Knowledge Transfer and Accommodation Effects in Multinational Corporations

The uniqueness of external knowledge

Research indicates that unique knowledge may develop in the periphery of the firm’s knowledge structure (Andersson, Forsgren, & Holm, 2002; Lyles & Schwenk, 1992; Zahra, Ireland, & Hitt, 2000). Related research shows that external knowledge in foreign subsidiaries may be valuable to MNCs (Doz, Santos, & Williamson, 2001; Frost, Birkinshaw, & Ensign, 2002; Ghoshal, 1987). For example, the MNC may take advantage of interfirm spillovers of technological knowledge (Jaffe, Trajtenberg, & Henderson, 1993) that are captured by subsidiaries, or extract knowledge from sophisticated customers and skilled competitors by having specialized subsidiaries in strategic locations (Asmussen, Pedersen, & Dhanaraj, 2009; Dunning, 1996; Porter, 1998). The Acer Aspire is a case in point—being designed by a local Silicon Valley firm called Frog Design, championed by an American product manager, and shaped by extensive inputs from local retailers and consumers (Bartlett & St. George, 1998). External knowledge of this kind is likely to be different from the knowledge of other MNC units because geographical proximity is conducive to knowledge spillovers (Jaffe et al., 1993) and most of the potential MNC recipients reside in other locations where they do not have proximate access to the knowledge environment of the focal subsidiary. ⁴

In contrast, internal knowledge is likely to somewhat overlap with preexisting knowledge in a potential receiving unit because both sender and recipient have previously received their knowledge from some of the same sources (such as MNC headquarters or other subsidiaries). Thus, internal knowledge, when transferred, may be relatively redundant in the context of the knowledge already possessed by the receiving unit. We thus argue that the uniqueness of the subsidiary’s knowledge in an MNC context tends to be larger, ceteris paribus, if that knowledge is acquired in the subsidiary’s external host country environment rather than from other MNC units. This enhances the benefit of knowledge transfer as it may stimulate innovation by increasing diversity of knowledge in the receiving unit (Page, 2007).

Knowledge fungibility

Although the preceding analysis suggests that external knowledge increases marginal knowledge transfer benefits (formally, B_KE > 0), other forces pull in the opposite direction. In particular, the value of transferring the knowledge of the subsidiary depends not only on the uniqueness of that knowledge but also on the degree to which it is specialized to its originating location. For example, production blueprints from a focal subsidiary may indeed be valuable to overseas MNC units, which can use them to manufacture products for their local customers—but only to the extent that the customers in those countries actually value the product described by the blueprint. This idea was first proposed by Teece (1982), who defined the fungibility of a resource as the extent to which the resource retains its value when applied in increasingly distant markets. Hence, highly fungible resources are those generalized, firm-specific capabilities that could potentially be used in a wide variety of markets (Teece, 1982; Wernerfelt & Montgomery, 1988). In an MNC context, the fungibility of the subsidiary knowledge stock should be an important determinant of the benefits of knowledge transfer.

Rugman and Verbeke (1992) stress the interaction of geographic distance and fungibility by distinguishing between location-bound (nonfungible) and non-location-bound (highly fungible) firm-specific advantages. Several factors may contribute to making a knowledge resource location bound. Often receiving units are poorly positioned to exploit knowledge simply because they are distant from the customers, suppliers, partners, or competitors for whom the knowledge was developed. This suggests that external knowledge is likely to be more difficult to apply in other MNC units, precisely because it has a more peripheral origin. It is less fungible than internal knowledge because it relates specifically to host country conditions and is, therefore, more context specific. External knowledge may therefore in fact be associated with lower benefits (B_KE < 0) and increased costs (C_KE > 0) of knowledge transfer.

Knowledge congruity

Furthermore, even if a focal subsidiary’s knowledge is highly fungible, it is not certain that potential recipient units possess the necessary complementary knowledge to absorb and utilize it (Cohen & Levinthal, 1990). This suggests that there is another aspect, in addition to the economic applicability of the knowledge to recipients’ markets, that matters to knowledge transfer benefits—a cognitive aspect that we refer to as the congruity of the subsidiary’s knowledge stock. Schulz (2003) noted that the relevance of a given knowledge stock to its recipient was a function of the extent to which the knowledge is related to and connected with the preexisting knowledge in the receiving unit. Consistent with that idea, we define knowledge congruity as the cognitive fit between the knowledge stocks of the focal subsidiary and those of potential recipients of that knowledge (e.g., HQ or a subsidiary in another location). One important antecedent to knowledge congruity is the combination of sources from where the knowledge of the subsidiary originates. If the knowledge originates within the MNC, it is likely to be reasonably congruous with that of a receiving MNC unit because both units have previously received their knowledge from some of the same sources and thereby have a foundation of shared knowledge on which to build. Conversely, external knowledge may be so far removed from the realities of the receiving unit that the personnel in this unit may not realize its value or fungibility or possess the necessary tacit complementary knowledge to exploit it (Yang et al., 2008) since the contexts of the sending and receiving subsidiaries differ. In other words, the receiving unit may not possess the relative absorptive capacity to utilize the external knowledge (Cohen & Levinthal, 1990; Lane & Lubatkin, 1998). Therefore, even if the knowledge reaches the intended recipient, it may lie dormant in that unit without being exploited to its full commercial potential, suggesting that B_KE < 0 also for this reason.

Similarly, the congruity of the subsidiary’s knowledge with the knowledge held in the recipient units may influence the costs of knowledge transfer. If the knowledge stock is mainly internal, the overlap between the knowledge of the sender and recipient units is likely to be high. This implies low knowledge transfer costs (Reagans & McEvily, 2003): As the two units hold a certain amount of knowledge in common, less codification should be required because the transferred knowledge is easily recognized and understood and fewer costly misunderstandings occur. In contrast, external knowledge elements are likely to have fewer overlaps with other knowledge assets in the MNC. Therefore, even though other MNC units might benefit from this knowledge, their relative absorptive capacity to assimilate it is limited (Cohen & Levinthal, 1990; Lane & Lubatkin, 1998). In turn, this means that the encoding and decoding of this knowledge may require more effort and the need for face-to-face contact may be higher. These factors suggest that the costs of knowledge transfer will be higher if the subsidiary possesses more external knowledge (C_KE > 0), an effect that compounds the effects of external knowledge on transfer benefits and reinforces its effects on fungibility.

Assimilation and accommodation as responses to external knowledge

Although the above arguments suggest an unambiguous amplifying effect of external knowledge on knowledge transfer costs, they offer conflicting predictions as to the effect of external knowledge on knowledge transfer benefits. In essence, external knowledge may make the focal subsidiary’s knowledge more unique and less applicable at the same time. The notion of a negative relationship between external knowledge and knowledge transfer benefits deserves elaboration as it may at first seem counterintuitive: How can more knowledge possibly lead to lower knowledge transfer benefits? Indeed, the tendency in the economics and management literatures is to see organizational knowledge stocks as cumulative, so that they can be augmented by acquiring additional external knowledge that stacks on top of the preexisting knowledge and adds to the value of that knowledge. Cognitive psychologists use the term assimilation to describe such a process, where recipients incorporate the learned knowledge into their existing knowledge stocks, which in turn are left largely unchallenged (Posner, Strike, Hewson, & Gertzog, 1982). Under knowledge assimilation, we would never expect to see a negative relationship between external knowledge and knowledge transfer benefits. In the worst case, acquired nonfungible and incongruous external knowledge would merely fail to increase knowledge transfer benefits further, but it should not destroy the benefits of transferring the subsidiary’s remaining knowledge, which would still be there, largely intact and valuable. ⁵ Therefore, there would always be, at worst, a flat relationship between external knowledge and marginal knowledge transfer benefits, so that B_KE ≥ 0.

However, the cognitive psychology literature also recognizes that assimilation is only one part of the learning process and that a fundamentally different reaction pattern, called accommodation, may also occur in knowledge recipients (Dole & Sinatra, 1998; Vosniadou & Brewer, 1987). In the process of accommodation, the acquisition of new knowledge makes the recipients alter some of their existing knowledge structures. This process is described in detail in the psychology literature as pertaining to individuals (e.g., Piaget & Brown, 1985), and accommodation to host country knowledge may thus happen in subsidiary individuals such as country managers and R&D personnel. However, we suggest that it is likely to also be embedded on a higher level—the foreign subsidiary itself—since organizational learning “results in associations, cognitive systems, and memories that are developed and shared by members of the organization” (Fiol & Lyles, 1985: 804). Furthermore, even though accommodation is to some extent a natural part of the learning process in individuals, it may be problematic in the MNC context because acquired host country knowledge may lead subsidiaries to reinterpret and, possibly, discard internal MNC knowledge. Cognitive elements (whether at the individual or subsidiary level) may instead be infused with host country norms for what are best practices (e.g., What is considered a superior technology?) and host country perceptions of causal relationships (e.g., What determines the performance of a complicated product?). This could make the knowledge less congruous with and fungible for potential MNC recipients who do not share these norms and perceptions and thereby lead to lower knowledge transfer. This suggests a possible explanation for Acer’s problems, where units outside of the United States struggled with manufacturing the Aspire (suggesting low congruity of the American innovation with the rest of the MNC) and felt compelled to make significant and cost-increasing adaptations to it for their own markets (suggesting low fungibility). Effectively, there is a risk that the dominant logic (Prahalad & Bettis, 1986) of the subsidiary moves closer to that of its knowledge sources in the host country and further away from the dominant logic of the MNC. Therefore, the internal legitimacy of the subsidiary may be reduced because of the trade-off between internal and external legitimacy (Kostova & Zaheer, 1999). This may lead to increased subsidiary isolation and thus to reduced knowledge transfer, as pointed out by Monteiro, Arvidsson, and Birkinshaw (2008).

Interaction of internal and external knowledge

As a result of the contrary effects of assimilation and accommodation, it is ambiguous how the external knowledge resources controlled by a subsidiary will affect the benefits of transferring knowledge to other MNC units. If the subsidiary merely assimilates external knowledge, more external knowledge can never be directly detrimental because the knowledge, at best, can be profitably exploited by receiving units and, at worst, is worthless to these units. In that case, we would expect that B_KE ≥ 0. On the other hand, if external knowledge causes accommodation to take place in the subsidiary’s internal knowledge stock, which thereby becomes less congruous and fungible, a negative relationship between external knowledge and marginal knowledge transfer benefits may exist, because the composite knowledge of the subsidiary becomes less applicable in receiving units, that is, B_KE < 0.

We argue that accommodation to external knowledge is more likely in subsidiaries that have low levels of internal knowledge, whereas subsidiaries with high levels of internal knowledge either reject new external knowledge or assimilate it by reshaping it to fit their existing internal knowledge structures. Hence, the interaction of external and internal knowledge is related to knowledge transfer benefits, such that the subsidiary’s “own internal know-how will increase the marginal returns to external knowledge acquisition strategies” (Cassiman & Veugelers, 2006: 68). There are three ways in which this can happen. First, for external knowledge to be combinable with knowledge in other MNC units, it has to be interpreted and formulated in such a way that it can be accessed by other units (Grant, 1996). This requires internal knowledge and may increase the congruity and fungibility of the subsidiary’s composite knowledge stock. In the absence of this interaction with internal knowledge, external knowledge may generate lower benefits in other parts of the MNC because it is too distant from the recipients’ preexisting knowledge. In other words, they may lack the absorptive capacity and the complementary knowledge to exploit it (as seemed to be the case for Acer’s market units in other regions). In contrast, having both internal and external knowledge in the technological knowledge stock of the subsidiary would be an ideal basis for intrafirm knowledge recombination (Ahuja & Katila, 2001) because the knowledge of the originating subsidiary is then neither too similar to the knowledge held by potential receiving units nor too different from it.

Second, internal and external knowledge resources can be complementary, so that they can be combined into a more valuable composite knowledge resource and subsequently transferred to other MNC units. A subsidiary might thus be able to combine technological stimuli from its host market with the proprietary technologies of the MNC before it transmits this knowledge to other MNC units. Arguably, if the U.S. subsidiary of Acer had possessed sufficient internal knowledge, it may have been able to push Frog Design to develop a product that fit not only with emerging consumer tastes but also with the manufacturing capabilities of Acer’s global organization.

Third, if the subsidiary possesses internal knowledge, it may be better able to screen the external environment (Arora & Gambardella, 1994; Cohen & Levinthal, 1989), enabling it to acquire external knowledge that is more fungible and congruous, and thereby more relevant to MNC recipients, in the first place. Internal knowledge can thereby empower the subsidiary as a “listening post” for the MNC—a role in which “receiver competence (assessing, filtering, and choosing information) and absorptive capacity (adapting inflows to fit firm-specific requirements) become crucial” (Mudambi & Navarra, 2004: 389). Taken together, these three mechanisms suggest a positive interaction between internal and external knowledge in the determination of the benefits of knowledge transfer, so that B_KEI > 0.

As a highly stylized example of this process, consider a subsidiary that possesses a knowledge element A. Furthermore, suppose this knowledge element is fungible and congruous and (in the absence of more external knowledge) would therefore be associated with knowledge transfer benefits. What would happen if the subsidiary then increases its external knowledge level by acquiring a host country knowledge element B and combining it with A? If it has sufficient internal knowledge, it will reinterpret this new knowledge in the MNC context, in turn making both A and B useful to other units (an assimilation process leading to a positive relationship between external knowledge and knowledge transfer benefits). If the subsidiary has only a moderate level of internal knowledge, it may not be able to reinterpret the new knowledge in the MNC context and B may therefore be useless to other units (while A is still useful, leading to a flat relationship). Finally, if the subsidiary possesses low levels of internal knowledge, it may be so vulnerable to external influence that the new knowledge B leads it to inadvertently reinterpret A in a way that is specific to the host country (perhaps by discarding aspects of A that are unrelated to the local technologies embodied in B). This accommodation process decreases the usefulness of A to other MNC units. In that case, an increase in external knowledge actually leads to a decrease in knowledge transfer benefits.

Operationalizing the model

How can such a tipping point be estimated from actual data? To make the model amenable to empirical testing, we adopt the simplest polynomial approximation that fulfills our theoretical assumptions. This enables us to write the net benefit of transferring knowledge out of subsidiary j,

N B j = g 1 E j K j + g 2 I j K j + g 3 E j I j K j − g 4 K j 2 + b ’ t j K j + μ j K j

where g and b are parameter vectors with g₃ > 0 (because of the complementarity between I and E) and g₄ > 0 (diminishing marginal net benefits ⁹ ), t _j is a vector of observed subsidiary-specific variables that affect the costs and benefits of knowledge transfer, and µ _j captures the net effect of all unobserved subsidiary-specific variables that affect these costs and benefits. ¹⁰ The first-order condition for this model is given by, ¹¹

g 1 E j + g 2 I j + g 3 E j I j − 2 g 4 K j * + b ’ t j + μ j = 0

which can be solved for a prediction of K^* _j :

K j * = g 1 2 g 4 E j + g 2 2 g 4 I j + g 3 2 g 4 E j I j + 1 2 g 4 b ’ t j + 1 2 g 4 μ j ≡ γ 1 E j + γ 2 I j + γ 3 E j I j + β ’ t j + 1 2 g 4 μ j

In this expression, γ _i ≡g_i / 2g₄, t _j represents the control variables, and ¹ —_2g4 µ_jis a subsidiary-specific intercept. If we denote the mean value of this intercept as ϕ, and the difference between the individual subsidiary’s intercept and this mean as ε_1j, the expected value of ε_1j is 0 and 1 2 g 4 μ j = φ + ε 1 j . Substituting this into Equation 7 yields an equation that has the advantage that it lends itself to direct empirical testing by regression analysis:

K j * = φ + γ 1 E j + γ 2 I j + γ 3 E j I j + β ’ t j + ε 1 , j

We can now estimate the level of internal knowledge that is required if external knowledge is to be beneficial. This level can be estimated by asking what level of I would ensure that (K*) _E > 0: ¹²

( K * ) E = γ 1 + γ 3 I ≡ g 1 2 g 4 + g 3 2 g 4 I > 0 ⇔ I > − γ 1 γ 3 ≡ − g 1 g 3 ≡ I ⌣

Thus, the level of internal knowledge has to be above −γ₁ / γ₃ before external knowledge has a positive effect on the knowledge transfer out of the subsidiary. This tipping point can be estimated with data on E, I, and K.

Endogenous regressors

Although we have treated internal and external knowledge as exogenous variables in our framework, they could in fact be endogenous if MNC managers are simultaneously determining where to locate their subsidiaries, what knowledge to source internally and externally, and how much of that knowledge will be transferred. Therefore, if we estimated Equation 8 using ordinary least squares (OLS), we would run the risk of obtaining biased results and, consequently, a biased tipping point. In principle, if we could identify and observe all variables affecting K, we would be able to control for these variables and thereby solve the endogeneity problem (controlling for observables). However, in practice, we cannot a priori rule out the presence of unobserved variables that may affect both I and E, as well as K. For example, suppose that the disturbances ε_1j were created by unobserved subsidiary-specific variables—such as tacit knowledge management capabilities—that affect both knowledge levels and knowledge transfer. In that case, I, E, and ε_1j would be positively correlated (because they are affected by the same unobserved capabilities), which would lead to inflated estimates of γ in Equation 8 and thereby to a biased estimate of the tipping point in Equation 9.

To correct this problem, we can replace the actual values of the endogenous regressors with instrumental variables or, preferably, with predicted values based on multiple instruments (Kennedy, 2003). First, we add the following two equations to our model,

I j = η 0 + η ’ n j + ε 2 j

E j = α 0 + α ’ a j + ε 3 j

where ε _j are disturbances and n _j and a _j are exogenous variables that are expected to influence internal and external knowledge acquisition, respectively (described below). Together, Equations 8, 10, and 11 constitute a recursive system of simultaneous equations. If the disturbance terms ε _jK are correlated, as they will be if there are unobserved variables affecting all three endogenous variables (Greene, 2008), then OLS coefficients would be biased (as revealed by a Hausman test of endogeneity). A better estimation method in that case is two-stage least squares (2SLS), in which the endogenous regressors I and E are regressed on all of the exogenous variables in the first stage and their predicted values (which, unlike the actual values, are uncorrelated with ε_1j) are used in Equation 8 in the second stage. Finally, one may apply seemingly unrelated regression (SUR) to the 2SLS results to allow the error terms of the three equations to correlate in a three-stage least squares (3SLS) model (Zellner & Theil, 1962).

Knowledge transfer

Davenport and Prusak (1998) argue that knowledge transfer involves not just the transmission of knowledge but also the practical application and use of the knowledge by the receiver. Accordingly our measure of (successful) knowledge transfer captures the use of the subsidiary knowledge in other MNC units. The subsidiaries were asked about the extent to which technological knowledge under their control had been of use to other MNC units on a 7-point Likert-type scale, where 1 was defined as to no use at all for other units and 7 was defined as very useful for other units. The transfer of technological knowledge was measured in terms of the average usefulness of three types of knowledge, research (basic and applied), development (of products and processes), and production (of goods and services), and this scale showed acceptable reliability with Cronbach’s alpha = .65.

Internal knowledge

The “internal knowledge” construct captures the knowledge developed through interaction with other MNC units. To measure the knowledge inputs from other MNC units, respondents were asked to assess the impact of various internal organizations on the development of the subsidiary’s competences, where 1 equaled no impact at all and 7 equaled very decisive impact. Internal technological knowledge was thus measured as the impact of internal MNC R&D units on the subsidiary’s competences.

External knowledge

The “external knowledge” construct (i.e., subsidiary-level knowledge built mainly from external knowledge inputs) captures the importance of external parties. The inputs from external partners were measured by asking respondents to assess the impact of various external organizations on the development of the subsidiary’s competences, where 1 equaled no impact at all and 7 equaled very decisive impact. External technological knowledge was then measured as the impact of external R&D units on the subsidiary’s competences.

Predictors of internal knowledge

As predictors of internal knowledge in Equation 10, we include a number of variables suggested by theory. First, we obtained measures of subsidiary dependency, which captures the extent to which the focal subsidiary is dependent on knowledge from other MNC units. It was measured by asking respondents, “What would be the consequences for the subsidiary if it no longer had access to the competencies of other MNC units?” (1 = no consequences and 7 = very significant consequences). We believe that a subsidiary that perceives itself as being highly dependent on other MNC units would also be more motivated to receive knowledge from these units. Second, we included the subsidiary’s mode of formation (0 = greenfield, 1 = acquisition). We would have liked a more fine-grained measure including, for example, information about joint venture partners, but this information was not available. We nevertheless believe our dummy variable captures an important distinction: A subsidiary that is acquired as an independent firm and subsequently integrated into the MNC may be more reluctant to receive knowledge from its parent and other subsidiaries than one that is designed and staffed by the MNC from inception, for example because of the not-invented-here syndrome (Katz & Allen, 1982). Third, we included the age of the subsidiary (in years), which captures the duration of time the subsidiary has been exposed to MNC policies and internal knowledge. Arguably, internal knowledge may be more readily absorbed when the subsidiary has well-developed channels, routines, and organizational structures for exchanging information with other units, and building such an infrastructure is subject to time compression diseconomies (Dierickx & Cool, 1989). The passing of time itself may also enable employees to become more strongly identified with the firm and develop trust in the internal network, which facilitates internal knowledge acquisition.

Predictors of external knowledge

Extant research has suggested several factors that may influence subsidiary acquisition of external knowledge. In particular, host country characteristics have been shown to be important predictors of subsidiary innovation (Almeida & Phene, 2004). We obtained several measures to gauge these characteristics and, especially, the quality of the knowledge in the external business environment, which might enhance the external technological opportunities (Cohen & Levinthal, 1989) of the subsidiary. First, we constructed measures of science per capita for each host country in which our subsidiaries are located. This was calculated as the number of scientific and technical journal articles published in the relevant host country (sourced via the World Development Indicators database from the World Bank), normalized by dividing with the population of each country. Second, although knowledge environments differ from country to country, they also vary across localities and industries within a country (Porter, 1990). We therefore also used Porter’s (1990) diamond model as a more fine-tuned measure of location effects. We asked respondents for their assessment of cluster factors along the following dimensions: availability of business professionals, availability of supply material, quality of suppliers, level of competition, government support, legal environment, and existence of research institutions (1 = very low quality and 7 = very high quality). Porter’s (1990) emphasis on the holistic nature of the model and the high intercorrelation between many of the items motivated us to construct a composite index (Cronbach’s α = .67). These cluster factors extend the science per capita measure as they contain both country-specific effects and industry-specific effects and some level of interaction between them. Third, we asked the respondents to evaluate the institutional environment in the host country. If institutional conditions are perceived as favorable, the subsidiary may be more willing to open up and engage actively with the surrounding environment, for example, collaborating with host country partners, without an expectation of disputes or intellectual property appropriation. This, in turn, would lead to better opportunities for external knowledge sourcing.

Knowledge transfer control variables

Finally, when predicting knowledge transfer in Equation 8 we included a number of variables related to the role of the subsidiary in the MNC network. When a subsidiary has a mandate as a strategic asset-seeking unit (Dunning, 1993), a competence-creating subsidiary (Cantwell & Mudambi, 2005), or a center of excellence (Frost et al., 2002; Holm & Pedersen, 2000), its knowledge is likely to be formally and informally recognized and valued by headquarters, by other units, and by subsidiary employees. This, in itself, may encourage the subsidiary to transfer more knowledge regardless of its actual knowledge levels. We therefore controlled for the MNC’s formal recognition of and dependency on the knowledge residing in the subsidiary, as well as the subsidiary’s own investments in and assessments of its competences. Formal recognition was measured by asking whether the indicated competencies of the subsidiary were formally recognized by the MNC headquarters. Respondents could answer yes or no to the following three activities: research, development, and production (the sum of which measured formal recognition of technological knowledge, from 0 to 3).

The MNC dependency on the knowledge of the focal subsidiary was measured using the following question: “What would be the consequences for other units in the foreign company if they no longer had access to the competencies of the subsidiary?” (1 = no consequences and 7 = very significant consequences). The subsidiaries’ own investments were measured by asking respondents to assess the level of investments in the subsidiary in the past 3 years, where 1 = very limited and 7 = substantial. Like formal recognition, the level of investments was assessed for the three technological activities (and averaged to determine investments in technological knowledge; Cronbach’s α = .69). In a similar way, the subsidiaries’ own assessment of their technological knowledge was captured by asking respondents to indicate the level of subsidiary competencies for these same activities (on a 7-point scale; Cronbach’s α for assessment of technological knowledge = .71). The subsidiaries vary in size, and their ability to absorb and exploit knowledge might be related to size, for example, in terms of employees. Therefore, we included a size measure of logarithm of total number of subsidiary employees as a control variable. Finally, we controlled for geographic distance with a dummy variable measuring whether the parent MNC was headquartered in Europe.

The correlation coefficients and descriptive data (mean values, standard deviations, and normality statistics) on all variables are provided in Table 3.

OPEN IN VIEWER

Table 3

Correlation Matrix

Variable	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15
1. Transfer of technological knowledge	1.00
2. Internal technological knowledge	.20	1.00
3. External technological knowledge	.27	.41	1.00
4. Subsidiary dependency	−.09	.26	.04	1.00
5. Subsidiary age (log)	−.09	.09	−.02	.19	1.00
6. Formation	.04	.07	.06	−.25	−.32	1.00
7. Institutional conditions	.11	.16	.25	.07	−.01	−.01	1.00
8. Cluster factors	.17	.15	.17	.07	.02	.03	.16	1.00
9. Science per capita	−.06	−.08	.12	−.04	−.02	.08	−.10	−.19	1.00
10. MNC dependency	.40	.17	.19	.18	.03	.04	.10	.15	−.12	1.00
11. Formal recognition technological knowledge	.47	.14	.19	−.12	−.08	.24	.07	.20	−.12	.31	1.00
12. Own technological competencies	.24	.04	.05	−.08	−.07	.09	−.02	.06	.06	.09	.27	1.00
13. Subsidiary investments in technology	.60	.19	.25	−.14	−.10	.24	.10	.14	.01	.28	.41	.16	1.00
14. Subsidiary employees (log)	.45	.09	.14	−.13	.01	.26	.10	.21	−.17	.23	.44	.17	.42	1.00
15. Headquartered in EU	−.04	−.04	−.01	−.06	−.01	.04	−.01	−.01	.01	−.01	−.01	−.02	−.04	−.12	1.00
M	2.01	2.86	2.10	4.51	2.63	0.58	2.61	4.31	0.08	3.26	0.89	5.41	2.46	4.71	0.73
SD	1.86	1.90	1.47	1.89	0.81	0.49	1.58	0.83	0.02	1.76	1.00	1.05	1.68	1.69	0.44
Min	0	1	1	1	0.69	0	1	0	0.05	1	0	1	0	0	0
Max	7	7	7	7	4.78	1	7	7	0.11	7	3	7	7	13.21	1
Kurtosis	−0.27	−0.87	0.92	−1.07	−0.45	−1.88	−0.13	2.83	−1.37	−0.84	−0.68	2.13	−0.79	−0.01	−0.93
Skewness	0.72	0.64	1.35	−0.34	0.21	−0.34	0.80	−0.67	−0.17	0.46	0.74	−1.00	0.18	0.23	−1.04

Note: N = 2,107. All correlations > .04 or < –.04 are significant at the 5% level.

The bivariate correlations between knowledge transfer and both internal and external knowledge are positive and highly significant. Five control variables (MNC dependency, formal recognition, own competencies, subsidiary investments, and subsidiary employees) are also positive and significantly related to transfer of knowledge. Surprisingly, the sixth control variable, EU headquarters, seems not to be related to any of the other variables in the data set. It is also noticeable that none of the correlation coefficients among the independent variables indicate the possibility of multicollinearity (i.e., r > .5). Our dependent variable, knowledge transfer, has a low mean (about 2), which implies technological knowledge that has been of low to moderate use to other units, but the standard deviation and skewness suggest a long right tail. This suggests that a substantial number of subsidiaries indicate that their knowledge is of no use at all to other units (1), whereas a few see their knowledge as more or less useful (2–6) or even very useful (7). In fact, 101 subsidiaries (about 5% of the sample) answered either 6 or 7 on the 7-point Likert-type scale. This is surprising but could be interpreted as an indication of a “star system” pertaining to reverse knowledge transfer, where a few subsidiaries claim a disproportionate share of top management attention (Blomkvist, Kappen, & Zander, 2010). Finally, we can see from Table 3 that internal knowledge (mean of about 2.9) seems to be more important to the subsidiaries than external knowledge (2.1), perhaps reflecting barriers that subsidiaries with limited absorptive capacity face when sourcing knowledge from their host markets.