Skip over navigation to the main content
Searching:9 databases
16 Recent searches
|
0 Selected items
|
My Research
|
Exit
Basic Search
Advanced
Publications
Browse
Preferences
English(Change Language)
– this link will open in a new windowHelp
ProQuest
Full text
Back to results
Previous(previous record)
Document 18 of 8891
Next(next record)
Add to selected items
Save to My Research
Email
Print
Cite
Export/Save
Tags
Share
Collapse panel
Other formats:Citation/Abstract
Full text – PDF (2 MB)?
ReferencesCited by (19)
More like this
See similar documents
Search with indexing terms
Subject
Organizational learning
Market entry
Research & development
R&D
Foreign investment
More…
Location
Japan
Search
Entry mode, organizational learning, and R&D in foreign affiliates: Evidence from Japanese firms
Belderbos, ReneView Profile
. Strategic Management Journal24. 3 (Mar 2003): 235-259.
Turn on hit highlighting for speaking browsers
Show highlighting
Abstract (summary)
Translate Abstract
This paper develops hypotheses concerning the role of entry mode and experience-based organizational learning as determinants of the R&D industry of foreign affiliates and tests these hypotheses on a sample of 420 Japanese manufacturing affiliates abroad. Entry mode has a major impact on R&D activities: the R&D intensities of acquired affiliates substantially exceed those in wholly owned greenfield affiliates, while the R&D intensities of minority owned ventures are higher if Japanese parent firms lack strong R&D capabilities at home. For greenfield operations, support is found for an incremental growth pattern of foreign R&D as a function of organizational learning and affiliate capability building. The results are consistent with the view that part of the explanation for Japanese firms’ relative lack of involvement in overseas R;D must be sought in their status as ‘latecomers’ in the establishment of overseas manufacturing networks.
LEC
LEC
Full Text
Translate Full text
Turn on search term navigation
Headnote
Received 13 January 2000
Final revision received 26 August 2002
Headnote
Key words: R;D; entry mode; organizational learning; Japanese firms
Headnote
This paper develops hypotheses concerning the role of entry mode and experience-based organizational learning as determinants of the R;D intensity of foreign affiliates and tests these hypotheses on a sample of 420 Japanese manufacturing affiliates abroad. Entry mode has a major impact on R;D activities: the R;D intensities of acquired affiliates substantially exceed those in wholly owned greenfield affiliates, while the R;D intensities of minority owned ventures are higher if Japanese parent firms lack strong R;D capabilities at home. For greenfield operations, support is found for an incremental growth pattern of foreign R;D as a function of organizational learning and affiliate capability building. The results are consistent with the view that part of the explanation for Japanese firms’ relative lack of involvement in overseas R&D must be sought in their status as ‘latecomers’ in the establishment of overseas manufacturing networks. At the same time, a number of Japanese firms have actively used foreign acquisitions and joint ventures to gain access to overseas technology and to establish overseas R&D capabilities at a faster pace. Copyright ?2002 John Wiley &Sons, Ltd.
INTRODUCTION
There is continuing interest in the internationalization of research and development (R&D) activities by multinational firms. Two basic motivations for the internationalization of R&D are distinguished: exploitation of the firm’s technology abroad through adaptation of technologies to local circumstances, and creation of technologies through access to overseas technology and know-how.1 Recent work has suggested an increasing importance of foreign R;D associated with a growing role of global technology creation. There is evidence that more foreign R;D sites are assigned the role of creators of basic technologies and developers of completely new products for world markets (e.g., Florida, 1997; Kuemmerle, 1997; Pearce and Singh, 1990; Gerybadze and Reger, 1999; Hakanson and Nobel, 1993a, 1993b).2 Driven by shortening of product life cycles, increased global competition and rapid technological developments, multinational firms need to ‘tap into’ centers of world excellence in given technological fields (Bartlett and Ghoshal, 1989). The challenge is to utilize local technological learning in geographically dispersed sites by communicating and integrating it into the firm’s global organization and leveraging it in other markets. This is the subject of an emerging literature on effective coordination and control processes in international R&D and the relationship with organizational learning (Brockhoff and Schmaul, 1996; DeMeyer, 1997; DeMeyer and Mizushima, 1989; Gassmann and von Zedtwitz, 1999; Ghoshal and Bartlett, 1988; Reger, 1999).
A common finding in the literature on international R&D is that Japanese multinational firms distinguish themselves from European and American firms by a particular reluctance to internationalize their R&D operations (Bartlett and Ghoshal, 1991; Behrman and Fischer, 1980; Cheng and Bolon, 1993; Patel, 1996, 1995; Pearce and Singh, 1990). For instance, in the late 1980s a mere 1 percent of U.S. patent grants to the 139 largest Japanese industrial multinational firms was based on research performed in overseas laboratories (Patel, 1995), compared to 8 percent for U.S. multinationals and 40-60 percent for European multinationals. Comparative analysis of R&D activities in large multinational firms has also indicated that Japanese firms operate much more recently established and smaller-scale laboratories compared to U.S. and European multinationals (Kuemmerle, 1999). Japanese firms furthermore display a preference to maintain relatively strong headquarter control over all corporate R&D activities including overseas R&D operations (Bartlett and Ghoshal, 1991; Behrman and Fischer, 1980; Cheng and Bolon, 1993; Gassmann and von Zedtwitz, 1999; Ghoshal and Bartlett, 1988; Reger, 1999). Two explanations have been put forward for these features of Japanese multinationals. A first explanation suggests that the limited scale of R&D operations abroad is simply a corollary of the ‘latecomer’ status of Japanese multinational firms. Their relatively late establishment of substantial overseas manufacturing operations since the second half of the 1980s has not allowed them to gain much experience in operating decentralized networks of manufacturing plants and to build up foreign R&D capabilities (Belderbos, 1997, 2001; Westney, 1996). Such international experience is seen as a prime source of organizational learning in multinational firms (e.g., Penrose, 1959; Barkema and Vermeulen, 1999; Kogut and Zander, 1993, 1995) allowing for more effective technology transfer to overseas affiliates and more effective management of foreign R&D operations. In this view, R&D activities abroad will expand more rapidly after firms have learned how to operate and integrate a network of dispersed manufacturing and R&D facilities abroad.
On the other hand, a second explanation for Japanese firms’ limited overseas R;D operations links this feature to the specific characteristics of Japanese R;D strategy and R;D management. Japanese firms have given priority to development and design for manufacturability and the associated speed of new product introduction (Westney, 1993). This focus requires intensive coordination and communication between strategic decisionmaking units at headquarters and R;D, marketing, and manufacturing functions, as well as collaboration with suppliers (Odagiri and Goto, 1993; Odagiri and Yasuda, 1996). To establish these close interlinkages, Japanese firms make intensive use of personnel rotation, person-toperson communication and informal control mechanisms to facilitate information exchange and to create common values and ideas at different levels and functions throughout the organization (Westney, 1994; Reger, 1999). Newly developed products are test manufactured in Japanese pilot plants (Fruin, 1992) and first sold on the sophisticated and quality-conscious Japanese market, which provides stimulus for continuous improvement (Porter, 1990). This orientation and organization of R;D requires a relatively strong geographical concentration of manufacturing, applied R;D, and engineering facilities near headquarters in Japan (Kenney and Florida, 1994; Westney, 1994). The strong integration of R;D with other functions of the firm is considered one of the factors behind the rise in Japanese firms’ technological capabilities and global competitiveness (e.g., Freeman, 1987; Wakasugi, 1992), but at the same time this system of R&D integration is inherently difficult to transfer abroad over long distances and cultural and language barriers.
Given the importance of foreign technological opportunities and the global competitive pressures facing Japanese firms, the question arises to what extent they have initiated changes in their R&D strategy and organization to enable a more decentralized and internationalized organization of R&D with a stronger focus on research and technology creation. Such changes in strategy may be expected given that the Japanese are strongly aware of the need to ensure access to foreign science and technology for building up corporate technological capabilities (Granstrand, 1999). There is some suggestion in the literature that the response of Japanese firms may involve the establishment of joint ventures with foreign partners and the acquisition of R&D-intensive foreign firms. Granstrand (1999) reported that R&D managers in large Japanese multinationals perceive their R&D strategy to rely relatively strongly on acquisitions, joint ventures, and collaboration with foreign universities, and Belderbos (2001) found evidence of an important role of acquired R&D units in patented overseas innovations by Japanese electronics firms.
In this paper we address two major research questions that emanate from this discussion. First, to what extent is the limited extent of foreign R&D by Japanese multinational firms due to their ‘latecomer’ status as multinational firms and the associated restricted learning opportunities they have had in effectively managing a dispersed network of foreign manufacturing and R&D activities? Second, given the perceived need to internationalize R&D, to what extent are acquisitions and collaborative ventures used to achieve a more rapid increase in foreign R&D capabilities? We seek to answer these questions through a systematic analysis of the determinants of the R&D intensity of a large sample of Japanese multinational firms’ overseas affiliates active in manufacturing industries. Informed by different theoretical perspectives among which transaction cost theory on foreign direct investment and entry mode choice and the evolutionary view of the multinational firm in relationship with organizational learning, we develop hypotheses with respect to the impact of entry mode and experience on the R;D intensity of foreign affiliates. The hypotheses are tested on a sample of 420 Japanese overseas affiliates, drawn from unpublished data in MITI’s benchmark survey on foreign investment for the fiscal year ending March 1993. This rich dataset also allows us to utilize a large set of controls for locational, industry, and other affiliate and parent firm characteristics.
Our paper is the first to analyze affiliate level data on R&D for a large and diverse sample of Japanese firms. As noted in surveys by Cheng and Bolon (1993) and Granstrand, Hakanson, and Sj”olander (1993), unavailability of data has kept the number of desegregated studies extremely limited. Most previous studies of overseas R&D expenditure (e.g., Hewitt, 1980; Hirschey and Caves, 1981; Kumar, 1996; Lall, 1979) substituted the industry for the firm, although it has been shown that industry-level results can often not be replicated in firm-level studies (Cheng and Bolon, 1993). Later studies have examined R&D internationalization at the level of the parent firm (e.g., Belderbos, 2001; Odagiri and Yasuda, 1996; Pearce, 1989). To date only a few studies of Swedish multinational firms (Fors, 1996; Hakanson and Nobel, 1993a, 1993b; Zejan, 1990) have systematically examined the determinants and characteristics of overseas R&D expenditures at the desegregated level of foreign affiliates.3
The remainder of this paper is organized as follows. In the next section we give an overview of the relevant literature and develop hypotheses concerning the determinants of the R&D intensity of overseas affiliates. After a description of the methodology and the dataset we present the statistical results. In the final part of the paper we discuss the results and provide conclusions.
THE DETERMINANTS OF THE R&D INTENSITY OF FOREIGN AFFILIATES: BACKGROUND AND HYPOTHESES
Background
In order to explain the roles of entry mode and organizational learning in foreign R&D activities, a number of theoretical perspectives and empirical research traditions are of relevance. Theoretical perspectives include Vernon’s product life cycle model of foreign direct investment, transaction cost theory, and the evolutionary view of the multinational firm. Relevant applied literature includes work on overseas R;D management, international mergers and acquisitions, and entry mode choice for foreign ventures. We will briefly review the relevant elements of these literatures below and draw on it in the formulation of our hypotheses.
Foreign R;D and foreign R;D management
Vernon’s product cycle theory regards R&D as the functional activity that multinational firms are only likely to transfer abroad in the later stages of their internationalization (Vernon, 1979). In product cycle theory, overseas R&D is motivated by the need to adapt products and production processes to local markets and resource conditions. Foreign R&D follows the expansion of overseas marketing and manufacturing activity and focuses on local improvements of relatively mature technologies developed in home country laboratories. R&D in overseas affiliates is conducted to allow better exploitation of the multinational firm’s technologies and is attracted by a large market potential for product adaptations and the productivity-enhancing potential of R;D applied to local manufacturing operations.
The transaction cost theory of foreign direct investment predicts the conditions under which the overseas exploitation of the firm’s proprietary technologies is conducted internally (through foreign direct investment), or externally through licensing contracts with local firms (e.g., Caves, 1995; Belderbos and Sleuwaegen, 1996). If the transfer of technology requires (transaction-)specific (R&D) investments to adopt and adapt technologies to overseas conditions, hold-up risks increase the cost of arm’s-length transactions. Asymmetric information concerning the characteristics and value of the technology between the licensor and licensee further reduces the return on external transfers and makes intrafirm transfers to foreign affiliates preferable. On the other hand, the same features of technology transfers imply that if they occur intrafirm to foreign affiliates they are not without costs either (e.g., Mansfield and Romeo, 1980). The extent to which technology is transferred to overseas affiliates has been shown to increase over time as the transfer costs decrease due to improved codification and standardization of the technology and reduced tacitness of knowledge content (Teece, 1977). The evolutionary view of the (multinational) firm further emphasizes the importance of the firm’s capability to learn from each transfer abroad and accumulate knowledge on how to apply tacit knowledge across borders in different geographic locations (Penner-Hahn, 1998; Kogut and Zander, 1993, 1995; Martin and Salomon, 2000). In this view, international experience is a prime source of organizational learning in multinational firms.4 In deciding on the internal transfer of proprietary assets abroad, the firm will not only take into account the direct costs and benefits in comparison with arm’s-length transactions but also that internal transfer and exploitation lead to the accumulation of critical organizational know-how and resources in the form of transfer management skills and know-how concerning the management of applications of the technology (Kogut and Zander, 1993, 1995). Multinational firms that have built up the required skills and routines are more likely to transfer proprietary technologies to, and adapt these technologies in, their foreign affiliates.
The literature on international R;D management has shown that the adaptation of coordination and control processes to efficiently manage a geographically dispersed network of R;D sites is extremely difficult and requires substantial trial and error and investments in new organizational routines. One issue compounding problems is that, given the differences in overseas R;D resources and environment, uniform procedures cannot be applied and affiliate-specific mechanisms are required (Nobel and Birkinshaw, 1998). In practice, few multinational firms yet operate global R;D collaboration systems among dispersed sites working on common projects. Firms rather attempt to minimize interdependencies among laboratories by letting them specialize in particular technological fields. A particularly difficult change is moving from a system of centralized management control to decentralized management necessary to provide sufficient autonomy to overseas R;D managers (DeMeyer and Mizushima, 1989). Brockhoff and Schmaul (1996), for instance, found that German multinational firms that operated a centralized ‘hub’ model with centralized coordination of overseas R;D perceived their R;D management system to perform better than firms that had started to operate a ‘network’ model with more decentralized control and distributed research tasks. Whereas the latter has the potential of increasing worldwide technological performance, leveraging locally generated knowledge by communicating research results and integrating it into global operations is still a major challenge for most multinational firms (Nobel and Birkinshaw, 1998; DeMeyer, 1997). At the same time, studies of the management of overseas R;D operations have reported evidence of an increasing role of foreign affiliates in research as creators of new technologies based on host countries’ technological strength and specialization (e.g., Kuemmerle, 1997; Florida, 1997; DeMeyer, 1997; Almeida, 1996; Branstetter, 2000; Frost, 2001). Empirical evidence has suggested that such assignment of more substantial R&D tasks and responsibilities to specific affiliates is a gradual process as a function of organizational learning and capability building in foreign affiliates (Hewitt, 1980; Hakanson and Nobel, 1993a, 1993b). The importance of such an evolution in subsidiary resources and tasks is also corroborated by an expanding literature on foreign subsidiary roles and dynamic development in affiliates driven by changes in the environment, headquarter assignment, and autonomous affiliate choices (e.g., Rugman and Verbeke, 2001; Jarillo and Martinez, 1990; Hood, Young, and Lal, 1994).
Although Japanese firms have overall adopted more centralized R&D management systems than U.S. and European firms and have been reluctant to assign substantial R&D tasks to foreign affiliates (Bartlett and Ghoshal, 1991; Cheng and Bolon, 1993; Reger, 1999), case study evidence has shown ample variation in management practices and strategies within the population of Japanese multinationals. For instance, Nippon Steel has traditionally focused strongly on R&D efficiency and has concentrated virtually all R&D and manufacturing activities in a few locations in Japan (Gassmann and von Zedtwitz, 1999). On the other hand, Sony has adopted a strong international orientation in marketing, manufacturing, and R&D early after the firm’s incorporation in response to the vision of its founders. In recent years it has led attempts to devolve decision-making power for R;D to overseas affiliates (Reger, 1999).
The relationship between R;D and entry mode
Previous studies examining the relationship between R;D and entry mode have treated R;D capabilities as one of the determinants of the choice between foreign acquisitions, joint ventures, and wholly owned greenfield entry (e.g., Hennart, 1991; Hennart and Park, 1992; Gomes-Cassares, 1989), rather than the consequences of entry mode choice for R;D operations.5 In the literature on entry mode choice, acquisitions are often viewed as a means to provide the acquiring firm with the opportunity to learn about technologies with which it is not familiar (e.g., Roberts and Berry, 1985), i.e., for technology sourcing. Internal growth through greenfield entry is chosen if the venture builds on existing technologies and competencies involving at least initially the exploitation of parent firm technological capabilities.6 Empirical studies have confirmed that entry by acquisition is more common than greenfield entry for firms with less distinct R;D capabilities or competitive advantages vis`a-vis foreign rivals (Hennart and Park, 1992; Yamawaki, 1993), suggesting that acquisitions are used as a means to acquire competitive resources. This is in line with other evidence that access to proprietary technologies and knowhow is an increasingly important motivation for acquisitions (e.g., Granstrand and Sj”olander, 1990; Chakrabarti, Hauschildt, and S”uverkrup, 1994; Granstrand et al., 1992; Chaudhuri and Tabrizi, 1999; Ahuja and Katila, 2001; Belderbos, 2001). Recent work has taken a resourcebased perspective to analyze the recombination and redeployment of (R;D) resources after acquisitions (Capron, Dussauge, and Mitchell, 1998; Ahuja and Katila, 2001). Capron et al. (1998) found substantial redeployment of R;D resources in a large sample of acquisitions in Europe and North America during 1988-92. They argue that reconfiguration of resources is often necessary to improve existing operations and sustain competitive advantages in response to environmental changes and increased competition. Acquisitions are favored over internal development if imitation of unique organizational resources of the target firm is difficult or if this would take a substantial period of time. Indeed, a crucial advantage of acquisitions is the speed of entry into new technological fields as rapid changes in the competitive environment do not allow firms enough time to develop new resources internally (Capron et al., 1998: 633; Chaudhuri and Trabizi, 2000). Organizations may also face strong internal constraints on developing new resources as innovative search routines are based on existing practices and restrain the technological domain of the firm’s activities. Barkema and Vermeulen (1999) suggest that foreign acquisitions can be an important means to overcome organizational learning boundaries and established routines and processes that have outlived their usefulness. Acquisitions may in this regard ‘revitalize’ firms, but at the same time they are likely to break with existing organizational learning practices and evolutionary development in (foreign) R&D operations and R&D management practice.
Another line of research has examined multinational firm’s foreign entry mode choice between joint ventures and wholly owned establishments. A large number of empirical studies have applied transaction cost theory to predict under what circumstances multinational firms would choose to establish foreign joint ventures with local partners rather than wholly owned affiliates (e.g., Gatignon and Anderson, 1988; Gomes-Cassares, 1989; Hennart, 1991; Mutinelli and Piscitello, 1998; PennerHahn, 1998; Nakamura and Yeung, 1994). These studies have generally confirmed the hypothesis that parent firms with advanced proprietary technologies planning to transfer technology to foreign ventures are more likely to choose to establish wholly owned affiliates rather than joint ventures. Controlled ventures are preferred because of the risk of opportunism on the side of the foreign partner and the potential loss of proprietary technology after it is transferred to the joint venture. Joint ventures are preferred if collaboration with the local partner provides complementary resources, which may consist of specific technologies or local market knowledge.
Summarizing, the diverse literature on foreign direct investment and foreign R;D, organizational learning, and entry mode choice suggests that a number of factors impact on the overseas R;D activities in foreign affiliates. Internationalization of R;D is likely to increase with the multinational firm’s experience-based organizational learning as a function of operating a decentralized and dispersed network of foreign affiliates and repeated international technology transfers. The extent of R&D operations is also likely to differ between individual affiliates due to a gradual development of affiliate-specific technological capabilities. Acquisitions may be used to gain access to distinct foreign technological capabilities but at the same time they are likely to break with past organizational routines and organizational learning processes. The establishment of foreign joint ventures can be motivated by the potential value creation of combining distinct R&D capabilities of the multinational and the local partner firm, but wholly owned greenfield entry is more likely to be chosen if the multinational firm possesses strong technological advantages which it plans to transfer abroad.
Hypotheses
We develop two sets of hypotheses concerning the impact of organizational learning and entry mode on the R&D intensity of foreign affiliates. Following the theoretical arguments above and previous work (e.g., Barkema and Vermeulen, 1999), we focus on international experience as the prime source of organizational learning relevant for international R&D operations. The first three hypotheses developed below are applicable to greenfield entries, among which are joint ventures, while the subsequent two hypotheses are applicable to acquisitions.
Greenfield entry
The notion of organizational learning as firms accumulate experience in operating a network of international manufacturing and marketing operations suggests that parent firms’ international experience has a positive impact on the decision to invest in foreign R;D operations. Product cycle theory, transaction cost theory on technology transfer, and the evolutionary view of the multinational firms all suggest an increase in transfer of the parent firm’s technology abroad and more extensive efforts in foreign affiliates to develop products to local market conditions as the firm gains experience in decentralized management of a dispersed network of affiliates. Firms that have been relatively early in setting up manufacturing operations abroad are more likely to have built up critical organizational know-how on international transfers of technology and tacit knowledge as well as the adaptation of technologies to different production and market environments (e.g., Kogut and Zander, 1995). These firms are also more likely to have learned how to use communication and control mechanisms facilitating the assignment of more complex tasks to overseas affiliates. This adaptation of corporate routines to international operations involves a gradual change in R&D coordination mechanisms allowing the assignment of greater R&D responsibilities to overseas affiliates (Belderbos, 2001; Hakanson and Nobel, 1993a, 1993b; Pearce, 1989; Zejan, 1990), which may also include more explorative research activities. Corroborating evidence is provided by Granstrand et al. (1993), who find that firms expanding overseas relatively late but rapidly typically operate less internationalized R&D operations. These relationships between international experience and incremental changes in technology transfer and international R&D coordination are characteristics of firms expanding through internal growth and greenfield entry, with overseas R&D operations building on or further developing the existing technology base of the parent.
Hypothesis 1: Parent firms with longer international experience through early investment in greenfield entries devolve greater R&D responsibilities to their overseas affiliates and operate affiliates with higher R&D intensities.
Organizational learning and gradual capability building in technology transfer and R&D also take place at the level of individual greenfield affiliates (Hewitt, 1980; Hakanson and Nobel, 1993a, 1993b). Affiliates that have been established to manufacture and sell products based on the parent firm’s technologies and product designs gradually obtain knowledge of local market characteristics and manufacturing conditions, which allows them to adapt product and processes to local conditions. As foreign affiliates gain manufacturing and design experience over time and establish credibility as competent and reliable adaptors of technologies, they are more likely to be assigned more complex tasks such as developing new products for local and regional markets (DeMeyer, 1997). In addition, they are more likely to be seen as reputable ‘insiders’ in local know-how sharing networks and have easier access to sources of knowledge creation in the host country (Frost, 2001). This in turn may foster an increasing role of affiliates in terms of local technology sourcing which may require a further commitment to applied or basic research. The implication is that affiliates that have been operating longer in a host country are more likely to have developed know-how on product and process adaptation, to have gained access to local know-how-sharing networks, and to have established themselves as a reliable factor in the firm’s manufacturing and R&D network. These characteristics are all associated with greater affiliate expenditures on R&D.
Hypothesis 2: Affiliate experience has a positive impact on the R&D intensity of greenfield affiliates.
The type of greenfield operations, wholly owned ventures or joint ventures, is expected to impact on the R&D operations of the affiliates in a more complex way. A stylized fact from empirical research on entry mode choice (e.g., Gatignon and Anderson, 1988; Gomes-Cassares, 1989; Hennart, 1991; Kogut, 1988; Mutinelli and Piscitello, 1998) is that parent firms with substantial R&D capabilities are more likely to choose to establish wholly owned affiliates. The main motivation for setting up wholly owned greenfield affiliates is to exploit the parent firm’s technology base in the foreign market. Wholly owned greenfield affiliates are preferred because the firm can exercise full management control, while joint ventures run the risk of opportunism on the side of the foreign partner and the potential loss of proprietary technology in case of technology transfer to the foreign venture. Additional evidence is provided by Ramachandran (1993), who found that the level of technology transfer to foreign ventures in India was significantly lower for minority owned or balanced stake joint ventures affiliates compared with wholly owned affiliates. This suggests that the scale of R;D activities in foreign affiliates, if these R;D activities are based on the parent firm’s proprietary technology base, is likely to be smaller in joint ventures than in wholly owned firms. This effect is greater, the greater the firms’ technological advantage.
On the other hand, if a firm establishes a joint venture abroad to benefit from complementary technological competencies of the local partner, the joint venture may be involved in more substantial R;D activities compared with the firm’s wholly owned affiliates. Joint ventures in this case are set up with the motivation of technology sourcing, to benefit from access to local technological know-how and R&D capabilities. In particular, if part of this know-how is tacit and particular to specific local firms it cannot readily be accessed through the establishment of a wholly owned affiliate. At the same time, studies have suggested that for firms wishing to be successful in technology sourcing activities, a degree of inhouse R&D is necessary to assess the value of, and to assimilate and absorb, technological information acquired externally (e.g., Veugelers, 1997; Cohen and Levinthal, 1990). There is previous evidence that technology sourcing has indeed been a motive for Japanese joint ventures abroad and that the importance of this motivation depends on relative R&D capabilities. Kogut and Chang (1991) found that Japanese firms set up joint ventures more frequently in those U.S. industries that were more R&D intensive than the comparable industries in Japan. Hennart, Roehl, and Zietlow (1999) concluded that Japanese joint ventures in the United States established in the 1980s were set up to benefit from the U.S. partner’s technological capabilities by building up new R;D resources.7
The above arguments and findings suggest that differences in the importance of R;D operations between joint ventures and wholly owned affiliates depend on the R;D capabilities of the investing firm and the related relative importance of technology sourcing or technology exploitation motivations. If a firm possesses important R;D capabilities it is likely to establish foreign affiliates to exploit its technologies abroad, and will establish wholly owned affiliates to this end. In case such firms do establish joint ventures they are less likely to transfer key technologies, and R;D activities of the ventures are likely to be relatively small in scale. These differences are most pronounced between wholly owned affiliates and minority owned joint ventures in which the parent firm does not possess management control and the risks of unwanted dissipation of proprietary technology are greatest. In contrast, for parent firms that lack important R;D capabilities the opposite relationship is expected to be present. They are more likely to set up joint ventures in order to get access to, and build on, the technological know-how of the local partner. This is likely to involve higher R;D intensities than in their wholly owed affiliates, which have more limited opportunities for technological exploitation given the relatively weak technology base of the parent.8 Again, these differences are likely to be most pronounced between wholly owned ventures and minority owned ventures.9 In case the local partner has important technological capabilities on which the joint venture is expected to build, the literature on entry mode choice suggests that the investing firm is less likely to be able to negotiate an equity stake giving it majority control (e.g., Gomes-Cassares, 1989; Nakamura and Yeung, 1994).10
Hypothesis 3: For parent firms with strong R;D capabilities, minority owned joint ventures have lower R;D intensities than wholly owned greenfield affiliates. For parent firms with weak R;D capabilities, the opposite pattern holds.
Entry through acquisition
In contrast with firms expanding through internal growth and greenfield entry strategies, an evolutionary development in corporate routines and decentralized management is not a characteristic of firms that rely on foreign acquisitions to expand abroad. Acquisitions are often chosen precisely to move away from existing bounds of organizational learning and technological knowledge in order to access complementary foreign R;D capabilities, R;D management practices, and organizational know-how (e.g., Barkema and Vermeulen, 1999). A major advantage of acquisitions is the speed at which the firm can access research capabilities highly embedded in overseas technological networks, in order to overcome the time-consuming and uncertain process of developing these capabilities and linkages through internal growth (Chaudhuri and Tabrizi, 1999; Capron et al., 1998). Hence, acquisitive multinational firms rely less on organizational learning processes that allow a gradual upgrading of foreign R;D responsibilities. The implication is that for acquisitive firms, previous international experience is not a determining factor of foreign R;D operations.
Hypothesis 4: A positive relationship between the parent firm’s international experience and foreign R&D does not exist for firms that expanded abroad through acquisitions.
Acquired foreign affiliates are expected to show different determinants of R&D activities than greenfield affiliates. A first difference occurs at the time of investment. Since the acquired affiliate in most cases will have an existing technology base, the affiliate by definition does not have to go through the time-consuming process of incrementally building up its knowledge base and organizational capabilities. At the time of investment, when the acquiring firm yet has to redeploy resources and to gain experience in controlling and operating the affiliate, acquired affiliates are for this reason likely to show greater R&D intensities than greenfield affiliates. Furthermore, the R&D intensity of acquired affiliates will be substantially higher if the main purpose of the acquisition is technology sourcing: to access and utilize the affiliate’s R;D capability and its linkages into the local system of innovation. Empirical studies have suggested that an important part of overseas R;D expansion by Swedish and Japanese firms (the latter since the late 1980s) has been due to acquisitions (Granstrand and Sj”olander, 1990; Belderbos, 2001). These studies have not made a distinction between acquisitions of a majority (controlling) stake in foreign firms and minority participations in the equity capital of foreign firms. On the one hand, majority stake acquisitions will give the acquiring firm the leverage to coordinate and integrate the existing R;D operations, and without this the benefits of R;D-intensive acquisitions may be limited. On the other hand, a minority stake may be sufficient to give the acquiring firm access to the technology developed by the foreign firm. For instance, in the electronics and biotechnology industries it is often observed that large firms take minority stakes in a range of highly R;D-intensive pioneering firms to gain preferential access to new technologies (e.g., Wortmann, 1990; Belderbos, 1997). Multiple minority participations in smaller R;D-intensive firms are often seen as investments in options to eventually acquire or get access to the technology in case it has proven to be of commercial value. For these reasons, both minority and majority stake acquisitions may be motivated by technology sourcing and may feature higher R;D intensities.
Although acquired affiliates are expected to have higher R;D intensities at the time of entry, the evolutionary perspective of gradual postentry growth in affiliate capabilities and responsibilities through recurrent investments in R;D is less appropriate for acquired affiliates. The acquired affiliate’s technology base when it comes under foreign control was developed independently and not as a function of the acquiring firm’s organizational learning. The years following the acquisition may involve efforts to integrate R;D operations into the existing network of the parent firm. On the one hand, if the motivation of the acquisition is to access local technology and know-how, the acquiring firm may provide larger financial resources to upgrade the acquired firm’s R&D function. On the other hand, R&D integration could lead to a reorganization of R&D operations involving a reduction in R&D where it duplicates existing work or does not fit into the R&D strategy of the acquiring firm (e.g., Hakanson, 1995; Belderbos, 2001). These contrasting possibilities probably explain the mixed empirical evidence on postacquisition developments in R&D spending. Hitt et al. (1991) found a postacquisition reduction in R&D intensity for merged groups and argued that acquisitions and R&D investment may be seen as substitutes by managers of the acquiring firms. Hall (1988), on the other hand, found no such effect in a large sample of acquisitions in the United States. These considerations lead to the conjecture that there is no systematic positive correlation between affiliate experience and R&D intensity in acquired affiliates. The implication is that over time the initial difference in R&D intensity between acquired and greenfield affiliates is likely to diminish, since we expect from Hypothesis 2 that greenfield affiliates gradually increase the scale of R&D activities.
Hypothesis 5: Acquired affiliates have higher R&D intensities than greenfield affiliates but this difference diminishes with affiliate experience. This holds for both minority stake and majority stake acquisitions.
DATA AND METHODS
Data
We draw our data from the fifth Benchmark Survey on Foreign Direct Investment held in 1993. The Benchmark Survey is an extensive survey among Japanese multinational firms held every 3 years by the Japanese Ministry of International Trade and Industry (MITI, 1994). This official survey is regulated under the Statistics Law of Japan and received a high response rate of 72.3 percent at the affiliate level in 1993. The responses are seen as representative and include large numbers of major multinational firms as well as small and mediumsized firms operating few or just one overseas affiliate. The survey received responses by 846 parent firms active in manufacturing industries, of which 658 firms operated at least one foreign manufacturing affiliate abroad. These 658 firms operated 2008 foreign manufacturing affiliates among them. Affiliates are included in the survey if the Japanese firm owns at least 10 percent of equity.11 R&D expenditures may either be expenditures on basic and applied research or expenditures on development and design of products and processes, but the survey data do not allow decomposition over the different types of R&D activities.12 Although firms are asked to enter a value for R&D expenditures for each affiliate, the response rate for this question is relatively low, reducing the sample of affiliates available for analysis. In addition, for a relatively large number of affiliates no information was available for one of the independent or control variables (primarily parent firm R&D and import and export intensities) reducing the sample for analysis to 420 affiliates owned by 199 parent firms. The total R&D expenditure of the 420 affiliates was 25.6 billion yen (about U.S. $240 million in 1992 exchange rates), out of a total foreign R&D expenditure of 52.9 billion yen recorded in the MITI survey for all manufacturing firms. Of the 420 affiliates, 215 reported zero R&D expenditure.
The affiliates are distributed over industries and regions in a roughly similar manner as the total population of affiliates.13 Table 1 shows the distribution over industries, regions, and mode of entry and ownership. The electronics and electrical engineering industry is divided into the less technology-intensive segments electrical machinery and home electrical appliances, and the more technology-intensive segments computers and telecommunication equipment, audio and video equipment, other electronic equipment, and parts and components. After this subdivision, the largest number of affiliates is present in the transport machinery industry (71), followed by the chemical industry (54), the metal industry (40), and general machinery (40). The average (unweighted) R&D intensity of the 420 affiliates is 0.78 percent. Affiliates’ R;D intensities are highest in the pharmaceutical industry, the rubber and tire industry, and computers and telecommunications, precision machinery, and other electronic equipment.14 The lowest affiliate R;D intensities are found in electrical machinery, building materials, and textiles. The regional distribution of affiliates shows that the largest numbers of affiliates are located in the Newly Industrialized Economies in Asia (NIEs), North America, and the ASEAN countries and China. In Western Europe and other countries fewer than 50 affiliates are located. North American and Western European affiliates have the highest average R;D intensity, at 1.34 and 1.44 percent, respectively. Least R;D intensive are affiliates in ASEAN countries and China (0.19 percent). The distribution of affiliates across entry modes shows that about a third (142) of the affiliates are newly established and wholly owned. Joint ventures are more numerous with 84 affiliates in which the Japanese firms have a majority stake and 101 affiliates in which Japanese firms have a balanced stake or minority ownership.15 Acquired affiliates have higher R;D intensities on average than greenfield affiliates: the average R;D intensity of majority acquired affiliates reaches 2.02 percent, and that of minority or balanced stake acquisitions 0.90 percent. We note, however, that one should be careful in drawing inferences from these unidimensional tabulations. Differences in R;D intensities across countries, for instance, are also determined by differences in industry mix of affiliates, the inclusion of acquisitions, etc. Only with multivariate analysis such as the analysis reported in the following sections can we gain more insights into the various factors affecting R;D intensities. The results will in a number of cases contrast with the simple inferences that one may be inclined to draw from Table 1 on the relationship between affiliate R;D intensity on the one hand, and industry, location, and entry mode and ownership on the other.
Method and definition of variables
Since the dependent variable, affiliate R;D intensity, is censored on the left-hand side at zero, ordinary least squares analysis gives inconsistent estimates (Greene, 1997). We therefore use a censored regression (Tobit) model to relate the R;D intensity of affiliate i, denoted for convenience as R;DINT i , to a vector of regressors xi. We define the underlying regression as:
R;DINT i = a-xi + ai (1a)
where a are the estimated coefficients and ai is a normally distributed error term. We observe:
R;DINT i = 0 I R;DINT i !U 0 (1b)
R;DINT i = R;DINT i I R;DINT i ; 0 (1C)
The estimated coefficients are marginal effects of the underlying regression and larger than the marginal effects of the regressors on the observed variable R;DINT i . To obtain the relevant marginal effects on R;DINT i, we have to multiply a by the probability of noncensoring: a[prob(RDINT i ; 0)].16
We test our hypotheses by including a number of determinants both at the affiliate and the parent firm level, while employing a large set of controls. The determinants and controls described by the vector xi are constructed from the information available in the MITI survey. Table 2 gives a description of the variables with their means and standard deviations and Appendix 1 contains the correlation coefficients between the variables.
Hypothesis 1 is tested by including the variable parent international experience (nonacquisitive firms): the log of the number of years since the Japanese firm established its first manufacturing affiliate abroad, provided that the firm has not chosen an acquisitive entry strategy. An acquisitive firm is defined as a firm that expanded abroad through at least one majority acquisition, while a nonacquisitive firm is defined as a firm that has no majority acquired foreign affiliates. We exclude minority (including balanced stake) acquisitions from the definition of acquisitive firm because the presence of minority stake acquisitions is a much less distinctive indicator of an acquisitive internationalization strategy. Minority stake acquisitions require only limited financial resource commitment and organizational integration and are often used to complement internal growth strategies by serving as options to expand in specific technological areas or markets. In contrast, the financial and organizational resources devoted to acquire and integrate majority stake or full acquisitions imply a major shift in internationalization strategy affecting areas of core technology and capabilities.17 A positive sign is predicted for parent international experience (nonacquisitive firms).18
Hypothesis 2 is tested by including the variable greenfield affiliate experience. This variable is the log of the number of years since the start-up year for greenfield affiliates. Hypothesis 2 suggests a positive sign. Hypothesis 3 is tested by including two variables. A dummy variable, minority joint venture, for minority owned or balanced stake joint ventures: ventures in which a local partner had an equity stake of 50 percent or more in 1993. The reference case for the minority joint venture dummy is wholly owned greenfield affiliate. We include balanced stake joint ventures with the category minority joint ventures because the main characteristic of balanced stake joint ventures remains that the Japanese firms cannot exercise full managerial control. Statistical tests confirmed that the R;D intensity of balanced stake joint ventures was not determined differently from the R;D intensity of minority owned joint ventures.19 A second variable is the interaction term of the Japanese parent firm’s R&D intensity (parent R&D intensity) as a measure of R&D capabilities, and minority joint venture. Hypothesis 3 suggests a positive impact of minority joint venture for parent firms with low R&D intensities and a negative impact for parent firms with high R&D intensities. Hence, a positive sign for minority joint venture is predicted but a negative sign for the interaction term parent firm R&D intensity x minority joint venture. For R&D-intensive parent firms the combined effect of minority joint venture and the interaction term parent firm R&D intensity x minority joint venture is predicted to be negative.
Hypothesis 4 is tested by including the variable parent international experience (acquisitive firms). This variable is the log of the number of years since the Japanese firm established its first manufacturing affiliate abroad, provided that the firm has followed an acquisitive foreign entry strategy as defined above. Hypothesis 4 predicts the absence of a significant effect for this variable. To test Hypothesis 5 we include three variables. A dummy variable, majority acquisition, takes the value 1 if the affiliate came under the control of the Japanese parent through an acquisition and if the affiliate was majority owned in 1993. A second variable minority acquisition takes the value 1 if the Japanese firm acquired a stake in an existing foreign firm but did not have majority control in 1993 (balanced stakes are included). Both majority acquisition and minority acquisition are expected to have a positive sign, but we do not have a strong prior concerning the relative magnitude of the two coefficients. A third variable, acquired affiliate experience, is the log of the number of years since acquisition of the affiliate. Hypothesis 5 predicts a nonsignificant impact of acquired affiliate experience on acquired affiliates’ R;D intensity. If Hypotheses 2 holds as well, this confirms the hypothesized decreasing difference between the R;D intensities of acquired and greenfield affiliates as a function of affiliate experience.
Control variables
We include an elaborate set of control variables of parent, affiliate, locational, and industry characteristics. A number of empirical studies have established a positive correlation between the scale of international manufacturing operations of firms and the extent to which they perform R;D abroad (e.g., Belderbos, 2001; Odagiri and Yasuda, 1996). This is consistent with the notion from Vernon’s product cycle theory that investments in foreign R&D to exploit the firm’s proprietary technologies are a function of the scale of foreign manufacturing operations. The greater the importance of overseas manufacturing and marketing operations, the greater the potential performance effects due to adaptation of products and process technologies. We include the explanatory variable parent scale of overseas manufacturing, the value of assets of all foreign manufacturing affiliates operated by the firm as a percentage of the assets of the firm in Japan, to control for this influence. The R;D intensity of the investing parent firm is likely to impact on the R;D intensity of foreign affiliates. If a parent firm produces technologically complex products and relies on R;D to gain a competitive advantage, this will also be reflected in the R;D intensity of affiliate operations. Belderbos (1997) and Zejan (1990) found a positive correlation between affiliate and parent firm R;D intensities, and Odagiri and Yasuda (1996) found such a relationship at the industry level. We include the variable parent R;D intensity. A third parent-related characteristic is the number of other affiliates owned by the parent firm in the same country. Frost (2001) argues that larger operations in a country provide greater legitimacy to the investing firm in local business communities and makes it easier to gain access to local technological knowledge-sharing networks and industry associations. This may promote further investments in technology scanning and explorative research. Larger operations may also provide the necessary scale to warrant more substantial investments in R;D capabilities to improve products and processes in local operations. On the other hand, if the Japanese parent firm follows a country-based strategy that involves a concentration of R;D in specifically designated affiliates (e.g., following the rule ‘one country one R&D site’), the presence of other affiliates may reduce the likelihood that the affiliate in question is assigned R;D responsibilities.20 We control for these potential effects of parent firms’ activities at the country level by including the variable other country affiliates, measuring the number of other manufacturing affiliates of the parent firm in the same country.
At the affiliate level, previous research has found a positive effect of the export intensity of affiliates on the scale of R&D activities (Hewitt, 1980; Zejan, 1990). High export intensities may imply that the affiliates have responsibility for regional or world markets in a product area rather than a more limited responsibility for the local market. Such affiliates may be more likely to possess substantial R&D expertise that allows them to function as ‘international adaptor’ sites or even as ‘international creator’ sites (Hakanson and Nobel, 1993b; Nobel and Birkinshaw, 1998). We include the export intensity of the affiliate as a control variable. We also include a dummy variable for majority joint ventures to control for potential differences in R&D intensities between such joint ventures and wholly owned ventures. Another affiliate-level control variable is affiliate size. A certain level of sales may be required to justify investments in development for the local market, and scale economies in research may favor R&D in larger affiliates. Larger affiliates may also be better able to bargain with headquarters about allocation of R&D resources. However, the evidence on the effect of firm size on affiliate R&D intensity is mixed. Kumar (1996) and Zejan (1990) found weakly positive effects of affiliate size on R&D intensity, but results in Odagiri and Yasuda (1996) and Belderbos (1997) suggested that R&D expenditures only increase proportionally with affiliate sales. The industrial organization literature on the relationship between firm size and R&D investment has also by and large failed to find evidence that larger firms are more R&D intensive (Cohen and Klepper, 1996). In order to control for possible scale effects, we include the variable affiliate size (the value of affiliate sales). Another characteristic of affiliates that may affect their R&D intensity is the reliance on imported components and materials as well as imported final goods. Affiliates that have a more accentuated role as a distributor and assembler rather than as integrated manufacturer will exhibit higher levels of intermediates and final goods import. These affiliates are also less likely to engage in substantial (adaptive) R&D. Some previous empirical studies have found a negative correlation between the reliance on imports and affiliate R&D (Odagiri and Yasuda, 1996; Zejan, 1990). We control for possible differences in the weight of the affiliates’ manufacturing and distribution roles by including import intensity, the ratio of total imports to sales of the affiliate.
Finally, we include a set of controls for industry and location. Various characteristics related to the industry affect the need to perform R;D abroad: the maturity of products, the degree of product differentiation and international differences in consumer tastes, the degree of reliance on science and basic research, the existence of multiple centers of excellence and the scope for technological learning abroad, and government regulations which vary by industry (for instance, regulations on clinical trials commonly in place in the pharmaceutical industry). Locational characteristics such as the size of the market, specific regulations requiring product adaptation, the available technological capabilities and excellence of the science base, and the degree of intellectual property rights protection available impact on the R;D intensity of affiliates (e.g., Fors, 1998; Kumar, 1996). Not all these influences are easily quantifiable and controlled for in a set of independent variables. Since the focus of the present study is on parent and affiliate characteristics, we control for industry and locational factors by including a set of industry and country dummies. We distinguish 17 industries and 20 countries or regions. For all countries in which more than four affiliates were present a separate dummy was included.
RESULTS
Hypotheses
The results of the Tobit model explaining the R;D intensity of foreign manufacturing affiliates are presented in Table 3. Parent international experience has the predicted positive sign for nonacquisitive firms as predicted by Hypothesis 1, and the coefficient is significant at the 10 percent level.21 The calculated marginal effect of international experience implies that 8 more years of experience leads to a 0.25 percent point increase in affiliate R;D intensity.22 This is a relatively strong impact considering the average affiliate R;D intensity of 0.79 percent. The results support Hypothesis 2: greenfield affiliate experience has a positive and significant (10% level) effect on the R;D intensity of the affiliates. The calculated marginal effect of greenfield experience implies that 9 more years of affiliate experience leads to a 0.20 percent points higher affiliate R;D intensity. The coefficient of the minority joint venture dummy is positive and highly significant, while the interaction term of parent firm R;D intensity and minority joint venture is negative and significant. These two results confirm Hypothesis 3: minority joint ventures show high R;D intensities compared with wholly owned greenfield affiliates if the Japanese parent firm has weak R;D capabilities, but lower R;D intensities if the Japanese firm possesses strong R;D capabilities. This pattern is illustrated in Figure 1, which depicts affiliate R;D intensity as a function of parent R;D intensity for wholly owned affiliates and minority owned joint ventures. The estimated coefficients imply that minority owned joint venture affiliates have a lower R;D intensity than wholly owned greenfield affiliates in case parent firm R;D intensity exceeds 8.0 percent.
Support is also found for the presence of a contrasting set of determinants of R;D intensities in case of entry through acquisition. The coefficient of parent international experience for acquisitive firms is not significantly different from zero in accordance with Hypothesis 4. The dummies for both majority stake acquisitions and minority stake acquisitions are positive and significant. The size of the coefficients suggests an important role for acquisitions in the expansion of Japanese R;D abroad. The marginal effects imply that an acquired affiliate has a 1.5 (majority stake acquisition) or 1.2 (nonmajority stake acquisition) percent points higher R;D intensity than a wholly owned greenfield affiliate. At the same time, acquired affiliate experience, in contrast with greenfield affiliate experience, has no significant impact on affiliate R;D intensity. These findings confirm the prediction of Hypothesis 5 that acquisitions show higher R;D intensities than greenfield affiliates, but that this difference diminishes with experience. The magnitude of this decrease over time is, however, relative small. The estimated coefficients imply that it would take a wholly owned greenfield affiliate more than 65 years to reach the same level of R;D intensity as majority acquired affiliates. If one also takes into account the additional positive effect on affiliate R;D of parent international experience for nonacquisitive firms, then roughly 30 years of parent international experience and greenfield affiliate experience would offset the acquisition effect. This order of magnitude of differences between acquisitions and greenfield entries is consistent with the notion that access to the R;D capabilities of foreign firms has been an important motivation for Japanese firms’ acquisitions.
Control variables
Among the control variables, the scale of the parent’s foreign manufacturing operations has a significantly positive impact, suggesting that firms for which overseas manufacturing and overseas markets are more important commit more R;D resources abroad. Parent firm R;D also has a positive and significant effect on the R;D intensity of foreign affiliates. The marginal effect of parent firm R;D does not indicate that this effect is very large though: a 1 percent point increase in the parent’s R&D ratio leads to a 0.08 percent point increase in affiliate R&D intensity. Export intensity of the affiliate has a positive and significant impact, which may suggest that export-oriented affiliates are assigned greater R&D responsibilities including development of products for export markets. The coefficient for majority owned joint ventures is positive but not significant, suggesting that these affiliates show no systematically different R&D intensities compared with the reference group, wholly owned greenfield affiliates. Affiliate size has a negative but insignificant effect such that the suggestion that R&D expenditures rise proportionally with affiliate sales cannot be rejected. This result is in accordance with previous findings on the relationship between firm size and R&D investment (Cohen and Klepper, 1996). The import intensity of the affiliate neither has a significant effect on R&D. The number of other affiliates operated by the parent firm in the country has a negative sign, consistent with the notion of the assignment of R&D tasks to one or few affiliates in a country, but the coefficient is insignificant. The industry dummies show that affiliate R&D intensity is significantly higher in chemicals and communication and computers, compared to the textile industry (the omitted dummy). The country dummies explain a substantial additional share of the variation in R&D intensity across affiliates. With Thailand as the reference group, the highest significant coefficients are estimated for the United Kingdom, North America, the Netherlands, and other Western European countries, followed by Brazil and Taiwan. The lowest coefficients are for China, and, perhaps surprisingly, Hong Kong.
Alternative specifications
We estimated the model under a number of alternative specifications to examine the robustness of the empirical results. Given the relatively large number of affiliates that did not engage in R&D, the question arises whether affiliates that a
