gupta;2000;knowledge flows

8/4/2019 Gupta;2000;Knowledge Flows

1/24

Strategic Management JournalStrat. Mgmt. J., 21: 473496 (2000)

KNOWLEDGE FLOWS WITHIN MULTINATIONAL

CORPORATIONS

ANIL K. GUPTA1* and VIJAY GOVINDARAJAN2

1The Robert H. Smith School of Business, The University of Maryland, CollegePark, Maryland, U.S.A.2The Amos Tuck School of Business, Dartmouth College, Hanover, NewHampshire, U.S.A.

Pursuing a nodal (i.e., subsidiary) level of analysis, this paper advances and tests an overarchingtheoretical framework pertaining to intracorporate knowledge transfers within multinationalcorporations (MNCs). We predicted that (i) knowledge outflows from a subsidiary would be

positively associated with value of the subsidiarys knowledge stock, its motivational dispositionto share knowledge, and the richness of transmission channels; and (ii) knowledge inflows into

a subsidiary would be positively associated with richness of transmission channels, motivationaldisposition to acquire knowledge, and the capacity to absorb the incoming knowledge. These predictions were tested empirically with data from 374 subsidiaries within 75 MNCs headquar-tered in the U.S., Europe, and Japan. Except for our predictions regarding the impact ofsource units motivational disposition on knowledge outflows, the data provide either full or

partial support to all of the other elements of our theoretical framework. Copyright 2000John Wiley & Sons, Ltd.

In recent years, researchers in organization theory(Levitt and March, 1988), economics (Nelson andWinter, 1982), as well as strategic management(Prahalad and Hamel, 1994; Schendel, 1996) have

identified organizational learning as one of themost important subjects for scholarly inquiry.Aimed at further deepening our understanding ofa key topic within this broad area viz., intra-firm flows of organizational knowledge, this paperreports the results of a theoretical and empiricalinvestigation into the determinants of internalknowledge transfers within multinational corpo-rations. The following four observations underliethe motivations for this study.

First, every firm constitutes a bundle of knowl-edge. As a corollary of the resource-based viewof the firm (Barney, 1991; Penrose, 1959; Wer-

Key words: knowledge flows, multinational corpo-rations, subsidiaries*Correspondence to: Anil K. Gupta, The Robert H. Smith

School of Business, The University of Maryland, CollegePark, MD 20742, U.S.A.

Received 27 August 1997Copyright 2000 John Wiley & Sons, Ltd. Final revision received 1 August 1999

nerfelt, 1984), this observation is now so widelyaccepted 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 particularlyimportant to note that, of all possible resourcesthat a firm might possess, its knowledge base hasperhaps the greatest ability to serve as a source ofsustainable differentiation and hence competitiveadvantage (Dierickx and Cool, 1989; Lippmanand Rumelt, 1982).

Second, the primary reason why MNCs existis because of their ability to transfer and exploitknowledge more effectively and efficiently in theintra-corporate context than through external mar-ket mechanisms. This internalization of intan-gible assets argument, originally advanced byHymer (1960), has been subjected to numerousconfirmatory empirical tests and is now widelyaccepted as the received theory on why MNCsexist (Buckley and Casson, 1976; Caves, 1971,1982; Ghoshal, 1987; Kindleberger, 1969; Porter,1986; Teece, 1981). Of course, external markets


2/24

474 A. K. Gupta and V. Govindarajan

continue to become more open, efficient, andglobal on an ongoing basis. Notwithstanding theincreasing sophistication of external markets, theyremain relatively ineffective mechanisms forknowledge transfer on at least two grounds: one,bulk of the specialized knowledge of any firm

exists in a tacit and thereby non-tradeable form;two, market-based transfers of knowledge areoften associated with negative externalities suchas involuntary expropriation and the risk of cre-ating a new competitor.

Third, the notion that MNCs exist primarilybecause of their superior ability (vis-a-vismarkets) to engage in internal knowledge transferdoes not in any way imply that such knowledgetransfers actually take place effectively andefficiently on a routine basis. In perhaps the onlystudy to date on the actual costs of cross-border

knowledge transfers, Teece (1981: 84) examineda sample of 26 technology transfer cases andreported that [T]he resource cost of internationaltransfer is nontrivial. Transfer costs ranged from2.25 percent to 59 percent of total project costswith a mean of 19.16 percent. The tacitnessor causal ambiguity of knowledge is one of themost widely recognized barriers to its transfer andreplication (Lippman and Rumelt, 1982; Polanyi,1966; Zander and Kogut, 1995). Levinthal andMarch ( 1993), Simon (1991), Szulanski (1996)and others have suggested additional barriers to

knowledge transfer e.g., barriers rooted in moti-vational dispositions and absorptive capacity.

Finally, notwithstanding the criticality of inter-nal knowledge transfers within MNCs, with somenotable exceptions (e.g., Ghoshal and Bartlett,1988 and Zander and Kogut, 1995), very littlesystematic empirical investigation into the deter-minants of intra-MNC knowledge transfers hasso far been attempted. As Ghoshal, Korine, andSzulanski (1994: 97) have observed, A numberof publications emphasize the importance of inter-unit communication for effective MNC man-agement

%

but in none of them is the constructoperationalized or measured, nor are the factorsthat influence such communication empiricallyexplored.

Building on these observations, the primaryobjective of this paper is to advance the state ofour theoretical as well as empirical understandingof the determinants of intra-MNC knowledgetransfers. Data for this study were collecteddirectly from the presidents of 374 subsidiaries

Copyright 2000 John Wiley & Sons, Ltd. Strat. Mgmt. J., 21: 473496 (2000)

belonging to 75 major MNCs headquartered inthe U.S., Japan, and Europe. In order to ensurereliability, data on the most critical variables(pertaining to knowledge transfers) were collectedalso from the immediate HQ-level superiors ofthe presidents of a large subset of the sampled

subsidiaries; further, the tests for the hypotheseswere conducted after controlling for the possibleeffects of the parent corporations country-of-origin, the resource characteristics of the parentcorporations industry, and the nature of the sub-sidiarys operations.

THE PHENOMENON OF INTEREST

Because MNCs are complex multi-dimensionalentities, knowledge flows within such enterprisesoccur not only along multiple directions but also

across multiple dimensions, e.g., the flow ofinformation pertaining to the Brazilian subsidi-arys financial performance over the last quarterto corporate headquarters, the transfer of packag-ing technology from a Swedish factory to one inIndia, or the transfer of customer service skillsfrom a Japanese subsidiary to one in the U.S. Inthis study, we focus on the transfer of largelyprocedural types of knowledge (e.g., productdesigns, distribution know-how, etc.) but not onthe transfer of largely declarative types of knowl-edge (e.g., monthly financial data). In other

words, this study focuses on the transfer of knowl-edge 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 andGovindarajan (1991), and Hedlund (1994) havesuggested, knowledge transfers within the MNCtake place within the context of an interorgani-zational network of differentiated units. Thus,flows of knowledge through the network can bestudied from at least three different levels ofanalysis: nodal (i.e., a focus on the behavior ofindividual units), dyadic (i.e., a focus on the jointbehavior of unit pairs), and systemic (i.e., a focuson the behavior of the entire network). Given thehighly complex nature of the phenomenon underinvestigation and the relative dearth of previousempirical work on it, in this study, we havechosen to limit our investigation to the nodallevel. More specifically, we focus on individualsubsidiaries only and examine the determinantsof knowledge flows in each of the following


3/24

Knowledge Flows within Multinational Corporations 475

four domains: (i) knowledge outflows to peersubsidiaries, (ii) knowledge outflows to the parentcorporation, (iii) knowledge inflows from peersubsidiaries, and (iv) knowledge inflows from theparent corporation.

THEORY

An overarching theoretical framework

As Krone, Jablin, and Putnam (1987) haveobserved in their review of communicationtheory, even though different communicationscholars have focused more (or less) heavily ondifferent elements of the communication process,virtually all of them recognize the following asthe basic elements of any two-person communi-cation: a message, a sender, a coding scheme,

a channel, transmission through the channel, adecoding scheme, a receiver, and the assignmentof meaning to the decoded message. Consistentwith these ideas from communication theory, weconceptualize knowledge flows (into or out of asubsidiary) to be a function of the following fivefactors: (i) value of the source units knowledgestock, (ii) motivational disposition of the sourceunit, (iii) existence and richness of transmissionchannels, (iv) motivational disposition of the tar-get unit, and (v) absorptive capacity of the targetunit. Barriers or facilitators to the transfer of

knowledge can manifest themselves in any or allof these five factors:

(a ) Value of source units knowledge stock.Knowledge flows across units are not costfree (Teece, 1981). We also know thatdifferent resources have different levels ofvalue (Barney, 1991). Thus, the greaterthe value of a subsidiarys knowledgestock for the rest of the MNC, the greaterwould be its attractiveness for other units.This idea is broadly consistent with theconcept of relative advantage in theliterature dealing with diffusion of inno-vations which has argued that the adoptionrate of an innovation is positively relatedto its relative advantage (Rogers, 1995).This idea has not yet been applied to theexamination of interunit knowledge trans-fers within multinational corporations.Within such corporations, we visualize theknowledge stock of any subsidiary as com-


posed of both duplicative as well as non-duplicative knowledge. The presence ofnon-duplicative knowledge is a necessary,although not sufficient, condition for suchknowledge to be of value to other units.Thus, we would anticipate that knowledge

outflows from a subsidiary are likely tobe high when the subsidiarys knowledgestock is non-duplicative as well as relevantfor the rest of the global network.

(b) Motivational disposition of the source unit.As Cyert (1995) has suggested, an organi-zational unit with uniquely valuable know-how is likely to enjoy an informationmonopoly within the corporation. Thisreality coupled with the fact that powerstruggles are a ubiquitous phenomenon inany organization (Pfeffer, 1981) implies

that at least some units will view uniquelyvaluable know-how as the currencythrough which they acquire and retain rela-tive power within the corporation. Levittand March (1988: 331) have observedsimilarly that In many (but not all) situ-ations%diffusion of experience has nega-tive consequences for organizations thatare copied. Therefore, we anticipate thatfactors which would enhance the moti-vational disposition of the source unit toshare its knowledge with other units within

the MNC are likely to counterbalance anyhoarding tendencies and thereby to havea positive impact on the magnitude ofknowledge outflows.

(c ) Existence and richness of transmissionchannels. As would be expected, and asdemonstrated empirically by Ghoshal andBartlett (1988) in the domain of MNCs,knowledge flows cannot occur without theexistence of transmission channels. Beyondmere existence, we would expect otherproperties of transmission channels to alsoaffect the extent of knowledge flows themost notable such property would be therichness/bandwidth of communicationlinks, as captured in aspects such as infor-mality, openness, and density of communi-cations (Daft and Lengel, 1986; Gupta andGovindarajan, 1991; Jablin, 1979; Tush-man, 1977).

(d) Motivational disposition of the target unit.The Not-Invented-Here (NIH) syndrome


4/24


is well-known and also has been the sub- ject of scholarly inquiry (Katz and Allen,1982). There are at least two drivers ofthe NIH syndrome: (i) ego-defense mecha-nisms (Allport, 1937; Sherif and Cantrill,1947) which can lead some managers to

block any information that might suggestthat others are more competent than theyare, and (ii) power struggles within organi-zations (Pfeffer, 1981) which can leadsome managers to try to downgrade thepotential power of peer units by pretendingthat the knowledge stock possessed bythese peer units is not unique and valuable.In short, unless counterveiling forces arepresent, the NIH syndrome can act as amajor barrier to the inflows of knowledgeinto any focal unit. These counterveiling

forces can manifest themselves in severalforms: the relative paucity of the focalunits knowledge stock, incentives thatincrease subsidiary managers eagerness tolearn from peer units, or coercive pressuresfrom corporate headquarters.

(e ) Absorptive capacity of the target unit.Even when exposed to the same environ-ment and even when there are insignificantdifferences in the desire to acquire newknowledge, individuals and organizationsmay differ in their absorptive capacity

i.e., in their ability to recognize the valueof new information, assimilate it, andapply it to commercial ends (Cohen andLevinthal, 1990: 128). There are at leasttwo reasons why absorptive capacity maydiffer across organizations: (i) the extentof prior related knowledge, and (ii) theextent of inter-unit homophily of thereceiving unit vis-a-vis the sending unit.Prior related knowledge is importantbecause it shapes the filters through whichthe organization differentiates betweenmore vs. less relevant signals and alsobecause it determines the organizationsability to internalize and assimilate themore valued signals (Cohen and Levinthal,1990). On the other hand, homophily i.e., the degree to which two or moreindividuals who interact are similar in cer-tain attributes, such as beliefs, education,social status, and the like (Rogers, 1995:1819) is important because when the


interacting individuals share commonmeanings, a mutual subcultural language,and are alike in personal and social charac-teristics, the communication of new ideasis likely to have greater effects in termsof knowledge gain, attitude formation, and

overt behavior change (Rogers, 1995: 19;see also Lazarsfeld and Merton, 1964).

Figure 1 presents a schematic diagram of theoverarching framework developed in this section.From the perspective of the nodal level ofanalysis being pursued in this study, this frame-work can be translated into the following sixpropositions:

Proposition 1: Ceteris paribus, the value of a

subsidiarys knowledge stock will be positively

associated with outflows of knowledge fromthat subsidiary.

Proposition 2: Ceteris paribus, the moti-

vational disposition of a subsidiary to share

its knowledge with other units will be positively

associated with outflows 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 outflows ofknowledge 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 inflows of

knowledge into that subsidiary.

Proposition 5: Ceteris paribus, the moti-

vational disposition of a subsidiary to

seek/accept knowledge from other units will

be positively associated with inflows of knowl-

edge into that subsidiary.

Proposition 6: Ceteris paribus, the capacity

of a subsidiary to absorb incoming knowledge

from other units will be positively associated

with inflows of knowledge into that subsidiary.

In the rest of this section, we operationalizethe constructs underlying these propositions and


5/24


Figure 1. Determinants of intra-corporate knowledge outflows from and inflows to foreign subsidiaries: Anoverarching theoretical framework

develop more concrete and empirically testablehypotheses.

Value of source units knowledge stock

We argued earlier that, in order for a sourceunits knowledge to be of value to other units,the source unit must (i) create non-duplicativeknowledge on its own, and (ii) this non-

duplicative knowledge must be of relevance forthe rest of the global network. Based on thisreasoning, we operationalize the construct ofvalue of knowledge stock in terms of the follow-ing three variables: mode of entry, subsidiarysize, and the economic level of the host countryrelative to that of the home country.

Mode of entry. As Caves (1982), Root (1987)and others have pointed out, an MNC may entera foreign country through one of several modes greenfield operations, strategic alliances, oracquisitions. Since our study focuses only onfully- or majority-owned subsidiaries, we examinehere the impact of greenfield vs. acquisitionmodes only. At a general level, we can visualizeevery subsidiary to consist of three bundles ofknowledge: duplicative knowledge, non-duplicative knowledge that is relevant only in thelocal environment, and non-duplicative knowledgethat is relevant also for other units within theglobal network.

As the literature on foreign direct investment


has argued and demonstrated (Hennart and Park,1993), the less the overlap between existingcorporate know-how and the know-how requiredto succeed in a host market, the greater theprobability of acquisition as the mode of entry.Thus, relative to greenfield subsidiaries, acquiredsubsidiaries on average can be expected to havea knowledge stock that is less duplicative vis-a-vis the knowledge stock of the rest of the corpo-

ration. It is true that only a subset of the non-duplicative knowledge would be of relevance forthe global network. However, since the pool ofnon-duplicative knowledge would be higher foracquired subsidiaries as compared to greenfieldsubsidiaries, it is likely that acquired subsidiariesshould have a larger pool of relevant knowledgeto offer to the global network than greenfieldsubsidiaries. Based on these arguments, Proposi-tion 1 can be operationalized in the form ofthe following two empirically testable hypotheses:ceteris paribus, relative to greenfield operations,

acquired subsidiaries will engage in greater

knowledge outflows to peer subsidiaries (H1a)

and to the parent corporation (H1a).

Subsidiary size. We anticipate that the typicalMNC would discourage investment of a subsidi-arys resources in the reinvention of knowledgethat exists elsewhere in the global network. Thus,we would expect that a subsidiarys ownresources would generally be directed at the cre-ation of non-duplicative knowledge. Since larger


6/24


subsidiaries will have a greater pool of resourcesdedicated to the creation of new knowledge, itfollows that subsidiary size should have a positiveimpact on the ability of the subsidiary to offernon-duplicative knowledge to the rest of thecorporation. Clearly, not all of the non-duplicative

knowledge generated by a subsidiary would haveglobal relevance; however, a subset of suchknowledge will. These arguments yield the fol-lowing additional operationalizations of Proposi-tion 1: ceteris paribus, the larger the size of asubsidiary, the greater will be the knowledge

outflows from that subsidiary to peer subsidiaries

(H1b) and to the parent corporation (H1b).

Relative economic level. Countries differ intheir levels of economic advancement. If we makethe straightforward assumption that most, perhapsall, societies around the world strive to increase

(rather than merely maintain or decrease) theirlevels of economic advancement, then it followsthat, on average, more advanced countries arelikely to serve as trend-setters and the sources oftechnological, marketing, as well as managerialknow-how to a greater extent than less advancedcountries. In other words, in the intracorporatecontext, on average, a focal unit is likely to viewthe knowledge stock of another unit located inan economically more advanced country relativeto itself as more valuable than that of a unitlocated in a relatively less advanced country.

These arguments also yield the following oper-ationalization of Proposition 1: ceteris paribus,the higher the level of the host countrys eco-

nomic development relative to the home country,

the greater will be the knowledge outflows from

that subsidiary to the parent corporation (H1c).

Since our empirical study was conducted at thenodal level of analysis, we did not collect anydata regarding knowledge flows between specificinter-subsidiary dyads. Accordingly, in the abovehypothesis, we have focused only on the relativeeconomic level of the focal subsidiary vis-a-visthe parent corporation and not on that vis-a-vis other specific subsidiaries. Thus, we neitheradvance nor test any inter-subsidiary hypothesespertaining to relative economic level.

Motivational disposition of the source unit

We posit that the extent to which the subsidiarypresident is rewarded for improvements in theperformance of a network of subsidiaries (rather


than just the focal subsidiary) would be a majordeterminant of motivation to share knowledgewith other subsidiaries. Based on this reasoning,we operationalized the construct of motivationaldisposition in terms of the subsidiary vs. corpo-rate focus (i.e., nodal vs. network optimization

focus) of the incentive system for the subsidi-ary president.Incentive focus. As Salter (1973) suggested

and as Gupta and Govindarajan (1986) and Pitts(1974) demonstrated, the incentive bonus for adivision/subsidiary general manager may belinked solely to the performance of the focal unit,solely to the performance of several units, or tosome combination of the two. As these authorshave argued, the greater the need to motivate aunit general manager to focus on system-wideoptimization as distinct from local optimization,

the better it is to link the incentives to theperformance of a cluster of units. These argu-ments result in the following operationalizationsof Proposition 2: ceteris paribus, the greater theextent to which a subsidiary presidents bonus is

network-focused rather than subsidiary-focused,


that subsidiary to peer subsidiaries (H2a) and to

the parent corporation (H2a).

Existence and richness of transmission

channelsAs communications theory informs us (Daft andLengel, 1986; Krone et al., 1987), transmissionchannels can be both formal and informal.Accordingly, we operationalize the construct oftransmission channels in terms of two mecha-nisms: one formal (viz., formal integrativemechanisms) and one informal (viz., corporatesocialization mechanisms).

Formal integrative mechanisms. Galbraith(1973) and Nadler and Tushman (1987) identifiedliaison positions, task forces, and permanent com-mittees as some of the key formal structuralmechanisms for integrating multiple units of anorganization. It is easy to see that the greater theextent to which a subsidiary is linked to the restof the global network through such integrativemechanisms, the greater would be the density ofcommunication interface between the subsidiaryand other units, thereby contributing positively tomedia richness (Daft and Lengel, 1986). Thus,focusing on knowledge outflows from the subsidi-


7/24


ary, we can now operationalize Proposition 3 interms of the following concrete hypotheses: cet-eris paribus, the greater the reliance on formal

mechanisms (liaison personnel, task forces, per-

manent committees) to integrate a subsidiary with

the rest of the MNC, the greater will be the

knowledge outflows from that subsidiary to peersubsidiaries (H3a) and to the parent corporation

(H3a). Similarly, focusing on knowledge inflowsinto the subsidiary, we can also operationalizeProposition 4 in terms of the following testablehypotheses: ceteris paribus, the greater thereliance on formal mechanisms (liaison person-

nel, task forces, permanent committees) to inte-

grate a subsidiary with the rest of the MNC, the

greater will be the knowledge inflows into that

subsidiary from peer subsidiaries (H4a) and from


Corporate socialization mechanisms. Corpor-ate socialization mechanisms refer to thoseorganizational mechanisms which build inter-personal familiarity, personal affinity, and conver-gence in cognitive maps among personnel fromdifferent subsidiaries (Edstrom and Galbraith,1977; Van Maanen and Schein, 1979). Greaterinterpersonal familiarity and personal affinity canbe expected to increase the openness of communi-cation between the interacting parties. Further, asDaft and Lengel (1986) have suggested, personaland more open communication increases the rich-

ness of communication channels. Thus, we wouldargue that greater participation in corporatesocialization mechanisms would have a positiveimpact on the richness of transmission channelsbetween the focal subsidiary and other units.

In this study, we separate lateral from verti-cal socialization mechanisms. Examples of theformer would be: job transfers to peer subsidiariesand participation in multi-subsidiary executiveprograms; similarly, examples of the latter wouldbe: job transfers to corporate headquarters andparticipation in corporate mentoring programs(Ghoshal and Bartlett, 1988). Focusing now onknowledge outflows from the focal subsidiary, wecan advance the following additional oper-ationalizations of Proposition 3: ceteris paribus,the greater the lateral socialization of a subsidi-

ary president, the greater will be the knowledge

outflows from that subsidiary to peer subsidiaries

(H3b); further, ceteris paribus, the greater the

vertical socialization of a subsidiary president,



that subsidiary to the parent corporation (H3b).

Similarly, focusing now on knowledge inflowsinto the focal subsidiary, we can advance thefollowing additional operationalizations of Propo-sition 4: ceteris paribus, the greater the lateralsocialization of a subsidiary president, the greater

will be the knowledge inflows 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 inflows into that subsidiary from the

parent corporation ( H4b).

Motivational disposition of the target unit

We argued earlier that a subsidiarys motivationaldisposition to acquire/accept knowledge fromother units within the enterprise would be a func-

tion of (i) incentives that increase subsidiarymanagers eagerness to learn, (ii) the relativepaucity of the subsidiarys knowledge stock,and/or (iii) coercive pressures from corporateheadquarters. Based on this reasoning, we oper-ationalized the construct of motivational disposi-tion of the target unit in terms of three variables:subsidiary vs. corporate focus of the incentivesfor the subsidiary president (a determinant ofeagerness to learn), relative economic level (adeterminant of the paucity of local knowledgestock), and HQ-subsidiary decentralization (a

determinant of coercive pressures).Incentive focus. Unlike the case of knowledge

outflows where the required motivational disposi-tion can be characterized as eagerness to helpothers, in the case of knowledge inflows, therequired motivation would be characterized aseagerness to learn and to help oneself. Wewould argue that, other things being equal, sub-sidiary personnel would be more eager to learnin those contexts where the linkage betweenincentives and the subsidiarys own capabilitiesis tighter rather than weaker i.e., in contextswhere incentives are linked more tightly to thefocal subsidiarys own performance than to theperformance of a cluster of subsidiaries. This isso because, unlike cluster-based incentives, whichcan create free-rider problems, subsidiary-basedincentives would create a stronger disposition tolearn from any and all sources. These argumentsyield the following operationalizations of Proposi-tion 5: ceteris paribus, the greater the extent towhich a subsidiary presidents bonus is subsidi-


8/24


ary-focused rather than network-focused, the

greater will be the knowledge inflows into that

subsidiary from peer subsidiaries (H5a) and from


Relative economic level. Paralleling our dis-cussion on this variable in the context of knowl-

edge outflows, we expect that, other things beingequal, the lower the level of economic advance-ment of the host country (i.e., where the sub-sidiary is located) vis-a-vis the home country(i.e., where the parent is located), the more eagersubsidiary personnel would be to learn from theparent corporation. They are likely to perceivethe knowledge stock of the parent as relativelymore valuable and, thus, are likely to regardknowledge inflows as a potential source of com-petitive advantage against other players in thelocal market. Knowledge inflows into such sub-

sidiaries may also be facilitated by explicit publicpolicy regimes that mandate technology inflowsas the condition for allowing MNCs access tothe local market; as an example, this is illustratedwell by the recent decisions of the Chinesegovernment (Smith and Hamilton, 1995: 2).These arguments suggest the following additionaloperationalization of Proposition 5: ceteris par-ibus, the lower the level of the host countrys

economic development relative to the home coun-

try, the greater will be the knowledge inflows

into the subsidiary from the parent corporation

(H5b). As discussed in the context of knowledgeoutflows, given our nodal level of analysis, weneither advance nor test any hypotheses pertainingto the relative economic levels of subsidiary pairs.

Headquarters-subsidiary decentralization.The concept of decentralization (or its obversei.e., centralization) has had a long history ofresearch in organization theory (see Ford andSlocum, 1977 for an extensive review). Even inthe domain of research on MNCs, scholars haveargued that centralization is one of the funda-mental dimensions of organization design(Egelhoff, 1988: 129). Our expectations of alinkage between decentralization and knowledgeinflows into a subsidiary parallel the broaderarguments of DiMaggio and Powell (1983),echoed also by Levitt and March (1988), thatcoercion is one of the major (but not the sole)drivers of inter-organizational isomorphism. Inthe MNC context also, similar arguments havebeen advanced by many scholars (e.g., Gates andEgelhoff, 1986; Ghoshal and Bartlett, 1988).


These arguments yield the following additionaloperationalization of Proposition 5: ceteris par-ibus, the lower the decentralization of decision-

making authority to a subsidiary, the greater will

be the knowledge inflows into that subsidiary

from the parent corporation (H5c). Since the con-

struct of decentralization pertains to parent-subsidiary relationships only, we advance nohypotheses pertaining to the impact of decentrali-zation on knowledge inflows from peer subsidiaries.

Absorptive capacity of the target unit

We argued earlier that the absorptive capacity ofa subsidiary would be a function of (i) its fa-miliarity with the incoming knowledge, and (ii)interunit homophily. Based on this reasoning, weoperationalized the construct of absorptive

capacity in terms of the following two variables:mode of entry (a determinant of the subsidiarysex-ante familiarity with the corporate-wideknowledge base) and the proportion of localnationals vs. expatriates within the subsidiarystop management team (a measure of the inter-unit homophily of subsidiary managers).

Mode of entry. Literature on foreign directinvestment (see e.g., Hennart and Park, 1993) hasargued theoretically and demonstrated empiricallythat the less the overlap between existing corpo-rate know-how and the know-how required to

succeed in a host market, the greater the prob-ability of acquisition as the mode of entry. Thus,as we discussed earlier, relative to greenfieldoperations, acquired subsidiaries are more likelyto have a non-duplicative knowledge base vis-a-vis the parent corporation. Building on Cohenand Levinthals (1990) arguments regarding thedeterminants of absorptive capacity, it followsthat, on average, the novelty of acquired subsidia-ries knowledge base should also imply a lowerabsorptive capacity for intra-corporate knowledgerelative to the case with greenfield subsidiaries.Based on these arguments, we can now oper-ationalize Proposition 6 in terms of the followingconcrete hypotheses: ceteris paribus, relative togreenfield operations, acquired subsidiaries will

engage in less knowledge inflows from peer sub-

sidiaries (H6a) and from the parent corporation

(H6a).1

1An anonymous reviewer has pointed out that, at first glance,the two hypotheses under H6 might appear inconsistent with


9/24


Proportion of local nationals in the subsidi-

arys top management team. Several studieshave indicated that national background accountsfor significant differences in managerial perspec-tives (e.g., Tung, 1982; Zeira, 1986). Accord-ingly, 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 homophilybetween the subsidiary and the rest of the corpo-ration. Building on Rogers arguments (1995),we would expect that inter-unit homophily islikely to be positively associated with absorptivecapacity. This is so because greater homophilyimplies a greater commonality in language sys-tems as well as in the meanings assigned to theartifacts of communication. Thus, on average,subsidiaries with a greater proportion of local

nationals within the TMT can be expected tohave lower absorptive capacity for incomingknowledge from the rest of the corporate network.These arguments yield the following oper-ationalization of Proposition 6: ceteris paribus,the greater the proportion of local nationals

within the subsidiarys top management team,

the less will be the knowledge inflows 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 outflows;an implicit assumption underlying this prediction was thatsuch knowledge would be absorbed by the receiving units.However, H6 argues that unfamiliarity with incoming knowl-edge would reduce absorptive capacity among the receivingunits. Thus, H1 could not be true unless mode of entry hasa different effect on flows from the subsidiary to the MNCthan it does on flows from the MNC to the subsidiary. Webelieve this to be the case. The roots of this differing effect liein the following two observations: One, the typical acquisitionrepresents a voluntary event for the acquiring MNC but aninvoluntary event for the acquired subsidiary; thus, the willing-ness of the acquiring MNC to integrate the new knowledgeof the acquired unit should, on average, be greater than the

willingness of the acquired unit to integrate the new knowl-edge of the acquiring MNC. Two, the typical MNC wouldhave much greater experience at acquiring and integratingnew units than the typical unit would have in being acquiredand integrated; accordingly, on average, the acquirers abilityto digest new knowledge should be greater than that of theacquired unit.2We should note that nationality structure of a subsidiarysTMT has the potential to affect knowledge inflows not onlythrough its impact on absorptive capacity but also through itsimpact on richness of transmission channels between the localsubsidiary and the rest of the global network. It does appearlikely that, on average, relative to local nationals, expatriatesshould have stronger and longer-tenured social ties with man-


Control variables

Country of origin. There already exists a largebody of both theoretical and empirical literaturedealing with the fact that country of origin hasa major impact on the propensities of MNCs vis-

a-vis the choice of global strategies, organi-zational structures and control systems, as wellas internal corporate cultures (e.g., Bartlett andGhoshal, 1989; Egelhoff, 1984; Franko, 1976;Porter, 1994; Yip, Roos and Johansson, 1994).Accordingly, all of our hypotheses were testedafter controlling for the effect of country-of-originof the MNC.

Industry resource characteristics. As dis-cussed earlier, economic theory posits that MNCscome to be primarily because external marketsare less effective and efficient at knowledge trans-

fer than intracorporate mechanisms (Caves, 1982;Hymer, 1960; Kindleberger, 1969). Empiricaltests of this theory have consistently shown thatindustries characterized by greater degrees ofknowledge intensities (industries with higher R&D-to-sales-ratios and/or higher advertising-to-salesratios) tend to be more global than other indus-tries (e.g., Goedde, 1978; Grueber, Mehta, andVernon, 1967; Horst, 1972). Accordingly, wedeemed it important that, in testing our hypoth-eses, we control also for the potential effects ofthree resource characteristics of the MNCs indus-

try: R&D intensity, fixed asset intensity, andadvertising intensity (Collis and Ghemawat,1994).

Nature of subsidiarys operations. It is wellaccepted that foreign subsidiaries will often varyin the scope of value chain activities includedwithin their operations (Porter, 1986). The pres-

agers at corporate HQ and in other subsidiaries. In fact, asthe correlations in Table 1 indicate, there does exist a strongnegative correlation (0.59, p 0.001) between proportion oflocal nationals in the subsidiarys TMT and vertical corporate

socialization. Thus, as pointed out by an anonymous reviewerand the consulting editor, the question arises as to whether,in the context of our study, TMT nationality might be abetter proxy for another factor ( such as richness of trans-mission channels) rather than absorptive capacity. We believethat this would indeed be a serious concern if we did nothave any direct measures of socialization mechanisms asone of the hypothesized antecedents of knowledge inflows.However, as captured in H3b, H3b, H4b, and H4b, we dotest for the direct effect of socialization mechanisms onknowledge inflows. Thus, in a multivariate regression context,any remaining impact of TMT nationality on knowledgeinflows is likely to be due primarily to absorptive capacityrather than transmission channel considerations.


10/24


ence or absence of any particular activity withinthe subsidiarys operations can be expected toshape the nature of the subsidiarys interactionswith the rest of the corporation and, thus, thenature of knowledge inflows into and outflowsfrom the subsidiary. Accordingly, all of our

hypotheses were tested after controlling also forthe potential effects of two dummy variables:presence of a primary upstream activity (i.e., R&Dand/or manufacturing) and presence of a primarydownstream activity (i.e., marketing and sales).

METHOD

Sample

Data for this study were collected through acombination of questionnaire surveys and second-

ary sources. The following steps guided the devel-opment of the questionnaire instrument: (i) inter-views with subsidiary presidents and corporate-level executives in six MNCs to understand andclarify the phenomenon of interest, (ii) a reviewof previous research to locate, wherever possible,measures that would appropriately capture theconstructs under study, and (iii) a pretesting ofthe questionnaire for clarity and relevance throughface-to-face interviews with four subsidiary presi-dents (two American and two non-American).

The pre-tested questionnaires were mailed to

the heads (variously titled as presidents, manag-ing directors, or general managers) of 987 foreignsubsidiaries of major MNCs headquartered in theU.S. (407 subsidiaries of 19 MNCs), Japan (270subsidiaries of 41 MNCs), and Europe (310 sub-sidiaries of 15 MNCs). Subsidiary presidentswithin Japanese MNCs received both an Englishand a Japanese language questionnaire; initialinterviews with the European companies indicatedthat only the English-language questionnairewould suffice. The U.S. sample was drawn fromthe list of the largest U.S.-based MNCs containedin the International Directory of Corporate Affili-ations (National Register, 1991); this was alsothe approach used for developing a list of subsidi-aries for 9 out of the 15 European MNCs. In thecase of the other 6 European MNCs, the list ofsubsidiaries was drawn up in cooperation withthe senior-most corporate executive in charge ofstrategic planning, an approach also used in thecase of all of the Japanese MNCs. Given theconstraints of time and funding and given the


need to obtain access, it was not possible to usea random sample either from the entire universeof MNCs or from the entire subset of MNCsheadquartered in the U.S., Europe, and Japan.Nonetheless, given the diversity of industries inwhich the sampled firms operate (food products,

industrial machinery, computers, telecommuni-cations, pharmaceuticals, automobiles, chemicalproduction, electronics, consumer durables, con-sumer nondurables, etc.), there is no prima faciereason to expect any systematic bias in the find-ings from subsidiaries within these firms.

A personalized cover letter accompanying eachquestionnaire explained the purpose of the studyand provided assurances regarding the confiden-tiality of collected data. In order to minimizeresponse bias, the participants were also providedwith pre-addressed envelopes to enable them to

return the completed questionnaires directly tothe researchers without any risk of perusal byothers in their firms. A total of 374 questionnaires(38 percent) were returneda response rate thatcompares very favorably with past survey-basedresearch studies in the strategic management area.The number of respondents for U.S., Japan, andEuropean MNCs were 117 (28 percent), 112 (41percent), and 145 (46 percent), respectively. Totest for inter-rater reliability on the most criticalvariables in the study (knowledge outflows andinflows), we were also able to get responses on

these knowledge flow variables from the immedi-ate corporate-level direct-report superiors of 89of the responding subsidiary presidents.

For the sample, median worldwide revenuesand median number of total employees for theparent firms were $5.8 billion and 32,100 respec-tively; at the subsidiary level, the median numberof employees per subsidiary was 350. These fig-ures pertain to 1991, the year in which the surveydata were collected.

Measures

A summary of how the independent variables aswell as the control variables were measured iscontained in the Appendix. Wherever possible,we used standard well-established research instru-ments with minor changes in wording to adaptthe instrument to the multinational context. Givenbelow are details pertaining to how the fourvariables central to this study knowledge out-flows to peer subsidiaries (KO-S), knowledge


11/24


outflows to the parent corporation (KO-P),knowledge inflows from peer subsidiaries (KI-S),and knowledge inflows from the parent corpo-ration ( KI-P) were measured.

As stated earlier, in this study, we focus onthe transfer of largely procedural types of knowl-

edge (e.g., product designs, distribution know-how, etc.) but not on the transfer of largelydeclarative types of knowledge (e.g., monthlyfinancial results). Knowledge flow data were col-lected on the following seven items: (1) market-ing know-how, (2) distribution know-how, (3)packaging design/technology, (4) productdesigns, (5) process designs, (6) purchasingknow-how, and (7) management systems andpractices. For each of these seven items, thesubsidiary president was asked to indicate on a7-point scale (ranging from not at all to a

very great deal) the extent to which the subsidi-ary engaged in transfers of knowledge andskills in each of the following four directions:(1) provides knowledge and skills to sister sub-sidiaries, (2) provides knowledge and skills toparent corporation, (3) receives knowledge andskills from sister subsidiaries, and (4) receivesknowledge and skills from the parent corpo-ration. For each of these knowledge flow direc-tions, responses across the seven items were aver-aged to yield composite measures of KO-S, KO-P, KI-S, and KI-P. For these four variables, the

means, the standard deviations, and Chronbachalpha 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 scaleused to measure knowledge flows, the meanvalues of the four types of knowledge flows inour sample (2.36, 2.39, 2.21, and 3.75) may atfirst glance appear low. However, as clarifiedabove, it should be noted that this study hasfocused on the transfers of largely proceduralknowledge (i.e., know-how) rather than on thetransfers of largely declarative knowledge (e.g.,operational data). Given the tacit rather thancodified nature of much procedural knowledge,we would expect the mean levels of knowledgetransfers in this arena to be on the lower ratherthan higher side. It also should be noted that,as would be expected in the case of hierarchicalorganizations, pairwise t-tests revealed thatknowledge inflows from the parent to focal sub-sidiaries (KI-P) were significantly greater (at


p 0.001) than each of the other three typesof knowledge flows.

Given the perceptual nature of these knowl-edge flow measures and given their centralityfor our study, we deemed it critical that theybe tested also for inter-rater reliability. Towards

this goal, we were able to get responses onthe same knowledge flow variables from theimmediate corporate-level direct report superiorsof 89 of the subsidiary presidents. Each superiorcompleted a separate questionnaire containingthe subsidiarys name for each of the sampledsubsidiaries reporting to him/her. This question-naire used exactly the same seven dimensionsof knowledge. For each of these seven items,the superior was asked to indicate on a 7-pointscale (ranging from not at all to a very greatdeal) the extent to which he/she expected the

named subsidiary to engage in transfers ofknowledge and skills in each of the followingtwo directions: (1) provides knowledge andskills to the rest of the corporation, and (2)receives knowledge and skills from the rest ofthe corporation. For each of these two knowl-edge flow directions, responses across the sevenitems were averaged to yield composite meas-ures of expected knowledge outflows from thesubsidiary (Chronbach alpha = 0.82) andexpected knowledge inflows into the subsidi-ary (Chronbach alpha = 0.81) respectively. For

these 89 subsidiaries, this corporate-level meas-ure of expected knowledge outflows from thesubsidiary correlates at 0.23 (p 0.05) withthe average of KO-S and KO-P; similarly, thecorporate-level measure of expected knowledgeinflows into the subsidiary correlates at 0.38(p 0.001) with the average of KI-S andKI-P. Given these positive correlations in thedata from subsidiary presidents and theirimmediate superiors, in a context where theyare typically separated by a geographic distanceof thousands of miles, we believe that our meas-ures of KO-S, KO-P, KI-S, and KI-P can bedeemed as reliable.

Table 1 contains the matrix of zero-order corre-lations among these and all other variables uti-lized in this study. As this table indicates, theaverage correlation among the four knowledgeflow variables is 0.32 implying that the four typesof knowledge flows are distinct, albeit related,variables not only conceptually (Gupta and Gov-indarajan, 1991) but also empirically.


12/24


Table1Zero-ordercorrelation

coefficientsamongallvariablesun

derstudy

X1

X2

X3

X4

X5

X6

X7

X8

X9

X10

X11

X12

X13

X14

X15

X16

X17

X1

KO-S

X2

KO-P

***0.54

X3

KI-S

***

***

0.58

0.33

X4

KI-P

***

***

0.00

0.19

0.25

X5

R&D

Intensity

*

**

**

0.12

0.07

0.15

0.15

X6

FixedAssetIntensity

**

***

0.04

0.03

0.02

0.15

0.28

X7

AdvertisingIntensity

***

**

***

***

0.20

0.14

0.07

0.02

0.21

0.39

X8

Upstream

Activities1

**

***

0.14

0.08

0.05

0.08

0.16

0.06

0.08

X9

Downstream

Activities2

***

***

**

***

0.25

0.22

0.13

0.06

0.05

0.03

0.06

0.27

X10

ModeofEntry3

***

*

***

0.09

0.03

0.01

0.32

0.12

0.02

0.04

0.16

0.04

X11

SubsidiarySize

***

**

***

***

*

0.22

0.14

0.01

0.06

0.20

0.00

0.00

0.37

0.09

0.03

X12

RelativeEconomic

**

***

***

*

**

*

Level

0.08

0.15

0.05

0.36

0.22

0.10

0.13

0.04

0.03

0.35

0.06

X13

IncentiveFocus4

*

*

*

***

0.08

0.02

0.12

0.06

0.13

0.05

0.04

0.01

0.01

0.13

0.02

0.18

X14

FormalInteg

***

***

***

***

*

**

Mechanisms

0.29

0.24

0.22

0.21

0.06

0.03

0.01

0.01

0.01

0.12

0.09

0.02

0.15

X15

LateralSocializnMech.

***

**

***

**

*

*

**

***

***

*

**

***

0.25

0.15

0.21

0.14

0.12

0.08

0.11

0.13

0.20

0.08

0.22

0.11

0.15

0.21

X16

VerticalSocializn

***

**

***

***

***

***

***

*

**

*

***

Mech.

0.22

0.14

0.25

0.25

0.21

0.19

0.20

0.09

0.08

0.21

0.06

0.18

0.08

0.00

0.03

X17

HQ-Sub

*

***

*

***

*

**

*

Decentralization

0.06

0.06

0.04

0.13

0.17

0.06

0.13

0.25

0.04

0.11

0.16

0.03

0.13

0.05

0.08

0.06

X18

LocalNationalsin

***

**

***

**

**

***

***

***

***

**

*

***

**

TMT

0.19

0.14

0.22

0.14

0.14

0.18

0.23

0.22

0.22

0.08

0.02

0.12

0.10

0.01

0.06

0.59

0.12

11=

Subsidiaryhasanupstream

activity(R&D

and/ormanufacturing);0

=

Subsidiaryhasnoupstream

activity.

21=

Subsidiaryhasadownstream

activity(marketingandsales);0=

Subsidiaryhasnodownstream

activity.

3Modeofentry:1=

Acquisition;

0=

Greenfield.

4Highervaluessignifythattheincentivesystem

ismorenetwork,rather

thansubsidiary,focused.

*one-tailp

0.05;**one-tailp

0.01;***one-tailp

0.001



13/24


RESULTS

We have four dependent variables (KO-S, KO-P,KI-S, or KI-P) and a set of hypotheses pertainingto each of these variables. Each set of thesehypotheses was tested through a series of multi-

variate OLS regressions: first, we entered the fourcontrol variables pertaining to country-of-origin;second, we entered the three control variablespertaining to industry resource characteristics;third, we entered the two control variables per-taining to nature of subsidiary operations; finally,we entered the independent variables hypothe-sized as the determinants of that particular typeof knowledge flows. Tables 2 through 5 containthe results of these regression analyses.

Table 2. Determinants of knowledge outflows to peer subsidiariesdependent variable = Knowledge outflows to peer subsidiaries (KO-S)

Independent Variables Hypothesized Standardized Beta CoefficientsRelationship

Equation 1 Equation 2 Equation 3 Equation 4

Japan 0.233*** 0.244*** 0.184** 0.153**U.K. 0.041 0.003 0.003 0.061Sweden 0.046 0.031 0.043 0.080Finland 0.065 0.058 0.045 0.019

R&D Intensity 0.069 0.046 0.032Fixed Asset Intensity 0.086 0.078 0.001Advertising Intensity 0.128* 0.130* 0.155*Upstream Activities1 0.070 0.005Downstream Activities2 0.195*** 0.187***P1: Value of Knowledge StockMode of Entry3 H1a (+) 0.127**Subsidiary Size H1b (+) 0.169**P2: Motivational DispositionIncentive Focus4 H2a (+) 0.003P3: Transmission ChannelsFormal Integrative Mechanisms H3a (+) 0.256***Lateral Socialization Mechanisms H3b (+) 0.100*R2 0.072 0.095 0.139 0.256

d.f. 4,335 7,332 9,330 14,325F 6.50*** 4.99*** 5.91*** 8.00***R2 0.072 0.023 0.043 0.117d.f. 4,335 3,332 2,330 5,325F 6.50*** 2.84* 8.33*** 10.27***

11 = Subsidiary has an upstream activity (R&D and/or manufacturing); 0 = Subsidiary has no upstream activity.21 = Subsidiary has a downstream activity (marketing and sales); 0 = Subsidiary has no downstream activity.3Mode of entry: 1 = Acquisition; 0 = Greenfield.4Higher 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.


Knowledge outflows to peer subsidiaries

Table 2 presents the results of regression analysesto test our predictions regarding the impact ofvalue of knowledge stock (P1), motivationaldisposition (P2), and transmission channels (P3)

on knowledge outflows to peer subsidiaries.Value of knowledge stock. In the contextof knowledge outflows to peer subsidiaries, weoperationalized this construct in terms of mode ofentry and subsidiary size. The results in Table 2(equation 4) support both of the resulting hypoth-eses. More specifically, knowledge outflows topeer subsidiaries are higher in the case of (i)subsidiaries that were acquired rather than setup as greenfield operations (beta for mode of


14/24


Table 3. Determinants of knowledge outflows to the parent corporationDependent variable = Knowledge outflows to the parent corporation (KO-P)



Japan 0.127* 0.141* 0.095 0.088U.K. 0.089 0.062 0.063 0.038Sweden 0.018 0.035 0.030 0.021Finland 0.020 0.092 0.079 0.135R&D Intensity 0.043 0.031 0.002Fixed Asset Intensity 0.089 0.081 0.063Advertising Intensity 0.120 0.120 0.129*Upstream Activities1 0.002 0.033Downstream Activities2 0.181*** 0.174***P1: Value of Knowledge StockMode of Entry3 H1a (+) 0.063Subsidiary Size H1b (+) 0.121*Relative Economic Level H1c (+) 0.169**P2: Motivational Disposition

Incentive Focus4

H2a (+) 0.018P3: Transmission ChannelsFormal Integrative Mechanisms H3a (+) 0.208***Lateral Socialization Mechanisms H3b (+) 0.073R2 0.033 0.054 0.084 0.162d.f. 4,322 7,319 9,317 15,311F 2.75* 2.58** 3.23*** 4.02***R2 0.033 0.020 0.030 0.078d.f. 4,322 3,319 2,317 6,311F 2.75* 2.30 5.28** 4.84***

11 = Subsidiary has an upstream activity (R&D and/or manufacturing); 0 = Subsidiary has no upstream activity.21 = Subsidiary has a downstream activity (marketing and sales); 0 = Subsidiary has no downstream activity.3Mode of entry: 1 = Acquisition; 0 = Greenfield.4Higher 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 (betafor subsidiary size = 0.169, p 0.01; thus, H1bis supported).

Motivational disposition. In the context ofknowledge outflows to peer subsidiaries, we oper-ationalized this construct in terms of the networkvs. subsidiary focus of the incentive system forthe subsidiary president. The results in Table 2(Equation 4) do not support the resulting hypoth-esis (H2a).

Transmission channels. In the context ofknowledge outflows to peer subsidiaries, we oper-ationalized this construct in terms of formal inte-grative mechanisms and lateral socializationmechanisms. The results in Table 2 (Equation 4)support both of the resulting hypotheses. More


specifically, knowledge outflows to peer subsidi-aries are higher in the case of (i) subsidiariesthat are integrated more tightly with the rest ofthe corporation through formal mechanisms (betafor formal integrative mechanisms = 0.256,p 0.001; thus, H3a is supported), and (ii) sub-sidiaries whose presidents have been involved inlateral socialization mechanisms with peer sub-sidiaries to a greater extent (beta for lateralsocialization mechanisms = 0.100, p 0.05;thus, H3b is supported).

Knowledge outflows to the parent

corporation

Table 3 presents the results of regression analysesto test our predictions regarding the impact of


15/24


Table 4. Determinants of knowledge inflows from peer subsidiariesdependent variable = Knowledge inflows from peer subsidiaries (KI-S)



Japan 0.333*** 0.362*** 0.349*** 0.231**U.K. 0.087 0.166** 0.164** 0.169**Sweden 0.064 0.084 0.084 0.087Finland 0.088 0.216** 0.212** 0.195**R&D Intensity 0.202*** 0.200*** 0.164**Fixed Asset Intensity 0.098 0.095 0.062Advertising Intensity 0.029 0.030 0.032Upstream Activities1 0.017 0.026Downstream Activities2 0.063 0.055P4: Transmission ChannelsFormal Integrative Mechanisms H4a (+) 0.167***Lateral Socialization Mechanisms H4b (+) 0.110*P5: Motivational DispositionIncentive Focus3 H5a () 0.015

P6: Absorptive CapacityMode of Entry4 H6a () 0.071Local Nationals in TMT H6b () 0.101R2 0.085 0.115 0.119 0.167d.f. 4,341 7,338 9,336 14,331F 7.90*** 6.29*** 5.04*** 4.73***R2 0.085 0.030 0.004 0.048d.f. 4,341 3,338 2,336 5,331F 7.90*** 3.88** 0.69 3.80**

11 = Subsidiary has an upstream activity (R&D and/or manufacturing); 0 = Subsidiary has no upstream activity.21 = Subsidiary has a downstream activity (marketing and sales); 0 = Subsidiary has no downstream activity.3Higher values signify that the incentive system is more network, rather than subsidiary, focused.4Mode of entry: 1 = Acquisition; 0 = Greenfield.*p 0.05

**p 0.01***p 0.001 For t-tests, these are one-tail values.

value of knowledge stock (P1), motivationaldisposition (P2), and transmission channels (P3)on knowledge outflows to the parent corporation.

Value of knowledge stock. In the context ofknowledge outflows to the parent corporation, weoperationalized this construct in terms of modeof entry, subsidiary size, and relative economiclevel. The results in Table 3 (Equation 8) supporttwo of the resulting three hypotheses. More speci-fically, knowledge outflows to the parent corpo-ration are higher in the case of (i) subsidiariesthat are larger in size (beta for subsidiarysize = 0.121, p 0.05; thus, H1b is supported),and (ii) subsidiaries that are located in countrieswith a higher level of economic advancementrelative to the country of the parent corporation(beta for relative economic level = 0.169,p 0.01; thus, H1c is supported). There was no


support for our prediction regarding the impactof mode of entry on KO-P (H1a).

Motivational disposition. In the context ofknowledge outflows to the parent corporationalso, we operationalized this construct in termsof the network vs. subsidiary focus of the incen-tive system for the subsidiary president. Theresults in Table 3 (Equation 8) do not supportthe resulting hypothesis (H2a).

Transmission channels. In the context ofknowledge outflows to the parent corporation, weoperationalized this construct in terms of formalintegrative mechanisms and vertical socializationmechanisms. The results in Table 3 (Equation 8)support one of the two resulting hypotheses. Morespecifically, knowledge outflows to the parentcorporation are higher in the case of subsidiariesthat are integrated more tightly with the rest of


16/24


Table 5. Determinants of knowledge inflows from the parent corporationdependent variable = Knowledge inflows from the parent corporation (KI-P)



Japan 0.020 0.042 0.062 0.073U.K. 0.245*** 0.278*** 0.276*** 0.086Sweden 0.077 0.044 0.040 0.011Finland 0.319*** 0.501*** 0.506*** 0.256***R&D Intensity 0.052 0.063 0.014Fixed Asset Intensity 0.151** 0.153** 0.095Advertising Intensity 0.205** 0.206** 0.168**Upstream Activities1 0.046 0.009Downstream Activities2 0.039 0.033P4: Transmission ChannelsFormal Integrative Mechanisms H4a (+) 0.182***Vertical Socialization Mechanisms H4b (+) 0.119*P5: Motivational DispositionIncentive Focus3 H5a () 0.097*

Relative Economic Level H5b () 0.209***HQ-Subsidiary Decentralization H5c () 0.086*P6: Absorptive CapacityMode of Entry4 H6a () 0.165***Local Nationals in TMT H6b () 0.009R2 0.120 0.180 0.184 0.298d.f. 4,325 7,322 9,320 16,313F 11.11*** 10.10*** 8.01*** 8.31***R2 0.120 0.060 0.004 0.114d.f. 4,325 3,322 2,320 7,313F 11.11*** 7.83*** 0.74 7.28***

11 = Subsidiary has an upstream activity (R&D and/or manufacturing); 0 = Subsidiary has no upstream activity.21 = Subsidiary has a downstream activity (marketing and sales); 0 = Subsidiary has no downstream activity.3Higher values signify that the incentive system is more network, rather than subsidiary, focused.4Mode of entry: 1 = Acquisition; 0 = Greenfield.*p 0.05**p 0.01***p 0.001 For t-tests, these are one-tail values.

the corporation through formal mechanisms (betafor formal integrative mechanisms = 0.208,p 0.001; thus, H3a is supported). There wasno support for our prediction regarding the impactof vertical socialization mechanisms on KO-P(H3b).

Knowledge inflows from peer subsidiaries

Table 4 presents the results of regression analysesto test our predictions regarding the impact oftransmission channels (P4), motivational disposi-tion (P5), and absorptive capacity (P6) on knowl-edge inflows from peer subsidiaries.

Transmission channels. In the context ofknowledge inflows from peer subsidiaries, weoperationalized this construct in terms of formal


integrative mechanisms and lateral socializationmechanisms. The results in Table 4 (Equation 12)support both of the resulting hypotheses. Morespecifically, knowledge inflows from peer subsidi-aries are higher in the case of (i) subsidiariesthat are integrated more tightly with the rest ofthe corporation through formal mechanisms (betafor formal integrative mechanisms = 0.167,p 0.001; thus, H4a is supported), and (ii) sub-sidiaries whose presidents have been involved inlateral socialization mechanisms with peer sub-sidiaries to a greater extent (beta for lateralsocialization mechanisms = 0.110, p 0.05;thus, H4b is supported).

Motivational disposition. In the context ofknowledge inflows from peer subsidiaries, weoperationalized this construct in terms of the net-


17/24


work vs. subsidiary focus of the incentive systemfor the subsidiary president. The results in Table 4(Equation 12) do not support the resultinghypothesis (H5a).

Absorptive capacity. In the context of knowl-edge inflows from peer subsidiaries, we oper-

ationalized this construct in terms of mode ofentry and proportion of local nationals in thesubsidiarys top management team. The results inTable 4 (Equation 12) do not support the resultinghypotheses (H6a and H6b).

Knowledge inflows from the parent

corporation

Table 5 presents the results of regression analysesto test our predictions regarding the impact oftransmission channels (P4), motivational disposi-

tion (P5), and absorptive capacity (P6) on knowl-edge inflows from the parent corporation.

Transmission channels. In the context ofknowledge inflows from the parent corporation,we operationalized this construct in terms of for-mal integrative mechanisms and vertical sociali-zation mechanisms. The results in Table 5(Equation 16) support both of the resultinghypotheses. More specifically, knowledge inflowsfrom the parent corporation are higher in the caseof (i) subsidiaries that are integrated more tightlywith the rest of the corporation through formal

mechanisms (beta for formal integrativemechanisms = 0.182, p 0.001; thus, H4a issupported), and (ii) subsidiaries whose presidentshave been involved in vertical socializationmechanisms 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 ofknowledge inflows from the parent corporation,we operationalized this construct in terms of thenetwork vs. subsidiary focus of the incentivesystem for the subsidiary president, relative eco-nomic level, and HQ-subsidiary decentralization.The results in Table 5 (Equation 16) support allthree of the resulting hypotheses. More speci-fically, knowledge inflows from the parent corpo-ration are higher in the case of (i ) subsidiarieswhose presidents operate under more subsidiary-focused, rather than network-focused, incentives(beta for incentive focus = 0.097, p 0.05;thus, H5a is supported), (ii) subsidiaries that arelocated in countries with a lower level of eco-


nomic advancement relative to the country of theparent corporation (beta for relative economiclevel = 0.209, p 0.001; thus, H5b issupported), and (iii) subsidiaries that are givenless decision-making autonomy by corporateheadquarters (beta for HQ-subsidiary

decentralization=

0.086, p

0.05; thus, H5c

is supported).Absorptive capacity. In the context of knowl-

edge inflows from the parent corporation, weoperationalized this construct in terms of modeof entry and proportion of local nationals in thesubsidiarys top management team. The results inTable 5 (Equation 16) support only the first ofthe two resulting hypotheses. More specifically,knowledge inflows from the parent corporationare higher in the case of subsidiaries that wereset up as greenfield 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 hasinvestigated both theoretically and empirically thedeterminants of intra-MNC knowledge flow pat-terns. While previous studies have focused morenarrowly on selected facets of intra-MNC knowl-edge transfer e.g., tacitness of know-how (Teece,

1977; Zander and Kogut, 1995), and normativeintegration and inter-subsidiary communication(Ghoshal and Bartlett, 1988), this study hasadvanced and adopted a more comprehensivetheoretical approach. Building on communicationtheory, we have argued that a complete mappingof the knowledge transfer process requires atten-tion to all of the following five major elements:(i) value of the knowledge possessed by thesource unit, (ii) motivational disposition of thesource unit regarding the sharing of its knowl-edge, (iii) the existence, quality, and cost oftransmission channels, (iv) motivational disposi-tion of the target unit regarding acceptance ofincoming knowledge, and (v) the target unitsabsorptive capacity for the incoming knowledge.

Further, unlike previous studies on intra-MNCknowledge transfers, we have conducted separateexaminations of knowledge flows that occur lat-erally among peer subsidiaries and those whichoccur hierarchically between a subsidiary and theparent corporation. Given the ongoing devolution


18/24


of authority and responsibility from the center tothe subsidiaries and the ability of informationtechnology to enable direct communication amongsubsidiaries, we would agree with Bartlett andGhoshal (1989), Hedlund (1994), Martinez andJarillo (1989), and others that direct inter-

subsidiary interactions are becoming increas-ingly important.Utilizing the overarching theoretical framework

and the broad propositions depicted in Figure 1,we advanced a set of hypotheses for each ofthe following four types of knowledge transfercontexts: (i) knowledge outflows to peer subsidi-aries, (ii) knowledge outflows to the parent corpo-ration, (iii) knowledge inflows from peer subsidi-aries, and (iv) knowledge inflows from the parentcorporation. These hypotheses were tested withdata collected from the presidents of 374 subsidi-

aries of 75 MNCs headquartered in the U.S.,Europe, and Japan. All hypotheses were testedafter controlling for the effects of country-of-origin, the resource characteristics of the MNCsindustry, and the nature of the subsidiarys oper-ations.

Commentary on the results

As can be seen from Tables 25 (across-tablecomparions of R2 and R2 values as well as thenumber of significant beta coefficients), our data

had the greatest success in uncovering the deter-minants of KI-P i.e., knowledge inflows to focalsubsidiaries from the parent corporation. In thiscontext, it may be useful to recall our earlierobservation that, for the sample as a whole, ofthe four types of knowledge flows, the magnitudeof KI-P was significantly greater than that ofeach of the other three types of flows. Thesetwo empirical observations lead us to draw thefollowing conjectures: (i) Of the four types ofknowledge flows examined in this study, the typi-cal MNC has perhaps had the longest experiencein undertaking knowledge outflows from thecenter to the units; (ii) Notwithstanding the factthat MNCs are indeed becoming heterarchies(Hedlund, 1994) i.e., integrated complex networkswith significant devolution of authority andresponsibility to the subsidiaries, the parentcorporation continues to serve as the most activecreator and diffuser of knowledge within thecorporation; and (iii) MNCs greater experiencein managing knowledge outflows from the parent


to the subsidiaries has also made them moresystematic (as distinct from stochastic orexperimental) in managing these particulartypes of knowledge flows.

Focusing now on the empirical validity of ouroverarching theoretical framework, we also note

from Tables 25 that the results support ourexpectations regarding the importance of four ofthe five main constructs underlying this frame-work. More specifically, the results provide eithercomplete or partial support to our predictionsregarding the impact of value of knowledge stockand transmission channels on knowledge out-flows; similarly, they also provide either completeor partial support to our predictions regardingthe impact of transmission channels, motivationaldisposition to acquire knowledge, and absorptivecapacity on knowledge inflows. However, they

do not provide any support to our predictionsregarding the impact of motivational dispositionto share knowledge with other units on knowledgeoutflows. There are at least two possible expla-nations for this lack of support: (i) a subsidiarysmotivational disposition to share knowledge maydepend not only on the incentive system but alsoon other variables not examined in this study,and/or (ii) in the knowledge transfer process, themotivation of the target unit to acquire knowledgemay be far more important than the motivationof the source unit to share its knowledge. An

examination of the validity of any of these orother possible explanations must await futureresearch.

At a more micro-level, a closer examination ofthe 8 hypotheses (out of the total of 23hypotheses) that were not supported reveals that3 pertained to incentive focus, 2 to mode ofentry, 2 to proportion of local nationals insubsidiarys TMT, and 1 to vertical sociali-zation mechanisms. Alternatively stated, resultsfailed to support 3 out of the 4 hypotheses dealingwith incentive focus, 2 out of the 4 dealing withmode of entry, 2 out of the 2 dealing withproportion of local nationals in subsidiarys TMT,and 1 out of the 2 dealing with vertical sociali-zation mechanisms. There are at least three pos-sible explanations for this lack of support: (i)logical errors in developing the hypotheses,and/or (ii) substitution effects among the inde-pendent variables, and/or (iii) irreducible noisein the data. Our conjecture at this stage wouldbe that the last two explanations represent the


19/24


more likely scenario. Nonetheless, any definitiveexplanations for the lack of support also mustawait future research.

Limitations of the study

We can identify three major limitations of thisstudy. First, since every MNC is a network(Ghoshal and Bartlett, 1990), all intra-MNCknowledge transfers take place in the contextof the network. As contrasted with dyadic orsystemic approaches to the examination of net-work-related phenomena, we conducted ourexamination at the nodal level of analysis the simplest level feasible. In the next subsectionfocusing on directions for future research, weidentify some of the important questions that werenot explored by us but which can be examined

through future work that looks at knowledgetransfers from a dyadic or a systemic perspective.

Second, despite the fact that, in this study, wefocused on largely procedural types of knowledgewhich, on average, tends to be more tacit thandeclarative knowledge, we neither measured norexplored the impact of degrees of tacitness. Not-withstanding the pioneering studies of Teece(1977), Zander and Kogut (1995), and others,empirical research into how degrees of tacitnessaffect the knowledge transfer process is still inits infancy.

Finally, the third major limitation of this studyhas to do with the use of perceptual instrumentsto measure the extent of knowledge outflows andinflows. Barring the case of certain types ofcodifiable technology transfers (as in the caseof technology licenses), this is a methodologicalchallenge that researchers have yet to overcome.In our view, researchers face at least two hurdlesin measuring the extent of knowledge transfersthrough objective data: (i) Unlike transfers ofcodified knowledge, the transfers of tacit knowl-edge leave at best partial objective traces thatcould be measured by an external researcher; and(ii) Because transfers of tacit knowledge tendto be slow, any real-time investigation of thisphenomenon would often require the researcherto undertake a multi-year study of each transfer;by way of example, note that Zander and Kogut(1995) reported that, in their sample, the mediantime to transfer was five years and, without cor-recting for censored observations, the average waseight years. It is also instructive to note that,


notwithstanding their excellent access to theMNCs being studied, even Ghoshal and Bartlett(1988: 382) felt compelled to observe: Col-lecting objective level measures for the relativelylarge number of variables for meaningful sta-tistical 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 knowl-edge by organizations in general and, by MNCs inparticular, constitutes one of the most importantsubjects for research in the fields of organizationtheory (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 earlystages and empirical work is almost literally atthe stage of infancy. Thus, although we view thecontributions of this study as important, in lightof future possibilities, we view them as at bestmodest. There are several promising directionsfor future research.

First, we believe that the payoffs from futureinvestigations at the dyadic and/or systemic levels

are likely to be high. At the dyadic level ofanalysis, at least two of the important issues toinvestigate would be: (i) the impact of bilateralhomophily (Lazarsfeld and Merton, 1964) ondispositions to engage in outflows and inflows,and (ii) the importance of reciprocity i.e., isAs disposition to share its knowledge with Bdependent on Bs disposition to share its knowl-edge with A? At the systemic level of analysis,some of the important issues would be: (i) theimpact of a units network centrality on the extentof knowledge outflows as well as knowledgeinflows, (ii) the impact of network density onthe overall magnitude of knowledge flows throughthe network, and (iii) the impact of global com-petitive intensity faced by the MNC on the mag-nitude and the directionality of knowledge flows.

A second line of productive inquiry would beto compare and contrast what we would termas complementary vs. substitutive knowledgetransfer contexts. By complementary contexts, werefer to the transfer of knowledge along different


20/24


stages in the companys value chain e.g., thetransfer of technical knowledge from the develop-ment laboratories to the factories and the market-ing units and the transfer of market knowledgefrom the field back to the facories and the labora-tories; in these instances, knowledge transfers

occur in contexts where the source and the targetunits possess complementary knowledge stocks.In contrast, substitutive knowledge transfer con-texts can be said to exist when the source andthe target units engage in identical or similaractivities (e.g., two laboratories, or two factories,or two sales units) and the transfer involves theimposition of the source units superior knowhowover that of the targets allegedly inferiorknowhow. We would expect that the motivationaldispositions of both the source and the targetunits are likely to be radically different in the

case of complementary vs. substitutive knowl-edge transfers.

A third line of productive inquiry would be adeeper application and examination of the over-arching framework advanced in this paper. Thereare many other possible determinants of the valueof a source units knowledge stock e.g., theresource base of the unit, the internal organizationof the unit, and the competitive environment inthe host country. Similarly, there are many otherpossible determinants of motivational dispositionsto engage in inflows or outflows e.g., personal

characteristics of subsidiary managers such as ageor locus of control, their organizational commit-ment, and so forth. This is also true for the otherelements in our model viz., transmission channelsand absorptive capacity. In the case of trans-mission channels, the impact of communicationmechanisms including the use of electronic mediais an obvious topic for future research. Similarly,future investigations into how absorptive capacityof a receiving unit is affected not merely by itsexisting knowledge stock but also by its internalorganization are likely to yield valuable insights.For example, building on Cohen and Levinthal(1990), it should be useful to examine the impactof intra-subsidiary communication as well as asubsidiarys activism at knowledge creation onits capacity to absorb incoming knowledge.

A fourth line of productive inquiry would beto go deeper into the question of tacitness. Itseems to us that, while the conceptual literatureon how the tacitness of knowledge affects itstransfer is notable for its abundance, systematic


empirical investigations into how tacit knowledgegets tranferred and the extent to which its transferdoes or does not require ex ante codification isall too rare. Thus, our advocacy would be tourge greater efforts towards empirical rather thanconceptual studies on the topic of tacitness.

Finally, a productive line of inquiry would alsobe to examine the joint (i.e., interactive) effectsof capability, motivation, and transmission chan-nels on knowledge flow patterns. Given thatresearch on knowledge flows within MNCs isstill in its infancy, in this study, we focusedexclusively on the main effects of these con-structs. Nonetheless, since the results of this studylend support to the validity of our framework, alogical next step would be to develop and testmore complex theoretical models.

ACKNOWLEDGEMENTS

Paritial funding support for this study was pro-vided by the Center for International BusinessEducation and Research (The University of Mary-land at College Park), The Amos Tuck Schoolof Business Administration (Dartmouth College),and The International Management ResearchInstitute (Tokyo). The authors have benefitedfrom comments on earlier versions of this paperfrom Robert Burgelman, Ranjay Gulati, Lee Pres-

ton, M. Susan Taylor, as well as participants atthe 1997 Strategy Conference at Stanford Univer-sity.

REFERENCES

Allport, G. W. (1937). Personality: A PsychologicalInterpretation. Holt, New York.

Barney, J. (1991). Firm resources and sustained com-petitive advantage, Journal of Management, 17,pp. 99120.

Bartlett, C. A. and S. Ghoshal (1989). Managing across Borders: The Transnational Solution. Harvard Busi-ness School Press, Boston, MA.

Buckley, P. J. and M. Casson (1976). The Future ofthe Multinational Enterprise: Holmes & Meier, NewYork, pp. 6684.

Caves, R. E. (1971). International corporations: Theindustrial economics of foreign investment, Eco-nomica, 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


21/24


capacity: A new perspective on learning and inno-vation, Administrative Science Quarterly, 35,pp. 128152.

Collis, D. and P. Ghemawat (1994). Industry analysis:Understanding industry structure and dynamics. InL. 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 Institutes1995 International Conference on High PerformanceGlobal Corporations. Excerpted in Global View,Newsletter of the Carnegie Bosch Institute forApplied Studies in Management, The Carnegie-Mellon University.

Daft, R. L. and R. H. Lengel (1986). Organizationalinformation requirements, media richness, and struc-tural design, Management Science, 32, pp. 554571.

Dierickx, I. and K. Cool (1989). Asset stock accumu-lation 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 collec-tive rationality in organizational fields, AmericanSociological Review, 48, pp. 147160.

Edstrom, A. and J. R. Galbraith ( 1977). Transfer ofmanagers as a coordination and control strategy inmultinational organizations, Administrative ScienceQuarterly, 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 MultinationalEnterprise: 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 BigBusiness. Greylock, Stamford, CT.

Galbraith, J. R. (1973). Designing Complex Organiza-tions. Addison-Wesley, Reading, MA.

Gates, S. R. and W.

gupta;2000;knowledge flows

Documents