Overseas R&D center and innovation performance: The view of host country’s location advantages

全文

(1)#1569265189. OVERSEAS R&D CENTER AND INNOVATION PERFORMANCE: THE VIEW OF HOST COUNTRY’S LOCATION ADVANTAGES YUAN-CHIEH CHANG Institute of Technology Management, National Tsing Hua University, 101, Sec 2, Kuang-Fu Rd., 30013, Hsinchu, Taiwan yucchang@mx.nthu.edu.tw. TING-LIN LEE APIBM, National University of Kaohsiung, 700, Kaohsiung University Rd., Nanzih District, 81148. Kaohsiung, Taiwan linda_lee@nuk.edu.tw. WEN-CHIANG CHIENG Institute of Technology Management, National Tsing Hua University, 101, Sec 2, Kuang-Fu Rd., 30013, Hsinchu, Taiwan martin@iii.org.tw. TZU-MIN CHIN Institute of Technology Management , National Tsing Hua University, 101, Sec 2, Kuang-Fu Rd., 30013, Hsinchu, Taiwan g9673506@oz.nthu.edu.tw. 1 .

(2) Introduction Follow the trends that product life cycle is shortening, customers’ demand is getting complicated, and market competition is getting keener, for a multinational enterprise (MNE), It has become the most important strategy that speeding up to transfer new products abroad for manufacturing and marketing. However, if manufacturing and R&D facilities were located in different sites could cause negative impacts on technology transfer and the proceeding innovation and improvement on existing process technologies. Thus, for shortening time-to-market of new product development and offering technical service to overseas markets, MNEs shift their R&D facilities abroad from the home countries and locate them near the markets. In addition to exploit technologies from the home countries to adapt for local market needs, although the main functions of R&D remain in headquarters of MNEs, the scarcity of high value R&D resources such as talent, know-how, and research-based networks forces MNEs to set up R&D sites in technology-advanced regions to retrieve such resources. According to Patel and Pavitt (1999), within MNEs which get approvals of intellectual property right in the US, innovation from overseas R&D sites outperform innovation from R&D units in the headquarters. More recently, the desire to acquire technologies, especially from the emerging economies, has been a major motive for MNEs to locate R&D facilities abroad (Kuemmerle, 1999). As for the motives for MNEs to set up their R&D facilities/centers outside the home countries from a host country’s perspective, This study attempts to employ the location advantages which based on Dunnings’ framework of ownership-location-internalization (OLI ) advantages (Dunning, 1988) as the sources of impacts on innovation performances of such MNEs’ R&D centers in host countries, due to most literature put focus on impacts of host countries’ location advantages on motives of MNEs to deploy R&D resources abroad but rarely studied the relationship between innovation performance and location advantages. Among those emerging countries that urge to attract R&D investment from MNEs which based in advanced countries, Taiwan has become an appropriate example to explore issues of R&D globalization. Taiwan has been famous for her rich engineering talent pool and solid semiconductor and ICT industry clusters. In addition, Taiwan’s high-quality research-based networks and government incentives had also made MNEs start to shift more and more R&D resources in Taiwan in recent years. For instance, G. R. Barrett, CEO of Intel, announced to set up its “Intel Innovation Center” in Taiwan in 2003 for Taiwan’s solid technological foundation and strong global competitiveness in ICT-related fields. Intel used to set up its applied technology supporting center in Taiwan mainly for offering technology services to its clients in Taiwan. The “Intel Innovation Center” could properly explain evolutions of MNEs’ R&D sites in Taiwan since Taiwan’s general R&D capacity was getting stronger. Therefore, this study aims to study impacts of Taiwan’s location advantages on innovation performance of the MNEs’ R&D centers in Taiwan by means of employing the system dynamics approach, and furthermore, to find out which location advantage exhibits the highest importance to innovation performance of MNEs’ R&D centers. The rest of this paper is organized as follows. Section 2 discusses trends and motives of internationalization of corporate R&D, four major types of location advantages of host countries, and innovation performance of MNEs’ R&D centers in host countries, followed by a review of literature. Section 3 details the research method, including explaining the concept of system dynamics approach and steps of constructing the system model architecture. The construction of the model, testing of the constructed model, and simulation of future government policy strategies are presented in Section 4. Section 5 concludes this study.. Literature Review. 2.1 Trends and motives of R&D internationalization Conventionally, corporate R&D was located in the home countries. But since the 1980s, internationalization of corporate R&D has become a major trend (Paoliz and Guerciui, 1997; Patel and Pavitt, 1999; Guellec et al., 2001). Due to seeking for the best R&D resource allocation, MNEs has started to integrate their overseas R&D sites into their global R&D networks and thus make those R&D sites interdependent and 2 .

(3) interactive (Chiesa, 1996; Gerybadze and Reger, 1999). Kuemmerle (1999) argued that MNEs started to build up overseas R&D centers to absorb knowledge resources that are embedded in different places and to retrieve local technologies and talent, Kumar (2001) argued that overseas R&D activities are deeply affected by location variables such as market scale, technological base, resource and capability of host countries. Hegde and Hicks (2008) also pointed out that market size of host countries will result in different involvement of US-based MNEs. As for motives and types of MNEs global R&D activities, Kuemmerle (1999) proposed two types of global R&D centers such as home-based exploiting (HBE) and home-based augmenting (HBA). HBE seeks to exploit firm-specific capabilities, indicates that the knowledge flow is from home-based R&D sites to home-based exploiting R&D sites. HBA seeks to access to unique resources in host countries, indicates the knowledge flow is from home-based augmenting R&D centers to home-based R&D sites. In general, the local market characteristics and knowledge base owned by host countries are the main reasons for MNEs to set up their R&D centers abroad (Le Bas and Patel, 2007). In addition to market and technology factors, lifting up quality of human resources is one of the major motives for MNEs to internationalize R&D activities. Anand and Kogut (1997) argued that seeking for competitive resources is the major motive that causes R&D internationalization. Competitive resources are not limited to tangible resources such as land, energy and raw material, rather than that, knowledge and high quality human resources are more important. That is, those MNEs which plan to conduct R&D activities abroad will expect and require that host countries can offer such resources which can enhance their competitive advantages.. 2.2 Location advantages of host countries According to Dunning’s framework of OLI advantages (Dunning, 1988), location advantages owned by host countries stand for strategic importance of host countries to MNEs. This study identifies and categorizes the location advantages into four major types. They are (1) local industry clusters; (2) local talent pool; (3) local research-based networks; and (4) local government support. 2.2.1 Local industry clusters While considering set up overseas R&D centers, MNEs will take local industrial comparative advantages of host countries into account. The main reason is that MNEs’ overseas R&D centers can enjoy spillover effects from local industry clusters which represent local industrial comparative advantages. Birkinshaw and Hood (2000) argued that the roles of overseas subsidiaries should be coped with the influential effects of local industry clusters. Zander (1999) also pointed out that MNEs will learn and accumulate technological capability by means of interacting with local technological resources embedded in local industry clusters. In a research of Swedish-based MNEs’ overseas location selection exhibited that industrial comparative advantages are the main decision factors (Fors and Zejan, 1966). 2.2.2 Local talent pool In doing foreign direct investment, MNEs will seriously consider quality of local human resources. Nachum (2000) argued that in considering location comparative advantages, the main factors for MNEs to decide where to set up their R&D centers had been departed from merely richness of tangible manufacturing resources to containing intangible assets such as culture, human capital, and local institutions. Kumar (2001) focused on the US-based and Japanese-based MNEs’ overseas R&D sites selection criteria and argued that rich R&D talent can effectively attract MNEs’ R&D investment. Westney (1992) also found that the richer industrial R&D talent of host countries, the higher R&D intensity of MNEs’ overseas R&D centers. 2.2.3 Local research-based networks Local research-based networks in host countries are also one of the major motives of R&D internationalization. Some studies indicated that externalities caused by knowledge interflows among local universities, national labs and innovation competitors within host countries enhance attractiveness to MNEs (Kogut and Chang, 1991; Florida, 1997; Nachum, 2000). 3 .

(4) 2.2.4 Local government support R&D activities are usually based upon continuously capital, human resources and facilities inputs. Therefore, a supportive R&D environment which offers basis for R&D activities is one of the key success factors of R&D performance. Host countries which can offer such supportive R&D environment, including intellectual property right protection, tax deduction, etc., will definitely attract MNEs to invest R&D resources. Dunning (1988) argued that many foreign R&D investment made by pharmaceutical firms are based upon strict regulations requirements of host countries. Taggart (1998) also pointed out that local government incentives are one of the major motives for MNEs to allocate R&D resources in host countries.. 2.3 Innovation performance of overseas R&D centers For MNEs’ overseas R&D centers, the main activities are focused on new product development and technical services. One of the major missions for overseas R&D is to retrieve required technologies and knowledge and to exploit them into processes of developing new products (Ronstadt, 1978; Hood and Young, 1982; Hedlund and Dag, 1990; Kuemmerle, 1997). Another major mission for overseas R&D is to offer technical services to other functional departments such as manufacturing and marketing departments within the MNEs. Brown and Sevenson (1998) argued that R&D activities form a system which includes five stages such as input, process system, output, access system, and outcome. In measuring innovation performance of overseas R&D centers, output and outcome could be proper measures.. Research Method The concept of system dynamics was proposed by MIT Professor Jay W. Forrester and colleagues in 1956. Forrester felt that the whole world is a cyclical dynamic system interacting through causal relationships, where the current state results in certain actions that are determined by existing decisions. These actions will cause the state to change, will influence new decisions, creating a cyclic causal system. Over time unpredictable effects will influence the cycle. The goal of this study is to investigate how Taiwan's advantages influence the establishment of R&D centers in Taiwan by MNEs, and also examine how these advantages influence the performance of R&D centers established in Taiwan by foreign enterprises. Because these processes involve decisions, actions, and changes of state, this study uses system dynamics to gain an understanding of time-varying system changes. To facilitate the achievement of these goals, the following section contains an examination and discussion of literature concerning the definition, theoretical basis, beliefs and characteristics, and basic elements of system dynamics.. 3.1 System dynamics 3.1.1 Definition of system dynamics Forrester (1961) points out that system dynamics is a means of studying enterprises' internal systems and constructing models for improving organizational structures and guiding formulation of strategies; system dynamics can be used in conjunction with verbal descriptions, experience, on-site observations, and any other usable information to construct mathematical models, and, via computer simulations, can show how factors such as organizational system structures, policies, and time delay can influence organizational growth and stability (Forrester, 1961). Senge (1990) defines system dynamics in a way similarly to that of the foregoing scholars, and suggestions that system dynamics can be used as a tool helping decision-makers to solve complex dynamic internal and external problems. In summary, this study concludes that system dynamics is a research method used in management science that can be used to analyze an enterprise's complex internal/external time-varying problems, and, through modeling, can also simulate implementation of various policies, assess the effect of policies on enterprise performance, and provide the enterprise with recommendations for improvement and target points for policies. From its definition, we know that system dynamics can analyze internal and external variables involving the 4 .

(5) environment and policies, etc., and it can effectively solve complex time-varying problems that ordinary quantitative measurement models cannot handle. Systems constructed using system dynamics contain elements with the following attributes: 1. Specific goals 2. There are least two elements 3. Each element has a certain definition and restrictions, and is self-adjusting 4. Elements interact with the external environment 5. Have an auto-correction effect 3.1.2 Characteristics of system dynamics In contrast with other areas of social science research, system dynamics research methods possess the following features (Han, 2002): 1. System dynamics is good at handling cyclical problems: System dynamics can analyze and simulate regular phenomena in the macroeconomic environment, such as the economic cycle and inflation, and can determine the reasons for problems. 2. System dynamics is good at handling long-term problems: System dynamics models can be used to observe system behavior via long-range simulations. 3. System dynamics allows research to be performed when data is lacking: Because system dynamics models are based on causal feedback loops, when there are multiple loops, and this ensures that system behavior is not sensitive to most variables during modeling, system behavior will style display identical states within the range of tolerance. 4. System dynamics is good at handling high-order, nonlinear, time-varying problems: System dynamics can be used to analyze large quantities of data via computer simulations, which can facilitate application to the solution of high-order, nonlinear equations and complex time-varying problems. 5. Systems dynamics can be used to perform simulations of specific conditions: System dynamics models can not only be used to perform dynamic simulations, but also to perform simulations of specific conditions involving structural states, policy intervention requirements, and parameters in order to forecast future trends. The greatest advantage of system dynamics is the elimination of experimental risk. Because decision-makers do not face cost increases due to possible policy failure when employing system dynamics, they can readily test various types of policies, and can continuously retest the assumptions implicit in each process step. By determining how to move the fulcrum in order to appropriately manipulate leverage, the cost of trial and error can be reduced and unnecessary investment avoided. 3.1.3 Basic model construction concepts System dynamics seeks to investigate real-world system. Systems are composed of interactions and mutual dependences, and a system may be a constituent part of a larger system to which it is subordinate. Because of this, a single system cannot be all-encompassing; a boundary must be marked out for it. When defining the scope of a system, in order to simulate real-world phenomena, system dynamics employs the two basic elements of stock and flow, which express how the quantities of things increase or decrease during dynamic processes. All dynamic system models must therefore take these two concepts as their basis if they are to effectively simulate the behavior of real systems. In addition, stock and flow are established under causal relationships and feedback loops, and causal relationships and feedback loops must be present in order to construct the elements of stock and flow. 3.1.4 Model construction steps According to Sterman (2000), while the modeling process possesses a high degree of originality, and each person will have a different modeling style and method, successful modelers must abide by identical successful procedures. Sterman emphasizes that these procedures do not necessarily involve linear steps, rather a repeated process of correction. New ideas generated in any step may influence other steps, giving rise to new hypotheses and models. 1. Define problem boundaries: Find problems of interest and major variables, and determine time units that can express the behavioral structure of the main variables. 5 .

(6) 2.. 3. 4. 5.. Establish dynamic hypotheses: Use an endogenous point of view concerning the influencing behavior of the feedback structure to establish hypotheses concerning the chief problems of concern, and establish a causal feedback diagram. Establish a system dynamics model: Establish a model that can be used to test the correctness of the hypotheses. This model must express the problem's characteristic structure and decision-making rules. Repeated model testing: Repeatedly test the model to determine its sensitivity and reality, and thereby obtain the model with the best credibility and validity. Perform policy design and assessment: Use the model to test decisions that may occur in the real world, and provide the results to decision-makers as a basis for decisions.. 3.2 R&D center system model 3.2.1 System model architecture The basic concepts of system dynamics imply that, since a system's boundaries must be defined, an appropriate choice of factors influencing the real system must be made when constructing a model architecture; key factors actually affecting the system must be taken into consideration, but factors with little influence on system behavior may be ignored. This study only investigated four major location advantages in looking at MNEs’ R&D centers in Taiwan, and did not take into consideration economic indicators such as exchange rates, interest rates, GDP, or the unemployment rate. As a result, macroeconomic environmental factors are not taken into consideration in this study's dynamic models. Taiwan’s four major types of location advantages were determined via an examination of the literature: local talent pool, local industry clusters, local research-based networks, and local government support. This study takes these four major advantages as premises, and constructs a model of the relationship between the four advantages and R&D centers (Fig. 1). This model is then used to explore how these advantages increase the revenue of MNEs’ R&D centers in Taiwan, the R&D capability of the parent companies, and the R&D centers’ willingness to continue to increase investment in R&D in Taiwan.. Subsidies from Taiwan government. Expatriate R&D personnel. Locally-hired R&D personnel. Internal R&D. MNEs’ R&D centers in Taiwan. R&D collaboration. New. Output. products. value. Research institutes. Firms. Universities. Figure 1 Framework of operations of MNEs’ R&D centers in Taiwan. 3.2.2 Data collection In order to enhance Taiwan’s R&D competitiveness, the government of Taiwan implemented the “Multinational Innovative R&D Centers in Taiwan" (MRDC) Program” and urged to encourage MNEs to establish R&D centers in Taiwan. In system dynamics, data is needed to perform simulations and assess models. 6 .

(7) The data used in this study was obtained mainly from the MOEA's MRDC Program database, and any data later found to be lacking was obtained through interviews with high-level managers of MNEs’ R&D centers in Taiwan. Statistics from the MRDC Program database revealed that 38 foreign R&D centers had an average of 62 personnel. Other relevant statistical data is shown in Table 1 and Table 2. Table 1 Home countries of MNEs’ R&D centers in Taiwan. Home country US Japan Germany Netherland. Home country Singapore Italy Great Britain Sweden. 21 6 3 2. 2 2 1 1. Source: MRDC Program database, MOEA, Taiwan Table 2 Start years of MNEs’ R&D centers in Taiwan. Start year 2002 2003 2004 2005. Start year 2006 2007 2008. 5 8 9 6. 1 5 4. Source: MRDC Program database, MOEA, Taiwan. With regard to the interviews, since most information concerning MNEs’ R&D centers in Taiwan is confidential, high-level managers at the R&D centers of Microsoft and DuPont were asked using open questions for the information needed for modeling. Apart from collecting needed data, the interviews also confirmed whether the model conformed to reality.. Model Construction, Testing, Simulation, and Policy Analysis This section describes the construction of the model in this study, testing of the constructed model, and simulation of future government policy strategies. Model construction is explained in the first part of this section, model testing intended to increase model credibility is explained in the second part, and basic simulations are described in the third part. When basic data was lacking, reasonable assumptions were used in simulations to observe system trends. The fourth part consists of policy simulations; in accordance with a review of the literature, the model and possible future government policy strategies were used to perform simulations, and the possible results examined.. 4.1 Model construction 4.1.1 Analysis of causal loops in the model This study took MNEs’ R&D centers in Taiwan as its focus, and investigated the effect of the causal loops established on the basis of the four major advantages of Taiwan specified in the second section on the functioning of R&D centers. R&D centers' operating models encompass both internal R&D and R&D collaborations. With regard to internal R&D, after receiving funding from the MRDC Program, multinational firms can hire outstanding local talent from Taiwan, and the parent company can also send R&D personnel to take part in R&D activities. The MRDC Program can therefore boost the revenue of MNEs’ R&D centers in Taiwan. With regard to R&D collaborations, because MNEs’ R&D centers in Taiwan receive subsidies from the government, and because of the advantages of Taiwan's industries, universities, and research institutes--such as a complete electronics industry value chain, more than 150 universities and colleges to provide the manpower needed for industrial development, and research institutes such as the Industrial Technology Research Institute (ITRI) and Institute for Information Industry (III), which have pioneered the development of many new technologies--MNEs’ R&D centers in Taiwan allocate a certain percentage of their funds for collaborations with domestic industries, universities, and research institutes. The many product innovations and improvements 7 .

(8) that have emerged from collaborative research have done much to increase the revenue of MNEs' R&D centers. 4.1.2 Explanation of model-related factors This study compiled Taiwan's four major location advantages and innovation performance assessment indicators from the literature. (Table 3) Table 3 Taiwan's location advantages and innovation performance assessment indicators. Location advantages Local government support. Local talent pool. Local industry clusters. Local research-based networks Innovation performance. Assessment indicators Protection of intellectual property rights Enhancement of Patent Law Tax deduction R&D subsidy from government Volume of R&D talent Quality of R&D talent Talent for specific industry field Low cost of R&D talent Collaborative R&D Technology clusters and R&D capability of host country Possess of unique technologies Collaborative R&D Inter-flows of knowledge Offering of technical services New product development Patent Publications Process improvement Facts/knowledge. Source: Synthesized by this study. The model in this study assigns a key role to total R&D budget, which is assumed to constitute government subsidies and a certain percentage of the R&D center's product sales revenue. As far as this budget is concerned, MNEs’ R&D centers in Taiwan determine how their funding is allocated, such as what percentage of the budgets will go for personnel, collaborative research, rental expenses, and equipment costs. Because this study sought to develop a model in line with the internal/external R&D perspective, emphasis was placed on R&D centers' personnel and collaborative research expenses. With regard to internal R&D, after a multinational firm's R&D center in Taiwan appropriates funds for personnel expenses, it will then decision how many new personnel to hire that year. With regard to external R&D, after a multinational firm's R&D center in Taiwan appropriates funds for collaborative research expenses, it will then decide on the basis of its development strategy what percentages of its collaboration budget to assign to collaboration with industry, universities, and research institutes. In summary, this model is based on cash flow, and employs number of products as a proxy for the performance of a multinational firm's R&D center in Taiwan. While this study considered the conceptual indicators in Table 4-01, since system dynamics requires measurable indicators, this study compiled four major location advantages and innovation performance assessment indicators from the literature, and selected one indicator to assess each location advantage. The following is a detailed description of this approach: (i). Local government support The MOEA drafted the "Multinational Innovative R&D Centers in Taiwan" (MRDC) program in 2002 in order to increase Taiwan's competitiveness by promoting the establishment of innovative R&D centers in Taiwan by foreign enterprises. This program has sought to achieve tangible benefits by recruiting overseas technical manpower, promoting R&D investment, boosting the linkage between scientific research and industrial technology in Taiwan, and helping Taiwan secure international R&D resources. Because the MRDC program provides subsidies to MNEs’ R&D centers in Taiwan, this study used "subsidies from the Taiwan government" as an indicator for local governments support. (ii) Local talent pool 8 .

(9) In-house R&D requires outstanding R&D manpower, and outstanding R&D manpower includes both locally-hired personnel and personnel supplied by headquarters. With regard to locally-hired personnel, R&D centers tend to obtain their local talent from subsidiaries in Taiwan. Newly-hired personnel are a minority of these personnel, and turnover is low. Locally-hired personnel make a significant contribution to R&D centers' product development work, however. Since the personnel qualifications and fields in Table 4-01 are somewhat abstract, and system dynamics requires the use of measurable indicators, this study employed "number of locally-hired R&D personnel" as an indicator for local talent pool. In addition, because of this study's focus on cash flow, the "number of locally-hired R&D personnel" is derived from "R&D personnel expenses" and "annual salary per R&D personnel." (iii) Local industry clusters In order to minimize R&D investment risk, obtain information on the newest technology development trends, develop new technologies and new products derived from new technologies, and improve existing products, R&D centers may collaborate with other local firms, universities, and research institutes. But because collaboration with other local firms is a conceptual indicator, and is difficult to quantify in system dynamics, the study used "expense of collaboration with local firms" as an indicator for area industry clusters. As a consequence, due to the central role of cash flow in this study, the indicator "expense of collaboration with local firms" was derived from "expense of local R&D collaboration" and "expense of collaboration with local firms as a percentage of total collaborative research expenses." (iv) Local research-based networks University instructors are firms' chief partners in industry-academic collaboration, which enables firms to take advantage of schools' existing equipment and plentiful research manpower, and achieves greater economic efficiency by coupling academic basic research with industrial applications research. The primary goal of industry-academic collaboration is to achieve a closer fit between academic theory and corporate needs, while boosting industry's technological standards. Collaboration with research institutes and academic institutions allows firms to use those units' R&D manpower and R&D equipment, which can save a vast amount of expenses. Since collaboration with research institutes and academic institutions is a conceptual indicator, and cannot be readily measured in system dynamics. This study therefore derived "expense of collaboration with research institutes" and "expense of collaboration with universities" from "expense of local R&D collaboration," "expense of collaboration with local research institutes as a percentage of total collaborative research expenses," and "expense of collaboration with local universities as a percentage of total collaborative research expenses." (v) Innovation performance Chang et al. (2005) found that MNEs’ R&D centers in Taiwan do not necessarily focus on basic research and disruptive R&D; rather, they emphasize short- and medium-term R&D activities, which primarily consist of new product development work. Since MNEs’ R&D centers in Taiwan are chiefly concerned about product development, and are not exploratory R&D centers, this study consequently used "number of new products" as the assessment variable for innovation performance. These new products may be either all-new inventions or innovations, or modifications of existing products, and may have been developed either internally or through external collaboration. In the case of internal R&D, because personnel assigned by an overseas corporate headquarters and the R&D center's existing and newly-hired R&D personnel all contribute significantly to the R&D center's product development, this study took "number of new products developed through internal R&D" as an assessment variable. Nevertheless, since not all R&D center personnel support the internal development of products, this study added the variable "internal R&D ratio," and derived the "number of new products developed through internal R&D" as the multiplication product of "internal R&D ratio" and "total number of R&D center personnel." As for external R&D, collaboration between MNEs’ R&D centers in Taiwan and other local firms, universities, or research institutes does not always result in the successful development of products, and the goal of some collaborative projects is not to develop products, but rather to develop technologies or provide technical support. Because collaboration for the purpose of product development makes a significant 9 .

(10) contribution to the product development efforts of MNEs’ R&D centers in Taiwan, this study took "number of new products developed through R&D collaboration" as an assessment variable. But since not all collaborative projects necessarily yield new products, the number of new products developed through R&D collaboration was derived as the product of "number of collaborative projects" and "commercialization ratio." Furthermore, in keeping with this study's cash flow orientation, the number of collaborative projects was derived by dividing the "expense of local R&D collaboration" by the "average expense of each collaborative project." Final, the successful development of new products either internally or through collaborative projects enables MNEs’ R&D centers in Taiwan to sell more new products, which increases the R&D centers' revenue. As a consequence, this model takes "revenue derived from new products" as an assessment variable. 4.1.3 System dynamics model functions (i) Level functions Please refer to the Appendix for the detailed structure of the model as a whole. The model is chiefly composed of two levels: "total R&D budget" and "total number of R&D center personnel." The equations used to obtain these levels are explained as follows: The first level of the model is total R&D budget, which is influenced by "subsidies from the Taiwan government," R&D center "output value," and R&D center "R&D budget as a percentage of output value." Since the model in this study is developed with an internal/external R&D perspective, emphasis is placed on "R&D center personnel expenses" and "expense of local R&D collaboration." Because the simulation was run starting in 2003, the initial value of the total R&D budget was the total R&D budget in 2003 (=28). The mathematical model is as follows: Total R&D budget = R&D budget - R&D expenditures Initial value = 83.43 (NT$1 million) (Equation 1) R&D budget = subsidies from the Taiwan government (time) + output value * R&D budget as a percentage of output value (Equation 2) R&D expenditures = 28 (Equation 3) The second level of the model is "total number of R&D center personnel", which is influenced by "locally-hired R&D personnel" and "expatriate R&D personnel." The initial value was the number of personnel in 2003 (=16). With regard to the outflow of R&D personnel, the interviews revealed that R&D center personnel had a low turnover rate, and since most of turnover consists of the transfer of personnel between R&D centers and subsidiaries, actual turnover is extremely low. This study included all personnel leaving an R&D center, regardless of whether they were transferred to a subsidiary, within the outflow of R&D center personnel. This percentage was roughly 10% of the total number of R&D center personnel (16*10%=1.6, rounded to 2). The mathematical model is as follows: Total number of R&D center personnel= personnel inflow - personnel outflow Initial value = 16 (Equation 4) Personnel inflow = locally-hired R&D personnel + expatriate R&D personnel (Equation 5) R&D personnel outflow = 2 (Equation 6) (ii) Explanation of constant setting The constants used in the model are for the variables "R&D budget as a percentage of output value," "expatriate R&D personnel," "annual salary per person," "average expense of each collaborative project," "average number of products developed per person," and "internal R&D ratio." Because the values of these variables remained very similar from 2003 to 2008, they are set as constants according to the actual values. 10 .

(11) Because there was no statistical information concerning final variable—"internal R&D ratio"—this variable was set on the basis of the results of the interviews. (iii) Explanation of auxiliaries This study's data period was from 2003 to 2008. During these six years, the numbers of MNEs’ R&D centers in Taiwan submitted applications for R&D project subsidies were 8, 9, 6, 1, 5, and 4 respectively. It can be seen that the number of centers was high at first. But because most applications in 2006 and 2007 were made at the end of the year, the total amount of subsidies fell sharply in 2007, and only began to rebound in 2008. In addition, since most projects implemented by R&D centers are three-year projects, many R&D centers may still be implementing past projects, and therefore cannot apply for new project subsidies. Furthermore, since the interviews revealed that the government plans to increase its science and technology budget by 10% annually, it was assumed in the simulation for 2009 through 2013 that the government subsidy amount will grow at a 10% annual rate during these years. Secondly, with regard to expense of collaboration as a percentage of total collaborative research expenses, the percentage during most of the six years ranged from 20% to 25%. Because the value did not change much during the remaining years, the value for 2008 was employed in the simulation for 2009 through 2013. Thirdly, with regard to expense of collaboration with local universities as a percentage of total collaborative research expenses, due to the fact that collaboration between R&D centers and universities may take the form of either product development or technological research, this percentage reflects R&D centers' strategies, and may change unpredictably. As a consequence, the value for 2008 was employed in the simulation for 2009 through 2013. Fourthly, with regard to expense of collaboration with local research institutes as a percentage of total collaborative research expenses, because most collaborative projects with research institutes seek to develop all-new products or technologies, the increase in this percentage indicates that R&D centers are increasingly pursuing this option. This percentage exhibits particularly large increase in 2008 due to the fact that having completed three-year projects and gained a better understanding of Taiwan's industry environment, many R&D centers began committing a greater share of their budget to the development of all-new products or technologies in their next round of projects. As a consequence, the value for 2008 was employed in the simulation for 2009 through 2013. Fifthly, with regard to expense of collaboration with local firms as a percentage of total collaborative research expenses, the interviews indicated that most of these projects sought to improve existing products. Although early R&D center projects were of this type, as the R&D centers increasingly focused on the development of all-new products or technologies, this percentage gradually leveled off and fell. The relatively sharp drop in 2008 was chiefly attributable to the squeezing out effect as R&D centers increased their collaborative projects with research institutes. As a consequence, the value for 2008 was employed in the simulation for 2009 through 2013. Sixthly, with regard to the commercialization ratio of collaborative projects, the ratio remained high throughout all six years. As previously mentioned, MNEs’ R&D centers in Taiwan are gradually placing more emphasis on the development of all-new products and technologies, and this trend caused the increase in the commercialization ratio to slow dramatically in 2008, although it remained quite high. As a result, the value for 2008 was employed in the simulation for 2009 through 2013. Seventhly, with regard to the average value generated by each product, the slowing of the commercialization ratio may have reduced output value, which reduced the average value generated by each product. As a result, the value generated by each product in 2008 was lower than during the previous years, although total value was certainly still considerable. As a consequence, the value for 2008 was employed in the simulation for 2009 through 2013. Finally, with regard to R&D personnel expenses as a percentage of the R&D centers' total budgets, the interviews revealed that because personnel at MNEs’ R&D centers in Taiwan have a low turnover rate, there are few newly-hired personnel, and the total number of personnel at the R&D centers remained roughly constant. But since MNEs’ R&D centers in Taiwan are placing growing emphasis on the development of all-new products and technologies, the total number of R&D personnel began increasing in 2008. As a result, the value for 2008 was employed in the simulation for 2009 through 2013.. 11 .

(12) 4.2 Model verification System dynamics models chiefly focus on long-term trends, and do not place particular emphasis on the value of any one variable. Although model parameters and model verification consequently do not require strict statistical testing, there are methods for determining the validity of system dynamics models. Because past research on the topic investigated in this study did not employ system dynamics, and because the research subjects were not in existence for a long time, there were certain limitations on data access. Nevertheless, because model structure is very important for this research topic, the parameter value test, unit consistency test, and extreme value test were used to verify the model and confirm its validity. 4.2.1 Parameter value test This study compared the model structure and variables with the real system as a parameter value test. Since the chief goal of this study was to analyze the effect of Taiwan's location advantages on the performance of MNEs’ R&D centers in Taiwan, the model's structure and variables had to reflect the real system's structure and parameters. As for the overall structure of the model, the interviews indicated that support from the government was the main reason multinational enterprises decided to establish R&D centers in Taiwan. Due to support from the government, multinational enterprises' headquarters felt confidence in Taiwan's investment environment, and were therefore willing to let their subsidiaries in Taiwan establish R&D centers. Apart from this, such advantages as Taiwan's industry clusters, the number and effectiveness of local research institutes, and the depth of the local talent pool also encouraged multinational enterprises to locate R&D centers in Taiwan. The interviewees also expressed that all of these advantages also affected the performance of their R&D centers in Taiwan. In addition, the interviews further suggested that the multinational enterprises' decisions to establish R&D centers in Taiwan were not primarily influenced by such macroeconomic factors is exchange rate, interest rate, GDP, or unemployment rate. This study's model did not take macroeconomic factors into consideration, and there was consequently little difference between the system architecture and the model. Models are simplifications of the real world, and not the real world itself. As a result, not all the variables included in this model actually affected the performance of MNEs’ R&D centers in Taiwan. On the other hand, some of the variables proposed in this study constituted important influencing factors--such as government subsidies, collaboration with other local firms, universities, and research institutes, and the deep local talent pool--and these variables all appeared in the model. In summary, the structure of this model was similar to the operating structure of existing foreign R&D centers in Taiwan, and included important measurable variables actually influencing R&D centers. Because of this, the model met the test's validity standard. 4.2.2 Unit consistency test Because units were not inconsistent during the modeling and simulation process, the unit consistency of the model did not exhibit any problems. 4.2.3 Extreme value test Although this study lacked real historical data allowing the recurrence of behavior to be tested, the model structure was inspected at all times during the modeling process, and extreme value testing of model behavior was performed to determine whether simulation results were stable. Extreme value test involved substitution of extreme values for different variables, and observing whether the model had any abnormal responses. After testing the maximum and minimum values of all variables, it was found that the extreme values did not influence the model's behavioral trends.. 4.3 Basic simulation Since there was no complete set of historical data to guide the basic simulation, the basic simulation in this study was based on initial value settings corresponding to reality. The simulation took years as units of time, and was conducted for six years. The results of the basic simulation are shown in Fig. 2, Fig. 3 and Fig. 4. It can be discovered from the simulation results that government subsidies and the R&D budgets and output values of MNEs’ R&D centers in Taiwan did not have identical trends during the period of 2003-2006. 12 .

(13) The relatively large government subsidies from 2003 to 2006 can be attributed to the relatively large number of early applicant firms, while falling subsidies after 2007 was due to a reduction in the number of R&D centers applying for project subsidies, which caused the total amount subsidies to gradually decrease. This decrease was not due to the government's reduction of the amount of subsidies.. Figure 2 Government subsidies and performance of R&D centers. The inconsistency between the trend of government subsidies and that of R&D center budget and output value can be attributed to the length of time needed for R&D centers to move from the preparatory stage to the operating and product sales stage. As a consequence, although government subsidies had a delayed effect on the performance of MNEs’ R&D centers in Taiwan, the simulation results suggested that subsidies still had a certain impact on R&D center performance. (Fig. 2). Figure 3 New products developed through R&D collaborations and different types collaborations. With regard to R&D collaborations, the simulation results revealed that there was a positive correlation between expense of collaboration with research institutes, expense of collaboration with local firms, expense of collaboration with local universities, and number of new products. Most examples of collaboration with local firms sought to improve existing products. The trend of this variable over the period of 2007-2008 requires a special explanation: The decline in the expense of collaboration with local firms in 2008 was caused by the fact that, having completed three-year projects and gained a better understanding of Taiwan's industry environment, many R&D centers began using funds from their next round of projects for new purposes, such as for the development of all-new products and technologies. In addition, one R&D center placed special emphasis on collaboration with research institutes, which caused the expense of collaboration with local firms to drop 13 .

(14) sharply in 2008. Since collaboration with universities and research institutes mostly takes the form of technological R&D, but R&D centers tended to focus their efforts on the improvement of existing products prior to 2007, there were relatively few cases of collaboration with universities and research institutes during that period. But as R&D centers completed their first three-year projects, they began to increase their collaboration with universities and research institutes, and expenditures on collaboration with universities rose significantly in 2008. (Fig. 3). Figure 4 New products developed through internal R&D and total R&D personnel. With regard to internal R&D, the simulation results reveal a positive correlation between the total number of R&D center personnel and number of new products. Because of their low turnover rates, R&D centers tend to maintain relatively stable numbers of personnel. Nevertheless, because R&D personnel make a significant contribution to the R&D centers' product development and output value, it is projected that the R&D centers will hire more new R&D personnel in the future. (Fig. 4). 4.4 Policy simulation This study used system dynamics to investigate the relationship between Taiwan's four major location advantages and the performance of MNEs’ R&D centers in Taiwan. In this section, specific variables are deleted to determine which advantage is the most important to the R&D centers, and the results of these simulations are then used to adjust the assumptions in section 4.1 before performing a new simulation. 4.4.1 Deleting specific variables Individual location advantage assessment variables are deleted in order to determine which advantage is the most important to MNEs’ R&D centers in Taiwan. The results of these simulations indicate that the lack of government subsidies would have the greatest effect on the performance of MNEs’ R&D centers in Taiwan. (Fig. 5) 4.4.2 Increasing the amount of government subsidies According to the interviews, the government plans to increase its science and technology budget by 10% annually. Because the simulations in the previous section revealed the importance of government subsidies to MNEs’ R&D centers, this study increased this budget growth rate to 30% to observe whether the performance of the R&D centers would change significantly. The simulation results suggest that an increase in government subsidies would have a significant effect on MNEs’ R&D centers in Taiwan, and would provide the R&D centers with more resources needed to develop new products. (Fig. 6). 14 .

(15) . Delete internal R&D. Delete government subsidies. Delete collaboration with local firms. Delete collaboration with local research-based networks. Figure 5 Comparisons of deleting specific location advantage. Figure 6 Increase the amount of government subsidies. 15 .

(16) 4.4.3 Policy simulation summary Because simulations involve temporal changes, MNEs’ R&D centers in Taiwan may be influenced by variables not considered in the policy simulation process. Nevertheless, since this study emphasized Taiwan's four major location advantages, its policy simulations only investigated the effect of the four major location advantages on the R&D centers' innovation performance. With regard to policy simulations, this study sequentially deleted the variable for each of the four major location advantages in order to determine which advantage is the most important to MNEs’ R&D centers in Taiwan. The simulation results revealed that only eliminating government subsidies would cause the performance of the R&D centers to decline. In contrast, eliminating Taiwan's advantages of industry clusters, a deep talent pool, and local research institutes only caused the R&D centers to exhibit ordinary, and not outstanding, performance. The two policy simulation aspects confirmed the importance of government subsidies to the performance of MNEs’ R&D centers in Taiwan. The results of this study's simulated increase in government subsidies similarly displayed that faster subsidy growth would have a significant influence on the performance of the R&D centers.. 4.5 Summary One of advantages of a model is that it can help decision-makers clearly see the crux of the issue, while also providing a policy laboratory in which the model structure or parameters can be adjusted to see effects of different policy changes. A model can thus avoid the waste of resources caused by an erroneous policy. Taking the model in this study as an example, government subsidies facilitate the operations of MNEs’ R&D centers in Taiwan by giving the R&D centers more resources with which to collaboration with local firms, universities, and research institutes, and enabling them to hire new R&D personnel. These two pathways can both boost the number of new products developed by the R&D centers. Nevertheless, although the favorable simulation results suggest that the future will be bright, the 2008 values were used for the graphical functions, and these values can be expected to change in the future. This may influence the number of new products developed by MNEs’ R&D centers in Taiwan. The results of policy simulations indicate that increasing government subsidies will have a significant influence on the performance of MNEs’ R&D centers in Taiwan. If the government hopes to improve the competitiveness of Taiwan's R&D environment, it can undertake policy planning and implementation in this area.. Conclusions. 5.1 Conclusions and policy implications Under the trends of R&D internationalization, Taiwan had successfully attracted MNEs to invest R&D resources due to advantages of local industry clusters, local talent pool, local research-based networks and local government support of the host country. This study attempts to explore impacts of Taiwan’s location advantages on innovation performance of the MNEs’ R&D centers in Taiwan by means of employing the system dynamics approach, and furthermore, to find out which location advantage exhibits the highest importance to innovation performance of MNEs’ R&D centers. After data collection process and computer simulations, the operations of MNEs’ R&D centers in Taiwan are identified, the factors which affect innovation performance of these R&D centers are clarified, and the influences of Taiwan’s location advantages on MNEs’ R&D centers are observed. There are two major findings. First, from Section 4, this study argues that all four major location advantages urge MNEs to set up their R&D centers in Taiwan and present positive impacts on new product development of MNEs’ R&D centers in Taiwan. Among the four major types of location advantages, local government support and local talent pool will urge MNEs’ R&D centers to hire R&D personnel for internal R&D activities. In addition, local government support, local research-based networks and local industry clusters result in aggregation of collaborative research expenses, thus make influences on outputs of external R&D 16 .

(17) activities. Second, according to the simulation results, local government support plays the most important role compared to another three location advantages. The finding conforms to the results of interviews with high-level managers of MNEs’ R&D centers. That is, the government’s cash grant stands for supports from the host country, thus strengthen willingness for MNEs to set up R&D centers in Taiwan. As for the policy implications, the simulation results indicates that local government support presents the most influential effect, Taiwan government could consider to further strengthen incentives such as tax deductions and intellectual property right protection or lift up the subsidized percentage. Furthermore, the simulation results also indicate the importance of collaboration with local industries, universities and research institutes for MNEs’ R&D centers. Local industries and research-based networks could also further consider ways to enhance R&D capability. The government could offer support to such enhancement. At last, local talent pool plays quite important role for attracting MNEs’ overseas R&D. The government could consider to further enhancement of quality of local R&D personnel, for instance, through training programs or tighter linkages between industries and universities.. 5.2 Further research A system dynamics model can simplify the real world. Therefore, it cannot perfectly fit the operations in the real world. A system dynamics model needs to be continuously testified and verified. This study suggests four further research issues. First of all, as for data collection, this study is limited to retrieve data from database of MOEA’s MRDC program and interviews with some high-level managers of MNEs’ R&D centers due to confidentiality concerns. Not all R&D centers which get cash grants from the government in Taiwan are accessed. This study suggests a longer-term observation and survey to make the model parameters more realistic. Secondly, the system dynamics model in this study could be a bit more complicated and fit the real world. For instance, to divide the government’s cash grant into more detailed streams and to explore impacts of different subsidized items on innovation performance of MNEs’ R&D centers in Taiwan. Thirdly, this study doesn’t involve effects of R&D talent outflow. And fourthly, this study only generates cash flows and positive feedback loops in the model. Further research is recommended to involve negative feedback loops by introducing human resource flows, technology flows, and information flows, etc.. 17 .

(18) Appendix A. +. Subsidies from Taiwan Gov't. RD expenditure. RD budget +. +. <Time>. <Time> RD budget as a percentage of output value. Expense of collaboration with local firms as a percentage of total collaborative research expenses. Expense of local RD collaboration. <Time>. + Expense of collaboration with local firms. <Time>. +. + Expense of collaboration with universities. Expense of collaboration with research institutes. Number of collaborative projects. <Time> Output value derived from new products. Annual salary per person. Expenses of local RD collaboration as a percentage of total RD budget. Expense of collaboration with local research institutes as a percentage of total collaborative research expenses. Average output value of each new product. +. 18 . Expatriate RD personnel +. RD personnel expenses as a percentage of total RD budget. Total RD budget. +. Locally hired RD personnel. RD personnel expenses. <Time>. +. Personnel inflow. Total number of RDC personnel. Expense of collaboration with local universities as a percentage of total collaborative research expenses. <Time> Number of new products developed through RD collaboration. Personnel outflow. + Number of new products developed through internal RD. Internal RD ratio. +. <Time> +. Commer cialization ratio Average expense of each collaborative project. Average number of products developed per person. Number of new products.

(19) References 1. 2. 3. 4.. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21.. 22. 23. 24.. 25. 26. 27. 28. 29.. Anand, J and Kogut, B (1997). Technological Capabilities of Countries: Firm Rivalry and Foreign Direct Investment. Journal of International Business Studies, Vol. 28, 445-465. Birkinshaw, J and Hood, N (2000). Characteristics of foreign subsidiaries in industry clusters. Journal of international business studies, first quarter, Vol. 31(1), 141-154. Brown, G and Sevenson, A (1998). Measuring R&D Productivity. Research Technology Management, Nov-Dec, Vol. 30(5), 11-15. Chang, YC, CT Shih and YL Wei (2005). Hooking up the local and innovation system: A study of MNE R&D Centers in Taiwan. Unpublished paper presented at Stanford SPRIE-STPRC workshop: Greater China Innovation Capabilities: Opportunities and Challenges, Tsing-Hua University, Beijing. Chiesa, V (1996). Managing the internationalization of R&D. IEEE Transactions on Engineering Management, Vol. 43(1), 7-23. Dunning, JH (1988). Multinationals, Technology and Competitiveness. London: Unwin Hymm. Florida, R (1997). The globalization of R&D: results of a survey of foreign-affiliated R&D laboratories in the USA. Research Policy, Vol. 26(1), 85-103. Forrester, JW (1961). Industrial Dynamics. Cambridge Mass: The M.I.T. Press. Fors, G and M Zejan (1996). Overseas R&D by multinationals in foreign centers of excellence, Working paper no. 111, The Economic Research Institute, Estocolmo: Stockholm School of Economics. Gerybadze, A and G Reger (1999). Globalization of R&D: Recent Changes in the anagement of Innovation in Transnational Corporations. Research Policy, Vol. 8(2-3), 251-274. Guellec, D and B van Pottelsberghe (2001). The internationalization of technology analysed with patent data. Research Policy, Vol. 30, 1253–1266. Han, C. (2002) System Dynamics: A Key for Exploring Dynamic Complexity. Taipei: Charles C. Han. Hedlund, G and R Dag (1990). Action in heterarchies: New approaches to managing the MNC. In Managing the Global Firm, C. A. Bartlett, Doz, Y. and Hedlund, G. (Eds.), Orlando, Florida: Harcourt. Hegde, D and D Hicks (2008). The maturation of global corporate R&D: Evidence from the activity of U.S. foreign subsidiaries. Research Policy, Vol. 37, 390-406. Hood, W and S Young (1982). FIND Dunning J.H. Multinationals, Technology and Competitiveness. London, Unwin Hymm, Chap7. Keegan, WJ (1995). Global Marketing Management. 5th Ed, New Jersey: Prentice-Hall. Kogut, B and SJ Chang (1991). Technological capability and Japanese foreign direct investment in the United States. The Review of Economics and Statistics, Vol. 73(3), 401-413. Kuemmerle, W (1999a). The drivers of foreign direct investment into research and development: an empirical investigation. Journal of International Business, Vol. 30(1), 1-24. Kuemmerle, W (1999b). Foreign investment in industrial research in the pharmaceutical and electronics industrial-results from a survey of multinational firms. Research Policy, Vol. 28(23), 179-193. Kumar, N (2001). Determinants of location of overseas R&D activity of multinational enterprises: the case of US and Japanese corporations. Research Policy, 30, 159-174. Le Bas, C and P Patel (2007). The determinants of home-base-augmenting and home-base-exploiting technological activities: some new results on multinationals locational strategies. Science and Technology Policy Research, Vol. 164. Nachum, L (2000). Economic Geography and the Location of TNCs: Financial and Professional Service FDI to the USA. Journal of International Business Studies, Vol. 31(3), 367-385. Paoliz, M and S Guercini (1997). R&D internationalisation in the strategic behaviour of the firm. Steep discussion paper no. 39, Sussex: Science Policy Research Unit, University of Sussex. Pavitt, K and P Patel (1999). Global corporations and national systems of innovation: who dominates whom?. In Innovation Policy in a Global Economy, Archibugi, D, J Howell and J Michie (ed.), pp.94–119. Cambridge University Press, Cambridge. Ronstadt, RC (1978). International R&D: The Establishment and Evolution of R&D Abroad by Seven U.S. MNCs. Journal of International Business Studies, Spring-Summer, 7-74. Serapio, J and H Donald (1999). Globalization of industrial R&D: an examination of foreign direct investment in R&D in the United States. Research Policy, Vol. 28, 303-316. Senge, PM (1990). The fifth discipline: the art and practice of the learning organization. Doubleday Press, New York. Sterman, JD (2000). Business Dynamics: Systems Thinking and Modeling for a Complex World. Boston, MA: McGraw-Hill. Taggart, JM (1998). Strategy shift in MNC subsidiaries. Strategic Management Journal, Vol. 19, 663-681. 19 .

(20) 30. Westney, E (1992). Organizational change and the internationalization of R&D, In Transforming Organization, Kochan TA and M Useem (eds.), New York: Oxford University Press. 31. Zander, I (1999). How do you mean “global”? An empirical investigation of innovation networks in the multinational corporation. Research Policy, Vol. 28, 195-213.. 20 .

(21)