Our goal is to provide a framework explaining why network governance emerges and thrives. To do so, we integrate transaction costs analysis (TCE) and social network theory. Similar to TCE, we see governance forms as "mechanism[s] for exchange" (Hesterly, Liebeskind & Zenger, 1990:404). In the TCE perspective, three exchange conditions — uncertainty, asset specificity, and frequency — determine which governance form is more efficient. Environmental uncertainty triggers adaptation, which is the "central problem of economic organization," because environments are rarely stable and predictable (Williamson, 1991:278). Asset specific (or customized) exchanges involve unique equipment, processes, or knowledge developed by participants to complete exchanges. This intensifies coordination between parties. Customization combined with uncertainty requires safeguarding exchanges through reducing behavioral uncertainty, which can range from honest disagreements to opportunism (Hesterly & Zenger, 1993). Frequency is important for three reasons. First, frequency facilitates transferring tacit knowledge in customized exchanges, especially for specialized processes or knowledge. Second, frequent interactions establish the conditions for relational and structural embeddedness which provide the foundation for social mechanisms to adapt, coordinate, and safeguard exchanges effectively. Third, frequent interactions provide cost efficiency in using specialized governance structures (Williamson, 1985: 60).

Many of our arguments are based on TCE logic. For a governance form to emerge and thrive, it must address problems of adapting, coordinating, and safeguarding exchanges more efficiently than other governance forms (Williamson, 1991). Less efficient modes of organizing are at a comparative disadvantage and will be selected against in the long run. However, we move beyond TCE in three ways. First, we identify the specific forms of uncertainty and asset specificity that give rise to network governance. Second, we extend TCE by incorporating task complexity (Powell, 1990; Powell, Kogut & Smith-Doerr, 1996) into the explanation of governance form. This is important because it moves the theory beyond a dyadic focus. Third, we show how Williamson’s notion of frequency, which is underspecified and underdeveloped in TCE, provides a link with social network constructs of relational and structural embeddedness (Granovetter, 1992, 1985; Uzzi, 199a, 1996b). Based on TCE and Powell’s work (1990), we identify four conditions necessary for network governance to emerge and thrive (see Figure 1): demand uncertainty with stable supply, customized exchanges high in human asset specificity, complex tasks integrating diverse specialists, and frequent exchanges among parties comprising the network. We discuss these in greater detail below:




How Interaction of Exchange Conditions lead to Structural

Embeddedness and Social Mechanisms in Network Governance


Product Demand Uncertainty with Stable Supply

Environmental uncertainty (also called state uncertainty) refers to an inability to predict future events (Milliken, 1987). The source of this uncertainty can come from suppliers, customers, competitors, regulatory agencies, unions, or financial markets (Miles & Snow, 1978). Understanding the sources of uncertainty is important since these influence what governance form is used to coordinate and safeguard exchanges. Research on environmental uncertainty and governance form shows that even modest levels of supply uncertainty combined with predictable product demand entice firms to vertically integrate (Helfat & Teece, 1987) whereas customers’ demand uncertainty makes vertical integration for firms risky due to obsolescence (Balakrishnan & Wernerfelt, 1986; Mariotti & Cainarca, 1986) or seasonality (Acheson, 1985).

Under conditions of demand uncertainty, firms disaggregate into autonomous units, primarily through outsourcing or subcontracting (Mariotti & Cainarca, 1986; Snow, Miles & Coleman, 1992; Zenger & Hesterly, forthcoming). This decoupling (Aldrich, 1979: 325-26) increases flexibility—the ability to respond to a wide range of contingencies—because resource bundles, now exchanged or rented rather than owned, can be cheaply and quickly reallocated to meet changing environmental demands. For example, the network structure of the textile industry in Prato, Italy, enhanced firms’ ability to respond quickly to changes in fashion (Piore & Sable, 1984: 215). In Japanese automobile keiretsu, decoupling enhanced organizational flexibility as parties learned from one another which reduced lead time and improved quality for new models (product line (Nishigushi, 1994).

Network governance is found in industries with high levels of demand uncertainty but a relatively stable supply of labor such as film, fashion, music, high technology, and construction. Demand uncertainty is generated by unknown and rapid shifts in consumer preferences. This is exemplified in the film industry where it is unclear what makes a film a hit with an audience. "Who knows what the public wants to see? ....I defy anyone to tell me up front how much a picture is going to make — or how much it is going to lose" says David Picker, who as President of United Artists was in charge of the studio’s movie selection (Baker & Firestone, 1972: 29-30).

Demand uncertainty is also generated by rapid changes in knowledge, or technology which results in short product lifecycles and makes the rapid dissemination of information critical (Barley, Freeman & Hybels, 1992; Garud & Kumaraswamy, 1995; Powell & Brantley, 1992; Robertson & Langlois, 1995). In high technology industries such as biotech and semiconductors, new products and technologies leap frog prior products and technologies, leaving participants scrambling to catch up or get left behind.

Demand uncertainty is generated by seasonality which makes vertical integration inefficient as in construction (Stinchcombe, 1959) or the Maine lobster industries (Acheson, 1985). In Maine lobster trapping seasonal fluctuations and wide swings in market prices make predicting both catches and revenues difficult. The region relies on network structures of small firms and individual fishermen rather than vertically integrated firms (Acheson, 1985). In essence, demand uncertainty with stable supply provides conditions amenable to networks and markets but inimical to hierarchies.

Customized Exchanges High in Human Asset Specificity

Customized (or asset specific) exchanges create dependency between parties. For example, if the buyer decides not to purchase the customized product or service, the seller cannot sale or transfer the product or service easily to another (Williamson, 1985). Customizing products or services increases demands for coordination between parties. It also raises concerns about how to safeguard these exchanges since customizing products or services makes both seller and buyer more vulnerable to shifts in markets. Cutomization in conjunction with demand uncertainty increases behavioral uncertainty in two ways: parties may disagree about what the initial customized exchange involved or whether parties will fulfill their initial, agreed upon, obligations now that circumstances have changed. With customized goods or services, exchange parties may try to reduce their dependency on one another. For example, in the mechanical engineering region of Lyons, both clients and subcontractors devised methods to reduce dependency stemming from customized investments such as restricting sales and having the client purchase specialized tools or dies (Lorenz, 1988).

Customization of products or services is common among firms in a network (Miles & Snow, 1992: 55). This form of customization involves human asset specificity (e.g., culture, skills, routines, teamwork acquired through "learning-by-doing", Williamson, 1985) because it is derived from participants knowledge and skills as in semiconductors (Saxenian, 1990), movies (Faulkner, 1987), construction (Stinchcombe, 1959), and process and product improvements in the auto industry (Dyer, 1994; Nishigushi, 1994).

Customized exchanges with high levels of human asset specificity requires an organizational form that enhances cooperation, proximity, and repeated exchanges to effectively transfer tacit knowledge among parties. Cooperation among exchange parties is necessary because parties must work together to gain tacit knowledge. Since "assets" may quit the exchange or diminish their efforts, they are more dependent upon one another’s cooperation to complete the exchange (Coff, 1993). Proximity facilitates transferring tacit knowledge through "information rich" mediums such as face-to-face communication (Lengel & Daft, 1988; Nohria & Eccles, 1992). In the auto industry, resident engineers who are employed by one firm but work at another firm enhance the transfer of knowledge and routines improving product and process quality (Dyer, 1994; Nishigushi, 1992). Repeated exchanges allows tacit knowledge, which cannot be assimilated in short-term interactions, to be assimilated over time. Pisano (1989: 116), in his study of the biotechnology industry, finds that "knowledge about a particular partner and how to collaborate with that partner represents important relationship-specific capital... (which)...becomes deeper for collaborative arrangements encompassing multiple projects than for those involving a single project." Customized exchanges with high levels of human asset specificity are not effectively coordinated by market mechanisms and require either hierarchies or networks.

Demand uncertainty pushes firms toward disaggregation while customized, human asset specific exchanges intensifies the need for coordination and integration among parties. Network governace balances these competing demands by enhancing the rapid dissemination of tacit knowledge across firm boundaries. In Silicon Valley, networks facilitated the rapid deployment of tacit knowledge across semiconductor firms spurring new innovations and markets, creating new ventures and generating revenues ten times that of non-networked Route 128 firms (Saxenian, 1994).


Complex, Time Pressured Tasks

Task complexity refers to the number of different specialized inputs needed to complete a product or service. Task complexity creates behavioral interdependence (Pfeffer & Salancik, 1978:41) and heightens the need for coordinating activities Differing specialists and inputs may result from an increased scope of activities, number of business functions needed, number of products created, or number of differing markets served (Killing, 1988). Task complexity under time pressures makes coordinating through a series of sequential exchanges unfeasible. Time pressures are due to the need to reduce lead time in rapidly changing markets such as semiconductors, computers, film, and fashion, or from the need to reduce costs in highly competitive markets such as autos and architecture. Task complexity in conjunction with time pressures has led to team coordination where diversely skilled members work simultaneously to produce a good or service (Goodman & Goodman, 1976; Faulkner & Anderson, 1987; Van de Ven, Delbecq & Koenig, 1976). Teams coordinate activities through mutual adjustment (horizontal information flows and group meetings) which speeds information sharing among parties and reduces the time to complete complex tasks (Clark & Fujimoto, 1989; Imai, Nonaka & Takeuchi, 1985).

Network governance integrates multiple autonomous, diversely skilled parties under intense time pressures to create complex products or services. The need for speeding products and services to markets is a critical condition for networks (Powell, 1990). In the film industry, the approximate time for film production went from two years in the 1950s to six weeks in the 1970s (Jones & DeFillippi, 1996). Using networks and team coordination in the auto industry to enhance organizational capabilities (e.g., informal and frequent communication between upstream-downstream production units and between work levels) gave the Japanese a competitive advantage over Europeans and Americans who used sequential coordination (Clark & Fujimoto, 1989: 43); these reduced lead times and reduced costs were substantial — 17 hours to assemble a car for the Japanese versus 25 and 37 hours for American and Europeans (Clark & Fujimoto, 1989). Coriat (1995) also argues that automotive firms across the globe are moving toward network governance in an effort to achieve variety under intense time pressures.


Frequent Exchanges Among Parties

Frequency concerns how often specific parties exchange. Although frequent exchange is identified by Williamson (1985) as an important determinant of governance, it is typically "set aside" (p. 293) in TCE. Due to the costs of specialized governance structures, TCE assumes that they used only when there are recurring exchanges (Williamson, 1985: 60). We suggest, however, that frequent exchanges not only justify but enable using interfirm networks as an alternative governance form. Some degree of frequency is important with human asset specific exchanges because human asset specificity results from learning-by-doing (Williamson, 1991:281), "deepens" through continued interaction, and creates exchanges where the "identity" of the other matters (Williamson, 1991: 282). Human asset specificity requires recurring exchanges to transfer tacit knowledge among parties.

In addition, frequency transforms the orientation that parties have toward an exchange and the amount of informal control that can be exerted over exchanges. The frequency of dyadic exchanges are central to notions of embeddedness. Embeddedness explores how dyadic and the overall structure of relations influences economic action and outcomes (Granovetter, 1992). Even Williamson (1975) notes that "Repeated personal contacts across organizational boundaries support some minimum level of courtesy and consideration between the parties…In addition, expectations of repeat business discourage efforts to seek a narrow advantage in any particular transaction…Individual aggressiveness is curbed by the prospect of ostracism among peers, in both trade and social circumstances" (pp. 107-8).

In addition to frequency, reciprocity also results in embeddedness (Uzzi, 1996b). Williamson, although not discussing embeddedness, agrees that reciprocity "transforms a unilateral supply relationship into a bilateral one" (1985: 191) and that where parties share a similar "destiny" greater "mutual interest" is enhanced (1985:155). Thus, TCE logic and analysis is not antithetical to social network notions of of embeddedness.

Granovetter (1992) identifies two aspects of embeddedness: relational and structural. Relational embeddedness captures the quality of dyadic exchanges — the degree to which exchange parties consider one another’s needs and goals (Granovetter, 1992) and the behaviors exchange parties exhibit such as trust, confiding, and information sharing (Uzzi, 1996b). Uzzi’s (1996a, 1996b) recent work provides rich description and measures for illuminating the behavioral and attitudinal orientations of exchange parties in primarily dyadic exchanges or on members’ relational embeddedness. Structural embeddedness — the network’s overall structure or architecture — and how this influences behavior is not described by Uzzi, however. Structural embeddedness provides "more efficient information spread about what members of the pair are doing, and thus better ability to shape that behavior" (Granovetter, 1992: 35). Thus, structural embeddedness, which is discussed more fully in the next section, focuses on social control. This notion of structural embeddedness is akin to Williamson’s notion of "atmosphere" which also emphasizes social control by facilitating "informal group influences," (1975:99), group disciplinary actions, and stronger informal infrastructure (1975:104).

The importance of frequency and reciprocity and their influence on control exchanges are important common ground between TCE and social network theorists, although this common ground is rarely recognized by either. However, a point of difference is that while a social network perspective often takes social structures as a given, TCE is interested in identifying the conditions which give rise to alternative governance forms and the social mechanisms that are employed within them. "A successful social analysis," suggests Aldrich (1982), "cannot take social structures as given, but rather must be able to account for their origins and their persistence" (p. 282). Even Granovetter (1985) notes: "Finally, I should add that the level of causal analysis adopted in the embeddedness argument is a rather proximate one. I have had little to say about what broad historical or macrostructural circumstances have led systems to display the social-structural characteristics they have…" (p. 506). We suggest that by integrating TCE with social network theory, we can enhance our understanding of the origins and persistence of structural embeddedness and social mechanisms which allow network governance to emerge and thrive.


Interaction Effects of Exchange Conditions

No single exchange condition propels the emergence of network governance. A combination of specific conditions are required for network governance to emerge and thrive as an organizational form that offers comparative advantages over markets and hierarchies. These conditions involve high adaptation needs due to changing product demand, high coordination needs due to integrating diverse specialists in complex tasks, high safeguarding needs due to overseeing and integrating parties’ interests in customized exchanges. These exchange conditions inhibit market mechanisms from use because complex, customized exchanges are not adequately safeguarded or efficiently coordinated among parties. The exchange condition of flexibility needed in rapidly changing markets inhibits hierarchies from use, even though they facilitate complex, customized exchanges. Our central point is that network governance balances competing demands of these exchange conditions.

Exchange conditions of complex, customized tasks with recurrent interaction generate structural embeddedness. Complex tasks require that many parties interact to complete a product or service. This enhances the likelihood that mutual contacts evolve rather than strictly bilateral exclusive, exchanges. Customized processes and knowledge intensifies the need for coordinating and safeguarding exchanges among parties and enhances the frequency of interaction so tacit knowledge can be shared. These exchange conditions provide the impetus for the emergence of structural embeddedness. Structural embeddedness creates the foundation for social mechanisms to adapt, coordinate, and safeguard customized, complex exchanges effectively. In industries with these exchange conditions, we should see network governance emerging and thriving more frequently. From this, we derive the following proposition.

Proposition 1: The interaction of exchange conditions — demand uncertainty with stable inputs, customized goods/services requiring high levels of human asset specificity, complex tasks requiring diverse specialists, and frequent exchanges — promotes structural embeddedness among exchange parties.