thesis.eur.nl thesis fleur de groot.docx  · web viewsocial network formation and clusters: a...

106
Social network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park Fleur Louise de Groot Student Number 335947 Master of Entrepreneurship, Strategy and Organization 0

Upload: others

Post on 27-May-2020

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

Social network formation and clusters: A two-way job mobility network analysis of the Leiden Bio

Science Park

Fleur Louise de Groot

Student Number 335947Master of Entrepreneurship, Strategy and OrganizationErasmus School of Economics, RotterdamSupervised by Sandra PhlippenCo-read by Completed December 2010

0

Page 2: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

Executive summary

This study investigates the degree to which a high tech cluster is embedded in a local

network, the characteristics that a well-functioning cluster network possesses, the

different types of actors involved in cluster networks and the central position of specific

actors in a cluster network. It proposes a new methodology in cluster network analysis

which includes both scientific and higher management team member employees in the

network analysis simultaneously, based on the importance of a combination of scientific

and market knowledge in high technology industries. Specific network properties are

linked to increased network performance and cluster success. The cluster network is

analyzed using job mobility networks of scientists and higher management team

members in the Leiden Bio Science Park in the Netherlands. The results imply that

successful clusters are highly embedded in local networks. The cluster network possesses

qualities that are inherent to well-functioning networks, which has positive implications

for cluster performance. The theoretical and empirical results support the inclusion of

management team members in future analyses and support the use of policy measures to

stimulate job mobility in clusters.

1

Page 3: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

Table of Contents

Executive summary.....................................................................................................11. Introduction..................................................................................................................32. Theoretical Background...............................................................................................7

2.1 The Embeddedness of Networks...............................................................................82.2 Network Topology...................................................................................................152.3 Types of Actors in Cluster Networks......................................................................182.4 Central Actors in a Network....................................................................................23

3. Methodology..............................................................................................................243.1 Local Embeddedness Measures...............................................................................273.2 Topological Characteristics of Networks................................................................283.3 Types of Actors........................................................................................................323.4 Central Actors: Bridging and Brokerage.................................................................33

4. Data Collection..........................................................................................................354.1 Scientist Network.....................................................................................................374.2 Management Team Member Network.....................................................................39

5. Results and Analysis..................................................................................................405.1 Level of Embeddedness of the LBSP Scientist Network........................................405.2 Topology of the LBSP Scientist Network...............................................................435.3 Type of Actor Networks..........................................................................................465.4 Central Actors in Combined Network.....................................................................48

6. Conclusions................................................................................................................506.1 Level of Embeddedness...........................................................................................506.2 Topology of a Successful High-tech Cluster...........................................................516.3 Type of Actors in Cluster Networks........................................................................526.4 Central Roles in High-tech Cluster Networks.........................................................53

7. Limitations.................................................................................................................55Acknowledgements....................................................................................................58Appendix....................................................................................................................59Bibliography..............................................................................................................60

2

Page 4: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

1. Introduction

High-technology (high-tech) clusters are a popular agenda item for local and national

governments world-wide. Clusters can be defined as a geographically concentrated group

of firms active in similar or closely connected technologies and industries with both

horizontal and vertical linkages (Dahl, 2002). Resources are invested into developing

high-tech clusters as it is widely believed that the clustering of companies, educational-

institutes and knowledge institutes stimulates an innovative climate which in turn will

benefit the economy, education and the employment opportunities1. More specifically

they are said to catalyze economic transformation, drive growth, enhance stability and

provide a chance for economic success (Mallet, 2004; Porter, 1998). High-tech clusters

are for example supported on a European level as part of the ambitious Lisbon strategy

set out in 2000, which aims to make the EU the most competitive economy in the world

by 2010.

Considering the economic importance yet relative scarcity of successful high-tech

clusters, a large amount of literature has probed into the underlying dynamics that

influence their formation and performance. The existence of ‘social networks’ in clusters

has emerged as an important factor in the determination of general cluster success

(Breschi & Lissoni, 2003; Gulati et. al.; 2000). Social networks provide benefits for

clusters by providing ‘social capital’ (Inkpen & Tsang, 2005; Uzzi & Gillespie; 2002).

Social capital can be defined as the ‘aggregate of resources embedded within, available

through, and derived from the network of relationships possessed by an individual or

organization’ (Inkpen & Tsang, 2005). A firm or individual can thus benefit from being

embedded in a social network through improved access to resources and information.

A central mechanism through which social networks in clusters form is through the

mobility of employees between firms (Dahl, 2002; Franco & Filson, 2000; Klepper,

2002). This ‘job mobility’ consists of employees moving between existing firms as well

1 Leiden Bioscience Park website http://www.leidenbiosciencepark.nl/about_leiden_bsp accessed on 12-09-2010

3

Page 5: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

as leaving firms to create start-ups within the cluster. As these mobile employees remain

related to their former colleagues, while at the same time forming new relations within

their new company, a ‘job mobility network’ is created. The evident importance of job

mobility for cluster success has in turn sparked the interest of scholars in the functioning

and dynamics of the labor markets and job mobility networks in clusters (Casper, 2007;

Casper & Murray, 2005; Higgings & Gulati, 2003).

While there is qualitative and quantitative evidence that looks into the effects that labor

market variations have on micro level job mobility networks, limited evidence is

available on the specific mechanisms at work in job mobility networks of clusters on a

micro level. Exceptions are the studies by Casper and Murray (2005) and Casper (2007).

The former looks into the effects that macro-level institutions have on micro-level job

mobility network dynamics in the successful bioscience clusters of Cambridge and

Munich, while the latter looks into the development of the San Diego bioscience cluster.

These studies find that in all three cases well developed job mobility networks exist, but

contain limitations in terms of data set completeness, the types of actors included in the

networks and a bias towards pro-publishing firms.

Considering the limitations of the studies by Casper and Murray (2005) and Casper

(2007), the general lack of research in this area and the economic importance of clusters,

it is of interest to verify previous results and to refine and extend previous methods. In

order to do so this study will perform a job mobility network analysis of a bioscience

cluster based on four main research questions. These research questions are;

1. To what degree is a successful high-tech cluster embedded in a local network?

2. What are the characteristics of the topology of a successful high-tech cluster

network?

3. What type of actors are involved in the functioning a high-tech cluster network?

4. What type of actors play a central role in the functioning of a high-tech cluster

network?

4

Page 6: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

Improving and extending previous research on job mobility networks in successful

clusters by answering the above questions can provide additional evidence for policy

makers that job mobility is a potentially valuable tool for stimulating cluster growth and

performance. More detailed information about the empirical characteristics of job

mobility networks can help identify target areas or actors for which specific policies can

be designed.

In order to attach any kind of value to a network measure one can either compare the

network over time or one can compare it to a real or random network. In this study the

research questions will be answered by analyzing the job mobility network of scientists

and management team (MT) members in the Leiden Bio Science Park (LBSP) in the

Netherlands. The networks will be analyzed by comparing selected network measures to

previous studies and/or to randomly generated networks. This research aims to provide

additional empirical evidence for the existence of dynamic job mobility networks in

successful clusters. The first goal is hence to provide improved evidence that the LBSP as

successful cluster is embedded in a local network (cluster network). The second aim is to

identify and demonstrate the topological characteristics of well-functioning cluster

networks. Thirdly this study aims to support the inclusion of different types of actors in

social network analyses and recommends a change in current cluster network analysis

methodology. The fourth aim is to illustrate the importance of central actors in a network

and to identify their characteristics in a cluster network that includes both scientists and

MT members.

The study will be structured as follows. First a theoretical framework will be built in

Chapter 1 which supports the research questions. Next the method of data collection will

be presented in Chapter 2. Chapter 3 will explain the data collection process, after which

the results of the analyses will follow in Chapter 4. The LBSP scientist job mobility

network will be presented in Sections 4.1 and 4.2, and a comparison will be made with

the networks as found in the bioscience clusters of Cambridge and Munich. In Section 4.3

the MT member job mobility network will be presented. In Section 4.4 the central

employees in the combined network will be identified. Lastly in Chapter 5 the

5

Page 7: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

conclusions will be presented and the implications of the results and limitations of the

research methods and theory will be discussed.

6

Page 8: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

2. Theoretical Background

There is increasing recognition in literature that social capital is the key to building and

sustaining a successful high-tech cluster. A ‘successful cluster’ can be interpreted in a

broad sense, meaning that the firms in the cluster are increasing in number, surviving,

growing or making a profit. Talented managers and scientists provide high-tech firms

with the human and social capital which forms the foundation of their innovative

capabilities and performance (Becker, 1962; Casper & Murray, 2005; Higgings & Gulati,

2003). Social network theory suggests that it is especially the process of interaction

between skilled individuals within a cluster which determines cluster success. Social

capital is thus formed through social networks. Specifically of importance are said to be

the informal links between scientists, engineers and managers. These informal links raise

the innovative capacity of a high-tech cluster through disseminating technological and

market intelligence (Casper, 2007).

A social network can form based on any type of linkage between individuals and firms,

wherefore many different types of cluster networks have been analyzed. Next to job

mobility networks, previous studies have looked at co-patenting networks, formal

alliance networks, R&D networks and personal networks for example (Johannisson,

1998; Owen-Smith & Powell, 2004; Porter et. al., 2005; Zaheer & George, 2004). This

study will consider a job mobility network as theory and research suggests that a strong

link exists between the social network of an individual within a cluster and their job

mobility network (Casper, 2007; Casper & Murray, 2005; Almeida & Kogut, 1999). A

related observation is that firms embedded within regions with a decentralized culture of

high mobility and knowledge diffusion have a regional advantage over firms who are not

(Casper, 2007; Herrigel, 1993; Sabel, 1992; Saxenian, 1994; Storper, 1997).

Based on this research it is suggested that the existence of a high level of job mobility

and a well developed job mobility network in a cluster creates various forms of social

capital. Social capital will be considered to be any benefit that a firm derives from being

‘embedded’ in a social network, as is implied by the definition. As ‘job mobility’ is a

7

Page 9: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

prerequisite for a ‘job mobility network’, it should be noted that they are two distinct

concepts. The benefits that a cluster derives from job mobility in general and the benefits

that it can derive from job mobility networks specifically will therefore be addressed

separately in Section 2.1, which concerns the benefits for a cluster of being ‘embedded’

in a network.

It is also interesting to look at the specific topological characteristics of a network. How

can the structure of a network influence the formation and performance of the network in

terms of creating social capital? Different types of network structures will be described in

Section 2.2. As mentioned, a social network is formed by relations between any

individuals of firms within a cluster. Why do cluster network studies include only one

type of actor, often the scientists? Do managers contribute equally to the formation of

social capital? Sector 2.3 will support the inclusion of both scientists and managers in

cluster network analysis. In the combined network of scientists and managers there will

always be some actors who perform more important functions than others. Is one type of

actor more central than the other, and what function do these central actors play in the

overall network? The theory concerning central actors will be addressed in Section 2.4.

2.1 The Embeddedness of Networks

While there are many definitions of embeddedness, this paper will use Granovetter’s

original formulation that embeddedness refers to the notion that all economic behavior is

embedded in social context (Granovetter, 1973). Specifically structural embeddedness

will be considered, which captures the extent to which an entity is entrenched in a

network of relationships (Grewal et. al., 2006). Based on this definition of structural

embeddedness, a well developed job mobility network with a high level of connections

can be directly related to a high level of embeddedness of firms and individuals in a

cluster network.

In recent literature embeddedness has been considered mostly as an inter-organizational

phenomenon where the level of embeddedness depends on the quality of the relationships

between firms (Almeida & Kogut, 1999; Uzzi, 1996). This paper supports an individual

8

Page 10: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

approach however, where a job mobility network is used as a proxy for analyzing the

level of local embeddedness in the social network (Casper & Murray, 2005). This

individual approach as opposed to a firm-based approach is based on the individualistic

nature of social capital, job mobility and social networks. The underlying thought is that a

high level of inter-firm mobility of employees within a cluster leads to an extensive job

mobility network where many individuals are acquainted with one another through past

or current employment. These connections between individuals simultaneously form

connections between firms and thus connections within the cluster. As one type of

connection automatically constitutes the other type of connection, there will be no

differentiation made throughout the paper between the embeddedness in a network of

individuals, firms or clusters. Each constitutes the same idea.

Now that the theory behind the method of analysis has been explained, it must be further

clarified why the level of embeddedness is of importance for a high-tech cluster. Next to

the theories about the formation of social capital that have been mentioned, another

theory that considers some of the benefits and processes involved in the embedding of a

firm in a network is the theory of embedded local economic growth. It states that local

economies are “islands of superior productivity, integrated into a global mosaic of

production that brings the reward of sustainable local accumulation” (Taylor, 2005). This

superior productivity is said to be the result of the complex process of ‘embedding’,

which Taylor referred to as “the incorporation of firms into place-based networks

involving trust, reciprocity, loyalty, collaboration, co-operation and a whole raft of

untraded interdependencies” (Taylor, 2005). This process of embedding creates social

capital, which in turn fosters the creation of products, services, processes and ideas that

are not appropriated by individual actors but are collectively shared in the system

(Leborgne & Lipietz, 1992).

Other benefits generated by being embedded in a network that have been mentioned are

that it will provide improved opportunities for learning, access to technologies and

resources, increased legitimacy and an opportunity to improve the firms its competitive

position (Dyer & Singh, 1998; McEvily & Zaheer, 1999; Nohria & Eccles, 1992). Three

9

Page 11: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

specific mechanisms involved in the embedding of a cluster network that can lead to such

benefits will be considered in more detail in Section 2.1.2 to 2.1.4, namely the function of

a network as labor pool, the safety net function of the network and the overlap between

professional networks and social networks. First of all however Section 2.1.1 will support

an important general benefit of job mobility for the cluster.

2.1.1 Job Mobility as Mechanism for Knowledge Spill-overs and Knowledge Diffusion

A knowledge spill-over is an exchange or movement of ideas amongst individuals or

between firms in which there is no transaction involved. Knowledge spill-overs can

benefit firms in terms of leading to innovation and growth. They are most likely to occur

in a specialized industry where firms are located in close proximity to one another and

when these firms are involved in highly innovative, high-tech industries such as in a bio

science cluster. This is so because the type of knowledge involved in such industries is

highly tacit and therefore requires face-to-face interaction for transfer. Tacit knowledge

by definition is non-codifiable and cannot be formalized (Audretsch & Feldman, 2003).

Casper and Murray (2005) mention that knowledge spill-overs can take place through

various mechanisms, including through job mobility, spontaneous social interaction

between employees and planned interaction. Social connections are thus a requirement in

order for knowledge to spill-over.

The difference between knowledge spill-overs and knowledge diffusion is important to

note here. In relation to job mobility, knowledge diffusion refers to knowledge that is

embodied in an employee which is included in the economic value involved in switching

jobs, such as the salary of the employee and the recruitment costs. Knowledge diffusion

is also highly valuable for the performance of a cluster and is often a driving force behind

head hunting and acquiring new talented employees with a valuable knowledge base. A

knowledge spill-over here is the ‘extra’ knowledge that such a new employee brings to a

firm that is not included in their economic value.

The reason that knowledge spill-overs and diffusion are strongly linked with job mobility

is because “ideas are embedded in the minds of individuals” (Feldman, 2000), and

10

Page 12: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

movement of these individuals between firms allows employers to benefit from the

knowledge that has been accumulated throughout their careers with other firms (Dahl,

2002). Knowledge therefore flows between companies with the mobility of employees.

This includes both the movement of an employee between 2 existing firms and the

situation in which an employee creates a start-up. Both knowledge spill-overs and

knowledge diffusion are thus related to cluster benefits in terms of human capital as

opposed to social capital.

The existence and importance of knowledge diffusion in relation to job mobility is well

supported by literature. Almeida & Kogut (1999) for example analyze data on the inter-

firm mobility of patent holders and empirically show that the inter-firm mobility of

engineers influences the local transfer of knowledge and that this flow of knowledge is

embedded in regional labor networks. The importance of the knowledge that employees

involved in a start-up have accumulated in their parent firm is shown to be a very

important feature in a paper by Klepper (2002), who finds that the success of a new firm

in the US automobile industry is to a large degree determined by the experience and

background of the founder.

Evidence supporting the existence of knowledge spill-overs in relation to job mobility on

the other hand is not as straight forward. The concept of knowledge spill-overs itself has

been criticized strongly by Breschi and Lissoni (2001) who describe it as a ‘black box’.

They state that the evidence supporting knowledge spill-overs is ambiguous, that

scientists over-interpret this theoretical concept and that research does not go into the

mechanisms through which knowledge spill-overs actually occur. Dahl (2002) embraces

this criticism in his paper on the mobility of engineers in Danish knowledge clusters and

shows that job mobility is in fact a mechanism through which knowledge spill-overs

occur. A collection of other literature also supports the role that job mobility plays in the

creation of knowledge spill-overs and considers it to be a central mechanism fueling

rapid innovation (Almeida & Kogut, 1999; Casper, 2005; Saxenian, 1994). Evidence in a

study on patenting and job mobility by Kim & Marschke (2005) supports that the

11

Page 13: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

turnover of scientific personnel can be an indication of technological knowledge

diffusion and spill-overs.

Job mobility as mechanism for knowledge diffusion and spill-overs may specifically be

important for the LBSP because the other routes through which knowledge can be

transferred (spontaneous social interaction between employees and planned interaction)

are not likely to be fruitful. The main reason for this was mentioned by park manager

Annelies Hoenderkamp, who said that even when other firms might not benefit from the

information, many firms prefer to keep their innovative ideas a secret and thus

consciously prevent knowledge spill-overs and diffusion from occurring (Hoenderkamp,

2009). Another reason why job mobility may specifically be an important knowledge

spill-over and diffusion mechanism for the LBSP is because about 80% of the current

companies are spin-offs, which were started mostly by current or former employees of

the Leiden University Medical Centre (LUMC). The high number of spin-offs itself is

evidence that there may be a significant amount of knowledge spilling over and diffusing

through job mobility, which is contributing to the growth and success of the cluster.

2.1.2 Job Mobility Network as Labor Pool

An extensive job mobility network can be highly valuable for a cluster by functioning as

a labor pool from which firms can recruit employees. Both the size and the quality of the

potential labor pool can influence high-tech cluster performance and growth. The degree

to which firms have access to such a pool of talented highly specialized employees is

crucial for high-technology firms (Becker, 1962; Higgins & Gulati, 2003). This coincides

with the statement that the structure and dynamics of the local labor market have

important performance implications for high-tech clusters (Almeida & Kogut, 1999;

Audretsch & Feldman, 1996).

According to Saxenian (1994) one of the reasons behind the success of the Silicon Valley

cluster was the strong regional labor market for engineers, scientists and workers. People

in the area were mobile within the spatial boundaries of the region, but were not

necessarily bound to a certain company and would switch jobs relatively easily and often.

12

Page 14: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

This kept both the firms in the area and the employees competitive and operating on the

edge of technology, which improved the overall innovative capacity of the region.

On top of having access to a pool of talented employees, the job mobility network can

provide valuable information to a company on the level of expertise and talent of a

potential future employee. A current employee can recommend a former colleague for a

new job based on a positive previous experience of co-employment. An employee may

even be recruited based on positive information provided by a third party.

An example of a situation in which the job mobility network functions as a labor pool is

when a university scientist is involved in the startup of a spin-off company and

subsequently recruits former colleagues to join his team. Next to the fact that the founder

is aware of the existence of his former colleagues as potential employees, the chance that

these former colleagues will join a possibly risky new venture is larger when they have

some social connection to the potential new employer. This is because through previous

co-employment or even through second-degree ties, a certain degree of mutual trust can

form between them which can overcome part of the information asymmetry involved.

That direct ties and second-degree ties can overcome information asymmetry in

economic transactions is generally supported by empirical evidence as found in

Granovetter (1985), Gulati (1995), Shane and Cable (2002) and Uzzi (1996).

2.1.3 Job Mobility Network as Safety Net

In order to be able to have a high level of job mobility, next to a pool talented employees

being available, the success of the firms in the cluster is in part determined by their

ability to entice skilled managers and employees to leave lucrative and often safe jobs to

join a new firm. This is specifically relevant for start-up ventures, as these usually entail

the highest levels of risk, while at the same time being crucial for the innovativeness and

growth of a cluster. According to Casper (2007) the reason why risk-adverse scientists

and managers are willing to leave their jobs for risky new ventures is because within

successful clusters the embeddedness of individuals within social networks makes it safe

to do so.

13

Page 15: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

Casper (2007) highlights the importance of a job mobility network as a safety net on

which individuals can rely in case a new venture fails and they lose their job. The easier it

is for an individual to be re-employed locally, the more likely they may be to switch to a

failure prone job. This results in a ‘recycling mechanism’ of employees (Brahami &

Evans, 1999; Casper, 2007). A relatively smooth recycling mechanism was also observed

by Saxenian (1994) in the Silicon value cluster, who mentioned that “moving from job to

job in Silicon Valley was not as disruptive of personal, social, or professional ties as it

could be elsewhere”. Casper (2007) even suggests that if such a safety-net does not exist

it can seriously hinder the success of a cluster and could be a reason why clusters fail,

even if they have reached a sufficient size for survival.

A career affiliation network as safety-net in combination with the existence of a talented

labor pool goes hand-in-hand with a higher level of job mobility. If the labor market

functions smoothly it enables firms to alter their research strategy quicker through hiring-

and-firing when necessary. Firms within the cluster can thus react to market changes

faster than competitors outside of the cluster. Such organizational flexibility can provide

firms with a considerable competitive advantage which can ultimately contribute to

overall cluster performance.

2.1.4 Job Mobility Networks and Social Networks

The last mechanism through which a job mobility network and the related level of

embeddedness can influence cluster success is through the strong connection between the

job mobility network or the ‘professional’ network of an individual and their social

network or friendship network. Individuals can form social connections through

professional connections, by for example meeting contacts of colleagues at social events.

The opposite can also occur, as a social connection can also lead to a professional

connection. This happens for example when the founder of new company recruits a

fellow researcher whom he knows from college or from a social club outside of the

cluster. Also considering the fact that individuals tend to live at a reasonable distance

from their work, it is likely that an overlap exists between the social and professional

14

Page 16: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

network of an individual. These social network connections would increase the total

number of ties in the network within a cluster if they were to be identified and included.

Social networks of people play a similar role in stimulating knowledge spill-overs and

job-mobility within clusters as job mobility networks do. There is evidence that the social

network of top-managers of a company in terms of the network size, the strength of

connections and the distance between actors, is positively related to firm performance in

high-tech industries (Collins & Clark, 2003). Saxenian (1994) mentions that social

networks in Silicon Valley increased labor mobility across firms and by doing so created

an additional mechanism of knowledge diffusion. Casper (2007) links sustainable

networks in clusters to the existence of dense social networks across key personnel.

Breschi and Lissoni (2001) on a similar note highlight the degree to which social and

professional contacts overlap and mention that ‘epistemic proximity may arise from

shared work or study experiences, or former cooperation efforts that required face-to-face

contacts and a high degree of socialization… Although highly dispersed in space,

members of these epistemic communities share more jargon and trust among them than

with any outsiders, no matter how spatially close’.

2.2 Network Topology

While the embeddedness of a high-tech cluster from an individual perspective can reveal

much about the network and labor market dynamics, the topology of a network can also

provide insightful characteristics from a more holistic point of view. The topology of a

network refers to the physical layout pattern of interconnections between actors. There

are many different types of measures that can be taken into consideration, several of

which are straight forward in their implications for the functioning of a social network.

These measures will be mentioned in the methodology section. Two topological

characteristics of social networks that do require theoretical background as to their

interest for the analyses are the core-periphery model and the small-world network

phenomenon.

15

Page 17: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

2.2.1 Core-periphery Model

While the benefits of clusters are well-researched, the process through which clusters

become sustainable remains to be further researched and developed. One study which

does consider cluster development over time based on career history data is by Casper

(2007). He finds evidence that there is a strong core of employees who were central to the

development of the San Diego bio tech cluster. This core group evolved as the scientists

had all worked together at one of the first companies in the cluster, which went bankrupt.

Most continued their careers in the cluster by starting new ventures or joining existing

companies.

One way in which such a core can thus develop is based on historic reasons, as a core of

actors may have been part of the original network present at beginning of the formation

of the cluster, as was the case in San Diego. This type of structure in the development of

a cluster is known as a ‘backbone’ as discussed by Casper (2007). Multiple studies

support the idea that a small group of valuable employees of one of the oldest firms in a

network can become the founders of new ventures within the cluster and in this way

become central in both the process of development and the current functioning of the

network (Casper, 2007; Feldman, 2003; Sorensen, 2005).

The core can function as a catalyst in the emergence of a network, as it entails a group of

initially tied actors to which new network members can latch on to, which promotes the

growth of a cohesive network (Casper, 2007; Powell et. al, 2004). These core members

are hence crucial to the embedding process in these networks.

Another study related to the development of clusters by Feldman (2003) considers the

idea that ‘existing firms may serve as anchors that establish skilled labor pools,

specialized intermediate industries and provide knowledge spillovers for new technology

intensive firms in the region’. She suggests that one large and successful firm can form

the basis of the direction of specialization and of the growth of a cluster. Similar to the

‘backbone’ theory of Casper, she supports the idea that such an anchor firm can supply a

16

Page 18: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

cluster with potential entrepreneurs who can use the knowledge of their parent firm to

create start ups. This supports the existence of a core-periphery structure in successful

high-tech cluster networks

It will be interesting to see whether the LBSP scientist network has evidence of a central

core of scientists who were perhaps part of such a backbone structure as mentioned in

cluster development literature. In an ideally locally embedded network however, where

all ties are equally dispersed, such a backbone or small core should not be visible. That it

is not visible however does not need to mean that it does not exist.

2.2.2 Small-world Phenomenon

In order to measure the performance of a network, an interesting method has been

developed by Watts (1999) and others based on the analysis of real-world networks,

known as the ‘small-world’ method. They noted that in large real-world networks, there

is often a structural pattern that seems paradoxical. On the one hand they found that in

general the path length, or average number of links along the shortest paths between

actors in a network, is relatively short, regardless of how large a network is. A well-

known example is the “6 degree of freedom” principle, which states that in order to get

from any one person in the world to another one must pass through only 6 connections.

On the other hand it turns out that there is a large degree of clustering visible locally,

where people in the same geographical area such as in neighborhoods are all connected to

one another. This means that the density in such neighborhoods is much higher than if

connections had been formed there at random. This is paradoxical, as the former states

that we are connected to everyone in the world through short distances, while the latter

suggests that at the same time we live in a narrow social world, where everyone knows

everyone. Examples of small-world networks are power grid networks and friendship

networks.

When a network has small-world properties it is considered to be a pareto optimal

equilibrium as it means that the network is both robust and efficient (Wilwhite, 2001).

17

Page 19: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

Wilwhite (2001) shows this by proving that there are significantly lower search and

negotiation costs involved in a hybrid model of economic exchange when the network

possesses small-world characteristics. It is thus an ideal network structure for the spread

of information and hence can positively influence the speed of innovation within a

network. He suggests that there are even ‘private incentives for such a system to arise’,

which for cluster networks can be interpreted to mean that it is worth investing resources

into stimulating the formation of a small-world network structure in order to improve the

performance of the network. Ultimately a small-world structure can thus contribute to a

well-functioning network which can positively influence cluster success.

2.3 Types of Actors in Cluster Networks

Now that the benefits of being embedded in a cluster network based on job mobility and

the theories of well-developed networks have been discussed, a step-back will be taken

and the actors involved in a high-tech cluster network will be considered. Who is actually

responsible for the formation and performance of a cluster network? Most literature

focuses on the role of scientific employees in the transfer of knowledge and the formation

of labor mobility networks. The characteristics of scientists in the labor market will

therefore be considered first in Section 2.3.1.

It can be argued however that next to scientists, another type of employee of high-tech

firms also plays an important role in the formation and functioning of a cluster network,

namely the higher management team (MT) members. Why is the job mobility networks

of scientist usually analyzed while senior managers and (star) scientists both can be

assumed to play an important role in the determination of firm success? MT members

strongly influence the strategy and direction of a firm and are responsible for bringing a

product to the market successfully. This can be said to be just as important for firm

success as the initial innovation development by scientists. This also holds for high tech

firms involved in non-R&D firms, such as manufacturing.

The importance of MT members in cluster networks will be further elucidated in Section

2.3.2. It will be argued that the MT members can influence the level of embeddedness

18

Page 20: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

and the functioning of the network in the same way as scientists can. This includes a role

in the process of knowledge spill-overs, the process of labor pool formation, their

involvement in the formation of a safety net and the importance of their social network

connections.

2.3.1 Scientists

Current literature mainly focuses on the scientific employees of firms when analyzing

networks (Casper et. al, 2005; Saxenian, 1994). This is as mentioned because scientists

are suggested to play an important role in knowledge spill-overs and knowledge diffusion

due to the high level of tacitness of the knowledge involved in their high-technology

functions. Working on the edge of technology, their knowledge may be unique and solely

transferrable to another firm by the scientist switching jobs. Their knowledge tends to be

highly specific and crucial in the process of innovation.

An interesting question that follows is whether scientific employees in high-tech

industries are highly mobile between firms or not. Saxenian (1994) noticed that within

the Silicon Valley scientists were indeed highly mobile and that this contributed to the

performance of the cluster. She suggests that scientists can be so highly dedicated to their

field of study and to advancing technology that they may not be loyal to their firms, but

rather to science. She observed that scientists would switch jobs easily as long as they

were provided with better research facilities, a higher salary or more research funds for

example. Ackers (2004) notes that it is the nature of scientific jobs that leads to a high

mobility of scientists, as she says that career progression in scientific research ‘demands a

very high level of mobility’.

As scientists are likely to be mobile, the question arises whether they are mobile locally

(within the cluster) and/or non-locally. Job mobility tends to be geographically bound as

individuals are generally reluctant to move. An employee is therefore likely to prefer a

new job close to home as opposed to having to move to a new city or even country. The

geographic pull-factor is stronger in case a scientist has a partner, has children, has family

19

Page 21: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

in the area or has a well-established career (Acker, 2004). It can therefore be expected

that scientists may be mobile especially within the cluster.

There is also evidence that scientists are motivated to be mobile globally based on

institutional factors. Research by Dickson (2003) on the migration of Italian scientists

supports that they are not motivated to switch firms based on financial gains, but are

rather concerned with the general funding of science and the influence that this has on

their ability to perform their job. Dickson (2003) says that scientists ‘leave their home

countries … to seek an environment in which they can “work effectively with enthusiasm

and support”. Other institutional push- and pull factors include the degree of contractual

insecurity, the cost of living, pensions and social benefits such as healthcare and

childcare (Acker, 2004; Dickson, 2003).

Is the institutional setting in the Netherlands on a national level a potential push- or pull-

factor? Are there specific push- or pull factors of the LBSP on a local level? These are

interesting questions for further research, but cannot be addressed in this paper based on

the local scope and micro-level of analysis. The institutional setting will however be

considered as possible influential factor in the determination of local job mobility levels.

It can lead to national differences between job mobility levels in clusters. It is therefore

interesting to gauge the mobility of scientists not only compared to other clusters, but

also compared to employees from the same cluster and thus with the same institutional

setting.

2.3.2 Management Team Members

In the network analyses of high-tech industries, the role of the MT seems often

overlooked. Senior managers as opposed to other non-scientific employees specifically

play an important role in the relationships that companies have with the environment in

determining the direction and strategy of the company for example. They hence clearly

have an influence on the success of the company. If they play such an important role for

the firm, are they not also likely to be important in the formation and functioning of the

cluster network? Casper (2007) supports the investigation of other types of actor

20

Page 22: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

networks and performs a job mobility analysis of senior managers. He provides evidence

that senior managers form dense social networks within the San Diego bio science

cluster. Similar to scientists, MT members are expected to be mobile locally, as their

movement is geographically bound while career advances or entrepreneurial ventures

often call for a job switch. In the following paragraphs it will be argued that MT

members play a similar role in the creation of social capital within a cluster as scientists

do, and that ultimately a cluster network is more complete once the two networks are

combined.

Firstly it can be argued that contrary to intuition, MT members can also be involved in

knowledge spill-overs. They can play a role in the spill-over of both tacit knowledge and

generic knowledge. Tacit knowledge is a broad concept, which includes not only

scientific knowledge but also technical knowledge, which can be held in managers in the

form of either unparalleled previous experience in a unique line of business or by a

unique combination of scientific and industry knowledge for example. Technical

knowledge can be highly specific and tacit in that it is often transmitted in “the jargon of

a much closer and restricted community (an ‘epistemic community’). Members of the

community learn it by joining it to practical experience, and cannot transmit it to any

outsider by informal means” (Breschi & Lissoni, 2001). Most managers in the LBSP also

have a scientific background and are likely to also understand at least an important part of

the scientific jargon next to their technical and managerial knowledge.

As the definition of a knowledge spillover is the exchange of ideas amongst individuals

without an economic transaction, ‘generic’ management knowledge can also spillover.

General management knowledge of MT members may specifically be valuable because it

may not be as generally applicable as one might expect. As mentioned such a highly

specific and unique industry like the bio technology (bio tech) industry involves a high

degree of jargon and specific knowledge, which is also used in the general management

knowledge of such a company. The generic knowledge is thus often combined with tacit

knowledge in the form of specific jargon as used in the epistemic community, and in this

way can be valuable as a spill-over.

21

Page 23: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

Next to knowledge spill-overs, the related concept of knowledge diffusion is also a

valuable aspect of MT member job mobility. The unique qualities that an MT member

possesses are valuable assets for firms which they are willing to pay a price for. An MT

member who for example has the knowledge and experience of setting up a bio-tech

company has an exclusive combination of generic and tacit knowledge, which when used

in another company or start-up can contribute to cluster success.

Secondly, next to knowledge spill-overs, MT members are expected to be involved in the

same cluster-wide benefits of being embedded in a job mobility network as scientists are.

MT members are also a part of the valuable labor pool within the cluster as they may

have a unique combination of understanding of the industry, scientific knowledge and

management know-how as mentioned. Having an experienced founder with the relevant

industry know-how can clearly benefit a new venture. Such new ventures do not just

benefit from the knowledge base of their managers, but also derive benefits from their

professional and social networks. Sorensen (2005) for example suggests that founders of

high-tech firms commonly recruit friends and former colleagues. Higgins and Gulati

(2003) find that recruiting talented senior managers is in fact strongly linked to the

success of bio tech firms, which also links back to the importance of MT member job

mobility and the availability of a labor pool.

Thirdly, the earlier mentioned high level of job mobility of MT members and their unique

knowledge base are intertwined with the function of the network of an MT member as a

safety net in case of unemployment. It is vital for the development of a high-tech cluster

that MT members willingly leave safe jobs to found new ventures, or start new ventures

next to another job. A safety net increases the chances that MT members will be willing

to engage in such risky entrepreneurial new ventures. The safety net of MT members in

part is already determined by the nature of board functions, as MT members are often a

member of multiple boards at a given time. If one of the firms in which a multi-board

member is involved fails, he or she will not be left unemployed. A large percentage of

MT members in the LBSP dataset are members of multiple boards at the same time.

22

Page 24: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

Fourthly, again similar to scientists, MT members are likely to have a strong overlap

between their professional and social networks, as supported by Sorensen (2005). This

will strengthen and expand the overall network in which they are embedded. These four

factors suggest that MT members perform an important function in the formation and

performance of a cluster network and hence of the overall cluster. It will therefore be

interesting to see whether the characteristics of the scientist and MT networks differ or

not, considering these theoretical similarities. The overall importance of MT members in

a local cluster network as theorized supports the notion that they ought to be included in

high-tech cluster network analysis.

2.4 Central Actors in a Network

The novel combined network of the scientist network and the MT network can give

interesting clues as to the structure and characteristics of a more complete cluster

network. As the combined network consists of 2 possibly distinct types of epistemic

communities, these networks may not be fully integrated. Based on the importance of the

combination of scientific and industry knowledge in high tech industries, it could be

predicted that those individuals with the highest degree of aptitude in combining the

scientific and the management knowhow will occupy more central positions in the

network than others.

Interestingly enough there are several members in the data sets who belong to both the

scientific data set based on previous patents and the MT member data set based on

current management positions. Theoretically actors in such a position could be potential

bridge-builders between the two types of networks and could have a broker position

within the network. Such a broker can control the flow of information through the

network, making this a potentially powerful and vital position in the network. At the

same time such a position can be a single point of failure, as part of the network may not

be able to reach one another if the actor drops out of the network.

23

Page 25: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

The existence of a link between the scientists and MT members is highly valuable as it is

crucial for the success of the bio tech industry that highly innovative scientific

discoveries are successfully introduced to the market. This requires skilled managers with

industry knowledge, an industry-wide network and a combination of scientific and

management experience, which are thus indispensible in high-tech industries. Boschma

(2005) for example supports that a combination of technical and market knowledge is

valuable specifically for the process of knowledge spill-overs. He states that is it not

sufficient for a high-tech firm to simply possess the often highly tacit knowledge, but that

is crucial for the absorption process of knowledge as described by Cohen and Levinthal

(1990) that a firm also possesses the technical and market competencies in dealing with

the specific market and technology related to that knowledge (Boschma, 2005). These

competencies may thus be embodied in the scientist-managers, and proof of a broker

position could support the dependence of the network upon them.

Another mechanism through which these broker functions are of importance to the

overall network is through their contribution to the desirable small-world properties as

mentioned in Section 2.2.2. Actors with a broker function contribute to a low path length

and a high cluster coefficient based on the nature of a broker function. The central

function as a broker can also depend simply on the job history of an individual and on

their personal level of mobility as opposed to the type of function the individual has. The

position in the network of these management-scientists will therefore be analyzed.

3. Methodology

In order to study a high-tech cluster network the bio tech industry has been chosen as a

relevant industry as it is known for the high levels of technology involved and the speed

of innovation and change. Bio tech concerns the commercialization of scientific

discoveries related to genetic engineering, which is a relatively new industry with a high

level of potential and global competitiveness. Significant levels of resources are

employed to stimulate new firm development and bioscience cluster formation. It is also

an interesting industry as there are few firms actually profitable and because a strong link

has been suggested to exist between those companies that are successful and the degree to

24

Page 26: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

which they are linked to external networks of various kinds (Bathelt et. al., 2002; Porter,

1998; Powell et al., 1996).

In order to investigate the job mobility network in a cluster, the Leiden Bio Science Park

will be used as a case study. The LBSP is a useful and appropriate case as it is home to

more than half of all biomedical life science companies in the Netherlands. The city of

Leiden, the Province of South-Holland and the Dutch Ministry of Economic Affairs

consider the LBSP to be the life science hotspot of the Netherlands2. The centre also has

recently won the prize for best business park of the Netherlands in 2009. Leiden won the

prize for its daring choice to specialize in biomedical life science and for making a

success out of it.

The LBSP itself has recognized the importance of micro-level job mobility as they

mentioned that they have recently made plans to facilitate labor-mobility through the

creation of a labor-pool to “retain workforce and talent, provide jobs and improve

employability”3. As the LBSP is a successful cluster, and as successful clusters have been

linked to the existence of extensive job mobility networks in previous studies, this paper

is biased towards finding a well developed job mobility network. This is not a problem

however as the main aim of this study is not to prove the existence of such a network but

to further explore the underlying dynamics and characteristics of such job mobility

networks.

Employee careers are a vital instrument through which mobility, accumulated social

capital and human capital are established (Baker, 2000; Uzzi & Lancaster, 2003; Burton

et. al. 2003, Casper, & Murray, 2005). A career does not only encompass employment

with a traditional firm, but also includes co-patenting activities and faculty positions. In

combination with the informal connections established between employees, these

activities contribute to the human capital that individuals take to firms (Murray, 2004;

Casper & Murray, 2005). The job mobility network for scientists is therefore defined as

2 http://www.leidenbiosciencepark.nl/about_leiden_bsp3 Presentation by Annelies Hoenderkamp

25

Page 27: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

all previous employment at the LBSP, including faculty positions and co-patenting

activities.

A job mobility network is a social network which can be defined as a bounded set of

connected social units. The boundary of the network is defined by the firms in the LBSP,

as only current and former employees of the park are included in the data sets. The

‘connectedness’ of these actors is of a binary type, where a connection is established

based on whether people have worked for the same company, or whether they have not.

Employees are assumed to know one another regardless of whether they were employed

at a company at the same time, which is not seen as a problem based on three reasons.

Firstly the majority of the data covers an approximate 10 year time-span, wherefore the

likelihood that 2 individuals worked within the same firm is relatively large. Secondly,

many jobs in the data set are academic jobs, which tend to be long term affiliations,

sometimes even for life. This increases the likelihood that 2 actors worked for the same

institution at the same time. Thirdly, as the institutions and departments of the university

have been separated to the finest degree, there is again an increase in the likelihood that 2

actors worked there simultaneously. Even if actors did not work simultaneously, there is

a reasonable chance that they do know each other directly or indirectly.

The network will be looked at from an individual actor point of view as well as from a

holistic point of view, as local connections of actors can be important for understanding

the social behavior of the whole population. Firstly the method used to analyze the

scientist network will be clarified on an individual (embeddedness) and holistic level

(topology). Then the method for the analysis of the MT member network will be

explained briefly, as it is similar to the method used for the scientist network. Next the

method for the overall network analysis of central actors will be illuminated. For all 3

networks, the structure and characteristics can be best made visible through a network

visualization which will be made using the Netdraw application of UCINET. The random

networks have been constructed using the ‘Erdos Renyi’ random network function in

Ucinet and Netdraw.

26

Page 28: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

3.1 Local Embeddedness Measures

As opposed to the standard approach to embeddedness which takes the firm as unit of

analysis, this paper will consider inter-personal relationships as the basis of the cluster

network as described. As well-developed and successful cluster, the LBSP is expected to

be strongly locally embedded in the network. There is hence expected to be a high level

of mobility amongst scientist and managers alike. There are various approaches to

characterizing the extent and form of the embedding of actors in a network. Since there is

not one way of indexing the degree of embedding, multiple approaches can be used

(Hanneman & Riddle, 2005).

A network can be either compared over time, to other networks or to randomly generated

networks. The local embeddedness measures selected, which include the size of a

network, the distribution of components and the network density, will be compared to

those of the Munich and Cambridge clusters as analyzed by Casper and Murray (2005).

This is done because network data for these clusters is readily available and can

demonstrate relative values of these measures, as they are not indexed values. These

relative values can reflect important social conditions for embedding and will next be

considered in further detail.

3.1.1 Main Component Connectedness

The distribution of components looks at the total number of nodes in a network and how

many of those nodes are linked versus how many are not linked. Casper and Murray

(2005) call this the ‘degree of connectivity’. It is measured by the percentage of people

connected to the main component of the network, which is the largest group of connected

nodes. If the network is fully connected, a message that starts anywhere can eventually

reach everyone. If in one population most actors are embedded in at least one dyad, while

in the other population there are many isolated actors, the social structures are likely to

vary (Hanneman & Riddle, 2005). A study by Owen Smith and Powell (2004) found that

in the bio tech industry about half of the network members are generally in the main

component. Casper (2007) found that in the San Diego bio tech cluster 95% of the senior

managers was linked to the main component by the time the cluster had fully developed.

27

Page 29: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

In a perfectly locally embedded network the actors are expected to all be connected to the

main component, as opposed to being isolated. Based on theory and earlier findings it is

thus expected that a relatively high percentage of actors in the LBSP scientist network is

connected to the main component.

3.1.2 Density

The density of the network measures the amount of interconnectedness per actor.

For binary data, density is the ratio of the number of adjacencies that are present divided

by the number of pairs, so the number of actual ties in proportion to the possible ties

within the network. The problem with the existing measures of density is that they are

size-dependent, which is why it is important that the number of nodes for the LBSP

scientist data is similar to Cambridge and Munich, to allow for comparison. The relative

density of a network can give insights into for example the speed at which information

diffuses among nodes, and the extent to which actors have high levels of social capital. In

a strongly embedded cluster network with a high level of mobility, the actors are

expected to have a relatively high level of interconnectedness and the level of density is

expected to be high.

3.2 Topological Characteristics of Networks

The topology of a network considers a top down perspective. This is valuable in order to

get an understanding and description of the network population as a whole based on the

way in which individual actors are constrained by the texture of their relations to one

another (Hanneman & Riddle, 2005). In order to get an understanding of the topology of

a high-tech cluster, several measures and characteristics will be considered which will be

described in the following sections. This includes the degree of network centralization,

the core-periphery structure, the path length, the cluster coefficient and the small-world

properties of a network.

3.2.1 Network Centralization

28

Page 30: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

Network centralization is also known as global centrality, which measures the degree to

which a network is focused around a few central nodes (Scott, 1991). Freeman (1979)

based the measure of global centrality around the ‘closeness’ of actors and it is expressed

in terms of the distances amongst various points. Other definitions of network

centralization include “the degree to which relations are guided by formal hierarchy” and

“the degree to which an inter-organizational network is dominated by a few places"

(Irwin & Hughes, 1992). It is equivalent to the variance in network ties per actor. When

the variance in the number of network ties per actor is low, no actor enjoys substantially

more ties than any other actor and therefore no actor is more central than any other.

Conversely, when the variance in the number of network ties per actor is high, some

actors have proportionately more ties and therefore are more central than others. In a

strongly embedded network, optimally functioning network the ideal situation would be

that all actors enjoy a roughly equal number of ties to one another. If all nodes in the

network have the same degree centrality than the network centralization is 0.

3.2.2 Core-periphery Model

The core-periphery structure of a high-tech cluster will be considered as it can give clues

about the hierarchical structure of ties within a network as well as about the process of

development. A high-tech cluster network may for example consist of a small core of

closely interconnected actors surrounded by a periphery of actors whom are loosely

connected to this main core through a low number of direct or indirect ties. The

difference between the inter-connectedness of the core versus the loosely connected

periphery can for example indicate the presence of a hierarchy in the network or the

existence of a ‘backbone’, where one group of actors (the core) is more important than

another in the formation and functioning of the network.

While historic development data of the LBSC cluster is beyond the scope of this research,

the current measures do partially include such a historical perspective based on the fact

that the position of an actor in the network depends on the number of ties an actor has,

which is related to the number of jobs an actor has had, which can be expected to be

related to the number of years an actor has been in the network and the age of an actor.

29

Page 31: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

At the other end of the spectrum lies a network which has no clear visible core and where

all ties are relatively equally dispersed. In such a case, where there is a very large core

and only a small or virtually no periphery, it can be argued that it is beneficial for the

overall functioning of the network. Any given actor in the network can more easily reach

another and knowledge can be transferred quickly and efficiently. Such an equally

dispersed network can be related to a high level of embeddedness. It is hence expected

that the LBSP scientist network does not have a strong core-periphery structure, unlike

the structure found in the Cambridge and Munich clusters. All scientists are expected to

ideally be on a similar level, without there being a strong hierarchy present in the

network. There is thus also not expected to be any evidence of a backbone structure in the

scientist network. The structure will be analyzed primarily based on a visualization of the

network topology.

3.2.3 Path Length

The path length, or geodesic distance, is determined by the average distance between all

pairs of nodes. It shows the network interconnectedness, so how far each actor is from

another actor as a source of information. When a network is highly connected it suggests

that there is a system through which information is likely to reach everyone and in a

relatively short amount of time. A short average path length also increases the chances of

information reaching other actors. The path length found is compared to the path length

that would results if the links between actors had been formed at random as it is size

dependent. If the network is highly locally embedded, the path length is expected to be

significantly lower than the random path length. Path length is also an indication of the

diameter of the network, and a smaller diameter compared to a given number of nodes

can indicate a higher level of embeddedness and performance.

It should be noted however that an extremely short path length may however not always

be a desirable characteristic for a network to posses. This can indicate that the actors in a

social network are highly similar in terms of for example their knowledge bases, which

can reduce the possibility of knowledge diffusion and spillovers (Cowan, 2004). Based

30

Page 32: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

on these characteristics, the path length for the LBSP scientist network is expected to be

relatively short, but not extremely short.

3.2.4 Cluster Coefficient

The cluster coefficient measures the degree to which a network can sustain connected

when nodes are removed from the network. This robustness will be high when nodes are

organized into cliques, where everyone knows everyone. When one person drops out of

such a clique, the remaining members will all remain connected. In assessing the degree

of clustering, it is thus useful to look at the direct neighborhood of actors. It can also be

useful to compare the cluster coefficient to the overall density.

Two different measures can be used. The overall graph clustering coefficient is the

average of the densities of the neighboring areas of all actors in the network. The

weighted cluster coefficient on the other hand gives a weight based on the neighborhood

densities proportional to their size. Actors with larger neighborhoods will receive more

weight in calculating the average density. As networks larger in size are generally (but

not always) less dense than smaller networks, the weighted average clustering coefficient

is normally less than the un-weighted cluster coefficient (Hanneman & Riddle, 2005).

If a high degree of clustering is visible in the network it can be an indication that some

groups of individuals are relatively more embedded in the network than others and that

embeddedness occurs very locally in certain neighborhoods. A high level of clustering

may also be an indication of the existence of a certain degree of hierarchy within the

network. A high clustering coefficient thus does not indicate a high level of

embeddedness for the whole cluster but it instead indicates strong local differences

between groups of actors. This measure can hence be related to the core-periphery

structure mentioned earlier.

3.2.5 Small-world Analysis

The small-world characteristics of a network can be measured by comparing the actual

path length and cluster coefficient value to the values of a network of the same size and

31

Page 33: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

density if it had been randomly formed. It must be noted that there are 2 different types of

structures of a network that can lead to the small-world properties of a high cluster

coefficient and a low path length. These structures vary considerably and will have a

different degree of centralization. One structure involves a network consisting of several

highly clustered groups of actors which are isolated except for being connected through

single direct paths, which leads to a low level of degree centrality. This is similar to a

hub-and-spoke type of network. The second way is through a network which consists of

more evenly spread out clustered groups which are close to one another and connected

through various actors, thus having a high degree of centralization. The former may have

a stronger form of an ‘achilles heel’, where few paths are crucial in order to keep the

clusters of actors connected and thus keep path length short. For the robustness of the

network the latter is hence the ‘stronger’ type of network. Whether either one type of

structure implies a higher level of embeddedness in the network however is debatable.

Considering that a lower level of centralization as discussed earlier can be related to a

higher level of embeddedness, the ‘achilles heel’ network is the expected small-world

structure for the LBSP scientist network.

3.3 Types of Actors

Once the network of the scientists has been analyzed, the network of the MT members

will be considered based on their importance in high-tech cluster networks as supported

by theory. The two networks will be compared to one another based the measures

mentioned for the scientists, with a focus on the differences between the two types of

networks.

Ideal for comparison would be if the MT network consisted of exactly the same number

of actors as the scientist network. As this was not possible based on limitations of the size

of the MT member data base, the size dependent measures will be taken into account. It is

expected that the MT member network exhibits similar characteristics in the level of

embeddedness and topology, as the scientist network as they are expected to play similar

roles in the micro-dynamics of job mobility network formation.

32

Page 34: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

3.4 Central Actors: Bridging and Brokerage

The structure of the combined network is expected to posses the characteristics of a well

developed and well functioning type of network. In this combined network some actors

are expected to perform a bridging function between the 2 types of actor networks.

Actors with the ‘special’ characteristic of being included in both the scientist and MT

member database could play a central role in the overall network.

Whether the manager-scientists do in fact play a bridge-building function can be tested

using the betweenness centrality measure, which is defined as the number of times that

any node needs a given node to reach another node by the shortest path (Freeman, 1979).

Actors in such a broker position must have a high level of betweenness centrality, as the

two types of actors are expected to be indirectly connected to one another through a

broker actor. Tested will be whether the 10 scientist-managers in the dataset on average

have a higher level of betweenness centrality than the rest of the scientists and managers.

Ideally a highly embedded cluster network would have few actors with such an above

average level of betweenness centrality, as it may be better for the stability of the network

to not be dependent upon a few central actors. The betweenness centrality of these

scientist-managers in the network will be considered as opposed to the historical

development of the cluster. As those who have been in the network longer are more likely

to have had multiple jobs and are thus more likely to be central to the network, part of the

historical aspect does influence this measurement by definition as mentioned. Related to

this historical aspect, it is important to note that there may be intervening variables which

influence the level of betweenness centrality. As shown in Diagram 1, the age of an actor

can influence both the measure of betweenness centrality and the chance that an actor is

both a scientist and an MT member.

33

Page 35: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

Scientist+MT Member

Actor Age

Betweenness Centrality

Diagram 1: Age as intervening variable

The former is because an older actor is more likely to have had a larger number of jobs

within the cluster, and the latter is due to the fact that an older actor is more likely to have

started as a scientists and then to have become an MT member later on during their

career. The fact that these measures are all interrelated can distort the measure as the

highest levels of betweenness may hence be the result of these actors being the oldest

actors in the network as opposed to having a broker position due to their function. As age

is not included in the database however this will not be corrected for so the results will be

biased.

A similar bias may occur based on a possible relation between the level of betweenness

centrality of actors in the network and the number of patents (or publications) that an

actor has. An actor with a high number of patents is more likely to have worked in the

cluster for a long time and with many other firms and scientists. This is reinforced by the

fact that multiple patents included in the dataset were joint projects of two or more firms.

The number of patents is again also directly related to the age of an actor and the

possibility of being a scientists-MT member, as older actors are more likely to have had

an extensive scientific career during which they have patented inventions, which is a time

consuming process in the bio tech industry. While the number of patents is available, it

has been decided that it is not in line with the qualitative nature of this analysis to go into

a quantitative analysis of the influence that the number of patents has on the level of

betweenness centrality of an actor.

34

Page 36: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

4. Data Collection

The data necessary for the analysis of the research questions consist of a scientist data

base and an MT member data base. First of all the generally applicable data collection

processes will be explained, after which the specific data sets for the scientists and MT

members will be considered. The data collection has been based on a list of companies as

provided by the LBSP management, which includes all companies that were located at

the park between 1985 and 2009. Several companies included in the list have ceased

business, have moved locations or have either merged or been acquired, which has been

consistently dealt with in the data sets. The most recent form of each firm within the park

has been used as variable. It is not considered to be a problem that based on this method

an employee of an acquired firm is considered to have worked for the acquiring firm, due

to the likelihood that employees are transferred to the new firm. Such merged and

acquired firms also turned out to be uncommon in the data set.

The University of Leiden and LUMC in some research may be considered as one firm,

but due to the size of both the university and the LUMC it is not realistic to assume that

employees from various departments are in contact with one another. Instead the choice

has been made to separate the departments as much as possible in order to avoid any bias,

including the separate institutes related to the University of Leiden. This resulted in a list

of 128 firms, institutes and departments.

In the process of building the two datasets of the job histories of the scientists and the MT

members, the number of jobs that an actor had within the cluster and the order in which

the actors had these jobs was crucial to determine. First it was important to determine

what a ‘job hop’ constitutes, and how ties are formed. Ties between individuals in the

dataset are created through employment, publication or patenting with a common firm.

Under this rule of tie-formation, ties linking individuals are only formed through job

mobility. As an employee switches jobs he or she is expected to stay in contact with

former colleagues while forming new ties at the new firm. Co-patenting and co-

publication are included as job hops because there has likely been a high level of

35

Page 37: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

interaction between such actors over an extended period of time, similar to if the actors

had been colleagues.

The order in which actors had the jobs was important to determine because the most

current jobs of actors had to be excluded to avoid bias towards current networks as is

applied in the method by Casper & Murray (2005). This is based on an argument by

Newman & Park (2003) that “within affiliation networks coefficient correlations are

biased upwards as groups of individuals are selected into the network on the basis of a

current joint affiliation, such as working for the same company” (Casper & Murray,

2005). By including only prior functions and by excluding any links created based on

current functions, this problem is circumvented.

For an actor to be included in the network they must thus have one visible ‘job hop’ next

to their most recent function within the park which adds up to 3 cluster jobs. If an actor

did not have at least 3 jobs within the cluster throughout their career they were excluded

from the dataset. Ties are assumed to last indefinitely. There is some evidence that a 5 to

10 year time-frame is more realistic (Casper, 2007; Uzzi & Spiro, 2005), though this is

debatable as colleagues can remain tied (friends) for life. Especially in the modern era of

social networking websites it has become easier to stay or get into contact with former

colleagues, so a decay-factor will not be included.

In case an actor worked for 2 companies at the same time, the job first started counted as

the oldest held job of the 2. The order of jobs was inserted based on 1 being the current

job, 2 being the previous job and so forth. In case an employee no longer worked for at

the LBSP in 2009 they were still included in the dataset. Both of the datasets built are

assumed to be incomplete in terms of including all scientists and MT members and thus

consist of a large sample as opposed to full networks. The MT member network however

is assumed to be reasonably complete as the database as provided initially was

constructed using a complete list of firms in the LBSP as provided by an employee of the

LBSP itself. The scientist network on the other hand was based on patent data, which

creates a bias toward pro-patent firms, research intensive firms and ‘star’ scientists. This

36

Page 38: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

is not considered to be a problem however as the scientist samples are intended to include

the most important scientists only and are designed so as to contain a minimal level of

bias in other areas, as will be explained in the specific data collection sections.

After the deletion of the most recent jobs of the actors, a 2-mode matrix was built of the

individuals versus the firms using 0 to indicate no prior employment and 1 to indicate

prior employment. This matrix was converted into a 1-mode matrix of individual versus

individual using Ucinet, where 1 means that the actors have worked at a common firm.

All ties are assumed to be reciprocal. The network visualizations have been made using

the 1-mode networks in Netdraw and the measures have been calculated using Ucinet.

4.1 Scientist Network

The scientist database was based on a patent data base as retrieved from the European

Patent Office. The OECD ‘regionalized’ this data by dividing it into the region of

inventors and applicants. From the rough dataset the firms within the ‘Leiden &

Bollenstreek’ area were selected (code NL331) and the companies within the park were

handpicked from this list. This resulted in a list of 634 scientists with patents at 20

different firms. The patent data is an appropriate method to identify the scientists most

relevant for the formation and functioning of the cluster network as ‘star’ scientists are

expected have the most influence on cluster performance and success.

To check the completeness of this data, it was checked by hand whether the other firms

on the list of firms at the LBSP applied for any patents. Some had done so through

foreign offices. The patents were only included if a Dutch inventor was involved as it can

be assumed that some of the knowledge will have come from the Dutch firm to which the

inventor is related. This provided a partial job history of these scientists.

Next the double entries were eliminated and the scientists were sorted based on the

number of total patents. This was done because scientists with the highest number of

patents are expected to be the star-scientists who play the most prominent roles in the

37

Page 39: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

innovation and success of the cluster. In case scientists had the same number of patents

they were sorted in alphabetical order based on their last name.

This database was added to and double checked by searching for the job histories of each

scientist using all possible online sources and databases. This included networking

websites such as LinkedIn, company websites, publishing databases, patent databases,

personal websites, university websites, and newspaper articles. This was done so as to

obtain the most complete job history record possible and to avoid any bias towards for

example jobs with pro-patenting firms or the largest firms. Next to these sources a

snowball sampling method was also used to randomly double check who scientists had

worked together with by for example going through the list of patent applicants listed

with a patent and co-publishing authors.

The job histories of scientists were checked in descending order of the number of patents

until a total of 71 scientists had been identified with a minimum of 3 jobs within the

LBSP. This was done in order to have the same number of actors as the Cambridge

cluster so as to be able to optimally compare these two cluster networks. It turned out that

71 out of a 124 scientists researched had a minimum of 3 verifiable functions with a firm

in the LSBP. As the most recent functions of the scientists were excluded, it appeared

that many new spin-off companies were not included in the final data base, which is a

possible source of bias towards older firms and employees.

Another criticism is that some jobs could have been missed in this process as there might

not be any online evidence connecting scientists to a firm. The last criticism is that the

patent data does not lead to a representative sample of scientists from the cluster, but

rather displays the cream of the crop when it comes to scientists. The resulting network

therefore may not be representative of the cluster network of all scientists or employees.

This implicates that caution must be taken when generalizing the results. Overall

however it is believed that the final data base gives a relatively unique and complete

picture of the job mobility of the most important scientists within a high-tech cluster.

38

Page 40: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

4.2 Management Team Member Network

In order to build a data base of the job histories of the MT members in the LBSP a ready-

made data base was used as basis, which was provided by the LBSP foundation. This

included a list of 163 MT members working within the park in 2009 and only included

firms in their job histories that were located at the LBSP at that moment. MT members

are defined as senior managers, chief executives, chief scientific officers, chief financial

officers, vice presidents or any other senior employees listed as senior management by

the firm. In order to check the completeness of the MT member list and the job histories

as provided all possible online resources were used. Often company websites gave

complete overviews of the current higher management, but previous functions were more

difficult to identify. Each MT member was thoroughly researched by hand, which lead to

many additions to the original data set. A snowball sampling method was again also

applied to double-check the completeness of the database. This resulted in a list of 49 MT

members out of 163 who could be related to a minimum of 3 firms within the LBSP

cluster. It is a limitation that the original database was limited in size, as it would have

improved the possibility of comparison if the MT member network had consisted of 71

actors like the scientists network.

39

Page 41: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

5. Results and Analysis

The results will now be presented along the lines of the four main research questions and

in the same order as the methodology sections. The first results will thus be of the level of

embeddedness of the scientist network, second will be the topology of the scientist

network, third the comparison of the scientist and MT member networks and lastly the

central actors in the combined network.

5.1 Level of Embeddedness of the LBSP Scientist Network

Based on the visualization of the scientist network as shown in Diagram 2, the LBSP job

mobility network appears to be extensive and well developed. The scientists are highly

mobile between firms at the park, especially when compared to the Cambridge and

Munich cluster networks which are shown in Diagram 3 and 4. The LBSP network

displays a far higher and more evenly spread level of interconnectedness and thus of

embeddedness.

40

Page 42: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

Diagram 2: Visualization LBSP scientists (71 actors)

Diagram 3: Visualization Cambridge scientists (71 actors) (Source: Casper & Murray (2005) “Cambridge career affiliation network” pp. 63)

41

Page 43: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

Diagram 4: Visualisation Munich (82 actors) (Source: Casper & Murray (2005) “Munich career affiliation network”, pp. 64)

In order to further asses the level of embeddedness the number of actors connected to the

main component and the density of the network were calculated. Of the 71 scientists in

the LBSP network, 67 are connected to the main component as shown in Table 1. This is

94.4% of all actors, which is a very high number and almost identical to the 95% as

found by Casper (2007) in the San Diego bio science cluster. This level is higher than the

73% of Cambridge and the 64% found for the Munich cluster. The density of the LBSP

scientist network as shown in Table 2 was 0.194, which indicates that 19.4% of all

possible ties are formed within the network. This is therefore neither a very dense nor a

very sparse network. In comparison to a network of the exact same size such as

Cambridge however the LBSP does have a higher level of density, as the Cambridge

network has a sparse level of interconnectedness with only 6.8% of all possible ties

formed. The extremely high percentage of actors connected to the main component and

the relatively higher level of density indicate that the LBSP is strongly embedded in a

cluster network.

Important to note is that in the scientist network, 32 of the total of 128 firms, departments

and institutions in the park are included in the final data base which excludes current jobs

of scientists. There is thus a strong bias towards certain departments, which are mainly

associated with the Leiden University and the LUMC as predicted based on the fact that

most scientists start their careers with the university.

Table 1: Main component connectedness

Total number not in main comp

Number of isolates

Largest (% of total)

Cambridge 17 13 52 (73%)

Munich 21 13 56 (64%)

LBSP Sc. 4 4 67 (94.4%)

Table 2: Overall Network Density

42

Page 44: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

Network Density

Cambridge 0.068

Munich 0.068

LBSP Sc 0.194

5.2 Topology of the LBSP Scientist Network

Based on the visualization in Diagram 2, the LBSP scientist network does not appear to

have a small or strong core of actors. Instead it appears to have a very large core which is

not focused around a few central actors but is relatively evenly spread out. The periphery

is hence very small or non-existent and there is no clear back-bone visible. This is very

different from the networks in Cambridge and Munich, which consist of small central

cores that are strongly interconnected and are surrounded by a loosely connected

periphery.

The visual image of the network is supported by various network measures. The level of

degree centrality found as presented in Table 3 is 37.7%, which coincides with the

generally central structure of the large core. The level of centralization is higher than in

Cambridge (20%) but not very high overall. There thus are some actors present in the

network who are more central than others and who have relatively more ties, but the

variance in the number of ties per actor remains relatively low. This indicates that there

may be a stronger hierarchic structure present in the LBSP cluster compared to the

Cambridge cluster, but that the overall power within the LBSP network is more evenly

spread out over the actors.

The average path length of the scientist network shows that the network is not randomly

generated but that it is likely to be influences by social dynamics. An actor must pass

through 2.08 actors on average to reach any other actor, compared to 2.312 for the

randomly generated network. This is not a large difference, but when comparing to

Cambridge with a path length of 3.701 it indicates a relatively efficient network for the

transfer of information. The diameter of the network is hence also smaller than the

43

Page 45: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

Cambridge and Munich networks which is positive for network performance. The

network would be more efficient if it had a lower path length, but this is perhaps not

realistic as there will always be some actors located in less central positions based on

fewer connections due to for example a difference in age. A very short path length as

mentioned may in fact not even always be a desirable characteristic for a network to

posses.

Next to the network being efficient, the cluster coefficient suggests that the LBSP

network is also highly robust. The cluster coefficient is 0.824 compared to 0.126 for a

randomly generated network, as presented in Table 4. Cambridge and Munich showed

similar high levels of ‘cliqueness’. This indicates that some groups of actors may be more

embedded in the network than other groups. As the LBSP however has a higher level of

density than Cambridge (0.194 > 0.068) while having a similar cluster coefficient, the

LBSP can be said to be more evenly distributed.

That the LBSP network appears to be a well developed and well functioning network is

supported by the evidence that the network possesses small world characteristics. There

was both a high cluster coefficient and a relatively low path length. Other social networks

that have been labeled ‘small world’ found path lengths above their random path length,

similar to Cambridge, averaging between 2.5 and 3.5 (Kogut & Walker, 2001).

Considering the structure of the small world network, the LBSP does not appear to have

the structure of an ‘achilles heel’ as mentioned nor the strongest type of network, based

on a medium level of network centralization. It rather seems to hold the middle ground

between these 2 small-world structures. The existence of small world properties in a

medium dense network as opposed to a dense network again suggests that the formation

of the network is not a random process but that there are social networks driving the

characteristics of the network structure (Casper, 2007).

44

Page 46: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

Table 3: Degree of centrality / network centralization

Mean node centrality S.D. Network centralization

Cambridge 4 4.6 21.3%

Munich 3 7.0 19.8%

LBSP Sc. 13.57 9.2 37.7%

Table 4: Path Length

Actual network Random comparison network of same size and density

Path length Path length

Cambridge 3.701 2.312

Munich 3.578 2.085

LBSP Sc. 2.080 2.312

Table 5: Cluster Coefficient

Actual network Random comparison network of same size and density

Cluster Coefficient Cluster Coefficient

Cambridge 0.855 0.126

Munich 0.835 0.148

LBSP Sc. 0.824 0.126

Table 6: Small World Analysis

CC Path Length Random CC RandomPath Length

LBSP Sc. 0.824 2.080 0.126 2.312

LBSP MT 0.650 2.561 0.148 2.085

LBSP Sc+MT 0.835 2.259 0.126 2.312

45

Page 47: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

5.3 Type of Actor Networks

There are clear differences visible between the scientist and MT member networks,

though the MT member network does contain some of the properties of a well-

functioning and well-developed network. While it is difficult to compare the 2 based on a

difference in the number of actors in the networks, the MT member network appears to

be more similar to the networks found in Cambridge and Munich than to the scientist

network. Comparatively few actors are connected to the main component (53.5%

<94.4%). This is similar however to the proposed industry average of 50% as suggested

by Owen Smith & Powell (2004). There are 13 actors who are isolated from the main

component but who are connected in pairs and in a small group. This can be driven by the

low number of interconnectedness in general as MT members may not be as mobile as

scientists, but it can also be driven by the low number of firms included in the dataset due

to the low number of actors (49<71). The relatively low number of actors decreases the

chance that actors are connected to the main component through common employment.

The network is sparse with 6.3% of all possible ties actually formed compared the 19.4%

for the scientist network. Considering that larger networks by definition become sparser,

the MT member network is relatively even sparser that the scientists network. This all

indicates that the MT member network is relatively less embedded than the scientist

network.

The MT member network is not guided by a few central actors compared to the scientist

network, with a level of centralization of 17.9% compared to 37.7% for the scientists.

The core of the network is smaller compared to the total number of actors in the network,

but also appears to be evenly spread out. A small periphery of about 7 actors who are

only connected to the core through 2 or 3 ties is visible. There appear to be 3 small cores

that are connected through only a few actors. This structure is reminiscent of the ‘achilles

heel’ structure as mentioned in the small-world network structures. The path length of

2.561 and cluster coefficient of 0.065 compared to a random path length of 3.062 and

cluster coefficient of 0.016 do indicate that the network possesses small world properties.

46

Page 48: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

As explained this is beneficial for network robustness and efficiency and thus for overall

network performance.

The larger number of ties of the scientists suggests that they have a higher level of social

capital compared to the MT members. While the MT member is also robust and efficient,

it is not as extensive as the scientist network. The networks are difficult to compare

however based on a difference in the number of actors. It would strengthen the analysis if

the MT network size could be increased. Another influential factor may the fact that a

larger sample of firms is included in the MT member data base, as 42 out of 128 firms are

included in the final data set compared to 32 out of 128 for the scientist data base, despite

the larger number of actors in the scientist data base. This leads to a more spread out and

less biased selection of firms for the MT members, which can lead to lower clustering

and density values for example. The comparison and the characteristics of the MT

network overall however do still offer support for the inclusion of MT members in the

network analysis as there is evidence that there is a considerable degree of job mobility

and proof of social network dynamics.

Table 7: MT Member Network Characteristics

Network Measure

Main Component Connectedness 26/49 (53.5%)

Density 0.063

Network Centralization 17.9%

Path Length (random) 2.561 (3.062)

Cluster Coefficient (random) 0.065 (0.016)

47

Page 49: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

Diagram 5: Visualization LBSP MT (49 actors)

5.4 Central Actors in Combined Network

When combining the network of the scientists with the MT members it resulted in a

network that appears highly similar to the scientist network. The network seems to be

well developed and strongly embedded, with 83.6% connected to the main component

and a density of 0.117. The actors who are both scientist and manager have been

highlighted to display their position in the network. Some clearly occupy central

positions as predicted while others have relatively few connections and are located on the

periphery of the cluster as opposed to connecting groups.

The average betweenness centrality of the 10 scientist-managers supports this observation

as displayed in Annex 1, as it is 60.8 compared to an average of 48 for the other 100

scientists and managers. Compared to the maximum betweenness centrality that was

found of 574.5, this is not a high average, which therefore offers limited support for the

notion that scientist-managers perform a brokerage function in the combined network.

The average of 60.8 is pulled up by the few scientist-managers who do occupy central

positions, as there is a high level of variance present in the betweenness centrality of the

48

Page 50: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

10 scientist-managers. For the top 10 actors in the combined network the average

betweenness centrality is 334, and only 1 scientists-manager is included, as visible in

Annex 2. This can be translated into the fact that 70% of the top 10 broker positions are

occupied by scientists compared to 61% if the actors had been randomly picked. Based

on this evidence scientists thus seem to occupy the most central positions in the network,

though this evidence is not convincing. These results could have been influenced by the

fact that scientists have more connections based on the larger data set, but might also be

influenced by other variables such as the age of an actor. Whether the scientists in the

database however are older than the MT members is beyond the scope of this research.

Diagram 6: Visualization combined Scientist and Management network (110 actors)

Table 8: Betweenness Centrality

Mean S.D. Max

LBSP Sc + MT (100) 48.0 105.9 574.5

SC & MT (10) 60.8 83.4 198.4

49

Page 51: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

6. Conclusions

6.1 Level of Embeddedness

There is strong support for the notion that the LBSP as successful cluster is highly

embedded in a cluster network, as expected. The network possesses efficient and robust

properties wherefore information can easily and quickly reach all scientists in the

network. Especially compared to other clusters the LBSP has favorable network

properties. There also appears to be evidence that networks formation is guided by social

network dynamics based on the non-random formation of the job mobility network of

scientists. Underlying reasons are that the boundaries of the network are determined by a

select group of firms, which in turn determine connectedness between individuals, which

creates patterns in the structure and topology of the network.

The level of job mobility of the scientists overall was surprisingly high. It turned out to

be much higher than for the MT members and previously analyzed clusters. This high

level of job mobility indicates a high level of creation of social capital, which is

beneficial for cluster success. It also indicates that the LBSP is likely to possess

characteristics that fuel job mobility, which could include various types of push- and pull

factors such as a tolerance to job mobility, positive institutional factors, the availability of

career opportunities and/or a high degree of cooperation between firms within the LBSP.

Based on the theories and results it appears as though there may be an upward spiraling

mechanism at work, where a high level of job mobility plays a role in the increase in

cluster innovation and success, which in turn may further induce job mobility increases

based on new opportunities. A network thus seems to be self-reinforcing through job

mobility, which again creates positive externalities for the cluster.

The high level of mobility found can also be considered as evidence for the existence of a

high degree of knowledge spill-overs and diffusion. The fact that there is such a high

degree of mobility can also indicate that there is a sufficiently large potential labor pool

within the cluster in case you include current cluster employees in the potential labor pool

as suggested. In case the potential labor pool is defined as employees not currently

50

Page 52: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

employed by a firm in the LBSP however, than it could imply that there is perhaps even a

relative scarcity of suitable outside employees from which LBSP firms can recruit, as was

suggested by Annelies Hoenderkamp. The LBSP firms may rely mostly on current park

employees to fill vacancies and start new ventures, which would lead to a high level of

local mobility, as found. This can be connected to the existence of a ‘recycling’

mechanism of employees as suggested by Casper (2007) and Saxenian (1994). Whether

such a shortage of scientists is truly the case ought to be further investigated by analyzing

the degree to which new functions within the LBSP are filled by employees with previous

jobs at the LBSP.

6.2 Topology of a Successful High-tech Cluster

The characteristics of the topology of the network of the LBSP were mostly as predicted

by theory, and indicate that the LBSP network has beneficial properties for network

performance. There is some hierarchy present in the network, though the power is

relatively evenly distributed. The presence of hierarchy is not as ideal for the transfer of

information as a non-hierarchic network, but is inevitable based on the individual

differences in the extent of scientist careers. The fact that the hierarchy is comparatively

average whilst simultaneously evenly distributed is therefore considered to be a positive

characteristic of the LBSP network.

The large evenly distributed core and small periphery of scientists also support the high

level of embeddedness and relative low degree of hierarchy in the network, which is

positive for network performance and hence cluster performance. This structure

optimizes the role of individuals in the spread of knowledge which positively impacts the

spread of innovation in the cluster. In comparison to the Cambridge and Munich clusters,

the scientist network displayed very different properties. This can mean that scientists in

the LBSP are embedded to a higher degree but may also be related to a difference in the

data collection or completeness of the data for Munich and Cambridge or it may be

driven by a difference in institutional factors for example.

51

Page 53: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

No evidence was found to support the theory of development of a cluster as proposed by

Casper (2007). This could indicate that the San Diego case, where the management of a

failed company formed the basis of a successful bio science cluster network, was a rare

event. It is more likely to be the case that such a backbone is present, but is not visible

due to the strong development of the network around the core actors over time. A

historical development study of the LBSP scientist network could clarify whether the

latter is the case. The scientist network further displays ideal properties for network

performance as small-world characteristics are present and the power distribution

between groups within the cluster is not highly skewed.

6.3 Type of Actors in Cluster Networks

There appeared to be sufficient theoretical evidence that the type of actors that can

influence cluster performance through social networks include both scientists and MT

members. The networks of these two types of actors however vary more than expected.

Interestingly the MT member network is actually more similar to the Cambridge and

Munich networks than to the scientist network. The fact that the MT network

characteristics differ from the scientist network can indicate that a difference in function

can influence the level and dynamics of mobility. Are managers really less mobile and

less embedded in a cluster network than scientists? In order to extend the support for this

claim further research is necessary.

The difference in the two networks also stems from the difference in the number of actors

and possibly from a difference in the number of firms included in the final data set. The

fact that the MT member data set includes a larger sample of firms than the scientist data

set means that it is less biased, and could be an important reason why the network is not

as highly interconnected as the scientist network.

The difference in the networks found however can also be considered as support for the

inclusion of MT members in overall cluster network analysis, as perhaps they are in fact

part of a different type of epistemic community and have different social networks than

scientists do. MT members undoubtedly play an important role in the growth and

52

Page 54: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

performance of clusters. Regardless of the reasons behind their lower levels of mobility,

there is theoretical and empirical evidence that supports the importance of management

functions in the job mobility network dynamics of a cluster. Just as the scientist network

the formation of the MT member network is non-random. The MT network is both robust

and efficient and hence possesses several of the properties mentioned in theory related to

an ‘ideal’ network. This is all evidence that social networks drive the formation of the

network, which provides the means for social capital to be created which positively

influences cluster success.

The theoretical and empirical support for the importance of senior management members

in cluster network analysis has important implications for future cluster network analysis.

By only including scientists in the analysis of cluster networks, an important group of

actors is left out who have a significant amount of influence on the network itself, the

functioning thereof and the overall performance of the firms in a cluster. By only

considering the senior management networks, the scientists are left out, while they

undoubtedly play an important role in cluster success. The interaction between, and the

integration of these two types of actors is vital for cluster success by combining scientific

and market knowledge. The scientists are dependent upon the MT members and vice

versa. Their networks therefore ought to be combined in order to provide an accurate

picture of the social dynamics within a cluster which influence cluster success.

6.4 Central Roles in High-tech Cluster Networks

The combined network displayed well-developed characteristics as predicted, and the MT

network and scientist network appear to be integrated. While the expectation was that

scientist-managers occupy central positions in the combined network, evidence to support

this claim was relatively weak. It appeared that percentage wise more scientists

performed bridge-building functions instead, though not to a convincing degree. Theory

and empirical evidence do offer support that central actors can perform important

functions as brokers in a network. They contribute to the small-world characteristics of

the network, keeping it robust and efficient in function. It however remains unclear

whether such a central role is related to the function of an actor or whether it can be

53

Page 55: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

attributed to personal characteristics. It is possible that a founder of new firms in the

cluster is a reason why an actor is a central broker in a network, as opposed to the

function of scientist- manager leading to such a central position. The causality is unclear,

and the many limitations of this analysis cloud the interpretation of these results,

wherefore further analysis is recommended.

The overall implications of the results of the analysis are that while the job mobility

networks of scientists and MT members are well developed, it may be worth further

developing through stimulating mobility. This study provides evidence that policy

makers can potentially influence cluster performance and success through influencing job

mobility networks. Based on these results the LBSP should definitely implement their

plans to create and facilitate a labor pool and could for example further support the

collaboration between scientists of different firms in joint projects. It is important that the

policy measures do not solely focus on scientist-mobility but that they also take higher

management team member mobility into consideration based on the importance of the

commercialization of science and entrepreneurial new ventures. On a national level the

LBSP provides a superb example of a well developed high-tech cluster network, which

adds to the existing evidence that a well developed network tends to be part of the

formula for a successful high tech cluster. The network dynamics in a cluster may be an

important factor in explaining the variation in success of high-tech clusters.

The job mobility network characteristics as identified in the LBSP case may partially be

generalized to apply to other high-tech clusters. As mentioned however, it is important to

realize that there are possible institutional forces at play and that the scientist network

includes only star-scientists and not a random sample of employees.

54

Page 56: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

7. Limitations

Various limitations in the methodology and data collection impact the measures and

interpretation of the results, some of which have been mentioned earlier. An important

concern which could be improved upon is the fact that the MT member network consists

of fewer actors than the scientist network, which influenced the network characteristics

and hindered comparison. Another possible driver behind the differences between the

scientists network, the MT network, the Cambridge network and the Munich network is

that the scientist network also included co-patenting and co-publishing activities as

shared job experiences even in case the patent was issued under only one of the firm

names for which a scientist may not have worked for on a daily basis. This is justified for

the scientist network analysis by the fact that co-publishing and co-patenting does

increase the chances that employees have worked closely together and have formed

durable ties, but could lead to an artificial intensification of network connectedness and

hence could lead to difficulties in directly comparing the networks. This is because MT

members are generally not involved in co-patenting and co-publishing activities.

Another concern is that it is highly likely that a few large firms influence the

characteristics of the network. Not nearly all the firms present in the LBSP are

represented. It would be highly interesting for future research to construct a scientist

database based on the database of firms within the LBSP as opposed to using a patent

database. Ideally all scientists of the cluster should be included, or a random sample

should be selected. This would reduce the bias towards pro-patent and larger firms and

toward the star scientists. While this is extremely labor intensive, this would decrease the

level of bias in the scientist network to an unparalleled degree.

Excluding all current positions of actors is also likely to have had a profound impact on

the structural characteristics of network as it excluded many positions in new ventures

and younger firms for example. It thus created bias towards older firms and actors,

specifically towards the university. Perhaps the reason for excluding the current positions

can be reconsidered or worked around. Another issue in the overall methodology is that

55

Page 57: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

no decay of ties was included. It may specifically have impacted the scientist network if

a decay of ties were to be included, as the MT member job histories in general were more

recent than the scientist career histories. It would also have decreased the bias towards

university-based ties, as these tend to be the oldest ties in the data base. This is based on

the fact that more scientists than MT members started their career at the LBSP with the

university and had longer career spans within the cluster in general.

As mentioned another limitation involves the reasons why some actors are more central

in the network than others. A central position is linked to both the level of

interconnectedness and the relative position of an actor in a network, which may be

related to other factors such as actor age, the number of patents an actor has or other

personal characteristics as opposed to the type of function. A higher age is not only

related to a higher chance of being both a scientist and a manager, but also means a

higher chance of more job hops. As it remains unclear what factors exactly influence the

central position of an actor in a network, it could be interesting to investigate exactly

what type of people perform these bridging functions in the network. This could for

example be analyzed by for example including variables such as the exact function

description, actor age, the number of patents, the number of working years, the level of

education (PHD) and the size of the firms for which the actor has worked. Based on the

list of limitations and suggestions for further research one can conclude that there is room

for improvement and further research in this area in general.

The theoretical section also raised an interesting question of whether scientists are locally

or globally mobile. By investigating the mobility of scientists within and outside of the

LBSP more insight can be given as to the motivational factors in the determination of

mobility of employees, which can be interesting for policy makers.

Next to methodological limitations a theoretical limitation should also be mentioned.

There is an interesting study by Taylor (2005) which questions the validness of the theory

of embedded local economic growth in general. Taylor suggests that the model is

overdrawn in that it does not sufficiently incorporate “the imperatives of capitalism, the

56

Page 58: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

impact of unequal power relations and the exigencies of time”. He suggests that it is

overdrawn in that it ‘fetishes proximity, promotes the chaotic concept of ‘institutional

thickness’ and labours under the limitations of the equally chaotic concept of ‘social

capital’” (Taylor, 2005). These limitations undermine the entire theoretical foundation of

this study, yet have not found general support from the wider scientific community. They

are thus interesting to take into consideration in future studies but do not provide enough

theoretical and empirical leverage to impede future research in this area.

57

Page 59: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

Acknowledgements

I would like to thank various people without whom this study would not have been

possible. I would like to thank my supervisor Sandra Phlippen for sparking my interest in

the topic of clusters during the seminar ‘Governance, Clusters and Networks’, for

organizing a group of students who could write their thesis on this topic, for setting up

the project with the Leiden Bio Science Park and for her guidance and critical comments

throughout the process. I especially want to thank her for continuing to supervise me

while I was in Australia. I want to thank Harmen Jousma and his colleagues at the

Leiden Bio Science Park for their practical and theoretical insights and for providing the

data bases of the firms and management team members. Lastly I want to congratulate and

thank my fellow group members for their contributions and would like to wish them the

best of luck for the years to come.

58

Page 60: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

Appendix

Annex 1: Betweenness Centrality Scientist-MT membersActor Name BetweennessBout, Abraham 169.217De Haan, Peter 0.000Ijzerman, Ad 169.876Melief, Kees 0.000Platenburg, Gerard 23.430Spaink, Herman Pieter 46.939Stegmann, Antonius, Johannes, Hendrikus

0.000

Strijker, Rein 0.000van Deutekom, Judith Christina Theodora

0.000

van Wezel,Gilles 198.449Mean 60.791

Annex 2: Betweenness Centrality Top-10 Actors Combined NetworkActor Name Betweenness Type of ActorDrijfhout, Jan, Wouter 574.474 ScientistHeyneker, Herbert, L. 450.313 ScientistHooykaas, Paul Jan Jacob 426.468 ScientistVan Boom, Jocobus Hubertus 405.757 ScientistValerio, Domenico 346.195 ScientistHennink, Wilhelmus Everhardus 296.768 ScientistYallop,Christopher 264.000 MT memberStolpe,Onno 202.585 MT membervan Wezel,Gilles 198.449 BothSchilperoort, Robbert Adriaan 176.696 ScientistMean 334.171

59

Page 61: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

Bibliography

Ackers, L. (2004). ‘Moving People and Knowledge: The Mobility of Scientists within the

European Union’. Retrieved from http://www.liv.ac.uk/ewc/docs/Ackers-

paper03.2004.pdf on 24-11-2010.

Almeida, P., and Kogut, B. (1999). ‘Localization of Knowledge and the Mobility of

Engineers in Regional Networks’. Management Science, 45 (7).

Audretsch, D. B. and M. P. Feldman (1996). ‘R&D Spillovers and the Geography of

Innovation and Production’. American Economic Review, 86(3) 630-640.

Baker, W. E. (2000). ‘Achieving success through social capital: Tapping the hidden

resources in your personal and business networks’. San Francisco: Jossey-Bass Inc

Pub.

Bathelt, H., Malmberg, A., and Maskell, P. (2002). ‘Clusters and Knowledge: Local Buzz

and Global Pipelines and the Process of Knowledge Creation’. DRUID working

paper No 02-12.

Becker, G. S. (1962). ‘Investment in Human Capital: A Theoretical Analysis’. The

Journal of Political Economy, 70 (5) 9-49.

Boschma, (2005). ‘Proximity and Innovation: A Critical Assessment’. Regional Studies,

39 (1) 61-74. 

Brahami, H., and Evans, S. (1999). ‘Flexible Re-cycling and High-Technology

Entrepreneurship’. California Management Review, 37 62-88.

Breschi, S., and Lissoni, F. (2001). ‘Knowledge spillovers and local innovation systems:

a critical survey’. Industrial and Corporate Change, 10 975–1005.

60

Page 62: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

Burton, D. M., Beckman, C.M., and O’Reilly, C. (2003). ‘Early teams: The impact of

team demography on VC financing and going public’, Journal of Business

Venturing,

22 (2) 147-173.

Casper, S., and Murray, F. (2005). ‘Careers and Clusters: Analyzing career

network dynamics of biotechnology clusters’. Journal of Engineering and

Technology Management, 22 (1-2) 51-74.

Casper, S. (2007). ‘How do technology clusters emerge and become sustainable? Social

network formation and inter-firm mobility within the San Diego biotechnology

cluster’. Research Policy, 36 (4) 438-455.

Cohen, W.M., and Levinthal, D.A. (1990) ‘Absorptive capacity: a new perspective on

learning an innovation’. Administrative Science Quarterly, 35 128–152.

Collins, C.J., and Clark, K.D. (2003). ‘Strategic Human Resource Practices, Top

Management Team Social Networks, and Firm Performance: The Role of Human

Resource Practices in Creating Organizational Competitive Advantage’. The

Academy of Management Journal, 46 (6) 740-751.

Cowan, R. (2004). ‘Network structure and the diffusion of knowledge’. Journal of

Economic Dynamics and Control, 200.

Dahl, M.S. (2002). ‘Embedded Knowledge Flows Through Labor Mobility in Regional

Clusters in Denmark’. Paper presented at DRUID Summer Conference on

‘Industrial Dynamics of the New and Old Economy - who is embracing whom?’,

June 2002.

61

Page 63: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

Dickson, D. (2003). ‘Mitigating the Brain Drain is a Moral Necessity’. Science and

Development Network, Retrieved from www.scidev.net/editorials/index on 22-11-

2010.

Dyer, J. H., and Singh, H. (1998). ‘The relational view: Cooperative strategy and sources

of interorganizational competitive advantage’. Academy of Management Review, 23

660-679.

Feldman, M.S. (2000). ‘Organizational routines as a source of continuous change’.

Organization Science, 11 (6) 611-629.

Franco A.M., and Filson, D. (2000). ‘Knowledge diffusion through employee mobility’.

Claremont Colleges Working Papers, 2000.

Freeman, L.C. (1979). ‘Centrality in social networks conceptual clarification’. Social

Networks, 1 (3) 215-239.

Granovetter, M. (1973). ‘The Strength of Weak Ties’. American Journal of Sociology, 78

May.

Granovetter, M. (1985). ‘Economic Action and Social Structure: the problem of

Embeddedness’. American Journal of Sociology, 91 481-510.

Grewal, R., Lilien, G.S., and Mallapragada, G. (2006). ‘Location, location, location: How

network embeddedness affects project success in open source systems’.

Management Science, 2006.

Gulati, R. (1995). ‘Does familiarity breed trust? The implications of repeated ties for

contractual choice in alliances’. Academy of Management Journal, 38: 85–112.

62

Page 64: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

Gulati, R. (2003). ‘Which ties matter when? The contingent effects of interorganizational

partnerships on IPO success’. Strategic Management Journal, 24 127–144.

Gulati, R., Nohria, N., and Zaheer, A. (2000). ‘Strategic networks’. Strategic

Management Journal, 21 203-215.

Hanneman, R.A., and Riddle, M. (2005). ‘Introduction to social network methods’.

Riverside, CA: University of California. Online book retrieved from

http://faculty.ucr.edu/~hanneman/nettext/ on 12-9-2010

Herrigel, G. (1993). ‘Power and the Redefinition of Industrial Districts: The Case of

Baden-Württemberg’, The Embedded Firm, London: Routledge. 227 – 251.

Higgins, M., and Gulati, R. (2003). ‘Getting Off to a Good Start: The Effects of Upper

Echelon Affiliations on Interorganizational Endorsements’. Organization Science,

14 244-263.

Hoenderkamp, A. (2009). Presentation on the Bio Science Cluster in Leiden, November

2010.

Inkpen, A.C., and Tsang, E. (2005). ‘Social capital, networks and knowledge tranfer’.

Academy of Management Review, 30 (1) 146–165.

Irwin, M.D., and Hughes, H.L. (1992). ‘Centrality and the Structure of Urban Interaction:

Measures, Concepts, and Applications’. Social Forces, 71.

Jaffe, A., Trajtenberg, M., and Handerson, R. (1993). ‘Geographic localization of

knowledge spillovers as evidenced by patent citations’. The Quarterly Journal of

Economics, 108 (3) 577-598.

63

Page 65: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

Johannisson, B. (1998). ‘Personal networks in emerging knowledge-based firms: spatial

and functional patterns’. Entrepreneurship & Regional Development.

Kim, J., and Marschke, G. (2005). ‘Labor Mobility of Scientists, Technological

Diffusion, and the Firm's Patenting Decision’, RAND Journal of Economics, 36 (2)

298-317.

Klepper, S. (2002). ‘Personal networks in emerging knowledge-based firms: spatial and

functional patterns’. Industrial and Corporate Change, 11 (4) 645-666.

Kogut, B., and Walker, G. (2001). ‘The small world of Germany and the durability of

national networks’. American Sociological Review, 66 (3) 317-335.

Leborgne, D., and Lipietz, A. (1992). Conceptual Fallacies and open Questions on post-

Fordism’. In M. Stoper and A.J. Scott (ed), Pathways to Industrialization and

Regional Development, London: Routledge, 332-348.

Mallet, J. G. (2004). ‘Silicon Valley North: The Formation of the Ottawa Innovation

Cluster’, in Professor Howard Thomas (ed.) Silicon Valley North, Emerald Group

Publishing Limited 9 21-31.

McEvily, B., and Zaheer, A. (1999). ‘Bridging ties: a source of firm heterogeneity in

competitive capabilities’. Strategic Management Journal, 20 (12) 1133–1156.

Murray, F. (2004). ‘The role of inventors in knowledge transfer: sharing in the laboratory

life’. Research Policy, 33 (4) 643-659.

Newman, N.E.J., and Park, J. (2003). ‘Why social networks are different from other types

of networks’. Physics Review E68 1-9.

64

Page 66: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

Nohria, N., and Eccles, R. (1992). ‘Introduction: Is the network perspective a useful way

of studying organizations?’ Harvard University Press, Cambridge.

Owen-Smith, J., and Powell, W. (2004). ‘Knowledge Networks as Channels and

Conduits: The Effects of Spillovers in the Boston Biotechnology Community’.

Organization Science 51, 5-21.

Porter, M.E. (1998). ‘Clusters and the New Economics of Competition’. Harvard

Business Review, 6 77-90.

Porter, K., Whittington, K.B., and Powell, W. (2005). ‘The Institutional Embeddedness

of High-Tech Regions: Relational Foundations of the Boston Biotechnology

Community’. In Breschi and Malerba: Clusters, Networks and Innovation, Oxford

University Press 261-296.

Powell, W., White, D., Koput, K. Owen-Smith, J. (2004). ‘Network dynamics and field

evolution: the growth of inter-organizational collaboration in the life sciences’.

American Journal of Sociology, 110 1132-1205

Powell. W., Koput, K., and Smith-Doerr, L. (1996). ‘Inter-organizational Collaboration

and the locus of Innovation: Networks of learning in Biotechnology’.

Administrative Science Quarterly, 41 116-145.

Sabel, C. (1992). ‘Studied Trust: Building New Forms of Cooperation in a Volatile

Economy’. Explorations in Economic Sociology, edited by R. Swedberg. New

York: Russell Sage Foundation 104-44.

Saxenian, A. (1994). ‘Regional Advantage: Culture and Competition in Silicon Valley

and Route 128’. Harvard University Press, Cambridge.

65

Page 67: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

Scott, J. (1991). ‘Social Network Analysis: A Handbook’. Sage Publications, Newbury

Park.

Shane, S., and Cable, D. (2002). ‘Network Ties, Reputation, and the

Financing of New Ventures’. Management Science, 48 (3) 364–381.

Sorensen, E. (2005). ‘The democratic problems and potentials of network governance’.

European Political Science, 4 348–357  

Storper, M. (1997). ‘The regional world: territorial development in a global economy’.

The Guilford Press, New York.

Taylor, M. (2005). 'Embedded local growth: a theory taken too far? As in R.A. Boschma

& R. Kloosterman (eds.) ‘Learning from clusters’. GeoJournal Library, 80 69-88.

Uzzi, B., and Gillespie, J. J. (2002). ‘Knowledge spillover in corporate financing

networks: Embeddedness and the firm’s debt performance’. Strategic Management

Journal, 23 595–618.

Uzzi, B. (1997). ‘Social structure and competition in interfirm networks: The paradox of

embeddedness’. Administrative Science Quarterly, 42 35–67.

Uzzi, B., and Lancaster, R. (2003). ‘Relational Embeddedness and Learning: The Case of

Bank Loan Managers and Their Clients’. Management Science, 49 (4) 383-399 

Uzzi, B. and Spiro, J. (2005). ‘Collaboration and creativity: the small world problem’.

American Journal of Sociology, 11, 447-504.

Watts, D. (1999). ‘Networks, dynamics, and the small world phenomenon’. American

Journal of Sociology, 105 (2) 493-537.

66

Page 68: thesis.eur.nl Thesis Fleur de Groot.docx  · Web viewSocial network formation and clusters: A two-way job mobility network analysis of the Leiden Bio Science Park. Fleur Louise de

Wilhite, A. (2001). ‘Bilateral Trade and Small-World Networks’. Computational

Economics, 18 49-64.

Zaheer, A., and George, V. P. (2004). ‘Reach out or reach within? Performance

implications of alliances and location in biotechnology’. Managerial and Decision

Economics, 25 437–452.

67