Innovation and Knowledge Diffusion in the Global Economy [PDF]

Innovation and Knowledge Diffusion in the Global Economy A thesis presented by Jasjit Singh to The Department of Business Economics in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the subject of Business Economics Harvard University Cambridge, Massachusetts April 2004

© 2004 – Jasjit Singh All rights reserved.

Innovation and Knowledge Diffusion in the Global Economy Thesis Chair: Professor Tarun Khanna

Author: Jasjit Singh Abstract

The first part of this dissertation studies two questions regarding the role of multinational firms (MNCs) in knowledge diffusion: (1) How actively do overseas subsidiaries of MNCs exchange knowledge with organizations from their host country? (2) To what extent do these subsidiaries facilitate bi-directional knowledge flow between the MNC home base and the host country? These questions are analyzed using citation data for over half a million patents from 4,400 firms and organizations from six countries. A novel regression framework using choice-based sampling is used to estimate the probability of knowledge flow. The results suggest that there are significant bi-directional knowledge flows between MNCs and their host countries, but MNCs contribute less to host country knowledge than they gain from it. However, the exact pattern varies significantly across countries and sectors, depending on the knowledge-intensity of foreign direct investment. The second part of this dissertation examines if collaborative networks among individuals explain two patterns of knowledge diffusion: (1) geographic localization of knowledge flows, and (2) easier transmission of knowledge within firms than between firms. Collaborative links among individuals are inferred using a “social proximity graph” constructed from patent collaboration data for more than one million inventors. The existence of a direct or indirect collaborative tie is found to be associated with a greater probability of knowledge flow, with the probability increasing with the directness

iii

of the tie. Controlling for collaborative ties significantly reduces the estimated impact of geographic co-location and firm boundaries on the probability of knowledge flow. In fact, conditional on the existence of close collaborative ties, geographical co-location and firm boundaries have no additional effect on the probability of knowledge flow. The third part of this dissertation analyzes innovation in emerging and newly industrialized economies, with the emphasis being on Asian economies. In particular, I use patent data to study how the overall and sector-level innovative capabilities of Taiwan, Korea, Hong Kong, Singapore, India and China have evolved over the past 30 years. I also study the relative importance of foreign multinationals, business groups, individuals, domestic firms and research institutes in innovation, and the concentration of innovative activity.

iv

Acknowledgements

I am extremely grateful to my thesis committee – Professors Tarun Khanna, Joshua Lerner and Richard Caves – for their constant guidance and support. I have also been fortunate to get an opportunity to work closely with Professors Ken Corts, Ananth Raman and V.G. Narayanan, from whom I have learnt the nuts and bolts of research. I am also thankful to Professors George Baker, Jerry Green and Lee Fleming for their constant encouragement and help over the years. It has been wonderful to be a part of the Boston academic community. I have learnt a lot from the faculty and fellow students at Harvard, MIT and Boston University. I am also grateful for detailed feedback and close mentoring from several people in the broader academic community, who helped me immensely even though they barely knew me to start with and had little to gain in return. While space constraints keep me from acknowledging them individually, I am indebted to each one of them! My parents Sarvajit Singh and Harmohinder Kaur have been my greatest source of strength. They inspired me to be an academic, and encouraged me to hang in there even on occasions when the journey looked rough. My wife Pia, little boy Pawan, and his soon-to-be-born sibling (“B2B2”) have helped make my PhD dream a reality through their endless love and support, and have brought a joyful balance to my life. I would also like to thank my mother-in-law Lisbeth, who helped us out when we were overwhelmed by the time pressures of having our first baby. And I am most fortunate to have a fatherin-law like Claes, who gave me confidence and even volunteered to proofread my thesis!

v

Table of Contents Chapter 1: Introduction ....................................................................................................... 1 Chapter 2: Multinational Firms and Knowledge Diffusion: ............................................... 6 1. Introduction................................................................................................................. 6 2. Hypotheses.................................................................................................................. 9 3. Data on Patent Citations and Multinational Ownership ........................................... 12 4. Preliminary Analysis................................................................................................. 17 5. Citation-Level Regression Methodology.................................................................. 21 6. Results....................................................................................................................... 26 7. Further Issues in Using USPTO Patent Citations ..................................................... 42 8. Discussion and Concluding Remarks ....................................................................... 44 Appendix 2.1. A Note on Choice-Based Sampling and WESML ............................... 47 Chapter 3: Collaborative Networks as Determinants of Knowledge Diffusion Patterns.. 51 1. Introduction............................................................................................................... 51 2. Hypotheses................................................................................................................ 54 3. Patent Data ................................................................................................................ 59 4. Empirical Methodology ............................................................................................ 63 5. Results....................................................................................................................... 72 6. Limitations ................................................................................................................ 82 7. Conclusion ................................................................................................................ 84 Chapter 4: Technological Dynamism in Asia................................................................... 87 1. Introduction............................................................................................................... 87 2. Comparing innovation across countries: methodology............................................. 91 3. Comparing innovation across countries: results ....................................................... 92 4. Sector-level analysis of innovation: methodology.................................................... 96 5. Sector-level analysis of innovation: results ............................................................ 102 6. Comparing type of innovators: methodology ......................................................... 110 7. Comparing type of innovators: results.................................................................... 112 8. Concluding thoughts ............................................................................................... 123 References....................................................................................................................... 125

vi

Chapter 1: INTRODUCTION

This dissertation studies technological innovation and knowledge diffusion. Motivating my research is the belief that acquisition of knowledge and management of innovation are critical for economic success, both for firms and for regions. Therefore, better understanding of these phenomena would lead to better prescriptions for firms in formulating their technology strategies, and for regions and countries in making policies governing technology transfer, innovation, and both incoming and outgoing investment. The ease with which knowledge diffuses has important implications for growth (Grossman and Helpman, 1991). However, even though ideas are intangible in nature, empirical evidence shows that they do not flow freely across regional and firm boundaries. Two patterns of knowledge diffusion have been identified. First, knowledge flows are geographically localized (Jaffe, Trajtenberg and Henderson, 1993). Second, knowledge flow is easier within firm boundaries than between firms (Kogut and Zander, 1992). This dissertation studies two different aspects of these patterns. The first paper studies how, because of easier flow of knowledge within firm boundaries, multinational firms (MNCs) can help overcome geographic constraints on knowledge flow and enable international diffusion of knowledge. The second paper studies how direct and indirect collaborative links between individuals are a key mechanism giving rise to the above knowledge flow patterns in the first place. Governments around the world continue to spend huge resources to attract MNCs, at least partly in the hope of knowledge gains from them. However, literature on how foreign direct investment (FDI) contributes to knowledge diffusion still remains

1

fragmented and inconclusive. My first paper (titled “Multinational Firms and Knowledge Diffusion: Evidence Using Patent Citation Data”) extends existing research on role of MNCs in knowledge diffusion. Related literature in international economics largely emphasizes uni-directional knowledge flows from foreign MNCs to host country domestic firms. However, as the strategy and international business literature has established, FDI can also be a channel through which domestic technology can fall into the hands of foreign competitors. Therefore, except for countries that have little unique technology of their own, it is important to consider bi-directional knowledge flows in studying net gains from FDI. The potential “leakage” of domestic knowledge through FDI is a particularly real issue for technologically advanced countries, which are the focus of my first paper. I find that knowledge flows from host countries to MNCs are about as intense as those between domestic entities, showing that MNCs are able to tap into local sources of knowledge just as much as the domestic entities are. On the other hand, knowledge flows back from MNC subsidiaries to their host countries are weaker. In other words, on an average, MNCs do not seem to contribute as much to local knowledge as they gain from it. However, this pattern differs across industries and countries depending on knowledgeintensity of local investment by foreign MNCs. I also find that subsidiaries of foreign MNCs, especially those from the same home country, are particularly good at learning from each other. Turning to cross-border knowledge flows, I find MNCs to be far better than markets at transferring knowledge across international borders, with knowledge flow being as intense from a foreign subsidiary to the MNC home base as from the home base to the foreign subsidiary. I also find that greater overseas innovation by an MNC leads

2

not just to direct learning by its foreign subsidiaries, but also to increase in its home base’s absorptive capacity for foreign knowledge. While the study summarized above focuses on measurement of knowledge flows, the second paper (titled “Collaborative Networks as Determinants of Knowledge Diffusion Patterns”) digs deeper into the mechanisms behind such knowledge flows. Numerous factors, including informal networks, institutions, norms, language, culture, incentives, and other formal and informal mechanisms might affect the ease with which knowledge diffuses. However, this paper explores the extent to which the observed knowledge diffusion patterns can be accounted for simply by the fact that people within the same region or firm have close collaborative links that might facilitate flow of complex knowledge. In particular, I analyze if collaborative ties between inventors help account for the effect of geographic co-location and firm boundaries on the probability of knowledge flow between individual inventors of U.S. patents. I allow for the possibility that direct and indirect ties could matter to a different extent. For example, if an individual X has a direct collaborative relationship with individual Y, and Y has a direct tie with Z, Z might learn indirectly about X’s work through his tie with Y. To measure the directness of collaborative ties among over a million inventors in the U.S. patent database, I construct a “social proximity graph” based on information about the team of inventors for each individual patent. This graph allows me to derive a measure of “social distance” between inventors. This data is then used to explore the extent to which collaborative links are important for knowledge diffusion. Collaborative ties are found to be crucial for knowledge flow, with the probability of

3

knowledge diffusion between two teams of inventors being inversely related to the “social distance” between them. Even more interestingly, I find that collaborative networks are useful in explaining why knowledge flows tend to be concentrated within firms and regions. The effect of being in the same region or the same firm on probability of knowledge flow falls significantly once collaborative networks are accounted for. In fact, conditional on having close collaborative ties, geographical co-location and firm boundaries have little effect on probability of knowledge flow. In contrast, for patent pairs with only indirect collaborative ties or no collaborative ties at all, geographic co-location and firm boundaries continue to be associated with greater probability of knowledge flow, possibly because of other kinds of formal and informal mechanisms influencing intra-regional and intra-firm knowledge flow. The first two papers described above also make important methodological contribution to the literature on knowledge diffusion. While patent citations are an imperfect measure of knowledge diffusion, they are widely used in research as a way to directly capture micro-level knowledge flow. Following this literature, the papers discussed above also use patent citations to measure micro-level knowledge flows. However, the methodology used here is entirely new. Jaffe, Trajtenberg and Henderson (1993) pioneered a widely-used statistical technique that tries to correct for factors other than knowledge spillovers that might determine distribution of technological activity, and hence the pattern of patent citations. However, Thompson and Fox-Kean (2004) have shown that existing application of this technique often leads to over-estimation of knowledge flows. To address this, I propose a novel citation-level regression approach

4

that estimates the probability of micro-level knowledge flow between innovating teams using a novel regression framework based on choice-based sampling (Manski and Lerman, 1977). As described in detail later, the resulting weighted maximum likelihood approach helps address some methodological concerns regarding existing use of citations for measuring knowledge diffusion. The third paper in this dissertation, titled “Technological Dynamism in Asia” (joint work with Ishtiaq P. Mahmood), compares the extent and composition of innovation in six Asian economies – Korea, Taiwan, Hong Kong, Singapore, India and China. Using patent data from the past three decades, it shows how Korea and Taiwan have transitioned to a level and quality of innovation comparable with world leaders, while Singapore and Hong Kong have only recently started to move in that direction. The findings suggest that the “Asian Tigers”, often studied as a homogenous bunch, actually differ substantially in the extent to which, and the mechanisms through which, innovation is responsible for economic growth in recent decades.

5

Chapter 2: MULTINATIONAL FIRMS AND KNOWLEDGE DIFFUSION: Evidence Using Patent Citation Data 1. Introduction Innovation and knowledge diffusion play a critical role in economic growth, with growth rates being highly sensitive to how easily knowledge diffuses (Romer, 1990; Grossman and Helpman, 1991; Eaton and Kortum, 1999). While economists once believed that ideas should be costless to transport, recent empirical literature has established that knowledge spillovers are geographically localized (Jaffe, Trajtenberg and Henderson, 1993; Audretsch and Feldman, 1996; Branstetter, 2001; Keller, 2002). Foreign direct investment can play an important role in overcoming this geographic constraint on the diffusion of knowledge (Caves, 1974; Aitken and Harrison, 1999; Branstetter, 2000).1 Governments around the world continue to spend huge resources to attract multinational firms (MNCs), at least partly in the hope of knowledge gains from them. However, literature on how foreign direct investment (FDI) contributes to knowledge diffusion still remains fragmented and inconclusive. Existing literature largely emphasizes uni-directional knowledge flows from foreign MNCs to host country domestic firms. However, while FDI can lead to knowledge flows for the domestic players, it can also be a channel through which domestic technology can fall into the hands of foreign competitors. Therefore, except for countries that have little unique technology of their own, it is important to consider bi-directional knowledge flows in studying net gains from FDI. The potential

1

Multinational activity is not the only way in which global economic activity can contribute to knowledge diffusion. Trade can also play an important role (Coe and Helpman, 1995), but is not studied in this paper.

6

“leakage” of domestic knowledge through FDI is a particularly real issue for technologically advanced countries, which are the focus of this paper. For example, Dalton and Shapiro (1995) say, “Rapid growth of foreign R&D in the US has led to concerns about an erosion of US technology leadership… Some observers have questioned the quality of the research effort by foreign companies. They have argued that US research centers of foreign companies are merely ‘listening posts’ that focus on technology scanning.” A central goal of my paper is study the extent to which this concern is valid. It is hard to measure knowledge spillovers directly. Therefore, several studies have tried to estimate the effect of FDI on productivity of domestic firms (Caves, 1974; Aitken and Harrison, 1999). A challenge in doing so, however, has been separating knowledge spillover effects of FDI from its effect on competition (Caves, 1996; Chung, 2001; Chung, Mitchell and Yeung, 2003). An alternate empirical approach, which I follow in this paper, is to measure knowledge diffusion using patent citation data. While patent citations are an imperfect measure of knowledge diffusion and also make it hard to separate true externalities from intentional knowledge transfer (Peri, 2003), they are widely used in research as a way to directly capture micro-level knowledge flows (Jaffe and Trajtenberg, 2002). I measure bi-directional knowledge flows between MNC subsidiaries and domestic players, and also between MNC home base and host countries, using data on citations made by over half a million patents originating from 4,400 MNCs and domestic organizations in the US, Japan, Germany, France, UK and Canada. In its use of patent data in studying role of MNCs, the current paper builds upon Almeida (1996), Branstetter (2000) and Frost (2001), while placing

7

much more emphasis on bi-directional knowledge flows, and looking at cross-country and cross-sector differences in the observed patterns. My findings suggest that there are significant bi-directional knowledge flows between MNCs and their host countries, but that MNCs contribute less to host country knowledge than they gain from it. For intra-national knowledge flows, my specific findings are: (1a) Knowledge flows from domestic entities to local subsidiaries of foreign MNCs are as strong as those between domestic entities; (1b) Knowledge flows from MNC subsidiaries to domestic entities are weaker on an average, with the pattern differing across sectors and countries depending on R&D-intensity of FDI; (1c) MNC subsidiaries are particularly good at learning from each other. For knowledge flows across borders, I find that: (2a) MNCs are as good at transferring knowledge from their subsidiaries to their home base as from the home base to the subsidiaries; (2b) More intense innovative activity by MNC subsidiaries increases bi-directional knowledge flow between the host country and the MNC home base, with the gains being larger for the MNC home base than for the host country’s domestic players. This paper also makes a methodological contribution to use of patent citation data in measuring knowledge spillovers. Jaffe, Trajtenberg and Henderson (1993) pioneered a widely-used statistical technique that tries to correct for factors other than knowledge spillovers that might affect technological specialization of regions, and hence the pattern of patent citations. However, Thompson and Fox-Kean (2004) have shown that existing application of this technique often leads to over-estimation of knowledge flows. To address this, I propose a novel citation-level regression approach that estimates the probability of citation between any two patents using a choice-based sampling approach

8

(Manski and Lerman, 1977). In addition, I use a combination of econometric techniques as well as additional robustness checks using European Patent Office (EPO) data to address concerns about using data from US Patent Office (USPTO) for international comparison. The rest of the paper is organized as follows. Section 2 presents my formal hypotheses. Section 3 describes the patent citation data and my subsidiary-parent database. Section 4 presents preliminary analysis of knowledge flows between MNCs and domestic organizations. Section 5 describes my citation-level regression framework. Section 6 presents results on role of MNCs in both intra-national and cross-border knowledge flows. Section 7 addresses concerns regarding use of USPTO data in measuring international knowledge diffusion. Section 8 offers concluding thoughts. 2. Hypotheses For international knowledge diffusion to be an interesting issue to study, the first fact to establish is that knowledge does not automatically transmit across countries. While previous work has found empirical support for geographic localization of knowledge spillovers (e.g., Jaffe, Trajtenberg and Henderson, 1993), recent work raises issues that could have led to over-estimation of this phenomenon (Thompson and FoxKean, 2004). Therefore, I revisit the following hypothesis using a new methodology that addresses the above concerns. Hypothesis 1. The probability of knowledge flow within a country exceeds that between different countries, even after controlling for technological specialization of countries. MNCs can facilitate international knowledge diffusion through their ability to transmit knowledge more effectively than would be possible through market-mediated

9

mechanisms (Hymer, 1976; Buckley and Casson, 1976). While the transaction cost literature suggests that this happens through decreased opportunism within a firm (Williamson, 1985; Ethier, 1986; Teece, 1986), other research shows social networks and a firm’s internal organization to transmit complex and tacit knowledge as the mechanisms (Hedlund, 1986; Bartlett and Ghoshal, 1989; Kogut and Zander, 1993; Nohria and Ghoshal, 1997). Distinguishing between these two is beyond the scope of this paper, but I do formally test the following hypothesis on intra-MNC knowledge flows: Hypothesis 2. The probability of cross-border knowledge flow within an MNC exceeds that between different firms, even after controlling for the relative technological proximity of different divisions within the same MNC. A central argument of this paper is that looking at uni-directional knowledge flows from an MNC subsidiary to its host country misses the point that knowledge could also flow from the host country to the MNC subsidiary (Almeida, 1996; Frost, 2001), and from the subsidiary to the MNC home base (Hedlund, 1986; Bartlett and Ghoshal, 1989). My next task therefore is to empirically establish the presence of such bi-directional knowledge flows: Hypothesis 3. There are significant knowledge flows in both directions between an MNC subsidiary and its host country. Hypothesis 4. There are significant knowledge flows in both directions between an MNC subsidiary and its home base. Existing literature also suggests that intra-national knowledge flows are particularly strong between different foreign MNC subsidiaries located in the same

10

country (Head, Ries and Swenson, 1995; Feinberg and Majumdar, 2001; Feinberg and Gupta, 2003), which I verify next: Hypothesis 5. There are significant knowledge flows between local subsidiaries of different foreign MNCs. Next, I examine the relative strength of different knowledge flows. If local subsidiaries of foreign MNCs are involved in knowledge-intensive activities like advanced research or innovative product development, we might expect greater knowledge spillover benefits to the host country. Existing evidence suggests, however, that even MNC subsidiaries doing R&D often focus on adaptation of their parent firm’s products for the local markets (Mansfield, Teece and Romeo, 1979), or on being “listening posts” to monitor local technological developments (Almeida, 1996; Florida, 1997; Frost; 2001). Surveys by Kuemmerle (1999) and Frost, Birkinshaw and Ensign (2002) reveal that, while the number of MNC subsidiaries doing advanced research has been increasing, such cases still comprise only a minority. Raising further concerns about the benefits from FDI is the adverse selection in the “knowledge intensity” of overseas operations of MNCs. Kogut and Chang (1991) find that a disproportionately large fraction of Japanese FDI in the US is restricted to industries where the Japanese MNCs lag behind their US counterparts. Similarly, Shaver and Flyer (2000) and Chung and Alcacer (2002) find that technologically advanced MNCs are less likely to locate sophisticated facilities overseas and, when they do, are likely to locate them far from domestic players to prevent their technology from being copied. Cantwell and Janne (1999) find that foreign subsidiaries of even technologically advanced MNCs focus on the specific technologies where these MNCs

11

lag behind. All of this raises concerns that host countries might lose more from “leakage” of domestic knowledge to MNCs than gain in the form of knowledge spillovers from MNCs, a hypothesis I directly test in this paper. Hypothesis 6. The probability of knowledge flow from the host country to an MNC subsidiary exceeds that from the MNC subsidiary to the host country. Extending the above logic, the relative extent of knowledge flows from the host country to MNCs should be most intense in settings where the domestic firms do more “knowledge-intensive” work than the MNC subsidiaries. This can be tested by seeing how the pattern of bi-directional knowledge flows varies with the relative R&D intensity (i.e., the ratio of R&D to total production) for domestic firms and MNC subsidiaries. Hypothesis 7. The probability of knowledge flow from the host country to MNC subsidiaries is particularly great in countries and sectors where the R&D intensity of MNC subsidiaries is significantly lower than that of the host country. Finally, if foreign subsidiaries of an MNC serve as listening posts for the home base, these subsidiaries should improve the absorptive capacity of the MNC home base for knowledge produced in the host countries. This gives the final hypothesis: Hypothesis 8. The relative probability of knowledge flow from a host country to a foreign MNC’s home base is greatest when the MNC’s local subsidiaries are most active in knowledge-related activities. 3. Data on Patent Citations and Multinational Ownership 3.1. Patent Citations as Measure of Knowledge Flow Patent citations leave behind a trail of how a new innovation potentially builds upon existing knowledge. An inventor is legally bound to report relevant “prior art”, with

12

the patent examiner serving as an objective check. Unlike academic papers, there is usually an incentive not to include superfluous citations, as that might reduce the scope of one’s own patent. There are, however, two factors that add noise to citations as a measure of knowledge flow. First, citations might be included by the inventor for strategic reasons (e.g., to avoid litigation). Second, a patent examiner might add citations to patents that the original inventor knew nothing about. Recent studies comparing citation data with inventor surveys show that the correlation between patent citations and actual knowledge flow is indeed high, but not perfect (Jaffe and Trajtenberg, 2002; Duguet and MacGarvie, 2002). The defense given in the common research use of patent citations is that use of citations is okay in large-sample studies as long as the noise does not bias the results of interest. Note that viewing patent citations as being correlated with knowledge flows is not the same as claiming that patents themselves are the mechanism behind these knowledge flows. Consider the analogy that a PhD student may cite research papers of his advisor, even though knowledge gained by working closely with the advisor could be much more than what could be captured in the advisor’s papers. 3.2. Data from US Patent Office (USPTO) Since patents from different patent offices are not comparable to each other, it is common practice to use data from a single patent granting country like US (Jaffe and Trajtenberg, 2002) or UK (Lerner, 2002) to standardize the measure of innovation for research purposes. Following this practice, I use a data set on US patents, constructed by merging data from the US Patent Office (USPTO) with an enhanced version made available by Jaffe and Trajtenberg (2002). A major issue in using patent data is that only some of the innovations are patented (Levin, Klevorick, Nelson and Winter, 1987), with

13

systematic differences across countries and sectors in their likelihood to file for USPTO patents. Since this makes counts of patents and patent citations misleading as raw measures, I only estimate the probability of knowledge flow between two innovations that do end up as patents, without claiming that these comprise all the innovations. Following standard practice, the country of residence of the inventors is taken as the country where an innovation takes place. In order to ascertain whether it originated from a domestic organization or from the local subsidiary of a foreign MNC, I check whether the “home country” of the assignee organization is the same as the country of innovation. As mergers and acquisitions are a potential issue in defining the home country, I restrict my analysis to patents in a narrow time window between 1986 and 1995 as I use various data sources from around 1990 for constructing the parentsubsidiary database. I examine patents by inventors from six leading economies: US, Japan, Germany, France, UK and Canada. The number of patents from these countries for the period 1986-1995 is about 0.9 million, or about 91% of all USPTO patents (Table 2.1, column 1).2 About 83% of these patents are owned by firms or organizations (as opposed to individuals), and are the ones of interest here (Table 2.1, column 2). 3.3. Multinational Data A crucial step in the data analysis was identifying whether an assignee firm has its home base in the country of innovation (e.g., IBM in the US), or if it is a local subsidiary of a foreign MNC (e.g., IBM in Germany). Unfortunately, the patent database has about 175,000 assignee names, and it is impossible to match all assignees to their parents. For 2

Since the remaining countries account for less than 10% of the USPTO patents, I found that adding more countries did not change the aggregate results, and was not useful for extending individual country results. So I dropped these to keep the number of citing and cited country fixed effects manageable in my econometric model.

14

Table 2.1: Overview of patent data Country

United States Japan Germany France United Kingdom Canada Subtotal 6 countries Other countries Total worldwide

Total patents Total number Assigned Fraction of 1986-95 in of assigned patents with patents from NBER patents clean parent multinational database information subsidiaries (1) (2) (3) (4) 546,824 418,045 287,787 8.5% 217,313 212,427 183,870 2.1% 74,041 67,154 45,869 19.5% 29,791 27,120 17,289 20.4% 26,631 23,968 15,131 40.3% 20,700 13,015 5,697 50.0% 915,300 761,729 555,643 9.0% 94,924 73,115 38,402 27.3% 1,010,224 834,844 594,045 10.2%

15

example, there is no systematic rule as to whether patents originating from researchers based in a German subsidiary of IBM would be listed under the parent firm “IBM” or a separate assignee “IBM Germany” (or a name from which it is even harder to infer that this is a subsidiary of IBM).3 To construct my parent-subsidiary database, I inspected about 10,000 assignees as follows. First, Compustat-based parent firm identifiers (from 1989) from Jaffe and Trajtenberg (2002) were used to match around 4,600 patent assignees to 2,500 parent firms. Second, Stopford’s Directory of Multinationals (1992) was used to match around 2,800 additional assignees with 200 parent firms. Third, using USPTO assignee information, keyword search and the Internet, about 400 government-affiliated bodies, 550 research institutes and 450 universities worldwide were identified. Finally, the ownership of another 1,000 major patent assignees was checked using a combination of Who Owns Who directories (1991) and data from company web sites. As Table 2.1 shows, the above steps account for about 556,000 patents, which is about 73% of all assigned patents. The remaining patents were dropped.4 About 9% of all patents arise from foreign MNC subsidiaries, though the fraction varies a lot across countries (Table 2.1, column 4).5 Although this variation is interesting in itself, exploring it is beyond the scope of this paper.

3

To avoid the situation in which a company could not be identified with a unique parent, I define an assignee to be an MNC subsidiary when a foreign firm has a majority stake in it. For cases where two firms had a 50-50 stake, I broke the tie in favor of the first firm. See Mowery, Oxley and Silverman (1996) or Gomes-Casseres, Jaffe and Hagedoorn (2003) for an in-depth study of alliances. 4

The main results reported below continue to hold if, instead of dropping any of the remaining assignees, I included them as independent entities, with the home country calculated as the country where most of its patents originate. 5

These numbers approximately equal estimates for the fraction of national R&D coming from MNC subsidiaries in these countries, as reported by OECD (1998). This serves as an additional validation for my dataset construction.

16

4. Preliminary Analysis Innovations in similar technologies are likely to be located in the same region, often for reasons other than potential knowledge spillovers. Therefore, to avoid overestimation of the localized knowledge spillover effect, it is important to control for the geographic distribution of technological activity. Jaffe, Trajtenberg and Henderson (1993) suggest a “matching” approach that takes this into account by defining the appropriate benchmark as being the citation frequency from the original patents to randomly drawn patents with similar technological and temporal characteristics as the originally cited patents. 4.1. The Matching Approach Existing studies typically use a 3-digit technological classification for the matching methodology suggested by Jaffe, Trajtenberg and Henderson (1993). However, Thompson and Fox-Kean (2004) show that this is not detailed enough to prevent overestimation of localized knowledge flows (Thompson and Fox-Kean, 2004). To overcome this issue, I start by using the 9-digit subclass information available from USPTO. Since this detailed classification consists of around 150,000 sub-classes, I am able to have a much finer control for geographic distribution of technological activity. Following standard practice, all citations for which either the original or the control patent involved a self-cite from an organization to itself were excluded from the sample. Since the time lag between two patents is also an important determinant of the probability of citation, the final sample only included those cited patents for which a control patent could be found with an application year within one year of the original. This leads to dropping about half of the citations from the original data, an issue I revisit in the next section.

17

To examine evidence for knowledge flows from MNC subsidiaries to domestic organizations, I examine if the fraction of MNC patents (i.e., patents originating from local subsidiaries of foreign MNCs) is higher in the set of patents cited by domestic organizations than in the set of control patents. The t-statistic used to formally test this is given by t M →D =

p M → D − p ′M → D p M → D (1 − p M → D ) p ′M → D (1 − p ′M → D ) + ND ND

where pM→D is the ratio of number of actual citations from domestic organizations to MNC subsidiaries to the total number of citations (ND) made by domestic entities, and p’M→D is the analogous ratio for the control citations. I similarly compute the t-statistics to test for domestic-to-multinational (D→M) knowledge flows. 4.2. Results from Matching Table 2.2(a) gives analysis of localized knowledge diffusion from local subsidiaries of foreign MNCs to domestic organizations (M→D flows). Column (1) gives the total number of citations made by domestic organizations, and columns (2) and (3) respectively give the number and fraction of these made to patents by local subsidiaries of foreign MNCs. Columns (4) and (5) report the same analysis for patent pairs obtained by replacing each original cited patent by its control patent. Column (6) reports the difference of proportions from columns (3) and (5), and column (7) shows that a t-test rejects their equality. Column (8) gives the ratio of the two proportions (which I call the M→D index). The overall M→D index of 1.13 indicates that the probability of knowledge flow from a patent by an MNC subsidiary to a domestic

18

Table 2.2(a): Knowledge diffusion from MNC subsidiaries to domestic organizations (M→D) Actual Citations (1) Country

Total citations by domestic

(2)

Control Citations (3)

(4)

Citations by %Citations domestic to by domestic mult sub to mult sub

Comparison (5)

(6)

(7)

(8)

Citations by %Citations domestic to by domestic mult sub to mult sub

(3) - (5)

t-ratio

(3)/(5)

United States Japan Germany France United Kingdom Canada

430,262 245,441 27,326 12,727 7,895 3,536

17,010 2,082 658 124 197 32

3.95% 0.85% 2.41% 0.97% 2.50% 0.90%

15,136 1,879 542 101 149 15

3.52% 0.77% 1.98% 0.79% 1.89% 0.42%

0.44% 0.08% 0.42% 0.18% 0.61% 0.48%

10.7 3.2 3.4 1.5 2.6 2.5

1.12 1.11 1.21 1.23 1.32 2.13

Total

727,187

20,103

2.76%

17,822

2.45%

0.31%

11.9

1.13

19

Table 2.2(b): Knowledge diffusion from domestic organizations to MNC subsidiaries (D→M)

Actual Citations (1) Country

Total citations by mult sub

(2)

Control Citations (3)

(4)

Citations by %Citations mult sub to by mult sub domestic to domestic

Comparison (5)

(6)

(7)

(8)

Citations by %Citations mult sub to by mult sub domestic to domestic

(3) - (5)

t-ratio

(3)/(5)

United States Japan Germany France United Kingdom Canada

41,272 5,156 10,841 3,856 9,689 3,457

22,590 2,464 1,302 166 220 38

54.73% 47.79% 12.01% 4.30% 2.27% 1.10%

18,799 2,083 985 114 274 25

45.55% 40.40% 9.09% 2.96% 2.83% 0.72%

9.19% 7.39% 2.92% 1.35% -0.56% 0.38%

26.5 7.6 7.0 3.2 -2.5 1.6

1.20 1.18 1.32 1.46 0.80 1.52

Total

74,271

26,780

36.06%

22,280

30.00%

6.06%

24.9

1.20

20

patent is 13% more likely than for two geographically random patents with similar technological and temporal characteristics. In Table 2.2(b), a similar approach shows significant knowledge flows from domestic organizations to local subsidiaries of foreign MNCs (D→M flows). The magnitude of the D→M index (1.20) is found to be even larger than the M→D case discussed above. Thus, not only does the localization of knowledge diffusion result still hold, the extent of knowledge diffusion is even stronger than the M→D case. In other words, MNC subsidiaries are better at gaining knowledge from domestic organizations than the latter are at gaining knowledge from the former. I will test this claim formally using my regression framework below.

5. Citation-Level Regression Methodology In addition to the 3-digit vs. 9-digit technological classification issue that I have already addressed above, Thompson and Fox-Kean (2004) point out two other challenges in using the matching approach. First, dropping observations with imperfect matches can lead to a systematic bias in the measured knowledge flow patterns. Second, while the matching approach focuses on the “primary” technological classification, most patents also have several “secondary” technology classes and subclasses, with the primary versus secondary distinction not necessarily being a true reflection of a patent’s fundamental characteristics. The matching approach does not capture the fact that technological relatedness of patents could show up as an overlap along any of their subclasses, and not just as their primary class or subclass being the same.

21

To overcome these challenges, I use a citation-level regression framework to model the probability of citation between two patents. Imagine that the probability that a patent K cites a patent k is given by a “citation function” P(K, k). My interest lies in studying how P(K, k) differs with characteristics of the cited and citing players. Among the explanatory variables, I include dummy variables for all dimensions along which I would have ideally liked to do the matching. This gives the flexibility of using multiple control variables to better control for propensity to cite even in cases where good matches do not exist.6 5.1. Choice-Based Sampling Since the number of potentially citing and cited patents can be of the order of a million, the number of all possible dyads (K, k) can be of the order of a trillion. In principle, one could take a random sample of patent dyads from the population of all possible dyads. One could then define a binary variable y that equals 1 if the citation actually takes place, and 0 otherwise, and estimate the citation function by assuming that it can be approximated using a logistic functional form. In other words, the dichotomous dependent variable y would be taken as a Bernoulli outcome that takes a value 1 for observation i with the probability

Pr( y = 1 | x = xi ) = Λ ( xi β ) =

1 1 + e − xi β

where xi is the vector of covariates and β is the vector of parameters to be estimated. However, an estimation approach based on random sampling of patent pairs is not 6

Some regression-based studies use an aggregate number of citations as the dependent variable. These models include a measure of “average technological distance” between sets of citing and cited patents using only a 2 or 3-digit technology classification. So the issue of bias remains because of within-set heterogeneity: sets with technologically closer patents have more frequent citations and also greater colocation of patents.

22

practical because citations between random pairs of patents are very rare: there are only about seven actual citations for every one million potential citations, making estimation impossible even with very large samples. From an informational point of view, it would be desirable to have a higher fraction of observations with y = 1 in the sample. This can be achieved by a “choicebased” sampling procedure that deliberately oversamples the patent pairs with y = 1.7 In this approach, the sample is formed by taking a fraction α of the population’s dyads with y = 0, and a fraction γ of the dyads with y = 1, α being much smaller than γ. Since this stratification is done on the dependent variable, however, using the usual logistic estimates would lead to a selection bias. A technique that overcomes this problem is the weighted exogenous sampling maximum likelihood (WESML) estimator suggested by Manski and Lerman (1977). The central idea is to explicitly recognize the difference in sampling of 0’s and 1’s by weighting each term in the log likelihood function by the inverse of the ex ante probability of inclusion of the corresponding observation in the sample. In other words, each sample observation is weighted by the number of elements it represents from the overall population in order to make the choice-based sample “simulate” a random exogenous sample. The WESML estimator is obtained by maximizing the following weighted “pseudo-likelihood” function

ln Lw =

1

γ

∑ ln(Λ i ) +

{ yi =1}

1

n

∑ ln(1 − Λ i ) = − ∑ wi ln(1 + e (1−2 yi ) xi β )

α { y = 0}

i =1

i

where wi = (1 / γ ) y i + (1 / α )(1 − y i ) . In addition, the appropriate estimator of the

7

Please see appendix 2.1 for technical details of the methodology discussed here. For a general discussion of choice-based sampling, see Amemiya (1985, pp. 319-338), Greene (2003, p. 673) or King and Zeng (2001). Sorenson and Fleming (2001) have also used this technique for predicting patent citations.

23

asymptotic covariance matrix is White’s robust “sandwich” estimator used for pseudomaximum likelihood estimation. Further, since the same citing patent can occur in multiple observations, the standard errors should be calculated without assuming independence across these observations. 5.2. Sample Construction Since robust standard errors can be quite large for weighted logit estimation (Green, 2003, p. 673), I use relatively large samples to ensure statistically meaningful analysis. In addition, I improve the efficiency of estimation through stratification on technological characteristics of the citing and cited patents. In other words, each actual citation is matched with “control citations” with the same 3-digit technology classes for the citing and cited patents. The weighted likelihood function described above has to be generalized by defining the weight attached to a y = 0 observation as the reciprocal of the ex ante probability of a y = 0 population pair with the same respective technological cell (i.e., the combination of technological classes for the citing and cited patents) being selected into the sample. I define the population of possible citations as all pairs of citing and cited patents in my data (over half a million patents from 1986-1995) such that the citing year does not come before the cited year. The sample used in regression analysis was drawn from this population as follows: First, all actual citations (y = 1) were included in the sample, except for self-citations from a geographical division of an organization to itself. Each of these “ones” was then matched with multiple potential citations (y = 0) that have the same “cell” as defined by the characteristics of the actual citation. This was done while making sure that no self-citation from a geographical division of an organization was

24

included among the control citations either. This led a sample of 5.57 million actual and potential citations. 5.3. Control Variables for Probability of Citation As the time lag between the citing and cited patents increases, the citation probability is known to increase initially and then fall beyond a certain point (Jaffe and Trajtenberg, 2002). To control for this, I introduce dummy variables for the number of years of lag between the citing and cited patents. In addition, since the patent citation rate may change over time, additional dummy variables are used to capture the citing year fixed effects. Since patents in different industry categories have different propensities to cite others, I also include fixed effects for the broad technological category of the citing patent, as defined in the Jaffe and Trajtenberg (2002) patent database. The next issue is that innovators in different countries might have a different propensity to cite patents registered with the USPTO. For example, a US patent filed by a European inventor might not necessarily cite a USPTO patent for an innovation, but might instead cite the corresponding European Patent Office (EPO) patent for that innovation. In order to avoid possible biases arising from this, all regressions include citing country fixed effects. A later section uses EPO data to carry out additional robustness checks comparing propensity to cite for MNCs and domestic firms within the same country. Patents that are technologically similar have a higher probability of citation. Existing patent citation literature typically compares the 3-digit technological class information on the citing and cited patents to control for this. However, this can lead to bias estimates since there can be large heterogeneity in technology within a 3-digit class.

25

For example, the 3-digit class “Aeronautics” includes 9-digit classes as diverse as “Spaceship control” and “Aircraft seat belts” (Thompson and Fox-Kean, 2004). To take this into account, I define dummy variables for the same broad technological category (1 out of 6), the same technological subcategory (1 out of 36), the same 3-digit primary class (1 out of 418) and the same 9-digit primary class (1 out of 150,000). Further, since the designation of a subclass as “primary” can sometimes be ad hoc, I also include a dummy variable that captures whether at least one of the secondary subclasses of a patent is the same as one of the primary or secondary subclasses for the other patents. While there is a chance that even these technology controls are not perfect, these are the most fine-grained level possible with USPTO data, and are much more detailed than the coarse controls used in most studies.

6. Results 6.1. Intra-Region and Intra-MNC Knowledge Flows (Hypotheses 1 and 2) Table 2.3 gives a summary of relevant variables used in the regressions. The results of weighted logit regressions (WESML) appear in Table 2.4, where the dependent variable is 1 for patent pairs that have a citation, 0 otherwise. Column (1) reproduces the empirical “fact” that knowledge flows are particularly strong within the same country and the same MNC. These effects, however, may partly result from technological specialization of regions and firms (Jaffe, Trajtenberg and Henderson, 1993). This is found to indeed be the case in column (2), where including controls at the 3-digit classification level reduces the estimated effects for within same country

26

Table 2.3: Summary of variables used for regressions analysis Same tech category

Indicator variable that is 1 if both the citing and the potentially cited patent belong to the same broad industry category (one of 6) as defined in the Jaffe and Trajtenberg (2002) database

Same tech subcategory

Indicator variable that is 1 if both the citing and the potentially cited patent belong to the same broad technical subcategory (one of 36) as defined in the Jaffe and Trajtenberg (2002) database

Same primary tech class

Indicator variable that is 1 if both the citing and the potentially cited patent belong to the same 3-digit primary technology class (one of about 450) as defined by USPTO

Same primary subclass

Indicator variable that is 1 if both the citing and the potentially cited patent belong to the same 9-digit primary technology subclass (one of about 150,000) as defined by USPTO

Secondary subclass overlap

Indicator variable that is 1 if at least one of the secondary 9-digit subclasses of one patent is the same as a primary or secondary subclass of the other patent in the dyad

Within same country

Indicator variable that is 1 if the citing and cited patents originate from inventors located in the same country

Within same MNC Indicator variable that is 1 if the citing and cited patents are from two divisions (located in different countries) of the same MNC D→D

Indicator variable that is 1 if both the citing and potentially cited patent belong to the same country, with assignees for both being domestic players in the country

D→M

Indicator variable that is 1 if both the citing and potentially cited patent belong to the same country, with assignee for the former being a local subsidiary of a foreign multinational and for the latter being a domestic player

M→D

Indicator variable that is 1 if both the citing and potentially cited patent belong to the same country, with assignee for the former being a domestic player and for the latter being a local subsidiary of a foreign multinational

M→M

Indicator variable that is 1 if both the citing and potentially cited patent belong to the same country, with assignees for both local subsidiaries of foreign multinationals

S→H

Indicator variable that is 1 if citing patent is from the home base of an MNC and the cited patent is from a foreign subsidiary (located abroad) of the same MNC

H→S

Indicator variable that is 1 if citing patent is from the local subsidiary of a foreign MNC and the cited patent is from the home base (located abroad) of the same MNC

Log(1 + number of patents that originate in the same country as the potentially cited Presence of citing assignee in patent and are assigned to the citing entity) cited country Presence of cited Log(1 + number of patents that originate in the same country as the citing patent and assignee in citing are assigned to the potentially cited entity) country Scale of citing assignee

Log(number of worldwide patents for 1980-99 that are assigned to the citing entity)

Scale of cited assignee

Log(number of worldwide patents for 1980-99 that are assigned to the cited entity)

27

Table 2.4: Intra-national and intra-MNC knowledge flows (1)

(2)

(3)

Within same country

0.672** (0.009) [3.83]

0.578** (0.005) [3.29]

0.520** (0.009) [2.96]

Within same MNC

3.291** (0.110) [18.76]

2.110** (0.026) [12.03]

1.825** (0.050) [10.40]

1.148** (0.011)

1.108** (0.012)

Same tech subcategory

1.246** (0.014)

1.218** (0.015)

Same primary tech class

3.243** (0.011)

1.930** (0.015)

Technological relatedness: Same tech category

Same primary subclass

2.282** (0.028)

Secondary subclass overlap

4.111** (0.012)

Number of observations

5,577,206

5,577,206

5,577,206

A weighted logit regression is used The dependent variable is 1 if there is a citation between two patents, 0 otherwise Robust standard errors in parentheses, with clustering on citing patent Marginal effects in square brackets after multiplication with 1,000,000 Fixed effects used for technological category, country of citing patent, citing patent year and time lag ** significant at 1%; * significant at 5%

28

and within same MNC. Column (3) addresses the concern, raised by Thompson and FoxKean (2004), that commonly used controls just for the 3-digit technological class are not sufficient. In particular, this specification controls for additional similarity along 9-digit primary technological classification as well as overlap of secondary technological classes between the citing and cited patents. The results show that, though absence of detailed controls was indeed leading to the biases, the estimates for within same country and within same MNC still remain significant. While statistical significance is not a surprise given the large sample size, let us now check for economic significance. The marginal effects are reported in square brackets, after multiplying by a million for readability.8 Since the predicted citation rate between two random patents is found to be about 5.70 in a million, the marginal effect of 2.96 for within same country suggests that patents from different organizations within the same country are about 52% more likely to have a citation than are otherwise similar patents from different organizations in different countries. Similarly, the marginal effect of 10.4 for within same MNC shows that patents from different international divisions of the same MNC are around 3 times as likely to have a citation than are those from different organizations in different countries, a finding consistent with that of Gomes-Casseres, Jaffe and Hagedoorn (2003). 6.2. Details of Intra-National Knowledge Flows (Hypotheses 3, 4, 5 and 6) Table 2.5 breaks up the within same country knowledge flows into 4 types: between domestic entities (D→D), from domestic entities to local subsidiaries of The marginal effect of a variable j is given by βj Λ’(xβ). From the logit form, it is easy to show that this equals βj Λ(xβ)[1-Λ(xβ)]. One can then substitute either the mean predicted probability or the population mean for Λ(xβ) for getting an estimate of the marginal effect. I report the former. The latter estimate is typically slightly higher in value.

8

29

Table 2.5: Break-up of intra-national and intra-MNC knowledge flows Within same country D→D

0.525** (0.010) [2.99]

D→M

0.521** (0.032) [2.97]

M→D

0.366** (0.030) [2.09]

M→M

0.768** (0.096) [4.38]

Within same MNC S→H

1.796** (0.080) [10.24]

H→S

1.848** (0.061) [10.53]

Observations

5,577,206

ID→M / ID→D IM→D / ID→D IM→M / ID→D

0.99

IM→D / ID→M IH→S / IS→H

0.70**

0.70** 1.46**

1.03

A weighted logit regression is used The dependent variable is 1 if there is a citation between two patents, 0 otherwise Robust standard errors in parentheses, with clustering on citing patent Marginal effects in square brackets after multiplication with 1,000,000 Controls for technological similarity of citing and cited patent included in regression, but not shown Fixed effects used for technological category, country of citing patent, citing patent year and time lag ** significant at 1%; * significant at 5% (In case of ratios, whether statistically different from 1 is tested)

30

foreign MNCs (D→M), from MNC subsidiaries to domestic entities (M→D) and between MNC subsidiaries (M→M). Figure 2.1 illustrates these definitions for clarity. The reference category is the cross-border inter-organizational knowledge flows, compared with which D→D knowledge flow probability is found to be greater by 3.0 in a million, D→M probability is greater by 3.0 in a million, M→D probability is greater by 2.1 in a million and M→M probability is greater by 4.4 in a million. Given that the average citation rate between two random patents is 5.7 in a million, all four kinds of intra-national knowledge flow effects are quite large in relative magnitude. The fact that M→D and D→M flows are both positive and significant, with the latter exceeding the former, is consistent with the earlier findings using matching (Table 2.2). Table 2.5 also breaks down the within same MNC category into two subcategories: knowledge flows from a foreign subsidiary of an MNC to its home base (S→H), and from its home base to the foreign subsidiary (H→S). The comparable (and statistically indistinguishable) estimates suggest that the probability with which a patent from a foreign subsidiary cites one from the MNC’s home base is about the same as that with which a patent from the home base cites one from the subsidiary. This is consistent with a view of MNCs as a “learning organization”, where subsidiaries not only build upon the knowledge of the home base but also contribute to further learning (Kogut and Zander, 1993; Dunning, 1993). The bottom of the table reports the relative magnitude and statistical comparison of different estimates. The coefficient for M→D flows is 30% smaller than for D→M flows, as indicated by the ratio βM→D / βD→M of 0.7. A test of equality

31

Intel [Home Base]

D →D

IBM

NEC

[Home Base]

[Home Base]

M →D

Sony

M →M

D →M

S →H H →S

NEC

USA Figure 2.1: Six kinds of knowledge flows

32

IBM

Japan

of βM→D and βD→M is rejected at the 1% significance level. Similarly, M→D flows are statistically smaller than the D→D flows (by 30%). D→M flows, on the other hand, are not any weaker in strength than D→D flows. Thus, the intensity of knowledge flows from domestic organizations to MNC subsidiaries is statistically no different from that between domestic organizations themselves. There is little evidence that MNC subsidiaries face a “liability of foreignness” (i.e., are unable to tap into the localized knowledge exchange in a country). To summarize, while MNCs are as good at learning from domestic organizations as domestic organizations are at learning from each other, MNCs contribute somewhat less to local learning.9 It is interesting to note that multinational subsidiaries are also really good at learning from each other, with the M→M estimate being much greater than that for even D→D or D→M knowledge flow. This is consistent with previous findings on knowledge spillovers between MNC subsidiaries (Head, Ries and Swenson, 1995; Feinberg and Majumdar, 2001; Feinberg and Gupta, 2003). In analysis not reported here, I found the M→M effect to be driven largely by the probability of knowledge flow being very high between foreign subsidiaries of MNCs from the same home country. 6.3. Cross-Country Differences in Bi-directional Knowledge Flows (Hypothesis 7) What is the underlying mechanism for the result that knowledge flows from the host countries to the MNCs exceed those back from the MNCs to the host countries?

9

In order to rule out the possibility the result is due to knowledge flows from domestic universities/research labs to MNC subsidiaries, I included separate dummy variables for whether the D→M flows were originating from domestic firms or domestic universities/research labs. I found that the D→M flows originating from domestic firms are actually slightly higher rather than lower than the D→M flows from domestic universities/research labs.

33

To dig deeper into this issue, I repeat the above analysis for the six individual countries. In Table 2.6, I interact each of the six indicator variables discussed earlier with dummy variables for countries. I find evidence of strong intra-national knowledge flows in all countries. The aggregate finding that D→M knowledge flows are stronger than M→D knowledge flows holds true for the US, Japan and Germany.10 The equality of the twoway flows cannot be rejected for France and Canada, while the trend actually reverses for the UK. One explanation for this pattern is that the domestic firms and organizations in the US, Japan and Germany are, on an average, technologically more advanced than the average subsidiary of a foreign multinational based there, and therefore have much less to learn from the latter. R&D data from OECD (1998) supports this explanation: the R&D intensity (i.e., R&D/production) of domestic firms and foreign MNCs differs most in Germany and Japan, with the domestic R&D intensity being almost twice of that for MNC subsidiaries. It is therefore no surprise that the disparity between D→M and M→D flows is also highest for these two countries. Likewise, the fact that UK is the only country where D→M knowledge flows are significantly weaker than M→D knowledge flows is consistent with the fact that UK is the only country where the R&D intensity of MNCs exceeds that of domestic players.

10

Thus, though Japanese firms gain by investing in the US, US firms also gain by investing in Japan, giving no evidence of Japanese firms being worse overall at sharing knowledge, a finding consistent with Spencer (2000).

34

Table 2.6: Intra-national and intra-MNC knowledge flows in different countries Country of origin of citing patent US Within same country D→D 0.517** (0.013)

Japan

Germany

France

UK

Canada

0.535** (0.016)

0.503** (0.042)

0.526** (0.089)

0.688** (0.141)

1.406** (0.173)

D→M

0.491** (0.037)

0.579** (0.081)

0.941** (0.114)

0.700** (0.148)

0.281* (0.109)

0.865** (0.213)

M→D

0.371** (0.032)

0.255* (0.103)

0.461** (0.082)

0.719** (0.149)

0.670** (0.143)

1.015** (0.245)

M→M

0.695** (0.120)

1.357** (0.354)

0.633** (0.235)

1.738** (0.338)

0.934** (0.167)

1.061** (0.309)

S→H

1.925** (0.107)

1.771** (0.212)

1.153** (0.204)

1.357** (0.192)

1.920** (0.211)

2.383** (0.292)

H→S

1.607** (0.115)

2.097** (0.251)

2.203** (0.145)

1.964** (0.120)

1.644** (0.095)

2.177** (0.100)

Country fixed effect

-

-0.384** (0.014)

-0.319** (0.021)

-0.248** (0.018)

-0.064 (0.038)

-0.022 (0.028)

ID→M / ID→D IM→D / ID→D IM→M / ID→D

0.95 0.72** 1.34

1.08 0.48** 2.54*

1.87** 0.92 1.26

1.33 1.37 3.30**

0.41* 0.97 1.36

0.62* 0.72 0.75

IM→D / ID→M IH→S / IS→H

0.76** 0.83*

0.44** 1.18

0.49** 1.91**

1.03 1.45**

2.38* 0.86

1.17 0.91

Within same MNC

A weighted logit regression is used The dependent variable is 1 if there is a citation between two patents, 0 otherwise Robust standard errors in parentheses, with clustering on citing patent Controls for technological similarity of citing and cited patent included in regression, but not shown Fixed effects used for technological category, country of citing patent, citing patent year and time lag ** significant at 1%; * significant at 5% (In case of ratios, whether statistically different from 1 is tested)

35

6.4. Cross-Sector Differences in Bi-directional Knowledge Flows (Hypothesis 7) To investigate the heterogeneity in knowledge flows further, I now look at cross-sector differences since learning-related incentives for location choice are greater for technologies where new knowledge plays an important role (Audretsch and Feldman, 1996). In particular, when locating abroad can lead to learning, both industry laggards and leaders have an incentive to open overseas subsidiaries. On the other hand, when the learning opportunities are small compared with potential leakage of their own technology, the leaders have less incentive to locate abroad. To explore this, I now break down analysis of innovations originating in the US into six broad technology categories.11 The sample used in Table 2.7 includes only the citing patents from the US. I interact each of the six indicator variables discussed earlier with dummy variables for technological categories. Although this coarse technological classification surely hides heterogeneity within technological categories, some interesting patterns still emerge. First, “Drugs & Medical” and “Chemical”, two of the most R&D intensive sectors, show high levels of knowledge exchange among all players. This is consistent with Chung and Alcacer (2002), who suggest that these are sectors where not just the foreign industry laggards but also industry leaders actively locate advanced facilities in the US. For example, all foreign pharmaceutical firms invest heavily in R&D in the US in order to keep abreast with the latest developments in a sector that involves discrete product innovation and a long uncertain product innovation process: R&D intensity for Pharmaceuticals is 10.5% for MNC subsidiaries, which is even higher 11

I would have liked to repeat the sector-level analysis for other individual countries, and for a finer sector classification, but the smaller resulting sample sizes for patents by MNC subsidiaries made that impractical.

36

Table 2.7: Knowledge flows for different sectors in the U.S. Technological category of citing patent Chemical

Computers & Drugs & Communicatio Medical

Electrical & Mechanical Electronic

Other

0.390** (0.029)

0.650** (0.021)

0.671** (0.068)

0.438** (0.025)

0.251** (0.028)

0.826** (0.055)

D→M

0.401** (0.065)

0.687** (0.056)

0.645** (0.185)

0.420** (0.082)

0.151 (0.102)

0.587** (0.112)

M→D

0.400** (0.063)

0.390** (0.064)

0.650** (0.103)

0.100 (0.079)

0.169* (0.073)

0.760** (0.121)

M→M

0.492* (0.208)

0.745** (0.184)

1.633** (0.228)

0.401 (0.358)

-0.124 (0.285)

1.749** (0.239)

S→H

1.861** (0.231)

1.780** (0.147)

2.270** (0.406)

1.747** (0.249)

2.504** (0.252)

1.895** (0.488)

H→S

1.875** (0.212)

1.024** (0.190)

2.351** (0.336)

1.638** (0.275)

2.052** (0.290)

1.461* (0.656)

Category fixed effect

-

0.900** (0.027)

-0.725** (0.059)

0.511** (0.029)

0.612** (0.030)

-0.372** (0.048)

ID→M / ID→D IM→D / ID→D IM→M / ID→D

1.03 1.03 1.26

1.06 0.60** 1.15

0.96 0.97 2.43**

0.96 0.23** 0.92

0.60 0.67 -0.49

0.71** 0.92 2.12**

IM→D / ID→M IH→S / IS→H

1.00 1.01

0.57** 0.58**

1.01 1.04

0.24** 0.94

1.12 0.82

1.29 0.77

Within same country D→D

Within same MNC

A weighted logit regression is used The dependent variable is 1 if there is a citation between two patents, 0 otherwise Robust standard errors in parentheses, with clustering on citing patent Controls for technological similarity of citing and cited patent included in regression, but not shown Fixed effects used for technological category, country of citing patent, citing patent year and time lag ** significant at 1%; * significant at 5% (In case of ratios, whether statistically different from 1 is tested)

37

than the 6.5% figure for domestic firms (OECD, 1998). Since MNC subsidiaries in these sectors are quite advanced, it is natural that the issue of weak M→D flows resulting from adverse selection in the technological competence of subsidiaries would not exist in these sectors. Two individual sectors where D→M knowledge flows are indeed significantly stronger than M→D knowledge flows are “Computers & Communication” and “Electrical & Electronics”. This is consistent with Chung and Alcacer’s (2002) finding that FDI in these sectors is dominated by industry laggards. For example, R&D intensity for Computers is 4.5% for MNC subsidiaries and 13.5% for domestic firms in the US (OECD, 1998). This is also consistent with Florida’s (1997) finding that 37% of the MNC subsidiaries in the US for these sectors have a “listening post” role, as opposed to only 17% in “Chemicals” and 25% in “Drugs & Medical.” For the “Mechanical” category, all three kinds of localized knowledge flows involving MNC subsidiaries are weaker than D→D flows, possibly because it is not a particularly knowledge-intensive sector. 6.5. Cross-Border Citations between Different Firms (Hypothesis 8) The above analyses study intra-national, inter-firm knowledge flows (D→D, D→M, M→D and M→M) and cross-border, within-firm knowledge flows (S→H and H→S). Taken together, the two show that MNC subsidiaries are an intermediary for cross-border, inter-firm knowledge flow. I now look for possible direct effect of an MNC’s subsidiary activity on the probability of cross-border citation between different firms (i.e., between host country domestic players and the MNC home base). Two caveats should be made: First, this is a very strong test. While an increased

38

probability of cross-border citation between different firms suggests intense knowledge flow, a zero effect does not indicate an absence of such knowledge flow since knowledge flowing indirectly through a subsidiary need not result in crossborder citation between different firms. Second, the findings are based on a crosssectional comparison, without modeling the endogeneity of the decision to locate overseas. I define the “presence” of the citing assignee in the cited country as the logarithm of the number of patents originating from its subsidiary in the cited country. This can be seen as a measure of its local absorptive capacity (Cohen and Levinthal, 1989). Similarly, I define the “presence” of the cited assignee in the citing country as the logarithm of the number of patents originating from its subsidiary in the citing country. In addition to the control variables already discussed above, additional controls used are the logarithm of worldwide patenting by the citing assignee and by the cited assignee. This ensures that the foreign presence variables do not simply pick up overall scale effects, which would arise if larger assignees systematically differ in the propensity to cite or be cited. Since I am now interested only in cross-border patent citations between different players, all patent pairs from the same firm or the same country are now dropped. The regression results are reported in Table 2.8. The negative estimate for the global scale of the citing assignee suggests that larger firms rely much less on external sources of knowledge, perhaps because they build more upon their own internal knowledge. Similarly, the positive estimate for the global scale of the cited assignee

39

Table 2.8: Effect of MNC subsidiary activity on cross-border citations Presence of citing assignee in cited country

0.030** (0.004) [0.16]

Presence of cited assignee in citing country

0.011** (0.004) [0.06]

Scale of citing assignee

-0.012* (0.006) [-0.06]

Scale of cited assignee

0.031** (0.005) [0.17]

Observations

3,027,928

A weighted logit regression is used The dependent variable is 1 if there is a citation between two patents, 0 otherwise Robust standard errors in parentheses, with clustering on citing patent Marginal effects in square brackets after multiplication with 1,000,000 Controls for technological similarity of citing and cited patent included in regression, but not shown Fixed effects used for technological category, country of citing patent, citing patent year and time lag ** significant at 1%; * significant at 5%

40

suggests that patents from larger firms have a greater likelihood of being cited by other firms. As discussed above, the variables of most interest to us are the presence of the citing assignee in the cited country, and that of the cited assignee in the citing country. The marginal effects of these variables can be interpreted follows. A 1% increase in inventive activity by a foreign MNC’s local subsidiary increases the citation probability by the foreign MNC’s home base to the host country’s domestic players by 3% (recall that regressions use log of presence, hence the percentage interpretations). In contrast, there is only a 1.1% increase in citation probability by the host country’s domestic players to the foreign MNC’s home base when the MNC’s local innovative activity goes up by 1%. Thus, though increased MNC activity is associated with increased cross-border patent citations in both directions, the asymmetry found in intra-national citations exists even for cross-national patent citations: the MNC home base gains more in terms of inter-organizational knowledge spillovers from its overseas investments than the domestic players in the host country do. These findings are consistent with similar results found in more specialized settings by Branstetter (2000) for Japanese FDI in the US, and Globerman, Kokko and Sjöholm (2000) for inward and outward FDI for Sweden. Further, when I analyzed the data separately for the six countries, increased presence of citing MNC in cited country had a positive and significant effect on citation probability in five of the six countries: US, Japan, France (at 10% significance level), UK and Canada. On the other hand, increased presence of potentially cited MNC in the citing country had a positive and significant effect in only 2 countries:

41

Japan and Canada (at 10% level). Once more, this suggests that the latter result is weaker than the former.

7. Further Issues in Using USPTO Patent Citations All regressions in this paper include country fixed effects to control for systematic cross-country differences in propensity to cite USPTO patents. However, this does not resolve a related concern that MNC subsidiaries and domestic organizations even within the same country might differ in their propensity to cite USPTO patents. In particular, patents from MNC subsidiaries might have a systematically different tendency to cite USPTO patents and instead cite a patent representing the same innovation but registered with another country’s patent office.12 To look into this possibility, I examined citations made to both USPTO and European Patent Office (EPO) patents by a random sample of 1,612 USPTO patents from 1995, about half of them originating in domestic organizations and the other half in MNC subsidiaries. For each patent in the sample, I identified if one or more cited EPO patents did have equivalent USPTO patents that could equivalently have been cited, and therefore represent “missing citations” in USPTO data. The mapping from EPO to USPTO patents was done using the “OECD Triadic Patent Families” database, which has information on patents filed for the same innovation at both USPTO and EPO. The results are summarized in Table 2.9. The mean number of citations to USPTO patents by a patent from the above sample was 5.85, while the mean number

12

Since USPTO patents provide patent protection only in the US, a patent needs to be separately applied for in Europe for protection there.

42

Table 2.9: Frequency of USPTO and EPO citations by a USPTO patent

Citing patents from all countries All assignees Domestic MNC (N=1,612) (N=810) (N=802)

Citing patents from US Domestic MNC (N=436) (N=369)

Citing patents not from US Domestic MNC (N=374) (N=433)

Mean number of citations to USPTO patents

5.84

5.68

6.00

6.75

6.95

4.42

5.19

Mean number of citations to EPO patents

1.12

0.83

1.41

0.77

1.42

0.89

1.41

Mean number of citations to EPO patents with "equivalent" US patents in the OECD triadic database

0.32

0.22

0.43

0.24

0.39

0.21

0.46

43

of citations to EPO patents was 1.13. A large fraction of the EPO citations did not have an equivalent USPTO patent, hence do not reflect any bias in the estimate of probability of citation between just the innovations captured by USPTO patents. The mean number of citations to EPO patents that do have equivalent USPTO patents, which really gives the number of “missing citations” described above, was only 0.32 per patent. The missing citations are thus quite small in number compared with citations that do get made to USPTO patents. Further, as Table 2.9 shows, the average number of missing citations per patent from MNC subsidiaries (0.43) is a little higher than those for domestic organizations (0.22). This holds both in the sub-sample of patents originating in the US (0.39 for MNC subsidiaries and 0.24 for domestic organizations), and for those that originate elsewhere (0.46 for MNC subsidiaries and 0.21 for domestic organizations). In either sub-sample, the missing citation bias therefore is in the direction of underestimating the extent of localized knowledge diffusion to MNCs more than to domestic organizations. In other words, if we could correct for this bias in the previous analysis, it would slightly strengthen the main result that probability of D→M knowledge flow exceeds that of M→D knowledge flow.

8. Discussion and Concluding Remarks Much of the recent debate on globalization has centered on whether MNCs contribute as much as they gain from their host countries. To address one aspect of this broad issue, I study how the extent of knowledge flows from MNCs to a host country compares with knowledge acquisition by MNCs from the host country. Analysis of patent

44

citation data reveals that, while local subsidiaries of foreign MNCs help a country gain access to knowledge originating in foreign firms, they also cause domestic technology to fall into the hands of foreign competitors. Thus, knowledge spillovers from inward FDI, particularly in countries that possess valuable technology of their own, are not free – they come at the cost of significant “leakage” of domestic knowledge. Knowledge flows from domestic organizations to MNCs are found to significantly exceed those from MNCs to domestic organizations for three of the six largest economies (US, Japan and Germany), and two of the six broad technological categories (“Computers & Communications” and “Electrical & Electronics”). The above patterns are consistent with a hypothesis that net knowledge flows from foreign MNC subsidiaries to domestic players are strongest in countries and industries where MNC subsidiaries are involved in knowledge-intensive activities. For the policy maker, it implies that not just the magnitude of FDI but also its level of sophistication should be considered in pursuing knowledge spillovers. Policies should focus on attracting FDI that is technologically sophisticated, and on sectors where the host country is a technological laggard. Further, the findings suggest that outward FDI might sometimes be more effective than inward FDI for acquiring knowledge originating abroad. Thus, instead of only promoting inward FDI and discouraging outward FDI, a country might gain from encouraging its domestic firms to also seek out foreign sources of knowledge. There are three caveats to any policy interpretation of my results. First, knowledge diffusion effects are only a part of the overall welfare effects of MNCs. Second, patent citation data does not allow separate measurement of knowledge transfers

45

(which are planned, priced and paid for) and knowledge spillovers (which are unintended externalities). Third, endogeneity of the MNC’s decision of whether and where to locate overseas is not incorporated in my model. The focus of this paper has been developed countries, partly because patent data is not as meaningful a source of information for developing countries. In particular, knowledge spillovers in developing countries lead less often to radical innovation and more often result in adoption of existing technologies. Also, since domestic organizations are rarely as advanced as foreign MNCs, the learning effect in developing countries might be weaker for MNCs and stronger for domestic organizations. But the general point made in the paper should still apply: not only the magnitude but also the knowledge content of investments by foreign MNCs affects the possibility of knowledge spillovers. Different kinds of MNC activity, like state-of-theart R&D or production facilities versus simple assembly operations, might have different implications for knowledge flows. Future research on FDI should therefore focus less on just measurement of knowledge spillovers, and more on the conditions needed for and the mechanisms driving such spillovers.

46

Appendix 2.1. A Note on Choice-Based Sampling and WESML In samples where the fraction of y=1 observations (the “ones”) is very small, the information content is much greater in the ones rather than the zeroes. To see this, recall that the asymptotic covariance matrix for the MLE for logit is given by (see Greene, 2003, p. 672)

n ' ∑ Λ i (1 − Λ i ) xi xi   i =1 

−1

If the logit model has some explanatory power, Λi is larger (i.e. closer to 0.5 for rare events) when yi =1. Thus Λi(1-Λi) is larger, implying that having a higher fraction of 1’s in the sample would reduce variance. Choice-based sampling tries to achieve this by over-sampling on the “ones” from the population. The sample is formed by taking a fraction α of the population’s dyads with y = 0, and a fraction γ of the dyads with y = 1, where α is much smaller than γ. The probability of a citation conditional on the dyad being in the sample flows from Bayes’ rule:

Λ'i =

γ Λi γ = = γ Λ i + α (1 − Λ i ) γ + αe − βX

1

i

1+ e

γ  −(ln   + βX i ) α 

The extra term ln(γ/α) in the exponent leads to a bias. However, since the functional form is still logistic, a simple estimation strategy is to simply subtract ln(γ/α) from the estimate for the constant term of the usual logit. The efficiency of the correction, however, depends crucially on the logit functional form not being misspecified (Manski and Lerman, 1977; Cosslet, 1981). An alternate method, which is not as sensitive to

47

model misspecification, is the weighted exogenous sampling maximum likelihood (WESML) estimator suggested by Manski and Lerman (1977). The WESML estimator is obtained by maximizing the following weighted “pseudo-likelihood” function:

1

ln Lw =

γ

∑ ln(Λ i ) +

{yi =1}

1

n

∑ ln(1 − Λ i ) = − ∑ wi ln(1 + e (1−2 yi ) xi β )

α {y =0} i

i =1

where wi = (1 / γ ) y i + (1 / α )(1 − y i ) . In other words, each sample observation is weighted by the number of elements it represents from the overall population in order to make the choice-based sample “simulate” a random exogenous sample. Here is some intuition on why WESML works: Let the joint probability density be g(x,y) for the sample, and g*(x,y) for the population. Let the fraction of elements with y = j be f(j) in the sample, and f*(j) in the population (j = 0,1). Let n and N be sample size and population size respectively, and nj and Nj be the number with y = j. Using conditional probability rules, g ( x, j ) = Pr( x | y = j ) f ( j ) =

g * ( x, j ) f ( j ) g * ( x, j )(n j / n) N / n = = g * ( x, j ) f * ( j) Nj /N w( j )

where w(j) = Nj/nj is the reciprocal of the sampling rate for observations with y = j. Let P(yi) be the probability of y = yi conditional on x = xi in the population. Then, the expected value of the weighted likelihood function is

  n E ln Lw = ∫  ∑ w( y i )[ln P ( y i )]g ( x, y i )dx   i =1

  N /n = ∑  ∫ w( y i )[ln P( y i )] g * ( x, y i )dx  w( y i ) i =1   n

=

N n

 n  [ln P( y i )]g * ( x, y i )dx ∫  ∑ i =1  48

Thus, ignoring the constant scaling factor N/n, the expected value of the weighted log likelihood equals the expected log likelihood for the same sample resulting through random exogenous sampling from the population. As shown formally in Amemiya (1985, section 9.5.2), this ensures consistency of WESML estimation. The choice-based WESML procedure described above can be extended to allow “matched samples”. This involves taking all actual citations (y=1) and matching each of these with k “control citations” (y=0) along a dimension z (e.g., the “cells” indexed by the vector combination of the citing technological class and cited technological class). Without loss generality, denote the values z can take as 1, 2, …, T. For a matching-based sampling design, it is easier to think of not just y but (z, y) as the dependent variable. In forming the likelihood function, I will use the result that

Pr( z = z i and y = j | x = xi ) = Pr( z = z i | xi ) Pr( y = j | z = z i and x = xi )

= Pr( z = z i | xi ) Pr( y = j | x = xi ) The second equality assumes that the vector x includes all information about z that affects citation outcome y, i.e., x is a sufficient statistic for z. The log-likelihood function for estimation using an exogenous random sample of size n would therefore be n

ln L = ∑ ln[Pr( z = z i and y = y i | xi )] i =1

n

= ∑ {y i ln[Pr( z = z i | xi )Λ ( xi β )] + (1 − y i ) ln[Pr( z = z i | xi )(1 − Λ ( xi β ) )]} i =1

This forms the basis for deriving the pseudo-likelihood function for choice-based sampling. Each log likelihood function term has to be weighted by the inverse of the probability that the corresponding population element will be included in the sample. To

49

derive these weights, denote the number of elements with z = t and y=j as ntj for the sample and Ntj for the population. Matching ensures that, from each cell, I pick all elements with y=1 and k times as many elements with y=0. In other words, nt1 = Nt1 and nt0 = kNt1. Also, since Ntj is known, the probability ptj of a population element with z = t and y = j getting selected in our sample is easily calculated as pt1= nt1/Nt1=1 and pt0= nt0/Nt0 = kNt1/Nt0 for all values of t. Denoting wtj = 1/ptj, the weighted likelihood function ln(Lw) for choice-based sampling is the given by

∑ {y w n

i =1

i

zi 1

}

ln[Pr( z = z i | xi )Λ( xi β )] + (1 − yi ) wzi 0 ln[Pr( z = z i | xi )(1 − Λ( xi β ) )] n

(

= C − ∑ wi ln 1 + e (1− 2 yi ) Xβ

)

i =1

where wi = y i wzi 1 + (1 − y i ) wzi 0

n

and C = ∑ wi ln[Pr( z = z i | xi )] i =1

Since C is independent of β, it can be ignored in the maximum likelihood procedure. Thus, a weighted logit estimation can be used, where the weights of the observations are now given by wi. Unlike the simple WESML with random sampling from the y=0 observations, the weights now depend not just on the value of y but also on the cell that the observations falls into.

50

Chapter 3: COLLABORATIVE NETWORKS AS DETERMINANTS OF KNOWLEDGE DIFFUSION PATTERNS 1. Introduction The ease with which knowledge diffuses has important implications for innovation and growth (Grossman and Helpman, 1991). However, even though ideas are intangible in nature, empirical evidence shows that they do not flow freely across regional and firm boundaries. Two patterns of knowledge diffusion have been identified. First, knowledge flows are geographically localized (Jaffe, Trajtenberg and Henderson, 1993). Second, knowledge flow is easier within firm boundaries than between firms (Kogut and Zander, 1992). This paper studies collaborative networks among individuals as the mechanism driving both these patterns of knowledge diffusion. Numerous factors, including informal networks, institutions, norms, language, culture, incentives, and other formal and informal mechanisms might also affect the ease with which knowledge diffuses. However, this paper studies the extent to which the observed knowledge diffusion patterns can be accounted for simply by the fact that people within the same region or firm have close collaborative links that might facilitate flow of complex knowledge. In particular, I analyze the extent to which direct and indirect collaborative ties between inventors help account for the effect of geographic colocation and firm boundaries on the probability of knowledge flow between individual inventors of U.S. patents. Following previous research, I use patent citations to measure these micro-level knowledge flows. The probability of knowledge flow is estimated using a novel regression framework based on choice-based sampling (Manski and Lerman,

51

1977). This approach helps address some methodological concerns regarding existing use of citations for measuring knowledge diffusion (Thompson and Fox-Kean, 2004). A rich literature in sociology studies information flow through interpersonal networks (Ryan and Gross, 1943; Coleman, Katz and Mendel, 1966; Granovetter, 1973; Burt, 1992; Rogers, 1995). However, different kinds of networks might be effective for transmitting different kinds of information. For example, in their study of transmission of complex technical knowledge from publicly funded research to private pharmaceutical firms, Cockburn and Henderson (1998) conclude: “It is important that these researchers [of private firms] be active collaborators with public sector researchers. Reading the journals, attending conferences, even being an active player on the informal network of information transfer within the industry are insufficient” (p. 163). Motivated by their findings, I rigorously examine a large dataset to investigate the extent to which diffusion of complex technical knowledge can be explained by collaborative ties between individuals. My analysis allows the possibility that direct and indirect ties could matter to a different extent. For example, if an individual X has a direct collaborative relationship with individual Y, and Y has a direct tie with Z, Z might learn indirectly about X’s work through his tie with Y. To measure the directness of collaborative ties among over a million inventors in the U.S. patent database, I construct a “social proximity graph” based on information about the team of inventors for each individual patent. This graph allows me to derive a measure of “social distance” between inventors. Three recent papers are particularly related to this study. Stolpe (2001) uses patent data to test if direct collaborative links between individuals lead to knowledge diffusion, but does not find empirical support for this in the specific setting of liquid crystal display

52

technology. Agrawal, Cockburn and McHale (2003) show that patents by inventors who move from one geographic region to another continue to be cited by former collaborators from their original region, reflecting that direct ties resulting from past collaborations can continue to be a mechanism for knowledge flow even across regions. Breschi and Lissoni (2002) find the association between patent citations and geographic co-location in Italy to be greater for socially connected patent teams, suggesting that there might be important interaction effects between geographic co-location and collaborative links. I build upon this stream of research by using a much larger dataset and improved methodology to study the impact of both direct and indirect collaborative ties on micro-level knowledge flows, and by further extending the analysis to study if these collaborative ties help explain observed patterns of intra-regional and intra-firm knowledge flow. My analysis reveals that collaborative networks have a strong influence on knowledge diffusion, with direct collaborative ties being more effective than indirect ties. Further, the effect of being in the same region or the same firm on probability of knowledge flow falls significantly once collaborative networks have been accounted for. In fact, conditional on having close collaborative ties, geographical co-location and firm boundaries have little effect on probability of knowledge flow. In contrast, for patent pairs with only indirect collaborative ties or no collaborative ties at all, geographic colocation and firm boundaries continue to be associated with greater probability of knowledge flow, possibly because of other kinds of formal and informal mechanisms influencing intra-regional and intra-firm knowledge flow. The paper is organized as follows. Section 2 motivates my formal hypotheses. Section 3 describes the patent citation data as well as the data on inventors. Section 4

53

introduces my citation-level regression framework for estimating probability of knowledge flow, and also describes how I measure collaborative ties using a “social proximity graph”. Section 5 reports the empirical findings. Section 6 discusses limitations of this study. Section 7 offers implications and concluding thoughts.

2. Hypotheses This analysis in this paper is comprised of three main parts, as summarized in Figure 3.1 and detailed in the formal hypotheses appearing in this section. The first part is to formally establish the “fact” that intra-regional and intra-firm knowledge flow is more intense than that across regions and firms. The second part is to test the extent to which existence and directness of collaborative links between individuals determines the probability of knowledge flow between them. The third part, which forms the crux of this paper, is to combine the results from the first two parts in order to examine the extent to which collaborative networks explain the more intense knowledge flow within regions and firms. While previous work has found empirical support for geographic localization of knowledge flows (e.g., Jaffe, Trajtenberg and Henderson, 1993), recent work raises methodological concerns that could have led to over-estimation of this phenomenon in existing research (Thompson and Fox-Kean, 2004). Therefore, before trying to explain intra-regional knowledge flows, I first test if the result does hold even when using a new approach (explained later) that addresses some of these concerns. Hypothesis 1. The probability of knowledge flow within a region exceeds that between different regions, even after controlling for technological specialization of regions.

54

Same region

Greater probability of knowledge flow

Same firm

Figure 3.1(a): Hypotheses 1 and 2

Greater probability of knowledge flow

Close collaborative links between indivduals

Figure 3.1(b): Hypotheses 3 and 4

Same region Same firm

Close collaborative links between indivduals

Figure 3.1(c): Hypotheses 5 and 6

55

Greater probability of knowledge flow

The second pattern of knowledge diffusion that I study is that firms transmit knowledge more effectively than would be possible through a market-mediated mechanism (Kogut and Zander, 1992). Before examining collaborative networks as a possible driver for this, I formally reproduce this result by testing the following hypothesis: Hypothesis 2. The probability of knowledge flow within a firm exceeds that between different firms, even after controlling for technological specialization of firms. Mobility of individuals has been shown to be one mechanism through which knowledge gets acquired by existing firms (Saxenian, 1994; Almeida and Kogut, 1999; Rosenkopf and Almeida, 2003) as well as start-ups (Klepper, 2001; Gompers, Lerner and Scharfstein, 2002). However, even in the absence of direct mobility of individuals, information and knowledge can diffuse through interpersonal networks (Zander and Kogut, 1995; Zucker, Darby and Brewer, 1998; Shane and Cable, 2002; Stuart and Sorenson, 2003; Uzzi and Lancaster, 2003). This paper focuses specifically on interpersonal ties that arise either from direct collaboration between inventors or indirect links between them through other inventors they both have links with. The next hypothesis is that such links do indeed matter for transmission of knowledge. Hypothesis 3. The probability of knowledge flow is greater between inventors with a direct or indirect collaborative tie than between inventors that are not connected in the collaborative network. Direct and indirect ties might have different implications for transmitting knowledge. Granovetter (1973) emphasizes that ties providing access to non-redundant information might be more valuable. While indirect ties provide non-redundancy, and

56

hence might be more efficient for transmission of simple codifiable information, direct ties are potentially more useful for transferring knowledge that is complex and not easily codified (Ghoshal, Korine and Szulanski, 1994; Uzzi, 1996; Hansen, 1999). The codified part of such knowledge (e.g., the subset of knowledge behind an innovation that gets codified as a patent description) may represent just the “tip of the iceberg”, with the remaining knowledge being “tacit” (Polanyi, 1966; Nelson and Winter, 1982; Kogut and Zander, 1992). Transmission of such knowledge may need close interaction between individuals (Allen, 1977; Nonaka, 1994; Szulanski, 1996). In addition, direct relationships might also induce more trust, improving willingness of individuals to share knowledge (Tsai and Ghoshal, 1998; Levin and Cross, 2003). Transmission of complex technical knowledge should therefore become more difficult as the “social distance”, or the number of intermediaries needed to pass knowledge from the source to the destination, increases. This suggests the following hypothesis: Hypothesis 4. The probability of knowledge flow between individuals is a decreasing function of the social distance between them. Now I come to the main hypotheses of interest, which is to study the extent to which the results from Hypotheses 1 and 2 can be explained by the collaborative networks from Hypotheses 3 and 4. Sorenson and Stuart (2001) show that geographical localization of venture capital investments is a result of localized flow of information regarding investment opportunities, which in turn results from localized interpersonal ties in the venture capital community. Analogously, I test if the correlation between geographic co-location and knowledge flow can be explained by the fact that

57

collaborative networks are more likely to exist between people from the same region, as given by the following formal hypothesis: Hypothesis 5. Controlling for collaborative networks leads to a significant drop in the effect of geographic co-location of inventor teams on the probability of knowledge flow between them. The alternate hypothesis is that geographic concentration of knowledge flows is driven not by collaborative networks but by other mechanisms such as informal interaction (“ideas in the air”) or region-specific factors like local infrastructure, institutions, regional publications, communication channels, norms, culture and government policies. Analogous to studying why intra-regional knowledge flows are strong is the question of why knowledge flows are stronger within firms than between firms. Like Simon (1991) and Grant (1996), I take individuals as the unit of analysis for studying knowledge flows even within organizations. Kogut and Zander (1992) describe firms as “social communities in which individual and social expertise is transformed into economically useful products and services by the application of a set of higher-order organizing principles” (p. 384). However, applying a unified network framework to both inter-firm and intra-firm knowledge flows implies that studying “higher-order organizing principles” is beyond the scope of this paper. However, I do explore how much of a firm’s ability to transfer knowledge between its employees can be explained simply by the fact that it is a tightly knit “social community” in the specific sense of having a dense collaborative network. This gives my final hypothesis:

58

Hypothesis 6. Controlling for collaborative networks leads to a significant drop in the effect of firm boundaries on the probability of knowledge flow between two teams of inventors. The alternate hypothesis here might be that intra-firm knowledge flows are driven not by collaborative networks of individuals but by other mechanisms such as informal interactions within organizations, organizational learning routines, confidentiality-related barriers, legal obstacles or incentive issues associated with firm boundaries.

3. Patent Data 3.1. Patent Citations as Measure of Knowledge Flow My dataset on US patents was constructed by merging data from the US Patent Office (USPTO) with an enhanced version made available by Jaffe and Trajtenberg (2002). Despite several challenges, patents are perhaps the best available measure of innovation for large-sample research (Griliches, 1990). A major issue with using patent data is that only some of the innovations are patented (Levin, Klevorick, Nelson and Winter, 1987). Since this makes counts of patents and patent citations misleading as raw measures, I only estimate the probability of knowledge flow between two innovations that do end up as patents, without claiming that these comprise all the innovations. Patent citations leave behind a trail of how a new innovation potentially builds upon existing knowledge. An inventor is legally bound to report relevant “prior art”, with the patent examiner serving as an objective check. Unlike academic papers, there is usually an incentive not to include superfluous citations, as that might reduce the scope of one’s own patent. There are, however, two factors that add noise to citations as a measure

59

of knowledge flow. First, citations might be included by the inventor for strategic reasons (e.g., to avoid litigation). Second, a patent examiner might add citations to patents that the original inventor knew nothing about. Recent studies comparing citation data with inventor surveys show that the correlation between patent citations and actual knowledge flow is indeed high, but not perfect (Jaffe and Trajtenberg, 2002; Duguet and MacGarvie, 2002). The defense given for the common use of patent citations for research is that use of citations should be appropriate in large-sample studies as long as the noise does not bias the results of interest. Note that viewing patent citations as being correlated with knowledge flows is not the same as claiming that patents themselves are the mechanism behind these knowledge flows. Consider the analogy that a PhD student may cite research papers of his advisor, even though knowledge gained by working closely with the advisor could be much more than what could be captured in the advisor’s papers. Since I would like to distinguish between knowledge flows within and between firms, the data had to be cleaned to correctly identify the firm associated with each patent. This was a non-trivial exercise because a firm’s patents may be listed under the name of one of its subsidiaries. Through a process described in chapter 2 in detail, I performed parent firm identification using a combination of available Compustat-based parent firm identifiers, Stopford’s Directory of Multinationals, Dun and Bradstreet’s Who Owns Whom directories and Internet sources. About 3,000 major firms were identified in the process, and this paper studies patents filed by these firms during 1986-95.13

13

I restricted the sample to 1986-95 since the parent-subsidiary match used data sources from around 1990. The 3,000 firms account for about half of all patents. The rest are scattered among individuals and 165,000 firm and non-firm organizations. Non-firm entities were not included to keep the inter-firm vs. intra-firm comparison clean.

60

To study the effect of geographic co-location on probability of knowledge flow, a “region” was defined as one of the states in the U.S. While I would have liked to study knowledge flows at an even finer geographic unit of analysis, data constraints allowed me to study localization of knowledge flows only at the level of the state. Also, I focus only on innovations arising in the U.S. because my dataset does not have clean state-level information for other countries. 3.2. Inventors Each patent includes the name and address of each of its individual inventors. A challenge in using this data, however, is correctly identifying when two different records refer to the same person. To this end, I use information on the first, middle and last names of inventors, and on the technological characteristics of their patents. I experimented with several methods to avoid too many “false positives” (different individuals being incorrectly identified as being the same) and too many “false negatives” (different records of the same inventor being incorrectly identified as having two different inventors). As a reasonable compromise, I finally arrived at an algorithm that identified two records as having the same inventor if and only if the following three conditions held: 1. The first and last names matched exactly. 2. The middle initials, if available, were the same. 3. When the middle initial field was blank in at least one of the two records, the records also overlapped on at least one of their technology "subcategories". The “subcategory” definition in the last condition is taken from Jaffe and Trajtenberg (2002), who divide the 418 US patent classes into 38 different

61

subcategories. Using only the first two conditions would have identified around 1.3 million distinct inventors. The third condition makes the matching criteria more stringent, leading to around 1.7 million inventors. I tried to rule out more “false positives” by requiring the finer patent class itself to overlap, or looking for an overlap of patent citations across patents. However, using either of these extra conditions led to too many "false negatives", since the overlap across records of the same inventor turned out to be lower than I had expected. I also considered requiring an additional match for street address and/or assignee firm, as used by Fleming, Colfer, Marin and McPhie (2004). However, I decided against it because interaction of collaborative links with geography and firm boundaries is a central focus of this paper, so using geography or firm identity for matching might bias these results. Also, as Fleming, Colfer, Marin and McPhie (2003) find, forcing these requirements would make the match too conservative, an issue they handle by not requiring the requirements for uncommon last names. There would, irrespective of the algorithm used, definitely be some errors in any matching process. However, unless there is a reason to believe that the matching is producing systematic errors, it should lead to an attenuation bias that only understates the effect of collaborative networks on probability of knowledge diffusion. Therefore, any effect I find for collaborative networks could be interpreted as a lower bound for its real effect.

62

4. Empirical Methodology Imagine that the probability that a patent K cites a patent k is given by a “citation function” P(K, k). Our interest lies in estimating what drives this probability. One could define a binary variable y that equals 1 if the citation actually takes place, and 0 otherwise, and estimate the citation function by assuming that it can be approximated using a logistic functional form. 4.1. Choice-Based Sampling As already discussed in section 5 of Chapter 2, a WESML estimator based on choice-based sampling (Manski and Lerman, 1977) is again appropriate for estimating the probability that there is a citation between any two patents. Once more, since technological similarity of two patents is a strong determinant of the probability of citation, estimation efficiency can be improved by matching each citing pair in the sample with a set of “control pairs” such that the citing and cited patent in each control pair belong to the same respective technology class as those in the original citing pair.14 The WESML approach again can be generalized by defining the weight attached to a y = 0 observation to be the reciprocal of the ex ante probability of a y = 0 population pair with the same technological characteristics being selected into the sample. In addition, I assigned each actual citation (i.e., y = 1 observation) a weight of one since all actual citations were included in the sample. This procedure led to a sample with over 2.5 million observations.

14

Sorenson and Stuart (2001) use a similar research design for estimating probability of venture capital funding.

63

4.2. Control Variables for Probability of Citation As the time lag between the citing and cited patents increases, the citation probability is known to increase initially and then fall (Jaffe and Trajtenberg, 2002). To control for this, my regressions use fixed effects for the difference between the application years of the patents. In addition, I also use fixed effects to capture systematic differences in citation rates over time. Further, I include fixed effects for the technological category of the citing patent to capture cross-sector differences in citation rates. Another key concern is that technologically similar patents have a greater probability of citation. Existing patent citation literature typically compares the 3-digit technological class of the citing and cited patents to control for this. However, this can lead to biased estimates, since there can be large heterogeneity in technology even within a 3-digit class. For example, the 3-digit class “Aeronautics” includes 9-digit subclasses as diverse as “Spaceship control” and “Aircraft seat belts” (Thompson and Fox-Kean, 2004). To take this into account, I define dummy variables for the same broad technological category (1 out of 6), the same technological subcategory (1 out of 36), the same 3-digit primary class (1 out of 418) and the same 9-digit primary class (1 out of 150,000). Further, since the designation of a subclass as “primary” can sometimes be ad hoc, I also include a dummy variable that captures whether at least one of the secondary subclasses of a patent is the same as one of the primary or secondary subclasses for the other patent. While there is a chance that even these technology controls are not perfect, these are the most fine-grained level possible with

64

USPTO data, and are much more detailed than the coarse controls used in most existing studies.15 4.3. Measuring Social Distance between Innovating Teams In order to measure the existence and directness of collaborative ties between inventors, I define “social distance” as the number of intermediaries needed to pass knowledge from the source to the destination. This is analogous to measuring “degrees of separation” in recent work on the “small worlds” phenomenon (Watts and Strogatz, 1998; Newman, 2001). In using collaboration data (e.g., on a patent, research paper, project, etc.), it is standard practice to assume that an observed collaboration marks the beginning of a tie between the individuals, which persists beyond the recorded collaboration (Stolpe, 2001; Breschi & Lissoni, 2002; Agrawal, Cockburn and McHale, 2003; Fleming, Colfer, Marin and McPhie, 2003). I follow this convention here. Data on inventors and inventing teams can be represented using an “affiliation matrix” A = {aij}, where aij is “1” if the ith inventor is on the collaborating team for the jth patent, “0” otherwise (Wasserman and Faust, 1994). Figure 3.2 gives an example, with 7 inventors A, B, C, D, E, F and G, and 7 patents P1, P2, P3, P4, P5, P6 and P7. A value of “1” for element (A, P1) and “0” for element (C, P1), for example, implies that A is one of the inventors for patent P1, but C is not. The first step for studying collaborative links between inventors is to construct a “social proximity graph”. The graph for year t includes as nodes all innovations 15

Some regression-based studies use the number of citations as the dependent variable (e.g., Jaffe and Trajtenberg, 2002). These models include a measure of “average technological distance” between citing and cited sets of patents using only a 2 or 3-digit technology classification. So the issue of bias remains: sets with a greater fraction of patent pairs with the same 9-digit technology have a greater probability of citations, and also more co-location of patents.

65

Innovating Team (Patent) Inventor P1 P2 P3 P4 P5 P6 P7 A 1 1 0 0 0 0 0 1 0 0 1 0 0 0 B C 0 1 1 0 0 0 0 D 0 0 1 0 1 0 0 E 0 0 0 0 1 0 1 F 0 0 0 0 1 0 0 G 0 0 0 0 0 1 1 Year

1986 1987 1988 1989 1989 1989 1990

Figure 3.2: An affiliation network

66

made by year t, with an edge between patenting teams X and Y if and only if the two teams have a common inventor.16 For example, in Figure 3.3(a), there is a common inventor A between teams for patents P1 and P2, which Figure 3.4 represents as a social distance of “0” for P1 → P2. Any two patents not linked via a common inventor might still be linked through other inventors. For example, in Figure 3.3(b), knowledge from P1 can flow to P3 indirectly via the path P1 → P2 → P3 (i.e., by being passed from A to C, with A and C having a collaborative link as evidenced by P2). To measure the closeness of such collaborative links, the social distance between any two such teams can be defined as the number of intermediate nodes on the minimum path (the geodesic) between the two. Thus the social distance is “1” for P1 → P3. Since knowledge flows are meaningful only from an innovation that happens earlier to one that happens later, social distance need not be defined for P2 → P1, P1 → P1, P2 → P2, etc., as indicated in Figure 3.4. Now consider Figure 3.3(c). The above definition suggests a social distance of “1” for P2 → P4, since there is a path P2 → P1 → P4. Does this make sense even though P1 precedes P2 in time? If the year of their recorded collaboration were literally the only time when knowledge passed between the inventors, the application year of every intermediate patent on the minimum path would have to exceed that of the one preceding it, and there would be no path of knowledge flows from P2 to P4. However, as discussed earlier, since a recorded collaboration between A and B is interpreted as the beginning of a collaborative tie between the two, B (who is the 16

The “Small Worlds” literature (Watts and Strogatz, 1998; Newman, 2001) uses nodes to represent individuals instead of teams, with edges between individuals that have collaborated. For this paper, it is more natural to define the collaborating teams as nodes since measured knowledge flows are from one team to another.

67

68

Patent P2 (1987) A

Inventors: A, C

Patent P1 (1986) Inventors: A, B

Figure 3.3(a): Social proximity graph for 1987

Patent P2 (1987)

C

Inventors: A, C

A

Patent P3 (1988) Inventors: C, D

Patent P1 (1986) Inventors: A, B

Figure 3.3(b): Social proximity graph for 1988

Patent P2 (1987) A

C

Inventors: A, C

Patent P3 (1988)

D

Inventors: C, D

Inventors: D, E, F Patent P4 (1989)

B

Patent P1 (1986)

Patent P5 (1989)

Inventor: B

Inventors: A, B

Patent P6 (1989) Inventor: G

Figure 3.3(c): Social proximity graph for 1989

Patent P2 (1987) A Patent P1 (1986)

C

Inventors: A, C

Patent P3 (1988) Inventors: C, D

B

Inventors: A, B

D

Patent P5 (1989) Inventors: D, E, F Patent P4 (1989)

Patent P7 (1990)

Inventor: B

Inventor: E, G

Patent P6 (1989) Inventor: G

Figure 3.3(d): Social proximity graph for 1990

69

E

G

Source Team

Destination Team P1 P2 P3 P4 P5 P6 P1 . 0 1 0 2  P2 . . 0 1 1  P3 . . . 2 0  P4 . . . . 3  P5 . . . 3 .  P6 . . .   . P7 . . . . . .

P7 3 2 1 4 0 0 .

Figure 3.4: Social distance between nodes

70

inventor for P4) can build upon knowledge of P2 that she may gain through her ties with A. Thus knowledge can flow “backwards” along the link P1 → P2, and then on to the link P2 → P4. Likewise, knowledge from P3 could be passed by C to A, and then further from A to B through the chain of ties P3 → P2 → P1 → P4, making the social distance P3 → P4 to be “2”. The social proximity graph changes over time. I use separate social proximity graphs for t=1986 through t=1995 to cover all the years for which I analyze knowledge flows. To measure social distances for innovating teams from year t, we need to use a graph of collaborative ties already in place by t. For example, the correct value of social distance from P3 to P6 is infinity (since P6 took place in 1989, and P3 and P6 are not even in the same connected component in 1989) and not “2” (as an incorrect interpretation of the 1990 graph might suggest).17 There are two practical issues in using the social distance measure as defined above. First, it imposes a rigid functional form assumption and potentially mixes “apples and oranges” into a single cardinal measure (e.g., the common inventor case with distance=0 and the past collaboration case with distance=1). Second, because of the large graph size, computing exact pair-wise social distances is practically impossible.18 Fortunately, it is still practical to classify all observations into five mutually exclusive and exhaustive categories based on whether the social distance is 0, 17

I construct the graph for year t using all collaborations from the first year in my data (1975) until year t. Since the social distance measure might not be comparable across years, I use year fixed effects. An alternate approach could be to use a rolling time window, e.g., use collaborations from year t-7 to t in defining the graph for year t.

18

Wasserman and Faust (1994) suggest computing pair-wise distances by defining element xij of a matrix X as 1 if there is an edge between nodes i and j, 0 otherwise. The distance between i and j is then the smallest number p such that the pth power matrix of X (i.e., p-1 multiplications of X into itself) has a nonzero entry (i, j). Unfortunately, this and other similar approaches become impractical for very large graphs (Cormen, Leiserson and Rivest, 1990).

71

1, 2, any finite value greater 2, or infinity (i.e., no social links).19 As Table 3.1 shows, I capture the first four cases as categorical variables common inventor, past collaboration, common past collaborator and indirect social link, with the no social link case being the reference category in all regressions.

5. Results 5.1. Intra-region and intra-firm knowledge flows Table 3.1 gives a summary of variables used in the regressions. Table 3.2 formally tests Hypotheses 1 and 2 (i.e., that knowledge flows are particularly strong within the same region or the same firm). The weighted logit framework described above is used to estimate the probability of citation between patents, with the dependent variable being 1 when a patent pair has a citation, 0 otherwise. Column (1) finds positive and significant estimates for within same region and within same firm. However, this could result simply from technological specialization of regions and firms (Jaffe, Trajtenberg and Henderson, 1993). As column (2) shows, including controls for technological relatedness (at the level of 3-digit technological class) between patents reduces the estimated coefficients for within same region and within same firm. However, Thompson and Fox-Kean (2004) have shown that even the 3digit technological controls, though extensively used in existing literature, are insufficient. To address this, column (3) uses additional controls based on a detailed 9digit primary and secondary technological classification of patents. The estimates for 19

I explicitly find out all pairs with a social distance of 0, 1 or 2 by calculating the first three power matrices mentioned above, since these matrices are sparse and computationally manageable. I then distinguish between having a more indirect social link and no social link by identifying all connected components of a graph.

72

Table 3.1: Definition of variables Within same region

Indicator variable that is 1 if the citing and cited patents originate from inventors located in the same region, i.e., the same state within US

Within same firm

Indicator variable that is 1 if the citing and cited patents are owned by the same parent firm

Same tech category

Indicator variable that is 1 if both the citing and the potentially cited patent belong to the same broad industry category (one of 6) as defined in the Jaffe and Trajtenberg (2002) database

Same tech subcategory

Indicator variable that is 1 if both the citing and the potentially cited patent belong to the same broad technical subcategory (one of 36) as defined in the Jaffe and Trajtenberg (2002) database

Same primary tech class

Indicator variable that is 1 if both the citing and the potentially cited patent belong to the same 3-digit primary technology class (one of about 450) as defined in the US Patent classification system

Same primary subclass

Indicator variable that is 1 if both the citing and the potentially cited patent belong to the same 9-digit primary technology subclass (one of about 150,000) as defined in the US Patent classification system

Secondary subclass overlap

Indicator variable that is 1 if at least one of the secondary 9-digit subclasses of one patent is the same as a primary or secondary subclass of the other patent in the dyad

Common inventor

Indicator variable that is 1 if there is at least one common inventor between the citing and the cited patents. This corresponds to social distance of 0.

Past collaboration

Indicator variable that is 1 if there is no common inventor between the two patents, but at least one inventor of the citing patent has collaborated with an inventor of the cited patent in the past. This corresponds to social distance of 1.

Common past collaborator

Indicator variable that is 1 if neither of the above two hold, but there is a common collaborator who has worked with an inventor of the citing patent and an inventor of the cited patent in the past. This corresponds to social distance of 2.

Indirect network link

Indicator variable that is 1 if none of the above three cases hold, but the two patents still belong to the same connected component of the social proximity graph. This corresponds to social distance of >2 but finite.

73

Table 3.2: Intra-region and intra-firm knowledge flows (1) 1.413** (0.051) [16.96]

(2) 1.050** (0.017) [12.60]

(3) 0.798** (0.020) [9.58]

3.781** (0.060) [45.37]

2.622** (0.022) [31.46]

2.217** (0.025) [26.60]

1.176** (0.026)

1.173** (0.023)

Same tech subcategory

1.161** (0.029)

1.105** (0.029)

Same primary tech class

2.637** (0.023)

1.545** (0.030)

Within same region

Within same firm

Technological relatedness: Same tech category

Same primary subclass

1.793** (0.043)

Secondary subclass overlap

3.688** (0.020)

Number of observations

2,528,764

2,528,764

2,528,764

A weighted logit regression is used The dependent variable is 1 if there is a citation between two patents, 0 otherwise Robust standard errors in parentheses, with clustering on citing patent Marginal effects in square brackets after multiplication with 1,000,000 Fixed effects for technological category, application year and time lag ** significant at 1%; * significant at 5%

74

within same region and within same firm fall further, but still remain significant. Since statistical significance is not a surprise given the large sample size, I now turn to the magnitude of these effects. The marginal effects for the weighted logit model are shown in square brackets in column (3) of Table 3.2, after being multiplied by a million for readability.20 The predicted citation rate between two random patents turned out to be about 12 in a million. Therefore, the reported marginal effect of 9.58 for within same region implies that patents from the same region are 80% more likely to have a citation than are otherwise similar patents from different regions. Similarly, the marginal effect of 26.6 for within same firm implies that patents from the same firm are over 3 times as likely to have a citation than are patents from different firms. 5.2. Effect of social distance on knowledge flows As discussed earlier, Table 3.1 defines common inventor, past collaboration, common past collaborator and indirect social link as dummy variables to capture a social distance of 0, 1, 2 and > 2 (but finite). If two patents belong to the same connected component in the social proximity graph, exactly one of these dummy variables is 1. Table 3.3 reports summary statistics for these variables. For the entire sample, the fraction of pairs belonging to the same connected component is 64.7% for pairs with citations, and only 48.9% for pairs with no citation, consistent with the hypothesis that connectedness leads to greater probability of citation. The inequality continues to hold true for the sub-sample without self-citations by firms, where the fraction of pairs

For logit, the marginal effect of a variable j can be shown to be βj Λ(xβ)[1-Λ(xβ)]. I substitute the mean predicted probability for Λ(xβ) into this expression in order to get an estimate of the marginal effect.

20

75

Table 3.3: Summary statistics

Entire sample Citations Controls (N=552,427) (N=1,976,337)

No self-citations by firms Citations Controls (N=349,251) (N=1,881,299)

Common inventor (Social distance = 0)

0.1512

0.0033

0.0132

0.0001

Past collaboration (Social distance = 1)

0.0593

0.0036

0.0079

0.0004

Common past collaborator (Social distance = 2)

0.0343

0.0052

0.0085

0.0011

Indirect social link (Social distance > 2 but finite)

0.4024

0.4767

0.5133

0.4775

Any social link

0.6472

0.4888

0.5429

0.4791

An entry in this table represents mean value of the variable for the corresponding row in the subset of the population indicated in the corresponding column.

76

belonging to the same connected component is 54.3% for pairs with citations, and only 47.9% for pairs with no citation. Table 3.4 reports regression analysis to test Hypotheses 3 and 4 (i.e., the impact of collaborative links on probability of patent citation). As a comparison of columns (1) and (2) shows, controlling for technological relatedness of patents is again important since teams with collaborative links are also more likely to be technologically related. Therefore, column (2) represents the regression specification of choice. The joint hypothesis that the social distance measures do not matter is easily rejected even at the 1% significance level, with a χ2(4) statistic of 8351.1. Consistent with Hypothesis 3, collaborative links seem to matter since estimates for common inventor, past collaboration, common past collaborator and indirect social link are all positive and significant. Note that the reference group for comparison is patent pairs that are not connected at all. Since statistical significance could again result from large sample sizes, I now show that these effects are also large in magnitude. The marginal effects for column (2) can be interpreted as follows: If two patents are trivially related via a common inventor (social distance = 0), the probability of citation is about 5 times as much as that for unrelated patents. More interestingly, if they are related via a past collaboration (social distance = 1), the probability of citation is still about 3.8 times as much. Similarly, if they are related only via a common past collaborator (social distance = 2), the probability of citation is about 3.2 times. Finally, if none of these cases occur but there still exists an indirect collaborative link between two patents, the probability of citation is about 15% greater than for unrelated patents. A statistical test of equality of estimates of different

77

Table 3.4: Effect of social distance on probability of citation between patents (1) 8.820** (0.078) [105.84]

(2) 4.002** (0.060) [48.02]

Past collaboration (Social distance = 1)

6.741** (0.162) [80.89]

2.859** (0.055) [34.31]

Common past collaborator (Social distance = 2)

5.210** (0.089) [62.52]

2.228** (0.054) [26.74]

Indirect social link (Social distance > 2 but finite)

0.212** (0.019) [2.54]

0.151** (0.012) [1.81]

Common inventor (Social distance = 0)

Technological relatedness: Same tech category

1.260** (0.021)

Same tech subcategory

1.172** (0.026)

Same primary tech class

1.660** (0.027)

Same primary subclass

1.638** (0.048)

Secondary subclass overlap

3.653** (0.021)

Number of observations

2,528,764

2,528,764

A weighted logit regression is used The dependent variable is 1 if there is a citation between two patents, 0 otherwise Robust standard errors in parentheses, with clustering on citing patent Marginal effects in square brackets after multiplication with 1,000,000 Fixed effects for technological category, application year and time lag ** significant at 1%; * significant at 5%

78

social measures was easily rejected. Thus, consistent with Hypothesis 4, the probability of citation falls as the social distance for a pair of patents increases. 5.3. Collaborative Networks and Patterns of Knowledge Flows In this section, I test Hypotheses 5 and 6 (i.e., that knowledge flows are more intense within the same region and the same firm because social distances are smaller). In other words, I explore the extent to which denser collaborative networks can be seen as the mechanism driving more intense knowledge flows within regions and firms. The analysis appears in Table 3.5. For easy comparison, column (1) reproduces the intra-region and intra-firm results from column (3) of Table 3.1. Column (2) adds the social distance measures to the econometric model. Upon doing so, the coefficient estimate for within same region drops from 0.798 to 0.603, with its marginal effect falling from 9.58 in a million to 7.24 in a million. In other words, once social distance has been controlled for, the incremental effect of geographic co-location on probability of citation falls from 79.8% to 60.3%.21 Likewise, the coefficient estimate for within same firm drops from 2.217 to 1.809, with the marginal effect falling from 26.6 in a million to 21.7 in a million. Put differently, once social distance has been controlled for, the incremental effect of being in the same firm on probability of citation falls from 222% to 181%. To summarize, controlling for collaborative ties diminishes the result of localized knowledge flows as well as more intense intra-firm knowledge flows. Not only is the decrease non-trivial in magnitude for both cases, it is also found to be statistically

21

Normally, in non-linear models, one should only compare marginal effects and not coefficient estimates across models. However, for rare events, the marginal effect βj Λ(xβ)[1-Λ(xβ)] can be approximated as βj Λ(xβ), making βj directly interpretable as the fractional change in probability of citation when binary variable j goes from 0 to 1.

79

Table 3.5: Does social distance explain intra-region and intra-firm knowledge flows? (1) 0.798** (0.020) [9.58]

(2) 0.603** (0.022) [7.24]

(3) 0.697** (0.033) [8.36]

2.217** (0.025) [26.60]

1.809** (0.027) [21.71]

2.079** (0.049) [24.95]

Common inventor (Social distance = 0)

2.096** (0.065)

4.509** (0.245)

Past collaboration (Social distance = 1)

1.017** (0.062)

2.998** (0.177)

Common past collaborator (Social distance = 2)

0.469** (0.065)

2.382** (0.101)

Indirect social link (Social distance > 2 but finite)

0.098** (0.013)

0.147** (0.013)

Within same region

Within same firm

Within same region * Common inventor

-0.714** (0.197)

Within same region * Past collaboration

-0.686** (0.124)

Within same region * Common past collaborator

-0.700** (0.102)

Within same region * Indirect social link

-0.030 (0.043)

Within same firm * Common inventor

-2.115** (0.199)

Within same firm * Past collaboration

-1.748** (0.182)

Within same firm * Common past collaborator

-1.747** (0.121)

Within same firm * Indirect social link

-0.278** (0.056)

Number of observations

2,528,764

2,528,764

A weighted logit regression is used The dependent variable is 1 if there is a citation between two patents, 0 otherwise Technological relatedness controlled for Robust standard errors in parentheses, with clustering on citing patent Marginal effects in square brackets after multiplication with 1,000,000 Fixed effects for technological category, application year and time lag between patents ** significant at 1%; * significant at 5%

80

2,528,764

significant.22 Recall that a social distance of 0 represents the case of a common inventor between the cited and the citing teams. To verify that the results are not driven just by this case, analysis not reported here dropped all patent pairs with a social distance of 0 from the sample. The findings continued to hold. In other words, knowledge flows were still strong within the same region or the same firm, and introducing control variables for social distance of 1, 2 and >2 (but finite) still led to a large and statistically significant drop in estimates for within same region and within same firm. To investigate the effect of collaborative ties further, I now consider the possibility that direct and indirect ties need not operate similarly for transferring knowledge. In other words, there might be interaction effects between social distance and geographic co-location as well as between social distance and firm boundaries. Since column (3) includes both these sets of interaction variables, the “main effects” for within same region and within same firm now have to be interpreted as the effects for the case when the citing and cited patents are not connected at all. Interestingly, the interaction effects for within same region with common inventor, past collaboration or common collaborator are all almost equal in magnitude but opposite in sign to the main effect, so the two almost cancel out. In other words, conditional on the social distance being small (i.e., 0, 1 or 2), geographical co-location has almost no effect on citation probability. In fact, a formal hypothesis that these effects are 0 could not be rejected. On the other hand, for patents that are connected only with larger social distances or not connected at all, 22

To test statistical significance, the coefficients of within same region in columns (1) and (2) were interpreted as means of samples drawn from normally distributed populations. A t-test was then used to test the hypothesis that the two means could arise from the same population. An analogous test was done for within same firm.

81

geographic co-location continues to affect citation probability significantly. An explanation might be that, for teams with no close ties apparent from collaboration data on patents, there might still exists other ties that are both geographically concentrated and beneficial for knowledge flow. These could, for example, be collaborations that did not lead to patents, and hence did not get captured in patent data. These could also be fundamentally different kinds of professional and social interaction, such as meeting at conferences and professional get-togethers, or even at golf clubs and coffee shops. Analogously, the interaction effects for within same firm with common inventor, past collaboration or common collaborator are all comparable in magnitude and opposite in sign to the main effect for within same firm. In other words, conditional on the social distance being small (i.e., 0, 1 or 2), being in the same firm also has very small net effect on citation probability. Once more, a formal hypothesis that the effect is 0 for the case of social distance of 0 or 1 could not be rejected. Although the hypothesis that being within the same firm matters even at a social distance of 2 could not be rejected, the net magnitude (0.332) is much smaller than the net magnitude (1.801) for social distance greater than 2 or that (2.079) for unrelated teams. In other words, once social distance has been controlled for, being in the same firm matters only when the social distance is not small. Once more, this might simply be a result of collaborations not captured in patent data, or of alternate mechanisms for intra-firm information flow.

6. Limitations This paper studies knowledge diffusion through a collaborative network of individual inventors, and explores direct and indirect collaborative ties as a mechanism

82

behind knowledge flows usually associated with geographic co-location and firm boundaries. By including all inventor teams that have patented since 1975, the boundary specification and network sampling issues that plague smaller-scale studies on networks are avoided. Also, analyzing knowledge flows among a far larger sample than any similar study helps make the findings more generalizable. All this, however, is not without cost. The first issue is the usual concern of patents being imperfect as a measure of innovation, and patent citations being imperfect as a measure of knowledge flow. Also, only a subset of collaborative links between people gets captured in a patent-based network. In this paper, I have tried to address or at least discuss these concerns to the extent possible. However, I acknowledge that there might still be unresolved issues, and that there would be value in replicating such a study using other data sources like surveys or firm archives. However, collecting alternate data that give the ability of conducting studies of this scale is a big challenge. A computational cost of working with a large-scale network is the difficulty of using more sophisticated network-related measures. For example, while I study directness of links using my “social distance” measure, I do not consider frequency of interaction, decay of social links over time, and team size and characteristics. Also, though I make the distinction between direct and indirect ties in knowledge diffusion, I do not study the role of “structural holes” (Burt, 1992; Ahuja, 2000). Another methodological issue, which applies to most papers that take network ties as given, is that network ties might actually arise endogenously as a result of deliberate investment in tie formation by rational actors (Coleman, 1988; Glaeser, Laibson and Sacerdote, 2002). If people have a higher likelihood of deliberately cultivating collaborative links in exactly those settings where

83

they expect more knowledge flows, regression estimates might overstate the true influence of collaborative links on knowledge flows. An emphasis in this paper is that collaborative networks are important for transfer of know-how both within firms (Kogut and Zander, 1992) and between firms (von Hippel, 1988). Adopting a network perspective at the individual level allows me to study both of these in a single framework. However, this does not do full justice to a more sophisticated view of “organizational knowledge” (Levitt and March, 1988; Huber, 1991; Kogut and Zander, 1992; Nonaka, 1994). Also, patent citations could be more common within firms partly because a firm does not lose anything by making superfluous citations to its own patents. The most conservative interpretation of my results would therefore be to view the within same firm dummy merely as a control variable, and to read this paper as only studying intra-regional knowledge flows. In results not reported here, all results regarding collaborative networks and intra-regional knowledge flows continue to hold even if within-firm data points are simply dropped. 7. Conclusion This paper shows that collaborative networks have an important influence on knowledge diffusion, and that the probability of knowledge diffusion increases with the directness of collaborative ties between individuals. Even more interestingly, collaborative networks are found to be an important mechanism behind two knowledge diffusion patterns: geographic localization of knowledge flows and stronger intra-firm knowledge flows. The analysis in this paper has important implications for knowledge management. It shows that interpersonal networks remain key to management of complex knowledge,

84

despite the growing emphasis on formal knowledge management systems. Further, consistent with Cockburn and Henderson (1998), it shows the importance of a specific kind of interpersonal links – those arising from close collaborations between individuals rather than only casual interaction between them. A caveat for acquiring knowledge from outside the firm is that collaborative links with outsiders can lead to not just knowledge inflows but also knowledge outflows from a firm, so the net effect might differ in different situations (see Chapter 2). The specific finding that geographic co-location has little extra effect in cases of direct collaborative ties suggests that geographic constraints on flow of knowledge can be overcome by fostering collaborative links across regions. A firm might gain more knowledge from collaborative links with people even in different regions than by just locating in a high-tech region per se without developing such links. Similarly, from the point of view of a policy-maker, enticing the most advanced firms to open a local division may not be enough for knowledge spillovers to local firms if collaborative networks between the two do not get established. Again, there might be much to be gained through explicit cultivation of collaborative networks, for example, through joint projects. The findings on intra-firm knowledge flows have important implications as well. For example, firm boundaries per se need not constrain knowledge flow if strong collaborative links can be established with outsiders. Even mergers or acquisitions might not be sufficient for knowledge to flow if the employees of the two former firms cannot be made to work closely. On the other hand, not going to that extreme and just relying on alliances and joint ventures for knowledge transfer might be enough as long as they can

85

be managed to result in close collaborative ties between key people from the two sides, an argument consistent with findings of Mowery, Oxley and Silverman (1996), Rosenkopf and Almeida (2003), and Gomes-Casseres, Jaffe and Hagedoorn (2003). The result that collaborative networks can help overcome geographic distances is particularly important for developing countries. These countries could take an active approach towards learning from others by tapping into foreign collaborative networks. In particular, overseas movement of people (“brain drain”) need not always be bad. Consistent with Saxenian (2002), governments could actively set up incentives and mechanisms for their well-trained emigrants to continue to maintain close professional links with the professionals back home. Likewise, overseas location of R&D facilities by local companies might not be all that bad if they can serve as “bridges” to get access to the most advanced knowledge available internationally.

86

Chapter 4: TECHNOLOGICAL DYNAMISM IN ASIA23

1. Introduction Over the past few decades, Asian economies like South Korea, Taiwan, Hong Kong and Singapore have achieved high growth rates (see Table 4.1). Proponents of the “accumulation” view of growth (e.g., Krugman, 1994; Young, 1995; Collins and Bosworth, 1996) argue that this is merely the result of high savings and investments that have made it possible for these countries to better use technologies inherited from the world's technological leaders. In contrast, proponents of the “assimilation” view (e.g., Dahlman, 1994; Hobday, 1995; Nelson and Pack, 1998; Kim, 1998) insist that the critical source of growth in East Asia has been productivity growth resulting from the learning, entrepreneurship and innovation that these economies have gone through, which has made not only adoption of foreign technologies but also development of indigenous technologies possible. In this paper, we investigate the extent of innovation in East Asia. While doing so obviously does not conclusively settle the assimilation versus accumulation debate, evidence of substantial increase in innovation-related capabilities lends some support to the plausibility of the assimilation view. We examine patent data to study if these economies have built indigenous technological and entrepreneurial capabilities. Most of previous literature using patent data has focused on patenting activity of developed countries (e.g. US and Western European countries) because the extent of patenting from 23

This chapter is based on joint work with Ishtiaq P. Mahmood, which previously appears as a paper by the same title in Research Policy, Vol 32, No 6, 2003, pp 1031-1054. It is reproduced here with permission from Elsevier.

87

Table 4.1: Annualized real GDP growth rate (%): 1970-99 Recipient Countries

1970-74

1975-79

1980-84

1985-89

1990-94

1995-99

Newly Industrialized Economies Taiwan (ROC) South Korea Hong Kong Singapore

n/a 8.2 6.7 9.6

n/a 7.2 12.0 8.5

6.7 8.1 5.7 6.3

9.2 10.0 7.6 8.5

7.1 7.5 5.3 9.2

4.6 3.1 1.4 4.3

Emerging Asian Economies India China Indonesia Malaysia Thailand

3.2 5.2 7.8 7.2 5.8

5.4 5.5 7.9 8.6 8.0

5.4 10.8 5.7 5.2 5.4

6.4 7.7 7.1 6.9 10.3

5.2 12.1 7.8 9.5 8.6

5.0 6.7 0.0 3.1 -0.3

6.3 10.3 3.1 -1.1 3.0

7.1 6.7 3.0 7.3 2.5

2.0 1.2 -2.4 1.1 -0.9

1.7 2.1 -0.3 6.8 2.8

1.6 3.2 6.7 8.7 3.5

4.1 1.3 2.9 3.4 -0.2

Emerging Latin American Economies Mexico Brazil Argentina Chile Venezuela

Calculations based on data from World Development Indicators and EIU Country Data

88

other countries was often too small to be considered statistically meaningful. However, in the past two decades, many other countries have also started to patent heavily, opening up an opportunity for more research using patent data. We find that Taiwan, South Korea, Hong Kong and Singapore now have a much higher US patenting activity than the emerging economies both in Asia (India, China, Indonesia, Malaysia and Thailand) and in Latin America (Mexico, Brazil, Argentina, Chile and Venezuela). The results are most dramatic for Taiwan and Korea, though less so for Hong Kong and Singapore. Taiwan and Korea appear to be far ahead of Hong Kong and Singapore in innovation, indicating that the “Asian Tigers” might actually differ in the extent of innovation and hence possibly in the mechanisms that have led to their rapid growth. It appears that Taiwan already saw a surge in patenting activity in the late 1980s, while the rapid increase in patenting is primarily a 1990s phenomenon for South Korea. Hong Kong, Singapore and India have also recently begun to increase the extent of their US patenting, though the remaining emerging economies in our sample do not show any evidence of significantly exceeding the average overall growth rates in patenting. All the results mentioned here continue to hold even if we account for differences in exports across countries. Sector-level analysis sheds additional light on innovation in Asia. The areas of specialization for any given country are found to be somewhat persistent, evolving only slowly over time. Both Korea and Taiwan have managed to gradually shift more towards fast-growing industries. Even though Korea has been a little behind Taiwan in the aggregate patenting activity, it has been quicker in making a transition to fast-growing industries and also achieving a higher degree of specialization. Unlike Korea and Taiwan,

89

Hong Kong and Singapore have seen a fall in the overall degree of specialization, even though they have also managed a transition towards the fast-growing sectors. We also compare the sources of innovation across the Asian economies. We find that the relative contribution to innovation by multinational subsidiaries has been highest in Singapore and India, minimal in Taiwan and Korea, and something in between for Hong Kong and China. Business groups have been behind more than 80% of the patents arising from Korea in the 1990s, while their contribution in Taiwan has been less than 4%. The importance of individual inventors seems to be declining over time across all countries. However, they still own 59% of the recent patents in Taiwan but a mere 7% in Korea. Individual inventors are also important in Hong Kong and China, but not so much for Singapore and India. We also study how concentration of innovative activity differs across different economies by calculating the fraction of the country’s patents held by its top 50 assignees. This number is found to be the highest for Korea (85%), followed by Singapore (70%), India (63%), Hong Kong (32%), Taiwan (26%) and finally China (24%). The paper is divided into the following sections. In section 2, we summarize our data and methodology for comparing innovation across countries. In section 3, aggregate data for the past three decades is used to compare the newly industrialized countries with other emerging countries in Asia and Latin America. The remaining sections focus on detailed study of innovation in six Asian economies — four of them being newly industrialized countries (Korea, Taiwan, Singapore, and Hong Kong) and two being emerging economies (India and China). The other Asian economies are not included in this detailed analysis because of the relatively small number of patents they have, making

90

detailed analysis statistically uninteresting for these countries. Sections 4 and 5 present sector-level analysis of innovation in the six Asian countries. Sections 6 and 7 study the role of multinational subsidiaries, business groups, domestic firms, government-affiliated institutes and individual players in innovation, and examine the degree of concentration of patenting activity. Section 8 offers concluding thoughts.

2. Comparing innovation across countries: methodology Both patents and R&D expenditure data are commonly used indicators of innovation. The absence of uniform international accounting standards as well as unavailability of detailed R&D data makes R&D data analysis impractical for our purposes. An alternative is to use patent data. However, patent counts from different patent offices are not comparable to each other because of different patent breadths, patenting costs, approval requirements and enforcement rules for patenting in different countries. A common remedy is to use patent data from a single patent granting country like US to standardize the unit of innovation, making cross-country comparisons possible. Since the US is the largest and technologically most advanced market in the world, any sufficiently big invention being patented anywhere with a global market in mind is likely to be patented in the US as well. Over the past two decades or so, the increasing number of patents taken out by the countries in Asia and Latin America now allows us to do statistically meaningful analysis. While patenting data does not always capture the cumulative and incremental aspect of learning and innovation (Amsden and Hikino, 1994), it still is perhaps the best means of making large-scale comparisons of innovation (Pavitt, 1988b; Griliches, 1990).

91

Our dataset, which includes successful applications registered with the US Patent Office (USPTO) during 1970-1999, was obtained by combining data obtained directly from USPTO with an enhanced dataset by Jaffe and Trajtenberg (2002). We divide the entire period of thirty years into six consecutive five-year periods based on the grant year (1970-74, 1975-79, ..., 1995-99) in order to reduce the erratic year-to-year variation.

3. Comparing innovation across countries: results Table 4.2(a) summarizes the trends in US patents granted to inventors based in several Asian and Latin American economies from 1970 to 1999. This helps us compare the newly industrialized countries in Asia (Taiwan, South Korea, Hong Kong and Singapore) with other emerging economies in Asia (India, China, Indonesia, Malaysia and Thailand) and Latin America (Mexico, Brazil, Argentina, Chile and Venezuela). As the data indicate, the overall patenting activity of the NICs had been quite low during the earlier part of this time period, but has gone up substantially in recent years relative to the trend in aggregate worldwide patenting as well as that of emerging economies in Asia and Latin America. The growth in patenting has been much more dramatic for Taiwan and South Korea than for Hong Kong and Singapore, suggesting that former in particular have experienced a massive increase in innovative capabilities. As Table 4.2(b) indicates, the countries in our sample differ substantially in the extent of foreign exports. It can be argued that the incentive of inventors from a country to patent abroad would depend on the extent to which they participate in world markets. Therefore, one fear in reading too much into raw patent counts from Table 4.2(a) is that the extent of US patenting might

92

Table 4.2(a): USPTO patents granted to each country's inventors: 1970-99 1970-74

1975-79

1980-84

1985-89

1990-94

1995-99

1

176

397

1,772

5,271

12,366

South Korea

24

43

91

424

2,890

11,366

Hong Kong

59

75

113

177

279

570

Singapore

21

9

20

47

148

499

Emerging Asian Economies India

83

67

40

64

126

316

China

61

2

7

129

239

332

Indonesia

19

5

5

10

26

18

Malaysia

2

13

6

13

43

89

Thailand

4

3

7

11

15

56

243

246

191

202

189

257

86

100

110

156

260

353

126

113

100

82

109

183

Chile

22

20

12

18

32

44

Venezuela

36

35

50

103

121

145

367,943

322, 385

309, 387

398,816

484,223

623,999

Recipient Countries Newly Industrialized Economies Taiwan (ROC)

Emerging Latin American Economies Mexico Brazil Argentina

Total Worldwide

93

Table 4.2(b): Country exports: 1970-99 Recipient Countries Newly Industrialized Economies Taiwan (ROC) South Korea Hong Kong Singapore

Emerging Asian Economies India China Indonesia Malaysia Thailand Emerging Latin American Economies Mexico Brazil Argentina Chile Venezuela

1970-74

1975-79

1980-84

1985-89

1990-94

1995-99

n/a

n/a

87.6 (21.5%) 125.0 (88.9%) 87.1 (121.9%)

181.0 (28.9) 184.0 (87.5%) 178.3 (171.6%)

256.7 (52.9%) 289.0 (34.3%) 311.0 (95.6%) 304.3 (192.9%)

399.8 (54.4%) 470.6 (35.7%) 561.0 (123.0%) 390.4 (187.6)

477.3 (45.0%) 549.5 (27.9%) 838.0 (139.3%) 595.1 (185.8)

697.1 (48.0%) 983.4 (37.3%) 1002.8 (137.0%) 823.6 (174.5)

23.6 (4.0%) 15.7 (2.9%) 41.1 (20.0%) 35.1 (40.0%) 27.3 (18.1%)

45.8 (6.4%) 32.8 (4.8%) 75.8 (25.6%) 61.7 (49.0%) 43.6 (20.3%)

52.5 (6.0%) 84.8 (8.6%) 118.0 (28.0%) 94.4 (54.4%) 65.8 (22.5%)

70.9 (6.2%) 204.0 (12.4%) 127.4 (22.9%) 140.4 (62.6%) 121.5 (29.7%)

136.0 (9.2%) 506.0 (20.1%) 215.2 (26.5%) 272.0 (79.8%) 243.0 (36.9%)

226.0 (11.3%) 932.0 (22.4%) 344.1 (32.9%) 507.9 (103.3%) 409.8 (48.7%)

53.1 (8.1%) 108.9 (7.5%) 59.2 (6.7%) 16.3 (13.8%) 62.6 (25.7%)

84.4 (9.6%) 151.9 (7.1%) 77.6 (7.9%) 27.2 (22.9%) 73.5 (24.7%)

171.0 (14.6%) 254.9 (10.2%) 77.7 (7.5%) 30.8 (21.2%) 73.3 (25.0%)

221.3 (18.2%) 295.1 (9.9%) 101.4 (10.0%) 56.7 (31.9%) 75.1 (24.2%)

234.1 (16.4%) 300.0 (9.6%) 89.0 (7.7%) 78.9 (30.8%) 111.8 (30.7%)

493.0 (13.9%) 297.6 (8.1%) 145.8 (10.2%) 104.5 (28.6%) 105.7 (26.8%)

Calculations based on data from World Development Indicators and EIU Exports measured in billions of constant 1995 US$ The numbers in parentheses indicate exports as a percent of the country's total GDP.

94

Table 4.2(c): USPTO patents granted per billion constant 1995 US$ of exports Recipient Countries

1970-74

1975-79

1980-84

1985-89

1990-94

1995-99

Newly Industrialized Economies Taiwan (ROC) South Korea Hong Kong Singapore

n/a 0.27 0.47 0.24

n/a 0.24 0.42 0.05

1.55 0.31 0.36 0.07

4.43 0.90 0.32 0.12

11.04 5.26 0.33 0.25

17.73 11.56 0.57 0.61

Emerging Asian Economies India China Indonesia Malaysia Thailand

3.52 3.89 0.46 0.11 0.15

1.46 0.06 0.07 0.05 0.07

0.76 0.08 0.04 0.07 0.11

0.90 0.63 0.08 0.08 0.09

0.93 0.47 0.12 0.06 0.06

1.40 0.36 0.05 0.11 0.14

Emerging Latin American Economies Mexico Brazil Argentina Chile Venezuela

4.58 0.79 2.13 1.35 0.58

2.91 0.66 1.46 0.74 0.48

1.12 0.43 1.29 0.39 0.68

0.91 0.53 0.81 0.32 1.37

0.81 0.87 1.22 0.41 1.08

0.52 1.19 1.26 0.42 1.37

95

simply reflect different size of the economies or different export orientation rather than genuine differences in innovativeness. In order to control for this possibility, we carry out a robustness check suggested by Archibugi and Pianta (1998) by dividing each country's number of US patents by their exports, giving us the normalized patenting numbers reported in Table 4.2(c). Even after controlling for differences in foreign exports, we find that Taiwan and Korea turn out to be far ahead of the rest in recent years.

4. Sector-level analysis of innovation: methodology Aggregate patent data hide important sector-level details of innovation. The assessment of national capabilities and performance in specific fields of technology is important because technological progress, particularly within a specific paradigm, seems to proceed cumulatively along the "technological trajectories" defined by the paradigm (Dosi, 1982; Archibugi and Pianta, 1992). The path dependency and the cumulative nature of technology together imply that a nation’s technological capabilities are likely to be in the technological neighborhood of previous successes, a claim that is corroborated by evidence provided by Pavitt (1988a) and Cantwell (1989). In the context of developed countries, it has been shown that analysis of technological convergence at the aggregate level can be very misleading, and only a sector-level analysis gives a clear picture of differences in technological capabilities of a country (Soete, 1987; Guerrieri and Milana, 1998; Patel and Pavitt, 1998; Archibugi and Pianta, 1998; Laursen, 1999). With this in mind, we focus on identifying the fields in which different Asian countries have an advantage or weakness relative to their overall scientific and technological activities. 4.1. Definition of sectors

96

In coming up with our definition for industries, we used 3-digit SIC codes as a starting point, but aggregated some of these up to give a total of only 33 sectors. We felt that 33 sectors was a reasonable trade-off between the richness of sectoral data and the number of patents per sector as a reliable measure of innovativeness in that sector. Our entire list of sectors, along with its mapping to SIC codes, appears in Table 4.3. We also want to classify the sectors in order to help capture the “quality” of national patterns of technological specialization. In an approach analogous to Archibugi and Pianta (1992), we sort the 33 sectors in decreasing order of their patenting growth rate. The top 11 sectors are classified as "fast-growing" sectors, the next 11 as "medium-growing" sectors and the last 11 as "slow-growing" sectors. The complete list of sectors according to the classification for each of these periods appears in Table 4.4. 4.2. Measuring sector-level specialization A general problem with using raw patent counts is that sectors vary in the propensity to patent (Scherer, 1983). Also, the raw numbers are obviously sensitive to our choice of sector definitions. We follow previous research (e.g., Soete, 1987; Archibugi and Pianta, 1992) in using a “relative technological advantage” (RTA) index that measures the relative distribution of a country’s inventive activity in each field. Formally, the RTA index for country i in sector j is defined as the ratio of country i’s share of total world patents in sector j to country i’s share of total world patents, i.e.,



 



RTAij ≡  n / ∑ n  /  ∑ n ij / ∑ ∑ n ij  ij ij  j i j  i  

where

n ij

is the number of patents of country i in sector j. By definition, this index equals

1 if the country holds the same share of worldwide patents in a given technology as in the

97

Table 4.3: List of industries Name

SIC Code(s)

Food, Other Related Products & Beverages Textiles, Apparel, Leather and Footwear Basic Industrial chemicals (org & inorg) Plastic materials and synthetic resins Agricultural chemicals Soaps, detergents, cleaners, perfumes, cosmetics Paints, varnishes, lacquers, enamels Miscellaneous chemical products Drugs and medicine Petroleum, Natural Gas & Related Products Rubber and Plastic Products Stone, class, glass and non-metal minerals Ferrous and Non-ferrous metals Fabricated metal products Engines and turbines Farm and garden machinery and equipment Metal working machinery and equipment Computers and office Special industry machinery, except metal working Other non-electric machinery and equipment Electric industrial machinery & equipment Electric household appliances Electric misc apparatus and supplies Electronics, Radio, TV, Comm Motor vehicles and other motor vehicle equipment Guided missiles and space vehicles and parts Ship and boat building and repairing Railroad equipment Motorcycles, bicycles, and parts Misc transport equipment and ordinance Aircraft and parts Professional and scientific equipment Other manufactured products

98

20 22, 23, 31 281, 286 282 287 284 285 289 283 29 30 32 33 34 (ex.3462,3463,348) 351 352 354 357 355 353, 356, 358, 359 361, 362, 3825 363 364, 369 365, 366, 367 371 376 373 374 375 379, 348 372 38 99

Table 4.4 (a): Sectors sorted by decreasing growth rate of all US patents (1980-89)

Top 11 (FastGrowing)

Middle 11 (MediumGrowing)

Bottom 11 (SlowGrowing)

1980-84

1985-89

Computers and office Petroleum, Natural Gas & Related Electric household appliances Agricultural chemicals Drugs and medicine Professional and scientific equipment Aircraft and parts Engines and turbines Electric industrial machinery Stone, class, glass and non-metal minerals Plastic materials and synthetic resins Rubber and Plastic Products Electric misc apparatus and supplies Electronics, Radio, TV, Comm Textiles, Apparel, Leather,Footwear Soaps, detergents, cleaners, perfumes, cosmetics Motor vehicles and equipment Fabricated metal products Farm and garden machinery Miscellaneous chemical products Other non-electric machinery Other manufactured products Railroad equipment Food, Other Related Products & Beverages Paints, varnishes, lacquers, enamels, and allied products Motorcycles, bicycles, and parts Special industry machinery, except metal working Metal working machinery and equipment Ferrous and Non-ferrous metals Guided missiles and space vehicles and parts Misc transport equipment and ordinance Basic Industrial chemicals Ship and boat building and repairing

Computers and office Guided missiles and space vehicles Electronics, Radio, TV, Comm Motorcycles, bicycles, and parts Ship and boat building and repairing Motor vehicles and other motor vehicle equipment Professional and scientific equipment Drugs and medicine Other manufactured products Electric industrial machinery & equipment Misc transport equipment and ordinance Agricultural chemicals Aircraft and parts Metal working machinery Fabricated metal products Electric misc apparatus and supplies Soaps, detergents, cleaners, perfumes, cosmetics and toiletries Other non-electric machinery Ferrous and Non-ferrous metals Food, Other Related Products & Beverages Electric household appliances Rubber and Plastic Products Textiles, Apparel, Leather and Footwear Engines and turbines Special industry machinery, except metal working Stone, class, glass and non-metal minerals Plastic materials and synthetic resins Miscellaneous chemical products Petroleum, Natural Gas & Related Products Farm and garden machinery and equipment Railroad equipment Basic Industrial chemicals Paints, varnishes, lacquers, enamels, and allied products

99

Table 4.4 (b): Sectors sorted by decreasing growth rate of all US patents (1990-99) Top 11 (FastGrowing)

Middle 11 (MediumGrowing)

Bottom 11 (SlowGrowing)

1990-94 Computers and office Drugs and medicine Plastic materials and synthetic resins Electronics, Radio, TV, Comm Electric misc apparatus and supplies Paints, varnishes, lacquers, enamels, and allied products Professional and scientific equipment Soaps, detergents, cleaners, perfumes, cosmetics and toiletries Rubber and Plastic Products Stone, class, glass and non-metal minerals Agricultural chemicals Basic Industrial chemicals Other manufactured products Food, Other Related Products & Beverages Farm and garden machinery and equipment Guided missiles and space vehicles and parts Miscellaneous chemical products Ship and boat building and repairing Motor vehicles and other motor vehicle equipment Ferrous and Non-ferrous metals Aircraft and parts Misc transport equipment and ordinance Special industry machinery, except metal working Motorcycles, bicycles, and parts Other non-electric machinery and equipment Fabricated metal products Engines and turbines Textiles, Apparel, Leather and Footwear Electric industrial machinery & equipment Railroad equipment Metal working machinery and equipment Electric household appliances Petroleum, Natural Gas & Related

100

1995-1999 Computers and office Drugs and medicine Electronics, Radio, TV, Comm Soaps, detergents, cleaners, perfumes, cosmetics and toiletries Agricultural chemicals Electric industrial machinery & equipment Electric misc apparatus and supplies Professional and scientific equipment Textiles, Apparel, Leather and Footwear Other manufactured products Motorcycles, bicycles, and parts Motor vehicles and other motor vehicle equipment Miscellaneous chemical products Electric household appliances Rubber and Plastic Products Stone, class, glass and non-metal minerals Special industry machinery, except metal working Basic Industrial chemicals Aircraft and parts Other non-electric machinery and equipment Fabricated metal products Paints, varnishes, lacquers, enamels, and allied products Food, Other Related Products & Beverages Farm and garden machinery and equipment Engines and turbines Railroad equipment Ship and boat building and repairing Metal working machinery and equipment Misc transport equipment and ordinance Ferrous and Non-ferrous metals Guided missiles and space vehicles and parts Petroleum, Natural Gas & Related Plastic materials and synthetic resins

aggregate, and is below (above) 1 if there is a relative weakness (strength). This allows cross-sectional as well as longitudinal comparison of relative technological strengths and weaknesses of countries. 4.3. Measuring overall degree of technological specialization As a country slowly diversifies out of sectors associated with abundant endowments of the conventional factors of production like textiles, mining and food processing towards advanced sectors like machinery, transportation and chemicals, their overall specialization might fall initially (Bell and Pavitt, 1993; Amsden and Hikino, 1994). However, as they eventually approach the technological frontier, the need for internal or external economies of scale in R&D suggests that the country would start to specialize on a narrow set of new industries. Thus, a country’s technological specialization could be expected to first decline and then rise as it moves from traditional to more high tech sectors. In order to measure how evenly or unevenly the patenting activities of a given country are distributed across all the sectors, we follow previous literature in using the Chi-square index, which is defined as

 χ = ∑  p − p wj i j   ij  2

2    / p wj   

where j is the sector, pwj is the percentage of total world patents in class j and pij is the percentage of patents held by country i in sector j. The more diverse a country is in relative sectoral strengths and weaknesses, the greater the value of Chi-square. Since the Chi-square indices are calculated on the country’s percentage distribution and not levels

101

of activities across sectors, they make cross-country comparisons in specialization meaningful.

5. Sector-level analysis of innovation: results Table 4.5 reports the top five sectors in terms of RTA as well as the overall Chisquare index for each time period for six Asian economies: Taiwan, South Korea, Hong Kong, Singapore, India and China.24 We start by making some general observations based on Table 4.5. First, we see that the countries are quite different in their areas of specialization, and these areas tend to be persistent for each country in the short run. Second, countries differ in their degree of overall specialization, and the degree of specialization evolves differently over time for different countries. For Taiwan, Singapore and Hong Kong, the degree of specialization (as measured by the Chi-Squared index in Table 4.5) seems to have steadily fallen over time, consistent with the theory of natural evolution of a “latecomer industrializing economy” as it makes the transition from a borrower to an innovator of technology (Amsden, 1989). Interestingly, South Korea does not show this pattern - instead, it shows an increase in the degree of specialization from the 1980s to 1990s (though the degree of specialization is somewhat lower in the late 1990s compared with early 1990s). India and China have both maintained relatively stable degrees of specialization, though the degree of specialization for India has been consistently higher (between 1.9 and 2.7) than for China (between 0.2 and 0.4).

24

We exclude Indonesia, Malaysia and Thailand because of the their low levels of patenting at the sector level. Additionally, data for 1970s and early 1980s has small sample sizes even for the selected countries (especially China, Singapore and India), and should therefore be interpreted with caution. In the 1990s, however, the sample sizes become sufficiently large for us to have more confidence in sector-level analysis using patent data.

102

Table 4.5(a): Chi-square index and top 5 RTA sectors for Taiwan & South Korea

Taiwan

South Korea

1980-84

(N=397, Chi-Sq=0.75) Motorcycles, bicycles, and parts (4.1) Other manufactured products (3.3) Fabricated metal products (2.3) Electric household appliances (2.0) Electric misc apparatus & supplies (1.4)

1985-89

(N=1772, Chi-Sq=0.74) Motorcycles, bicycles, and parts (5.2) Other manufactured products (2.7) Fabricated metal products (2.7) Electric misc apparatus & supplies (2.3) Electric household appliances (1.9)

1990-94

(N=5271, Chi-Sq=0.64) Motorcycles, bicycles, and parts (6.5) Other manufactured products (2.7) Fabricated metal products (2.4) Electric misc apparatus & supplies (2.2) Electric household appliances (1.8)

1995-99

(N=12366, Chi-Sq=0.46) Motorcycles, bicycles, and parts (6.0) Electric misc apparatus & supplies (2.1) Other manufactured products (2.1) Fabricated metal products (1.9) Electronics, Radio, TV, Comm (1.6)

(N=91, Chi-Sq=0.37) Ship and boat building and repairing (3.8) Electric misc apparatus and supplies (2.4) Other manufactured products (2.3) Basic Industrial chemicals (1.6) Fabricated metal products (1.5) (N=424, Chi-Sq=0.35) Electric household appliances (3.6) Motorcycles, bicycles, and parts (3.1) Ship and boat building and repairing (3.0) Other manufactured products (1.9) Electric industrial machinery & equip (1.8) (N=2890, Chi-Sq=0.84) Electronics, Radio, TV, Comm (3.0) Electric household appliances (2.4) Computers and office (1.6) Electric industrial machinery & equip (1.0) Electric misc apparatus and supplies (.8) (N=11366, Chi-Sq=0.60) Electric household appliances (3.1) Electronics, Radio, TV, Comm (2.5) Electric industrial machinery & equip (1.2) Computers and office (1.1) Other non-electric machinery and equip (1.0)

N indicates the number of US patents granted to the country in the particular period. The numbers in parentheses indicate the RTA value for each sector.

103

Table 4.5(b): Chi-square index and top 5 RTA sectors for Hong Kong & Singapore

1980-84

Hong Kong (N=113, Chi-Sq=1.16) Electric misc apparatus and supplies (4.0) Other manufactured products (3.8) Motorcycles, bicycles, and parts (2.5) Railroad equipment (1.7) Computers and office (1.5)

1985-89

(N=177, Chi-Sq=0.82) Electric household appliances (5.1) Electric industrial machinery & equip (2.9) Other manufactured products (2.8) Electric misc apparatus and supplies (2.2) Railroad equipment (1.4)

1990-94

(N=279, Chi-Sq=0.92) Electric household appliances (3.9) Electric industrial machinery & equip (3.8) Other manufactured products (2.8) Electric misc apparatus and supplies (2.5) Fabricated metal products (1.4) (N=570, Chi-Sq=0.74) Electric household appliances (4.1) Other manufactured products (3.2) Electric industrial machinery & equip (2.3) Electric misc apparatus and supplies (2.3) Ship and boat building and repairing (1.9)

1995-99

Singapore (N=20, Chi-Sq=8.26) Misc transport equip & ordinance (28.6) Ship and boat building & repair (17.4) Food, Related Products & Beverages (6.6) Electric misc apparatus and supplies (2.7) Engines and turbines (2.2) (N=47, Chi-Sq=1.48) Farm/garden machinery & equipment (8.5) Misc transport equip & ordinance (4.8) Metal working machinery & equip (3.3) Electric household appliances (2.6) Other non-electric mach & equip (2.4) (N=148, Chi-Sq=1.15) Ship and boat building & repair (4.6) Electronics, Radio, TV, Comm (3.2) Computers and office (2.4) Farm/garden machinery & equip (1.6) Miscellaneous chemical products (1.4) (N=499, Chi-Sq=0.66) Petroleum, Gas & Related Prod (2.8) Electronics, Radio, TV, Comm (2.4) Food, Related Products & Beverages (1.9) Electric industrial machinery & equip (1.8) Electric household appliances (1.8)

N indicates the number of US patents granted to the country in the particular period. The numbers in parentheses indicate the RTA value for each sector.

104

Table 4.5(c): Chi-square index and top 5 RTA sectors for China & India

1980-84

1985-89

1990-94

1995-99

India

China

(N=40, Chi-Sq=1.92) Motorcycles, bicycles, and parts (9.6) Stone, class, glass, non-metal minerals (5.0) Agricultural chemicals (4.7) Ferrous and Non-ferrous metals (4.4) Miscellaneous chemical products (4.2) (N=64, Chi-Sq=2.66) Soaps, detergents, cleaners, perfumes, cosmetics and toiletries (8.0) Drugs and medicine (7.7) Agricultural chemicals (6.9) Railroad equipment (3.8) Plastic materials and synthetic resins (3.3) (N=126, Chi-Sq=2.17) Basic Industrial chemicals (5.2) Drugs and medicine (5.0) Agricultural chemicals (4.8) Plastic materials and synthetic resins (3.7) Ferrous and Non-ferrous metals (2.4)

(N=7, Chi-Sq=5.71) Motorcycles, bicycles, and parts (41.1) Farm/garden mach & equipment (10.4) Engines and turbines (4.2) Aircraft and parts (4.1) Other manufactured products (3.7) (N=129, Chi-Sq=0.31) Motorcycles, bicycles, and parts (7.0) Electric misc apparatus & supplies (2.8) Misc transport equip & ordinance (2.4) Ferrous and Non-ferrous metals (2.0) Drugs and medicine (1.9) (N=239, Chi-Sq=0.22) Ferrous and Non-ferrous metals (3.0) Miscellaneous chemical products (2.1) Electric misc apparatus & supplies (2.0) Basic Industrial chemicals (2.0) Petroleum, Gas & Related Prod (1.8) (N=332, Chi-Sq=0.41) Miscellaneous chemical products (3.6) Basic Industrial chemicals (2.8) Ship/ boat building and repairing (2.6) Agricultural chemicals (2.2) Drugs and medicine (2.1)

(N=316, Chi-Sq=2.45) Basic Industrial chemicals (6.6) Drugs and medicine (4.3) Plastic materials and synthetic resins (3.3) Agricultural chemicals (3.3) Soaps, detergents, cleaners, perfumes, cosmetics and toiletries (2.6)

N indicates the number of US patents granted to the country in the particular period. The numbers in parentheses indicate the RTA value for each sector.

105

5.1. South Korea As Table 4.5(a) shows, the top five RTA sectors have changed completely between 1980-84 and 1995-99 for Korea. However, this change has been gradual as there has been a significant overlap in the top five lists between any two adjacent periods. This suggests that country-specific factors prevent rapid change in areas of specialization, though these areas do change over a sufficiently long period. During 1980-84, none of the top five RTA sectors for South Korea appears in the "Fast Growing Industries" list for patenting activity as defined in Table 4.4. In contrast, during 1995-99, four of the top five RTA sectors for Korea are drawn from the fast growing industries list. This is consistent with the explanation given by Hobday (1995) that Korea has only recently developed strong technological capabilities because of increased exposure to foreign markets and competition through increased exports in the 1970s and 1980s. Chi-square values over time for Korea reveal that the overall degree of technological specialization is much higher in the 1990s than in the 1980s. The increasing value of the Chi-square index suggests that Korea has been making the transition from a scale-intensive phase to a technology-intensive phase of development (Bell and Pavitt, 1993). When we examine this finding in light of Korea’s sectoral patterns of specialization in Table 4.5(a), this seems to be a plausible conclusion. The “Heavy and Chemical Industries” drive was initiated by President Park in the 1970s to enhance Korea’s self-sufficiency in industrial raw materials and to upgrade its industrial structure from being labor-intensive to being capital-intensive stage. Special legislation singled out six strategic industries--steel, petrochemicals, nonferrous metals, shipbuilding, electronics, and machinery--to receive support, including tax incentives, subsidized

106

public services, and preferential financing. This was followed by industrial policies of the subsequent regimes that emphasized the development of specialized industries such as semiconductors and electronics. The patenting growth for Korea as reported in Table 4.1 and the specialization outcomes reported in Table 4.5(a) seem consistent with these policy measures. 5.2. Taiwan Unlike Korea, the areas where Taiwan has focused have remained remarkably consistent during the past twenty years. This once more highlights that country-specific drivers of technological specialization are indeed quite stable. As reported in Table 4.5(a), four out of the top five RTA sectors have remained the same from 1980-84 to 1995-99. The most notable change that took place is in "Electronics, Radio, TV and Communications", where the RTA value has gone up from 0.8 during 1980-84 to 1.6 in 1995-99. Taiwan's top RTA industry has remained "Motorcycles, Bicycles & Parts", where its RTA has in fact steadily increased from 4.1 in 1980-84 to 6.0 in 1995-99. Comparing Taiwan's and Korea's top RTA lists, we find that the two have specialized in different sectors, with "Electronics, Radio, TV and Communications" being the only common sector. During period 1980-84, only one of the top five RTA sectors for Taiwan appears in the "Fast Growing Industries" list for patenting activity as defined in Table 4.4. In contrast, during 1995-99, three of the top five RTA sectors for Taiwan are drawn from the fast growing industries list. Taiwan, like Korea, seems to have developed stronger technological capabilities in areas with high overall percentage rate of increase worldwide. However, just like it lags behind Korea in the level of technological

107

complexity, it also seems to lag behind Korea a little in its focus on the fast-growing industries. Chi-square values over time for Taiwan reveal that the overall degree of technological specialization is just marginally lower in the 1990s than in the 1980s. This result is consistent with the evidence of relatively consistent profiles of RTA for the past twenty years. Since the l980s, an important beneficiary of the government’s industrial policies in Taiwan has been the information and communication science sector. In addition to low interest loans, investment credits, and favorable tariff rates for imported computer components, the government has established research institutes to facilitate the generation of new technology and the diffusion of existing technology. By 1990, Taiwan had become the sixth largest producer of computers in the world. This may explain why "Electronics, Radio, TV and Communications" is a part of the top five RTA sectors in Taiwan. 5.3. Singapore and Hong Kong From Tables 4.2(a) and 4.2(c), it appears that the patenting activity in Singapore and Hong Kong has consistently been much lower than in South Korea and Taiwan. Singapore and Hong Kong have not been as innovative as these other newly industrialized economies, indicating much weaker technological capabilities. Therefore, the innovative performance of the so-called "Asian Tigers" is actually quite different, indicating that the drivers of growth have also been different. The number of patents for Singapore and Hong Kong has been particularly small during the earlier periods, making a detailed sector-level analysis relatively meaningful only for the 1990s, which shall be the focus of our discussion.

108

Table 4.5(b) shows how the top five RTA sectors have evolved for Singapore and Hong Kong over time. Unlike Korea and Taiwan, areas of high RTA seem to change substantially in Singapore and Hong Kong from one period to the next. For example, the only industry that appears in Singapore's top-five list for RTAs for both 1990-94 and 1995-99 is "Electronics, Radio, Television and Communications". There is, however, a clear move from relatively low-tech areas in the 1980s to high-tech areas in the 1990s. Although Singapore appears to have developed relative specialization in electronics and other high technology areas, a large fraction of Singapore's patenting activity continues to actually be a result of multinationals rather than domestic entities, as discussed later in this paper. Chi-square values for Singapore and Hong Kong reveal that the overall degree of technological specialization has been consistently falling over time. This is similar to the trend observed in the context of developed countries wherein countries move from niche positions to much broader bases of innovation during the transition phase. Compared with the case of Singapore, the top five RTAs have been slightly more stable over time for Hong Kong. There is a fair bit of overlap in specialization of Hong Kong and Singapore, though Singapore has developed a leadership in electronics as well as electrical goods and Hong Kong focuses on just a wider variety of electrical goods. 5.4. India and China Table 4.2(a) reveals that, although India and China are still not very large players in US patenting, they have shown a substantial surge in patenting in the 1990s. However, as Table 4.2(c) shows, this increase begins to appear smaller for India and actually negative for China once we normalize for increase in foreign trade. Since the number of patents is not too large, it is perhaps not worthwhile trying to read too much into the time

109

trends in RTAs reported in Table 4.5(c). It seems worth noting, however, that both Indian and China seem to be building up substantial innovative capabilities in all kinds of chemicals as well as drugs and medicine. Additionally, India seems to be quite strong in plastic materials and synthetic resins.

6. Comparing type of innovators: methodology Next, we turn to comparing sources of innovation across the Asian economies. In particular, we want to document the fraction of innovation arising from multinational subsidiaries, business groups, individual inventors and other domestic firms and organizations in each of these countries.25 Given the differences in the national systems of innovation across different countries (Freeman, 1993), we expect the composition of the set of innovators to vary substantially across countries as well. Business groups are known to play an important role in the overall economic activity of Asian economies (Khanna, 2000; Khanna and Rivkin, 2001). Therefore, we try to study their specific contribution to patenting. We were able to obtain data on business groups for Korea, Taiwan and India, so we classified all domestic patent assignees from these countries into whether they had a group affiliation or not.26 This enabled us to calculate the fraction of patents arising from business groups for these countries. We also

25

Ideally, we would have liked to break up the components of “other domestic firms and organizations” that are for-profit firms and non-profit research institutes. Unfortunately, since both of these are listed as “Non-government organization” in the US patent data, this is a non-trivial exercise. While US patent data does sometimes separately list patents assigned to governments, the numbers of these are trivial since they do not include research institutes. For this reason, we have simply included them in the “other domestic firms and organizations” category.

26

We used two datasets for business group data: one was the dataset used in Khanna and Rivkin (2001) kindly made available to us by Tarun Khanna and the other was data we downloaded from the web site of the Center for International Data at UC Davis (http://data.econ.ucdavis.edu/international).

110

study role of individuals in innovation. For our purposes, patents assigned to individuals are those that are marked as either “individual” or “unassigned” in the US patent data. Next, we turn to calculating the fraction of patents attributable to local subsidiaries of foreign multinationals. In order to determine whether a given patent originates from the local subsidiary of a foreign multinational, we check whether the home country of the assignee organization is the same as the country of the first inventor. A crucial step in building the dataset was therefore identifying whether an assignee firm had its home base in the country of patenting, or if it was part of a foreign firm.27 To achieve this, we undertook the following extensive data cleaning exercise. First, we used Compustat-based CUSIP numbers (from year 1989) included in the database by Jaffe and Trajtenberg (2002) to make sure that the subsidiaries of companies that have CUSIP numbers are correctly matched to their respective corporate parents identified using the same CUSIP number. Next, we used Stopford’s (1992) directory of 428 largest multinationals to manually associate all their major subsidiaries correctly with the corporate parent. Finally, for every remaining assignee, we calculated the home country as the country in which maximum numbers of patents originated for that assignee. We also study the list of top 50 players for each of the six countries considered here. This has several goals: First, it helps identify important individual players for innovation. Second, it gives an idea of the role of non-profit research institutes versus forprofit domestic firms since both of them show up simply as “domestic firms & organizations” in US patent database. Third, calculation of the fraction of patents held by 27

We defined the subsidiary as being a company in which the multinational has a majority stake. While one can argue that even a “high enough” minority stake can give a multinational enough control over a foreign company, we wanted to avoid the situation in which a company could not be identified with a unique parent. For cases where two multinationals had exactly 50-50 stake in a company, we broke the tie by assuming it was a part of the multinational whose name appeared first in the joint venture.

111

the top 50 players helps identify the extent to which innovative activity in a country is concentrated among a few players rather than dispersed among many players in the economy.

7. Comparing type of innovators: results Table 4.6 gives the composition of the set of innovators in the six Asian economies we study. Consistent with previous research (e.g. Hobday, 1995; Chen and Sewell, 1996; Kim, 1998; Choung, 1998), we find that business groups or chaebols have played a key role in developing Korea's innovative capabilities. About 81% of all Korean patents arose from business groups. In contrast, the fraction attributable to business groups is less than 4% for the case of Taiwan. On the other hand, individual inventors own a mere 7% of the patents coming from Korea but as much as 59% of the patents from Taiwan. Industrial policies seem to have played an important role in shaping the innovative fabric of these countries. Unlike Korea, where large business groups dominate, Taiwan’s national system of innovation has a much greater role for small and medium sized enterprises (SME).28 Individual inventors are also relatively important in China (40%) and Hong Kong (31%), though less so in India (18%) and Singapore (10%). Singapore has relied quite heavily on multinationals, which account for 46% of

28

Based on analysis of a dataset for 1994-2000 (with a different industry classification) obtained from CHI Research, we find that institutes in Taiwan focus on areas such as “Biotechnology,” “Plastics, Polymers, & Rubbers,” etc. SMEs are dominant in industries such as, “Motor Vehicle & Parts,” “Other Transportation Equipment,” “Textiles & Apparels,” “Miscellaneous Machinery,” etc. In terms of absolute patent numbers, SMEs are most productive in “Semiconductors & Electronics” with 1,111 patents (31.41% of the patents), “Computers & Peripherals” with 249 patents (28% of the patents), and “Electronics Appliances & Components” with 261 patents (28% of the patents). Interestingly, in the field of “Semiconductors & Electronics,” MNEs dominate with 1,830 patents (52% of total).

112

Table 4.6: Break-up of patenting activity by inventor type Economy Taiwan

Korea

Hong Kong

Singapore

India

China

Period 1970-79 1980-89 1990-99 1970-79 1980-89 1990-99 1970-79 1980-89 1990-99 1970-79 1980-89 1990-99 1970-79 1980-89 1990-99 1970-79 1980-89 1990-99

Multinationals 2.9% 1.9% 1.9% 14.7% 2.5% 0.8% 26.1% 17.3% 16.6% 50.0% 19.7% 45.7% 54.5% 48.1% 29.6% 14.5% 14.4% 17.2%

Business groups 0.0% 0.5% 3.5% 2.9% 31.4% 80.7% 0.6% 6.5% 11.1% -

113

Individuals Domestic firms & orgs 87.7% 9.4% 87.0% 10.6% 59.0% 35.6% 69.1% 13.2% 47.3% 18.8% 6.8% 11.7% 45.5% 28.4% 31.5% 51.2% 30.7% 52.7% 43.3% 6.7% 47.0% 33.3% 9.6% 44.7% 24.7% 20.1% 22.2% 23.1% 18.3% 41.0% 76.8% 8.7% 39.6% 46.0% 40.1% 42.7%

all patents arising from Singapore in the 1990s.29 In analysis not reported in Table 4.6, it appears that the relative role of domestic entities is beginning to go up – only 59 of the 148 patents for 1990-94 were granted to domestic entities, while 287 of the 499 patents in 1995-99 were owned by domestic entities. Thus, it seems that recent adoption of a more R&D-oriented policy by the government is helping Singapore to begin developing strong indigenous innovative capabilities as well. Unlike Singapore, Hong Kong seems to have been less reliant on foreign multinationals for the patenting originating from inventions done there, with multinationals accounting for only 17% of the patents. Instead, the innovative landscape in Hong Kong is dominated by small and medium sized enterprises.30 The emergence of Hong Kong’s SMEs sector dates back to the 1950s, when Hong Kong’s entrepot trade with China was stopped. Most of the local enterprises began as small family ventures and therefore fostered the reinvestment of all revenues back into the business itself. The local government also provided several agencies like Hong Kong Productivity Council to facilitate the development of local industries, which helped increase the innovative capacity of SMEs (Hobday, 1995). The results from Table 4.6 highlight that innovation in Taiwan and Korea has been almost exclusively the result of innovation by domestic entities, with multinational 29

Analysis based on CHI research data reveals that local entities --mostly research institutes or government backed SME --constituted 81% of the total 253 patents in “Semiconductors & Electronics” and 94% of the 17 patents in Biotechnology during 1994-2000. On the other hand, multinationals in Singapore were the main source of innovation in “Electrical Appliances & Components” and “Telecommunications Equipment”. However, there has been an increase in the share of patents held by local entities in industries traditionally dominated by MNEs. For instance, 90% of the 20 patents in “Telecommunications Equipment” industry over 1986-1993 went to multinationals while the 68% of 121 patents for 1994-2000 went to multinationals. 30

Our analysis based on CHI research data suggests the industries in Hong Kong where small and medium sized enterprises have been the main source of patenting include “Other Industries,” “Industrial Process Equipment,” “Office Equipment & Cameras”, and “Electric Appliances & Components”.

114

subsidiaries being responsible for less than 2% of the patents in the past two decades. Also, multinationals seem somewhat important in India (30%) but less so for China (17%).31 So there is an enormous variation in the relative role of subsidiaries of foreign multinationals in innovation in different countries. 32 Table 4.7 lists the top 50 patent holders from each of the six countries considered here. The lists illustrate our analysis above. For example, the Taiwanese list is dominated by “Other Domestic Firms or Organizations”, the Korean list is dominated by business groups, Singapore list is dominated by “Foreign Multinationals or Organizations”, and the Hong Kong, India and China lists are a combination of “Domestic Firms or Organizations” and “Foreign Multinationals or Organizations”. An additional insight from these lists is that research institutes play an important role in innovation in most countries. Industrial Technology Research Institute and National Science Council in Taiwan, Electronics and Telecommunications Research Institute and Korea Institute of Science and Technology in Korea, Hong Kong University of Science and Technology in Hong Kong, National University of Singapore in Singapore, Council of Scientific and Industrial Research in India and Tsinghua University in China are examples of important patent holders from their respective countries. Therefore, it appears that public research

31

For both India and China, multinational enterprises are the dominant source of patenting in “Computer & Peripherals” and “Telecommunications Equipment” while domestic entities that have been responsible for most of the patenting in “Chemicals”.

32

Among other countries that we discussed in the aggregate analysis but have not included in the detailed analysis, foreign multinationals subsidiaries are most important for innovation in Malaysia, somewhat important in Brazil, Mexico and Argentina, and least important in Thailand, Chile and Venezuela.

115

Table 4.7(a): Top 50 patent winners for Taiwan (1970-1999) Assignee Name Industrial Technology Research Inst. United Microelectronics Corporation Taiwan Semiconductor Manufacturing Co. National Science Council Vanguard International Semiconductor Winbond Electronics Corp. Hon Hai Precision Ind. Co., Ltd. Mosel Vitelic, Incorporated Acer Peripherals, Inc. Texas Instruments Inc Acer Incorporated Macronix International Co., Ltd. Holtek Microelectronics Inc. Mustek Systems, Inc. Umax Data Systems Inc. Silitek Corporation Primax Electronics Ltd. United Semiconductor Corp. Greenmaster Industrial Corp. Etron Technology, Inc. Powerchip Semiconductor Corp. Tong Lung Metal Industry Co., Ltd. Behavior Tech Computer Corp. E. Lead Electronic Co., Ltd. Delta Electronics Inc. Development Center For Biotechnology Hwa Shin Musical Instrument Co., Ltd. Enlight Corporation Inventec Corporation Fu Tai Umbrella Works, Ltd. Shin Jiuh Corp. Taiwan Fu Hsing Industrial Co., Ltd. Duracraft Corporation Shin Yeh Enterprise Co., Ltd. Quarton, Inc. China Textile Institute Must Systems, Inc. Chung Cheng Faucet Co. Ltd. Chicony Electronics Co., Ltd. Institute Of Nuclear Energy Research Kalloy Industrial Co., Ltd. Compal Electronics, Inc. China Steel Corporation Pan-International Industrial Corporati Food Industry Research And Development Teh Yor Industrial Co., Ltd. Silicon Integrated Systems Corp. Formosa Saint Jose Corporation Yuan Mei Corp. Foxconn International, Inc. Total patents for top 50 assignees Other patents Overall total 1970-99 for Taiwan Fraction of patents held by top 50 assignees

Affiliation Domestic Firm Or Org Domestic Firm Or Org Domestic Firm Or Org Domestic Firm Or Org Domestic Firm Or Org Walsin Lihua Group Domestic Firm Or Org Pacific Electric Wire & C Acer Group Foreign Multinational Acer Group Domestic Firm Or Org Domestic Firm Or Org Domestic Firm Or Org Umax Group Liton Enterprise Group Domestic Firm Or Org Domestic Firm Or Org Domestic Firm Or Org Domestic Firm Or Org Umax Group Domestic Firm Or Org Domestic Firm Or Org Domestic Firm Or Org Domestic Firm Or Org Domestic Firm Or Org Domestic Firm Or Org Domestic Firm Or Org Domestic Firm Or Org Domestic Firm Or Org Domestic Firm Or Org Domestic Firm Or Org Foreign Multinational Domestic Firm Or Org Domestic Firm Or Org Domestic Firm Or Org Domestic Firm Or Org Domestic Firm Or Org Domestic Firm Or Org Domestic Firm Or Org Domestic Firm Or Org Domestic Firm Or Org Domestic Firm Or Org Domestic Firm Or Org Domestic Firm Or Org Domestic Firm Or Org Domestic Firm Or Org Domestic Firm Or Org Domestic Firm Or Org Foreign Multinational

116

Patent Count 1,229 946 752 367 301 216 107 85 70 60 56 55 48 47 47 44 40 36 31 29 28 27 26 25 24 22 22 21 21 20 19 19 18 17 17 17 16 16 15 15 15 15 13 13 13 12 12 12 12 12 5,100 14,883 19,983 25.5%

Table 4.7(b): Top 50 patent winners for Korea (1970-1999) Assignee Name Samsung Electronics Co., Ltd. Daewoo Electronics Company, Ltd. Hyundai Electronics Industries Co., Ltd. Goldstar Company, Ltd. LG Semicon Co., Ltd. LG Electronics Inc. Electronics And Telecommunications Res. Hyundai Motor Co., Ltd. Gold Star Electron Co., Ltd. Samsung Display Devices Co., Ltd. Korea Institute Of Science And Tech. Samsung Electron Devices Co., Ltd. Samsung Aerospace Industries, Ltd. Samsung Electro-Mechanics Co., Ltd. Korea Advanced Institute Of Science Korea Research Institute Of Chem. Tech. Korea Telecommunication Authority Samsung Heavy Industries, Co., Ltd. Lucky Ltd. LG Industrial Systems Co., Ltd. Kia Motors Corp. SKC Limited Daewoo Telecom Co., Ltd. Daewoo Heavy Industries Co., Ltd Pohang Iron & Steel Co., Ltd. Mando Machinery Corp. Ltd. Korea Atomic Energy Research Institute Agency For Defence Development LG Chemical Ltd. Korea Kumho Petrochemical Co., Ltd. Kwangju Electronics Co., Ltd. Samsung Semiconductor & Telecom. Kolon Industries Inc. Sindo Ricoh Co., Ltd. Toray Industries Inc. Samsung Heavy Industry Co., Ltd. Yukong Limited Orion Electric Co., Ltd. Anam Industrial Co., Ltd. Sunkyong Industries Co., Ltd. Cheil Industries, Inc. Pacific Corporation Cheil Foods & Chemicals, Inc. Dong Kook Pharmaceutical Co., Ltd. Anam Semiconductor, Inc. Medison Co., Ltd. Volvo Construction Equipment Korea Co. Korea Chemical Co., Ltd. Samsung Corning Co., Ltd. Korea Institute Of Machinery & Metals Total patents for top 50 assignees Other patents Overall total 1970-99 for Korea Fraction of patents held by top 50 assignees

Affiliation Samsung Group Daewoo Group Hyundai Group LG Group LG Group LG Group Domestic Firm or Org Hyundai Group LG Group Samsung Group Domestic Firm or Org Samsung Group Samsung Group Samsung Group Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Samsung Group LG Group LG Group Kia Group Sunkyong Group Daewoo Group Daewoo Group POSCO Group Halla Group Domestic Firm or Org Domestic Firm or Org LG Group Kumho Group Samsung Group Samsung Group Kolon Group Domestic Firm or Org Foreign Multinational or Org Samsung Group Sunkyong Group Daewoo Group Anam Group Sunkyong Group Samsung Group Pacific Group Domestic Firm or Org Domestic Firm or Org Anam Group Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Samsung Group Domestic Firm or Org

117

Patent Count 5,350 1,008 931 892 696 566 397 347 252 243 238 214 131 124 105 100 96 71 68 65 62 51 42 36 35 30 29 27 25 25 24 23 23 22 20 19 19 18 17 16 16 16 14 13 13 12 12 11 11 10 12,585 2,253 14,838 84.8%

Table 4.7(c): Top 50 patent winners for Hong Kong (1970-1999) Assignee Name Astec International, Ltd. Johnson Electric S.A. Johnson Electric Industrial Manuf. Motorola Inc W. Haking Enterprises Limited The Hong Kong University Of Science & Tech. World-Wide Stationery Manufacturing Co China Pacific Trade Ltd. Chiaphua Industries, Ltd. Playart Limited Polycity Industrial Ltd. Arco Industries Ltd. Solar Wide Industrial Limited T. K. Wong & Associates Limited Pentalpha Enterprises Ltd. Leco Stationery Manufacturing Co., Ltd Outboard Marine Corp John Manufacturing Limited Mego Corp. Mr. Christmas, Incorporated Asm Assembly Automation Ltd. The Chinese University Of Hong Kong The Hong Kong Polytechnic University Wing Shing Products (Bvi) Co. Ltd. Alza Corp Computer Products Inc Windmere Corp Achiever Industries Limited G. E. W. Corporation Limited International Quartz Ltd. Meyer Manufacturing Company Limited Payview Limited Tradebest International Corporation United Chinese Plastics Products Co. Pacusma Co. Ltd. East Asia Services Ltd. Addway Engineering Limited Conair Corp General Electric Company Polaroid Corp Recoton Corp Tiger Electronics, Inc. Timex Corporation Concord Camera Corp. Heep Tung Manufactory Limited Kwoon Kwen Metal Ware Company Limited Maxpat Trading & Marketing Refined Industry Company Limited Simatelex Manufactory Company Limited Sonca Industries Limited Total patents for top 50 assignees Other patents Overall total 1970-99 for Hong Kong Fraction of patents held by top 50 assignees

Affiliation Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Foreign Multinational or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Foreign Multinational or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Foreign Multinational or Org Domestic Firm or Org Foreign Multinational or Org Foreign Multinational or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org

118

Patent Count 44 33 24 23 20 15 14 12 12 11 10 10 8 7 7 7 7 6 6 6 5 5 5 5 5 5 5 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 3 3 3 3 3 3 403 870 1,273 31.7%

Table 4.7(d): Top 50 patent winners for Singapore (1970-1999) Assignee Name Chartered Semiconductor Manufacturing Hewlett-Packard Co National University Of Singapore Texas Instruments Inc Motorola Inc Thomson SA Molex Inc Tritech Microelectronics International Matsushita Electric Industrial Co Ltd Philips SGS-Thomson Microelectronics (Pte) Ltd Sun Industrial Coatings Private Ltd. Tritech Microelectronics, Ltd. Chartered Industries Of Singapore Priv Institute Of Microelectronics Nestec, S.A. Berg Technology, Inc. Seagate Technology Siemens Aktiengesellschaft Eastern Oil Tools Pte, Ltd. Singapore Computer Systems Limited Institute Of Microelectronics Sunright Limited Advanced Systems Automation Limited Apple Computer Inc Du Pont Advanced Materials Technologies Pte Lt Enteron, L.P. United Technologies Corp Whitaker Corporation Creative Technology Limited Varta Batterie A.G. Sumitomo Chemical Company, Limited Nortrans Shipping And Trading Far East Abb Vetcogray Inc. Litton Industries Black & Decker Corp Chevron Rmt, Inc. Thomas & Betts Corp Symtonic Sa Rhone Poulenc Industries Hitachi Chemical Company, Ltd. Toshiba Corporation Sandvik Multiscience System Pte. Ltd. Port Of Singapore Authority Singapore Institute Of Standards And I Aztech Systems Ltd. Matsushita Refrigeration Industries Total patents for top 50 assignees Other patents Overall total 1970-99 for Singapore Fraction of patents held by top 50 assignees

Affiliation Domestic Firm or Org Foreign Multinational or Org Domestic Firm or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Domestic Firm or Org Foreign Multinational or Org Foreign Multinational or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Foreign Multinational or Org Foreign Multinational or Org Domestic Firm or Org Domestic Firm or Org Foreign Multinational or Org Foreign Multinational or Org Domestic Firm or Org Foreign Multinational or Org Foreign Multinational or Org Domestic Firm or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Foreign Multinational or Org

119

Patent Count 122 43 35 35 28 23 23 21 18 11 9 8 8 7 7 6 6 6 5 5 5 5 5 5 5 5 4 4 4 4 4 3 3 3 3 3 3 3 3 3 2 2 2 2 2 2 2 2 2 2 523 221 744 70.3%

Table 4.7(e): Top 50 patent winners for India (1970-1999) Assignee Name Council Of Scientific And Industrial Research Hoechst Ciba-Geigy Corporation Ranbaxy Laboratories Limited Unilever NASA Texas Instruments Inc Dr. Reddy'S Research Foundation Lupin Laboratories Limited Indian Explosives Ltd. General Electric Company National Institute Of Immunology Monsanto Co. Panacea Biotec Limited Iowa India Investments Company Limited Indian Oil Corporation, Ltd. Union Carbide Corp Elf Aquitaine Cadbury India Limited Indian Petrochemicals Corporation Ltd. Gem Energy Industry Limited Aktiebolaget Astra Procter & Gamble Fiberstars, Inc. Xerox Corp Novartis (Sandoz) Forschungszentrum Julich Gmbh Licentia Patent-Verwaltungs-Gmbh Boots Company Plc Imperial Chemical Industries Zeneca Limited All India Institute Of Medical Science Hawkins Cookers Limited Iel Limited Indian Space Research Organisation Karamchand Premchand Private Limited Sree Chitra Tirunal Inst. For Medical National Chemical Laboratory The Chief Controller, Research And Dev GEC Westinghouse Electric Corp American Cyanamid Co Analog Devices Avnet Inc Johnson & Johnson Mobil Sri International United States Of America, Air Force University Of California University Of Minnesota Total patents for top 50 assignees Other patents Overall total 1970-99 for India Fraction of patents held by top 50 assignees

Affiliation Domestic Firm or Org Foreign Multinational or Org Foreign Multinational or Org Ranbaxy Group Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Dr. Reddy's Group Lupin Group Domestic Firm or Org Foreign Multinational or Org Domestic Firm or Org Foreign Multinational or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Foreign Multinational or Org Foreign Multinational or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org Foreign Multinational or Org

120

Patent Count 141 45 38 20 19 18 17 10 9 8 8 7 7 6 4 4 4 3 3 3 3 3 3 3 3 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 439 257 696 63.1%

Table 4.7(f): Top 50 patent winners for China (1970-1999) Assignee Name China Petrochemical Development Corp. United Microelectronics Corporation Tsinghua University Autry Industries, Inc. Industrial Technology Research Inst. China Petrochemical Corporation Fujian Institute Of Research North China Research Institute Of Elec. Peking University Shanghai Institute Of Biochemistry Taiho Pharmaceutical Company Limited Acer Incorporated Beijing Research Institute Of Chem. Chinese Academy Of Medical Sciences Huazhong Institute Of Technology Institute Of Physics, Chinese Academy Shanghai Institute Of Organic Chemistry Tianjin University CSL Opto-Electronics Corp. Nan Kai University Central Iron & Steel Research Inst. Bayer Leco Stationery Manufacturing Co., Ltd Beijing Polytechnic University China Metallurgical Import & Export Co. China National Seed Corporation Jilin University Of Technology Luoyang Petrochemical Engineering Corp Qing-Yang Machine Works Research Institute Of Petroleum Proces Science & Technic Department Of Dagang Shanghai Lamp Factory Institute Of Materia Medica Chinese Building Technology Services University Of Electronic Science And Tech. South China University Of Technology Research Institute Of Petroleum Proc. Traditional Chinese Medicine Research Dalian Institute Of Chemical Physics University Of Science And Technology Shanghai Yue Long Nonferrous Metals Ltd. Vasomedical, Inc. Panzhihua Iron And Steel (Group) Co. Wonder & Bioenergy Hi-Tech International Pacific Sources, Inc. Fushun Research Institute Of Petroleum Plastic Advanced Recycling Corp. Institute Of Materia Medica, An Inst. Liaohe Petroleum Exploration Bureau Jiangsu Goodbaby Group, Inc. Total patents for top 50 assignees Other patents Overall total 1970-99 for China Fraction of patents held by top 50 assignees

Affiliation Domestic Firm or Org Foreign Multinational or Org Domestic Firm or Org Foreign Multinational or Org Foreign Multinational or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Foreign Multinational or Org Foreign Multinational or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Foreign Multinational or Org Foreign Multinational or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org Domestic Firm or Org

121

Patent Count 26 21 10 8 7 5 4 4 4 4 4 4 3 3 3 3 3 3 3 3 3 3 3 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 188 582 770 24.4%

institutes have played an important role not just in assimilating and diffusing foreign technology but also in generating new ideas. For example, KIET (Korea Institute of Electronics Technology) started out as a “demonstration laboratory” for showing the efficient implementation of complex imported production processes such as integrated circuit wafer fabrication. With the development of private R&D, ETRI (Electronics and Telecommunications Research Institute), which evolved from KIET, shifted its focus from technology transfer and applied R&D to basic research and innovation. Similarly, Singapore’s National Technology Plan and National Science and Technology Board made major investments to fund R&D and increase the number of local researchers in the 1990s, which may account for the increase in patents during the late 1990s by institutes such as the National University of Singapore and domestic SMEs affiliated with it. For China and also India to some extent, the top 50 inventors list seems to have a disproportionately high number of research institutes and government-affiliated organizations, indicating that private-sector R&D and innovation has not developed much yet in these countries. We can also calculate the fraction of the country’s patents held by its top 50 assignees in order to get a measure of how concentrated innovative activity is in different economies. This number is found to be the highest for Korea (85%), followed by Singapore (70%), India (63%), Hong Kong (32%), Taiwan (26%) and finally China (24%). This is not surprising, given that economic activity in Korea and Singapore is dominated largely by large players (whether domestic or multinational) while that in Taiwan and China is dominated by individuals and SMEs.

122

8. Concluding thoughts We have used US patent data to study innovation in Asian economies. Our results are consistent with prior evidence (Dahlman, 1994; Rausch, 1995; Choung, 1998) that there has been a rise in technological capability over time in East Asian economies, and dramatically so for Korea and Taiwan. Another key finding of our paper is that the emerging economies are quite heterogeneous bunch in their technological capabilities. In particular, they differ a lot in extent of patenting, areas of specialization and driving players behind innovation. We demonstrate that the newly industrialized countries have achieved leadership even in sectors that are on the frontier of technological progress, and are not specializing in just the more mature sectors where the developed countries might not compete in anymore. Further, the areas of specialization for each country have evolved very slowly over time. Thus, our analysis extends previous research that reached analogous conclusions in study of patenting activity by developed countries (e.g. Patel and Pavitt, 1998; Archibugi and Pianta, 1998). More generally, it contributes to the literature that shows that the sources and areas of technological specialization are heavily dependent on the individual national systems of innovation (Lundvall, 1992; Nelson, 1993; Edquist, 1997; Freeman and Soete, 1997). Previous research has established that wide differences in nations have led to a great deal of variation across countries in the economic role played by multinationals, business groups, individuals, private firms and government institutes. Our analysis of patent data is consistent with this finding. For example, while large-scale conglomerates like Samsung, Daewoo, Hyundai and LG Group dominate innovation in Korea, innovation in Taiwan and Hong Kong is a result of domestic individuals and independent

123

firms and that in Singapore is heavily influenced by foreign firms. We find innovative activity to be most concentrated in Korea, fairly concentrated in Singapore and much less concentrated in Taiwan and Hong Kong. While the data and analysis presented in this paper do not conclusively settle the accumulation versus assimilation debate, we feel that they do make new and interesting contribution to the discussion. While Korea and Taiwan are now definitely two of the world's leading innovators, Singapore and Hong Kong do not seem to have made any such transition yet (though the recent trends are promising). This may partially be explained by the fact that while the former two have been taking aggressive policy steps to develop indigenous technological capabilities, the latter two have been quite content (until recently) in importing foreign technologies rather than making cutting-edge innovations themselves. An important lesson is that the "Asian Tigers" are actually a heterogeneous bunch, and different mechanisms could be behind economic success in different countries. While the evidence in this paper informally suggests that innovation might play an important role in growth, more needs to be done to address this problem formally. Important contributions have already been made in studying this subject (e.g. see the excellent discussions and references in Archibugi and Jonathan Michie, 1998; Archibugi, Howells and Michie, 1999; Laursen, 2000). However, most research has focused only on developed countries, leaving room for further research on innovation in other parts of the world. We hope that our paper will be useful in motivating further research in this area.

124

REFERENCES Agrawal, A., I. Cockburn and J. Mchale. 2003. Gone but not forgotten: labor flows, knowledge spillovers, and enduring social capital. National Bureau of Economic Research Working Paper 9950. Ahuja, G. 2000. Collaboration networks, structural holes, and innovation: a longitudinal study. Administrative Science Quarterly, 45: 425-455. Aitken, B. and A. Harrison. 1999. Do domestic firms benefit from foreign investment? Evidence from Venezuela. American Economic Review 89: 605-618 Allen, T.J. 1977. Managing the Flow of Technology. Cambridge, MA: MIT Press. Almeida P. 1996. Knowledge sourcing by foreign multinationals: patent citation analysis in the U.S. semiconductor industry. Strategic Management Journal 7: 155-165. Almeida, P. and B. Kogut. 1999. The localization of knowledge and the mobility of engineers in regional networks. Management Science 45(7), 905-917. Amemiya, T. 1985. Advanced Econometrics. Harvard University Press, Cambridge. Amsden, A.H. 1989. Asia’s Next Giant: South Korea and Late Industrialization. Oxford University Press. Amsden, A.H. and T. Hikino. 1994. Project execution capability, organizational know-how and conglomerate corporate growth in late industrialization. Industrial and Corporate Change 3(1): 111-148. Archibugi, D., J. Howells and J. Michie (Eds.). 1999. Innovation Policy in a Global Economy. Cambridge University Press. Archibugi, D. and J. Michie (Eds.). 1998. Trade, Growth and Technical Change. Cambridge University Press. Archibugi, D. and J. Michie 1998. Trade, growth, and technical change: what are the issues? In: Archibugi, D. and J. Michie (Eds.), Trade, Growth and Technical Change. Cambridge University Press. Archibugi, D. and M. Pianta. 1992. The Technological Specialization of Advanced Countries: A Report to the EEC on International Science and Technology Activities. Kluwer Academic Publishers. Archibugi, D. and M. Pianta. 1998. Aggregate convergence and sectoral specialization in innovation: evidence for industrial countries. In: Archibugi, D. and J. Michie (Eds.), Trade, Growth and Technical Change. Cambridge University Press.

125

Audretsch, D.B. and M.P. Feldman. 1996. R&D spillovers and the geography of innovation and production. American Economic Review, 86(3): 630-640. Bartlett, C.A. and S. Ghoshal. 1989. Managing Across Borders: The Transnational Solution. Harvard Business School Press: Boston, MA. Bell, M. and K. Pavitt. 1993. Technological accumulation and industrial growth: contrasts between developing and developed countries. Industrial and Corporate Change 2(2). Branstetter, L. 2000. Is foreign direct investment a channel of knowledge spillovers? Evidence from Japan’s FDI in the United States. National Bureau of Economic Research Working Paper # 8015. Branstetter, L. 2001. Are Knowledge Spillovers International or Intranational in Scope? Microeconometric Evidence from the U.S. and Japan. Journal of International Economics 53: 53-79. Breschi, S. and F. Lissoni . 2002. Mobility and social networks: localised knowledge spillovers revisited. Mimeo. Buckley, P.J. and M.C. Casson. 1976. The Future of the Multinational Enterprise. London: Holmes & Meier. Burt, R.S. 1992. Structural Holes: The Social Structure of Competition. Harvard University Press: Cambridge, MA. Cantwell, J. 1989. Technological Innovation and Multinational Corporations. Oxford: Basil Blackwell. Cantwell, J.A. and O. Janne. 1999. Technological globalisation and innovative centres: the role of corporate technological leadership and locational hierarchy. Research Policy 28:119-144. Caves, R.E. 1974. Multinational firms, competition and productivity in hostcountry markets. Economica 41: 176-193. Caves, R.E. 1996. Multinational Enterprise and Economic Analysis. Cambridge University Press (Second Edition). Cheng-Fen, C. and G. Sewell. 1996. Strategies for technological development in South Korea and Taiwan: the case of semiconductors. Research Policy 25 (5), 759-783. Choung, J.Y. 1998. Patterns of innovation in Korea and Taiwan. IEEE Transactions on Engineering Management 45 (4), 357-365. Chung, J.S. 1986. National Policies for Developing High Technology Industries-International Comparisons: Korea. Westview Press, London.

126

Chung, W. 2001. Identifying technology transfer in foreign direct investment: influence of industry conditions and investing firm motives. Journal of International Business Studies, 32 (3): 211-229. Chung, W. and J. Alcacer. 2002. Knowledge seeking and location choice of foreign direct investment in the United States. Management Science 48(12):1534-1554. Chung, W., W. Mitchell and B Yeung. 2003. Foreign direct investment and host country productivity: The American automotive component industry in the 1980s. Journal of International Business Studies. 34(2): 199-218. Cockburn, I.M. and R.M. Henderson. 1998. Absorptive capacity, coauthoring behavior, and the organization of research in drug discovery. Journal of Industrial Economics, 46(2): 157-182. Coe, D.T. and E. Helpman. 1995. International R&D spillovers. European Economic Review 39: 859-887. Cohen, W. and D. Levinthal. 1989. Innovation and learning: The two faces of R&D. Economic Journal 99: 569-596. Coleman, J.S. 1988. Social capital in the creation of human capital. American Journal of Sociology 94: S95-S120. Coleman, J.S., E. Katz and H. Menzel. 1966. Medical Innovation. New York: Bobbs-Merrill. Collins, S.M. and B. Bosworth. 1996. Economic growth in East Asia: accumulation versus assimilation. Brookings Papers on Economic Activity 2, 135-204 Cormen, T.H., C.E. Leiserson and R. L. Rivest. 1990. Introduction to Algorithms. MIT Press, Cambridge, MA. Dahlman, C. 1994. Technology strategy in East Asian developing economies. Journal of Asian Economics 5, 541-572. Dalton, D.H. and M.G. Shapiro. 1995. Globalizing Industrial Research & Development. Office of Technology Policy, U.S. Dept of Commerce. Dosi, G. 1982. Technological paradigms and technological trajectories: a suggested interpretation of the determinants and directions of technical change. Research Policy 11 (3), 147-162. Duget, E. and M. Macgarvie. 2002. How Well Do Patent Citations Measure Knowledge Spillovers? Mimeo. Dunning, J.H. 1993. Multinational Enterprises and the Global Economy. Addison-Wesley.

127

Eaton, J. and S. Kortum. 1999. International Patenting and Technology Diffusion: Theory and Measurement. International Economic Review 40: 537-570. Ethier, W. 1986. The multinational firm. Quarterly Journal of Economics 101: 805-834. Fagerberg, J. 1987. A technology gap approach to why growth rates differ. Research Policy 16: 87-99. Feinberg, S. and S.K. Majumdar . 2001. Technology spillovers from foreign direct investment in the Indian pharmaceutical industry. Journal of International Business Studies 32(3): 421-438. Feinberg, S. and A.K. Gupta. 2003. Knowledge spillovers and the assignment of R&D responsibilities to foreign subsidiaries. Strategic Management Journal. Fleming, L., L. Colfer, A. Marin and J. Mcphie . 2003. Why the valley went first: agglomeration and emergence in regional inventor networks. Mimeo. Florida, R. 1997. The globalization of R & D: results of a survey of foreignaffiliated R&D laboratories in the USA. Research Policy 26(1): 85-103. Freeman, C. and L. Soete. 1997. The Economics of Industrial Innovation. MIT Press, Cambridge, MA. Frost, T.S. 2001. The geographical sources of foreign subsidiaries’ innovations. Strategic Management Journal 22: 101-123. Frost, T.S., J.M. Birkinshaw and P.C. Ensign. 2003. Centers of excellence in multinational firms. Strategic Management Journal 23: 997-1018. Ghoshal, S., H. Korine and G. Szulanski. 1994. Interunit communication in multinational corporations. Management Science 40: 96-110. Glaeser, E.L, D. Laibson and B. Sacerdote. 2002. The economic approach to social capital. Economic Journal. Globerman, S., A. Kokko and F. Sjöholm. 2000. International technology diffusion: evidence from Swedish patent data. Kyklos 53: 17-38. Gomes-Casseres, B., A.B. Jaffe and J. Hagedoorn. 2003. Do alliances promote knowledge flows? Mimeo. Gompers, P., J. Lerner and D. Scharfstein. 2002. Entrepreneurial spawning: public corporations and the genesis of new ventures, 1986-1999. Mimeo. Granovetter, M.S. 1973. The strength of weak ties. American Journal of Sociology. 78: 1360-1380.

128

Grant, R.M. 1996. Toward a knowledge-based theory of the firm. Strategic Management Journal, 17: 109-122. Greene, W. 2003. Econometric Analysis. Prentice Hall, 5th Edition. Griliches, Z. 1990. Patent statistics as economic indicators: a survey. Journal of Economic Literature 28: 1661-1797. Grossman, G. and E. Helpman. 1991. Innovation and Growth in the World Economy, Cambridge, MA: MIT Press. Guerrieri, P. and C. Milana. 1998. High-technology industries and international competition. In: Archibugi, D. and J. Michie (Eds.), Trade, Growth and Technical Change. Cambridge University Press. Hansen, M.T. 1999. The search-transfer problem: the role of weak ties in sharing knowledge across organization subunits. Administrative Science Quarterly, 44: 82-111. Head, K., J. Ries and D. Swenson . 1995. Agglomeration benefits and location choice: evidence from Japanese manufacturing investments in the United States. Journal of International Economics, 38: 223-247. Hedlund, G. 1986. The hypermodern MNC: A heterarchy? Human Resource Management 25(1): 9-35. Hellmann, T. 2002. When do Employees become Entrepreneurs? Working Paper 1770, Graduate School of Business, Stanford University. Hikino, T. and A.H. Amsden. 1994. Staying behind, stumbling back, sneaking up, soaring ahead: late industrialization in historical perspective. In: Baumol, W.J. Nelson, R.R. Edward N.W. Convergence of Productivity: Cross-National Studies and Historical Evidence. Oxford University Press. Hobday, M. 1995. Innovation in East Asia. Edward Elgar Publishing Ltd. Huber, G.P. 1991. Organizational learning: the contributing processes and the literatures. Organization Science, 2(1): 88-115. Hymer, S.H. 1976. The International Operations of National Firms: A Study of Direct Investment. MIT Press, Boston, MA. Jaffe, A.B. and M. Trajtenberg. 2002. Patents, Citations & Innovations: A window on the knowledge economy. MIT Press, Cambridge, MA. Jaffe, A.B., M. Trajtenberg and R. Henderson. 1993. Geographic localization of knowledge spillovers as evidenced by patent citations. Quarterly Journal of Economics 434: 578-598.

129

Keller, W. 2002. Geographic localization of international technology diffusion. American Economic Review 92(1). Khanna, T. 2000. Business groups and social welfare in emerging markets: existing evidence and unanswered questions. European Economic Review 44, 748-761. Khanna, T. and J.W. Rivkin. 2001. Estimating the performance effects of business groups in emerging markets. Strategic Management Journal 22, 45-74. Kim, J. and L. Lau. 1994. The sources of economic growth of the East Asian newly industrialized countries. Journal of the Japanese and the International Economies 8 (3), 235-271. Kim, L. 1998. From Imitation to Innovation: Dynamics of Korea’s Technological Learning. Harvard Business School Press, Boston. King, G. and L. Zeng. 2001. Logistic regression in rare events data. Political Analysis 9(2): 137-163 Klepper, S. 2001. Employee startups in high-tech industries. Industrial and Corporate Change, 10:639-674. Kogut, B. and S. J. Chang. 1991. Technological Capabilities and Japanese Foreign Direct Investment in the United States. The Review of Economics and Statistics 73 (3): 401-413. Kogut, B. and U. Zander. 1992. Knowledge of the firm, combinative capabilities, and the replication of technology. Organization Science. 3 (3): 383-397. Kogut, B. and U. Zander. 1993. Knowledge of the firm and the evolutionary theory of the multinational corporation. Journal of International Business Studies. 24(4), pp. 625-645. Krugman, P. 1994. The myth of the Asian miracle. Foreign affairs. 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: 179-193. Laursen, K. 1999. The impact of technological opportunity on the dynamics of trade performance. Structural Change and Economic Dynamics 103 (4), 341-357. Laursen, K. 2000. Trade Specialization, Technology and Economic Growth: Theory and Evidence from Advanced Countries. Edward Elgar Pub. Lerner, J. 2002. 150 years of patent protection. American Economic Review Papers and Proceedings, 92: 221-225.

130

Levin, D. and R. Cross. 2003. The strength of weak ties you can trust: the mediating role of trust in effective knowledge transfer. Management Science, Forthcoming. Levin, R., A. Klevorick, R. Nelson and S. Winter. 1987. Appropriating the returns from industrial research and development. Brookings Papers on Economic Activity 3: 783-820. Levitt, V. and J.G. March. 1988. Organizational Learning. Annual Review of Sociology. 14: 319-340. Lundvall, B. (Ed.). 1992. National Systems of Innovation: Towards a Theory of Innovation and Interactive Learning. London, Pinter. Mansfield, E., D. Teece and A. Romeo. 1979. Overseas research and development by US-based firms. Economica 46(182): 187-196. Manski, C.F. and S. R. Lerman. 1977. The estimation of choice probabilities from choice based samples. Econometrica 45(8): 1977-88. Mowery, D.C., J.E. Oxley and B.S. Silverman. 1996. Strategic alliances and interfirm knowledge transfer. Strategic Management Journal 17: 77-91. Nelson, R. and S. Winter. 1982. An Evolutionary Theory of Economic Change. Harvard University Press: Cambridge, MA. Nelson, R.R. (Ed.). 1993. National Innovation System. New York. Oxford University Press. Nelson, R.R. and H. Pack. 1998. The Asian miracle and modern growth theory. Policy Research Working Paper No. 1881, Development Research Group, The World Bank. Newman, M.E.J. 2001. The structure of scientific collaboration networks. Proceedings of National Academy of Science 98: 404-409. Nohria, N. and S. Ghoshal. 1997. The Differentiated network: Organizing Multinational Corporations for Value Creation. Jossey-Bass Publishers, San Francisco. Nonaka, I. 1994. A dynamic theory of organizational knowledge creation. Organization Science, 5(1): 14-37. OECD. 1998. Internationalisation of Industrial R&D: Patterns and Trends. Pack, H. 1992. Technology gaps between industrial and developing countries: Are there dividends for latecomers? In: Summers, L. Shah, S. (Eds.), Proceedings of the World Bank Annual Conference on Development Economics. The World Bank.

131

Patel, P. and K. Pavitt. 1998. Uneven and divergent technological accumulation among advanced countries: evidence and a framework of explanation. In: Archibugi, D. and J. Michie(Eds.), Trade, Growth and Technical Change. Cambridge University Press. Pavitt, K. 1988a. International patterns of technological accumulation. In: Hood, N. And Vahlne, J.E. (Eds.), Strategies in Global Competition. London: Croom Helm. Pavitt, K. 1988b. Uses and abuses of patent statistics. In: Van Raan, A. (Ed.), Handbook of Quantitative Studies of Science and Technology. Amsterdam, Elsevier. Peri, G. 2003. Knowledge Flows, R&D Spillovers and Innovation. Mimeo. Polanyi, M. 1966. The Tacit Dimension. London: Routledge & Kegan Paul. Porter, M.E. 1990. The Competitive Advantage of Nations. Free Press. Rausch, L.M. 1995. Asia's new high-tech competitors: an SRS report. National Science Foundation 95-309 Rogers, E.M. 1985. Diffusion of Innovations. New York: Free Press. Romer, P.M. 1990. Endogenous Technological Change. Journal of Political Economy 98 (5): S71-S102. Rosenkopf, L. and P. Almeida. 2003. Overcoming local search through alliances and mobility. Management Science 49(6). 0751-0766. Ryan, B. and N. Gross. 1943. The diffusion of hybrid seed corn in two Iowa communities. Rural Sociology, 8(1): 15-24. Saxenian, A.L. 1994. Regional Advantage: Culture and Competition in Silicon Valley and Route 128. Cambridge: Harvard University press. Saxenian, A.L. 2002. Transnational communities and the evolution of global production networks: the cases of Taiwan, China and India. Industry and Innovation, 9(3): 183-202. Scherer, F. M. 1983. The propensity to patent. International Journal of Industrial Organization 1: 107-128. Shane, S. and D. Cable . 2002. Network Ties, Reputation, and the Financing of New Ventures. Management Science 48 (3): 364-381. Shaver, J.M. and F. Flyer. 2000. Agglomeration economies, firm heterogeneity, and foreign direct investment in the United States. Strategic Management Journal 21: 1175-1193. Simon, H.A. 1991. Bounded rationality and organizational learning. Organization Science, 2: 125-134. 132

Soete, L. 1987. The impact of technological innovation on international trade patterns: the evidence reconsidered. In: Freeman, C. (Ed.), Output Measurement in Science and Technology. North Holland, Amsterdam. Sorenson, O. and L. Fleming. 2001. Science and the diffusion of knowledge. Working paper 02-095, Harvard Business School. Sorenson, O. and T.E. Stuart. 2001. Syndication networks and the spatial distribution of venture capital investments. American Journal of Sociology, 106(6): 154688. Spencer, J.W. 2000. Knowledge flows in the global innovation system: do U.S. firms share more scientific knowledge than their Japanese rivals? Journal of International Business Studies 31(3): 521-530. Stolpe, M. 2001. Mobility of research workers and knowledge diffusion as evidenced in patent data the case of liquid crystal display technology. Kiel Working Paper No. 1038. Stopford, J.M. 1992. Directory of Multinationals. Stockton Press: New York. Stuart, T. and O. Sorenson . 2003. The geography of opportunity: spatial heterogeneity in founding rates and the performance of biotechnology firms. Research Policy 32: 229-253. Szulanski, G. 1996. Exploring internal stickiness: impediments to the transfer of best practice within the firm. Strategic Management Journal, 17: 27-43. Teece, D.J. 1986. Transaction cost economics and multinational enterprise. Journal of Economic Behavior and Organization 7: 21-45. Thompson, P. and M. Fox-Kean. 2004. Patent citations and the geography of knowledge spillovers: a reassessment. American Economic Review, forthcoming. Tsai, W. and S. Ghoshal . 1998. Social capital and value creation: the role of intrafirm networks. Academy of Management Journal, 41: 464-476. Uzzi, B. 1996. The sources and consequences of embeddedness for the economic performance of organizations: the network effect. American Sociological Review, 61: 674-698. Uzzi, B. and R. Lancaster. 2003. Relational embeddedness and learning: The case of bank loan managers and their clients. Management Science, 49: 383-399. Von Hippel, E. 1988. The Sources of Innovation, Cambridge: MIT Press. Wasserman, S. and K. Faust . 1994. Social Network Analysis: Methods and Applications. Cambridge University Press.

133

Watts, D.J. and S. Strogatz. 1998. Collective dynamics of small world networks. Nature. 393: 440-442. Williamson, O.E. 1985. The Economic Institutions of Capitalism. The Free Press. Young, A. 1995. The tyranny of numbers: confronting the statistical realities of the East Asian growth experience. Quarterly Journal of Economics 110 (3), 641-680. Zander, U. and B. Kogut . 1995. Knowledge and the speed of the transfer and imitation of organizational capabilities: an empirical test. Organization Science, 6: 76-91. Zucker, L.G., M.R. Darby and M.B. Brewer . 1998. Intellectual human capital and the birth of u.s. biotechnology enterprises. American Economic Review 88 (1): 290-306.

134