Paper to be presented at DRUID18 Copenhagen Business School, Copenhagen, Denmark June 11-13, 2018

Transcription

1 Paper to be presented at DRUID18 Copenhagen Business School, Copenhagen, Denmark June 11-13, 2018 Multinational companies knowledge strategies in R&D internationalization: A comparison of emerging versus advanced economy multinationals Philip J. Steinberg University of Groningen Innovation Management & Strategy p.j.steinberg@rug.nl Vivien D. Procher University of Wuppertal Jackstaedt Center of Entrepreneurship & Innovation Research procher@uni-wuppertal.de Diemo Urbig University of Wuppertal Jackstaedt Center of Entrepreneurship & Innovation Research urbig@uni-wuppertal.de Christine Volkmann University of Wuppertal UNESCO-Chair of Entrepreneurship and Intercultural Management volkmann@wiwi.uni-wuppertal.de Abstract Emerging economy multinational companies (EMNCs) are increasingly internationalizing their R&D to catch-up with advanced economy multinational companies (AMNCs) technologies. We examine EMNCs versus AMNCs R&D internationalization strategies, specifically their focus on knowledge exploitation and knowledge exploration. Our empirical findings based on a large multi-country dataset show that EMNCs compared to AMNCs are more likely to follow a knowledge exploration and less likely to follow a knowledge exploitation strategy. Moreover, we show that technological leadership constitutes a firmlevel contingency to explain EMNC-AMNC differences in knowledge exploitation but not in knowledge exploration. While there are substantial differences in strategies within the groups of EMNCs respectively AMNCs, these differences are not moderated by technological leadership, and may, thus, be related to other mechanisms.

2 Multinational companies knowledge strategies in R&D internationalization: A comparison of emerging versus advanced economy multinationals Abstract: Emerging economy multinational companies (EMNCs) are increasingly internationalizing their R&D to catch-up with advanced economy multinational companies (AMNCs) technologies. We examine EMNCs versus AMNCs R&D internationalization strategies, specifically their focus on knowledge exploitation and knowledge exploration. Our empirical findings based on a large multi-country dataset show that EMNCs compared to AMNCs are more likely to follow a knowledge exploration and less likely to follow a knowledge exploitation strategy. Moreover, we show that technological leadership constitutes a firm-level contingency to explain EMNC-AMNC differences in knowledge exploitation but not in knowledge exploration. While there are substantial differences in strategies within the groups of EMNCs respectively AMNCs, these differences are not moderated by technological leadership, and may, thus, be related to other mechanisms. Keywords: Emerging economy multinationals, innovation and R&D, internationalization, knowledge exploration, knowledge exploitation, technological position 1

3 1. Introduction The global presence of emerging economy multinational companies (EMNCs) has dramatically increased in the last two decades (UNCTAD, 2005, 2015), making them viable competitors to advanced economy multinational companies (AMNCs). Nowadays, not only AMNCs but also EMNCs place more attention on the internationalization of R&D-related activities (EU, 2006, 2016). While AMNCs earlier internationalization was often driven by exploiting existing firm-specific advantages (Rugman, 1981) and ownership advantages (Dunning, 1988), it is argued that recent EMNC internationalization serves as a strategy to overcome technolocial disadvantages by gaining access to distant knowledge (Awate, Larsen, and Mudambi, 2015). Thus, through internationalization EMNCs may eventually catch up with firms at the technological frontier (Awate, Larsen, and Mudambi, 2012; Bell and Pavitt, 1993; Brandl and Mudambi, 2014; Mathews, 2002; Mathews, Hu, and Wu, 2011). Within their internationalization efforts, we, thus, expect EMNCs compared to AMNCs to place greater emphasis on knowledge exploration, which aims at sourcing advanced foreign knowledge for global use, rather than knowledge exploitation, whereby MNCs leverage upon their existing knowledge. While initial empirical evidence from qualitative indepth case studies supports the notion that EMNCs and AMNCs differ in their internationalization strategies (Awate et al., 2015), we lack larger quantitative evidence that these differences also hold on average for larger numbers of firms and beyond specific pairs of advanced and an emerging economies (Cuervo- Cazurra and Ramamurti, 2014; Ramamurti, 2012). While mostly considered national flagships in their domestic market, EMNCs often lack AMNCs technological capabilities. 1 EMNC internationalization serves as a means of catching- 1 It is important to outline, that we are elaborating on the average EMNC, which should be somewhat more infant than AMNCs, which are on average more mature. 2

4 up with more innovative AMNCs (Awate et al., 2012; Bell and Pavitt, 1993; Brandl and Mudambi, 2014; Mathews, 2002; Mathews et al., 2011) while AMNC internationalization is rather driven by the opportunity to exploit firm-specific capabilities abroad (Rugman, 1981). EMNCs aim at accessing strategic assets and advanced knowledge to overcome latecomer disadvantages regarding production and innovation capabilities (Awate et al., 2012, 2015; Guillén and Garcia-Canal, 2009; Luo and Tung, 2007). Subsequently, EMNCs can apply these newly acquired strategic assets to increase their competitiveness at home, but also to compete in global markets (Hennart, 2012). While the knowledge gap in production and output capabilities can be closed rather quickly, the envisaged innovation catch-up is a big challenge for EMNCs (Awate et al., 2012). This raises the question whether infant EMNCs strategic-asset-seeking motives to invest abroad translate into different R&D internationalization strategies compared to mature AMNCs. Based on a comparative study of two MNCs from the wind turbine industry, Awate et al. (2015) find that, the Indian EMNC has a stronger focus on competence accessing and creation than on competence exploitation in their R&D internationalization compared to the Danish AMNC. Consequently, EMNCs foreign subsidiaries are the major source of innovation and not the EMNC headquarters (Awate et al., 2015; Giuliani et al., 2014). While Awate et al. (2012, 2015) provide first tentative evidence that the R&D internationalization process is fundamentally different for EMNCs and AMNCs, we lack information on the generalizability of those findings to other emerging and other advances economies (Rugman and Nguyen, 2014). We address this gap by studying the R&D internationalization strategies of 374 EMNCs and AMNCs from China, India, the US and Germany. 3

5 Research on EMNCs often implicitly characterizes EMNCs as being infant and lagging behind in terms of their output and innovation capabilities while AMNCs are mature with a superior technological base. Hence, EMNCs motive of technological catch-up results from their lack of technological knowledge and innovation (cf., Awate et al., 2015; Cuervo-Cazurra and Ramamurti, 2014), which, however, is solely based on their emerging market origin. While this might hold true for the average firm from emerging economies, we should take into account the large variation in the technological leadership position of EMNCs. For example, Chinese firms like Alibaba (e-commernce), Huawei (IT hardware) and Tencent (IT services) can be considered to be technological leaders within their industry. Consequently, we analyze whether the R&D internationalization strategy of EMNCs versus AMNCs is contingent on firm s technological position. We expect that EMNCs, which are technological leaders within their industry, will focus less on knowledge exploration and more on knowledge exploitation, thereby mitigating EMNC-AMNCs differences in their R&D internationalization strategies. Our paper provides three important contributions to the literature on R&D internationalization of multinational companies. First, we provide large-scale quantitative empirical analyses on the R&D internationalization strategies of EMNCs in comparison with AMNCs, and thereby allow for testing the generalization of previously conducted comparative case studies (Awate et al., 2012, 2015). We support the view that EMNC internationalization is not a pure reiteration of AMNC internationalization, that is, EMNCs R&D internationalization strategy is much less driven by the eagerness to leverage on existing firm-specific advantages but rather to overcome related deficiencies. Second, we take scholarly calls into account to consider the importance of home country effects when analyzing EMNC internationalization (e.g., Cuervo-Cazurra and Ramamurti, 2014; Ramamurti, 2012). Our dataset includes firms from two rather distinct 4

6 emerging economies, namely India and China, and two also distinct advanced economies, namely the US and Germany, enabling us to also examine the country heterogeneity in both groups. We find significant differences not only across but also within the group of EMNCs and AMNCs and demonstrate that the gap between EMNCs and AMNCs with respect to the importance of knowledge exploration versus knowledge exploitation substantially depends on the chosen pair of countries, being largest for China versus Germany and smallest for India versus the United States. Third, following IB scholars stressing the need to incorporate boundary conditions (Hernandez and Guillén, 2018; Luo and Wang, 2012), we identify a firm s technological position within its industry as a contingency factor when comparing EMNCs and AMNCs focus on knowledge exploitation. We document that EMNCs on average focus less on knowledge exploitation within their R&D internationalization than their AMNC counterparts, but if EMNCs are technological leaders, they exploit their knowledge to the same extent as technological leaders from advanced economies. Interestingly, we cannot identify a similar condition for an MNC s focus on knowledge exploration. We discuss potential explanations. 2. R&D internationalization of EMNCs 2.1. Growing importance of EMNCs In the 21 st century, the number of EMNCs has dramatically increased, partly sparked by promarket reforms in the emerging markets in the 1980s and 1990s (Cuervo-Cazurra and Ramamurti, 2014). As an illustration, the 2017 Fortune Global 500 listing 2, representing the top 500 companies worldwide regarding revenue, included as much as 109 Chinese and 7 Indian companies, compared to 132 companies from the United States. 3 Just a decade ago, in 2007, this 2 See (accessed ). 3 Note however, Rugman and Nguyen (2014) discuss that in the year 2012 only five of these Chinese companies were actually true MNCs, following Rugman s (1981) definition of a MNC. They challenge the view that EMNCs 5

7 list contained only 25 Chinese and 6 Indian companies. These data point towards a shift in economic activity towards emerging economies. Especially the quadruple-increase of Chinese firms within a decade is an impressive catch-up. A similar picture emerges when looking at the 2016 EU Industrial R&D Investment Scoreboard, which monitors the worldwide trends in corporate R&D. Among the 50 global R&D investors (by R&D intensity) there are three companies from China (30. Baidu, 33. Huawei and 37. ZTE) (EU, 2016) 4, compared to three companies from Germany and 15 companies from the US. Thus, these EMNCs have become serious competitors for AMNCs (Guillén and García-Canal, 2013; Ramamurti and Singh, 2009; Williamson and Zeng, 2009; Williamson et al., 2013). The striking growth of EMNCs sparked interest among scholars because theoretical predictions for MNC internationalization being triggered by exploitative motives seemed not fully suitable to explain the foreign investment behavior and internationalization of EMNCs. Despite their aggressive internationalization (Ramamurti, 2012) and the cannibalization of global market shares from AMNCs (Awate et al., 2012; Azevedo et al., 2016; Brandl and Mudambi, 2014), most EMNCs can still be classified as infant MNCs or late-movers with weaker technologies (Awate et al., 2012; Cuervo-Cazurra and Genc, 2008; Ramamurti, 2016), when compared to incument AMNCs, which exploit their technology superiority built at home and through internationalization (Luo and Tung, 2007; Ramamurti, 2016). Considering their technological inferiority, it is unclear how EMNCs can catch-up with the technologies, innovations, and expertise of Western firms (Chari, 2015; Williamson, 2014). If these firms want are already largely present on the world market, stating also that most firms from India in the Fortune Global 500 are not MNCs. Many Chinese MNCs listed in the Fortune Global 500 are still state-owned enterprises (SOEs). 4 The "EU Industrial R&D Investment Scoreboard" (the Scoreboard), published annually since 2004, monitors top EU based R&D investing companies benchmarking them with top R&D investors located in other parts of the world. (EU, 2016, p. 7). The Top 50 R&D companies include firms from the US (23), the EU (15 incl. Germany 3), Japan (4), Switzerland (3), China (3), South Korea (1) and Taiwan (1) (see p.46-51). 6

8 to compete on equal terms on global markets with developed country firms, they have to improve their resources and capabilities (Gaur and Kumar, 2009), especially related to technology and innovation (Awate et al., 2012, 2015). Previous research shows that one of the main drivers for EMNC internationalization is the access to knowledge and resources they do not have in their domestic markets (Mathews, 2006). Therefore, the aggressive EMNC internationalization is often evaluated as an attempt to overcome economic and institutional deficiencies of their home countries (Aulakh, 2007; Cuervo-Cazurra and Ramamurti, 2014; Mathews, 2002) and as a strategy to catch-up with mature AMNC competitors (Awate et al., 2012, 2015; Brandl and Mudambi, 2014; Williamson, 2014). Consequently, EMNCs are most likely to have different internationalization strategies, especially regarding their R&D internationalization, than their counterparts from advanced economies (Luo and Tung, 2007) Role of knowledge exploration and knowledge exploitation in EMNC catch-up As R&D is a key source of knowledge and technology, we focus on the two most prominent strategies related to knowledge when firms internationalize their R&D activities through subsidiaries or external partners. 5 First, firms can leverage on their existing knowledge via knowledge exploitation (von Zedtwitz and Gassmann, 2002; Kuemmerle, 1999). 6 In this case, firms exploit the knowledge that is allocated and orchestrated around the headquarters to build their firm-specific advantage (cf., Rugman, 1981). MNCs can typically scale on their knowledge generated at home by adapting it to the local peculiarities of foreign markets (Kuemmerle, 1999; von Zedtwitz and Gassmann, 2002). Second, in the last decades, firms have increasingly 5 While some studies like Cantwell and Mudambi (2005) focus on mandates of subsidiaries, we widen the definition to include also external foreign R&D activities (cf., e.g., Lewin et al., 2009) since these activities can also be motivated by the exploitation or exploration of knowledge. 6 Please note that strategies or motives for R&D internationalization have been labeled differently in the literature. The knowledge exploitation strategy, characterized by predominant knowledge flows from the headquarters to foreign subsidiaries and foreign partners, has also been labeled home-base exploitation (Kuemmerle (1999), competence-exploiting (Cantwell and Mudambi, 2005) or market-driven R&D internationalization (von Zedtwitz and Gassmann, 2002). 7

9 recognized the potential to source knowledge from foreign destinations (Cantwell, 1989, 1995; Cantwell and Mudambi, 2005; Kogut and Chang, 1991; Pearce, 1999), a strategy labeled knowledge exploration 7 : Firms explore and source knowledge from foreign markets that eventually enriches their knowledge base and thus contributes to their firm-specific advantages and competitiveness (Cantwell and Mudambi, 2005; Kuemmerle, 1999; von Zedtwitz and Gassmann, 2002). EMNCs may close the technology gap to their AMNC competitors by sourcing knowledge and technology from foreign locations that provide market-specific advantages, which are not available in their home country. As Awate et al. (2015) outline, EMNC headquarters often have lower knowledge competencies than their foreign subsidiaries, while in case of AMNC they observe the opposite. This means that EMNCs headquarters catch up with the knowledge and technology of their foreign R&D subsidiaries or R&D partners, making them net users of knowledge, while foreign subsidiaries are knowledge providers (Awate et al., 2015). With respect to R&D internationalization strategies, EMNCs can be expected to strive for knowledge exploration with its possibilities to source and transfer knowledge from foreign locations and, to be able to catch up with AMNC, they need to do so even more than AMNCs, which leads to our Hypothesis 1a: Hypothesis 1a. For their international R&D activities, EMNCs (vs. AMNCs) focus more on a knowledge exploration strategy. Due to their more advanced technological basis, AMNCs are likely to have more opportunities and, consequently, may also be more engaged in knowledge exploitation through 7 Alternative labels have been used to describe this strategy, which is characterized by predominant knowledge flows from the foreign destinations to the headquarters or overall MNE knowledge base, e.g., competence-creating (Cantwell and Mudambi, 2005), home-base augmenting (Kuemmerle, 1999) or technology-driven R&D internationalization (von Zedtwitz and Gassmann, 2002). 8

10 the transfer of existing knowledge from domestic headquarters to foreign subsidiaries (Awate et al., 2015). EMNCs, in contrast, are less likely to develop superior technologies at their home base (Cuervo-Cazurra and Genc, 2008) and, hence, will simultaneously be less likely to follow a knowledge exploitation strategy when internationalizing their R&D, which forms our Hypothesis 1b: Hypothesis 1b. For their international R&D activities, EMNCs (vs. AMNCs) focus less on a knowledge exploitation strategy. The above theorizing for Hypotheses 1a and 1b relates to absolute levels of the importance that firms assign to both the knowledge exploration and the knowledge exploitation strategy. Since knowledge exploration and knowledge exploitation are not necessarily mutually exclusive strategies (cf., Kedia, Gaffney, and Clampit, 2012), it is also important to consider how firms balance both strategies. Firms can choose a unique strategic orientation regarding their knowledge-related R&D internationalization strategies in which they opt for a certain relative balance between knowledge exploration and knowlegde exploitation. For example, MNCs could have a strong focus on both, the exploitation of exisiting knowledge and the exploration of new knowledge and technologies (cf., von Zedtwitz and Gassmann, 2002), resulting in a strategy of equally balancing competence-exploitation (e.g., exploiting knowledge) and competencecreation (e.g., exploring knowledge) (cf., Cantwell and Mudambi, 2005). Since, EMNC are constraint in their use of a knowledge exploitation strategy (Cuervo-Cazurra and Genc, 2008), such a balance of exploiting and exploring knowledge at high absolute levels is more likely to be available for AMNCs. While a balancing strategy does not imply that AMNCs engage less than EMNCs in knowledge exploration, overall the relative importance of knowledge exploration vis- 9

11 à-vis knowledge exploitation can be expected to be more pronounced for EMNCs in comparison with AMNCs (see Hypothesis 1c). Hypothesis 1c. For their international R&D activities, EMNCs (vs. AMNCs) assign a higher relative importance of a knowledge exploration strategy over a knowledge exploitation strategy The relative technological position as moderator of EMNC-AMNCS differences Previous qualitative research suggests that the difference between EMNCs and AMNCs is partly driven by the firm-specificity of technological inferiority (Awate et al., 2015) or in other words, the company s technological capability. A firm s technological advancement usually determines its technological position within an industry (Grimpe and Sofka, 2016; Salomon and Jin, 2010; Steinberg, Procher, and Urbig, 2017), 8 within a country (Mahmood and Rufin, 2005), or worldwide (Kumar and Russell, 2002). The technological frontier represents the maximal available technological opportunity at a given time (Grimpe and Sofka, 2016; McCain, 1977). Chung and Alcácer (2002) suggest that technologically lagging firms can use their international investments to catch up with leading firms and thereby reduce the technology gap. In line with this argumentation, Blalock and Gertler (2009) claim that, as a consequence, laggards have more opportunities to improve their technological position and, therefore, gain more from exposure to new technologies and knowledge. Thus, laggards experience stronger incentives and are, therefore, more likely to be actively seeking for and exploring technology through international investments (Chung and Alcácer, 2002). Leaders, in contrast, will have more opportunities to exploit their technologically advanced knowledge in foreign places. Applied to the context of the comparison between EMNCs and AMNCs, this implies that an MNC s relative technological position influences its strategic focus on knowledge exploration, 8 A firm s relative technological position within an industry is strongly overlapping with a firm s technological position relative to competitors in its main sales-market since the main sales-market usually determines the firm s primary industry within industry classification, such as ISIC, or NACE. 10

12 respectively, knowledge exploitation. Consequently, Ramamurti (2012) and Ramamurti and Hillemann (2018) emphasize the relevance of considering the development stage of an MNC. While EMNCs have usually been classified as infant MNCs and incumbent AMNCs as mature MNCs, we expect that the difference between EMNCs and AMNCs in knowledge exploration and knowledge exploitation will be less pronounced if a firm is a technological leader relative to its competitors in its respective main sales market, and more pronounced if it is a laggard. In other words, technological leaders from emerging markets can opt for similar strategies as incumbent AMNCs. As an example, Huawei, being a relative technological leader in telecommunications is probably no longer dependent on an innovation catch-up strategy and can even exploit its technological expertise through internationalization. We, therefore, hypothesize that a firm s relative technological leadership (i.e., being close to the technological frontier) will have a negative moderating effect on the relation between EMNCs (versus AMNCs) and knowledge exploration (both absolute and relative to knowledge exploitation) and a positive moderation effect on the relation between EMNCs (versus AMNCs) and knowledge exploitation: Hypothesis 2a. The positive relationship between EMNCs (vs. AMNCs) and a knowledge exploration strategy is less positive for EMNCs closer to the technological frontier. Hypothesis 2b. The negative relationship between EMNCs (vs. AMNCs) and a knowledge exploitation strategy is less negative for EMNCs closer to the technological frontier. Hypothesis 2c. The positive relationship between EMNCs (vs. AMNCs) and the relative importance of a knowledge exploration strategy over a knowledge exploitation strategy is less positive for EMNCs closer to the technological frontier. 11

13 3. Method 3.1. Research design We base our analyses on primary data collected through a large quantitative survey on international R&D activities of 500 firms from China, India, the United States and Germany, which we conducted in conjunction with EY (Ernst & Young) in These primary data were merged with and carefully validated based on secondary data derived from Bureau van Dijk s Orbis database. The survey, which we refer to as EY Global Innovation Survey (EYGIS), focuses on MNCs R&D activities and related management practices and was simultaneously carried out in four countries. Besides information on the national and international set-up of R&D activities, participants were asked about several national and global characteristics of their company, allowing us to use a range of relevant control variables. Furthermore, our key variables related to knowledge exploration and exploitation cannot be derived from official balance sheet data or nation-wide innovation surveys like the Community Innovation Survey (CIS). The survey questions are aligned to well-known surveys, following standards such as the Frascati and Oslo Manual (OECD, 2002, 2005) where applicable. Before data collection, the survey was discussed in an iterative process with industry experts holding similar positions as the later respondents to ensure that each question is comprehensive. The original questionnaire was constructed in German and was then translated into English and Chinese by professional service providers. Due to the many different languages and dialects in India, the main part of the interview in India was held in English while the welcome address and farewell were held in local languages. The survey addressed MNCs that carried out internal or external R&D abroad, and are active within the information and communication technologies (ICT) or manufacturing industries, 12

14 excluding pharmaceuticals 9, with a focus on firms with at least 50 million USD sales worldwide. The data were collected at the parent company level. Based on a corresponding selection of firms form Bureau van Dijk s Orbis database, firms were approached until the envisioned sample size of 125 responses (irrespective of whether or not all questions were answered) for each country was achieved. The survey was conducted via structured computer-aided telephone interviews by specialized service providers with fluent language speakers in April and May Survey respondents were senior executives responsible for the enterprise s R&D. The short time span in which the survey was conducted offers an advantage over surveys with longer time ranges because potential influences of changes in economic conditions over time are minimized. The survey was implemented together with EY as a professional partner, which provided us easier access to potential interview partners. The overall response rate for the survey was 48% and the average number of calls until success was Before conducting the interviews, information on the MNCs industry classifications based on Bureau van Dijk s Orbis database were merged. To increase the reliability of the data, information from the survey on MNC sales and numbers of employees were carefully crosschecked with the associated information from Orbis. When larger differences occurred, a second telephone interview was conducted to validate key information. Considering the steps taken to increase validity, we expect the data to have comparable reliability as nation-wide innovation surveys, such as the CIS surveys. For a sample of 374 firms, we have information on all our variables of interest (113 from China, 63 from India, 93 from the US and 105 from Germany). The average domestic 9 The pharmaceutical industry was excluded due to the special role of clinical drug trials (which are partly declared as R&D expenditures) and related regulatory issues. 13

15 employment is 2,472 employees, and the average global revenue is 895 million USD. The manufacture of machinery and equipment (15.0%) and the manufacture of computer, electronic and optical products (11.8%) represent the largest industries in our sample Country setting With this study we aim at larger quantitative evidence to complement previous qualitative studies and acknowledge that country differences may confound the identified EMNC effect if we focused on only one emerging and one advanced economy (Hoskisson et al., 2000). We selected China and India as two economically very relevant though very different emerging economics, e.g., with respect to the business culture and specific industry policies, and the US and Germany as two economically very relevant but again culturally rather different advanced economies. Even though there are no strict criteria on how to define emerging market and developing economies (International Monetary Fund, 2016), an emerging market economy is usually characterized by low to middle income per capita, rapid economic development, and being in the process of moving from a closed economy to a free-market system (Arnold and Quelch, 1998; Cuervo-Cazurra and Ramamurti, 2014). In line with this characterization and consistent with previous research (Arnold and Quelch, 1998; Brandl and Mudambi, 2014; Luo and Zhang, 2016), we consider China and India as emerging economies, and the US and Germany as advanced economies. With these four countries, we can collect a sufficient number of firms within each economy, but also have substantial diversity within the group of MNCs from emerging and from advanced economies. Figure 1 illustrates some prominent economic indicators for our selected four countries. China is ranked second among the four countries regarding total GDP (Panel 1A), experiencing a massive economic growth over the last 15 years. When normalizing the GDP by capita (Panel 14

16 1B) a different picture emerges, which reveals that China and India are still economic laggards compared to advanced economies like the US and Germany Figure 1 goes about here Following trade liberalization in China and India exports and imports have been steadily increasing since the late 20 th century (Panels 1C and 1D). India has first implemented radical economic reforms in 1991, followed by further reforms to loosen trade barriers, such as cuts of tariffs in 2004 and China has also undergone serious economic reforms, starting as early as 1978 with opening markets for FDI, followed by further reforms in the late 1980s until the early 2000s with a focus on privatization, lifting of price controls and decreasing protectionism. As a result, both economies have benefitted from market liberalization with increasing trade shares and GDPs. Regarding the total amount of exports, China has even taken the lead among the four countries in the last years. Regarding GDP and trade, at least China is catching up with Germany and the US, but the difference between the two advanced economies and the emerging economies remains striking when looking at key innovation indicators, such as the patent applications at the European Patent Office (EPO) for the four countries (Panel 1E). The US had, by far, the most applications and also has experienced a strong increase in recent years (42,597 applications in 2015). Germany follows at a steady rate of about 25,000 applications per annum. China (5,728 applications in 2015) and India (577 applications in 2015) are both fundamentally lagging behind, but experience a strong increase (which especially holds for China). This increase may be partly explained by China s strong increase in public and private R&D expenditures relative to its GDP 15

17 in the last years (see Panel 1F). In terms of relative input for innovation, China is slowly but steadily catching up with Germany and the US and might well reach their relative spending in the next years. Remarkably, India s relative R&D expenditures are stagnating below 1% of its GDP. One reason might be that India has no state policy that pushes towards innovation as immensely as in China Variables Dependent variables To capture the extent to which firms follow a knowledge exploration strategy in their R&D internationalization, firms indicated whether access to specialist knowledge or technologies and access to qualified personnel was not important (1), important (2) or very important (3) to them. 10 The items to indicate an exploration strategy are coherent with items from surveys such as the Eurostat survey on international organization and sourcing of business activities in 2012 and the Swiss Innovation Survey 2011 conducted by ETH Zuerich, and the items relate to the key characteristics of exploration strategies as indicated by von Zedtwitz and Gassmann (2002). We use the average of responses to these items to construct our variable indicating a firm s knowledge exploration strategy. The scale s internal reliability is sufficiently high as indicated by coefficient alpha of To capture the extent to which firms engage in a knowledge exploitation strategy in their R&D internationalization, we proceed identically to the operationalization for knowledge exploration. Evaluations of the motives access to new markets and adaptation of products to the foreign sales market were used to operationalize the knowledge exploitation strategy. As before, these items are consistent with the surveys mentioned above. We use the average of 10 Please note, that the scale follows the Eurostat survey. Additionally, our survey was conducted via computeraided telephone interviews and in such cases scales with too many alternative responses can be confusing and incomprehensible to participants. 16

18 responses to these items to construct our variable. The scale s internal reliability is sufficiently high as indicated by a coefficient alpha of Relative knowledge exploration reflecting the relative importance of knowledge exploration over knowledge exploitation strategies was constructed as the importance of the knowledge exploration strategy normalized by the sum of the importance of both, knowledge exploration and exploitation strategies Key explanatory variables EMNC is a dummy variable taking the value of 1 if the firm is a multinational company from an emerging economy (China or India) and 0 if the firm is a multinational enterprise from an advanced economy (the United States or Germany). Technological leadership is a variable measuring the innovativeness of a firm compared to its competitors in its primary sales market (cf., Makino, Lau, and Yeh, 2002). On a five-point scale from 1 (very rarely first) to 5 (very often the first) participants evaluated whether their firm compared to its competitors in their primary sales market in the period 2013 until 2015 were innovative first-movers when launching new products, services, business processes, technologies, etc. (Covin, Slevin, and Schultz, 1994; Hansen, Shrader, and Monllor, 2011). R&D intensity relative to a firm s industry is often used as a proxy for technological leadership (Grimpe and Sofka, 2016; Salomon and Jin, 2010; Steinberg et al., 2017). In our specific case, where we compare EMNCs and AMNCs, this variable would not be a suitable proxy. EMNCs, in an effort to catch up, may dramatically increase their R&D intensity, as input into the innovation process, but might not have yet achieved sufficient innovation output and a sufficient knowledge base to actually be considered as technology leaders (see also comparison of Figure 1E and 1F). 17

19 In this very reasonable case, the R&D intensity as a proxy for technological leadership would wrongly suggest that these firms are already technological leaders Control variables While our hypotheses only relate to differences between MNC from emerging versus MNCs from advanced economies, we also explore country-related heterogeneity within each of the two groups. We construct two contrast codes for INDIA (vs. China) and USA (vs. Germany), the contrast codes are +1 and -1 for the two contrasted countries (e.g., India = +1 and China = -1) and 0 for the two remaining countries. While the EMNC dummy together with two dummies controlling for the differences within each group (e.g., dummies for India and Germany) would be econometrically equivalent, the contrast coding allows us to directly interpret the coefficient estimated for the EMNC dummy as the average effect of emerging versus advanced country MNCs, averaged at the country level (Cohen et al., 2013). For illustrations, we calculate countryspecific coefficients based on appropriate linear combinations of the EMNC average effect and the estimated within-group contrast (e.g., the effect associated with China would be the coefficient of EMNC minus the coefficient of the contrast code). To further reduce the risk of spurious results caused by, e.g., differences in MNCs general input into R&D, we include R&D capacity. We follow Berchicci (2013) and measure R&D capacity as the number of domestic R&D employees divided by the number of total domestic employees (see also Cassiman and Veugelers, 2002). We additionally control for firm performance. Respondents in the survey were asked how satisfied the company s top management was with the general performance of the company in The variable ranges from 1 (very dissatisfied) to 5 (very satisfied). 18

20 To control for spurious results due to correlated strategies for R&D internationalization, we control for a third major strategy, which is cost- and not knowledge-related (cf., Cantwell and Mudambi, 2005). Cost strategy is a multi-item measure including survey items on the importance of Cost savings due to lower personnel expenses, Cost savings due to lower tax burden or tax incentives abroad, Greater funding opportunities for R&D abroad and Other savings, which were rated on the same scale as exploitation and exploration strategies. We use the average of responses to these items to construct the variable. The internal reliability is sufficiently high as indicated by coefficient alpha of To control for potential learning effects from internationalization experience, we include international R&D experience and export share. As Ramamurti (2012) states, international R&D experience is an important variable, since effects that may seem to be related to the country destination (as the EMNC effect) are in fact attributable to international experience of MNCs. Based on the available data we included a categorical variable with 1-5 years, 6-10 years, years and > 15 years of international R&D experience. We do not have a category no experience, because some international R&D activity was a pre-requisite for participating in the survey. We include a firm s export share to control for the general importance of foreign markets to a firm. Export share is a categorical variable indicating the proportion of a firm s total domestic revenue that is attributable to exports, classifying firms into no export, 1-25%, 26-50%, 51-75% and >75%. We include further firm-specific control variables that may be related to firms international R&D activities. Firm age is a variable capturing the firm s age as the natural logarithm of the mid-point of five-year intervals. Firm size is measured by the natural logarithm of the number of employees. Since some unique features of EMNCs may not necessarily be associated with their 19

21 home country but rather with their industry of operation (Brandl and Mudambi, 2014; Ramamurti, 2012), we also control for any remaining industry effects by including industry fixed effects. Following OECD categorization for the manufacturing industry (OECD, 2011) we include dummy variables for high-tech, medium-high-tech, medium-low-tech and low-tech manufacturing industries, as well as one for firms from ICT industries. 4. Results Table 1 reports descriptive statistics for our sample and correlations between the dependent variables and all other variables (see Appendix A for the complete correlation matrix). Small to moderate correlations and variance inflation factors below 5.0 in all models indicate no problems with multicollinearity (Belsley, Kuh, and Welsch, 2005). Table 1 reveals that the association of EMNC with knowledge exploration is positive, while the association with exploitation is the opposite (negative). A similar picture emerges for firm age, with younger firms being rather associated with exploration and older firms with exploitation. Interestingly, firm size is related to knowledge exploration but not to exploitation. The cost strategy is positively related to both knowledge strategies but much stronger for exploitation, such that firms following a cost strategy place also more importance on exploitation rather than exploration strategies. This is in line with the reasoning that often a focus on cost is associated with a development towards exploitation (cf., Cantwell and Mudambi, 2005) Insert Table 1 about here Turning to our hypothesis tests, Table 2 reports results of hierarchical ordinary least squares regression (with clustered standard errors correcting individual-level correlations over time) that 20

22 test Hypotheses 1a, 1b and 1c. Analyses were run for an MNC s focus on a knowledge exploration strategy (Models 1-3), knowledge exploitation strategy (Models 4-6), and relative knowledge exploration (Model 7-9). Each initial model focuses on the control variables including a firm s relative technological position (i.e., technological leadership). In Model 1 firm size, R&D capacity, performance satisfaction, and technological leadership are positively and statistically significantly correlated with knowledge exploration. The findings for technological position and performance seem to contradict the idea that especially laggards are focusing on exploration to catch-up (Chung and Alcácer, 2002), and may support the perspective that many firms simultaneously engage in competence exploration and exploitation (Cantwell and Mudambi, 2005). This perspective is further supported by the finding that the relationship between the two knowledge-related strategies of knowledge exploration and exploitation is positive and statistically significant in Models 1 and 4, respectively. In Model 4, firm age is positively and statistically significantly correlated with knowledge exploitation, reflecting the fact that mature MNC may exploit their technological superiority by internationalization (Ramamurti, 2016). Moreover, cost strategy is positively and statistically significantly correlated with knowledge exploitation, reflecting a potential association between the two strategies (cf., Cantwell and Mudambi, 2005). The findings for Model 7 corroborate the findings from Model 1 and 4, with statistically significant relations of relative knowledge exploration with firm size (positive), firm age and cost strategy (negative) Insert Table 2 about here

23 Model 2, 5 and 8 introduce the EMNC dummy to test our Hypotheses 1a, 1b, and 1c. Being an EMNC is significantly related to exploration (positive) and to exploitation (negative), as well as to relative knowledge exploration (positive). Hence, we support our hypotheses stating that EMNCs, in comparison with AMNCs, have a stronger focus on a knowledge exploration strategy and a weaker focus on a knowledge exploitation strategy when internationalizing their R&D activities. Being an EMNC, on average, increases the importance of the knowledge exploration strategy by 26% of one standard deviation (Model 2) and decreases the importance of the knowledge exploitation strategy by 21% of one standard deviation (Model 5) in our regression analyses. 11 Our analyses so far have pooled firms from the two emerging countries and those from the two advanced economies. Differences between emerging and advanced firms may not apply to all pairs of emerging and advances economies but may result from particularities of selected pairs. To explore the heterogeneity within the groups of emerging respectively advanced economies, we include the contrast codes of INDIA (vs. China) and USA (vs. Germany) (Models 3, 6, 9). While the estimated coefficients do not seem to be large enough to actually dominate the effect of EMNC versus AMNC, we observe substantial and statistically significant differences particularly for knowledge exploration in both groups and for relative knowledge exploration between EMNCs. Based on the EMNC dummy and the contrast codes, Figure 2 graphically illustrates the estimated levels (and related confidence intervals) for the three dependent variables for all four countries. We observe that Chinese MNCs have the strongest focus on exploration and weakest focus on exploitation in our sample, both in absolute and in relative terms. India, however, does not differ that much from the AMNCs and, in particular, with respect 11 Due to standardizing our dependent variables, these percentages can be directly derived from the coefficients reported in Tables 2 and 3. 22

24 to absolute and relative knowledge exploration, it does not differ from the US. While one might be tempted to argue that India does not explore enough to be able to catch up, the relatively high relative importance of knowledge exploration for the US versus Germany, might also indicate that US-based firms are more likely to balance knowledge exploration and exploitation at comparable levels (cf., Cantwell and Mudambi, 2005) and, thus, to keep high levels of exploration despite their technological advancement. In sum, while the gap between EMNCs and AMNCs does not seem to be substantially reversed, nevertheless, its magnitude is highly sensitive to the selected pairs of emerging and advanced economies and might, in some circumstance, even virtually disappear. While this might be counted as evidence against a systematic difference between EMNCs and AMNCs, it is important to note that this might only imply that besides an EMNC effect there is additional country-specific variation that overlaps with the EMNC effect. Through our following analyses based on moderation effects, we will demonstrate that the latter, i.e. the overlapping of different effects, might indeed be relevant Insert Figure 2 about here In Table 3 (Models 10, 13, and 16) we report estimations that additionally include the interaction of EMNC with technological leadership, which allow testing our Hypotheses 2a, 2b, and 2c. To simplify interpretation, Figure 3 graphically illustrates the moderating effects; it plots the estimated difference between EMNC and AMNC and the related confidence interval conditioned on the level of technological leadership. For all strategy variables, we observe that the magnitude of the effect is smaller when comparing technological leaders; however, this moderating effect is small and statistically not significant for the absolute levels of knowledge 23

25 exploration. Hence, we cannot support Hypothesis 2a. However, we find unambiguous support for Hypothesis 2b, stating that the difference between EMNCs and AMNCs regarding knowledge exploitation within their R&D internationalization is negative but decreases in magnitude the more technologically advanced an MNC is. Similarly, we observe that the relative focus on exploration is more positive for EMNCs compared to AMNCs, but decreases for technologically more advanced firms, which supports our Hypothesis 2c. In fact, these latter two moderation effects are as strong as to virtually make the EMNC-AMNC gap disappear for technologically very advanced firms. Our earlier analyses revealed that a substantial part of strategic differences is explained by heterogeneity within the groups of EMNCs and AMNCs (see Models 3, 6, and 9). We have also seen that the difference between EMNCs and AMNCs with respect to knowledge exploitation (and the relative importance of knowledge exploration) can indeed be explained by firms differences in technological leadership, which supports catch-up-based explanations. With Models 12, 15 and 18, which are reported in Table 3, we now explore to what extent the previously mentioned within-group effects are moderated by technological leadership. Our regression analyses do not reveal any indication that within-group effects are moderated by the technological position. Hence, in contrast to the between-group effect and, in particular, for exploitation and the relative importance of exploration over exploitation, these within-group effects are not likely to be driven by mechanisms related to catching up, but possibly by other country peculiarities, e.g., the institutional context. Hence, we isolate effects that are specific to the difference between EMNCs and AMNCs rather than to differences between countries in general. 24

26 Insert Table 3 and Figure 3 about here Discussion In this article, we theorize on differences between EMNCs and AMNCs R&D internationalizing strategies. We base our reasoning for a stronger focus on knowledge exploration of EMNCs R&D internationalization, in comparison to AMNCs, on the premise that EMNCs strive to catch up with the technologies and expertise of more mature AMNC competitors. Given that EMNCs, on average, lack firm-specific technology advantages due to their weak knowledge base at home (Awate et al., 2012; Cuervo-Cazurra and Genc, 2008; Ramamurti, 2016), we theorize that EMNCs will be less focused on knowledge exploitation, because they cannot yet leverage on strong existing innovation capabilities when they internationalize their R&D. We further theorize that technological leadership, as an important firm-level contingency (Makino et al., 2002), diminishes the differences between EMNCs and AMNCs regarding knowledge exploration and knowledge exploitation. We base our analyses on a relatively large sample of 374 MNCs from four different countries (China, India, the United States, and Germany) that allow us to test our theorizing of EMNC- AMNC differences and also explore firm-level heterogeneity and country-level heterogeneity within the group of EMNCs and AMNCs, respectively. We find that EMNCs (versus AMNCs) are more likely to focus on a knowledge exploration strategy and less likely to focus on a knowledge exploitation strategy when internationalizing their R&D. Furthermore, we find that in case of knowledge exploitation the difference between EMNCs and AMNCs is less pronounced and even almost absent for technological leaders compared to technological laggards. Though 25