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 father-

in-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 knowledge-

intensity 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 Fox-

Kean, 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 parent-

subsidiary 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 Total patents 1986-95 in

NBER database

Total number of assigned

patents

Assigned patents with clean parent information

Fraction of patents from multinational subsidiaries

(1) (2) (3) (4)United States 546,824 418,045 287,787 8.5%Japan 217,313 212,427 183,870 2.1%Germany 74,041 67,154 45,869 19.5%France 29,791 27,120 17,289 20.4%United Kingdom 26,631 23,968 15,131 40.3%Canada 20,700 13,015 5,697 50.0%Subtotal 6 countries 915,300 761,729 555,643 9.0%Other countries 94,924 73,115 38,402 27.3%Total worldwide 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 over-

estimation 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 over-

estimation 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

D

DMDM

D

DMDM

DMDMDM

Npp

Npp

ppt

)1()1( →→→→

→→→ ′−′

+−

′−=

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 Control Citations Comparison(1) (2) (3) (4) (5) (6) (7) (8)

Country Total citations by

domestic

Citations by domestic to

mult sub

%Citations by domestic to mult sub

Citations by domestic to

mult sub

%Citations by domestic to mult sub

(3) - (5) t-ratio (3)/(5)

United States 430,262 17,010 3.95% 15,136 3.52% 0.44% 10.7 1.12Japan 245,441 2,082 0.85% 1,879 0.77% 0.08% 3.2 1.11Germany 27,326 658 2.41% 542 1.98% 0.42% 3.4 1.21France 12,727 124 0.97% 101 0.79% 0.18% 1.5 1.23United Kingdom 7,895 197 2.50% 149 1.89% 0.61% 2.6 1.32Canada 3,536 32 0.90% 15 0.42% 0.48% 2.5 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 Control Citations Comparison(1) (2) (3) (4) (5) (6) (7) (8)

Country Total citations by

mult sub

Citations by mult sub to

domestic

%Citations by mult sub to domestic

Citations by mult sub to

domestic

%Citations by mult sub to domestic

(3) - (5) t-ratio (3)/(5)

United States 41,272 22,590 54.73% 18,799 45.55% 9.19% 26.5 1.20Japan 5,156 2,464 47.79% 2,083 40.40% 7.39% 7.6 1.18Germany 10,841 1,302 12.01% 985 9.09% 2.92% 7.0 1.32France 3,856 166 4.30% 114 2.96% 1.35% 3.2 1.46United Kingdom 9,689 220 2.27% 274 2.83% -0.56% -2.5 0.80Canada 3,457 38 1.10% 25 0.72% 0.38% 1.6 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

ββixii e

xxxy −+=Λ===

11)()|1Pr(

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 co-location 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 “choice-

based” 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

{ } { }∑∑∑=

−

==

+−=Λ−+Λ=n

i

xyi

yi

yiw

ii

ii

ewL1

)21(

01

)1ln()1ln(1)ln(1ln β

αγ

)1)(/1()/1( iii yyw

where −+= αγ . 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 pseudo-

maximum 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

Presence of citing assignee in cited country

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

Presence of cited assignee in citing country

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

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** 0.578** 0.520**

(0.009) (0.005) (0.009)[3.83] [3.29] [2.96]

Within same MNC 3.291** 2.110** 1.825**(0.110) (0.026) (0.050)[18.76] [12.03] [10.40]

Technological relatedness: Same tech category 1.148** 1.108**

(0.011) (0.012)

Same tech subcategory 1.246** 1.218**(0.014) (0.015)

Same primary tech class 3.243** 1.930**(0.011) (0.015)

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 usedThe dependent variable is 1 if there is a citation between two patents, 0 otherwiseRobust standard errors in parentheses, with clustering on citing patentMarginal effects in square brackets after multiplication with 1,000,000Fixed 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 Fox-

Kean (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

8 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.

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

D→M / D→D 0.99

M→D / D→D 0.70**

M→M / D→D 1.46**

M→D / D→M 0.70**

H→S / S→H 1.03

A weighted logit regression is usedThe dependent variable is 1 if there is a citation between two patents, 0 otherwiseRobust standard errors in parentheses, with clustering on citing patentMarginal effects in square brackets after multiplication with 1,000,000Controls for technological similarity of citing and cited patent included in regression, but not shownFixed 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 sub-

categories: 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

D →MM →D

D →DIBM[Home Base]

USA

NEC

Intel[Home Base]

SonyM →M

NEC[Home Base]

Japan

S →H

H →S IBM

D →MM →D

D →DIBM[Home Base]

USA

NEC

Intel[Home Base]

SonyM →M

NEC[Home Base]

Japan

S →HS →H

H →SH →S IBM

Figure 2.1: Six kinds of knowledge flows

32

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 two-

way 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 patentUS Japan Germany France UK Canada

Within same country D→D 0.517** 0.535** 0.503** 0.526** 0.688** 1.406**

(0.013) (0.016) (0.042) (0.089) (0.141) (0.173)

D→M 0.491** 0.579** 0.941** 0.700** 0.281* 0.865**(0.037) (0.081) (0.114) (0.148) (0.109) (0.213)

M→D 0.371** 0.255* 0.461** 0.719** 0.670** 1.015**(0.032) (0.103) (0.082) (0.149) (0.143) (0.245)

M→M 0.695** 1.357** 0.633** 1.738** 0.934** 1.061**(0.120) (0.354) (0.235) (0.338) (0.167) (0.309)

Within same MNC

S→H 1.925** 1.771** 1.153** 1.357** 1.920** 2.383**(0.107) (0.212) (0.204) (0.192) (0.211) (0.292)

H→S 1.607** 2.097** 2.203** 1.964** 1.644** 2.177**(0.115) (0.251) (0.145) (0.120) (0.095) (0.100)

Country fixed effect - -0.384** -0.319** -0.248** -0.064 -0.022(0.014) (0.021) (0.018) (0.038) (0.028)

D→M / D→D 0.95 1.08 1.87** 1.33 0.41* 0.62* M→D / D→D 0.72** 0.48** 0.92 1.37 0.97 0.72 M→M / D→D 1.34 2.54* 1.26 3.30** 1.36 0.75

M→D / D→M 0.76** 0.44** 0.49** 1.03 2.38* 1.17 H→S / S→H 0.83* 1.18 1.91** 1.45** 0.86 0.91

A weighted logit regression is usedThe dependent variable is 1 if there is a citation between two patents, 0 otherwiseRobust standard errors in parentheses, with clustering on citing patentControls for technological similarity of citing and cited patent included in regression, but not shownFixed 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 patentChemical Computers &

CommunicatioDrugs & Medical

Electrical & Electronic

Mechanical Other

Within same country D→D 0.390** 0.650** 0.671** 0.438** 0.251** 0.826**

(0.029) (0.021) (0.068) (0.025) (0.028) (0.055)

D→M 0.401** 0.687** 0.645** 0.420** 0.151 0.587**(0.065) (0.056) (0.185) (0.082) (0.102) (0.112)

M→D 0.400** 0.390** 0.650** 0.100 0.169* 0.760**(0.063) (0.064) (0.103) (0.079) (0.073) (0.121)

M→M 0.492* 0.745** 1.633** 0.401 -0.124 1.749**(0.208) (0.184) (0.228) (0.358) (0.285) (0.239)

Within same MNC

S→H 1.861** 1.780** 2.270** 1.747** 2.504** 1.895**(0.231) (0.147) (0.406) (0.249) (0.252) (0.488)

H→S 1.875** 1.024** 2.351** 1.638** 2.052** 1.461*(0.212) (0.190) (0.336) (0.275) (0.290) (0.656)

Category fixed effect - 0.900** -0.725** 0.511** 0.612** -0.372**(0.027) (0.059) (0.029) (0.030) (0.048)

D→M / D→D 1.03 1.06 0.96 0.96 0.60 0.71** M→D / D→D 1.03 0.60** 0.97 0.23** 0.67 0.92 M→M / D→D 1.26 1.15 2.43** 0.92 -0.49 2.12**

M→D / D→M 1.00 0.57** 1.01 0.24** 1.12 1.29 H→S / S→H 1.01 0.58** 1.04 0.94 0.82 0.77

A weighted logit regression is usedThe dependent variable is 1 if there is a citation between two patents, 0 otherwiseRobust standard errors in parentheses, with clustering on citing patentControls for technological similarity of citing and cited patent included in regression, but not shownFixed 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 cross-

border citation between different firms. Second, the findings are based on a cross-

sectional 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 0.030**assignee in cited country (0.004)

[0.16]

Presence of cited assignee 0.011**in citing country (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 usedThe dependent variable is 1 if there is a citation between two patents, 0 otherwiseRobust standard errors in parentheses, with clustering on citing patentMarginal effects in square brackets after multiplication with 1,000,000Controls for technological similarity of citing and cited patent included in regression, but not shownFixed 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 Citing patents from US Citing patents not from USAll assignees Domestic MNC Domestic MNC Domestic MNC

(N=1,612) (N=810) (N=802) (N=436) (N=369) (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-the-

art 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)

1

1

')1(−

=

Λ−Λ∑

n

iiiii xx

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:

)(ln

'

1

1)1( i

i XX

ii

ii

ee β

αγβαγ

γαγγ

+

−

−

+

=+

=Λ−+Λ

Λ=Λ

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:

{ } { }∑∑∑=

−

==

+−=Λ−+Λ=n

i

xyi

yi

yiw

ii

ii

ewL1

)21(

01)1ln()1ln(1)ln(1ln β

αγ

)1)(/1()/1( iii yyw −+=

where αγ .

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,

),(*)(

//

)/)(,(*)(*

)(),(*)()|Pr(),( jxgjwnN

NNnnjxg

jfjfjxgjfjyxjxg

j

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

dxyxgyPywLE i

n

iiiw ),()]()[ln(ln

1∫ ∑

=

=

∑ ∫=

=

n

ii

iii dxyxg

ywnNyPyw

1),(*

)(/)]()[ln(

dxyxgyPnN

i

n

ii ),(*)]([ln

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()|Pr()|Pr( iiiiii xxandzzjyxzzxxjyandzz ========

)|Pr()|Pr( iii xxjyxzz ====

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

iiii xyyandzzL

1)|Pr(lnln

[ ] ( )[ ]{ }∑=

Λ−=−+Λ==n

iiiiiiiii xxzzyxxzzy

1)(1)|Pr(ln)1()()|Pr(ln ββ

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

[ ] ( )[ ]{ }∑=

Λ−=−+Λ=n

iiiiziiiizi xxzzwyxxzzwy

ii1

01 )(1)|Pr(ln)1()()|Pr(ln ββ

( )∑=

−+−=n

i

Xyi

iewC1

)21(1ln β

[ ]∑=

==−+=n

iiiizizii xzzwCwywyw

ii1

01 )|Pr(ln and)1(where

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 co-

location 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 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 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

Close collaborative links between indivduals

Greater probability of knowledge flow

Figure 3.1(b): Hypotheses 3 and 4

Same regionClose collaborative links between indivduals

Greater probability of knowledge flow

Same firm

Figure 3.1(c): Hypotheses 5 and 6

55

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 0B 1 0 0 1 0 0 0C 0 1 1 0 0 0 0D 0 0 1 0 1 0 0E 0 0 0 0 1 0 1F 0 0 0 0 1 0 0G 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)

Inventors: A, CA

Patent P1 (1986)

Inventors: A, B Figure 3.3(a): Social proximity graph for 1987

Patent P2 (1987)

Inventors: A, CA

Patent P1 (1986)

Inventors: A, B

Patent P3 (1988)

Inventors: C, D

C

Figure 3.3(b): Social proximity graph for 1988

Patent P2 (1987)

Inventors: A, CA

Patent P1 (1986)

Inventors: A, B

Patent P3 (1988)

Inventors: C, D

C

Patent P4 (1989)

Inventor: B

Patent P5 (1989)

Inventors: D, E, F

Patent P6 (1989)

Inventor: G

B

D

Figure 3.3(c): Social proximity graph for 1989

Patent P2 (1987)

Inventors: A, CA

Patent P1 (1986)

Inventors: A, B

Patent P3 (1988)

Inventors: C, D

C

Patent P4 (1989)

Inventor: B

Patent P5 (1989)

Inventors: D, E, F

Patent P6 (1989)

Inventor: G

B

D

Patent P7 (1990)

Inventor: E, G

E

G

Figure 3.3(d): Social proximity graph for 1990

69

Sour

ce T

eam

Destination TeamP1 P2 P3 P4 P5 P6 P7

P1 . 0 1 0 2 3P2 . . 0 1 1 2P3 . . . 2 0 1P4 . . . . 3 4P5 . . . 3 . 0P6 . . . . 0P7 . . . . . . .

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 non-zero 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 3-

digit technological controls, though extensively used in existing literature, are

insufficient. To address this, column (3) uses additional controls based on a detailed 9-

digit 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) (2) (3)Within same region 1.413** 1.050** 0.798**

(0.051) (0.017) (0.020)[16.96] [12.60] [9.58]

Within same firm 3.781** 2.622** 2.217**(0.060) (0.022) (0.025)[45.37] [31.46] [26.60]

Technological relatedness: Same tech category 1.176** 1.173**

(0.026) (0.023)

Same tech subcategory 1.161** 1.105**(0.029) (0.029)

Same primary tech class 2.637** 1.545**(0.023) (0.030)

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 usedThe dependent variable is 1 if there is a citation between two patents, 0 otherwiseRobust standard errors in parentheses, with clustering on citing patentMarginal effects in square brackets after multiplication with 1,000,000Fixed 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

20 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.

75

Table 3.3: Summary statistics

Entire sample No self-citations by firmsCitations Controls Citations Controls(N=552,427) (N=1,976,337) (N=349,251) (N=1,881,299)

Common inventor 0.1512 0.0033 0.0132 0.0001(Social distance = 0)

Past collaboration 0.0593 0.0036 0.0079 0.0004(Social distance = 1)

Common past collaborator 0.0343 0.0052 0.0085 0.0011(Social distance = 2)

Indirect social link 0.4024 0.4767 0.5133 0.4775(Social distance > 2 but finite)

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) (2)Common inventor 8.820** 4.002**(Social distance = 0) (0.078) (0.060)

[105.84] [48.02]

Past collaboration 6.741** 2.859**(Social distance = 1) (0.162) (0.055)

[80.89] [34.31]

Common past collaborator 5.210** 2.228**(Social distance = 2) (0.089) (0.054)

[62.52] [26.74]

Indirect social link 0.212** 0.151**(Social distance > 2 but finite) (0.019) (0.012)

[2.54] [1.81]

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 usedThe dependent variable is 1 if there is a citation between two patents, 0 otherwiseRobust standard errors in parentheses, with clustering on citing patentMarginal effects in square brackets after multiplication with 1,000,000Fixed 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) (2) (3)Within same region 0.798** 0.603** 0.697**

(0.020) (0.022) (0.033)[9.58] [7.24] [8.36]

Within same firm 2.217** 1.809** 2.079**(0.025) (0.027) (0.049)[26.60] [21.71] [24.95]

Common inventor 2.096** 4.509**(Social distance = 0) (0.065) (0.245)

Past collaboration 1.017** 2.998**(Social distance = 1) (0.062) (0.177)

Common past collaborator 0.469** 2.382**(Social distance = 2) (0.065) (0.101)

Indirect social link 0.098** 0.147**(Social distance > 2 but finite) (0.013) (0.013)

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 2,528,764

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

80

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) n/a n/a 6.7 9.2 7.1 4.6 South Korea 8.2 7.2 8.1 10.0 7.5 3.1 Hong Kong 6.7 12.0 5.7 7.6 5.3 1.4 Singapore 9.6 8.5 6.3 8.5 9.2 4.3 Emerging Asian Economies

India 3.2 5.4 5.4 6.4 5.2 5.0 China 5.2 5.5 10.8 7.7 12.1 6.7 Indonesia 7.8 7.9 5.7 7.1 7.8 0.0 Malaysia 7.2 8.6 5.2 6.9 9.5 3.1 Thailand 5.8 8.0 5.4 10.3 8.6 -0.3 Emerging Latin American Economies

Mexico Brazil Argentina Chile

6.3 10.3 3.1

-1.1

7.1 6.7 3.0 7.3

2.0 1.2

-2.4 1.1

1.7 2.1

-0.3 6.8

1.6 3.2 6.7 8.7

4.1 1.3 2.9 3.4

Venezuela

3.0 2.5 -0.9 2.8 3.5 -0.2

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

Recipient Countries

1970-74 1975-79 1980-84 1985-89 1990-94 1995-99

Newly Industrialized Economies

Taiwan (ROC) 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

Emerging Latin American Economies

Mexico Brazil Argentina Chile

243

86

126

22

246

100

113

20

191

110

100

12

202

156

82

18

189

260

109

32

257

353

183

44

Venezuela

36

35

50

103

121

145

Total Worldwide

367,943

322, 385

309, 387

398,816

484,223

623,999

93

Table 4.2(b): Country exports: 1970-99

Recipient Countries

1970-74 1975-79 1980-84 1985-89 1990-94 1995-99

Newly Industrialized Economies

Taiwan (ROC) n/a n/a 256.7 (52.9%)

399.8 (54.4%)

477.3 (45.0%)

697.1 (48.0%)

South Korea 87.6 (21.5%)

181.0 (28.9)

289.0 (34.3%)

470.6 (35.7%)

549.5 (27.9%)

983.4 (37.3%)

Hong Kong 125.0 (88.9%)

184.0 (87.5%)

311.0 (95.6%)

561.0 (123.0%)

838.0 (139.3%)

1002.8 (137.0%)

Singapore 87.1 (121.9%)

178.3 (171.6%)

304.3 (192.9%)

390.4 (187.6)

595.1 (185.8)

823.6 (174.5)

Emerging Asian Economies

India 23.6 (4.0%)

45.8 (6.4%)

52.5 (6.0%)

70.9 (6.2%)

136.0 (9.2%)

226.0 (11.3%)

China 15.7 (2.9%)

32.8 (4.8%)

84.8 (8.6%)

204.0 (12.4%)

506.0 (20.1%)

932.0 (22.4%)

Indonesia 41.1 (20.0%)

75.8 (25.6%)

118.0 (28.0%)

127.4 (22.9%)

215.2 (26.5%)

344.1 (32.9%)

Malaysia 35.1 (40.0%)

61.7 (49.0%)

94.4 (54.4%)

140.4 (62.6%)

272.0 (79.8%)

507.9 (103.3%)

Thailand 27.3 (18.1%)

43.6 (20.3%)

65.8 (22.5%)

121.5 (29.7%)

243.0 (36.9%)

409.8 (48.7%)

Emerging Latin American Economies

Mexico Brazil Argentina Chile

53.1 (8.1%) 108.9

(7.5%) 59.2

(6.7%) 16.3

(13.8%)

84.4 (9.6%) 151.9

(7.1%) 77.6

(7.9%) 27.2

(22.9%)

171.0 (14.6%)

254.9 (10.2%)

77.7 (7.5%)

30.8 (21.2%)

221.3 (18.2%)

295.1 (9.9%) 101.4

(10.0%) 56.7

(31.9%)

234.1 (16.4%)

300.0 (9.6%)

89.0 (7.7%)

78.9 (30.8%)

493.0 (13.9%)

297.6 (8.1%) 145.8

(10.2%) 104.5

(28.6%) Venezuela

62.6 (25.7%)

73.5 (24.7%)

73.3 (25.0%)

75.1 (24.2%)

111.8 (30.7%)

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) n/a n/a 1.55 4.43 11.04 17.73 South Korea 0.27 0.24 0.31 0.90 5.26 11.56 Hong Kong 0.47 0.42 0.36 0.32 0.33 0.57 Singapore 0.24 0.05 0.07 0.12 0.25 0.61 Emerging Asian Economies

India 3.52 1.46 0.76 0.90 0.93 1.40 China 3.89 0.06 0.08 0.63 0.47 0.36 Indonesia 0.46 0.07 0.04 0.08 0.12 0.05 Malaysia 0.11 0.05 0.07 0.08 0.06 0.11 Thailand 0.15 0.07 0.11 0.09 0.06 0.14 Emerging Latin American Economies

Mexico Brazil Argentina Chile

4.58 0.79 2.13 1.35

2.91 0.66 1.46 0.74

1.12 0.43 1.29 0.39

0.91 0.53 0.81 0.32

0.81 0.87 1.22 0.41

0.52 1.19 1.26 0.42

Venezuela 0.58 0.48 0.68 1.37 1.08 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.,

∑ ∑∑∑≡i j ijnijn

i ijnijnijRTAj

///

where 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

nij

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

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

98

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

1980-84 1985-89 Top 11 (Fast- Growing)

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

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 Middle 11 (Medium- Growing)

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

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

Bottom 11 (Slow- Growing)

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

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)

1990-94 1995-1999 Top 11 (Fast- Growing)

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

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

Middle 11 (Medium- Growing)

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

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 Bottom 11 (Slow- Growing)

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

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

100

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

∑

−=

j wjpwjpijpi /22χ

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 Chi-

square 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)

(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)

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)

(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)

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)

(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)

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=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

Hong Kong

Singapore

1980-84 (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)

(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)

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)

(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)

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=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)

1995-99 (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)

(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

India China

1980-84 (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=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)

1985-89

(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=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)

1990-94 (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=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)

1995-99 (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=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 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 for-

profit 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 Period Multinationals Business groups Individuals Domestic firms & orgsTaiwan 1970-79 2.9% 0.0% 87.7% 9.4%

1980-89 1.9% 0.5% 87.0% 10.6%1990-99 1.9% 3.5% 59.0% 35.6%

Korea 1970-79 14.7% 2.9% 69.1% 13.2%1980-89 2.5% 31.4% 47.3% 18.8%1990-99 0.8% 80.7% 6.8% 11.7%

Hong Kong 1970-79 26.1% - 45.5% 28.4%1980-89 17.3% - 31.5% 51.2%1990-99 16.6% - 30.7% 52.7%

Singapore 1970-79 50.0% - 43.3% 6.7%1980-89 19.7% - 47.0% 33.3%1990-99 45.7% - 9.6% 44.7%

India 1970-79 54.5% 0.6% 24.7% 20.1%1980-89 48.1% 6.5% 22.2% 23.1%1990-99 29.6% 11.1% 18.3% 41.0%

China 1970-79 14.5% - 76.8% 8.7%1980-89 14.4% - 39.6% 46.0%1990-99 17.2% - 40.1% 42.7%

113

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 Affiliation Patent CountIndustrial Technology Research Inst. Domestic Firm Or Org 1,229United Microelectronics Corporation Domestic Firm Or Org 946Taiwan Semiconductor Manufacturing Co. Domestic Firm Or Org 752National Science Council Domestic Firm Or Org 367Vanguard International Semiconductor Domestic Firm Or Org 301Winbond Electronics Corp. Walsin Lihua Group 216Hon Hai Precision Ind. Co., Ltd. Domestic Firm Or Org 107Mosel Vitelic, Incorporated Pacific Electric Wire & C 85Acer Peripherals, Inc. Acer Group 70Texas Instruments Inc Foreign Multinational 60Acer Incorporated Acer Group 56Macronix International Co., Ltd. Domestic Firm Or Org 55Holtek Microelectronics Inc. Domestic Firm Or Org 48Mustek Systems, Inc. Domestic Firm Or Org 47Umax Data Systems Inc. Umax Group 47Silitek Corporation Liton Enterprise Group 44Primax Electronics Ltd. Domestic Firm Or Org 40United Semiconductor Corp. Domestic Firm Or Org 36Greenmaster Industrial Corp. Domestic Firm Or Org 31Etron Technology, Inc. Domestic Firm Or Org 29Powerchip Semiconductor Corp. Umax Group 28Tong Lung Metal Industry Co., Ltd. Domestic Firm Or Org 27Behavior Tech Computer Corp. Domestic Firm Or Org 26E. Lead Electronic Co., Ltd. Domestic Firm Or Org 25Delta Electronics Inc. Domestic Firm Or Org 24Development Center For Biotechnology Domestic Firm Or Org 22Hwa Shin Musical Instrument Co., Ltd. Domestic Firm Or Org 22Enlight Corporation Domestic Firm Or Org 21Inventec Corporation Domestic Firm Or Org 21Fu Tai Umbrella Works, Ltd. Domestic Firm Or Org 20Shin Jiuh Corp. Domestic Firm Or Org 19Taiwan Fu Hsing Industrial Co., Ltd. Domestic Firm Or Org 19Duracraft Corporation Foreign Multinational 18Shin Yeh Enterprise Co., Ltd. Domestic Firm Or Org 17Quarton, Inc. Domestic Firm Or Org 17China Textile Institute Domestic Firm Or Org 17Must Systems, Inc. Domestic Firm Or Org 16Chung Cheng Faucet Co. Ltd. Domestic Firm Or Org 16Chicony Electronics Co., Ltd. Domestic Firm Or Org 15Institute Of Nuclear Energy Research Domestic Firm Or Org 15Kalloy Industrial Co., Ltd. Domestic Firm Or Org 15Compal Electronics, Inc. Domestic Firm Or Org 15China Steel Corporation Domestic Firm Or Org 13Pan-International Industrial Corporati Domestic Firm Or Org 13Food Industry Research And Development Domestic Firm Or Org 13Teh Yor Industrial Co., Ltd. Domestic Firm Or Org 12Silicon Integrated Systems Corp. Domestic Firm Or Org 12Formosa Saint Jose Corporation Domestic Firm Or Org 12Yuan Mei Corp. Domestic Firm Or Org 12Foxconn International, Inc. Foreign Multinational 12Total patents for top 50 assignees 5,100Other patents 14,883Overall total 1970-99 for Taiwan 19,983Fraction of patents held by top 50 assignees 25.5%

116

Table 4.7(b): Top 50 patent winners for Korea (1970-1999) Assignee Name Affiliation Patent CountSamsung Electronics Co., Ltd. Samsung Group 5,350Daewoo Electronics Company, Ltd. Daewoo Group 1,008Hyundai Electronics Industries Co., Ltd. Hyundai Group 931Goldstar Company, Ltd. LG Group 892LG Semicon Co., Ltd. LG Group 696LG Electronics Inc. LG Group 566Electronics And Telecommunications Res. Domestic Firm or Org 397Hyundai Motor Co., Ltd. Hyundai Group 347Gold Star Electron Co., Ltd. LG Group 252Samsung Display Devices Co., Ltd. Samsung Group 243Korea Institute Of Science And Tech. Domestic Firm or Org 238Samsung Electron Devices Co., Ltd. Samsung Group 214Samsung Aerospace Industries, Ltd. Samsung Group 131Samsung Electro-Mechanics Co., Ltd. Samsung Group 124Korea Advanced Institute Of Science Domestic Firm or Org 105Korea Research Institute Of Chem. Tech. Domestic Firm or Org 100Korea Telecommunication Authority Domestic Firm or Org 96Samsung Heavy Industries, Co., Ltd. Samsung Group 71Lucky Ltd. LG Group 68LG Industrial Systems Co., Ltd. LG Group 65Kia Motors Corp. Kia Group 62SKC Limited Sunkyong Group 51Daewoo Telecom Co., Ltd. Daewoo Group 42Daewoo Heavy Industries Co., Ltd Daewoo Group 36Pohang Iron & Steel Co., Ltd. POSCO Group 35Mando Machinery Corp. Ltd. Halla Group 30Korea Atomic Energy Research Institute Domestic Firm or Org 29Agency For Defence Development Domestic Firm or Org 27LG Chemical Ltd. LG Group 25Korea Kumho Petrochemical Co., Ltd. Kumho Group 25Kwangju Electronics Co., Ltd. Samsung Group 24Samsung Semiconductor & Telecom. Samsung Group 23Kolon Industries Inc. Kolon Group 23Sindo Ricoh Co., Ltd. Domestic Firm or Org 22Toray Industries Inc. Foreign Multinational or Org 20Samsung Heavy Industry Co., Ltd. Samsung Group 19Yukong Limited Sunkyong Group 19Orion Electric Co., Ltd. Daewoo Group 18Anam Industrial Co., Ltd. Anam Group 17Sunkyong Industries Co., Ltd. Sunkyong Group 16Cheil Industries, Inc. Samsung Group 16Pacific Corporation Pacific Group 16Cheil Foods & Chemicals, Inc. Domestic Firm or Org 14Dong Kook Pharmaceutical Co., Ltd. Domestic Firm or Org 13Anam Semiconductor, Inc. Anam Group 13Medison Co., Ltd. Domestic Firm or Org 12Volvo Construction Equipment Korea Co. Domestic Firm or Org 12Korea Chemical Co., Ltd. Domestic Firm or Org 11Samsung Corning Co., Ltd. Samsung Group 11Korea Institute Of Machinery & Metals Domestic Firm or Org 10Total patents for top 50 assignees 12,585Other patents 2,253Overall total 1970-99 for Korea 14,838Fraction of patents held by top 50 assignees 84.8%

117

Table 4.7(c): Top 50 patent winners for Hong Kong (1970-1999) Assignee Name Affiliation Patent CountAstec International, Ltd. Domestic Firm or Org 44Johnson Electric S.A. Domestic Firm or Org 33Johnson Electric Industrial Manuf. Domestic Firm or Org 24Motorola Inc Foreign Multinational or Org 23W. Haking Enterprises Limited Domestic Firm or Org 20The Hong Kong University Of Science & Tech. Domestic Firm or Org 15World-Wide Stationery Manufacturing Co Domestic Firm or Org 14China Pacific Trade Ltd. Domestic Firm or Org 12Chiaphua Industries, Ltd. Domestic Firm or Org 12Playart Limited Domestic Firm or Org 11Polycity Industrial Ltd. Domestic Firm or Org 10Arco Industries Ltd. Foreign Multinational or Org 10Solar Wide Industrial Limited Domestic Firm or Org 8T. K. Wong & Associates Limited Domestic Firm or Org 7Pentalpha Enterprises Ltd. Domestic Firm or Org 7Leco Stationery Manufacturing Co., Ltd Domestic Firm or Org 7Outboard Marine Corp Foreign Multinational or Org 7John Manufacturing Limited Domestic Firm or Org 6Mego Corp. Foreign Multinational or Org 6Mr. Christmas, Incorporated Foreign Multinational or Org 6Asm Assembly Automation Ltd. Domestic Firm or Org 5The Chinese University Of Hong Kong Domestic Firm or Org 5The Hong Kong Polytechnic University Domestic Firm or Org 5Wing Shing Products (Bvi) Co. Ltd. Domestic Firm or Org 5Alza Corp Foreign Multinational or Org 5Computer Products Inc Foreign Multinational or Org 5Windmere Corp Foreign Multinational or Org 5Achiever Industries Limited Domestic Firm or Org 4G. E. W. Corporation Limited Domestic Firm or Org 4International Quartz Ltd. Domestic Firm or Org 4Meyer Manufacturing Company Limited Domestic Firm or Org 4Payview Limited Domestic Firm or Org 4Tradebest International Corporation Domestic Firm or Org 4United Chinese Plastics Products Co. Domestic Firm or Org 4Pacusma Co. Ltd. Domestic Firm or Org 4East Asia Services Ltd. Domestic Firm or Org 4Addway Engineering Limited Domestic Firm or Org 4Conair Corp Foreign Multinational or Org 4General Electric Company Foreign Multinational or Org 4Polaroid Corp Foreign Multinational or Org 4Recoton Corp Foreign Multinational or Org 4Tiger Electronics, Inc. Foreign Multinational or Org 4Timex Corporation Foreign Multinational or Org 4Concord Camera Corp. Foreign Multinational or Org 4Heep Tung Manufactory Limited Domestic Firm or Org 3Kwoon Kwen Metal Ware Company Limited Domestic Firm or Org 3Maxpat Trading & Marketing Domestic Firm or Org 3Refined Industry Company Limited Domestic Firm or Org 3Simatelex Manufactory Company Limited Domestic Firm or Org 3Sonca Industries Limited Domestic Firm or Org 3Total patents for top 50 assignees 403Other patents 870Overall total 1970-99 for Hong Kong 1,273Fraction of patents held by top 50 assignees 31.7%

118

Table 4.7(d): Top 50 patent winners for Singapore (1970-1999) Assignee Name Affiliation Patent CountChartered Semiconductor Manufacturing Domestic Firm or Org 122Hewlett-Packard Co Foreign Multinational or Org 43National University Of Singapore Domestic Firm or Org 35Texas Instruments Inc Foreign Multinational or Org 35Motorola Inc Foreign Multinational or Org 28Thomson SA Foreign Multinational or Org 23Molex Inc Foreign Multinational or Org 23Tritech Microelectronics International Domestic Firm or Org 21Matsushita Electric Industrial Co Ltd Foreign Multinational or Org 18Philips Foreign Multinational or Org 11SGS-Thomson Microelectronics (Pte) Ltd Domestic Firm or Org 9Sun Industrial Coatings Private Ltd. Domestic Firm or Org 8Tritech Microelectronics, Ltd. Domestic Firm or Org 8Chartered Industries Of Singapore Priv Domestic Firm or Org 7Institute Of Microelectronics Domestic Firm or Org 7Nestec, S.A. Foreign Multinational or Org 6Berg Technology, Inc. Foreign Multinational or Org 6Seagate Technology Foreign Multinational or Org 6Siemens Aktiengesellschaft Foreign Multinational or Org 5Eastern Oil Tools Pte, Ltd. Domestic Firm or Org 5Singapore Computer Systems Limited Domestic Firm or Org 5Institute Of Microelectronics Domestic Firm or Org 5Sunright Limited Domestic Firm or Org 5Advanced Systems Automation Limited Domestic Firm or Org 5Apple Computer Inc Foreign Multinational or Org 5Du Pont Foreign Multinational or Org 5Advanced Materials Technologies Pte Lt Domestic Firm or Org 4Enteron, L.P. Domestic Firm or Org 4United Technologies Corp Foreign Multinational or Org 4Whitaker Corporation Foreign Multinational or Org 4Creative Technology Limited Domestic Firm or Org 4Varta Batterie A.G. Foreign Multinational or Org 3Sumitomo Chemical Company, Limited Foreign Multinational or Org 3Nortrans Shipping And Trading Far East Domestic Firm or Org 3Abb Vetcogray Inc. Foreign Multinational or Org 3Litton Industries Foreign Multinational or Org 3Black & Decker Corp Foreign Multinational or Org 3Chevron Foreign Multinational or Org 3Rmt, Inc. Foreign Multinational or Org 3Thomas & Betts Corp Foreign Multinational or Org 3Symtonic Sa Foreign Multinational or Org 2Rhone Poulenc Industries Foreign Multinational or Org 2Hitachi Chemical Company, Ltd. Foreign Multinational or Org 2Toshiba Corporation Foreign Multinational or Org 2Sandvik Foreign Multinational or Org 2Multiscience System Pte. Ltd. Domestic Firm or Org 2Port Of Singapore Authority Domestic Firm or Org 2Singapore Institute Of Standards And I Domestic Firm or Org 2Aztech Systems Ltd. Domestic Firm or Org 2Matsushita Refrigeration Industries Foreign Multinational or Org 2Total patents for top 50 assignees 523Other patents 221Overall total 1970-99 for Singapore 744Fraction of patents held by top 50 assignees 70.3%

119

Table 4.7(e): Top 50 patent winners for India (1970-1999) Assignee Name Affiliation Patent CountCouncil Of Scientific And Industrial Research Domestic Firm or Org 141Hoechst Foreign Multinational or Org 45Ciba-Geigy Corporation Foreign Multinational or Org 38Ranbaxy Laboratories Limited Ranbaxy Group 20Unilever Foreign Multinational or Org 19NASA Foreign Multinational or Org 18Texas Instruments Inc Foreign Multinational or Org 17Dr. Reddy'S Research Foundation Dr. Reddy's Group 10Lupin Laboratories Limited Lupin Group 9Indian Explosives Ltd. Domestic Firm or Org 8General Electric Company Foreign Multinational or Org 8National Institute Of Immunology Domestic Firm or Org 7Monsanto Co. Foreign Multinational or Org 7Panacea Biotec Limited Domestic Firm or Org 6Iowa India Investments Company Limited Domestic Firm or Org 4Indian Oil Corporation, Ltd. Domestic Firm or Org 4Union Carbide Corp Foreign Multinational or Org 4Elf Aquitaine Foreign Multinational or Org 3Cadbury India Limited Domestic Firm or Org 3Indian Petrochemicals Corporation Ltd. Domestic Firm or Org 3Gem Energy Industry Limited Domestic Firm or Org 3Aktiebolaget Astra Foreign Multinational or Org 3Procter & Gamble Foreign Multinational or Org 3Fiberstars, Inc. Foreign Multinational or Org 3Xerox Corp Foreign Multinational or Org 3Novartis (Sandoz) Foreign Multinational or Org 2Forschungszentrum Julich Gmbh Foreign Multinational or Org 2Licentia Patent-Verwaltungs-Gmbh Foreign Multinational or Org 2Boots Company Plc Foreign Multinational or Org 2Imperial Chemical Industries Foreign Multinational or Org 2Zeneca Limited Foreign Multinational or Org 2All India Institute Of Medical Science Domestic Firm or Org 2Hawkins Cookers Limited Domestic Firm or Org 2Iel Limited Domestic Firm or Org 2Indian Space Research Organisation Domestic Firm or Org 2Karamchand Premchand Private Limited Domestic Firm or Org 2Sree Chitra Tirunal Inst. For Medical Domestic Firm or Org 2National Chemical Laboratory Domestic Firm or Org 2The Chief Controller, Research And Dev Domestic Firm or Org 2GEC Foreign Multinational or Org 2Westinghouse Electric Corp Foreign Multinational or Org 2American Cyanamid Co Foreign Multinational or Org 2Analog Devices Foreign Multinational or Org 2Avnet Inc Foreign Multinational or Org 2Johnson & Johnson Foreign Multinational or Org 2Mobil Foreign Multinational or Org 2Sri International Foreign Multinational or Org 2United States Of America, Air Force Foreign Multinational or Org 2University Of California Foreign Multinational or Org 2University Of Minnesota Foreign Multinational or Org 2Total patents for top 50 assignees 439Other patents 257Overall total 1970-99 for India 696Fraction of patents held by top 50 assignees 63.1%

120

Table 4.7(f): Top 50 patent winners for China (1970-1999) Assignee Name Affiliation Patent CountChina Petrochemical Development Corp. Domestic Firm or Org 26United Microelectronics Corporation Foreign Multinational or Org 21Tsinghua University Domestic Firm or Org 10Autry Industries, Inc. Foreign Multinational or Org 8Industrial Technology Research Inst. Foreign Multinational or Org 7China Petrochemical Corporation Domestic Firm or Org 5Fujian Institute Of Research Domestic Firm or Org 4North China Research Institute Of Elec. Domestic Firm or Org 4Peking University Domestic Firm or Org 4Shanghai Institute Of Biochemistry Domestic Firm or Org 4Taiho Pharmaceutical Company Limited Foreign Multinational or Org 4Acer Incorporated Foreign Multinational or Org 4Beijing Research Institute Of Chem. Domestic Firm or Org 3Chinese Academy Of Medical Sciences Domestic Firm or Org 3Huazhong Institute Of Technology Domestic Firm or Org 3Institute Of Physics, Chinese Academy Domestic Firm or Org 3Shanghai Institute Of Organic Chemistry Domestic Firm or Org 3Tianjin University Domestic Firm or Org 3CSL Opto-Electronics Corp. Domestic Firm or Org 3Nan Kai University Domestic Firm or Org 3Central Iron & Steel Research Inst. Domestic Firm or Org 3Bayer Foreign Multinational or Org 3Leco Stationery Manufacturing Co., Ltd Foreign Multinational or Org 3Beijing Polytechnic University Domestic Firm or Org 2China Metallurgical Import & Export Co. Domestic Firm or Org 2China National Seed Corporation Domestic Firm or Org 2Jilin University Of Technology Domestic Firm or Org 2Luoyang Petrochemical Engineering Corp Domestic Firm or Org 2Qing-Yang Machine Works Domestic Firm or Org 2Research Institute Of Petroleum Proces Domestic Firm or Org 2Science & Technic Department Of Dagang Domestic Firm or Org 2Shanghai Lamp Factory Domestic Firm or Org 2Institute Of Materia Medica Domestic Firm or Org 2Chinese Building Technology Services Domestic Firm or Org 2University Of Electronic Science And Tech. Domestic Firm or Org 2South China University Of Technology Domestic Firm or Org 2Research Institute Of Petroleum Proc. Domestic Firm or Org 2Traditional Chinese Medicine Research Domestic Firm or Org 2Dalian Institute Of Chemical Physics Domestic Firm or Org 2University Of Science And Technology Domestic Firm or Org 2Shanghai Yue Long Nonferrous Metals Ltd. Domestic Firm or Org 2Vasomedical, Inc. Domestic Firm or Org 2Panzhihua Iron And Steel (Group) Co. Domestic Firm or Org 2Wonder & Bioenergy Hi-Tech International Domestic Firm or Org 2Pacific Sources, Inc. Domestic Firm or Org 2Fushun Research Institute Of Petroleum Domestic Firm or Org 2Plastic Advanced Recycling Corp. Domestic Firm or Org 2Institute Of Materia Medica, An Inst. Domestic Firm or Org 2Liaohe Petroleum Exploration Bureau Domestic Firm or Org 2Jiangsu Goodbaby Group, Inc. Domestic Firm or Org 2Total patents for top 50 assignees 188Other patents 582Overall total 1970-99 for China 770Fraction of patents held by top 50 assignees 24.4%

121

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 host-country 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 foreign-affiliated 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 inter-firm 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): 1546-88.

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

134

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.