Acknowledgments 

This research has been made possible by a grant from the Carnegie Bosch Institute. 
We are grateful to Professor Jaeyong Song, School of Business, Columbia University for assisting in the research. 
We wish to thank the participants of seminars at the Strategic Management Society 1997 in Barcelona, Spain and the Carnegie Bosch Institute Conference on Knowledge Management in Rome, Italy for their suggestions and comments. 

  
Are international corporations superior to markets and alliances in facilitating the flow of knowledge between countries? Despite widespread acknowledgement of the superior efficiency of the firm in international knowledge transfer, the theory remains underdeveloped, and empirical support is conspicuous by its absence. This paper has two primary goals. First, to use patent citation data to compare the relative performances of firms, alliances, and markets in the transfer of technological knowledge between countries. Second, to investigate the reasons for the superior capability of the international corporation in facilitating cross-border knowledge flows by examining the mechanisms through which international firms manage international technology transfer. Our findings confirm the superior performance of firms over both alliances and markets as conduits for the flow of knowledge between countries. A more detailed examination of the experiences of five large semiconductor firms suggests that this superiority is the result of its ability to utilize a wide range of knowledge transfer mechanisms flexibly and in combinations with one another, and to embed these transfer mechanisms within a social context that enhances their effectiveness. 
 

International corporations have long been viewed as playing a central role in the development, transfer and exploitation of technology. They have been regarded as the primary conduits for the flow of technology among the advanced industrialized nations and from the industrialized to the developing world. Government policies to attract inward direct investment including generous investment grants and taxation allowances have been based upon the assumption that such investment by multinational companies brings with it an inflow of technology and know-how that can contribute substantially to a nation's rate of technological progress. 

The prominent role of international corporations in international technology flows is a result of the inefficiency of markets as institutions for the transfer of technology, and knowledge more generally. Markets are inefficient due to transaction costs and weak appropriability. As a result, technology intensive industries tend to display high levels of multinationality among companies. Economies of scale and scope in technology and other knowledge assets encourage companies to exploit knowledge across multiple national markets, typically using direct investment as their preferred strategy. The ability to exploit innovations worldwide provides an incentive for international corporations to invest in R&D and other knowledge generating activities. 

Yet, despite a broad consensus concerning the role of the international corporation in the flow of knowledge between countries, little is known, either about the effectiveness of international corporations in technology transfer compared with alternative institutional mechanisms, or about the mechanisms which international corporations use to transfer knowledge. This paper has three main goals: 

• To develop our theoretical understanding of the relative efficiencies of firms as compared with alternative institutional forms, namely markets and alliances, to transfer knowledge internationally.
• To use patent citation data to test the effectiveness of international firms as compared with markets and alliances in transferring technological knowledge across national borders.
• To gain understanding of the mechanisms which international forms use to transfer technological knowledge across countries.

The Knowledge Diffusion Model 

Knowledge has long played a leading role in the theory of the international corporation. The idea of foreign direct investment being driven by firms' technological advantages was pioneered by Stephen Hymer (1959), and was a central theme in analyses of the expansion into Europe by U.S. multinationals (Dunning, 1958; Servan-Schrieber, 1968). These technological advantages included not only technological innovations in products and processes but also superior management systems. The analysis was refined by Caves (1971) who identified the exploitation of indivisible, firm-specific assets as the driving force of multinational expansion. Buckley and Casson (1976) and Teece (1986) extended the 'theory of the firm' to the international dimension noting that excess capacity in technology and other firm-specific resources were an insufficient basis for explaining foreign direct investment, the existence of transaction costs in the exploitation of these assets was a necessary condition for internalization within the multinational corporation. Implicit in this analysis of the multinational corporation is a unidirectional view of knowledge flows. Knowledge is developed within the corporation's home base and is subsequently diffused world wide through a firm's subsidiaries (See Figure 1). 

[Figure 1]

Knowledge Accessing and Integration by the International Corporation 

Over the past decade and a half, this "knowledge diffusion" view of internationalization has given way to a more sophisticated analysis of the role of knowledge within the multinational corporation (MNC). Recent thinking in international strategy views the MNC not just as an exploiter of home country knowledge abroad, but as a vehicle for integrating knowledge located in different parts of the world. This view recognizes that the knowledge-base of the MNC is the result of knowledge created internally by the firm though its own research and development, and knowledge acquired from outside the firm. This outside knowledge is acquired not just within the home base, but from all the locations where the MNC does business. Thus, the MNC creates value from knowledge both from exploiting economies of scale and scope in knowledge from deploying its knowledge assets in multiple geographical markets, and from acquiring and integrating knowledge from different locations. 

In most industries, this knowledge accessing by MNCs in any industry is focused on a few locations. A feature of industrialization, throughout history and as evident today as ever, is the tendency for companies within a particular industry to concentrate around a few locations (agglomeration). Common to Alfred Marshall's (1920) analysis of industrial districts and the more recent analyses by Michael Porter (1991) of patterns of national competitive advantage, is the recognition of the advantages of proximity in the flow of knowledge between firms. If industry-specific knowledge tends to be localized and if each industry has multiple geographical clusters throughout the world, then the multinational enterprise has an advantage over national enterprises in its ability to access knowledge in multiple locations. 

Recent research looks at the multinational as a dispersed network of national subsidiaries which derives advantages by leveraging knowledge throughout its multinational system. This view of the knowledge integrating multinational can be seen in Perlmutter's (1969) seminal work on the 'geocentric' firm and is later captured in Bartlett and Ghoshal's (1989) 'transnational' and Hedlund's (1994) 'hypermodern heterarchy'. The internationalization of technology development implied by the knowledge leveraging model of the MNC is confirmed by a considerable weight of empirical evidence. Across all industry sectors, companies have allocated an increasing proportion of their R&D budgets outside their home countries (Mansfield, Teece & Romero, 1979; Pearce, 1989; Florida, 1997). Studies of the evolution of MNCs have shown that the establishment of overseas R&D facilities to be a critical stage in the development strategies of MNCs' foreign subsidiaries (Hakanson, 1981). Figure 2 depicts the international corporation as an institution for accessing, integrating and leveraging knowledge. 

[Figures 2 ]

The theory of the multinational corporation has directed attention to the superior efficiency of firms over markets in conducting international transactions. This view of the economic system as dichotomized between firms and markets has been required substantial amendment as a result of the growing importance of collaborative arrangements between firms. These arrangements are closer and more stable than arms-length market contracts but are not internalized within individual firms. The distinguishing feature of these "hybrid" forms(Powell, 1987) is their use of "relational" or "incomplete" contracts as a basis for economic coordination (Williamson, 1991). We use the term cross-border strategic alliances to refer to collaboration involving firms in different countries coordinating their activities and sharing resources in relationships which are not completely defined by market contracts. There has been considerable recent interest in the role of strategic alliances in sharing technology and developing new products ( Badaracco, 1991; Nohria and Eccles, 1992; Hagedoorn, 1993). Such hybrid arrangements, whether national or cross-border, have the potential to combine the advantages of both markets and hierarchies: combining the flexibility of market arrangements with the potential for managed economic activity to avoid the transaction costs associated with market contracting. 

While the role of knowledge, technology in particular, has always been central to the theory of the international corporation, the past few years have seen a surge of knowledge management and the role of knowledge in the theory of the firm, both of which have had substantial implications for the theory of the international firm. 

Contributions from Knowledge-based Theory 

Central to the emerging "knowledge-based view of the firm" is the recognition that different types of knowledge have very different implications for economic organization. The most important distinction is between tacit and explicit knowledge. Explicit knowledge, primarily information, factual and scientific knowledge, is capable of articulation and therefore of communication. The costs of transferring explicit knowledge are low and, with the advance of information technology, falling rapidly. Tacit knowledge (or "know-how"), on the other hand, is revealed only in its application. It is not capable of articulation and is characterized by idea that "we know more than we can tell" (Polanyi, 1962). It s transfer is slow and costly, its acquisition requiring careful observation and extensive learning-by-doing. 

Although explicit knowledge is capable of articulation and therefore of communication, its transfer across markets is impeded by severe problems of appropriability. Unlike most commodities, ownership of knowledge is non-rivalrous in nature: knowledge may be sold to a buyer without the seller giving up ownership (Arrow, 1962). Second, the seller cannot establish the value of knowledge without revealing it, and once revealed, the customer can avoid paying for it. Market transactions are only capable of effectively appropriating the returns to explicit knowledge where unambiguous property rights can be established (e.g. through patents and copyrights). The widespread failure of markets in the transfer and integration of knowledge has led Demsetz (1991) to conclude that market are efficient in transferring knowledge only where knowledge is encapsulated in goods and services such that the purchaser's use of the good or service is independent of the knowledge required to produce the good or service. 

Tacit knowledge, or "know-how," is less subject to such risks, primarily because of its limited transferability. Since it is revealed only in its application and is embodied within individuals, it can be transferred by markets either through contracts for its application or through employment contracts which transfer the individual. However, particularly in the case of international transfers, these contracts tend to be subject to significant transaction costs. 

It would appear, therefore, that international firms possess clear advantages in the transfer of knowledge across borders. In the case of explicit knowledge, firms can utilize their internal information systems to transfer explicit knowledge at low costs while using secrecy and other mechanisms to permit appropriability of their explicit knowledge. In the case of tacit knowledge, firms have the opportunity to transfer personnel at lower costs than can international labor markets, and can facilitate tacit knowledge transfer both through training and through the conversion of tacit into explicit knowledge (Nonaka and Takeuchi, 1995). 

However, while firms can avoid many of the costs associated with market transactions, we must recognize that international knowledge transfer within firms is still a complex and costly process. Several studies have explored these costs. The preponderance of evidence reveals two important facts (a) international technology transfer is not a smooth or easy process - it is often incomplete or unsuccessful and (b) the efficiency of technology transfer improves with experience. A detailed investigation into the composition of these transfer costs was undertaken by Teece (1977). Transfer costs comprised (a) pre-engineering technological exchanges, (b) engineering costs associated with transferring the product or process design, (c) R&D personnel costs associated with the transfer, (d) adaptation costs and costs of solving unexpected problems, and (e) pre-start-up and start-up costs including the costs of training and debugging. Teece also pointed out that the or cost of these technology transfers varied with the capability of the firm (experience), the relationship between the transferor and recipient, home and host country factors, and the age of technology. It is interesting to note that the costs of transfer did not appear to be influenced substantially by whether or not the international transfer was within the same country or across separate firms (whether or not linked by joint venture agreements). 

A series of papers by Kogut and Zander (Kogut & Zander, 1992a, 1992b, 1996; Zander & Kogut, 1995), have explored the role of the firm in managing knowledge. Kogut and Zander have placed particular attention on the ways in which the properties of knowledge influence the knowledge management processes within the firm. In a survey of innovations by Swedish companies, Kogut and Zander (1992b), found that the more complex, the less codifiable, and the less teachable the knowledge embodied in an innovation, the more likely that the firm would choose direct investment over licensing to exploit that innovation overseas. 

Kogut and Zander's theory of the multinational corporation rests less on the role of the transactions costs of markets, and more upon the multinational firm's superior efficiency as an organizational vehicle by which to transfer this, often tacit, knowledge across borders (Kogut and Zander, 1992b: 1). They argue that the more tacit is the knowledge base of a company, then the greater the advantage of internalization over licensing in exploiting that knowledge. 

Knowledge Management Processes 

To go further in developing insight into the relative efficiencies of alternative institutional forms, we need to understand more fully how knowledge creates value within the economic system. So far we have focused upon knowledge transfer, but why is knowledge is transferred, to whom, and for what purposes? 

The knowledge-based literature distinguishes two categories of knowledge activities, those which involve increasing the organization's stock of knowledge, which Spender (1992) refers to as knowledge generation, and those which involve the utilization of that stock of knowledge, which Spender calls knowledge application. The knowledge-diffusion model of the MNC adopts a simplified view of both knowledge generation and knowledge application: knowledge generation is equated with knowledge creation through R&D and other internal knowledge generating processes, knowledge application is equated with knowledge diffusion--the international transfer of technology and know-how. The knowledge-leveraging model of the MNC implies a more complex view of knowledge management processes within the firm. The key elements of the following discussion are summarized in Table 1. 
  

Knowledge generation 

Knowledge creation               Creating new (mainly technical and scientific) knowledge through research 

Learning-by-doing                  Accumulating know-how (especially that relating to processes through 
                                                          the accumulation of experience ("moving down the learning curve") 

External learning                   The transfer for of both explicit and tacit knowledge from outside the firm's boundaries 

Knowledge application: 
      Knowledge replication            Duplicating existing organizational knowledge (typically in a different location) in order 
                                                        to exploit economies of scale and scale in knowledge utilization 
        Knowledge integration            Combining different types of knowledge to transform inputs into outputs 
 

(a) Knowledge generation 

The knowledge-based literature has heavily focuses around issues of knowledge generation. Traditionally, this has meant knowledge creation, primarily through invention and innovation driven by research and development. The dominant paradigm envisages a linear development process that goes from scientific discovery to the embodiment of science in inventions, to the commercialization of invention through innovation, and the subsequent diffusion of the innovation. In recent years interest has shifted to firms' acquiring knowledge through organizational learning. The interest in organizational learning has grown out of the work by March (Levitt & March, 1988; March 1991) and Argyris (Argyris & Schon, 1978; Argyris, 1993), among others, and has featured contributions by Senge (1990), Quinn (1992), and Nonaka (1994). 

(b) Knowledge application 

Traditional organizational and management theory has implicitly been concerned with the effective application of knowledge, in particular, the how the structures, systems, and management techniques of an organization can be designed to maximize the efficiency with which it applies its existing body of knowledge. However, this literature has taken little account of the epistemological assumptions that underpin it. For example, scientific management implicitly assumes a particular distribution of knowledge between the manager and the worker, and the theory of bureaucracy implicitly assumes that knowledge is capable of moving vertically up and down the organization. 

Among practicing managers there is increasing recognition that the major sources of value creation from knowledge management are likely to be in knowledge application rather than knowledge generation. As one of our interviewees observed: "In terms of our ability to boost our bottom line, the biggest gains are likely to come from increasing the effectiveness with which we access and utilize the know-how that already exists within the company." 

Value creation through knowledge application depends upon two key factors: the extent to which the firm is able to fully utilize its knowledge assets through replicating them, and the extent to which it can integrate its knowledge into the production of goods and services, Let us briefly consider each of these processes. 

Efficiency of knowledge utilization concerns the extent to which economies of scale in knowledge are exploited through its replication. These economies of scale derive from two key properties of knowledge: first, it is not exclusionary in use, it can transferred to another person without its possessor being deprived of it, second, it can be transferred at a lower cost than the original cost of creating it. Economies of scale are especially important in information and scientific knowledge which can typically be transferred at a very low cost. In the case of know-how, transfer costs are much higher, since tacit knowledge transfer requires learning-by-doing. As we shall see, exploiting economies of scale and scope in tacit knowledge normally requires that it can be converted into some form of explicit knowledge that can then be readily disseminated throughout the organization (Nonaka and Takeuchi, 1995).. Unless individuals' tacit knowledge can be taught to others or transformed into organizing principles, "the craft shop is forever simply a shop" (Kogut & Zander 1992a: 390). If knowledge cannot be converted into explicit form, then its transfer is much more complex and difficult. Replication requires comparatively long periods of training and learning-by-doing which normally involves the transfer of personnel. Either moving the learners to the experts, or transferring the experts to coach and develop the learners. Thus, in replicating highly complex processes, such as the fabrication of advanced integrated circuits, the building of the plant and the installation of the required equipment and information technology pose few problems: the critical difficulties are in replicating the highly complex set of organizational routines involved in the various stages of semiconductor manufacturer each of which is based upon high levels of individual and collective know how. 

These economies of scale in knowledge are reinforced by the fact that, unlike most other resources, knowledge expands rather than depreciates when it is used. Thus, in utilizing the same knowledge in multiple markets, the MNC also expands the breadth and depth of this knowledge.. 

The replication of existing knowledge in different geographical locations is the essence of the knowledge dissemination process in the MNC. The international expansion of U.S. industrial corporations such as Singer, International Harvester, Ford, and DuPont during the early decades of the 20th century was driven by the desire to replicate product and process technologies and management capabilities in multiple markets. Expansion through geographical replication is the fundamental feature of McDonalds' strategy. Within semiconductors, knowledge replication is especially important in disseminating process technology across multiple plants. However, because so much of the relevant knowledge is tacit and is embodied in complex sets of routines, replication presents a huge challenge. Polanyi's (1962) observation of the failure of Hungarian light bulb plant using standard Western technology and equipment to produce a single flawless bulb is apposite in this regard. 

At the same time, international exploitation of knowledge is not simply about replication, it is also about bringing together different types of knowledge from different locations. As Demsetz (1991) recognizes, a fundamental dilemma for economic organization is to reconcile the efficiencies of specialization in knowledge creation with the need to combine many different types of knowledge in order to produce goods and services. The solution lies in some process of knowledge integration which permits individuals to apply their specialized knowledge to the production of goods and services, while preserving the efficiencies of specialization in knowledge acquisition. 

Despite the emphasis which much of the literature places on learning, both at the individual and organizational level, one person learning what another person knows tends to be a costly way of integrating different areas of specialist knowledge, especially when the knowledge involved is tacit. Hence, a key challenge for any system of production is to establish mechanisms for knowledge integration which support more efficient means of knowledge integration. 

An advantage of firms over markets is their ability to use mechanisms for knowledge integration that are not so readily available to markets. Firms are "social communities in which individual and social expertise is transformed into economically-useful products and services by the application of a set of higher-order organizing principles" (Kogut & Zander, 1992a: 384). The literature points to three main mechanisms for achieving knowledge integration within firms: 

• Direction provides a "low-cost method of communicating between specialists and the large number of persons who are either non-specialists or specialists in other fields (Demsetz, 1991: 172). Firms convert sophisticated specialized knowledge into directives, rules, and operating procedures that can be imposed through authority-based relationships. Product knowledge is integrated through CAD software that comprises a set design rules embodying the tacit knowledge of a large number of engineers, designers, and other experts. Process knowledge is transferred through operating manuals that comprise a complex set of standard operating procedures. ]

• Organizational routines involve complex patterns of coordination that permit different specialists to integrate their knowledge into the production of goods and services while preserving the efficiencies of knowledge specialization (Nelson and Winter, 1982). Organizational routines involve individuals developing complex sequences of actions and interactions such that each is able to coordinate their inputs of expertise, but with little in the way of explicit collaboration or communication. An important role of the firm is providing the continuity of association and geographical proximity and sense of identification among organizational members that supports the development and operation of routines (Kogut & Zander, 1996). A critical issue for the geographically dispersed firm is whether routines can extend beyond a single location. The key is likely to be the existence of a common "organizational platform" comprising social structure, culture, and communication systems.

• Problem solving is a much more communication-intensive form of knowledge integration that is utilized for non-routine types of knowledge integration. Group problem solving involves knowledge integration among a group of individuals for the purpose of resolving problems characterized an absence of applicable rules or routines. In such situations, the organization must use interaction-intensive mechanisms to pool the knowledge of relevant organizational members in order to decide a plan of action. Typically, problem solving operates in tandem with other forms of knowledge integration. Most business processes involve coordination based upon rules, directives and routines, with group problem solving used to make decisions when unfamiliar circumstances emerge. The transition from routine to problem-solving mode is can be highly disruptive, especially in the event of crisis (Hutchins, 1991). Novel design problems and the production challenges of 12 inch wafers, and sub-micron circuitry are likely to require substantial knowledge integration through problem solving modes.

The distinction between knowledge generation and knowledge application is useful analytically in terms of recognizing how knowledge creates value, and distinguishing and understanding the key processes of knowledge management within the firm. However, it must not obscure the fact that the processes of knowledge generation and application are closely linked. The ability to transfer knowledge depends, in part, upon the absorptive capacity of the recipient. This capacity is a function of the recipient's level of knowledge and motivation, both of which depend upon the extent to which the recipient also active in knowledge creating activities. Cohen and Levinthal (1990) show that learning and innovation require the same kinds of technological capabilities and that the two tend to occur as joint processes. 

The close relationship between the processes of knowledge creation and use suggest that these need to be studied as joint processes. This close complementarity is especially important given the tendency for innovation within firms to follow long-term trajectories in which knowledge creation and transfer occur simultaneously. Almeida and Rosenkopf (1997) have described this joint process through which knowledge is created, transferred and further developed within the firm as one of "knowledge building." In knowledge building investments in knowledge creation build the absorptive capacity necessary for knowledge transfer to occur and result in the process where the fusion of the received knowledge with the existing knowledge base provides the impetus for further knowledge creation. This process of knowledge building in which knowledge generation and knowledge application are joint processes is a central theme of our research and is the focus of our empirical testing. 

An obvious implication of this discussion is that the processes of knowledge management in the international economy are complex. New knowledge is being created through R&D, while localized clusters of industry knowledge are created through learning-by-doing and by localized information. Creating value from this internationally dispersed knowledge requires not only replicating existing knowledge in multiple locations, but also integrating the knowledge available in different locations. The implication we draw is that the firm has substantial advantages from its ability to utilize many its organizational characteristics that are not present in either markets or alliances. For example: 

• Firms have access to authority based relationships which facilitate coordination

• Firms are characterized by continuity of association which is conducive to socialization of its members, the building of trust, and the establishment of a common culture and sense of identity, all of which are conducive to the types of knowledge-based processes we have identified above.

At the same time, our analysis suggests that strategic alliances between firms in different countries are likely to more effective in transferring knowledge than pure arms-length relationships. This superiority rests in the collaborative nature of strategic alliances which can help to avoid opportunism through building trust, allow for the effective appropriation of the returns to knowledge, and the potential for deploying a range of management practices which are not easy to establish through arms-length market contracting, 

As a result, we offer the following hypotheses: 

Hypothesis 1: Multinational firms are superior to markets in transferring knowledge across borders. 

Hypothesis 2: Multinational firms are superior to alliances in transferring knowledge across borders. 

Hypothesis 1: Alliances are superior to markets in transferring knowledge across borders. 

The Research Site 

To test the above hypotheses and to explore the mechanisms of cross-border knowledge transfer, we chose the semiconductor industry as our research site. The semiconductor industry is a particularly appropriate arena within which to study international patterns of technology development by MNCs since it is the apotheosis of a knowledge-based industry. The industry had its inception on December 23rd, 1947,when John Bardeen and Walter Brattain demonstrated their experimental transistor to William Shockly at Bell Laboratories. Since then the industry has remained at the leading edge of scientific discovery pushing continually at the limits of the physical sciences not just in electronics, but also in quantum physics, electomagmetics, optics, lasers, metallurgy and materials sciences, chemistry, and lithography. 

The tendency for semiconductor innovation to be strongly science-driven would seem, at first face, to be consistent with a "knowledge diffusion" approach to multinational strategy, with R&D centralized within the firm's home base, and product and process innovations being exploited globally. While elements of this model this model can be discerned among the leading semiconductor companies up until the early 1980s, by the 1990s all the leading companies had moved towards much greater internationalization of their technology development. Table 2 shows the international distribution of research, design and fabrication activities among some of the leading semiconductor producers. 

Table 2. The International Distribution of Semiconductor Activities among Leading Companies 

                                                    North America                            Europe /Middle East                Asia 

Fujitsu 
Research                                                 CA                                                                               Japan 
Design                                                     CA, TX                                                                        Japan 
Fabrication/Assembly                               CA, TX                         Germany (j-v)                           Japan, Singapore 

Intel 
Research                                                 CA, OR, AZ                 Israel 
Design                                                    CA, OR, AZ                 Israel 
Fabrication/Assembly                              CA, OR, AZ, WA         Ireland, Israel                             Malaysia 
NM, Puerto Rico Philippines 

Motorola 
Research                                               TX, CA 
Design                                                  TX, CA, NC                 France, UK, Israel                        Japan 
Fabrication/Assembly                            TX, CA, NC                 France, UK,                                 Japan, S. Korea, 
                                                           Mexico                         Germany                                      Malaysia 

National Semiconductor 
Research                                              CA, ME, TX,                Israel, UK                                    India 
Design                                                 CA, CO, GA                 Israel, Germany                            India 
                                                           WA, MA, AZ                Netherlands 
Fabrication/Assembly                           CA, ME, TX                 Israel, UK                                     Malaysia, China 
                                                                                                                                                Philippines 

Philips 
Research                                            NY                              Netherlands, UK, 
                                                                                            Germany, France 
Design                                                CA                             Netherlands, UK                           Taiwan 
                                                                                           Germany, France 
Fabrication/Assembly                         CA, NM                       Netherlands, UK                            Taiwan, Thailand, 
                                                                                           Germany, France                          Philippines, HK 
                                                                                           Switzerland 
Siemens 
Research                                           NJ                              Germany, Austria 
Design                                              CA                              Germany                                       Singapore 
Fabrication/Assembly                        CA                              Germany, Austria                           Singapore, Malaysia 

Texas Instruments 
Research                                          TX, CA                                                                            Japan 
Design                                             TX, CA, VT                  Italy                                              Japan, Singapore 
Fabrication/Assembly                       TX, CA, VT                  Italy, Germany                                Japan, Malaysia, 
                                                                                         Israel, Netherlands                          Singapore, Taiwan 

The major evidence of internationalization in technology development is in the increased dispersion of semiconductor design activities among the companies. In 1997, all the companies were engaged in designing semiconductors in all three of the major regional bases of the industry: North America, Europe and Asia. Although research has tended, for the most part, to remain highly geographically concentrated in each company's home base, several of the companies had established semiconductor research facilities overseas. This was particularly evident among European and Asian companies establishing research labs in the U.S. In addition, several U.S. companies had established research labs in Europe and Israel (IBM, National, Intel) and Japan (Texas Instruments). 

Yet, ownership of facilities gives little indication of the extent of inventive activity. A better performance indicator is output of patents. Table 3 classifies the U.S. semiconductor patent applications by each of the companies according the location of the inventor. Table 3 shows that, while patenting activity remains heavily concentrated in each company's home region (and home country), the trend over time has been for increased patenting activity outside the home base. While these data strongly support increased internationalization of invention, it is difficult to draw implications regarding either the level of overseas versus domestic invention or cross-sectional differences between companies. 

                      N. America     Europe     Japan     Other Asia     Rest of World     Total no. of patents 

Fujitsu 
    1981-85            0                     0            83             0                     0                                     83 
    1986-90            6                     0           126            2                     0                                    131 
    1991-95            12                   6           222            8                     0                                    248 

Intel 
    1981-85        310                 5             0               0                    0                                     315 
    1986-90        420               10             0               2                   12                                     444 
    1991-95        915                36            6             12                   42                                   1,111 

Motoro 
    1981-85         164                4            0                4                    0                                     172 
    1986-90         290              10            2                4                    8                                     314 
    1991-95         340              18            8              10                   14                                     390 

National Semi. 
    1981-85          333             4             1                0                     1                                        338 
    1986-90          383             0             2                0                     6                                        391 
    1991-95          809            17            3                3                    46                                        878 

Philips 
    1981-85            6          96              0                 0                     0                                      112 
    1986-90           16         144             0                 2                      0                                     162 
    1991-95           44         207             0                11                     1                                      263 

Siemens 
    1981-85             7          110             0                 0                     0                                       117 
    1986-90           28          155             0                 2                     0                                       185 
    1991-95           38          221             0                 8                     0                                       267 

Texas Instruments 
    1981-85          387              5              2                0                      0                                     394 
    1986-90          521             16             4                1                      2                                     544 
    1991-95          820              22            48              11                      5                                    906 

 Note: The data covers semiconductor design and fabrication patents assigned to the above companies. The location of each patent is determined by the location of the inventor(s) as given in the patent filing. 

Source: US Parent Office data from Lexis/Nexus. 

Internationalization of Technology Development 

The pattern of development we observe seems consistent with those that are typical for multinational firms. Initially, the companies' research and most of their design efforts were concentrated within their home countries. Later, fabrication plants were located in overseas countries based upon market proximity and access considerations, and the quest for subsidies, tax breaks, and skilled labor. The result was heavy investment in fabrication plants in east and south-east Asia where all these factors exerted a powerful pull. The final phase has been the internationalization of technology development activities. For most of the leading semiconductor companies, the mid-to-late 1980s and early 1990s saw the establishment of design and, in some cases, research activities in overseas locations. Prior to that time, some Asian and European semiconductor producers had located research and design facilities in the U.S. -- attracted primarily by U.S. leadership in semiconductor research. During the late 1980s and early 1990s, however, there was a much greater move to decentralize knowledge creating activities in order to access localized knowledge in different locations in the world. 

Our discussions with company executives pointed to five major factors encouraging the internationalization of technology development among the semiconductor firms: 

• The role of customer industries. For all the companies, the most important factor driving the dispersion of research and design activities was the desire to be close to leading customers. Computers, consumer electronics, defense, medical electronics and telecommunications have provided the market need that is a driving force behind product innovation in semiconductors. Since these consuming industries are concentrated in different locations, they have encouraged the geographical dispersion of each company's technology development.. For example, as the world center for the computer industry, Silicon Valley has attracted firms in the area of microprocessors and other computer chips. Japanese leadership in consumer electronics, video technologies, and electronic games has attracted most of the non-Japanese semiconductor firms supplying these industries to Japan. The prominence of European companies in telecommunications equipment (wireless communications in particular) has made Europe a center for communications products. The importance of accessing the knowledge base of customer industries has increased as more of the functionality of the final product is built into integrated circuits.

• Universities have also been important in providing a localized source of technological knowledge and attracting firms wishing to access that knowledge base. The role of Stanford University in the development of Silicon Valley is legendary. Cambridge University in England, University of Texas in Austin and Dallas, MIT in Cambridge and the Indian Institute of Science in Bangalore have all had important influences on local semiconductor innovation.

• Mergers, acquisitions, and joint venture have also caused the international dispersal of companies' technology development. These have included U.S. companies acquisitions of European firms, and European and Japanese companies acquiring U.S. semiconductor firms. National Semiconductor's design activity of communication chips in Sweden arose from the initial acquisition of a Swedish company. The semiconductor research and design undertaken by IBM at its facility at Essones, France was influenced by technology joint ventures with European companies.

• Cost reduction through accessing low cost technical resources has been influential in relation to certain counties. For example, the availability of well trained engineers and other technical experts at comparatively low wage cost has been a significant factor encouraging semiconductor companies to locate design and development activities in Israel and India. Government policies, including subsidies and tax concessions, have also influenced relative costs and encourages development as well as fabrication in countries such as Malaysia and U.K.

• Economies from operating across multiple time zones. Several companies recognized the potential for reducing new product development cycles through using design teams located in different time zones. The concept of design teams operating in different time zones working sequentially to permit continuous 24-hour design and development activity. This involves an Asian team transferring its design files to a European or Israeli team, and the European team handing off to a North American team. Although an attractive concept for many semiconductor companies, in practice, few have made substantial progress in this direction. The general view of the managers we spoke to was that the problems of coordination outweighed any time compression benefits. Achieving internationally integrated product development was also hampered by the absence of worldwide licensing agreements for design software The principal economy from multiple time zones appeared to be in better utilization of IT hardware.

This view of firms' multinational strategies seeking access to localized pools of technological knowledge finds considerable support from earlier work by Almeida in the semiconductor industry. Almeida and Kogut (1996) document the phenomenon of knowledge localization in the semiconductor industry by observing, not only the geographical concentration of patenting activity, but, more significantly, by revealing the localization of patent citations. They thus capture the processes through which technology is passed between firms and show that inter-firm knowledge building is strongly linked to geographical proximity. Hence, knowledge in the semiconductor industry tends to remain localized within regions like Silicon Valley. These localized clusters of technological knowledge act as magnets to overseas companies seeking access to these knowledge clusters. Almeida (1996) showed that multinational enterprises in semiconductors are engaged, not only in international knowledge dissemination and application, but also in international accessing and integrating knowledge. Again using patent citation data, Almeida found that the U.S. subsidiaries of foreign semiconductor companies draw heavily upon their local knowledge bases in terms of citing the patents of local companies. 

For the purposes of testing or hypotheses regarding the relative efficacy of alternative institutional forms in knowledge transfer, the semiconductor industry is especially suitable because of the presence of all three institutional forms. Thus, in additional to the extensive internationalization in terms of foreign direct investment, all the leading firms in the industry have long histories of strategic alliances ranging from formal joint ventures to informal collaborative arrangement with both domestic and foreign partners. Among over 1,800 alliances that were recorded within the semiconductor industry between 1961 and 1989, almost 1,200 involved US firms. Many of these arrangements have taken the form of technology exchange and technology sourcing agreements. Griffen (1989) found that most of these agreements resulted in an outflow of technology from the US in semiconductors. 

Patent Citation Data 

Since the pioneering work of Scherer (1965), patent data have been commonly used by economists to illuminate the process of innovation and to evaluate its relationship to technological and economic development. Patent data have received so much attention because they are systematically compiled, have detailed information and are available continuously across time. We use patent data to shed light on the knowledge building patterns of semiconductor firms across borders. 

A patent document contains a host of information including citations to other patents. The list of citations for each patent are arrived at through a uniform and rigorous process applied by the patent examiner as a representative of the patent office. The patent applicant and his or her lawyer are obliged by law to specify in the application any and all of 'the prior art' of which he or she is aware. The list of patent citations so compiled is available on the patent document, along with information on the patenting firm, inventor, geographic location, and technology types. Thus through patent documents, one can infer both organizational and technological influences on a particular invention and thus track knowledge building across people, firms, geographic regions and countries, and time. 

It would be inappropriate to claim that each and every patent citation represents knowledge building, as some citations may be introduced to distinguish the invention from dissimilar ones, or to protect the firm from litigation. While acknowledging this noise in the citation process, we still believe that due to the rigorous and uniform process applied during citation compilation by the patent examiner (unlike the process for academic citations), patent citations allow us to observe overall tendencies of the interfirm knowledge building process and its location in technological, temporal, and geographic space, which can then be traced to the variety of mechanisms associated with this process. The US patent database is useful to examine international knowledge flows since (a) every major player in the semiconductor industry patents extensively under this system, and (b) the system of citations is applied uniformly across firms regardless of national origin. 

The Samples 

We compare the patterns of citations to matched patents belonging to three sets of firms. 

• Sample A (the Multinational Firm Sample) consists of patents belonging to US subsidiaries of international corporations with recorded patenting activity both in the US and in their home country (Japan, Taiwan, Korea, Singapore, Italy, France, Germany, the Netherlands and the UK). There are 24 such international corporations. We selected up to 10 patents, filed before 1991 for each of these firms. The total number of patents selected was 146 (since several of the firms had only a few patents invented in their US subsidiaries).

• Sample B (the Alliance Sample) consists of patents belonging to the domestic plants of US firms. The US firms were chosen such that (a) they have a prior strategic alliance with the corresponding international firm in Sample A and (b) they themselves do not have a subsidiary in the home country of the firm in sample A. Thus if the first patent in Sample A belongs to Siemens in Silicon Valley, the corresponding patent from sample B could be from Valid Logic. Valid Logic has an alliance with Siemens, Germany, but does not have a plant in Germany. The patent from Valid Logic also was chosen in the same technology class and the filed in the same year as the Siemens patent. Thus Sample B is made up of 146 patents corresponding to those in Sample A.

• Sample C (the Market Sample) consists of patents having the same technology class and filed year as the corresponding patent in Sample A & B, but assigned to firms in the US with no formal links to the international corporation or its home country. Thus, continuing the example from the previous paragraph, a patent belonging to Zilog Corp. could be randomly selected since the firm is neither allied with Siemens, nor does it have a plant in Germany. Of course, the total number of patents in Sample C is 146. Thus we have three matched samples.

Statistical Testing 

Let P_A be the frequency probability that the patent from Firm A in the US (international corporation) is cited in its home country. Let PB and PC be the corresponding frequency probabilities that the patents from Firm B (alliance partner) and Firm C (independent firm in the US) are cited in the same country. Assuming binomial distributions, the null hypotheses are 

Ho1: PA = PB, Ho2: PA = PC, Ho3: PB = PC, 
and the alternate hypotheses are 

Ha1: PA > PB, Ha2: PA > PC, Ha3: PB > PC, 
The t statistic for the first hypothesis is calculated as follows 

tAB = (PA-PB)/[(PA(1-PA)+PB(1-PB))/n]0.5 

The 't' statistic tests the difference between two independently drawn binomial proportions. A positive significant value of Student's t indicates support of the proposition. 

Results 

We conducted the t-tests described above in two ways. The first test was conducted at the firm level of analysis. Thus if the patent in Sample A belonged to Siemens, a successful case of cross-border knowledge transfer within the firm was indicated whenever the citing firm was Siemens, Germany. For Sample B, too, a citation by Siemens, Germany, indicated knowledge transfer through the prevailing alliance. For Sample C, a citation by Siemens, Germany indicated knowledge transfer through the market. 

The results of the tests can be seen in Table 4A. The t-tests for all three hypotheses are positive and significant. The multinational firm can be seen to be the most effective institutional arrangement for transferring knowledge abroad. Alliances, while inferior to the firm are superior to the market in their knowledge transfer ability. However, the significance of these results could be diluted by the possibility of a self-citation bias. The number of citations by Siemens, Germany of its US subsidiary could be influenced by the motivation for the firm to cite itself more often than necessary. (It should be noted that this tendency is held in check by the application of rigorous procedures by the patent examiner). 

To counter this possible bias, we also conduct the same t-tests by using the region as the level of analysis. The three regions defined are Japan, Europe and Rest of Asia (Taiwan, Korea, and Singapore). Thus if the patent in Sample A belongs to Siemens, a successful case of cross-border knowledge transfer within the firm is indicated whenever the citing firm is located in Europe. The knowledge transfer is interpreted as having taken place in two stages - first to Germany by the multinational firm and then locally in Germany. Correspondingly, for Sample B, too, a citation within Europe, indicates knowledge transfer through the prevailing alliance to Germany, followed by local diffusion. For Sample C, a citation within Europe indicates knowledge transfer through the market. For all three samples, to remove any possible bias due to self-citations, we omitted any citations by the parent firm (in this example Siemens). 

[A] Firm level analysis of international knowledge transfer 
  
 

	Sample A	Sample B	Sample C	t-value
Hypothesis 1	.4726  (.1618)	.1164  (.0370)		2.16**
Hypothesis 2	.4726  (.1618)		.0410  (.0164)	2.67***
Hypothesis 3		.1164  (.0370)	.0410  (.0164)	1.93*

[B] Regional level analysis of international knowledge transfer 
 

	Sample A	Sample B	Sample C	t-value
Hypothesis 1	1.3082  (.2403)	.6575  (.1261)		2.65***
Hypothesis 2	1.3082  (.2403)		.8013  (.1991)	1.95**
Hypothesis 3		.6575  (.1261)	.8013  (.1991)	-0.64

  

Note: 

1) Number of observations for each sample is 146. 

2) Numbers indicate mean value of matching citations per observation. 

3) Numbers in Parenthesis are standard errors. 
 
The results of the t-tests are shown in Table 4B. There is support for the first two hypotheses indicating the superiority of the multinational firm over both alliances and the market in facilitating cross-border knowledge transfer. However, we do not find support for the superiority of alliances over the market in transferring knowledge across borders. Thus the results of the t-tests conducted at the regional level and the firm level are consistent regarding the superiority of the multinational firm as a facilitator of cross-border knowledge flows. There is however less consistency with the results comparing alliances and the market. The source of these weaker findings for alliances may be two-fold. 

First, while setting up Sample B, we considered all alliances between international firms and domestic firms that may have resulted in knowledge flows. We included joint ventures of various types, sourcing arrangements, technology licensing agreements etc. However, not all these alliances may have been set up with the intention of knowledge exchange, and this could have contributed to the weaker results of knowledge flows through alliances. 

Secondly, patent data though extremely useful in tracking knowledge building across firms, does not capture pure knowledge replication. If a firm only replicates the knowledge of another this knowledge is not patentable. Hence, when alliances result in replication of knowledge across firms (and not knowledge building), the knowledge transfer would not be captured by patent data. 
  

Having established that international corporations are more effective conduits for cross-border technology flows than either alliances or arms-length market transactions, let us seek further insight into the mechanisms by which companies integrate and exploit knowledge across national borders. Our purpose here is not simply to identify and classify knowledge transfer media, but to understand the relative efficacy of different mechanisms and media in terms of the characteristics of the knowledge to be transferred, and ways in which that knowledge is being used to create value within the company. Based upon the discussion in Sections 2 and 3, we view the choice of transfer mechanism as depending upon: 

• The characteristics of the knowledge being transferred

• The use to which the knowledge is being put within the firm

• The organizational context, such as the breadth of knowledge dissemination (in terms of the number of individuals to whom the knowledge is being transferred).

        Internal consultants         Highly tacit, complex and specialized knowledge. Uneconomic to train. 

        Short visits                       Permit observation that assists replication of routines. 
                                               Also fosters interpersonal relationships that supports more socially-arid communication modes 

2. Electronic data exchange Codifiable information with highly standardized format and low ambiguity. 
                                               Efficient for both limited and wide dissemination 

3.Electronic mail                   Though suited to primarily to information transfer, characterized by versatility with regard to 
                                               format and information types and increasingly capable of expressing individuality and subtlety 

4. Groupware                        Characterized by versatility in its ability to transfer and integrate many types of information and 
                                               to provide platform for integration of different individuals' tacit knowledge. 

5. Telephone                         Bilateral voice communication extendible through conference calling. Interactively 
                                              provides opportunities for expressing highly complex and partially implicit knowledge. 

6. Fax                                   Rapid bilateral transfer of small quantities of written or graphical information 

7. Video conferencing         Video capability permits richer inter-personal context conducive to joint problem-solving 
                                            processes 

8. Written reports & manuals Suitable for transfer of complex explicit knowledge which is not readily codifiable into 
                                            simple data 

9. Face-to-face meetings    Rich media for knowledge transfer. Versatility and multiplicity of embodied communication 
                                            modes permit transfer of know-how and contextually-embedded explicit knowledge 

10. Training seminars and courses Effective for transfer of broad-based dissemination of complex explicit knowledge. 
                                            Also useful in building common culture and interpersonal relationships. 

11. Specialist knowledge transfer groups 

            Communities-of-practice   Individuals with shared specialist knowledge maintaining close regular communication unaffected by external and  internal organizational boundaries where commonality of knowledge and norms permits very rich exchanges of know-how. 

            Communities of interest     Individuals with shared interest engaged in bilateral and multilateral information exchange 

12.. Rules, procedures and directives    Knowledge transfer through the translation of functional and general management know-how into simple   rules,  directives, procedures and instructions implemented through authority-based relationships. 

13. Modular integration     The decomposition of a complex system (e.g. products or processes) into a loosely-coupled modular form that permits different organizational members or departments to integrate their specialist knowledge without actual transfer of the   knowledge itself. Examples include CAD-based modular product designs. 

Our principal findings related to the characteristics of the different knowledge transfer mechanisms with regard to four sets of factors: 

• Their capacity for transferring different types of knowledge---explicit and tacit knowledge in particular.

• Their breadth of communication, in terms of their capacity for communicating among a few or many individuals.

• Their richness of communication .

• Their formality.

Knowledge Transfer Capacity in Relation to Explicit and Tacit Knowledge 

Probably the single factor most important in distinguishing different knowledge transfer mechanisms is their capacity for the efficient transfer of different types of knowledge. Where knowledge is so tacit that it is almost impossible to transfer, even with long periods of observation, training, and practice---then there may be no alternative than to transfer knowledge by transferring the experts who possess this knowledge to wherever they are needed. This may involve long-term expatriate assignments, or if the knowledge required is highly specialized and is required only for specific short-term tasks may be offered by geographically-mobile internal consultants. Personnel transfer also takes place in the opposite direction; when it appears that knowledge in transferable in relatively short time (a few weeks or a few months), then the desired recipients of the knowledge may be assigned to the location where the specialist knowledge exists (this is especially prevalent in the case of replicating process knowledge, where learning is best achieved through observation of the process in operation. 

At the other end of the spectrum, highly codifiable information is capable of being transferred at exceptional speed and low cost through electronic transfer. 

The greatest challenge to companies, the semiconductors producers being no exception, is to find ways in which complex, valuable tacit knowledge can be both replicated and integrated quicker and at lower cost. Replication requires either substantial investments in training (both on the job and through seminars and training programs), or the conversion of tacit knowledge into wholly or partially explicit form using rules, procedures and directives. The process of replication is likely require the use of "rich" modes of communication which we shall discuss further below. 

For some processes, the most important being new product development, the difficulties of transferring specialized tacit knowledge can be circumvented by mechanisms which permit knowledge to be integrated through its applications but not necessarily transferred between individuals and departments. The attraction of modular design systems linked by a common CAD software of some other form of standardized structure, is that separate units can apply their specialized (typically tacit) knowledge to specific parts of the product with the need for extensive cross-learning. Although integrated circuits are "tightly-coupled" rather than "loosely-coupled" systems, is still nevertheless possible to disaggregate the design process into modules with some degree of local autonomy. 

Breadth of Dissemination 

The different knowledge transfer mechanisms also differ according to their scope for transferring knowledge to a few or to many organizational members. Some communication modes are inherently limited to bilateral or small-group transfers (e.g. telephone communications), or involve diminishing productivity as the numbers of participants increase (e.g. face-to-face meetings) or they have the potential to reach all organizational members as easily as they can a reach a few (e.g. e-mail). Figure 3 displays the different knowledge transfer mechanisms in relation to their suitability for transferring explicit and tacit knowledge, and breadth of dissemination. 

[Figure 3 ]

Communication media have been distinguished according to their richness in terms of the complexity of language they permit, the extent of information supported, the flexibility of format, the extent they permit personalization (Daft and Lengl, 1986; Daft and Wiginton, 1979), and the extent of interactivity they permit (Weick, 1979). Richer media such as face-to-face conversations obviously allow a deeper exchange of knowledge, and permit clarifications and questioning. These media are especially useful when the context of the knowledge transfer in unclear or the knowledge matter is complex. Leaner media like letters, telephone conversations and faxes are perhaps cheaper and less time consuming but they do not facilitate the same level of knowledge exchange. Thus for a more challenging process of knowledge building rich media are probably the most useful for knowledge transfer. Rich mechanisms of communication can assist not only in the transfer of complex explicit knowledge, but can also help communicate salient aspects of tacit knowledge too. Rich use of language including the use of metaphor, intonation, emotion, and gesture can help convey key aspects of experiential know-how. 

Our interviewees emphasized the importance of rich ("broad bandwidth") communication media for problem solving activities and for knowledge transfer involving experience-based, mainly tacit knowledge. 

The richness of a knowledge transfer mechanism depends not only on the technical characteristics of the media, but also on the context within which the communication takes place, in particular the social relationships between the communicating individuals. The effectiveness of communities of practice in transferring knowledge are less a result of the modes of communication used as of the relationships between the parties. Almost all the design engineers we spoke to confirmed the importance of personal relationships. In communicating with colleagues in Germany, one Siemens engineer working in California observed that before a three week visit to Germany, his e-mail communications with his German counterparts were unproductive. It was establishing personal contact which was the key to providing both the motivation and the social context for more successful knowledge transfer. 

Formality 

The different knowledge transfer mechanisms vary considerably according to the degree of formality they involve and hence their suitability for different types of knowledge transfer. For most transfers of explicit knowledge, a high degree of formality is necessary for successful transfer. One limitation upon the successful knowledge transfer and integration of product development processes across counties has been a lack of standardized rules for the structure and format of data files, and other aspects of standardization and compatibility. Ultimately, all communication requires commonality of language which, to some extent implies a measure of formality. All the companies pointed to different national languages as a key problem in securing improved knowledge transfer. Separate languages reinforced by distinct national cultures was a key force encouraging local autonomy and adding to the costs of multinational integration. Texas Instruments experience with its Tokyo semiconductor research center is particularly interesting in this regard. Commonalties of language also correspond to rules and standards with regard to data and their transfer. The ability to achieve internationally-integrated chip design depends upon the design units in each country agreeing to a common design methodology. For activities involving problem solving and the transfer of tacit knowledge where the problem and the relevant knowledge cannot be readily structured in a precise manner, informal knowledge transfer mechanism---informal meetings and exploratory telephone conversations---were found to be particularly effective. 

Figure 4 shows the various knowledge transfer mechanism in relation to the criteria of richness and 

[Figure 4]

The Companies' Experiences 

In responding to the potential for better utilization of knowledge, all the firms we contacted were making efforts to improve the efficiency with which they transferred knowledge across borders. However, what emerged was a somewhat fragmented approach to international knowledge transfer. While all companies had developed sophisticated systems for internal communication and data exchange, these systems were not within a structure that integrated multiple knowledge transfer mechanisms and related the overall design to the goals of knowledge transfer. In particular, there was only limited evidence of hierarchically-integrated international knowledge transfer mechanisms from top-level planning of international technology strategy down to piecemeal problem solving and exchanges of experiences between individual engineers. 

There was also limited explicit awareness that different types of knowledge and different knowledge management processes require different knowledge transfer mechanisms. The major thrust of the semiconductor companies' efforts orientated around investments in electronic communication ranging from enhanced intranets for e-mail, data transfer, and "groupware" (e.g. Lotus notes) to common databases and data formats, and common design tools and design methodologies. Only limited attention was given to the problem of transferring and integrating tacit knowledge. 

The evidence from engineers pointed to the important of multiple, complementary modes of knowledge transfer. At the level of individual engineers, the backbone of the system for technical communications is the e-mail system. E-mail provides the first point of contact for the individual to identify where knowledge is located and what complimentary knowledge development activities are being undertaken elsewhere. In the case of design projects, e-mail communication between engineers at different locations occurs at more than daily frequency. E-mail was seen as advantageous in terms of speed of transmission, preparation and reply time and because it can be used across time zones. 

However, the productivity of e-mail use was also dependent upon other forms of communication---notably personal visits for face-to--face contact and the use of telephone as a richer medium for probing and clarifying technical information exchanges and engaging in problem solving activities. 

Although most knowledge transfer and communications mechanisms were found to be useful, there was surprisingly little support for the value of video-conferencing. Several commentators regarded visual images as more of a distraction than a support to knowledge exchange. Several engineers found the use of video conferencing was more formalized and therefore not conducive to the relaxed interchange of ideas. The primary value of video conferencing was for group meetings. 

A common observation among the interviewees, both executives and engineers, was that on its own, information and communications technology were not enough. The impersonal nature of electronic communication meant that its effectiveness was enhanced with a social context. A critical element in the effectiveness of communication appeared to be the mutual confidence and respect between the individuals communicating. 

As far as personnel movement is concerned, there were a variety of views and practices across the companies. Prior research on semiconductor engineers has shown that movement from one place to another, even across firms, allows engineers to exploit often tacit knowledge in new locations (Almeida and Kogut, 1996). Hence personnel transfer is an important mechanism for facilitating the movement of tacit knowledge. However, personnel transfer is also the most costly of knowledge transfer mechanisms. Personnel transfer within firms also causes disruption for individuals and work units. Among most of the companies, personnel transfers for short periods of learning and collaboration were being expanded, while long-term overseas assignments were being discouraged. Among IBM, Siemens and TI, the general policy was that expatriate employees should be limited to key personnel and to limited-term assignments. Two considerations appeared to be dominant: first, that each national subsidiary should be staffed by nationals of that country as a means of ensuring national identity, second, the costs of expatriate employees were very high. Conversely, National Semiconductor sees its goal of internationalizing its semiconductor design and development as requiring significant relocations of key technical staff. As part of its goal of building design teams regardless of geographical location it has adopted the policies of, first, rotating staff between design sites, and, second, advertising senior technical positions on a world-wide basis without country preferences. 

Where collaborative projects span several locations, participating engineers made visits to the overseas location quarterly, or in some cases, monthly. Interviewees reported these visits to be invaluable in identifying who possessed what type of knowledge, and building personal relationships. Despite the importance of these meetings, there was little attempt by companies to manage knowledge transfer through face-to-face meetings as part of their systems for knowledge transfer and integration. Even research and product development budgets tended to be implicit rather than explicit over allocation of funds for visits. 

Some of the most interesting opportunities relate to knowledge integration whereby local units and experts are able to input their specialist know-how into projects, but without extensive cross-learning. Through modularized design projects, each national team works on a particular module within a loosely coupled structure and coordinates with other teams on interface issues only. In knowledge management terms, such an approach is efficient since it reconciles the efficiencies of specialization in knowledge creation with the combination of a broad span of knowledge. 

Our paper makes three contributions to international management thought: 

1. We have extended existing ideas concerning the international corporation as a "transnational," a "heterarchy," and an "international integrated network" by specifying more precisely than hitherto the processes through which the international corporation creates value from knowledge. In particular we recognize the complementarity of knowledge generation and knowledge application, and identify the combined process of "knowledge building" which we as fundamental to value creation in the international firm. By outlining the processes through which the international corporation generates and applies knowledge, we can gain a much deeper understanding of the nature and requirements of cross-border knowledge transfer.
2. We provide a direct test of the relative efficacy with which multinational corporations, cross-border strategic alliances, and arms-length, market-based relationships support the process of cross-border knowledge building. Using patent citation data, our analysis provides strong support for the superior performance of international firms over both alliances and markets in transferring knowledge internationally.
3. Our exploration of the mechanisms for cross-border knowledge transfer helps us to appreciate the complexity of this issue and the extent of the management challenge which multinational corporations face in managing international knowledge transfer mechanisms. Our findings provide insight into why multinational corporations are superior to alliances and market contracts in this activity due to the range of mechanisms that are available to them, their capacity for using these mechanisms flexibly and in multiple permutations, and their ability to support these mechanisms by a common organizational culture, a rich social structure, and by authority-based relations. As far as normative implications are concerned, our findings point to the need to match knowledge transfer mechanisms to the process of knowledge generation/application being undertaken, to the characteristics of the knowledge being transferred, and to the organizational context. Our finding point to the advantages of complementary knowledge transfer mechanism, and to the critical role of social relationships. The greatest challenges relate to the transfer of deeply-embedded tacit knowledge and to the creation of mechanisms, which facilitate the integration at a single point specialist knowledge dispersed across multiple locations.

Almeida, Paul, 1996, "Knowledge sourcing by foreign multinationals: Patent citation analysis in the US semiconductor industry", Strategic Management Journal, Special Issue on "Knowledge and the Firm', 7: 155-165. 

Almeida, Paul and Bruce Kogut, 1996, "Technology and geography: The localization of knowledge and the mobility of patent holders", Working Paper, The Huntsman Center for Global Competition and Innovation, The Wharton School, Philadelphia, PA. 

Almeida, Paul and Lori Rosenkopf, 1997, "Inter-firm knowledge building by semiconductor start-ups: The role of alliances and mobility", Working paper, The Wharton School, Philadelphia, PA. 

Argyris, C., 1993, Knowledge for action: Overcoming the Barriers to Organizational Change, Jossey-Bass, San Francisco. 

Argyris, C., and D. A. Schon, 1978, Organizational Learning, Addison-Wesley, Reading MA. 

Arrow, K., 1962, "Economic welfare and the allocation of resources for invention". In National Bureau of Economic Research, The Rate and Direction of Inventive Activity, Princeton University Press, Princeton, pp. 609-625. 

Badaraccio, J.L., The Knowledge Link, Harvard Business School Press, Boston, 1991. 

Bartlett, C.A. and Ghoshal, S., 1989, Managing Across Borders: The transnational solution, Harvard Business School Press, Boston, MA. 

Baldwin, C.Y. and K. B. Clark, 1998, Design Rules: The Power of Modularity, Free Press, New York. 

Buckley, P. and M. Casson, 1976, Alternatives to the Multinational Enterprise, Macmillan, London. 

Caves, R.E., 1971, "International corporations: The industrial economics of foreign investment," Economica, 38, pp. 1-27. 

Cohen, Wesley and Daniel Levinthal, 1990, "Absorptive Capacity: A New Perspective on Learning and Innovation", Administrative Science Quarterly, 35, pp. 128-152. 

Daft, R.L. and R.H. Lengl, 1986, "Organizational information requirements, media richness and structural design," Management Science, 32, pp. 554-571. 

Daft, R.L. and J.C. Wiginton, 1979, "Language and Organization", Academy of Management Review, 4, pp. 179-191. 

Demsetz, H., 1991, "The Theory of the Firm Revisited". In Williamson and Winter (eds.) The Nature of the Firm, Oxford University Press, New York, pp. 159-178. 

Dunning. J.H., 1958, American Investment in British Manufacturing Industry, George Allen and Unwin, London. 

Florida, R., 1997, "The globalization of R&D: Results of a survey of foreign affiliated R&D laboratories in the USA." Research Policy, 26, pp. 85-103. 

Ghoshal, Sumantra, 1987, "Global Strategy: An Organizing Framework," Strategic Management Journal, 8, pp. 425-440. 

Grant, R.M. 1996. Prospering in dynamically-competitive environments: organizational capability as knowledge integration, Organization Science, 7, pp. 375-387. 

Hagedoorn, J. ,1993. Understanding the rationale for strategic technology partnering: Interorgnizational models of cooperation and sectoral differences. Strategic Management Journal, 14, pp. 371-386. 

Hakanson, L., 1981, "Organization and evolution of foreign R&D in Swedish MNCs." Geografiska Annaler, 63B, pp. 47-56. 

Hedlund, G., 1994,"A model of knowledge management and the N-form corporation", Strategic Management Journal 15 (special Issue) Summer, pp. 73-90. 

Hutchins, E., 1991, "Organizing work by adjustment," Organization Science, 2, pp. 71-87. 

Hymer, S., 1957, The international operations of national firms: A study of direct investment, PhD Dissertation, MIT. 

Kogut, Bruce, 1989, "A Note on Global Strategies", Strategic Management Journal, 10, pp. 383-389. 

Kogut, B. & Zander, U. 1992a. Knowledge of the firm, combinative capabilities, and the replication of technology, Organization Science, 3: 383-397. 

Kogut, B. & Zander, U. 1992b. Knowledge of the firm, technology transfer, and the theory of the multinational corporation, Stockholm School of Economics, Institute of International Business, December. 

Kogut, B. & Zander, U. 1996. What firms do? Coordination, identity and learning. Organization Science, 7, pp. 502-518. 

Levitt, B, and J.G. March, 1988, "Organizational learning." Annual Review of Sociology, 14, pp. 319-340. 

Mansfield, E., D.J. Teece, and A. Romero, 1979, "Overseas R&D by US-based firms," Economica, 46, pp. 187-196. 

March, J.G., 1991, "Exploration and exploitation in organizational learning." Organization Science, 2, pp. 71-87. 

Marshall, Alfred, 1920, Industry and Trade, Macmillan, London. 

Nelson, R. and S. Winter, An Evolutionary Theory of Economic Change, Belknap, Cambridge, 1982. 

Nonaka, I., and H. Takeuchi, 1995, The Knowledge Creating Firm, Oxford University Press, New York. 

Nohria, N. & Eccles, R.G. (eds.) 1992. Networks and Organizations: Structure, Form, and Action. Boston: Harvard Business School Press. 

Pearce, R.D., 1989, The Internationalization of R&D by Multinational Enterprises, St. Martins Press, 1989. 

Polyani, M., 1962 , Personal Knowledge: Towards a Post-Critical Philosophy, University of Chicago Press, Chicago. 

Porter, Michael, 1990, "The Competitive Advantage of Nations", Harvard Business Review, March-April, pp. 73-100. 

Quinn, J.B., 1992, Intelligent Enterprise, Free Press, New York. 

Sanchez, R. and J. T. Mahoney, 1996, "Modularity, flexibility and knowledge management in product and organizational design." Strategic Management Journal, 17 (Winter Special Issue), pp. 63-76. 

Senge, P., 1990, The Fifth Dimension, Century Business, London. 

Servan-Schrieber, J., 1968, The American Challenge, Athenaeum, New York. 

Spender, J.-C., 1992, "Limits to learning from the west." International Executive, 34, pp. 389-410. 

Szulanski, G., 1996, "Exploring Internal Stickiness: Impediments to the Transfer of Best Practice within the firm", Strategic Management Journal, 17, pp. 27-44. 

Teece, D.J., 1986, "Transaction cost economics and multinational enterprise", Journal of Economic Behavior and Organization, 7, pp. 21-45. 

Weick, Karl, 1979, The Social Psychology of Organizing, 2^nd, ed., Reading MA: Addison-Wesley. 

Winter, Sidney G., 1987, "Knowledge and Competence as Strategic Assets", in D. Teece (ed.), The Competitive Challenge, Harper & Row, New York. 

Zander, U., 1992, Exploiting a Technological Edge - Voluntary and Involuntary Dissemination of Technology, Institute of International Business. 

Zander, U. and B. Kogut, 1995. Knowledge and the speed of the transfer and imitation of organizational capabilities: An empirical test. Organization Science, 6, pp. 76-92.