Strategic Management Journal

Strat. Mgmt. J., 21: 473496 (2000)

KNOWLEDGE FLOWS WITHIN MULTINATIONAL CORPORATIONS

ANIL K. GUPTA1* and VIJAY GOVINDARAJAN2
1

The Robert H. Smith School of Business, The University of Maryland, College Park, Maryland, U.S.A. 2 The Amos Tuck School of Business, Dartmouth College, Hanover, New Hampshire, U.S.A.

Pursuing a nodal (i.e., subsidiary) level of analysis, this paper advances and tests an overarching theoretical framework pertaining to intracorporate knowledge transfers within multinational corporations (MNCs). We predicted that (i) knowledge outows from a subsidiary would be positively associated with value of the subsidiarys knowledge stock, its motivational disposition to share knowledge, and the richness of transmission channels; and (ii) knowledge inows into a subsidiary would be positively associated with richness of transmission channels, motivational disposition to acquire knowledge, and the capacity to absorb the incoming knowledge. These predictions were tested empirically with data from 374 subsidiaries within 75 MNCs headquartered in the U.S., Europe, and Japan. Except for our predictions regarding the impact of source units motivational disposition on knowledge outows, the data provide either full or partial support to all of the other elements of our theoretical framework. Copyright 2000 John Wiley & Sons, Ltd.

In recent years, researchers in organization theory (Levitt and March, 1988), economics (Nelson and Winter, 1982), as well as strategic management (Prahalad and Hamel, 1994; Schendel, 1996) have identied organizational learning as one of the most important subjects for scholarly inquiry. Aimed at further deepening our understanding of a key topic within this broad area viz., intrarm ows of organizational knowledge, this paper reports the results of a theoretical and empirical investigation into the determinants of internal knowledge transfers within multinational corporations. The following four observations underlie the motivations for this study. First, every rm constitutes a bundle of knowledge. As a corollary of the resource-based view of the rm (Barney, 1991; Penrose, 1959; WerKey words: knowledge ows, multinational corporations, subsidiaries
*Correspondence to: Anil K. Gupta, The Robert H. Smith School of Business, The University of Maryland, College Park, MD 20742, U.S.A.

nerfelt, 1984), this observation is now so widely accepted as to have become almost axiomatic (Grant, 1996; Huber, 1991; Kogut and Zander, 1992; Nelson and Winter, 1982; Nonaka, 1994). In the context of this paper, it is particularly important to note that, of all possible resources that a rm might possess, its knowledge base has perhaps the greatest ability to serve as a source of sustainable differentiation and hence competitive advantage (Dierickx and Cool, 1989; Lippman and Rumelt, 1982). Second, the primary reason why MNCs exist is because of their ability to transfer and exploit knowledge more effectively and efciently in the intra-corporate context than through external market mechanisms. This internalization of intangible assets argument, originally advanced by Hymer (1960), has been subjected to numerous conrmatory empirical tests and is now widely accepted as the received theory on why MNCs exist (Buckley and Casson, 1976; Caves, 1971, 1982; Ghoshal, 1987; Kindleberger, 1969; Porter, 1986; Teece, 1981). Of course, external markets
Received 27 August 1997 Final revision received 1 August 1999

Copyright 2000 John Wiley & Sons, Ltd.

474

A. K. Gupta and V. Govindarajan

belonging to 75 major MNCs headquartered in the U.S., Japan, and Europe. In order to ensure reliability, data on the most critical variables (pertaining to knowledge transfers) were collected also from the immediate HQ-level superiors of the presidents of a large subset of the sampled subsidiaries; further, the tests for the hypotheses were conducted after controlling for the possible effects of the parent corporations country-oforigin, the resource characteristics of the parent corporations industry, and the nature of the subsidiarys operations. THE PHENOMENON OF INTEREST Because MNCs are complex multi-dimensional entities, knowledge ows within such enterprises occur not only along multiple directions but also across multiple dimensions, e.g., the ow of information pertaining to the Brazilian subsidiarys nancial performance over the last quarter to corporate headquarters, the transfer of packaging technology from a Swedish factory to one in India, or the transfer of customer service skills from a Japanese subsidiary to one in the U.S. In this study, we focus on the transfer of largely procedural types of knowledge (e.g., product designs, distribution know-how, etc.) but not on the transfer of largely declarative types of knowledge (e.g., monthly nancial data). In other words, this study focuses on the transfer of knowledge that exists in the form of know-how rather than on the transfer of knowledge that exists in the form of operational information. As Ghoshal and Bartlett (1990), Gupta and Govindarajan (1991), and Hedlund (1994) have suggested, knowledge transfers within the MNC take place within the context of an interorganizational network of differentiated units. Thus, ows of knowledge through the network can be studied from at least three different levels of analysis: nodal (i.e., a focus on the behavior of individual units), dyadic (i.e., a focus on the joint behavior of unit pairs), and systemic (i.e., a focus on the behavior of the entire network). Given the highly complex nature of the phenomenon under investigation and the relative dearth of previous empirical work on it, in this study, we have chosen to limit our investigation to the nodal level. More specically, we focus on individual subsidiaries only and examine the determinants of knowledge ows in each of the following
Strat. Mgmt. J., 21: 473496 (2000)

continue to become more open, efcient, and global on an ongoing basis. Notwithstanding the increasing sophistication of external markets, they remain relatively ineffective mechanisms for knowledge transfer on at least two grounds: one, bulk of the specialized knowledge of any rm exists in a tacit and thereby non-tradeable form; two, market-based transfers of knowledge are often associated with negative externalities such as involuntary expropriation and the risk of creating a new competitor. Third, the notion that MNCs exist primarily because of their superior ability (vis-a-vis markets) to engage in internal knowledge transfer does not in any way imply that such knowledge transfers actually take place effectively and efciently on a routine basis. In perhaps the only study to date on the actual costs of cross-border knowledge transfers, Teece (1981: 84) examined a sample of 26 technology transfer cases and reported that [T]he resource cost of international transfer is nontrivial. Transfer costs ranged from 2.25 percent to 59 percent of total project costs with a mean of 19.16 percent. The tacitness or causal ambiguity of knowledge is one of the most widely recognized barriers to its transfer and replication (Lippman and Rumelt, 1982; Polanyi, 1966; Zander and Kogut, 1995). Levinthal and March (1993), Simon (1991), Szulanski (1996) and others have suggested additional barriers to knowledge transfer e.g., barriers rooted in motivational dispositions and absorptive capacity. Finally, notwithstanding the criticality of internal knowledge transfers within MNCs, with some notable exceptions (e.g., Ghoshal and Bartlett, 1988 and Zander and Kogut, 1995), very little systematic empirical investigation into the determinants of intra-MNC knowledge transfers has so far been attempted. As Ghoshal, Korine, and Szulanski (1994: 97) have observed, A number of publications emphasize the importance of interunit communication for effective MNC management%but in none of them is the construct operationalized or measured, nor are the factors that inuence such communication empirically explored. Building on these observations, the primary objective of this paper is to advance the state of our theoretical as well as empirical understanding of the determinants of intra-MNC knowledge transfers. Data for this study were collected directly from the presidents of 374 subsidiaries
Copyright 2000 John Wiley & Sons, Ltd.

Knowledge Flows within Multinational Corporations

four domains: (i) knowledge outows to peer subsidiaries, (ii) knowledge outows to the parent corporation, (iii) knowledge inows from peer subsidiaries, and (iv) knowledge inows from the parent corporation.

475

THEORY
An overarching theoretical framework As Krone, Jablin, and Putnam (1987) have observed in their review of communication theory, even though different communication scholars have focused more (or less) heavily on different elements of the communication process, virtually all of them recognize the following as the basic elements of any two-person communication: a message, a sender, a coding scheme, a channel, transmission through the channel, a decoding scheme, a receiver, and the assignment of meaning to the decoded message. Consistent with these ideas from communication theory, we conceptualize knowledge ows (into or out of a subsidiary) to be a function of the following ve factors: (i) value of the source units knowledge stock, (ii) motivational disposition of the source unit, (iii) existence and richness of transmission channels, (iv) motivational disposition of the target unit, and (v) absorptive capacity of the target unit. Barriers or facilitators to the transfer of knowledge can manifest themselves in any or all of these ve factors: (a) Value of source units knowledge stock. Knowledge ows across units are not cost free (Teece, 1981). We also know that different resources have different levels of value (Barney, 1991). Thus, the greater the value of a subsidiarys knowledge stock for the rest of the MNC, the greater would be its attractiveness for other units. This idea is broadly consistent with the concept of relative advantage in the literature dealing with diffusion of innovations which has argued that the adoption rate of an innovation is positively related to its relative advantage (Rogers, 1995). This idea has not yet been applied to the examination of interunit knowledge transfers within multinational corporations. Within such corporations, we visualize the knowledge stock of any subsidiary as comCopyright 2000 John Wiley & Sons, Ltd.

posed of both duplicative as well as nonduplicative knowledge. The presence of non-duplicative knowledge is a necessary, although not sufcient, condition for such knowledge to be of value to other units. Thus, we would anticipate that knowledge outows from a subsidiary are likely to be high when the subsidiarys knowledge stock is non-duplicative as well as relevant for the rest of the global network. (b) Motivational disposition of the source unit. As Cyert (1995) has suggested, an organizational unit with uniquely valuable knowhow is likely to enjoy an information monopoly within the corporation. This reality coupled with the fact that power struggles are a ubiquitous phenomenon in any organization (Pfeffer, 1981) implies that at least some units will view uniquely valuable know-how as the currency through which they acquire and retain relative power within the corporation. Levitt and March (1988: 331) have observed similarly that In many (but not all) situations%diffusion of experience has negative consequences for organizations that are copied. Therefore, we anticipate that factors which would enhance the motivational disposition of the source unit to share its knowledge with other units within the MNC are likely to counterbalance any hoarding tendencies and thereby to have a positive impact on the magnitude of knowledge outows. (c) Existence and richness of transmission channels. As would be expected, and as demonstrated empirically by Ghoshal and Bartlett (1988) in the domain of MNCs, knowledge ows cannot occur without the existence of transmission channels. Beyond mere existence, we would expect other properties of transmission channels to also affect the extent of knowledge ows the most notable such property would be the richness/bandwidth of communication links, as captured in aspects such as informality, openness, and density of communications (Daft and Lengel, 1986; Gupta and Govindarajan, 1991; Jablin, 1979; Tushman, 1977). (d) Motivational disposition of the target unit. The Not-Invented-Here (NIH) syndrome
Strat. Mgmt. J., 21: 473496 (2000)

476

A. K. Gupta and V. Govindarajan

interacting individuals share common meanings, a mutual subcultural language, and are alike in personal and social characteristics, the communication of new ideas is likely to have greater effects in terms of knowledge gain, attitude formation, and overt behavior change (Rogers, 1995: 19; see also Lazarsfeld and Merton, 1964). Figure 1 presents a schematic diagram of the overarching framework developed in this section. From the perspective of the nodal level of analysis being pursued in this study, this framework can be translated into the following six propositions: Proposition 1: Ceteris paribus, the value of a subsidiarys knowledge stock will be positively associated with outows of knowledge from that subsidiary. Proposition 2: Ceteris paribus, the motivational disposition of a subsidiary to share its knowledge with other units will be positively associated with outows of knowledge from that subsidiary. Proposition 3: Ceteris paribus, the existence and richness of transmission channels linking a subsidiary to other units within the MNC will be positively associated with outows of knowledge from that subsidiary. Proposition 4: Ceteris paribus, the existence and richness of transmission channels linking a subsidiary to other units within the MNC will be positively associated with inows of knowledge into that subsidiary. Proposition 5: Ceteris paribus, the motivational disposition of a subsidiary to seek/accept knowledge from other units will be positively associated with inows of knowledge into that subsidiary. Proposition 6: Ceteris paribus, the capacity of a subsidiary to absorb incoming knowledge from other units will be positively associated with inows of knowledge into that subsidiary. In the rest of this section, we operationalize the constructs underlying these propositions and
Strat. Mgmt. J., 21: 473496 (2000)

is well-known and also has been the subject of scholarly inquiry (Katz and Allen, 1982). There are at least two drivers of the NIH syndrome: (i) ego-defense mechanisms (Allport, 1937; Sherif and Cantrill, 1947) which can lead some managers to block any information that might suggest that others are more competent than they are, and (ii) power struggles within organizations (Pfeffer, 1981) which can lead some managers to try to downgrade the potential power of peer units by pretending that the knowledge stock possessed by these peer units is not unique and valuable. In short, unless counterveiling forces are present, the NIH syndrome can act as a major barrier to the inows of knowledge into any focal unit. These counterveiling forces can manifest themselves in several forms: the relative paucity of the focal units knowledge stock, incentives that increase subsidiary managers eagerness to learn from peer units, or coercive pressures from corporate headquarters. (e) Absorptive capacity of the target unit. Even when exposed to the same environment and even when there are insignicant differences in the desire to acquire new knowledge, individuals and organizations may differ in their absorptive capacity i.e., in their ability to recognize the value of new information, assimilate it, and apply it to commercial ends (Cohen and Levinthal, 1990: 128). There are at least two reasons why absorptive capacity may differ across organizations: (i) the extent of prior related knowledge, and (ii) the extent of inter-unit homophily of the ` receiving unit vis-a-vis the sending unit. Prior related knowledge is important because it shapes the lters through which the organization differentiates between more vs. less relevant signals and also because it determines the organizations ability to internalize and assimilate the more valued signals (Cohen and Levinthal, 1990). On the other hand, homophily i.e., the degree to which two or more individuals who interact are similar in certain attributes, such as beliefs, education, social status, and the like (Rogers, 1995: 1819) is important because when the
Copyright 2000 John Wiley & Sons, Ltd.

Knowledge Flows within Multinational Corporations

477

Figure 1. Determinants of intra-corporate knowledge outows from and inows to foreign subsidiaries: An overarching theoretical framework

develop more concrete and empirically testable hypotheses. Value of source units knowledge stock We argued earlier that, in order for a source units knowledge to be of value to other units, the source unit must (i) create non-duplicative knowledge on its own, and (ii) this nonduplicative knowledge must be of relevance for the rest of the global network. Based on this reasoning, we operationalize the construct of value of knowledge stock in terms of the following three variables: mode of entry, subsidiary size, and the economic level of the host country relative to that of the home country. Mode of entry. As Caves (1982), Root (1987) and others have pointed out, an MNC may enter a foreign country through one of several modes greeneld operations, strategic alliances, or acquisitions. Since our study focuses only on fully- or majority-owned subsidiaries, we examine here the impact of greeneld vs. acquisition modes only. At a general level, we can visualize every subsidiary to consist of three bundles of knowledge: duplicative knowledge, nonduplicative knowledge that is relevant only in the local environment, and non-duplicative knowledge that is relevant also for other units within the global network. As the literature on foreign direct investment
Copyright 2000 John Wiley & Sons, Ltd.

has argued and demonstrated (Hennart and Park, 1993), the less the overlap between existing corporate know-how and the know-how required to succeed in a host market, the greater the probability of acquisition as the mode of entry. Thus, relative to greeneld subsidiaries, acquired subsidiaries on average can be expected to have ` a knowledge stock that is less duplicative vis-avis the knowledge stock of the rest of the corporation. It is true that only a subset of the nonduplicative knowledge would be of relevance for the global network. However, since the pool of non-duplicative knowledge would be higher for acquired subsidiaries as compared to greeneld subsidiaries, it is likely that acquired subsidiaries should have a larger pool of relevant knowledge to offer to the global network than greeneld subsidiaries. Based on these arguments, Proposition 1 can be operationalized in the form of the following two empirically testable hypotheses: ceteris paribus, relative to greeneld operations, acquired subsidiaries will engage in greater knowledge outows to peer subsidiaries (H1a) and to the parent corporation (H1a). Subsidiary size. We anticipate that the typical MNC would discourage investment of a subsidiarys resources in the reinvention of knowledge that exists elsewhere in the global network. Thus, we would expect that a subsidiarys own resources would generally be directed at the creation of non-duplicative knowledge. Since larger
Strat. Mgmt. J., 21: 473496 (2000)

478

A. K. Gupta and V. Govindarajan

than just the focal subsidiary) would be a major determinant of motivation to share knowledge with other subsidiaries. Based on this reasoning, we operationalized the construct of motivational disposition in terms of the subsidiary vs. corporate focus (i.e., nodal vs. network optimization focus) of the incentive system for the subsidiary president. Incentive focus. As Salter (1973) suggested and as Gupta and Govindarajan (1986) and Pitts (1974) demonstrated, the incentive bonus for a division/subsidiary general manager may be linked solely to the performance of the focal unit, solely to the performance of several units, or to some combination of the two. As these authors have argued, the greater the need to motivate a unit general manager to focus on system-wide optimization as distinct from local optimization, the better it is to link the incentives to the performance of a cluster of units. These arguments result in the following operationalizations of Proposition 2: ceteris paribus, the greater the extent to which a subsidiary presidents bonus is network-focused rather than subsidiary-focused, the greater will be the knowledge outows from that subsidiary to peer subsidiaries (H2a) and to the parent corporation (H2a). Existence and richness of transmission channels As communications theory informs us (Daft and Lengel, 1986; Krone et al., 1987), transmission channels can be both formal and informal. Accordingly, we operationalize the construct of transmission channels in terms of two mechanisms: one formal (viz., formal integrative mechanisms) and one informal (viz., corporate socialization mechanisms). Formal integrative mechanisms. Galbraith (1973) and Nadler and Tushman (1987) identied liaison positions, task forces, and permanent committees as some of the key formal structural mechanisms for integrating multiple units of an organization. It is easy to see that the greater the extent to which a subsidiary is linked to the rest of the global network through such integrative mechanisms, the greater would be the density of communication interface between the subsidiary and other units, thereby contributing positively to media richness (Daft and Lengel, 1986). Thus, focusing on knowledge outows from the subsidiStrat. Mgmt. J., 21: 473496 (2000)

subsidiaries will have a greater pool of resources dedicated to the creation of new knowledge, it follows that subsidiary size should have a positive impact on the ability of the subsidiary to offer non-duplicative knowledge to the rest of the corporation. Clearly, not all of the non-duplicative knowledge generated by a subsidiary would have global relevance; however, a subset of such knowledge will. These arguments yield the following additional operationalizations of Proposition 1: ceteris paribus, the larger the size of a subsidiary, the greater will be the knowledge outows from that subsidiary to peer subsidiaries (H1b) and to the parent corporation (H1b). Relative economic level. Countries differ in their levels of economic advancement. If we make the straightforward assumption that most, perhaps all, societies around the world strive to increase (rather than merely maintain or decrease) their levels of economic advancement, then it follows that, on average, more advanced countries are likely to serve as trend-setters and the sources of technological, marketing, as well as managerial know-how to a greater extent than less advanced countries. In other words, in the intracorporate context, on average, a focal unit is likely to view the knowledge stock of another unit located in an economically more advanced country relative to itself as more valuable than that of a unit located in a relatively less advanced country. These arguments also yield the following operationalization of Proposition 1: ceteris paribus, the higher the level of the host countrys economic development relative to the home country, the greater will be the knowledge outows from that subsidiary to the parent corporation (H1c). Since our empirical study was conducted at the nodal level of analysis, we did not collect any data regarding knowledge ows between specic inter-subsidiary dyads. Accordingly, in the above hypothesis, we have focused only on the relative economic level of the focal subsidiary vis-a-vis the parent corporation and not on that vis-avis other specic subsidiaries. Thus, we neither advance nor test any inter-subsidiary hypotheses pertaining to relative economic level. Motivational disposition of the source unit We posit that the extent to which the subsidiary president is rewarded for improvements in the performance of a network of subsidiaries (rather
Copyright 2000 John Wiley & Sons, Ltd.

Knowledge Flows within Multinational Corporations

ary, we can now operationalize Proposition 3 in terms of the following concrete hypotheses: ceteris paribus, the greater the reliance on formal mechanisms (liaison personnel, task forces, permanent committees) to integrate a subsidiary with the rest of the MNC, the greater will be the knowledge outows from that subsidiary to peer subsidiaries (H3a) and to the parent corporation (H3a). Similarly, focusing on knowledge inows into the subsidiary, we can also operationalize Proposition 4 in terms of the following testable hypotheses: ceteris paribus, the greater the reliance on formal mechanisms (liaison personnel, task forces, permanent committees) to integrate a subsidiary with the rest of the MNC, the greater will be the knowledge inows into that subsidiary from peer subsidiaries (H4a) and from the parent corporation (H4a). Corporate socialization mechanisms. Corporate socialization mechanisms refer to those organizational mechanisms which build interpersonal familiarity, personal afnity, and convergence in cognitive maps among personnel from different subsidiaries (Edstrom and Galbraith, 1977; Van Maanen and Schein, 1979). Greater interpersonal familiarity and personal afnity can be expected to increase the openness of communication between the interacting parties. Further, as Daft and Lengel (1986) have suggested, personal and more open communication increases the richness of communication channels. Thus, we would argue that greater participation in corporate socialization mechanisms would have a positive impact on the richness of transmission channels between the focal subsidiary and other units. In this study, we separate lateral from vertical socialization mechanisms. Examples of the former would be: job transfers to peer subsidiaries and participation in multi-subsidiary executive programs; similarly, examples of the latter would be: job transfers to corporate headquarters and participation in corporate mentoring programs (Ghoshal and Bartlett, 1988). Focusing now on knowledge outows from the focal subsidiary, we can advance the following additional operationalizations of Proposition 3: ceteris paribus, the greater the lateral socialization of a subsidiary president, the greater will be the knowledge outows from that subsidiary to peer subsidiaries (H3b); further, ceteris paribus, the greater the vertical socialization of a subsidiary president, the greater will be the knowledge outows from
Copyright 2000 John Wiley & Sons, Ltd.

479

that subsidiary to the parent corporation (H3b). Similarly, focusing now on knowledge inows into the focal subsidiary, we can advance the following additional operationalizations of Proposition 4: ceteris paribus, the greater the lateral socialization of a subsidiary president, the greater will be the knowledge inows into that subsidiary from peer subsidiaries (H4b); further, ceteris paribus, the greater the vertical socialization of a subsidiary president, the greater will be the knowledge inows into that subsidiary from the parent corporation (H4b). Motivational disposition of the target unit We argued earlier that a subsidiarys motivational disposition to acquire/accept knowledge from other units within the enterprise would be a function of (i) incentives that increase subsidiary managers eagerness to learn, (ii) the relative paucity of the subsidiarys knowledge stock, and/or (iii) coercive pressures from corporate headquarters. Based on this reasoning, we operationalized the construct of motivational disposition of the target unit in terms of three variables: subsidiary vs. corporate focus of the incentives for the subsidiary president (a determinant of eagerness to learn), relative economic level (a determinant of the paucity of local knowledge stock), and HQ-subsidiary decentralization (a determinant of coercive pressures). Incentive focus. Unlike the case of knowledge outows where the required motivational disposition can be characterized as eagerness to help others, in the case of knowledge inows, the required motivation would be characterized as eagerness to learn and to help oneself. We would argue that, other things being equal, subsidiary personnel would be more eager to learn in those contexts where the linkage between incentives and the subsidiarys own capabilities is tighter rather than weaker i.e., in contexts where incentives are linked more tightly to the focal subsidiarys own performance than to the performance of a cluster of subsidiaries. This is so because, unlike cluster-based incentives, which can create free-rider problems, subsidiary-based incentives would create a stronger disposition to learn from any and all sources. These arguments yield the following operationalizations of Proposition 5: ceteris paribus, the greater the extent to which a subsidiary presidents bonus is subsidiStrat. Mgmt. J., 21: 473496 (2000)

480

A. K. Gupta and V. Govindarajan

These arguments yield the following additional operationalization of Proposition 5: ceteris paribus, the lower the decentralization of decisionmaking authority to a subsidiary, the greater will be the knowledge inows into that subsidiary from the parent corporation (H5c). Since the construct of decentralization pertains to parentsubsidiary relationships only, we advance no hypotheses pertaining to the impact of decentralization on knowledge inows from peer subsidiaries. Absorptive capacity of the target unit We argued earlier that the absorptive capacity of a subsidiary would be a function of (i) its familiarity with the incoming knowledge, and (ii) interunit homophily. Based on this reasoning, we operationalized the construct of absorptive capacity in terms of the following two variables: mode of entry (a determinant of the subsidiarys ex-ante familiarity with the corporate-wide knowledge base) and the proportion of local nationals vs. expatriates within the subsidiarys top management team (a measure of the interunit homophily of subsidiary managers). Mode of entry. Literature on foreign direct investment (see e.g., Hennart and Park, 1993) has argued theoretically and demonstrated empirically that the less the overlap between existing corporate know-how and the know-how required to succeed in a host market, the greater the probability of acquisition as the mode of entry. Thus, as we discussed earlier, relative to greeneld operations, acquired subsidiaries are more likely to have a non-duplicative knowledge base vis-avis the parent corporation. Building on Cohen and Levinthals (1990) arguments regarding the determinants of absorptive capacity, it follows that, on average, the novelty of acquired subsidiaries knowledge base should also imply a lower absorptive capacity for intra-corporate knowledge relative to the case with greeneld subsidiaries. Based on these arguments, we can now operationalize Proposition 6 in terms of the following concrete hypotheses: ceteris paribus, relative to greeneld operations, acquired subsidiaries will engage in less knowledge inows from peer subsidiaries (H6a) and from the parent corporation (H6a).1
1

ary-focused rather than network-focused, the greater will be the knowledge inows into that subsidiary from peer subsidiaries (H5a) and from the parent corporation (H5a). Relative economic level. Paralleling our discussion on this variable in the context of knowledge outows, we expect that, other things being equal, the lower the level of economic advancement of the host country (i.e., where the subsidiary is located) vis-a-vis the home country (i.e., where the parent is located), the more eager subsidiary personnel would be to learn from the parent corporation. They are likely to perceive the knowledge stock of the parent as relatively more valuable and, thus, are likely to regard knowledge inows as a potential source of competitive advantage against other players in the local market. Knowledge inows into such subsidiaries may also be facilitated by explicit public policy regimes that mandate technology inows as the condition for allowing MNCs access to the local market; as an example, this is illustrated well by the recent decisions of the Chinese government (Smith and Hamilton, 1995: 2). These arguments suggest the following additional operationalization of Proposition 5: ceteris paribus, the lower the level of the host countrys economic development relative to the home country, the greater will be the knowledge inows into the subsidiary from the parent corporation (H5b). As discussed in the context of knowledge outows, given our nodal level of analysis, we neither advance nor test any hypotheses pertaining to the relative economic levels of subsidiary pairs. Headquarters-subsidiary decentralization. The concept of decentralization (or its obverse i.e., centralization) has had a long history of research in organization theory (see Ford and Slocum, 1977 for an extensive review). Even in the domain of research on MNCs, scholars have argued that centralization is one of the fundamental dimensions of organization design (Egelhoff, 1988: 129). Our expectations of a linkage between decentralization and knowledge inows into a subsidiary parallel the broader arguments of DiMaggio and Powell (1983), echoed also by Levitt and March (1988), that coercion is one of the major (but not the sole) drivers of inter-organizational isomorphism. In the MNC context also, similar arguments have been advanced by many scholars (e.g., Gates and Egelhoff, 1986; Ghoshal and Bartlett, 1988).
Copyright 2000 John Wiley & Sons, Ltd.

An anonymous reviewer has pointed out that, at rst glance, the two hypotheses under H6 might appear inconsistent with
Strat. Mgmt. J., 21: 473496 (2000)

Knowledge Flows within Multinational Corporations

Proportion of local nationals in the subsidiarys top management team. Several studies have indicated that national background accounts for signicant differences in managerial perspectives (e.g., Tung, 1982; Zeira, 1986). Accordingly, the greater the proportion of local nationals (i.e., the lower the proportion of expatriates) within the subsidiarys top management team (TMT), the lower would be the homophily between the subsidiary and the rest of the corporation. Building on Rogers arguments (1995), we would expect that inter-unit homophily is likely to be positively associated with absorptive capacity. This is so because greater homophily implies a greater commonality in language systems as well as in the meanings assigned to the artifacts of communication. Thus, on average, subsidiaries with a greater proportion of local nationals within the TMT can be expected to have lower absorptive capacity for incoming knowledge from the rest of the corporate network. These arguments yield the following operationalization of Proposition 6: ceteris paribus, the greater the proportion of local nationals within the subsidiarys top management team, the less will be the knowledge inows into that subsidiary from peer subsidiaries (H6b) and from the parent corporation (H6b).2
the two hypotheses under H1. In H1, we predicted that, because of their large non-duplicative knowledge base, acquired subsidiaries would exhibit high knowledge outows; an implicit assumption underlying this prediction was that such knowledge would be absorbed by the receiving units. However, H6 argues that unfamiliarity with incoming knowledge would reduce absorptive capacity among the receiving units. Thus, H1 could not be true unless mode of entry has a different effect on ows from the subsidiary to the MNC than it does on ows from the MNC to the subsidiary. We believe this to be the case. The roots of this differing effect lie in the following two observations: One, the typical acquisition represents a voluntary event for the acquiring MNC but an involuntary event for the acquired subsidiary; thus, the willingness of the acquiring MNC to integrate the new knowledge of the acquired unit should, on average, be greater than the willingness of the acquired unit to integrate the new knowledge of the acquiring MNC. Two, the typical MNC would have much greater experience at acquiring and integrating new units than the typical unit would have in being acquired and integrated; accordingly, on average, the acquirers ability to digest new knowledge should be greater than that of the acquired unit. 2 We should note that nationality structure of a subsidiarys TMT has the potential to affect knowledge inows not only through its impact on absorptive capacity but also through its impact on richness of transmission channels between the local subsidiary and the rest of the global network. It does appear likely that, on average, relative to local nationals, expatriates should have stronger and longer-tenured social ties with manCopyright 2000 John Wiley & Sons, Ltd.

481

Control variables Country of origin. There already exists a large body of both theoretical and empirical literature dealing with the fact that country of origin has a major impact on the propensities of MNCs visa-vis the choice of global strategies, organizational structures and control systems, as well as internal corporate cultures (e.g., Bartlett and Ghoshal, 1989; Egelhoff, 1984; Franko, 1976; Porter, 1994; Yip, Roos and Johansson, 1994). Accordingly, all of our hypotheses were tested after controlling for the effect of country-of-origin of the MNC. Industry resource characteristics. As discussed earlier, economic theory posits that MNCs come to be primarily because external markets are less effective and efcient at knowledge transfer than intracorporate mechanisms (Caves, 1982; Hymer, 1960; Kindleberger, 1969). Empirical tests of this theory have consistently shown that industries characterized by greater degrees of knowledge intensities (industries with higher R&D -to-sales-ratios and/or higher advertising-to-sales ratios) tend to be more global than other industries (e.g., Goedde, 1978; Grueber, Mehta, and Vernon, 1967; Horst, 1972). Accordingly, we deemed it important that, in testing our hypotheses, we control also for the potential effects of three resource characteristics of the MNCs industry: R&D intensity, xed asset intensity, and advertising intensity (Collis and Ghemawat, 1994). Nature of subsidiarys operations. It is well accepted that foreign subsidiaries will often vary in the scope of value chain activities included within their operations (Porter, 1986). The presagers at corporate HQ and in other subsidiaries. In fact, as the correlations in Table 1 indicate, there does exist a strong negative correlation (0.59, p 0.001) between proportion of local nationals in the subsidiarys TMT and vertical corporate socialization. Thus, as pointed out by an anonymous reviewer and the consulting editor, the question arises as to whether, in the context of our study, TMT nationality might be a better proxy for another factor (such as richness of transmission channels) rather than absorptive capacity. We believe that this would indeed be a serious concern if we did not have any direct measures of socialization mechanisms as one of the hypothesized antecedents of knowledge inows. However, as captured in H3b, H3b , H4b, and H4b , we do test for the direct effect of socialization mechanisms on knowledge inows. Thus, in a multivariate regression context, any remaining impact of TMT nationality on knowledge inows is likely to be due primarily to absorptive capacity rather than transmission channel considerations.
Strat. Mgmt. J., 21: 473496 (2000)

482

A. K. Gupta and V. Govindarajan

need to obtain access, it was not possible to use a random sample either from the entire universe of MNCs or from the entire subset of MNCs headquartered in the U.S., Europe, and Japan. Nonetheless, given the diversity of industries in which the sampled rms operate (food products, industrial machinery, computers, telecommunications, pharmaceuticals, automobiles, chemical production, electronics, consumer durables, consumer nondurables, etc.), there is no prima facie reason to expect any systematic bias in the ndings from subsidiaries within these rms. A personalized cover letter accompanying each questionnaire explained the purpose of the study and provided assurances regarding the condentiality of collected data. In order to minimize response bias, the participants were also provided with pre-addressed envelopes to enable them to return the completed questionnaires directly to the researchers without any risk of perusal by others in their rms. A total of 374 questionnaires (38 percent) were returneda response rate that compares very favorably with past survey-based research studies in the strategic management area. The number of respondents for U.S., Japan, and European MNCs were 117 (28 percent), 112 (41 percent), and 145 (46 percent), respectively. To test for inter-rater reliability on the most critical variables in the study (knowledge outows and inows), we were also able to get responses on these knowledge ow variables from the immediate corporate-level direct-report superiors of 89 of the responding subsidiary presidents. For the sample, median worldwide revenues and median number of total employees for the parent rms were $5.8 billion and 32,100 respectively; at the subsidiary level, the median number of employees per subsidiary was 350. These gures pertain to 1991, the year in which the survey data were collected. Measures A summary of how the independent variables as well as the control variables were measured is contained in the Appendix. Wherever possible, we used standard well-established research instruments with minor changes in wording to adapt the instrument to the multinational context. Given below are details pertaining to how the four variables central to this study knowledge outows to peer subsidiaries (KO-S), knowledge
Strat. Mgmt. J., 21: 473496 (2000)

ence or absence of any particular activity within the subsidiarys operations can be expected to shape the nature of the subsidiarys interactions with the rest of the corporation and, thus, the nature of knowledge inows into and outows from the subsidiary. Accordingly, all of our hypotheses were tested after controlling also for the potential effects of two dummy variables: presence of a primary upstream activity (i.e., R&D and/or manufacturing) and presence of a primary downstream activity (i.e., marketing and sales).

METHOD
Sample Data for this study were collected through a combination of questionnaire surveys and secondary sources. The following steps guided the development of the questionnaire instrument: (i) interviews with subsidiary presidents and corporatelevel executives in six MNCs to understand and clarify the phenomenon of interest, (ii) a review of previous research to locate, wherever possible, measures that would appropriately capture the constructs under study, and (iii) a pretesting of the questionnaire for clarity and relevance through face-to-face interviews with four subsidiary presidents (two American and two non-American). The pre-tested questionnaires were mailed to the heads (variously titled as presidents, managing directors, or general managers) of 987 foreign subsidiaries of major MNCs headquartered in the U.S. (407 subsidiaries of 19 MNCs), Japan (270 subsidiaries of 41 MNCs), and Europe (310 subsidiaries of 15 MNCs). Subsidiary presidents within Japanese MNCs received both an English and a Japanese language questionnaire; initial interviews with the European companies indicated that only the English-language questionnaire would sufce. The U.S. sample was drawn from the list of the largest U.S.-based MNCs contained in the International Directory of Corporate Afliations (National Register, 1991); this was also the approach used for developing a list of subsidiaries for 9 out of the 15 European MNCs. In the case of the other 6 European MNCs, the list of subsidiaries was drawn up in cooperation with the senior-most corporate executive in charge of strategic planning, an approach also used in the case of all of the Japanese MNCs. Given the constraints of time and funding and given the
Copyright 2000 John Wiley & Sons, Ltd.

Knowledge Flows within Multinational Corporations

outows to the parent corporation (KO-P), knowledge inows from peer subsidiaries (KI-S), and knowledge inows from the parent corporation (KI-P) were measured. As stated earlier, in this study, we focus on the transfer of largely procedural types of knowledge (e.g., product designs, distribution knowhow, etc.) but not on the transfer of largely declarative types of knowledge (e.g., monthly nancial results). Knowledge ow data were collected on the following seven items: (1) marketing know-how, (2) distribution know-how, (3) packaging design/technology, (4) product designs, (5) process designs, (6) purchasing know-how, and (7) management systems and practices. For each of these seven items, the subsidiary president was asked to indicate on a 7-point scale (ranging from not at all to a very great deal) the extent to which the subsidiary engaged in transfers of knowledge and skills in each of the following four directions: (1) provides knowledge and skills to sister subsidiaries, (2) provides knowledge and skills to parent corporation, (3) receives knowledge and skills from sister subsidiaries, and (4) receives knowledge and skills from the parent corporation. For each of these knowledge ow directions, responses across the seven items were averaged to yield composite measures of KO-S, KOP, KI-S, and KI-P. For these four variables, the means, the standard deviations, and Chronbach alpha values respectively were as follows: KO-S (2.36, 1.25, 0.89), KO-P (2.39, 1.20, 0.87), KI-S (2.21, 1.27, 0.92), and KI-P (3.75, 1.59, 0.89). Given the 1-to-7 range of the 7-point scale used to measure knowledge ows, the mean values of the four types of knowledge ows in our sample (2.36, 2.39, 2.21, and 3.75) may at rst glance appear low. However, as claried above, it should be noted that this study has focused on the transfers of largely procedural knowledge (i.e., know-how) rather than on the transfers of largely declarative knowledge (e.g., operational data). Given the tacit rather than codied nature of much procedural knowledge, we would expect the mean levels of knowledge transfers in this arena to be on the lower rather than higher side. It also should be noted that, as would be expected in the case of hierarchical organizations, pairwise t-tests revealed that knowledge inows from the parent to focal subsidiaries (KI-P) were signicantly greater (at
Copyright 2000 John Wiley & Sons, Ltd.

483

p 0.001) than each of the other three types of knowledge ows. Given the perceptual nature of these knowledge ow measures and given their centrality for our study, we deemed it critical that they be tested also for inter-rater reliability. Towards this goal, we were able to get responses on the same knowledge ow variables from the immediate corporate-level direct report superiors of 89 of the subsidiary presidents. Each superior completed a separate questionnaire containing the subsidiarys name for each of the sampled subsidiaries reporting to him/her. This questionnaire used exactly the same seven dimensions of knowledge. For each of these seven items, the superior was asked to indicate on a 7-point scale (ranging from not at all to a very great deal) the extent to which he/she expected the named subsidiary to engage in transfers of knowledge and skills in each of the following two directions: (1) provides knowledge and skills to the rest of the corporation, and (2) receives knowledge and skills from the rest of the corporation. For each of these two knowledge ow directions, responses across the seven items were averaged to yield composite measures of expected knowledge outows from the subsidiary (Chronbach alpha = 0.82) and expected knowledge inows into the subsidiary (Chronbach alpha = 0.81) respectively. For these 89 subsidiaries, this corporate-level measure of expected knowledge outows from the subsidiary correlates at 0.23 (p 0.05) with the average of KO-S and KO-P; similarly, the corporate-level measure of expected knowledge inows into the subsidiary correlates at 0.38 (p 0.001) with the average of KI-S and KI-P. Given these positive correlations in the data from subsidiary presidents and their immediate superiors, in a context where they are typically separated by a geographic distance of thousands of miles, we believe that our measures of KO-S, KO-P, KI-S, and KI-P can be deemed as reliable. Table 1 contains the matrix of zero-order correlations among these and all other variables utilized in this study. As this table indicates, the average correlation among the four knowledge ow variables is 0.32 implying that the four types of knowledge ows are distinct, albeit related, variables not only conceptually (Gupta and Govindarajan, 1991) but also empirically.
Strat. Mgmt. J., 21: 473496 (2000)

484

Table 1 Zero-order correlation coefcients among all variables under study

X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X15 X16 X17

X1 X2 *** 0.54 *** 0.58 *** 0.33 *** 0.19 0.07 0.02 0.07 0.08 0.06 0.03 0.02 0.00 0.10 0.05 0.03 0.04 0.01 0.04 0.04 0.06 0.08 0.05 ** 0.13 0.01 0.01 0.06 *** 0.32 0.04 * 0.09 0.03 0.01 0.01 *** 0.20 0.08 0.06 *** 0.18 0.04 *** 0.22 *** 0.27 *** 0.16 *** 0.37 0.02 *** 0.39 0.03 ** 0.14 0.08 *** 0.22 *** 0.25 ** 0.15 ** 0.15 ** 0.15 *** 0.28 *** 0.21 *** 0.16 0.00 * 0.12

KO-S KO-P

X3

KI-S

X4

KI-P

Copyright 2000 John Wiley & Sons, Ltd.

X5

R&D Intensity

X6

Fixed Asset Intensity

X7

Advertising Intensity

A. K. Gupta and V. Govindarajan

X8

Upstream Activities1

X9

Downstream Activities2

0.04 *** 0.20 ** 0.14 *** 0.25 0.09 *** 0.22 0.08 0.06 *** 0.36 0.00 * 0.13

X10

Mode of Entry3

X11

Subsidiary Size

X12

X13

Relative Economic Level Incentive Focus4

0.03 ** 0.14 ** 0.15

0.05 * 0.12 *** 0.20 *** 0.22 * 0.13 0.03 *** 0.35 * 0.13 * 0.12

0.06 0.02 0.09 *** 0.22 0.08 *** 0.21 * 0.11 0.08 0.06 ** 0.16 0.02 *** 0.18 ** 0.15 ** 0.15 0.02 * 0.11 *** 0.18 *** 0.21 0.00 0.05 0.03 ** 0.12 0.01 0.03 0.08 0.06 0.06 *** 0.59 ** 0.12

X14

X15 0.08 *** 0.19

X16

0.02 *** 0.24 ** 0.15 ** 0.14 0.06 ** 0.14 0.04 *** 0.22 0.06 * 0.12 *** 0.21 *** 0.17 ** 0.14 0.01 * 0.11 *** 0.20 * 0.13 *** 0.23

0.05 * 0.12 *** 0.22 *** 0.21 *** 0.25 0.06 *** 0.21 ** 0.14 *** 0.25 * 0.13 ** 0.14 0.01 ** 0.13 * 0.09 *** 0.25 *** 0.22

X17

X18

0.08 Formal Integ *** Mechanisms 0.29 Lateral Socializn Mech. *** 0.25 Vertical Socializn *** Mech. 0.22 HQ-Sub Decentralization 0.06 Local Nationals in *** TMT 0.19

0.08 * 0.13 * 0.10

Strat. Mgmt. J., 21: 473496 (2000)

1 = Subsidiary has an upstream activity (R&D and/or manufacturing); 0 = Subsidiary has no upstream activity. 1 = Subsidiary has a downstream activity (marketing and sales); 0 = Subsidiary has no downstream activity. Mode of entry: 1 = Acquisition; 0 = Greeneld. 4 Higher values signify that the incentive system is more network, rather than subsidiary, focused. *one-tail p 0.05; **one-tail p 0.01; ***one-tail p 0.001

Knowledge Flows within Multinational Corporations RESULTS

We have four dependent variables (KO-S, KO-P, KI-S, or KI-P) and a set of hypotheses pertaining to each of these variables. Each set of these hypotheses was tested through a series of multivariate OLS regressions: rst, we entered the four control variables pertaining to country-of-origin; second, we entered the three control variables pertaining to industry resource characteristics; third, we entered the two control variables pertaining to nature of subsidiary operations; nally, we entered the independent variables hypothesized as the determinants of that particular type of knowledge ows. Tables 2 through 5 contain the results of these regression analyses. Knowledge outows to peer subsidiaries

485

Table 2 presents the results of regression analyses to test our predictions regarding the impact of value of knowledge stock (P1), motivational disposition (P2), and transmission channels (P3) on knowledge outows to peer subsidiaries. Value of knowledge stock. In the context of knowledge outows to peer subsidiaries, we operationalized this construct in terms of mode of entry and subsidiary size. The results in Table 2 (equation 4) support both of the resulting hypotheses. More specically, knowledge outows to peer subsidiaries are higher in the case of (i) subsidiaries that were acquired rather than set up as greeneld operations (beta for mode of

Table 2. Determinants of knowledge outows to peer subsidiaries dependent variable = Knowledge outows to peer subsidiaries (KO-S) Independent Variables Hypothesized Relationship Equation 1 Japan U.K. Sweden Finland R&D Intensity Fixed Asset Intensity Advertising Intensity Upstream Activities1 Downstream Activities2 P1: Value of Knowledge Stock Mode of Entry3 Subsidiary Size P2: Motivational Disposition Incentive Focus4 P3: Transmission Channels Formal Integrative Mechanisms Lateral Socialization Mechanisms R2 d.f. F R2 d.f. F
1 2 3

Standardized Beta Coefcients Equation 2 0.244*** 0.003 0.031 0.058 0.069 0.086 0.128* Equation 3 0.184** 0.003 0.043 0.045 0.046 0.078 0.130* 0.070 0.195*** Equation 4 0.153** 0.061 0.080 0.019 0.032 0.001 0.155* 0.005 0.187*** 0.127** 0.169** 0.003 0.256*** 0.100* 0.256 14,325 8.00*** 0.117 5,325 10.27***

0.233*** 0.041 0.046 0.065

H1a (+) H1b (+) H2a (+) H3a (+) H3b (+) 0.072 4,335 6.50*** 0.072 4,335 6.50*** 0.095 7,332 4.99*** 0.023 3,332 2.84* 0.139 9,330 5.91*** 0.043 2,330 8.33***

1 = Subsidiary has an upstream activity (R&D and/or manufacturing); 0 = Subsidiary has no upstream activity. 1 = Subsidiary has a downstream activity (marketing and sales); 0 = Subsidiary has no downstream activity. Mode of entry: 1 = Acquisition; 0 = Greeneld. 4 Higher values signify that the incentive system is more network, rather than subsidiary, focused. *p 0.05 **p 0.01 ***p 0.001 For t-tests, these are one-tail values.
Copyright 2000 John Wiley & Sons, Ltd. Strat. Mgmt. J., 21: 473496 (2000)

486

A. K. Gupta and V. Govindarajan

Table 3. Determinants of knowledge outows to the parent corporation Dependent variable = Knowledge outows to the parent corporation (KO-P) Independent Variables Hypothesized Relationship Equation 5 Japan U.K. Sweden Finland R&D Intensity Fixed Asset Intensity Advertising Intensity Upstream Activities1 Downstream Activities2 P1: Value of Knowledge Stock Mode of Entry3 Subsidiary Size Relative Economic Level P2: Motivational Disposition Incentive Focus4 P3: Transmission Channels Formal Integrative Mechanisms Lateral Socialization Mechanisms R2 d.f. F R2 d.f. F
1 2 3

Standardized Beta Coefcients Equation 6 0.141* 0.062 0.035 0.092 0.043 0.089 0.120 Equation 7 0.095 0.063 0.030 0.079 0.031 0.081 0.120 0.002 0.181*** Equation 8 0.088 0.038 0.021 0.135 0.002 0.063 0.129* 0.033 0.174*** 0.063 0.121* 0.169** 0.018 0.208*** 0.073 0.162 15,311 4.02*** 0.078 6,311 4.84***

0.127* 0.089 0.018 0.020

H1a (+) H1b (+) H1c (+) H2a (+) H3a (+) H3b (+) 0.033 4,322 2.75* 0.033 4,322 2.75* 0.054 7,319 2.58** 0.020 3,319 2.30 0.084 9,317 3.23*** 0.030 2,317 5.28**

1 = Subsidiary has an upstream activity (R&D and/or manufacturing); 0 = Subsidiary has no upstream activity. 1 = Subsidiary has a downstream activity (marketing and sales); 0 = Subsidiary has no downstream activity. Mode of entry: 1 = Acquisition; 0 = Greeneld. 4 Higher values signify that the incentive system is more network, rather than subsidiary, focused. *p 0.05 **p 0.01 ***p 0.001 For t-tests, these are one-tail values.

entry = 0.127, p 0.01; thus, H1a is supported), and (ii) subsidiaries that are larger in size (beta for subsidiary size = 0.169, p 0.01; thus, H1b is supported). Motivational disposition. In the context of knowledge outows to peer subsidiaries, we operationalized this construct in terms of the network vs. subsidiary focus of the incentive system for the subsidiary president. The results in Table 2 (Equation 4) do not support the resulting hypothesis (H2a). Transmission channels. In the context of knowledge outows to peer subsidiaries, we operationalized this construct in terms of formal integrative mechanisms and lateral socialization mechanisms. The results in Table 2 (Equation 4) support both of the resulting hypotheses. More
Copyright 2000 John Wiley & Sons, Ltd.

specically, knowledge outows to peer subsidiaries are higher in the case of (i) subsidiaries that are integrated more tightly with the rest of the corporation through formal mechanisms (beta for formal integrative mechanisms = 0.256, p 0.001; thus, H3a is supported), and (ii) subsidiaries whose presidents have been involved in lateral socialization mechanisms with peer subsidiaries to a greater extent (beta for lateral socialization mechanisms = 0.100, p 0.05; thus, H3b is supported). Knowledge outows to the parent corporation Table 3 presents the results of regression analyses to test our predictions regarding the impact of
Strat. Mgmt. J., 21: 473496 (2000)

Knowledge Flows within Multinational Corporations

Table 4. Determinants of knowledge inows from peer subsidiaries dependent variable = Knowledge inows from peer subsidiaries (KI-S) Independent Variables Hypothesized Relationship Equation 9 Japan U.K. Sweden Finland R&D Intensity Fixed Asset Intensity Advertising Intensity Upstream Activities1 Downstream Activities2 P4: Transmission Channels Formal Integrative Mechanisms Lateral Socialization Mechanisms P5: Motivational Disposition Incentive Focus3 P6: Absorptive Capacity Mode of Entry4 Local Nationals in TMT R2 d.f. F R2 d.f. F
1 2

487

Standardized Beta Coefcients Equation 10 0.362*** 0.166** 0.084 0.216** 0.202*** 0.098 0.029 Equation 11 0.349*** 0.164** 0.084 0.212** 0.200*** 0.095 0.030 0.017 0.063 Equation 12 0.231** 0.169** 0.087 0.195** 0.164** 0.062 0.032 0.026 0.055 0.167*** 0.110* 0.015 0.071 0.101 0.167 14,331 4.73*** 0.048 5,331 3.80**

0.333*** 0.087 0.064 0.088

H4a (+) H4b (+) H5a () H6a () H6b () 0.085 4,341 7.90*** 0.085 4,341 7.90*** 0.115 7,338 6.29*** 0.030 3,338 3.88** 0.119 9,336 5.04*** 0.004 2,336 0.69

1 = Subsidiary has an upstream activity (R&D and/or manufacturing); 0 = Subsidiary has no upstream activity. 1 = Subsidiary has a downstream activity (marketing and sales); 0 = Subsidiary has no downstream activity. 3 Higher values signify that the incentive system is more network, rather than subsidiary, focused. 4 Mode of entry: 1 = Acquisition; 0 = Greeneld. *p 0.05 **p 0.01 ***p 0.001 For t-tests, these are one-tail values.

value of knowledge stock (P1), motivational disposition (P2), and transmission channels (P3) on knowledge outows to the parent corporation. Value of knowledge stock. In the context of knowledge outows to the parent corporation, we operationalized this construct in terms of mode of entry, subsidiary size, and relative economic level. The results in Table 3 (Equation 8) support two of the resulting three hypotheses. More specically, knowledge outows to the parent corporation are higher in the case of (i) subsidiaries that are larger in size (beta for subsidiary size = 0.121, p 0.05; thus, H1b is supported), and (ii) subsidiaries that are located in countries with a higher level of economic advancement relative to the country of the parent corporation (beta for relative economic level = 0.169, p 0.01; thus, H1c is supported). There was no
Copyright 2000 John Wiley & Sons, Ltd.

support for our prediction regarding the impact of mode of entry on KO-P (H1a). Motivational disposition. In the context of knowledge outows to the parent corporation also, we operationalized this construct in terms of the network vs. subsidiary focus of the incentive system for the subsidiary president. The results in Table 3 (Equation 8) do not support the resulting hypothesis (H2a). Transmission channels. In the context of knowledge outows to the parent corporation, we operationalized this construct in terms of formal integrative mechanisms and vertical socialization mechanisms. The results in Table 3 (Equation 8) support one of the two resulting hypotheses. More specically, knowledge outows to the parent corporation are higher in the case of subsidiaries that are integrated more tightly with the rest of
Strat. Mgmt. J., 21: 473496 (2000)

488

A. K. Gupta and V. Govindarajan

Table 5. Determinants of knowledge inows from the parent corporation dependent variable = Knowledge inows from the parent corporation (KI-P) Independent Variables Hypothesized Relationship Equation 13 Japan U.K. Sweden Finland R&D Intensity Fixed Asset Intensity Advertising Intensity Upstream Activities1 Downstream Activities2 P4: Transmission Channels Formal Integrative Mechanisms Vertical Socialization Mechanisms P5: Motivational Disposition Incentive Focus3 Relative Economic Level HQ-Subsidiary Decentralization P6: Absorptive Capacity Mode of Entry4 Local Nationals in TMT R2 d.f. F R2 d.f. F
1 2

Standardized Beta Coefcients Equation 14 0.042 0.278*** 0.044 0.501*** 0.052 0.151** 0.205** Equation 15 0.062 0.276*** 0.040 0.506*** 0.063 0.153** 0.206** 0.046 0.039 Equation 16 0.073 0.086 0.011 0.256*** 0.014 0.095 0.168** 0.009 0.033 0.182*** 0.119* 0.097* 0.209*** 0.086* 0.165*** 0.009 0.298 16,313 8.31*** 0.114 7,313 7.28***

0.020 0.245*** 0.077 0.319***

H4a (+) H4b (+) H5a () H5b () H5c () H6a () H6b () 0.120 4,325 11.11*** 0.120 4,325 11.11*** 0.180 7,322 10.10*** 0.060 3,322 7.83*** 0.184 9,320 8.01*** 0.004 2,320 0.74

1 = Subsidiary has an upstream activity (R&D and/or manufacturing); 0 = Subsidiary has no upstream activity. 1 = Subsidiary has a downstream activity (marketing and sales); 0 = Subsidiary has no downstream activity. 3 Higher values signify that the incentive system is more network, rather than subsidiary, focused. 4 Mode of entry: 1 = Acquisition; 0 = Greeneld. *p 0.05 **p 0.01 ***p 0.001 For t-tests, these are one-tail values.

the corporation through formal mechanisms (beta for formal integrative mechanisms = 0.208, p 0.001; thus, H3a is supported). There was no support for our prediction regarding the impact of vertical socialization mechanisms on KO-P (H3b). Knowledge inows from peer subsidiaries Table 4 presents the results of regression analyses to test our predictions regarding the impact of transmission channels (P4), motivational disposition (P5), and absorptive capacity (P6) on knowledge inows from peer subsidiaries. Transmission channels. In the context of knowledge inows from peer subsidiaries, we operationalized this construct in terms of formal
Copyright 2000 John Wiley & Sons, Ltd.

integrative mechanisms and lateral socialization mechanisms. The results in Table 4 (Equation 12) support both of the resulting hypotheses. More specically, knowledge inows from peer subsidiaries are higher in the case of (i) subsidiaries that are integrated more tightly with the rest of the corporation through formal mechanisms (beta for formal integrative mechanisms = 0.167, p 0.001; thus, H4a is supported), and (ii) subsidiaries whose presidents have been involved in lateral socialization mechanisms with peer subsidiaries to a greater extent (beta for lateral socialization mechanisms = 0.110, p 0.05; thus, H4b is supported). Motivational disposition. In the context of knowledge inows from peer subsidiaries, we operationalized this construct in terms of the netStrat. Mgmt. J., 21: 473496 (2000)

Knowledge Flows within Multinational Corporations

work vs. subsidiary focus of the incentive system for the subsidiary president. The results in Table 4 (Equation 12) do not support the resulting hypothesis (H5a). Absorptive capacity. In the context of knowledge inows from peer subsidiaries, we operationalized this construct in terms of mode of entry and proportion of local nationals in the subsidiarys top management team. The results in Table 4 (Equation 12) do not support the resulting hypotheses (H6a and H6b). Knowledge inows from the parent corporation Table 5 presents the results of regression analyses to test our predictions regarding the impact of transmission channels (P4), motivational disposition (P5), and absorptive capacity (P6) on knowledge inows from the parent corporation. Transmission channels. In the context of knowledge inows from the parent corporation, we operationalized this construct in terms of formal integrative mechanisms and vertical socialization mechanisms. The results in Table 5 (Equation 16) support both of the resulting hypotheses. More specically, knowledge inows from the parent corporation are higher in the case of (i) subsidiaries that are integrated more tightly with the rest of the corporation through formal mechanisms (beta for formal integrative mechanisms = 0.182, p 0.001; thus, H4a is supported), and (ii) subsidiaries whose presidents have been involved in vertical socialization mechanisms with corporate HQ to a greater extent (beta for vertical socialization mechanisms = 0.119, p 0.05; thus, H4b is supported). Motivational disposition. In the context of knowledge inows from the parent corporation, we operationalized this construct in terms of the network vs. subsidiary focus of the incentive system for the subsidiary president, relative economic level, and HQ-subsidiary decentralization. The results in Table 5 (Equation 16) support all three of the resulting hypotheses. More specically, knowledge inows from the parent corporation are higher in the case of (i) subsidiaries whose presidents operate under more subsidiaryfocused, rather than network-focused, incentives (beta for incentive focus = 0.097, p 0.05; thus, H5a is supported), (ii) subsidiaries that are located in countries with a lower level of ecoCopyright 2000 John Wiley & Sons, Ltd.

489

nomic advancement relative to the country of the parent corporation (beta for relative economic level = 0.209, p 0.001; thus, H5b is supported), and (iii) subsidiaries that are given less decision-making autonomy by corporate headquarters (beta for HQ-subsidiary decentralization = 0.086, p 0.05; thus, H5c is supported). Absorptive capacity. In the context of knowledge inows from the parent corporation, we operationalized this construct in terms of mode of entry and proportion of local nationals in the subsidiarys top management team. The results in Table 5 (Equation 16) support only the rst of the two resulting hypotheses. More specically, knowledge inows from the parent corporation are higher in the case of subsidiaries that were set up as greeneld operations rather than acquired (beta for mode of entry = 0.165, p 0.001; thus, H6a is supported).

DISCUSSION
Pursuing a nodal level of analysis, this study has investigated both theoretically and empirically the determinants of intra-MNC knowledge ow patterns. While previous studies have focused more narrowly on selected facets of intra-MNC knowledge transfer e.g., tacitness of know-how (Teece, 1977; Zander and Kogut, 1995), and normative integration and inter-subsidiary communication (Ghoshal and Bartlett, 1988), this study has advanced and adopted a more comprehensive theoretical approach. Building on communication theory, we have argued that a complete mapping of the knowledge transfer process requires attention to all of the following ve major elements: (i) value of the knowledge possessed by the source unit, (ii) motivational disposition of the source unit regarding the sharing of its knowledge, (iii) the existence, quality, and cost of transmission channels, (iv) motivational disposition of the target unit regarding acceptance of incoming knowledge, and (v) the target units absorptive capacity for the incoming knowledge. Further, unlike previous studies on intra-MNC knowledge transfers, we have conducted separate examinations of knowledge ows that occur laterally among peer subsidiaries and those which occur hierarchically between a subsidiary and the parent corporation. Given the ongoing devolution
Strat. Mgmt. J., 21: 473496 (2000)

490

A. K. Gupta and V. Govindarajan

to the subsidiaries has also made them more systematic (as distinct from stochastic or experimental) in managing these particular types of knowledge ows. Focusing now on the empirical validity of our overarching theoretical framework, we also note from Tables 25 that the results support our expectations regarding the importance of four of the ve main constructs underlying this framework. More specically, the results provide either complete or partial support to our predictions regarding the impact of value of knowledge stock and transmission channels on knowledge outows; similarly, they also provide either complete or partial support to our predictions regarding the impact of transmission channels, motivational disposition to acquire knowledge, and absorptive capacity on knowledge inows. However, they do not provide any support to our predictions regarding the impact of motivational disposition to share knowledge with other units on knowledge outows. There are at least two possible explanations for this lack of support: (i) a subsidiarys motivational disposition to share knowledge may depend not only on the incentive system but also on other variables not examined in this study, and/or (ii) in the knowledge transfer process, the motivation of the target unit to acquire knowledge may be far more important than the motivation of the source unit to share its knowledge. An examination of the validity of any of these or other possible explanations must await future research. At a more micro-level, a closer examination of the 8 hypotheses (out of the total of 23 hypotheses) that were not supported reveals that 3 pertained to incentive focus, 2 to mode of entry, 2 to proportion of local nationals in subsidiarys TMT, and 1 to vertical socialization mechanisms. Alternatively stated, results failed to support 3 out of the 4 hypotheses dealing with incentive focus, 2 out of the 4 dealing with mode of entry, 2 out of the 2 dealing with proportion of local nationals in subsidiarys TMT, and 1 out of the 2 dealing with vertical socialization mechanisms. There are at least three possible explanations for this lack of support: (i) logical errors in developing the hypotheses, and/or (ii) substitution effects among the independent variables, and/or (iii) irreducible noise in the data. Our conjecture at this stage would be that the last two explanations represent the
Strat. Mgmt. J., 21: 473496 (2000)

of authority and responsibility from the center to the subsidiaries and the ability of information technology to enable direct communication among subsidiaries, we would agree with Bartlett and Ghoshal (1989), Hedlund (1994), Martinez and Jarillo (1989), and others that direct intersubsidiary interactions are becoming increasingly important. Utilizing the overarching theoretical framework and the broad propositions depicted in Figure 1, we advanced a set of hypotheses for each of the following four types of knowledge transfer contexts: (i) knowledge outows to peer subsidiaries, (ii) knowledge outows to the parent corporation, (iii) knowledge inows from peer subsidiaries, and (iv) knowledge inows from the parent corporation. These hypotheses were tested with data collected from the presidents of 374 subsidiaries of 75 MNCs headquartered in the U.S., Europe, and Japan. All hypotheses were tested after controlling for the effects of country-oforigin, the resource characteristics of the MNCs industry, and the nature of the subsidiarys operations. Commentary on the results As can be seen from Tables 25 (across-table comparions of R2 and R2 values as well as the number of signicant beta coefcients), our data had the greatest success in uncovering the determinants of KI-P i.e., knowledge inows to focal subsidiaries from the parent corporation. In this context, it may be useful to recall our earlier observation that, for the sample as a whole, of the four types of knowledge ows, the magnitude of KI-P was signicantly greater than that of each of the other three types of ows. These two empirical observations lead us to draw the following conjectures: (i) Of the four types of knowledge ows examined in this study, the typical MNC has perhaps had the longest experience in undertaking knowledge outows from the center to the units; (ii) Notwithstanding the fact that MNCs are indeed becoming heterarchies (Hedlund, 1994) i.e., integrated complex networks with signicant devolution of authority and responsibility to the subsidiaries, the parent corporation continues to serve as the most active creator and diffuser of knowledge within the corporation; and (iii) MNCs greater experience in managing knowledge outows from the parent
Copyright 2000 John Wiley & Sons, Ltd.

Knowledge Flows within Multinational Corporations

more likely scenario. Nonetheless, any denitive explanations for the lack of support also must await future research. Limitations of the study We can identify three major limitations of this study. First, since every MNC is a network (Ghoshal and Bartlett, 1990), all intra-MNC knowledge transfers take place in the context of the network. As contrasted with dyadic or systemic approaches to the examination of network-related phenomena, we conducted our examination at the nodal level of analysis the simplest level feasible. In the next subsection focusing on directions for future research, we identify some of the important questions that were not explored by us but which can be examined through future work that looks at knowledge transfers from a dyadic or a systemic perspective. Second, despite the fact that, in this study, we focused on largely procedural types of knowledge which, on average, tends to be more tacit than declarative knowledge, we neither measured nor explored the impact of degrees of tacitness. Notwithstanding the pioneering studies of Teece (1977), Zander and Kogut (1995), and others, empirical research into how degrees of tacitness affect the knowledge transfer process is still in its infancy. Finally, the third major limitation of this study has to do with the use of perceptual instruments to measure the extent of knowledge outows and inows. Barring the case of certain types of codiable technology transfers (as in the case of technology licenses), this is a methodological challenge that researchers have yet to overcome. In our view, researchers face at least two hurdles in measuring the extent of knowledge transfers through objective data: (i) Unlike transfers of codied knowledge, the transfers of tacit knowledge leave at best partial objective traces that could be measured by an external researcher; and (ii) Because transfers of tacit knowledge tend to be slow, any real-time investigation of this phenomenon would often require the researcher to undertake a multi-year study of each transfer; by way of example, note that Zander and Kogut (1995) reported that, in their sample, the median time to transfer was ve years and, without correcting for censored observations, the average was eight years. It is also instructive to note that,
Copyright 2000 John Wiley & Sons, Ltd.

491

notwithstanding their excellent access to the MNCs being studied, even Ghoshal and Bartlett (1988: 382) felt compelled to observe: Collecting objective level measures for the relatively large number of variables for meaningful statistical analysis represented enormous and, for us, insurmountable practical problems. Directions for future research As we observed at the beginning of the paper, the creation, diffusion, and absorption of knowledge by organizations in general and, by MNCs in particular, constitutes one of the most important subjects for research in the elds of organization theory (Levitt and March, 1988: Huber, 1991), strategic management (Prahalad and Hamel, 1994), evolutionary economics (Nelson and Winter, 1982), and international business (Buckley and Casson, 1976; Ghoshal and Bartlett, 1988; Kogut and Zander, 1993; Teece, 1977). Conceptual work in this area is still in the early stages and empirical work is almost literally at the stage of infancy. Thus, although we view the contributions of this study as important, in light of future possibilities, we view them as at best modest. There are several promising directions for future research. First, we believe that the payoffs from future investigations at the dyadic and/or systemic levels are likely to be high. At the dyadic level of analysis, at least two of the important issues to investigate would be: (i) the impact of bilateral homophily (Lazarsfeld and Merton, 1964) on dispositions to engage in outows and inows, and (ii) the importance of reciprocity i.e., is As disposition to share its knowledge with B dependent on Bs disposition to share its knowledge with A? At the systemic level of analysis, some of the important issues would be: (i) the impact of a units network centrality on the extent of knowledge outows as well as knowledge inows, (ii) the impact of network density on the overall magnitude of knowledge ows through the network, and (iii) the impact of global competitive intensity faced by the MNC on the magnitude and the directionality of knowledge ows. A second line of productive inquiry would be to compare and contrast what we would term as complementary vs. substitutive knowledge transfer contexts. By complementary contexts, we refer to the transfer of knowledge along different
Strat. Mgmt. J., 21: 473496 (2000)

492

A. K. Gupta and V. Govindarajan

empirical investigations into how tacit knowledge gets tranferred and the extent to which its transfer does or does not require ex ante codication is all too rare. Thus, our advocacy would be to urge greater efforts towards empirical rather than conceptual studies on the topic of tacitness. Finally, a productive line of inquiry would also be to examine the joint (i.e., interactive) effects of capability, motivation, and transmission channels on knowledge ow patterns. Given that research on knowledge ows within MNCs is still in its infancy, in this study, we focused exclusively on the main effects of these constructs. Nonetheless, since the results of this study lend support to the validity of our framework, a logical next step would be to develop and test more complex theoretical models.

stages in the companys value chain e.g., the transfer of technical knowledge from the development laboratories to the factories and the marketing units and the transfer of market knowledge from the eld back to the facories and the laboratories; in these instances, knowledge transfers occur in contexts where the source and the target units possess complementary knowledge stocks. In contrast, substitutive knowledge transfer contexts can be said to exist when the source and the target units engage in identical or similar activities (e.g., two laboratories, or two factories, or two sales units) and the transfer involves the imposition of the source units superior knowhow over that of the targets allegedly inferior knowhow. We would expect that the motivational dispositions of both the source and the target units are likely to be radically different in the case of complementary vs. substitutive knowledge transfers. A third line of productive inquiry would be a deeper application and examination of the overarching framework advanced in this paper. There are many other possible determinants of the value of a source units knowledge stock e.g., the resource base of the unit, the internal organization of the unit, and the competitive environment in the host country. Similarly, there are many other possible determinants of motivational dispositions to engage in inows or outows e.g., personal characteristics of subsidiary managers such as age or locus of control, their organizational commitment, and so forth. This is also true for the other elements in our model viz., transmission channels and absorptive capacity. In the case of transmission channels, the impact of communication mechanisms including the use of electronic media is an obvious topic for future research. Similarly, future investigations into how absorptive capacity of a receiving unit is affected not merely by its existing knowledge stock but also by its internal organization are likely to yield valuable insights. For example, building on Cohen and Levinthal (1990), it should be useful to examine the impact of intra-subsidiary communication as well as a subsidiarys activism at knowledge creation on its capacity to absorb incoming knowledge. A fourth line of productive inquiry would be to go deeper into the question of tacitness. It seems to us that, while the conceptual literature on how the tacitness of knowledge affects its transfer is notable for its abundance, systematic
Copyright 2000 John Wiley & Sons, Ltd.

ACKNOWLEDGEMENTS
Paritial funding support for this study was provided by the Center for International Business Education and Research (The University of Maryland at College Park), The Amos Tuck School of Business Administration (Dartmouth College), and The International Management Research Institute (Tokyo). The authors have beneted from comments on earlier versions of this paper from Robert Burgelman, Ranjay Gulati, Lee Preston, M. Susan Taylor, as well as participants at the 1997 Strategy Conference at Stanford University.

REFERENCES
Allport, G. W. (1937). Personality: A Psychological Interpretation. Holt, New York. Barney, J. (1991). Firm resources and sustained competitive advantage, Journal of Management, 17, pp. 99120. Bartlett, C. A. and S. Ghoshal (1989). Managing across Borders: The Transnational Solution. Harvard Business School Press, Boston, MA. Buckley, P. J. and M. Casson (1976). The Future of the Multinational Enterprise: Holmes & Meier, New York, pp. 6684. Caves, R. E. (1971). International corporations: The industrial economics of foreign investment, Economica, 38, pp. 127. Caves, R. E. (1982). Multinational Enterprise and Economic Analysis. Cambridge University Press, Cambridge, U.K. Cohen, W. M. and D. A. Levinthal (1990). Absorptive
Strat. Mgmt. J., 21: 473496 (2000)

Knowledge Flows within Multinational Corporations

capacity: A new perspective on learning and innovation, Administrative Science Quarterly, 35, pp. 128152. Collis, D. and P. Ghemawat (1994). Industry analysis: Understanding industry structure and dynamics. In L. Fahey and R. M. Randall (eds.), The Portable MBA in Strategy. John Wiley, New York, pp. 171194. Cyert, R. M. (1995). Management of knowledge. Keynote address at the Carnegie Bosch Institutes 1995 International Conference on High Performance Global Corporations. Excerpted in Global View, Newsletter of the Carnegie Bosch Institute for Applied Studies in Management, The CarnegieMellon University. Daft, R. L. and R. H. Lengel (1986). Organizational information requirements, media richness, and structural design, Management Science, 32, pp. 554571. Dierickx, I. and K. Cool (1989). Asset stock accumulation and sustainability of competitive advantage, Management Science, 35, pp. 15041514. DiMaggio, P. J. and W. W. Powell (1983). The iron cage revisited: Institutional isomorphism and collective rationality in organizational elds, American Sociological Review, 48, pp. 147160. Edstrom, A. and J. R. Galbraith (1977). Transfer of managers as a coordination and control strategy in multinational organizations, Administrative Science Quarterly, 22, pp. 248263. Egelhoff, W. G. (1984). Patterns of control in U.S., U.K., and European multinational corporations, Journal of International Business Studies, 15(3), pp. 7383. Egelhoff, W. G. (1988). Organizing the Multinational Enterprise: An Information-processing Perspective. Ballinger, Cambridge, MA. Ford, J. D. and J. W. Slocum (1977). Size, technology, environment, and the structure of organizations, Academy of Management Review, 2, pp. 561575. Franko, L. G. (1976). The European Multinationals: A Renewed Challenge to American and British Big Business. Greylock, Stamford, CT. Galbraith, J. R. (1973). Designing Complex Organizations. Addison-Wesley, Reading, MA. Gates, S. R. and W. G. Egelhoff (1986). Centralization in headquarters-subsidiary relationships, Journal of International Business Studies, 17(2), pp. 7192. Ghoshal, S. (1987). Global strategy: An organizing framework, Strategic Management Journal, 8(5), pp. 425440. Ghoshal, S. and C. A. Bartlett (1988). Creation, adoption, and diffusion of innovation by subsidiaries of multinational corporations, Journal of International Business Studies, 19, pp. 365388. Ghoshal, S. and C. A. Bartlett (1990). The multinational corporation as an interorganizational network, Academy of Management Review, 15, pp. 603625. Ghoshal, S., H. Korine and G. Szulanski (1994). Interunit communication in multinational corporations, Management Science, 40(1), pp. 96110. Goedde, A. G. (1978). U.S. multinational manufacturCopyright 2000 John Wiley & Sons, Ltd.

493

ing rms: The determinants and effects of foreign investment, Ph.D. dissertation, Duke University. Grant, R. M. (1996). Toward a knowledge-based theory of the rm, Strategic Management Journal, Winter Special Issue, 17, pp. 109122. Grueber, W., D. Mehta and R. Vernon (1967). The R&D factor in international trade and international investment of U.S. industries, Journal of Political Economy, 75, pp. 2037. Gupta, A. K. and V. Govindarajan (1986). Resource sharing among SBUs: Strategic antecedents and administrative implications, Academy of Management Journal, 29, pp. 695714. Gupta, A. K. and V. Govindarajan (1991). Knowledge ows and the structure of control within multinational corporations, Academy of Management Review, 16(4), pp. 768792. Hedlund, G. (1994). A model of knowledge management and the N-form corporation, Strategic Management Journal, Summer Special Issue, 15, pp. 7390. Hennart, J.-F. and Y.-R. Park (1993). Greeneld vs. acquisitions: The strategy of Japanese investors in the United States, Management Science, 39, pp. 10541070. Hofstede, G. H. (1967). The Game of Budget Control. van Gorcum, London. Horst, T. (1972). Firm and industry determinants of the decision to invest abroad: An empirical study, Review of Economics and Statistics, 54, pp. 258266. Huber, G. (1991). Organizational learning: The contributing processes and the literatures, Organization Science, 2(1), pp. 88115. Hymer, S. H. (1960). The international operations of national rms: A study of direct foreign investment, Ph.D. dissertation, Massachusetts Institute of Technology. Jablin, F. M. (1979). Superior-subordinate communication: The state of the art, Psychological Bulletin, 5, pp. 12011222. Katz, R. and T. J. Allen (1982). Investigating the Not Invented Here (NIH) syndrome: A look at the performance, tenure and communication patterns of 50 R&D project groups, R&D Management, 12, pp. 719. Kindleberger, C. P. (1969). American Business Abroad: Six Lectures on Direct Investment. Yale University Press, New Haven, CT. Kogut, B. and U. Zander (1992). Knowledge of the rm, combinative capabilities, and the replication of technology, Organization Science, 3(2). pp. 383397. Kogut, B. and U. Zander (1993). Knowledge of the rm and the evolutionary theory of the multinational corporation, Journal of International Business Studies, 24, pp. 625645. Krone, K. J., F. M. Jablin and L. L. Putnam (1987). Communication theory and organizational communication. In K. J. Krone, L. L. Putnam, K. H. Roberts, and L. M. Porter (eds.), Handbook of Organizational Communication: An Interdisciplinary Perspective. Sage, Newbury Park, CA, pp. 1840. Lazarsfeld, P. F. and R. K. Merton (1964). Friendship as social process: A substantive and methodological
Strat. Mgmt. J., 21: 473496 (2000)

494

A. K. Gupta and V. Govindarajan

Sherif, M and H. Cantrill (1947). The Psychology of Ego Involvements, Social Attitudes and Identications. Wiley, New York. Simon, H. A. (1991). Bounded rationality and organizational learning, Organization Science, 2(1), pp. 125134. Smith, C. S. and D. P. Hamilton (18 December 1995). Chinas demands for technology sharing trouble American ofcials, executives, The Asian Wall Street Journal, p. 2. Szulanski, G. (1996). Exploring internal stickiness: Impediments to the transfer of best practice within the rm, Strategic Management Journal, Winter Special Issue, 17, pp. 2744. Teece, D. J. (1977). Technology transfer by multinational rms: The resource cost of transferring technological know-how, The Economic Journal, 87, pp. 242261. Teece, D. J. (1981). The market for know-how and the efcient international transfer of technology, Annals, AAPSS, 458, pp. 8196. Tung, R. L. (1982). Selection and training procedures of U.S., European, and Japanese multinationals, California Management Review, 25(1), pp. 5771. Tushman, M. L. (1977). Communications across organizational boundaries: Special boundary roles in the innovation process, Administrative Science Quarterly, 22, pp. 587605. Van Maanen, J. and E. H. Schein (1979). Toward a theory of organizational socialization. In B. M. Staw (ed.), Research in Organizational Behavior, Vol. 1. JAI Press, Greenwich, CT, pp. 209264. Vancil, R. F. (1980). Decentralization: Managerial Ambiguity by Design. Financial Executives Research Foundation, New York. Wernerfelt, B. (1984). A resource-based view of the rm, Strategic Management Journal, 5(2), pp. 171180. World Bank (1995). World Development Report. Oxford University Press, Oxford, UK. Yip, G. S., J. Roos and J. K. Johansson (1994). Effects of nationality on global strategy in major American, European, and Japanese MNCs, CIBER working paper No. 9402, Anderson Graduate School of Management, University of California at Los Angeles. Zander, U. and B. Kogut (1995). Knowledge and the speed of the transfer and imitation of organizational capabilities, Organization Science, 6(1), pp. 7692. Zeira, Y. (1986). Management development in ethnocentric multinational corporations, California Management Review, 18(5), pp. 3442.

analysis. In M. Berger et al. (eds.) Freedom and Control in Modern Society. Octagon, New York, pp. 2363. Levinthal, D. A. and J. G. March (1993). The myopia of learning, Strategic Management Journal, Winter Special Issue, 14, pp. 95112. Levitt, B. and J. G. March (1988). Organizational learning, Annual Review of Sociology, 14, pp. 319340. Lippman, S. A. and R. P. Rumelt (1982). Uncertain imitability: An analysis of interrm differences in efciency under competition, Bell Journal of Economics, 13, pp. 418438. Martinez, J. I. and J. C. Jarillo (1989). The evolution of research on coordination mechanisms in multinational corporations, Journal of International Business Studies, 20, pp. 489514. Miller, D., M. F. R. Kets de Vries and J. M. Toulouse (1982). Top executive locus of control and its relationship to strategy-making, structure, and environment, Academy of Management Journal, 25, pp. 237253. Nadler, D. A. and M. L. Tushman (1987). Strategic Organization Design. Scott, Foresman, New York. National Register (1991). International Directory of Corporate Afliations: 19901991. Wilmette, IL. Nelson, R. R. and S. G. Winter (1982). An Evolutionary Theory of Economic Change. Belknap Press, Cambridge, MA. Nonaka, I. (1994). A dynamic theory of organizational knowledge creation, Organization Science, 5(1), pp. 1437. Penrose, E. T. (1959). The Theory of the Growth of the Firm. Basil Blackwell, Oxford. Pfeffer, J. (1981). Power in Organizations. Pitman, Marsheld, MA. Pitts, R. A. (1974). Incentive compensation and organization design, Personnel Journal, 53, pp. 338344. Polanyi, M. (1966). The Tacit Dimension. Routledge & Kegan Paul, London. Porter, M. E. (1986). Competition in global industries: A conceptual framework. In M. E. Porter (ed.), Competition in Global Industries. Harvard Business School Press, Boston, MA, pp. 1560. Porter, M. E. (1994). Global strategy: Winning in the world-wide marketplace. In L. Fahey and R. M. Randall (eds.), The Portable MBA in Strategy. John Wiley, New York, pp. 108141. Prahalad, C. K. and G. Hamel (1994). Strategy as a eld of study: Why search for a new paradigm?, Strategic Management Journal, Summer Special Issue, 15, pp. 516. Rogers, E. M. (1995). Diffusion of Innovations. Free Press, New York. Root, F. R. (1987). Entry Strategies for International Markets. Lexington Books, Lexington, MA. Salter, M. S. (1973). Tailor incentive compensation to strategy, Harvard Business Review, 49(2), pp. 94102. Schendel, D. (1996). Editors introduction to the 1996 Winter Special Issue: Knowledge and the rm, Strategic Management Journal, Winter Special Issue, 17, pp. 14.
Copyright 2000 John Wiley & Sons, Ltd.

APPENDIX
Measurement of variables Independent variables Mode of entry. Subsidiary presidents were asked to indicate whether their subsidiary became a part
Strat. Mgmt. J., 21: 473496 (2000)

Knowledge Flows within Multinational Corporations

of this corporation as a result of an acquisition/merger (coded as 1) or whether the subsidiary was created as a greeneld operation (coded as 0). Summary statistics on this variable are: mean = 0.42, s.d. = 0.49). Subsidiary size. This variable was measured in terms of the number of employees in the subsidiary (mean = 908, s.d. = 1552). In order to dampen the high variability in size and achieve a more normal distribution, the natural logarithm of the number of employees was used to indicate subsidiary size in our analyses. Relative economic level. This variable was computed by dividing the per capita income for the host country (where the subsidiary is located) by that for the home country (the country-of-origin of the parent corporation). For each country, data on per capita income (i.e, gross national product per capita adjusted for purchasing power parity) were obtained from the World Development Report (World Bank, 1995). Summary statistics on this variable are: mean = 0.81, s.d. = 0.39. Incentive focus. Based on Gupta and Govindarajan (1986) and Salter (1973), the following question was posed to the subsidiary presidents: Your annual incentive bonus may depend solely on your subsidiarys performance or solely on the performance of a group of subsidiaries or some combination of both. Please indicate below how your incentive bonus was actually determined for the most recent year. Your answers should total 100%: (1) percentage of your incentive bonus that was based on your subsidiarys performance; (2) percentage of your incentive bonus that was based on the performance of a cluster of subsidiaries. Responses to the second item were used as a measure of the extent to which the incentive system was network-focused rather than subsidiary-focused (mean = 17.55, s.d. = 30.67). Formal integrative mechanisms. Based on Galbraith (1973), Nadler and Tushman (1987), and Miller, Kets de Vries, and Toulouse (1982), this variable was measured through a 3-item Likert-type 7-point scale (ranging from used rarely to used very frequently) that asked respondents to indicate the extent to which their subsidiary used liaison personnel, temporary task forces, and permanent teams to coordinate decisions and actions with sister subsidiaries. The nal measure was a weighted average of
Copyright 2000 John Wiley & Sons, Ltd.

495

responses to the three items where the most complex mechanism (permanent teams) was given a weight of 3, the intermediately complex mechanism (temporary task forces) was given a weight of 2, and the least complex mechanism (liaison personnel) was given a weight of 1. Summary statistics on this variable are: mean = 2.92, s.d. = 1.53. Lateral socialization mechanisms. This measure was adapted from Ghoshal and Bartlett (1988). Respondents were asked to provide yes or no answers to the following two questions: (1) Have you worked for one or more years in other subsidiaries of this corporation? and (2) Have you participated in executive development programs involving participants from several subsidiaries? For each respondent, the total count of yes responses was treated as a measure of participation in lateral socialization (mean = 1.08, s.d. = 0.77). Vertical socialization mechanisms. This measure also was adapted from Ghoshal and Bartlett (1988). Respondents were asked to provide yes or no answers to the following two questions: (1) Have you worked for one or more years at corporate headquarters in this corporation? (2) Do you have a mentor at corporate headquarters? For each respondent, the total count of yes responses was treated as a measure of participation in vertical socialization mechanisms (mean = 0.95, s.d. = 0.84). Headquarters-subsidiary decentralization. Following Vancil (1980), each respondent was provided with the following list of nine strategically relevant decisions: (i) formulation of your subsidiarys annual budget; (ii) discontinuing a major existing product or product line; (iii) investing in major plant and equipment to expand capacity for existing products; (iv) developing a major new product line; (v) increasing (beyond budget) the level of expenditure for advertising and promotion; (vi) changing the selling price on a major product or product line; (vii) increasing (beyond budget) the level of expenditure for research and development; (viii) buying from an outside vendor when the items required could be supplied by another unit of the country; and (ix) increasing (beyond budget) the number of personnel employed by your subsidiary. Using an approach similar to Hofstede (1967), for each of these decisions, each respondent was asked to indicate, on the following 5-point Likert scale,
Strat. Mgmt. J., 21: 473496 (2000)

496

A. K. Gupta and V. Govindarajan

advertising intensity (i.e., advertising expenses to sales ratio). All raw data were obtained from Standard & Poors Compustat PC+ Database and were averaged for two years: 1990 and 1991. Utilizing these raw data, the three measures of industry resource characteristics were computed as follows. First, we identied the dominant industry group at the 2-digit SIC code level for the parent corporation. Second, utilizing industrylevel data, for each 2-digit industry group, we computed the proportion of revenue contributed by each 4-digit industry segment within that industry group. Third, for each of these 4-digit industry segments, we computed measures of R&D intensity, xed asset intensity, and advertising intensity. Finally, using the proportion of revenues contributed by each 4-digit industry segment to its 2-digit industry group as weights, we then computed weighted average measures of these three resource characteristics at the 2-digit industry group level. These measures, computed at the level of the parent MNCs dominant industry group, were then applied to all of the subsidiaries in our sample belonging to that particular MNC. For the sample, summary statistics on these three industry resource characteristics are: R&D intensity (mean = 0.03, s.d. = 0.02), xed asset intensity (mean = 0.36, s.d. = 0.21), and advertising intensity (mean = 0.03, s.d. = 0.02). Nature of subsidiary operations. We measured subsidiary operations through two dummy variables: upstream activities and downstream activities. The variable upstream activities was coded as 1 if the subsidiary performed a primary upstream operation (R&D and/or manufacturing); otherwise, this variable was coded as 0. Similarly, the variable downstream activities was coded as 1 if the subsidiary performed a primary downstream operation (marketing and sales); otherwise this variable was coded as 0. The raw data for these two variables were obtained by asking each subsidiary president to provide yes or no answers to the following three questions: (i) Does your subsidiary have one or more research and development facilities?, (ii) Does your subsidiary have one or more manufacturing facilities? and (iii) Does your subsidiary have one or more marketing and sales facilities? Summary statistics on the two dummy variables are: upstream activities (mean = 0.75, s.d. = 0.43) and downstream activities (mean = 0.84, s.d. = 0.37).
Strat. Mgmt. J., 21: 473496 (2000)

the typical inuence that they had in affecting the outcome of the decision: (1) your opinion not asked but decision is explained to you; (2) proposal by superior, your opinion is asked and it carries little weight; (3) proposal by superior, your opinion is asked and it carries a lot of weight; (4) proposal by you, decision made jointly by you and your superior; and (5) proposal by you, followed by consultation with superior, with your opinion prevailing. Responses on the 9 questions were averaged to create an index of headquarters-subsidiary decentralization (Chronbach alpha = 0.86). Higher values on this measure indicate higher decentralization (mean = 4.04, s.d. = 0.74). Proportion of local nationals in the subsidiarys top management team. For managers heading each of seven positions, the subsidiary presidents were asked to indicate the nationality of each particular person on a four-point scale: local national, home country expatriate, third country expatriate, and not applicable implying that there was nobody heading such a position. The instrument also explained the precise meanings of these terms. The seven positions were: subsidiary president, head of marketing, head of manufacturing, head of R&D, head of nance, controller, and head of human resources. The percentage of applicable positions that were headed by local nationals was regarded as a measure of the extent to which the subsidiary top management team was localized (range = 0 to 100; mean = 63.87; s.d. = 38.48). Control variables Country-of-origin. Each MNC in this sample was headquartered in one of the following ve countries: U.S., Japan, U.K., Sweden, and Finland. Treating the U.S. as the base case, dummy variables were created for each of the other four countries of origin. For example, in the case of Japanese MNCs, the variable Japan was given a value of 1; in the case of non-Japanese MNCs, this variable was given a value of 0. A similar approach was followed for U.K., Sweden, and Finland. Industry resource characteristics. For each subsidiary, measures of industry resource characteristics were computed at the level of the parent corporations dominant industry group along three dimensions: R&D intensity (i.e., R&D expenses to sales ratio), xed asset intensity (i.e., net physical plant and equipment to sales ratio), and
Copyright 2000 John Wiley & Sons, Ltd.