Exploring supplier-supplier innovations within the Toyota supply network
Potter, Antony; Wilhelm, Miriam
Published in:
Journal of Operations Management
DOI:
10.1002/joom.1124
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below.
Document Version
Publisher's PDF, also known as Version of record
Publication date: 2020
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
Potter, A., & Wilhelm, M. (2020). Exploring supplier-supplier innovations within the Toyota supply network: A supply network perspective. Journal of Operations Management, 66(7-8), 797-819.
https://doi.org/10.1002/joom.1124
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.
R E S E A R C H A R T I C L E
Exploring supplier
–supplier innovations within the Toyota
supply network: A supply network perspective
Antony Potter
1|
Miriam Wilhelm
21Management Science, Alliance
Manchester Business School, University of Manchester, Manchester, UK
2Faculty of Economics and Business,
University of Groningen, Groningen, The Netherlands
Correspondence
Antony Potter, Management Science, Alliance Manchester Business School, University of Manchester, Room 3.102, Booth Street West, Manchester, UK, M15 6PB.
Email: antony.potter@manchester.ac.uk Handling Editor: Subodha Kumar, Sriram Narayanan, Fabrizio Salvador
Abstract
This article investigates the development of supplier–supplier innovations
that occur when two firms that are part of the same supply network co-patent a new product. This study unravels how the structure of the
supply network influences each firm's ability to form supplier–supplier
innovations with other network members. Specifically, we investigate how
supplier degree centrality influences the generation of supplier–supplier
innova-tions, and the extent to which this relationship is moderated by the structural embeddedness of firms in the supply network. Using data from the Toyota supply
network, the results reveal that a firm's ability to co-develop supplier–supplier
innovations with other network members depends on its number of ties and their direction within the supply network. Although betweenness centrality has no sig-nificant moderating effect, closeness centrality, and embeddedness in small world clusters negatively moderate the relationship between supplier degree centrality
and supplier–supplier innovations. Additionally, the number of manufacturing
plants a firm operates in Japan strengthens the positive effect supplier degree
cen-trality has on the development of supplier–supplier innovations.
K E Y W O R D S
betweenness centrality, closeness centrality, degree centrality, small world clusters, supplier– supplier innovation, Toyota Motor Corporation
1
|
I N T R O D U C T I O N
As internally generated innovations are not sufficient to compete in dynamic markets, firms are seeking new ways to leverage their supply networks to co-develop innova-tions (Gao, Xie, & Zhou, 2015; Hong & Hartley, 2011; Narasimhan & Narayanan, 2013). The integration of sup-pliers and customers into internal product development efforts is well documented (Clark & Fujimoto, 1991; Law-son, Krause, & Potter, 2015; Rost, 2011; Zhou, Zhang, Sheng, Xie, & Bao, 2014), and researchers also have begun
to unravel how the structural characteristics of supply net-works can influence the generation of innovations (Narasimhan & Narayanan, 2013). The majority of studies in this field have focused on the relationship between sup-plier degree centrality and the number of innovations cre-ated by firms (Chae, Yan, & Yang, 2020; Gao et al., 2015; Narasimhan & Narayanan, 2013). Yet another route also is pertinent, and arises when firms in the same supply
net-work co-develop supplier–supplier innovations, an effort
that may particularly depend on a firm's embeddedness in the supply network (Hong & Hartley, 2011; Kamath &
DOI: 10.1002/joom.1124
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.
© 2020 The Authors. Journal of Operations Management published by Wiley Periodicals LLC. on behalf of the Association for Supply Chain Management, Inc.
Liker, 1994; Kim, 2014). With this study, we therefore seek to contribute to extant literature on supply network-enabled innovation by asking how the structural character-istics of a supply network influence the development of
supplier–supplier innovations.
Research using social network theory suggests that the network ties between firms within a supply network can play an important role during the innovation process (Bellamy, Ghosh, & Hora, 2014; Gao et al., 2015; Schil-ling & Phelps, 2007). These network ties can be defined as the ties that connect a firm to its upstream suppliers and downstream customers within a supply network (Kim, Choi, Yan, & Dooley, 2011). Anecdotal evidence shows that network ties can lead to the diffusion of valu-able knowledge and operational practices (Choi, Wu,
Ellram, & Koka, 2002; Dyer & Nobeoka, 2000;
Wilhelm, 2011) and thus to enhanced product develop-ment performance (Hong & Hartley, 2011; Kamath & Liker, 1994). In particular, supplier degree centrality, which measures the number of network ties a firm has with other suppliers and customers in the supply net-work, is often associated with a greater degree of innova-tion by firms (Bellamy et al., 2014; Gao et al., 2015). Building upon this literature we anticipate that supplier degree centrality has a positive effect on the
co-development of supplier–supplier innovations within a
supply network setting.
Furthermore, previous studies of supply network-enabled innovations also indicate that the structural embeddedness of firms within a supply network can influ-ence the way they develop new products (Kim, 2014; Narasimhan & Narayanan, 2013). In particular, researchers often measure the structural position of firms in a supply
network according to their betweenness centrality
(Carnovale & Yeniyurt, 2015; Kim et al., 2011), closeness centrality (Borgatti & Li, 2009; Carnovale, Yeniyurt, & Rogers, 2017), and their embeddedness in small world clus-ters (Kito, Brintup, New, & Tsochas, 2014; Sharma, Kumar, Yan, Borah, & Adhikary, 2019). Expanding this research, we propose that the relationship between supplier degree
centrality and the formation of supplier–supplier
innova-tions can be influenced by the structural position of firms within the supply network.
Within this article, we explore these relationships using data from firms within Toyota's supply network
related to their supplier–supplier innovations over a
4-year period from 2015 to 2018. We offer further empirical evidence for the important contribution supplier degree
centrality makes to the co-development of supplier–
supplier innovations. More specifically, we show that the number and direction of ties can have differential effects
on the occurrence of supplier–supplier innovations
between firms across the supply network. Firms that manage a large portfolio of upstream ties with suppliers
tend to be more capable at generating supplier–supplier
innovations, but in contrast, firms with fewer down-stream ties to key customers seem to enjoy more success
in co-developing supplier–supplier innovations in a
col-laborative manner.
Moreover, we unravel how the structural position of firms within the supply network, in terms of their
betweenness centrality, closeness centrality, and
embeddedness in a small world cluster, influences the relationship between supplier degree centrality and the
co-development of supplier–supplier innovations. We
investigate whether betweenness centrality strengthens the positive effect of supplier degree centrality on
sup-plier–supplier innovations due to the unique influence of
network brokers that have a large degree of power and control within a supply network. Even if firms in broker positions (i.e., with a high betweenness centrality) have an important role in managing the flow of materials across supply networks, we do not find evidence that they enjoy any advantages in terms of using their network ties
to generate supplier–supplier innovations. While we
expected that firms with high closeness centrality can use their network ties to rapidly absorb the latest knowledge, our results show that the positive relationship between
supplier degree centrality and supplier–supplier
innova-tions is stronger among firms in more remote posiinnova-tions, on the periphery of the supply network (i.e., with a low closeness centrality).
In line with our expectations, we find that small world clusters weaken, rather than strengthen, the
rela-tionship between supplier degree centrality and supplier–
supplier innovations. This is because firms embedded in
small world clusters can foster supplier–supplier
innova-tions with fewer network ties, but firms outside of these clusters need a larger number of network ties to absorb new knowledge and co-develop innovations within sup-pliers. Finally, we unravel the geographical complexity of the supply network by studying how the location of firms' manufacturing plants influences these relationships. Thus, we go beyond traditional supply network analysis by adding a geographical dimension that captures the number of manufacturing plants each firm operates within Japan. In our study setting, the number of manufacturing plants a firm operates in Japan helps to strengthen the effect supplier degree centrality has on
supplier–supplier innovations.
With this evidence, our study contributes to research into supply network-enabled innovation in three main ways. First, we build on research by Bellamy et al. (2014) and Gao et al. (2015) that has explored the relationship between supplier degree centrality and firm-level innova-tions, to determine how supplier degree centrality might
affect the co-development of supplier–supplier
downstream ties are not equally important for creating
supplier–supplier innovations across inter-firm
bound-aries. Instead, upstream ties with suppliers generate more opportunities for firms to jointly innovate with other net-work members, but downstream ties with customers have negative effects. Second, we find that a firm's net-work position in terms of betweenness centrality, close-ness centrality, and embeddedclose-ness in a small world cluster seems to weaken the positive effect supplier
degree centrality has on the generation of supplier–
supplier innovations. Instead, upstream ties with
sup-pliers are more effective at developing supplier–supplier
innovations when firms are in remote positions in the periphery of the supply network and outside small world
clusters. This finding suggests that supplier–supplier
innovations may evolve in a decentralized manner, resulting from horizontal network ties that connect remote firms in the periphery of the supply network, challenging the assumption that innovations among sup-pliers are being top-down or centrally coordinated. Third, beyond the structural characteristics of supply networks, firms with more manufacturing plants in Japan appear to have network ties that are more capable of generating
supplier–supplier innovations. By accounting for the
den-sity of manufacturing activity in Japan, we introduce a geographical component to considerations of how struc-tural characteristics influence supply network-enabled innovations. Finally, our post hoc analysis reveals that these findings are more pronounced among firms that are developing eco-innovations for new green vehicle architectures in the automotive industry.
2
|
T H E O R E T I C A L B A C K G R O U N D
2.1
|
Supplier
–supplier innovations
Asanuma (1985) described how Japanese automakers convinced competing suppliers to collaborate on product
development and obtain supplier–supplier innovations.
In a continuation of this tradition, many Japanese auto-makers adopt a two-vendor policy and encourage both suppliers to collaborate to design and create new prod-ucts, while also offering protections of each party's status (Kamath & Liker, 1994). Choi et al. (2002) report that
Daimler–Chrysler also encourages suppliers of
instru-ment panels and seats to jointly undertake NPD activi-ties, as part of its buyer-initiated Tech Teams program.
Supplier–supplier innovations occur when a firm
co-develops a patented new product with another organi-zation within an automaker's supply network. Their prac-tical relevance is growing, yet research on this topic lags their real-world existence. In a notable exception, Hong
and Hartley (2011) define supplier–supplier connections
as an approach in which “…the buyer expects first-tier
suppliers to coordinate their activities, communicate directly with each other and make mutual adjustments in component designs without the buyer's direct
involve-ment” (p. 45). The result may be more innovations, with
fewer delays and higher product quality, because the sup-pliers share new knowledge that can improve the inter-face between components and reduce production defects (Choi et al., 2002; Choi & Krause, 2006). However,
supplier–supplier innovations seldom occur in isolation
and instead require considerations of the firm's structural embeddedness in the overall supply network (Choi & Kim, 2008; Kim, 2014; Wilhelm, 2011). Structural
embeddedness refers to how “…a supplier's performance
depends on how it environs itself with other companies
(i.e., its suppliers and customers)” (Choi & Kim, 2008,
p. 5). However, this element of network theory has not been applied systematically to study how a firm's embeddedness in a supply network might influence its ability to develop innovations with other suppliers.
2.2
|
Network theory
Network theory, originating from sociology literature, focuses on the ties of different nodes within a network (Burt, 2004; Granovetter, 1973; Uzzi, 1996). Borgatti and Halgin (2011) summarize the main characteristics of net-work theory as follows: First, the position of nodes, their ties, and the structure of the network explain organiza-tional outcomes, rather than the attributes of the nodes. Second, ties between nodes provide vital conduits for the transfer and diffusion of knowledge within a network. Nodes with more ties and that are centrally positioned in a network receive more knowledge flows than nodes on the periphery of the network with only a few connec-tions. Third, a network is both a sociological construct and a mathematical object, such that its structural char-acteristics can influence how its nodes interact.
Researchers increasingly have applied network theory to the study of supply networks, often using concepts and measures from social network analysis to depict a supply network as the number of firms (nodes), connected together through ties to other members of the supply
net-work (Borgatti & Li, 2009; Galaskiewicz, 2011;
Kim, 2014; Leenders & Dolfsma, 2016). These ties are based on the flow of materials, components, and modules between firms within the supply network (Basole, 2016; Bellamy et al., 2014; Kim et al., 2011). For this study, we focus only on firms that are suppliers to the focal auto-maker (Toyota), through its supplier association, because such firms play important roles for the co-development of innovations (Aoki & Lennerfors, 2013; Dyer & Nobeoka, 2000).
Research in supply network-enabled innovations has contributed to enhancing our understanding of how the structure of the supply network influences the generation of innovations in different industries (Narasimhan & Narayanan, 2013). However, research has yet to explore how supplier degree centrality affects the development of
supplier–supplier innovations, and how this relationship is
influenced by the structural embeddedness of firms within the supply network (Choi & Kim, 2008; Kim, 2014). Con-sequently, we seek to unravel how the relationship
between supplier degree centrality and supplier–supplier
innovations is moderated by the supplier's betweenness centrality (Carnovale & Yeniyurt, 2015; Kim et al., 2011), closeness centrality (Borgatti & Li, 2009; Carnovale et al., 2017), and embeddedness in small world clusters (Kito et al., 2014; Sharma et al., 2019), as well as the num-ber of manufacturing plants in Japan. Figure 1 provides an overview of this conceptual framework and each of our research hypotheses.
3
|
H Y P O T H E S E S D E V E L O P M E N T
3.1
|
Supplier degree centrality as an
antecedent of supplier–supplier
innovations
Using network theory, several studies indicate that firms with more network ties to other organizations within a supply network tend to be more innovative (Bellamy et al., 2014; Gao et al., 2015; Schilling & Phelps, 2007). Degree centrality captures the number of network ties a firm has with other members in the supply network, and it is often regarded as an important antecedent of firm-level
innovation outcomes (Ahuja, 2000; Leenders &
Dolfsma, 2016; Tsai, 2001). For our research context, sup-plier degree centrality refers to the number of upstream
ties with suppliers and downstream ties with customers that the firm has within a supply network (Kim et al., 2011). Building upon previous studies, we expect that the number of upstream ties (Bellamy et al., 2014; Clark & Fujimoto, 1991; Cusumano & Takeshi, 1991) and down-stream ties (Prahalad & Ramaswamy, 2004; Urban & Von Hippel, 1988) a firm manages within a supply network will
influence the way in which it co-develops supplier–
supplier innovations. In particular, the integration of mul-tiple components requires large amounts of collaboration
in product development across the supplier–supplier
inter-face to ensure the smooth functioning of the new product (Hong & Hartley, 2011; Takeishi & Fujimoto, 2001).
Previous studies have shown that when firms have many upstream ties they will often involve and integrate their suppliers into the NPD process to benefit from shorter development times, lower development costs, and higher product quality (Clark & Fujimoto, 1991; Hand-field, Ragatz, Petersen, & Monczka, 1999; Lawson et al., 2015). For example, firms that act as system inte-grators typically maintain ties with different suppliers to access new knowledge and foster inter-firm problem solv-ing dursolv-ing the innovation process (Hobday, Davies, & Prencipe, 2005; Kim et al., 2011). In a similar vein, firms with multiple downstream ties with customers frequently participate in NPD activities with their key clients, so they have more opportunities to jointly develop new
products in a collaborative manner (Prahalad &
Ramaswamy, 2004; Urban & Von Hippel, 1988). Building on these studies, we predict that supplier degree central-ity is equally beneficial for developing inter-firm innova-tions with various members of a supply network. Upstream and downstream ties that span a supply net-work may act as important channels for the formation of
inter-firm NPD projects and the development of
supplier–supplier innovations. We therefore propose the
following research hypothesis:
Supplier-Supplier Innovations Supplier Betweenness Centrality H1+ Supplier Degree Centrality Supplier Closeness Centrality Number of plants located in Japan Small World Clusters H2 + H3 + H4 - H5 + F I G U R E 1 Conceptual framework
Hypothesis 1 Supplier degree centrality is positively
asso-ciated with more supplier–supplier innovations.
3.2
|
Moderating effect of supplier
betweenness centrality
The ability to leverage such network ties for the purpose
of developing supplier–supplier innovations also depends
on the firm's structural embeddedness in the supply net-work (Borgatti & Li, 2009; Choi & Hong, 2002; Kim, 2014). A frequently studied measure of a firm's net-work position is its betweenness centrality, which is often defined as the number of times a firm acts as a link along the shortest path of all combinations of pairs of nodes within the supply network (Freeman, 1979; Kim et al., 2011; Marsden, 2002). Notably, firms with high betweenness centrality are often characterized as brokers that control flows of materials and products among dif-ferent organizations throughout a supply network (Carnovale & Yeniyurt, 2015; Choi & Wu, 2009). Prior research suggests that brokers are often characterized by their high betweenness centrality that enables them to control knowledge flows between different organizations within the supply network (Owen-Smith & Powell, 2004). In an innovation setting, we suggest that betweenness centrality strengthens the positive effect of supplier degree
centrality on the co-development of supplier–supplier
innovations due to the unique role played by brokers. Due to their power and prominence within the supply network, such brokers will have network ties that are more effective
at absorbing new knowledge and generating supplier–
supplier innovations with a range of stakeholders (Leenders & Dolfsma, 2016). Broker firms with higher betweenness centrality accordingly might be able to make more effective use of their network ties as they will be able to use their power and control within the supply network to encourage suppliers and customers with complemen-tary technologies to co-develop innovations with them. For example, Toyota Boshoku Corporation, an important broker in the Toyota supply network has successfully undertaken multiple inter-firm innovation projects with their suppliers and customers that rely on complementary technologies. In contrast, firms that are not brokers likely do not have reputations as important network partners though, so they might not be able to exert the necessary
control to leverage their network ties to foster supplier–
supplier innovations (Carnovale & Yeniyurt, 2015). We therefore propose the following hypothesis:
Hypothesis 2 Supplier betweenness centrality positively moderates the relationship between supplier degree
centrality and supplier–supplier innovations.
3.3
|
Moderating effect of supplier
closeness centrality
In the supply network literature, closeness centrality refers to how close a firm is to all other organizations in a supply net-work beyond the ones it has direct connects to (Freeman, 1979; Kim et al., 2011). It is therefore often used to identify the degree to which a firm has the independence and freedom from the influence of other members of the supply network. Firms with high closeness centrality are often reg-arded as navigators that can quickly reach other organiza-tions within the supply network, including those with which they are not directly connected, using relatively few path lengths within the supply network (Kim et al., 2011; Marsden, 2002). These centrally positioned firms often have the freedom to efficiently absorb knowledge from hard-to-reach parts of the network (Costenbader & Valente, 2003; Sal-man & Saives, 2005), which they then can use for their own innovation activities (Leenders & Dolfsma, 2016).
By extension, we argue that the positive effect of
sup-plier degree centrality on the co-development of supsup-plier–
supplier innovations will be stronger among navigator firms that have a high closeness centrality. In particular, due to their central position in the supply network, these navigator firms can use their upstream and downstream ties more
effectively for the joint development of supplier–supplier
innovations. With their fewer path lengths and shorter sup-ply chains, they are in a better network position to use their network ties to rapidly absorb the latest knowledge, with lit-tle risk of information distortions (Kim et al., 2011). Naviga-tors also have access to knowledge sources outside their network ties with their suppliers or customers, which may be particularly beneficial if the new product being designed relies on other materials, components, and modules in the supply network (Takeishi & Fujimoto, 2001). Firms in more remote, peripheral positions in the supply network instead may face greater difficulties using their network ties to
effectively generate supplier–supplier innovations, due to
their limited and restricted access to knowledge from differ-ent network partners (Carnovale, Rogers, & Yeniyurt, 2016; Schilling & Phelps, 2007). Building upon these arguments, we put forward the below research hypothesis:
Hypothesis 3 Supplier closeness centrality positively moderates the relationship between supplier degree
centrality and supplier–supplier innovations.
3.4
|
Moderating effect of supplier
embeddedness in small world clusters
Rather than clearly defined supplier tiers within a pyra-mid structure, the Toyota supply network containsmultiple, small world clusters (Kito et al., 2014), such that the firms in these clusters have more links con-necting them to one another than to other firms in the network (Schilling & Phelps, 2007). That is, firms within small world clusters have multiple ties among themselves but few ties with organizations outside the cluster that still constitute the wider supply network (Sharma et al., 2019; Watts & Strogatz, 1998). The ties of a firm with other members of a small world cluster help facili-tate efficient knowledge sharing, but ties to more periph-eral members in the supply network grant access to new knowledge from outside the cluster (Galaskiewicz, 2011; Schilling & Phelps, 2007; Sharma et al., 2019). For instance, the average clustering coefficient measure indi-cates the extent to which a firm clusters with neighboring firms, into a tightly knit group that contains multiple ties (Basole, 2016; Sharma et al., 2019).
However, a firm's embeddedness in a small world clus-ter actually might weaken, rather than strengthen, the
rela-tionship between supplier degree centrality and supplier–
supplier innovations. Firms within small world clusters already benefit from intensified knowledge flows, so they might require only a few network ties to absorb new
knowl-edge and co-develop supplier–supplier innovations (Dyer &
Nobeoka, 2000; Sharma et al., 2019). In contrast, firms that have not joined small world clusters need a lot of network ties to absorb all the new knowledge they need to jointly
create supplier–supplier innovations. Yet these firms also
might achieve greater freedom to leverage their upstream and downstream ties to foster innovations with different organizations from other parts of the supply network. In this sense, firms enjoy the independence and autonomy to choose the most creative supplier or innovative customer with which to design new products, without being confined to the few members of the small world cluster. Therefore, we developed the following research hypothesis.
Hypothesis 4 Supplier embeddedness in a small world cluster negatively moderates the relationship between
supplier degree centrality and supplier–supplier
innovations.
3.5
|
Moderating effect of the number of
manufacturing plants in Japan
Network centrality metrics often treat the firm as a sin-gle node, thereby oversimplify the real-life complexity of supply networks. The literature on international
manufacturing networks (Ferdows, Vereecke, &
Meyer, 2016; Rudberg & Olhager, 2003) has pointed out that a firm is usually comprised of multiple facili-ties that are spread over different geographical
locations. Thus, beyond the traditional structural char-acteristics of the supply network, we also add a geo-graphical dimension that captures the density of manufacturing plants that each firm operates within Japan. Although globalization has encouraged many firms to establish manufacturing plants in low-cost locations and emerging markets, Japanese automotive component manufacturers also have retained and expanded their manufacturing plants within Japan (Aoki & Lennerfors, 2013; Dyer & Nobeoka, 2000; Pot-ter & Graham, 2018). In fact, the majority of first-tier suppliers (78%) and second-tier suppliers (65%) in the Toyota supply network are located in Japan (Kito et al., 2014), and many of these firms manage large domestic networks of manufacturing plants. In partic-ular, maintaining a large number of manufacturing plants within Japan helps to foster greater synergies with local suppliers related to alternative fuel technol-ogies (Belderbos, Cassiman, Faems, Leten, & van Looy, 2014; Potter & Graham, 2018). Thus, local manufacturing plants may have a more strategic role, reflecting the typically tight integration of manufactur-ing and research capabilities through local network ties with the automaker (Aoki & Lennerfors, 2013; Morgan & Liker, 2006).
We contend that the number of manufacturing plants a firm operates in Japan will help to strengthen the posi-tive effect supplier degree centrality has on the formation
of supplier–supplier innovations. In particular, firms
with a large Japanese manufacturing base are able to use
their extensive manufacturing capabilities to help
develop the engineering and research capabilities of local suppliers as well (Dyer & Nobeoka, 2000; Lawson et al., 2015). Large automotive component manufacturers even operate so-called mother plants (i.e., lead factories) and showcase facilities in Japan that maintain a reposi-tory of the firm's manufacturing and research knowhow,
which is often used to help upgrade local suppliers’ R&D
capabilities (Cheng, Farooq, & Johansen, 2015; Dyer & Nobeoka, 2000; Ferdows et al., 2016). Firms in the
Japa-nese automotive industry with multiple domestic
manufacturing plants also tend to engage in personnel exchanges with suppliers and customers, to share new knowledge, train workers, experiment with prototypes, and jointly invest in R&D, which helps to ensure that
their network ties are used to develop supplier–supplier
innovations (Aoki & Lennerfors, 2013; Dyer &
Nobeoka, 2000; Todo, Matous, & Inoue, 2016). Accord-ingly, we anticipate that firms with a large manufactur-ing presence in Japan possess local network ties that are more effective at knowledge sharing and co-developing
supplier–supplier innovations. This leads us to the
Hypothesis 5 The number of manufacturing plants located in Japan positively moderates the
relation-ship between supplier degree centrality and supplier–
supplier innovations.
4
|
R E S E A R C H M E T H O D O L O G Y
4.1
|
Research context: Toyota supply
network
Increasing trends toward collaboration between Japanese automakers and suppliers during innovation processes have led to a substantial rise in the number of co-patents
within supply networks (Borgstedt, Neyer, &
Schewe, 2017; Konno, 2007; Potter & Graham, 2018). For this study, we consider 219 firms within the Toyota sup-ply network that manufacture the materials, components, and modules that go into Toyota's vehicles. Based on our analysis, all these firms are first-tier suppliers that supply parts to Toyota and are members of its supplier associa-tion (Kyohokai). The Toyota supplier associaassocia-tion aims to
enhance members’ cooperation and knowledge sharing
with Toyota and one another, through regular meetings and workshops (Toyota, 2016). Members of Kyohokai are Toyota's most important material suppliers, accounting for more than 90% of Toyota's purchasing (Sako, 1996). We therefore focus our analysis on the product innova-tions developed by material suppliers within the Kyohokai part of the supplier association. However, due to data limitations, we are unable to incorporate firms in the lower tiers of the supply network as this is beyond the scope of our study (Kito et al., 2014).
We use secondary data sources to measure the vari-ables. The Japan Patent Office (JPO) is one of the main sources of data we use in our study and we use its patent data to collect information on the number of co-patents recorded by each firm and other Toyota suppliers within the supply network over a 4-year period from 2015 to
2018 (i.e., t0to t+4). By studying this period, we can
evalu-ate the direction of causality and investigevalu-ate whether
sup-plier degree centrality at t0 is an antecedent of later
supplier–supplier innovations. In addition, this time
period corresponds to the 4 years following the launch of the fourth-generation Toyota Prius hybrid vehicle and the Toyota Mirai vehicle that is a hydrogen fuel cell vehi-cle that began mass production in 2015. The measures of supplier degree centrality, betweenness centrality, close-ness centrality, small world clusters, number of plants in
Japan, and control variables all refer to 2015 (t0), so that
we can make inferences about causality. To measure the centrality and small world cluster variables, we use the Marklines database, which records transactions by firms
in the global automotive industry. It contains information about 40,000 firms, making it one of the most detailed databases of automakers and their suppliers. Kito et al. (2014) also used the Marklines database to establish the detailed structure of the Toyota supply network. We also cross-referenced the data on network ties with other secondary data sources such as the S&P Capital IQ data-base of supplier and customer ties in the automotive industry. Finally, we gather data from company websites to record the number of plants in Japan owned by each firm in the Toyota supply network.
4.2
|
Measures
4.2.1
|
Dependent variable: Supplier
–
supplier innovations
Patents are a widely used measure of the innovation performance of different organizations (Artz, Norman,
Hatfield, & Cardinal, 2010; Liu, Yeung, Lo, &
Cheng, 2014). To capture the degree of inter-firm inno-vation, an alternative measure focuses on the total num-ber of co-patents a firm generates during a specific period (Kim & Song, 2007; Potter & Graham, 2018). Within this study, we use patent data from the Japan Platform for Patent Information (JPPI) database that is managed by the Japan Patent Office. Specifically, we use this database to collect patent data on the number of co-patents registered by different firms within the Toyota supply network. We began by measuring how many co-patents each firm in the Toyota supply network had reg-istered that also included one other Toyota supplier as a
co-assignee from 2015 (t0) to 2018 (t+4). We only focus
on co-patents with each firm and another Toyota sup-plier as registered co-assignees. In other words, we
mea-sure supplier–supplier innovations by focusing exclusively
on the co-patents that are registered by two suppliers within the Toyota supply network. Therefore, we do not include co-patents that are registered with the automaker as this is beyond the scope of our study. Then we cross-referenced these co-patents with other secondary data sources, checking that they appeared in the Derwent Inno-vation Index and Google patents database. We recorded
the total, such that our measure of supplier–supplier
inno-vation reflects the total number of other Toyota suppliers
each firm in the Toyota supply network had co-patented with during the 4-year study window. Overall, we find that
firms often develop supplier–supplier innovations with
their immediate suppliers as they account for 20% of all
supplier–supplier innovations in the supply network,
whereas immediate customers only constitute 13%. In
innovations involve the firm collaborating with its immedi-ate suppliers or customers to jointly design and pimmedi-atent an innovative new product. This also indicates that the
major-ity of supplier–supplier innovations represent
collabora-tions between firms and the other members of the Toyota supply network with whom they do not have network ties.
4.2.2
|
Explanatory variables: Supplier
indegree and outdegree centrality
Similar to Kito et al. (2014), we identify the company name of each firm in the Marklines database, then record the total number of upstream ties and down-stream ties it has in the Toyota supply network. With this information, we developed a supplier indegree
cen-tralityvariable that measures the total number of
sup-pliers from which each firm sources materials,
components, parts, and modules, within the Toyota supply network. For example, supplier indegree cen-trality is the number of direct ties that flow to a firm (node) from its suppliers in the supply network. Adapting the approach used by Kim et al. (2011) and Basole (2016), this can be defined as:
DCi=
X
j
nij
where nijis equal to 1 if there is a direct tie that flows to
nifrom nj, and equal to 0 otherwise (Freeman, 1979; Kim
et al., 2011).
Then we created a separate supplier outdegree
cen-tralityvariable that indicates the number of customers
in the Toyota supply network to which each firm sells its products (Basole, 2016; Kim et al., 2011). With this detailed data, our empirical analysis spans two levels, and we test our hypotheses using data about supplier indegree centrality and supplier outdegree centrality separately. Therefore, we can evaluate whether our results only apply to specific interactions between a firm and its Toyota suppliers (i.e., supplier indegree centrality) or its customers within the Toyota supply network (i.e., supplier outdegree centrality). As we are interested in the network ties among the firms in the Toyota supply network, we do not include data on the ties between each firm and the focal automaker (Toyota) for the network measures used in this study.
4.2.3
|
Moderator: Betweenness
centrality
Previous studies have measured betweenness centrality by focusing on how often a firm appears on the shortest
paths between other nodes in the network (Carnovale & Yeniyurt, 2015). Following Basole (2016), betweenness centrality is measured as follows:
BCi=
X
q,j≠iPq,i,j
where pq,i,jis the proportion of shortest paths that occur
between q and j that occur through i in the supply net-work. In a similar manner to Basole (2016) we use our data on network ties together with Gephi network
soft-ware1to measure the betweenness centrality of each firm
within the Toyota supply network.
4.2.4
|
Moderator: Closeness centrality
Closeness centrality is often defined as the sum of dis-tances from all other firms within the network, in which distance from one firm to another is regarded as the length (in links) of the shortest path (Carnovale et al., 2017). According to Basole (2016) and Carnovale et al. (2017), closeness centrality can be expressed as:
CCi=
Xn
i= 1d pi, pj
In the above equation, d(pi,pj)represents the number
of ties in the shortest path that connect piand pj. As we
are studying a supply network with directional ties between firms, we use the harmonic version of closeness centrality. Specifically, harmonic closeness centrality cap-tures the sum of the inverted distances and is expressed in the following equation:
HCCi= X i≠j 1 d pi, pj
Consequently, firms with a high harmonic closeness centrality will be centrally positioned, whereas firms with a low harmonic closeness centrality will be in remote positions in the periphery of the supply network (Carnovale et al., 2016).
4.2.5
|
Moderator: Supplier
embeddedness in a small world cluster
Small world clusters were originally proposed by Watts and Strogatz (1998) and can be identified when a firm has a high average clustering coefficient within a supply network (Sharma et al., 2019). Specifically, the average clustering coefficient can be used to measure the extentto which a firm's network partners form a cluster within the supply network (Basole, 2016; Sharma et al., 2019). For example, Kito et al. (2014) use average clustering coefficients to help analyze small world clusters within the Toyota supply network. Similar to Basole (2016), our measure of the average clustering coefficient is expressed in the following manner:
ACCi=
np
niðni−1Þ=2
In the above equation np represents the number of
ties that occur among all nidirect partners p of each firm
i (Basole, 2016). In particular, a high average clustering
coefficient indicates that a firm is closely tied to its adja-cent firms within a small world cluster in comparison to a situation where all the ties are randomly distributed (Sharma et al., 2019).
4.2.6
|
Moderator: Number of plants in
Japan
We used data gathered from company websites to iden-tify the addresses of the individual manufacturing plants
that each firm operates in 2015 (t0). We then determined
the number of plants in Japan owned by each firm in the Toyota supply network. We cross-referenced these data with secondary data sources such as Marklines, S&P Cap-ital IQ, and company annual reports, to verify that the firm owned and operated each manufacturing plant.
4.2.7
|
Control variables
We controlled for organizational factors that might influ-ence the frequency of co-patenting (Artz et al., 2010). Firm
size refers to the total number of employees working for
each organization, obtained from the S&P Capital IQ data-base and company websites (log). Firm age is the number of years since the firm was established (log). For tional governance, we capture whether the firm is interna-tionally listed with headquarters outside Japan (1 = yes). Overall, the majority of firms (93.62%) are Japanese owned with a domestic headquarters, and only a small number of Toyota's suppliers are internationally listed. Previous stud-ies indicate that R&D centers influence innovation pro-cesses, so we measured the number of research centers, institutes, and laboratories owned by the firm (log).
Two other variables capture different product charac-teristics. First, Hong and Hartley (2011) and Salvador and Villena (2013) demonstrate that module systems have a significant effect on how firms develop innovations. We used detailed data from a list, published by Toyota, of
suppliers and the products they manufacture for the automaker to identify whether each firm manufactures a modular system, sub-system, assembly, or sub-assembly (Toyota, 2016). In a database of all components provided by each firm to Toyota, we recorded a value of 1 if the description of a component included a keyword related to a module system, such as module, sub-assembly,
assembly, sub-system, or system. We also
cross-referenced these data with the Marklines database and company websites to confirm that the firm supplied mod-ule systems. Second, using data from Toyota's (2016) list of suppliers and components, we determined if each firm provided parts, components, or products for Toyota's
engine system(1 = yes).
Automakers and suppliers have invested in green technologies, so we also checked whether the results might be more pronounced among green technology sup-pliers. Similar to Borgstedt et al. (2017) and Potter and Graham (2018), we used data from the JPO and searched patent titles, abstracts, and codes registered by different firms to identify if they had developed any eco-innovations used in vehicles that were powered by hybrid, electric, or hydrogen fuel cell powertrains during a 4-year period. This dichotomous variable, Green
Tech-nology Supplier(GTS), measures whether a firm had
pat-ented an eco-innovation in green technology (1 = yes). To capture the effects of distinct keiretsu networks of interlocking ownership in Japan, we created another dichotomous variable, Toyota investment, which mea-sures whether a firm had received financial investment
from the automaker. To address supplier–supplier
gover-nance, we also used a dichotomous variable called
sup-plier investment that captures whether another Toyota
supplier had invested financially in each firm in the sup-ply network (coded 1). Both these measures reflect data from the S&P Capital IQ database, Toyota's financial
accounts, the suppliers’ annual reports, and company
websites. Long-term relationships influence the develop-ment of innovations in inter-firm collaborations, so we used data from Sako (1996) that identify whether a firm was a member of the Toyota supply network in 1980, then construct a dichotomous variable to represent its
long-term relationship with Toyota (coded 1 for yes).
Finally, the degree of competition varies across compo-nent markets in the automotive industry. Therefore, we measured market competition as the number of major competitors that produced the products manufactured by the firm, using data from automotivenews.com.
4.3
|
Empirical analysis
The systematic investigation of the data relied on multi-variate ordinary least square (OLS) regression techniques
(Woolridge, 2016). As mentioned, the empirical analysis proceeds at two levels in the Toyota supply network, so we used a two-stage approach, by testing the hypotheses with data about supplier indegree centrality, then repeat-ing the analysis with information gathered about supplier outdegree centrality. The Variance Inflation Factors (VIF) statistics indicate that multicollinearity is not a concern; we provide a correlation matrix in Table 1. Fol-lowing guidelines by Ketokivi and McIntosh (2017) and Woolridge (2016), we check for endogeneity using two-stage least squares (2SLS) regression with instrumental variables. In addition, we also test whether our results are more pronounced among firms developing
eco-innovations for new green vehicle architectures
(Appendix).2
5
|
E M P I R I C A L R E S U L T S
5.1
|
Supplier indegree centrality: Main
and moderator effects
In Table 2, we present the empirical findings in
sequen-tial order: Models 1–4 reveal the effects of supplier
inde-gree centrality, whereas Models 5–7 pertain to the
contributions of supplier outdegree centrality. Model 2 offers support for the positive effect of supplier indegree
centrality on supplier–supplier innovations (B = 0.470;
p< .000), and Model 4 indicates that betweenness
cen-trality has no significant moderating effect on this
rela-tionship (B = −0.219; p = .132). Contrary to our
expectations, closeness centrality had a significant nega-tive moderating effect on the relationship between
sup-plier indegree centrality and supplier–supplier
innovations (B = −0.571; p < .017). During our initial
moderation analysis, we began by creating equations for
each relationship at−1 standard deviation (low) and +1
standard deviation (high) values of the explanatory and moderator variables to investigate the moderating effects using simple slope statistics. Next, we used conditional interaction plots to undertake a more detailed examina-tion of the relaexamina-tionships at the 10th percentile (low) and 90th percentile (high) values of the explanatory variable, together with the moderator values at the 25th percentile (low) and 75th percentile (high) values. This approach helps us to illustrate the nature of the relationships within the interaction plots without the potential influ-ence of outlier firms. When firms have high closeness centrality, we find no significant relationship between
supplier indegree centrality and supplier–supplier
inno-vations (B =−0.358; p = .610), but if their closeness
cen-trality is low, a significant positive relationship emerges, albeit at a 90% significance level (B = 1.612; p < .071).
The interaction plot affirms the positive relationship
between supplier indegree centrality and supplier–
supplier innovations only occur when there is a low degree of closeness centrality (Figure 2).
Regarding the moderating effect of small world clus-ters, in line with our expectations, we find that they exert a negative moderating effect by weakening the
relation-ship between supplier indegree centrality and supplier–
supplier innovations (B =−0.787; p < .004). The results
from the moderation analysis identify that supplier inde-gree centrality has no statistically significant effect on
supplier–supplier innovations when firms are located
within small world clusters (B = −2.207; p < .113). In
contrast, when firms are positioned outside of small world clusters a significant moderating effect occurs that strengthens the relationship between supplier indegree centrality and the co-development of innovations across
supplier–supplier boundaries (B = 3.461; p < .001). This
moderating effect is further evident in the interaction plot in Figure 3 where the positive relationship between
supplier indegree centrality and supplier–supplier
inno-vations emerges only if the level of small world clustering is low.
Finally, the results in Model 4 of Table 2 demonstrate that the number of manufacturing plants in Japan has a positive moderating effect on the ability of supplier
inde-gree centrality to generate supplier–supplier innovations
(B = 0.412; p < .000). In the moderation analysis, firms with many manufacturing plants in Japan experience a positive and significant relationship between supplier
indegree centrality and supplier–supplier innovations
(B = 1.421; p < .048), but no significant relationship exists for firms with a small number of manufacturing
plants in Japan (B =−0.167; p = .809). To illustrate this
finding, the interaction plot in Figure 4 reveals a positive relationship only when firms have many manufacturing plants in Japan.
5.2
|
Supplier outdegree centrality: Main
and moderator effects
Supplier outdegree centrality is not significantly
associ-ated with supplier–supplier innovation in Model 5
(B =−0.105; p = .109), though it seemingly generates a
significant negative effect in Model 6 (B = −0.212;
p< .001). Whereas supplier indegree centrality appears
to be a positive antecedent of supplier–supplier
innova-tions, our results suggest supplier outdegree centrality has a negative effect. We next check for moderating effects (Model 7, Table 2) but do not find any evidence of
significant moderation by betweenness centrality
TABLE 1 Correlation matrix Variables Min Max Mean Stand ard dev iation 1 23456 7891 0 1 1 1 2 1 3 1 4 1 5 1 6 1 7 1 8 1. Firm size 21.00 0 320, 725.000 13,8 14.292 36,5 58.931 1.000 2. Firm age 4.000 162.000 75.41 1 26.15 8 0.041 (0.544) 1.000 3. International governance 0.000 1.000 0.064 0.245 − 0.017 (0.809) 0.016 (0.81 1) 1.000 4. R&D centers 0.000 15.000 0.950 2.405 0.306 *** (0.000) 0.022 (0.74 5) 0.249*** (0.00 0) 1.000 5. Module system 0.000 1.000 0.073 0.261 0.202 ** (0.003) 0.029 (0.67 4) 0.070 (0.30 2) 0.243*** (0.000) 1.000 6. Engine system 0.000 1.000 0.146 0.354 − 0.056 (0.414) 0.137* (0.04 2) − 0.002 (0.97 2) − 0.064 (0.343) − 0.116 † (0.086) 1.000 7.
Green technology supplier
0.000 1.000 0.648 0.479 0.357 *** (0.000) − 0.054 (0.43 1) 0.114 † (0.09 2) 0.328*** (0.000) 0.133 * (0.049) − 0.074 (0.273) 1.000 8. Toyota investment 0.000 1.000 0.192 0.395 0.208 ** (0.002) − 0.082 (0.22 8) − 0.127 † (0.06 0) − 0.063 (0.354) 0.130 † (0.054) − 0.037 (0.583) 0.116 † (0.08 7) 1.000 9. Supplier investment 0.000 1.000 0.292 0.456 0.178 ** (0.008) − 0.102 (0.13 4) 0.037 (0.58 3) − 0.021 (0.758) 0.090 (0.186) − 0.038 (0.572) 0.137* (0.04 3) 0.274*** (0.00 0) 1.000 10. Lon g-term relationship 0.000 1.000 0.233 0.424 0.131 † (0.054) − 0.042 (0.53 6) 0.033 (0.63 1) − 0.003 (0.962) − 0.030 (0.657) − 0.075 (0.269) 0.225** (0.00 1) 0.116 † (0.08 7) 0.074 (0.279) 1.000 11. Marke t competition 1.00 12.000 2.269 2.168 0.167 * (0.013) 0.110 (0.10 4) 0.189** (0.00 5) 0.226** (0.001) 0.200 ** (0.004) 0.068 (0.316) 0.218** (0.00 1) 0.245*** (0.00 0) 0.105 (0.122) 0.046 (0.499) 1.000 12. Suppli er Indegree centrality 0.000 139.000 6.132 15.12 7 0.288 *** (0.000) − 0.034 (0.61 8) 0.055 (0.42 1) 0.246*** (0.000) 0.135 * (0.046) 0.032 (0.643) 0.209** (0.00 2) 0.280*** (0.00 0) 0.049 (0.471) − 0.014 (0.835) 0.241 *** (0.000) 1.000 13. Suppli er Outdegree centrality 0.000 26.000 2.986 4.119 − 0.172* (0.011) − 0.025 (0.71 1) − 0.126 † (0.06 2) − 0.183 ** (0.006) 0.091 (0.182) 0.005 (0.947) − 0.226** (0.00 1) 0.112 † (0.09 9) 0.031 (0.643) − 0.030 (0.662) − 0.130 † (0.055) − 0.134* (0.048) 1.000 14. Be tweenness centrality 0.000 6,457.616 241.4 89 719.8 44 0.163 * (0.016) − 0.736 (0.27 8) − 0.004 (0.95 2) 0.224** (0.001) 0.149 * (0.028) 0.003 (0.964) 0.091 (0.18 2) 0.200** (0.00 3) 0.054 (0.429) 0.048 (0.478) 0.149 * (0.028) 0.729 *** (0.000) 0.126 † (0.06 4) 1.000 15. Clo seness centrality 0.000 0.848 0.357 0.184 0.106 (0.119) − 0.021 (0.76 0) 0.022 (0.74 5) 0.091 (0.180) 0.179 ** (0.008) 0.071 (0.296) 0.055 (0.41 7) 0.280*** (0.00 0) 0.086 (0.204) 0.024 (0.721) 0.178 ** (0.008) 0.457 *** (0.000) 0.426*** (0.00 0) 0.390*** (0.00 0) 1.000 16. Sm all world clusters 0.000 1.000 0.240 0.173 − 0.132 † (0.051) 0.107 (0.11 4) 0.063 (0.35 4) 0.002 (0.973) − 0.032 (0.634) 0.161 * (0.017) − 0.011 (0.87 7) 0.038 (0.57 6) − 0.069 (0.311) 0.041 (0.551) 0.023 (0.739) − 0.179** (0.008) 0.114 † (0.09 1) − 0.157 * (0.02 0) 0.201** (0.003) 1.000 17. Num ber of plants locate d in Japan 0.000 39.000 5.500 5.900 0.386 *** (0.000) 0.006 (0.92 9) 0.001 (0.98 9) 0.339*** (0.000) 0.091 (0.182) − 0.150* (0.026) 0.272*** (0.00 0) 0.008 (0.91 1) 0.079 (0.244) 0.168* (0.013) 0.058 (0.391) 0.248 *** (0.000) − 0.044 (0.51 6) 0.341*** (0.00 0) 0.157* (0.020) − 0.071 (0.294) 1.000 18. Suppli er – supplier innovations 0.000 24.000 1.671 3.281 0.414 *** (0.000) − 0.021 (0.75 7) − 0.071 (0.29 8) 0.248*** (0.000) 0.162 * (0.016) − 0.037 (0.582) 0.245*** (0.00 0) 0.155* (0.02 2) 0.098 (0.147) 0.125 † (0.066) 0.180 ** (0.008) 0.542 *** (0.000) − 0.171* (0.01 1) 0.541*** (0.00 0) 0.192** (0.004) − 0.131 † (0.054) 0.487 *** (0.000) 1.000 Note: In addition, our results indicate that multicollinearity is not a concern, as the Variance Inflation Factor (VIF) average (mean) score is 1.59, which is below the threshold value of 10. *** p < .001; ** p < .01; * p < .05; † p < 0.10.
TA BLE 2 The antecedent and moderating factors that influence supplier –supplier innovations (2015 –2018) Variables Model 1 Mod el 2 Model 3 Model 4 Variables Model 5 Model 6 Model 7 Control variables Control variables Firm size 0.309*** (0.000) 0.228*** (0.000) 0.176** (0.004) 0.266*** (0.000) Firm size 0.294*** (0.000) 0.186** (0.003) 0.193** (0.002) Firm age − 0.037 (0.561) − 0.013 (0.817) − 0.013 (0.798) − 0.016 (0.744) Firm age − 0.037 (0.555) − 0.06 (0.912) − 0.005 (0.924) International governance − 0.124 † (0.059) − 0.137* (0.018) − 0.102 † (0.058) − 0.089 † (0.080) International governance − 0.131* (0.046) − 0.101 † (0.060) − 0.090 † (0.090) R&D centers 0.139 † (0.053) 0.065 (0.308) − 0.022 (0.714) − 0.057 (0.319) R&D centers 0.131 † (0.067) − 0.035 (0.571) − 0.064 (0.293) Module system 0.048 (0.465) 0.043 (0.452) 0.052 (0.332) 0.077 (0.135) Module system 0.065 (0.322) 0.056 (0.299) 0.086 (0.123) Engine system 0.004 (0.953) − 0.024 (0.672) 0.012 (0.818) 0.033 (0.488) Engine system 0.01 (0.932) 0.013 (0.804) 0.012 (0.812) Green technology supplier 0.054 (0.440) 0.022 (0.721) 0.018 (0.759) 0.051 (0.346) Green technology supplier 0.037 (0.603) 0.020 (0.734) 0.032 (0.595) Toyota investment 0.036 (0.604) − 0.081 (0.196) − 0.036 (0.536) − 0.078 (0.159) Toyota investment 0.051 (0.465) − 0.001 (0.986) − 0.000 (0.999) Supplier investment 0.012 (0.859) 0.042 (0.462) 0.026 (0.627) 0.030 (0.555) Supplier investment 0.015 (0.811) 0.015 (0.784) − 0.007 (0.891) Long-term relationship 0.068 (0.289) 0.105 † (0.065) 0.051 (0.331) 0.035 (0.483) Long-term relationship 0.070 (0.273) 0.029 (0.583) 0.044 (0.395) Market competition 0.090 (0.190) 0.040 (0.511) 0.061 (0.279) 0.069 (0.184) Mark et competition 0.078 (0.257) 0.045 (0.430) 0.046 (0.412) Main effect Main effect Supplier indegree centrality – 0.470*** (0.000) 0.307*** (0.000) 0.191 (0.363) Supplier outdegree centrality − 0.105 (0.109) − 0.212*** (0.001) − 0.385 † (0.084) Moderator effects Moderator effects Betweenness centrality –– 0.238** (0.003) 0.149 (0.291) Betweenness centrality – 0.425*** (0.000) 0.446*** (0.000) Closeness centrality –– − 0.121 †(0.050) − 0.203** (0.009) Closeness centrality – 0.041 (0.528) 0.200 (0.407) Small world clusters –– 0.036 (0.517) − 0.291* (0.018) Small world clusters – 0.000 (0.997) 0.060 (0.449) Number of plants located in Japan –– 0.269*** (0.000) 0.178** (0.003) Number of plants located in Japan – 0.249*** (0.000) 0.161* (0.014) Interaction effects Interaction effects Supplier indegree centrality × Betweenness centrality –– – − 0.219 (0.132) Supplier outdegree centrality × Betweenness centrality –– 0.077 (0.522) Supplier indegree centrality × Closeness centrality –– – − 0.571* (0.017) Supplier outdegree centrality × Closeness centrality –– 0.218 (0.450) Supplier indegree centrality × Small world clusters –– – − 0.787** (0.004) Supplier outdegree centrality × Small world clusters –– 0.081 (0.302)
p = .450), or small world clusters (B = .081; p = .302). Only the number of plants located in Japan has a signifi-cant, negative moderating effect on the relationship
TA BLE 2 (Continued) Variables Model 1 Mod el 2 Model 3 Model 4 Variables Model 5 Model 6 Model 7 Supplier indegree centrality × Number of plants located in Japan –– – 0.412*** (0.000) Supplier outdegree centrality × Number of plants located in Japan –– − 0.222*** (0.001) Adjusted R 2 0.183 0.366 0.460 0.544 Adjusted R 2 0.189 0.458 0.478 Overall model F 5.43*** 11.46*** 12.62*** 13.99*** Overall model F 5.23*** 12.49*** 10.97*** N 219 219 219 219 N 219 219 219 Note: Dependent variable: Number of supplier –supplier innovations during the 4-year period from 2015 to 2018. Additionally, Sharma et al. (2019) suggest that small world clusters can some-times be detected by their path lengths within the supply network. Therefore, in a separate analysis, we replaced our measure of average clustering co efficients with a variable called path lengths and found similar results. *** p < .001; ** p < .01; * p < .05; †p < .10.
F I G U R E 2 The moderating effect of closeness centrality
between supplier outdegree centrality and supplier–
supplier innovations (B =−0.222; p < .001). The results
from the moderation analysis reveal that when firms have many manufacturing plants located within Japan a negative relationship emerges between supplier
out-degree centrality and the intensity of supplier–supplier
innovations (B =−2.200; p < .009). By comparison, when
firms have a small number of manufacturing plants in
Japan there is no moderating effect (B = −0.326;
p= .655). The interaction plot in Figure 5 confirms that
the number of plants located in Japan has a negative moderating effect on how supplier outdegree centrality
influences supplier–supplier innovations.
5.3
|
Endogeneity
5.3.1
|
Supplier indegree centrality
In accordance with recommendations by Ketokivi and McIntosh (2017) and Woolridge (2016), we used a 2SLS regression with instrumental variables to check for endo-geneity. The instrumental variable, external indegree cen-trality, measures the total number of ties each firm has to suppliers outside the Toyota supply network. Kim et al. (2011) and Kito et al. (2014) show that the way a firm manages its ties in a supply network often depends on the
size and scope of its external supply base. For example, firms with a high external indegree centrality likely manu-facture products that require them to integrate materials from external sources, together with different components from suppliers within the supply network. Furthermore, firms that know how to source from different external sup-pliers likely have sufficient purchasing capabilities to man-age their upstream ties with suppliers within the supply network too. A second instrumental variable, external out-degree dependence, instead captures the percentage of product ties with external customers, outside the Toyota supply network. By taking a percentage measure, we can capture the relative importance of external customers to each firm. With a large external customer base, firms likely adopt an external orientation in sourcing their mate-rials, components, and parts from a wider range of sup-pliers so that they can meet the unique demands of different automakers (Kim et al., 2011). Therefore, we expect that external outdegree dependence will be associ-ated with externally oriented firms that source from rela-tively fewer suppliers within the supply network. Furthermore, as external indegree centrality and external outdegree dependence help to capture the orientation of firms towards external partners they are unlikely to
influ-ence the co-development of supplier–supplier innovations
within the supply network. F I G U R E 4 The moderating effect of the number of plants
located in Japan (supplier's indegree centrality)
F I G U R E 5 The moderating effect of the number of plants located in Japan (supplier's outdegree centrality)
To be valid instrumental variables, external indegree centrality and external outdegree dependence must relate significantly to supplier indegree centrality. Therefore, in the first stage of the 2SLS estimation, we include supplier indegree centrality as a dependent variable and regress it against the control and instrumental variables (Bellamy et al., 2014). The results in Model 1 of Table 3 reveal that external indegree centrality (B = 0.278; p < .000) and
external outdegree dependence (B = −0.259; p < .000)
both have significant effects on the dependent variable, supplier indegree centrality. As suggested by Bellamy et al. (2014), we also check the validity of the
instrumental variables by applying over-identifying
restriction tests. Although it is not possible to test the sta-tistical independence of the instrumental variables from the error terms for the dependent variable, we can use this test of over-identifying restrictions to confirm their relevance (Bellamy et al., 2014). The null hypothesis states that we have valid instrumental variables, so an insignificant p value in the Sargan and Bassman tests would indicate that we can use these instrumental
vari-ables to test for the presence of endogeneity
(Woolridge, 2016). Both the Sargan (p = .694) and Bas-sman (p = .703) tests are statistically insignificant, in T A B L E 3 Investigating endogeneity using 2SLS regression with instrumental variables
Variables Model 1 supplier Indegree centrality (OLS) Model 2 supplier– supplier innovations (2SLS) Variables Model 3 supplier Outdegree centrality (OLS) Model 4 supplier– supplier innovations (2SLS)
Control variables Control variables
Firm size 0.117†(0.075) 0.817*** (0.000) Firm size −0.014 (0.802) 0.938*** (0.000) Firm age −0.089 (0.128) −0.060 (0.740) Firm age 0.018 (0.715) −0.123 (0.541) International
governance
−0.079 (0.219) −1.795** (0.017) International governance
0.013 (0.805) −1.821* (0.032) R&D centers 0.139* (0.037) 0.268 (0.211) R&D centers −0.026 (0.646) 0.412†(0.071) Module system −0.028 (0.643) 0.558 (0.430) Module system 0.106* (0.042) 0.981 (0.231) Engine system 0.051 (0.384) −0.161 (0.750) Engine system 0.055 (0.274) 0.060 (0.915) Green technology
supplier
0.030 (0.654) 0.202 (0.631) Green technology supplier
−0.042 (0.459) 0.167 (0.727) Toyota investment 0.188** (0.004) −0.455 (0.422) Toyota investment 0.091†(0.096) 0.509 (0.367) Supplier investment −0.048 (0.427) 0.255 (0.532) Supplier investment 0.040 (0.438) 0.132 (0.772) Long-term relationship −0.049 (0.406) 0.747†(0.085) Long-term
relationship
0.117** (0.023) 0.553 (0.248) Market competition 0.112†(0.079) 0.167 (0.400) Market competition −0.066 (0.221) 0.226 (0.303)
Explanatory variable Explanatory
variable Supplier indegree centrality – 1.197** (0.008) Supplier outdegree centrality – −0.594†(0.056)
Instrumental variables Instrumental variables External indegree centrality 0.278*** (0.000) – Internal outdegree dependence 0.715*** (0.000) – External outdegree dependence −0.259*** (0.000) – Japan manufacturing dependence −0.127* (0.023) – R2 0.348 0.392 R2 0.523 0.229 N 219 219 N 219 219
Note:Supplier–supplier innovations is the dependent variable in Models 1, 3, 4, and 6. Supplier indegree centrality is the dependent variable in Model 2 and Supplier outdegree centrality is the dependent variable in Model 5. The measure of supplier–supplier innovations focuses on the number of supplier–supplier innovations during the 4-year period from 2015 to 2018. ***p < .001; **p < .01; * p < .05;†p< .10.
support of the validity of the instrumental variables. Fur-thermore, the partial F statistic suggests they are strong rather than weak (Woolridge, 2016).
In the second stage of the 2SLS estimation procedure
supplier–supplier innovations is used as the dependent
variable and is regressed against supplier indegree central-ity as the explanatory variable, as well as the predicted values from the first-stage estimation as independent vari-ables (Bellamy et al., 2014). The results in Model 2 in Table 3 reveal that supplier indegree centrality relates
pos-itively to supplier–supplier innovations (B = 1.197;
p< .008). According to the Durbin–Wu–Hausman (DWH)
post-estimation test of endogeneity, our results are not influenced by endogeneity concerns. The significance
levels for the Durbin score (p = .396) and Wu–Hausman
test (p = .412) are greater than 0.10, so we do not reject the null hypothesis that our variables are exogenous (Bellamy et al., 2014). Overall, the relationship between
supplier indegree centrality and supplier–supplier
innova-tions does not appear subject to endogeneity concerns.
5.3.2
|
Supplier outdegree centrality
We also explored potential endogeneity in the
relation-ship between supplier outdegree centrality and supplier–
supplier innovations, with a similar method. The instrumental variable internal outdegree dependence mea-sures the percentage of each firm's product ties that are with customers from the supply network, so we can cap-ture the internal orientation of each firm toward cus-tomers within the supply network. We expect that firms for which a greater share of their total customer base is within the Toyota supply network will supply products to many network members. As a second instrumental vari-able, we use Japan manufacturing dependence, which measures the percentage of each firm's total manufactur-ing plants located in Japan. In the highly competitive Japanese automotive industry, firms that are more depen-dent on Japan as a manufacturing location may sell their products to a wider variety of Japanese automakers and suppliers. Therefore, we anticipate that Japan manu-facturing dependence is negatively associated with sup-plier outdegree centrality, because firms with a large proportion of manufacturing plants in Japan will focus on selling their products to the wide range of Japanese automakers and customers outside the Toyota supply net-work. Consequently, as these instrumental variables are concerned with the orientation of the firm towards an internal customer base and its dependence on Japan as a manufacturing location they are unlikely to have a
signif-icant effect on the co-development of supplier–supplier
innovations across the supply network.
With supplier outdegree centrality as the dependent variable (Model 3), the first stage of the 2SLS regression indicates its positive association with internal outdegree dependence (B = 0.715; p < .000), whereas Japan manufacturing dependence is negatively related to it
(B =−0.127; p < .023). In the over-identifying restrictions
tests, neither the Sargan (p = .310) nor Bassman (p = .325) test is statistically significant at a 95% level, so we appear to have valid instrumental variables (Bellamy et al., 2014). The partial F statistic suggests that these instrumental var-iables are not weak either (Woolridge, 2016).
In the second stage, we regress supplier–supplier
innovations as the dependent variable on supplier out-degree centrality and the predicted values from the first-stage estimation (Bellamy et al., 2014). As we detail in Model 4 in Table 3, supplier outdegree centrality is
nega-tively related to supplier–supplier innovations
(B =−0.594; p < .056), at a 90% significance level. The
Durbin score (p = .278) and Wu–Hausman test (p = .294)
are both insignificant. These results suggest that the rela-tionship between supplier outdegree centrality and
supplier–supplier innovations is not unduly affected by
endogeneity issues.
6
|
D I S C U S S I O N
With this study, we contribute to supply network-enabled innovation literature by studying the phenomenon of
supplier–supplier innovation (Hong & Hartley, 2011;
Narasimhan & Narayanan, 2013). Network ties to upstream suppliers and downstream customers can have different implications for the co-development of
sup-plier–supplier innovations. Moreover, we find that the
positive effect of supplier indegree centrality on supplier–
supplier innovations can be moderated by how firms are structurally embedded within the supply network (Kim, 2014). While betweenness centrality does not exert any significant moderating effects, closeness cen-trality, and embeddedness in small-clusters seem to have a substituting rather than complementary effects on the relationship between supplier indegree centrality
and supplier–supplier innovations. Beyond the structural
characteristics of the supply network, the geographical locations of plants also matters, and the number of manufacturing plants a firm operates in Japan is found to strengthen the positive effect of supplier indegree
central-ity on supplier–supplier innovations, whereas it
accentu-ates the negative effect of supplier outdegree centrality. We summarize the main empirical findings in Table 4 and discuss them in more detail next.
As our findings reveal, supplier indegree centrality
