Knowledge for Growth Industrial Research & Innovation (IRI)

Transcription

1 Knowledge for Growth Industrial Research & Innovation (IRI) R&D without patents? Foreign R&D activities of European Enterprises in Central and Eastern Europe and Asia: a Patent Analysis CONTRIBUTED PAPER FOR THE 2007 CONFERENCE ON CORPORATE R&D (CONCORD) Structure and Strategy for Corporate R&D File name: Author: Status: Last updated: Organisation: dachs_weber v1.doc Bernhard Dachs, Matthias Weber <Draft> <DATE> Austrian Research Centers ARC, Department of Technology Policy, Vienna, Austria Page 1 of 20

2 TABLE OF CONTENTS Abstract: Background Patents as a Measure for the Internationalisation of R&D Data Empirical Results Descriptive Analysis Multivariate Analysis Operationalisation Estimation Results R&D without Patents?...16 Page 2 of 20

3 Abstract: There is a growing concern in European S&T policy debates about the relocation of R&D activities from Western Europe to Central and Eastern Europe and Asia. Newspaper headlines and magazine reports highlight that European MNEs increasingly set up R&D units in China, India or other low wage countries. As a consequence, many see a danger that this internationalisation of R&D may erode technological competences in the home countries. This paper examines if there is empirical evidence that would back up these fears. We employ patent data to analyze the R&D activities of European (EU27) enterprises outside of their home countries. To cover the most recent developments, our data set only includes patent applications received by the European Patent Office (EPO) between January 2000 and Mid 2007, regardless if the patent has been granted or not. Similar studies that employ granted patents usually cannot deliver recent results because of considerable time lags between patent application and grant. Out analysis indicates that EU27 enterprises indeed perform a considerable amount of R&D activities outside of their home countries. Central and Eastern Europe, the CIS or Asia, however, still account for only small fraction of overseas R&D activities of EU27 enterprises. Despite large investments in RTD reported by various sources, these countries yield only few inventions applied for a patent by European enterprises so far. The most important R&D locations of European enterprises besides their home countries are still the EU15, EFTA countries and North America. The current internationalisation of R&D can therefore be characterized by a deepening of the intra European division of labour in R&D and a strengthening of R&D activities of European enterprises in North America. We argue that this pattern can be explained by the economic and scientific conditions host countries offer to European enterprises. A multivariate regression analysis of the determinants of cross border patents provides support for this argument. The number of cross border patents in a host country increases with its relative and absolute market size and with higher levels of overall R&D intensity in the country. Distance between home and host country is negatively related to the number or cross border patents. Many potential host countries including China and India still offer too few incentives that may outweight the disadvantages of distance. The positive correlation between income per capita and the extend of foreign R&D activities may also point to the fact that low wages are overrated as the main determinant of foreign R&D. Moreover, R&D productivity in terms of the ratio of patent inventions to industry financed R&D expenditure is still considerably lower in China and India than in the EU15 or in North America. Page 3 of 20

4 1 Background An important aspect of the ongoing integration of the world economy is the internationalisation of research and development. Enterprises not only produce and sell, but also increasingly develop goods and services outside of their home countries. The internationalisation of R&D is not a new phenomenon, but has gained momentum in recent years. A considerable part of this acceleration can be attributed to the activities of multinational enterprises (MNEs). Enterprises increasingly organize their research and development (R&D) in a more decentralized way with geographically distributed R&D and innovation units and a vivid MNE internal exchange of knowledge and information (Zanfei 2000; von Zedtwitz and Gassmann 2002). The literature discusses two basic motives of R&D internationalisation (Kuemmerle 1999; Kumar 2001; le Bas and Sierra 2002; von Zedtwitz and Gassmann 2002; Narula and Zanfei 2005): First, overseas R&D activities of enterprises focus on adapting existing products to local needs and supporting local production for the host markets ( assetexploiting strategy). The internationalisation of R&D is therefore a by product of the internationalisation of production and sales. Second, enterprises wish to create new knowledge and technologies abroad, which is facilitated by excellent knowledge infrastructure and the presence of knowledge spillovers from universities and enterprises in the host country ( asset augmenting strategy). India and The People s Republic of China (referred to as China in the paper) have come into focus as host countries for the R&D activities of North American and European MNEs in particular. Jiatao Li and Deborah Yue report 378 foreign R&D labs in China, a considerable number of them presumably operated by European MNEs (Li and Yue 2005, Table 1). Von Zedtwitz (2004) identifies 199 distinct foreign owned R&D centres established by foreign companies in China in early 2004 (most of them set up in 2002 and 2003, however). He cites other sources that speak of more than 400 foreign owned R&D units in China (von Zedtwitz 2004, Table 1). Most recently, Walsh (2007) cites Chinese sources that speak of as many as 750 foreign owned R&D centres. We find similar reports for other parts of the world, for example for Central and Eastern Europe. Dalia Marin (2004) did an extensive survey of German and Austrian affiliates in Central and Eastern Europe, South Eastern Europe, and the Commonwealth of Independent States (CIS). She finds that 15% of all employees at German and 12% at Austrian affiliates in these countries work in R&D. Both Austrian and German subsidiaries in this region already exhibit a higher skill intensity than their parent companies (Marin 2004, Tables 10 and 11). This paper tests some assumptions which have come up in these recent public debates of internationalisation. Today, many people fear that overseas R&D activities of European MNEs may be not an expansion, but a substitute for similar activities at home country. Well trained engineers and scientists in India, China and Central and Eastern Europe offer comparable expertise to that of their US or Western European colleagues at considerably lower cost. Page 4 of 20

5 As a consequence, high skilled jobs in design, innovation, research and development are regarded to be in danger of being relocated. A survey by the Association of German Chambers of Industry and Commerce (DIHK) reports that a third of all German enterprises already invest in research and development activities abroad and nearly half of these enterprises have in turn reduced their R&D efforts in Germany (Rose and Treier 2005, p. 8 and 10). We will examine the extend European MNEs perform R&D activities outside of their home countries with patent data. Our approach is simple: R&D activity manifests itself in patents. If European enterprises perform extensive R&D activities outside of their home countries, the results of such activities should be reflected in a large number of cross border patents applied for by these enterprises with residents from countries other than their home countries as their inventors. Moreover, we examine the determinants of R&D internationalisation with multivariate regression analysis. The analysis covers both, the old and the new EU member states, referred to as EU15 and EU12 in the text. 2 Patents as a Measure for the Internationalisation of R&D Authors have employed various methods to measure the internationalisation of R&D. One of the most fruitful is the usage of patent data. A patent is an intellectual property right issued to protect a technological invention. By granting these rights to inventors, the patent system enhances the approbability of inventions and therefore stimulates the creation of novelty (Griliches 1990). But the patent system is not only a legal institution. Patents are widely used a indicators of research and development activities at the level of industries and individual firms (Jaffe and Trajtenberg 2002). One feature makes patents very useful for studies of the internationalisation of technology in particular: since a patent protects both the owner s and the inventor s rights, it contains information on the location of the applicant (owner) and on the inventor s place of residence. By comparing these two locations, one can derive a measure for the ownership of crossborder patents 1, which can be used as an indicator for the internationalization of R&D activities (Guellec and van Pottelsberghe de la Potterie 2004). We speak of a cross border patent when the applicant and at least one inventor reside in different countries. An example for a cross border patent is given in Figure 1. Here, Siemens, located in Munich, Germany, applies for a patent based on an invention of Austrian inventors. The Austrian inventors may be doing contract research for the German enterprise, or work for the Austrian affiliate of the company. Cross border patents can take two different forms: first, patents applied in a particular country A and invented in another country B. This case illustrates the degree organisations in country A are engaged in R&D activities in country B. Second, patents invented in 1 Cross border patents should not be confused with the application of a patent at different patent offices. Page 5 of 20

6 country A and owned by foreign applicants located in country B, which shows the degree R&D activities in country A are controlled by organisations located in country B. Figure 1: European Patent Document Source: European Patent Office Since we are interested in the shares of foreign based inventions on the EU s patent stock, we will only deal with the first variant of cross border patents to measure foreign located R&D activities of country A. We calculate the indicator of foreign R&D activities as the fraction of all patents invented abroad and owned by a particular country on the country s overall patent applications. Patents are weighted by inventor country and applicant country shares; if a patent application is invented by a German and a Chinese inventor, it counts as 0.5 patent inventions for each country. The same rule is applied to patent applications. Page 6 of 20

7 3 Data Our analysis draws on patent documents filed with the European Patent Office (EPO) between 1 st Jan, 2000 and the end of June, The data only include patent applications, regardless if the patent has been granted or not. The following graph (Figure 2) illustrates the patenting process at the EPO and helps to understand the advantages and shortcomings of our approach. The EPO collects patent applications received via direct application at the EPO, via the national patent offices and international Patent Cooperation Treaty (PCT) applications for which the EPO acts as designated local office (see European Patent Office 2004 for details on the patent process). The PCT application is becoming increasingly popular as a means to protect an invention in a large number of countries by a single international application. The EPO publishes these patent applications 18 months after the date of filing or the earliest priority date unless the applicant requests earlier publication. The time of publication is also the moment when a patent application enters out database. We only use patent applications, not granted patents. The reason for this choice are the time consuming examination and opposition procedures which, on average, take 44 months, but may also take longer (Dernis and Guellec 2001). Figure 2: A simplified Illustration of the Patenting Process at the European Patent Office National PO EPO PCT Filing at the European Patent Office (EPO) Publication (18m after filing) Substantive Examination Grant (on average after 44m) Source: European Patent Office (2004) 2 Patent documents have been retrieved from EPO s database published on the internet, Data has been retreived on June, 27, Page 7 of 20

8 Granted patents are therefore not useful to study the most recent developments in R&D internationalisation. It is very likely that a number of these patent applications will be withdrawn by the applicants, or do not pass the examination by the EPO. For the purpose of this study, we regard this drawback of patent applications as acceptable compared to the disadvantage of outdated data in the case of granted patents. We count the patent documents according to their priority date which denotes the first filing of a patent in any country worldwide. The priority date is usually considered to be closest to the real invention date (Hinze and Schmoch 2004). Some documents have a priority date prior to the application date, because the application at the EPO is not their first application worldwide. Documents with a priority date before 2000 have been removed from the database. Another selection criterion for the database is the location of the applicant. We only consider patent documents where at least one applicant resides in the EU27. If a patent document lists applicants and/or inventors from two or more countries, the patent is divided between the countries according to the relative share of the applicants/inventors from each country. The total number of patent applications for a county is therefore not necessarily an integer. Table 1 reports the number of patent documents in the database and their distribution between the largest applicant countries. The decreasing number of patent applications in 2004 and 2005 is due to 18 month delay in publication. The six largest application countries account for more than 80% of all patent applications. Applications from the EU12 have a share of below 1%. Table 1: Number of EPO patent applications by EU 27 countries and their geographical distribution, Year Total Country of the applicant Total Share ,264 GERMANY 113, % ,829 FRANCE 39, % ,902 UNITED KINGDOM 21, % ,837 NETHERLANDS 21, % ,650 ITALY 20, % ,419 SWEDEN 11, % 261, , % Source: European Patent Office, own calculations, data retrieved Page 8 of 20

9 4 Empirical Results 4.1 Descriptive Analysis Previous research (le Bas and Sierra 2002; Verspagen and Schoenmakers 2004; Criscuolo et al. 2005; OECD 2006, p. 34) has shown that, as a general trend, the number of cross border patents has increased in most countries world wide since the early 1990s. The share of foreign inventions on total national patent applications varies considerably between countries. Most patent applications, however, are still based on domestic inventions and the linkages to the domestic knowledge bases are strong in the words of Patel and Pavitt (1991), technology is an important case of non globalisation. Our data confirms these results for the years after The share of patent applications based on foreign R&D activity is 12.6% for the whole EU27. It increased from 11.7% (2000) to 14.5% (2003) and decreased to its initial value of 11.7% in The degree of R&D internationalisation has increased notably in Belgium, Denmark, Sweden or the Netherlands, while remained in stable in Germany and even decreased in the United Kingdom. It is therefore difficult to say if R&D internationalisation in Europe has increased or not during the period 2000 to 2005, since the value for 2005 is based on preliminary numbers due to the publication delay. If there was an increase, it was only a modest one. The numbers also show that there was no relocation of R&D activity from Europe to other parts of the world since both, inventions inside and outside the EU27 grew at comparable rate. The EU27 average, however, covers a lot of heterogeneity at the member state level. Its value is strongly influenced by the large countries Germany, Italy, France and the UK (Figure 3). Small countries such as Luxembourg, Ireland, Cyprus and Malta, in contrast, exhibit very high shares of foreign R&D activities on total patent applications. In most cases this can be explained by the high concentration of these patenting activities on a small number of domestic multinational enterprises with activities abroad. Large member states, in contrast, show low share of foreign R&D activity, as can be seen in Figure 3. This indicates that the technological competencies of enterprises located in these countries are still largely based on domestic, and not foreign R&D activity. The only exception is Sweden, a large country in terms of overall patent applications (see Table 1). A closer look at individual patent documents reveals that the high share of Sweden can largely be explained by the global R&D strategies of large Swedish Multinationals such as Eriksson AB or Astra Zeneca AB. These two multinational companies account for the bulk of Swedish patent applications of foreign R&D activities. What is the role of Asia and the EU12 countries as host countries for overseas R&D activities of EU15 and EU12 countries? It is a very limited one, as can be seen in Figure 4 which shows the geographical distribution of foreign inventions applied for patent by organisations located in the EU27. Countries are sorted by their number of foreign inventions, with Germany as the largest applicant of foreign patent inventions at the bottom. Page 9 of 20

10 Figure 4 reveals that foreign R&D activities in the EU12, the CIS and Asia and made only a limited contribution to the internationalisation of European R&D activities. Together, these regions only account for only 5.4% of all cross border patents of the EU27. The share of EU12 locations on all foreign inventions applied is highest in Slovakia (56% of total), Greece (28%), and the Czech Republic (22%). Overall integration of the EU12, however, is poor. Only 10% of all foreign inventions applied by EU12 organisations come from another EU12 country. Figure 3: Share of patent applications at the EPO based on foreign R&D activities as a share of total patent applications owned by country residents, Source: European Patent Office, own calculations The main driver of the internationalisation of R&D activity in Europe was not an increase of R&D activity in Asia or the EU12. At first, European R&D internationalisation is driven by foreign R&D activities in other European countries. The EU27 and other European countries (particularly Switzerland) are by far the most important target countries for foreign R&D activities of EU27 enterprises. Every second foreign invented patent application applied by a EU15 enterprise comes from another EU15 country. The internationalisation of European R&D, therefore, is mainly a further integration of European research. A second major host region of overseas R&D activities of EU27 companies is America, in particular the US and Canada. The share of United States on all foreign inventions of EU27 organisations is 26%. Canada is a larger host location for foreign inventions of European enterprises than China. German, French, and Dutch enterprises together account for two thirds of total EU applications invented in the US. The share of applicants from the EU12 is below one percent. The data also reveal a higher degree of internationalisation between neighbouring countries than the relative sizes of the countries would suggest. The strongest of these effects can be observed between the Czech Republic and Slovakia and between Austria Page 10 of 20

11 and Germany. Germany hosts nearly half of all Austrian cross border patents, while the corresponding proportion of Austria on all German foreign inventions is 13%, which is higher than the share of France. Smaller effects between neighbouring countries can also be observed between Belgium, Luxemburg and the Netherlands, and in the Nordic states, but only within the EU12 and the EU15, not between countries from the EU12 and EU15. We will further examine this effect in Figure 4: Share of different regions on all foreign inventions applied by EU27 organisations, SLOVAKIA BALTIC STATES BULGARIA, ROMANIA CYPRUS, MALTA PORTUGAL SLOVENIA GREECE CZECH REPUBLIC POLAND HUNGARY LUXEMBOURG IRELAND SPAIN DENMARK AUSTRIA BELGIUM FINLAND SWEDEN ITALY NETHERLANDS UNITED KINGDOM FRANCE GERMANY EU15 EU12 EUROP GUS AMERICAS AFRICA, M EAST, OCEANIA JAPAN ASIA 0% 25% 50% 75% 100% Source: European Patent Office, own calculations Share on all foreign inventions To sum up, patent statistics do not give evidence of large scale foreign R&D activities by EU27 enterprises in Central and Eastern Europe, the CIS or Asia. Therefore, fears that relocations of R&D activities already taking place seem to be exaggerated. We can therefore clearly reject the assumption that these countries have become important locations for foreign R&D activities for EU27 companies. This result, however, creates new questions, in particular a) the determinants that create the observed geographical distribution, and b) an explanation for the obvious contradiction between our finding and the large number of reports of foreign R&D activity in Asia cited in section 1. We will deal with these issues in the final two sections of the paper. 4.2 Multivariate Analysis We will now apply a simple econometric model to investigate what determines the pattern of foreign R&D activity identified in the previous section. The model reflects the basic motives and determinants of the internationalisation of corporate R&D discussed in the literature (Kuemmerle 1999; Kumar 2001; le Bas and Sierra 2002; von Zedtwitz and Gassmann 2002; Narula and Zanfei 2005). Moreover, it follows the gravity model approach to explain international trade flows (Leamer and Levinsohn 1995; Disdier and Page 11 of 20

12 Head 2004) by incorporating distance between host and home country as an explanatory variable. The literature distinguishes two strategies regarding R&D internationalisation: First, overseas R&D activities of enterprises focus on adapting existing products to local needs and supporting local production for the host markets ( asset exploiting strategy). High income and a large local market can therefore pose a major incentive for enterprises to start R&D activities in a particular country. As a consequence, we expect a positive association between the number of applications and relative and market size of the host country, measured by total GDP and GDP per capita. Second, enterprises wish to create new knowledge and technologies abroad, which is facilitated by excellent knowledge infrastructure and the presence of knowledge spillovers from universities and enterprises in the host country ( asset augmenting strategy). A very general proxy for the scientific and technological capacity of the host country is its overall R&D intensity, measured by the share of aggregate R&D expenditure on GDP. We expect a positive association between the number of patent applications and R&D intensity. In addition, we include two controls for the home country, the total GDP and home country R&D intensity in the estimation, since we estimate the absolute, not the relative number of cross border patents. Previous research has also pointed out that geographical proximity between host and home country results in higher levels of cross border patenting activity (Guellec and van Pottelsberghe de la Potterie 2001). This distance effect in foreign R&D activities reflects the costs associated with a decentralisation of innovative activity (described by von Zedtwitz and Gassmann 2002; Sanna Randaccio and Veugelers 2003; Gersbach and Schmutzler 2006): enterprises loose scale and scope economies in R&D and innovation in the home country; it becomes more difficult to keep knowledge secret when its production is geographically dispersed; moreover, the diffusion of knowledge within the MNE becomes a more difficult task, because exchange is no longer one way, but includes reverse knowledge transfer between subsidiaries and the parent enterprises as well as between the affiliates themselves. These diseconomies of distance may be further increased by diminishing cultural proximity between distant countries. Trust, for example, results from long lasting economic relations between countries. Generally, cultural proximity may help to overcome the liability of foreignness (Zaheer 1995; Sofka 2006) and facilitate the access to foreign markets and localized knowledge in the host countries for MNEs. The innovation systems of neighbouring countries may also share some similarities such as the time horizon of the agents, the role of trust or the way authority is expressed that make foreign located R&D activities easier (Lundvall 1998, p. 409 f; Drejer 2000, p. 379) Operationalisation The dependent variable is the number of foreign patents invented in country B (the host country) and applied for by country A (the home country) for each year from 2000 until Home countries are all EU27 countries, while we excluded all host countries with Page 12 of 20

13 less than five patent inventions between 2000 and This results in 613 distinct home and host country pairs. As independent variables, we employ total GDP and GDP per capita for all host countries in Mio. USD based on purchasing power parity (PPP) and total R&D expenditure as percentage of GDP at current prices. GDP data has been delivered by the April 2007 edition of IMF s World Economic Outlook Database 3. Data on total R&D intensity has been taken from the 2007/1 edition of OECD s Main Science and Technology Indicators (OECD 2007a) and from the online database of the UNESCO Institute for Statistics 4. We have complete data on GDP and R&D intensity for 565 of the country pairs, which cover over 95% of total foreign patents of the EU27. The remaining 48 pairs are dropped from the analysis. Geographical distance between countries is measured in two steps: first, we define a centre for each country which is the point with the longest distance to all boarder points of the country. We then calculate distances between these centres for all countries in our analysis. To examine the determinants of foreign R&D activity, we estimate the following model: ln pat = α k + β 1 ln GDP t + β 2 ln GDP cap + β 3 rd + β 4 ln dist + β 5 ln GDP t _ hom e + β 6 rd _ hom e + ε GDP t and GDP cap denote total GDP and GDP per capita of the host country; GDP t _home identifies the GDP of the home country. All values are in their natural logarithms. Rd and rd_home are the R&D intensities of the host and home country. Dist is the distance between home and host country. GDPs and distances are included in their logarithmic form. Additional to this specification, we estimate a second model which distinguishes the effects of geographical proximity from the effects of cultural proximity and economic integration. It includes all variables of the first model and two additional dummy variables. The first dummy variable is language, which identifies country pairs with a common official language. We assume that a common language can be a powerful cultural advantage for enterprises from a neighbouring country because it facilitates market access, makes adoptions less costly and enables communication between the parent enterprise in the home country and the affiliate. Therefore, a common language may reinforce neighbourhood effects. We expect a positive sign of the coefficient, which would imply that a common language stimulates foreign R&D activity beyond geographical proximity. The second dummy variable, EU15, denotes country pairs where all two partners are members of the EU15. The European integration in the Common Market during the 1990s has proved to be very powerful in stimulating intra EU foreign direct investment and trade; we assume that a similar integration effect can also be observed in foreign R&D activity, which would manifest in a positive sign of the coefficient of the dummy variable Page 13 of 20

14 4.2.2 Estimation Results Patent counts are usually estimated with a count model (for example a Poisson model) to account for the fact that patent counts are usually integers (an example is Baudry and Dumont 2006). In our case, however, the dependent variable is not an integer since patents can be applied for by two or more entities from different countries which count as fractions. We therefore employ panel regression with random effects since the data relates to different years. The results of the regression are given in Table 2. Column (1) reports results without the dummy variables language and eu15, column (2) delivers results with the two dummies. A total R squared of 0.46 and 0.49 indicates that our model can explain nearly half of the total variance of foreign R&D activity. The coefficients are highly significant and show the expected signs: The level of foreign R&D activity is positively related with relative and absolute GDP of the host country, which illustrates the role of market size and demand in the host country as major incentives for innovative activities abroad. Table 2: Regression Results (1) (2) lnpat lnpat lngdp_t 0.576*** 0.549*** (0.039) (0.039) lngdp_cap 0.379*** 0.273*** (0.098) (0.099) rd 0.260*** 0.273*** (0.063) (0.061) lndist 0.470*** 0.393*** (0.052) (0.057) lngdp_t_home 0.378*** 0.363*** (0.038) (0.038) rd_home 0.560*** 0.557*** (0.056) (0.055) language 1.364*** (0.21) eu * (0.13) Constant 6.509*** 6.000*** (1.10) (1.08) Observations Number of id R squared within R squared between R squared overall Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 Page 14 of 20

15 A higher level of R&D intensity in the host country is associated with higher levels of foreign R&D activity. On average, an increase of one percentage point in host country R&D intensity results in a 0.26 percentage point increase in the level of foreign R&D activity. This reflects that the scientific potential including spillovers between the host country and foreign owned enterprises is another major incentive for foreign R&D activities. Moreover, foreign R&D activity decreases with distance between host and home country, as indicated by the negative sign of lndist. Similar effects of distance are also found for trade flows by a number of studies (Leamer and Levinsohn 1995). We conclude that even in intangible activities such as R&D, information and communication technologies have not lead to the death of distance. We further investigate this distance effect in column 2 where we report results for the enlarged model. Two additional dummy variables identify countries with a common language and country pairs where both countries are EU15 member states. We see that the coefficients of both variables are significant although the EU15 dummy is only significant at the 10% error level. Both dummies do not greatly increase explanatory power of the model (as can be seen by the R squared, which only increases by less than 0.03), but considerably enhance our understanding of the distance effect. We can expect 1.3% more cross border patents between countries which a common language compared to two other countries with distance and all other economic factors incorporated in the model equal. This points to the importance of cultural factors for innovative activity and the location of overseas R&D activity in particular. A higher number of cross border patents can also be expected between EU15 members. All other things equal, the number of foreign R&D activities is higher between two EU15 members than between a EU15 member and a non EU15 country. This points to the fact that the Single Market not only fostered economic, but also scientific integration in Europe. Compared to culture, however, economic integration is a much less powerful explanation of foreign R&D activity. An interesting perspective on the determinants of R&D internationalisation offers a comparison of the coefficients of home and host country GDP and R&D intensity. The coefficient of home R&D intensity is nearly two times that of host R&D intensity. In the case of GDP, the coefficient of host country GDP is larger than that of home country GDP. This indicates, at first, that the internationalisation of R&D is rather driven by market opportunity in the host country than by the size of the host country. Large home countries in economic terms, such as Germany, Japan or the US, tend to have a quite low share of foreign invented patents compared to some small countries such as Sweden, the Netherlands, Belgium and Finland. The home country, however, matters in terms of R&D intensity. A high R&D intensity indicates the presence of R&D intensive industries, which have a higher likelihood to engage in foreign R&D activities than sectors with a lower R&D intensity. This supply factor of enterprises willing to go abroad with R&D activities is more important than locational factors at the host country for explaining the R&D internationalisation of a country. Page 15 of 20

16 5 R&D without Patents? This paper investigated the extend enterprises from the EU27 countries perform R&D activities outside their home countries and the determinants for these overseas activities. To measure internationalisation, we calculated an indicator of cross border patent activity by comparing the location of the inventor and the location of the applicant in patent documents. The patent data we used for the analysis includes all applications to the European Patent Office between 2000 and Our results do not deliver evidence for large scale overseas R&D activities of EU27 enterprises in Central and Eastern Europe, China or India. European multinational enterprises internationalized their R&D activities considerably during the 1990s. The most important host countries for these overseas activities, however, are clearly in Western Europe and North America. High incomes, a large market size, and a high scientific and technological level seem to be stronger attractors than low wages for R&D activities. This observation is confirmed by the results of a multivariate regression analysis. The number of cross border patents in a host country increases with its relative and absolute market size and with higher levels of overall R&D intensity in the host country. Distance between home and host country is negatively related to the number or cross border patents. How can we explain this obvious contradiction between patent data and the results of empirical studies cited in the introduction? If we exclude some common sense limitations of patent data such as the gap between application and final grant and the fact that the propensity to patent varies with size, sector, and other enterprise characteristics (Griliches 1990; Patel and Pavitt 1995; Cohen et al. 2000), two explanations remain: First, following the results of our multivariate analysis, we can conclude that economic and scientific conditions in Central and Eastern Europe or Asia still offer too few incentives for foreign R&D activities that may outweight the disadvantages of distance. Despite their considerable absolute market size and rapid growth in the last years, China, India or other countries still considerably lack behind the EU15 in terms of GDP per capital. These countries are the poorest in our sample in terms of GDP per capita, and still need time to become attractive host markets that would justify R&D investments. The positive correlation between income per capita and the extend of foreign R&D activities may also point to the fact that low wages are overrated as the main determinants of foreign R&D. The same is true for scientific capabilities. R&D expenditure have increased in China, India and other countries in recent years; R&D productivity in terms of the ratio of domestic patent inventions to industry financed R&D expenditure, however, is still considerably lower than in the EU15 or in North America (Baudry and Dumont 2006, p. 325; OECD 2006, p. 13). Innovation capabilities of the emerging economies are (not yet) on a par with those of OECD countries, which explains the low number of cross border patents from these countries. Page 16 of 20

17 A second explanation relates to the goals of foreign sponsored R&D in these countries. European MNEs mainly focus on adapting existing products to local needs and supporting local production for the host markets, this strategy may yield only very few inventions worth to be protected by an EPO patent. A look at applications at patenting activity the Chinese Patent Office CIPO provides some evidence for this assumption (Sun 2003; Sun 2004; OECD 2007b). Foreign companies are very active at the CIPO and account for a considerable share of total patenting activity, both with patents based on foreign and Chinese inventions. These patents, however, do not show up in EPO patent counts, because they are only filed in China. Part of this story may be also the fact that R&D investments of EU15 and US enterprises in these countries still are relatively young and take time to become productive. Our results do not imply that the countries in question will never develop into important locations for overseas R&D activities of European enterprises. Since Central and Eastern Europe and Asia are main target countries for FDI of EU27 companies, it seems very likely that these companies will also build up R&D capacities there. This, however, takes time since corporate innovation processes are inherently uncertain, complex, and cumulative and characterized by continuing interaction with the firm s environment. Therefore, many engagements in foreign located R&D of European enterprises today may be mainly seen as mainly adapting existing products and as a commitment to the host country; however, in the future we will envisage rising shares of these countries in the portfolio of R&D locations of EU enterprises. Page 17 of 20

18 6 References Baudry, M., and B. Dumont (2006). Comparing firms triadic patent applications across countries: Is there a gap in terms of R&D effort or a gap in terms of performances? Research Policy, 35: Cohen, W. M., R. R. Nelson, and J. Walsh (2000). Protecting Their Intellectual Assets: Appropriability Conditions and Why U.S. Manufacturing Firms Patent (or Not), NBER Working Paper, Vol Cambridge [Mass]. Criscuolo, C., R. Narula, and B. Verspagen (2005). Role of home and host country innovation systems in R&D internationalisation: a patent citation analysis. Economics of Innovation and New Technology, 14(5): Dernis, H., and D. Guellec (2001). Using Patent Counts for Cross country Comparisons of Technology Output. STI Review(27): Disdier, A. C., and K. Head (2004). The Puzzling Persistence of the Distance Effect on Bilateral Trade, UBC Economics Department Discussion Paper, Vol. 21. Vancouver. Drejer, I. (2000). Comparing Patterns of Industrial Interdependence in National Systems of Innovation A Study of Germany, the United Kingdom, Japan and the United States. Economic Systems Research, 12(3): European Patent Office (2004). How to get a European Patent. Guide for applicants. patent office.org/legal/guiapp1/pdf/g1en_net.pdf Gersbach, H., and A. Schmutzler (2006). Foreign Direct Investment and R&D Offshoring, Socioeconomic Institute University of Zurich Working Paper, Vol Zürich. Griliches, Z. (1990). Patent Statistics as Economic Indicators: A Survey. Journal of Economic Literature, 28(4): Guellec, D., and B. van Pottelsberghe de la Potterie (2001). The Internationalisation of Technology Analysed with Patent Data. Research Policy, 30(8): Guellec, D., and B. van Pottelsberghe de la Potterie (2004). Measuring the Internationalisation of The Generation of Knowledge. In H. F. Moed, W. Glänzel, & U. Schmoch (eds.), Handbook of Quantitative Science and Technology Research: Dordrecht: Kluwer Academic Publishers. Hinze, S., and U. Schmoch (2004). Opening the Black Box. In H. F. Moed, W. Glänzel, & U. Schmoch (eds.), Handbook of Quantitative Science and Technology Research: Dordrecht: Kluwer Academic Publishers. Jaffe, A. B., and M. Trajtenberg (2002). Patents, citations, and innovations: A window on the knowledge economy. MIT Press: Cambridge and London. Kuemmerle, W. (1999). Foreign Direct Investment in Industrial Research in the Pharmaceutical and Electronics Industries Results from a Survey of Multinational Firms. Research Policy, 28(2 3): Kumar, N. (2001). Determinants of Location of Overseas R&D Activity of Multinational Enterprises: The Case of US and Japanese Corporations. Research Policy, 30(1): Page 18 of 20

19 le Bas, C., and C. Sierra (2002). 'Location versus Home Country Advantages' in R&D Activities: Some Further Results on Multinationals' Locational Strategies. Research Policy, 31(4): Leamer, E. E., and J. Levinsohn (1995). International Trade Theory: The Evidence. In G. E. Grossman, & K. Rogoff (eds.), Handbook Of International Economics, Vol. 3: Amsterdam, New York and Oxford: Elsevier, North Holland. Li, J., and D. R. Yue (2005). Managing Global Research and Development in China: Patterns of R&D Configuration and Evolution. Technology Analysis & Strategic Management, 17(3): Lundvall, B. A. (1998). Why Study National Systems and National Styles of Innovation? Technology Analysis & Strategic Management, 10(4). Marin, D. (2004). A Nation of Poets and Thinkers Less So with Eastern Enlargement? Austria and Germany, Discussion Paper, Department of Economics, University of Munich, Vol Munich. Narula, R., and A. Zanfei (2005). Globalisation of Innovation: The Role of Multinational Enterprises. In J. Fagerberg, D. C. Movery, & R. R. Nelson (eds.), The Oxford Handbook of Innovation: Oxford: Oxford University Press. OECD (2006). Compendium of Patent Statistics Paris: Organisation for Economic Cooperation and Development. OECD (2007a). Main Science and Technology Indicators 1/2007. Paris: Organisation for Economic Co operation and Development. OECD (2007b). OECD Reviews of Innovation Policy CHINA. Synthesis Report. Paris: Organisation for Economic Co operation and Development. Patel, P., and K. Pavitt (1991). Large Firms in the Production of the World's Technology: An Important Case of "Non Globalisation". Journal of International Business Studies, 22(1): Patel, P., and K. Pavitt (1995). Patterns of technological activity. In P. Stoneman (ed.), Handbook of Innovation and Technological Change: Oxford: Blackwell. Rose, G., and V. Treier (2005). FuE Verlagerung: Innovationsstandort Deutschland auf dem Prüfstand. Sanna Randaccio, F., and R. Veugelers (2003). Global Innovation Strategies of MNEs: Implications for Host Economies. In J. Cantwell, & J. Molero (eds.), Multinational Enterprises, Innovative Strategies and Systems of Innovation: Cheltenham: Edward Elgar. Sofka, W. (2006). Innovation Activities Abroad and the Effects of Liability of Foreignness: Where it Hurts, ZEW Working Paper, Vol Mannheim. Sun, Y. (2003). Determinants of foreign patents in China. World Patent Information, 25: Sun, Y. (2004). Spatial distribution of patents in China. Regional Studies, 34(5): Verspagen, B., and W. Schoenmakers (2004). The Spatial Dimension of Patenting by Multinational Firms in Europe. Journal of Economic Geography, 4(1): von Zedtwitz, M. (2004). Managing foreign R&D laboratories in China. R&D Management, 34(4): Page 19 of 20

20 von Zedtwitz, M., and O. Gassmann (2002). Market versus Technology Drive in R&D Internationalization: Four different Patterns of Managing Research and Development. Research Policy, 31(4): Walsh, K. A. (2007). China R&D: A High Tech Field of Dreams. Asia Pacific Business Review, 13(3): Zaheer, S. (1995). Overcoming the Liability of Foreignness. Academy of Management Journal, 38(2): Zanfei, A. (2000). Transnational Firms and the Changing Organisation of Innovative Activities. Cambridge Journal of Economics, 24(5): Page 20 of 20