How Firm Strategies Influence the Architecture of Transaction
Transcrição
How Firm Strategies Influence the Architecture of Transaction
How Firm Strategies Influence the Architecture of Transaction Networks Jianxi Luo Daniel E. Whitney Carliss Y. Baldwin Christopher L. Magee Working Paper 11-076 Copyright © 2011 by Jianxi Luo, Daniel E. Whitney, Carliss Y. Baldwin, and Christopher L. Magee Working papers are in draft form. This working paper is distributed for purposes of comment and discussion only. It may not be reproduced without permission of the copyright holder. Copies of working papers are available from the author. How Firm Strategies Influence the Architecture of Transaction Networks Revision and Resubmission to Industrial and Corporate Change Jianxi Luo* Massachusetts Institute of Technology [email protected] Daniel E. Whitney Massachusetts Institute of Technology [email protected] Carliss Y. Baldwin Harvard Business School [email protected] Christopher L. Magee Massachusetts Institute of Technology [email protected] * Corresponding author January 27, 2011 How Firm Strategies Influence the Architecture of Transaction Networks Abstract In the context of business ecosystems, hierarchy is an architectural property that refers to the degree to which transactions proceed in a single direction, from “upstream” to “downstream.” It is often assumed that a unidirectional flow of goods in a value chain implies a corresponding hierarchy in the transaction networks of firms participating in the chain. However, this is an untested hypothesis: in fact, little is known about whether hierarchy varies across transaction networks, and, if so, what causes such variation. In this study, we apply network-based methods to define and measure the degree of hierarchy in interfirm transaction networks in two industry sectors in Japan: automotive and electronics. Our empirical results show that the electronics sector exhibits a much lower degree of hierarchy than the automotive sector due to the existence of numerous interfirm transaction cycles. Transaction cycles in turn can arise when a subset of firms adopt the strategy of vertically permeable boundaries. Such firms are vertically integrated in the sense of participating in multiple stages of the value chains, but their internal upstream units also sell into and downstream units buy from intermediate markets. Our comparative analysis suggests that firms elect the strategy of vertically permeable boundaries when they face low transaction costs and high rates of product innovation, but at the same time believe there are knowledge complementarities between different stages of the value chain. Vertically permeable boundaries allow such firms to take advantage of cross-division knowledge complementarities while maintaining the competitiveness of upstream units through their participation in intermediate markets. Keywords: industry architecture; transaction networks; firm boundaries; vertical integration; knowledge complementarities; innovation; hierarchy; modularity THE ARCHITECTURE OF TRANSACTION NETWORKS 1 Introduction Transactions are the most basic form of interfirm relationship and a fundamental unit of economic analysis (Commons, 1943; Williamson, 1981). Inspired by the seminal work of Coase (1937) and Williamson (1975, 1981, 1985), over the last three decades, a great deal of scholarly work has focused on transactions. Transaction cost economics (Coase, 1937; Williamson, 1975; 1981; 1985), knowledge-based theories of the firm (Grant, 1996; Teece et al., 1997), and modularity (Baldwin and Clark, 2000; Baldwin, 2008) have been used as theoretical lenses to examine the characteristics of transactions and related strategies such as “make or buy.” Most research on transactions has focused on single firms or industries or bilateral relationships between customers and suppliers (Nishiguchi, 1994; Argyres, 1996; Argyres and Liebeskind, 1999; Helper et al., 2000; Hoetker, 2006; Parmigiani, 2007; Nagaoka et al., 2008; Novak and Stern, 2009). Yet many technical systems have multi-step value chains, while others give rise to complementary firms producing different parts of a large system. Thus a growing body of research has been conducted at the level of a sector in which firms collectively innovate, design, and produce a coherent set of system products (Dosi, 1988; Langlois and Robertson, 1992; Malerba and Orsenigo, 1993; 1996; 1997; Teece et al., 1994; Baldwin and Clark, 2000; Brusoni and Prencipe, 2001; Malerba, 2002; Jacobides, 2005; Dalziel, 2007; Nakano and White, 2007; Adner and Kapoor, 2010). Increasing attention has also been paid to industry architecture and dynamics (Baldwin and Clark, 2000; Jacobides, 2005; Jacobides and Billinger, 2006; Stefano et al., 2009). However, despite this prior work, we still know very little about how transaction patterns vary across sectors or what factors influence the strategic choices of firms in such sectors. A business ecosystem is a group of firms linked by transactions and complementarities that together provide complex products and related services to end users (Iansiti and Levien, 2004; Adner and Kapoor, 2010). Recently, business ecosystems have emerged as an important form of economic organization, in many cases replacing the vertically integrated corporations studied by Chandler (1962, 1977) and Williamson (1975, 1985). Industrial firms participate in ecosystems through transactions and, by their individual strategic choices, collectively form transaction networks. In this paper, we explore the architecture of transaction networks in order to shed light on firms’ strategies in different business ecosystems. (In what follows, we will use the terms “sector” and “ecosystem” interchangeably.) 1 THE ARCHITECTURE OF TRANSACTION NETWORKS In addition to the above-cited work on transactions, there is a growing literature aimed at understanding interfirm network structures in general and their effects on firm performance (Gulati, 1995, 1998; Sorenson and Stuart, 2001; Rosenkopf and Schilling, 2007; Schilling and Phelps, 2007). Various kinds of network ties between firms, such as alliances (Gulati, 1995, 1998; Stuart, 1998; Rosenkopf and Schilling, 2007), social and financial ties (Sorenson and Stuart, 2001) and skill relatedness (Neffke and Henning, 2010) have been studied. However, the methods used in these studies cannot be directly applied to transaction network analysis, because the links in alliance and social networks are non-directed. In contrast, transactions form directed links: goods flow from Firm A to Firm B. Networks with directed links, such as transaction networks, require different methods of analysis, which is one of the contributions of this paper. To begin the work of exploring transaction networks, this paper addresses four questions: 1) Do the hierarchies of interfirm transaction networks vary across different ecosystems? In the context of business ecosystems, “hierarchy” is an architectural property that refers to the degree to which transactions proceed in a single direction, from “upstream” to “downstream.” In prior work Nakano and White (2007) conjectured that a unidirectional flow of goods in a production process implies a corresponding hierarchy in transactions between firms, but their data was limited to a single ecosystem. In this paper, we apply a new network-based metric developed by Luo and Magee (2011) to transaction data from the Japanese automotive and electronics sectors. The empirical results show that the pure hierarchy hypothesis does not always hold: specifically, the electronics sector is only partially hierarchical due to the wide presence of interfirm transaction cycles. This part of the research contributes to the industry architecture literature on a macro level by constructing a measure of hierarchy that can be applied to any interfirm transaction network and by showing that different networks do indeed display different degrees of hierarchy. 2) What practices explain the differences in hierarchy across these two ecosystems? To explain the variation in hierarchy across ecosystems, we investigated the theoretical determinants of hierarchy in a transaction network. We show that violations of pure hierarchy arise only if a subset of firms participate in multiple stages of their value chains, and their downstream internal units buy from and upstream internal units sell to intermediate markets. Such firms have vertically permeable boundaries in the sense 2 THE ARCHITECTURE OF TRANSACTION NETWORKS defined by Jacobides and Billinger (2006). A significant number of the largest electronics firms have adopted the strategy of vertically permeable boundaries. In contrast, the largest automotive firms in Japan concurrently source components from internal and external suppliers (Parmigiani, 2007), but their internal component divisions do not sell to external customers. 3) How do firms’ strategies influence hierarchy in different ecosystems? To better understand these differences in strategy and behavior, we conducted interviews with managers at key firms in the two sectors. We found that the electronics firms with vertically permeable boundaries faced short product life cycles and low transaction costs, creating pressures towards vertical disintegration. However, the firms’ managers also believed that there were significant knowledge complementarities across stages of the electronics value chains, which caused them to want to retain in-house component divisions (and related capabilities). The internal component divisions in turn honed their capabilities by competing with more specialized suppliers in intermediate markets. 4) What environmental factors explain the differences in the largest firm’s strategies in each ecosystem? The contrasting strategies of the largest firms in the automotive and electronic sectors in turn arose because the two groups of firms face different environmental conditions. As indicated, electronics firms face short product life cycles, low levels of asset specificity, and low transaction costs, whereas automotive firms face relatively long product life cycles, high levels of asset specificity and high transaction costs. These differences, which arise from differences in the modularity of the underlying products and processes, change the relative benefits and costs associated with the strategy of vertically permeable boundaries. Differences in modularity in turn can be traced back to the basic physics of automobiles vs. electronic products and processes. This article proceeds as follows. Section 2 reviews the relevant literature. Sections 3 and 4 introduce our measurement methods and describe our empirical results. Section 5 explains how the observed differences in the two transaction networks can be traced back to differences in the participating firms’ strategies and underlying technological constraints. Section 6 concludes by discussing this paper’s contributions, limitations of the analysis, and future research directions. 3 THE ARCHITECTURE OF TRANSACTION NETWORKS 2 Literature Review: Hierarchy in Industry Architecture We are interested in measuring variations in industry architecture across different ecosystems, and explaining this variation in terms of firms’ strategic choices and technological constraints. In this section, we briefly review three strands of related literature involving: (1) industry architecture, (2) hierarchy, and (3) the influence of technology on industry structure. 2.1 Industry Architecture Prior research has investigated industry structures in terms of horizontal concentration versus diversification within a single layer or industry (Penrose, 1959; Chandler, 1962; Teece, 1982; Davis and Duhaime, 1992), and vertical integration and disintegration (Baldwin and Clark, 2000; Sturgeon, 2002; Nagaoka et al., 2008; Helfat and Campo-Rembado, 2009). Single industry studies often investigate horizontal divisions of labor, revenue, or assets by firm or region. Many are based on theories as to how and why firms differ in terms of their resources and capabilities (Penrose, 1959; Wernerfelt, 1984; Barney, 1991; Teece et al., 1997). Studies of vertical integration/disintegration in turn investigate the boundaries of the firm, the division of activities between customer and supplier, and the nature of contracts. These analyses are often based on transaction cost economics (Williamson, 1975; 1981; 1985) or knowledge-based theories of the firm (Grant, 1996; Teece et al., 1997). Most prior work has not addressed the possibility that numerous complementary goods will be combined into larger systems supplied by many different firms. That is, many end-user products are not designed and produced within a single industry with stable boundaries, but by a network of firms connected by transactions in intermediate markets whose boundaries co-evolve over time (Brandenburger and Nalebuff, 1996; Langlois and Robertson, 1992; Jacobides, 2005). Such networks, made up of firms in different sub-industries that supply complementary goods, are sometimes called “sectors” (Malerba, 2002). More recently, the term “ecosystem” has come into widespread use (Iansiti and Levien, 2004; Evans, Hagiu and Schmalensee, 2006; and Adner and Kapoor, 2010). To avoid confusion, we consider a group of firms making similar products (for example, firms making engines or semiconductor chips) to form an “industry.” We call a collection of industries linked by transactions a “sector” or “ecosystem,” using these last two terms interchangeably. For example, the automobile sector (or ecosystem) includes firms making whole automobiles, plus those making sub-systems such as engines or interiors, components 4 THE ARCHITECTURE OF TRANSACTION NETWORKS such as pistons or seats, and materials such as glass, plastic, and steel. Transactions serve to coordinate activities across firm boundaries within a sector (Christensen and Rosenbloom, 1995; Jacobides and Winter, 2005). “Industry architecture” has been defined by Jacobides et al. (2006) as a somewhat stable but evolving set of relationships used to organize a given set of production and innovation processes.1 These relationships set the patterns through which labor and assets are divided in a sector between different types of industry participants, and the associated set of “rules and roles” that guide their behavior in the short and intermediate run. The necessary level of analysis for understanding industry architectures is the sectoral network of firms, comprising a group of related industries, rather than a supplier/customer pair or a single industry. An increasing number of studies at the sector level have explored how the internal boundaries of industries change, how intermediate markets emerge (Langlois and Robertson, 1992; Jacobides, 2005; 2006; Jacobides et al., 2006), and how sectors differ in terms of their technological bases, innovation patterns, and economic performance (Nelson and Winter, 1982; Dosi, 1988; Malerba and Orsenigo, 1993, 1996; 1997; Castellacci, 2007). However, our understanding is limited in terms of how industry architectures vary across sectors, and how technological constraints and/or strategic choices contribute to such variation. In particular, quantitative empirical descriptions of overall industry architectures are still lacking. 2.2 Hierarchy As indicated, hierarchy is an architectural property that refers to extent to which transactions proceed in a single direction, from “upstream” to “downstream.”2 At the sector level, existing studies often observe or suggest hierarchical industry architectures (Coase, 1937; Malerba, 2002; Dalziel, 2007; Nakano and White, 2007) in which production processes are organized into sequential stages (Coase, 1937; Abernathy et al, 1983; White, 2002a; Jacobides, 2005). Firms that perform higher level tasks depend upon firms that perform lower level tasks (Dalziel, 2007), a form of sequential interdependence (Thompson, 1967). According to Harrison 1 Industry architecture generally focuses on firms and the transactional relationships between them. In contrast, classic supply chain analysis focuses on inventories or plants and shipments between them. Clearly the two views are complementary. 2 This is different from the meaning of “hierarchy” in organization theory, where it refers to a command-and-control type of decision structure (e.g., Williamson, 1975). The difference between a “flow”, i.e., transaction, hierarchy and a “containment”, i.e., organizational, hierarchy are discussed in the next section. 5 THE ARCHITECTURE OF TRANSACTION NETWORKS White (2002b:87), “production markets” show “persistent directionality in continuing flows of intermediate goods” in which “only a niche within an industry establishes you in a line of business.” In contrast, such hierarchy does not exist in a pure market, such as a stock market, where traders can both buy from and sell to one another.3 Nakano and White (2007) offer empirical evidence that the network of firms connected by supplier-customer relationships in the Tokyo industrial district exhibits a strict hierarchical architecture. Based on this evidence, they further hypothesized that hierarchy is a general property of production markets because firms in such markets tend to become entrenched in their positions and roles as buyers and sellers over time. In contrast, firms in financial markets constantly change roles, so that an enduring institutionalized role structure does not exist (Nakano and White, 2007). Nakano and White’s analysis neglected the role of firm strategy, however. For reasons of strategy, some firms might not adhere to fixed positions and roles: in dynamic markets, such commitments might be disadvantageous in the long run. For example, Foxconn, the largest Taiwanese original design manufacturer (ODM) of personal computers, supplies finished computers directly to personal computer makers, such as Dell and Apple, but it also produces and sells many connecters, cables, and printed circuit boards to other suppliers upstream in the personal computer manufacturing sector. Foxconn plays multiple (and changing) roles, and as a result, its vertical position in the value chains is ambiguous. If a sector includes many firms like Foxconn, the transaction flows in the sectoral network may not be purely hierarchical, but instead partially hierarchical or even non-hierarchical. If hierarchy in a transaction network can exist along a continuum, then the appropriate question to ask is not whether a transaction network is hierarchical or not, but is it more or less hierarchical than others. Answering this question requires new methods of analysis. A general approach to quantifying hierarchy in networks has been recently developed in related work (Luo and Magee, 2011). In section 3, we introduce this metric and in section 4 we use it to measure the variable degree of hierarchy in two interfirm transaction networks. 3 In Williamson (1975)'s Markets and Hierarchy, the market is assumed to be a non-hierarchical space, and it represents the opposite to hierarchy as a means of organizing economic activity. Following Williamson, the term “hierarchy” is often used to mean relationships within enterprises or firms where command and control replace market transactions. 6 THE ARCHITECTURE OF TRANSACTION NETWORKS 2.3 The Influence of Technology on Industry Structure The extent to which a given transaction network is hierarchical is the collective result of firms’ strategic choices within the constraints of the technologies they use. Technology has been considered as a factor influencing the structure of alliance networks (Rosenkopf and Schilling, 2007) and technological relatedness networks (Neffke and Henning, 2008). Christensen and Rosenbloom (1995) and Murmann and Frenken (2006) have also suggested that hierarchical supply relationships between firms may mirror the hierarchical structure of their products (comprising of systems, subsystems, components, and so on). In addition, Paprzycki (2005), based upon several descriptive cases in the Japanese electronics sector in the 1990s, argues that the previously hierarchical supplier-customer relationships collapsed to some degree, when more modular and open product architectures emerged. He reasoned that firms at the time had to actively search outside their well-aligned hierarchical value chains in order to access external resources or capabilities required by new architectures (Penrose, 1959; Wernerfelt, 1984; Teece et al., 1997). Thus technological changes toward more modular architectures can cause intermediate markets to emerge. Formerly vertically integrated firms might then find that their internal divisions are no longer competitive in certain stages of the value chain (Jacobides, 2005). Their choice is then to exit from such stages or increase the competitiveness of their internal divisions. The end result of this process could be a set of purely focused firms or a set of vertically integrated firms with permeable boundaries. It is often suggested that a higher level of modularity in a technical system is associated with a higher level of vertical disintegration. Industry studies have addressed this issue in two ways (Colfer and Baldwin, 2010). First, some scholars have asked, do industries vertically disintegrate when modular product architecture becomes the dominant design (Suarez and Utterback, 1995; Utterback, 1996)? Conversely, does an industry vertically integrate when an integral dominant design emerges? Transaction cost economics predicts such “mirroring” between product architecture and transactions (Williamson, 1981; Henderson and Clark, 1990; Baldwin, 2008), but firms’ strategies generally go beyond simple transaction cost logic. Thus Helfat (1997) has suggested that vertically integrated firms can take advantage of knowledge complementarities in building long-term dynamic capabilities (Teece and Pisano, 1994). She supported her argument by documenting that oil companies with complementary knowledge and physical assets undertook more R&D on synthetic coal conversion in response to 7 THE ARCHITECTURE OF TRANSACTION NETWORKS rising oil prices during the 1970s. In a dynamic model of alternating modular and integral architectures (Fine, 1998; Fine and Whitney, 1999), Helfat and Campo-Rembado (2009) further argued that even when modular architectures are advantageous, forward-looking firms still might choose to remain vertically integrated in order to maintain their capabilities over the technology cycle. Thus vertically integrated firms would co-exist with vertically specialized firms if the former perceived that the future value of their dynamic capabilities outweighed the short-term disadvantages of vertical integration. Numerous empirical studies have attempted to correlate modularity with vertical integration/disintegration, in industries including stereo systems (Langlois and Robertson, 1992), computers (Baldwin and Clark, 2000; Hoetker, 2006), banking products and services (Jacobides, 2005), and bicycles (Galvin and Morkel, 2001; Fixson and Park, 2008). Overall, the results have been mixed. On balance it appears that modularity is a necessary but not sufficient condition to explain vertical disintegration in an industry or sector (Colfer and Baldwin, 2010). In section 5, we will argue that differences in the modularity of products and processes lead firms to select different strategies with respect to vertically permeable boundaries. These differences in turn explain why the two transaction networks we studied differ in terms of their empirically measured degrees of hierarchy. In the next two sections, we explain how we measure hierarchy and describe our empirical results. 3 Measuring Hierarchy in Industry Architecture A hierarchy is a generic structure, in which levels are asymmetrically ranked according to a specific type of relation (Ahl and Allen, 1996; Luo and Magee, 2011). A flow hierarchy arises when there is directional movement through a series of stages. For example, if firm B purchases a good from firm A, processes and adds value to the good and then sells the new good to firm C, the goods flow from A to B to C in a sequence of value-adding stages. The order of stages essentially determines the direction of the flows of goods. Such flows are fundamental to the organization of the economic system, and (at a high level of aggregation) are reflected in national input-output tables (Leontief, 1951).4 4 Flow hierarchies differ from organizational hierarchies. Specifically, an organizational hierarchy is containment hierarchy, also known as a “nested hierarchy” (Ahl and Allen, 1996; Simon, 1962; Christensen and Rosenbloom, 1995; Murmann and Frenken, 2006) or a “hierarchy of inclusion” (Wilson, 1969). In a containment hierarchy, lower 8 THE ARCHITECTURE OF TRANSACTION NETWORKS Interfirm transaction networks can be represented as directed graphs in which the nodes are firms and the links are transactions. Such networks include cycles and sequences (Thompson, 1967). A pure flow hierarchy has only sequences, and all flows are unidirectional, as shown in the example networks A, B and C in Figure 1. Thus a pure flow hierarchy generalizes Thompson’s notion of “sequential interdependence” (Thompson, 1967, p. 54). Cycles violate the principle of strict hierarchy because flows can come back to their origin. Networks D, E, F in Figure 1 embed cycles to different extents. The firms in such networks display some degree of “reciprocal interdependence” either direct or indirect (Thompson, 1967, p.55). 1 1 3 2 4 5 6 1 3 2 7 4 5 A) 6 3 2 7 4 5 B) 1 7 C) 2 1 6 2 3 5 3 1 3 2 4 6 8 4 5 6 7 7 9 D) E) 5 4 F) Figure 1 Examples of different flow hierarchies Directionality is the fundamental principle of a flow hierarchy. It follows that we can measure the degree of hierarchy in a network by capturing to what extent it contains or does not contain cycles. Our hierarchy metric (h) is calculated as the percentage of links that are not included in any cycle: m ∑e i h= i =1 (1) m where m is the number of links in the network and ei=0 if link i is in a cycle and 1 levels lie within or are aggregated into upper levels, and upper levels contain lower levels. Complex products like airplanes are often viewed as containment hierarchies, because they are made up of subsystems, which contain sxmaller components and parts (Tushman and Murmann, 1998). 9 THE ARCHITECTURE OF TRANSACTION NETWORKS otherwise. In general, this metric is advantageous in its clarity and ease of computation in comparison to other potential metrics. It has wide applicability in other network systems, such as organizations, teams, and products.5 In some applications, it is useful to weight the links by, for example, the volume of flows. However, in this paper we focus on unweighted networks because our empirical data includes incomplete information about weights of all the links we draw. In addition, this metric only counts whether a link is involved in any cycle, but does not take into account the length of cycles. Completely tracing cycle sizes is computationally difficult when networks are large and adds little insight. We tested the flow hierarchy metric on the simple networks shown in Figure 1. The results are given in Table 1. Networks A, B, and C, in which all flows proceed in one general direction and no cycle exists, are purely hierarchical, thus h=1. Networks D and E are partially hierarchical and their degrees of hierarchy are 0.71 and 0.40 respectively. Network F is a pure cycle, thus h=0. Table 1 Degrees of hierarchy in the example networks in Figure 1 Networks in Figure 1 A B C D E F Hierarchy Degree (h) 1 1 1 0.71 0.40 0 Architectural Regime Purely Hierarchical Purely Hierarchical Purely Hierarchical Partially Hierarchical Partially Hierarchical Purely Cyclic Table 1 shows that the flow hierarchy metric potentially characterizes interfirm transaction networks into three canonical architectures: (1) Purely hierarchical (single-directional transaction flow), h=1; (2) Purely cyclic (every firm is involved in cycles), h=0; (3) Partially hierarchical (sequence and cycle are combined), 0<h<1. In the next section, we apply the hierarchy metric, as well as network visualization tools, to transaction data from two industry sectors in Japan, and show that one sector is substantially more hierarchical than the other. 5 For more details on this metric, including the algorithm to calculate it for large-scale networks, comparisons with other metrics, and its applications to other kinds of networks, see Luo (2010) and Luo and Magee (2011). 10 THE ARCHITECTURE OF TRANSACTION NETWORKS 4 Data and Empirical Results 4.1 Data We extracted supplier-customer transactional relationship data from the series data books “The Structure of the Japanese Auto Part Industry” and “The Structure of the Japanese Electronics Industry,” which are based on regular surveys by Dodwell Marketing Consultants. The company directories in these two data books provide information on the major customers and suppliers for each firm. Such information makes it possible to extract “who-supplies-whom” type of connections between firms,6 and to build multi-tier sectoral transaction networks. The data books are only available in hard copy and had to be manually entered into an electronic database. We used the data books published in 1983, 1993, and 2001, but we believe the data actually represents the situation approximately two to three years before the publishing year, because the publications were refreshed every two to three years. Table 2 lists the largest firms included in our data by revenue. Largest 10 Firms in Electronics Network Largest 10 Firms in Automotive Network Table 2 List of the largest firms included in the network data Company Year Ending Toyota Motor Nissan Motor Honda Motor Mitsubishi Motors Mazda Motor Isuzu Motors Suzuki Motor Fuji Heavy Industries Daihatsu Motor Hino Motors Hitachi Matsushita Electric Industrial* Toshiba Sony NEC Fujitsu Mitsubishi Electric Canon Sanyo Electric Sharp June 1993 March 1993 March 1993 March 1993 March 1993 October 1993 March 1993 March 1993 March 1993 March 1993 March 1992 March 1992 March 1992 March 1992 March 1992 March 1992 March 1992 December 1991 November 1991 March 1992 Revenue (Billion Yen) 9,031 3,897 2,695 2,615 2,191 1,199 1,053 873 785 632 7,766 7,450 4,722 3,915 3,744 3,422 3,343 1,869 1,616 1,555 * Matsushita Electric Industrial was renamed to Panasonic Corporation in 2008 6 We do not have the detail on the specifics of individual transactions. 11 THE ARCHITECTURE OF TRANSACTION NETWORKS For each sector in a specific year, we constructed a directed network in which nodes are firms and links are supplier-customer transactional relationships. The transactions indicated are compensated transactions of physical products, excluding services and intellectual property. Basic network statistics, including Number of Firms (n), Number of Transactional Relationships (m), and Average Degree7 (k=m/n), are shown in Table 3. The automotive transaction networks contain more nodes and links and have a higher average degree than the electronics transaction network in 1993. Table 3 Network descriptive statistics Network Attributes Japanese Automotive Sector Japanese Electronics Sector Year 1983 1993 2001 1993 Number of Firms (n) 356 679 627 227 Number of Transactional Relationships (m) 1480 2437 2175 648 Average Degree (k=m/n) 4.157 3.589 3.469 2.855 Existing studies of interfirm alliance networks have shown skewed degree distributions, that is, a small number of firms have a large number of alliance partners. It is conjectured that firms with larger size, higher prestige, and stronger architectural control8 tend to attract a greater number of alliances than smaller and less prestigious firms (Stuart, 1998; Rosenkopf and Schilling, 2007). Similarly, our interfirm transaction data (see Table 4) indicate that the largest firms (according to Table 2) and the firms that control product architectures have the highest numbers of suppliers in both the automotive and electronics networks. 7 In graph theory, the degree of a node means the number of nodes connected to it. In a directed network, there are two types of degrees applying to a single node: in-degree (number of nodes connected to it) and out-degree (number of nodes it connects to). Average in-degree and out-degree of a network are equal. 8 Architectural control refers to “the ability of a firm(s) to define specifications for both the individual subsystems of a product as well as the integration of these subsystems to form the end product.” (Rosenkopf and Schilling, 2007) 12 THE ARCHITECTURE OF TRANSACTION NETWORKS Table 4 The most connected firms in automotive and electronics interfirm transaction networks in 1993 Top 10 firms with the highest number of suppliers Electronics Sector Automotive Sector Company Mitsubishi Motors Nissan Motor Honda Motor Toyota Motor Mazda Motor Isuzu Motors Fuji Heavy industries Suzuki Motor Daihatsu Motor Hino Motors Hitachi Toshiba NEC Sony Fujitsu Mitsubishi Electric Matsushita Electric Industrial Sharp Sanyo Electric Victor Company of Japan Number of Suppliers 226 176 169 166 157 135 127 125 99 98 52 40 38 36 34 33 30 23 15 15 Top 10 firms with the highest number of customers* Number of Customers 0 0 0 0 0 0 0 0 0 0 17 26 18 3 12 7 27 3 3 1 Company Du Pont Japan Fuji Bellows Bando Chemical Industries Sumitomo Electric Ind. Hashimoto Forming Ind. Tokai Rubber Industries Meiwa Industry Nishikawa Rubber Nisshinbo Industries Ogihara Matsushita Electric Industrial Toshiba NEC Hitachi Alps Electric TDK Fujitsu Yokogawa-Hewlett-Packard Nippon Motorola Intel Japan k.k. Number of Suppliers 0 0 0 4 3 3 2 0 0 0 30 40 38 52 1 1 34 5 1 0 Number of Customers 18 17 15 12 11 11 11 11 11 11 27 26 18 17 15 13 12 10 10 10 * “Customers” are within sector and do not include end-users. However, there is a notable difference between two sectors in that the automotive assemblers have no customers within the sector while some of the largest electronics firms (such as Matsushita Electric, Toshiba, NEC, Hitachi, and Fujitsu) have the highest numbers of customers in the sector. In other words, the largest purchasers and suppliers of electronics components are the same firms. Meanwhile, in the automotive sector, the firms with the largest number of suppliers normally have no customers within the sector, and vice versa. With these basic facts in hand, we now analyze the two transaction networks using standard network tools plus our hierarchy metric. In sub-sections below, we present graphical visualizations, matrix visualizations, hierarchy calculations, and an analysis of embedded cycles for the two transaction networks. 4.2 Graphical Visualization We used Netdraw, a leading social network visualization software (Borgatti, 2002), to create graphical pictures of the interfirm transaction networks in the automotive and electronics sectors in a comparable year (1993). The visualizations are shown in Figure 2. They allow us to 13 THE ARCHITECTURE OF TRANSACTION NETWORKS see that the automotive network has more nodes and links and that both networks contain a number of “hubs” (nodes with many links). But the diagrams do not reveal the presence of cycles or hierarchy in the two networks. A) Automotive Sector B) Electronics Sector Figure 2 Japanese interfirm transaction networks in 1993 4.3 Matrix Visualization Matrices are better than graphs at revealing flow hierarchies in networks. In design theory, the Design Structure Matrix (Steward, 1981; Eppinger et al., 1994) is used to examine the dependencies between design elements in a square matrix. These elements are often design tasks, design parameters, or the actual components in a designed or engineered system, such as software (MacCormack et al., 2006) and automotive brake systems (Eppinger et al., 1994). In contrast to the design processes that are the focus of most DSM research, interfirm transaction networks are not designed centrally: instead they are self-organizing systems that emerge from numerous, decentralized “make or buy” decisions by the constituent firms. A square matrix can still be used to examine the overall pattern of economic interdependencies between firms, but calling it Design Structure Matrix is inaccurate. Instead, we call it Dependency Structure Matrix (DSM). Figure 3 shows DSM representations for the automotive and electronics interfirm transaction networks in Japan in 1993. The elements on both axes are firms listed in the same order, and the dots represent dependencies. If firm j is a customer of firm i, i.e., firm j depends on firm i for its supplies, we put a dot in the cell (i, j) of the matrix. For example, in the automotive DSM, dot (359, 524) indicates that Nippon Denso (firm 524) is a customer of Arai 14 THE ARCHITECTURE OF TRANSACTION NETWORKS Seisakusho (firm 359). In the electronics DSM, dot (147, 124) indicates that Omron (firm 124) is a customer of Matsushita Electric Industrial (firm 147–now renamed Panasonic). 0 50 100 150 200 250 300 350 400 450 500 550 600 650 0 0 50 100 150 200 0 Arai Seisakusho 50 Matsushita Electric Industrial (Panasonic) Partially Hierarchical Hierarchical 100 150 50 200 250 100 300 350 Omron 400 150 450 500 550 Nippon Denso 200 600 650 A) Automotive Sector B) Electronics Sector Figure 3 Dependency Structure Matrices for Japanese interfirm transaction networks in 1993. The small boxes drawn inside the DSMs encapsulate strong components (Wassserman and Faust, 1994), in which all nodes are on cycles with each other. The automotive sector DSM shows there is only one strong component in the automotive network, and its size is (3 nodes, 3 links). In contrast, in the electronics network, there are four strong components, and their sizes are (84 nodes, 254 links), (3 nodes, 4 links), (2 nodes, 2 links), and (2 nodes, 2 links), respectively. The dots in a row indicate how many suppliers the firm in that row has. In DSM A, the dense bottom rows are the large car manufacturers like Toyota and Nissan. No such dominance appears in B. In the DSMs in Figure 3, firms are ordered according to their visibilities,9 thus firms on connected cycles will be grouped together (MacCormack et al., 2010). In the automotive DSM, almost all dots are below the main diagonal, indicating that this network is extremely hierarchical. In contrast, the electronics DSM has many dots above the diagonal, indicating that many firms participate in transaction cycles. Furthermore, most of these cycles are intertwined together in one strongly connected component that encapsulates 84 nodes (37% of the firms) and 254 links (39% of the transactional relationships). Firms in the large component buy from and sell to each other in complex patterns. In general, the comparison of the two DSMs in Figure 3 reveals significant differences between the two sectors in terms of their hierarchical degree. We now use the Luo and Magee (2011) metric, described above, to quantify the differences. 9 “Visibility” is the count of all the direct and indirect dependencies a node possesses with other nodes. 15 THE ARCHITECTURE OF TRANSACTION NETWORKS 4.4 Hierarchy Measurement We computed the hierarchy metric, h, for the interfirm networks of the Japanese electronics production sector in 1993 and the Japanese automotive production sector in 1983, 1993, and 2001. Table 5 repeats the descriptive statistics of Table 3 (for convenience) and calculates the degree of hierarchy (h) for each network. The comparison of these two sectors in one year (1993) shows that the electronics production sector (h=0.5957) is quantitatively much less hierarchical than the automotive production sector (h=0.9988), due to the presence of many transaction cycles. Furthermore, the degree of hierarchy in the automotive sector in Japan did not change much over time and remained high from the early 1980s to the early 2000s. Table 5 Empirical Measurement Results Network Attributes Japanese Automotive Sector Japanese Electronics Sector complete remove 10 network largest firms Year 1983 1993 2001 1993 1993 Number of Firms (n) 356 679 627 227 216 Number of Transactional Relationships (m) 1480 2437 2175 648 221 Average Degree (k=m/n) 4.157 3.589 3.469 2.855 1.023 Number of Firms in Cycles (nc) 4 3 2 91 13 Number of Links in Cycles (mc) 4 3 2 262 14 Degree of Hierarchy (h=1- mc/ /m) 0.9973 0.9988 0.9991 0.5957 0.9367 many* 7 two-node cycles Cycle Tracking 2 two-node 1 three-node 1 two-node cycles cycles cycle * In the 1993 electronics sector network, there were 51 two-node cycles, 12 three-node cycles, 92 four-node cycles, 107 five-node cycles, 598 six-node cycles, and many larger cycles. There were four strong components in which all nodes are on cycles with each other, and their sizes are (84 nodes, 254 links), (3 nodes, 4 links), (2 nodes, 2 links), and (2 nodes, 2 links), respectively. These four strong components are boxed in the electronics sector DSM in Figure 3. In the automotive sector, only one or two small cycles are found in the network in any year. These are shown in Figure 4. 16 THE ARCHITECTURE OF TRANSACTION NETWORKS A) The only two cycles found in 1983 Arakawa Auto Body (Araco) • major products: car assembly (trucks, buses, specialized vehicles, etc) 46%, interior parts 28%, other auto parts 6%, other items 20% • this company changed its name to “Araco” in the early 1990s This link disappeared after 1983 Toyota Auto Body Fuji Kiko • major products: car assembly 84% (passenger cars 45%, commercial vehicles 31%, trucks 8%), auto parts, etc. 16% • major products: steering columns, safety belts, transmission parts, pedals, engine parts, reclining seat equipment, suspension system parts, etc. 99%, agricultural machinery parts, other items 1% This link disappeared after 1983 Ikeda Bussan • major products: seating 78.3%, interior accessories 14.4%, chemical products 6%, other items 1.3% B) The only cycle found in 1993 Araco • major products: vehicle assembly (trucks, buses, specialty vehicles, etc) 50%, seating, door trims, and roof linings 50% • customers: Toyota 98.4%, Daihatsu motor 0.2%, Toyota Auto Body 0.1% Toyota Auto Body • major products: car assembly 84% (passenger cars 45%, commercial vehicles 31%, trucks 8%), auto parts, etc. 16% • customers: Toyota Motor, Toyota Tsusho, Gifu Auto Body Industry this link disappeared after 1993 Gifu Auto Body Industry • major products: bodies for trucks, specialty vehicles 64%, pressed auto parts (seat adjuster, radiator baffles, door trims) 31%, others, 5% • customers: Toyota Motor 90%, Takashimaya Nippatsu Kogyo 3%, Toyota Shatai 1%, Dahatsu Motor 1%, Araco Toyota Auto Body Co Ltd (TA) acquired the vehicle manufacturing and sales business of Araco Corp (AR), a manufacturer of automotive seat cover, and a unit of Toyota Motor Corp (TM) – announced on October 1st, 2004 Toyota Auto Body Co Ltd (TA) acquired the remaining 89.09% interest of Gifu Auto Body Co Ltd, a manufacturer of automobile and truck bodies – announced on October 1st, 2007 C) The only cycle found in 2001 Delta Kogyo K.K. • major products: seating 68%, shift levers 11%, sun visors 3%, ashtrays 1%, door parts 1% In 1983 and 1993 data, this link did not exist In 1983, this link did not exist Toyo Seat • major products: seating for automobiles 95.4%, exhaust pipes 0.9%, others 3.7% Figure 4 All the cycles found in the automotive interfirm transaction networks in 1983, 1993, and 2001 17 THE ARCHITECTURE OF TRANSACTION NETWORKS The transactional links in these cycles are of only minor volume10, that is, they are “weak ties” in the network analysis term (Granovetter, 1973). These cycles also did not last long. For instance, in the 2000s, Toyota Auto Body acquired the other two firms (Araco and Gifu Auto Body Industry) involved in the only cycle found in the 1993 network, so this cycle no longer exists. In brief, the automotive transaction network is very hierarchical, and thus is clear which firms are “upstream” and which are “downstream.” In contrast, approximately 40% of the interfirm transactional relationships in the electronics sector in 1993 were involved in cycles. In the electronics network, there were 51 two-node cycles, 12 three-node cycles, and many larger cycles. Figure 5 presents two examples extracted from the data. A) An example of 2-node cycle PC, Server, Systems Shindengen Electric Mfg Fujitsu Components, Power Units B) An example of 3-node cycle PCB Assemblies Matsushita Electric Industrial (Panasonic) CMK Components Matsushita Electric Works Materials for boards Figure 5 Example cycles found in the electronics interfirm transaction network Note: Fujitsu owned a 7.2% share of Shindengen Electric Mfg in 1992. Matsushita Electric Industrial owned a 32.5% share of Matsushita Electric Works, and Matsushita Electric Works owned a 3.6% share of CMK in 1992. Information on what was transacted comes from our interviewees at Fujitsu and Panasonic, respectively, based on their knowledge of the firms’ business in the early 1990s. Fujitsu purchased components and power units from Shindengen Electric Manufacturing for the integration of its personal computer, server and system products, and then also supplied such products to Shindengen Electric. Matsushita Electric Industrial (now Panasonic) sold 10 The Dodwell data for the auto part industry has some but incomplete information about the portion of procurement from each of the major suppliers that a customer firm lists. Fortunately, we can find such information for the firms involved in these four cycles in the automotive networks, but not for all the firms. 18 THE ARCHITECTURE OF TRANSACTION NETWORKS components to Matsushita Electric Works, which sold materials for making electronic boards to CMK. CMK in turn was a supplier of PCB (printed circuit board) assemblies to Matsushita Electric Industrial. We found that most of the cycles in the electronics sector included at least one large diversified firm, such as Panasonic or Fujitsu. These large firms apparently play a major role in forming cycles of varied sizes. We tested this hypothesis by removing the largest 10 firms by revenue (see Table 2) from the network. In the new network without the largest 10 firms, only 14 out of 221 links were on cycles (i.e. h=0.9367), while only 13 of 216 firms participated in cycles. This indicates that most of the cycles in the electronics sector were bridged by a small number of large firms. In brief, the electronics interfirm transaction network is only partially hierarchical. Inside the strongly connected component of the network (the large central block in Figure 3B), it is not clear which firms are “upstream” and which are “downstream.” This transaction network thus provides a counterexample to Nakano and White’s (2007) hypothesis that hierarchy is a general property of production networks. In the next section, we will analyze the strategies and technological constraints of the largest automotive and electronics firms, to see how these may have affected the formation of transaction cycles. 5 Understanding Hierarchy in Industry Architecture As stated above, industry architecture is not designed centrally but emerges from the interaction of individual firms’ strategies and behaviors, subject to their social and technical environments. In this section, we conduct a theoretical analysis to understand how transaction cycles result from firms’ strategic choices. Our analysis is based on the basic properties of networks augmented by information from interviews conducted in 2009 with managers at three firms in the automotive sector (Denso, Sumitomo Light Metal, and Panasonic Automotive Systems), and six firms in the electronics sector (Fujitsu, Panasonic, Casio-Hitachi, Seiko-Epson, Sony, and Takashima Sangyo). In particular, the interviews revealed significant differences between the two sets of firms in terms of their strategies, organizational capabilities, product architectures and innovation dynamics.11 11 Details about the interviewees and the firms are reported in Luo (2010). 19 THE ARCHITECTURE OF TRANSACTION NETWORKS 5.1 Transaction Cycle and Vertically Permeable Boundary We first explore how transaction cycles emerge from firms’ strategies in a generic self-organizing transaction network. On the one hand, if all firms in a sector are vertically integrated and do not buy or sell goods in intermediate markets, there will be no (intermediate) transactions, and perforce no cycles. This is the classic Chandlerian model of vertical integration (Chandler, 1962; 1977). Alternatively, if all firms are purely focused on a single stage of the value chain, then the interfirm transaction flows will mirror the uni-directional flow of goods through the stages of production. There will be many transactions, but no cycles. Finally, firms may practice “concurrent sourcing,” obtaining components from both internal units and external suppliers (Parmigiani, 2007). However, if such firms only purchase and do not sell components to other firms in their sector, then again, there can be no cycles. Cycles are possible only if some firms are present in multiple stages of the value chain and simultaneously downstream internal units buy components from and upstream internal units sell components to other firms in the sector. This is the case for Foxconn (discussed above), Fujitsu and Panasonic (shown in Figure 5). One of our interviewees described a cycle of transactions at a large electronics manufacturer in Japan.12 This firm (Firm A) makes package substrates, chipsets, and whole systems. Internally, its substrate unit transfers its products to the chipset unit which in turn transfers products to the systems unit. But the substrate unit also sells products to Firm B, a specialized chipset maker, while the systems unit purchases chipsets from that firm. Thus Firm A has a vertically permeable boundary as described by Jacobides and Billinger (2006). It allows goods to flow from division to division within the firm, but at the same time, internal divisions are allowed, encouraged, or required to buy from and sell to external suppliers and customers. In the case described by our interviewee, a transaction cycle is formed between Firms A and B as shown in Figure 6. Firm A has a vertically permeable boundary, while Firm B, which is present in only one stage of the value chain, does not. 12 The firms’ names have been disguised at the request of the interviewee. 20 THE ARCHITECTURE OF TRANSACTION NETWORKS Firm A Systems Chipsets Firm B Subsystems Subsystems Package Substrates Components Market Transaction Internal Transfer Firm Intermediate Products / Processes Figure 6 Vertically permeable boundary and interfirm transaction cycle. Figure 7 shows another way in which vertically permeable boundaries can give rise to interfirm transaction cycles. Here Firms C and D have internal divisions that participate in upstream and downstream stages of different value chains within the same sector. For example, Firm C might make PCB (Printed Circuit Board) (a subsystem) and TV sets (a system), while Firm D makes flat panel displays (a subsystem) and computers (a system). In this stylized example, there are no product flows between the subsystem and system divisions within each firm, but both firms are present in different stages of technologically related value chains computers and TVs). (We might call such firms “vertically diversified.”) If Firm C sells PCB to Firm D and purchases flat panel displays from it, then a transaction cycle exists between the two firms. Both firms are present in multiple stages of their value chains and both buy from and sell into intermediate markets, hence, by our definition, both have vertically permeable boundaries.13 13 Interfirm transaction cycles can also arise across sectors if some firms adopt a strategy of unrelated diversification. The incidence of cross-sector cycles depends on the prevalence of business groups made up of technologically unrelated units in the economy. Investigating such patterns is an interesting topic for future research. 21 THE ARCHITECTURE OF TRANSACTION NETWORKS Computers Television Sets Firm C Firm D System 2 Subsystem s for 1 Market Transaction System 1 PCB Flat Panel Display Firm Subsystems for 2 Intermediate Products / Processes Figure 7 Vertically permeable boundaries and interfirm transaction cycle: a different example. Jacobides and Billinger (2006) described how a European clothing manufacturer, facing increased competition, made its vertical boundaries permeable. In adopting this strategy, the firm sought to: (1) improve transparency and the ease of monitoring performance of its business units; (2) gain operational efficiencies; (3) improve resource allocation; (4) facilitate innovation; and (5) nurture dynamic capabilities. However, this paper focused on a single firm, thus did not consider the potential impact of vertically permeable boundaries on the surrounding transaction network. Our analysis of the Japanese electronics sector shows that when a few large firms create vertically permeable boundaries, it becomes possible for transaction cycles to emerge, and hierarchy may be reduced as a result. Any interfirm transaction cycle must include at least one firm whose vertical boundary is permeable in both directions: that is, a firm whose downstream units buy and upstream units sell components in intermediate markets. Our data reveal that the largest electronics firms satisfy this condition, whereas the largest automotive manufacturers do not (Whitney, 2007; MacDuffie, 2008). Although some of the largest automotive manufacturers (e.g. Toyota, Nissan) concurrently source components from internal units and external suppliers, such firms almost never sell components made by their upstream units to external customers. (See Table 4.) What explains this difference in the strategies of the largest firms in the two sectors? On first glance, one possible explanation for vertically permeable boundaries might be the keiretsu business culture in Japan. A keiretsu is a group of companies with long-time interlocking 22 THE ARCHITECTURE OF TRANSACTION NETWORKS business relationships and shareholdings. For example, Denso Corporation is a long-time keiretsu member of Toyota. Indeed many external suppliers are keiretsu members of the largest assemblers in both sectors, and the presence of keiretsu may be related to the practice of concurrent sourcing, which is found in both sectors. However, interfirm transaction cycles only took place widely among keiretsu firms in the electronics sector, not in the automotive sector. Thus, the keiretsu business culture, which is present in both sectors, cannot explain their different degrees of hierarchy. Below we offer an explanation that rests on well-documented technological differences between the automotive and electronics sectors and perceived knowledge complementarities within the electronic sector. 5.2 The Impact of Modularity While our two sectors are culturally similar, they differ substantially in the technological dimensions of their products and processes. In particular, firms in the automotive and electronics sectors have strategically embraced different degrees of modularity in their major products (Fujimoto, 2007; MacDuffie, 2008). In two survey papers, Fujimoto (2007) and Schilling and Phelps (2007) asked experts for their subjective evaluations of modularity versus integrality of different industrial goods. In both studies, experts agreed that most electronic and electrical products are far more modular than automobiles. When products and processes are modular, it is relatively easy to decompose and recombine components (Simon, 1962; Whitney, 1996; Baldwin and Clark, 2000; Schilling, 2000). In contrast, integral products exhibit many strong functional or physical interdependencies between components, hence are hard to break apart (Simon, 1962). Modularity in turn has been shown to affect both transaction costs and the rate of innovation in technical systems (Baldwin and Clark, 2000; Sturgeon, 2002; Baldwin, 2008; Colfer and Baldwin, 2010). Integrality of the Automotive Sector. As indicated, contemporary automobiles are integral products in which components and subsystems are highly interdependent. Takeishi and Fujimoto (2001) observed that, in the Japanese automotive industry, the functions assigned to individual parts of an automobile have become increasingly complex, and the need for structural or functional coordination has increased commensurately. MacDuffie (2008) argued that the integrality of contemporary automobiles is due to a number of systemic requirements, including 23 THE ARCHITECTURE OF TRANSACTION NETWORKS energy efficiency, emissions, noise, vibration, safety, stability, driving feel, design, and cost, that must be met in order to attract consumers and satisfy regulators. Interdependencies in automotive designs demand a specific tailored response by the supplier that designs the component and its interfaces. In general major automotive components, such as piston rings and mufflers, cannot be designed without detailed knowledge of the products in which they will be used and the other components with which they interact. In the automotive sector, component designs are specific to the systems in which the components are used. Such component specificity, also called “synergistic specificity” by Schilling (2000), gives rise to Williamsonian asset specificity in the following way (Williamson, 1975; 1981). In order to guarantee that components will function in a given system, product designs must be specifically tailored. As a result, contracts between firms are relational and “hand-in-glove.” Both suppliers and customers must invest in skills, assets and resources (including knowledge) that are valuable only in the context of their specific relationship (Asanuma, 1989; Baker et al., 2002). Indeed the automotive suppliers we interviewed confirmed that they had to tailor their product designs and production processes to the particular requirements of their customers, and invest in deep, ongoing relationships. Such relationships have continuously improved quality, cost and delivery in the automotive sector (Sako, 1992; Nishiguchi, 1994; Helper et al., 2000). Consistent with Parmigiani’s (2007) concept of concurrent sourcing, automotive manufacturers often procure intermediate products from external suppliers, in order to benchmark or discipline internal units, or for other reasons (Fine and Whitney, 1999). However, they rarely sell from their own vertically integrated component divisions to other automakers (or even the suppliers of other automakers). External sales are not particularly attractive for two reasons. First the need to design components specifically for customers’ systems means that external sales do not increase economies of scale or lower manufacturing costs for component divisions. Second the fact that external customers are also competitors raises concerns about excessive vulnerability, particularly the potential leakage of intellectual property—in both directions (Henkel and Baldwin, 2010). In a nutshell, system integrality leads to component and asset specificity in the automotive sector. Such specificity reduces the scale benefits and increases the transaction costs of having upstream internal units sell into intermediate markets. For these reasons, the major automotive firms have boundaries that are permeable in one direction only. They source 24 THE ARCHITECTURE OF TRANSACTION NETWORKS components from external suppliers, but they do not sell components to external customers. In the absence of vertically permeable firm boundaries, interfirm transaction cycles do not emerge. This contrasts with the practice of major electronics firms, discussed next. Modularity of the Electronics Sector. In contrast to automobiles, since the 1980s, electronics products, e.g., computers, communications, and consumer electronics, have been designed as modular systems with largely standardized components (Whitney, 1996; Baldwin and Clark, 2000; MacDuffie, 2008). Modularity in turn has implications for both transaction costs and rates of innovation in this sector, which affect the benefits and costs of vertically permeable boundaries. First, with respect to transaction costs, because of their standardized interfaces, the behavior of electronic components does not change when they are used in different systems, as long as some design rules are obeyed. Thus in contrast to automobiles, the design and production of many electronic components, such as memory chips and batteries, can be conducted without detailed knowledge of the products in which they are used (Whitney, 1996). This low level of component specificity results in low asset specificity between suppliers and customers (Williamson, 1975; 1981). Arm’s-length contracts and/or spot-market transactions are economical in this setting and this in turn allows intermediate markets to form. There is no need for the long-term, hand-in-glove relational contracts that are common in the automotive sector. Second, the modularity of electronics allows independent or unsynchronized development activities, which in turn lead to high rates of innovation in modules (Henderson and Clark, 1990; Teece, 1996; Baldwin and Clark, 2000). In this respect, Koh and Magee (2008) showed that information technologies achieved much higher rates of performance increase than energy technologies over the past 100 years. High rates of innovation in modules result in short product life cycles for both components and systems that combine components in new ways (Klepper, 1997; Tushman and Murmann, 1998). Indeed all of the major electronics manufacturers that we interviewed identified “short product life cycles” as the major challenge facing their business. (In contrast, our interviewees in the automotive sector viewed technological innovations as rare and their sector as slow-paced.) Short product life cycles create volatile demand—what customers want this year is not the same as last year. Such conditions favor specialized component suppliers with larger production scales and higher product development speeds than may be found in the internal 25 THE ARCHITECTURE OF TRANSACTION NETWORKS component divisions of vertically integrated firms. Thus, as our interview results indicate, large electronics firms, which were formerly vertically integrated, now procure components from external suppliers (including the component divisions of other vertically integrated firms) if the latter can offer better performance, price, and quality than their internal divisions. Paprzycki (2005) also observed that since the 1990s, Japanese electronics firms have increasingly outsourced components from independent suppliers or the component divisions of competitors. At the same time, the internal component divisions of such firms strive to sell components to external customers in order to increase scale economies, use capacity more effectively, and benchmark their efficiency and product quality to industry standards. For instance, a large vertically integrated electronics firm that we interviewed operates a so-called Industrial Marketing Group at corporate headquarters, which focuses on promoting component product sales in intermediate markets. Such efforts are not observed in the automotive sector. Thus the vertical boundaries of the largest Japanese electronics firms are permeable in both directions. Such firms not only procure components from external suppliers but simultaneously sell components through the intermediate industrial markets. As shown in the previous section, such firms account for the vast majority of cycles in the sector’s transaction network: removing the ten largest firms increases the network’s degree of hierarchy (h) from below 60% to over 93%. If scale economies are crucial and transaction costs are low in the electronics sector, why do the largest electronics firms not move further in the direction of vertical disintegration? In other words, why do they continue to maintain internal component divisions? The rationale offered by our interviewees was essentially that knowledge complementarities across vertical stages of the value chains gave their firms dynamic capabilities, enabling them to create new systems products and respond rapidly to changing market conditions (Teece and Pisano, 1994; Teece et. al., 1997). The capabilities cited involved the ability to engage in architectural innovation (in addition to modular innovation) in order to generate a variety of new and distinctive system products (Whitney, 1996). That is, the largest electronics firms continually strive to create differentiated system products including PCs, cameras, mobile devices, and TVs, for end-user markets. Historically, the Japanese electronics sector as a whole has grown based upon its success in utilizing pre-existing component technologies to create successful system products, such as radios and Walkmans (Nakayama et al., 1999). 26 THE ARCHITECTURE OF TRANSACTION NETWORKS According to our interviewees, knowledge of diverse component-level technologies is an important resource for the rapid development of architecturally novel systems. Maintaining and managing such knowledge resources in house allows the firms to share and deploy these resources more effectively than if they had to be procured outside the firm. This strategic choice is supported by the fact that in electronics, component-level knowledge is generic and coherent (Teece et al., 1994). The same basic knowledge of electronics and electrical engineering can be used across computing, communications, and consumer electronics applications. Neffke and Henning (2010) show empirically that firms are much more likely to diversify into skill-related industries than into skill-unrelated industries. In this respect, one can view the presence of diversified system groups and their component divisions in large electronics firms as a form of skill-related diversification. One of our interviewees, a senior manager at a large electronics firm, confirmed that their pursuit of systems innovations was a partial explanation of his firm’s preference for partial vertical integration coupled with vertically permeable boundaries, as opposed to the simpler strategy of full vertical disintegration. Another interviewee claimed that his firm’s core competence was the capability to leverage component-level resources and knowledge to create diverse system products for end-user markets. At the same time, the component divisions maintained their own capabilities and realized scale economies by competing in intermediate markets via vertically permeable boundaries. In a nutshell, facing short product life cycles, low transaction costs, and low asset specificity, but convinced of knowledge complementarities between system and component divisions, a number of large Japanese electronics firms have adopted a strategy of partial vertical integration with vertically permeable boundaries. According to several interviewees at these firms, a robust stream of innovative system products is made possible through the firms’ control of complementary assets, knowledge and capabilities with respect to components. Vertically permeable boundaries in turn increase the viability of the component divisions by exposing them to competition. This strategic choice by a relatively small number of large firms in the electronics sector has led to the emergence of cycles and a lower degree of hierarchy in this sector’s interfirm transaction network than in the automotive sector’s network. 27 THE ARCHITECTURE OF TRANSACTION NETWORKS 5.3 Physical Limits to Modularity As explained in the previous section, different degrees of modularity in automotive and electronic products affect asset specificity, transaction costs, and the rate of innovation in the two sectors. But modularity is also to some extent a strategic choice (Baldwin and Clark, 2000; Henkel and Baldwin, 2010). Thus why are the technologies of one sector more modular than the other? Whitney (1996) suggests that strategic choices of modularity are subject to physical limits. Specifically, in an automobile, significant energy is processed, and significant power is involved in the functioning and interaction of components. Indeed high power is needed to fulfill the automobile’s basic function—to transport humans and goods of a given mass at high speeds. High-powered systems, like automobiles, incur difficult-to-anticipate side effects, such as heat and vibration, which can only be addressed through interdependent, iterative design processes. In contrast, electronics products, such as computers, phones and other devices, process information via low-powered signals. Lower power causes less severe and less frequent side effects, hence the decomposition of low-powered systems into modular components is relatively easy (Simon, 1962). Furthermore, many forms of information, such as interfaces, can be standardized. Standardized interfaces in turn support information hiding between different parts of a system, which is a pre-requisite of modularity (Parnas, 1972; Baldwin and Clark, 2000). From these arguments, it follows that strategic initiatives to pursue modularity and standardization may face inherently more difficult technical challenges in the automotive sector than in the electronics sector (Whitney, 1996; Takeishi and Fujimoto, 2001; MacDuffie, 2008). Thus we can trace a line from the technological properties of artifacts (high or low power), to product architecture (integral or modular) to firms’ sourcing strategies (concurrent sourcing or vertically permeable boundaries) to the structure of interfirm transaction networks (purely or partially hierarchical). However, firms’ beliefs and perceptions enter this chain of logic as well— at the point where the firms formulate their strategies. While low transactions costs and short product life cycles are consequences of modularity, and higher transaction costs and longer life cycles are consequences of integrality, how best to prosper in these environments is open to debate. Our analysis reveals and our interviews confirm that the major Japanese electronics firms have by and large rejected purely focused strategies in favor of partial vertical integration combined with vertically permeable 28 THE ARCHITECTURE OF TRANSACTION NETWORKS boundaries. Their espoused theory is that knowledge complementarities between component and systems designs allow them to generate a stream of novel systems products, while vertically permeable boundaries allow their component division to realize economies of scale and stay close to the state of the art. In the automotive sector, our analysis reveals and our interviews confirm that the major automakers have rejected both pure vertical integration and vertically permeable boundaries in favor of concurrent sourcing and long-term relational contracts with external suppliers. Their espoused theory is that such relationships support component specificity and have continuously improved quality, cost and delivery in the sector. They did not explicitly consider vertically permeable boundaries as a potential strategy, possibly because the scale benefits are low and potential hazards in terms of intellectual property leakage are too high. 6 Conclusion In this research, we have empirically measured the variation in hierarchy in the industry architectures of the automotive and electronics sectors in Japan, and explained this variation in terms of product modularity and firm strategy. This research hopefully points the way to new approaches to analyzing industry architectures and understanding the strategic choices of firms participating in business ecosystems. Business ecosystems, that is, groups of firms linked by transactions and complementarities that jointly provide complex products to end users, are an increasingly important form of economic organization. At the same time, transactions are the most basic form of interfirm relationship. However, as indicated, comparatively little is known about how individual transactions between firms in business ecosystems become aggregated into interfirm transaction networks. This paper has sought to address this gap, and in the process has made four contributions to the literature on strategy and industry architecture. First, we defined “flow hierarchy,” showed how the concept applied to transactions, and applied a metric that can be used to measure the hierarchy of transaction networks. Second, we used standard network tools in conjunction with the hierarchy metric to analyze transaction data from the Japanese automotive and electronic sectors and showed that they were significantly different. Third, we traced the observed differences in transaction network structure to differences in the largest firms’ strategies with 29 THE ARCHITECTURE OF TRANSACTION NETWORKS respect to vertical integration, concurrent sourcing and vertically permeable boundaries. Fourth, we showed how the firms’ strategic choices were conditioned by the modularity of their products and influenced by their managers’ beliefs about how best to nurture and maintain the firms’ dynamic capabilities. Like all studies, this one has significant limitations. First, because of the difficulty of obtaining good transactions data, we are able to analyze only two sectors in the same country in close years. If data for other sectors with different technological bases, cultures (e.g. American, European, Chinese), and in different stages of evolution were collected, our empirical results and our theoretical reasoning would be better tested. However, it is difficult to collect and compile sector-wide data on transactions: many firms are unwilling to share information on their suppliers and customers. We hope that as interest in industry architecture, ecosystems and transaction networks rises, new data sources will become available. A second limitation of this study pertains to our understanding of firms’ strategies. We interviewed a relatively small number of managers in the two sectors. In terms of positions, they were well-placed to understand their firms’ strategies, and their accounts were consistent. However, they may have been expressing personal views and not representing their own firm’s strategic intent. It is also possible that the firms’ espoused strategies were simply ex post justifications of their path-dependent histories. Related to the problem of path-dependence, the network structures we observed may have been transient, non-equilibrium patterns. This is more likely to be a problem in the electronics industry, for which we have only a snapshot year (1993), than in the automotive industry, where we have three observations spanning almost twenty years (1983-2001). We relied on the literature (Paprzicki, 2005) and our interviews (conducted in 2009) to show that interfirm transaction cycles still exist in this sector. However, the Japanese electronics sector may become more hierarchical as it continues to evolve and mature. For example, some large Japanese electronics firms seem to be loosening their ties with internal component divisions and forming joint ventures, such as Renesas,14 to focus on component technologies. If this trend toward vertical disintegration persists, the Japanese electronics interfirm transaction network will 14 Renesas Technology Corporation was spun off from Hitachi (55%) and Mitsubishi (45%) Electric in 2003, and aimed to make the research, development, and production of semiconductors more economic and efficient (Renesas News Release, April 1, 2003). In 2009, NEC electronics and Renesas agreed to merge and establish the world’s third largest semiconductor firm (Renesas News Release, April 27, 2009). 30 THE ARCHITECTURE OF TRANSACTION NETWORKS become increasingly hierarchical in the future. This paper opens up several avenues of potential future research. First and foremost is the opportunity to seek out new data sources which permit the study of interfirm transaction networks across diverse sectors and over time. In this regard, we think it is especially fruitful to link formal network metrics and tools with qualitative research on strategic reasoning within firms. Simple theories based on technology alone would not have predicted that vertically permeable boundaries are an important dimension of strategy for large Japanese electronics firms nor that transaction cycles would be prominent feature of that sector’s transaction network. Interviews with key managers were needed to elucidate plausible reasons for these empirically documented patterns. In conclusion, we hope this paper may be seen as an invitation to further explore transaction patterns in business ecosystems in order to gain a deeper understanding of the linkages between technology and innovation, firm strategies and capabilities, and industry architecture and evolution. 31 THE ARCHITECTURE OF TRANSACTION NETWORKS Acknowledgments We thank Michael Jacobides, Joel Moses, Oliver de Weck, Joel Sussman, Daniel Roos, Margaret Dalziel, William Mitchell, Lee Branstetter, and Susan Helper, whose insights helped us refine this research, as well as seminar participants at Carnegie Mellon University and at the Academy of Management professional development workshop on Innovation, Firm, and Ecosystem. The authors also thank the International Motor Vehicle Program and the MIT-Portugal Program at MIT for financial support. We particularly thank Takahiro Fujimoto, Daniel Heller, Masanori Yasumoto, and other researchers at the Manufacturing Management Research Center at the University of Tokyo for generous support and help with the field work and data collection in Japan. Finally, the authors thank the executives and managers of the investigated firms for their time, patience, valuable input, and the spirit of knowledge sharing that has allowed a deep understanding of their strategies and the related context and rationales. The authors alone are responsible for any errors and oversights. References Abernathy, W. J., K. B. Clark and A. M. Kantrow (1983), Industrial Renaissance: Producing a Competitive Future for America. Basic Books: New York. Adner, R. and R. Kapoor (2010), ‘Value creation in innovation ecosystems: how the structure of technological interdependence affects firm performance in new technology generations,’ Strategic Management Journal, 31, 306-333. Ahl, V. and T. F. H. Allen (1996), Hierarchy Theory: A Vision, Vocabulary and Epistemology. New York: Columbia University Press. Argyres, N. S. (1996), ‘Evidence on the role of firm capabilities in vertical integration decisions,’ Strategic Management Journal, 17, 129-150. Argyres, N. S. and J. P. Liebeskind (1999), ‘Contractual commitments, bargaining power and governance separability: Incorporating history into transaction cost theory,’ Academy of Management Review, 24(1), 49-63. Asanuma, B. (1989), ‘Manufacturer-supplier relationships in Japan and the concept of relation-specific skill,’ Journal of the Japanese and International Economies, 3(1): 1-30. Baker, G., R. Gibbons and K. J. Murphy (2002), ‘Relational contracts and the theory of the firm,’ The Quarterly Journal of Economics, 117, 39-84. Baldwin, C. Y. and K. B. Clark (2000), Design Rules, Volume 1: The Power of Modularity. MIT Press: Cambridge, MA. 32 THE ARCHITECTURE OF TRANSACTION NETWORKS Baldwin, C. Y. (2008), ‘Where do transactions come from? Modularity, transactions, and the boundaries of firms,’ Industrial and Corporate Change, 17(1), 155-195. Barney, J. (1991), ‘Firm resources and sustained competitive advantage,’ Journal of Management, 17(1): 99-120. Borgatti S. P. (2002), NetDraw: Graph Visualization Software. Analytic Technologies: Lexington, MA. Brandenburger, A. M. and B. J. Nalebuff (1996). Co-opetition, New York: Doubleday. Brusoni, S. and A. Prencipe (2001), ‘Unpacking the black box of modularity: Technologies, products and organizations,’ Industrial and Corporate Change, 10(1):179-205. Castellacci, F. (2007), ‘Technological regimes and sectoral differences in productivity growth,’ Industrial and Corporate Change, 16(6), 1105–1145 Chandler, A. D. (1962), Strategy and Structure. MIT Press: Cambridge, MA. Chandler, A. D. (1977) The Visible Hand: The Managerial Revolution in American Business, Cambridge, MA: Harvard University Press. Christensen, C. M. and R. S. Rosenbloom (1995), ‘Explaining the attacker's advantage: technological paradigms, organizational dynamics, and the value network,’ Research Policy, 24, 233-257 Clauset, A., C. Moore and M. E. J. Newman (2008), ‘Hierarchical structure and the prediction of missing links in networks,’ Nature, 453, 98–101. Coase, R. H. (1937), ‘The nature of the firm,’ Economica, 4(4), 386-405. Colfer, L. and C. Y. Baldwin (2010), ‘The mirroring hypothesis: theory, evidence and exceptions,’ Harvard Business School Working Paper #10-058. Dalziel, M. (2007), ‘A systems-based approach to industry classification,’ Research Policy, 36, 1559-1574. Davis, R. and I. M. Duhaime (1992), ‘Diversification, vertical integration, and industry analysis: new perspectives and measurement,’ Strategic Management Journal, 13(7), 511-524. Dosi, G. (1988), ‘Sources, procedures, and microeconomic effects of innovation,’ Journal of Economic Literature, 26(3), 1120-1171. Eppinger, S. D., D. E. Whitney, R. P. Smith and D. A. Gebala (1994), ‘A model-based method for organizing tasks in product development,’ Research in Engineering Design, 6, 1-13. Evans, D. S., A. Hagiu and R. Schmalensee (2006) Invisible Engines: How Software Platforms Drive Innovation and Transform Industries, Cambridge, MA: MIT Press. Fine, C. H. (1998), Clockspeed: Winning Industry Control in the Age of Temporary Advantage, Reading MA: Perseus Press. Fine, C. H. and D. E. Whitney (1999), ‘Is the make-buy decision a core competence?’ Muffatto, M. and K. Pawar (eds.), Logistics in the Information Age, Servizi Grafici Editoriali, Padova, Italy, pp. 31-63. Also available at http://esd.mit.edu/esd_books/whitney/pdfs/make_buy.pdf 33 THE ARCHITECTURE OF TRANSACTION NETWORKS Fixson, S. K. and J-K Park (2008), ‘The power of integrality: linkages between product architecture, innovation, and industry structure,’ Research Policy, 37(8), 1296-1316. Fujimoto, T. (2007), ‘Architecture-based comparative advantage—a design information view of manufacturing,’ Evolutionary and Institutional Economics Review, 4, 55-112. Galvin, P. and A. Morkel (2001), ‘The effect of product modularity on industry structure: the case of the world bicycle industry,’ Industry and Innovation, 8(1), 31-47. Granovetter, M. (1973) ‘The strength of weak ties,’ American Journal of Sociology, 78(6), 1360-1380. Grant, R. M. (1996), ‘Toward a knowledge-based theory of the firm,’ Strategic Management Journal, 17, 109-122. Gulati, R. (1995), ‘Does familiarity breed trust? The implications of repeated ties for contractual choice in alliances,’ Academy of Management Journal, 38(1), 85-112. Gulati, R. (1998), ‘Alliances and networks,’ Strategic Management Journal, 19(4), 293-318. Helfat, C. E. (1997), ‘Know-how and asset complementarity and dynamic capability accumulation: the case of R&D,’ Strategic Management Journal, 18(5), 339-360. Helfat, C. E. and M. Campo-Rembado (2009), ‘Integrative capabilities, vertical integration, and innovation over successive technology life cycles,’ Working Paper, Tuck School of Business at Dartmouth. Helper, S., J. P. MacDuffie and C. Sabel (2000), ‘Pragmatic collaborations: advancing knowledge while controlling opportunism,’ Industrial and Corporate Change, 9(3), 443 488. Henderson, R. M. and K. B. Clark (1990), ‘Architectural innovation: the reconfiguration of existing product technologies and failure of established firms,’ Administrative Science Quarterly, 35, 9-30 Henkel, J. and C. Y. Baldwin (2010), ‘Modularity for value appropriation: How to draw the boundaries of intellectual property,’ Harvard Business School Working Paper #11-054 (November). Hoetker, G. (2006), ‘Do modular products lead to modular organizations?’ Strategic Management Journal, 27:501-518. Iansiti, M. and R. Levien (2004), The Keystone Advantage: What the New Dynamics of Business Ecosystems Mean for Strategy, Innovation, and Sustainability, Boston: Harvard Business School Press. Jacobides, M. G. (2005), ‘Industry change through vertical disintegration: how and why markets emerged in mortgage banking,’ Academy of Management Journal, 48(3), 465-498. Jacobides, M. G. (2006), ‘The architecture and design of organizational capabilities,’ Industrial and Corporate Change, 15(1), 151-171. Jacobides, M. G. and S. Billinger (2006), ‘Designing the boundaries of the firm: from “make, buy, or ally” to the dynamic benefits of vertical architecture,’ Organization Science, 17(2), 249-261. 34 THE ARCHITECTURE OF TRANSACTION NETWORKS Jacobides, M. G., T. Knudsen and M. Augier (2006), ‘Benefiting from innovation: value creation, value appropriation and the role of industry architectures,’ Research Policy, 35, 1200-1221. Jacobides, M.G. and S.G. Winter (2005), ‘The co-evolution of capability and transaction costs: explaining the institutional structure of production,’ Strategic Management Journal, (26)5, 395-413. Koh, H. and C. L. Magee (2008), ‘A functional approach for studying technological progress: extension to energy technology,’ Technological Forecasting & Social Change, 75, 735-758. Klepper, S. (1997), ‘Industry life cycles,’ Industrial and Corporate Change, 6(1): 145-181 Langlois, R. N. and P. L. Robertson (1992), ‘Networks and innovation in a modular system: lessons from the microcomputer and stereo component industries,’ Research Policy, 21, 297-313. Leontief, W.W. (1951) The Structure of the American Economy, 1919-1939, 2nd ed., Oxford University Press: New York. Luo, J. (2010), Hierarchy in Industry Architecture: Transaction Strategy under Technological Constraints, Doctoral Dissertation, Massachusetts Institute of Technology. Luo, J. and C. L. Magee (2011), ‘Detecting evolving patterns of self-organizing networks by flow hierarchy measurement,’ Complexity, in print, doi 10.1002/cplx.20368 MacCormack, A., J. Rusnak and C. Y. Baldwin (2006), ‘Exploring the structure of complex software designs: an empirical study of open source and proprietary code,’ Management Science, 52(7), 1015-1030. MacCormack, A., J. Rusnak and C.Y. Baldwin, (2010) ‘The Architecture of Complex systems: Do Core-periphery Structures Dominate,’ Harvard Business School Working Paper, #10-059. MacDuffie, J. P. (2008), ‘Technological and organizational barriers to modularity: Persistent integrality in the global automotive industry,’ International Motor Vehicle Program Working Paper, Version 1.5, January 10, 2008 Malerba, F. (2002), ‘Sectoral systems of innovation and production,’ Research Policy, 31, 247-264. Malerba, F. and L. Orsenigo (1993), ‘Technological regimes and firm behavior,’ Industrial and Corporate Change, 2(1), 45-71. Malerba, F. and L. Orsenigo (1996), ‘Schumpeterian patterns of innovation are technology-specific,’ Research Policy, 25, 451–478. Malerba, F. and L. Orsenigo (1997), ‘Technological regimes and sectoral patterns of innovative activities,’ Industrial and Corporate Change, 6(1), 83-118. Murmann, J. P. and K. Frenken (2006), ‘Toward a systematic framework for research on dominant designs, technological innovations, and industrial change,’ Research Policy, 35, 925-952. Nagaoka, S., A. Takeishi and Y. Norob (2008), ‘Determinants of firm boundaries: empirical 35 THE ARCHITECTURE OF TRANSACTION NETWORKS analysis of the Japanese auto industry from 1984 to 2002,’ Journal of the Japanese and International Economies. 22(2), 187-206. Nakano, T. and D. R. White (2007), ‘Network structures in industrial pricing: the effect of emergent roles in Tokyo supplier-chain hierarchies,’ Structure and Dynamics, 2(3), 1. Nakayama, W., W. Boulton and M. Pecht (1999), ‘The Japanese electronics industry,’ Boca Raton, Fla.: Chapman & Hall/CRC Press. Neffke, F. and M. Henning (2010), ‘Skill relatedness and firm diversification,’ DRUID Summer Conference 2010 on Opening Up Innovation: Strategy, Organization and Technology, Imperial College London Business School, June 16~18 Nelson, R. and S. Winter (1982), ‘An evolutionary theory of economic change,’ The Belknap Press of Harvard University Press: Cambridge, MA. Nishiguchi, T. (1994) Strategic industrial sourcing: the Japanese advantage. Oxford University Press, USA. Novak, S. and S. Stern (2009), ‘Complementarity among vertical integration decisions: Evidence from automobile product development,’ Management Science, 55(2): 311-322. Paprzycki, R. (2005), Interfirm networks in the Japanese electronics industry. RoutledgeCurzon: London and New York. Parmigiani, A. (2007), ‘Why do firms both make and buy? An investigation of concurrent sourcing,’ Strategic Management Journal, 28, 285–311. Parnas, D. L. (1972), ‘On the criteria to be used in decomposing systems into modules,’ Communications of the ACM, 15: 1053-1058. Penrose, E. (1959), The Theory of the Growth of the Firm. Basil Blackwell: Oxford. Rosenkopf, L. and M. A. Schilling (2007), ‘Comparing alliance network structure across industries: observations and explanations,’ Strategic Entrepreneurship Journal, 1, 191-209. Sako, M. (1992), Prices, quality, and trust: inter-firm relations in Britain and Japan, Cambridge University Press: Cambridge, UK. Schilling, M. A. (2000), ‘Towards a general modular systems theory and its application to interfirm product modularity,’ Academy of Management Review, 25, 312-334. Schilling, M. A. and C. C. Phelps (2007), ‘Interfirm collaboration networks: The impact of large-scale network structure on firm innovation,’ Management Science, 53(7): 1113-1126. Simon, H. A. (1962), ‘The architecture of complexity,’ Proceedings of the American Philosophical Society, 106, 467-82. Sorenson, O. and T. E. Stuart (2001), ‘Syndication networks and the spatial distribution of venture capital investments,’ American Journal of Sociology, 106(6), 1546-1588 Stefano, B., M. G. Jacobides and A. Prencipe (2009), ‘Strategic dynamics in industry architectures and the challenges of knowledge integration,’ European Management Review, 6, 209-216 Steward, D. V. (1981), ‘The design structure system: A method for managing the design of complex systems,’ IEEE Transactions on Engineering Management, 28, 71-74. 36 THE ARCHITECTURE OF TRANSACTION NETWORKS Sturgeon, T. J. (2002), ‘Modular production networks: a new American model of industrial organization,’ Industrial and Corporate Change, 11(3), 451-496. Stuart T. E. (1998), ‘Network positions and propensities to collaborate: an investigation of strategic alliance formulation in a high-technology industry,’ Administrative Science Quarterly 43 (3), 668–698. Suarez, F. and J. M. Utterback (1995), ‘Dominant designs and the survival of firms’, Strategic Management Journal, 16(6), 415-430. Takeishi, A. and T. Fujimoto (2001), ‘Modularisation in the auto industry: Interlinked multiple hierarchies of product, production, and supplier systems,’ International Journal of Automotive Technology and Management, 1(4), 379-396. Teece, D. J. (1982), ‘Towards an economic theory of the multiproduct firm,’ Journal of Economic Behavior and Organization, 3, 39-63. Teece, D. J., R. Rumelt, G. Dosi, and S. Winter (1994), ‘Understand corporate coherence: theory and evidence,’ Journal of Economic Behavior and Organization, 23, 1-30. Teece, D. J. and G. Pisano (1994), ‘The dynamic capabilities of firms: An introduction,’ Industrial and Corporate Change, 3, 537–556. Teece, D.J. (1996), ‘Firm organization, industrial structure, and technological innovation,’ Journal of Economic Behavior and Organization, 31, 193-224. Teece, D. J., Pisano, G. and A. Shuen (1997), ‘Dynamic capabilities and strategic management,’ Strategic Management Journal, 18(7), 509-533. Thompson, J. D. (1967), Organization in Action. Chicago: McGraw-Hill. Tushman, M. L. and J. P. Murmann (1998), ‘Dominant designs, technology cycles and organizational outcomes,’ Research in Organizational Behavior, 20, 231-266 Utterback, J. M. (1996), Mastering the Dynamics of Innovation, Harvard Business School Press. Wasserman, S. and K. Faust (1994), Social Network Analysis: Methods and Applications. Cambridge University Press: Cambridge; New York. Wernerfelt, B. (1984), ‘The resource-based view of the firm,’ Strategic Management Journal, 5(2), 171-180 Williamson, O. E (1975), Markets and Hierarchy. Free Press: New York Williamson, O. E. (1981), ‘The economies of organization: the transaction cost approach,’ American Journal of Sociology, 87, 548-577. Williamson, O. E. (1985), The Economic Institutions of Capitalism: Firms, Markets, Relational Contracting. Free Press: New York. Wilson, D. (1969), ‘Forms of hierarchy: a selected bibliography,’ In L. L. Whyte, A. G. Wilson and D. Wilson (eds.), Hierarchical Structures, American Elsevier: New York, 287-314. White, H. C. (2002a), Markets from Networks: Socioeconomic Models of Production. Princeton, NJ: Princeton University Press. White, H. C. (2002b), ‘Businesses mobilize production through markets: parametric modeling of 37 THE ARCHITECTURE OF TRANSACTION NETWORKS path-dependent outcomes in oriented network flows,’ Complexity, 8(1), 87-95. Whitney, D. E. (1996), ‘Why mechanical design cannot be like VLSI design,’ Research in Engineering Design, 8(3), 125-138. Whitney, D. E., Heller, D. A., Higashi, H. and M. Fukuzawa (2007). ‘Production engineering as system integrator? A research note based on a study of door engineering and assembly at Toyota Motor Corporation,’ MMRC Discussion Paper #169, the University of Tokyo. 38
Documentos relacionados
50 Key Trends for 2010
economies, we know that the ratio of workers to retirees is shrinking - reducing the pool of pension funds available to serve a rising level of demand. Governments, businesses, the media, society a...
Leia maisThe norms of entrepreneurial science: cognitive effects - ONI
advance as well as invention of devices. These activities involve sectors of the university, such as basic science departments, that heretofore, in principle, limited their involvement with industr...
Leia mais