You are on page 1of 18

Journal of Operations Management 29 (2011) 194211

Contents lists available at ScienceDirect

Journal of Operations Management


journal homepage: www.elsevier.com/locate/jom

Structural investigation of supply networks: A social network analysis approach


Yusoon Kim a, , Thomas Y. Choi b , Tingting Yan b , Kevin Dooley b
a b

Department of Management, Marketing, and Logistics, College of Business Administration, Georgia Southern University, Statesboro, GA 30460, United States Department of Supply Chain Management, W. P. Carey School of Business, Arizona State University, Tempe, AZ 85287, United States

a r t i c l e

i n f o

a b s t r a c t
A system of interconnected buyers and suppliers is better modeled as a network than as a linear chain. In this paper we demonstrate how to use social network analysis to investigate the structural characteristics of supply networks. Our theoretical framework relates key social network analysis metrics to supply network constructs. We apply this framework to the three automotive supply networks reported in Choi and Hong (2002). Each of the supply networks is analyzed in terms of both materials ow and contractual relationships. We compare the social network analysis results with the case-based interpretations in Choi and Hong (2002) and conclude that our framework can both supplement and complement case-based analysis of supply networks. 2010 Elsevier B.V. All rights reserved.

Article history: Received 26 October 2009 Received in revised form 5 November 2010 Accepted 10 November 2010 Available online 18 November 2010 Keywords: Supply networks Supply chain management Second-tier suppliers Social network analysis Network structure Structural analysis Network indices

1. Introduction Supply chain management has focused on linear relationships of buyers and suppliers (Cox et al., 2006; Zhu and Sarkis, 2004). While a linear perspective may be useful for planning certain mechanical aspects of transactions between buyers and suppliers, it fails to capture the complexity needed to understand a rms strategy or behavior, as both depend on a larger supply network that the rm is embedded in (Choi and Kim, 2008). A rms supply network consists of ties to its immediate suppliers and customers, and ties between them and their immediate suppliers and customers, and so on (Cooper et al., 1997; Croxton et al., 2001). In the past decade there has been increased discussion of the benets of adopting a network perspective in supply chain management research (Choi et al., 2001; Lazzarini et al., 2001; Lee, 2004; Wilding, 1998). From a supply network perspective, the relative position of individual rms with respect to one another inuences both strategy and behavior (Borgatti and Li, 2009). In this context, it becomes imperative to study each rms role and importance as derived from its embedded position in the broader relationship structure (Borgatti and Li, 2009; DiMaggio and Louch, 1998). For example, Burkhardt and Brass (1990) and Ibarra (1993) claim that power and inuence derive from a rms structural position in its surrounding network. Others have linked network position to such issues as

Corresponding author. Tel.: +1 912 478 2465; fax: +1 912 478 2553. E-mail address: yusoonkim@georgiasouthern.edu (Y. Kim). 0272-6963/$ see front matter 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.jom.2010.11.001

innovation adoption (e.g., Burt, 1980; Ibarra, 1993), brokering (e.g., Pollock et al., 2004; Zaheer and Bell, 2005), and creating alliances (e.g., Gulati, 1999). To date, there have been few studies of real-life supply networks, due to the difculties in obtaining data. The studies of real networks that have been done have relied on qualitative methods to derive theoretical and practical insights (e.g., Harland et al., 2001; Jarillo and Stevenson, 1991). While qualitative interpretations have their merits, their validity is threatened by a researchers bounded rationality, which includes the difculty to conceptualize complex phenomena such as networks. Thus in this paper we propose to analyze the structural characteristics of supply networks using a formal, quantitative modeling approachsocial network analysis (Borgatti and Li, 2009; Grover and Malhotra, 2003; Harland et al., 1999). We will show how social network analysis can both supplement and compliment more traditional, qualitative interpretation methods when analyzing cases involving supply networks. Social network analysis (SNA) has recently gained acceptance among scholars for its potential to integrate the operations and supply management eld with other branches of management science (Autry and Grifs, 2008; Borgatti and Li, 2009; Carter et al., 2007). According to Borgatti and Li (2009), SNA concepts are particularly suitable for studying how patterns of inter-rm relationships in a supply network translate to competitive advantages through management of materials movement and diffusion of information. To date, SNA has not been applied in an empirical study of real supply networks; in fact there is a general paucity of SNA appli-

Y. Kim et al. / Journal of Operations Management 29 (2011) 194211

195

cations in operations and supply management, with only a few exceptions (e.g., Carter et al., 2007; Choi and Liker, 1995). This is largely because there is lack of conceptual clarication as to how the key SNA metrics (e.g. centrality) can be theoretically interpreted in the context of supply networks. Therefore in this study we link different SNA metrics at the node- or rm-level to specic roles in a supply network. We consider supply networks based on both materials ow and contractual relationships. The metrics yield six supply network related constructs: supply load, demand load, operational criticality, inuential scope, informational independence, and relational mediation. Different network-level SNA metrics are also linked to their implications for supply network performance. We apply our framework to real supply network data derived from three published case studies of automotive supply networks (Choi and Hong, 2002). In that study the authors created empirically three complete network maps of the center console assembly for Honda Accord, Acura CL/TL, and DaimlerChrysler Grand Cherokee. In the present paper, we convert the network data from Choi and Hong (2002) into matrix forms and analyze them using the software UCINET 6 (Borgatti et al., 2002). These quantitative results are then interpreted using our theoretical framework. Finally, we discuss our quantitative SNA results comparing to the qualitative ndings of Choi and Hong (2002) and consider the implications.

1991; Nishiguchi, 1994). Other studies focused on developing taxonomies of supply networks (Harland et al., 2001; Lamming et al., 2000; Samaddar et al., 2006). More recently, Borgatti and Li (2009) have highlighted the salience of SNA to study supply networks. In fact, there have been a few studies in the operations and supply management eld that used or promoted the use of SNA. Choi and Liker (1995) used SNA to investigate the implementation of continuous improvement activities in automotive supplier rms. Carter et al. (2007) provided an example of the application of SNA in a logistics context. Autry and Grifs (2008) applied the concept of social capital, framed as part of social network theory, to supply chain context. However, still lacking in such studies is a theoretical framework that relates social network theory to supply network dynamics and the comprehensive application of SNA to studying supply networks. In the following section, we provide a brief overview of SNA, focusing on the key metrics useful for investigating and explaining phenomena within supply networks. 2.2. Social network analysis (SNA) A network is made up of nodes and ties that connect these nodes. In a social network, the nodes (i.e., persons or rms) have agency in that they have an ability to make choices. With its computational foundation in graph theory (Cook et al., 1998; Kircherr, 1992; Li and Vitnyi, 1991), SNA analyzes the patterns of ties in a network. Naturally, SNA has been used to study community or friendship structure (Kumar et al., 2006; Wallman, 1984) and communication patterns (Koehly et al., 2003; Zack and McKenney, 1995). It has been adopted to explore the spreading of diseases (e.g., Klovdahl, 1985) and diffusion of innovation (e.g., Abrahamson and Rosenkopf, 1997; Valente, 1996). In organization studies and strategic management, scholars have used it to investigate corporate interlocking directorships (Robins and Alexander, 2004; Scott, 1986) and network effects on individual rms performance (e.g., Ahuja et al., 2009; Burkhardt and Brass, 1990; Gulati, 1999; Jensen, 2003; Rowley et al., 2005; Stam and Elfring, 2008; Uzzi, 1997). Operations and supply management scholars have also noted the methodological potential of SNA. For instance, Choi et al. (2001) stated that one could approach the study of supply networks from the social network perspective. Ellram et al. (2006) acknowledged social network theory as a useful tool to study inuence in supply chains. Carter et al. (2007) identied SNA as a key research method to advance the elds of logistics and supply chain management. More recently, Borgatti and Li (2009) and Ketchen and Hult (2007) echoed such sentiments. They have also recognized the difculty of collecting network-level data in supply networks but argued its imperativeness for operations and supply management to be integrated with other management disciplines. According to Borgatti and Li (2009), a more systematic adoption of SNA will be instrumental in exploring behavioral mechanisms of entire supply networks. A SNA approach allows us to better understand the operations of supply networks, both at the individual rm level and network levelhow important the individual rms are, given their positions in the network and how the network structure affects the individual rms and performance of the whole network. Social network scholars (Everett and Borgatti, 1999; Freeman, 1977, 1979; Krackhardt, 1990; Marsden, 2002) have developed a range of network metrics at the node- or network-level to characterize the dynamics inside a social network. 2.3. Key network metrics Network metrics can be calculated at two levelsthe node level and network level. Node-level metrics measure how an individual node is embedded in a network from that individual nodes

2. Literature review 2.1. Supply networks Supply networks consist of inter-connected rms that engage in procurement, use, and transformation of raw materials to provide goods and services (Lamming et al., 2000; Harland et al., 2001). The relatively recent incorporation of the term network into supply chain management research represents a pressing need to view supply chains as a network for rms to gain improved performance, operational efciencies, and ultimately sustainable competitiveness (Corbett et al., 1999; Dyer and Nobeoka, 2000; Kotabe et al., 2002). Therefore, it is increasingly important to analyze the network structure of supply relationships. In the operations and supply management eld, a complex system perspective has been used as a theoretical lens for describing supply networks. Wilding (1998) studied dynamic events in supply networks through what he referred to as supply chain complexity triangle (p. 599). Choi et al. (2001) conceptualized supply networks as a complex adaptive system (CAS). Surana et al. (2005) proposed how various complex systems concepts can be harnessed to model supply networks. Pathak et al. (2007) discussed the usefulness of CAS principles in identifying complex phenomena in supply networks. Others have examined supply networks from a strategic management perspective. Greve (2009), using supply networks in the maritime shipping industry, studied whether technology adoption is more rapid in centrally located network positions. Mills et al. (2004) suggested different strategic approaches to managing supply networks depending on whether a rm is facing upstream or downstream and whether it is seeking its long-term or short-term position in the supply network. Methodologically, simulation models have been used to study hypothetical supply networks (Kim, 2009; North and Macal, 2007; Pathak et al., 2007). Others have studied real-world supply networks using the case study approach (Jarillo and Stevenson, 1991; Nishiguchi, 1994; Choi and Hong, 2002). Scholars in the industrial marketing have developed descriptive models of supply networks (Ford, 1990; Hkansson, 1982, 1987; Hkansson and Snehota, 1995). Descriptive case studies in this genre illustrate how companies such as Benetton, Toyota, or Nissan attained competitive advantage through their supply networks (Jarillo and Stevenson,

196

Y. Kim et al. / Journal of Operations Management 29 (2011) 194211

perspective. In this study, we focus on three types of node-level metricsdegree, closeness, and betweenness centrality. Networklevel metrics compute how the overall network ties are organized from the perspective of an observer that has the birds eye view of the network. The network-level metrics we consider are network density, centralization, and complexity. 2.3.1. Node-level metrics Identifying the key actors in a social network is one of the primary uses of SNA (Tichy et al., 1979; Wasserman and Faust, 1994). The concept of centrality is fundamental to node-level network metrics (Borgatti and Everett, 2006; Borgatti and Li, 2009). Centrality reects the relative importance of individual nodes in a network. A nodes central position in a social network has a signicant impact on its and others behaviors and well-beings (Mizruchi, 1994). Centrality has been associated with social status (Bonacich, 1972; Freeman, 1979), power (Coleman, 1973), and prestige (Burt, 1982). There are different types of centrality metrics and they identify nodes that are important, in different aspects. Most prominent are degree centrality, closeness centrality, and betweenness centrality (Everett and Borgatti, 1999; Krackhardt, 1990; Marsden, 2002). Of these, the most straightforward is degree centrality. This concept builds on an observation that the more links a node has the more central it iswhen a node is connected to a large number of other nodes, the node has high degree centrality. Due to its greater connectedness with other nodes, a node with high degree centrality would necessarily be more visible in the network (Freeman, 1979; Marsden, 2002). Another centrality concept is closeness centrality. As the term suggests, this metric focuses on how close a node is to all the other nodes in the network beyond ones that it is directly connected to. A node is central if it can quickly reach all the others, and that is why closeness centrality includes indirect ties. This centrality is usually associated with nodes autonomy or independence in social networks (Freeman, 1979; Marsden, 2002)a node with high closeness centrality has more freedom from others inuence and higher capacity for independent actions. Such nodes become less reliant on other nodes. Betweenness centrality measures how often a node lies on the shortest path between all combinations of pairs of other nodes. The more a given node connects nodes that would otherwise be disconnected, the more central that node isother nodes are dependent on this node to reach out to the rest of the network. This metric focuses on the role of a node as an intermediary and posits that this dependence of others makes the node central in the network. As such, the betweenness centrality usually denotes a nodes potential control or inuence in the network (Marsden, 2002). A node with high betweenness centrality has a great capacity to facilitate or constrain interactions between other nodes (Freeman, 1979). 2.3.2. Network-level metrics SNA also yields metrics concerning the structure of the overall network, such as network density, network centralization, and network complexity. Network density refers to the number of total ties in a network relative to the number of potential ties. It is a measure of the overall connectedness of a network (Scott, 2000)a network in which all nodes are connected with all other nodes would give us a network density of one. Network centralization captures the extent to which the overall connectedness is organized around particular nodes in a network (Provan and Milward, 1995). Conceptually, network centralization can be viewed as an extension of the node-level centrality (Freeman, 1979)if a network had such a highly centralized structure that all connections go through few central nodes, then that network would be high on network centralization. The network

with highest possible centralization is one with a star structure, wherein a single node at the center is connected to all other nodes and these other nodes are not connected to each other. Likewise, the lowest centralization occurs when all nodes have the same number of connections to others. Network centralization and network density are complementary. Whereas centralization is concerned with the distribution of power or control across the network, density reects network cohesiveness. A network that has every node connected with everyone else would have a highest possible density (i.e., density of one). This network would be a highly cohesive network but would have a diffuse and distributed control structure. Network complexity is dened as the number of dependency relations within a network (Frenken, 2000, p. 260) and thus would depend on both the number of nodes in the network and the degree to which they are interlinked (Frenken, 2000; Kauffman, 1993). In the context of a supply network, complexity relates to the collective operational burden born by the members in the network (Choi and Krause, 2006). For instance, a large number of units in a system is likely to entail high coordination cost (Kim et al., 2006; Provan, 1983). Further, if these units are highly interdependent, then the collective operational burden would be high and thus more complex at the system level. Network complexity is related to network density and network centralization. First, more complex networks require higher operational burden (Lokam, 2003; Pudlk and Rdl, 1992, 1994). Second, network density is conceptually linked with network complexity because a denser network requires more effort to build and maintain (Marczyk, 2006). Finally, network centralization is associated with network complexity because the highest coordination costs require when every node is connected to all other nodes (i.e., a network with the least centralization) (Pudlk et al., 1988). 3. Conceptual framework for analyzing supply networks 3.1. Two types of supply network There are a number of different ways in which ties can be established between rms in the supply network. For example, a tie might be established between two rms if they were collaborating on a new product development or if they had overlapping board membership or belonged to the same trade organization. In this paper we focus on two types of ties that Choi and Hong (2002) collected data for in their study. Firms can be linked because of the delivery and receipt of materials, or they can be linked through a contractual relationship (Choi and Hong, 2002). In a tree-like structure of materials ow (Berry et al., 1994; Chopra and Sodhi, 2004; Hwarng et al., 2005), the network describes which supplier delivers to which customer. The other type of network is based on contractual relationships. Often, when a buying company wants to control the bill of materials, it engages in directed sourcing, wherein it establishes a contract with a second- or third-tier supplier and directs the top-tier supplier to receive materials from them (Choi and Krause, 2006; Chopra and Sodhi, 2004; Park and Hartley, 2002). In this context, materials ow occurs between two rms who do not have a contract and vice versa. These two types of supply networks, although based on the same set of nodes, can have different network structures and, therefore, different logics and implications (Borgatti and Li, 2009). 3.2. Supply network constructs 3.2.1. Firm-level constructs We now consider the key node-level SNA metrics and discuss how they can be used to interpret different roles in supply net-

Y. Kim et al. / Journal of Operations Management 29 (2011) 194211 Table 1 Node-level centrality metrics and their implications for supply networks. Network type Centrality metrics Supply network constructs Conceptual denitions Implication for central nodes Rolea Materials ow Indegree centrality Supply load The degree of difculty faced by a Integrator rm in managing incoming material ows from the upstream rms The degree of difculty faced by a Allocator rm in dealing with demands from the downstream rms The extent to which a rm impacts Pivot the nal assemblers operational performance in terms of product quality, coordination cost and overall lead-time. The extent to which a rm has an impact on operational decisions or strategic behavior of other rms in the supply network The extent to which a rm has freedom from the controlling actions of others in terms of accessing information in the supply network The extent to which a rm can intervene or has control over interactions among other rms in the supply network

197

Description To put together or transform different parts into a value-added product and ensure it functions well To distribute limited resources across multiple customers, focusing on scale economies To facilitate or control the ows of supply across the whole network

Key capabilities System integration Design/development Architectural innovation Process/manufacturing Quality management Component innovation Out-bound logistics Risk management In- and out-bound logistics Cross-functional integ. Contract management SRM/CRM

Outdegree centrality

Demand load

Betweenness centrality

Operational criticality

Contractual relationship

Degree centrality

Inuential scope

Closeness centrality

Informational independence

Coordinator To reconcile differences of network members and align their opinions with the greater supply network goals Navigator To explore, access, and collect various information with greater autonomy in the supply network

Information acquisition Strategic alignment with OEM

Betweenness centrality

Relational mediation

Broker

To mediate dealings between network members and turn them into its own advantage

Information processing Strategic alignment with OEM

Network role given high centrality.

works. Table 1 offers an overview of key centrality metrics, the corresponding supply network constructs, and their implications for network roles in the context of modeling supply networks. We propose this new framework for the interpretation of the SNA metrics in the supply network context. To illustrate these constructs, we rst discuss the calculation of key SNA metrics and the essential properties of each. Then, we integrate each key SNA metric separately with the two types of supply networks (i.e., materials ow and contractual relation). We should note that the supply network based on materials ow is directional, whereas the supply network based on contractual relationship is non-directional as legal obligations are mutually agreed and enacted. 3.2.1.1. Degree centrality in supply network. Degree centrality is measured by the number of direct ties to a node. Degree centrality CD (ni ) for node i(ni ) in a non-directional network is dened as: CD (ni ) =
j

other rms on their operations or decisions as the rm has more direct contacts with others (Cachon, 2003; Cachon and Lariviere, 2005; Ferguson et al., 2005). In contrast, nodes with low degree centrality are considered peripheral in the same network. If a node is completely isolated (i.e., zero degree), then removing this node from the network has virtually no effect on the network. Therefore, a rm who has more contractual ties in the network garners a broad range of inuence on others, and at the same time such a rm would often be required to reconcile conicting schedules or interests between others. For the nal assembler, for instance, it would make sense to align with suppliers with high degree centrality. In a directional network of materials ow, the focus is either on the ow initiated (out-degree) or ow received (in-degree). For instance, out-degree centrality of a node is dened as: CD (ni ) = xi+ . g1

xij =
j

xji

where xij is the binary variable equal to 1 if there is a link between ni and nj but equal to 0 otherwise (Freeman, 1979; Glanzer and Glaser, 1959; Nieminen, 1973; Proctor and Loomis, 1951; Shaw, 1954). To account for the impact of network size g, degree centrality is normalized as the proportion of nodes directly adjacent to ni : CD (ni ) = CD (ni ) . g1

For comparison purposes, in this study, we convert normalized degree centrality to a 0100 scale by multiplying by 100. A high degree centrality points to where the action is in a network (Wasserman and Faust, 1994, p. 179). Freeman (1979) describes it as reecting the amount of relational activities, and such activities make the nodes with high degree more visible. For instance, in a non-directional contractual relationship network, the degree centrality refers to the extent to which the rm inuences

In-degree centrality and out-degree centrality indicate the size of the adjacent upstream tier and downstream tier, respectively. A high in-degree or out-degree can capture transactional intensity or related risks for a rm (Powell et al., 1996). In a materials ow network, in-degree centrality for a rm can reect the degree of difculty faced by the rm when managing the incoming material ows. In other words, this metric measures the rms operational load coming from the upstream suppliers. A rm with high indegree centrality may serve the role of an integrator, as they are tasked with organizing and incorporating a range of parts from various suppliers to maintain the overall integrity of the product or service (Parker and Anderson, 2002; Violino and Caldwell, 1998). Such members in a supply network are instrumental and vital in carrying out the architectural or technical changes in the current product (Henderson and Clark, 1990; Iansiti, 2000). Out-degree centrality relates to the rms level of difculty in managing the needs of customers. The more direct customers there are in downstream, the more challenging it is for the rm to ensure on-time delivery, cost-effective inventory, and order

198

Y. Kim et al. / Journal of Operations Management 29 (2011) 194211

management for their customers. The number of direct customers is thus positively associated with the operational load related to demand integration and resource allocation (Frohlich and Westbrook, 2002). In a materials ow supply network, a rm with high out-degree centrality tends to be a common supplier to multiple downstream rms. Such supplier can economize and capitalize on its own internal resources as it aggregates demands from a range of customers (Nobeoka, 1996). Further, this rm is more likely than others to gain access to proprietary assets or information of its customer rms. This rm is in the best position to allocate or channel production or technical information to others in the network (Cassiman and Veugelers, 2002). 3.2.1.2. Closeness centrality in supply network. The calculation of closeness centrality is based on geodesic distance d(ni , nj ) the minimal length of a path between two nodes ni and nj (Hakimi, 1965; Sabidussi, 1966). In this study, closeness centrality is considered only in contractual relationship networks, as shown in Table 1. In a directed network (e.g., materials ow), the geodesic(s) from ni to nj may not be the same as the one(s) from nj to ni , or there can be two geodesics between two non-adjacent nodes. In the case of supply networks, this does not make physical sense. Therefore, typical node closeness is dened as:

ni s betweenness is then simply the sum of the probabilities that the node lies between other nodes. The betweenness reaches the maximum when ni falls on all geodesics and has a minimum of zero when ni falls on no geodesics. We normalize it to a value between 0 and 100: CB (ni ) = CB (ni ) 100. [(g 1)(g 2)/2]

CC (ni ) =
j=1

d(ni , nj )

where

nodes. At a maximum, the index equals (g 1)1 , which happens when the node is adjacent to all other nodes. When all the other nodes are not reachable from the node in question, the index reaches its minimum value of zero. The index can be normalized by multiplying CC (ni ) by g 1. The value then ranges between 0 and 1 regardless of network size (Beauchamp, 1965). In this study, the normalized index is converted to a 0100 scale. Nodes with high closeness need not much rely on others for relaying information or initiating communications (Bavelas, 1950; Beauchamp, 1965; Leavitt, 1951). This metric, in a supply network context, thus can represent the extent to which a rm can act autonomously and navigate freely across the network to access resources in a timely manner. Such a rm has comparatively shorter supply chains, both upstream and downstream. Shorter chains translate into less distortion of information and better ability to access reliable information (e.g., demand forecasts, supply disruption) in a timelier manner (Lee et al., 1997; Chen et al., 2000). Such accessibility to high-quality information increases the rms capability to match supply and demand (Cachon and Fisher, 2000), resulting in less inventory and lower operational costs (Lee et al., 2000). 3.2.1.3. Betweenness centrality in supply network. Betweenness centrality appears under both types of networks. A rm can lie between a pair of non-adjacent rms either along their materials ow or contractual relationship. The intermediary will have different effects on the rms it links, whether directionally or non-directionally. Measuring betweenness centrality begins with an assumption that a connection between two nodes, nj and nk , follows their geodesics. Therefore, betweenness centrality can be expressed as (Freeman, 1977): CB (ni ) =
j<k

g d(ni , nj ) j=1

is the total distance between ni and all other

The betweenness can be viewed as indicating how much gatekeeping ni does for the other nodes (Borgatti and Everett, 2006; Freeman, 1980; Spencer, 2003). Gatekeeping occurs because a node on geodesic can control the ows of materials or communication (Marsden, 2002). When applying to materials ow networks, rms with high betweenness act as a hub or pivot that transmits materials along the supply chains, and betweenness centrality relates to the extent to which a rm potentially affects the downstream rms daily operations (e.g., lead time) and eventually the performance (e.g., nal product quality) of the whole network. For instance, if a rm with high betweenness transmits materials to a wrong place or does not respond to changes in demand in a timely manner, it can easily lead to supply disruptions (Chopra and Sodhi, 2004). Similarly, the effects of poor-quality outputs from these rms can easily infect the broader supply network, interfering with normal product ows (Kleindorfer and Saad, 2005). Therefore, operational hiccups caused by such rms can surely hamper the functioning of the entire supply network (Hendricks and Singhal, 2005). Considering the signicance of negative impacts such members can have, it would be prudent of the nal assembler to ensure high or, at least, consistent operational performance of these rms (Hendricks and Singhal, 2003). In a contractual relationship network, the metric can denote the extent to which a rm can affect the interactions among others in the same supply network. A rm with high betweenness centrality mediates many pathways and thus can either facilitate or interfere with the network communications. The social network literature suggests that a node linking dense regions of relationships enjoys the benets of non-redundant information to increase its control over others (Burt, 1992, 1998). Supply network research also postulates that a buyer can enjoy the increased sourcing leverage when it lies between two disconnected, competiting suppliers (Choi and Wu, 2009; Wu and Choi, 2005). For instance, the buyer can play two rival suppliers off each other to drive down the purchasing price. 3.2.2. Network-level constructs We now discuss the key network-level metrics. Table 2 summarizes the theoretical interpretation of the metrics and their implications for network performance in the context of supply networks. 3.2.2.1. Supply network centralization. Recall that CD (ni ) is nodelevel degree centrality, and CD (n ) is its maximum value in the i network. Then, a general denition for network centralization is (Freeman, 1979): CD =
g [C (n ) CD (ni )] i=1 D . g [C (n ) CD (ni )] i=1 D

max

gjk (ni ) gjk

where gjk is the total number of geodesics linking the two nodes, and gjk (ni ) is the number of those geodesics that contain ni . The

Given g nodes in the network, the denominator reduces to (g 1)(g 2). The value of CD reaches the maximum value of 1 when one node is connected with all other g 1 nodes, and the others interact only with this node. Its minimum value of 0 occurs when all degree centrality values are equal. In supply networks, centralization can refer to how much power or control the core rms exercise over other network members (Choi and Hong, 2002). In this study, besides centralization based on degree, two other centralization

Y. Kim et al. / Journal of Operations Management 29 (2011) 194211 Table 2 Network-level metrics and their implications for supply networks. Network type Network-level metrics Conceptual denition in supply networks Implication of overall network structurea

199

Characteristics Materials ow Centralization The extent to which particular focal rms control and manage the movement of materials in a supply network Operational authority (e.g., power to make decisions on materials ow) concentrated in few central rms Centralized decision implementation process More rms engaged in the delivering and receiving of materials More steps required to move the materials along

Performance implications High level of controllability in production planning Low level of operational effectiveness at the network-level (i.e., more time taken to reach a decision and take actions on issues at a local level) Low level of operational efciency at the network level (i.e., longer lead time from the most upstream to the nal assembler or more parts for the same product function) High level of controllability in product design, product quality, and/or cost management Low level of responsiveness to or more time for resolution on issues occurring at a local level Low level of robustness or high degree of vulnerability to supply disruptions (i.e., more time to channel information and a higher likelihood of information distortion across a supply network)

Complexity

The amount of collective operational burden born by the member rms in a supply network

Contractual relationship

Centralization

The extent to which particular focal rms exercise bargaining power or relationship management control over other rms in a supply network The amount of load on the supply network as a whole that requires relationship coordination

Lack of interactions between central and peripheral rms in a supply network Decoupled relationships between rms at different tiers

Complexity

More rms involved in transferring information Active interactions at a local level Slow relaying communications from downstream to the nal assembler

Implications given high metric score.

indices are also usedones based on closeness and betweenness centrality. Further, there are other proxy measures of centralization used in this study. They are multiple indices of density that involve the core and periphery sub-groups in a network (see Table 6). When a network is partitioned into two clusters, a core cluster appears among nodes that are densely connected together and a periphery is formed among nodes that are more connected to core members than to each other (Borgatti and Everett, 2000; Luce and Perry, 1949). For instance, in Fig. 1, there are 19 rms in the core group (see Table 6) around Honda and CVT who appear at the center. The rest appears in the periphery.

incur a greater amount of coordination load, compared to a situation where there is only one top-tier rm. Therefore, a complex supply networks would be associated with large network size, large core size, high network density, high core density, high periphery density, high CTP density, and high PTC density. Note that in the contractual relation supply networks, the PTC and CTP densities are identical, since every link in the network is non-directional and the adjacency matrix representing this network is symmetric. 4. Research methodology 4.1. Data source Choi and Hong (2002) (hereafter, denoted as C&H) reported three supply networks from raw materials suppliers to a nal assembler involved in the production of an automobile center console assembly. The three product lines represented were Honda Accord, Acura CL/TL, and DaimlerChrysler (DCX) Grand Cherokee. Using an inductive case study approach, the authors derived propositions regarding the behavioral characteristics of supply networks. Table 3 provides a review of this particular work. In our analysis, we include all the rms in the supply network as identied in C&Hthey are direct suppliers and parts brokers, stretching from raw materials suppliers to the nal assembler. As indicated before, each supply network contains two different types of network informationone pertaining to materials ow and another based on contractual relationships. These two different types of network data yield a total of six supply networksthree based on directional materials ow and three based on nondirectional contractual relationships. 4.2. Data analysis The network information from C&H is converted into a binary adjacency matrix (Wasserman and Faust, 1994) that has rms representing both the rows and columns of the matrix. For instance, cell (i,j) would equal 1 if the rms i and j were linked either by materials ow or contractual relationship, and would be 0 otherwise. Supply networks may yield adjacency matrices that are

3.2.2.2. Supply network complexity. Supply network complexity refers to the load on the network as a whole that requires coordination (Choi and Hong, 2002). While the general state of the literature regarding the property of complexity at the network level is still emerging (Butts, 2001a,b; Everett, 1985; Freeman, 1983), we adopt the idea put forth by Kauffman (1993) and Frenken (2000). They propose that network complexity can be indicated by the number of nodes and degree of interdependency among nodes in a given network. Therefore, we use two types of SNA output metricssizetype and density-typeto represent the number of supply network members and the level of connectedness among them, respectively. The size-type outputs are shown in network size and core size, and the density-types include network density, core density, periphery density, core-to-periphery (CTP) density, and peripheryto-core (PTC) density. Network size relates to the average path length among nodes in the network (Ebel et al., 2002). More rms in a network translate into more steps and more time needed to complete the same task, whereby creating a higher likelihood of the supply being interrupted en route and higher collective burden born at the system level (Frenken, 2000). Likewise, between the two networks of identical size, more links imply a higher probability that the functioning of the individual nodes in the network is likely to be impeded by others, leading to a greater coordination load on the whole network (Choi and Krause, 2006). For instance, if an OEM has two top-tier suppliers, the rm would necessarily

200

Y. Kim et al. / Journal of Operations Management 29 (2011) 194211

Fig. 1. Materials ow network for Accord.

symmetric (i.e., non-directional) or asymmetric (i.e., directional), depending on the nature of the linkages. As noted earlier, a materials ow network is directional and thus asymmetric, while a contractual relationship network is non-directional and thus symmetric. Once generated, the adjacency matrices are imported into UCINET 6 and are used as inputs for network analysis (Borgatti et al., 2002). UCINET is a comprehensive software package for the analysis of social network data. It has been one of the most widely accepted SNA tools for conducting the structural analysis of interorganizational networks (e.g., Gulati, 1995, 1999; Human and Provan, 1997; Rowley et al., 2005; Ahuja et al., 2009). The program contains dozens of network analytic methods such as centrality measures, subgroup identication, role analysis, elementary graph analysis, and permutation-based statistical analysis. While performing SNA, UCINET can create network visualizations. A visualization of each of the six supply networks, also known as a sociogram, is shown in Figs. 16. 5. Results 5.1. Node-level results Tables 4 and 5 list key rms in the two types of supply networks. We identify key rms based on their centrality values. Tables 4 and 5 build on Table 1. The supply network constructs shown on the top row come from Table 1, and centrality computations are conducted on the corresponding centrality metrics shown in Table 1. As indicated below each table, there is a cut-off point for each supply network construct (e.g., 10 for in-degree, 6 for out-degree). The cut-off point is determined based on one rule: when there is

a noticeable drop-off in the score, the previous score constitutes the threshold. In all cases except one, there are multiple key rms. The exception is out-degree centrality for the materials ow type of DCXs supply network. Every node in the network, except for the OEM, has only one customer, showing the same value on outdegree centrality; consequently, there was no threshold value. In Tables 4 and 5, the number shown in parenthesis next to a rm name represents the centrality score. 5.2. Network-level results Tables 68 show SNA results at the network level. Tables 6 and 7 focus on centralization metrics, respectively, for directional materials ow and non-directional contractual relationships. Table 8 summarizes all complexity metrics for both types of networks. In Table 6, various network-level indicators are shown across three different supply networks. Beginning with network size and density, individual node-level centrality scores are averaged for each supply network. Then, the three network centralization scores are listed. Up to this point, all values reect network-level attributes. Below the network-level values, Table 6 lists values at the group level. It rst shows the size of core group and its density (see Section 3.2.2.1 on supply network centralization for a discussion on core and peripheral groups). It then moves on to listing other group-level measures. Addressing contractual relationship supply networks, Table 7 is constructed much the same way. Since this supply network type is non-directional, network measures shown on the left-side column are slightly different from those of Table 6, as discussed under Table 1. Also note that most of the network-level metrics are in normalized form, which allows us to compare them across the three different supply networks.

Y. Kim et al. / Journal of Operations Management 29 (2011) 194211 Table 3 Summary of case data from Choi and Hong (2002). Network measures Product type

201

Honda Accord Centralization Two rms, CVT and JFC, are top-tier suppliers to Honda Several second- or third-tier suppliers (e.g., Emhart, Garden State, and Miliken) directly selected by Honda Some third-tier suppliers directly selected by CVT, based on Hondas core supplier list Hondas penchant for centralized control when it comes to the product design and supplier selection

Acura CL/TL One top-tier supplier, Intek, a complete integrator of this supply network Honda engaging in directed sourcing at the second, third, and even fourth tiers Intek likewise engages in directed sourcing by selecting its own suppliers and even their suppliers suppliers, based on Hondas core supplier list Directed sourcing generally for high-priced or strategic items Hondas centralized control of the product design activities 76 entities in the network: 1 rst-tier, 20 second-tier, 28 third-tier, 17 fourth-tier, 9 fth-tier, and 1 sixth-tier suppliers The coupling between Honda and Intek based on their shared history may reduce the level of complexity The decoupling between Intek and JFC, a second-tier supplier of the critical subassembly, may further the complexity Hondas effort to centralize second-tier suppliers may increase complexity of the network as a whole

DaimlerChrysler (DCX) Grand Cherokee Textron as the sole top-tier supplier that integrates parts and subassemblies Textron-Farmington and Leon Plastics appear as two key second-tier suppliers Textron assumes the leading role in designing console Directed sourcing occurs only on a limited basis

Complexity

All together, 50 network entities: 2 rst-tier, 21 second-tier, 18 third-tier, 7 fourth-tier, and 2 fth-tier suppliers Majority of the suppliers at the second-tier level Four different nature of businesses in the networkmanufacturing companies, raw materials suppliers (e.g., GE Plastics), distribution centers (e.g., Iwata Bolt), and trading houses (e.g., Honda Trading) Reciprocal relationship between CVT and JFC, two top-tier suppliers, contributing to either reduction or increase of network complexity depending on the relational nature

41 entities: 2 rst-tier, 10 second-tier, 22 third-tier, and 7 fourth-tier suppliers At the top-tier level, Textron is engaged in assembly work and also acts as a conduit for a part from Leon as it ships the front console mat directly to the DCX plant with Textrons label No reciprocal relations among suppliers As per DCXs recommendation, Textron has consolidated the second-tier suppliers, leading to reduced number of second-tier suppliers and subsequently reduced complexity

Fig. 2. Materials ow network for Acura.

202

Y. Kim et al. / Journal of Operations Management 29 (2011) 194211

Fig. 3. Materials ow network for Grand Cherokee.

Fig. 4. Contractual relationship network for Accord.

Y. Kim et al. / Journal of Operations Management 29 (2011) 194211

203

Fig. 5. Contractual relationship network for Acura.

Fig. 6. Contractual relationship network for Grand Cherokee.

204 Table 4 List of key rms based on materials ow network. Supply loada Accord Acura DCX
a b c d

Y. Kim et al. / Journal of Operations Management 29 (2011) 194211

Demand loadb CVT (15), C&C (7.4), JFC (7.4), GE (7.4), Yamamoru (7.4), Industry Products (7.4) Iwata Bolt (9.1), Tobutsu (6.1), Arkay (6.1), Twist (6.1), Milliken (6.1), Garden State (6.1), Select Ind. (6.1) None

Operational criticalityc CVT (13), Emhart (2), Yamamoru (1.7), Fitzerald (1.7), JFC (1.3) Intek (3), Arkay (1.7) Textron (3.8), Leon Plastics (2.5)

CVT (59 ), JFC (15), HFI (11) Intek (58), Arkay (21), Select Ind. (12) Textron (65), Leon Plastics (31)

Firms with in-degree > 10. Firms with out-degree > 6. Firms with betweenness > 1.0. Centrality score.

Table 5 List of key rms based on contractual relationship network. Inuential scopea Accord Acura DCX
a b c

Informational independenceb CVT (57), Honda (53), Yamamoru (40) Intek (62), Honda (56), Arkay (44), Select Ind. (43), Tobutsu (41), HFI (40) Textron (72), Leon Plastics (58) Daimler (46)

Relational mediationc CVT (79), Honda (64), Emhart (21) Yamamoru (15), Fitzerald (14) Intek (77), Honda (63), Select Ind. (14), Iwata Bolt (12), Arkay (10) Textron (88), Leon Plastics (53) Daimler (15)

CVT (52), Honda (30), Yamamoru (15) Intek (45), Honda (36), Arkay (18), Select Ind. (15) Textron (62), Leon Plastics (35)

Firms with degree > 15. Firms with closeness > 40. Firms with betweenness > 10.

Table 6 Network-level results for materials owa supply networks. Network measures Product type Accord Network size (rms) Network density Average in-degree Average out-degree Average betweenness Centralization (in-degree) Centralization (out-degree) Centralization (betweenness) Core group size (rms) Core density Core to periphery (CTP) density Periphery to core (PTC) density Periphery density
a

6. Interpretation of results In this section, we recapitulate the SNA results shown in Tables 48 with reference to the supply network constructs developed in this study (see Tables 1 and 2). We provide network dynamics implications of the node-level results rst and then those of the network-level results. A summary of the SNA results at the node- and network-level is shown, respectively, in Tables 9 and 10. 6.1. Node-level implications 6.1.1. Key rms in the materials ow supply networks Table 4 compares groups of rms across supply load, demand load, and operational criticality (see Table 1 for denitions). CVT, a rst-tier supplier in Accord supply network, appears highly central, showing the highest scores on all three columns. In other words, CVT assumes the most operational burden on both the supply side and demand side. This rm is tasked with integrating multiple parts into a product, which also means the rm can make the most of its resources by pooling customer demands and the related risks. CVT is also the pivotal player in the movement of materials. Without this rm, the entire supply chain would be disrupted. In contrast, we observe that another top-tier supplier of Accord, JFC, is not as central. Its centrality scores are markedly lower than those of CVT, and there are other second- (i.e., C&C, Emhart, and Yamamoru) and third-tier suppliers (i.e., Fitzerald) who appear more central than JFC. This is because most suppliers supplying to JFC also serve CVT but not the other way around (see Fig. 1). Intek, the top-tier supplier for Acura, appears most central under both supply load and operational criticality. The bulk of network resources ow into and through this rm. However, unlike CVT in Accord network, Intek does not appear central under demand load. This is because the rm primarily receives materials (see Fig. 2). In fact, Iwata Bolt, a second-tier supplier for Acura, comes rst under demand load. This simply means that this rm delivers to a relatively large number of buying rms, which implies that this supplier has leverage in allocating its internal resources across multiple customers. Another noteworthy nding is that Arkay, a second-tier supplier, is the only rm that ranks high on all the three centrality metrics. Without conducting SNA, Arkays central role in the Acura supply network may very well be overlooked.

Acura 34 0.037 3.654 3.654 0.231 0.556 0.056 0.029 23 0.059 0.000 0.043 0.000

DCX 27 0.037 3.704 3.704 0.234 0.641 0.001 0.038 4 0.250 0.000 0.250 0.000

28 0.046 4.630 4.630 0.809 0.567 0.106 0.128 19 0.067 0.006 0.064 0.000

Represented by asymmetric matrix.

Table 8 re-organizes some information from Tables 6 and 7. It lists values for the select indicators of network complexitythey represent the degree of interdependency among rms. Network size is listed as the rst indicator. Network density is then listed in both materials ow and contractual relationships networks. Then, group-level indicators are listed in both types of supply networks.
Table 7 Network-level results for contractual relationshipa supply networks. Network measures Product type Accord Network size (rms) Network density Average degree Average closeness Average betweenness Centralization (degree) Centralization (closeness) Centralization (betweenness) Core group size (rms) Core density CTP or PTC density Periphery density
a

Acura 34 0.066 6.595 37.747 5.375 0.413 0.513 0.738 6 0.467 0.179 0.000

DCX 27 0.074 7.407 41.959 5.778 0.585 0.641 0.854 3 0.667 0.333 0.000

28 0.074 7.407 35.716 7.407 0.479 0.459 0.748 17 0.125 0.048 0.036

Represented by symmetric matrix.

Y. Kim et al. / Journal of Operations Management 29 (2011) 194211 Table 8 Key indicators for network complexity. Network size (rms) Materials ow network Contractual relationship network

205

Network density Accord Acura DCX 28 34 27 0.046 0.037 0.037

Core size 19 23 4

Core density 0.067 0.059 0.250

CTP density PTC density Network density 0.006 0.000 0.000 0.064 0.043 0.250 0.074 0.066 0.074

Core size 17 6 3

Core density 0.125 0.467 0.667

Periphery density 0.036 0.000 0.000

PTC density 0.048 0.179 0.333

There are a comparatively less number of central rms in DCXs supply network. The implication is that the structure of the DCX network is simpler (see Fig. 3) than those of Honda and Acura. For one, there are no rms listed under demand load. This is because every supplier in this network has only one customer, including Textron and Leon, a top-tier and a key second-tier supplier, respectively. These two suppliers appear under both supply load and operational criticality. Both rms engage in value-adding activities by integrating parts and facilitating their ows. The supply streams in this supply network take place primarily through Textron or Leon. 6.1.2. Key rms in contractual relationship supply networks In Table 5, CVT is again prominent on all centrality metrics in Accord supply network. This rm appears as most inuential on the operation of the contractual relationship supply network, just as it does in the materials ow network. Nonetheless, there are a few notable differences. First, Honda does not appear at all in Table 4, but in this network based on contractual relationships, Honda emerges quite visibly (second to CVT) on all three columns. This is because Honda maintains a contractual relationship with many of its second- and third-tier suppliers (see Fig. 4). Second, JFC, a top-tier supplier who appears in all three centrality metrics in Table 4, is gone in Table 5. In other words, when it comes to managing contracts, Honda emerges as central and JFC disappears. Clearly, JFC is more isolated in the contractual relationship network. For Acura supply network, Intek appears yet again as most central, while Honda emerges as central also. Thus, Intek looks like most inuential in the contractual relation network and none could bypass Intek to connect with Honda. The network position allows Intek to take control of information and communication ows. One supplier for Acura that appears in Table 5 but did not in Table 4 is HFI. HFI is a lone third-tier supplier that SNA picked up as being a key rm under Informational Independence. This is largely because
Table 9 Node-level overview. Materials ow network Accord CVT, a 1st-tier suppliers, is most central, and assumes the most operational burden on both supply and demand sides JFC, another 1st-tier supplier, is not as much central as CVT Honda appears not central in this network HFI and C&C, two 2nd-tier suppliers, need to handle high degrees of supply load and demand load, respectively Two 2nd-tier suppliers (Emhart and Yamamoru), and one 3rd-tier (Fitzerald) are also central as a go-between along the materials ow Acura Intek, the sole 1st-tier supplier, is most central under both supply load and operational criticality Arkay, a 2nd-tier supplier, is central on every centrality metric Iwata Bolt, a 2nd-tier supplier, is most central under demand load Honda is virtually out of sight in this network Textron, the sole 1st-tier supplier, and Leon, a 2nd-tier supplier, are most central under both supply load and operational criticality No central rm under demand load Daimler is rather central only under supply load

HFI does business with other central rms such as Intek and Arkay, and this is how it stays in the loop (see Fig. 5). Unlike Accord and Acura, the list of rms that appear in Table 5 for DCX shows little change. There were two rms (Textron and Leon) in Table 4 and the same rms appear again in Table 5. The only exception is Daimler. Compared to the materials ow network, the OEM is more prominent in the contractual relationship network (Table 5), and this is due to its direct links with two third-tier suppliers, Irwin and E.R. Wagner (see Fig. 6). Daimler thus has leverage over the relationships between these two suppliers and Textron, the top-tier supplier. 6.2. Network-level implications We now turn to discussing the dynamics at the network level. Characterization we make below pertains to the whole supply networks based on Tables 6 and 7. 6.2.1. Characteristics of the materials ow supply networks In Table 6, Accords supply network shows a comparatively high density compared to the other two networks of Acura and DCX. Accords supply network also features relatively high average scores on the key centrality metrics. Particularly, on average betweenness, Accords lead is substantial. It implies that rms in this supply network are more engaged in both delivering and receiving materials than rms in other supply networks. It also means that there are more steps required to move the materials along. From an operational standpoint, it might indicate that this network provides less efciency (e.g., longer lead time, more parts used for the same function) as it imposes more managerial attention on the rms in a central position. Looking at centralization scores for Accord, indegree score stands out, suggesting the inow of materials is concentrated in a small group of rms in the supply network. We also note a rather large discrepancy in the scores

Contractual relationship network CVT is most central under all three measuresoperational exibility, managerial independence, and relational control Honda emerges as the close second to CVT on all centrality metrics JFC is extinct and becomes isolated in this network Yamamoru, a 2nd-tier suppliers, emerges as central under managerial independence Emhart, a 2nd-tier supplier, is central under relational control

Intek is again most central on all three centrality Honda is the close second to Intek on every centrality metric HFI, a 3rd-tier supplier, emerges as key under managerial independence due to its ties with other key suppliers Two 2nd-tier suppliers, Arkay and Select Industries, rank consistently high on all three centrality metrics Little change from materials ow network Textron and Leon are two most central on every centrality metrics Daimler comes next but by a large margin on all three metrics No other rms, than the three rms, appear as central in this network

DCX

206 Table 10 Network-level overview. Materials ow network

Y. Kim et al. / Journal of Operations Management 29 (2011) 194211

Contractual relationship network Relatively high overall density Highest average betweenness but lowest average on closeness Largest core group with low density Relatively high periphery density Comparatively low PTC density There are more interactions overall among more members Rather complex at the network level Relatively less centralized Largest overall membership but with lowest overall density Lowest average betweenness score More tightly coupled core group No interactions among peripheral rms Network activities mostly concentrated around the core group Relatively high PTC density Comparatively more centralized around the smaller core group Comparatively less complex at the network level Highest average scores on closeness centrality Highest centralization indices Smallest core group with very high density No interactions among peripheral rms By far higher PTC density Majority of network activities centers around the core group Peripheral rms engage only in supplying to the core rms Most centralized and least complex at the network level

Accord Comparatively high overall density Highest average score on all three centrality metrics Relatively high centralization across all three types, and substantial lead on average betweenness score Much higher indegree centralization than the other two types No connectivity among peripheral rms Little reciprocity between the core and peripheral rms (much higher PTC density than CTP density) Peripheral rms engage solely in supplying to core rms Acura Comparatively large overall membership but with low density Relatively low average scores on all the three centrality metrics Comparatively low centralization indices Very large core group with very low density Virtually no materials ows among peripheral rms No reciprocity between the core and the periphery rms Network activities concentrated around the core group Comparatively less complex at the network level Smallest membership with relatively low density Comparatively high indegree centralization, but quite low outdegree centralization Smallest and tightly knit core group No materials ows among peripheral rms Largest discrepancy between PTC and CPT density among three SNs Peripheral rms engage exclusively in supplying to the core rms Relatively more centralized and least complex at the network level

DCX

of CTP density and PTC density, which signies little reciprocity between the core and peripheral rms. Further, the far-off rms do not interact at all, as demonstrated in the periphery density of 0, and this is true for all product types. In other words, the peripheral rms engage solely in supplying to the core rms. Acuras supply network has comparatively large membership but low overall density. The three average centrality scores are relatively low. Acuras supply network, compared to Accords, has less number of links and the overall steps required to get things done are not as many, which may indicate higher operational efciency. Further, based on centralization scores, Acuras supply network appears as less centralized than Accords. At the local level, the core group has very large membership but with relatively low density. Since there are more rms in the core group, it suggests that the power in the network is more spread out; the more at structure of operational authority again may be an indication that this network works more efciently (e.g., less time expended to make a decision on issues at a local level). DCXs supply network has the smallest membership, and the overall density is also relatively low. The centralization index based on in-degree is comparatively high, as is the case with Accord and Acura. However, notable difference occurs with out-degree centralization. It is quite low, indicating that most of the materials ow out to few common dominant rms, and this observation is also supported by the small size of the core group. There is also a huge discrepancy between CTP density and PTC density, which simply means that the majority of materials ow links in the network is concentrated on a small number of rms. As expected, these rms in the core group are tightly knit, as evidenced by a high core density. Such simple structure can provide high operational efciency at the network-level (e.g., shorter lead time from upstream suppliers to the nal assembler); however, if multiple issues were to happen simultaneously they could overwhelm the few central players and could require much more time for resolution. 6.2.2. Characteristics of contractual relationship supply networks The density for Accord is much higher in Table 7 than it was in Table 6. This is because contracts can jump across several tiers. As

expected, the same thing happens for Acura and DCX as well. In terms of centrality metrics, Accords supply network shows relatively low average closeness but high average betweenness scores. Such a structure may be less responsive or more susceptible to supply disruptions. It would possibly take more time channeling information and there is a higher chance that information becomes distorted on its way along the chains as more rms get involved in transferring it. Therefore, such structure is likely to be less robust or less effective when it comes to coping with supply disruptions. By the same token, the structure would provide greater complexity at the network level for Accord, as also evidenced by Accords relatively large core group size (see Table 7). Further, it has relatively high periphery density, which further indicates that the network is complex because there are more interactions going on even among peripheral members. Still, more contacts among members at the local level might facilitate identifying, if any, supply issues occurring locally. Acuras supply network shows relatively low overall density but with large membership, which correspond with less number of contractual links overall. Regarding average betweenness, this supply network shows the lowest score, indicating that this network needs a smaller number of channels to get things done. Comparatively, therefore, this supply network appears as more efcient, for instance, in managing such issues as supply disruptions because communications at the network level can be comparatively faster and more organized than those of Accords, which is also supported by Acuras comparatively more tightly knit core group and zero periphery density. Interestingly, DCXs supply network shows the highest average closeness score. In other words, the rms in the DCXs network are more readily reachable from each other, indicating that information can travel faster across the network. To put it differently, the network structure is more conducive to the centralized control by dominant actors. As might be expected, this supply network features the highest centralization indices among all three supply networks. There is additional evidence for DCXs high centralization at the network levelthe majority of the activities in the supply network seem to center around a very small group of rms (i.e.,

Y. Kim et al. / Journal of Operations Management 29 (2011) 194211

207

three core rms) that are highly interwoven together (i.e., the highest core density of 0.667). Further, the rms in the periphery, with no interactions among them, focus on catering to the core rms needs, evidenced by high PTC density. Because network information tends to spread relatively fast and converge at a small group of dominant actors, the network as a whole would be comparatively more effective and robust when it comes to dealing with supply disruptions. Particularly, active interactions between core and periphery rms would further enhance such capability of the supply network.

7. Discussion 7.1. Comparisons between SNA results and C&H study 7.1.1. Overlapping and divergent results One of the main ndings of C&H was the three OEMs varying degrees of centralized control over their supply networks. The SNA results conrm this. In particular, the nal assemblers practice of directed sourcing is captured in the contractual relationship network structure. For instance, the high values in Hondas various centralities and overall density in the contractual relationship network, compared to those in the materials ow network, is clearly attributable to the added links that represent Hondas directed sourcing practice involving its second- and third-tier suppliers. Another nding shared by both studies is the relational salience of those tertiary-level suppliers in the network that are sourced directly by OEMs. All of such suppliers (e.g., Emhart for Accord and Iwata Bolt for Acura) emerge as visible in the contractual relationship network, through their exhibiting high scores on the various centrality metrics or becoming a member of the core group in their respective supply networks. Divergent results between the two studies relate largely to network-level properties such as network centralization and complexity. First, C&H describe Hondas two supply networks as more centralized than DCXs. However, SNA suggests the opposite (see Tables 6 and 7). In evaluating network centralization, C&H actually take the perspective of the nal assemblers (i.e., Honda and DCX). They present the argument that Honda is more centralized compared to DCX because it has more direct ties with its suppliers (i.e., top-tier as well as second- and third-tier suppliers)Honda has more centralized control of its supply networks. However, SNA, in contrast, looks at how central all rms are in the supply network, not just the nal assembler. SNA evaluates the relative node-level centrality scores of all the network members to arrive at the indicators of network centralization. The two studies also diverge when considering which supply network is most complex. C&H suggest that Acuras network is most complex. This judgment is based on the network-level physical attributes (e.g., total number of entities, average geographical distance between companies) and qualitative evidence regarding the lack of shared history and the perceived level of decoupling among members. Contrarily, SNA points to Accords network as being most complex. This is because SNA focuses on how individual rms and their relationships are connected to one another at the network level. For instance, SNA considers various aspects of interdependence among members in the network, such as network density, core density, periphery density, and PTC density. The two studies, as such, draw different conclusions on some aspects of supply network properties. Nonetheless, we want to caution that this does not mean one is more accurate; rather, we want to say that they just focus on different aspects of the same phenomenonthe case approach focuses on contextual information, whereas SNA operates on numerical breakdown of data on relative positions of members.

7.1.2. What C&H offer but SNA does not C&Hs qualitative approach offers a contextually rich picture of network dynamics. For instance, they make statements about the network structure by drawing on such observations as Hondas strong penchant toward centralized policy with respect to supplier selection and product design and DCXs practice of delegating authority to the rst-tier supplier as to who will be second-tier suppliers and how to design the console. Further, the case method can provide more detailed accounts of how the supply networks operate and behave. For instance, in the Hondas supply networks, the second-tier suppliers selected directly by Honda tend to be less cooperative with the top-tier supplier, which contributes to furthering complexity at the network level; in the DCXs network, Daimler commissions the top-tier supplier to consolidate the second-tier suppliers to reduce operational complexity. Such ndings are context-specic and would be very difcult for SNA to capture. Also, C&H offer some propositions representing the overarching principles of the supply networks, derived from the qualitative data. For instance, the study observes, Formalized rules, norms, and policies lead to the varying degrees of centralization in the supply network . . . (p. 488); The cost consideration represents the most salient force that shapes the emergence of the supplynetwork structure (p.488); and A centralized approach to supply network involves a common list of core suppliers and the design activities are tightly controlled by the nal assembler (p. 489). Only from case-based qualitative studies could such propositions be compiled. SNA would be unable to capture such contextually rich information. 7.1.3. What SNA offers but C&H do not Simply, SNA offers many quantitative metrics that qualitative approaches cannot. By analyzing the structural characteristics of supply networks, SNA brings us new intriguing results that would likely be overlooked by qualitative methods. First, by producing various network metrics, from node- to group- and to networklevel, SNA facilitated a comprehensive analysis of supply networks. For instance, SNA evaluated differing roles of the individual nodes and their relative importance with respect to others in the same network (see Tables 4 and 5). Second, SNA allowed for a comparative analysis of two different network structuresmaterials ow and contractual relationship. Between the two different network structures, we have observed some divergent results even on the same network metrics (e.g., density, betweenness centrality). Those discrepancies, as noted earlier, come from the fact that the two structures are constructed based on different types of relational connection. Thus, it is not proper to say that one type of link is a more accurate depiction of a given network than the other; but rather the two different types of network information should be considered jointly to fully understand a supply network. Further, SNA enabled a group-level analysis by partitioning each supply network into two structurally distinct clusterscore and periphery sub-groups. The core-periphery analysis in fact facilitated assessing network-level properties across different supply networks (i.e., network centralization and network complexity). 7.2. Academic contributions Our goal in this paper has been to introduce SNA as a means to analyze the structure of supply networks and draw theoretical conclusions from such analysis. Our framework translates key SNA metrics into the context of supply networks, and discusses how roles of individual supply network members vary depending on their relative structural position in the network. Subsequently, we suggested a guideline as to how to identify central nodes and eval-

208

Y. Kim et al. / Journal of Operations Management 29 (2011) 194211

uate them differently. Central rms require possessing a particular set of capabilities corresponding to the roles they assume in the network (see Table 1). For instance, rms with high in-degree centrality should focus on developing a capability in system integration or product architectural innovation (Parker and Anderson, 2002; Violino and Caldwell, 1998); rms with high betweenness centrality may be in a better position to engage in supply risk management. Thus, it would be prudent for a buying company (e.g., OEMs), when selecting or developing a supplier, to consider these issues. We hope that the theoretical framework of this study would be instrumental in facilitating future supply network research adopting SNA approach. The papers methodological contribution is two-fold. First, this study demonstrates the value of SNA in studying supply networks. SNA considers all member rms in a given supply network to determine which rms are most important, in what aspect, to the operation of the whole network. Capitalizing on computating power, SNA can generate various analytic outputs reecting either individual- or group-level behavioral dynamics, which in fact facilitate gaining a more comprehensive and systematic view of network dynamics. Second, applying the widely accepted networklevel analytical concepts (i.e., network density, centralization, and core-periphery), SNA can complement qualitative methods in capturing the structural intricacy of the whole network in a more objective way. As has been demonstrated, SNA has considerable potential for enhancing our studies of supply networks (Borgatti and Li, 2009; Carter et al., 2007) and can effectively complement qualitative methods. 7.3. Managerial contributions Based on C&Hs data, our study brings to the fore the salience of two types of supply networksmaterials ow and contractual relationship. We propose that managers consider these two types for any given supply network, as we have demonstrated how the two networks organize and behave differently. For instance, in Acuras supply networks, the size of the core group becomes much larger when based on materials ow than the contractual relationship (see Tables 6 and 7 for comparison). Also, managers should note that there can be different sets of key rms between the two types of supply network (see Tables 4 and 5). One rm that does not appear as central in one type (e.g., HFI in the Acura network) may be a key player in another. Depending on which type of link to focus on, individual suppliers position of importance and the strategic roles will vary. For instance, the key rms in a materials ow network can have a considerable effect on the operational quality of overall supply network, affecting lead time, product quality, OEMs inventory level, or stockout costs (Bourland et al., 1996). Key suppliers in a contractual relation network could facilitate the timely indentication or resolution of those system-level operational problems and other supply disruption risks (Lee, 2002). Further, it may be prudent for a manufacturing rm to identify central second- or third-tier suppliers using SNA. Some of these suppliers become a key player by being linked to more visible other key rms in the supply network. In other words, some tertiarylevel suppliers emerge as important because they are vital to other more prominent suppliers in supply networks. We anticipate these second- and third-tier suppliers that previously went unnoticed will play a more signicant role in future. As the issue of supply chain scalability takes the center stage for safety and sustainability, large nal assemblers are moving toward identifying and managing key tertiary-level suppliers. Collecting complete supply network data and applying SNA, as we have done in this paper, may serve as a useful approach. In general, having a pictorial rendition of a supply network will be useful to managers. SNA can help generate network sociograms

(see Figs. 16). As a visual embodiment of relationship patterns in supply networks, these sociograms can be instrumental in attaining a realistic picture of networking patterns and the dynamics. Just as all graphs, network drawings can help save search efforts, facilitate recognition, and provide interesting new perspectives and insights into supply networks. Also, SNA provides a methodological frame for collecting and organizing data, which will be useful for planning and monitoring changes in the operation of supply networks. The position of a node in the network affects the opportunities and constraints of that node and of others (Gulati et al., 2000; Rowley, 1997). 7.4. Limitations and future directions Our study represents a very rst step in theorizing and empirically investigating supply networks using SNA concepts. We acknowledge that our study is limited in ways that suggest opportunities for future research. First, our analysis is conned to a specic automobile module (i.e., center console assembly). Any one supplier in the supply network might be involved in several overlapping supply networks across different product lines. A suppliers role based on one supply network will look quite different from that derived by considering the multiple supply networks together it is a member of. Therefore, the central roles a supplier plays in our analysis should be qualied to the single product line. It would not be reasonable to consider the results of our analysis as a general statement regarding that supplier. In a similar vein, supply networks are considered basically egocentriccentered around a focal actor (Hkansson and Ford, 2002; Mizruchi and Marquis, 2006). The three supply networks studied here were also mapped based on information obtained from the nal assemblers. Therefore, any possible effect each suppliers extended network can have on the rms strategic importance to the OEM could not be captured in our analysis. For instance, one second tier supplier to Honda may have a tie to other OEMs. If such extended ties were also counted, certain centrality metrics (e.g., betweenness) for the supplier might have shown different scores from those based on the egocentric network, whereby placing the supplier in a different strategic position with respect to Honda. Such egocentric network approach, albeit considered a reliable substitute for complete (sociocentric) network data (Marsden, 2002), may not be enough to provide a full understanding or potential of a given supplier, embedded within the larger social network (Mizruchi and Marquis, 2006). Third, in quantifying the inter-rm ties, we did not consider the variances in strength. All the links considered in our analysis were treated as having the same weight, while the link an OEM has with the rst-tier suppliers should involve more intensive information exchanges (i.e., kanban system) or a greater amount of materials (i.e., larger contract size) than those with the second-tier rms, for instance. Also, we viewed supply networks based on the materials ow and contract connections. However, certainly there are many other relational connection types that can be considered in supply networks, such as ownership, technology dependence, intellectual property, and risk sharing. Network ties could be represented by the number of joint programs or of shared patents, level of trust, or perceived transactional risks. Future studies therefore can incorporate the relative strength of supply ties using SNA as the method can effectively illustrate networks with weighted links (Borgatti and Li, 2009; Battini et al., 2007). Exchange ties involving a multi-level interface will have differential impact compared to other comparable supply ties based only on a single type of transaction. We note that most supply networks are considered a scalefree network, whose degree distribution closely follows a power law (Albert and Barabsi, 2002; Pathak et al., 2007). That is, most

Y. Kim et al. / Journal of Operations Management 29 (2011) 194211

209

nodes have very few links and only a small number of nodes (e.g., core rms) have many connections. Future studies may apply the scale-free network metrics to studying supply networks, such as clustering coefcient and characteristic path length. Clustering coefcient measures the degree to which nodes in a network tend to cluster together around a given node (Barabsi et al., 2002), and it can inform us of how suppliers behave with respect to the nal assembler at both the local and the global level. For instance, it can tell us how suppliers would come together for better coordination, based on some governance mechanism involving an OEM. Indicating the system-level closeness, characteristic path length can assist in evaluating whether a given supply network is optimally designed (Braha and Bar-Yam, 2004; Lovejoy and Loch, 2003). Given a supply network, it can be of considerable interest to know how the path length compares to the best or worst possible conguration for networks with the same number of nodes and lines. This can provide implications for how effectively the network is designed and how robust it can be to possible supply disruptions. Finally, SNA could be applied to advancing existing theories regarding the structure or topology of supply networks. A range of SNA metrics can serve as a useful means in this effort. Such network variables as density and various centralities could be applicable to characterizing typological archetypes of supply network structures, eventually leading to the development of a portfolio of contingent approaches to supply management. In conclusion, we hope that this paper can serve as a call to other operations and supply management researchers regarding the importance of framing supply chains as networks and continuing to develop useful supply network indices. We hope to see more researchers taking advantage of the usefulness of SNA for untangling and understanding the complex phenomena embedded in supply networks. References
Abrahamson, E., Rosenkopf, L., 1997. Social network effects on the extent of innovation diffusion: a computer simulation. Organization Science 8 (3), 289309. Ahuja, G., Polidoro Jr., F., Mitchell, W., 2009. Structural homophily or social asymmetry? The formation of alliances by poorly embedded rms. Strategic Management Journal 30 (9), 941958. Albert, R., Barabsi, A.L., 2002. Statistical mechanics of complex networks. Reviews of Modern Physics 74, 4797. Autry, C.W., Grifs, S.E., 2008. Supply chain capital: the impact of structural and relational linkages on rm execution and innovation. Journal of Business Logistics 29 (1), 157174. Barabsi, A.L., Jeong, H., Nda, Z., Ravasz, E., Schubert, A., Vicsek, T., 2002. Evolution of the social network of scientic collaborations. Physica A: Statistical Mechanics and its Applications 311 (34), 590614. Battini, D., Persona, A., Allesina, S., 2007. Towards a use of network analysis: quantifying the complexity of Supply Chain Networks. International Journal of Electronic Customer Relationship Management 1 (1), 7590. Bavelas, A., 1950. Communication patterns in task-oriented groups. Journal of the Acoustical Society of America 22, 271282. Beauchamp, M.A., 1965. An improved index of centrality. Behavioral Science 10, 161163. Berry, D., Towill, D.R., Wadsley, N., 1994. Supply chain management in the electronics product industry. International Journal of Physical Distribution and Logistics Management 24 (10), 2032. Bonacich, P., 1972. Factoring and weighting approaches to status scores and clique identication. Journal of Mathematical Sociology 2, 113120. Borgatti, S.P., Everett, M.G., 2000. Models of core/periphery structures. Social networks 21 (4), 375395. Borgatti, S.P., Everett, M.G., 2006. A graph-theoretic perspective on centrality. Social Networks 28 (4), 466484. Borgatti, S., Everett, M., Freeman, L., 2002. UCINET 6 for Windows: Software for Social Network Analysis. Analytic Technologies, Inc., Natick, MA. Borgatti, S.P., Li, X., 2009. On social network analysis in a supply chain context. Journal of Supply Chain Management 45 (2), 521. Bourland, K.E., Powell, S.G., Pyke, D.F., 1996. Exploiting timely demand information to reduce inventories. European Journal of Operational Research 92 (2), 239253. Braha, D., Bar-Yam, Y., 2004. Information ow structure in large-scale product development organizational networks. Journal of Information Technology 19, 244253.

Burkhardt, M.E., Brass, D.J., 1990. Changing patterns or patterns of change: the effects of a change in technology on social network structure and power. Administrative Science Quarterly 35, 104127. Burt, R.S., 1980. Innovation as a structural interest: rethinking the impact of network position on innovation adoption. Social Networks 2 (4), 327355. Burt, R.S., 1982. Toward a Structural Theory of Action. Academic Press, New York. Burt, R.S., 1992. Structural Holes: The Social Structure of Competition. Harvard University Press, Cambridge. Burt, R.S., 1998. The gender of social capital. Rationality and Society 10 (1), 546. Butts, C.T., 2001a. An axiomatic approach to network complexity. Journal of Mathematical Sociology. Butts, C.T., 2001b. The complexity of social networks: theoretical and empirical ndings. Social Networks 23 (1), 3172. Cachon, G.P., 2003. Supply chain coordination with contracts. In: de Kok, A.G., Graves, S.C. (Eds.), Handbooks in Operations Research and Management Science, vol. 11. Elsevier, Boston, pp. 229340. Cachon, G., Fisher, M., 2000. Supply chain inventory management and the value of shared information. Management Science 46 (8), 10321048. Cachon, G.P., Lariviere, M.A., 2005. Supply chain coordination with revenue-sharing contracts: strengths and limitations. Management Science 51 (1), 3044. Carter, C.R., Ellram, L.M., Tate, W., 2007. The use of social network analysis in logistics research. Journal of Business Logistics 28 (1), 137169. Cassiman, B., Veugelers, R., 2002. Complementarity in the innovation strategy: Internal R&D, external technology acquisition and cooperation. Discussion paper series No. 3284. Centre for Economic Policy Research, Cambridge. Chen, F., Dresner, Z., Ryan, J.K., Simchi-Levi, D., 2000. Quantifying the bullwhip effect in a simple supply chain: the impact of forecasting, lead times, and information. Management Science 46 (3), 436443. Choi, T.Y., Dooley, K.J., Rungtusanatham, M., 2001. Supply networks and complex adaptive systems: control versus emergence. Journal of Operations Management 19 (3), 351366. Choi, T.Y., Hong, Y., 2002. Unveiling the structure of supply networks: case studies in Honda, Acura, and DaimlerChrysler. Journal of Operations Management 20 (5), 469493. Choi, T.Y., Liker, J.K., 1995. Bringing Japanese continuous improvement approaches to U.S. manufacturing: the roles of process orientation and communications. Decision Sciences 26 (5), 589620. Choi, T.Y., Krause, D.R., 2006. The supply base and its complexity: implications for transaction costs, risks, responsiveness, and innovation. Journal of Operations Management 24 (5), 637652. Choi, T.Y., Kim, Y., 2008. Structural embeddedness and supplier management: a network perspective. Journal of Supply Chain Management 44 (4), 513. Choi, T.Y., Wu, Z., 2009. Triads in supply networks: theorizing buyersuppliersupplier relationships. Journal of Supply Chain Management 45 (1), 825. Chopra, S., Sodhi, M.S., 2004. Managing risk to avoid supply-chain breakdown. MIT Sloan Management Review 46 (1), 5362. Coleman, J.S., 1973. The Mathematics of Collective Action. Aldine, Chicago. Cook, W.J., Cunningham, W.H., Pullyblank, W.R., Schrijver, A., 1998. Combinatorial Optimization. Wiley, New York. Cooper, M.C., Lambert, D.M., Pagh, J.D., 1997. Supply chain management: more than a new name for logistics. The International Journal of Logistics Management 8 (1), 114. Corbett, C.J., Blackburn, J.D., Van Wassenhove, L.N., 1999. Partnerships to improve supply chains. Sloan Management Review 40 (4), 7182. Cox, A., Sanderson, J., Watson, G., 2006. Supply chains and power regimes: toward an analytic framework for managing extended networks of buyer and supplier relationships. Journal of Supply Chain Management 37 (2), 2835. Croxton, K.L., Garca-Dastugue, S.J., Lambert, D.M., Rogers, D.S., 2001. The supply chain management processes. The International Journal of Logistics Management 12 (2), 1336. DiMaggio, P., Louch, H., 1998. Socially embedded consumer transactions: for what kinds of purchases do people most often use networks? American Sociological Review 63, 619637. Dyer, J.H., Nobeoka, K., 2000. Creating and managing a high-performance knowledge-sharing network: the Toyota case. Strategic Management Journal 21 (3), 345367. Ebel, H., Davidsen, J., Bornholdt, S., 2002. Dynamics of social networks. Complexity (Special Issue: Complex Adaptive Systems: Part I) 8 (2), 2427. Ellram, L.M., Tate, W.L., Carter, C.R., 2006. Product-process-supply chain: an integrative approach to three-dimensional concurrent engineering. International Journal of Physical Distribution & Logistics Management 37 (4), 305330. Everett, M.G., 1985. Role similarity and complexity in social networks. Social Networks 7, 353359. Everett, M.G., Borgatti, S.P., 1999. The centrality of groups and classes. Journal of Mathematical Sociology 23 (3), 181201. Ferguson, R.J., Paulin, M., Bergeron, J., 2005. Contractual governance, relational governance, and the performance of interrm service exchanges: the inuence of boundary-spanner closeness. Journal of the Academy of Marketing Science 33 (2), 217234. Ford, I.D., 1990. Understanding Business Markets: Interactions, Relationships and Networks (Ed.). Academic Press, San Diego, CA. Freeman, L.C., 1977. A set of measures of centrality based on betweenness. Sociometry 40, 3541. Freeman, L.C., 1979. Centrality in social networks: conceptual clarication. Social Networks 1, 215239.

210

Y. Kim et al. / Journal of Operations Management 29 (2011) 194211 the US and Japanese automotive industries. Strategic Management Journal 24 (4), 293316. Krackhardt, D., 1990. Assessing the political landscape: structure, cognition, and power in organizations. Administrative Science Quarterly 92, 142154. Kumar, R., Novak, J., Tomkins, A., 2006. Structure and evolution of online social networks. In: Proceedings of the 12th ACM SIGKDD International Conference , Philadelphia, PA, USA, pp. 611617. Lamming, R.C., Johnsen, T.E., Zheng, J., Harland, C.M., 2000. An initial classication of supply networks. International Journal of Operations & Production Management 20 (6), 675691. Lazzarini, S.G., Chaddad, F.R., Cook, M.L., 2001. Integrating supply chain and network analyses: the study of netchains. Journal on Chain and Network Science 1 (1), 722. Leavitt, H.J., 1951. Some effects of communication patterns on group performance. Journal of Abnormal and Social Psychology 46, 3850. Lee, H.L., 2002. Aligning supply chain strategies with product uncertainties. California Management Review 44 (3), 104119. Lee, H.L., 2004. The triplea supply chain. Harvard Business Review 83, 102112. Lee, H.L., Padmanabhan, V., Whang, S., 1997. Information distortion in a supply chain: the bullwhip effect. Management Science 43 (4), 546558. Lee, H.L., So, R., Tang, C.S., 2000. The value of information in a two-level supply chain. Management Science 46 (5), 626643. Li, M., Vitnyi, P.M.B., 1991. Combinatorics and Kolmogorov complexity. In: Proceedings of the Sixth Annual IEEE Conference on Structure in Complexity Theory. Springer Verlag, Berlin, New York. Lokam, V.S., 2003. Graph complexity and slice functions. Theory of Computing Systems 36, 7188. Lovejoy, W.S., Loch, C.H., 2003. Minimal and maximal characteristic path lengths in connected sociomatrices. Social Networks 25, 333347. Luce, R.D., Perry, A.D., 1949. A method of matrix analysis of group structure. Psychometrika 14, 95116. Marczyk, J., 2006. Measuring and tracking complexity. In: Proceedings of the Sixth International Conference on Complex Systems , June 2530. New England Complex Systems Institute, Boston, MA, USA. Marsden, P.V., 2002. Egocentric and sociocentric measures of network centrality. Social Networks 24 (4), 407422. Mills, J., Schmitz, J., Frizelle, G., 2004. A strategic review of supply networks. International Journal of Operations & Production Management 24 (9/10), 10121036. Mizruchi, S.M., 1994. Social network analysis: recent achievements and current controversies. Acta Sociologica 37 (4), 329343. Mizruchi, S.M., Marquis, C., 2006. Egocentric, sociocentric, or dyadic? Identifying the appropriate level of analysis in the study of organizational networks. Social Networks 28, 187208. Nieminen, J., 1973. On the centrality in a directed graph. Social Science Research 2, 371378. Nishiguchi, T., 1994. Strategic Industrial Sourcing. Oxford University Press, Oxford/New York. Nobeoka, K., 1996. Alternative component sourcing strategies within the manufacturer-supplier network: benets of quasi-market strategy in the Japanese automobile industry. Kobe Economic and Business Review 41, 6999. North, M.J., Macal, C.M., 2007. Managing Business Complexity: Discovering Strategic Solutions with Agent-based Modeling and Simulation. Oxford University Press, New York. Park, S., Hartley, J.L., 2002. Exploring the effect of supplier management on performance in the Korean automotive supply chain. Journal of Supply Chain Management 38 (2), 4653. Parker, G.G., Anderson, E.G., 2002. From buyer to integrator: the transformation of the supply-chain manager in the vertically disintegrating rm. Production and Operations Management 11 (1), 7591. Pathak, S.D., Day, J.M., Nair, A., Sawaya, W.J., Kristal, M.M., 2007. Complexity and adaptivity in supply networks: building supply network theory using a complex adaptive systems perspective. Decision Sciences 38 (4), 547580. Pollock, T.G., Porac, J.F., Wade, J.B., 2004. Constructing deal networks: brokers as network architects in the US IPO market and other examples. The Academy of Management Review 29 (1), 5072. Powell, W.W., Koput, K.W., Smith-Doerr, L., 1996. Interorganizational collaboration and the locus of innovation: networks of learning in biotechnology. Administrative Science Quarterly 41 (1), 116145. Proctor, C.H., Loomis, C.P., 1951. Analysis of sociometric data. In: Jahoda, M., Deutsch, M., Cook, S.W. (Eds.), Research Methods in Social Relations. Dryden Press, New York, pp. 561586. Provan, K.G., 1983. The federation as an interorganizational network. Academy of Management Review 8 (1), 7989. Provan, K.G., Milward, H.B., 1995. A preliminary theory of network effectiveness: a comparative study of four community mental health systems. Administrative Science Quarterly 40 (1), 133. Pudlk, P., Rdl, V., 1992. A combinatorial approach to complexity. Combinatorica 14, 221226. Pudlk, P., Rdl, V., 1994. Some combinatorial-algebraic problems from complexity theory. Discrete Mathematics 136, 253279. P., 1988. Graph complexity. Acta Informatica 25, Pudlk, P., Rdl, V., Savicky, 515535. Robins, G., Alexander, M., 2004. Small worlds among interlocking directors: network structure and distance in bipartite graphs. Computational & Mathematical Organization Theory 10, 6994.

Freeman, L.C., 1980. The gatekeeper, pair-dependency and structural centrality. Quality and Quantity 14 (4), 585592. Freeman, L.C., 1983. Spheres, cubes, and boxes: graph dimensionality and network structure. Social Networks 5, 139156. Frenken, K., 2000. A complexity approach to innovation networks. The case of the aircraft industry (19091997). Research Policy 29 (2), 257272. Frohlich, M.T., Westbrook, R., 2002. Demand chain management in manufacturing and services: web-based integration, drivers and performance. Journal of Operations Management 20, 729745. Glanzer, M., Glaser, R., 1959. Techniques for the study of group structure and behavior. I. Analysis of structure. Psychological Bulletin 56, 317331. Greve, H.R., 2009. Bigger and safer: the diffusion of competitive advantage. Strategic Management Journal 30 (1), 123. Grover, V., Malhotra, M.D., 2003. Transaction cost framework in operations and supply chain management research: theory and measurement. Journal of Operations Management 21 (4), 457473. Gulati, R., 1995. Social structure and alliance formation patterns: a longitudinal analysis. Administrative Science Quarterly 40, 619652. Gulati, R., 1999. Network location and learning: the inuence of network resources and rm capabilities on alliance formation. Strategic Management Journal 20 (5), 397420. Gulati, R., Nohria, N., Zaheer, A., 2000. Strategic networks. Strategic Management Journal 21 (special issue), 203215. Hkansson, H. (Ed.), 1982. International Marketing and Purchasing of Industrial Goods: An Interaction Approach. Wiley, Chichester. Hkansson, H. (Ed.), 1987. Industrial Technological Development: A Network Approach. Croom Helm, London, UK. Hkansson, H., Ford, D., 2002. How should companies interact in business networks? Journal of Business Research 55 (2), 133139. Hkansson, H., Snehota, I., 1995. Developing Relationships in Business Networks. Routledge, London, England. Hakimi, S.L., 1965. Optimum locations of switching centers and the absolute centers and medians of a graph. Operations Research 12, 450459. Harland, C.M., Lamming, R.C., Cousins, P.D., 1999. Developing the concept of supply strategy. International Journal of Operations & Production Management 19 (7), 650673. Harland, C.M., Lamming, R.C., Zheng, J., Johnsen, T.E., 2001. A taxonomy of supply networks. The Journal of Supply Chain Management 37 (4), 2127. Henderson, R., Clark, K., 1990. Architectural Innovation: the reconguration of existing product technologies and the failure of established rms. Administrative Science Quarterly 35, 930. Hendricks, K.B., Singhal, V.R., 2003. The effect of supply chain glitches on shareholder value. Journal of Operations Management 21 (5), 501522. Hendricks, K.B., Singhal, V.R., 2005. An empirical analysis of the effect of supply chain disruptions on long-run stock price performance and equity risk of the rm. Production and Operations Management 14 (1), 3552. Human, S., Provan, K., 1997. An emergent theory of structure and outcomes in smallrm strategic manufacturing networks. Academy of Management Journal 40 (2), 369403. Hwarng, H.B., Chong, C.S.P., Xie, N., Burgess, T.F., 2005. Modelling a complex supply chain: understanding the effect of simplied assumptions. International Journal of Production Research 43 (13), 28292872. Iansiti, M., 2000. How the incumbent can win: managing technological transitions in the semiconductor industry. Management Science 46 (2), 169185. Ibarra, H., 1993. Network centrality, power, and innovation involvement: determinants of technical and administrative roles. Academy of Management Journal 36, 471501. Jarillo, J.C., Stevenson, H.H., 1991. Co-operative strategies: the payoffs and the pitfalls. Long Range Planning 24 (1), 6470. Jensen, M., 2003. The role of network resources in market entry: commercial banks entry into investment banking, 19911997. Administrative Science Quarterly 48 (3), 466497. Kauffman, S.A., 1993. The Origins of Order: Self-Organization and Selection in Evolution. Oxford University Press, New York. Ketchen Jr., D.J., Hult, G.T.M., 2007. Bridging organization theory and supply chain management: the case of best value supply chains. Journal of Operations Management 25 (2), 573580. Kim, W.S., 2009. Effects of a trust mechanism on complex adaptive supply networks: an agent-based social simulation study. Journal of Articial Societies and Social Simulation 12 (3), 4. Kim, T.Y., Oh, H., Swaminathan, A., 2006. Framing interorganizational network change: a network inertia perspective. Academy of Management Review 31, 704720. Kircherr, W., 1992. Kolmogorov complexity and random graphs. Information Processing Letters 41 (3). Kleindorfer, P.R., Saad, G.H., 2005. Disruption risk management in supply chains. Production and Operations Management 14 (1), 5368. Klovdahl, A.S., 1985. Social networks and the spread of infectious diseases: the AIDS example. Social Science & Medicine 21 (11), 12031216. Koehly, M.L., Peterson, K.S., Watts, G.B., Kempf, Kri, K.G.K., Vernon, W.S., Gritz, R.E., 2003. A social network analysis of communication about hereditary nonpolyposis colorectal cancer genetic testing and family functioning. Cancer Epidemiology, Biomarkers & Prevention 12, 304313. Kotabe, M., Martin, X., Domoto, H., 2002. Gaining from vertical partnerships: knowledge transfer, relationship duration and supplier performance improvement in

Y. Kim et al. / Journal of Operations Management 29 (2011) 194211 Rowley, T.J., 1997. Moving beyond dyadic ties: a network theory of stakeholder inuences. The Academy of Management Review 22 (4), 887910. Rowley, T.J., Baum, J.A.C., Greve, H.R., Rao, H., Shipilov, A.V., 2005. Time to break up: social and instrumental antecedents of rm exits from exchange cliques. Academy of Management Journal 48 (3), 499520. Sabidussi, G., 1966. The centrality index of a graph. Psychometrika 31, 581603. Samaddar, S., Nargundkar, S., Daley, M., 2006. Inter-organizational information sharing: the role of supply network conguration and partner goal congruence. European Journal of Operational Research 174 (2), 744765. Scott, J.P., 1986. Capitalist Property and Financial Power. Brighton, Wheatsheaf. Scott, J., 2000. Social Network Analysis: A Handbook. Sage Publications, London. Shaw, M.E., 1954. Group structure and the behavior of individuals in small groups. Journal of Psychology 38, 139149. Spencer, J.W., 2003. Global gatekeeping, representation, and network structure: a longitudinal analysis of regional and global knowledge-diffusion networks. Journal of International Business Studies 34 (5), 428442. Stam, W., Elfring, T., 2008. Entrepreneurial orientation and new venture performance: the moderating role of intra- and extra-industry social capital. The Academy of Management Journal 51 (1), 97111. Surana, A., Kumara, S., Greaves, M., Raghavan, U.N., 2005. Supply chain network: a complex adaptive systems perspective. International Journal of Production Research 43, 42354265. Tichy, N.M., Tushman, M.L., Fombrun, C., 1979. Social network analysis for organizations. Academy of Management Review 4 (4), 507519.

211

Uzzi, B., 1997. Social structure and competition in inter-rm networks: the paradox of embeddedness. Administrative Science Quarterly 42, 3567. Valente, T.W., 1996. Social network thresholds in the diffusion of innovations. Social Networks 18 (1), 6989. Violino, B., Caldwell, B., 1998. Analyzing the integrators. InformationWeek 709 (November), 4569. Wallman, S., 1984. Eight London Households. Tavistock, London. Wasserman, S., Faust, K., 1994. Social Network Analysis: Methods and Applications. Cambridge University Press, Cambridge. Wilding, R., 1998. The supply chain complexity triangle: uncertainty generation in the supply chain. International Journal of Physical Distribution & Logistics 28 (8), 599616. Wu, Z., Choi, T.Y., 2005. Suppliersupplier relationships in the buyersupplier triad: building theories from eight case studies. Journal of Operations Management 24 (1), 2752. Zack, M.H., McKenney, J.L., 1995. Social context and interaction in ongoing computer-supported management groups. Organization Science 6 (4), 394422. Zaheer, A., Bell, G.G., 2005. Beneting from network position: rm capabilities, structural holes, and performance. Strategic Management Journal 26 (9), 809825. Zhu, Q., Sarkis, J., 2004. Relationships between operational practices and performance among early adopters of green supply chain management practices in Chinese manufacturing enterprises. Journal of Operations Management 22 (3), 265289.

You might also like