Working Papers | Carnegie Bosch Institute | Carnegie Mellon University

THE MANAGEMENT OF KNOWLEDGE IN INTERNATIONAL ALLIANCES: THE ROLE OF COLLABORATIVE PROCESS

Andrew C. Inkpen

The primary question examined in this paper is how firms exploit the knowledge potential of their international alliances. Using data from a longitudinal study of joint ventures between North American and Japanese firms, several collaborative processes emerged as the basis for accessing and exploiting alliance knowledge.
......
Viewing alliances as learning opportunities provides an alternative to mutual alliance value creation. Alliances can provide firms with access to the embedded knowledge of other organizations. This access creates the potential for firms to internalize partner skills and capabilities. Huber (1991) referred to this process as grafting, the process by which organizations increase their store of knowledge by internalizing knowledge not previously available within the organization. In an alliance, two or more organizations are brought together because of their complementarity and their differences. The differences in partner skill areas are the fuel for learning. Whether or not the differences are identified and internalized determines whether learning occurs.
While still rather small, there is a growing body of theoretical research (Kogut, 1988; Westney, 1988) and empirical studies (Dodgson, 1993; Hamel, 1991; Inkpen & Crossan, 1995; Simonin & Helleloid, 1993) addressing the issue of alliances as mechanisms for organizational learning. This paper focuses on the management of alliance knowledge as a key element in the learning process. Given the premise that alliances create learning opportunities for the alliance partners, how do the partners exploit the knowledge potential of their alliances? How is knowledge transformed from an alliance context to a partner context? What is the process used by alliance partners to restructure, recontextualize, and amplify alliance knowledge? Why is the knowledge creation process more effective in some alliances than others. These questions are explored through a longitudinal field study of North American-based joint ventures (JVs) between North American and Japanese firms.
CONCEPTUAL BACKGROUND
The concept of organizational learning refers to the development of skills, knowledge, and associations between past actions, the effectiveness of those actions, and future actions (Fiol & Lyles, 1985). The development of firm skills through a learning process involves the interpretation of past experiences and strategy choices as a basis for present and future actions (Porter, 1991). The knowledge generated through learning supports a firm's ability to understand the consequences of past actions and respond to environmental stimuli. Thus, effective learning results in an enhancement of an organization's skills and capabilities (Levinthal, 1991).
Mintzberg (1990) suggested that strategic initiatives may be left on their own to develop or flounder or they may be championed by managers higher up in the organization who integrate them with elements of existing strategy. The initiatives create experiences, actions, and strategic choices which provide the basis for learning. The focus of this study is on a particular strategic initiative - the formation of an alliance. The alliance experience can be the action that "triggers" learning because it provides new stimuli that may force changes in the mental maps of the organization (Nonaka & Johansson, 1985). An underlying assumption is that managers have some understanding of the causal relationships associated with knowledge, action, and outcomes.
The outcome of the learning process is the capacity for organizational action. When individual knowledge is integrated into a collective knowledge base or organizational memory, the stored information from an organization's history can be retrieved and translated into action (Walsh & Ungson, 1991). Action is represented by the incorporation of managerial experiences into the activities of organizations (Nelson & Winter, 1982). The translation of new knowledge into action is the basis for creating new skills that underpin firm competitive advantage. Thus, as an organization learns, it strengthens and possibly renews its core competence. In turn, core competencies can be seen to represent the collective learning in the organization (Prahalad & Hamel, 1990).
Organizational Knowledge Learning is a dynamic process involving various organizational levels and actors (Inkpen A.C. and Corssan M., 1995). Although much of the learning literature addresses the product or content of learning, of equal importance is the process of learning. A focus solely on content ignores the complex cognitive and behavioral changes that must occur before a learning "outcome" can be identified. To capture the dynamic nature of learning, Nonaka (1994) developed the concept of a spiral of knowledge creation. In the spiral, knowledge moves upward in an organization, starting at the individual level, moving up to the group level, and then to the firm level. As the knowledge spirals upward in the organization, it may be enriched and amplified as individuals interact with each other and with their organizations.
The question of whether or not organizations learn is controversial. Focusing on knowledge creation may provide a more valid foundation for understanding how knowledge travels and changes within organizations (Hedlund & Nonaka, 1993). Clearly, organizations are repositories of knowledge. Whether or not they learn is not the focal issue in this research. The important question is how individual and group interactions contribute to organizational knowledge creation.
JV Knowledge
Knowledge, as opposed to information, is an organized flow of information that is anchored in the commitment and beliefs of its holder (Nonaka, 1994).1 In the JV context, learning and internalization of JV-derived knowledge by a JV parent can be viewed from three perspectives. First, JV knowledge might be used in the design and management of other JVs and alliances (see Lyles, 1988). Simonin and Helleloid (1993) refer to this as collaborative know-how. Second, parent firms may seek access to other firms' knowledge and skills but will not necessarily wish to internalize the knowledge in their own operations. As Hamel (1991) pointed out, knowledge that is embodied only in the specific outputs of the JV has no value outside the narrow terms of the collaborative agreement.
Third, and the focus of this research, knowledge from a JV can be used by the parent company to enhance its own strategy and operations. The acquisition of this type of knowledge, called output knowledge by Westney (1988), has been suggested as one of GM's objectives in its JV with Toyota (Keller, 1989). GM hoped to learn about the efficient production of small cars and transfer its knowledge to GM plants. The focus on knowledge useful to the parent is consistent with the identification of learning as a possible motive for the formation of JVs.
Knowledge Management Following Nonaka (1994) and analogous to the innovation diffusion process (e.g., Tushman & Scanlon, 1981), JV knowledge creation is viewed as a multi-stage process. The first stage begins with the formation of the JV and partner interactions. The interactions and the managers' exposure to partner knowledge may lead to the recognition of partner skill differences embodied in the JV operation. The nature and extent of managerial interactions and the situational factors surrounding the JV should influence knowledge creation that occurs at this stage.
Knowledge acquired from outside the organization can be used strategically only to the extent that it is distributed and interpreted within the organization (Jelinek, 1979). Thus, a second stage involves the internalization of knowledge acquired by individual managers into the parent's collective knowledge base. For this integration to occur, the parents must engage in efforts to transfer partner skill-related knowledge from the JV to the parent. The intensity of the parent firm's learning efforts reflects the degree to which the parent is actively trying to internalize the skills and capabilities of its partner.
This study focuses on the learning efforts and mechanisms used by JV parent firms to capitalize on alliance knowledge. An important assumption is that JV partners have knowledge that at least in part, should be considered valuable by the other partner(s). Thus, the mechanisms used to gain access to partner knowledge and the role played by the collaborative process between the partners are fundamental elements in alliance knowledge management.
METHODS
This study is based on a two stage research design. The first stage established the industry context and basis for the selection of cases for longitudinal study. The second stage used an open-ended approach of grounded theory building (Glaser & Strauss, 1967) to examine the process of alliance knowledge management. This paper is based primarily on the second stage case studies, although the industry, JV, and partner contextual data from stage one were critical in interpreting the case study data.
Because this research was exploratory, I used an inductive process to study knowledge management and creation. The study was exploratory in that key constructs were tentatively identified a priori, but the specific relationships among the constructs and the dimensions of the constructs were not known. For example, technology sharing as a knowledge creation process was identified in the first research stage, but the components and specific types of technology sharing processes were unknown before the study. As well, a preliminary identification of the organizational conditions supporting knowledge creation was possible prior to data collection but the important linkages with the learning processes emerged from the case studies.
Stage One: Context Definition
The initial research stage was designed to provide contextual understanding of the alliance learning issues and to gain a cross-sectional perspective on the basic dimensions of alliance learning. Two research questions were explored: Do alliance parents recognize and seek to exploit alliance learning opportunities? and 2) What organizational and strategic factors play a role in the learning process? The alliance form studied was the JV. For the study, a JV was defined as a means of performing activities in combination with one or more firms instead of autonomously. A JV occurs when two or more distinct firms (the parents) pool a portion of their resources within a separate jointly owned organization. This definition excludes other forms of cooperative agreements such as licensing, distribution and supply agreements, research and development partnerships, or technical assistance and management contracts.
A sample of North American-Japanese JVs located in North America provided the empirical base.2 The primary data collection method for stage one was field interviews with senior managers involved in JV management. Interviews were conducted with 58 managers associated with 40 two-partner JVs. The majority of managers held positions such as JV president or JV general manager and were either employed by the American partners or appointed by the American partners to senior management positions in the JVs.
All JVs were suppliers to the automotive industry and only one had less than 50 percent of its sales to automotive customers. Most were direct suppliers to the automotive assemblers (i.e., tier-one suppliers). With two exceptions, all JVs were startup or greenfield organizations. In terms of ownership, 17 ventures were 50-50, in 15 ventures the Japanese partners had majority equity, and in eight ventures the American partners had majority equity.
Stage Two: Longitudinal Case Studies Stage one of the research confirmed the existence of important learning opportunities for the American JV parents and provided the foundation for an emerging model of alliance knowledge management. Various knowledge management mechanisms were identified. A linkage between JV performance and learning efforts initiated emerged, as did the important role of top management commitment to the alliance and the learning opportunity. The second stage of research was aimed at understanding in detail the knowledge management process that emerged in stage one. An emphasis on process suggested the need for a longitudinal approach that provided deeper and more extensive access to the individuals involved in collaborative exchange.
A multiple case study design was used, based on theoretic replication (Yin, 1989) because the choice of cases was directed by the emerging theory from stage one. To gain some efficiencies in data collection and to capitalize on established industry contacts, the five cases selected were a subset of those from stage one. Several criteria were used to select the cases, with the objective of finding variance across several dimensions. Of particular interest was the learning potential created by the JVs. This factor was important because it influenced the knowledge management mechanisms initiated by the JV parents and the motivation of the American parents to exploit the learning potential. Knowledge management mechanisms initiated were based on a first cut examination of processes including rotation of personnel, regular visits and tours by parent company executives, information sharing between the JV and the American parent, and senior American parent management involvement in JV activities. JV management indicated if senior JV management came from the American parent or if outsiders were hired.
Differences in JV performance, partner history, and the source of JV management were other criteria used in the selection of cases (see Table 1 for case selection criteria). As well, the JVs selected were still in existence because it is better to initiate historical study before the outcomes of strategic change processes become known (Van de Ven, 1992).
Building on the data collected in the first stage, site visits and interviews began in May 1993 and continued to September 1994. The interviews in this stage were usually 1.5 to 2 hours in length, although a few were half day or more.

Data Analysis For all interview data, data reduction began immediately following the interview and helped bring the raw data into a manageable form. Within a 24-hour period, the detailed interview write-ups were completed. The interview write-ups summarized the interviews in a consistent and logical manner. The main objective of the write-ups was a "product intelligible to anyone, not just the fieldworker" (Miles & Huberman, 1984: 50). All write-ups were reviewed for omissions and clarity problems with follow-up data collected if necessary.
As the research progressed the categorized data, write-ups, and memos were examined for emerging patterns, themes, and processes that might account for the frequency and absence or presence of data categories. An objective of exploratory research is the discovery of new categories of data that emerge out of the data rather than having been decided prior to data collection and analysis (Patton, 1987). For example, a pattern that emerged early in the study was a relationship between learning intent and knowledge management.
Knowledge creation processes: As indicated earlier, the first stage resulted in the identification of several processes considered fundamental to knowledge creation. These processes provided an important input to the second stage case studies. As the research progressed, the classification of these processes evolved into the categories discussed in the following section.
Assessing relationships: The analysis yielded data on variables such as performance, prior partner history, and top management involvement in JV management. Over the several years of data collection, relationships emerged between knowledge creation processes and organizational conditions that either supported or created barriers to knowledge creation. Propositions based on these relationships are introduced later in the paper.
RESULTS
Several collaborative processes emerged as the basis for accessing and exploiting alliance knowledge. As an example, Table 2 provides a summary of knowledge management processes in one of the cases. Similar data were collected for each of the cases. Then, using the tentative constructs identified in stage one, four critical processes were identified and expanded on in the current study: technology sharing, interorganizational interaction, personnel movement, and strategic integration. Table 3 classifies the cases on the strength of each process. Each of the processes is discussed in detail.
Examples of Knowledge Management Processes in One JV
1. The American parent (AP) has studied various aspects of the JV's operation, including its use of employee involvement programs, kaizen teams, and its scheduling system. The AP has also studied some of the JV's process innovations, one of which the JV considers proprietary.
2. Several JV Manages were promoted to positions within the AP. One manager was promoted into an AP staff training position at AP HQ. Several engineers have also been promoted.
3. AP senior management is committed to the JV and to an Asian connection. The JV is the strongest Asian connection. The president of the AP has a very close relationship with the former Japanese parent chairman.
4. The AP has formed "teams" to share information on forming and fabrication. The idea is that representatives from the various AP plants (the JV included) will get together regularly and share information.
5. The AP has set up what it calls "gatekeepers" which are units of the company responsible for certain aspects of manufacturing. The gatekeeper is expected to be available to all units of the company on the specific process or technology. The JV was asked to be the gatekeeper for JIT.
6. The AP has several engineers temporarily working in Japan in the Japanese parent organization.
7. The AP and the Japanese parent have initiated a joint engineering project. A piece of manufacturing equipment will be made by the Japanese partner in Japan, with an American engineer visiting Japan during the project period. The equipment will be installed in an AP plant and the AP will pay the JP for the developmental work.
8. More than 15 employees in the JV have visited Japan.

Technology Sharing In all cases, the JVs were manufacturing products functionally similar to those of the American parent. In four cases, the Japanese partners were responsible for implementing the manufacturing process, installing the equipment, and supplying the product technology. Consequently, the JVs provided four American partners with a unique opportunity to study a new, state-of-the art manufacturing organization that would not have been possible without a collaborative relationship. In the fifth case, the American partner had responsibility for the manufacturing process.
There were various explicit mechanisms in place to facilitate technology sharing between the JV partners. The most evident approach was through meetings between JV and parent managers. In one case, monthly meetings were held, with the location alternating between the JV and one of the American parent plants. In attendance at the meetings were plant managers, heads of quality control, R&D managers, the VP manufacturing at the American parent head office, and several senior JV managers. In addition, quarterly R&D meetings were held involving the JV and American parent. In another case, the JV plant manager proposed a training program focused on JIT (just-in-time) and manufacturing process that could be taken to other American parent plants. The proposal was under consideration.
Access to partner technology skills also occurred through direct linkages between Japanese and American partners. In two cases, there were regular visits by American parent personnel to Japanese parent facilities. Consistent with Hedlund and Nonaka's (1993) argument that Western firms find it difficult to undertake activities not fitting prevailing notions of what the company is about, an American parent president expressed frustration at the lack of tangible output from these visits.
Our engineers go to Japan and come back with some good ideas but nothing ever happens. They are too protective of their technology and way of doing things. It drives me crazy when I visit a Japanese partner plant. They are doing the same things we are with one third the employees. I tell our people here but they can't do it.
To capitalize on the Japanese partner's fabrication knowledge and ability to operate with fewer equipment operators, the American president invited several Japanese engineers to the United States to train parent engineers. The Japanese engineers brought very detailed equipment designs that would allow the American firm to replicate their manufacturing process. When no visible progress was made on designing new equipment, the American president decided to contract the design and manufacturing of the equipment to the Japanese partner. An American engineer would be sent to Japan to learn about the equipment so it could be installed in the United States.
In another case, the partners signed a very broad global technology agreement. Both partners agreed to be completely open in sharing both product and manufacturing technology. For product technology there were explicit terms on licensing and royalties. For manufacturing technology there were no established financial terms. For example, the American parent may ask to borrow a Japanese partner engineer for a few weeks. When this had happened in the past there was never any financial considerations involved because "it all comes out in the wash." The American partner recognized the need for reciprocal commitment (Gulati, Khanna, & Nohria, 1994) and tried to make the technology sharing a two-way relationship:
When we give something to the Japanese partner, they will return it ten fold. If we are not coming up with anything, they will not give us anything in return.
Not all the American parents were interested in access to Japanese partner technology. Worried about inroads its Japanese competitor was making with U.S. domestic automakers, a Japanese partner offered to share its manufacturing technology with its American partner. The Japanese partner had developed some proprietary process technology and was willing to share it at no cost. The technology was used in the JV and therefore, was very visible to American partner managers. The offer was communicated in a written memo from a JV manager to the American partner president. The American firm never followed up on the offer. Why was the offer refused? One JV manager's opinion was that "the people from the American parent do not want to learn because they see the JV as an upstart."
Interorganizational Interaction The JV-parent interorganizational domain plays a key role in knowledge management. In addition to the technology sharing initiatives discussed above, various interorganizational mechanisms can create the social context necessary to bring JV knowledge into a wider arena. As well, the institutionalization of day-to-day interaction can help shape the nature of the collaborative process.
One means for parent managers to learn about their JVs was through visits and tours of JV facilities. Most of the JV managers were convinced that many of the differences embodied in the JV were visible and that parent managers would appreciate the differences if they would spend more time in the JV. However, visits were not always utilized effectively, as a JV manager explained:
Plant managers have been invited and some have visited. However, the American parent organization is so lean that these people have little time to invest in learning. . . . A group of 1st line supervisors spent 2 weeks in the JV. They spent time learning about the JV systems and took videos, notes, etc. back to the parent. They went back to the parent plants and nothing happened. . . . The Japanese partner, on the other hand, sends many people to the JV with a learning objective. They are not afraid to ask questions and spend a lot of time in the JV doing that. There are always Japanese people visiting, both from Japanese parent divisions and from Japanese parent world headquarters. It is not always clear what they are here for. Sometimes they just observe, other times they ask a lot of questions.
In several cases, the JV relied extensively on the American parent for various services, such as purchasing, accounting, and human resource management. The JV paid a fee for the services. The result was that certain American parent managers had no choice but to be involved in the JV. It also meant that employees might not differentiate a JV plant from a parent plant because the two used the same systems. In one case, the JV made extensive use of the American parent laboratory facilities.
Customer-supplier relationships between the JV and the American parent also created a basis for extensive, although not always amicable, interaction. In one case, the JV acted as both a supplier and a customer for the American parent. Neither relationship was considered satisfactory, although presumably it was an excellent source of knowledge for the American parent. As a customer, the JV had so many quality problems with the American parent's products that most of the business was shifted to an outside firm. As a supplier, there were also problems. In one instance, the parent asked the JV to carry out a special order because they were behind in their deliveries. The JV refused the business because of concerns about the product quality. The reaction from the American parent was "those people are too inflexible and going too far with the quality issue."
Meetings between managers was also a means of interacting and exchanging information. In one case, monthly sales meetings were held involving all the American parent divisions and the JV. These meetings were designed to ensure that divisions were not competing against each other and to find out how the divisions (including the JV) could work together.
Individual knowledge and perspectives remain personal unless they are amplified and articulated through social interaction (Nonaka, 1994). Thus, JV-parent interactions provide the basis for Brown and Duguid's (1991) concept of evolving communities of practice. Communities emerge not when the learners absorb abstract knowledge but when the learners become "insiders" and acquire the particular community's subjective viewpoint and learn to speak its language. In several cases, JV managers referred to parent managers as "transformed" because the parent and JV managers both saw the learning potential in the JVs. From Brown and Duguid's (1991) perspective, a community of practice had emerged. In another case, a community began emerging because "over time the JV has become grudgingly accepted as more people have been exposed to the JV. Now, there is high regard for what is going on."
Personnel Movement
Personnel rotation can be considered a process of organizational reflection (Hedlund & Nonaka, 1993) and a means of mobilizing personal knowledge. Rotation helps members of an organization understand the business from a multiplicity of perspectives, which in turn makes knowledge more fluid and easier to put into practice (Nonaka, 1994: 29). The rotation of managers through JV positions may encourage the "bleedthrough" of ideas from the venture to the parent (Harrigan, 1985). In this study, rotation meant a two way movement of personnel between the JV and parent. If there is only one way movement, such as from the parent to the JV, this is not considered rotation.
Interestingly, none of the cases studied had an explicit process of rotation between the JV and the parent. However, in four cases, there was an informal system of personnel movement between the organizations. In the fifth case, the parent and the JV had made an agreement that the JV would not "poach" any more people from the parent.
In one case, the American parent had promoted several JV managers to positions within the American parent. One manager was promoted into a parent staff training position at parent HQ. Several engineers also were promoted. In four cases, senior management in the JV were transferred to the JV when the JV was formed. The careers of these managers were considered closely linked to the American parent and not just the JV. In one JV, the Chief Operating Officer of the JV came from the American parent to act as mentor for the younger JV management. This manager will eventually return to the American parent. In another JV, two plant managers have spent time in the JV and then returned to plant management positions in American parent plants.
A constraint on rotation was the attitude of the Japanese parent. In one case, the Japanese parent preferred that JV personnel not move to the American parent. The Japanese parent saw the JV as distinct and separate from the American parent. Despite this concern, the American parent has moved personnel from the JV to the parent. Another constraint is that JV personnel do not always want to return to the American parent.
Strategic Integration
The strategic integration is the process by which an alliance strategy is linked with the strategy of the parent organization (Harrigan & Newman, 1990). Cohen and Levinthal (1990) suggested that effective integration at the learning interface will augment an organization's learning capacity. Thus, a JV perceived as peripheral to the parent organization's strategy will likely yield fewer opportunities for the transfer of alliance knowledge to the parent. A JV viewed as important may receive more attention from the parent organization, leading to substantial parent-JV interaction and a greater commitment of resources to the management of the collaboration. This supports Hamel's (1991) argument that receptivity to learning is enhanced if the parent and its alliance are closely related.
To maximize exposure to partner knowledge, alliance partners must go beyond the narrow confines of the JV agreement. In two cases, the JV functioned like a related division of the American parent, with the parent's emphasis on managing the partner relationship, not just the JV itself. According to the president of one of the American parents:
The JV is treated exactly the same as our other divisions. The JV participates in all our meetings and all of the JV's salaried employees have the same benefits as their counterparts at other divisions. This makes it easier to move people back and forth between the JV and parent.
In this case, the relationship between the partners was getting much tighter. The JV started off strictly as a transplant supplier and relatively independent of its American parent, relying extensively on the Japanese partner for product technology and marketing support. Over the years, the JV became less independent as ties between the two partners increased. A similar situation existed in another case. When the JV was formed it was initially presented as a Japanese company to the transplant customers. The JV evolved into a much less "Japanese" firm and through its American parent's contact, had developed a substantial amount of business with domestic customers. The objective, remarked a JV manager, was for both the JV and the parents to benefit.
MANAGING THE KNOWLEDGE CREATION PROCESS
The previous section considered the question of how firms accessed JV knowledge. The processes discussed are viewed as the outcome of the decision to engage in knowledge management. A firm involved in an alliance has a choice as to whether or not knowledge creation processes should be initiated. The firm also has a choice as to the resources and effort that should be devoted to alliance knowledge creation. Given the premise that alliances can be effective learning vehicles, why is the alliance knowledge creation process more effective in some cases than others? Why are some firms more willing to initiate knowledge creation efforts than others? Addressing this "why" question is the focus of this section. As the basis for identifying propositions, this section explores what Nonaka (1994: 27) described as "organization-wide enabling conditions that promote a favorable climate for effective knowledge creation." The absence of these condition creates obstacles to knowledge creation. In addition to two of the conditions identified by Nonaka, redundancy and creative chaos, the roles of intent, trust, managerial commitment, and performance myopia are considered.
Intent
The collaborative objectives of the JV partners have been identified as a key element in alliance learning. As Hamel, Doz, and Prahalad (1989: 138) asserted, "It is self-evident: to learn, one must want to learn. . . . Western companies must become more receptive [to the benefits of alliance learning]." Hamel (1991), in his inductive model of alliance learning, argued that learning intent was one of the critical learning determinants. When a firm seeks to internalize knowledge from its JV, it can be said to have a learning intent. Hamel (1991: 89-90) defined an alliance learning intent as "a firm's initial propensity to view collaboration as an opportunity to learn".
The findings in this study suggest that initial learning intent may have little impact on the effectiveness of knowledge creation efforts. This is not to suggest that learning intent is unimportant. If a learning intent is associated with the formation of a JV, a parent firm may enter more actively into the search for knowledge. However, if the initial learning intent is not correctly focused, knowledge management efforts may be ineffective. In several cases in the stage one study, the American firm did not have an initial learning intent until skill discrepancies became obvious and unavoidable. For example, an American firm that had prided itself on its high quality product status found its quality lacking when it attempted to supply a Japanese transplant automaker:
The American partner was considered a high quality domestic supplier and had a Q1 [highest quality] rating from Ford. However, in the JV we quickly discovered that to deal with the Japanese you have to be world class; we were only American class. We initially had problems meeting our Japanese customers' quality standards.
In the second stage case studies, JV A (in Tables 1 and 3) had a very explicit technology learning objective. As Table 3 shows, knowledge management initiatives in JV A for three of the four categories were classified as low. In this case, management did not have a clear understanding of its partner's areas of competency, probably because of causal ambiguity (Reed & DeFillippi, 1990) about the Japanese partner's competitive advantage. As the American partner increased its understanding of its Japanese partner, its learning intent weakened along with efforts to manage alliance knowledge.
In JV E, the situation was almost the reverse. The American parent was interested in forming a JV primarily to gain access to the Japanese transplant market. When negotiations to form the JV were started, American parent management made it clear that they were only willing to be involved if they managed the JV. According to the JV president, "we have a quality reputation which we should be able to carry over to the JV." But, after working together for several years, American parent management realized that alliance knowledge could be important to their firm and greater effort was made to gain access to the JV operations and JV partner knowledge. The American parent formed its JV with a weak learning intent that grew stronger with exposure to the JV partner. Thus,
Proposition 1: The parent firms' learning intent at the time of JV formation will be positively related to the initiation of knowledge creation efforts in the JV's initial years of existence but over time, the relationship between initial intent and knowledge creation efforts will dissipate as a new learning intent takes shape.
Trust
Trust between JV partners has been identified as an important element in collaborative relationships (Beamish, 1988; Dodgson, 1993). Breakdowns in the value creation process in JVs often stem from a lack of trust between partners (Borys & Jemison, 1989). Trust between the partners reflects the belief that a partner's word or promise is reliable and that a partner will fulfill its obligations in the relationship. In the knowledge creation process, an atmosphere of trust should contribute to the free exchange of information between committed exchange partners since the decision-makers do not feel that they have to protect themselves from the others' opportunistic behavior (Jarillo, 1988). Without trust, information exchanged may be low in accuracy, comprehensiveness, and timeliness (Zand, 1972).
Trust between the partners appeared to be both a function of top management involvement in the relationship and a history of cooperation prior to the formation of the JV. In one case, a JV manager suggested that the high trust relationship between the "patriarchs" in each partner was critical to the partner relationship. In another case, in response to a question as to the single most important factor in ensuring an enduring partner relation, the American parent president indicated that a long history of cooperation was essential. If there is no history of stable ties and interrelationships between the partners, initial states of trust will be precarious and will contribute to the liability of newness that characterizes joint ventures (Fichman & Levinthal, 1991).
Mutual trust was also important between the JV and the parent as a basis for sharing and cooperating. In both stages of the research, JV managers often indicated that the JVs were viewed by parent middle managers with distrust. In three of the cases studied in the second stage, the JV-parent relationship had evolved into a high trust relationship. In one of the cases, there was a high level of distrust about the nature of the relationship and the motives of the two organizations, as illustrated by the following JV manager statement:
The American parent typically screws up and asks the JV to smooth things over. They [the American parent] cannot meet a commitment. We have helped them lots of times, what have they done for us?
Proposition 2: The level of trust between the JV partners will influence the exchange of information between the partners.
Proposition 2a: The level of trust between the JV and its parent will influence the exchange of interorganizational knowledge.
Managerial Commitment
Hamel (1991) defined a receptive learning environment as one in which top management of the parent firms is enthusiastic about learning and willing to be "taught" by their partners. Top management commitment to the JV and involvement in JV management is viewed as a critical element in exploiting alliance knowledge. If management adopts a passive approach, they may not be in a position to adapt to changes in collaborative roles and circumstances (Gulati, Khanna, & Nohria, 1994).
Top management creates the "organizational intentions" by asking questions on behalf of the entire organization and creating the challenges for intellectual growth (Nonaka, 1994). In one case, the CEO of the American parent joined the parent shortly after the JV was formed. In the JV's initial years, there was a moderate amount of ideas shared between the two firms, primarily because the JV was formed as an offshoot of a licensing agreement between the Japanese and American partners.
Upon joining the American parent, the CEO found a deteriorating relationship between the JV and the American parent. To improve communication, regular "differences meetings" between the two sides were set up. For example, one issue discussed was the American parent's role in performing some intermediate manufacturing for the JV. JV management accused the American parent of poor quality and high prices. After a few meetings, the CEO stopped attending and no more meetings were held. According to the CEO:
The two sides are too different. They are like oil and water. The JV is in a different business. They do not have traditional customer relationships and they make a handful of parts compared to us. Everything takes so long to get done there. They are experts at nit-picking.
From the JV president's perspective, the American parent was aware that there were technology differences between the two firms.
When the American parent people come to the plant they can see the differences but they tend to rationalize them: you have new machines, you have only one customer, etc. The real problem is that their management does not have to deal with the same customer demands as us.
Despite the poor communications between the JV and its American parent, the relationship between the partners was sound. Learning may have been occurring because as Fichman and Levinthal (1991) argued, the persistence of interfirm relationships can be viewed as the result of a sorting process in which firms are learning about each other. Board meetings were amicable and neither partner was actively seeking to end the relationship. A lack of top management commitment was also seen in another case, as indicated by a comment from a JV manager.
The top American partner people come in once or twice a year. They are impressed with the venture and will go back to headquarters and tell their managers: go do this Japanese stuff. The problem is they do not back it up with support. For example, the first plant manager was transferred to Europe and told to "do the Japanese stuff." He put together a proposal that would cost $200,000. The plan died at that point.
Proposition 3: The commitment of top management to the strategic value of the JV will have a positive impact on the degree to which knowledge creation is perceived as a legitimate collaborative activity.
Performance-Induced Myopia
Knowledge creation processes in organizations must cope with confusing experience (Levinthal & March, 1993). One such experience for JV parents was the assessment of JV performance. In the first stage of the research, several managers in the American parent companies pointed to the poor financial performance of the JVs as evidence that learning was not occurring, or could not occur. More generally, a preoccupation with short-term issues was a common characteristic of the American partners. Although it is too simplistic to describe Japanese management as long-term oriented and American management as short-term oriented, the Japanese partner firms in this study appeared to focus on customer satisfaction and product quality rather than profit-based performance. Consistent with other studies, (e.g. Doyle, Saunders, & Wong, 1992), the Japanese firms seemed less constrained by issues of share price and impatient boards of directors than their American counterparts.
Because the American partners were heavily focused on financial performance issues, knowledge creation may have been a secondary and less tangible concern. While North Americans focused on the bottom line, the Japanese focused on improving productivity, quality, and delivery. For American managers, it was difficult to conceive that learning could be occurring in the face of poor performance. Consequently, there was a reluctance to commit to or even try out proposals generated at the JV level. This finding is consistent with Levinthal and March's (1993) argument that organization learning oversamples successes and undersamples failures. As a result, learning processes tend to eliminate failures and sustained experimentation becomes difficult.
Levinthal and March (1993) suggested that learning is a balancing act between the competing goals of developing new knowledge and exploiting current competencies. This balancing act can be seen in the paradoxical challenge faced by JV general managers. On the one hand, general managers were charged with generating an adequate financial return for the American parents; on the other hand, they were expected to act as the conduit for the parent's knowledge creation initiative. A focus on one objective detracted from the other. More importantly, when either learning or performance were less than satisfactory, there were implications for the assessment of the other objective. Thus, while poor performance can act as a barrier to knowledge creation, unexploited learning opportunities may lead to perceptions of unsatisfactory JV performance.
Proposition 4: Poor JV performance will act as a barrier to the initiation of knowledge creation efforts.
Redundancy
Redundancy means the conscious overlapping of company information, activities, and management responsibilities (Nonaka, 1994). Redundancy encourages frequent dialogue and communication. Based on this notion, the knowledge creation mechanisms discussed in the previous section involved elements of redundancy. Much of the discussion revolved around concepts such as sharing, interaction, and integration, all of which imply the transfer of knowledge between individuals.
Managerial tolerance for redundancy was not consistent across the cases. In one case, the regular attendance of JV managers at meetings involving parent division managers could have been seen as redundant given that the JV was initially formed with a narrow mandate to supply one transplant firm. In another example, the American parent president realized that the parent had to make a large commitment in managerial time when the JV was formed if the JV was going to be successful. The result was a JV closely integrated with the parent's strategy and a clear overlapping of roles.
In a low tolerance case, the JV general manager actively promoted the JV as a training ground for parent managers. With the exception of a few instances, the parent was unwilling to incur the minimal expense of sending key parent managers to the JV on a regular basis to experience the JV firsthand. This type of action may have been seen as wasteful and not directly associated with successful JV management. However, as Nonaka (1990) suggested, allowing individuals to enter each others' areas of operation promotes the sharing and articulating of individual knowledge, which may lead to problem generation and knowledge creation. This study found that the Japanese parents frequently took the opportunity to send Japan-based managers to visit the JV, probably because of a greater tolerance for redundancy and because in Japanese firms life-long learning is an explicit element in the career path of Japanese managers (Hedlund & Nonaka, 1993).
Proposition 5: Parent firms will tend to reject the value of knowledge creation efforts when parent managers have a low tolerance for information redundancy.
Creative Chaos
Chaos is created naturally when an organization faces a crisis, such as a rapid decline in performance (Nonaka, 1994). Chaos can also occur when differences or discrepancies disrupt normal routines. Chaos increases tension within the organization and focuses attention on forming and solving new problems. The job of managers in the knowledge creating company is to orient the chaos toward knowledge creation by providing managers with a conceptual framework that can be used to interpret experience (Nonaka, 1991).
The JVs between Japanese and American firms in the automobile supply industry were largely formed in the late 1980s. For many suppliers, this was a time of chaos. With the domestic automakers under pressure from Japanese firms, many suppliers found their traditional customer base shrinking. In one case, the problems in the auto industry strengthened the JV-parent relationship, as a JV manager explained in 1991:
With the downturn in the auto industry, the JV is now starting to beat the other parent plants. They are losing money and the JV is clearly superior in terms of quality and efficiency. The American parent can no longer ignore the differences between the JV and the NA plants.
However, in several of the other cases, there was a great deal of suspicion at the middle management level about why the JVs were formed. According to one JV manager:
There are still people with the attitude "these guys from Japan are not going to show me how to run a JIT plant." It is still hard for Americans to admit that there may be something worth learning from Japanese firms.
The impact of crisis-induced chaos on knowledge creation is difficult to assess. A crisis associated with serious financial problems may not lead to the type of managerial reflection that Nonaka suggested. However, if chaos can be invoked or manipulated creatively by top management, it can be a powerful motivation. In one case, the JV participated in corporate level meetings with other parent divisions. By showing superior quality indicators to parent plant managers, the JV manager was able to send a very powerful signal. In fact, by treating the JV as a related division and encouraging interaction, managers were in a much better position to experience the kind of discrepancies that challenge what is taken for granted. In contrast, a situation described earlier involved conflict over the role of the JV as a parent supplier and customer. While this provided an excellent opportunity to leverage the resultant chaos, parent management chose to use the experience as an excuse for lessening interaction between the parent and JV.
Proposition 6: Discrepancies that disrupt normal routines can motivate parent and JV managers to engage in knowledge exchange.

CONCLUSION
Bounding the Study
This study's findings document knowledge management processes in international alliances and propose new relationships among alliance context, collaborative process, and learning. However, the distinct characteristics of the alliances studied may limit generalizations to other alliances. The industry setting for the study was the automotive industry. Using a single industry with a homogeneous set of organizations imposes certain constraints. In particular, theory development is restricted to limited domain or middle-range theories (Pinder & Moore, 1980) and generalizability is confined to other industries sharing similar structural characteristics. However, a single industry offers greater control over market and environmental peculiarities, an important consideration in exploratory research.
The four key knowledge management processes discussed should be critical regardless of the industry context. However, the nature of the processes and the type of knowledge involved will differ. For example, in four of the five cases, the Japanese partner was responsible for the manufacturing process and product technology, which obviously influenced the type of technology knowledge that could be exchanged between the partners. As well, industry conditions can influence the learning intent and managerial commitment. When this research began in 1990, the U.S. automotive industry was under serious attack by Japanese firms. Suppliers had to cope with new competitors and their traditional customers were losing market share. By 1994, the situation was very different. The domestic auto industry was recapturing some of its market share and suppliers were reaping the benefits. The learning imperative that existed in 1990 was no longer as explicit. This highlights one of the problems with cross sectional research, namely that firms and industries are in constant evolution. Longitudinal research includes the evolutionary patterns in the underlying research context.
The study did not attempt explicitly to link knowledge management with organizational performance. However, a tentative conclusion based on Tables 1 and 3 is that successful knowledge creation and management are positively related to performance. For example, JV A was the poorest performing JV and had the least active knowledge management processes. JV C was classified as high performance and three of four knowledge management activities were also classified as high. In the case of JV A, poor performance was clearly acting as a barrier to the initiation of knowledge management efforts. For JV C, the early profitability of the JV eliminated doubts about the JV's viability, which in turn helped foster greater interaction between the JV and the parent.
A related issue is the question of which knowledge management mechanisms are most important. Table 3 shows five different knowledge management strategies. Does a firm need to be good at all the mechanisms to create knowledge or will an "unbalanced" approach to knowledge management also work? The answer depends on the type of knowledge sought (e.g., tacit vs. articulated) and the strategic value attached to JV knowledge. No two firms will attach the same value to JV knowledge and therefore, may adopt very different knowledge management strategies. An important area for further empirical research is the integration of a typology of organizational knowledge with knowledge management processes. This area has been explored conceptually by Nonaka and Hedlund (1993) but was beyond the scope of the current empirical research.
Summary
Organizations must not only process information but also create new information and knowledge. This paper explored how organizations involved in international alliances can use their alliance experience as the basis for processing and creating knowledge. There are two underlying assumptions. The first is that there can be a significant payoff in cooperating, namely knowledge creation. The second is that knowledge creation is not confined to identifiable points of input-output sequences, but is in fact far more continuous, haphazard, and idiosyncratic. It may occur unintentionally and it may occur even if success cannot be assessed in terms of objective outcomes. Knowledge creation is a dynamic process involving interactions at various organizational levels and an expanding community of individuals that enlarge, amplify, and internalize the alliance knowledge.
NOTES 1. An example might help in the differentiation of knowledge and information. In explaining how the JV differed from its American parent, a JV manager said that there were simple things, such as shipping exactly 7200 parts not 7201. The fact that 7200 parts were shipped is information; the deeper meaning associated with customer satisfaction on quality delivery is knowledge.
2. Although there were two Canadian firms in the sample, for brevity, future references to the sample of firms will use American rather than North American.

REFERENCES
Beamish, P. W. 1988. Multinational Joint Ventures in Developing Countries. London: Routledge.
Borys, B. & Jemison, D. B. 1989. Hybrid arrangements as strategic alliances: Theoretical issues in organizational combinations. Academy of Management Review, 14: 234-249.
Brown, J. S. & Duguid, P. 1991 Organizational learning and communities of practice: Towards a unified view of working, learning, and organization. Organization Science, 2: 40-57.
Cohen, W. M. & Levinthal, D. A. 1990. Absorptive capacity: A new perspective on learning and innovation. Administrative Science Quarterly, 35: 128-152.
Dodgson, M. 1993. Learning, Trust, and Technological Collaboration. Human Relations, 46: 77-95.
Doyle, P., Saunders, J. & Wong, V. 1992. Competition in global markets: A case study of American and Japanese competition in the British market. Journal of International Business Studies, 23: 419-442.
Fiol, C. M. & Lyles, M. A. 1985. Organizational Learning. Academy of Management Review, 10: 803-813.
Fichman, M. & Levinthal, D. A. 1991. Honeymoons and the liability of adolescence: A new perspective on duration dependence in social and organizational relationships. Academy of Management Review, 16: 442-468.
Glaser, B. G. & Strauss, A. L. 1967. The Discovery of Grounded Theory: Strategies for Qualitative Research. New York: Aldine.
Gulati, R., Khanna, T. & Nohria, N. 1994. Unilateral Commitments and the Importance of Process in Alliances. Sloan Management Review, 22 (Spring): 61-69.
Hamel, G. 1991. Competition for Competence and Inter-Partner Learning within International Strategic Alliances. Strategic Management Journal, 12 (special issue): 83-104.
Hamel, G., Doz, Y. L., & Prahalad, C. K. 1989. Collaborate with your Competitors--and Win. Harvard Business Review, 67, 2: 133-139.
Harrigan, K. R. 1985. Strategies for Joint Ventures. Lexington, MA: Lexington Books.
Harrigan, K. R. & Newman, W. H. 1990. Bases of Interorganizational Co-operation: Propensity, Power, Persistence. Journal of Management Studies, 27: 417-434.
Hedlund, G. & Nonaka, I. 1993. Models of Knowledge Management in the West and Japan. In P. Lorange, B. Chakravarthy, J. Roos, & A. Van de Ven (Eds.) Implementing Strategic Processes: Change, Learning, and Co-Operation, Oxford: Basil Blackwell, 117-144.
Huber, G. P. 1991. Organizational Learning: The Contributing Processes and a Review of the Literature. Organization Science, 2: 88-117.
Inkpen, A.C. & Crossan, M. 1995. Believing Is Seeing: Joint Ventures and Organizational Learning, Journal of Management Studies, Forthcoming.
Jarillo, J. C. 1988. On Strategic Networks. Strategic Management Journal, 9: 31-41.
Jelinek, M. 1979. Institutionalizing Innovations: A study of Organizational Learning Systems. New York: Praeger.
Keller, M. 1989. Rude Awakening: The Rise, Fall, and Struggle for Recovery of General Motors. New York: William Morrow.
Kogut, B. 1988. Joint Ventures: Theoretical and Empirical Perspectives. Strategic Management Journal, 9: 319-322.
Levinthal, D. A. 1991. Organizational Adaptation and Environmental Selection: Interrelated Processes of Change. Organization Science, 2: 140-145.
Levinthal, D. A. & March, J. G. 1993. The Myopia of Learning. Strategic Management Journal, 14: 95-112.
Lyles, M. A. 1988. Learning Among JV-sophisticated Firms. In F. Contractor & P. Lorange, (Eds.), Cooperative Strategies in International Business. Toronto: Lexington Books, 301-316.
Miles, M. B. & Huberman, A. M. 1984. Qualitative Data Analysis: A Sourcebook of New Methods. Newbury Park, CA: Sage.
Mintzberg, H. 1990. Strategy Formation: Schools of Thought. In J. Frederickson, (Ed.), Perspectives of Strategic Management. New York: Harper Business, 105-235.
Nelson, R. R. & Winter, S. G. 1982. An Evolutionary Theory of Economic Change. Cambridge, MA: Harvard University Press.
Nonaka, I. 1994. A Dynamic Theory of Organizational Knowledge. Organization Science, 5: 14-37.
Nonaka, I. 1991. The Knowledge-Creating Company. Harvard Business Review, 69, 6: 96-104.
Nonaka, I. 1990. Redundant, Overlapping Organizations: A Japanese Approach to Managing the Innovation Process, California Management Review, 32, 3: 27-38.
Nonaka, I. & Johansson, J. K. 1985. Organizational Learning in Japanese Companies. In R. Lamb & P. Shrivastava, (Eds.), Advances in Strategic Management, Vol. 3. Greenwich, CT: JAI Press, 277-296.
Patton, M. Q. 1987. How to use Qualitative Methods in Evaluation. Newbury Park, CA: Sage.
Pinder, C. C. & Moore, L. F. 1980. The Inevitability of Multiple Paradigms and the Resultant Need for Middle-Range Analysis in Organizational Theory. In C. Pinder & L. Moore, (Eds.), Middle Range Theory and the Study of Organizations. Boston: Martinus Nijhof, 87-100.
Porter, M. E. 1991. Towards a Dynamic Theory of Strategy. Strategic Management Journal, 12 (Special Issue Winter): 95-117.
Prahalad, C. K. & Hamel, G. 1990. The Core Competence of the Corporation. Harvard Business Review, 68, 3: 79-91.
Reed, R. & DeFillippi, R. J. 1990. Causal Ambiguity, Barriers to Imitation, and Sustainable Competitive Advantage. Academy of Management Review, 15: 88-102.
Simonin, B. L. & Helleloid, D. 1993. Do Organizations Learn? An Empirical Test of Organizational Learning in International Strategic Alliances. In D. Moore, (Ed.), Academy of Management Best Paper Proceedings 1993.
Tushman, M. & Scanlon, T. J. 1981. Boundary-Scanning Individuals: Their Role in Information Transfer and their Antecedents. Academy of Management Journal, 24: 289-305.
Van de Ven, A. H. 1992. Suggestions for Studying Strategy Process: A Research Note. Strategic Management Journal, 13 (special issue): 169-188.
Walsh, J. P. & Ungson, G. R. 1991. Organizational Memory. Academy of Management Journal, 16: 57-91.
Westney, D. E. 1988. Domestic and Foreign Learning Curves in Managing International Cooperative Strategies. In F. Contractor & P. Lorange, (Eds.), Cooperative Strategies in International Business. Toronto: Lexington Books, 339-346.
Yin, R. K. 1989. Case Study Research: Design and Methods, revised edition, Newbury Park, CA: Sage.
Zand, D. E. 1972. Trust and Managerial Problem Solving. Administrative Science Quarterly, 17: 229-239.