Copyright page:

© 2013

Xiaoyu Pu

ALL RIGHTS RESERVED

MNC SUBUNIT KNOWLEDGE SOURCING AND COMPETENCE CREATING

ACTIVITIES – A DYNAMIC VIEW OF SUBUNIT EVOLUTION

By

XIAOYU PU

A Dissertation submitted to the

Graduate School-Newark

Rutgers, The State University of New Jersey

in partial fulfillment of the requirements

for the degree of

Doctor of Management

Graduate Program in

Rutgers Business School

written under the direction of

Professor Dr. John A. Cantwell

and approved by

_________________________

_________________________

_________________________

_________________________

Newark, New Jersey

May, 2013

ii

ABSTRACT OF THE DISSERTATION

MNC Subunit Knowledge Sourcing and Competence Creating Activities – A Dynamic

View of Subunit Evolution

By XIAOYU PU

Dissertation Director:

Professor Dr. John Cantwell

The innovative activities of multinational corporation (MNC) operations overseas can be

represented as two types: either competence exploiting (CE) – exploiting the core

competence base of the parent group – or competence creating (CC) – creating new

competencies that were not already among the strengths of the relevant parent company.

To a large extent, the share of these two types of activities determines and reflects a given

subunit’s strategic role within its MNC. This research examines (1) the patterns of MNC

subunits’ knowledge sourcing in terms of the technological and geographical dispersion

of knowledge sources; (2) the extent to which MNC subunits’ technological fields of

expertise are distinct from those of their parent companies, and how this technological

distinctiveness is related to their knowledge sourcing patterns; and (3) how MNC

subunits’ profiles of CC and CE activities (in terms of their overall technological distance

from their parent companies, and the degree to which they are engaged in CC versus CE

activities) evolve over time, reflecting the evolution of their knowledge creating role and

status within their international group. Attention is focused on the heterogeneity of firm-

specific evolutionary paths in the patterns of knowledge accumulation that support CC

iii

activities, controlling for the industry-specific determinants, location-specific factors, and

MNC group structural influences on such technological trajectories.

This study proposes a dynamic model in which the extent to which a subunit is likely to

take up CC activities is influenced by the technological and geographical dispersion of

that subunit’s knowledge sourcing. The results show a consistently positive relationship

between the technological and geographical dispersions of knowledge sourcing, an effect

that is moderated by the extent of subunit specialization in general purpose technology

(GPT) fields, and the geographical proximity between dispersed knowledge sources. We

also find a positive relationship between the technological dispersion of knowledge

sourcing and the technological distinctiveness of subunits. However, the geographical

dispersion of knowledge sources has a negative relationship with subunit technological

distinctiveness. A typology of subunit strategic roles is proposed, based upon the

evolutionary trajectory of a subunit’s share of CC activities and its technological distance

from its parent company.

iv

Preface Acknowledgements

There is no easy word to describe this long journey of PhD studies. In this most important

part of my life, I experienced the ups and downs just like most of my fellow PhD

colleagues – the excitement of pursuing knowledge, the humbleness of encountering

great minds, the stress and struggling of facing challenges, and the indescribable feeling

when six years worth of work is finally ready to be presented. This was indeed the most

fruitful six years, during which I achieved many goals I had set up in the beginning, and

none of these achievements would have come true without the tremendous help, support,

and encouragement from many others who influenced me in one way or another.

I would first like to show the deepest appreciation to my committee chair Dr. John

Cantwell. I studied with Dr. Cantwell since the first year in my PhD program at Rutgers

Business School. His guidance in my pursuit of scholarship, his insights on issues I

encounter in research, as well as his patient and persistent assistant in every aspect of my

study, made this dissertation possible. Even during this last year, when I have started my

job as an assistant professor with a full teaching schedule while working on the

dissertation at the same time, Dr. Cantwell offered to meet and give me guidance on

weekends and even holidays, for which I cannot thank him enough.

My committee member Dr. Farok Contractor gave me insights on the theorization of this

dissertation, with his comments I was able to improve my overall research framework; Dr.

Michelle Gittelman challenged my data construction and research methodology, with her

insights I developed further understanding on patent data; Dr. Lucia Piscitello encouraged

v

me to improve theory building, and to examine further relationship among key variables.

These are just a snapshot of a long list of valuable comments and suggestions my

committee provided to me, for which I greatly thank them.

Aside from my dissertation committee, there are many people helped me along this

journey. Dr. George Farris was one of the first who guided me when I was working as his

teaching assistant. The Technology Management Research Center he directed provided

me funding for research and travels to conferences. I also received generous funding from

Eastern Asia Research Center directed by Dr. Cantwell, as well as Department of

Management and Global Business at Rutgers Business School. Dr. Nancy DiTomaso who

chaired our MGB department, and our secretary Dawn Gist, had always offered their kind

assistance during my years in the program.

In the meanwhile, I need to thank the countless numbers of colleagues and friends, whose

friendship and kindness have made my journey a pleasant one. Among all others, Anke

Pippenbrink, Robert McNamee, Ranfeng Qiu, Jun Li, Hao-hsuan Chiu, and Shengsheng

Huang had offered me incredible help and support.

Finally, I couldn’t have done this without the support from my loving family. My

wonderful husband Ritchie Kim gave me unconditional love and support in all possible

ways. Thank you for staying by my side, your love keeps me going. Mom and dad, thank

you for your support for all these years, hope I have made you proud!

vi

TABLE OF CONTENTS

ABSTRACT OF THE DISSERTATION ........................................................................................ ii

Preface ............................................................................................................................................ iv

TABLE OF CONTENTS ................................................................................................................ vi

LIST OF TABLES ........................................................................................................................ viii

LIST OF FIGURES ........................................................................................................................ ix

1. Introduction ............................................................................................................................. 1

2. Data Overview ........................................................................................................................ 11

2.1 Data Description ............................................................................................................ 11

2.2 Key concepts .................................................................................................................. 19

2.3 Sample Data from Dow Chemical, DuPont, and Novartis .............................................. 26

3. MNC Subunit Knowledge Sourcing ........................................................................................ 35

3.1 Knowledge based view of MNC ..................................................................................... 35

3.2 Local and Remote Knowledge Sourcing ......................................................................... 39

3.3 Knowledge Sourcing Patterns and Subunit Innovation ................................................. 43

3.4 Framework ..................................................................................................................... 51

3.5 Methodology .................................................................................................................. 51

3.6 Results and Discussion ................................................................................................... 53

4. Competence Exploiting and Competence Creating ............................................................... 58

4.1 Competence Exploiting vs. Competence Creating ......................................................... 58

4.2 The Relationship between Knowledge Sourcing Patterns and Competence Creating .. 60

4.3 Framework ..................................................................................................................... 67

4.4 Methodology .................................................................................................................. 68

4.5 Results and Discussion ................................................................................................... 71

5. The Evolution of Subunit Roles .............................................................................................. 79

5.1 Typology Based on Knowledge Sourcing Pattern .......................................................... 79

5.2 Typology Based on Innovation Pattern – CC intensity vs. sub distance ........................ 86

5.3 Evolution of Subunit ....................................................................................................... 91

6. Conclusion .............................................................................................................................. 95

vii

APPENDICES ............................................................................................................................. 100

Appendix A1 – Firm List ........................................................................................................... 100

Appendix A2 - Country and patent distribution list ................................................................. 105

Appendix A3 – Geographic distribution of patents – by parent companies and by foreign subunits .................................................................................................................................... 107

Appendix A4 – 31 Technological Fields and Corresponding Patent Class Code ...................... 108

Appendix B – Table of Correlation for Key Variables in Study 1 .............................................. 111

Appendix C – Table of Correlation for Key Variables in Study 2 .............................................. 112

REFERENCES ............................................................................................................................ 114

CURRICULUM VITAE .............................................................................................................. 121

viii

LIST OF TABLES

Table 1 Geographic distribution of patents ................................................................................... 17Table 2 Technologic distribution of patents .................................................................................. 18Table 3 Distribution of Dow Chemical patents .............................................................................. 29Table 4 DuPont patent distribution ............................................................................................... 30Table 5 Novartis patent distribution .............................................................................................. 32Table 6 Hypothesis 1 Model with all data ...................................................................................... 53Table 7 Hypothesis 1 Model without host country sources .......................................................... 54Table 8 Hypothesis 1 Model without home country sources ........................................................ 54Table 9 Results for H2-H5, Study 1 ................................................................................................ 56Table 10 Results for Study 2 - All Data ........................................................................................... 72Table 11 Results for Study 2 - Large Subunit Sample .................................................................... 73

ix

LIST OF FIGURES

Figure 1 Overall Framework of study ............................................................................................... 9Figure 2 Geographic distribution of parent company locations .................................................... 15Figure 3 Dow Chemical patents invented by subunits ................................................................... 30Figure 4 DuPont patents invented by subunits ............................................................................. 31Figure 5 Novartis patent invented by parent and subunits ........................................................... 33Figure 6 Framework of Study 1 - Knowledge sourcing patterns .................................................... 51Figure 7 Conceptual Framework of Study 2 ................................................................................... 68Figure 8 Subunit typology by knowledge sourcing directions ....................................................... 81Figure 9 Subunit Typology by Competence Creativeness .............................................................. 87Figure 10 Subunit Distance and CC Intensity (All Data) ................................................................. 90Figure 11 Subunit Distance and CC Intensity (Large Subunits) ...................................................... 90Figure 12 Categorization of Subunit Evolution Patterns ................................................................ 91Figure 13 Subunit Evolution Patterns ............................................................................................ 92

1

1. Introduction

Knowledge, or in a lot of cases its application form, technology, is the driving force of

world development. Creation of new knowledge is often times the result of accumulation

and combination of existing knowledge. Because of the “stickiness” of knowledge, the

transfer process could be in fact difficult and time-consuming (Szulanski, 1994).

Researches on the nature of knowledge have shed light on organization’s capability of

creating, transferring, absorbing, and recombination of knowledge (Polanyi, 1966; Winter,

1987; von Hippel, 1988; Kogut and Zander, 1992; Nonaka, 1994; Szulanski, 2000; Gupta

and Govindarajan, 2000).

Traditional views of multinational corporation (MNC) consider the internal (most of the

time top-down) transfer of knowledge to be the dominating type in regards to subunit

knowledge accumulation. Due to the very nature of knowledge – being consist of both

explicit and tacit components, it is considered to be organizationally embedded, which

makes it difficult for inter-firm instead of intra-firm knowledge transfer (Kogut, 1988).

Knowledge transfer of this type relies largely on an intra-firm network. Gupta and

Govindarajan (1991, 2000) constructed typologies of MNC subsidiary roles based on the

transfer of knowledge throughout this inter-firm network. However, later studies have

shown that it is possible and necessary for external knowledge transfer. For instance,

Mowery and Oxley (1996) argued that the forming of strategic alliance and the building

of absorptive capacity could facilitate inter-firm transfer.

2

Although the traditional view of MNC knowledge transfer does shed light on one way

that MNCs tend to organize their innovation systems, it doesn’t fully appreciate the

newer development of subunit roles. Over the past few decades, foreign subunits are no

longer contend to remain in a status of a simple manufacturing center; new knowledge

can be created in subunits, innovations can be initiated in subunits, further along these

innovative technology can be transferred to elsewhere within the MNC, including the

headquarters and other subunits. This process has gained grown importance since it was

first spotted by researchers on MNC knowledge transfer and subsidiary external

embeddedness in the 1990s (Gupta and Govindarajan, 1991; Andersson and Forsgren,

1996).

In this relatively new development of views of MNC knowledge transfer, subsidiaries are

considered to rely, more heavily than before, on an inter-firm network, which involves

actors outside of an MNC’s firm boundary. Subsidiaries draw on external knowledge

sources aside from their own MNC intra-firm networks, expanding their knowledge

horizon, extending their technological trajectories into new sectors, and hence bring in

newly combined knowledge to create its unique capability. Since a company usually has

limited resource to be allocated on strategic development, it is sometimes difficult for

such subsidiaries to give strategic focus on both the intra- and inter-firm knowledge

networks. Moreover, aside from the firm affiliation of knowledge sources, the

geographical boundary plays a role in the decision of knowledge sourcing (Pearce, 1989;

Cantwell, 1995). Although modern development of information and communication

technology has vastly enabled and speeded up ways of communication, giving knowledge

3

seekers greater ability to seek for various knowledge sources from dispersed geographical

locations, it is still widely believed that the transfer of technological knowledge happens

more effectively and efficiently when the physical distance between the originator and

recipient is relatively close. Meanwhile, as the advancement of science and technology,

new development of technology requires more and more comprehensive complex

knowledge that often combines disciplines from previously unrelated fields. Research has

shown that in order to cope with uneven technological development levels and

unpredictable product interdependency, firms tend to extend and diversify their

knowledge to areas more than what’s needed for their production, thus firms tend to have

a larger patent footprint than their actual product footprint (Brusoni, Prencipe and Pavitt,

2001). In such cases knowledge sources may become more dispersed into different

technological fields.

It is known that large MNCs tend to have multiple operational units overseas, sometimes

more than one subsidiary could be set up in a single foreign host country. From a

business standpoint this could make sense when that particular host country provides a

wide range of geographical locations suitable for foreign investment, and these

investments could fall into different divisions of operation, which makes setting up

multiple subsidiary entities necessary; or for certain legal or political reasons MNC needs

to divide up their foreign operation into different functional companies, sometimes these

registered companies may even only exist for the purpose of keeping the right paperwork.

For these various reasons, it has been difficult for research on subsidiary level of MNC to

match the company activities in one particular geographical location to the real

4

subsidiaries of the MNC. In this study, an MNC subunit is defined as an operational unit

of an MNC outside its home country. Instead of looking at the organizational structural

sense of an MNC’s foreign entity, the focus of this study is on the innovative activities of

the MNC that are taking place outside the home country of the parent company. Parent

company, in this study, refers to the part of MNC that conducts innovative activities in

the home country. Thus we have our first research question based on the concepts of

MNC subunit and its knowledge sourcing:

Research Question 1: what are the patterns of knowledge sourcing of an MNC subunit

with respect to its technological dispersion, geographic dispersion, and level of sourcing

concentration in its two main sources – home country and host country?

This question examines the precedents, or input, of a firm’s knowledge accumulation,

which further indicates its own knowledge creation or innovation. One would ask then,

what could be the impact of knowledge sourcing on the outcome of innovation? In the

following study we focus on the outcome of innovation and the relationship between the

input and outcome. Following previous works on competence creating (CC) and

competence exploiting (CE) MNC subsidiaries (Cantwell & Mudambi, 2005), the

outcome of innovation is examined here in the form of either CC or CE types of activities.

CE types of activities indicate that the subunit is following the technological trajectory of

the MNC group and further developing the firm’s strong fields of technology. CE

activities themselves are nevertheless innovative activities, just that they don’t emphasize

the novelty of these innovations from the subunit’s parent company. CC types of

5

activities focus on creating competence in technological areas that are new to the

corporate group, or in technological areas that use to be a relatively weak area for the

group. Subunits focusing on CC types of activities have a better chance of creating future

competence for the MNC group, hence might gain a strategic importance within the

group. But one has to note that heavily investing on CC activities too distant from the

corporate group’s core technological trajectory may result in the subunit or particular

business unit falling out of the loop.

In examining the capability of competence creating, this part of the study focuses on the

ability of a subunit to not just be innovative, but also be innovative in a way that’s

different from its parent company. This distinction could be externally driven, such as a

mandate from the parent company, or a local specialization providing available pool of

knowledge that’s related but different from the MNC group’s area of expertise; it can also

be internally driven – for instance a subunit’s management sees the local opportunity and

has the autonomy to investigate into something new and different. This technological

novelty could mean two things to subunits’ knowledge sourcing: for one thing, the new

threads of technology development would push for a different direction of knowledge

sourcing from the parent company, so that it can provide necessary foundations for

distinctive innovation; for another, wider sourcing of knowledge would in return

reinforce the capability of subunit to create distinct innovations compared to its parent

company. This study defines the term technological distinctiveness as the extent to

which a subunit’s level of technological specialization in a given field is distinct from its

parent company. Hence we have our second research questions:

6

Research Question 2: how does a subunit’s knowledge sourcing pattern influence its

local competence creating capability, hence the technological distinctiveness from its

parent company?

Although we study the impact of knowledge input on competence creating or competence

exploiting activities, we have to note that composition of CC or CE activities would in

turn influence the strategic role of the focal subunit, which may result in a shift of MNC

mandate on this subunit, hence reshape its knowledge sourcing pattern. For instance, a

subunit may start with a competence exploiting mandate, which indicates that its

knowledge sourcing pattern follows those that dedicate on exploiting MNC competency,

with a relatively low level of technological diversification; this in turn reinforces this

subunit’s strength in a concentrated area, making it harder to create competency that’s

new to the corporate group. For this reason, we cannot simply view the relationship

between knowledge input and output as a static one. Over time, their mutual influence

should demonstrate a co-evolutionary pattern that could be captured by examining their

relationship longitudinally.

There are two dimensions of subunits’ innovative activity: CC intensity – the extent to

which a subunit’s innovative activities are composed of CC as opposed to CE, and

subunit technological distance – the extent to which a subunit’s innovations are

technologically distant from its parent company’s innovations. The combination of these

two dimensions demonstrates a subunit’s positioning in competence creativeness in its

international corporation group. Our third research questions tries to look at the

evolutionary pattern of subunit strategic roles based on these two dimensions:

7

Research Question 3: if knowledge sourcing and competence creation are two mutually

interdependent processes, is there a typology of evolutionary trajectories followed by the

paths of technological growth of subunits?

Chemical industry, as one of the traditionally considered knowledge-intensive industries,

rely its growth heavily on the accumulation and recombination of knowledge. This

dissertation has chosen the general chemical industry as the object of research, because of

its fully matured development as an industry over hundreds of years, yet widely spread

into various sub-sectors of technologies with some of them being newly emerging within

the past few decades. Its historically consistent concentration on technology development

has brought in life not only in depth research of fields that are closely related to

traditional chemical industry, but also streams of newly emerged knowledge in fields that

are expanded around the main sectors, such as pharmaceutical, biotechnology, etc.

Research has shown that compared with many other industries, chemical industry

consistently emphasizes on the protection of intellectual property by the means of

patenting; the analysis of patenting behavior in chemical industry can to a large extent

reflect the actual advancement of technology.

Therefore this dissertation is going to focus on large multinational corporations in the

broadly defined chemical industry, including general chemicals, petroleum and refining,

pharmaceuticals, and biotechnology. The studies use USPTO patent data to analyze the

relationship between technology fields, citations, and geographic location of inventors.

Three studies are designed to work out the theoretical framework of subunit knowledge

sourcing pattern and competence creation activities:

8

1) Study one is an explorative study with the purpose of understanding the patterns

of MNC subunit knowledge sourcing, theorizing and examining the relationships

between a subunit’s technological diversification degree, and the technological

diversification and geographic dispersion of their knowledge sources, controlling

for the effect of intra-firm and inter-firm sources – whether the sources are from

different parts of the MNC group or originated external to the firm.

2) Study two further explores the influence of patterns of MNC subunit knowledge

sourcing on competence creativeness of subunits. I examine on technological field

level the extent to which a subunit’s activity is co-specialized compared to its

parent company, then analyze to which extent this co-specialization is influenced

by the knowledge sourcing patterns found in study one. This study looks more

closely at each subdivision of the broadly defined chemical industry,

distinguishing petroleum, pharmaceutical and biotech, from general chemicals.

3) Study three uses time serious analysis to examine the evolutionary pattern

between knowledge sourcing pattern and competence creativeness of MNC

subunit. Meanwhile, this study creates a typology of MNC knowledge sourcing

structure and a typology of strategic roles of MNC subunits.

Together, the three studies in this dissertation create a framework for a better

understanding of MNC subunit knowledge sourcing pattern and competence creating

activities. A comprehensive overview of this framework is demonstrated in Figure 1.

9

Figure 1 Overall Framework of study

Contributions of this dissertation are three fold. Theoretically, the studies contribute in

areas of MNC knowledge accumulation and competence creating by establishing new

connection between the two and further on identifying evolutionary trajectory of MNC

subunit strategic role development. Empirically, the studies created new ways of studying

MNC subunit innovation, as well as a clear typology of subunit role based on its

competence creativeness. It also examines the relationship between subunit knowledge

sourcing and innovative activity, which could lead to further development of study on

these two interdependent characteristics of subunit strategy. Finally, methodologically,

the studies propose a combination of two dimensions – subunit technological distance

and CC activity intensity – to indicate strategic positioning of the subunit within an MNC

international network, this method takes into consideration of the composition of subunit

10

overall innovativeness and its novelty compared to parent company, which gives us a

comprehensive view of subunit competence creativeness.

In the next chapter, I will describe the construction and characteristics of my main dataset,

as well as some key variables that’s been introduced in this chapter. A case example of

Dow Chemical is given to demonstrate the utilization of data. The following three

chapters will be dedicated to studies one, two and three, respectively; then I will conclude

the dissertation with some final discussions in the last chapter.

11

2. Data Overview

In this chapter, I will describe the construction of my dataset, provide definitions of

several key concepts, and set an example with a typical firm in my sample – Dow

Chemical.

2.1 Data Description

Patent dataset construction

The studies in this dissertation are based on innovative activities of multinational

corporations in Chemical industry represented on patent data granted by US Patent and

Trademark Office (USPTO). The reason to choose USPTO patent data to reveal

innovative activities is twofold. For one thing, US patent data is comprehensive, granted

to assignees across the world, which represents the majority of innovative activities

conducted by all types of individuals and organizations in any country. Under the current

trend of fast movement towards globalization, it becomes increasingly important for

companies, especially MNCs, to obtain recognition from a universally accepted

intellectual property protection mechanism, and US patenting system fits this need. As a

result, although the basis of the patent office is in US, but it in fact represents patenting

activities from all over the world. The latest statistical data provided by USPTO shows

that among all of the 3,433,074 patents granted during the period of 01/01/1987 –

12/31/2011, about 52% or so are US originated patents, the other 48% or so are all from

foreign origins. Another reason is that patent data, although some would argue is a post

ante way of examining innovative behavior, and it’s inevitable that there are noises of

12

patenting behavior that are not solely for the purpose of protecting intellectual property

hence being an accurate proxy to innovative activities, studies have shown that as a

objective dataset patent data has the advantage of providing universally comparable

indicators of both innovative activity of firms and connections of these innovative

activities in terms of their accumulation paths (demonstrated by the citations of patents).

Discussion on the using of patent data has led to a specific issue in regards to patent

continuation. Continuation patent application is a unique situation to the US patent

system, it indicates situations when a patent application is a parent to one or more

applications; the continuation applications have the same disclosure as the earlier field

parent application and they claim the benefit of the filing date of the earlier application

(Eisenberg, 2000). One of the key conditions of filing continuation application is that at

least one inventor must be in common between the parent application and the

continuation application. Although it is arguable that continuation helps protecting

“pioneering inventors” in fields with high uncertainty, economists have been long

criticizing the abuse of continuation patents (Lemley & Moore, 2003). In this study, it is

concerned that in a lot of cases that there might be multiple patents in one “patent family”

filed as continuation patents but in fact all belonging to the same invention project.

Therefore, within a dataset of a company’s patent data, there might be ones that belong to

one family, which could result in high interdependence between patents. However,

previous study has shown that the type of continuation mainly used in chemical industry

(which has the character of being R&D intensive) is called Continuation-In-Part (CIP).

CIP patents tend to cover more “valuable” invention compared to the other types of

13

continuations (Hedge, Mowery & Graham, 2007)1

The patent data records in this dissertation come from mainly three sources: NBER US

patent database, patent database developed by Dr. John Cantwell, and some records

pulled directly from the USPTO patent database. The data consists of US patent granted

to 147 large corporations in the general chemical industry between 1976 and 2006.

General chemical industry as the object of research because of its fully matured

development as an industry over hundreds of years, yet widely spread into various sub-

sectors of technologies with some of them being newly emerging within the past few

decades. Its historically consistent concentration on technology development has brought

in life not only in depth research of fields that are closely related to traditional chemical

industry, but also streams of newly emerged knowledge in fields that are expanded

around the main sectors, such as pharmaceutical, biotechnology, etc. Research has shown

that compared with many other industries, chemical industry consistently emphasizes on

the protection of intellectual property by the means of patenting; the analysis of patenting

. In other words, due to the specific

characteristic of chemical industry, continuation patents contain unique new information

that differs from the “parent patent”. Although there might still be interdependence

between continuation patents and their parent patent, to some extent it represents the

continuous natures of innovative activities (along with other efforts besides innovation by

patent attorneys). Further discussion will be brought up again towards the end of this

dissertation in Chapter Six.

1 There are three types of continuations – Continuation Application (CAP), Continuation-In-Part (CIP), and Division. CAP must have the same disclosure as the original application; CIP includes a substantial portion or all of the original application but adds new disclosure to it; Division claims one part of the invention from the original application (Hedge et al., 2007).

14

behavior in chemical industry can to a large extent reflect the actual advancement of

technology.

One of the difficulties of using company patent data from a general patent database is the

attribution of patents to firms. Although each patent has one or multiple assignee

indicating its affiliation with an organization, these assignee codes are not a reliable

source of identifying the actual subunit where the patent is invented. Some firms may

have multiple assignees for their patents due to policy reasons; some may attribute all

patents to their parent company even when the patents are in fact invented in a foreign

location. In order to prevent attributing patent related innovative activities to the wrong

location, and to select the right patents and attribute them to my target firms’ foreign

operations, there are several steps to take. First, I identified the largest firms in the

general chemical industry from Fortune 500 listings and according to records of historical

trends. Only companies with a consistent record of high performance throughout my

research time period are chosen into the sample.2

2 Some firms are later dropped from the analysis sample in studies one and two if they are relatively new (have patenting records for less than 10 years during the researching period of 1976-2006), or have very insignificant patenting activities throughout the period (less than 50 patents granted during the entire thirty-one years).

However, these patents from newer or

smaller firms are counted into the total number for the analysis of chemical industry

innovative activity technology and geographic distribution. The selection of industry

includes firms focusing on general chemical, pharmaceutical and biotechnology,

petroleum refining and related sectors. Countries of origin of these firms include

developed nations like US (71), Germany (8), UK (10), Japan (28), etc., and developing

15

nations like Israel (1), Taiwan (2).3

A list of these companies is available in Appendix A1.

Figure 2 demonstrates the distribution of these 147 firms’ home country (parent company)

locations.

Figure 2 Geographic distribution of parent company locations

Once the target firms are identified, the second step is to map out all affiliates of these

firms. I used the resource of Who Owns Whom to manually input all names of affiliates

for my target firms. Then a match is run between these company names with the standard

company name data on the NBER database, hence extracting out all possible assignee

codes that’s associated with target firms. Each target firm may have various number of

assignee codes, depending on the firm’s policy on patent ownership. However, the

location of these assignees (as provided in the NBER dataset) doesn’t correctly reflect the

location of the invention (innovative activity). The purpose of extracting these assignee

3 The numbers in the parentheses reflects the number of firms in those countries.

Parent Company Distribution

USA

Germany

UK

Italy

France

Japan

Netherlands

Belgium

Switzerland

Denmark

16

codes associated with our target firms is solely to further extract all patents assigned to

these assignees, hence our 147 firms.

Therefore, the third step is to extract all patents assigned to the assignee numbers

identified in step two, which allows us to obtain the full patent dataset of our target firms.

From here on, I was able to extract all 286,667 patents that are assigned to my target 147

firms in the general chemical industry from the NBER patent database.

Then, the fourth step is to identify for each of these patents the first inventor’s country

location. Location of first inventor is a better indicator of the actual location of innovative

activity. This geographical data is extracted from Dr. Cantwell’s proprietary patent

database. The geographic location of first inventors in those patents attributable to

international corporate groups can then be used to examine the geographical distribution

of the technological activity of these firms (Cantwell, 1995). Each patent invented in a

certain country is an indicator of innovative activities by a firm in that country. Subunit,

in this research, is defined as the patent-country combination indicating MNC operation

in a country, which means the country location of the patent’s first inventor indicates the

host country location of the subunit. Table 1 demonstrates a distribution of patents

invented in these countries. Column two shows the overall patent distribution (patents

invented in those countries), column three and four indicates the number of parent

companies in each country and the patents invented by these parent companies; column

five and six indicates the number of foreign subunit in each country and the patent

invented by those subunits. A more comprehensive distribution table of patent numbers

in each country location (including both subunit patents and parent patents distribution)

17

can be found in Appendix A2. Pie chart figures demonstrating the distribution of patents

invented by parent companies and foreign subunits across geographical locations can be

found in Appendix A3.

Table 1 Geographic distribution of patents

Country Patents Parent Patents by

Parent Subunit Patents by

Subunit USA 153,916 71 130,854 62 23,062 Germany 40,341 8 25,768 93 14,573 UK 13,613 10 7,819 88 5,794 Italy 1,445 N/A N/A 65 1,445 France 9,646 7 7,965 70 1,681 Japan 47,776 28 45,240 69 2,536 Netherlands 1,083 2 460 60 623 Belgium 2,727 1 485 57 2,242 Switzerland 8,445 6 7,876 52 569 Denmark 1,167 2 992 30 175 Canada 2,427 2 67 71 2,360 Australia 526 2 79 57 447 South Korea 674 2 151 16 523 Other Countries 2,881 6 506 465 2,375 Total 286,667 147 228,262 1255 58,405

Technological fields of patents

The unit of analysis is technological fields (studies one and two) – innovative activities

conducted by a certain MNC subunit in particular technological fields – and MNC

subunits (study three) – the combination of parent-country indicates an MNC’s operation

unit in a certain country. The primary field of technological activity of each patent is

derived from the US patent class system, and these fields are grouped into 56

18

technological sectors (Cantwell and Andersen, 1996). This grouping is established on the

basis of the class system, providing economic meanings for each of these fields. Due to

the limitation of industry sample, some fields among these 56 are not as highly

represented as others; as a result an analysis across these 56 fields of technological

sectors would somehow demonstrate a skewed result. Therefore this study further

grouped these 56 fields into 31 respective ones in order to have meaningful representative

numbers for each technological field so that there is a comparable structure of data

distribution. Table 2 demonstrates a brief description of each technological field and

patent frequencies of these fields in this research. A detailed description of the

relationship between 31 technological fields and the corresponding patent class and sub-

class codes can be found in Appendix A4.

Table 2 Technologic distribution of patents

tech31 Field Freq. 1 Food and tobacco products 1,735 2 Distillation processes 738 3 Inorganic chemicals 4,139 4 Agricultural chemicals 5,035 5 Chemical processes 13,636 6 Photographic chemistry 9,461 7 Cleaning agents and other compositions 15,082 8 Synthetic resins and fibres 37,678 9 Bleaching and dyeing 3,130

10 Other organic compounds 48,999 11 Pharmaceuticals and biotechnology 57,747 12 Other chemicals and Related - disinfecting, preserving, textiles and explosives 815 13 Metallurgical processes 2,488 14 Miscellaneous metal products 4,080 15 Chemical and allied equipment 10,947 16 Paper making apparatus 1,808 17 Assembly and material handling equipment 1,694

19

18 Other specialised machinery 3,553 19 Other general industrial equipment 2,813 20 Mechanical engineering nes 3,870 21 Electrical devices and systems 2,146 22 Other general electrical equipment 4,427 23 Office equipment and data processing systems 2,431 24 Electrical equipment nes 4,290 25 Transport equipment 896 26 Rubber and plastic products 4,056 27 Non-metallic mineral products 14,143 28 Coal and petroleum products 6,579 29 Photographic equipment 2,417 30 Other instruments and controls 14,019 31 Other manufacturing and non-industrial 1,815

Total

286,667

It is noticeable in Table 2 that the patents used in this study are highly concentrated in

chemical related technological fields, and it also shows a concentration on some non-

chemical fields that are considered general purpose technology (GPT, shown in bold

letters in Table 2). This distribution is as expected with our selection of industry as well

as the utilization of GPT fields.

2.2 Key concepts

Knowledge Sourcing and Patent Citations

The use of patent citation as an indicator of knowledge flow has been questioned by a

serious of research since there is a large portion of citations added by the examiner rather

than by inventors of the patents (Alcacer & Gittelman, 2006; Alcacer, Gittelman &

Sampat, 2009). One cannot argue that those examiner-added citations can be a good

indicator of knowledge “flow”. This study, however, is not trying to measure the actual

“flow” of knowledge. The definition of knowledge sourcing in this dissertation is the

connection of a subunit with direct or indirect sources of knowledge components that

20

helps in the process of recombination and generating new knowledge. This concept of

knowledge sourcing emphasizes the implicit connection between different knowledge

bodies; therefore patent citation - even with examiner added ones – is a good indicator of

knowledge sourcing. If there is a citation linkage between two patents, it means there is a

direct or indirect influence from one to another.

Technology – Diversification, Dispersion, Distinctiveness, and Distance

There is three “D”s about technology – technological diversification, technological

dispersion, and technological distinctiveness – that will be used throughout this

dissertation. It is important to clarify the meanings of each one of these terms before we

move on to the actual analysis.

Technological diversification and technological dispersion have similar constructions

from the point of data. However, the meanings behind them are different. Technological

diversification focuses at the innovative activity itself. By engaging in diversified

technological activity, firms are actively expanding their operations across different types

of productive activity (Cantwell and Piscitello, 2000). Technological diversification is

therefore considered a means of exploring and accumulating corporate competency. From

the data point of view, technological diversification indicates the extent to which a

subunit’s innovative activities (patents) are wide spread across various technological

fields. Technological dispersion, on the other hand, focuses on the dispersed

technological composition of knowledge sources. With respect to the data, technological

dispersion indicates the degree of knowledge sources being widespread across different

21

technological fields. While technological diversification focuses at the subunit itself, it is

an intrinsic measure of the subunit’s strategic intension in regards to fields of

specialization, different fields among a subunit’s expertise could be interrelated to one

another, and there is a meaningful combination of these fields; technological dispersion

focuses at the extrinsic measure of a subunit’s knowledge sources, it is merely a

demonstration of fields that could contribute to a subunit’s technological innovation, and

these fields are not necessarily all interrelated with one another.

That being said, the actual measurement of these two concepts is the same – both are

measured by the reverse of Herfindahl-Hirschman index (HHI) of concentration. HHI

was first used in 1951 to analyze the concentration within the steel industry, calculated by

adding the squares of firm’s market shares in per cent (Weinstock, 1982).

𝐻𝐻𝐻𝐻𝐻𝐻 = �𝑆𝑆𝑖𝑖2𝑛𝑛

𝑖𝑖=1

Where si is the market share of firm i measured in percentage points. This definition

implies that the shares of the larger firms are given greater weight than those of the

smaller companies. The result of this original form of HHI should vary between 0 and 1,

a higher value of HHI indicates higher level of concentration in the market. Taking from

this point of analysis, if the si in the calculation is share of technological innovation

instead of market share, then this index could be used to demonstrate the technological

concentration of given industry. Then we deduct this result from 1, which gives us

reversed result of technological concentration. This could then be used to indicate the

22

diversification/dispersion of technology – depending on whether we are examining the

knowledge bodies themselves or the knowledge sources. Hence, my calculation of

technological diversification/dispersion is:

𝑇𝑇𝑇𝑇𝑇𝑇ℎ𝐷𝐷𝑖𝑖𝐷𝐷𝑖𝑖 = 1 − �(𝑃𝑃𝑖𝑖𝑖𝑖∑ 𝑃𝑃𝑖𝑖𝑖𝑖𝑖𝑖

)2

31

𝑖𝑖=1

In which TechDivi indicates the technological diversification of subunit i, Pij is the

number of patents a subunit invented in technological field j, and ∑ 𝑃𝑃𝑖𝑖𝑖𝑖𝑖𝑖 sums up the total

number of patents in all field invented by this subunit.

For technological dispersion, the formula is:

𝑇𝑇𝑇𝑇𝑇𝑇ℎ𝐷𝐷𝑖𝑖𝐷𝐷𝐷𝐷𝑖𝑖 = 1 − �(𝑃𝑃𝑖𝑖𝑖𝑖∑ 𝑃𝑃𝑖𝑖𝑖𝑖𝑖𝑖

)2

31

𝑖𝑖=1

Since this is for knowledge sourcing data, we want to examine the technological

dispersion of knowledge sources of a subunit’s activity in each field, so the analysis is at

technological field level. Among all the citations by a subunit’s patenting activity in each

field, we construct the similar calculation method as technological diversification, only

this time Pij on the right hand side of the equation indicates number of cited patents in

technological field j.

Technological distinctiveness, the main variable used in study 2, is another concept

that’s completely different from the previous two technological related terms. The notion

23

of technological distinctiveness comes from the idea of creating a variable to examine the

extent to which a subunit’s specialization in certain technological field is different from

its parent company, which can serve as an indicator of that subunit’s competence creating

activities. Before I explain the rationale behind the construction of such variable, I need

to first introduce a concept that is needed in understanding the mechanism behind my

measure – RTA.

Revealed Technological Advantage (RTA) is an index that demonstrates the level of

technological specialization for one unit of observation (a subunit, a firm, or even a

geographic location – country or region) in a given technological field, it was first

pioneered in the work of Soete (1980) and then further developed by Patel and Pavitt

(1987, 1991) and Cantwell (1989, 1995). The RTA index of a firm in a particular

technological field is given by the firm’s share of patenting in that field divided by its

share of patenting in all sectors and is defined as follows:

𝑅𝑅𝑇𝑇𝑅𝑅𝑖𝑖𝑖𝑖 = (𝑃𝑃𝑖𝑖𝑖𝑖 /� 𝑃𝑃𝑖𝑖𝑖𝑖 )/(� 𝑃𝑃𝑖𝑖𝑖𝑖𝑖𝑖

/��𝑃𝑃𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖

)𝑖𝑖

In which is the total number of patents of firm i in field j. The index varies around unity,

such that values greater than one suggest a firm’s comparative advantage in the field of

activity in question relative to other firms in the same industry, while values less than one

are indicative of a position of comparative disadvantage (Cantwell and Piscitello, 2000).

When this is calculated at a subunit level, the unit of analysis is then changed from firm

to subunit – a firm’s activities within a foreign host country. In the meanwhile, if we

enlarge the unit of analysis to a country level, then this index could be used to reveal

24

comparative advantage or disadvantage of a country’s activities in certain technological

fields.

While RTA index illustrates the technological advantage/disadvantage positions of a

subunit in certain fields, we still need to apply a mechanism to compare the positioning of

both a subunit and its parent company, which then leads to our operationalization of

technological distinctiveness – the extent to which a subunit’s degree of specialization in

a certain technological field is distinct from its parent company. To put it short, we want

a method to effectively compare the RTA between a subunit and its parent company in

each field. Here we took a methods used by Cantwell and Santangelo (2002) that’s

designed to assess the co-presence or absence of a matching degree of specialization

between the two groups of firms in each patent class. This method draws on the measure

of intra-industry trade across sectors in the international trade literature (Grubel and

Lloyd, 1971, 1975). The Grubel-Lloyd index generates a variable GLi by comparing the

export and import of good i:

Where Xi measures the export, and Mi measures the import of a good i. If GLi = 1, there is

only intra-industry trade, whereas if GLi = 0, there is only inter-industry trade. Cantwell

and Santangelo (2002) further developed this model as a measure of co-presence of a

matching degree of specialization between the two groups of firms (k) (where k = 1 or 2)

in a particular ICT patent class (c) in each region (r):

25

Crc = 2 min (RTA1rc, RTA2rc) / (RTA1rc + RTA2rc); 0 ≤ Crc ≤ 1

In my study, however, to examine technological distinctiveness, I use the reverse of co-

specialization, which is:

𝑇𝑇𝐷𝐷𝑖𝑖𝑖𝑖 = 1 −𝑅𝑅𝑇𝑇𝑅𝑅𝑖𝑖𝑖𝑖 + 𝑅𝑅𝑇𝑇𝑅𝑅𝐷𝐷𝑖𝑖 − |𝑅𝑅𝑇𝑇𝑅𝑅𝑖𝑖𝑖𝑖 − 𝑅𝑅𝑇𝑇𝑅𝑅𝐷𝐷𝑖𝑖 |

𝑅𝑅𝑇𝑇𝑅𝑅𝑖𝑖𝑖𝑖 + 𝑅𝑅𝑇𝑇𝑅𝑅𝐷𝐷𝑖𝑖=

|𝑅𝑅𝑇𝑇𝑅𝑅𝑖𝑖𝑖𝑖 − 𝑅𝑅𝑇𝑇𝑅𝑅𝐷𝐷𝑖𝑖 |𝑅𝑅𝑇𝑇𝑅𝑅𝑖𝑖𝑖𝑖 + 𝑅𝑅𝑇𝑇𝑅𝑅𝐷𝐷𝑖𝑖

Where 𝑇𝑇𝐷𝐷𝑖𝑖𝑖𝑖 denotes the technological distinctiveness of subunit i in field j, 𝑅𝑅𝑇𝑇𝑅𝑅𝑖𝑖𝑖𝑖 is the

revealed technological advantage of subunit i in field j, whereas 𝑅𝑅𝑇𝑇𝑅𝑅𝐷𝐷𝑖𝑖 represents the

revealed technological advantage of subunit i’s parent company in field j.

However, this measure of distinctiveness doesn’t specify the direction of distinction – a

large number (closer to one) of 𝑇𝑇𝐷𝐷𝑖𝑖𝑖𝑖 could indicate either the subunit is distinctively

more specialization in the field than its parent company, or the parent company is

distinctively more specialized than this subunit. In order to capture the effect of subunit’s

competence creativeness, one more step is needed to identify subunits that are more

specialized than their parent companies in the focal fields. Hence in actual data

processing, it is necessary to screen out situations when 𝑅𝑅𝑇𝑇𝑅𝑅𝑖𝑖𝑖𝑖 < 𝑅𝑅𝑇𝑇𝑅𝑅𝐷𝐷𝑖𝑖 , and

subsequently set the value of 𝑇𝑇𝐷𝐷𝑖𝑖𝑖𝑖 to zero, indicating that this subunit has no distinct

specialization in this field compared to its parent company.

Finally, the term technological distance between a subunit and its parent company

brings the analysis up to a firm/subunit level again. The calculation of this variable uses

the Jaffe (1986) approach which borrows the concept of Euclidean vector calculation

using (1 – Cosin Similarity) to construct a measure for distance:

26

𝑇𝑇𝑇𝑇𝑇𝑇ℎ 𝐷𝐷𝑖𝑖𝐷𝐷𝐷𝐷𝐷𝐷𝑛𝑛𝑇𝑇𝑇𝑇 = 1 −∑(𝑆𝑆𝑆𝑆𝑆𝑆 × 𝑃𝑃𝐷𝐷𝑃𝑃𝑇𝑇𝑛𝑛𝐷𝐷)

�∑𝑆𝑆𝑆𝑆𝑆𝑆2 × �∑𝑃𝑃𝐷𝐷𝑃𝑃𝑇𝑇𝑛𝑛𝐷𝐷2

Where Sub and Parent represents the number of patents in field j granted to the subunit

and parent, respectively; the ∑ in this equation is calculating the sum of patents

through all technological fields.

2.3 Sample Data from Dow Chemical, DuPont, and Novartis

This section of the paper will depict an overall picture of three selected chemical industry

firms: Dow Chemical (US), DuPont (US), and Novartis (Switzerland), to illustrate the

basic data composition of this research.

Overview of the three firms

The Dow Chemical Company is an international chemical enterprise, whose center of

operations lies in Midland, Michigan, USA. Dow Chemical is known for the

manufacturing of specialty chemicals, advanced materials, agrosciences and plastics.

Dow was founded in 1897 by Canadian chemist Herbert Dow who created a new method

of extracting bromine. Over the next twenty years, Dow was effective a rapidly

diversifying its product line and supplied many of the war materials during World War I,

which were previously imported from Germany. Dow expanded internationally in 1942

with the organization of Dow Chemical of Canada and followed suit in Japan in 1952

with Asahi Dow, it’s first subsidiary overseas and Dow Europe in Zurich, Switzerland.

By 1965, with additional plants created in Latin America, sales outside the U.S

constituted nearly 25 percent of total sales. By 1973, Dow became the first foreign

27

industrial company listed on the Tokyo Stock Exchange and sales outside of the U.S.

reached 47 percent. Currently, 5000 products are manufactured at 188 sites in 36

countries across the globe with a focus on six operating segments: Electronic and

Functional Materials, Coatings and Infrastructure Solutions, Agricultural Sciences,

Performance Materials, Performance Plastics and Feedstocks/Energy.

E. I. du Pont de Nemours and Company, more often referred to as DuPont, started their

business in the explosives industry from the very beginning of 20th century. Its

development experienced six phases: the founding period from 1902 to 1911, the

centralization and diversification from 1911 to 1921, decentralization in the period of

1921 to 1935, Nylon and new Nylon era from 1935 to 1960, the maturation period in the

1960s, and then the rethinking of corporate strategy since the 1970s (Hounshell, 1989,

1992). In the sense of the technological sectors, after starting from explosives, DuPont

soon diversified into other fields such as synthetic fibers and photographic chemicals,

with the driving force of the development of Nitrocellulose technology. In research

organizing, DuPont had a unique strategy of a general experimental laboratory – namely,

the Experimental Station – whose job was to pick up or create opportunities for

innovation in all other divisions in the corporation. This distinction between division labs

and the general lab was DuPont’s strategy of catching up with IG back to the early 20th

century, and which was proved to be successful – at least in the sense of facilitating the

technological diversification of DuPont. In more recent years, the company ranked 66th

28

in the Fortune 500 on the strength of nearly $28 billion in revenues and $1.8 billion in

profits.4

Novartis is a pharmaceutical company that has its center of research and development

based in Basel, Switzerland, with other key research sites located in Horsham, UK;

Vienna, Austria; Tsukuba, Japan; East Hanover, New Jersey, USA. Although the firm

was formed in 1996, its innovative activities dated far back in the 19th century. Merged

from Ciba-Geigy and Sandoz Laboratories, Novartis inherited long innovative history of

J.R. Geigy, CIBA, and Sandoz.

5

Dow Chemical patent description

With an emphasis on innovation, Novartis not only

vigorously develop its in-house R&D capability, but also seeks for global collaboration

possibilities. Their strategy for innovation is “focused diversification”, by which they

mean a concentration on pharmaceutical, biomedical innovation while encouraging

diversification into related areas. Novartis is now ranked as one of the top pharmaceutical

companies in the world, along with Pfizer, GlaxoSmithKline, Sanofi-Aventis, and Merck.

Dow Chemical and its affiliated subunits have been granted a total number of 12193

patents during the years 1976-2006 in the USPTO system, among which a majority

11568 (95%) are invented in the US, the rest 625 (5%) patents are invented by its 24

foreign subunits distributed around the globe. It is obvious that Dow Chemical clearly has

an innovation policy of home base concentration. A distribution of these patents in

different subunits is demonstrated in Table 3. As shown in the table, aside from its parent

4 "Fortune 500: 1955–2005.". CNN. Retrieved September 19, 2011 5 Adapted from "Company history", corporate website (Novartis), novartis.com

29

company, most of the innovative activities of Dow Chemical is concentrated in

developed countries such as Germany, UK, Netherlands, Canada, etc. Figure 3 illustrates

the distribution of patents invented by subunits.

Table 3 Distribution of Dow Chemical patents

Host Country of Subunit Number of Patents Germany 88

UK 86 Italy 13

France 48 Japan 46

Netherlands 81 Belgium 62

Switzerland 58 Sweden 3

Denmark 5 Ireland 1 Spain 6

Greece 2 Canada 89

Australia 5 India 4 Brazil 1 Chile 1

Colombia 2 South Korea 7

Taiwan 2 South Africa 9

Other Latin America 2 Other Asia 4

30

Figure 3 Dow Chemical patents invented by subunits

DuPont Patents

The patents from DuPont also demonstrate a concentrated pattern on its home country the

US. Among the 12969 patents granted during the period 1976-2006, about 90% were

invented by the parent company, with 1229 (about 10%) invented in foreign locations, as

shown in Table 4 and Figure 4.

Table 4 DuPont patent distribution

Host Country of Subunit Number of Patents Germany 283

UK 100 Italy 8

France 45 Japan 149

Netherlands 104 Belgium 53

Switzerland 74 Sweden 2

Denmark 1

Germany

UK

Italy

France

Japan

Netherlands

Belgium

Switzerland

Sweden

Denmark

31

Spain 3 Luxembourg 17

Austria 15 Norway 2

Czechoslovakia 1 USSR 4

Canada 311 Australia 15

New Zealand 1 Brazil 4 Israel 9

Argentina 1 Mexico 1

Venezuela 1

Figure 4 DuPont patents invented by subunits

DuPont has its foreign innovative activities concentrated mainly in Germany and Canada,

followed by other developed countries like Japan, UK, and Netherlands. To some extent,

compared to Dow Chemical, DuPont puts more emphasis on developing centers of

excellence in foreign locations. Dow Chemical, on the other hand, emphasizes a more

Germany

UK

Italy

France

Japan

Netherlands

Belgium

Switzerland

Sweden

Denmark

32

even development across its main foreign R&D locations. It is later shown in the analysis

of competence creativeness, in Chapter Four, that although an equal number of subunits

from both companies has reported competence creating activities, more of these subunits

from DuPont have indicated a continuous strength over the years in competence creating,

and these activities demonstrate a continuous trajectory of DuPont subunit’s areas of

expertise. For example, the subunit of DuPont in Canada has shown over the years a

continuous specialization in technological fields 13, 18 and 19, which represents areas

like metallurgical processes, specialised machinery, and other industry equipments. These are

not DuPont’s key areas of expertise as a corporate group, but their Canadian subunit was

able to develop unique strength in these areas to offer complementary technology

development for the firm.

Novartis

Novartis’s patent distribution is different from Dow Chemical and DuPont. Given that

this is not an US based firm, it is understandable then that Novartis doesn’t have as high a

concentration in US hosted innovation as the other two. Among its total number of 9148

patents in the period of 1976-2006, about 54% (4942) of them are invented in its home

country Switzerland, while the rest are invented in foreign subunits. Table 5 and Figure 5

depicts the numbers and shares of patent distribution.

Table 5 Novartis patent distribution

Host Country of Subunit Number of Patents USA 2160

Germany 894 UK 663

33

Italy 8 France 202 Japan 41

Netherlands 14 Belgium 6

Switzerland 4942 Sweden 7

Denmark 7 Spain 2

Austria 35 Norway 2 Canada 34

Australia 25 India 20 Brazil 2 Israel 1 Chile 1

Mexico 2 South Korea 1

China 1 Other Countries 3

Figure 5 Novartis patent invented by parent and subunits

USA

Germany

UK

Italy

France

Japan

Netherlands

Belgium

Switzerland

Sweden

34

Besides its home country, Novartis has extensive innovative activities recorded by

patents in US, Germany, UK, and France. Aside from regular R&D facilities in these

countries, it is discussed before that Novartis has set up key research sites in countries

outside of its home location. Some newer research centers are not demonstrating a strong

impact from the current dataset, due to the cut up time point of 2006. However, as can be

observed in these charts, in comparison with the previous two firms, Novartis has a more

evenly dispersed geographic composition of innovative activities. One of the research

concentrations of this dissertation is to examine whether this type of structure could

increase the chance of developing CC subunits. Results in later chapters are going to

answer this question. Interestingly, as shown in Chapter Five, Novartis is indeed one of

the few firms that have shown trends of subunit evolution over the years, although

different subunits tend to display distinct patterns of evolution, due to various external

and internal driving forces for innovation.

35

3. MNC Subunit Knowledge Sourcing

3.1 Knowledge based view of MNC

Theories of firm in regards to forms of hierarchy or market to organize production and

related economic behavior were discussed by the school of transaction cost economists

(Coase, 1937; Williamson, 1975). In the resource based view, the organizations are

considered to gain sustainable competitive advantages from the possession of a series of

resources, and hence achieve greater long-term performance (Penrose 1959; Barney 1986;

Grant 1991). Conventional approaches consider general sources of competitive advantage,

namely advantages in cost and differentiation (Porter 1985). Further empirical efforts

clarified the relationship between the organizations’ resources and capabilities, indicating

that when resources are scarce, they become valuable and this leads to temporary

competitive advantages (Schmalensee, 1983; Henderson and Cockburn, 1994; Powell and

Dent-Micallef, 1997; Spanos and Lioukas, 2001; and Powell, Lovallo and Caringal,

2006). If organizations can protect their main resources from imitation, transfer or

substitution, these competitive advantages can become sustainable (Hulland 2004), so

confirming the strategic decisions taken by the firm (Parker and Russell 2004).

Following the lead of Resource Based View, studies in the past few decades have

turned their focus onto a more specific type of resource – knowledge. Scholars have

investigated the concept of knowledge using various perspectives and methods. Polanyi

(1966) brought the notion of “tacit” knowledge versus “explicit” knowledge into

attention of the academia world. Since then the discussion about characteristic of

knowledge has been developed around this notion. Winter (1987) suggested that

36

knowledge could be understood in terms of four dimensions – (a) tacit-articulate, (b)

observable-not observable, (c) complex-simple, and (d) element in a system-independent.

Among these dimensions the one that emphasizes the tacitness nature of knowledge is the

most researched on by different strands of studies. Kogut and Zander distinguished

knowledge as information and know-how (Kogut & Zander, 1992). Information is

“knowledge which can be transmitted without loss of integrity once the syntactical rules

for deciphering it are know”, while know-how is “the accumulated practical skill or

expertise that allows one to do something smoothly and efficiently” (Kogut & Zander,

1992; von Hippel, 1988). Kogut and Zander used the term “combinative capabilities” to

describe the ability of the firm to "generate new combinations of existing knowledge"

and "to exploit its knowledge of the unexplored potential of the technology" (Kogut and

Zander, 1992, P.391). Nonaka’s notion of dynamic interaction between two dimensions

of knowledge transfer extended the thoughts of tacit vs. explicit knowledge, bringing

linkages between the transformations (“knowledge conversion”) of the two types of

knowledge and transfers among different entities that possess knowledge (Nonaka, 1994).

Indeed knowledge is inevitably considered a resource that is crucial for firm

technological capability development (Grant, 1996; Teece et al, 1997). The knowledge

based view of organizations take a “static” perspective in suggesting knowledge as a

strategic asset of an organization (Grant, 1996), where dynamic capabilities discusses the

creation of new knowledge through the interaction of existing tacit and explicit

knowledge (Teece et al, 1997). Upon the distinction between explicit and tacit

knowledge, scholars emphasizes that a globally dispersed R&D operations provide

37

MNCs with competitive advantages not available in single-country centralized R&D

operations (Dunning, 1995), and that such a competitive advantage is based on how

efficiently MNCs share knowledge across HQs and R&D subsidiaries (Gupta &

Govindarajan, 2000). In the context of within a multinational corporation network, Gupta

and Govindarajan (2000) identified five determinants of intra-corporate knowledge flows

(including outflows and inflows), namely the value of knowledge stock, the motivation

disposition to share/acquire knowledge, the existence and richness of transmission

channels, and the absorptive capacity.

In order to avoid confusion caused by the ambiguity of the term “network” and its

wide use across different fields, it is necessary to clarify the definition “network” and

“knowledge sourcing”. In the domain of social science, the origins of the concept of

network refers to a social structure made up of individuals (or organizations) called

“nodes”, which are tied (connected) by one or more specific types of interdependency,

such as friendship, kinship, common interest, financial exchange, or relationships of

beliefs, knowledge or prestige. Clearly, the term itself has a variety of meanings. The

research on organization studies, however, often restrict the term to an extent which

stresses that the formation of a network requires actual connections (most of the time in

the forms of personal communication or interaction) between the nodes. Such definitions

can be found in studies on business network, knowledge flows, technology diffusion, etc.

(Ingram & Roberts, 2000; Owen-Smith & Powell, 2004; Hansen & Lovas, 2004; Bell &

Zaheer, 2007).

38

Following this stream of research, the concept of “knowledge networks” is widely

used in studies on the first level of social networking activities, sometimes

interchangeable with the terms “knowledge flow”, or “knowledge transfer” (Best, 2001;

Britton, 2004; Van Geenhuizen, 2008). However, this study is looking at these network

connections of knowledge reflected on a different level – in this more abstract level of

knowledge network, the “nodes” are entities (usually the organizations) with a stock of

knowledge that’s a combination information and knowhow from different technological

fields, and the “ties” are knowledge connections among different entities in the forms

such as sharing similar knowledge base, applying related technologies, or building upon

one another. This level of connection does not necessarily require actual interpersonal

relationships, but instead it indicates the relatedness of knowledge between organizations,

mainly in the formal or informal manners of acknowledging the related knowledge that a

certain knowledge creation activity is based upon. Therefore, in order to distinguish my

work from that of many social scientific studies of knowledge network, I define this type

of connection as “knowledge sourcing” – the extent to which the knowledge of two

organizations (subunit level) is connected with one another in the process of production

and innovation via the sourcing of knowledge. The sourcing activity itself can be the

result of either direct or indirect knowledge transfer, but the focus here is the cumulative

expression of knowledge relationships. There may or may not be direct actual knowledge

“flows” between a subunit and its knowledge sources, but rather it reflects the inevitable

connection between a subunit’s innovative activities and the sources of knowledge that

accumulated to enable these activities.

39

Within an MNC network, the knowledge sourcing includes connections between

subunits and MNC headquarter, and among subunits in the same MNC group. This type

of knowledge sourcing is by definition international; it reflects the way a multinational

corporation is organized, as well as the pattern of innovative activities in various subunits

within that MNC. Outside of an MNC network, a subunit’s knowledge sourcing could

include connections with local actors, and with any other external sources from

international locations.

3.2 Local and Remote Knowledge Sourcing

The conceptualization of international integration of MNC has been the center of debates

among IB researchers. Earlier findings pioneered by the “product life cycle” model

(Vernon, 1966) suggest that foreign subsidiaries mainly exploit proprietary advantages

abroad generated in home base by the parent firm and thus technological capabilities in

foreign countries were concentrated on the adaptation of products to meet the particular

needs of local market (Ronstadt, 1977; Behrman & Fischer, 1980; Hakanson & Nobel,

1993; Porter, 1990).

More recently, an increasing stream of resource-based and network-based

theoretical and empirical research on multinational firm challenge the traditional view on

MNC value creation. According to them, the MNC is not just an exploiter of home

country knowledge abroad, but a vehicle integrating knowledge from different parts of

word (Ghoshal & Bartlett, 1991; Birkinshaw & Hood, 1998). The emphasis on the

internationally integration strategies at the MNC group level has also shifted from the

40

conventional demand-side potential (as markets) to the supply side potential (innovation

and technology) (Cantwell, 1992, 1993).

In the new context, from Bartlett and Ghoshal (1989)’s suggestion on

“transnational’-form MNCs, to Hedlund (1994)’s proposition of “N-form” corporations,

an increasing strand of literature suggests that MNCs are incrementally sourcing and

generating knowledge from various locations (Pearce, 1989; Cantwell, 1995) and sharing

them across the organization. This knowledge sharing is facilitated by the emerging

ability of multinationals to integrate knowledge across related technologies and across

geographic boundaries (Zander, 2002). In this sense, FDI is interpreted as a mechanism

through which firms seek and develop capabilities on a global basis (Kogut & Chang,

1991; Teece, 1992).

In terms of managing the so called “internal network (intra-firm)” MNCs,

Hedlund and Ridderstrale (1995) observed a closer integration in the international value-

added activities of MNCs in the 1990s and afterwards, and pointed out that the

managerial challenge has shifted from controlling bilateral relations between one

technology creating headquarters and many implementing subsidiaries to a more complex

integrated network in which each units have their own creative capabilities. Zander and

Solvell (2002) also suggested that the integrated but increasingly heterogeneous

innovation structure encourages increasing reciprocity in inter-unit relations. This is

because to the extent that these fields of local knowledge accumulation are themselves

complex or systemic, they may need to draw on wider international sources as well for

41

the complementary branches of expertise required, and these supporting lines of

knowledge may either come from headquarters or from other entities of the parent group.

This integrated network allows the subsidiaries generate innovations based on the

stimuli and resources from host countries (Prahalad & Doz, 1987; Barlett & Ghoshal,

1989; Cantwell, 1992, 1993; Dunning, 1998; Nobel & Birkinshaw, 1998). This

international integration on the one hand can facilitate high subsidiary involvement in the

formulation and implementation of the company’s R&D strategy, and on the other hand

will minimize the “duplication of effort” problems (Hakanson & Zander, 1988).

The embeddedness of MNC subsidiaries in their local environment has been

uncovered and emphasized in the past few years (Uzzi, 1996, 1997; Birkinshaw et al.,

2002). Uzzi (1997) suggested that embeddedness is a logic of exchange that promotes

economies of time, integrative agreements, Pareto improvements in allocate efficiency,

and complex adaptation. Economic behaviors are considered to be embedded in the social

structure (Granovetter, 1985). Yet, the system embeddedness of knowledge, in turn, is

affecting organization structure. System embeddedness is the extent to which the

knowledge is affected by the system or context in which it is embedded. It emphasizes on

the social system (Birkinshaw, Nobel, & Ridderstrale, 2002). It is implied that some

knowledge is much more sensitive to its social and physical context than other

knowledge. For example, Birkinshaw et al. (2002) find out that Ericsson’s software

development is undertaken according to well-established procedures that are relatively

easily replicated in different settings; therefore the sensitivity to physical location and

social context is low. On the other hand, Alfa Laval’s R&D in milking machines and

42

separators has always been undertaken in specialized locations that have dedicated

physical infrastructure with functional activities nearby; hence their underlying

knowledge is highly sensitive to location, and would be very costly to move.

Andersson and Forsgren have completed a series of research on the subsidiary

embeddedness issue (Andersson & Forsgren, 1996; Andersson, Forsgren, & Holm, 2001;

Andersson, Bjorkman, & Forsgren, 2005). They define subsidiary embeddedness as the

extent to which individual relationships in the local market can serve as a source of

knowledge (Andersson et al, 2005). As the level of subsidiary embeddedness increases,

the potential of control from headquarter is higher as it attempts to integrate the

subsidiary better into the overall corporate strategy (Andersson & Forsgren, 1996,

Andersson et al, 2005). They argue that the external technical embeddedness is positively

affecting the subsidiary’s expected market performance, and thus increasing the

subsidiary’s importance for MNC competence development, using absorptive capacity

theory (Andersson et al., 2001).

Strategic alliances also facilitate external knowledge sourcing. Different types of

alliances ranging from “relational” contracts to equity joint venture all have impact on

organizational learning (Contractor & Lorange, 2002). Previous research looked at how

knowledge is managed in strategic alliances (Inkpen, 2002; Martin and Salomon, 2002),

how knowledge is transferred across partners (Mowery et al., 1996; Simonin, 2004), and

how knowledge is acquired from the parents by the joint venture itself (Lyles and Salk,

1996), with an emphasize on how organizations learn from their partners and develop

new competencies through their collaborative efforts (Inkpen, 2002).

43

3.3 Knowledge Sourcing Patterns and Subunit Innovation

When analyzing patterns of knowledge sourcing, there are mainly two dimensions that

can distinguish the composition of knowledge sources – technological areas and

geographical locations. The former looks at different areas of technologies that the

knowledge sources are composed from – whether they are focused on certain particular

fields or dispersed from different areas; the latter examines the geographical locations of

these knowledge sources – whether the sources are geographically concentrated or

scattered across the world. Technically speaking, these two dimensions are not

necessarily correlated to one another. However, on the one hand, the complexity of

technological accumulation usually induces a more proactive search across geographical

boundaries for relative or complementary knowledge sources, whereas on the other hand,

a cross-regional dispersed composition of knowledge sources could more than likely

bring in a wide variety of technological expertise. Therefore I propose that there is a

positive relationship between technological dispersion and geographical dispersion of

knowledge sources for any MNC subunit. The more geographically dispersed their

knowledge sourcing is, the more likely these knowledge sources would be wide spread

across technological fields; the more diverse their technological fields are, the more

likely they are sourcing from a wide range of geographical locations.

The relationship between technological and geographical dispersion of knowledge

sources, however, could be affected by the level of concentration on home country

sourcing or host country sourcing. High concentration on home country sourcing could

result in a relatively low geographical dispersion of knowledge sources, but as a subunit

44

seeking for knowledge sources from its parent company, the technological dispersion of

these sources could still be high if the parent company has a diversified range of

technological activities.

Hypothesis 1: There is a positive relationship between technological dispersion and

geographical dispersion of knowledge sources, but the relationship is moderated by level

of concentration in citation in home or host countries.

Technological dispersion of knowledge sources is determined by the nature of technology

that the focal subunit is working on; usually the more complex the technology is, the

more diverse of knowledge it needs to source from. The development of science and

technology determines that recombination and creation of new knowledge requires an

often cross-sector combination of existing knowledge sources. Although historical study

of technological changes has shown a relatively stable country-specific path dependency

of technology trajectory (Pavitt, 1982; Vertova, 1999), there are empirical evidences

demonstrating an increasingly diversified technological development along with a more

complex and complementary knowledge accumulation (Cantwell and Piscitello, 2000;

Cantwell and Vertova, 2004).

In comparison of the diversification of a subunit’s technological profile and its

knowledge source technological dispersion, I propose that there is likely to be a positive

relationship between those two. Although the causality can be argued in both directions,

this study focuses at the direction from source to outcome. Therefore I suggest:

45

Hypothesis 2: The technological diversification of an MNC subunit’s knowledge

innovation is positively related to the technological dispersion of its knowledge sourcing.

When we are examining the overall effect of technological dispersion of knowledge

sourcing on innovation diversification, one of the areas that we should pay attention to is

the composition of general purpose technology (GPT) in areas of specialization of a

subunit. GPT fields are characterized by the potential for pervasive use in a wide range of

sectors (Rosenberg, 1982; Bresnahan and Trajtenberg, 1992). Some technologies such as

non-electrical machinery (Rosenberg, 1976), instrument and controls, chemical processes,

and computing (Granstrand et, al., 1997) have been found actively mobilized in a wide

range of firms in different industries, and thus to belong to GPTs.

More recently, scholars further characterized the term “general purpose” with not only a

wide range of users, but also the technological cumulativeness, dynamism and

complementarity innovations (Bresnahan and Trajtenberg, 1995; Helpman and

Trajtenberg, 1998). In more recent years, a new ICT-based (Information and

Communication Technologies) technology paradigm has emerged (Granstrand, et al,

1992; Oskarsson, 1993; Patel and Pavitt, 1991). ICTs are then considered an advanced

type of GPTs (Cantwell and Santangelo, 2000). Firms as main actors in these fields tend

to reinforce the development of GPTs to support an even more widely dispersed network

of differentiated creativity; and research has shown a positive relationship between firm’s

involvement in GPT and the technological diversification and geographical

diversification (internationalization) of the firm’s innovation network (Qiu, 2011).

Therefore I hypothesize:

46

Hypothesis 2a: The relationship between technological diversification of an MNC

subunit’s knowledge innovation and technological dispersion of its knowledge sourcing is

positively moderated by the share of general purpose technology fields in the technology

developed by the subunit.

The second dimension to describe knowledge sourcing characteristic is its geographic

dispersion. Geographic dispersion is defined by the level of knowledge sources being

scattered about over a range of geographic locations. When examining the geographic

dispersion of knowledge sources, we are looking at subunits’ sourcing from both their

“internal” and “external” knowledge “networks”. Research shed light on geographic

factors and their influence on knowledge transfer in the past decade (Rosenkopf &

Almeida, 2003; Hansen & Lovas, 2004; Bell & Zaheer, 2007). There is a strong positive

link found between geographic proximity and the sharing of knowledge (Saxenian, 1994;

Allen, 1997); it is difficult to access and transfer knowledge from distant sources

comparing to local knowledge sources (Jaffe et al., 1993; Almeida and Kogut, 1999). In

this perspective, when we are examining the intensity of connections within a certain

knowledge network, the geographic distribution of participants in the network must be

taken into consideration. However, due to the different natures of an MNC subunit’s

“internal” and “external” networks, the geographic dispersion of connections in those two

networks can have different influences on the corresponding network connection

intensities. In the meanwhile, both the connection intensity and geographic dispersion can

be influenced by the industrial characteristics – namely the “pool” of accessible

knowledge in a certain field at a location.

47

Generally speaking, the geographic dispersion of knowledge connectedness

within an MNC largely is supposed to reflect the intensity of connection between

headquarter and subunit, or among different subunits. For a typical MNC subunit, a

geographically concentrated distribution of internal knowledge connections most likely

indicates the strong tie between itself and its headquarter, while a geographically

dispersed internal connection pattern demonstrates the possible even connection between

itself and different subunits in other locations, may or may not include the headquarter.

That being said, as a deduction from hypotheses one and two, one would expect a

geographically dispersed knowledge sourcing could more than likely result in a higher

level of technological diversification of subunit’s innovative activities. Therefore, we

have the following hypothesis:

Hypothesis 3: The technological diversification of an MNC subunit’s knowledge

innovation is positively influenced by the geographic dispersion of its knowledge

sourcing.

Although the sourcing of knowledge is known to be increasingly dispersed worldwide,

geographic proximity still plays an important role in determining the extent of knowledge

spillover. Research has shown that the externality of knowledge has the strongest effect

when both the sender and receiver (either intentionally or unintentionally) are

geographically proximate to one another (Jaffe et al., 1993; Almeida and Kogut, 1999).

Despite the advancement of transportation, communication and information technology,

it is still hard to engage in efficient technological knowledge transfers at a distance.

When geographic locations are treated equal to one another, it is difficult to capture the

48

effect of geographical distance on knowledge sourcing. Therefore we propose a

moderated model which distinguishes the geographic locations of knowledge sourcing by

treating intra-region and inter-region knowledge sources differently. Knowledge sourcing

from a country within the same geographic region indicates high geographical proximity,

while sourcing from a country outside the geographic region indicates low geographical

proximity.

Hypothesis 3a: The relationship between technological diversification of an MNC

subunit’s knowledge innovation and geographic dispersion of its knowledge sourcing

may be moderated by the general geographic proximity of knowledge sources to the

subunit.

The extent of external embeddedness of subunits to its host country (local environment)

could contribute to the composition of fields of technological specialization. As discussed

before, level of local embeddedness is shown to be related to MNC strategies such as

parental control, subunit mandate, subunit autonomy, etc. A highly embedded subunit is

expected to reflect in its technological profile the characteristics of the local technological

expertise. Although it may be intuitive to deduct that the level of technological

diversification of subunit is positively related to the level of sourcing dispersion in the

local environment, it is possible that this effect may be strong enough to suppress the

influence of knowledge sources from outside the host country – in a case where the local

embeddedness is at a high level. By controlling the effect of knowledge sourcing from

host country, we can then observe the international knowledge sourcing effect of a

subunit and its influence on the subunit’s technological diversification.

49

To note here, by excluding the effect of host country knowledge sourcing, we can also

find this as a proxy of the knowledge sourcing from internal sources of a subunit’s own

international multinational group, with the noise of sourcing from external sources that’s

not in the local context.

First, we hypothesize that there is a positive relationship between a subunit’s

technological diversification and its knowledge sourcing technological dispersion outside

its host country. The more dispersed into different fields its knowledge sources are, the

more likely it is to develop diversified technological innovation.

Hypothesis 4a: The technological dispersion of international knowledge sourcing from

outside the host country has a positive impact on the MNC subunit’s technological

diversification.

Then if we take a look at the geographic composition of knowledge sources outside the

host country, following the analysis of geographic dispersion and technological

diversification, it is expected that there is also a positive relationship between these two

variables. Excluding the host country effect, the more geographically dispersed a

subunit’s knowledge sources are, the more likely it is to develop diversified technological

innovation.

Hypothesis 4b: The geographical dispersion of international knowledge sourcing from

outside the host country has a positive impact on the MNC subunit’s technological

diversification.

50

Aside the effect of host country as a geographical location where subunits tend to heavily

source their knowledge from, home country location is another knowledge source that

should be taken into consideration. A subunit sourcing heavily from its home country (in

this case most likely from its parent company, which is the operation of its MNC group in

the home location) could in effect result in an overall technological and geographical

dispersion skewed towards its home country sourcing. It might actually be sourcing from

dispersed technological fields, but if the home country sourcing is mainly drawn from

certain fields, the effect of technological dispersion would be diminished; or, it might be

intentionally concentrating on seeking for sources from certain technological fields, but

due to dispersed sourcing effect from home country (many of the times a result of

sourcing from parent group that establishes the basis of the subunit’s technological

innovation which is intrinsically dispersed) the concentration effect could be diminished.

Therefore by excluding home country sourcing, we may be able to see a different result

showing the knowledge sourcing pattern of subunits. Hence I hypothesize:

Hypothesis 5a: The technological dispersion of external knowledge sourcing from outside

the home country has a positive impact on the MNC subunit’s technological

diversification.

Similarly, if a subunit has a large share of citation from its home country, it’s difficult to

determine whether its knowledge sources through the entire global knowledge base is

geographically concentrated or dispersed. As a lot of subunit tend to rely on its parent

company or home country basis of knowledge stock, it is understandable to observe a

relatively high share of home country originated knowledge sources. By excluding the

51

home country sourcing, we will then be able to capture the subunit’s effort of knowledge

sourcing from the rest of the world. The following hypothesis examines this effect:

Hypothesis 5b: The geographical dispersion of external knowledge sourcing from outside

the home country has a positive impact on the MNC subunit’s technological

diversification.

3.4 Framework

Figure 6 demonstrates a framework for this part of my study.

Figure 6 Framework of Study 1 - Knowledge sourcing patterns

3.5 Methodology

The study uses step-wise regression models to analyze the relationship between

technological and geographical dispersions of knowledge sources, and their impact on

subunit technological diversification. Data is constructed at MNC subunit level, including

US patents of 157 large chemical industry firms granted between 1976 and 2006. In order

to minimize the effect of small number problem, a five-year window is constructed to

aggregate patent numbers into a higher level, creating data panels through the 31-year

period.

52

Key Variables

Technological Diversification – The extent to which a subunit’s innovative activities

(patents) are wide spread across various technological fields (see Chapter 2 for detailed

definition).

Technological Dispersion – The extent to which knowledge sources is composited from

various technological fields (see Chapter 2 for detailed definition).

Geographical Dispersion – The extent to which knowledge sources is composited from

various geographical locations (see Chapter 2 for detailed definition).

Non-host Tech Disp – Technological dispersion of knowledge sources excluding those

from host country.

Non-host Geo Disp – Geographical dispersion of knowledge sources excluding those

from host country.

Non-home Tech Disp – Technological dispersion of knowledge sources excluding those

from home country.

Non-home Geo Disp – Geographical dispersion of knowledge sources excluding those

from home country.

GPT Share – Share of technological innovation that falls into the category of GPT fields.

Inter Region Share – Share of knowledge sources that are from a different geographical

region than the host country of subunit.

53

Moderators and Control Variables

Home Share – Share of knowledge sources from home country of the MNC group.

Host Share – Share of knowledge sources from host country of the subunit.

Host Country Tech Div – Degree of technological diversification of the host country

where the subunit is located.

US-Home – Dummy variable indicating whether the MNC firm is US originated.

US-Host – Dummy variable indicating whether the subunit is located in US.

The table of correlation of key variables is shown in Appendix B.

3.6 Results and Discussion

Hypothesis 1 examines the relationship between technological dispersion and

geographical dispersion of subunit knowledge sources, with the effect being moderated

by shares of home-country or host-country sources. Three separate sets of models are

analyzed for hypothesis 1 to capture the effect of (1) all knowledge sources moderated by

home country sourcing share; (2) knowledge sources outside the host country moderated

by host country sourcing share; (3) knowledge sources outside the home country

moderated by home country sourcing share. The regression results of these three sets of

models are shown in Tables 6, 7 and 8:

Table 6 Hypothesis 1 Model with all data

Tech Dispersion 1

2

3 Geo Dispersion 0.51 *** 0.481 *** 0.408 ***

Home Share

0.064 *** 0.009

54

Geo Dispersion X Home Share

0.115 *** _cons 0.251 *** 0.248 *** 0.289 ***

R-sq 0.254

0.17

0.173 chi2 5741.7 *** 2771.2 *** 2792.7 ***

Table 7 Hypothesis 1 Model without host country sources

Non-host Tech Disp 4

5

6 Non-host Geo Disp 0.45 *** 0.464 *** 0.07 ***

Host Share

-0.139 *** -0.487 *** Non-host Geo Disp X Host Share

0.762 ***

_cons 0.215 *** 0.006 *** 0.46 ***

R-sq 0.19

0.216

0.26 chi2 4462.3 *** 3738.61 *** 4581.2 ***

Table 8 Hypothesis 1 Model without home country sources

Non-home Tech Disp 7

8

9 Non-home Geo Disp 0.468 *** 0.471 *** 0.088 ***

Home Share

-0.176 *** -0.491 *** Non-home Geo Disp X Home Share

0.701 ***

_cons 0.19 *** 0.275 *** 0.45 ***

R-sq 0.2242

0.276

0.306 chi2 4752.2 *** 4309.01 *** 5011.8 ***

From these results we can see that there is indeed a positive relationship between

technological dispersion and geographical dispersion of knowledge sources across all

samples of data. This positive relationship, however, is reduced when a partial sample is

taken – the coefficients shown in model 4 and model 7 as two main effects are lower than

the coefficient in model 1. In the all-data sample model (Table 6 Model 3), a positive

55

effect of share of home country sources is demonstrated. It shows that when there is a

large share of home country sourcing, there is a stronger positive relationship between

technological dispersion and geographical dispersion – which is consistent with the result

in the sample excluding home country sourcing (Table 8 Model 9), the coefficient

becomes smaller in this model, and when the share of home country sourcing is taken

into consideration, it negatively affects the relationship between technological dispersion

and geographical dispersion. The result in Table 7 (Model 6) tells a similar story, except

this time the location excluded from the sample is on host country.

Hypothesis 2 to 5 examines the relationship between subunit’s technological

diversification and its knowledge sourcing patterns. The results (shown in Table 9)

support hypotheses 2, 3, 4a, 4b, 5a, and 5b. However, some interesting results are found

in the test of hypotheses 2a and 3a. There is a negative moderating effect of GPT share on

the relationship between subunit technological diversification and technological

dispersion of knowledge sources. Although GPT share itself is positively influencing

technological diversification (which is consistent with the argument of general purpose

technology), the interaction term of technological dispersion of sourcing and subunit GPT

share shows a negative coefficient. It indicates that subunits with a high concentration on

innovation in GPT fields are themselves intrinsically diversified due to the effect of

investment on GPT, then the effect of overall knowledge sourcing technological

dispersion is not that significant given the case.

56

Table 9 Results for H2-H5, Study 1

Tech Div H2 H2a H2a H3 H3a H3a H4a H4b H5a H5bUS_home -0.74 *** -181 *** -0.08 *** -0.073 *** -0.073 *** -0.079 *** -0.08 *** -0.08 *** -0.073 *** -0.075 *** -0.072 *** -0.076 ***US_host 0.115 *** 0.178 *** 0.101 *** 0.099 *** 0.99 *** 0.11 *** 0.111 *** 0.11 *** 0.11 *** 0.11 *** 0.11 *** 0.11 ***Host Share 0.005 -0.024 0.022 0.026 ** 0.024 ** 0.033 *** 0.004 0.009 0.019 ** 0.01 0.015 0.02Home Share -0.004 0.87 *** 0.009 0.009 0.009 0.028 ** 0.026 * 0.026 * 0.001 0.008 0.007 -0.001_cons 0.685 *** 0.514 0.63 *** 0.56 *** 0.539 *** 0.622 *** 0.656 *** 0.647 *** 0.656 *** 0.669 *** 0.655 *** 0.669 ***Host Tech Div 0.81Tech Disp 0.088 *** 0.088 *** 0.13 ***GPT Share 0.143 *** 0.187 ***Tech Disp X GPT Share -0.087 ***Geo Disp 0.075 *** 0.063 *** 0.079 ***Inter Region Share -0.03 *** 0.001Geo Disp X Inter Region Share -0.06 *Non-host Tech Disp 0.05 ***Non-host Geo Disp 0.025 ***Non-home Tech Disp 0.053 ***Non-home Geo Disp 0.026 ***

R-sq 0.07 0.311 0.11 0.153 0.153 0.078 0.077 0.077 0.091 0.074 0.093 0.075chi2 321.56 *** 490.1 *** 588.4 *** 989.7 *** 1004 *** 394.2 *** 406.9 *** 413.6 *** 447.96 *** 346.5 *** 464.43 *** 349.6 ***

57

For geographical dispersion, the direct effect on technological diversification is a positive

one, just as proposed in hypothesis 3. However, the effect of share of inter-regional

knowledge sourcing is shown to have a negative impact on technological diversification.

The rationale behind this is that when subunits are sourcing their knowledge from a wide

range of geographical locations, they tend to be more concentrated on the targeting

technological fields. In other words, when subunits have a higher level of concentration

on their specialized technological fields, rather than being diversified into various fields,

they tend to source their knowledge more focused and more purposely from the entire

knowledge base around the globe, instead of being geographically bounded. Whereas

when subunits focuses more on knowledge sources from geographically proximate

locations, they tend to draw on all the potential areas of expertise from these locations,

hence more likely to develop a technologically diversified profile of innovation activities.

58

4. Competence Exploiting and Competence Creating

4.1 Competence Exploiting vs. Competence Creating

The theory of organizational capability represents an extension and synthesis of the

contributions from the knowledge based view of organizations, based upon the idea that

the essence of organizational capability is the integration of individuals’ specialized

knowledge (Grant, 1996).

The notion of “exploitation” and “exploration” was first brought into attention in

organizational learning theories (March, 1991). The distinction between the two draws on

March’s (1991, p. 85) view of exploitation as “the refinement and extension of existing

competencies, technologies, and paradigms” and exploration as “experimentation with

new alternatives that have returns that are uncertain, distant, and often negative.” Based

on the distinction between the two concepts, studies of a wide range of management

fields have shown that exploitation and exploration require substantially different

structures, processes, strategies, capabilities, and cultures to pursue and may have

different impacts on firm adaptation and performance (He & Wong, 2004). In general,

exploration is associated with organic structures, loosely coupled systems, path breaking,

improvisation, autonomy and chaos, and emerging markets and technologies.

Exploitation is associated with mechanistic structures, tightly coupled systems, path

dependence, routinization, control and bureaucracy, and stable markets and technologies

(Ancona et al., 2001; Brown and Eisenhardt, 1998; Lewin et al., 1999).

59

A firm’s competence refers to the knowledge, skills, and related routines that

constitute its ability to create and deliver superior customer value (Day 1994).

Competence exploitation refers to the tendency of a firm to invest resources to refine and

extend its existing product innovation knowledge, skills, and processes, which aims to

greater efficiency and reliability of existing innovation activities; while in contrast,

competence exploration refers to the tendency of a firm to invest resources to acquire

entirely new knowledge, skills, and processes, with the objective of attaining flexibility

and novelty in product innovation through increased variation and experimentation

(March, 1991; Atuahene-Gima, 2005).

When an MNC establishes a foreign subsidiary of a resource-seeking or market-

seeking kind, the initial effort in the traditional view of foreign direct investment (FDI) is

to exploit the competence that’s already established by the parent group in the home

country base. Technological knowledge required in the foreign production of this kind is

mostly created at home, and then transferred via the internal international corporate

network to foreign production locations (Cantwell & Piscitello, 2000; Cantwell &

Mudambi, 2005). As the increasing complexity of technology and internationalization of

production development brought attention to the competence-seeking or knowledge-

seeking activities of foreign investment, the competence-based theory of firm treats firm

as an institution where capabilities are constructed through continuous internal learning;

the major issue of a firm is then no longer the exploitation of established competence, but

rather how to create new competence in a geographically dispersed and technologically

60

diversified system leveraging not only the MNC internal resources but also the

technology base of the host location (Cantwell, 1991; Cantwell & Piscitello, 2000).

4.2 The Relationship between Knowledge Sourcing Patterns and Competence Creating

The types of innovative activities – whether they are competence creating (CC) or

competence exploiting (CE) – that an MNC subunit conducts can be determined mainly

by several factors – the parent group pressure, the locational pressure, or the subunit

autonomy. MNC parent group makes the initial definition of subunit strategic position –

the purpose of establishing a certain subunit in the specific market can be local market

oriented, local resource (natural or labor) oriented or local competence oriented. The first

two types of subunits are designed mainly to exploit the established parent group

competence base in a new location, while the third type seeks to create new competence

that contributes back to the MNC group. Locational characteristics such as industry

expertise, high quality human resource availability, and policy to FDI play an important

part in such choice making of an MNC group. Because the creation of competence

demonstrates the capability of a subunit to innovate into new profitable technological

fields, which in turn earns the subunit better position within the MNC network, and hence

more favorable resource allocation by the parent group, subunit management are

encouraged to seek to achieve a mandate of CC by initiating local knowledge seeking or

new competence creating (Cantwell & Mudambi, 2005).

In order for subunits to achieve a CC mandate, they need to first have a strong

capability of innovation per se; then these innovative activities need to be focused on

61

exploring new inventions that are novel to the parent MNC group. These activities

require knowledge connection and recombination from different sources – from the

parent group to build a strong innovation base, and from the external sources to explore

new areas of innovative capability. Since MNC subunits themselves can be consist of a

wide variety of operations, it is difficult to define a subunit clearly as a CC or CE type.

For instance, a subunit with a CE mandate can establish competence in certain areas that

is new to the parent group – even that isn’t the core competence or that subunit; on the

contrary, a subunit with a CC role can also manage some operations that are generally

based on parent group competence, with little creative inventions in areas new to the

home base. It is more appropriate then to examine the individual innovative activities of a

subunit, instead of treating all things as a whole in one large overseas operation. On the

subunit level, we can then look at the CC activity intensity (the share of competence

creating activities among all innovative activities within a subunit) and CC activity distance

(the technological distance between competence creating activities of a subunit and the closest

areas of expertise of its parent group).

The intensity of CC activity has much to do with the knowledge sourcing and

recombination from all possible sources. High embeddedness in either the MNC

international network or the external local network (or even international) both facilitates

the subunit’s capability of creating new competence. Especially when the embeddedness

in both networks are high, the subunit can have more possible knowledge sources to learn

and recombine, which means higher potential to create new competence. This reflected

62

on the abstract level of network means high knowledge connectedness from either the

internal sources or the external sources can both promote CC activity intensity.

On the level of individual technological fields, the extent to which a subunit’s

specialization in that field is different from the specialization degree of its parent

company can be captured by technological distinctiveness, as introduced in Chapter 2.

This concept takes a further step from simply identifying a subunit’s innovative activities

in one field to be CC or CE, by quantifying how much is the subunit’s degree of

specialization different from that of its parent company. Technological distinctiveness is

an index that demonstrates the composition of CC or CE activities comparing the subunit

and its parent company.

The factors that can influence the degree of technological distinctiveness, and

therefore influence the competence creativeness of a subunit, can be various. Externally,

the mandate from parent company and the technological influence from the host country

environment could have an impact on the subunit’s technological distinctiveness in

certain fields; internally, the management autonomy, the strategy chosen, and the

intellectual capability of human resource could impact a subunit’s technological

distinctiveness in certain fields. Needless to say the parent company’s technological

expertise plays a very important role in defining the subunit’s “distinct” areas of expertise.

Since this concept is a comparison between the levels of specialization of the two entities,

simply demonstrating a strong specialization in a certain field doesn’t necessarily indicate

that the subunit has a high technological distinctiveness in this field. If the level of

63

specialization of subunit and parent are both high, the technological distinctiveness of the

subunit in this field might actually be relatively low.

From the MNC’s point of view, the structure of innovation network for the MNC

group and the positioning of a subunit in this network could be two of the important

factors demonstrating the mandate that’s assigned to the subunit. Structural concentration

of the innovation network usually results in a concentration of innovative activities in the

home bases. This form of MNC structure is consistent with the traditional views of

multinational firms, where research and innovations are mostly initiated from the parent

company, and then transferred to other parts of the international group. Within this type

of structure, foreign subunits are usually conducting incremental innovative activities

around the areas of expertise of the parent company; exploration into new areas of

technology is not supported by the innovation network structure, therefore I propose:

Hypothesis 1a: There is a negative relationship between an MNC’s structural

concentration level of innovative activity and the subunit’s degree of technological

distinctiveness from its parent company.

However, this negative effect could be changed of a subunit has a relatively high share of

innovative activity among all international locations of its MNC group. That is to say,

even if there is a high level of geographical concentration on innovative activities for an

MNC, if the concentration is not limited in the MNC home location (parent company),

but other foreign locations also play an important role in the innovation network, then

there could be a positive relationship between the degree of geographical concentration

64

and the technological distinctiveness of subunit in certain fields. In the meanwhile, if a

subunit is one of the most important innovative locations for its MNC group, it is more

likely that this subunit develops a strong basis of innovation capability, then it is more

likely to diversify its technological expertise into some fields that are not necessarily the

fields of specialization for the parent company. Therefore I propose the following

hypothesis:

Hypothesis 1b: There is a positive relationship between a subunit’s share of innovative

activity in its international corporate group and its degree of technological

distinctiveness from its parent company.

The relationship between a subunit’s knowledge sourcing pattern and technological

distinctiveness could be a mutually interdependent one. It is hard to argue whether it is

because a subunit’s mandate from its parent company that strategized it to perform a

competence creating role, then the subunit focuses its knowledge sourcing into dispersed

technological fields and geographical locations to facilitate its diversification into new

and novel technological fields; or it could be that the subunit has access to dispersed

sources of knowledge due to location specific advantages and human resource factors,

then these readily available knowledge sources could reinforce its specialization in

certain technological fields that are not inherited from specialized field of its parent

company. Although this causality could be argued either way, it is believed that there is a

relationship between a subunit’s knowledge sourcing pattern and the composition of its

technological specialization.

65

As we discussed in Chapter three, the knowledge sourcing pattern of a subunit can be

described by two dimensions – the technological dispersion and geographical dispersion

of its knowledge sources. The former looks at different areas of technologies that the

knowledge sources are composed from – whether they are focused on certain particular

fields or dispersed from different areas; the latter examines the geographical locations of

these knowledge sources – whether the sources are geographically concentrated or

scattered across the world. If a subunit is sourcing from dispersed areas of technological

fields, it is more likely to develop innovative activities in a diversified range of

technology fields (as the result demonstrated in Chapter Three). Among these fields, it is

then more likely to create a new area of expertise that’s different from that of its parent

company. Therefore I hypothesize:

Hypothesis 2a: There is a positive relationship between a subunit’s technological

dispersion of its knowledge sources and its degree of technological distinctiveness from

its parent company.

Similarly, if a subunit has a wide range of knowledge sources across various geographical

locations, it is more likely to develop a higher degree of technological diversification.

Therefore, higher level of geographical dispersion of knowledge sourcing in a certain

field could result in a higher possibility of technological distinctiveness in that field.

Again, I hypothesize a positive relationship between these two:

66

Hypothesis 2b: There is a positive relationship between a subunit’s geographical

dispersion of its knowledge sources and its degree of technological distinctiveness from

its parent company.

The next set of hypotheses is drawn upon the notion of a subunit’s knowledge sourcing

pattern being influenced on its integration with its home country or its embeddedness

with its host country.

It is arguable that high embeddedness in host country could many times result in a

subunit diversifying its technology into areas of expertise of the host country location,

which are not always consistent with its parent company’s areas of expertise. Therefore,

if a subunit’s knowledge sourcing is heavily concentrated on its host country location, it

could dilute the relationship between knowledge sourcing pattern and technological

distinctiveness of the subunit. Therefore I exclude the knowledge sources from subunits’

host countries, and propose the following hypotheses with the expectation of stronger

positive relationships than those in hypotheses 2a and 2b:

Hypothesis 3a: Excluding host country sourcing, there is a positive relationship between

a subunit’s technological dispersion of its knowledge sources and its degree of

technological distinctiveness from its parent company.

Hypothesis 3b: Excluding host country sourcing, there is a positive relationship between

a subunit’s geographical dispersion of its knowledge sources and degree of technological

distinctiveness from its parent company.

67

If a subunit is concentrating its knowledge sourcing heavily from its home country,

especially in the cases where the level of MNC integration is high, it is more likely that

the subunit relies on its home country / parent company technological expertise; hence it

is less likely to develop innovative activities into areas that are distinct from its parent

company. More specifically, the share of subunit concentration on home country sources

should have a negative impact on its technological distinctiveness (tested in model as

control variable). When the sources from home country is excluded from the sample, or

when we focus on knowledge sources outside a subunit’s home country, we should

therefore expect a stronger positive relationship between its technological dispersion of

knowledge sources and degree of technological distinctiveness from its parent company,

and a stronger positive relationship between its geographical dispersion of knowledge

sources and the degree of technological distinctiveness from its parent company. The

corresponding hypotheses are:

Hypothesis 4a: Excluding home country sourcing, there is a positive relationship between

a subunit’s technological dispersion of its knowledge sources and degree of

technological distinctiveness from its parent company.

Hypothesis 4b: Excluding home country sourcing, there is a positive relationship between

a subunit’s geographical dispersion of its knowledge sources and degree of technological

distinctiveness from its parent company.

4.3 Framework

The following framework in Figure 7 summarizes the conceptual model of Study 2.

68

4.4 Methodology

The study uses step-wise GLS regression models. Data is constructed at MNC subunit

level, including US patents of 147 large chemical industry firms granted between 1976

and 2006. In order to minimize the effect of small number problem, a five-year window

is constructed to aggregate patent numbers into a higher level, creating data panels

through the 31-year period. The measurements of key variables are listed as follows:

Dependent Variable:

Technological Distinctiveness

Technological Dispersion

MNC Innovation Concentration

Geographical Dispersion

Subunit Share of Innovation

Firm Structure

Subunit Knowledge Sourcing Pattern

Home Country Sourcing

Host Country Sourcing

H1

H2

H3 H4

Figure 7 Conceptual Framework of Study 2

69

Technological Distinctiveness – The extent to which a subunit’s technological

specialization in a given field is different from that of its parent company (see Chapter

Two for detailed definition and measurement description).

Independent Variables:

Firm Innovation Concentration (FirmCon) – MNC group’s level of concentration of its

innovative activities. This variable is measured with similar mechanism from the HHI

index of concentration, by aggregating the square of share of innovative activities of all

subunits in an MNC group. If 𝑃𝑃𝑖𝑖 indicates the number of patents in subunit i, then the

formula for FirmCon is:

𝐹𝐹𝑖𝑖𝑃𝑃𝐹𝐹𝐹𝐹𝐹𝐹𝑛𝑛 = � �𝑃𝑃𝑖𝑖∑ 𝑃𝑃𝑖𝑖𝑖𝑖

�2

𝑖𝑖

Subunit Share (SubShare) – share of a subunit’s patent within its MNC group.

Technological Dispersion – The extent to which knowledge sources is composited from

various technological fields (see Chapter Two for detailed definition and measurement

description).

Geographical Dispersion – The extent to which knowledge sources is composited from

various geographical locations (see Chapter Two for detailed definition and measurement

description).

Non-host Tech Disp – Technological dispersion of knowledge sources excluding those

from host country.

70

Non-host Geo Disp – Geographical dispersion of knowledge sources excluding those

from host country.

Non-home Tech Disp – Technological dispersion of knowledge sources excluding those

from home country.

Non-home Geo Disp – Geographical dispersion of knowledge sources excluding those

from home country.

Control Variables:

GPT Share – Share of technological innovation that falls into the category of GPT fields.

Inter Region Share – Share of knowledge sources that are from a different geographical

region than the host country of subunit.

Home Share – Share of knowledge sources from home country of the MNC group.

Host Share – Share of knowledge sources from host country of the subunit.

US-Home – Dummy variable indicating whether the MNC firm is US originated.

US-Host – Dummy variable indicating whether the subunit is located in US.

RTA Host – The revealed technological advantage of a subunit’s host country in a given

field. This variable controls for the host country’s technological specialization effect on

subunit technological distinctiveness.

71

Host Type – Dummy variable indicating type of host country: developed country or

developing country.

Firm size – Total number of patents of a given MNC. The variable of firm size uses the

log result of patent number.

Industry – Dummy variables designed to subdivide the general chemical industry into

subcategories based on a MNC group’s main activity: chemical, pharmaceutical and

biology (phar), petroleum and energy (petr).

4.5 Results and Discussion

The results of Study Two are shown in the following Table 10. Most of the models are

shown to be fit, while a few models are indicating a relationship between the independent

variable and dependent variable that’s in the opposite direction as proposed in my

hypotheses. In order to minimize the effect of small number problem, as well as to test

the robustness of the methodology, I then selected the sample with several cut-off points

of subunit patent number. Table 11 reports the results of analysis with a cut-off point of

50, which means all subunits in this sample has more than 50 patents in the

corresponding five-year window.

72

Table 10 Results for Study 2 - All Data

1 2 3 4 5 6 7 8 9

US_home 0.031 0.031 -0.044 * -0.057 ** -0.031 -0.049 ** -0.043 * -0.049 ** -0.046 *

US_host 0.011

0.01

-0.033

-0.06 ** -0.023

-0.044

-0.031

-0.047 * -0.035 hs_share -0.069 * -0.08 * -0.043

0.012

-0.114 ** -0.024

-0.048

-0.031

-0.035

hm_share -0.14 *** -0.125 *** -0.143 *** -0.116 *** -0.207 *** -0.136 *** -0.15 *** -0.121 *** -0.139 ***

RTA_host 0.013 *** 0.013 *** 0.013 *** 0.014 *** 0.013 *** 0.013 *** 0.013 *** 0.013 *** 0.013 ***

host_type -0.065

-0.061

-0.068

-0.073

-0.069

-0.074

-0.068

-0.08

-0.068 log size -0.109 *** -0.108 *** -0.099 *** -0.105 *** -0.1 *** -0.107 *** -0.099 *** -0.106 *** -0.1 ***

phar -0.036 ** -0.029

-0.014

-0.005

-0.015

-0.008

-0.014

-0.008

-0.014 petr 0.038

0.037

0.008

0.012

0.002

0.008

0.008

0.008

0.008

gpt_share

0.083 *** 0.052 ** 0.054 ** 0.052 ** 0.054 ** 0.052 ** 0.05 ** 0.052 **

intreg_share

-0.02

0.01

0.034

-0.008

0.003

0.01

0.001

0.01 _cons 0.878 *** 0.844 *** 0.665 *** 0.574 *** 0.78 *** 0.634 *** 0.671 *** 0.637 *** 0.658 ***

struct

0.299 *** 0.319 *** 0.3 *** 0.327 *** 0.298 *** 0.327 *** 0.302 ***

sub_share

0.003 *** 0.002 *** 0.003 *** 0.002 *** 0.003 *** 0.003 *** 0.003 ***

ced_td

0.138 *** ced_gd

-0.115 **

nhsced_td

0.086 *** nhsced_gd

-0.008

nhmced_td

0.093 *** nhmced_gd

0.01

R sq 0.047

0.052

0.092

0.1

0.093

0.094

0.092

0.095

0.091

Chi2 131.61 *** 146.59 *** 239.91 *** 271.71 *** 245.19 *** 255.51 *** 240.14 *** 258.05 *** 240.01 *** * p < .10

** p < .05 *** p < .01

73

Table 11 Results for Study 2 - Large Subunit Sample

1 2 3 4 5 6 7 8 9

US_home -0.083 * -0.055 -0.138 *** -0.15 *** -0.087 -0.139 *** -0.114 ** -0.139 *** -0.126 **

US_host -0.073

-0.064

-0.087 * -0.102 ** -0.044

-0.087 * -0.053

-0.091 * -0.075 hs_share 0.085

0.138 * 0.161 ** 0.2 *** -0.079

0.161 ** 0.109

0.154 ** 0.104 *

hm_share -0.067

-0.117

-0.116

-0.088 ** -0.299 *** -0.109

-0.224 ** -0.09

-0.155 *

RTA_host 0.024 ** 0.025 *** 0.025 *** 0.027 *** 0.027 *** 0.026 *** 0.026 *** 0.026 *** 0.026 ***

log size -0.129 *** -0.119 *** -0.098 *** -0.105 *** -0.098 *** -0.106 *** -0.092 *** -0.109 *** -0.094 ***

phar 0.074 ** 0.088 *** 0.094 *** 0.097 *** 0.092 *** 0.097 *** 0.093 *** 0.1 *** 0.092 ***

petr 0.096 ** 0.099 ** 0.045

0.052

0.024

0.042

0.058

0.042

0.052 gpt_share

0.171 *** 0.06

0.046

0.076

0.057

0.051

0.052

0.052 *

intreg_share

0.116 * 0.137 ** 0.16 ** 0.094

0.127 * 0.137 ** 0.12 * 0.152 **

_cons 0.788 *** 0.598 *** 0.458 *** 0.371 *** 0.77 *** 0.437 *** 0.563 *** 0.407 *** 0.512 *** struct

0.248 *** 0.27 *** 0.237 *** 0.271 *** 0.217 *** 0.287 *** 0.231 ***

sub_share

0.003 *** 0.003 *** 0.003 *** 0.003 *** 0.003 *** 0.003 *** 0.003 ***

ced_td

0.134 *** ced_gd

-0.32 ***

nhsced_td

0.067 nhsced_gd

-0.145 **

nhmced_td

0.407 *** nhmced_gd

-0.078

R sq 0.052

0.074

0.134

0.139

0.143

0.134

0.149

0.138

0.139 Chi2 63.48 *** 77.48 *** 137.19 *** 144.21 *** 151.1 *** 139.76 *** 146.84 *** 146.23 *** 140.37 ***

* p < .10 ** p < .05 *** p < .01

74

Most of the main effects are shown to be consistent between Table 10 and Table 11.

Hypothesis 1a is not supported by the data analysis in both samples (Model 3 in both

samples). In fact, there is a consistent strong positive relationship between corporate

innovation structural concentration and subunit’s technological distinctiveness, instead of

the negative one proposed in previous analysis. There are two possible explanations to

this result: (1) the firm’s structural concentration could be partially focused on the focal

subunit. In that case the subunit would demonstrate a high share of innovation in the

MNC group. (2) In some cases, if a firm has highly concentrated innovation structure, it

is therefore focusing its core areas of technology development in these concentrated

locations (usually the home country base), whereas the technological development in

other foreign locations could be diversified into various areas that are not the parent

company’s fields of specialization. As a result the technological distinctiveness of those

subunits that are conducting some level of R&D outside of the home country focus their

efforts on developing new areas of technological innovations for the MNC corporation.

The first explanation could be further tested with an examination of interaction effect

between the MNC group’s innovation structural concentration and the subunit’s share of

innovation within the group. The second explanation needs to examine the parent

company’s share of innovation to confirm that the concentration is indeed within the

home country base.

Hypotheses 2a, 3a, and 4a examine the relationship between technological dispersion of

knowledge sources and subunit’s technological distinctiveness. All of these hypotheses

are supported in the sample with all data (Models 4, 6, and 8), but the story of the large

75

subunit sample is a different one. Hypotheses 2a and 4a by Models 4 and 8 are supported,

while Model 6 shows insignificant result, which means considering knowledge sources

outside of host country, there isn’t a significant positive relationship between knowledge

source technological dispersion and the subunit’s technological distinctiveness compared

to its parent company. In other words, for large subunits, their knowledge sources’

diversified contributions to their competence creativeness mainly come from the sources

within their host countries. Another implication behind these results is that large subunits

tend to be more diversified in their technological fields than smaller subunits, therefore

the capability of them in developing novel areas of specialization is not that strongly

dependent on their knowledge sources’ technological dispersion. To put it another way,

when the level of technological diversification is constantly high, from the result of study

1 we can expect a similarly high technological dispersion level, and as these levels get

higher their numbers shown in the data gets closer to 1, resulting in very small variation

among observations. Therefore it’s difficult to capture the relationship between

technological dispersion and technological distinctiveness among large subunits. (2) The

insignificant result in Model 6 may indicate that for large subunits, the knowledge

sources from host country location play a significant role in building up their capacity of

developing novel innovations to the MNC group. When these sources are taken out of the

sample, the effect between the international sources’ technological dispersion and the

subunit’s technological distinctiveness becomes insignificant. This means that the

innovative activities that rely on knowledge sourcing around the globe are actually more

76

concentrated on enhancing the areas of expertise that are inherited from the parent

company.

Hypotheses 2b, 3b, and 4b, examine the relationship between a subunit’s knowledge

sourcing geographical dispersion and its technological distinctiveness. It is shown in both

samples that Model 5 is significant with a negative coefficient. This result is inconsistent

with the hypotheses, where I proposed a positive relationship between the two.

Interestingly, this is consistent with the results shown in the previously discussed models

for technological dispersion. Generally speaking, the more geographically dispersed of a

subunit’s knowledge sources, the less likely it is to develop a high level of technological

distinctiveness in that area. This explanation is in fact consistent with our previous

discussion about local embeddedness and competence creating. In the case of a highly

locally embedded subunit, especially in regions with high specialization of technological

fields, the subunit would show a more geographically concentrated pattern of knowledge

sourcing, and is more likely to develop technological specialization in fields that are not

its parent company’s expertise. With this understanding, it is then easier to understand the

unsupported results in Models 7 and 9 for Hypotheses 3b and 4b.

Note that the control variable of firm size shows a consistent negative effect on

technological distinctiveness of subunit. It is because the larger the firms are, the more

likely it is technologically diversified into a wide range of fields, in extreme cases the

number would be very close to 1. There for in such cases it is very difficult for subunits

to specialize in innovative activities that are not in areas of expertise of the MNC group.

Imagine the most extreme case – if the MNC group is so large that it has high level of

77

RTA in all fields that its subunits could possibly develop; then it would be extremely

hard for its subunits to bring in something new to the group!

Another notion is on the control variable of host country RTA. In both samples, there is a

consistently significant positive effect on subunit technological distinctiveness. The

explanation is that if a host country is specialized in a certain technological field, it’s

more likely for the subunit to learn from the host country and develop an expertise in this

field – this again fits the understanding of the importance of local embeddedness.

There are two more different effects between the all-data-sample and large subunit

sample. First, in large subunit sample, those in pharmaceutical industry tend to be more

likely to develop technological distinctiveness compared to others; however the same

effect isn’t found in all-data-sample. Note that our sample size selection is based on

patenting numbers during the given time period. One could therefore explain that for

large subunits, pharmaceutical companies tend to emphasize more on competence

creating, compared to companies in other sectors. However this effect can be substituted

if the sample is allowing smaller subunits from other industry to drive up their average

distinctiveness based on concentrated innovation efforts in areas away from their parent

companies. Second, the share of patents in GPT fields has a positive impact on

technological distinctiveness in the all-data-sample, but such effect seems to be canceled

out in the large subunit sample. The former supports the general argument of GPT fields,

it increases the possibility of firms to develop novel innovation. Therefore, a positive

relationship is shown. However, for large subunits, like discussed in the end of Chapter

Three, because they are likely to be intrinsically diversified due to their sizes, having a

78

high share of patents in GPT fields doesn’t have as much of an impact on its

distinctiveness.

These results in effect demonstrated that the capability of subunit to have fields of

technological distinctiveness is to some extent locally grounded. One of the important

indicators is a contrast between a subunit’s home and host country’s technological

environment – whether or not the subunit’s host country can provide them specialized

technology in some fields that the home country can’t provide. Hence a subunit’s

technological distinctiveness in some ways reflects the local technological distinctiveness

of host country location vis-à-vis the home country location.

79

5. The Evolution of Subunit Roles

This chapter looks at the evolution paths of subunit roles more closely focusing on the

heterogeneity of firm-specific evolutionary paths in the patterns of knowledge accumulation that

support CC activities, controlling for the industry-specific determinants of such technological

trajectories.

5.1 Typology Based on Knowledge Sourcing Pattern

Subunit evolutions are driven by both internal (initiated by subunits themselves) and

external factors (investment from parent company or other external forces) (Birkinshaw

& Hood, 1998). External forces largely shape the options of subunit, while it is the

subunit managers to take the initiative to respond to the external opportunities. The idea

that multinational subsidiaries are differentiated according to their technological

capabilities and roles can be traced back to Ghoshal and Bartlett’s study (1998) on

subunit tasks. They argued that the tasks of affiliates can be classified in three categories:

creation, adoption and diffusion. ‘Creation’ is to use subunit’s own technical and

managerial resources to respond to local circumstances; ‘adoption’ is to adopt innovation

developed by parent company or a central R&D facility, or other national subsidiaries of

the firm; and ‘diffusion’ is to diffuse their local innovations back to the parent company

or to other subsidiaries. Similarly, Pearce (1999) characterized MNC subsidiaries as

world product mandate and regional product mandate by distinguishing their subunit-

level capabilities. Almeida (Almeida, Dokko, & Rosenkopf, 2003) discussed the same

question by unbundling the process of knowledge management, defining this process as

80

search, transfer, and integration, and linking each of these components with subunit

capabilities.

Gupta and Govindarajan (1991) created a new typology of subsidiaries by

applying a knowledge flow based construct. They applied the magnitude and the

direction of knowledge flows of each subunit. (i.e. Subsidiaries are the provider or

receiver of knowledge). Four subunit roles are identified in their model: Global innovator,

integrated player, implementer, and local innovator. Similar categorization can be found

in Ambos and Reitsperger’s study (2004), which distinguishes between technological

mandate subsidiaries and task-related interdependence subsidiaries. Birkinshaw and

Morrison (1995) provided a three-fold typology which classifies subsidiaries as local

implementer, specialized contributor and world mandate. These studies are all based on

the characters and natures of intra-MNC knowledge flow. Cantwell and Janne (1999)’s

research made efforts around the same topic, but from a different approach. Their work

compared the technological interrelatedness of R&D activities in foreign centers with

those in domestic countries. In other words, the research activities by foreign facilities are

distinguished as either ‘replication’ or ‘diversification’.

The analysis of subunit roles in this paper, however, takes a different route. We

distinguish three types of subunit roles according to their relative degrees of involvement

in either the MNC network or the local network in which they operate. Namely the three

types of subsidiaries are international competence exploitation subsidiaries, local

competence exploration subsidiaries, and global competence creation subsidiaries.

Intuitively, there can be a trade-off effect between the subunit’s MNC integration and its

81

local embeddedness given the limited resource of management and investment focus.

However, as suggested by some researchers, in certain circumstances when the MNC

operates within (and sometimes contributes to) a fast changing environment, this trade-

off effect could be eliminated, and instead, a complementary relationship between the

involvements into the two networks could exist. Hence, a subunit can have local

competence creating activities such as knowledge transferring and innovation based on

its relations with local partners and resources, while increasingly relying on competence

exploiting exchanges within its MNC group. Together, they reinforce the subunit’s

position in both local and MNC networks.

A subunit’s involvement in each of the two networks can be demonstrated in a

two-two matrix (Figure 8). Using the degree of MNC integration as one dimension, and

the degree of local embeddedness as the other, each subunit can be allocated to a cell in

this metrics.

ICE

LCE

GCC

MNC Integration

Local Embeddedness

H

H

L

L

Figure 8 Subunit typology by knowledge sourcing directions

82

International Competence Exploitation subunit

A subunit appearing on the upper-left side indicates its higher degree of MNC integration

and lower degree of local embeddedness. It means that the subunit’s technological

activities are largely dependent on the parent group’s technological trajectory, while its

geographical position is simply a strategic choice of the MNC group. The mandate of

these subsidiaries is to exploit the existing competences of the parent group in a new

(foreign) location, which is usually local market oriented. We name this type of subunit

the International Competence Exploitation subunit (ICE). The function of ICE subunit

is therefore closely associated with the product life cycle model (Vernon, 1966), which

indicates the spreading of innovation from home country towards foreign markets.

This is the prevalent case in the 1950s-60s. In this period, large MNCs

internationalize their production in the seeking of foreign markets. In this case, the

linkage between a subunit and its parent group (both headquarter and other subsidiaries)

is very strong. Subsidiaries are under the centralized control of the parent group, and

inter-unit networking is encouraged for the benefit of the whole group’s global strategy.

The more integrated a subunit become into its MNC network, the more likely that the

subunit will be able to appreciate new developments in this intra-firm network, and

therefore the more easier it can acquire and assimilate new knowledge from this network.

Local Competence Exploration subunit

83

A subunit appearing on the lower-right side of the matrix in Figure 1, on the other hand,

indicates a higher degree of embeddedness into the local network, while a lack of

integration within the MNC network. This type of subunit is Local Competence

Exploration subunit (LCE). On the one hand, it actively interacts with its local business

partners, competitors, or even government and other research institutes, in regards to their

technological activities. On the other hand, it maintains only a weak connection with its

MNC group, and its technology developments aren’t necessarily following the path of the

whole MNC network. One typical and extreme example of this type of role is a newly

acquired subunit. This subunit would have an established business network in its local

environment, while its connection with headquarter and other parts of the MNC is very

weak.

In this case, the more embedded the subunit is in its local network, the more likely

it is able to acquire and assimilate new knowledge from this network. However, if the

local network and MNC network have different focuses on technology, the more locally

embedded the subunit is, the less likely it is to have an equally strong capacity to gain

new knowledge from the MNC network.

Global Competence Creation Subunit

Although historically, both ICE and LCE type of subsidiaries have their prevalence in

MNC strategies, neither of these types can serve the purpose of MNC internationalization

in the fast changing global economy nowadays. Given the globalization of technological

activities and knowledge accumulation, subsidiaries are not competitive if they are too

dependent on either network. The conflicts between the alternative networks in previous

84

stages that caused obstacles for knowledge exchange between parties are now less

significant as the opening and interaction of both networks take place in the dual-course

of subunit evolution (Cantwell, 2008).

For an international competence exploitation subunit, focusing on MNC mandates

would no doubt help establishing the advanced technological capability to fulfill the

manufacturing needs for the MNC group, which in turn allows the subunit to achieve a

strategically important position within the MNC group. However, lack of interaction with

local business partners or industrial clusters can result in the loss of emerging technology

advantages. Moreover, cluster economy can provide a local expertise group on certain

advantageous fields, feeding on the information interchange in this type of group could

help the subunit to benefit from local sources of knowledge, which is not necessarily

related closely to the original mandate of the subunit, but could be equally, if not more,

critical to the competition in both the local industry cluster and the expanding global

market.

For a local competence exploration subunit, on the other hand, the benefit of local

business network is fully appreciated. Clustering in specific locations which represent

“centers of excellence” of one or more industries could increase the chances of being

exposed to the most advantageous technology, so that subsidiaries can be focusing on the

local industry or market. However, since the local focus isn’t always consistent with

MNC’s strategic planning. Being overly committed to a local network and building up its

unique competitive advantage will cause a subunit to fall out of the strategy direction of

85

its MNC group, which can result in a weakened position within the MNC. In an extreme

case, this would result in spin-offs.

Hence, a more efficient type of subunit role emerged in the evolution path – the

Global Competence Creation subunit (GCC). Subsidiaries with a GCC role are not only

highly integrated within the MNC’s strategic blueprint, but also tightly embedded in its

local business network. When MNCs move towards a more closely integrated

international network, foreign technological activities carried out by affiliates are not

limited to adopting and developing existing competencies, but rather incrementally

targeting at new ones. Cantwell and Noonan (2007) proposed that the technological

activities of foreign subsidiaries located in Germany are not heavily concentrated in home

base exploiting-type activities, but on exploring-type activities.

Zander (1998) suggests that foreign unit activities are associated with a

significantly higher probability of entry into new and more distantly related fields of

technology, creating a long-term drift into new technological capabilities. In other words,

foreign subsidiaries on average are getting more focused on the ‘R’ in the R&D process

and evolving with the motivations to discover some new capabilities. In this case, an

increasing number of competency-creating subsidiaries will be less reliant upon

knowledge sourcing from their parent company or from other organizations in the home

country of their MNC corporate group. Their strategies are more likely to be locally

accumulative in some specialized fields of knowledge-building that are distinct

(diversified) from the competence base of their parent. In our dynamic model of subunit

role, this evolutionary process can be represented by the arrow from the upper-left side to

86

the upper-right side in Figure 1. It indicates that over time, ICE subsidiaries are shifting

towards GCC subsidiaries. They are combining the local expertise with their MNC

knowledge, creating subunit-specific competence by getting more embedded into their

local network. This type of subunit evolution is largely driven by subunit level initiates,

while enhanced by the support of MNC group and local network.

Meanwhile, the LCE subsidiaries are becoming more integrated into the MNC

network. In order to avoid the problem of repeated R&Ds in geographically dispersed

subsidiaries, and to take full advantages of different locations, MNC reinforce its control

over the LCE subsidiaries and hence increase their contribution to the parent group.

Driven largely by the MNC group, LCE subsidiaries are shifting into the role of GCC,

keeping their advantageous local connection but enhancing their competence creating

capability by closely cooperating with other parts of the MNC network.

Therefore, subsidiaries evolving towards a GCC role tend to achieve both

capability enhancement and charter establishment. They create unique competence by

interacting with partners in both MNC and local network, and hence develop a

strategically important position within the MNC group as an active innovator, while

embracing the benefits of a certain degree of autonomy in its location, which in turn

guarantees its knowledge absorbing capability in each of the networks.

5.2 Typology Based on Innovation Pattern – CC intensity vs. sub distance

Now if we take a look at the innovation patterns of foreign subunits, instead of their

knowledge sourcing compositions, it is understandable if we shift focus to the actual

87

competence creating / competence exploiting activities. As discussed in Chapter Four, the

composition of these two types of activity could determine a subunit’s strategic position

within the MNC group. In this part, I will demonstrate a different typology of subunit

roles based on the extent to which a subunit’s innovative activities are CC as opposed to

CE, and on the technological distance between the subunit and its parent company.

The first dimension is defined as subunit CC intensity. It is measured with the share of

competence creating activities among all innovative activities (CC+CE) within a subunit.

The construction of this variable takes two steps. First, I identified the innovative

activities in fields that are categorized as competence creating vis-à-vis competence

exploiting by comparing the RTA value of subunit i in field j (RTAij) to the RTA value of

subunit i’s parent company in field j (RTApj). If RTAij >1 and RTApj <1, field j is

considered a field of CC for subunit i. This comparison is done by all subunits in all

technological fields. Second, I calculate the number of patents that belong to these CC

fields for each subunit, and then divide this number by the total number of patents for

each corresponding subunit. The result is a percentage number indicating the CC

intensity of a subunit.

The second dimension is subunit distance – the technological distance between

innovative activities of a subunit and its parent company. A detailed definition and

calculation of this variable can be found in Chapter Two.

Combining these two dimensions generates another two-by-two matrix (as shown in

Figure 9).

Subunit Distance

H

Figure 9 Subunit Typology by Competence Creativeness

88

Subunits that are positioned in the lower left corner (Quadrant III of Figure 5) have both

low degree of CC intensity and low distance from its parent company. This type of

subunit is mostly miniature replica of the parent company in a foreign location; its

innovation is mainly incremental ones based upon the technological specialties of the

parent company. This fits the description of traditional subunits, which have a large share

of CE activities as opposed to CC.

Subunits positioned at the upper right corner (Quadrant I of Figure 5), on the other hand,

is the other extreme of the case. These subunits are not just highly innovative, but their

innovation is to a large extent composed of CC activities, and their portfolios of these

activities are significantly different from that of their parent companies’. These subunits

are of high strategic importance to an MNC group as they bring in new potentials to the

entire group, especially when some of these CC areas are matured, the group can then

recognize them as part of its new core competency, and therefore spread these specialties

across the MNC group’s international network. Their strategic role could be categorized

the ultimate competence creating type, and they are considered centers of excellence

within their MNC group’s network.

Subunits that fall under the category of high technological distance from its parent

company while having low CC intensity (Quadrant II of Figure 5) could be the ones that

has a technology portfolio that includes mostly in some certain areas of expertise that are

same as the parent company’s areas of specialization, but their technological field

diversification is highly skewed by a few outliner fields that don’t fall into the categories

of parent company specialization. These subunits are developing some niche applications

89

based on their local host country competency, while their main innovative activity is still

focused on the exploitation of MNC group specialties.

In the lower right quadrant (Quadrant IV of Figure 5), subunits are found to have high

degree of CC intensity with low technological distance compared to their parent

companies. These subunits are the ones of most interest in my typology. They

concentrate their innovative efforts on areas that are not particularly specialized by the

parent company. However, these areas are not completely unrelated or brand new to the

MNC group. Their technology portfolio shows consistency with the parent company, just

that they tend to concentrate more on the fields that are relatively new to the group. The

new knowledge combinations these subunits are making are close to the core competency,

and they conduct a more focused exploration compared to those with a CC role or a

Niche Application Role. We therefore categorize this type of subunit as “Core Base

Extension”. These subunits have the most potential to bring in new competencies to the

MNC group and incorporate these new areas of technology into the development of new

group level core competency.

Figures 10 and 11 show a distribution of subunits among these four categories across the

dimensions of subunit technological distance and subunit CC intensity. Figure 10 samples

the full dataset while Figure 11 focuses on large subunits with more than patents in a five

year window.

90

Figure 10 Subunit Distance and CC Intensity (All Data)

Figure 11 Subunit Distance and CC Intensity (Large Subunits)

91

Comparing these two diagrams, larger subunits seems to demonstrate a stronger positive

relationship between dimensions of subunit technological distance and subunit CC

intensity. Large subunits distribute more closely along the fitted regression line, while

smaller subunits tend to have more various types of positioning as shown on the diagrams.

5.3 Evolution of Subunit

To examine the trajectories of subunit role evolution, I use the typology developed in the

previous part with dimensions of subunit CC intensity and subunit technological distance.

By examining the change in each of these dimensions over different time periods, I am

able to construct a new model that distinguishes subunit evolution paths based on their

directions of movement. Figure 12 shows the categorization of these movements.

Figure 13 shows the scattered distribution of subunits’ evolution patterns. Since this chart

uses change of value of the two dimensions in Figure 5 as the axis, it is intuitive to use

Delta SC

Delta_SD

Figure 12 Categorization of Subunit Evolution Patterns

92

value 0 as the dividing cut-up point to distinguish types of subunit evolution pattern. I’m

using all data sample in this chart. The large subunits sample shows almost the same

results, except with a few less observations. The reason of this is that subunits have to

show a consistency in their overall competence creativeness throughout more than one

time window to be included in this result. Lots of subunits are not showing this

consistency; therefore it’s hard to capture their evolution patterns. Large subunits,

however, are more likely to show up repeatedly in this chart because of the capability of

conducting continuous CC activities.

Figure 13 Subunit Evolution Patterns

In Quadrant I of Figure 13, subunits are demonstrating an increase in both their

technological distance from their parent group and their CC intensity. These subunits are

probably those that are heavily embedded in host country local environment while at the

same time possess a favorable mandate from the MNC group to facilitate their further

development of innovative activities that are new to the MNC group. 42 subunits are

identified in QI, among which the most consistency is shown by the subunit of the Swiss

firm Novartis in the US. Novartis is a pharmaceutical company that has its center of

93

research and development based in Basel, Switzerland, with other key research sites

located in Horsham, UK; Vienna, Austria; Tsukuba, Japan; East Hanover, New Jersey,

USA. Their strategy for innovation is “focused diversification” (from Novartis website),

by which they mean a concentration on pharmaceutical, biomedical innovation while

encouraging diversification into related areas. It is shown in the data that Novartis US

started as a traditional type of subunit, with technological distance of 0.04 and CC

intensity of 0.04; in the following period this pair of numbers were increased to (0.09,

0.30), whereas towards the end of my observation period the numbers reached (0.11,

0.40). It is clear that Novartis US is on the trajectory of evolving into a category of more

emphasize on CC type of activities.

Quadrant II of Figure 13 includes subunits that are increasing in their technological

distance from the parent company, yet the level of CC activity intensity decreases

overtime. These subunits are moving into the direction of concentrating more on CE

activities, but their areas of niche application stays strong and even draws further away

from the parent company. A typical example of subunit in this quadrant is ExxonMobil’s

operation subunit in Canada.

Quadrant III is consisting of subunits that are decreasing on both technological distance

and CC intensity. Depending on the actually type of subunit, falling into this type of

evolutionary trajectory could mean two things: (1) the subunit is already well advanced in

both technological distance and CC intensity as a center of excellence, now it is time for

the competence created by this subunit to be transferred back to the parent company and

then further to the international network of MNC group – as a result the subunit itself

94

demonstrates a relatively lower level of CC intensity and technological distance overtime;

(2) the subunit is less of a center of excellence, and the field of specialization which is

new to the parent company is either no longer a specialty of the subunit itself, or

somehow adapted by the parent company. One typical example is US company Schering

Plough’s operation in Germany. Its subunit technological distance dropped from 0.033 to

0.004, while its CC intensity dropped from an outlining 0.949 all the way down to 0.387.

This is an extreme case due to a merger between the US company Plough Inc. and

Germany company Schering Corporation. In this case, the reduction of CC activity

presence is showing the effect of two companies corresponding and transmission

technological capabilities into one another.

Quadrant IV indicates an increase in CC intensity while the technological distance

decreases. This could be caused by a more focused effort on CC areas that are to some

extent related to the parent company’s fields of specialization – just like apples don’t fall

far from the tree. An example would be Novartis’s operation in Germany, which follows

perfectly to the company’s strategy – “focused diversification” as explained in previous

discussion.

95

6. Conclusion

The innovative activities of multinational corporation (MNC) operations overseas can be

represented as two types: either competence exploiting (CE) – exploiting the core

competence base of the parent group – or competence creating (CC) – creating new

competencies that were not already among the strengths of the relevant parent company.

To a large extent, the share of these two types of activities determines and reflects a given

subunit’s strategic role within its MNC. This research examines (1) the patterns of MNC

subunits’ knowledge sourcing in terms of the technological and geographical dispersion

of knowledge sources; (2) the extent to which MNC subunits’ technological fields of

expertise are distinct from those of their parent companies, and how this technological

distinctiveness is related to their knowledge sourcing patterns; and (3) how MNC

subunits’ profiles of CC and CE activities (in terms of their overall technological distance

from their parent companies, and the degree to which they are engaged in CC versus CE

activities) evolve over time, reflecting the evolution of their knowledge creating role and

status within their international group. Attention is focused on the heterogeneity of firm-

specific evolutionary paths in the patterns of knowledge accumulation that support CC

activities, controlling for the industry-specific determinants, location-specific factors, and

MNC group structural influences on such technological trajectories.

Three research questions were brought up in the Introduction chapter, now let’s revisit

every one of them:

96

Research Question 1: what are the patterns of knowledge sourcing of an MNC subunit

with respect to its technological dispersion, geographic dispersion, and level of sourcing

concentration in its two main sources – home country and host country?

Three studies were designed around patent and patent citation data of large multinational

corporation groups in the chemical industry to examine these three questions. For

Research Question 1, we find out that subunits tend to have a knowledge sourcing pattern

of increasing dispersion both in technological fields and in geographical locations. There

is a positive relationship between technological dispersion and geographical dispersion,

but the effect is moderated by the effect of share of home country sourcing and share of

host country sourcing. The patterns of knowledge sourcing have positive relationship

with the technological diversification of subunit, which in turn has an implication on the

influencing patterns discussed in Study Two.

Research Question 2: how does a subunit’s knowledge sourcing pattern influence its

local competence creating capability, hence the technological distinctiveness from its

parent company?

This part of study focuses on the technological distinctiveness as an indicator of

competency creativeness of subunits. We found out that the technological dispersion of a

subunit’s knowledge sourcing for activities in a certain field has a positive relationship

with the subunit’s technological distinctiveness of that field, while the geographical

dispersion of sourcing in a given field has a negative relationship with technological

distinctiveness. These findings confirms the relationship between knowledge sourcing

97

patterns and subunit technological distinctiveness, while supporting the streams of

current literature on subunit technological diversification, internationalization, and local

embeddedness.

Research Question 3: if knowledge sourcing and competence creation are two mutually

interdependent processes, is there a typology of evolutionary trajectories followed by the

paths of technological growth of subunits?

Study 3 examines this last question with exploratory data analysis. Two typologies of

subunit roles are identified, with the first one categorizing subunits based on their

knowledge sourcing patterns, and the second one categorizing subunits based on their

competence creativeness patterns. Based on the results from studies one and two, these

two typologies could be interconnected due to the interdependency between these two

sets of concepts. Four types of evolutionary patterns were recognized in the last section,

with regards to the second typology of subunit role proposed in that chapter – subunit

technological distance and subunit CC intensity.

Contributions of this dissertation are three fold. Theoretically, the studies contribute in

areas of MNC knowledge accumulation and competence creating by establishing new

connection between the two and further on identifying evolutionary trajectory of MNC

subunit strategic role development. Empirically, the studies created new ways of studying

MNC subunit innovation, as well as a clear typology of subunit role based on its

competence creativeness. It also examines the relationship between subunit knowledge

sourcing and innovative activity, which could lead to further development of study on

98

these two interdependent characteristics of subunit strategy. Finally, methodologically,

the studies propose a combination of two dimensions – subunit technological distance

and CC activity intensity – to indicate strategic positioning of the subunit within an MNC

international network, this method takes into consideration of the composition of subunit

overall innovativeness and its novelty compared to parent company, which gives us a

comprehensive view of subunit competence creativeness.

However, due to the nature of data (patent database) used in this study, it is very difficult

to identify real subsidiaries entities for analysis. Instead I’m using the firm-country

combination to indicate an MNC’s innovative activity in a foreign country. This may

cause some problems especially for a large country like the US, since there is huge

variation across regions in this country. A potential solution is to divide the country into

different regions, so that these regions can be comparable to other countries.

Another limitation of the data is the identification of patent affiliations of citations. The

current dataset doesn’t provide this information. If further research on be based on a

complete data of patent and corporate affiliation information, we can further examine the

relationship between inter- firm and intra- firm knowledge sourcing patterns, and their

impact on subunit innovation.

As discussed in Chapter Two, patent continuation could result in multiple count of single

invention, as well as high interdependence of different patents within one firm. If

invention level data could be acquired, it will avoid having this problem in research. One

of the suggestions for future improvement of this study is to rule out non-independent

99

patents by the same inventors within each firm. This process requires using the original

USPTO dataset to identify all inventor information, as well as patent continuation

information.

Future study could concentrate more on a comprehensive empirical analysis of the

evolutionary trajectory for subunits. Location industry conditions could be taken into

consideration when analyzing these evolutionary trends.

100

APPENDICES

Appendix A1 – Firm List

Firm Name Home Country 3M US Abbott Laboratories US Air Liquide France Air Products and Chemicals US Akzo Nobel Netherlands Albright and Wilson UK Allergan US Allied US American Cyanamid US American Home Products US Amgen US Asahi Kasei Japan Ashland US Astellas Pharma Japan AstraZeneca UK Avery Dennison US Avon Products US Basell Netherlands BASF Germany Bausch Lomb US Baxter International US Bayer Ag Germany Beecham Group UK Biogen Idec US BOC UK Boehringer Ingelheim Germany BP UK Bristol-Myers Squibb US Cabot US Celanese US Cephalon US CF Industries US Charbonnages de France France Chemische Werke Huls Germany Chemtura US Chesebrough Ponds US

101

Chevron Phillips Chemical US Ciba Specialty Chemicals Switzerland Clariant Switzerland Clorox US Cognis Germany Colgate-Palmolive US ConocoPhillips US CSL Australia Cytec Industries US Daicel Chemical Industries Japan Daiichi Sankyo Japan Dainippon Inks & Chemicals Japan Dainippon Sumitomo Pharma Japan Denki Kagaku Kogyo Japan Dow Chemical US Dow Corning Corporation US E.I.Du Pont De Nemours and Company US Eastman Chemical US Ecolab US Eisai Japan Eli Lilly US Enterprise Miniere et Chemique France Ethyl US ExxonMobil US Firm_name home_country Forest Laboratories US Formosa Chemicals & Fibre Taiwan Formosa Plastics Corporation Taiwan Fuji Photo Film Japan Genetech US Genzyme US Gilead Sciences US Glaxo UK GlaxoSmithKline UK H. Lundbeck Denmark Hanwha Chemical Korea Henkel Germany Hercules US Hexion US Huntsman US ICI UK

102

Int Minerals & Chemicals US Johnson Johnson US Johnson Matthey UK Kaneka Japan Kao Corporation Japan Kemira Finland King Pharmaceuticals US Koppers US Kuraray Japan Kyowa Hakko Kogyo Japan Lanxess Germany LG Chem Korea Lion Japan Lonza Switzerland Loreal France Lyondell Chemical US Marathon Oil US Merck US Mitsubishi Chemical Holdings Corporation Japan Mitsubishi Gas Chemical Japan Mitsubishi Petrochemicals Japan Mitsubishi Pharma Japan Mitsui Chemicals Japan Mitsui Petrochemical Japan Monsanto US Morton Norwich Products US Nalco US Norsk Hydro Norway Nova Chemicals Canada Novartis Switzerland Novo Nordisk Denmark Olin US Orica Australia PennWalt US Pfizer Inc US Potash Corp. of Saskatchewan Canada PPG Industries US Praxair US Procter & Gamble US Reckitt Colman UK Reichhold Chemicals US

103

Revlon US Rhodia France Rhone-Poulenc France Richardson Vicks US Roche/Sapac Switzerland Rohm and Haas US Rutgerswerke Germany Sanofi Aventis France Schering Plough US SCM US Sekisui Chemical Japan Sherwin Williams US Shin-Etsu Japan Shionogi Japan Shiseido Japan Showa Denko Japan Smith Kline Beckman US Solutia US Solvay Belgium Squibb US Stauffer Chemicals US Sterling Drug US Sumitomo Chemical Japan Syngenta Switzerland Taiho Pharmaceutical Co Japan Takeda Pharmaceutical Co Japan Teva Pharmaceutical Industries Israel Tosoh Japan Transammonia US Ube Industries Japan Union Carbide US Upjohn US Valero Energy US Valspar US W.R. Grace US Wacker-Chemie US Warner Lambert US Watson Pharmaceuticals US Wyeth US Yara International Norway

104

105

Appendix A2 - Country and patent distribution list

Country Patents Patents

by Parent Patents by

Subunit USA 153,916 130,854 23,062 Germany 40,341 25,768 14,573 UK 13,613 7,819 5,794 Italy 1,445

1,445

France 9,646 7,965 1,681 Japan 47,776 45,240 2,536 Netherlands 1,083 460 623 Belgium 2,727 485 2,242 Switzerland 8,445 7,876 569 Sweden 856

856

Denmark 1,167 992 175 Ireland 151

151

Spain 193

193 Portugal 3

3

Luxembourg 30

30 Greece 8

8

Austria 303

303 Norway 318 278 40 Finland 106 82 24 German Democratic Republic (1945-1989) 1

1

Hungary 37

37 Poland 10

10

Czechoslovakia 5

5 Yugoslavia 4

4

USSR 71

71 Canada 2,427 67 2,360 Australia 526 79 447 New Zealand 37

37

India 117

117 Brazil 96

96

Israel 178 140 38 Argentina 8

8

Chile 12

12 Colombia 6

6

Mexico 30

30 Panama 1

1

Peru 1

1

106

Venezuela 12

12 Philippines 17

17

South Korea 674 151 523 Taiwan 48 6 42 Turkey 1

1

South Africa 53

53 West Indies and Guianas 8

8

Bulgaria 1

1 East Indies 14

14

Other Latin America 9

9 Other Africa 11

11

China 38

38 Uruguay 1

1

Ecuador 1

1 Other Asia 11

11

Other Middle East 13

13 Other Europe (Monaco, Liechtestein, Iceland, Malta, Andorra, Albania, Cyprus, Greenland) 7

7

Hong Kong 31

31 Singapore 23

23

Total 286,667 228,262 58,405

107

Appendix A3 – Geographic distribution of patents – by parent companies and by foreign subunits Geographic distribution of patents invented by parent companies:

Geographic distribution of patents invented by subunits:

Patents by Parent

USA

Germany

UK

Italy

France

Japan

Netherlands

Belgium

Switzerland

Denmark

Patents by Subunit

USA

Germany

UK

Italy

France

Japan

Netherlands

Belgium

Switzerland

Denmark

108

Appendix A4 – 31 Technological Fields and Corresponding Patent Class Code

Tech31 Field Name Patent class (sub-class) 1 Food and tobacco products 127 (29-71), 131 (1-226, 291-999), 426 2 Distillation processes 201, 203 3 Inorganic chemicals 423 4 Agricultural chemicals 71, 504 5 Chemical processes 23, 51 (293-328), 55 (1-99), 62 (1-122), 95,

117, 134 (1-42), 156 (1-479, 600-668), 204 (1-192, 900-914), 205, 210 (501-982), 216, 260 (95, 684-708), 427, 432 (1-53), 518

6 Photographic chemistry 430 7 Cleaning agents and other

compositions 106, 252, 508, 510, 512, 516, 588

8 Synthetic resins and fibres 260 (1-94, 666-683, 709-999), 520,521, 522,523, 524, 525, 526, 527, 528

9 Bleaching and dyeing 8 10 Other organic compounds 260 (96-665), 530, 534, 536, 540, 544, 546,

548, 549, 552, 554, 556, 558, 560, 562, 564, 568, 570, 930, 987

11 Pharmaceuticals and biotechnology 424, 435, 436, 514, 800, 935 12 Other chemicals and Related -

disinfecting, preserving, textiles and explosives

422, 2, 36,245,289,450, 149

13 Metallurgical processes 29, 75, 148, 164 (1-148, 101-265), 419, 420 14 Miscellaneous metal products 3,4,7, 10, 16,24, 27, 30 (1-165, 167, 395-499,

501-999), 49, 63, 70, 108, 109, 124, 132, 135, 138, 150, 160, 182, 190, 206, 211, 215 (100-367), 220, 232, 248, 256, 267, 272, 279, 285, 292, 312, 383, 403, 411, 464, 623

15 Chemical and allied equipment 34, 51 (1-292, 329-999), 55 (100-999), 68, 96, 118, 134 (43-999), 156 (480-599, 699-999), 159, 196, 202, 209, 210 (1-500), 261, 366, 422 (44-999), 494, 502, 503

16 Paper making apparatus 53, 162, 229, 493 17 Assembly and material handling

equipment 186, 187 (1-28, 30-999), 193, 198, 212,224, 226, 242, 254 (134-999), 258, 271, 294, 402, 406, 410, 414, 901

18 Other specialised machinery 15, 30 (166, 168-394, 500), 79, 98, 100, 116, 133, 140, 141, 147, 157, 169, 194, 221, 222, 227, 254 (1-133), 277, 291, 300, 401, 425, 452

19 Other general industrial equipment 48, 91, 92, 110, 122, 126, 137, 165, 184, 185, 188, 192, 237, 239, 251, 303, 415, 416, 417, 418, 431, 432 (54-999)

109

20 Mechanical engineering nes 99, 127 (1-28), 131 (227-900), 59, 72, 76, 81, 82, 83, 163, 164 (149-999), 173, 225, 228 (1-100), 234, 266, 269, 308 (1-9, 11-245), 384, 407, 408, 409, 413, 474, 65 (138-999), 241 (132-999), 249, 56, 111, 130, 172, 278, 460, 37, 171, 404 (83-133), 166, 175, 299, 445, 12, 19, 26, 28, 38, 57, 66, 69, 87, 112, 139, 223, 101, 199, 270, 276, 281, 282, 283, 412, 462, 142, 144, 145, 235 (61-89; 419-434), 400

21 Electrical devices and systems 174, 200, 307 (1-199, 586-999), 308, 323, 327, 328, 330, 331, 333, 334, 335, 336, 337, 338, 339, 361, 363, 372, 439, 505

22 Other general electrical equipment 62 (123-999), 136, 204 (193-499), 219, 236, 290, 310, 318, 320, 322, 361 (433-436), 373, 388, 392, 429, 437, 438

23 Office equipment and data processing systems

235 (375-386, 400-418, 435-457), 360, 364, 365, 369, 371, 377, 700, 701, 702, 703, 704, 705, 706, 707, 708, 709, 710, 711, 712, 713, 714, 715, 716, 717, 718, 719, 720, 725, 726, 902

24 Electrical equipment nes 178, 179, 329, 332, 367, 370, 375, 379, 455, 340, 341, 382, 342, 343, 84, 181, 348, 358, 381, 386, 725, 313, 314, 315, 362, 357

25 Transport equipment 123, 180, 296, 244 (1-13, 15-999), 114 (1-19, 26-999), 440, 441, 104, 105, 213, 238, 246,191, 280, 293, 295, 298, 301, 305

26 Rubber and plastic products 152, 264 27 Non-metallic mineral products 52, 65 (1-137), 125, 215 (1-99), 241 (1-131),

428, 442, 501 28 Coal and petroleum products 44, 208, 585 29 Photographic equipment 354, 355, 396, 399 30 Other instruments and controls 33, 73, 74, 128, 177, 187 (29), 235 (1-60, 90-

374, 387-399, 458-999), 250, 324, 346, 347, 349, 350, 351, 352, 353, 356, 359, 368, 374, 378, 385, 398, 433, 475, 600, 601, 602, 604, 606, 607

31 Other manufacturing and non-industrial

60, 376, 976, 5, 217, 297, 6, 14, 17, 40, 42, 43, 47, 54, 86, 89, 102, 114 (20-25), 119, 168, 231, 244 (14), 273, 380, 404 (1-82, 134-999), 405, 434, 446, 449, 452, 463

110

111

Appendix B – Table of Correlation for Key Variables in Study 1

cng_td ced_td ced_gd hsced_td nhsce~td nhsce~gd hmced_td nhmce~td nhmce~gdhost_td US_home US_host hs_share hm_share gpt_sh~e ced_gp~e intreg~ecng_td 1.000ced_td 0.3019* 1.000ced_gd 0.1250* 0.4720* 1.000hsced_td 0.3045* 0.6020* -0.010 1.000nhsced_td0.2474* 0.6956* 0.2203* 0.4201* 1.000nhsced_gd0.1782* 0.2731* 0.4923* 0.3222* 0.4491* 1.000hmced_td 0.2547* 0.6154* 0.016 0.7595* 0.5653* 0.2415* 1.000nhmced_t 0.2847* 0.6801* 0.2190* 0.5470* 0.8566* 0.4200* 0.4170* 1.000nhmced_g0.1520* 0.2600* 0.4813* 0.2121* 0.3999* 0.8243* 0.2856* 0.4530* 1.000host_td 0.5127* 0.1912* 0.1536* 0.2127* 0.1909* 0.1600* 0.1600* 0.1980* 0.1446* 1.000US_home -0.009 0.0469* -0.1616* 0.1305* 0.008 0.0755* 0.2428* -0.0797* 0.1954* -0.0715* 1.000US_host 0.2693* 0.1911* -0.1248* 0.3869* 0.0605* 0.3127* 0.2670* 0.1417* 0.1712* 0.1680* 0.3598* 1.000hs_share 0.1216* -0.1080* -0.6862* 0.3488* -0.0763* 0.1450* 0.2062* -0.0314* -0.1383* 0.2042* 0.3026* 0.7044* 1.000hm_share -0.0551* -0.1984* -0.7135* 0.1422* -0.1323* -0.2388* 0.2481* -0.1648* 0.0525* -0.1281* 0.6662* 0.2887* 0.6159* 1.000gpt_share 0.1924* 0.0515* 0.0190* 0.0634* 0.0418* 0.0154* 0.0304* 0.0502* -0.0227* 0.025 -0.0465* 0.0327* 0.013 -0.0430* 1.000ced_gpt_s 0.0722* 0.1116* -0.0324* 0.1299* 0.1417* 0.0166* 0.1164* 0.1383* 0.008 0.020 0.008 0.014 0.002 0.008 0.0696* 1.000intreg_sha-0.2120* -0.2212* -0.0288* -0.2436* -0.0258* -0.2703* -0.1781* -0.0694* -0.1224* -0.1944* -0.2068* -0.5728* -0.7815* -0.2089* -0.0289* 0.009 1.000

112

Appendix C – Table of Correlation for Key Variables in Study 2

Dissim1 RTA_host logsize struct sub_sh~e ced_td ced_gd nhsce~td nhsce~gd nhmce~td nhmce~gd

T. Distinct 1

RTA_host 0.1249* 1 logsize -0.1620* 0.0075 1

FirmCon 0.0588* 0.0448* 0.1736* 1 sub_share 0.1499* -0.0253* -0.1153* 0.1046* 1

Tech Disp 0.0579* -0.0535* 0.0544* -0.0514* 0.0766* 1 Geo Disp 0.0113 -0.0157 0.0061 -0.0279* 0.0427* 0.4066* 1

nhsced_td -0.0173 -0.0441* 0.1173* -0.0704* 0.0645* 0.7421* 0.2016* 1 nhsced_gd -0.0545* -0.0027 0.0315* -0.1029* 0.0319* 0.2041* 0.5901* 0.3648* 1

nhmced_td 0.0063 -0.0641* 0.0386* -0.1851* 0.0904* 0.7019* 0.2233* 0.7020* 0.3306* 1 nhmced_gd -0.0431* -0.006 0.0611* -0.1061* 0.0303* 0.1969* 0.6105* 0.2909* 0.5996* 0.3858* 1

*: p < .05 Sample: All data

113

Distin RTA_host logsize FirmCon sub_sh~e Tech Disp Geo Disp nhsce~td nhsce~gd nhmce~td nhmce~gd

T. Distinct 1

RTA_host 0.0524* 1 logsize -0.1890* -0.0345 1

FirmCon 0.0410* 0.0137 0.1480* 1 sub_share 0.2544* -0.0224 -0.3173* 0.3736* 1

Tech Disp 0.0337 -0.0567* 0.0507* -0.0627* 0.0162 1 Geo Disp 0.0114 0.0381 0.0479* -0.0356 0.0226 0.3069* 1

nhsced_td -0.0707* -0.0283 0.0918* -0.0620* 0.025 0.6484* 0.1453* 1 nhsced_gd -0.1316* 0.0301 -0.0656* -0.1111* -0.0264 0.1557* 0.3483* 0.3964* 1

nhmced_td 0.0245 -0.0282 -0.0026 -0.1285* 0.018 0.7585* 0.0875* 0.6634* 0.2992* 1 nhmced_gd -0.0659* 0.035 0.1158* -0.0630* 0.0005 0.0804* 0.6169* 0.2434* 0.4214* 0.2030* 1

*: p < .05 Sample: Large subunits

114

REFERENCES

Almeida, P., Dokko, G., & Rosenkopf, L. 2003. Startup size and the mechanisms of external learning: Increasing opportunity but declining usefulness? Research Policy, 23(2): 301–315.

Ambos, B., & Reitsperger, W. 2004. Offshore centers of excellence: social control and success. Management International Review, 44(Special Issue 2): 51–65.

Andersson, U. 2003. Managing the transfer of capabilities within multinational corporations: the dual role of the subsidiary. Scandinavian Journal of Management, 19: 425-442.

Andersson, U., Bjorkman, I., & Forsgren, M. 2005. Managing subsidiary knowledge creation: the effect of control mechanisms on subsidiary local embeddedness. International Business Review, 14: 521-538.

Andersson, U., & Forsgren, M. 1996. Subsidiary embeddedness and control in the multinational corporation. International Business Review, 5(5): 487-508.

Andersson, U., Forsgren, M. & Holm, U. 2001. Subsidiary embeddedness and competence development in MNCs – a multi-level analysis. Organization Studies, 22(6): 1013-1034.

Andersson, U., Forsgren, M., & Holm, U. 2002. The strategic impact of external networks – subsidiary performance and competence development in the multinational corporation. Strategic Management Journal, 23:979-996.

Andersson, U., Forsgren, M., & Holm, U. 2007. Balancing subsidiary influence in the federative MNC: a business network view. Journal of International Business Studies, 38: 802-818.,

Andersson, U., Forsgren, M., & Pedersen, T. 2001. Subsidiary performance in multinational corporations: the importance of technology embeddedness. International Business Review, 10:

Alcacer, J. & Gittelman, M. 2006. Patent citations as a measure of knowledge flows: the Influence of examiner citations. The Review of Economics and Statistics, 88(4): 774–779.

Alcacer, J., Gittelman, M. & Sampat, B. 2009. Applicant and examiner citations in U.S. patents: An overview and analysis. Research Policy, 38(2): 415-427.

115

Asmussen, C.G., Pedersen, T. & Dhanaraj, C. 2009. Host-country environment and subsidiary competence: Extending the diamond network model. Journal of International Business Studies, 40: 42-57.

Atuahene-Gima, K. 2005. Resolving the capability-rigidity paradox in new product innovation. Journal of Marketing, 69:61-83.

Ahuja, G. 2000. Collaboration Networks, Structural Holes, and Innovation: A Longitudinal Study. Administrative Science Quarterly, 45(3): 25-455

Bell, G. & Zaheer, A. 2007. Geography, Networks, and Knowledge Flow. Organization Science, 18(6): 955-972

Birkinshaw, J., Nobel, R. and Ridderstrale, J., 2002. Knowledge as a contingency variable: do the characteristics of knowledge predict organization structure?, Organization Science, Vol.13, No.3, pp.274-289.

Brusoni, S., Prencipe A. and Pavitt, K. 2001. Knowledge Specialisation, Organizational Coupling and the Boundaries of the Firm: Why Firms Know More Than They Make?, Administrative Science Quarterly, 46 (4): 597-621.

Cantwell, J.A. 1991. The theory of technological competence and its application to international production, in D. McFetridge (ed.), Foreign Investment, Technology and Economic Growth, University of Calgary Press: Calgary.

Cantwell J.A. 1995. The globalisation of technology: what remains of the product cycle model? Cambridge Journal of Economics, 19(1): 155-174.

Cantwell J. A. and H. B. Andersen. 1996. A statistical analysis of corporate technological leadership historically, Economics of Innovation and New Technology, 4:211-234.

Cantwell J. A. 2009. Location and the multinational enterprise. Journal of International Business Studies, 40:35-41.

Cantwell J.A, & Mudambi R. 2005. MNC competence-creating subsidiary mandates. Strategic Management Journal 26(12): 1109-1128.

Cantwell J.A., Piscitello L. 2000. Accumulating technological competence - its changing impact on corporate diversification and internationalisation, 2000, Industrial and Corporate Change, 9(1): 21-51.

Cantwell J.A., Piscitello L. 2005. The recent location of R&D activities by large MNCs in the European regions: The role of spillovers and externalities, Regional Studies, 39: 1-16.

116

Cantwell J.A., Piscitello L. 2007. Attraction and deterrence in the location of foreign-owned R&D activities: the role of positive and negative spillovers, International Journal of Technological Learning, Innovation and Development, 1: 1-30.

Contractor, F. J. & Lorange, P. 2002. The growth of alliances in the knowledge-based economy. International Business Review. 11(4): 485-502.

Day, G. 1994. The capabilities of market-driven organizations. Journal of Marketing, 58 (October): 37–52.

Dunning, J. 1995. Reappraising the eclectic paradigm in an age of alliance capitalism, Journal of International Business Studies, 26(3): 461–492

Eisenberg, H. 2000. Patent law you can use: continuation patent applications. From website of Yale University Office of Cooperative Research.x

Forsgren, M., Holm, U., & Johanson, J. 2005. Transfer of knowledge in the embedded multinational - the role of shared values and business networks. in Managing the Embedded Multinational: A Business Network View, Northampton: Edward Elgar.

Gittelman, M., 2006, “National institutions, public-private knowledge flows, and innovation performance: a comparative study of the biotechnology industry in the US and France”, Research Policy, 35:1052-1068.

Ghoshal, S., & Bartlett, C. 1988. Creation, adoption, and diffusion of innovations by subsidiaries of multinational corporations. Journal of International business studies, 19(3): 365-388.

Ghoshal, S., & Bartlett, C. 1991. The multinational corporation as an interorganizational network. Academy of Management Review, 15(4): 603–625.

Granovetter, M. 2005. The impact of social structure on economic outcomes. Journal of Economic Perspectives, 19(1): 33-50

Grubel, Herbert G.; Lloyd, Peter J. (1971). "The Empirical Measurement of Intra-Industry Trade". Economic Record 47(4): 494–517.

Grubel, Herbert G.; Lloyd, Peter J. (1975). Intra-industry trade: the theory and measurement of international trade in differentiated products. New York: Wiley.

Gulati,R. 1999. Network location and learning: the influence of network resources and firm capanbilities on alliance formation. Strategic Management Journal, 20:397-420

Gupta, A. and Govindarajan, V., 1991. Knowledge flows and the structure of control within multinational corporations, The Academy of Management Review, 16(4):768-792

117

Gupta, A. and Govindarajan, V., 2000. Knowledge flows within multinational corporations, Strategic Management Journal, 21(4):473-496

He, Z. & Wong, P. 2004. Exploration vs. exploitation: an empirical test of the ambidexterity hypothesis. Organization Science, 15(4): 481-494

Hedge, D. Mowery, D., & Graham, S. 2007. Pioneers, submariners, or thicket-builders: which firm use continuations in patenting? NBER Working Paper No. 13153.

Hedlund, G. 1994. A model of knowledge management and the N-form corporation’, Strategic Management Journal, Summer Special Issue, 15: 73–90.

Hedlund, G., & Ridderstråle, J. 1995. International development projects: Key to competitiveness, impossible, or mismanaged. International Studies of Management and Organization, 25(1–2): 121–157.

Hounshell, D.A. (1992), "Continuity and change in the management of industrial research: the Du Pont Company, 1902-1980", chapter 8 in G. Dosi, R. Giannetti and P.A. Toninelli (eds.), Technology and Enterprise in a Historical Perspective, New York: Oxford University Press.

Inkpen, A.C. 2002. Learning, Knowledge Management, and Strategic Alliances: So Many Studies, So Many Unanswered Questions, in F. J. Contractor and P. Lorange ( eds.) Cooperative Strategies and Alliances, Elsevier Science: Amsterdam, pp: 267-289.

Jaffe, A. 1986. Technological Opportunity and Spillovers of R&D: Evidence from Firms' Patents, Profits and Market Value, American Economic Review, LXXVI (1986): 984-1001.

Jaffe, A., Trajtenberg, M., & Henderson, R. 1993. Geographic Localization of Knowledge Spillovers as Evidenced by Patent Citations, The Quarterly Journal of Economics, 108(3): 577-598.

Kim, L. 1998. Crisis construction and organizational learning: capability building in catching-up at Hyundai Motor. Organization Science, 9: 506-201.

Kogut, B. 1988. Joint ventures: Theoretical and empirical perspectives, Strategic Management Journal, 9(4): 3 19-332.

Kogut, B. 1991. Joint ventures and the option to expand and acquire, Management Science, 37(1):19-33.

Kogut, B. & Chang, S. 1991. Technological Capabilities and Japanese Foreign Direct Investment in the United States. The Review of Economics and Statistics, 73: 400-413.

118

Kogut, B. & Zander, U., 1992, Knowledge of the firm, combinative capabilities, and the replication of technology, Organization Science, 3(3): 383-397.

Kogut, B. & Zander, U., 1993, Knowledge of the firm and the evolutionary theory of the multinational corporation, Journal of International Business Studies, 24(4): 625-645.

Lemley, M. and Moore, K., 2004. Ending abuse of patent continuations. Boston University Law Review, 84.

Lyles, M.A. & Salk, J.E. 1996. Knowledge acquisition from foreign parents in international joint ventures: an empirical examination in the Hungarian context, Journal of International Business Studies. 27(5): 877-903.

Macher, J., 2006. Technological Development and the Boundaries of the Firm: A Knowledge-Based Examination in Semiconductor Manufacturing. Management Science, 52(6): 826-843

March, J. G. 1991. Exploration and exploitation in organizational learning. Organization Science. 2(1): 71–87

Martin, X. and Salomon, R. 2002. Knowledge transfer capacity and cooperative strategies: when should firms leverage tacit knowledge through licensing and alliances? In F. J. Contractor and P. Lorange (eds.) Cooperative Strategies and Alliances, Elsevier Science: Amsterdam, pp: 729-748.

Martin, X. and Salomon, R., 2003. Tacitness, learning, and international expansion: a study of foreign direct investment in a knowledge-intensive industry. Organization Science, 14(3): 297-311.

McEvily, B. & Zaheer, A. 1999. Briging Ties: a source of firm heterogeneity in competitive capabilities. Strategic Management Journal, 20: 1133-1156.

McEvily, B. & Marcus, A. 2005. Embedded Ties and the Acquisition of Competitve capabilities. Strategic Management Journal, 26: 1033-1055.

Mowery, D. C., & Oxley, J. E. 1995. Inward technology transfer and competitiveness: The role of national innovation systems. Cambridge Journal of Economics, 19: 67-93.

Mowery, D. C., Oxley, JE. & Silverman, BS. 1996. Strategic Alliance and Interfirm Knowledge Transfer. Strategic Management Journal, 17(winter special issue): 77-91

119

Mudambi, R. and Navarra, P., 2004. Is knowledge power? Knowledge flows, subsidiary power and rent-seeking within MNCs, Journal of International Business Studies, 35(5): 385-406.

Nobel, R., & Birkinshaw, J. 1998. Innovation in multinational corporations: control and communication patterns in international R&D operations. Strategic Management Journal, 19(5): 479-496.

Simonin, B. L. 2004. An Empirical Investigation of the Process of Knowledge Transfer in International Strategic Alliances. Journal of International Business Studies, 35(5): 407-427.

Pearce, R. 1999. The evolution of technology in multinational enterprises: the role of creative subsidiaries. International Business Review, 8: 125-148.

Phene, A., & Almeida, P. 2008. Innovation in multinational subsidiaries: the role of knowledge assimilation and subsidiary capabilities. Journal of International Business Studies, 39: 901-919.

Porter, M. 1990. The Competitive Advantage of Nations. Free Press, New York.

Prahalad, C., & Doz, Y. 1981. An approach to strategic control in MNCs. Sloan Management Review, 23(Summer): 5–13.

Ronstadt, R. C. 1977. Research and Development abroad by U.S. Multinationals. Praeger, New York.

Rowley, T., Behrens, D. & Krackhardt, D. 2000. Redundant Governance Structures: An analysis of structural and relational embeddedness in the steel and semiconductor industries. Strategic Management Journal, 21: 369-386

Szulanski, G. (1994). Intra-firm transfer of best practice project: Executive summary of the findings (Report) APQC.

Szulanski, G. 2000. The Process of Knowledge Transfer: A Diachronic Analysis of Stickiness. Organizational Behavior and Human Decision Processes. 82(1), May: 9-27

Uzzi, B. 1996. The sources and consequences of embeddedness for the economic performance of organizations: the network effect. American Sociological Review, 61(4): 674-698.

Uzzi, B. 1997. Social structure and competition in interfirm networks: the paradox of embeddedness. Administrative Science Quarterly, 42: 35-67.

120

Teece, D., Pisano, G., & Shuen, A. 1997. Dynamic capabilities and strategic management. Strategic Management Journal, 18(7): 509-533.

Tilton, J. E. 1971. International diffusion of technology: the case of semiconductors. Washington DC: Brookings Institution Press.

Uzzi, B. 1996. The sources and consequences of embeddedness for the economic performance of organizations: the network effect. American Sociological Review, 61(4): 674-698.

Uzzi, B. 1997. Social structure and competition in interfirm networks: the paradox of embeddedness. Administrative Science Quarterly, 42: 35-67.

Vernon, R. 1966. International investment and international trade in the product cycle. Quarterly Journal of Economics, 81: 190-207.

Williamson, O., 1985, The Economic Institutions of Capitalism, Chapter 4, pp.85-102.

Winter, S., 1987, “Knowledge and competence as strategic assets”, D. Teece, ed. The Competitive Challenges – Strategic for Industrial Innovation and Renewal. Ballinger, Cambridge, MA, pp.159-184.

Zander, I. 2002. The formation of international innovation networks in the multinational corporation: an evolutionary perspective. Industrial and Corporate Change, 11(2): 327-353.

Zander, I., & Solvell, O. 2000. Cross-border innovation in the multinational corporation. International Studies of Management and Organization, 30(2): 44-68.

121

CURRICULUM VITAE

EMPLOYMENT

Assistant Professor at SUNY Oswego 2012.8– Present Department of Marketing and Management, School of Business

EDUCATION

Rutgers Business School, PhD in Management: International Business 2007.9 – 2013.5

School of Management, Zhejiang University, Doctoral program: Management 2004.9 – 2007.6 Chu Kechen Honors College, Zhejiang University, Bachelor: Computer Science 2000.9 – 2004.6

DISSERTATION

MNC subunit knowledge sourcing and competence creating activities – a dynamic view of subunit evolution

Dissertation Committee: John Cantwell (chair), Farok Contractor, Michelle Gittelman, Lucia Piscitello

Dissertation Abstract:

This dissertation suggests a framework to examine the intra- and inter- firm knowledge sourcing behaviors of an MNC subunit, and how this influences the extent to which the subunit is being competence creative (CC) rather than competence exploitative (CE), and the study looks at how in turn the composition of CC and CE activities in a location influences the knowledge creating role and status of a subunit within its international group. I developed a dynamic model which proposes that the extent to which a subunit is likely to take up CC activities is influenced by the intensity and spread of that subunit’s connectedness of their processes of knowledge accumulation with various sources - whether in other parts of the focal MNC group or external to the firm. The focus of attention is placed on the heterogeneity of firm-specific evolutionary paths in the patterns of knowledge accumulation that support CC activities, controlling for the industry-specific determinants of such technological trajectories.

122

AWARDS

• Nominated for Best Dissertation Proposal Award in AIB 2011, Nagoya • Dissertation Fellowship 2011-2012 by Rutgers Graduate School • Technology Management Research Center fund for Summer Research Assistants,

2011 • Dean's Fund for Summer Ph.D. Research Assistants, 2010 • Competitive summer research scholarship, 2008

RESEARCH INTERESTS

• Intra- and Inter- firm knowledge networks of MNC • MNC subsidiary strategy, knowledge flow, subsidiary evolution, organizational

culture • Technological innovation: exploration versus exploitation, technology diffusion,

learning

PUBLICATIONS

Slone, R., Becker, S., Penton, P., Pu, X., McNamee, R., (Nov-Dec 2011). Managing Global R&D Networks. Research Technology Management, 54( 6 ), 59-61

Chen, J., Pu, X., & Shen, H. 2010. A Comprehensive Model of Technological Learning: Empirical Research on Chinese Manufacturing Sector. In The Rise of Technological Power in the South, Edt. Xiaolan Fu and Luc Soete. Palgrave Macmillan 2010.

WORKING PAPERS

Pu X and Cantwell J, Dual network patterns of MNC subunit knowledge sourcing – the legend, evolution, and new trend

Cantwell J, Piscitello L and Pu X, MNC subunit evolution – the role of technological relatedness and location openness

McNamee R and Pu X, Insights into MNC’s global knowledge networks: strategies and implementations (Research on Research project of IRI)

123

Huang S, McNamee R, Piepenbrink A and Pu X, Marshal vs. Jacobs: an intervention via multi-dimensional, dynamic characterization of clusters

CONFERENCE PRESENTATIONS

Pu X, (AIB 2011, Nagoya) MNC subunit knowledge sourcing and CC activities - a dynamic view of subunit evolution

Huang S, McNamee R and Pu X (AIB 2011, Nagoya) Marshall vs. Jacobs: an intervention via multi-dimensional, dynamic characterization of clusters

Pu X, (AIB 2010, Rio de Janeiro) Organization-culture fit: and insight into MNC subsidiary’s culture dimensions

Pu X and Qiu R, (AOM 2009, Chicago) The impact of network involvement on absorptive capacity: a dynamic model of subsidiary role evolution

Pu X, (CICALICS Workshop 2006, Chengdu) Research on the Mechanisms of Technology Diffusion in CoPS Innovation – A Review of Current Studies

Pu X, (GLOBELICS Academy 2006, Lisbon) Research on the Mechanisms of Technology Diffusion in CoPS Innovation: An IPR Protection Perspective.

TEACHING

• International Business (Summer 2009, Fall 2009, Spring 2010, Winter 2011, Summer 2011, Fall 2012)

• Principle of Management (Summer 2011, Fall 2012, Spring 2013) • Strategic Management (Spring 2013)

PROFESSIONAL SERVICES

• Reviewer for AOM (2007, 2008, 2009, 2011), AIB (2009, 2010, 2011, 2013) • Co-organizer of departmental seminar series 2009