities. The overall circumstances of the

64,938 technology collaboration relations in the Chinese automobile sector during the

period 1985–2010 have been assessed through several perspectives, e.g., descriptive sta-

tistics, longitudinal trends, and collaboration networks in this study. Considering the levels

of innovativeness of inventions, utility models and external designs are not the same

(Cheung and Lin 2004; Li 2012), and the heterogeneity and idiosyncratic features of six

technology domains of automobile sector, we analyzed the longitudinal trend of specific

collaborations separately by patent type and technology domain. More importantly, we

systematically analyzed the changing structure and evolution process of DD collaboration,

DF collaboration and UIR collaboration by measuring network properties and visualizing

the whole networks and their largest components.

One of our major findings suggests that the majority of collaborations are embedded in

invention and utility rather than in design. This fact implies that both individuals and

enterprises are more willing to make joint efforts to pursue R&D with a higher level of

innovativeness. Therefore, the open innovation mode is a preferable manner for innovators

to obtain advance knowledge and valuable opportunities from networking with external

exchange partners in the Chinese automobile sector. With the exception of the fuel domain,

the collaborations are distributed evenly in the other five technology domains (automobile

body, chassis, electrical device, engine, and others). This suggests that there is no ‘‘hot

Scientometrics (2014) 101:663–683 677

123

Page 16: Dynamic patterns of technology collaboration: a case study

topic’’ that attracts most R&D efforts in that area. Approximately 22 % of collaborations

are joint efforts made by enterprises, which indicates that there is still large room for

enterprises to search for open innovation partners in the near future. Considering the high

proportion of individual applicants in collaborative patents, we suggest policy makers

should launch specific regulations to adjust the procedures for patenting application,

assessment and registration and make an effort to adopt the standards used in developed

countries so that more analysis can be performed to support future policy designs. The

relatively small proportion of collaboration between organizations and individuals

(3.95 %) implies that it is difficult to bridge organization R&D with personal innovations,

and some institutional efforts should be made to improve the level of transformation of

personal R&D output into commercial usage.

The main result from the descriptive analysis of the longitudinal trend of collaborations

implies that the majority of collaborations were conducted after China entered the WTO in

2001. A more open environment has proven to be an encouragement for innovators to

engage in open innovation (Chesbrough and Crowther 2006; Laursen and Salter 2006).

Interestingly, we found that DD collaborations are interested in both inventions and utility

models, whereas DF collaborations only are interested in inventions rather than utility

models. This reflects that when Chinese indigenous enterprises collaborate with overseas

partners, they are only concerned about absorbing frontier technology via joint R&D

efforts. This is in line with the argument of the technology-seeking strategy in IB (inter-

national business) theory (Dunning 1988; Luo and Tung 2007). Moreover, we found that

the year 2001 was the turning point; from then on, a huge increase in DD and DF col-

laborations was detected. This interesting phenomenon implies that the technology gap

between Chinese indigenous enterprises and foreign counterparts is becoming smaller than

before, and therefore, more domestic enterprises have begun to collaborate with external

partners. A further analysis suggests that DD collaborations are interested in all technology

areas expect for fuel, whereas DF collaborations are more likely to focus on the automobile

body and chassis domains. This difference in interests implies that domestic enterprises

have a clear technology focus when collaborating with foreign counterparts.

Additionally, the DF collaborations in invention patents experienced a notable drop

since 2007. Two possible explanations for this phenomenon are follows. One reason is that

domestic enterprises have accumulated a certain level of R&D capabilities and they can

obtain interested knowledge through cooperating with indigenous exchange partners with

relatively less coordination costs than with foreign counterparts. Another reason may be

due to the Chinese government launching the ‘‘Medium- to Long-Term Plan for the

Development of Science and Technology’’ (MLP) in 2006, in which policy makers

emphasize the importance of cultivating independent R&D capabilities of domestic

innovators (Abrami et al. 2014; Cao et al. 2013). Chinese automobile manufacturers may

therefore have a propensity to enhance their independent innovation capabilities against

this background.

A further investigation of the evolution process of the DD network, the DF network,

and the UIR network provided many more details about the dynamic changing of net-

work structure during the period of 1985–2010. To better understand this evolution

process, we measured both node and network properties, e.g., number of nodes and links,

density, average path, K-core, clustering coefficient, and degree of centralization. In

addition, we visualized the whole network and the corresponding largest component for

each of the five snapshots. The findings suggest that although the DD collaboration

network is much larger than the DF network, the DF network is denser than the DD

network. In addition, the average path of the DF network is usually longer than that of

678 Scientometrics (2014) 101:663–683

123

Page 17: Dynamic patterns of technology collaboration: a case study

the DD network. This phenomenon may be caused by the fact that domestic automobile

enterprises could obtain limited technology support from collaborating with indigenous

enterprise but gain relatively more advance knowledge and valuable opportunities via

making joint R&D efforts with foreign enterprises. Moreover, the size of the largest

component of the DF network (454 in 2006–2010) is much larger than that of the DD

network (39 in 2006–2010). We identified the top three enterprises in terms of degree of

centrality within the DF network in the period of 2006–2010, i.e., Toyota Motor Co., Ltd.

(degree of 281), Honda Motor Co., Ltd. (degree of 121), and Hyundai Motor Co., Ltd.

(degree of 101). The first two are famous automobile manufacturers in Japan, and the

latter is a top vehicle manufacturer in Korea. Therefore, the largest component within the

DF collaboration network is actually a R&D alliance that centers on these foreign

enterprises. This indicates that the majority of Chinese automobile manufacturers have

absorbed advance technologies through collaboration with several top enterprises during

the last decade. The evolution process of the UIR network implies that first-tier uni-

versities, e.g., Tsinghua University, Zhejiang University and Jilin University, are the key

players in the UIR collaboration network. An increasing number of collaborations

between those key universities and enterprises are found in the UIR network, but both

the size and links of the largest component are much smaller than those in the DF

network. This suggests that more effort should be expended by knowledge-intensive

business services (KIBS) to facilitate commercialization of technology innovations made

by universities and research institutes (Yam et al. 2011).

This paper provides a fuller picture of the patent collaboration pattern in the Chinese

automobile industry. However, it is not without its limitations, which warrant future

studies. First, this paper focuses on the collaborations between organizations rather than

individuals as the SIPO patent records do not include any personal information except

individuals’ names. We are unable to distinguish between namesakes and further con-

struct collaboration network at the individual level. Future research might be in a good

position to re-examine some of these issues when more detailed data become available.

Second, we have visualized both the whole network and the largest component for

collaboration networks, but we have not analyzed the influential actors in the network,

namely those enterprises with the highest K-core value (Zhu and Guan 2013). Third, this

study did not discuss the small world network properties of these collaboration networks,

which is an avenue for future research. Finally, this study only focused on the patent

collaborations in the Chinese automobile industry. A comparison study of the patent

patterns and evolution processes with other industries or in cross-cultural settings, e.g.,

between China and European countries or the U.S., might reveal additional points of

interest.

Acknowledgments This research is funded by national Science Foundation of China (grant no.71302133;grant no.71233002), Youth Project of Ministry of Education, Humanities and Social Sciences PlanningFunding (grant no.13YJC790154), and the Sichuan University’s Special Research Program for the Philos-ophy Social Science from the Subordinate Universities of Ministry of Education’s Basic Research Foun-dation (SKYB201302; SKX201004).

Appendix

See Table 4.

Scientometrics (2014) 101:663–683 679

123

Page 18: Dynamic patterns of technology collaboration: a case study

References

Abrami, R. M., Kirby, W. C., & McFarlan, F. W. (2014). Why China can’t innovate. Harvard BusinessReview, 92, 107–111.

Ahuja, G. (2000). Collaboration networks, structural holes, and innovation: A longitudinal study. Admin-istrative Science Quarterly, 45, 425–455.

Allen, R. C. (1983). Collective invention. Journal of Economic Behavior & Organization, 4, 1–24.Baum, J. A., Shipilov, A. V., & Rowley, T. J. (2003). Where do small worlds come from? Industrial and

Corporate Change, 12, 697–725.Becker, W., & Dietz, J. (2004). R&D cooperation and innovation activities of firms—Evidence for the

German manufacturing industry. Research Policy, 33, 209–223.Bierly, P. E., Damanpour, F., & Santoro, M. D. (2009). The application of external knowledge: Organi-

zational conditions for exploration and exploitation. Journal of Management Studies, 46, 481–509.

Table 4 Descriptive statistics of selected collaboration data

Year Patentnumber

Collaborationrelations

DDcollaborationa

DFcollaborationa

UIRcollaborationa

1985 289 178 14 (7.87 %) 12 (6.74 %) 26 (14.61 %)

1986 211 420 31 (7.38 %) 30 (7.14 %) 20 (4.76 %)

1987 861 818 27 (3.30 %) 4 (0.49 %) 25 (3.06 %)

1988 1,282 1,298 42 (3.24 %) 8 (0.62 %) 34 (2.62 %)

1989 1,364 1,498 30 (2.00 %) 11 (0.73 %) 28 (1.87 %)

1990 1,785 1,606 37 (2.30 %) 0 (0.00 %) 21 (1.31 %)

1991 2,149 1,812 53 (2.92 %) 3 (0.17 %) 15 (0.83 %)

1992 2,582 1,784 182 (10.20 %) 0 (0.00 %) 25 (1.40 %)

1993 2,564 1,428 41 (2.87 %) 23 (1.61 %) 15 (1.05 %)

1994 2,684 1,020 52 (5.10 %) 27 (2.65 %) 25 (2.45 %)

1995 2,853 1,130 26 (2.30 %) 83 (7.35 %) 19 (1.68 %)

1996 3,219 1,140 43 (3.77 %) 88 (7.72 %) 41 (3.60 %)

1997 3,187 1,438 53 (3.69 %) 153 (10.64 %) 21 (1.46 %)

1998 3,519 1,374 49 (3.57 %) 104 (7.57 %) 17 (1.24 %)

1999 3,872 1,522 66 (4.34 %) 124 (8.15 %) 35 (2.30 %)

2000 4,631 1,846 67 (3.63 %) 174 (9.43 %) 82 (4.44 %)

2001 5,310 1,770 72 (4.07 %) 212 (11.98 %) 40 (2.26 %)

2002 7,464 2,849 101 (3.55 %) 355 (12.46 %) 81 (2.84 %)

2003 9,643 3,116 160 (5.13 %) 526 (16.88 %) 190 (6.10 %)

2004 11,482 3,648 186 (5.10 %) 993 (27.22 %) 161 (4.41 %)

2005 13,933 4,063 358 (8.81 %) 853 (20.99 %) 158 (3.89 %)

2006 17,718 4,387 516 (11.76 %) 1,038 (23.66 %) 274 (6.25 %)

2007 20,186 3,549 477 (13.44 %) 804 (22.65 %) 408 (11.50 %)

2008 25,177 6,028 953 (15.81 %) 755 (12.52 %) 462 (7.66 %)

2009 30,860 7,312 1,494 (20.43 %) 526 (7.19 %) 741 (10.13 %)

2010 36,942 7,904 1,518 (19.21 %) 508 (6.43 %) 545 (6.90 %)

Source Compiled by authors using the patent data of the Chinese automobile sectora Percentages in parenthesis are the shares of specific types of R&D collaborations over the total number ofcollaborations

680 Scientometrics (2014) 101:663–683

123

Page 19: Dynamic patterns of technology collaboration: a case study

Bloomberg. (2010). China ends US’s reign as largest auto market. http://www.bloomberg.com/apps/news?pid=newsarchive&sid=aE.x_r_l9NZE. Accessed 17 Sep 2013.

Borgatti, S. P., & Cross, R. (2003). A relational view of information seeking and learning in social networks.Management Science, 49, 432–445.

Borgatti, S. P., Everett, M. G., & Freeman, L. C. (2002). Ucinet for Windows: Software for social networkanalysis. Harvard, MA: Analytic Technologies.

Burt, R. S. (2004). Structural holes and good ideas. American Journal of Sociology, 110, 349–399.Butts, C. T. (2006). Exact bounds for degree centralization. Social Networks, 28, 283–296.Butts, C. T. (2008). Social network analysis: A methodological introduction. Asian Journal of Social

Psychology, 11, 13–41.Cao, C., Li, N., Li, X., & Liu, L. (2013). Reforming China’s S&T system. Science, 341, 460–462.Cassiman, B., & Veugelers, R. (2006). In search of complementarity in innovation strategy: Internal R&D

and external knowledge acquisition. Management Science, 52, 68–82.Cattani, G., & Ferriani, S. (2008). A core/periphery perspective on individual creative performance: Social

networks and cinematic achievements in the Hollywood film industry. Organization Science, 19,824–844.

Chesbrough, H. (2003). Open innovation: The new imperative for creating and profiting from technology.Cambridge, MA: Harvard Business Press.

Chesbrough, H., & Crowther, A. K. (2006). Beyond high tech: Early adopters of open innovation in otherindustries. R&D Management, 36, 229–236.

Cheung, K.-Y., & Lin, P. (2004). Spillover effects of FDI on innovation in China: Evidence from theprovincial data. China Economic Review, 15, 25–44.

Cho, T.-S., & Shih, H.-Y. (2011). Patent citation network analysis of core and emerging technologies inTaiwan: 1997–2008. Scientometrics, 89, 795–811.

Choi, S. B., Lee, S. H., & Williams, C. (2011). Ownership and firm innovation in a transition economy:Evidence from China. Research Policy, 40, 441–452.

Cohen, W. M., & Levinthal, D. A. (1989). Innovation and learning: the two faces of R & D. The EconomicJournal, 99, 569–596.

Cowan, R., & Jonard, N. (2004). Network structure and the diffusion of knowledge. Journal of economicDynamics and Control, 28, 1557–1575.

Dahl, M. S., & Pedersen, C. Ø. (2004). Knowledge flows through informal contacts in industrial clusters:myth or reality? Research Policy, 33, 1673–1686.

de Nooy, W., Mrvar, A., & Batagelj, V. (2005). Exploratory social network analysis with Pajek. New York:Cambridge University Press.

Dodgson, M. (2009). Asia’s national innovation systems: Institutional adaptability and rigidity in the face ofglobal innovation challenges. Asia Pacific Journal of Management, 26, 589–609.

Doloreux, D., & Shearmur, R. (2012). Collaboration, information and the geography of innovation inknowledge intensive business services. Journal of Economic Geography, 12, 79–105.

Dunning, J. H. (1988). Open R&D and open innovation: exploring the phenomenon. The eclectic paradigmof international production: a restatement and some possible extensions, 19, 1–31.

Enkel, E., Gassmann, O., & Chesbrough, H. (2009). Open R&D and open innovation: exploring the phe-nomenon. R&D Management, 39, 311–316.

Eslami, H., Ebadi, A., & Schiffauerova, A. (2013). Effect of collaboration network structure on knowledgecreation and technological performance: the case of biotechnology in Canada. Scientometrics, 97,99–119.

Fabrizio, K. R. (2009). Absorptive capacity and the search for innovation. Research Policy, 38, 255–267.Fleming, L., King, C., & Juda, A. I. (2007). Small worlds and regional innovation. Organization Science, 18,

938–954.Gao, X., Guan, J., & Rousseau, R. (2011). Mapping collaborative knowledge production in China using

patent co-inventorships. Scientometrics, 88, 343–362.Glanzel, W. (2001). National characteristics in international scientific co-authorship relations. Scientomet-

rics, 51, 69–115.Glanzel, W., & Schubert, A. (2001). Double effort = double impact? A critical view at international co-

authorship in chemistry. Scientometrics, 50, 199–214.Goncalves, E., & Almeida, E. (2009). Innovation and spatial knowledge spillovers: evidence from Brazilian

patent data. Regional Studies, 43, 513–528.Gulati, R. (1999). Network location and learning: The influence of network resources and firm capabilities

on alliance formation. Strategic Management Journal, 20, 397–420.

Scientometrics (2014) 101:663–683 681

123

Page 20: Dynamic patterns of technology collaboration: a case study

Handcock, M. S., Hunter, D. R., Butts, C. T., Goodreau, S. M., & Morris, M. (2008). statnet: Software toolsfor the representation, visualization, analysis and simulation of network data. Journal of StatisticalSoftware, 24, 1548.

He, J., & Hosein Fallah, M. (2009). Is inventor network structure a predictor of cluster evolution? Tech-nological Forecasting and Social Change, 76, 91–106.

Hong, W. (2008). Decline of the center: The decentralizing process of knowledge transfer of Chineseuniversities from 1985 to 2004. Research Policy, 37, 580–595.

Hong, W., & Su, Y.-S. (2013). The effect of institutional proximity in non-local university–industry col-laborations: An analysis based on Chinese patent data. Research Policy, 42, 454–464.

Howells, J., Gagliardi, D., & Malik, K. (2008). The growth and management of R&D outsourcing: evidencefrom UK pharmaceuticals. R&D Management, 38, 205–219.

Inkpen, A. C., & Tsang, E. W. (2005). Social capital, networks, and knowledge transfer. Academy ofManagement Review, 30, 146–165.

Laursen, K., & Salter, A. (2006). Open for innovation: The role of openness in explaining innovationperformance among UK manufacturing firms. Strategic Management Journal, 27, 131–150.

Li, X. (2011). Sources of external technology, absorptive capacity, and innovation capability in Chinesestate-owned high-tech enterprises. World Development, 39, 1240–1248.

Li, X. (2012). Behind the recent surge of Chinese patenting: An institutional view. Research Policy, 41,236–249.

Liu, W., & Dicken, P. (2006). Transnational corporations and ‘obligated embeddedness’: Foreign directinvestment in China’s automobile industry. Environment and Planning A, 38, 1229.

Luo, Y., & Tung, R. L. (2007). International expansion of emerging market enterprises: A springboardperspective. Journal of International Business Studies, 38, 481–498.

Motohashi, K., & Yun, X. (2007). China’s innovation system reform and growing industry and sciencelinkages. Research Policy, 36, 1251–1260.

Narula, R. (2004). R&D collaboration by SMEs: New opportunities and limitations in the face of global-isation. Technovation, 24, 153–161.

Newman, M. E. (2001). Scientific collaboration networks. I. Network construction and fundamental results.Physical Review E, 64, 016131.

Park, H. W., & Leydesdorff, L. (2010). Longitudinal trends in networks of university–industry–governmentrelations in South Korea: The role of programmatic incentives. Research Policy, 39, 640–649.

Saxenian, A. (1994). Regional advantage: Culture and competition in Silicon Valley and Route 128.Cambridge: Harvard University Press.

Schartinger, D., Rammer, C., Fischer, M. M., & Frohlich, J. (2002). Knowledge interactions betweenuniversities and industry in Austria: Sectoral patterns and determinants. Research Policy, 31, 303–328.

Schilling, M. A., & Phelps, C. C. (2007). Interfirm collaboration networks: The impact of large-scalenetwork structure on firm innovation. Management Science, 53, 1113–1126.

Schmiele, A. (2012). Drivers for international innovation activities in developed and emerging countries.The Journal of Technology Transfer, 37, 98–123.

Schrader, S. (1991). Informal technology transfer between firms: Cooperation through information trading.Research Policy, 20, 153–170.

Schubert, A., & Glanzel, W. (2006). Cross-national preference in co-authorship, references and citations.Scientometrics, 69, 409–428.

Sorenson, O. (2005). Social networks and industrial geography, entrepreneurships, the new economy andpublic policy (pp. 55–69). New York: Springer.

Sun, P., Mellahi, K., & Thun, E. (2010). The dynamic value of MNE political embeddedness: The case ofthe Chinese automobile industry. Journal of International Business Studies, 41, 1161–1182.

Swar, B., & Khan, G. F. (2014). Mapping ICT knowledge infrastructure in South Asia. Scientometrics, 99,117–137.

Uzzi, B. (1996). The sources and consequences of embeddedness for the economic performance of orga-nizations: The network effect. American Sociological Review, 61, 674–698.

Von Hippel, E. (1987). Cooperation between rivals: informal know-how trading. Research Policy, 16,291–302.

Von Hippel, E. (2007). The sources of innovation. New York: Oxford University Press.Walter, J., Lechner, C., & Kellermanns, F. W. (2007). Knowledge transfer between and within alliance

partners: Private versus collective benefits of social capital. Journal of Business Research, 60,698–710.

Wang, H. (2003). Policy reforms and foreign direct investment. The case of the Chinese automobileindustry. Journal of Economics and Business, 6, 287–314.

682 Scientometrics (2014) 101:663–683

123

Page 21: Dynamic patterns of technology collaboration: a case study

Wang, Y., Pan, X., Wang, X., Chen, J., Ning, L., & Qin, Y. (2013). Visualizing knowledge space: a casestudy of Chinese licensed technology, 2000–2012. Scientometrics, 98, 1935–1954.

Wang, Y., & Zhou, Z. (2013). The dual role of local sites in assisting firms with developing technologicalcapabilities: Evidence from China. International Business Review, 22, 63–76.

Wasserman, S., & Faust, K. (1994). Social network analysis: Methods and applications. Cambridge:Cambridge University Press.

White, D. R., & Jorion, P. (1996). Kinship networks and discrete structure theory: Applications andimplications. Social Networks, 18, 267–314.

Yam, R., Lo, W., Tang, E. P., & Lau, A. K. (2011). Analysis of sources of innovation, technologicalinnovation capabilities, and performance: An empirical study of Hong Kong manufacturing industries.Research Policy, 40, 391–402.

Zaheer, A., & Soda, G. (2009). Network evolution: The origins of structural holes. Administrative ScienceQuarterly, 54, 1–31.

Zheng, J., Zhao, Z.-Y., Zhang, X., Chen, D.-Z., & Huang, M.-H. (2013). International collaborationdevelopment in nanotechnology: A perspective of patent network analysis. Scientometrics, 98,683–702.

Zhou, P., Zhong, Y., & Yu, M. (2013). A bibliometric investigation on China–UK collaboration in food andagriculture. Scientometrics, 97, 267–285.

Zhu, W., & Guan, J. (2013). A bibliometric study of service innovation research: Based on complex networkanalysis. Scientometrics, 94, 1195–1216.

Scientometrics (2014) 101:663–683 683

123