Paper to be presented at the DRUID Summer Conference 2003 on

CREATING, SHARING AND TRANSFERRING KNOWLEDGE. The role of Geography, Institutions and Organizations.
Copenhagen June 12-14, 2003

Theme B: Knowledge Transfers within and between firms,

John A. Cantwell Department of Management and Global Business Rutgers Business School 111 Washington Street Newark NJ 07102-3027, USA Phone: 973-353-5050 Email: Ram Mudambi (Corresponding Author) Department of General and Strategic Management Fox School of Business & Management Speakman Hall (006-00) Temple University Philadelphia PA 19122, USA Phone: 215-204-2099 Email: Date of submission: May 12, 2003 Abstract Considerable evidence has been gathered that points to home-base augmenting knowledge creation by MNE subsidiaries through tapping into local centers for the development of technologies that are complementary to those created by the parent company in the home base of the MNE. However, the precise pattern of such competence-creation is as yet unclear. We develop a theoretical argument that highlights the role of both oligopolistic deterrence and location theoretic (pull and push) factors. The theory suggests that competence-creating and competence-exploiting subsidiaries will differ systematically in terms of the sources of their knowledge inflows. We test the model using US patents granted to UK subsidiaries of non-UK MNEs. Keywords: MNC-subsidiary knowledge strategies, patent analysis, networks JEL codes: O32, F23


Introduction In recent years, considerable evidence has been gathered that points to knowledge creation by MNE subsidiaries through tapping into local centers for the development of technologies that are complementary to those created by the parent company in the home base of the MNE, and so 'home base augmenting' (Kuemmerle 1999, Zander 1999). However, while considerable evidence documenting the importance of competencecreating subsidiaries has been amassed (in contrast to the more traditional type of purely competence-exploiting subsidiaries), the precise locational pattern of such competencecreation is as yet unclear. We do know that competence-creating subsidiaries are more likely to be located in sites with good local infrastructure, and especially with a good science base (Cantwell and Iammarino 2000, Cantwell and Piscitello 2002). It has been found as well that competence-creating subsidiaries are likely to develop a higher proportion of technologies that are non-primary for the firm's industry than does the parent company at home (Cantwell and Kosmopoulou 2002) - primary technologies being defined as those that are most directly related to the products of the firm's main line of business (LOB), such as chemical technologies for firms in the chemical industry. This is one of the key respects in which competence-creating subsidiaries complement or augment innovation in the home base. Subsidiary development is especially likely to be geared to the creation of non-primary technologies in all-round or 'higher order' centers of excellence, as opposed to more specialized centers in which foreign-owned subsidiary and indigenous company innovative strategies are more closely aligned (Cantwell, Iammarino and Noonan 2001). Yet while most MNEs may follow these general criteria in their locational decisions, the precise geographical


pattern of competence development in each of the leading MNEs in the same industry is usually very different in terms of the actual sites in which they each have competencecreating subsidiaries. In this paper we attempt to shed light on this issue by focusing on firm and industry effects as well as the nature of the knowledge itself. We develop a theoretical argument that highlights the role of both oligopolistic deterrence and location theoretic (pull and push) factors. Following Cantwell and

Kosmopoulou (2002), leading MNEs are less likely to locate competence creation related to the primary technologies of their main LOB in subsidiaries in the home locations of their major competitors. Subsidiaries in such locations will typically not be competencecreating. If these subsidiaries have knowledge-creating mandates, they will tend to be either (a) in technologies that run complementary to the parent’s main LOB or (b) focused on knowledge processes involving low levels of creativity. We might call this phenomenon complementary location that implies some geographical separation by the leading MNEs in a common industry. Conversely, it seems likely that subsidiaries receive competence-creating mandates in the primary technological fields of the parent MNE’s main LOB only when there is a sufficiently wide dispersion of other independent firms technologically active in that location. We might call this related phenomenon colocation by a wider range (or a cluster) of firms within an industry in a center of specialized expertise for the industry in question. The former phenomenon may be driven partly by strategies aimed at maximizing the rents generated from the MNE’s own knowledge creation in its home base, but more importantly by the strong gravitational pull upon or absorption of key resources when there is a single major dominant player in a site (knowledge as a private good). The latter is driven by strategies aimed at


The patents granted to the subsidiaries in the study (or to their parent groups for inventions that are attributable to research undertaken in the subsidiaries) are obliged to cite all earlier patents that were relevant to their own discovery. and still less on how these then fit into the overall technological strategies of the corporate groups of which they are part. The location of the inventor of the cited patent enables us to observe the geographical origins of the knowledge sources used. the focus of cluster research has typically been on the public good aspect of knowledge and inter-firm or inter-institutional spillovers or externalities. Maskell 2001).. Moreover. The geographical characteristics of the knowledge sourcing of these subsidiaries is measured using patent citations. By analyzing overall firm strategy in an international context. Our theory implies a set of testable hypotheses about the nature of knowledge-creation in MNE subsidiaries. Background Cantwell and Mudambi (2001) distinguished between competence-exploiting and competence-creating types of subsidiary. This distinction was not itself measured by or derived from the characteristics of the research and development (R&D) conducted by 4 .enhancing the MNE’s knowledge pool through learning from external networks (knowledge as a public good). We test these hypotheses using a data set for foreign-owned subsidiaries operating in UK engineering industries.g. relatively little has been said on the role of MNE subsidiaries in such clusters. enhanced opportunities for division of labor and a deeper and richer labor market. While co-location has been extensively studied in the context of clusters (e. through for example. and by using both dimensions of knowledge we are able to develop a more complete understanding of knowledge creation in MNE subsidiaries.

the subsidiary. it is necessary to follow a two-stage procedure in our investigation. Such subsidiary mandates tend to be the outcome of subsidiary evolution. whereas acquired competence-creating subsidiaries tended to enjoy higher R&D as they benefited to a far greater extent from the asset-seeking element of acquisition. and 5 . Birkinshaw and Ensign 2002). We wish to establish first. An especially significant example of this divergence in the innovative strategies of subsidiaries appears in the impact of whether the subsidiary was brought into its current group through acquisition. That is. the determinants of a competence-creating subsidiary mandate. Acquired competence-exploiting subsidiaries tended to have reduced levels of R&D as they suffered disproportionately from the post-acquisition elimination of duplication in their new group. Frost. are regarded as having attained a competencecreating status. those that have acquired a competencecreating status through an appropriate evolution as opposed to those that have not. in that they provide some independently creative contribution to their corporate group. or which generally allied to this have further some responsibility for international strategy development. and subsidiary evolution in turn depends upon a combination of and an interaction between local initiative and parent company assignment (Birkinshaw and Hood 1998. and so the principal drivers of R&D are also very different (Cantwell and Mudambi 2001). subsidiaries that have acquired a mandate to undertake some area of product development. but was rather determined by the output mandate of the subsidiary. With the background of this study in mind. it becomes clear that the kinds of R&D in which each type of subsidiary tends to be involved is very different. Having identified these subsidiary types.

then second. Now as suggested above the capacity of a subsidiary to form constructive external linkages in its local environment of the kind associated with beneficial knowledge spillovers will depend upon the structure of the relevant local industry. then a subsidiary would be far more constrained in its local network linkages. To consider an extreme case. Andersson and Forsgren 2000. Forsgren and Holm 2002). so providing a greater diversity of opportunities for network linkages and spillovers to each of the individual companies located in that area. if there were just one dominant local player in its own industry (the degree of local industrial concentration were very high). Research Hypotheses The modern MNE must combine the capacity for the cross-border integration of activity and the need for its subsidiaries to be responsive to their own local contexts (Prahalad and Doz 1987. In its turn. Bartlett and Ghoshal 1989). the greater is the local embeddedness of the subsidiary. It is now well known that the ability of a subsidiary to be not merely locally responsive. the forms of knowledge sourcing of local innovation upon which each type of subsidiary is most likely to rely. Andersson. There are two possible ways of thinking about this impact 6 . but to undertake creative initiatives to some degree independently of its corporate group. depends upon its capacity to form favorable external network linkages with other companies and institutions in its own local environment (Birkinshaw. Hood and Jonsson 1998. a conventional cluster effect is associated by definition with an inter-firm variety of local actors (or in other words a sufficiently low degree of local industrial concentration). the higher the likelihood that it will acquire a competence-creating mandate. Put the other way round.

Hypothesis 1: As industry concentration rises. using March's (1991) terms. to the detriment of the potential for creativity of a subsidiary located in the same vicinity. the probability of the subsidiary achieving a competence-creating mandate falls. The rationale for this argument has to do with the extent to which the technological effort of the subsidiary is geared towards exploration rather than exploitation. so it is liable to draw relatively more intensively on its own immediate external network. in that dominant local players discourage or prohibit the transfer of potentially useful knowledge to subsidiaries located adjacent to them. As a subsidiary becomes more independently technologically creative or explorative.of local industrial concentration on the scope for subsidiary local embeddedness. It has been established that the greater is the innovation scale of a subsidiary. So whether for reasons of oligopolistic deterrence or due to the gravitational pull of a more dominant local company. This is partly due to the role of local network embeddedness in the 7 . and thus on knowledge sources in its local area. and the more that it exercises some technical leadership in a field of innovation. The other is in terms more simply of the strong gravitational pull on the best local resources exercised by a dominant local actor. the likelier it is that its innovations will draw on knowledge that originates in the host country in which the subsidiary is located (Frost 2001). One is in terms of oligopolistic deterrence. the diversity and the extent of opportunities available to a subsidiary to draw on local knowledge resources and become independently creative will tend to be reduced as local industrial concentration rises. A higher local industrial concentration will therefore likewise reduce the likelihood of a subsidiary acquiring a competence-creating mandate within its group.

and that is certainly true. However. owing to their greater embeddedness in their local environment. once the first stage of mandate acquisition is disentangled. Previous studies have examined this issue from the perspective of the amount of knowledge that the subsidiary itself is capable of creating. It has also been shown that if a subsidiary's innovation is based on adapting its parent company's prior achievements (which it does normally for the purposes of the more effective local exploitation of those achievements). the subsidiary is more likely to draw upon technical ideas originating in the home country of its parent company (Frost 2001). which encourages the formation of reciprocal knowledge linkages or exchanges.exercise of subsidiary creativity mentioned already. if a subsidiary patent cites (draws upon) any earlier patent of its parent company then it is more likely that the other patents cited by the same subsidiary patent 8 . Those subsidiaries that have gained competence-creating mandates can be expected to rely relatively more on local knowledge sources. the underlying proposition can be formulated more directly. Thus. This implicitly supposes that subsidiaries that have acquired competence-creating mandates will tend to have higher levels of local research activity. whether in some particular technological field or in general. and to their correspondingly greater absorptive capacity for knowledge spillovers available in that environment. Hypothesis 2: Competence-creating subsidiaries have a higher propensity to source knowledge locally than do competence-exploiting subsidiaries. as a subsidiary becomes more independently technologically creative. That is. so its opportunity and capability to absorb knowledge from its local environment (its local absorptive capacity) rises. but it is also because a more creative subsidiary can provide more to other local actors in its vicinity.

whether from the parent company itself or from others located in the home country. to propose that a competence-exploiting subsidiary. Hypothesis 3a: Competence-exploiting subsidiaries have a higher propensity to source knowledge from the home country than do competence-creating subsidiaries. a competence-exploiting subsidiary is also likelier to draw on knowledge sources from other parts of its corporate group located outside its own host country. Hypothesis 3b: Competence-exploiting subsidiaries have a higher propensity to source knowledge from the parent firm in its home country than do competence-creating subsidiaries. is also likelier to draw on home country knowledge sources. Alternatively. As argued above (in hypothesis 1).will originate from invention in the home country (excluding of course others that also belong to the parent firm). which were more liable to have been found in the parent company's own home environment. This is because competence-exploiting innovation that builds upon or adapts the established expertise of the parent company is likelier to also draw on the complementary knowledge sources that supported the original competence. being likelier to conduct research that is adaptive of parent company competence. subsidiaries that are situated in a more concentrated local industry are less likely to acquire a competence-creating mandate. This argument too can easily be strengthened and generalized. since the subsidiary exploits competence generated by other parts of its group and not only by the parent company. Hypothesis 3c: Competence-exploiting subsidiaries have a higher propensity to source knowledge from other parts of their corporate group than do competence-creating subsidiaries. 9 .

it is generally necessary to be able to recruit from among the best technical personnel available in that location. Any oligopolistic deterrents are especially likely to be applied to subsidiaries that provide a more obvious potential competitive threat in terms of their innovative capacity. the lower is the propensity to source knowledge locally. competence-creating subsidiaries suffer more acutely from the adverse effects of the gravitational pull on the best resources exercised by a dominant player as the main pole of attraction in a location. This constraint on local knowledge sourcing in a more concentrated local industry is particularly likely to affect subsidiaries that do manage to attain a competence-creating mandate in such an environment. So it follows that a higher degree of local industrial concentration tends to inhibit local knowledge sourcing in general. and perhaps also due to strategies of oligopolistic deterrence on the part of local leaders. 10 . Hypothesis 4a: The more highly concentrated is the host country industry. simply because there is a lesser diversity of sources available. Reasoning in the same way. but that it has the most detrimental effect on the scope for local knowledge sourcing of competencecreating subsidiaries.since they have a less extensive and a less diverse base of opportunities to construct external network knowledge linkages. subsidiary innovation is less likely to draw on local knowledge sources in a more highly concentrated local industry. For a subsidiary to establish and maintain the high quality linkages that are required with local actors so as to create a distinctively local new stream of innovation that relies less on the parent group. and so tend to have less scope for local embeddedness in their environment. Probably more importantly still.

acquired competence-exploiting subsidiaries are likely to become less reliant on local 11 . Hoskisson. those in an acquired competence-creating subsidiary may be able to establish a case for themselves to have still freer rein over the development of local creativity and in the participation in local external networks that this requires. managers in a typical competenceexploiting subsidiary of the acquired company tend to be in a relatively weaker position to state their case. while managers in an acquired competence-exploiting subsidiary may find themselves obliged to become increasingly dependent upon technological knowledge sourced from elsewhere in their new group. and tend to be likely to obtain more resources to support their local innovative efforts following the acquisition (Cantwell and Mudambi 2001). Thus. Ireland and Harrison 1991). the reverse applies in the case of the acquisition of a competence-creating subsidiary. However. In particular. It seems reasonable to expect that this bifurcation of subsidiary types with respect to the impact on local innovation of acquisition is likely to affect as well the local sourcing strategies that accompany such subsidiary innovation. which is instead usually particularly valuable from an asset-augmenting purpose to the acquiring group. as in the integration process that follows acquisition there is an effort to reduce the duplication of innovative efforts in the new combined group (Hitt.Hypothesis 4b: The negative effect of host country industry concentration on the propensity to source knowledge locally is greater for competence-creating subsidiaries than for competence-exploiting subsidiaries. Further. and so in this event subsidiary managers are actually in a stronger position. It has been observed that acquisition tends on average to reduce R&D-intensity in the acquired group.

Hypothesis 5: Competence-exploiting subsidiaries that are an acquired part of their MNE group are less likely to source knowledge locally. 1995). Data. which will obviously influence both the likelihood of a subsidiary acquiring a competence-creating mandate. location-specific data and subsidiary-level data. It includes traditional mechanical engineering 12 . 34-35 and 73-10 under the 1992 UK Standard Industrial Classification code (Office of National Statistics 1992). A variety of control variables are allowed for in testing these hypotheses. We take account as well of the extent to which the host country (the UK) is a risky location compared to the home country of the parent company of the MNE. whereas competence-creating subsidiaries that are acquired are more likely to do so. which is granted to the least dynamic and more problematic locations. we control for the quality and dynamism of the location in which the subsidiary is sited. Estimation and Results The current study uses three levels of data: industry-level data. The engineering and engineering-related industry group roughly corresponds to subsections 24-32. Industry-level data are used mainly for classification purposes and were drawn from Dun & Bradstreet indexes (Dun & Bradstreet 1994. and the propensity to source knowledge locally. while competence-creating subsidiaries are likely to become more so.knowledge sources in their innovation. and the degree to which the subisidiary is externally oriented (outside the host country) in its activities. Locational quality and dynamism is measured inversely by whether or not the local area has obtained development area status. It is also important to control for the degree to which a subsidiary has come to exercise a greater measure of strategic independence. Most notably.

Knowledge flows are measured using patent data. 1993.firms. The sample frame yielded a preliminary list of 601 subsidiaries with personal contact names. supported by telephone and field interviews. presented in Table 1. The data are obtained from the US Patent and Trademark Office database. Almeida 1996. and take their citations of earlier US patents as the indicator of the sources of prior such knowledge upon which they have drawn in their own innovative efforts. Subsidiaries for which separate data for the parent firm Descriptive statistics related to all these variables are 13 . Following the reasoning and approach of Cantwell (1995). Frost 2001). The subsidiary-level data were derived from a large 1995 postal survey of foreign-owned firms in the UK. along with the source of the data. electrical engineering firms (that include electronics) and chemical engineering firms (that include the biotechnology sector). Data comparing location risk characteristics of the host country (the UK) with those in the companies’ home countries were drawn from the financial markets publication Euromoney. 1995). Location-specific data relate to the classification of the local area in terms of Regional Selective Assistance (RSA) program and are based on the relevant Department of Trade and Industry (DTI) assisted areas map (August. 1993). The use of patent citations as measures of knowledge flows is well established in the literature (Jaffe et al. Almeida and Kogut 1997. The Appendix includes definitions all the variables used in the estimation. we use US patenting by these UK subsidiaries as the best common comparable measure of their own contributions to technological knowledge. The sample frame for this survey was constructed from Dun & Bradstreet indexes (Dun & Bradstreet 1994. supplemented by the London Business School company annual report library.

In both cases. Two sets of late respondents were defined corresponding to those who responded after receiving the first reminder and those who responded after receiving the second faxed reminder (the first set includes the second). In order to improve the response rate. Overall. The final usable sample frame consisted of 568 subsidiaries. The survey was mailed out in two waves of 224 and 344 in March and April 1995. The comparisons were carried out using a χ2 test of independence. and 12 were unusable for various other reasons. Of these.were unavailable were deleted. The first (pilot) wave focused on entries into the Midlands region (the most successful region in the UK in terms of attracting FDI). Each set of late respondents was compared to the early respondents on the basis of six sample measures. while the second wave targeted entries into the rest of the country. Two reminders were faxed to the companies that had not yet responded ten and twenty-one days after the survey was mailed out.96%). Non-response bias was investigated with the widely used method suggested by Armstrong and Overton (1977). the questionnaire had to be short. leaving 225 (39. concise and of current interest or salient to the respondent (Heberlein and Baumgartner 1978). the responses from early and late respondents were virtually identical. 244 responses were received to the mail survey (42. This involved comparing early and late respondents. 7 were found to be UK-owned firms mistakenly identified as non-UK-owned firms. The response rate is well within the range expected for an unsolicited mail survey. 14 .61%) valid responses for evaluation.

each factor should account for the variation in at least one variable if it is to be considered useful from a data summarization perspective (Churchill 1995). There are 3 factors with eigenvalues greater than unity. Using a χ2 test of independence. 20 respondents were randomly selected and interviewed by telephone to confirm their survey responses. responsibilities’ (STRAT). responses from field interviews were found to be virtually identical to those obtained from the postal survey on the basis of four sample measures. The extent to which supplier decisions are made by the subsidiary (SUPPLY). Since the variation in each variable is unity after it has been standardized. several of them are highly correlated with one another. the extent to which the subsidiary has responsibility for hiring management staff (HIRE) and for the international marketing function (MKT) and the percentage of subsidiary top management from the host country (UK) all load heavily on this factor. Finally. on the basis of the varimax rotated factor loading matrix. Second. Summarizing the subsidiary-specific data Two problems arise in using most of the firm-specific variables.7% of total variance. The latent root criterion is used to determine the number of factors (or summary variables) extracted.Survey responses were tested for veracity by comparing postal responses to responses obtained from field interviews. 15 . The factor analysis results are presented in Table 2. several of them are categorical and/or ordinal. STRAT. A total of 28 field interviews were carried out. The first factor. These problems are addressed by running the problem variables through principal component factor analysis. First. explains 30. ‘external The three factors are termed ‘strategic (EXTERNAL) and ‘process orientation’ responsibilities’ (PROCESS).

and the level of knowledge flows occur conditional on state of the mandate. EXTERNAL. Knowledge inflows into the UK subsidiary are estimated in the second stage. This would explain the loading pattern that emerged. We use the heteroskedasticity- 16 . the three factors account for almost 77% of the variance of all the underlying variables. it became clear that a considerable amount of training that occurred at these subsidiaries is of the operational or process type. explains another 14. We assume that the decision process is sequential. PROCESS. We adopt an instrumental least squares approach (ILS). but rather a strategic choice.2% of total variance. In interview with managers at several of the responding firms.4% of total variance. with the lowest value in excess of 70% and the highest value near 90%. its export experience as a percentage of total tenure (EXPT) and the geographic scope of its output mandate (GSCOPE) are the variables that load heavily on this factor. MAND. explains 26. Estimating knowledge inflows into the subsidiary The competence-creating mandate. The third factor. The communalities of the individual variables are very high as well. so that the competence-creating strategy is selected first. where we utilize the estimates of the endogenous mandating decision from in the first stage. The subsidiary’s process engineering responsibilities in operations (PROC) and training (TRAIN) are the variables that load on this factor. We treat it as such and estimate it in the first stage of the estimation.The second factor. The percentage of the subsidiary’s output that is exported (WEXPORT). These estimates are presented in Table 3. Hence it is an endogenous variable. The mandate is not a ‘given’ variable. Overall. is specified to be endogenous to the firm (including the subsidiary).

The data provide support for Hypothesis 1 in that the level of concentration in the UK subsidiary’s industry (CR5) has a significant negative impact on the probability that it achieves a competencecreating mandate (MAND). have significantly higher levels of knowledge inflows from local sources (LCIT). This provides evidence in support of Hypotheses 3a. We can use the results in Table 3 to examine Hypothesis 1. These results are presented in Table 5. Finally. A higher level of concentration in the UK subsidiary’s industry (CR5) has a significant negative impact on the levels of knowledge inflows from local sources (LCIT). also reduces that probability that the subsidiary achieves a competence-creating mandate. Since we have no endogeneity problems here. we find that subsidiaries with competence-creating mandates (MAND). after accounting for the endogeneity of the mandating decision. 3b and 3c. These results are presented in Table 4. We find that the effect of the level of 17 . we use heteroskedasticitycorrected OLS to obtain the estimates. we perform sub-group analysis by estimating the results for competenceexploiting and competence-creating subsidiaries separately. This provides evidence in support of Hypothesis 2.corrected variance-covariance matrix to generate standard errors since the Breusch-Pagan test indicated the size of the UK subsidiaries influenced conditional variances. This provides support for Hypothesis 4a. from other home country units of the parent MNE (CHCIT) as well as from other worldwide units of the parent corporate group (COCIT). Such subsidiaries also have significantly lower knowledge inflows from the home country of the parent firm (HCIT). Of the control variables. with a poor local resource pool. Proceeding to Table 4. we find that location in a Development Area (R1).

Discussion and conclusions As is common this study builds upon. but not for competence-exploiting subsidiaries. 18 . However. the impact on local knowledge inflows is significant and negative for competence-creating subsidiaries. Competence- creating UK subsidiaries of US MNE parents source significantly more knowledge locally. consolidates and confirms the results of a number of other recent studies in this field. this finding has important managerial and policy implications. Clearly. No such effect is found for competence-exploiting subsidiaries.concentration in the UK subsidiary’s industry (CR5) on the levels of knowledge inflows from local sources (LCIT) is insignificantly different from zero for competenceexploiting subsidiaries. and if it does get one. Duration of UK operations (DT) has a significant positive effect on local knowledge inflows for competencecreating subsidiaries. and add something distinctively new to the literature. It is encouraging that the results are entirely consistent with those of previous investigations. It has been shown how higher local industrial concentration both reduces the likelihood of a subsidiary attaining a competence-creating mandate. Two aspects of our results are especially striking. it is interesting to note that acquisition entry (ACQUISITION) has significant but conflicting effects on local knowledge inflows for competence-exploiting and competence-creating subsidiaries. As had been anticipated. This provides support for Hypothesis 5. This provides support for Hypothesis 4b. still reduces the scope that a subsidiary has for local knowledge sourcing in its innovation strategy. The first of these is the role in local subsidiary knowledge creation of indigenous industrial concentration.

19 . upon the quality and dynamism of the location. upon the extent of its own external orientation.Second. Apart from the negative impact of host country industrial concentration. and upon its own duration. While these results are in line with our expectations. as opposed to competence-exploiting subsidiaries. local knowledge sourcing in competence-creating subsidiaries depends positively upon acquisition. it is striking just how divergent the different kinds of subsidiary are in the conditions of their local innovation. but positively affected by the extent of home country risk. there is a polarized divergence in the determinants of local knowledge sourcing in competence-creating. In contrast. local knowledge sourcing in competenceexploiting subsidiaries is negatively influenced by acquisition.

J. pp. pp. and Overton.A. 19. U. 2000.979-996. S. 40(4). and Iammarino.M. M. 1998. Sticky places in slippery space . pp. Boston. Estimating non-response bias in mail surveys. Small Business Economics. 305-334. Armstrong. J. Cantwell. pp. C. Pain. 19(1). Andersson. and Kosmopoulou. pp. The strategic impact of external networks: subsidiary performance and competence development in the multinational corporation. 2002. and Forsgren. Academy of Management Review. J. pp.. P and Kogut. Multinational subsidiary evolution: capability and charter change in foreign-owned subsidiary companies. 2001. 155165. V. S. Strategic Management Journal. and Ghoshal. London: Palgrave. Oxford: Pergamon.). Bartlett. and Holm. E. and Noonan. N.A. S. 221-241. N. 396-402..). Forsgren. Almeida. C.. B. in Inward Investment. Cantwell. S. 1996. Birkinshaw.. Regional Studies. 1995. N. and Hood. 2000.A. Birkinshaw. pp. P. In search of centre of excellence: network embeddedness and subsidiary roles in multinational corporations. M. The globalisation of technology: what remains of the product cycle model? Cambridge Journal of Economics. U. 23. Cantwell. 317-332. and Jonsson. pp.S. J. Iammarino.A.155-174.M. Strategic Management Journal. 773-795. Knowledge sourcing by foreign multinationals: patent citation analysis in the US semiconductor industry.References Almeida. Multinational corporations and the location of technological innovation in the UK regions. MA: Harvard Business School Press. 34(4). 329-350. Strategic Management Journal. What determines the internationalisation of corporate technology? in Critical Perspectives on Internationalisation. 1977. H. 23(4). Management International Review. Andersson. Havila. 2002. 21–31. T. pp. 1989.the location of innovation by MNCs in the European regions. M. 17. Journal of Marketing Research 14(3). Cantwell. J. Building firm-specific advantages in multinational corporations: the role of subsidiary initiative. 1998. J. Technological Change and Growth. pp. U. Exploration of technological diversity and the geographic localization of innovation.A. 9. and Håkanson. (ed. 20 . Forsgren. 1997. Hood. Managing Across Borders: The Transnational Solution. (eds. J.

Dun & Bradstreet. Business Register. T. 71-87. pp. pp. and Piscitello. Centers of excellence in multinational corporations. H.A. T.A. Churchill. 21 . and Ensign. R. Fort Worth: Dryden Press.Cantwell. R. 2001. Factors affecting response rates to mailed questionnaires: a quantitative analysis of the published literature. Dun & Bradstreet. P. 101-123. Frost.. 1995.E. and Kim. pp. and Harrison. Exploration and exploitation in organizational learning. Birkinshaw. 40(4). Heberlein.. 1978.A. Kuemmerle. M. Strategic Management Journal. J. J.D.G. Hoskisson. 1991. Jaffe. 1993.M. pp. Ireland. 23. 69-96.A. Towards a knowledge-based theory of the geographical cluster.A. Quarterly Journal of Economics. 30(1). 10(4). 767-798. 1995. J.. 1994.S. and Henderson. 921-943. Business Register. R.A. Dun & Bradstreet. Industrial and Corporate Change.. 1991. Journal of International Business Studies. A. pp. pp. Journal of International Management. Academy of Management Journal. Maskell. Strategic Management Journal. G. Effects of acquisitions on R&D inputs and outputs. R. 34(3). March. 1997. T. 2001. Cantwell. Trajtenberg.S. Hitt. L. 2002. 2002. M.. pp.E. J. The drivers of foreign direct investment into research and development: an empirical investigation. MNE competence-creating subsidiary mandates: an empirical investigation. 22. and Mudambi. Jr. 577-598. J. no. pp. 447-462. International diversification: effects on innovation and firm performance in product-diversified firms. 285. 58. 1995. 2001. University of Reading Discussion Papers in International Investment and Management. The location of technological activities of MNCs in European regions: the role of spillovers and local competencies. R. pp. as evidenced by patent citations. 997-1018. pp. 1-24. Marketing Research: Methodological Foundations. 8(1). Academy of Management Journal. Hitt. M. Organization Science. July.. The geographic sources of foreign subsidiaries' innovations. 6th ed.C. Hoskisson. and Baumgartner. 1994. American Sociological Review 43(4). R.. 2(1). P. Dun & Bradstreet. 1999. London.S. Frost. Geographical localization of knowledge spillovers. W. London. 693-706.

K. 1987. The UK Standard Industrial Classification of Economic Activities. Prahalad. How do you mean 'global'? An empirical investigation of innovation networks in the multinational corporation. Research Policy. I. The Multinational Mission: Balancing Local Demands and Global Vision.Office of National Statistics. 28(2-3). London: HMSO. 195-213. pp. Zander. New York: Free Press. 22 . and Doz. C. Y. 1999. 1992.

K. if UK subsidiary is in a Development area*** 0. if the UK subsidiary is in a Split Development/Intermediate area*** 0. if UK subsidiary is in a mechanical engineering and related industry 0. average. otherwise CR5 5 firm concentration ratio in UK subsidiary’s 4-digit industry Location variables RLOCRSK R1 R2 Relative country risk.Appendix Variable definitions VARIABLE Dependent variables MAND DEFINITION SOURCE Survey.). otherwise 1. if UK subsidiary is in an electrical engineering and related industry 0. supplemented by company annual reports US PTO database US PTO database US PTO database US PTO database 1. if UK subsidiary is in a chemical engineering and related industry 0. home country/host country (U. the UK subsidiary has achieved a competence-creating mandate* 0. otherwise LCIT UK subsidiary’s US patent citations of patents by other firms in the UK as percentage of total citations HCIT UK subsidiary’s US patent citations of patents assigned to other units in its home country CHCIT UK subsidiary’s US patent citations of patents assigned to other units of the parent MNC in its home country COCIT UK subsidiary’s US patent citations of patents assigned to other units of the parent MNC Industry variables ELEC 1. otherwise MECH 1. 1993-1994 1. otherwise 23 Business Register Business Register Business Register Census of Industry Euromoney** DTI DTI . otherwise CHEM 1.

otherwise Extent to which decisions on suppliers are Survey made in the UK (7 pt. supplemented by management (directors and above) from host company annual reports country (UK) Extent of responsibilities in the international Survey marketing function (7 pt. if parent firm HQ is in the US Survey. otherwise 1. otherwise 1. supplemented by company annual reports Variance of UK subsidiary’s rate of return on Survey. 1994 * MAND is generated on the basis of the functional scope of the UK subsidiary’s output mandate. supplemented by company annual reports 0. Likert scale) Percentage of UK subsidiary top Survey. supplemented by capital. Likert scale) Extent to which UK subsidiary has Survey responsibility for training in process engineering (7 pt. 1986-1994 company annual reports UK subsidiary turnover. supplemented by acquisition DTI data 0. Likert scale) UK subsidiary’s R&D/sales ratio. (2) UK and company annual reports mainland Europe. supplemented by duration of UK operations company annual reports Geographical scope of UK subsidiary’s Survey. supplemented by output mandate – (1) UK only. Likert scale) Extent to which UK subsidiary has Survey responsibility for hiring management staff (7 pt. otherwise Duration of UK subsidiary operations (years) Survey.Subsidiary variables RD FINRSK SALES DIVERS ACQUIRE DT USDUM JAPDUM SUPPLY HIRE TOPMGMT MKT WEXPORT EXPT GSCOPE PROC TRAIN Survey. 1994 (£million) Survey. supplemented by company annual reports 1. Likert scale) Exports as a percentage of UK subsidiary Survey. if operations in the UK are in parent’s Survey. (2) 24 . if parent firm HQ is in Japan Survey. if UK operations are the result of an Survey. supplemented by output company annual reports Years of exporting as a percentage of total Survey. supplemented by company annual reports 0. (3) Worldwide UK subsidiary’s process engineering Survey operational responsibilities (7 pt. Output mandates were categorized as: (1) Sales and service. supplemented by company annual reports 0. supplemented by main line of business @ company annual reports and DTI data 1.

(3) Manufacturing. (5) International strategy development. debt default. short-term and capital market finance. political risk. debt indicators.Assembly. A competence-creating mandate is operationalized as a subsidiary whose output mandate is either (4) or (5). (4) Product development. August 1993). 25 . credit ratings. This definition is based on Hitt et al (1997). access to bank. ** Euromoney risk index. which includes economic performance. and the discount on forfeiting *** Based on the Department of Trade and Industry (DTI) Assisted Areas map (revised. @ The parent firm’s main line of business is defined to be its largest non-UK sales segments whose cumulative contribution to the entropy index of diversification just exceeds 60%.

0.4089 0.4267 0.1113 4.6445 0.5050 0.7160 4.4957 0.1644 37.4042 0.5982 19.5845 1.Table 1 Summary statistics VARIABLE MAND LCIT HCIT CHCIT COCIT ELEC MECH CHEM CR5 RLOCRSK R1 R2 STRAT EXTERNAL PROCESS RD FINRSK SALES DIVERS ACQUIRE DT USDUM JAPDUM MEAN Dependent variables 0.6927 374.9827 8.2044 0.8889 0.4808 0.4307 15.2444 38.07464 18.1822 3.1599 327.4927 0.3073 Location variables 1.0830 0.0711 S.54223 25.0016 2.0403 0.D.2089 0.7378 24.1200 Subsidiary variables 0.4307 19.9484 0.0280 -0.6311 9.9981 1.2576 26 .6007 Industry variables 0.4709 0.4836 5.3257 1.0064 0.4074 0.7262 0.4089 0.7963 5.

* The eigenvalue for the 4th factor is 0.891 0.021 0.712 0. 27 .7206.262 Factor 3 PROCESS 0.204 0.026 –0.812 0.814 0.8875 0.330 –0.1784 2.164 0.802 0.102 2.6847 3.020 0.241 –0.735 6.804 0.004 3.902 –0.898 0.3802 0.3084 1.860 0.754 0.117 0.002 0.794 1.791 0.307 Factor Loadings Factor 2 EXTERNAL 0.0008 0.809 0.760 0.144 Communality 0.203 0.084 0.884 0.792 0.027 –0.Table 2 Factor analysis of subsidiary-specific qualitative variables: Varimax Rotation Variable SUPPLY HIRE TMT MKT WEXPORT EXPT GSCOPE PROC TRAIN Eigenvalue* Variance % Variance Factor 1 STRAT 0.018 0.4226 0.057 0.124 0.891 0.061 0.768 Loadings of variables associated with particular factors are shown in bold.

02) 0.906 (2.568) 0.88)* 0.0062 (2.17) -0.666 (1.31) DIAGNOSTICS Log-likelihood Restricted Log-Likelihood Likelihood Ratio Test: χ2(11) ‘p’ value Iterations -104.32) -0.0936 (0.1335 41.54) 0.36826 (0. ** Estimate significant at the 1% level.218 (2.579 (2.0540 (0.00001) 36 REGRESSOR CONSTANT R1 R2 STRAT CR5 RD ACQ DIVERS USDUM JAPDUM MECH ELEC Notes: * Estimate significant at the 5% level.28) 0.265 (1.0751 (0. MAND=0 (Subsidiary has no competence-creating mandate) PARAMETER ESTIMATE (‘T’ STAT) -0. 28 .185 (0.22)* -0.361 (1.0572 (1.4494 -125.55) -0.583)* -0.97)* -0.Table 3 Estimating the probability of a subsidiary competence-creating mandate: Maximum Likelihood Probit Estimates REGRESSAND: Binary variable: MAND=1 (Subsidiary has competence-creating mandate).

039 (0.41) -3.2711 0.8625 -942.305 (0.48 (2.046 (0.44) -0.311 (0.009 (1.50) 1.211 (0.13) -1.35) 0.27) 4.282 (2.596 (0.09 -928.30) -3.25) -0.66) -6.72) -1.51)* 1.94)° 0.58) -0.014 (1.41) -1.61) 8.816 (0.581 (0.123 (2.63) 5.47) -0.130 (0.036 (1.96) ° 0.2716 4. 29 .21) -1.51) -1.082 (1.01) -0.503 (0.79 (2.537 (2. ° Estimate significant at the 10% level.62) -0.22) 0.95) -0.036 (2.439 (0.22) 0.989 (1.27)* -0.570 (0.24) -4.41)* -0.62 (2.166 (2.41) -5 0.126 (1.617×10-6 (0.296 (1.56 -762.96) ° -0.356 (1.290 (0.041 (0.170 (0.783 (0.381 (2.476 (0.909 (0.09 (2.82)** DIAGNOSTICS 0.33)* 0.Table 4 Estimating the Extent of Subsidiary Knowledge Inflows Instrumental Least Squares Estimates REGRESSOR Constant ELEC Industry MECH variables CR5 RLOCRSK Location R1 variables R2 STRAT EXTERNAL PROCESS FINRSK SALES Subsidiary variables DIVERS ACQUIRE DT USDUM JAPDUM MAND Adj R F Stat (17.471 (0.005 (0.076 (2.04)* -11.159 (0.030 (0.254 (1.41) 8.21 -959.20) 0.353 (0.108×10-6 (0.146 (2. ** Estimate significant at the 1% level.190 (2.63)** 1.57) -0.3457 5.7628 -787.30) -2.11) -0.366 (0.23 5.38) 3.91) 0.24) 0.306 (1.80) 2.13)* 4.91) ° -0.13) -0.616 (0.773 (1.2862 Notes: White’s heteroskedasticity-corrected variance-covariance matrix used to generate ‘t’ statistics.25)* 0.2637 5.792×10 0.93) 39.330 (1.94) -22.574 (1.03) -0.2915 31.10) 0.17)* 0.330 (0.3593 -981.821 (0.019 (0.38) -0.38)* -2.93) ° -2.6837 -992.787 (0.58) 5.45) 2.0024 (0.801 (0.5988 -989.827 (0.50) -11.17) 0.132 (2. * Estimate significant at the 5% level.023 (0.125 (1.600 (1.96) ° -1.294×10-5 (0.41)* -38.79) ° -2.414 (0. 207) Log-Likelihood Restricted Log-Likelihood LCIT REGRESSAND HCIT CHCIT Parameter (‘t’ stat) 23.60)* 5.109 (0.568 (0.72) -1.197 (0.015 (1.75) ° 0.29) 0.8187 COCIT 31.52)* 1.52) 0.05) (1.55) -0.427 (0.73) 0.65) ° 9.

43) 7.36 (18.242 (1.9282 -247. 30 .21) -2.816 (2.46) -9.11) 12.502 (1.00) -5 0.972 (1.640 (1.1256 (0.62) 0.) Log-Likelihood Restricted Log-Likelihood REGRESSAND LCIT (MAND=0) LCIT (MAND=1) Parameter (‘t’ stat) 42.907 (1. * Estimate significant at the 5% level.309 (6.72) -4.54)* 19.64)** 2.27) DIAGNOSTICS 0.65) -8.26) 0.612 (0.61)* -0.29) -0.9945 (1.437 (0.076 (2.0429 (2.0258 (0.972 (0.41) 0.257 (0.71)** 2.59) -.76)** 0. ° Estimate significant at the 10% level. 151) 7.00354 (0.3946×10 -0.303 (1.469 (2.693 (2.Table 5 Estimating the Extent of Local Knowledge Inflows: Competence-exploiting vs.146 (0.076 (0.1345×10-4 (1.279 (0.08) 6.3461 5.0069 -740.07) 10.31) 6. Competence-creating Subsidiaries Ordinary Least Squares Estimates REGRESSOR Constant ELEC Industry MECH variables CR5 RLOCRSK Location R1 variables R2 STRAT EXTERNAL PROCESS FINRSK SALES Subsidiary variables DIVERS ACQUIRE DT USDUM JAPDUM Adj R F Stat (d.598 (0.4562 Notes: White’s heteroskedasticity-corrected variance-covariance matrix used to generate ‘t’ statistics.822 (0.56) 6.1198 (0.878 (2.01)* -2.56)* -1.15 (18.864 (2.89) 19.f.2372 0.79) 3.594 (2.85) -5.05) -0.08) 3.07) 0.557 (1.1473 (0.59)* 2.21) 0.0429 (1.1333 -205.70)** 0. ** Estimate significant at the 1% level.36) -728.

Sign up to vote on this title
UsefulNot useful