Multinational corporations (MNCs) have for a long time been conceptualized as organizations whose advantage is derived from their ability to acquire and utilize knowledge across borders (Almeida, Song, & Grant, 2002; Gupta & Govindarajan, 2000; Kogut & Zander, 1993; Mudambi, 2002) and who are superior to alternative organizational configurations in terms of transferring knowledge (Feinberg & Gupta, 2004; Hansen & Løvås, 2004; Kogut & Zander, 1992, 1993). While earlier studies have primarily focused on headquarters (HQ) as the focal unit of analysis, over the last few years much of the emphasis has shifted to analyzing the subsidiary, both as a recipient of knowledge from HQ and other subsidiaries and as a source of knowledge for HQ and peer subsidiaries. We refer to research on knowledge flows within MNCs where the subsidiary is positioned as the focal unit of analysis as "research on subsidiary knowledge flows in MNCs".
Research on knowledge flows in MNCs has grown considerably over the last decade or so, it has become more "fluid", prone to multiple definitions and measures of the same notions and constructs and exposed to contrary findings and conclusions. While studies of different aspects might be a natural feature of the development of any stream of literature, there is a potential danger in focusing on diversity to the detriment of building on existing and cumulative knowledge. We have reached a stage in theory development and testing that calls for a systematic classification. In order to understand discrepancies in existing explanations and analyses and to make sense of divergent (sometimes contrary) findings, mapping the existing literature on subsidiary knowledge flows within MNCs is a necessary and useful exercise. We analyze what has been achieved so far, identify existing gaps and suggest future research opportunities.
In order to present a comprehensive yet systematic review of the research on subsidiary knowledge flows in MNCs, we have narrowed the scope of our analysis in the following ways. First, while we acknowledge that there are emerging studies on knowledge transfers from MNC subsidiaries to external firms and from host countries to subsidiaries, we only review research on knowledge flows within the MNC. Second, we review research published only in top Management and in International Business journals. Top Management journals apply strict quality selection criteria and provide good examples for future scholars. International Business journals are important for our review as we specifically focus on MNCs. Third, we only review recent research, from 1996, when a few highly influential studies on knowledge and knowledge processes were published (e.g., Grant, 1996; Spender, 1996; Spender & Grant, 1996; Szulanski, 1996), to 2008 inclusive, and divide the 13 year period into two sub-periods, from 1996 to 2001 and from 2002 to 2008.
The rest of the paper is organized as follows. In the following section we describe the criteria applied in selecting the journals and identifying the articles that comprise our review. Then we discuss how the literature has evolved over the last 13 years. We then present our framework, which organizes the literature according to knowledge inflows and outflows, and according to four clusters of themes: knowledge characteristics, characteristics of actors involved in intra-MNC knowledge flows, characteristics of relationships between actors and outcomes of knowledge flows. Subsequently we present and discuss our findings in terms of the key accomplishments and the gaps we see in the existing literature on subsidiary knowledge flows. We conclude by commenting on the topics of examination and the theoretical and methodological underpinnings that have dominated research so far and suggest recommendations for future research.
Journal Selection, Article Identification and Temporal Trends in the Literature
Our starting point in selecting the top-tier management journals were the lists comprised by Gomez-Mejia and Balkin (1992), Tahai and Meyer (1999), Werner (2002) and Podsakoff, Mackenzie, Bachrach, and Podsakoff (2005). These lists are comprehensive and have been consistently utilized and cited in subsequent reviews (Kirkman, Lowe, & Gibson, 2006; Tsui, Nifadkar, & Ou, 2007). Our initial list included all journals listed in the above rankings and contained 32 journals. Because we only review academic research, we excluded California Management Review, Harvard Business Review, Long Range Planning, Organizational Dynamics, and Sloan Management Review. Besides Journal of International Business Studies, typically included among the top management journals, we also added the other main journals specializing in international business - International Business Review, International Studies of Management and Organization, Journal of International Management, Journal of World Business, and Management International Review.
To identify the articles dealing with subsidiary knowledge flows in MNCs, we conducted computer-assisted (ProQuest, ABI/Inform, ScienceDirect, and OVID) searches for "knowledge" in conjunction with "subsidiary" and "knowledge" in conjunction with "MNC" in titles and abstracts of papers published from January 1996 to December 2008 inclusive. We then extended the search to also include the term "subunit" as a number of papers use this term instead of "subsidiary". Following our objective to study knowledge flows, we searched, both electronically and manually, all identified articles to select those that have used at least one of following terms: knowledge "flows", "transfer", "sharing", "exchange", or "involvement". These are terms used in different articles or interchangeably in the same article to describe knowledge flows. As a subsequent step, we conducted manual search of articles in journals not available electronically over the entire search period. Finally, we checked the references of some of the latest articles on knowledge in MNCs to ensure our search was as complete as possible. The final result is 82 articles (listed in the Appendix) which deal with knowledge flows in MNCs and which we analyze in the next sections. Table 1 lists the number and percentage of articles by journal.
-----------------------------
Insert Table 1 about here
------------------------------
As evident from Table 1, 53 articles (65%) are published in International Business journals, with Management International Review, International Business Review and Journal of International Business Studies having published 10 or more articles each. The other 29 articles (35%) are published in management journals, with Academy of Management Journal, Organization Science and Strategic Management Journal having published 17 of the articles.
------------------------------
Insert Figure 1 about here
-------------------------------
As illustrated in Figure 1, research on subsidiary knowledge flows has been increasing steadily from 1996 to 2008, with an average increase of 25% in the first period and an average increase of 40% in the second period, with a clear boost from 2002 and with only minor short-term fluctuations in the two sub-periods. The sharp increase of publications in 2002 is probably a result of the seminal article by Gupta and Govindarajan (2000) on subsidiary knowledge flows. The temporal trends within the second period are slightly more volatile compared to the first period, with an initial decrease from 2002 to 2003 before a steady increase and peak of 12 articles in 2005. A subsequent slight decrease and stability for two years was followed by an increase of published research again in 2008.
ORGANIZING FRAMEWORK
We organize the literature in two main clusters: studies that examine knowledge inflows and studies that focus on knowledge outflows. We group existing research as belonging to the literature on knowledge inflows to MNC subsidiaries as long as the measures used suggest that MNC subsidiaries are the recipient of knowledge. We then categorize the articles within this cluster in two sub-clusters according to the source of knowledge - knowledge inflows from HQ (vertical inflows) and knowledge inflows from peer subsidiaries within the same MNC (horizontal inflows). We group the articles that treat MNC subsidiaries as the source of knowledge/technology/organizational practices as belonging to the literature on knowledge outflows from MNC subsidiaries. The articles within this group are also organized in two sub-groups depending on whether the recipient of subsidiaries' knowledge is the HQ (vertical outflows) or other subsidiaries within the MNC (horizontal outflows). A few studies (Buckley & Carter, 2002; Holtbrügge & Berg, 2004; Phene & Almeida, 2003) simultaneously examine knowledge in- and outflows so, they appear both under inflows and outflows in our classification. To reiterate, we examine how the four directions of knowledge flows in MNCs - vertical and horizontal inflows and vertical and horizontal outflows - have been treated in the literature.
We organize the articles in relation to four topics: a) knowledge characteristics, b) characteristics of actors involved in the intra-MNC knowledge flows, c) features of the relationships between the actors involved in the knowledge flows and d) outcomes of knowledge flows. This is similar to the way Argote, McEvily, and Reagans (2003) organized the research published on organizational learning and knowledge management: these authors classified the literature according to whether it addressed properties of knowledge, properties units, properties of the relationships between units, or knowledge management outcomes. Szulanski (1996) also used the first three of the above clusters.
RESULTS AND DISCUSSION
The results of our review are presented in Table 2. Before we discuss them, a few important remarks need to be made. First of all, several studies refer to flows from and to "other units in the MNC" or from and to "the rest of the MNC" without specifying whether they include HQ, subsidiaries or both (e.g., Björkman, Barner-Rasmussen, & Li, 2004; Holtbrügge & Berg, 2004; Minbaeva, Pedersen, Björkman, Fey, & Park, 2003). In such cases we list these studies as examining both horizontal and vertical knowledge flows. Second, while some research solely focuses on knowledge inflows (e.g. Schulz, 2003) or on knowledge outflows (e.g., Schulz, 2001), other studies do not specify the direction of the flow. In the latter cases we carefully examined the studies and scrutinized the measures to determine the direction of flows. Finally, when studies examine knowledge flows within the MNC, as well as with external actors in the host country (Almeida et al., 2002; Holtbrügge & Berg, 2004; Yamin & Otto, 2004; Almeida & Phene, 2004), we only include the examination of knowledge flows within MNCs.
------------------------------
Insert Table 2 about here
------------------------------
Eight of the studies which satisfied our selection criteria for inclusion in the review are not included in Table 2. Some of them are studies which use subsidiary knowledge flows to categorize subsidiaries' roles (Manea & Pearce, 2006; Manolopolous et al., 2007; Papanastassiou & Pearce, 1997) and network types (Tseng, Yu & Seetoo, 2002). We also excluded articles which functioned as overviews: a research agenda (Zander and Solvell, 2000), an introduction to a special issue, (Mudambi, 2002) and a review paper (Werner, 2002). Finally, we excluded Denrell, Arvidsson and Zander (2004) as their study examines the reliability of capability evaluations in MNC subsidiaries for effective knowledge management.
Characteristics of Knowledge
Compared to the other three clusters of themes, knowledge characteristics / types / dimensions have been the least studied in the literature on subsidiary knowledge flows. However, several characteristics have been considered, of which the tacit-explicit continuum is the most extensively examined. Variables closely related to the tacit-explicit continuum include drawability and demonstrability (Sunaoshi et al., 2005), knowledge complexity (Minbaeva, 2007; Riusala & Smale, 2007), knowledge non-specificity and non-availability (Minbaeva, 2007), teachability ( Riusala & Smale, 2007), and articulability and observability (Håkanson & Nobel, 2000). Our reading of the literature suggests that the tacit-explicit continuum has been unnecessarily re-labeled in various ways, whereas in essence the term denotes the extent to which the knowledge being transferred is codified into easier-to-retrieve and access forms. Knowledge codifiability has received particularly close examination, appearing in more than one study (Hansen, 1999; Hansen, 2002; Schulz, 2003; Riusala & Smale, 2007). It makes sense to focus on this variable when examining effects of the explicit-tacit continuum on other variables and relationships among those instead of suggesting "new" labels that, in fact, refer to the same dimension. There is a danger in inventing new terms which do not provide new insights and actually prevent the building up of knowledge on the basis of well established and recognized dimensional characteristics and notions.
Some of the literature is concerned with types of knowledge, rather than its internal dimensions. Technological knowledge is the most widely examined, although social knowledge (Buckley, Carter, Clegg, & Tan, 2005) and marketing knowledge (Papanastassiou & Pearce, 1997; Schlegelmilch & Chini, 2003; Holm & Sharma, 2006) have also been studied. We observe that when scholars study more than a single type of knowledge, for instance, how the different types are transferred in the MNC, these typologies and categorizations are not unified. Bhagat, Kedia, Harveston, and Triandis (2002), for example, distinguish between human, structured and social knowledge while Lee and Macmillan (2008) classify knowledge types into procedural and coordinative. Often the distinction between knowledge types and knowledge dimensions is blurred and some authors refer to types of knowledge when they, in fact, discuss dimensions (e.g., Almeida et al. (2002) refer to two 'types' of knowledge - codified versus experience-based tacit knowledge). We also find that the subsidiary knowledge flow literature is more relaxed in its treatment of the notion of knowledge types by including, for instance, human resource practices (Sparkes & Miyake, 2000) and organizational learning systems (Hong, Easterby-Smith, & Snell, 2006). We would rather treat these as mechanisms for achieving certain outcomes from managing specific knowledge processes, and not as knowledge types themselves.
In general, the literature has provided sufficient support for the fact that knowledge characteristics influence subsidiary knowledge flows in one way or another. We are in doubt, however, whether much more can be learnt about these flows by examining only knowledge characteristics and their influence on the occurrence and/or effectiveness of subsidiary knowledge flows. We argue, rather, that the field is in need of further examination about how knowledge characteristics influence relationships between actors engaged in knowledge flows, the interface between knowledge characteristics and relationships between the source and the recipient of knowledge and how such interface influences knowledge flows. There is an identifiable, albeit relatively weak, trend in the literature suggesting that one can gain important insights into subsidiary knowledge flows by integrating knowledge characteristics with other knowledge flow characteristics. Only a few studies - Hansen (1999), Hansen (2002), Levin and Cross (2004) and Szulanski et al. (2004) - have examined such combinations and we see potential in pursuing this line of inquiry. Hansen (1999) examined tie strength and its influence on the transfer of codified and non-codified knowledge and found that weak ties are especially strong in influencing the transfer of non-codified knowledge while strong ties positively influence the transfer of codified knowledge. Hansen (2002) found that project completion time is shorter when a team has higher number of direct relations and knowledge is codified. Levin and Cross (2004) concluded that competence-based trust is more important than other types of trust in the receipt of useful knowledge when knowledge is tacit rather than explicit. Similarly, Szulanski, Cappetta and Jensen (2004) examined and found support for the idea that causal ambiguity has a moderating effect on perceived trustworthiness and accuracy of knowledge transfer.
Knowledge characteristics have been examined more in relation to knowledge inflows (18 studies) than to knowledge outflows (13 studies). Furthermore, with the exception of those which adopt a dyadic level of analysis (Jensen & Szulanski, 2004; Szulanski et al., 2004; Szulanski, 1996), only four studies (Hansen, 2002; Levin & Cross, 2004; Lee & Macmillan, 2008; Pérez-Nordtvedt et al., 2008) focus on horizontal flows. The apparent lack of focus on knowledge characteristics with regard to knowledge outflows on the one hand and to horizontal flows on the other hand might be due to the traditional perception (and examination) of subsidiaries as subordinate and inferior to HQ within the MNC network. Although subsidiaries have been also regarded as Global Innovators (Gupta & Govindarajan, 1991, 1994; Harzing & Noorderhaven, 2006), International Creators (Nobel & Birkinshaw, 1998), Centers of Excellence (Andersson & Forsgren, 2000; Frost, Birkinshaw & Ensign, 2002), and possessing World Mandates (Birkinshaw & Morrison, 1995), the subsidiary knowledge flows literature still treats subsidiaries predominantly as recipients of knowledge from HQ, hence the bias towards examining primarily vertical knowledge inflows.
The period from 1996 to 2001 witnessed the treatment of some knowledge characteristics as moderating and interacting variables (Hansen, 2002; Levin & Cross, 2004; Szulanski et al., 2004). In the period after 2001, knowledge characteristics have been considered primarily as control variables (Björkman et al., 2004; Almeida & Phene, 2004; Persson, 2006; Monteiro, Arvidsson & Birkinshaw, 2008). This development indicates the importance of understanding the nature of knowledge characteristics in defining issues such as actors' characteristics and characteristics of the relationships between actors in determining MNC knowledge flows. We see a need for future studies which examine subsidiary knowledge flows in relation to actors' characteristics and also in relation to the characteristics of relationships between actors. Such studies should control for knowledge characteristics in order to accurately determine their effects on subsidiary knowledge flows.
Characteristics of Actors Involved in MNC Knowledge Flows
Actors' characteristics constitute the most examined theme in research on subsidiary knowledge flows, and the number of studies dealing with this theme has grown significantly in relation to both inflows and outflows. 51 studies address characteristics of actors: 35 in relation to knowledge inflows and 27 in relation to knowledge outflows. We observe an increase from six studies (prior to 2001) to 34 studies (after 2001) in relation to inflows and from four to 23 over the same period in relation to outflows. It is worth noting that only eight studies have examined both vertical and horizontal in- and outflows.
Several characteristics of actors have been examined which is a probable reflection of the fact that various actors can be engaged in knowledge flows (e.g. individuals, the subsidiary, the knowledge transfer network). The majority of studies examine characteristics of subsidiaries such as organizational culture, motivation to acquire knowledge, absorptive capacity, disseminative capacity, strategic importance, autonomy, size, and ownership structure. Characteristics of individual actors include their linguistic competence, career considerations, and (un)willingness to question rules. Characteristics of the networks to which subsidiaries belong (e.g. network size, strength, and path length) have also been analyzed. Our reading of the literature suggests that the emphasis on examining subsidiary characteristics occurs at the expense of network- and individual level studies. Given the importance social network theories have gained in the subsidiary knowledge flows literature (please refer to Table 1), it is strange that we have as yet seen no study of subsidiary knowledge flows at the network/group level with regard to their characteristics.
Among the various characteristics of actors, ownership structure (Buckley et al., 2005; Zhao & Luo, 2005; Cho & Lee, 2004; Holtbrügge & Berg, 2004), trustworthiness (Jensen & Szulanski, 2004; Levin & Cross, 2004; Szulanski et al., 2004) and motivation (Gupta & Govindarajan, 2000; Jensen & Szulanski, 2004) have attracted research attention whereas other characteristics (subsidiary autonomy, organizational culture) have remained peripheral. The most examined characteristic of actors is absorptive capacity. Nine studies have explicitly examined it: Szulanski (1996), Gupta and Govindarajan (2000), Tsai (2001), Minbaeva et al. (2003), Mahnke, Pedersen and Venzin (2005), Ambos, Ambos, and Schlegelmilch (2006), Madhok and Liu (2006), Minbaeva (2007) and Riusala and Smale (2007). While all nine studies agree on the definition of absorptive capacity, they measure it in nine different ways. We cannot find a meaningful explanation of why authors opt for different measures, but we emphasize that this is not a fruitful way to accumulate valuable knowledge on subsidiary knowledge flows in general and on absorptive capacity in particular.
Characteristics of Relationships between Actors
Relationships between actors involved in knowledge inflows have been extensively examined, with an increase from just four studies prior to 2001 to 30 studies after 2001. We observe a similar trend in relation to knowledge outflows: four studies in the first period and 23 in the second period. We should note that all studies on the characteristics of the relationship between actors that examine vertical and horizontal knowledge outflows also examine vertical and horizontal knowledge inflows. This signals an increasing awareness of the importance of the quality and nature of the relationships between actors in knowledge flow processes.
The literature has recently shifted its attention to view the transfer process between transferor and transferee and to examine the nature of the relationship between HQ and subsidiaries, as well as between subsidiaries themselves. Among different characteristics, trust is the most widely examined, with six studies focusing on it (Tsai & Ghoshal, 1998; Kostova & Roth, 2002; Buckley et al., 2005; Inkpen & Tsang, 2005; Li, 2005; Riusala & Smale, 2007). Even though Tsai and Ghoshal first examined trust in 1998 as a characteristic of the relationship between actors in knowledge flows, the subsequent five studies that have included trust only occur after 2001. Among the characteristics of vertical relationships are dependence on HQ (Kostova, 1999; Kostova & Roth, 2002), organizational identification with HQ (Kostova, 1999; Kostova & Roth, 2002; Vora & Kostova, 2007), and HQ's level of internationalization experience, host country experience, and marketing knowledge (Fang, Wade & Beamish, 2007). Among the characteristics of subsidiaries' horizontal relationships are inter-unit network position (Tsai, 2001), network size, strengths and length (Hansen, 2002; Hansen, Mors & Løvas, 2005) and the frequency and simplicity of interactions and communications (Adenfelt & Lagerström, 2008; Monteiro et al., 2008).
Studies that examine characteristics of relationships between actors focus primarily on knowledge inflows. We also observe a particular trend: studies that investigate knowledge inflows tend to focus on vertical flows while those that are concerned with knowledge outflows primarily study horizontal flows. Such a trend is possibly associated with the increasing attention towards viewing the MNC as an inter-organizational network (Ghoshal & Bartlett, 1990). We also notice that scholars tend to focus more on inter-subsidiary or inter-unit networks within the MNC, rather than including the HQ in their analyses, a practice which, we would argue, is incomplete and undesirable given the importance of HQ and of its roles and functions within the MNC.
Another characteristic of the relationship between transferor and transferee in the knowledge transfer process are distance-related variables - institutional distance, cultural distance and organizational distance - studied in relation to both vertical and horizontal knowledge flows. Kostova and Roth (2002) and Jensen and Szulanski (2004) examined institutional distance between subsidiaries and HQ, and between the focal subsidiary and peer subsidiaries, respectively. Jensen and Szulanski (2004) use Kogut and Singh's (1988) cultural distance index to measure institutional distance. This is confusing and misleading as other scholars use this well-regarded and highly criticized index to measure cultural distance. Two studies examine cultural distance, one conceptual (Schlegelmilch & Chini, 2003) and one empirical (Ambos et al., 2006); yet Ambos et al. (2006) combine Kogut and Singh's (1988) index with their own perceived measure of cultural distance. Organizational distance (Schlegelmilch & Chini, 2003; Ambos et al., 2006), technological distance (Zhao & Luo, 2005) and social distance (Engelhard & Nägele, 2003) have also been examined, but less frequently, as characteristics of the relationships between actors involved in subsidiary knowledge flows.
Outcomes of Knowledge Flows
As is the case with the literature addressing relationships between actors in knowledge flows, the studies that examine outcomes of knowledge inflows are unequally distributed in time. Prior to 2001, only seven studies were published in relation to knowledge inflows, and another seven on outcomes from knowledge outflows; whereas after 2001, there were 47 papers written on inflows and 37 on outflows. As with the other categories in our framework, more studies are devoted to knowledge inflows as compared to knowledge outflows.
The very occurrence of subsidiary knowledge flows has been treated as an outcome of knowledge flows. This initial trend extended to studying the frequency, diversity and scope of knowledge sharing within the MNC. Subsidiary knowledge flows have also been used to categorize subsidiaries according to roles (Harzing & Noorderhaven, 2006; Manolopolous et al., 2005, 2007; Manea & Pearce, 2006), subsidiary network types (Tseng et al., 2002), and parent-subsidiary links (Luo, 2003). Only a few studies have extended the occurrence of knowledge in- or outflows to other phenomena in the subsidiary or the MNC. While most outcomes of subsidiary knowledge inflows have been related to subsidiary innovation and performance (Tsai, 2001; Luo, 2003; Almeida & Phene, 2004; Mahnke et al., 2005; Fang et al., 2007; Lee & Macmillan, 2008), outcomes of outflows have been restricted to the occurrence, frequency, speed, scope, diversity and effectiveness of the flow.
VENUES FOR FUTURE RESEARCH
We have outlined several possibilities for future research when presenting and discussing the results of our review of the literature on subsidiary knowledge flows. We have outlined not only what has been achieved in terms of examining knowledge characteristics, characteristics of actors, characteristics of relationships between actors engaged in knowledge flows and outcomes of these flows, but have also argued how we think future research should develop. In the next sections we outline future research avenues that go beyond the single four categories that we have discussed.
Topics of Study and Flow Directions
Subsidiary knowledge inflows have received more research attention compared to knowledge outflows. The trend prevails in all categories of our organizing framework suggesting that subsidiaries as sources of knowledge for the entire MNC is still understudied and less understood. Subsidiaries have traditionally been perceived as receivers of knowledge and theorized about as inferior to HQ on several dimensions, including the stock and value of knowledge they possess. Only recently have subsidiaries begun to be regarded as contributors of knowledge for HQ and peer subsidiaries and as valuable and powerful players in the MNC network.
We argue that the literature can absorb more insights on subsidiary knowledge outflows, particularly on outcomes of these outflows. We recommend, however, that future research examines outcomes of subsidiary knowledge flows beyond categorizations and typologies of subsidiaries and beyond treating the occurrence of knowledge flows or knowledge sharing as an outcome of knowledge flows. Instead, the literature needs insights into how peer subsidiaries and HQ can benefit from subsidiary knowledge outflows, vertical as well as horizontal.
It is striking and far from satisfying that many of the studies we have reviewed treat vertical and horizontal flows in the same way or are ambiguous regarding the direction of the flow using terms such as "other units in the MNC" or "the rest of the MNC". Six out of 14 conceptual papers and 20 out of 68 empirical papers do not distinguish between vertical and horizontal flows and do not even acknowledge that flows can take place in different directions. This prevents researchers from accumulating findings that can assist in understanding the nature of and the mechanisms behind the different flows and, consequently, how these can be managed. We advocate against continuing this trend and suggest that future research on subsidiary knowledge flows is specific and explicit in terms of whether it examines vertical or horizontal flows. While both these directions generate significant advantages for the focal subsidiary as well as for the MNC, they also differ significantly (Schulz, 2001).
Theoretical Underpinnings
Not surprisingly, the theory most commonly utilized in studies on subsidiary knowledge flows is knowledge based theory. 25 articles have used this theory. Organizational learning theory and network and social network theories also have a prominent position appearing in 12 studies each. 10 studies, all published after 2001, have been conducted by viewing the MNC as a (differentiated) network, a concept first proposed by Bartlett and Ghoshal (1989). It is worth noticing that several of these studies have interpreted the term differently, from differentiated activities within the MNC network (the original idea behind the concept) to the very network itself being seen as differentiated. Social capital theory, first utilized by Tsai and Ghoshal (1998) to examine intra-MNC knowledge transfer, has only recently (after 2003) been actively integrated into the literature on subsidiary knowledge flows, appearing in seven studies. The resource based view is utilized in six papers while (co-)evolutionary theory of the MNC and institutional theory each form the theoretical basis of five studies. Organizational behavior based concepts (e.g. intra-firm politics, organizational identification, organizational communication), human resource management based concepts (e.g., human capital, top management teams, expatriation) and culture and language each appear in four papers.
Borrowing theories from related and more distant disciplines and bodies of knowledge can be fruitful in advancing the literature on subsidiary knowledge flows. The examination of the MNC has historically adopted insights from industrial organization theory, transaction cost theory, sociology, evolutionary theory of the firm, contingency theory, business network theory and institutional theory, to mention but a few. They, and many others, have provided powerful lenses to examine MNC related processes and phenomena. While some of these theories will, most probably, continue providing a useful theoretical platform for analyzing subsidiary knowledge flows, some less utilized ones (e.g., political science, psychology, economic geography, comparative sociology) are promising in terms of entering the theoretical terrain in the coming years.
Methods
By far most of the papers dealing with subsidiary knowledge flows in MNCs are empirical papers. Only 14 articles (17 % of the articles in our sample) are conceptual. Among those are an introduction to a special issue (Mudambi, 2002), a literature review paper (Werner, 2002) and a research agenda (Zander & Solvell, 2000). Excluding these three papers brings us to 11 papers published in a period of 13 years - somewhat humble progress in terms of theory development. The vast majority of these papers, however, are published after 2001which indicates movement in the right direction.
It is puzzling that only one conceptual study of the 14 included in our review, Kostova (1999), has been (partially) tested empirically later on, by Kostova and Roth (2002). Why is it that no more conceptual papers have attracted researchers to conduct empirical tests? We speculate that because some conceptual papers propose constructs and relationships at multiple levels of analysis (e.g., Bhagat et al., 2002), testing them empirically is more demanding. While we acknowledge difficulties and challenges in conducting multi-level research, we encourage future research to engage in such type of research - an issue which we return to in the next section.
As with International Business literature in general, the literature on subsidiary knowledge flows has extensively applied quantitative methods of inquiry. Of all 68 empirical papers in our review, one is based on mathematical modelling, 58 employ quantitative methods and only nine utilize qualitative methods, opting primarily for case studies. The majority of the qualitative studies focus on subsidiaries in China (Buckley, Clegg & Tan, 2006; Wang, Tong & Koh, 2004; Hong et al., 2006; Buckley et al., 2005), and other emerging countries like Russia (Engelhard & Nägele, 2003), Brazil and Mexico (Sparkes & Miyake, 2000). The Management journals are particularly focused on quantitative examinations of subsidiary knowledge flows, with only Journal of Management Studies having published qualitative papers. International Business journals prove to be more receptive towards qualitative studies on the topic, with Journal of World Business taking the lead in this area.
We argue that the literature is disadvantaged by not adopting qualitative methods to reach a deeper and more nuanced understanding and examination of subsidiary knowledge flows. If the trend of conducting solely quantitative studies continues, it is questionable as to how much more in-depth knowledge we can generate about the phenomenon. As a growing stream of research adopts a more sociological perspective of the MNC, conceptualizing it and empirically examining it in a way that emphasizes the role of individuals, the field can only benefit from deeper and contextualized qualitative accounts of subsidiary knowledge flows in the MNC.
Units and Levels of Analysis
The quantitative studies in our review adopt various units of analysis: transfers or dyads (Szulanski, 1996; Jensen & Szulanski, 2004; Szulanski et al., 2004; Barner-Rasmussen & Björkman, 2005; Ambos et al., 2006), teams or project members within subsidiaries (e.g., Zellmer-Bruhn & Gibson, 2006; Hansen et al., 2005) or patent citations (Almeida & Phene, 2004; Yamin & Otto, 2004; Frost & Zhou, 2005; Phene, Madhok & Liu, 2005). In a few instances, the same data has been utilized for the purpose of more than one article (Szulanski, 1996 and Jensen & Szulanski, 2004; Schulz, 2001 and Schulz, 2003; Minbaeva, 2007 and 2008). Scholars typically examine a single or only a few MNCs (e.g., Tsai, 2001; Tsai & Ghoshal, 1998; Sunaoshi et al., 2005; Hansen, 2002; Tsai, 2002). Additionally, 20 out of the 68 empirical studies focus on a single country and more than half (60%) focus on China or the USA. Not many studies utilize a two-country context - Brazil and Mexico (Sparkes & Miyake, 2000), USA and Denmark (Schulz, 2001, 2003), and Finland and China (Björkman et al., 2004; Fey & Furu, 2008). Only a few papers consider subsidiaries located in several countries within the same region. For instance, Grosse (1996) examines MNC subsidiaries in Latin America, Foss and Pedersen's (2002) study is conducted on the basis of subsidiaries in seven West European countries and Kostova and Roth (2002) consider ten countries in Europe, North America and Asia.
In line with these latter studies, we recommend examining subsidiary knowledge flows in a diversified geographical context. While China has been a popular focus of research, expanding the geographical coverage of subsidiary knowledge flows will provide further examinations of variables and relationships that have been subject to inconsistent findings and conclusions (for example, absorptive capacity and institutional distance). We also recommend that future studies shift away from examining subsidiaries of a single MNC and instead, focus on subsidiaries of several MNCs, or subsidiaries located in a single or a few contextually similar host countries. Additionally, we suggest a move away from examining subsidiaries located in the country of origin of the MNC (usually a Western developed country such as Germany, Sweden, or the USA). While such geographical coverage may lead to greater response rates because of typically easier negotiations and access, we encourage the research of subsidiaries located in one or a few contextually similar host countries to accurately explicate the role of factors in relation to subsidiary knowledge flows. Studying subsidiaries of a single MNC, or subsidiaries of MNCs located in a chosen home country provides limited insights. In relation to this, we encourage authors to pay careful consideration to the situational settings in which knowledge flows take place and to invest effort in understanding contextual conditions, both theoretically and empirically. We speculate that taking context seriously and explicitly integrating it into studying subsidiary knowledge flows can help in explaining (or even reconciling) mixed findings from previous research.
Almost all the studies in our sample are conducted at a single level of analysis, either individual (Riusala & Smale, 2007), team (Zellmer-Bruhn & Gibson, 2006), or (mostly) subsidiary level. Notable exceptions are Mudambi and Navarra (2004) who utilized firm and industry level data and Hansen et al. (2005) who analyzed within-team, inter-subsidiary and transfer networks data. We suggest that future examination is conducted at multiple levels of analysis. The field is also in need of more studies which investigate how the sharing of knowledge at the individual level can affect subsidiary-level and MNC-level phenomena. For instance, a fine-grained assessment of the attitudes, behavior and interactions of individual organizational members might trigger new insights about the success or failure of knowledge flows in subsidiaries and in MNCs. Similarly, in order to better understand the nature, intensity and quality of subsidiary knowledge flows, we need to assess the surrounding contexts (industry, market, technology, culture, institutions) in which these flows are embedded. It is worth considering using individual level data (e.g. cognitive style, motivation, tolerance towards ambiguity) as moderators when examining relationships between variables at higher levels of analysis.
The significance of the multi-level examination of concepts, processes and phenomena associated with subsidiary knowledge flows is theoretical, methodological and practical. Investigating the relationship between constructs at different levels captures much of the nested complexity of knowledge flows, enriching existing theories such as the knowledge based view of the firm and organizational learning theory. Methodologically, multi-level research on knowledge flows provides the opportunity for cross-level - i.e., bottom-up as well as top-down - examination of links and interactions, increasing internal validity and avoiding ecological or atomistic fallacy. Finally, managers are in need of guidelines and advice as to the mechanisms (and combinations thereof) with which they can work to steer the direction and intensity of knowledge flows depending on what they want to achieve.
CONCLUSION
Our aim in this work is to contribute to bringing clarity to a field which has been growing steadily over the last decade or so, and which is increasingly experiencing a diversity of definitions and measures of the same key constructs and different - often contrary - findings and conclusions. The specific objective of this paper is to classify and thoroughly review the literature on subsidiary flows in MNCs. We were able to organize existing studies along a few important dimensions by differentiating between knowledge inflows and outflows and between vertical and horizontal flows. In parallel to this, we also classified the literature into four categories: studies with a focus on knowledge characteristics, those with an emphasis on actors involved in the knowledge flows, those which concentrate on relationships between these actors and finally, those which highlight the outcomes of knowledge flows. Such an organizing framework allowed us to identify key accomplishments, existing gaps and possible future research directions.
Our study is not exempt from limitations. We only review literature in top Management journals and in International Business journals. While these are appropriate choices for several reasons, we do not cover studies published in other journals which have engaged seriously with knowledge flows in MNCs. For instance, Human Resource Management journals and Information Systems journals have published on knowledge flows and several papers there have used MNCs as a research context. Another important limitation is the fact that we restrict ourselves to analyzing only the literature on knowledge flows within MNCs. We realize that studies on subsidiary knowledge flows to and from external actors (e.g. firms, customers, suppliers, governmental agencies, research institutions in the host and home countries) are on the rise and can be a valuable addition to our work. Notwithstanding these limitations we believe that our work provides a much needed classification of a rich and rapidly growing literature that will facilitate the accumulation of further knowledge on subsidiary knowledge flows.
